- ~~~~~~~~~~~~~~~s, 1'i n ec too ,l.l-w. C . % ' ~~~~~~~~V.11cric R. 1,kniciv'-tnga, Brmce D}. 5Stnitti, arldt MitHs M.S w trf . - . - ; . s% { c . ) ~~1""r *4 ( oct NUAc . t T t.it s$ tM*n t ':s \ C '~~~~~~~~~for! 1- 1 I o. i 7 Wn .ats 1[r[>XT,, , ,,,. , , . ,, j , ,., ,, , ~~~and in.,,c2ial t rfitrin 114c i ; 5r~~~~~~~Sock N1lsr1 0 units of capital (gross of transactions costs) at t + j. Thus j represents the gestation length of capital investments in the technology with that index, and R1 represents the (gross) productivity of that technology. We assume further that if K, denotes the total capital stock available at t, K, is simply the sum of maturing capital investments produced through all technolo- gies. Thus, more specifically, all capital-produced by any investment technol- ogy-is perfectly substitutable as an input in final goods production. (The as- sumption that all capital, however produced, is perfectly substitutable in production is relaxed by Bencivenga, Smith, and Starr 1994.) Our assumptions here on capita' production technologies imply that capital investnents are com- pletely unproductive until they mature. This can be thought of as an Austrian model of investment. It is possible to alter the analysis so as to allow all capital investments to mature in one period but at the same time have capital produced using different technologies with different productive lifetimes. This, however, is a more complicated model, and we do not pursue it here. 248 THE WORLD ANK ECONOMIC REVIEW, VOL. 10. NO.2 Since agents are two-period lived, the use of any investment technology with j> 1 requires owners of capital in process (ciP) to transfer ownership of it in equity markets. This is true of CIP in all periods prior to maturity, so that own- ership of CIP is transferred through a sequence of holders in equity-or capital resale-markets. We are interested in considering how the costs of transacting in these markets affect capital accumulation and per capita income, the equilib- rium return on savings, the equilibrium choice of capital production technolo- gies, and welfare in a steady-state equilibrium. For simplicity, we assume a proportional transactions costs structure con- fronting agents who operate in equity markets. Our specific assumption is that tran iferring ownership of one unit of ap produced using technology i that has been in process for h periods (that is, which is j - b periods from maturity), consumes a1i' units of cp. Thus, after a sale of one unit of type (I, b) ciP, 1- ai units remain. Under this specification, transactions costs represent a pure re- source loss. We consider below the alternative case in which transactions costs represent a pure transfer to market makers (or a tax paid to the govermnent). Finally, we assume that when ap matures it is used in the production process and then depreciates completely. This assumption allows us to abstract from the existence of resale markets for mature-as opposed to maturing-capital. Trade Three kinds of transactions occur in this economy: capital and labor are rented in competitive factor markets, final output is bought and sold, and agents trade ownership of c[p in competitive equity markets. We focus on transactions costs in equity markets and c 4herwise keep the model as close to standard as possible (see Diamond 1965 and Azariadis 1992, ch. 13). Therefore, we assume that there are no costs associated with transactions in output or f,ctor markets. We also focus throughout on steady-state equilibria. We therefore omit time sub- scripts wherever possible. Let w denote the (steady-state value of) the real wage rate, and let r denote the rental rate on capital. In a competitive economy, factor prices will equal marginal products of the appropriate factor, so that (1) r=a (2) w=b. Each young agent earns the wage income w, all of which is saved. LetS denote savings by a representative young agent, measured in units of Cum. The only deci- sion confronting such an agent is how to allocate savings among various alterna- tive assets; the available assets are type j CIP ( = 1, . . ., J) of vintage b (h = 1, .. .,j- 1). Mature capital is simply rented to firns. Since there are no transac- tions costs in factor markets, dtis amounts to assuming that aii= 0 = 1, . . Bencivenga, Smitb, and Starr 249 Let SNM denote the amount of type i ar that is 6 periods old acquired by a representative agent. Then, for example, Sh° represents the amount of newly initiated investment in technology i, and Sit1 is the amount of type j cip acquired that will mature in one period. Similarly, let PF4 be the price-Ain units of current consumption-of one unit of technology cip that is b periods old. Since one unit of the final good invested in technology j at any date becomes one unit of tech- nology j CuP (by choice of units), Pf' = 1. Moreover, mature CIP is simply capital, which is rented to firms. As one unit of technologyj cii' yields R1 units of rentable capital on maturity, P;i = rR1 must hold. That is, the price of mature ca is simply the rental value of the associated capital. For j > 1 and 0 c b c pMb must be determined. Without loss of generality, we can assume that transactions costs are borne by sellers of c3p. Then, since each agent consumes only when old, the budget constraints confronting an individual agent are (3) 1, I pi_&Sib < ,;=1 h=O (X4) c< I;l Pi.blSish(mahbi>). Equation 3 imposes that the value of asset purchases does not exceed savings, and equation 4 asserts that old-age consumption is funded by the net-of- transactions-cost proceeds of asset sales. We also impose nonnegative asset hold- ings (no short sales) for all types of assets. Note that while agents take the transactions costs parameters a1" as given, the total quantity of transactions costs incurred is a choice variable that responds not just to the values ahb but to asset prices as well. Agents here care only about the rates of return received on assets, since our model contains no motive for diversification. Thus, all assets actually held in equilibrium must bear a common (gross) rate of return, net of transactions costs, which we denote by y. If technology j is actively used in a steady-state equilib- rium, then technology j cip of all possible times to maturity must be held by some agent. Thus, for allh = 0, . . .,j- 1, the gross rate of return on type (j, h) cIiP must equal y. In other words, then, if technology j is active in a steady-state equilibrium, (5) = (1 - ai"1) pi.1+lJpi.b for allb = O, ...,j-1. When rates of return on all capital investments in use are equated, each young agent is indifferent to the composition of his portfolio. The aggregate composi- tion of investment will, nevertheless, be determinate, as we will see shortly. 250 THE WORLD BANK ECONOMIC REVIEW, VOL 10 NO.2 Long-Maturity Investments The assumption that agents are two-period lived implies that the number of times any investment changes hands is simply its gestation length minus 1. As a consequence, long-maturity investments are traded more times than short- maturity investments. Although this is a necessary consequence of our assump- tions on agents' life cycles, we believe that it captures a feature that is generally true of reality. A more general but much more complicated formulation would allow for longer-lived agents who make endogenous choices about the holding periods for each of their investments. These holding periods would typically depend on the patterns of agents' incomes and expenditures, as well as on the structure of transactions costs. Indeed, we could additionally add the feature that agents' incomes and expenditures are subject to some uncertainty. In this case, the oc- currence of various shocks would motivate asset market transactions that agents could not perfectly predict in advance. In either of these more realistic (and hence complex) formulations, we would expect agents to have determinate asset demand functions. These demand functions would, however, necessarily depend on the structure of transactions costs. So long as transactions costs are greatest for assets of longer maturities, reductions in trans- actions costs will be conducive to the transfer of savings from shorter- to longer- maturity instruments. It is this transfer that is central to our results. We therefore do not believe that the introduction of uncertainty or of richer life-cycle savings behav- ior into the analysis would tend to alter our main conclusions. III. STEADY-STATE EQUILIBRIUM In order to describe the steady-state equilibrium capital stock, output level, and rate of return on savings, we need to know two things: first, which capital production technology (or technologies) will be in use in such an equilibrium; second, how savings will be divided among cIP of different gestation periods in this technology. The Equilibrium Choice of Investment Technology As intuition might suggest-and as Bencivenga, Smith, and Starr (1995, 1996) show formally-in equilibrium the capital production technology or technolo- gies in use maximize the internal rate of return on investnent, net of transac- tions costs. An investment of one unit in technology i at t yields ki units of capital at t + j, net of transactions costs, where _ i-I (6) Rz, -Ri (1 - abl ). M_O This capital is then rented at its competitive rental rate, r. With this 'point input, point output" technology, the internal rate of return on technology i is Bencivanga, Smith, and Starr 251 Vrk)llk. The equilibrium choice of capital production technology-in steady state-is that which maximizes this internal rate of return. Thus, if j' is the equilibrium choice of capital production technology, (7) j =argmax [(aRi)'Q. We assume throughout that j' is unique, which will generically be the case. The equilibrium rate of return on savings is equal to the internal rate of return on the equilibrium capital production technology. Therefore, in equilibrium, (8) y = (AR,411' holds. To summarize, in choosing which technology to use, agents care only about the intenal rate of return on investments, net of transactons costs. The costs of trans- acting in equity markets influence the equilibrium capital production technology through their influence on this rate of return. After characterizing the remainng aspects of an equilibrium, we will pursue the implications of this observation. The Capital Stock and the Composition of Savings The equilibrium level of the capital stock-as well as, by implication, the equilibrium level of output-depends very heavily on how savings is allocated between existing, but not yet mature, cIP and the initiation of new capital invest- ment. Purchases of the former represent equity holdings so, in effect, the divi- sion of young people's saving between equity and new investment determines the steady-state capital stock. Let oih denote the fraction of per capita saving (of young people) invested in purchasing technology j cap of vintage hi2 In a steady state, Oih = O holds for all technologies that do not maximize the internal rate of return on investment. Thus, only Oi>* > 0 holds. We now consider the equilibrium value of these weights. Since O1 sb is a fraction of total savings in the form of type (t, b) CIP, Ei;jh H E; (hb 9 = l- ( (9) ± 7 xe~..xr.1 i-l h=O b=O must hold. In addition, the market for type (it b) CIP must clear at each date. The demand for such cip is the fraction of savings invested in this form, Oie.b, times savings, w, divided by the price of type (ij, b) CP. In other words, the demand for such ap is given by Ojr-b wIpt'.%3 The supply of type (j*, b) cip is the amount of new capital investments in technology jl initiated h periods ago, less 2. If ali households were behaving identically, " = piik SMIw would hold. 3. Notice that6i'k w gives the value, in real terms, of the demand for type (j*,) aP. Division byr'-b converts this demand into units of ciP. 2S2 THE WORLD ANK ECONOMIC REVIEW. VOL 10, NO.2 the amount of CIP consumed by the transactions technology in the interim. Thus, the supply of type (j}, b) ap equals h-I h-I Oi'0 wfl (1(,,) P''0 =i 8'' rl (1 -a (=0 1=0 h-I since 1- H (1 - ao'1+l) of the initial CIp created has been lost in the transactions '=0 process. The market for type (j", hJ op clears, then, if Oj'.bW Pi'Jji',O -I -a.fl) (10) WI - PO'0w[ (1- t=0 We now observe that "-I (11) pi'J2 = pyr.o(prAh/pF '-)pr.-b-pr.h/-) (piJ/pI%O) y)b/J (1 oiV+I) r=O Substituting equation 11 into equation 10, we obtain (12) 0:t - I= ()hIO. Loosely speaking, equation 12 asserts that the demand for (and the supply of) cip of increasing maturity grows at the rate of interest. This is necessary in order for all vintages of op to bear a net-of-transactions-cost rate of return equal to the equilibrium interest rate. Equations 9 and 12 imply that t(13) oi'-° = (I - -y) J [1 - (lf)i*], In view of equation 8, equations 12 and 13 imply that the composition of sav- ings is determined entirely by the internal rate of return on investmnents in the equilibrium capital production technology. In particular, ecluation 13 describes how this rate of return determines the amount of new capital investment, and equation 12 governs how the remainder of agents' savings are allocated to the purchase of already-existing cIP in equity markets. The internal rate of return on savings depends on two factors: the marginal product of capital (a), and the net-of-transactions-cost productivity of the equi- librium investment technology (R.). We now investigate how changes in R.. influence capital accumulation. Let k denote the per capita capital stock in a steady-state equilibrium. Then (14) k =, 0''0w. Equation 14 simply notes that the steady-state equilibrium capital stock (per capita) equals the per capita initiation of new capital investments j* periods earlier (9i.° w), times the amount of capital produced, per unit invested, net of Bencivenga, Smith, and Starr 253 transactions costs (R1.). The linearity of final goods production implies that w = b. Equation 13 implies that VtO depends entirely on 'y, which in turn is the internal rate of return on technology j*. Finally, which technology maximizes the internal rate of return depends on the productivities of the capital invest- ment technologies, net of transactions costs. Thus, ultimately, the net-of- transactions-cost productivities of the various investment technologies, that is the values {Rk, 1 fully determine the steady-state capital stock. Later we will explore how the structure of transactions costs affects capital formation. Before doing so, however, we will characterize the equilibrium value of equity market activity. Equity Market Activity The real value of equity market transactions in each period-in per capita terms-is per capita saving less the real value of new capital investments ini- tiated. In particular, all savings-other than what goes into new capital invest- ments-is used to purchase existing ci in equity markets. Thus, the real value of net purchases in equity markets is given by w(1 - Oi8°). As before, this quan- tity is largely determined by the structure of transactions costs. [V. THE EFFECTS OF CHANGES IN TRANSACTIONS COSTS To investigate how changes in the level of transactions costs affect all aspects of a steady-state equilibrium, it is convenient to have a simple representation of transactions costs. To that end, we henceforth assume a constant proportional transactions cost structure: (15) i = a = (0,1); b 0,j. Consistent with our earlier discussion, we assume that there are no costs asso- ciated with capital rental or with the initiation of new capital investment. Thus we impose ai O = aiW = 0, for all j. Under these assumptions, (16) Rj = R( a - orr holds for all i. Our assumptions imply that a reduction in transactions costs has the largest proportional effect on A, for those investments with the longest maturities. The notion that transactions costs are most significant for long-maturity assets is certainly consistent with casual observation. For instance, The Wall Street Jour- nal of July 23, 1993, reported a bid/ask spread on three-month treasury bills of the previous day equal to 0.005 percent of price. For a thirty-year treasury bond, this spread was 0.062 percent of price, but for a thirty-year treasury strip (a pure discount instrument, equivalent to a long-term bill), the spread was 0.7 percent of price. Thus, transactions costs vary by a factor of 100 with maturity 254 THE WORLD RANK ECONOMIC REVIEW, VOL 1.E NO.2 alone, despite the fact that these observations ignore the obvious likelihood that a long-term instrument will be rolled over many more times during its lifetime than a short-term instrument. Given the specification of transactions costs in equation 15, technology j has a higher internal rate of return than technology j - 1 if and only if (17) 1-a Ž a(Rj_,)i1(Rj)'' is satisfied. Therefore, for a given set of technological parameters a and (Rl, R2, . . ., RI), the level of the transactions cost parameter a determines which tech- nology will be in use. Long-gestation technologies-which may intrinsically be highly productive-are also transactions intensive; that is, their ownership must be transferred many times. This will only be economical if transactions costs are sufficiently low. To put some structure on the parameters (R1, R2, . . ., Rj), we henceforth assume that (18) (Rji'+I1(RjI )i > (Rj_I)i1(Rj)1t- holds for all 1 > 2. The conditions imposed in expression 18 are essentially tech- nical: they guarantee that, for each capital production technology, some value of a exists at which that technology maximizes the internal rate of return to investment, net of transactions costs. We comment later on what happens if the assumption in 18 is relaxed. Expression 18 is compatible with a variety of different configurations of the values (RI, R2, . . ., RJ). For example, if J = 3, setting (RI, R2, R3) = (2, 1, 113) satisfies expression 18, as does setting (RI, R1, R3) = (1/3, 1, 2). The Dependence of Equilibrtium Values on Transactions Costs Here we consider how exogenous changes in the transactions cost parameter ax affect (a) the steady-stare capital stock, (b) the rate of return on savings, and (c) the volume of activirt in equity markets. In order to analyze these issues, we need to begin with the effect of a on the equilibrium maturity of capital invest- ments, j*. Suppose first that (19) 1 - a c a(R1)21R2 holds. Then expressions 17 and 18 imply thati* = 1; in other words, tr2nsactions costs are so high that technology 1-which requires no transactionr-raxixnizes the internal rate of return on investments, net of transactions costs. In this case, there is no equity market activity, and 91,0 = 1. The equilibrium per capita capital stock (see equation 14) is k = K1b, and the equilibrium return on savings is Rla. Bencivenga, Smitb, and Starr 255 Small reductions in a-and, more specifically, reductions that leave expres- sion 19 satisfied-do not change the equilibrium choice of capital production technology. Hence, there continues to be no equity market activity, and both the capital bcock and the real return on savings are unaffected. Suppose that a is now reduced enough so that expression 19 ceases to hold and, instead, (20) a(R2P/(R3)Y > 1 - a> a(R1)21R2 is satisfied. For this lower level of transactions costs, expressions 17 and 18 imply that j* = 2: transactions costs are now low enough for agents to view making longer-gestation investments that require some resale as economical. The equilibrium rate of return on savings is [aR2(I - a)]03; expression 20 guar- antees that this return exceeds aRl. Hence, the lower level of transactions costs has resulted in a higher equilibrium rate of return on savings. Once j* = 2 holds, it must be the case that 021 > 0 also obtains. In particu- lar, agen-ts must sell ownership claims in cIP that is not yet mature. As a consequence, (P.0 < 1. Thus, activity in equity markets diverts some saving away from capital formation. Indeed, the steady-state capital stock is now I = R2(1- a) 81'0b. If R2(1 - a) > RI holds, the reduction in transactions costs has two competing effects on the capital stock. First, the net-of-transactions- costs productivity of investment has risen [R2(1 - a) > RI holds], which is conducive to increased capital formation. However 02L0 c 01-0 = 1 also holds, so that some savings has been transferred from the initiation of new capital investment to the purchase of existing equity. Which effect dominates? The answer depends on the magnitude of al. Equation 13 implies that 0,O = 11 + [aR2(1 - a)105)-l Hence, R2(1 - a) 020 Ž R1 holds if and only if (21) R2(1 - a)I(l + [aR2(1 - a)]03s Ž RI. If equation 21 holds, transactions costs are low enough so that the use of tech- nology 2 actually leads to a higher steady-state capital stock than the exclusive use of technology 1. In this case, reductions in transactions costs (that is, more- efficient capital markets) lead to a higher long-run level of real activity. How- ever, if equation 21 fails--as is perfectly compatible with expression 20 being satisfied-then lower transactions costs divert enough saving away from capital formation into equity so that the long-run capital stock and the level of real activity are actually reduced. Thus to summarize, reductions in transactions costs that raise the equilibrium gestation period of capital investment must in- crease the real return on saving and the volume of equity market activity. How- 2S6 THE WORLD BANK ECONOMIC REVIEW, VOL. 10 NO.Z ever, the latter fact implies that long-run real activity may fall as a result of an increase in the efficiency of equity markets. Once transactions costs are low enough to satisfy expression 20, further reduc- tions in a that leave expression 20 satisfied have very unambiguous effects. The internal rate of return on investments (net of transactions costs) rises as transac- tions costs fall. The same is true for the value of equity market activity. It is also straightforward to show that R2(1 - a) 82,0 iS increased by such a reduction in transactions costs; as transactions costs fall, the steady-state capital stock and level of real activity necessarily increase. Even further reductions in a can cause expression 20 to be violated, and (22) a(R3)41(R4 )3> 1 - a > a(R)31(R3) to hold. Here expressions 17 and 18 imply that ji = 3. The effects of such a transition are exactly analogous to those of a transition from j" = 1 to ji = 2. The steady-state rate of return on savings must rise, as a reduction in transac- tions costs raises the internal rate of return (net of transactions costs) on all investment technologies. The volume of equity market activity also rises, or, in other words, 03.0 c 02-0 necessarily holds.4 The steady-state capital stock is given by k = R3(1 - a)2 03 0b. If R3(1 - a)2 > R2(1 - a), then again the implications for the capital stock are ambiguous. The reduction in transactions costs is itself conducive to capital formation, but, as before, savings are diverted away from initiation of new capital investment into the purchase of existing equity. For transactions cost levels that make technology 3 just marginally preferred to technology 2, the latter effect must dominate. The former effect may dominate for yet lower levels of transactions costs. As long as a satisfies expression 22, further reductions in a raise the internal rate of return on investrnent, increase the level of equity market activity, and raise R3(1 - a)2 03.0* Thus, the steady-state capital stock rises with reductions in the costs of financial transactions. As ax is reduced further, the equilibrium maturity of capital investments can continue to rise, and all of the factors we have noted previously will come into play. However, if 1 > a(Rj.4JI(Rj,t1 holds, low enough transactions costs imply that ji = J. Here further reductions in transactions costs cannot raise the gesta- tion period of investmnents, and therefore necessarily increase the capital stock. Although the increase in equity market activity means that less savings is being placed in new capital investment, this effect never dominates the increased effi- ciency of capital markets as long as j* is unaltered. When reductions in transactions costs increase i*, however, they result in discrete increases in the volume of equity market activity.5 Thus, even though the net-of-transactions-cost productivity of all capital investment technologies 4. See Bencivenga, Smith, and Starr (1996) for a formal proof of this assertion. S. Under the assumptions made here, i is nonincreasing in oL This resuk does not depend on any of the currentsimplifying assunptions. See Bencivenga, Smith, and Starr (1996) for a more general treatment. Bencivenga, Smitb, and Starr 2S7 rises, enough savings can be diverted away from new capital investment so that lower transactions costs are actually-over some rangedetrimental to capital formation. Admittedly, this result has been obtained by imposing a variety of assump- tions, for example the condition on the parameters (Rl, R2, . . ., RI) stated in expression 18. However, the same conclusions survive when many of our as- sumptions are generalized. For example, if expression 18 is relaxed for some capital production technologies, it will still be true that the equilibrium maturity length of capital investments is nonincreasing with transactions costs, and that reductions in transactions costs necessarily raise the rate of return on savings and increase the volume of equity market activity. The primary consequence of relaxing expression 18 is merely that, when expression 18 fails, some capital investment technologies could never be used in a steady-state equilibrium. In- deed, if expression 18 fails to hold for all capital production technologies, then either ji = 1 or j =J must hold. The former will obtr in for high enough transac- tions costs and the latter wiLl obtain for low enough transactions costs. Steady-State Welfare All young agents save their young-period wage, which, under our assump- tions, is independent of the capital stock. These agents then consume their inter- est income, so that steady-state welfare is proportional to the rate of return on savings. Transactions cost reductions always increase the internal rate of return on savings, net of transactions costs, and hence always raise this rate of return. Thus, transactions cost reductions necessarily increase steady-state welfare.' V. TRANSACrIONS Consm AS PURE RENTS In this section we analyze the same set of issues we did earlier under the assumption that transactions costs represent pure fees paid to a broker, a mar- ket maker, or the government. Thus, although the fees associated with transac- tions represent costs to equity market participants, these fees no longer repre- sent a social resource loss. For simplicity, we assume that the fees collected are simply rebated to old agents as a lump sum; we can think of this as corresponding to a situation where all old agents are given equal shares in a brokerage firm or where the government rebates the revenue it collects as a lump sum. The assumption that resources collected in the form of fees or taxes are re- bated to old agents prevents a transfer of the proceeds from those who bear these costs-by assumption, sellers (or old agents)-to those who do not-by assumption, buyers (or young agents). A transfer of resources from old to young agents would, under our assumptions, raise the aggregate savings rate and, in and of itself, constitute a stimulus to capital formation. Instead, we wish to isolate the effects of transactions costs alone. Therefore we rebate the proceeds 6. This asseron would not hold so generally if the wage rate depended on the capital stock. See Bendvenga, Smith, and Starr (1996). 2S8 TH4E WORLD BANK ECONOMIC REVIEW. VOL. 10. NO. 2 of fee or tax collections to those who pay the fees or taxes. Given our preference assumptions, the result is that savings patterns are unaltered by the existence of fees or taxes. If the transactions fees we analyze are interpreted as payments received by a broker or market-maker, we assume that none of the labor of these agents is diverted away from goods production when young. This reinforces the notion that there are (in this section) no social resource costs associated with the trans- actions process. If the fees paid are received by the government, then it is per- missible to think of them as negative, a situation corresponding to that in which the government subsidizes activity in capital markets. Many developing country governments do act to subsidize the formation of equity markets (Fry 1988); we analyze the consequences of this activity in this section. As we did in the previous section, we assume that there is a constant fee a c 1 levied on transactions. (We no longer impose a > 0.) We also assume-in consonance with our earlier formulation-that there are no fees associated with the initiation of new capital investments, or with transactions in factor markets. Thus, a10 = aiJ = 0, for all j. Steady-State Equilibrium Conditions The fact that transactions costs may not represent a social resource loss does not affect most of the steady-state equilibrium conditions. Indeed, these remain as above, with two exceptions. First, since transactions costs no longer consume capital, equation 14 must be replaced by (23) k = RK. O''0°b. Second, transactions costs no longer erode the supply of CI. Thus, equation 10 is supplanted by the condition (24) ihpjr- = U.; h = O,..., j-1. Note that equations 11 and 24 imply that (25) Oirb= VP0 [y(l -cx)]b; b = 0, 1, . ..,j -1. Equation 25 is equation 12, corrected for the fact that transactions costs now represent a transfer between agents, rather than a social resource loss. Equa- tions 9 and 25 now determine 61 , and equation 23 gives the steady-state equi- librium capital stock. As before, technology j bears a higher internal rate of return than technology - 1 if and only if equation 17 is satisfied. Thus, the level of transactions costs determines the equilibrium choice of investment technology, as well as the inter- nal rate of return on savings. From equations 9 and 25, the internal rate of return on savings determiines the fraction of savings going into the initiation of Ecncivenga, Smith, and St4rr 259 new capital investment (OP'°), which in turn determines the steady-state capital stock (k). Note that when transactions costs do not represent social resource costs, they affect the steady-state capital stock and output level only through their effects on the equilibrium maturity of capital investment (j*) and the port- folio weight (0i'.O) attached to new capital investment. We now wish to investigate how the steady-state capital stock, as well as other equilibrium quantities, vary with the transactions cost parameter ca. To simplify the exposition, we focus on the case 1 = 2. When there are only two technologies for producing capital, equation 17 implies that i* = 2 holds if and only if (26) 1 -a> a(R1)2/R2. Equation 26 describes how the equilibrium choke of capital production technol- ogy varies with a. When equation 26 fails,J* = 1 holds, as does el-0 = 1. Equation 23 implies that k = bRI. When equation 26 holds, ji = 2, and equations 9 and 25 imply that (27) 024] = (1 + [aR2I(1 - x)]0s1 In this case, equations 23 and 27 determine k. Evidently, for all values of a violating equation 26, the steady-state capital stock is independent of the level of transactions costs, there is no equity market activity, and the rate of return on savings is simply aRR. When a satisfies equa- tion 26, by contrast, j* = 2 holds. Now we can see that 820 c 1, so that equity markets are active, and that 02.0 is decreasing in aL Thus, when transactions costs are pure fees, a reduction in transactions costs raises the fraction of sav- ings invested in the initiation of new capital investments. As is apparent from equation 23, reductions in transactions costs-once equation 26 is satisfied- necessarily raise the steady-state capital stock. However, as (1 - a) transits from being just below a(R1)2/R2 to being just above it, the economy moves from a situation with 0190 = 1 to a situation with 0240 c 1. Hence, transitions in the equilibrium choice of investment technology will be associated with diversions of savings away from investment and, as be- fore, it is possible to show that if technology 2 is only slightly preferred to tech- nology 1, the steady-state capital stock will be smaller than it could be if tech- nology 1 were in use. Figure 1 depicts the steady-stare relationship between k and oL The steady-state equilibrium capital stock with j* = 2 is no less than that with j: = 1 if and only if (28) a < 1 - (aR21IR2) [R21(R2 - RI)]2 < 1 - (aR2,RJ). Expression 28 describes how low a must be in order for the use of technology 2 not to be associated with a reduction in the steady-state equilibrium capital stock. 260 TlE WORI.!) IIANKC iCONOMIC RIlVIitW, VOl.. 11. NO.2 Fzigure 1. The Steady-S/ate Relationship betwecen the Capital Stock and the Transactions COsts Parameter Capital stock, k \ k- hiR2 I + laR2itl-t)l"'5 bR1 -- bR,/11 +(RI/IR2)] 0 4' t 1 Transactons costs I - CaR 2IRZ) parmmeter, a a -1- (aR 2/R2)IR/(R2-Rf2 )I Note At low transactions costs, long-getation Investnent technologies are used. The lower the trsctons costs (forr - 2), the highier die capital stock. However, for transactions costs just below those that nmke tedinology 1 preferred to technology 2. the capital stock Is lower than It would be width P- 1. When a > 1- (aR) IR2),zf - 1. WVhen ac 1- (aR 2 /R2), r - 2. Here k is decreasing In a. Steady-State Welfare We want to know how the choice of c, which might reflect a conscious policy decision, affects the level of steady-state welfare. The analysis of this issue requires the consideration of two factors. First, since the real wage rate is just b, and since all young-period income is saved, one component of old-period consumption is simply yb. Moreover, since 'y = max {aR1,[aR2(1 - a)]0-5), the choice of a transactions fee can affect y. Second, the choice of a can affect the lump-sum transfer received by old agents. Recall that the real value of financial transactions per capita is given by b(1 - & °). In addition, all agents pay a fee of a per transaction, with transactions measured in real terms. Thus, the transfer received by an old agent in real terms is given by ab(1 - VA'), and steady-state welfare is then U = bmax faRl, [aR2(l - a)]°)5) + ab(1 - Oi'tO). if i* = 1, that is, if a > 1 - (aR21/R2), then no financial transactions occur, Oi"° = gl,0 = 1, and steady-state utility is just U1 = baRl. Altematively, if j = 2, some Bencivenga, Smith, and Starr 261 financial transactions occur (as described by equation 27), and transactions fees are paid. The net result is that steady-state welfare is described by the expression (29) U2(ax) b[aR2(1 - a)J0-5 + ab(l - 1/fl + [aR2/(1 - a)]°0j) = b(aR2j) 5 [t + [aR2(1 - a)]0°)/I(l - a)0-5 + (aR2)"3). It is straightforward but tedious to show that steady-state welfare is increasing in the cost of transacting-that is, U2 (a) > 0 holds-if and only if the internal rate of return on technology 2 is lower than the steadv-state rate of growth (that is, aR2 S 1 holds). When the latter condition is satisfied, higher transactions fees actually lead to higher steady-state welfare levels, so long as any secondary mar- ket transactions are undertaken. (Such transactions will be undertaken if and only if 1 - a > aRI/R2.) Such a result should be intuitive, since higher transac- tions costs in this case discourage low-return investments. In particular, then, if j = 2 and aR2 <1 hold, there is a socially excessive volume of equity market activity. It is therefore desirable to reduce the amount of this activity, as argued by Keynes and Tobin. However, there remains the issue of whether it is desir- able to drive the level of financial transactions to zero, which requires setting a > 1 - (aR1IR2). At levels of a this high, j* = 1, and there are no financial market transactions. We now describe when it is and is not desirable to eliminate secondary capital markets in the case aR2 < 1. We then show what happens when the internal rate of return on technology 2 exceeds the steady-state growth rate, which obtains when aR2 > 1. CAsE 1: aR2 < 1. In this case, U2(a) is maximized by setting a = 1 - (aR2IR2); this is the largest value of at consistent with j = 2. Evaluating equation 29 at this value of a yields (30) U2[1 - (aR2IR2)] = b(1 + aR1)/[1 + (R11R2)]. Then U2[l - (aR9R2)] Ž U1 = baR, holds if and only if (31) 1 > aR21JR2, or, that is, if and only it the value of a that maximizes steady-state welfare with j = 2 is positive. Thus, if aR2 1c and equation 31 holds, steady-state welfare is ma.cimized by setting a = 1- (aR21/R2) Ž 0, or, in other words, by setting fees as high as possible without eliminating secondary capital markets. In particular, when aR2 < 1 and equation 31 holds, there is a socially excessive volume of transactions because capital is overaccumulated, in the standard Diamond (1965) sense. It is then attractive from a policy perspective to reduce capital formation, which can be done by increasing transactions fees (see figure 1). However, tech- 262 THE WORLD BAN ECONOMIC REVIEW, VOL 10, NO.2 nology 2 still has a higher intenral rate of return than technology 1, and hence it should be used. Secondary capital markets are therefore required. When aR2 c 1 and equation 31 fails, the use of technology 2 is socially in- efficient in and of itself. Agents can be induced to use the more efficient technol- ogy by taxing equity market activity at a high enough rate and, indeed, one which eliminates this activity altogether. When aR2 < 1 holds, and if technology 2 is in use when cc = 0, our analysis indicates that there will be a socially excessive amount of financial market activ- ity. Under these conditions, the arguments made by Keynes and Tobin for tax- ing this activity would have merit. We might note that the context in which this situation arises is somewhat unconventional. The internal rate of return to physical investment is negative, but investrnent is still attractive because it represents the principal private means of saving. An implication of this observation is that there is, socially speaking, oversaving, which can be partially corrected by taxing financial transactions. C'ASE 2: aR2> 1. When aR2> 1 holds, U2(a) is decreasing in at, and is therefore maximized by setting a arbitrarily small. From equation 29, lim U2((a) = baR2. If R2> RI, then steady-state welfare maximization dictates setting j =2, and maxinally subsidizing equity market activity. In particular, technology 2 is more productive than technology 1, and incentives should be created to use it as in- tensively as possible. If, instead, RI > R2 holds, it is technology 1 that is socially most productive. In this event, transactions fees should be set high enough to discourage secondary market activity, alnng with the use of technology 2. That outcome can be attained, as in case 1, by setting a > 1 - (aR2j/R2). Note that the case for taxing secondary market activity need not rely on either technology's having a particularly low internal rate of returrn SUMMARY. As the examples just given indicate, it can be desirable to either subsidize or heavily tax agents transacting in equity markets. It will be optimal to confront agents undertaking such transactions with relatively heavy fees when aR2 c 1 or, in other words, when the internal rate of return on technology 2 (at a zero transactions cost level) is lower than the real growth rate of the economy. It will also be optimal to impose high fees in these markets when aR2 > 1 and R1> R2 both hold. Thus, even if the internal rate of return on technology 2 exceeds the rate of growth (- ith no transactions costs), it is undesirable to use technology 2 if it is less productive than techology 1. By contrast, when aR2 > 1 (the internal rate of return on technology 2 exceeds the growth rate) and R2 > R (technology 2 is more productive than technology 1), there is good reason to subsidize equity market activity. Bencivenga, Smith, and Starr 263 These observations suggest a criterion for determining when it is desirable to tax (or raise the costs faced by) equity market participants. A socially excessive volume of financial market transactions is undertaken in economies with rela- tively high levels of equity market activity (j* = 2, so that 02,0 < 1), and with real interest rates (gross of transactions costs) lower than the long-run real rate of growth of the economv (aR2 c 1). VI. SOME FINAL THOUGHTS How does the efficiency of an economy's capital resale or equity mar- kets-as measured by the costs of transacting in them-affect the economy's efficiency in producing physical capital and, through this channel, final goods and services? In order to propose an answer to this question, we have fol- lowed Hicks (1969) in emphasizing the role of equity markets in providing liquidity to holders of long-lived and inherently illiquid capital. As the effi- ciency of an economy's capital markets increases (that is, as transactions costs fall), the general effect is to cause agents to make longer-term, and hence more transactions-intensive, investments. The result is a higher rate of return on savings, as well as a change in its composition. These general equi- librium effects on the composition of savings cause agents to hold more of their wealth in the form of existing equity claims and to invest less in the initiation of new capital investments. As a result, a reduction in the resource losses suffered in the transactions process can cause the capital stock either to rise or to fall, and we have described conditions under which each situa- tion will obtain. However, a general point that bears emphasis is that a re- ducrion in transactions costs will typically alter the composition of savings and investment, and that any analysis of the consequences of such changes must take these effects into account. It has also often been proposed that a significant fraction of financial market transactions are socially unproductive, and that transactions taxes should be imposed to reduce the level of financial market activity. We have put forward a criterion for verifying when such taxes might be imposed, and we have indicated that to tax financial market transactions when long-gestation investments are relatively productive and when real interest rates are relatively high is not desirable. One topic that we have not addressed is the role for financial intermediation in the kind of environment we have considered. If intermediation does not allow transactions to be undertaken at a lower cost, then clearly it has no role here. However, if intermediaries can issue liabilities that are traded more cheaply than their underlying assets, then clearly there can be a role for banks. These banks would hold relatively illiquid, long-maturity assets, and would issue rela- tively liquid, short-maturity liabilities. This is obviously a natural function of banks. The integration of financial intermediaries into the analysis is an impor- tant topic for future investigation. 264 THE WORLD BANK ECONOMIC REVIEW. VOL 10, NO.2 REFERENCES The word "processed" describes informally reproduced works that may not be com- monly available through library systems. Antje, Raymond, and Boyan Jovanovic. 1992. 'Stock Markets and Development." Draft. New York University, Department of Economics, New York. Processed. Azariadis, Costas. 1992. Intertemporal Macroeconotics. London: Basil Blackwell. Bencivenga, Valerie R., and Bruce D. Smith. 1991. "Financial Intermediation and En- dogenous Growth." Review of Economic Studies 58(2, April):195-209. Bencivenga, Valerie R., Bruce D. Smith, and Ross M. Starr. 1994. "Liquidity of Secondary Capital Markets, Capital Accumulation, and the Term Structure of Asset Yields." Draft. Cornell University, Department of Economics, Ithaca, N.Y. Processed. . 1995. "Transactions Costs, Technological Choice and Endogenous Growth.' Journal of Economic Theory 67(1, October):153-77. . 1996. "Liquidity of Secondary Capital Markets: Allocative Efficiency and the Maturity Composition of the Capital Stock." Economic Theory 7(1):19-50. Bohm-Bawerk, E. von. 1891. Positive Theory of Capital. Translated by W. Smart. Lon- don: Macmillan. Boyd, John H., and Edward C. Prescott. 1986. 'Financial Intermediary Coalitions." Journal of Economic Theory 38(2):211-32. Cameron, Rondo. 1967. Banking in the Early Stages of Industrialization. New Yorlc Oxford University Press. Cooley, Thomas F., and Bruce D. Smith. 1992. 'Financial Markets, Specialization, and Learning by Doing." Draft. University of Rochester, Graduate School of Business, Rochester, N.Y. Processed. Diamond, Douglas W. 1984. 'Financial Intermediation and Delegated Monitoring.' Review of Economic Studies 51(3, July):393-414. Diamond, Peter. 1965. "National Debt in a Neoclassical Growth Model." American Economic Review 55(5):1126-50. Diaz-Alejandro, Carlos. 1985. "Good-Bye Financial Repression, Hello Financial Crash? Journal of Development Economics 19:1-24. Dickson, P. G. M. 1967. The Financial Revolution in England. London: Macmillan. Fry, Maxwell. 1988. Money, Interest, and Banking in Economic Development. Balti- more: Johns Hopkins University Press. Galbis, Vicente. 1979. "Inflation and interest Rate Policies in Latin America, 1967- 76.n IMF Staff Papers 26:334-66. Goldsmith, Raymond W. 1969. Financial Structure and Development. New Haven, Conn.: Yale University Press. Greenwood, Jeremy, and Boyan Jovanovic. 1990. "Financial Development, Growth, and the Distribution of Income."Journal of Political Economy 98(5, October):1076- 1107. Hicks, John. 1969. A Tbeory of Economic History. Oxford, U.K.: Clarendon Press. Keynes, John Maynard. 1964. The General Theory of Employment, Interest, and Money. New York: Harcourt, Brace and World. Bencidenga, S with, andStarr 2S Khatkhate, Deena R. 1988. "Assessing the Impact of Interest Rates in Less Developed Countries." World Development 16:577-88. King, Robert G., and Ross Levine. 1993a. "Finance, Entrepreneurship, and Growth: Theory and Evidence." journal of Monetary Economics 32(3, December):513-42. . 1993b. 'Finance and Growth: Schumpeter Might Be Right." QuarterlyJournal of Economics 108(3, August):717-38. Levine, Ross. 1991. "Stock Markets, Growth, and Tax Policy." Journal of Finance 46(4):1445-65. McKinnon, Ronald L 1973. Money and Capital in Economic Development. Washing- ton, D.C.: Brookings Institution. Shaw, Edward S. 1973. Financial Deepening in Economic Development. Oxford, U.K.: Oxford University Press. Tobin, James. 1994. "A Tax on International Currency Transactions." In United Na- tions, Human Development Report, 1994. New York: United Nations. van Wiinbergen, Sweder. 1985. "Macroeconomic Effects of Changes in Bank Interest Rates: Simulation Results for South Korea." Jornal of Development Economics 18:541-54. Williamson, Stephen D. 1986. "Costly Monitoring, Financial Intermediation, and Equi- librium Credit Rationing." Journal of Monetary Economics 18(2):159-79. THE WORLD BANK ECONOMIC REVIEW. VOL. 10. NO. 12 267-239 A Measure of Stock Market Integration for Developed and Emerging Markets Robert A. Koraiczyk A wide array of official capital controls across countries makes it difficult to per- form cross-sectional analysis of the effects of market segmentation. This artickle con- structs a measure of deviations from capital market integration that can be consis- tently applied across countries. It measures the deviations of asset returns from an equilibrium model of returns constructed assuming market integration. Applying the measure to stock returns from twenty-four national markets indicates that market segmentation tends to be much larger for emerging markets than for developed mar- kets, and that the measure tends to decrease over time. Along several dimensions, the proposed measure yields results that are consistent with reasonable priors about the relations between effective integration and explict capital controls, capital mar- ket development, and economic growth. In financially integrated markets, capital should flow across borders in order to ensure that the price of risk--the compensation investors receive for bearing risk-is equalized across assets. Conversely, if capital controls or other forces prevent free movement of capital across borders, then it is likely that different economies will demand different levels of compensation for risk. In some mar- kets, direct measures of the severity of capital controls are available. For ex- ample, some countries have dual classes of common equity. Restricted equity can be held only by domestic residents, but unrestricted equity can be held by both domestic and foreign investors. The price differential between restricted and unrestricted shares that have identical payoffs is a direct measure of the effects of capital controls (Hietala 1989; Bailey and Jagtiani 1994). Similarly, differences between official and black market exchange rates, between official and offshore interest rates, or between the market price and the net asset value of closed-end country mutual funds can be used to measure the effects of capital controls (Bonser-Neal and others 1990). A difficulty arises when attempting intercountry comparisons of the severity of capital controls because different countries may have different mechanisms Roberr A. Korajczyk is with the Department of Finance at Nortwestem University. This arricle was originally prepared for the World Bank conference on Stock Markets, Corporate Finance, and Economic Growth, held in Washington, D.C., February 16-17, 1995. The author would like to thank Eric Chang, Ash Demirgfii-Kunt, Ross Levine, Peter Montiel, participants at the confercnce, and two anonymous referees for helpful comments. 0 1996 The International Bank for Reconstruction and Development /THE WORLD BANK 267 268 THE WORLD BANK ECONOMIC REVIEW, VOL. 10, NO.2 for restricting capital movements. For example, a country that prohibits all for- eign investment does not have unrestricted shares whose prices can be com- pared to restricted shares. In addition, countries without any formal restrictions against foreign investment will not have restricted shares trading. Although the former case is ostensibly one of segmented markets and the latter case is one of integrated markets, there may be methods by which investors circumvent the restrictions in the former case, and there may be informal barriers that lead to actual segmentation in the latter case (such as less stringent accounting stan- dards or insider trading regulations). Given the difficulty of directly comparing the effects of the wide array of official capital controls across countries, a measure of deviations from capital market integration that can be consistently applied across countries is impor- tant for cross-sectional analyses of the effects of market segmentation. The ap- proach taken here is to measure deviations from integration by measuring the deviations of asset returns from an equilibrium model of returns constructed assuming market integration. Testing the law of one price (Lop) in financial markets requires a model that identifies the type of risk that is important to investors. The model used here is the International Arbitrage Pricing Theory (IAPT). An advantage of an approach that relies on asset prices or returns is that effective barriers to capital flows, regardless of their source, should lead to actual deviations from LOP. Statutory barriers to capital flows that are ineffective should not lead to pricing deviations. Ostensibly free markets with large nonstatutory barriers (such as large differentials in information costs) should exhibit pricing de- viations. A disadvantage of the L&T approach is that it relies on a particular specifica- tion of the asset pricing model. If the asset pricing model is incorrect, then pric- ing errors will be observed even when markets are integrated. Also, regime shifts, such as those that would occur when an economy moves from being segmented to integrated, will lead to changes in the asset pricing relation and to large short- term measured deviations from LOP. The next section contains a brief description of the asset pricing model. Sec- tion II relates pricing errors to the existence of deviations from the law of one price induced by market segmentation. Section m discusses estimation of pric- ing errors. Section IV addresses the effects of regime shifts. Section V describes the data. The techniques used to construct factor-mimicking portfolios are de- scribed in section VI. The empirical measures of deviations from the law of one price are described in section VII. Section VIII presents conclusions and sugges- tions for future work. I. THE MULTIFACTOR ASSET PRICING MODEL The logic behind the Arbitrage Pricing Theory (Arr; Ross 1976) and interna- tional extensions (Ross and Walsh 1983; Solnik 1983; Levine 1989; and Clyman, Koraiczyk 269 Edelson, and Hiller 1991) is that there are a small number of risks that are common to most assets, for which investors command risk premiums. Risk that is specific to one asset (or a small set of assets) is diversifiable and, therefore, investors do not demand compensation for this risk. The Case without Diversiflable Risk The arbitrage argument can be most easily illustrated in the case where there is no diversifiable, or idiosyncratic, risk. Assume that the realized returns on securities are given by the following linear factor model: (1) rit = 14,t + bi,1,1,t + - - . + bj,kt where rjJ denotes realized returns on asset j at time t, bj, is the sensitivity of asset j to the ith common source of risk, 8, is the realization of risk factor i in period t, and p,= E4vvr)j is the expected return on asset j. In this case where there is no asset- specific risk, there could be a riskless, costless arbitrage opporunity unless:1 (2) Iti, = Ao + b,.lXLi + . . . + bjk*k where ?,t is the return on a riskless asset and Xj, is the risk premium on the ith source of risk. More generally, expected returns could be expressed as (3) p4 = a- + ot + b11,X .t + . - - + b,kA*.t where oc1 represents the pricing error, or deviation of expected returns from the predictions of the multifactor asset pricing model. In this case, ac must equal zero for all j so that no arbitrage opportunities are possible. Let p' = (pi,,,. p ... p,4j, at= (a1, a2,..., c;), X' = (QL, Al, .. ., 4), and B = (1, b) where t is an n-vector of ones and b is an n x k matrix whose (;, z) element is bjji. In matrix notation, equation 3 can be expressed as: I= a + BA. The value of I that mininizes the pricing error (in terms of minimizing the sum of squared pricing errors) is A = (BBt'B'p and a = [I - B(' B)-' Blp, where I is an n x n identity matrix. Note that c'B = 0, so that a portfolio formed by choos- ing the portfolio weight on asset i to be a, is costless (since c'i = 0) and riskless (since a'b = 0, which implies that the portfolio has no exposure to the risk fac- tors). The expected return on the portfolio is aL =aa +a'B =a'a + 0 > O. 1. This requires the assumptions that there are more assets than sources of risk (In > k) and that the n x k matrix of sensitivities, b-where the Uf) element of b is X,,-has rank k. 270 THE WORLD tANK ECONOMIC REVIEW, VOL 10, NO. Thus, this portfolio is riskless and costless and has a strictly positive return. This is an arbitrage opportunity that will be exploited. In order to avoid arbitrage opportunities, the pricing relation given by equation 2 must hold. That is, a1 = 0 for all f in equation 3. The Case with Diversifuable Risk The expression for asset returns in equation 1 assumes that there are only k worldwide factors that influence all asset returns. To generalize this specifica- tion to include uncertainty that is asset specific, or diversifiable, returns will be expressed as: (4) r +., = + b,11,4 + . . * + bi.kSht + where yi,, is the uncertainty in asset j's returns that is not explained by the worldwide factors. Ross (1976) assumes that there are an infinite number of assets and that the asset-specific risks are uncorrelated across assets, that is, corr(;7, ,,.,,) = 0 for ji m. Ross notes that weaker conditions also imply that the risk embodied in the term s,, is diversifiable (Chamberlain and Rothschild 1983; Connor and Koraiczyk 1993). Because each asset has its own unique, or asset-specific, risk, it will not be possible to form riskless portfolios from a finite set of risky assets. However, an asymptotic arbitrage opportunity can be defined as one in which it is possible to construct a sequence of portfolios whose expected returns approach infinity and whose variance approaches zero as the number of assets, n, approaches infinity. The absence of such arbitrage opportunities implies that the sum of squared pricing deviations (oa + aj + ... + oa) must remain finite as n approaches infinity (Ross 1976; Huberman 1982). The fact that the sum of squared pricing deviations must remain finite implies (in an economy with an infinite number of assets) that most of the pricing errors must be small and that equation 2 holds as an approximation for most assets: (5) =t- ,t + bj A1,, + . . * + bi+k)kt- Further restrictions can be placed on the economy to get the pricing model to hold as an equality (Connor 1984; Constantinides 1989). I will assume that, under the null hypothesis of financial market integration, either such restric- tions hold or the approximation is good enough to ignore the approximation error in equation S. II. MARKEr SEGMENTAnON AND PRICING ERRoRs Although the method of estimating the risk factors is described more fully later, it is useful at this juncture to point out that capital market segmentntion prevents cross-market arbitrage and, therefore, prevents the prices of risk (vec- Korajczyk 271 tor .) from being equated across markets. Capital market segmentation will lead to pricing errors relative to risk factors constructed assuming capital market integration. This is illustrated by a hypothetical world consisting of two markets (a and b) that are influenced by the same single-world factor. That is, assets in each economy satisfy a one-factor pricing model. However, because the markets are segmented, the parameters of tie asset pricing model are different across markets. The expected returns on asset j in the two markets are given by 117t=24 + bt IM ILi,t = 24 + bl with 2424 and A, * A4b. However, the implied riskless return and world factor risk premium estimated by pooling the two markets together and assuming (in- correctly) that they are integrated will be (assuming the markets are of equiva- lent size and have the same distribution of sensitivities) X0 = (M + Xt)f2 and XI = (Ak: + A24) 12. That is, estimating an integrated model when the null is incor- rect will lead to estimated risk premiums that are weighted averages of the true segmented risk premiums. This implies that for economy a, the measured pric- ing deviation (relative to a model estimated assuming integration) of asset j is (6) a7 = (ao - o) + b, I( - X1) and for economy b, the measured pricing deviation of asset j is (7) -= (2b -5 ) + 1 - j1). Thus, the mispricing parameters, a, provide a direct measure of deviations from the LOP. Although the example assumes a single common factor, the results extend to any number of common factors, under the assumption that the LAPT would be the appropriate pricing model in an integrated world. That is, different prices of risk for common factors will lead to nonzero alphas relative to the IAL. Under certain conditions, the pricing errors relative to a factor model might actually miss market segmentation. This could occur if there are local factors in one or more countries that are priced locally (because of segmentation) but not priced in other countries. In addition, market segmentation could be missed if (1) the local factors are unrelated to asset returns in the other countries (that is, the factors are not common in that the sensitivities of nonlocal assets to the local factors are zero) and (2) the local factors are included in the asset pricing model. To illustrate this situation, consider a hypothetical world consisting of two markets (a and b) that are influenced by the same single-world factor. Coun- try b's assets are sensitive to a local factor that is priced in b but not in a. That is, assets in each economy satisfy the factor pricing models 272 T4E WORLD BANK ECONOMIC REVIEW. VOL. 10 NO.2 1<= + bl, ww K MI = + 0 JA where W and L stand for world and local factors. Here I have assumed that the two economies have the same price of risk for the world factor but different prices for the local factor. Applying a two-factor model to these economies us- ing the world and local factors will not reveal a pricing deviation even though the nonzero price of the local factor in country b is due to market segmentation. The pricing deviation will be undetected because assets in economy a have zero sensitivity to the local factor. Failure to reject integration in this case hinges on including the local factor in the model. If only the world factor, and not the local factor, is included in the model, then the pricing errors for country b's assets will be nonzero and should lead to a rejection of the market integration hypothesis. An alternative approach to testing the law of one price is to estimate the price of risk for different subsets of securities and test the hypothesis that the price of risk is equal across subsets. This is done withini a single country in Roll and Ross (1980), who test for the equality of the zero-beta return, X(, across subsets, and in Brown and Weinstein (1983), who test for the equality of all risk premiums. In an international setting, subsetting by country allows an estimate of country- specific prices of risk. This is done by Cho, Eun, and Senbet (1986); Gultekin, Gultekin, and Penati (1989); and Harvey (1991). III. ESTIMATION OF PRICING ERRoRs The pricing deviations discussed in sections I and II were expressed as dis- crepancies between an asset's true expected return and the expected return im- plied by the asset pricing model. However, the ex post return on the asset is observed, not the true expected returns on the asset. From equation 4, the asset's ex post return deviates from its expected return because of shocks from the common factors and asset-specific shocks. Let T be the number of time periods for observed asset returns; n the number of securities; rK the n x T matrix of excess returns on the assets; F the k x T matrix of realized factors plus risk premiums Fit = k + A,,0, where Fi, is the excess return on the portfolio that mimics factor i in period t; M the n x k matrix of sensitivities, or factor loadings; and £7 the n x k matrix of idiosyncratic (asset specific) returns. Equations 2 and 4 imply that (8) =b +e with E(Fs) = 0, E(s') = 0, and E(erl) = VP. The assumption of a factor structure and the asset pricing theory (equations 2 and 4) imply that there is a restriction on a multivariate regression of asset Koraick 273 returns on a constant and the excess returns on factor-mimicking portfolios, which is embodied in equation 8. The restriction is that the intercepts are jointly equal to zero. That is, in the multivariate regression (9) r' = a + b'F + e the vector of intercept terms, or, contains the pricing deviations. If markets are integrated and the multifactor asset pricing model describes asset expected re- turns, cx should be equal to zero. However, if risks are priced differently across economies, these pricing differences will lead to nonzero values of aW. Thus, one measure of financial integration is the size of the intercept terms in the multi- variate regression (equation 9). [V. AssEr PiuacNG DYAMIcs AND REGIME SHiFrs The theoretical pricing errors in equations 6 and 7 are derived assuming that each economy is in a steady-state segmented equilibrium each period. However, the recent trend in most markets is movement from segmented markets toward integrated markets. This trend implies that the asset pricing regimes will shift from segmented to integrated regimes and that the param- eters in equation 9 are likely to change through time. In the long run, in- creasing integration should lead to smaller pricing errors (zero pricing errors in the limit approaching complete integration). However, in the short run, measured pricing errors might be larger as asset prices change because of the changes in asset pricing regimes. Since the movement from a completely seg- mented market to a completely integrated market is rarely smooth, the asset pricing dynamics during the transition phase are difficult to characterize. In particular, if market participants anticipate the liberalization from a seg- mented to integrated market, asset expected returns in the transition period are not likely to be set according to models that assume complete segmenta- tion or complete integration. The appendix contains a simple numerical example of an unanticipated regime shift. Although the numerical results are clearly dependent on the numbers picked for the example, the fact still remains that shifts across pric- ing regimes are likely to cause changes in the parameters (a and b) and cause large measures of mispricing in the short run. Also, these shifts in pricing should occur as shifts in regimes become anticipated. Thus, the effects of regime shifts could be spread over a longer period as the market updates its assessment of changes in capital controls. The unconditional approach in equation 9 ignores shifts in the parameters and the transition dynamics. Clearly, if the nature of pricing errors was constant through time, the full time-series sample should be used to estimate a. Given the importance of regime shifts, I investigate the behavior of estimated mispricing over a se- quence of different time periods. This is an admittedly crude method of ac- 274 THE WORLD BANK ECONOMIC REVIEW, VOL. 10, NO.2 commodating nonstationarities such as regime shifts. However, it is a first step toward measuring levels of integration. Most empirical studies of market integration assume that the level of seg- mentation is constant over the estimation period. A notable exception is Bekaert and Harvey (1995). They propose a regime switching model in which markets stochastically move between segmented and integrated regimes. In addition, the probability of a regime shift is allowed to change as a function of predetermined instrumental variables. Within a regime, the model assumes that assets are priced as though agents expect to be in that regime forever. Although this does not admit an investigation of the pricing effects of anticipated regime shifts, it takes seriously the fact that regime shifts happen and that the transition probabilities vary through time. V. DATA SOURCES AND SUMMARY STATISTICS Historical monthly data on equity returns for individual stocks trading in twenty emerging markets are from the Emerging Markets Data Base provided by the International Finance Corporation (wc). The econornies covered by the data base are Argentina, Brazil, Chile, Colombia, Greece, India, Indonesia, Jor- dan, the Republic of Korea, Malaysia, Mexico, Nigeria, Pakistan, the Philip- pines, Portugal, Taiwan (China), Thailand, Turkey, Venezuela, and Zimbabwe. The set of emerging markets is geographically diverse as well as diverse in the severity of capital controls. The sample of developed equity markets includes stocks from Australia, Ja- pan, the United Kingdom, and the United States. Equity data sources for the developed markets are the Centre for Research in Finance at the Australian Graduate School of Management for Australia, the Japan Securities Research Institute for Japan, the London Share Price Data Base from the London Business School for the United Kingdom, and the Center for Research in Security Prices at the University of Chicago for the United States. The sample includes all assets traded on the Australian Stock Exchange, the New York and American Stock Exchanges, the first section or the Tokyo Stock Exchange, the London Stock Exchange, and the unlisted securities market in the United Kingdom. Monthly returns, adjusted for dividends and stock splits, are transformed into U.S. dollar returns using end-of-month exchange rates. The emerging mar- kets exchange rates are from the IFC's Emerging Markets Data Base, and the developed markets exchange rates are from MF (various issues). To compute excess returns, I use the U.S. Treasury Bill returns from Ibbotson Associates (1993). Data on equity returns and shares outstanding from the rFC's Emerging Mar- kets Data Base were screened for unusual values. Confirmations or corrections of these values were obtained from the iFc. When there is more than one ex- change rate system in place within a country, the IFc attempts to obtain a free market rate either from newspapers or from IFc correspondents in each market Korajczyk 27S (see IFC 1993). There are a few instances, earlier in the sample, when the ex- change rates seem too stable to be free market rates, given that black market rates seem to be varying over the same period. This does not appear to be an issue later in the sample. Construction of the Emerging Markets Data Base began in 1981. Firms were chosen at that time on the basis of 1980 data (Errunza and Losq 1985: 562). Although such a choice poses no particular problems for returns after 1980, there may be a survivorship bias induced for the returns before 1981. That is, firms that disappeared between 1975 and 1980, for example, would not be in- cluded in the data base. An actual portfolio strategy might have included those assets in the sample. As shown in table 1, eleven of the twenty emerging markets have data prior to 1981. Errunza and Losq (1985) investigate the issue of survivorship bias in a sample of eight of these emerging markets. They apply the selection criteria to assets as of the beginning of the sample, December 1975. They find that the overlap between this sarnple and the actual sample in the data base is between 53 and 85 percent. Seven companies that would have been in- cluded when applying the selection criteria in 1975 were not trading on the ex- changes in 1980. Errunza and Losq argue that the survivorship bias is small in the sample. Although the reported statistics about the overlap of the samples and delistings are suggestive, the extent of the survivorship bias is difficult to esti- mate without recreating each market's sample with a nonanticipatory inclusion rule over the period from 1975 to 1980. Even if this were done, there is another potential survivorship bias in that tlhe initial set of emerging markets was chosen on the basis of information available in 1980. There may have been markets that would have been included in 1975 that performed poorly between 197S and 1980 (that is, failed to emerge) and were thus not included in the sample. Random errors in the individual stock data will clearly induce noise into the estimated mispricing parameters. More systematic errors-such as errors in the exchange rate, which influence the calculated return on all assets in a given economy, or survival biases-will tend to cause false rejection of market inte- gration by inducing nonzero pricing errors in the factor model. Tables 1 and 2 provide some summary statistics on the emerging markets in the sample. The iFc's Emerging Markets Data Base does not include all of the stocks traded in the emerging markets. Rather, the data base consists of a sample of stocks from each market. The stocks are chosen on the basis of trading activ- ity, capitalization, and diversity across market sectors (see IFC 1993). On aver- age, the stocks in the IFc sample represent S7 percent of the total capitalization of the respective markets. As illustrated in table 1, the average number of firms ranges from 11 (Zimnba- bwe) to 66 (Indonesia). The capitalization (as of December 1992) of the stocks included in the data base ranges from $268 million (Zimbabwe) to $66 billion (Mexico and Korea).2 Average monthly turnover (volume for month t divided by 2. One billion is cqual to 1,000 million. 276 THE WORLD BANK LCONOMIC REVIEW. VOL 10, NO.Z Table 1. Summary Statistics for Emerging Markets Capitalization Trading volume as of as of Average Begitning Average December1992 December1992 turnover of sample number of (millions of (millions of (monthly Market period, firms U.S. dollars) U.S. dollars) percentage) Emerging market composite" 1984 590 401,998 16,S35 8.79 Europe and Middle East Greece 1975 1S 5,377 112 1.05 Jordan 1978 15 1,988 70 1.18 Portugal 198S 20 4,868 52 1.07 Turkey 1975 19 3,872 158 2.54 Latin America Latin America composite" 1984 176 135,638 3,936 3.24 Argentina 1975 23 14,293 1,112 3.41 Brazil 1975 33 23,200 803 3.40 Chile 1975 25 21,933 96 0.76 Colombia 1984 21 5,107 23 OA9 Mexico 1975 32 66,108 1,806 5.31 Venezuela 1984 14 4,997 96 2.00 Asia Asia compositeb 1984 325 249,191 12,204 10.99 India 1975 36 25,365 364 6.42 Indonesia 1989 66 8,661 260 3.85 Korea, Rep. of 197a 37 66,461 6,007 8.20 Malaysia 1984 52 47,941 773 1.11 Pakistn 1984 52 3,774 33 0.86 Philippines 1984 22 8,167 84 2.13 Taiwan (China) 1984 SO 60,454 3,172 23.66 Thailand 1975 17 28,368 1,877 5.38 Africa Nigeria 1984 18 797 1 0.0S Zimbabwe 1975 11 268 1 0.35 a. The sample begins at the end of December of the year indicated. b. Composites are value-weighted portfolios formed from the national market indices. Source: tFc's Emerginig Markets Data Base. the capitalization of the market in month t - 1) for the sample period is lowest for Nigeria at O.OS percent and highest for Taiwan (China) at 23.66 percent. Table 2 reports some statistics for the return distributions of the IFC emerg- ing market indexes. The average monthly rate of return in U.S. dollars is lowest for Indonesia (-1.02 percent) and highest for Argentina (5.66 per- cent). The variability of the index returns (standard deviation in table 2) is also quite high. Jordan has the smallest monthly standard deviation, 5.17 Korajezyk 277 Table 2. Monthly Return and Pricing Error Statistics for Emerging Markets Monthly return (percent) Standard Autocorrelationt Average prjcing Market Mean deviation coefficient, p,O errors, ab Emerging market composites 1.50 6.98 0.16 (1.53) Europe antd Middle East Greece 0.62 10.46 0.13 0.85 (1.89) Jordan 0.90 5.17 0.00 2.42 (0.00) Portugal 2.88 14.50 0.29 -10.43 (2.61) Turkey 3.15 21.44 0.23 -4.43 (1.97) Latin America Latin America compositec 2.60 11.21 0.24 (2AO) Argentina 5.66 30.00 0.05 -2.73 (0.77) Brazil 1.84 17.39 0.03 -31.29 (0.41) Chile 3.06 11.42 0.17 48.99 (2.41) Colombia 3.64 9.28 0.49 11.01 (4.79) Mexico 2.53 12.86 0.25 10.97 (3.53) Venezuela 2.68 13.66 0.27 2.43 (2.62) Asia Asia compositec 1.50 7.42 0.01 (0.13) India 1.68 7.86 0.08 -2.05 (1.13) Indonesia -1.02 9A0 0.28 -0.01 (1.71) Korea, Rep. of 1.77 9.34 0.00 3.15 (-0.02) Malaysia 1.15 7.61 0.05 2.50 (0.51) Pakistan 1.79 6.70 0.25 0.22 (2.45) Philippines 3.78 11.02 024 6.62 (3.32) Taiwan (China) 2.84 15.27 0.07 1.84 (0.72) Thailand 1.86 7.44 0.11 5.33 (1.63) Africa Nigeria 0.22 10.54 0.08 _A (0.83) Zimbabwe 0.65 9.86 0.14 21.70 (1.97) a. First-order autocorrelation coefficient of market index returns. t-statistics are in parentheses. b. rne-series average of bias-adjusted average squared pricing errors. c. Composites are value-weighted portfolios formed from the national market indices. Source: IFC's Emerging Markets Data Base. 278 THE WORLD SANK ECONOMIC REVIEW. VOL. 10, NO. 2 percent, while Argentina has the largest monthly standard deviation, 30 per- cent. By contrast, the S&P 500 portfolio has a monthly standard deviation of 4.46 percent for the period from January 1976 to December 1992 (Ibbotson Associates 1993). VI. CONSTRUCrnON OF FACTOR-pMIMICKING PoRTFouos In order to estimate the level of mispricing, " in equation 9, we need to have the matrix F. F contains the time series of excess returns on k portfolios whose innovations are perfectly correlated with the k sources of factor risk. In prac- tice, these excess return portfolios need to be estimated. To estimate the excess returns on the factor-mimicking portfolios, I use the asymptotic principal components technique of Connor and Korajczyk (1986, 1988). The asymptotic principal components procedure can easily accommodate the large number of stocks in the sample. The procedure as- sumes that the factor structure is as in equation 4; that the exact multifactor pricing relationship, equation 2, holds; that the conditional factor loadings, Cj/ are constant through time for most assets; and that the cross-sectional average asset-specific variance is constant through time. Let £Qn be the Tx T matrix defined by Qn = r"'r"In and FP the k x T matrix of the first k eigenvec- tors of "Qn (where PF is an estimate of F, and where r" and F are as defined in equation 8). Under the assumption that asset returns follow a k-factor model as in equation 4, Connor and Koraiczyk (1986) show that F" converges in probability to a nonsingular linear transformation of F as n goes to infinity. Because the sample of equity returns is large, I ignore the estimation error in FI% In order to use all available data in the sample, I employ an extension of the principal components technique from Connor and Koraiczyk (1988), which does not require that asset returns exhibit continuous time series of returns. This method is designed to avoid a common source of survivorship bias. Although these types of factor portfolios do not fully explain the pric- ing of international equities, they perform well relative to common alterna- tive models (Koraiczyk and Viallet 1989). I use the returns on all stocks from the twenty-four national stock markets to estimate the factor-mimicking portfolios. For an average month in the period from January 1976 to December 1992, 6,851 firms from the twenty-four mar- kets have available returns. An alternative approach to implementing international multiple factor mod- els is to specify, ex ante, the identity of the factors. This is the approach taken in Harvey (1995a, 1995b). The factors used in those papers include the return on the proxy for the world market portfolio, the return on a portfolio of currencies, and proxies for changes in commodity and agricultural prices (Harvey 1995a) plus proxies for oil price movements, inflation, and world business cycles (Harvey 1995b). Harvey (1995a) investigates conditional as well as unconditional factor models. Koraiczyk 279 VII. EMPIRICAL MEASURES OF DEVATIONS FROM THE LAW OF ONE PRICE As discussed in section III, the estimated lArT pricing errors (that is, the inter- cepts, at, in a regression of asset returns (in excess of a riskless asset) on the excess returns of a factor-mimicking portfolio) are a measure of segmentation. The estimation treats the pricing errors as constant over the sample period even though there have been significant liberalizations of capital controls in many economies. Therefore, I estimate pricing errors over a sequence of time periods and attempt to characterize the time-series behavior of the mispricing param- eters. Although it is somewhat schizophrenic to assume ac is fixed for estimation but then to look at the time series of a over different sample periods, the exer- cise should provide some information about the behavior of pricing errors (see Foster and Nelson 1991 for an analysis of such "rolling regressions"). My approach is as follows: 1. Estimate factor-mimicking portfolios using the asymptotic principal components procedure. The asymptotic principal components procedure uses data on all of the equities traded in the twenty-four markets studied here to estimate factor-mimicking portfolios for the entire sample period. 2. For each national market, estimate equation 9 for all stocks individually over rolling eighteen-month sample periods. This estimation will yield vectors of mispricing estimates, &q, where the time subscript denotes the sample period ove-r which the parameters are estimated. 3. Calculate a summary measure of the mispricing for each national market. In focusing on deviations (both positive and negative) of a from zero, a natu- ral measure of mispricing across the assets is the average squared mispricing coefficient, etwal/n. However, the regressions provide only an estimate of a", &", not the true value. The average squared values of the estimates, 6P' c/n, will converge to ara'In plus the average squared value of tile estimation error. Thus, eflrln will yield an upwardly biased estimate of a"' cr/n. However, the bias for asset i, E&a2 - acf], has an expected value equal to the variance of the intercept coefficient. Let vi denote the estimated variance of the regression intercept for asset i and let the n-vector of these variances for n assets be v". Given v", an adjusted average squared pricing error can be calculated as 6 = &"Zcln - 't/n, where t is an n-vector of ones. The quantity 6 will be called the average adjusted mispricing for the n assets. In the empirical analysis, I use estimates, v, which are corrected for conditional heteroskedasticity, as in White (1980). Under the null hypothesis that a" = 0, the expected value of A is zero. Thus, if capital markets are integrated and share the same set of pervasive risks, the average adjusted mispricing should be close to zero. This measure of mispricing should tend to be larger the more severe the barriers to free capital flows. In addition, the periods of transition from segmented to integrated markets should be associated with large average adjusted mispricing, as asset prices adjust to a different equilibrium level. 280 THE WORLD BAN ECONOMIC REVIEW, VOL 10, NO.2 Rather than emphasize formal statistical tests, I wish to characterize the cross- sectional and time-series characteristics of the estimated mispricing and relate the behavior of the measures to changes in capital controls in the various mar- kets. This characterization of the empirical properties of the rnispricing, or mar- ket segmentation, measures should provide some sense of the forces causing the measured deviations from LOP. Average adjusted mispricing is estimated for each of the twenty-four national markets. Since the severity of capital controls is likely to vary through time, I estimate a time series of O's, rather than estimate each economy's adjusted mispricing for the entire sample period. The time se- ries of O's is constructed by estimating A for each (overlapping) eighteen-month period in the sample. That is, data from January 1976 to June 1977 are used to estimate06w7= =&a &%/n - 4,77iIn, data from February 1976 to July 1977 are used to estimate 07,, and so on. The final period is July 1991 through Decem- ber 1992. All firms are included in the sample as long as they have at least fifteen monthly observations in the subperiod. The time-series averages of the values of It are reported in table 2. The aver- age adjusted mispricing, 0, is plotted for each national market in figure 1. Each graph also plots 0, for the United States as a reference point. The values of B, for Australia, Japan, and the United Kingdom are generally small. The largest de- viations from the value of zero occur for Australia, with values around -20 that occur around the 1987 stock market crash. Argentina begins in the late 1970s with very high values of Ot (around 300), that decline rapidly. There is a sharp rise in A, in 1986, followed by a sharp decline. The period from 1986 through 1987 coincides with increased invest- ment by foreign institutional investors. Beginning in the autumn of 1989, there is a period in which A, takes on large negative values. This latter period coincides with the beginning of a series of economic reforms in Argentina.3 The reforms include the State Reform Law (in September 1989), which announced-among other things-various privatizations, and the New Foreign Investnent Regime (in November 1989), which essentially opened the Argentine capital markets to foreign investors by eliminating restrictions on both foreign ownership (except in selected sectors) and the repatriation of capital. The values of A, for Brazil are particularly large in the period from 1985 through 1989. The largest deviations occur in 1986. This corresponds to a period in which the government announced the Cruzado Plan, which insti- tuted strict price controls on goods, wages, and official exchange rates. There is a short-lived boom in the stock market in which the IFC index of stocks doubles (in U.S. dollar terms) in the span of two months (from February to April). The boom is followed by a decline in the IFC index of stocks to their February levels by the end of the year. Large negative values of adjusted mispricing seem to accompany liberalizations in March 1987 (approval of 3. Sources of information on economic and political developments as wel as extant capital controls in emerging markets are Chuppe and Adkin (1992), IFC (1993), Park and Van Agrael (1993), Levine and Zervos (1994), and Bekaert (1995). Koraiezyk 281 Figure 1. AverageAdjustedMispricing Selected Markets, June 1977 to December 1992 Argentina Australia Average adjusted nispricing Avcrp adjusted mispdcing 350 -~~~~~~~~- 250 - 10- to 200 - 5 -20 1977798B1 83 8587 8991 197779 81 83 85 87 89 91 Brazil Chile Average 3duset niisprilg Average djusted mispricing 150 - 1L000 100 50 8005 0 6100V -50 400 -100 -20 -150- 200 1977 79 81 83 85 87 89 91 1977 79 81 83 85 87 89 91 Coombia Gree Averge adjusted mispricing Average adjusted misprcing 120- 1_00 . 100 1 80 60 -0 - 400 - 25020- -250I -3D0 - _ __ _ _ __ _ __ _ _ __ _ _ 1977 79 81838597 89 91 1977 79 8183 858775991 Colomia Cndreeia Average adjsted iniispricing Average adjusted nilprkcing -80 -10s 20- CoEy Uie Sie Ru oiUuso bebOLgae 282 THE WORLD BANK ECONOMIC REV, VOL. 10, NO.2 Figure 1. (continued) Japan Jordan Average adjusted mispricing Average adjusted mispalcing 35- 6 30 - 4- 25 - 2 - k A20- 1 20 r . -Y y - y 1 0 . -4 - V N5 0AR ., -6-I -8 -10 - 60 -15 1977 79 81 83 85 87 89 91 1977 79 81 83 85 87 89 91 Republic of Korea Malaysia Average adjusted mispricing Average adjusted mispriclng loo - ~~~~~~~~~~120 _lo -100 6077 79 81 3 8 8 89S1197 798183 5 7 8 980 60 - 40 - ~~~~~~~~~~40- 20 -20 IN 40 -40- -40- - __ __60__ _-60- 1977 79 8183 8587 8991 1977 79 8183 85 787991 Mexico Nigeria Average adjusted mispriding Average adjusted mispricing 200 -80 150o 60- 100. 40- 50 -20- -0- MA0 * V ~~~~~~~-20. -50 - ~~~~~~~~-40- -100 -60- -150-1 -80- 1977 79 81 838587 89 91 1977 79 8183 858789 91 Pakistan Philippines Average Adusted mispricing Average adjusted mnisprlclag 30 - ~~~~~~~~~350- 25j 3m -~~~~~~~~5 200- 15 - ~~~~~~~~~150- 10 100- 5 -50 -150 1977 79 8183 8587 8991 1977 79 81 83858978991 - Councry - Urnited States Koraiczyk 283 Portugal Taiwan (China) Averge adjusted mispricing Aver adjusted misp1iOng 50 - 140- a. ~~~~~~~~~120- 0 - ~~~~~~~~~~~100- -50- o -1000 7 40O -150 0 -200~~~~~~~~~~4 -250 -60 1977 79 81 85 85 87 89 91 1977 79 81 83 85 87 89 91 Thailand Turkey Average adjusted mispriing Avenge adjusted mispridng 120 - 150 100 - ~~~~~~~~~~100 so - 55° - i 80~~~~~~~~~0 60 -250 40 --0 20~~~~~~~~~~~-0 _ aO 3~~~~~~~~~~~~~50 1977 79 81 83 85 87 89 91 1977 79 81 83 85 87 89 91 United Kingdom Venezuela Average adjusced mispricing Avenge adusted mispricing 10 - 150 6 100 - 2-8 -1 vov A.wut3 lo r-7 I7 79 81 - 83 85W 87 B99A977 18 5S 99 -2 -6 ~~~~~~~~~~~-100 -lo ~~~~~~~~~~~-150 1977 79 81838587 89 91 1977 79 81 83858978991 Zimnbabwe Averge adjusted mispricmng 100- G~~~~~~~~~~~~~~ountry 60° W Ill 1lt - UnitedSates 4o0 -20- 1977 79 81 83 85 87 89 91 Source ICs Emerging Mlrkets Data Bar. 284 THE WORLD BANK ECONOMIC REVIEW. VOL. 10, NO.2 foreign investment trusts) and 1990 (when an interbank foreign exchange market is allowed). Chile has extremely large values of adjusted mispricing in 1977-78. There are smaller (but still large) values of mispricing in 1981 and 1987. Colombia shows a steady decline in %, for the periods ending March 1986 through March 1988. After that, 9, stays relatively close to zero until late 1990. The large values of G, in late 1990 and early 1991 may be associated with a move toward making the market 100 percent investable in February 1991. The values of B, are relatively close to zero for Greece until the mid-1980s. There is a large increase in mispricing in 1990. After several years of a socialist govemment (from 1981 to 1989) and a year in which two elections failed to produce a clear winning party, the conservative party was elected to power in April 1990. There is a 59 percent return to holding the portfolio of stocks in the IFc index in April 1990, followed by a 44 percent return in June 1990. Prices subsequently decline in late 1990. The average adjusted mispricing values for India are generally small except from 1985 to 1987 and in 1991-92. This later period includes a balance of payments crisis in mid-1991, followed by a series of reforms phasing in full convertibility of the rupee. Restrictions on institutional investment in Indian equities were loosened in 1992. In April 1992 it was disclosed that a number of banks were illegally investing funds in the Indian equity market The disclosure led to a sharp decline (approximately 40 percent) in the equity market. The time-series sample for Indonesia is rather short. Because of this, it is d:f cult to detect particular patterns in the average adjusted mnispricing. .:rdan exhibits some of the smallest absolute levels of average adjusted mispricing among the emerging markets. It also exhibits the lowest volatility and one of the lowest mean returns among the emerging markets (table 2). The largest values of adjusted mispricing occur in 1991-92. The Korean stock market exhibits relatively small values of adjusted mispricing except in the late 1970s and the mid-1980s, in spite of the fact that there were severe restrictions on foreign investment in Korean equities. In 1981 the first of a series of funds was offered through which foreign investors could invest in Korean securities (see Chuppe and Atkin 1992). From 1985 through 1987 addi- tional liberalization occurred. Additional Korean mutual funds were offered to international investors. In 1985, companies on the Korean stock exchange were granted authorization to raise capital in international bond markets and, as a result, gained access to equity capital through convertible bond issues. An over- the-counter market for unlisted stocks was opened in 1987. A government fund to stabilize stock prices was created in 1989 after the 1989 crash. The Malaysian stock market shows very large levels of mispricing in 1986 and early 1987. The period through late 1986 involved extensive liberalization of restrictions on capital inflows. The large values of adjusted mispricing might be due to large capital inflows at that time, although it is difficult to infer much from the short time series. Korajczyk 285 For the Mexican stock market, the average adjusted mispricing is relatively large until 1989. From 1989 through 1992 the average pricing errors are rela- tively low. Before 1989, restricted shares-which could be owned by foreign- ers-were typically restricted to below 50 percent of a firm's equity capital. In 1989, foreigners were allowed to hold up to 100 percent of a firm's equity in most industrial sectors. A trust fund was also established in 1989 to allow for- eign investors to buy (through the trust) previously restricted shares. The Nigerian stock market was essentially closed to foreign investment throughout the sample period. The average adjusted mispricing is large and vola- tile in the mid-198Os. The value of 0, declines to approximately 0 in the late 1980s with a jump to the 7-10 range in 1991. Pakistan has relatively small average mispricing throughout the sample pe- riod. There is a small jump in mispricing in 1991, which coincides with the lifting of restrictions on foreign investment. The Philippines shows large values of average mispricing from 1986 to 1989. This may reflect the price effects of inflows of capital following the ouster of President Marcos. After 1989 the average mispricing is generally small. The Portuguese stock market shows large values (first positive, then negative) of average mispricing from 1987 through 1989. This may have been caused by pricing effects of the Portuguese entry into the European Union (and the associ- ated elimination of barriers to foreign investments) followed by the October 1987 crash. After 1989 the estimated average adjusted mispricing is relatively small. The Taiwan (China) stock market shows generally large levels of estimated mispricing with no discernible trend, even though the period is one in which barriers to foreign investments were generally being lifted. Indirect investrnent was allowed through investment trust funds in 1982, with direct investment by foreign institutions following in 1991 (with a temporary halt in 1992). Average mispricing on the Thailand stock market is generally small in the 1980s. Larger average mispricing occurs in the late 1970s and early 1990s. The Istanbul Stock Exchange opened in 1986. The short time series of mispricing for Turkey does not show any pronounced trend, except the initial increase from very negative values. The Zimbabwe stock market shows generally high levels of adjusted mispricing. This is consistent with the fact that the market was closed to foreign investment throughout the sample period. VI. CONCLUSIONS AND SUGGESTIONS FOR FuTuRE WORK In this article I suggest a measure of the deviations from the law of one price across potentially segmented capital markets. This measure is applied to stock returns from twenty-four national markets (four developed markets and twenty emerging markets). The measure of market segmentation tends to be much larger for emerging markets than for the developed markets, a result consistent with 286 THE 'WORLD SANK ECONOMIC REVIEW. VOL 10, NO, 2 larger barriers to capital flows into or out of the emerging markets. The mea- sure often tends to decrease through time, a result that is consistent with grow- ing levels of integration. Large values of adjusted mispricing also occur around periods of economic turbulence and periods in which capital controls change significantly. Thus, the adjusted mispricing estimates measure not only the level of deviations from the LOP but also the revaluations inherent in moving from one regime to another. Relating the proposed measure of market integration to alternative measures of integration (as in Bekaert 1995), to measures of capital market development, or to ex post measures of economic growth would be useful for highlighting the advantages and disadvantages of this measure. Bekaert and Harvey (1995, figure 2) plot the estimated probability of being in the integrated regime. There are some interesting similarities and differences in the conclusions that could be drawn from their measure of integration and the adjusted mispricing plotted in figure 1. For example, Bekaert and Harvey (1995) show dramatic declines in the probability of India's stock market's being integrated in 1985 and 1992, corresponding to the periods in which there are large values of the adjusted mispricing parameter for India in figure 1. An ex- ample of a case in which the measures of integration seem to differ is Mexico. Bekaert and Harvey's estimate of the probability of integration is quite low in the post-1989 period. This is the period in which the adjusted mispricing esti- mate is the closest to zero for Mexico (figure 1). Thus, these alternative mea- sures of market integration seem to be highlighting different aspects of the mecha- nism generating expected returns. Demirgiiu-Kunt and Levine (1996) investigate the cross-sectional relation between mispricing and other indicators of capital market development. They find that mispricing (without the bias adjustment) is significantly negatively cor- related with the size (market capitalization) and trading volume of the respec- tive markets and is significantly positively related to market volatility and con- centration. Levine and Zervos (1994, 1995) find that the mispricing measure proposed here is negatively correlated with economic growth and that the levels of adjusted mispricing decline after liberalization of restrictions on capital flows. Thus, along several dimensions, the proposed measure of integration yields re- sults that are consistent with reasonable priors about the relations between ef- fective integration and explicit capital controls, capital market development, and economic growth. APPENDix. UNANTICIPATED AssEr PRICING REGIME Smrrs To illustrate the short-term effects of regime shifts, I will consider the some- what artificial but tractable example of a market that changes unexpectedly from being completely segmented from world markets to being completely inte- grated. Assume that under complete segmentation the economy's assets are priced by a domestic representative consumer with time-additive utility, Koraiczyk 287 ut= pU(Ct+) 5s-i where p reflects the consumer's rate of time preference, c,., is the consumer's consumption in period t + s, and u(-) is the per period utility of consumption. The pricing of assets in this multiperiod, multifactor world depends crucially on the comovements of the risk factors with the marginal utility of consumption (see Connor and Koraiczyk 1989 for details). Consider the following special case: the covariances between the representative consumer's marginal utility of consumption and the risk factors are constant Et[Si.z+1'(ct,jIu(cjJ = y,, and as- sets are expected to pay one unit of consumption each period, but their actual payoff depends on the risk factors. Then asset j will have a price equal to pilt = I _ p(I + bj,lyl + * + +bikYk)- To make the example concrete, assume that for the closed-economy (seg- mented market) case there are two risk factors (a world factor and a domestic factor) that are correlated with the marginal utility of consumption; the domes- tic representative investor has a time-preference parameter of 0.98 (p = 0.98); and the covarances between the representative consumer's marginal utility of consumption and the two risk factors-E,[81,t+5u'(c+s)Iu'(cj)] and E,[S2z+5u'(c,+jI u(c,)]-are -0.10 and -0.20, respectively. Asset j will have a price equal to: j,t = (99 [1 + b,j (-0.1) + b.2 (-0.20)]1 Thus, if asset j has b,j = 1.0 and bp = 0.5, then Pi,, = 39.20. Now assume that the market is opened to global investors and asset prices are determined by the preferences of a globally diversified representative consumer. The new param- eters are p = 0.98, EJt[,u'[c,.J)ut(c,)] = -0.10, and Et[,u,'(c2,5)Iu'(c4)] = 0.0. For example, the covariance between the domestic factor and the global repre- sentative investor's marginal utility might be zero because the small economy's domestic factor risk is diversifiable across economies. The unexpected shift from a segmented to an integrated economy leads to a change in price from $39.20 to $44.10, an immediate return of 12.5 percent. If the parameter p were to simul- taneously change from 0.98 to 0.99, the price of asset i would jump to $89.10, an immediate return of 127 percent. REFERENCES The word 'processed' escribes informally reproduced works that may not be com- monly available through library systems. 288 THE WORLD DANK ECONOMIC REVIFW, VOL. 10, NO.1 Bailey, Warren, and Julapa Jagtiani. 1994. "Foreign Ownership Restrictions and Stock Prices in the Thai Capital Market."Journal of Financial Economics 36(1, August):57- 87. Bekaert, Geert. 1995. "Market Integration and Investment Barriers in Emerging Equity Markets." The World Bank Economic Review 9(1, January):75-107. Bekaert, Gecrt, and Campbell R. Harvey. 1995. "Time-Varying World Market Integra- tion." Journal of Finance 50(2, Junej:403-44. Bonser-Neal, Catherine, Greggory Brauer, Robert Neal, and Simon Wheatley. 1990. "International Investment Restrictions and Closed-End Country Fund Prices." Jour- nal of Finance 45(2, June):523-47. Brown, Stephen J., and Mark I. Weinstein. 1983. 'A New Approach to Testing Asset Pricing Models: The Bilinear Paradigm." Journal of Finance 38(3, June):711-43. Chamberlain, Gary, and Michael Rothschild. 1983. "Arbitrage and Mean Variance Analysis on Large Asset Markets.' Econometrica 51(S, September):1281-1304. Cho, D. Chinhyung, Cheol S. Eun, and Lemma W. Senbet. 1986. "International Arbitrage Pricing Theory: An Empirical Investigation." Journal of Finance 41(2, June):313- 29. Chuppe, Terry M., and Michael Atkin. 1992. 'Regulation of Securities Markets: Some Recent Trends and Their Implications for Emerging Markets." Policy Research Working Paper 829. International Finance Corporation, Economics Department, Washington, D.C. Processed. Clyman, Dana R., Michael E. Edelson, and Randall S. Hiller. 1991. "International Arbitrage Pricing, Risk Premiums and Exchange Rate Drift." Working Paper 92- 019. Harvard University, Harvard Business School, Cambridge, Mass. Processed. Connor, Gregory. 1984. "A Unified Beta Pricing Theory." Journal of Economic Theory 34(1, October):13-31. Connor, Gregory, and Robert A. Korajczyk. 1986. 'Performance Measurement with the Arbitrage Pricing Theory: A New Framework for Analysis." Journal of Financial Economics 1S(March):373-94. =-. 1988. 'Estimating Pervasive Economic Factors with Missing Observations." Working Paper 34. Northwestern University, Department of Finance, Evanston, Ill. Processed. . 1989. "An Intertemporal Equilibrium Beta Pricing Model." Review of Finan- cial Studies 2(3):373-92. - -. 1993. "A Test for the Number of Factors in an Approximate Factor Model." journal of Fintnace 48(4, September):1263-91. Constantinides, George M. 1989. "Theory of Valuation: Overview and Recent Devel- opments." In S. Bhattacharya and George M. Constantinides, eds., Theory of Valu- ation: Frontiers of Modern Financial Thteory. Totowa NJ.: Rowman & Littlefield. Demirguii-Kunt, Ash, and Ross Levine. 1996. "Stock Market Development and Finan- cial Intermediaries: Stylized Facts." The World Bank Economic Review 10(2):291- 321. Errunza, Vihang R., and Etienne Losq. 1985. "The Behavior of Stock Prices on LDC Markets.' Journal of Banking and Finance 9:561-75. Foster, Dean P., and Daniel B. Nelson. 1991. "RoIling Regressions." Working Paper. University of Chicago, Graduate School of Business, Chicago. Processed. Koraiczyk 289 Gultekin, Mustafa N., N. Bulent Gultekin, and Alessandro Penati. 1989. 'Capital Con- trols and International Capital Market Segmentation: The Evidence from the Japa- nese and American Stock Markets." Journal of Finance 44(4, September):849-69. Haney, Campbell R. 1991. 'The World Price of Covariance Risk." Journal of Finance 46(1, March):111-57. %. 1995a. "Predictable Risk and Returns in Emerging Markets." Review of Finan- cial Studies 8(3, Fall):775-8 1i. -. 1995b. 'Risk Exposure of Emerging Equity Markets." The World Bank Eco- nomic Review 9(1, January):19-50. Hietala, Pckka T. 1989. "Asset Pricing in Partially Segmented Markets: Evidence from the Finnish Market." Journal of Finance 44(3, July):697-718. Huberman, Gur. 1982. "A Simple Approach to Arbitrage Pricing." Journal of Eco- nomic Theory 28(1, October):183-91. Ibbotson Associates. 1993. Stocks, Bonds, Bills, and Inflation. 1993 Yearbook. Chi- cago. IFC (International Finance Corporation). 1993. IFC Index Metbodology. Washington, D.C. imF (International Monetary Fund). Various issues. International Financial Statistics. WVashington, D.C. Korajczyk, Robert A., and Claude J. Viallet. 1989. 'An Empirical Investigation of In- ternational Asset Pricing." Review of Financial Studies 2(4):553-85. Levine, Ross. 1989. "An International Arbitrage Pricing Model with Prp Deviations." Economic Inquiry 27(0ctober):587-99. Levine, Ross, and Sara Zervos. 1994. "International Capital Flow Liberalization and Stock Market Development: A Cross-Country Event Study." Working Paper. World Bank, Policy Research Department, Washington, D.C. Processed. . 1995. "Stock Markets and Banks: Revving the Engines of Growth." Working Paper. World Bank, Policy Research Department, Washington, D.C. Processed. Park, Keith K. H., and Antoine W. Van Agtrnael, eds. 1993. The World's Emerging Stock Markets: Structure, Developments, Regulations and Opportunities. Chicago: Probus Publishing. Roll, Richard, and Stephen A. Ross. 1980. "An Empirical Investigation of the Arbitrage Pricing Theory." Journal of Finance 3S(5, December):1073-1103. Ross, Stephen A. 1976. "The Arbitrage Theory of Capital Asset Pricing." Journal of Economic Theory 13:341-60. Ross, Stephen A., and Michael M. Walsh. 1983. 'A Simple Approach to the Pricing of Risky Assets with Uncertin Exchange Rates." Research in International Business and Finance 3:39-54. Solnik, Bruno. 1983. "International Arbitrage Pricing Theory."Journal of Finance 38(2, May):449-57. White, Halbert. 1980. "A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity." Econometrica 48(3, May):817-38. THI! WORLD BANK ECONOMIC REVIlW, VOl. 1o, NO. 2i 291-321 Stock Market Development and Financial Intermediaries: Stylized Facts Asti Demirgiiq-Kunt and Ross Levine World stock markets are booming, and eni erging stock ntarkets account for a dis- proportionate share of this growth. Yet economists lack a common concept or mea- sure of stock market development. This article collects and compares a broad array of indicators of stock market and financial intermediary development, using data from forty-four developing and industrial countries during the period from 1986 to 1993. The empirical results exhibit wide cross-country differences for each indicator as well as intuitively appealing correlations between various indicators. The article constructs aggregate indexes and analyzes them to docunment the relationship be- tween the emergence of stock markets and the growth of financial intermediaries. It produces a set of stylized facts that facilitates and stimulates research into tihe links amorg stock markets, economic development, and corporate financing decisions. The growth and globalization of emerging stock markets are impressive. In 1994, emerging market capitalization was $1.9 trillion, compared to $0.2 trillion in 1985. Similarly, $39 billion flowed into emerging equity markets from abroad in 1994, compared with $0.1 billion in 1985.1 These developments have at- tracted the attention of academics, practitioners, and policymakers. Several stud- ies focus on measuring the benefits of holding a globally diversified portfolio (for example, see Harvey 1995 and De Santis 1993); and many countries are reforming regulations and laws to foster capital market development and at- tract foreign portfolio flows. Yet, economists have neither a common concept nor a common measure of stock market development. This article gives empirical content to the phrase "stock market development" by collecting and comparing a broader array of empirical indicators of stock market development than any previous study. Using data on forty-four develop- ing and industrial countries from 1986 to 1993, we examine different measures of stock market size, market liquidity, market concentration, market volatility, institutional development, and integration with world capital markets. Since each indicator suffers from statistical and conceptual shortcomings, we use a variety of indicators, which provide a more accurate depiction of stock markets 1. One billion is 1,000 million; one trillion is 1,000 billion. Ash Demirgiu-Kunt and Ross Levine are with the Policy Research Department at the World Bank. This aricle was originally prepared for the World Bank conference on Stock Markers, Corporate Finance, and Economic Growth, held in Washington, D.C, February 16-17,1995. 0 1996 The International Bank for Reconstruction and Development /IHm woRw &wK 291 292 TI-ll' WORLD IIIANK CONOMIC RrvILW. VOL. 10. NO. 2 than any single measuire. Furthermore, stock market development-like the level of economic development-is a complex and multifaceted concept. No single measure will capture all aspects of stock market development. Thus, our goal is to produce a set of stylized facts about various indicators of stock market devel- opment that facilitates and stimulates research into the links among stock mar- kets, economic development, and corporate financing decisions. After describing each of tne stock market development indicators, we exam- ine the relationships among them. We find enormous cross-country variation in the stock market indicators. For example, five countries have market capitaliza- tion to gross domestic product (GDP) ratiios greater than 1, and five countries have market capitalization to GDP ratios less than 0.10. We find attractive cor- relations among the indicators. For example, large stock markets are more liq- uid, less volatile, and more internationally integrated than smaller markets; coun- tries with strong information disclosure laws, internationally accepted accounting standards, and unrestricted international capital flows tend to have larger and more liquid markets; countries with markets concentrated in a few stocks tend to have smaller, less liquid, and less internationally integrated markets; and in- ternationally integrated markets are less volatile. Although many stock market development indicators are significantly corre- lated in an intuitively plausible fashion, the individual indicators produce differ- ent country rankings. Thus, to produce an assessment of the overall level of stock market development across countries, we produce indexes of stock mar- ket development that average together the information contained in the indi- vidual indicators. Developing aggregate indexes that summarize the extent of a country's stock market development in a single figure is especially helpful for analysts who are interested in making comparisons across countries. These in- dexes can be used in empirical studies linking stock market development and other economic phenomena,. as in Levine and Zervos (1996) and Demirgiiu- Kunt and Maksimovic (1996). We find that from 1986 to 1993 the most devel- oped stock markets in the world are in Japan, the United States, and the United Kingdom, and the most underdeveloped marklets are in Colombia, Venezuela, Nigeria, and Zimbabwe. The data suggest that HIong Kong, Singapore, the Re- public of Korea, Switzerland, and Malaysia have highly developed stock mar- kets; Turkey, Greece, Argentina, and Pakistan have underdeveloped markets. Furthernore, although richer countries generally have more developed stock markets than poorer countries, many markets labeled emerging are more devel- oped than those in France, the Netherlands, Australia, Canada, Sweden, and Norway. We use the assortment of stock market indicators to cvaluate which stock markets have been developing fastest over the last eight years. Using measures of size, liquidity, and international integration, Indonesia, Turkey, Portugal, and Venezuela stand out as the most rapidly developing markets in the world. This article documents the relationship between the various stock market indicators and measures of financial intermediary development. Since debt and Demirgi4-Kunt and Levine 293 equity are frequently viewed as alternative sources of corporate finance, stock markets and banks are sometimes viewed as alternative vehicles for financing corporate investments (see Demirgucr-Kunt and Maksimovic 1996). Conse- quently, we document the cross-country ties between stock market develop- ment and financial intermediary development. We use measures of the size of the banking system, the amount of credit going to private firms, the size of nonbank financial corporations, and the size of private insurance and pension companies. We find that most stock market indicators are highly correlated with the development and efficient functioning of banks, nonbank financial cor- porations, and private insurance companies and pension funds. Countries with well-developed stock markets tend to have well-developed financial intermedi- aries. Section I presents indicators of stock market ci2velopment and describes their theoretical relevance. Section II ranks countries using the different indicators of stock market development and studies the correlations among the indicators. Section III examines which countries have the fastest-developing stock markets. Section IV analyzes the links between stock market development and financial intermediary development. Section V summarizes the results. I. INDICATORS OF STOCK MARKET DEVELOPMENT A growing theoretical literature examines the relationship between particular attributes of stock markets and both economic growth and firms' financing de- cisions. For example, Devereux and Smith (1994) and Obstreld (1994) show that by facilitating risk sharing, internationally integrated stock markets affect saving decisions, the allocation of capital, and long-run economic growth rates. Greater risk diversification and liquidity have theoretically ambiguous effects on saving rates, however, because saving rates could fall sufficiently for en- hanced liquidity and risk diversification to lead to slower economic growth. Levine (1991) and Bencivenga, Smith, and Starr (1996) emphasize that stock market liquidity-the ability to easily trade securities-facilitates investments in longer-run, higher-return projects that involve more transactions. On stock market size, Pagano (1993) studies the increased risk-sharing benefits of larger stock markets due to thick market externalities. Besides stock market size, li- quidity, and integration with world capital markets, theorists have examined stock return volatility. For example, DeLong and others (1989) argue that ex- cess volatility in the stock market can hinder investment, and therefore growth, although there is considerable disagreement over the existence of excess volatil- ity in stock returns (see Shilier 1981). In terms of corporate finance, some theo- ries link stock market functioning with firms' financing and investment deci- sions. Pagano (1993) models the ties between risk-diversification and corporate financing decisions, while Boyd and Smith (1996) analyze complementarities between debt and equity financing for capital investments. Yet, as Demirgiui- Kunt and Maksimovic (1996) discuss, the effect of stock market development 294 THE WORLD BANK ECONOMIC REVIEW, VOL. 10, NO.2 on firms' financing decisions is thworetically inconclusive. Thus, theory provides a rich array of channels through which stock markets-market size, liquidity, integration with world capital markets, and volatility-may be linked to eco- nomic growth and corporate financing decisions. There is very little empirical evidence on the links among stock markets, eco- nomic development, and firms' corporate financing decisions. To facilitate em- pirical research, this article collects and compares a broad array of stock market indicators motivated by the above theoretical studies and constructs aggregate indexes of overall stock market development. Demirgiiu-Kunt and Maksimovic (1996) and Levine and Zervos (1996) use these indexes to examine the empiri- cal relationship among stock market development, firms' financing decisions, and long-run economic growth. The rest of this section presents and discusses an array of stock market devel- opment indicators. We focus on indicators identified by existing theoretical stud- ies. We describe measures of market size, market liquidity, market volatility, market concentration, asset pricing efficiency, regulatory and institutional de- velopment, and conglomerate indexes that aggregate the information contained in the individual measures. For developing countries, we use data from the In- ernational Finance Corporation's (Fc's)Emerging Markets Data Base. For in- dustrial countries, data are from M4organ Stanley Capital International (MscI). We also use macroeconomic data from IMF (various issues). The data cover the period from 1986 to 1993 for up to forty-four developing and industrial coun- tries. The appendix provides details of data construction and discusses cross- country comparability issues. Stock Market Size The market capitalization ratio equals the value of listed shares divided by GDP. Analysts frequently use the ratio as a measure of stock market size. In the rest of the article, we refer to this measure as market capitalization. In terms of economic significance, the assumption behind market capitalization is that mar- ket size is positively correlated with the ability to mobilize capital and diversify risk. For example, Pagano (1993) motivates his theoretical model by observing the great variation in market capitalization and in the number of listed compa- nies in different economies. As indicated in table 1, South Africa, Hong Kong, Malaysia, Japan, and Singapore all had market capitalization ratios greater than 1 from 1986 to 1993, while Nigeria, Argentina, Indonesia, Colombia, and Tur- key all had market capitalization ratios of less than 0.10 during the same period. We include statistics on the number of listed companies as an additional mea- sure of market size. Although marginal differences in the number of listed com- panies are uninformative, extreme values can be useful. It is not very interesting that Australia averaged 1,184 listed companies and Canada averaged 1,118 listed companies during the period from 1986 to 1993. But the fewer than 70 listed companies for Finland and Zimbabwe suggest that these countries have very limited markets (table 1). Similarly, the fact that in Indonesia, Turkey, and Por- Demirgfi-Kunt and Levine 295 tugal the number of listed companies grew at over 20 percent a year from 1986 to 1993 suggests rapid stock market development (see table 3 in section IHI). Liquidity Although economists advance many theoretical definitions of liquidity, ana- lysts generally use the term to refer to the ability to easily buy and sell securities. Since liquidity allows investors to alter their portfolios quickly and cheaply, it makes investment less risky and facilitates longer-term, more profitable invest- ments. Liquidity is an important attribute of stock market development because theoretically liquid markets improve the allocation of capital and enhance pros- pects of long-tenn economic growth. A comprehensive measure of liquidity would quantify all the costs associated with trading, including the time costs and un- certainty of finding a counterpart and settling the trade. Because we want to compare liquidity across countries and because data are very limited, we simply use two measures of realized stock tradiag. Total value tradedlGDP equals total shares traded on the stock market ex- change divided by GDP. The total value traded ratio measures the organized trading of equities as a share of national output, and should therefore positively reflect liquidity on an economywide basis. Japan, Hong Kong, Malaysia, the United States, and the United Kingdom all had total value traded/GDP ratios above 0.40, while in Pakistan, Zimbabwe, Colombia, and Nigeria, the total value traded/GDP ratio was about 0.01 from 1986 to 1993. The total value tradedl GDP ratio complements the market capitalization ratio. Although market capi- talization may be large, there may be little trading. For example, South Africa and Chile had above-average market capitalization but below-average total value traded/GDP (table 1). Together, market capitalization and total value traded/GDP inform us about market size and liquidity. A second measure of liquidity is the turnover ratio. Turnover equals the value of total shares traded divided by market capitalization. High turnover is often used as an indicator of low transactions costs. Korea and Germany (largely reflecting massive trading around reunification) had turnover ratios above 0.90, while Nigeria, Zimbabwe, and South Africa had turnover ratios below O.OS. The turnover ratio complements market capitalization. A small but active mar- ket will have small market capitalization but high turnover. For example, Nor- way and India had below-average market capitalization but above-average turn- over (table 1). Alternatively, South Africa's market capitalization to GDP ratio was the highest in the world, but its turnover ratio was one of the smallest. Turnover also complements total value traded/GDP. Although total value traded/ GDP captures trading compared with the size of the economy, turnover measures trading relative to the size of the stock market. Put differently, a small, liquid market will have a high turnover ratio but a small total value traded/GDP ratio. For example, there was not much equity trading in Brazil relative to the size of its economy, but Brazil's turnover ratio was high, reflecting a small but active stock market. By contrast, Malaysia had the third-highest market capitalization Table 1. Indicators of Stock Market Developmett, 1986-93 (annual average) Market Total value Number of listed Market Institutional APT ICAPM capitalizationa traded/GDPb cortpaniesc Turnoverd Volatility' concenirationf developments pricing errorh pricing error" Economy Rank Value Rank Value Rank Valute Rank Valtue Rank Valise Rantk Valte Rank Valte Rank Value Rank Vatlee Argentina 40 0.06 34 0.02 26 187 19 0.34 37 0.34 25 0.64 10 1.16 14 4.98 24 11.58 Australia 10 0.54 12 0.17 5 1,184 21 0.31 11 0.04 13 4.94 12 4.14 Austria 35 0.10 22 0.07 39 90 5 0.69 14 0.05 Belgium 18 0.36 28 0.04 27 182 35 0.12 6 0.04 Brazil 34 0.11 26 0.05 9 579 11 0.48 36 0.25 7 0.26 4 1.54 24 7.26 23 6.92 Canada 13 0.48 13 0.15 6 1,118 20 0.31 S 0.04 8 0.27 Chile I1 0.52 30 0.04 21 225 37 0.08 25 0.06 19 O.SO 5 1.52 17 5.S6 13 4.25 Colombia 38 0.07 40 0.01 40 87 38 0.07 23 0.06 26 0.74 11 1.16 19 5.62 15 4.82 Denmark 19 0.28 23 0.07 17 267 23 0.24 Finland 26 0.19 27 0.04 42 62 30 0.21 13 0.05 France 20 0.27 18 0.09 8 641 16 0.35 15 0.05 6 0.26 Germany 22 0.24 8 0.35 11 551 1 1.47 10 0.04 15 0.41 Greece 32 0.12 37 0.02 34 126 34 0.13 31 0.10 17 0.47 18 0.77 16 5.29 19 S.23 Hong Kong 2 1.36 2 0.59 14 318 12 0.44 India 31 0.16 25 0.06 2 4,614 9 0.50 24 0.06 3 0.22 8 1.34 7 3.33 7 2.89 Indonesia 39 0.06 36 0.02 38 91 27 0.23 17 0.96 9 3.68 8 3.03 Ireland 21 0.06 Israel 25 0.21 15 0.11 15 312 3 0.72 18 0.06 Italy 29 0.16 29 0.04 19 227 24 0.24 20 0.06 Japan 4 1.08 1 0.62 3 2,027 8 0.54 12 0.04 2 0.19 1 2.39 4 2.26 Jordan 9 0.57 14 0.13 36 103 29 0.22 7 0.04 23 0.59 12 1.16 2 2.5S 1 2.0S Korea, Rep. of 15 0.40 6 0.37 10 576 2 0.93 30 0.08 9 0.28 3 1.55 10 3.73 9 3.18 Luxembourg 23 205 Malaysia 3 1.28 3 0.46 16 291 26 0.24 17 0.05 12 0.36 1 1.63 11 3.90 5 2.45 Mexico 24 0.22 19 0.09 25 193 7 O.S6 32 0.10 10 0.36 2 1.61 21 5.94 21 5.77 Netherlands 12 0.49 11 0.21 18 239 14 0.41 3 0.03 New Zealand 16 0.39 24 0.06 20 226 32 0.17 16 0.05 Nigeria 41 0.04 41 0.00 33 127 41 0.01 21 O.S1 20 0.64 8 3.66 11 3.72 Norway 27 0.19 17 0.09 35 126 10 0.48 27 0.07 Pakistan 33 0.11 38 0.01 12 487 36 0.08 1 0.03 5 0.25 13 1.09 3 2.59 2 2.15 Philippines 23 0.24 31 0.04 30 152 28 0.23 29 0.08 22 0.52 9 1.32 15 5.26 16 4.90 Portugal 30 0.16 32 0.03 29 162 31 0.20 4 0.03 14 0.41 6 1.37 12 4.02 20 5.28 Singapore S 1.04 7 0.35 31 147 18 0.34 South Africa 1 1.54 21 0.08 7 700 39 O.OS Spain 21 0.25 20 0.08 13 383 17 0.35 19 0.06 Sweden 14 0.46 16 0.10 32 133 25 0.24 22 0.06 Switzerland 7 0.77 9 0.31 28 176 15 0.39 8 0.04 20 0.50 Taiwan (China) 24 197 34 0.15 13 0.40 16 0.98 20 5.68 14 4.54 Thailand 17 0.36 10 0.22 22 210 4 0.70 26 0.07 11 0.36 7 1.36 6 3.12 10 3.18 Turkey 37 0.08 33 0.03 37 91 22 0.28 35 0.17 18 0.50 Id 1.06 22 6.38 22 6.66 United Kingdom 6 0.92 5 0.41 4 1,932 13 0.44 9 0.04 4 0.24 5 2.94 6 2.56 United States 8 0.64 4 0.41 1 7,087 6 0.65 2 0.03 1 0.14 4 2.71 3 2.24 Venezuela 36 0.10 35 0.02 41 82 33 0.15 33 0.13 24 0.63 IS 1.00 23 6.67 17 5.15 Zimbabwe 28 0.18 39 0.01 43 57 40 0.03 28 0.07 16 0.44 19 0.66 18 5.57 18 5.18 Average 0.41 0.15 627 0.36 0.08 0.40 1.19 4.49 4.34 Number of economies 41 41 43 41 37 26 20 24 24 Note: For each indicator, the stock market development of each economy is ranked from high to low. Thus, for market capitalization, total value tradedlGDP, number of listed companies, turnover, and institutional development, the ranking by value of the indicator is from high to low. For volatility, market concentration,mAnpricingerror, and ICAPNI pricing error, the ranking by value of the indicator is from low to high. a. Market capitalization is the value of stocks divided by GDlP. b. Total value traded/doP is total value of traded shares divided by GDP. c. Number of companies listed represents the number of shares listed on the exchange. d. Turnover is given by total value traded divided by rnarket capitalization. e. Volatility is the twelve-month rolling standard deviation estimate based on market re:urns. f. Market concentration is the share of market capitalization held by the ten largest stocks. g. Institutional development is an average of institutional indicators as described in the text. h. APT and ICAPWI pricing errors are obtained from Korajczyk (1994). Source: Authors' calculations and Koraiczyk (1994). 298 THE WORLD BANK ECONOMIC REVIEW. VOL. 10. NO.2 and total value tradedlGDP ratios from 1986 to 1993, but it had below-average turnover (table 1). Thus, incorporating information on market capitalization, total value traded/GDP, and turnover provides a more comprehensive picture of development than any single indicator can provide. Concentration In some countries a few companies dominate the market. High concentration is not desirable because it may adversely affect the liquidity of the market. To measure the degree of market concentration, we compute the share of market capitalization accounted for by the ten largest stocks and call this measure concen- tration. The United States and Japan have very low concentration. The ten larg- est stocks account for less than 20 percent of the markets. In Venezuela, Argen- tina, and Colombia, where the concentration ratio averaged above 0.60 in the period from 1986 to 1993 (table 1), concentration is three times larger than that in the United States and Japan. Volatility We include a measure of stock market volatility, because volatility of stock returns is another attribute that has received significant attention in the litera- ture and is of great interest to practitioners. This indicator is a twelve-month, rolling, standard-deviation estimate based on market returns. We cleanse the return series of monthly means and twelve months of autocorrelations using a procedure defined by Schwert (1989). Greater volatility is not necessarily a sign of more or less stock market development. Indeed, high volatility could be an indicator of development, so far as revelation of information implies volatility in a well-functioning market (see, for example, Bekaert and Harvey 1995). Here we refer to 'less volatility" as reflecting "greater stock market development" for simplicity. As with the other indicators, there are great cross-country differ- ences in volatility. Volatility in Pakistan, the United States, and the Netherlands averaged about 0.03 from 1986 to 1993; volatility in Brazil and Argentina was above 0.25. Asset Pricing Academic researchers and market practitioners have devoted prodigious resources to measuring the degree of integration between national stock markets and the world market and to gauging whether markets price risk efficiently (see Bonser-Neal and others 1990; Cho, Eun, and Senbet 1986; Claessens, Dasgupta, and Glen 1995; Errunza and Losq 1989, 1985a, 1985b; Errunza and Senbet 1981; Errunza, Losq, and Padmanabhan 1992; Gultekin, Gultekin, and Penati 1989; Jorion and Schwartz 1986; Koraiczyk and Viallet 1989; Solnik 1974; Stehle 1977; and Wheatley 1988). Although a market need not be integrated into the world capital markets to be developed, ana- lysts generally refer to countries that are more integrated and that price risk more efficiently as more developed. Demirgl4-KuntandLevine 299 To measure asset pricing efficiency, we use estimates of asset pricing errors computed by Koraiczyk (1994, 1996). Unfortunately, the data only permit com- putation of these pricing errors for twenty-four countries. As argued in Koraiczyk and Viallet (1989), the capital asset pricing model (CAPM) and arbitrage pricing model imply that the expected return on each asset is linearly related to a bench- mark portfolio or linear combination of benchmark portfolios. In domestic ver- sions of these asset pricing models the benchmark portfolios include only secu- rities traded on the local exchange, but in the international versions the portfolios include all securities. If the models are correct, then the benchmark portfolio, or combination of portfolios, should explain all of the systematic expected returns on assets above the risk-free interest rate.2 Thus, we term systematic deviations of expected returns as pricing errors under the maintained hypothesis that the model is correct. Using different asset pricing models, Korajczyk (1994) com- putes the systematic deviation between actual returns and those implied by the models. The asset pricing theory (Arr) and international capital asset pricing model (IcAPM) compute pricing errors using an international arbitrage pricing model and international capital asset pricing model, respectively. Koraiczyk (1994) computes the extent of pricing error under the maintained hypothesis that the models are correct. We take the average of the absolute value of the pricing errors for the stocks in a country as a measure of capital market integration. Thus, under the maintained hypothesis, greater values of the APT and iCAPM measures reflect less international integration. Greater pricing errors may reflect poor information about firms, high transactions costs, and official barriers to international asset trading. We refer to greater pric- ing errors as indicating less stock market development. The APT and ICAPM pricing errors give similar country rankings. Brazil, Turkey, and Mexico had relatively large pricing errors, but the United States, Japan, Jordan, and Pa- kistan yielded lower pricing errors, which suggest a high level of interna- tional integration. These two pricing-error estimates-Anr and ICAPM-rely on the success of equilibrium models of asset pricing that investigators sometimes have rejected as good representations of the pricing of risk. However, these measures allow us to incorporate indicators, albeit imperfect indicators, of the ability of agents to diversify risk domestically and internationally. Furthermore, we analyze the evolution of the degree of integration between each domestic market and the world market over time. Regulatory and Institutional Indicators Regulatory and institutional factors may influence the functioning of stock markets (see Pagano 1993). For example, mandatory disclosure of reliable in- formation about firms and financial intermediaries may enhance investor par- 2. Since no asset is riskless in real terms, Korajczyk and Viallet (1989) rest the restrictions implied by a zero-beta asset. 300 THE WORLD BANK ECONOMIC REVIEW, VOL. 10, NO.2 ticipation in equity markets. Regulations that instill investor confidence in bro- kers and other capital-market intermediaries should encourage investment and trading in the stock market. To measure the institutional development of emerging stock markets, we use information provided by the IFC and construct seven regulatory-institutional in- dicators. The first indicator shows whether the firms that are listed in a stock market publish price-earnings information. We give a value of 0 or 1, where 1 indicates that the information is comprehensive and published internationally. The second indicator measures accounting standards. We assign values of 0, 1, or 2, for countries with poor, adequate, or good (internationally accepted) ac- counting standards. The third indicator measures the quality of investor protec- tion laws as judged by the IFC, where 0, 1, and 2 are used to indicate poor, adequate, or good investor protection laws. The fourth indicator shows whether the country has a securities and exchange commission. The fifth, sixth, and sev- enth indicators measure restrictions on dividend repatriation by foreign inves- tors, capital repatriation by foreign investors, and domestic investments by for- eigners. We assign values of 0, 1, and 2, indicating whether capital flows are restricted, have some restrictions, or are free, respectively. We compute an aver- age institutional development indicator, which simply averages the seven regu- latory-institutional indicators. These indicators are available on an annual basis from 1986 to 1993 for twenty developing countries. There is substantial variation across countries and indicators. For example, Jordan freely allowed international capital flows to cross its borders, but did not publish regular price-earnings information and had poor accounting standards. India had accounting standards of internationally accepted quality, but restricted capital inflows and the repatriation of capital and dividends. Nigeria tightly restricted capital flows over most of the period and did not publish price- earnings information on firms in a comprehensive and internationally accepted manner. In contrast, Malaysia, Mexico, Korea, Brazil, and Chile had very high institutional development indicators overall (table 1). Correlations between Various Izdicators of Stock Market Development Many stock market indicators are significantly correlated in an intuitively plausible fashion.3 First, market size is significantly positively correlated with total value traded/GcP and the average institutional indicator, and significantly negatively correlated with pricing error and volatility. Countries with big stock markets have less volatile, more efficient stock markets with a high volume of trading relative to GmP. Second, countries with highly concentrated markets have markets that are underdeveloped. Market concentration is significantly nega- tively correlated with market size and market liquidity, and significantly posi- tively correlated with pricing error. Third, countries that have stock markets which are more integrated internationally-as measured by low APT and ICAPM 3. We do not report the actual values here due to space constraints For these and for more detailed stanstics throughout the article, see Demirgii;-Kunt and Levine (1995). Dem;irgi-Kunt and Levine 307 values-have less volatile stock returns. Fourth, countries with well-developed regulatory and institutional systems, as defined by the IFC, tend to have large, liquid stock markets. Although many stock market development indicators are significantly cor- related in intuitively attractive ways, the correlation coefficients are frequently below 0.60. The correlations suggest that the different indicators capture different aspects of stock market development. For example, the correlation between the two measures of market liquidity, total value traded/Gmp and turnover is only 0.50. Thus, although the degree of trading relative to the size of the economy is significantly correlated with the degree of trading relative to the size of the market, the two liquidity measures do not move one for one. Instead, they provide complementary information about stock market liquidity. Therefore, to measure how well stock markets function in general, that is, to compute an index of overall stock market development, we need to incorporate the information contained in a broad selection of these indicators. IH. WHICH STOCK MARKETs ARE MOST DEVELOPED? Which stock markets are most developed overall? To answer this question, we construct four conglomerate indexes of stock market development that aggregate the information contained in the individual indicators. We then use these conglom- erate indexes to rank countries in terms of overall stock market development. The Indexes To compute the conglomerate indexes of stock market development, we average the means-removed values of particular stock market development indicators. To construct each index, we follow a two-step procedure. INDEX1 aggregates information on market capitalization, total value traded/cDP, and turnover. First, for each country, i, we compute the means-removed market capitalization, total value tradedlGDP, and turnover ratios. We define the means-removed value of variable X for country i as X(i)m = [X(i) - mean(X)] 1 {ABS[ mean(X)]), where the term in the denominator is the absolute value of the average value of X across all countries from 1986 to 1993. For the pricing-error measures (APT and iCAPM) and the market concentration measure, where larger numbers refer to less stock market development, we multiply the indicator num- bers by negative 1 before computing the means-removed values. Second, we take a simple average of the means-removed market capitalization, total value traded, and turnover ratios to obtain an overall index of stock market devel- opment, INDEX1.4 INDEX1 is calculated for forty-one countries (see table 2). mINEx2 is constructed in the same way. It aggregates information on the three 4. We computed prindpal components indexes of the indicators, which allow the data to choose the weights rather than taking a simple average. However, we do not report these indexes because the rankings they produce are very highly correlated with the indexes we report. 302 THE WORLD AN .ECONOMIC REVIEW, VOL 10. NO.2 Table 2. Aggregate Indexes of Stock Market Development, 1986-93 INDEXIN INDEX2b INDEX3c INDEX4d Economy Rank Value Rank Value Rank Value Rank Value Argentina 32 -0.59 15 -0.47 23 -0.87 15 -OS0. Australia 13 0.19 7 0.12 7 0.15 Austria 19 -0.15 Belgium 28 -0.47 Brazil 24 -0.29 11 -0.38 12 -0.37 10 -0.23 Canada 14 0.09 Chile 27 -0.46 12 -0.40 11 -0.34 13 -0.37 Colombia 40 -0.88 23 -0.71 22 -0.68 21 -0.73 Denmark 26 -0.37 Finland 30 -053 France 21 -0.21 Germany 3 1.38 Greece 36 -0.73 18 -0.61 17 -0.60 16 -0.52 Hong Kong 2 2.01 India 23 -0.26 9 -0.13 9 -0.11 7 -0.01 Indonesia 35 -0.71 17 -0.52 14 -0.48 Israel 15 0.08 Italy 29 -0.51 Japan 1 2.02 1 1.63 1 1.63 1 1.41 Jordan 16 -0.08 8 0.04 8 0.07 8 -0.06 Korea, Rep. of 6 1.05 4 0.84 4 0.8S 4 0.73 Malaysia 8 0.90 5 0.72 5 0.79 5 0.60 Me3ico 18 -0.14 10 -0.16 10 -0.17 9 -0.11 Netherlands 12 0.32 New Zealand 25 -0.33 Nigeria 41 -0.96 20 -0.67 21 -0.67 19 -O.9 Norway 20 -0.18 Pakistan 39 -0.82 16 -0.51 16 -0.49 11 -0.33 Philippines 31 -0.54 14 -0.43 13 -0.42 14 -0.40 Portugal 33 -0.61 13 -0.42 1S -0.49 12 -0.34 Singapore 7 1.04 South Africa 10 0.48 Spain 22 -0.25 Sweden 17 -0.10 Switzerland 9 0.75 Thailand 11 0.38 6 0.36 6 0.36 6 0.31 Turkey 34 -0.61 19 -0.61 19 -0.62 17 -0.54 United Kingdom 4 1.23 3 1.01 3 1.02 3 0.89 United States 5 1.21 2 1.01 2 1.03 2 0.94 Venezuela 37 -0.74 22 -0.68 18 -0.61 20 -0.66 Zimbabwe 38 -0.81 21 -0.67 20 -0.66 18 -0.56 Average 0.02 -0.07 -0.07 -0.05 Number of economies 41 23 23 21 Note: Details of the calculation of the indexes are discussed in the text. Definitions of the indicators are given in table 1. The ranking order, by index, is from high to low. The indexes represent averages duringthe period from 1986 to 1993. a. INDEX1 is the average of market capitalization, total value traded/GDP, and tumover. b. INDEX2 adds APT pricing error to INDExI. c. INDE3 adds icAPM pricing error to INDEXi. d. INDEC4 adds marcet concentration to iNDEx2. Source: Authors' calculations. Demirgfi4-Ku 'd Levinie 303 indicators used in INDEXI and APT pricing error to obtain an overall indicator of stock market development that incorporates international integration. INDEX2 includes only the twenty-tlree countries with APT estimates. INDEX3 combines INDExi with the icAPM pricing error. INDEX3 includes only the twenty-three coun- tries with ICAPM pricing-error estimates. INDEX4 averages the means-removed values of market capitalization, total value traded/GDP, turnover, APT pricing error, and market concentration. We compute this index only for the twenty- one countries with data on all five underlying indicators. Rankings of Stock Market Development Table 2 gives the country-by-country values and rankings for the four aggre- gate indexes. Although there are variations in country rankings, the indexes are very highly correlated, with correlation coefficients of 0.96. Thus, the various conglomerate indexes give very similar country rankings. Here we briefly sum- marize the results from table 2. Consider first INDEx4, which aggregates the largest number of individual stock market development indicators but has the fewest countries. The INDEX4 vari- able says that Japan, the United States, the United Kingdom, and Korea have the most developed stock markets when aggregating information on market size, liquidity, international integration, and market concentration. Colombia, Ven- ezuela, Nigeria, and Zimbabwe have the four lowest rankings in this twenty- one-country sample. Next, consider NDExl, which aggregates the least information but includes the most economies (forty-one) with data on all the underlying indicators. NDEX1 ranks Japan, Hong Kong, Germany, the United Kingdom, the United States, Korea, Singapore, and Malaysia as liaving very highly developed stock markets when aggregating information on market size and liquidity. mDExl implies that Nigeria, Colombia, Pakistan, and Zimbabwe have the least developed stock markets. As noted above, Germany's high ranking is strongly influenced by the tumultuous years surrounding reunification when there was an explosion of eq- uity transactions. If Germany's two years of exceptionally high trading are re- moved in computing its averages during the period from 1986 to 1993, Ger- many falls from the top ten. Although it is difficult to answer unambiguously the question of which stock markets are most developed, our evaluation of the indexes presented in table 2 suggests that the three most developed markets are in Japan, the United States, and the United Kingdom. The most underdeveloped markets are in Colombia, Venezuela, Nigeria, and Zimbabwe. Furthermore, the data suggest that Hong Kong, Singapore, Korea, Switzerland, and Malaysia have highly developed stock markets, and Turkey, Greece, Argentina, and Pakistan have underdeveloped markets. Note that there is a dose correspondence between income per capita and stock market development. Poorer countries have lower stock market development than richer countries on average. Also note that there are important exceptions. Fre- 304 Tlil' WORLD IANK BCONOMIC RVIE.W, VOI. 10, NO. Z quently, many markets termed emerging-such as Korea, Malaysia, and Thailand- are uniformly ranked higher than markets termed developed-such as France, the Nedterlands, Australia, Canada, Sweden, and many other European countries. III. WHICH STOCK MARKETs AmIE DEVELOPING Mon RAPIDLY? Which stock markets are developing most rapidly? To answer this question, we rank countries according to the growth rates of the individual indicators of stock market development. Growth Rates of Individual Indicators of Stock Market Development Table 3 presents the average annual growth rates of the individual indicators of stock market development from 1986 to 1993. Here we highlight three points. First, in terms of market size, Indonesia and Turkey boomed over this period, growing at average annual rates of more than 100 percent a year. As a bench- mark, market capitalization in the United States grew at 4 percent annually. At the other extreme, Finland, Japan, Germany, Sweden, New Zealand, and Italy saw their market capitalization ratios shrink from 1986 to 1993. Using another measure of market size, Indonesia, Turkey, Portugal, and Thailand saw the num- ber of listed companies grow at an annual rate of over 18 percent. Second, as measured by total value traded/CDP, Indonesia, Portugal, Turkey, Venezuela, and Greece experienced rapid liquidity growth (more than 200 per- cent), while Japan and Italy weathered rapid declines (-12 and -14 percent, respectively). As with total value tradedlGDp, the turnover measure of liquidity identifies Indonesia as the fastest-growing market in terms of liquidity. Third, some cross-country quandaries emerge from studying stock market growth. Consider, for example, the cases of Mexico and Portugal. Both coun- tries liberalized their capital markets and privatized public enterprises, and both countries experienced very rapid improvements in internatioinal integration (as measured by the APT pricing error). In terms of market volatility, Mexico saw rapid declines in return volatility as it liberalized its economy and privatized state enterprises. In contrast, stock return volatility in Portugal exploded as it liberalized its capital markets and privatized its public enterprises. Another note- worthy difference between the two countries is that while market concentration grew dramatically in Mexico, it shrunk steadily in Portugal. Growtb Rates of Aggregate Indexes of Stock Market Development Using individual stock market development indicators, we found it diffi- cult to assess which markets experienced the most rapid overall develop- ment. Thus, we now evaluate the growth rate of overall indexes of stock market development. In section II, the goal was to compare the level of stock market development across countries. Here, however, we seek to measure the growth rate of each country's level of overall stock market development. Consequently, we now use the growth rate of each country's stock market Demirgi4-KuntantdLevine 305 indicator. We average these growth rates to compute an overall index of stock market development. We construct INDEXGI, which aggregates information on market capitaliza- tion, total value traded/GDP, and turnover, by computing the average annual growth rate for each indicator for each country. We then take a simple average of the growth rates to obtain an overall index of stock market development for each country. This index allows us to examine the growth rate of each country's overall level of stock market development. mNDExG2 combines the growth rates of market capitalization, total value traded! GDP, tumover, and the Arr pricing-error measure. INDEXG2 includes only coun- tries with APT pricing-error estimates. INDEXC3 is similar to INDEXG2, except that INDEXG3 uses the ICAPM pricing-error estimates instead of the APT pricing- error estimates. Finally, INDEXG4 averages the annual growth rates of market capitalization, total value traded/GDP, turnover, APT pricing error, and market concentration. WFe compute this index only for the twenty-five countries with data on all five underlying indicators for the period from 1986 to 1993. Table 4 reports the aggregate indexes of overall q;;A market growth. The main findings are straightforward. Regardless of the index, Indonesia, Turkey, Portugal, and Venezuela experienced the most rapid overall stock market devel- opment over the eight years. Although these countries began the period with underdeveloped markets, other countries with similarly underdeveloped stock markets-such as Colombia, Pakistan, and Zimbabwe-did not enjoy the ex- plosive development experienced by Indonesia, Turkey, Portugal, and Venezuela. We investigated whether stock markets that were initially underdeveloped grew faster. There is some evidence in support of convergence. Markets that were initiaily small and illiquid grev faster and became more liquid. Markets that initially were volatile and priced risk poorly tended to grow larger but not necessarily more liquid. IV. Is STOcK MARKET DEVELOPMENT LINKED TO THE REST OF THE FiNANCIAL SYsTEM? Do countries with well-developed stock markets have well-developed banks and nonbanrk financial intermediaries? To address this question, we discuss four types of measures of financial intermediary development: financial system, banks, nonbank financial corporations, and insurance and pension companies. We look at correlations among the indicators. We then construct aggregate indexes of financial intermediary development, which we use to examine the correlation between stock muket development and financial intermediary development. Indicators of Financial Intermediary Development Here we discuss the siue of the financial system, the size aiid efficiency of the banking system, the sinz of nonbank financial corporations, and the size of pri- vate insurance and private pension funds. Table 3. Growvth Rates of Indicators of Stock Market Development, 1986-93 Number Market Total valhe of listed Market Institutional APT ICAPM capitalization tradedlGDP companies Trnrriover Volatility concentration development pricing error pricing error Economy Rank Valuc Rarrk Value Rank Value Rank Value Rank Value Rank Valute Ratnk Valhe Razk Valse Rank Value Argentina 3 0.87 8 1.18 41 -0.03 21 0.17 31 0.09 22 0.08 6 0.09 18 0.14 24 0.43 Australia 3: 0.02 34 0.08 32 0.00 29 0.06 5 -0.02 8 -0.01 10 0.01 Austria 12 0.37 6 1.48 13 0.06 S 0.91 26 0.04 Belgium 33 0.00 37 0.01 38 -0.02 3 1.54 8 -0.02 Brazil 16 0.30 21 0.34 37 -0.01 42 -0.11 24 0.04 21 0.07 14 0.04 S -0.03 13 O.OS Canada 35 0.00 38 0.01 30 0.01 32 0.05 22 0.03 2s 0.09 Chile 19 0.27 23 0.27 25 0.03 43 -0.11 16 0.01 1S 0.02 16 0.03 11 0.00 14 0.06 Colombia 11 0.42 15 0.54 35 -0.01 37 -0.03 35 0.15 18 0.05 10 0.05 17 0.09 23 0.27 Denmark 24 0.06 14 0.55 34 -0.01 13 0.38 Finland 36 -0.02 28 0.19 23 0.03 18 0.24 27 0.05 France 23 0.07 32 0.09 16 O.OS 30 0.06 33 0.10 19 0.06 Germany 38 -0.03 24 0.26 26 0.02 IS 0.30 2 -0.05 8 -0.02 Greece 9 0.51 S 2.50 17 0.05 I1 0.43 29 0.08 10 0.00 2 0.22 23 0.19 18 0.13 Hong Kong 26 0.06 22 0.31 9 0.09 17 0.25 India 15 0.32 29 0.16 29 0.02 40 -0.08 32 0.09 12 0.00 17 0.02 4 -0.06 8 0.00 Indonesia 1 1.89 1 17.74 1 0.37 1 1.82 20 -0.06 19 0.14 1 -0.26 Ireland 21 0.03 Israel 7 0.53 16 0.50 5 0.15 7 0.54 25 0.04 Italy 41 -0.10 41 -0.14 19 O.OS 22 0.16 6 -0.02 Japan 37 -0.03 40 -0.12 27 0.02 39 -0.07 28 0.06 2 -0.09 3 -0.10 3 -0.07 Jordan 20 0.12 12 0.58 33 0.00 19 0.24 7 -0.02 5 -0.05 12 0.04 24 0.26 20 0.16 Korea, Rep. of 17 0.28 18 0.43 8 0.09 36 -0.01 18 0.01 24 0.09 15 0.03 12 0.03 9 0.01 Luxembourg 6 0.12 2 1.66 Malaysia 14 0.34 7 1.31 10 0.09 12 0.40 3 -o.0s 3 -0.08 19 0.01 14 0.04 5 -_ ;'l Mexico 10 0.49 11 0.62 22 0.04 34 0.01 1 -0.06 26 0.23 13 0.04 1 -0.15 4 -0.07 Netherlands 32 0.01 31 0.13 21 0.04 4 1.39 12 -0.01 New Zealand 40 -0.05 26 0.20 43 -0.11 24 0.10 4 -0.03 Nigeria 21 0.10 2S 0.23 12 0.08 35 -0.01 9 0.00 4 0.11 9 -0.01 15 0.06 Norway 25 0.06 10 0.67 39 -0.03 9 0.4S 23 0.03 Pakistan 18 0.27 19 OAO 11 0.08 41 -0.09 11 -0.01 23 0.08 7 0.09 21 0.16 2 0.25 Philippines 5 0.61 13 0.57 18 0.05 38 -0.06 17 0.01 20 0.07 9 0.06 13 0.04 6 0.00 Portugal 8 0.51 2 3.25 3 0.20 14 0.3S 37 0.85 7 -0.03 18 0.02 2 -0.14 2 -0.26 Singapore 29 0.04 20 0.34 15 0.06 10 0.43 South Africa 28 0.04 30 0.13 20 0.04 27 0.08 Spain 34 0.00 33 0.09 24 0.03 25 0.10 20 0.03 Sweden 39 -0.05 39 -0.02 42 -0.05 26 0.09 30 0.08 Switzerland 22 0.07 35 0.05 14 0.06 16 0.29 19 0.02 17 0.02 Taiwan (China) 7 0.12 20 0.17 10 -0.02 6 -0.03 8 0.09 16 0.05 17 0.12 Thailand 6 0.57 9 0.76 4 0.18 31 0.05 IS 0.01 1 -0.12 11 0.04 22 0.17 16 0.09 Turkey 2 1.02 3 2.87 2 0.23 6 0.65 34 0.13 13 0.01 1 0.29 20 0.14 19 0.14 ,, United Kingdom 30 0.03 27 0.20 40 -0.03 23 0.10 14 0.00 14 0.02 7 -0.02 7 0.00 q United States 27 0.04 36 0.04 36 -0.01 33 0.02 9 -0.02 16 0.02 6 -0.02 11 0.01 Venezuela 4 0.66 4 2.77 31 0.01 8 0.48 36 0.27 11 0.00 3 0.17 1S 0.04 21 0.24 Zimbabwe 13 0.3S 17 0.4S 28 0.02 28 0.06 13 0.00 4 -0.07 5 0.11 10 -0.01 12 0.04 Average 0.27 1.02 0.05 0.31 0.05 0.02 0.07 0.04 0.06 Number of economies 41 41 43 43 37 26 20 24 24 Note. Growth rates are the average annual growth rates. Definitions of the indicators are given in table 1. For each indicator, economies are ranked by the rate of growth of stock market development, from high to low. Thus, for market capitalization, total value traded/Gtp, number of listed companies, turnover, and institutional development, the ranking by value of the indicator is from high to low. For volatility, market concentration, AFr pricing error, and icAP&M pricing error, the ranking by value of the indicator is from low to high. Source: Authors' calculations. 308 THE WORLD BANK LCONOMIC REVIEW, VOL. 10, NO.2 Table 4. Growth Rates of Aggregate Indexes of Stock Market Development, 1986-93 INDEXGI INDEXG2 INDEXG3 INDEXG4 Economy Rank Value Rank Value Rank Value Rank Value Argentina 7 0.74 6 0.52 7 0.45 S 0.54 Australia 36 0.05 21 0.04 21 0.04 Austria 6 0.92 Belgium 10 0.52 Brazil 24 0.18 15 0.14 15 0.12 16 0.12 Canada 38 0.02 24 -0.01 Chile 26 0.14 18 0.11 17 0.09 17 0.10 Colombia 18 0.31 12 0.21 14 0.17 12 0.22 Denmark 16 0.33 Finland 28 0.14 France 34 0.07 22 0.04 Germany 25 0.17 14 0.14 Greece 5 1.15 S 0.81 5 0.83 4 0.86 Hong Kong 22 0.20 India 29 0.14 16 0.12 16 0.10 18 0.10 Indonesia 1 7.15 1 5.33 1 5.43 Israel 9 0.52 Italy 40 -0.03 Japan 41 -0.07 23 -0.03 23 -0.04 25 -0.03 Jordan 17 0.31 13 0.17 12 0.19 9 0.25 Korea, Rep. of 21 0.23 14 0.17 13 0.17 13 0.15 Malaysia 8 0.68 7 0.50 6 0.52 6 0.53 Mexico 14 0.37 8 0.32 9 0.30 11 0.22 Netherlands 11 0.51 New Zealand 33 0.09 Nigeria 31 0.11 20 0.08 20 0.06 20 0.08 Norway 13 0.39 Pakistan 23 0.20 17 0.11 18 0.08 15 0.13 Philippines 1S 0.37 10 0.27 10 0.28 8 0.26 Portugal 3 1.37 3 1.06 3 1.09 2 1.04 Singapore 20 0.27 South Africa 32 0.09 Spain 35 0.06 Sweden 39 0.01 Switzerland 27 0.14 19 0.10 Thailand 1;! 0.46 9 0.30 8 0.32 7 0.37 Turkey 2 1.51 2 1.10 2 1.10 1 1.13 United Kingdom 30 0.11 19 0.09 19 0.08 21 0.08 United States 37 0.03 22 0.03 22 0.02 23 0.02 Venezuela 4 1.30 4 0.97 4 0.92 3 0.98 Zimbabwe 19 0.29 11 0.22 11 0.21 10 0.23 Average 0.53 0.55 0.54 0.31 Number of economies 41 23 23 25 Note: Growth rates of indexes are obtained by averaging the growth rates of different stock market indicators, depending on the index. Indexes are defined in table 2. The ranking, by growth rate of each index, is from high to low. Source: Authors' calculations. Demirgfi-KIunt and Levine 309 FINANCIAL SYSrEM. On the basis of work by King and Levine (1993), we use three measures of financial system development. The ratio of liquid liabilities of the financial intermediaries to GDP is M3 money supply divided by GDP. The ratio is a measure of the overall size of the formal financial system. If the size of the financial system is positively related to the provision of financial services, then this ratio should be a good indicator of the provision of financial intermediary services. The ratio of quasi-liquid liabilities to GDP is M3 money supply minus MI, divided by GDP. It subtracts narrow money from the liquid liabilities measure of financial intermediary size. Analysts sometimes use the quasi-liquid measure instead of liquid liabilities because Ml/GDP represents highly liquid bank depos- its and therefore may not be as closely associated with efficient financial inter- mediation as longer-term investments in financial intermediaries. The quasi- liquid measure focuses on longer-term liabilities. Liquid and quasi-liquid liabilities that finance government deficits may not reflect the provision of efficient financial intermediary services (such as acquir- ing information about firms, monitoring managers, and facilitating transactions and risk diversification). Therefore, we compute a third variable, domestic credit to private firms divided by GDP. Unfortunately, although IMF (various issues) classifies credit as "claims on the private sector," some of these claims in some countries include credit to public enterprises. Table 5 indicates that Hong Kong, Japan, and Switzerland had well-devel- oped financial systems as measured by liquid and quasi-liquid liabilities to GD? and domestic credit to private firms. In contrast, Argentina, Brazil, Mexico, Colombia, and Nigeria had underdeveloped financial systems as revealed by these three indicators. BAsKS. To measure the level of development of the banking system, we use the ratio of the total claims of deposit money banks to GDP. The three countries with the largest values for this indicator were Switzerland, Luxembourg, and Japan. At the other extreme, Nigeria, Argentina, and Venezuela had the lowest ratio of bank credit to GDP during the period from 1986 to 1993. We compute a measure of banking efficiency, which we call spread, that equals the difference between bank lending and borrowing rates. This measure may not accurately capture banking efficiency because the interest rate data may not accurately reflect borrowing and lending costs. The spread indicator will not provide accurate information on how weli banks monitor firm managers, nor will it capture government intervention in the banking system in a very informa- tive way. But the spread indicator is widely used and available across countries. We include it for completeness. For better measures of financial repression for a few select countries see Giovannini and De Melo (1993). According to the spread indicator, the banking systems of Switzerland, Canada, and the United King- dom were among the most efficient, whereas Argentina, Israel, and Turkey had the least efficient banks. Table 5. Indicators of Financial Intermediary Developmelt, 1986-93 (annual average) Assets of private Assets of private Quasi-liquid Domnestic credit Total claimis nonbank financial insurance a7d Liquid liabilities liabilities to private sector of deposit corporations to pension fIinds to GDPa to GDPb to GDP banks to GDP Spreadc GDP to GDP Economy Rank Value Rank Valte Rank Value Rank Value Rank Valie Rank Valure Rank Value Argentina 44 0.23 41 0.15 39 0.26 41 0.37 38 45.28 Australia 24 1.13 19 0.89 16 1.07 22 1.19 28 6.28 S 0.4S 8 0.35 Austria 9 1.72 7 1.44 11 1.36 4 2.39 Belgium 31 0.94 33 0.55 29 0.61 20 1.20 26 5.70 Brazil 43 0.26 42 0.14 38 0.29 36 0.51 Canada 21 1.27 14 0.97 24 0.86 28 0.93 2 1.38 6 0.42 6 0.48 w Chile 36 0.72 28 0.61 22 0.93 29 0.90 30 6.96 ° Colombia 41 0.47 38 0.28 40 0.2S 33 9.70 21 0.06 18 0.03 Denmark 23 1.19 27 0.62 21 0.98 21 1.20 23 5.35 1S 0.12 S 0.54 Finiand 25 1.10 21 0.82 8 1.60 14 1.60 12 3.55 12 0.21 9 0.33 France 16 1.36 20 0.87 6 1.77 8 2.00 34 10.57 10 0.28 11 0.20 Germany 17 1.34 15 0.94 5 1.80 5 2.16 22 S.15 10 0.33 Greece 12 1.54 9 1.21 33 0.45 27 0.95 31 7.19 Hong Kong 1 3.91 1 3.53 India 33 0.87 31 0.57 32 0.51 33 0.68 8 3.00 Indonesia 37 0.65 35 0.44 28 0.66 34 0.65 16 4,23 22 0.02 ireland 32 0.88 26 0.64 31 0.52 30 0.87 21 5.10 Israel 20 1.30 10 1.19 17 1.01 7 2.07 37 20.95 Italy 13 1,47 23 0.75 27 0.71 24 1.01 32 7.34 9 0.33 17 0.06 Japan 2 3.57 2 3.00 2 2.27 . 3 2.s8 11 3.31 7 0.43 Jordan 4 2.40 8 1.41 15 1.24 16 I.S2 24 5.56 17 0.08 16 0.07 Korea, Rep. of 29 0.96 22 0.78 19 0.99 2S 1.00 6 2.90 3 0.55 12 0.14 Luxembourg 6 2.36 2 2.59 4 2.31 Malaysia 8 1.89 5 1.51 12 1.33 13 1.61 5 2.68 8 0.39 14 0.10 Mexico 42 0.42 37 0.29 37 0.29 38 0.48 3S 13.76 18 0.08 19 0.02 Netherlands 10 1.61 11 1.16 9 1.53 9 1.97 29 6.92 25 0.00 1 1.08 New Zealand 27 1.03 29 0.61 23 0.92 23 1.10 17 4.49 Nigeria 40 0.48 40 0.23 42 0.24 42 0.33 19 4.60 19 0.08 Norway 22 1.26 30 0.61 14 1.27 15 1.57 1S 4.21 Pakistan 35 0.79 39 0.25 30 O.S5 32 0.70 23 0.01 22 0.00 Philippines 38 0.63 34 0.48 36 0.34 37 0.48 20 5.04 20 0.07 20 0.01 Portugal 14 1.47 16 0.93 25 0.84 17 1.49 27 5.96 Singapore 7 2.26 4 1.80 7 1.64 12 1.87 9 3.02 2 0.84 13 0.11 South Africa 26 1.06 24 0.72 26 0.74 31 0.78 10 3.20 Spain 1S 1.44 18 0.90 13 1.31 11 1.89 18 4.59 11 0.24 15 0.08 Sweden 30 0.96 20 0.98 18 1.41 2S 5.68 1 0.89 4 0.56 Switzerland 3 2.83 3 2.26 1 3.14 1 3.26 1 0.87 Taiwan (China) 5 2.38 6 1.49 4 1.80 6 2.10 Thailand 19 1.31 12 1.12 18 0.99 19 1.23 13 3.60 13 0.15 21 0.01 Turkey 39 0.61 36 0.41 35 0.36 35 0.54 36 19.50 24 0.01 United Kingdom 11 1.59 17 0.92 3 1.97 10 1.97 3 1.82 16 0.08 2 0.92 United States 18 1.33 13 0.99 10 1.42 26 0.99 7 3.00 4 0.33 3 0.67 Venezuela 34 0.80 32 0.55 34 0.40 40 0.45 7 0.40 Zimbabwe 28 0.96 25 0.70 41 0.24 39 0.45 14 3.90 14 0.13 Average 1.33 0.95 1.01 1.31 6.81 0.26 0.30 Number of economies 44 42 42 42 38 25 22 Note: The financial intermediary development of each economy is ranked from high to low. This ranking is shown by ranking the value of the indicator from high to low for liquid liabilities to GDP, quasi-liquid liabilities to GDP, domestic credit to private sector to GDP, total claims of deposit banks to GDP, assets of private nonbanks to GDP, and assets of private insurance and pension funds to GDP; for spread, the ranking by value of the indicator is from low to high. a. Liquid liabilities of the financial system are the M3 definition of money. b. Quasi-liquid liabilities are M3 minus MI money. c. The spread is the difference between bank lending and borrowing rates. Source: Authors' calculations. 312 TlE WORLD 1ANK ECONOMIC REVIEW, VOL. 10 NO.2 NONBANK FINANCIAL CORPORATIONS. We use the ratio of assets of private nonbank financial intermediaries to GDP to measure the size of nonbank financial corporations, such as finance companies, mutual funds, and brokerage houses. The four economies with the largest values for this indicator were Sweden, Singapore, Korea, and the United States. Indonesia, Pakistan, Turkey, and the Netherlands had very low values.5 INSURANCE AND PENSION COMPANIES. Finally, we use the ratio of assets of private insurance companies and pension funds to GDP to measure the size of private insurance and pension companies. The three countries with the largest values for this indicator were the Netherlands, the United Kingdom, and the United States. The Philippines, Thailand, and Pakistan had very low values. Correlations between Various Indicators of Financial Intermediary Developnment The measures of financial system size-the liquid, quasi-liquid, and domestic credit to private firms indicators-are very highly correlated. The correlation coefficients are 0.79 or higher and significant at the 0.01 level. The correlations between the indicators of the size of the financial system and indicators of the size of banks, private nonbank financial corporations, and private insurance and pension companies are not as strong. Although all of the correlations are positive, many are not significant. Furthermore, the correlation coefficient of those that are significant is frequently below 0.50. The dir ferent financial intermediary indicators give different country rankings of financial intermediary development. These differences reflect financial struc- tures across countries, that is, different combinations of financial intermediar- ies and financial markets that compose a country's financial system. Differences in financial structure may reflect legal differences. For example, countries with unmversal banking, as distinct from the more segregated legal and regulatory impediments of the United States, may develop different combinations of finan- cial intermediaries. The overall size of the financial system across countries with different financial structures, however, may be similar, as may be the provision of financial services to investors. For example, countries with big financial sys- tems have big banks and nonbank financial corporations, but the correlation between financial system size and private insurance and pension companies is not strong. Aggregate Indexes of Financial Intermediary Development Because we want to compare an overall measure of financial intermediary development with our aggregate indicators of stock market development, we construct conglomerate indexes of financial intermediary development. Using 5. We collected data on private nonbank financial corporations, insurance companies, and pension funds from individual country reports, including documents published by ministries of finance, central banks, and regulatory agencies. Demirgl4-Kunt and Levine 313 the same procedure for constructing conglomerate indexes discussed above, this section constructs three financial intermediary indexes. FINDExI averages the means-removed values of the ratio of liquid liabilities to GDP and the ratio of domestic credit to the private sector to GDP. FJNDEX2 averages the means- removed values of the ratio of liquid liabilities to GDP, the ratio of domestic credit to the private sector to GDP, the ratio of assets of private nonbank finan- cial corporations to GDP, and the ratio of assets of private insurance and pension funds to GDP. FNDEX3 combines the means-removed values of the ratio of total claims of deposit banks to GDP, the ratio of assets of private nonbank financial corporations to GDP, and the ratio of assets of private insurance and pension funds to GDP. Table 6 provides the country rankings and the values of these indexes from 1986 to 1993. The aggregate indexes of financial intermediary development are highly correlated, with correlation coefficients above 0.73 and P-values less than 0.01. The results in table 6 on RNDEx3-which aggregates iniormation on banks, private nonbank financial corporations, and private insurance companies and pension funds-suggest that the top five economies with the most developed financial intermediaries were Switzerland, Sweden, Luxembourg, Australia, and Singapore. The five countries with the least developed financial intermediaries were Colombia, Pakistan, the Philippines, Turkey, and Mexico. We prefer FINDEx3 to the other financial intermediary indexes because it combines information on particular financial intermediaries: banks, nonbank financial corporations, and insurance companies and pension funds. The other aggregate indexes mix infor- mation on particular intermediaries with information on liabilities that span across different types of intermediaries. Stock Market Development and Financial Intermediary Development Do countries with well-developed stock markets have well-developed banks and nonbank financial intermediaries? Table 7 presents the correlations between individual indicators of stock market development and individual indicators of financial intermediary development. Here we highlight three points. First, stock market size (market capitalization) and liquidity (as measured by total value traded/GDP) are positively correlated with all of the indicators of financial intermediary development. They are significantly correlated with all of the indicators of financial intermediary development except the ratio of the as- sets of private insurance and pension companies to GDP. Second, volatility is significandy negatively correlated with all the indicators of financial intermedi- ary development except the ratio of assets of private nonbank financial corpora- tions to GDP. Thus, countries with well-developed financial intermediaries, large banks, and large private insurance companies and pension funds tend to have less volatile stock markets. Third, APT and ICAPM pricing errors are negatively correlated with indicators of financial intermediary development. Countries with stock markets that are internationally integrated tend to have larger financial systems and banks than countries with less internationally integrated markets. 314 11-1EWORLDBANKECONOMICREVIIE.VOL lO,NO.I Table 6. Aggregate Indexes of Financial Intermediaries, 1986-93 FINDEXI' FINDEX2b FINDEX3' Economy Rank Value Rank Value Rank Value Argentina 42 -0.79 37 -0.72 Australia 11 0.23 4 0.75 Austria 20 -0.12 7 0.23 12 0.34 Belgium 29 -0.35 23 -0.06 Brazil 41 -0.75 34 -O.S8 Canada 17 -0.06 6 0.27 13 0.32 Chile 28 -0.29 28 -0.32 Colombia 40 -0.72 20 -0.78 43 -0.82 Denmark 21 -0.12 12 -0.02 19 0.01 Finland 13 0.12 10 0.03 20 0.01 France 8 0.31 9 0.09 18 0.06 Germany 9 0.30 14 0.31 Greece 26 -0.23 27 -0.30 India 30 -0.44 33 -0.48 Indonesia 32 -0.46 36 -0.72 Ireland 31 -0.45 29 -0.36 Israel 19 -0.07 10 0.54 Italy 22 -0.13 1S -0.17 26 -0.23 Japan 2 1.31 7 0.62 Jordan 6 0.42 14 -0.16 31 -0.45 Korea, Rep. of 24 -0.21 11 0.02 17 0.08 Luxembourg 3 0.94 Malaysia 10 0.29 8 0.10 21 0.00 Mexico 39 -0.71 19 -0.77 39 -0.77 Netherlands 7 0.34 4 0.53 6 0.65 New Zealand 23 -0.20 2S -0.19 Nigeria 38 -0.71 38 -0.72 Norway 15 0.03 15 0.16 Pakistan 33 -0.46 17 -0.72 42 -0.81 Philippines 37 -0.61 18 -0.73 41 -0.78 Portugal 18 -0.06 16 0.11 Singapore 4 0.56 1 0.70 5 0.68 South Africa 27 -0.23 30 -0.39 Spain 14 0.11 13 -0.15 24 -0.14 Sweden 25 -0.21 2 0.67 2 1.04 Switzerland 1 1.45 1 1.39 Taiwan (China) 3 0.64 11 0.51 Thailand 16 -0.02 16 -0.36 32 -0.48 Turkey 36 -0.59 40 -0.78 United Kingdom 5 0.45 S 0.53 9 O.SS United States 12 0.14 3 0.59 8 0.60 Venezuela 35 -0.52 22 -0.06 Ziimbabwe 34 -0.52 35 -0.59 Average -0.08 -0.00 -0.02 Number of economies 42 20 43 Note: Details of the calculation of the indexes are discussed in the text. The ranking order, bygrowth rate of each index, is from high to low. a. FINDEX1 is the average of the ratio of liquid liabilities (M3 money) to GDP and the ratio of domestic credit to the private sector to GDP. b. wNDEXZ is the average of the ratio of liquid liabilities (M3 money) to GDP, the ratio of domestic credit to the pri-vate sector to GDP, the ratio of the asscts of private nonbank institutions to GDP, and the ratio of assets of private insurance and pension funds to GDP. c. FINDEX3 is the average of the ratio of total claims of deposit banks to GDP, the ratio of the assets of private nonbank institutions to GDP, and the ratio of the assets of private insurance and pension funds to GDP. FINDEx3 does not include the last two terms if data are not available. Source: Authors' calculations. DemirgiJg-Kunt and Levine 31 S Table 7. Correlations between Indicators of Financial Intermediary and Stock Market Development, 1986-93 Financial intermediavry indicator Total Domestic Assets of claims of credit to Quasi- Assets private Liquid deposit private liquid of private insurance liabilities banks to sector liabilities nonbanks and pension Stock market indicator to GDP' GDP to GDP to GDP? to GDP funds to GDP Market capitalization Correlation 0.66 0.40 0.52 0.67 0.47 0.29 [0.001 10.01] [0.00] [0.00] [0.021 [0.20] Number of observations 41 40 40 40 2S 22 Total value traded/GDP Correlation 0.75 0.58 0.70 0.78 0.46 0.33 [0.00] [0.00] [0.00] [0.00] [0.02] 10.14] Number of observations 41 40 40 40 25 22 Turnover Correlation 0.18 0.42 0.38 0.22 0.27 0.11 [0-25] [0.01] [0.01] [0.16] [0.20] [0.611 Number of observations 41 40 40 40 25 22 APT pricing error Correlation -0.49 -0.48 -054 -OA5 -0.06 -0.40 [0.011 [0.02] [0.01] [0.03] [0.841 [0.201 Number of observations 24 24 24 24 16 12 ICAPM pricing error Correlation -0.51 -0.47 -055 -0.46 -0.23 -0.38 10.01] [0.02] [0.01] [0.02] [0.391 10.221 Number of observations 24 24 24 24 16 12 Volatility Correlation -0.41 -0.42 -0.40 -0.37 -0.12 -0.52 [0.01] [0.011 10.01] [0.031 [0.60] [0.021 Number of observations 37 37 37 36 21 20 Market concentration Correlation -0.24 -0.28 -0.32 -0.24 -0.42 -0.56 [0.24] [0.16] [0.11] [0231 [0.11] 10.04] Number of observations 26 26 26 26 16 14 Institutional development Correlation -0.05 0.21 0.26 0.04 0.42 0.31 [0.84] [0.37] [0.27] [0.861 [0.15] 10.20] Number of observations 20 20 20 20 13 8 Note: P-values are in brackets. Indicators of stock market development are defined in table 1. a. Liquid liabilities are the M3 definition of money. b. Quasi-liquid liabilities are M3 minus Ml money. Source: Auxhors' calculations. 316 Tmi WORLD IANK ECONOMIC RIVIEW. VOL. 10, NO.2 Table 8. Correlations between Aggregate Indexes of Financial Intermediary and Stock Market Development, 1986-93 Stoak market index Financial intermediary index INDEXI INDEX2 INDEX3 INDEX4 FINIDEXI Correlation 0.72 0.83 0.84 0.81 [0.001 [0.00] [0.001 [0.00] Number of observations 40 23 23 21 FINDEX2 Correlation 0.67 0.89 0.89 0.92 [0.00J 10.001 [0.00] 10.001 Number of observations 20 11 11 10 FINDEX3 Correlation 0.62 0.79 0.79 0.80 10.00] 10.00] 10.001 [0001] Number of observations 40 23 23 21 Note: P-values are in bmckets. The stock market indexes are defined in table 2, and the financial intcrmcdiarY indexes are defined in table 6. Details of the calcularion of the indexes are discussed in the text. Source: Authors' calculations. Using the conglomerate indexes of stock market development and the con- glomerate indexes of financial intermediary development, the strong positive correlation between stock market development and financial intermediary de- velopment emerges even more strongly. As shown in table 8, the aggregate in- dexes of stock market development are all significantly correlated with the ag- gregate indexes of financial intermediary development at the 0.01 level. Furthermore, measures of stock market pricing errors, as represented by APT and IcAPM, are positively correlated with banking inefficiency as measured by the interest rate spread (table 9). Stock market development (including mea- sures of pricing errors) and financial intermediary development (including mea- sures of banking efficiency) go hand in hand. These results are consistent with Boyd and Smith's (1996) model, where there is a role for both banking and equity markets as economies develop. Thus, with increases in per capita income and wealth, stock markets emerge and complement (but not replace) bank lend- ing. As economiies develop, their financial systems display a wide array of insti- tutions and markers. V. SUMMARY This article collected and summarized information on a wide assortment of indicators of stock market and financial intermediary development. To describe different characteristics of equity market development, we used measures of stock market size, liquidity, integration with w,vorld capital markets, volatility, DemirgifiKunt and Levine 317 Table 9. Correlations between Stock Market Pricing Errors and Financial Intermediary Inefficiency, 1986-93 APT ICAPM Indicator Spread' pricing error pricing error Spread Correlation 1.00 0.20 0.81 10.00] [0.391 [0.001 Number of observations 39 21 21 APT pricing error Correlation 1.00 0.68 (0.00] [0.00] Number of observations 24 24 ICAPM pricing error Correlation 1.00 [0.00] Number of observations 24 Note: P-values are in brackets. a. The spread is the difference between bank lending and borrowing rates. Source: Authors' calculations. concentration, and features of the regulatory system. To describe the develop- ment and structure of financial intermediaries, we used measures of the overall size of the financial intermediary sector, the allocation of credit, the spread be- tween borrowing and lending interest rates, and the size of particular types of financial intermediaries, such as banks, insurance companies, and pension funds. No single mneasure is the correct measure of stock market or financial interme- diary development. Indeed, each indicator may be the appropriate measure for a particular question. Consequently, this artide's major contribution is the collec- dion and comparison of a wide variety of indicators. The article constructs ag- gregate indexes of stock market and financial intermediary development that combine the information reflected in several individual indicators. There are enormous cross-country differences for each indicator of stock market development. For example, the ratio of market capitalization to GDP is greater than 1 Y; five countries and less than 0.10 in five countries. Even so, there are intuitively appealing correlations among the individual stock market indicators and between the stock market indicators and measures of financial intermnediary development. Big markets, for example, tend to be less volatile, more liquid, and less concentrated in a few stocks; internationally integrated markets tend to be less volatile; and institutionally developed markets tend to be large and liquid. Moreover, we find that across countries the level of stock mar- ket development is highly correlated with the development of banks, nornbank financial corporations, and insurance companies and private pension funds. When we compute conglomerate indexes of overall stock market develop- ment, plausible and educational patterns emerge. We find that the three most 318 TH wORWLD BANK ECONOMIC REVIEW. VOL. 10, NO. 2 developed markets are Japan, the United States, and the United Kingdom. The most underdeveloped markets are Colombia, Venezuela, Nigeria, and Zimbabwe. The data suggest that Korea, Switzerland, and Malaysia have highly developed stock markets, while Turkey, Greece, Argentina, and Pakistan have underdevel- oped markets. Furthermore, although richer countries generally have more de- veloped stock markets than poorer countries, many markets labeled emerging- such as Korea, Malaysia, and Thailand-are systematically more developed than markets labeled developed-such as France, the Netherlands, Australia, Canada, Sweden, and Norway. During the period from 1986 to 1993, some markets exhibit very rapid devel- opment in terms of size, liquidity, and international integration. Indonesia, Tur- key, Portugal, and Venezuela have experienced explosive developrnent. Future case studies into the underlying causes of and the economic consequences of this rapid development could yield valuable insights. In this article, the goal has not ;0Lven to test specific hypotheses rigorously. Rather, our objectives have been to compile and compare different indicators of stock market developnient, highlight some important correlations, and, most important, stimulate future research into the links between stock market devel- opment and economic development. APPENDIx. THE CROSS-COUNTRY CONPARABILTYr OF STOCK MARKET DATA The lFc began calculating emerging mnarket indexes in 1981. IFC selects stocks for inclusion in the indexes on the basis of three criteria: size, liquidity, and industry. The indexes include the largest and most actively traded stocks in each market, targeting 60 percent of total market capitalization at the end of each year. The index targets 60 percent of trading volume during the year. Size is measured by market capitalization, and liquidity is measured by the total value of shares traded during the year. Selection criteria used by Morgan Stanley Capital International (MscI) in creating industrial coLntry stock indexes are comparable to those of the IFC. In constructing the MSCI indexes, 60 percent coverage of the total market capitalization of each market is the primary objective. In contrast to the IFc indexes, Msa indexes have no secondary objective regarding volume of trading. Instead, they try to replicate the industrial composition of the local market and take a representative sample of large, medium, and small capitaliza- tion stocks. Msci uses liquidity as a consideration in choosing among the me- dium and small capitalization stocks. The IFC indexes represent value-weighted portfolios of the stocks in each market. Each stock is weighted by its market capitalization in the same way in which the MSCI country indexes are formed, using the chained Paasche method. Most of the stock market indicators compiled in this study are constructed using complete market information, and are fully comparable. For example, the market capitalization ratio is the value of all listed shares in the stock exchange divide ci by GDP in all countries. This is true for all our indicators except the Demirgi4-IKunt and Levine 319 volatility and asset pricing indicators, which use index information or individual stock prices for those indicators included in the indexes. For volatility and asset pricing, differences in constructing MSCI and IFC indexes may introduce a bias. However, as discussed above, the magnitude of this bias is likely to be small, and MSCI and IFC indexes have been used in cross-country empirical studies in the literature (see for example Bekaert and Haney 1995 and De Sandis 1993).6 6. This appendix is mostly based on wc r (1993) and Schmidt (1990). REFERENCES The word "processed" describes informally reproduced works that may not be com- monly available through library systems. Bekaert, Geert, and Campbell Harvey. '1995. 'Time-Varying World Market Integra- tion." Journal of Finance 5O(2):403-44. Bencivenga, Valerie R., Bruce D. Smith, and Ross Mvi Stan. 1996. "Equity Markets, Transactions Costs, and Capital Accumulation: An Illustration." The World Bank Economic Review 10(2)-.241-65. Bonser-Neal, Catherine, Greggory Brauer, Robert Neal, and Simon Wheatley. 1990. "International Investment Restrictions and Closed-End Country Fund Prices." Jour- nal of Finance 45(2):523-47. Boyd, John, and Bruce Smith 1996. "The Coevolution of the Real and Fmancial Sec- tors in the Growtih Process." The World Bank Economic Review 10(2):371-96. Cho, D. C., C. S. Eun, and L W. Senbet. 1986. "International Arbitrage Pricing Theory: A * Empirical Investigation." Journal of Finance 41(2, June):313-29. Claessens, Stijn, Susmita Dasgupta, and Jack Glen. 1995. "Return Behavior in Emerg- ing Stock Markets." The World Bank Economic Reviesw 9(1l):131-52. DeLong, J. B., Andrei Shleifer, Lawrence Summers, and R. J. Waldmann. 1989. "The Size and Incidence of the Losses from Noise Trading." Journal of Finance 44(3):681-96. Demirgiis-Kunt, Ash, and Ross Levine. 1995. "Stock Market Development and Finan- cial Intermediaries: Stylized Facts." Working Paper No. 1462. World Bank, Policy Research Department, Washington, D.C. Processed. Demirguc-Kunt, Ash, and Vojislav Maksimovic. 1996. "Stock Market Development and Finarncing Choices of Firms." The World Bank Economic Review 10(2):341- 69. De Santis, Giorgio. 1993. "Asset Pricing and Portfolio Diversification Evidence from Emerging Fmancial Markets." In Stiin Claessens and Sudarshan Gooptu, eds., Port- folio Investment in Developing Countries. World Bank Discussion Paper 228. Wash- ington, D.C. Devereux, Michael B., and Gregor W. Smith. 1994. "International Risk Sharing and Economic Growth.' Internationl Economic Review 35(4):535-50. Errunza, Vihang, and Etienne Losq. 1985a. "The Behavior of Stock Prices on LDc Mar- kets." Journal of Banking and Finance 9:561-75. 1985b. "International Asset Pricing under Mild Segmentation: Theory and Test." Journal of Finance 40(1, March):105-24. 320 THE WORLD BANK ECONOMIC REVIEW. VOL tO, NO.2 . 1989. 'Capital Flow Controls, International Asset Pricing and Investors' Wel- fare: A Multi-Country Framework." Journal of Finance 44(1):1025-38. Errunza, Vihang, and Lemma Senbet. 1981. "The Effects of International Operations on the Market Value of the Firm: Theory and Evidence," Journal of Finance 36(May):401-17. Errunza, Vihang, Etienne Losq, and Prasad Padmanabhan. 1992. "Tests of Integration, Mild Segmentation and Segmentation Hypotheses." Journal of Banking and Finance 16(5):949-72. Giovannini, Alberto, and Martha De Melo. 1993. "Government Revenue from Finan- cial Repression." American Eccnomic Review 4(September):953-63. Gultekin, Mustafa N., N. Bulent Gultekin, and Alessandro Penati. 1989. "Capital Con- trols and International Capital Market Segmentation: The Evidence from the Japa- nese and American Stock Markets." Jounal of Finance 44(September):849-69. Harvey, Campbell R. 1995. "The Risk Exposure of Emerging Equity Markets." The World Bank Economic Review 9(1):19-50. iuc (International Finance Corporation). 1993. IFC Index Methodology. Washington, D.C. w (International Monetary Fund). Various issues. International Financial Statistics. Washington, D.C. Jorion, Philippe, and Eduardo Schwartz. 1986. "Integration vs. Segmentation in the Canadian Stock Market." Journal of Finance 41(3):603-12. King, Robert G., and Ross Levine. 1993. "Finance and Growth: Schumpeter Might Be Right." Quarterly Journal of Economics 108(3):717-38. Koraiczyk, Robert A. 1994. "Measuring Integration of Developed and Emerging Stock Markets." Northwestern Llniversity, Department of Finance, Evanston, Ill. Processed. - % 1996. "A Measure of Stock Market Integration for Developed and Emerging Markets." The World Bank Economic Revew 10(2):267-89. Koraiczyk, Robert A., and Claude J. Viallet. 1989. "An Empirical Investigation of In- ternational Asset Pricing." Review of Financial Studies 2(4):553-85. Levine, Ross. 1991. "Stock Markets, Growth, and Tax Policy." Journal of Finance 46(4):1445-65. Levine, Ross, and Sara Zervos. 1996. "Stock Market Development and Long-Run Growth." The World Bank Economic Review 10(2):323-39. Obstfeld, Maurice. 1994. "Risk-Taking, Global Diversification, and Growth." Ameri- can Economic Review 85(5, December):1310-29. Pagano, Marco. 1993. "The Flotation of Companies on the Stock Market: A Coordina- tion Failure Model." European Economic Review 37:1101-25. Schmidt, Dana. 1990. Morgan Stanley Capital International (Mscr) Indices. New York: Morgan Stanley. Schwert, G. Wlliam. 1989. "Why Does Stock Market Volatility Change over rune?" Journal of Finance 44(5):1115-53. Shiller, Robert J. 1981. "Do Stock Prices Move Too Much to Be Justified- by Subse- quent Changes in Dividends?" American Economic Review 71:421-36. Solnik, B. E. 1974. "The International Pricing of Risk: An Empirical Investigation of the World Capital Market Structure." Journal of Finance 29:365-78. Demirgl4-Kunt and Levine 321 Stehie, Richard. 1977. "An Empirical Test of the Alternative Hypothesis of National and International Pricing of Risky Assets." journal of Finance 32(2):493-502. Wheatley, Simon. 1988. "Some Tests of International Equity Market Integration."Jour- nal of Financial Economics 21:177-213. THE WORLD BANK ECONOMIC REVIEW, VOL. 10, NO. 2t 323-339 Stock Market Development and Long-Run Growth Ross Levine and Sara Zervos Is the financial system important for economic growth? One line of research argues that it is not; another line stresses the importance of the financial system in mobiliz- ing savings, allocating capital, exerting corporate control, and easing risk manage- ment. Moreover, some tbeories provide a conceptual basis for the belief that larger, more efficient stock markets boost economic growth. This artide examines wbetber there is a strong empirical association between stock market development and long- run economic growtb. Cross-country growtb regressions suggest that the predeter- mined componen. of stock market development is positively and robustly associated with long-run economic growth. To assess whether stock markets are merely burgeoning casinos where more and more players are coming to place bets, or whether stock markets are impor- tandy linked to economic growth, this article reviews a diffuse theoretical litera- tuie and presents new empirical evidence. In terms of theory, a growing litera- ture argues that stock markets provide services that boost economic growth. Greenwood and Smith (forthcoming) show that large stock markets can lower the cost of mobilizing savings and thereby facilitate investment in the most pro- ductive technologies. Bencivenga, Smith, and Starr (1996) and Levine (1991) argue that stock market liquidity-the ability to trade equity easily-is impor- tant for growth. Although many profitable investments require a long-run com- mitment of capital, savers do not like to relinquish control of their savings for long periods. Liquid equity markets ease this tension by providing an asset to savers that they can quickly and inexpensively sell. Simultaneously, firms have permanent access to capital raised through equity issues. Moreover, Kyle (1984) and Holmstrom and Tirole (1993) argue that liquid stock markets can increase incentives for investors to get information about firms and improve corporate governance. Finally, Obstfeld (1994) shows that international risk sharing through internationally integrated stock markets improves resource allocation and can accelerate the rate of economic growth. Ross Levine is with the Policy Research Department at the World Bank, and Sara Zervos is with the Finance Department at Brunl University. This article was originally prepared for the World Bank conference on Stockl Markets, Corporate Fmiance, and Economic Growth, held in Washingon, D.C, February 16-17, 1995. Thc authors aclnowledge helpful advice from Mark Baird, John Boyd, Gerard Caprio, Ash Demirgui-Kunt, William Easterly, Michael Gavin, Robert Korajczyk, Lant Pritchetr Sergio Rebelo, William Schwert, Bruce Smith, Alan Stocknan, David Zervos, and two anonymous referees. 0 1996 The International Bank for Reconstruction and Development THE WORLD BANK 323 324 THe WORLD BANK ECONOMIC REVIEW. VOL. 10 NO.2 Theoretical disagreement exists, however, about the importance of stock markets for economic growth. Mayer (1988) argues that even large stock mar- kets are unimportant sources of corporate finance. Stiglitz (1985, 1994) says that stock market liquidity will not enhance incentives for acquiring informa- tion about firms or exerting corporate governance. Moreover, Devereux and Smith (1994) emphasize that greater risk sharing through internationally inte- grated stock markets can actually reduce saving rates and slow economic growth. Finally, the analyses of Shleifer and Summers (1988) and Morck, Shleifer, and Vishny (1990a, 1990b) suggest that stock market development can hurt eco- nomic growth by easing counterproductive corporate takeovers. We use cross-country regressions to examine the association between stock market development and economic growth. To conduct this investigation, we need measures of stock market development. Theory does not provide a unique concept or measure of stock market development, but it does suggest that stock market size, li'quidity, and integration with world capital markets may affect economic growth. Consequently, we use a conglomerate index of overall stock market development constructed by Demirgiiq-Kunt and Levine (1996).' More specifically, we use pooled cross-country, time-series regressions to evalu- ate the relationship between stock market development and economic growth. Using data on forty-one countries over the period from 1976 to 1993, we split the sample period, so that each country has two observations (data permitting) with data averaged over each subperiod. In the tradition of recent work (Barro 1991), we regress the growth rate of gross domestic product (GDP) per capita on a variety of variables designed to control for initial conditions, political stabil- ity, investment in human capital, and macroeconomic conditions. We then in- clude the conglomerate index of stock market development. Thus, we evaluate whether there is a relationship between economic growth and stock market de- velopment that is independent of other variables associated with economic growth. Our article builds on Atie and Jovanovic's (1993) study of stock market trad- ing and economic growth in two ways. First, we use indexes of stock marker development that combine information on stock market size, trading, and inte- gration. Second, we control for initial conditions and other factors that may affect economic growth in light of the evidence that many cross-country regres- sion results are fragile to changes in the conditioning information set (Levine and Renelt 1992). Thus, we gauge the robustness of the relationship between overall stock market development and economic growth to changes in the con- ditioning information set. We find a strong correlation between overall stock market development and -long-run economnic growth. After controlling for the initial level of GDP per capita, initial investment in human capital, political insta- bility, and measures of monetary, fiscal, and exchange rate policy, stock market development remains positively and significantly correlated with long-run eco- 1. When we ran the regressions using the other aggregate indexes in Demirgii;-Kunt and Levine (1996), the results were similar to those presented in table 1. Levine and ZervCs 325 nomic growth. The results are consistent with theories that imply a positive relationship between stock market development and long-run economic growth. The results are inconsistent with theories that predict no correlation or a nega- tive association between stock market development and economic performance. Cross-country growth regressions suffer from measurement, statistical, and conceptual problems. In terms of measurement problems, country officials some- times define, collect, and measure variables inconsistently across countries. Fur- ther, people with detailed country knowledge frequently find discrepancies be- tween published data and what they know happened. In terms of statistical problems, regression analysis assumes that the observations are drawn from the same population; yet vasdy different countries appear in cross-country regres- sions. Many countries may be sufficiently different to warrant separate analy- ses. Conceptually, we should interpret the coefficients from cross-country re- gressions cautiously. When averaging over long periods, many changes are occurring simultaneously: countries change policies, economies experience busi- ness cycles, and governments rise and fall. Thus, aggregation may blur impor- tant events and differences across countries. Analysts should extend this research by examining the time-series relationship between stock market development and economic growth. Also, cross-country regressions do not resolve issues of causality. Consequently, we should not view the coefficients as elasticities that predict the magnitude of the change in growth following a particular policy reform. Instead, the coefficient estimates and the associated t-statistics should be used to evaluate the strength of the partial correlation between stock market development and economic growth. These measurement, statistical, and conceptual problems, however, should not detract from the benefits that can accrue from cross-country comparisons. Elucidating cross-country empirical regularities between stock market develop- ment and economic growth will influence beliefs about this relationship and shape future theoretical and empirical research. Put differently, beliefs about stock markets and growth not supported by cross-country comparisons wili be viewed more skeptically than those views confirmed by cross-country regres- sions. Section I reviews the theoretical literature on the functioning of stock mar- kets and economic growth. Section II turns to the data and constructs a con- glomerate measure of stock market development. Section mI evaluates the strength of the empirical link between stock market development and long-run economic growth. Section IV summarizes the findings. I. THEioREncAL FRAmEwoRK Is the financial system important for economic growth? One line of research argues that the financial system is unimportant for economic growth; another line stresses the importance of the financial system in mobilizing savings, allo- cating capital, exerting corporate control, and easing risk management. Fur- 326 THE WORLD BANK ECONOMIC REVIEW, VOL 10, NO.2 thermore, some theories provide a conceptual basis for believing that larger, more efficient stock markets boost economic growth. In a recent survey of development economics, Stern (1989) does not mention the role of the financial system in economic growth. At the end of his review, Stern lists various issues that he did not have sufficient space to cover. Finance is not even included in the list of omitted topics. Similarly, a recent coilection of essays by the pioneers of development economics, including three Nobel prize- winners, does not describe the role of the financial system in economic growth (Meier and Seers 1984). Clearly, according to these economists, the financial system plays an inconsequential role in economic development. Furthermore, 1995 Nobel prizewinner Robert Lucas argues that economists frequently exag- gerate the role c'" -ancial factors in economic development (Lucas 1988). Such a view is not limited to the recent past; Robinson (1952) argues that the finan- cial system does not spur economic growth; financial development simply re- sponds to developments in the real sector. Thus, many influential economists give a very- minor role, if any, to the financial system in economic growth.2 In contrast, a prominent line of research stresses the role of the financial system in economic growth. Bagehot (1962), Schumpeter (1932), Cameron and others (1967), Goldsmith (1969), and McKinnon (1973) provide conceptual descriptions of how, and empirical examples of when, the financial system af- fects economic growth. Building on these seminal contributions, Gelb (1989), Ghani (1992), King and Levine (1993a, 1993b), and De Gregorio and Guidotti (199S) show that measures of banking development are strongly correlated with economic growth in a broad cross-section of countries. According to this vein of research, a well-functioning financial system is critical for sustained economic growth Besides evaluating the general importance of the financial system, this article provides empirical evidence regarding the growing debate concerning the spe- cific role of stock markets in economic growth. A burgeoning theoretical litera- ture suggests that the functioning of equity markets affects liquidity, risk diver- sification, acquisition of information about firms, corporate control, and savings mobilization. By altering the quality of these services, the functioning of stock markets can alter the rate of economic growth. Debate exists, however, over the sign of this effect. Specifically, some models suggest that stock market develop- ment has a negative effect on growth, while other models predict a positive relationship between stock market development and economic growth. Stock markets may affect economic activity through their liquidity. Many high-return projects require a long-run commitment of capital. Investors, how- ever, are generally reluctant to relinquish control of their savings for long peri- ods. Therefore, without liquid markets or other financial arrangements that pro- mote liquidity, less investment may occur in the high-return projects. Levine (1991) and Bencivenga, Smith, and Starr (1996) show that stock markets may 2. Many of these references are from Cbandavarkaes (1992) insightful discussion of financial and economic development. Leuine and Zermes 327 arise to provide liquidity: savers have liquid assets--such as equities-while firms have permanent use of the capital raised by issuing equities. Liquid stock mar- kets reduce the downside risk and costs of investing in projects that do not pay off for a long time. With a liquid equity market, the initial investors do not lose access to their savings for the duration of the investment project because they can quickly, cheaply, and confidently sell their stake in the company. Thus, more liquid stock markets ease investment in long-run, potentially more profit- able projects, thereby improving the allocation of capital and enhancing pros- pects for long-term growth. Theory is unclear, however, about the effects of greater liquidity on growth. Bencivenga and Smith (1991) show that by reduc- ing uncertainty, greater liquidity may reduce saving rates enough to slow growth. Risk diversification through internationally integrated stock markets is an- other vehicle by which stock market development may influence economic growth. Saint-Paul (1992), Devereux and Smith (1994), and Obstfeld (1994) demonstrate that stock markets provide a vehicle for diversifying risk. These models also show that greater risk diversification can influence growth by shift- ing investment into higher-return projects. Intuinvely, because projects with high expected returns also tend to be comparatively risky, better risk diversification through internationally integrated stock markets will foster investment in projects with higher returns. Again, however, theory suggests circumstances in which greater risk sharing slows growth. Devereux and Smith (1994) and Obstfeld (1994) show that reduced risk through internationally integrated stock markets can depress saving rates, slow growth, and r educe economic welfare. Stock markets may also promote the acquisition of information about firms (Grossman and Stiglitz 1980; Kyle 1984; Holmstrom and Tirole 1993). In larger and more liquid markets it will be easier for an investor who has gotten infor- mation to trade at posted prices. The investor will thus be able to make money before the information becomes widely available and prices change. The ability to profit from information will stimulate investors to research and monitor firms. Better information about firms will improve resource allocation and spur eco- nomic growth. Opinions differ, however, on the importance of stock markets in stimulating the acquisition of information. Stiglitz (1985, 1994), for example, argues that well-functioning stock markets quickdy reveal information through price changes. Quick public revelation will reduce-not enhance-incentives for expending private resources to obtain information. Thus, theoretical debate still exists on the importance of stock markets in enhancing information. Stock market development may also influence corporate controL. Diamond and Verrecchia (1982) and Jensen and Murphy (1990) show that efficient stock markets help mitigate the principal-agent problem. Efficient stock markets make it easier to tie manager compensation to stock performance. A closer link helps to align the interests of managers and owners. Furthermore, Laffont and Tirole (1988) and Scharfstein (1988) argue that takeover threats induce managers to maximize a firm's equity price. Thus, well-functioning stock markets that ease corporate takeovers can mitigate the principal-agent problem and promote effi- 328 The WORLD BANK ECONOMIC REVIEW. VOL. 10. NO. 2 cient resource allocation and growth. Opinion differs on this issue, too. Stiglitz (1985) argues that outsiders will be reluctant to take over firms because outsid- ers generally have worse information about firms than do owners. Thus, the takeover threat will not be a useful mechanism for exerting corporate control; stock market development, th-erefore, will not improve corporate control sig- nificantly (Stiglitz 1985). Moreover, Shleifer and Vishny (1986) and Bhide (1993) argue that greater stock market development encourages more diffuse owner- ship and this dfffusion of ownership impedes effective corporate governance. Finally, Shleifer and Summers (1988) note that by simplifying takeovers, stock market development can stimulate welfare-reducing changes in ownership and management. In terms of raising capital, Greenwood and Smith (forthcoming) show that large, liquid, and efficient stock markets can ease savings mobilization. By agglomerating savings, stock markets enlarge the set of feasible investment projects. Since some worthy projects require large capital injections and some enjoy economies of scale, stock markets that ease resource mobilization can boost economic efficiency and accelerate long-run growth. Disagreement exists, however, on the importance of stock markets for raising capital. Mayer (1988), for example, argues that new eq- uity issues account for a very small fraction of corporate investment. HI. MEAsuRES oF STocKC MARKET DEVELOPMENT Each theoretical model in the literature focuses on one characteristic of the functioning of stock markets, such as size, liquidity, or integration. Consequently, one research strategy is to evaluate empirically, characteristic by characteristic, the predictions from each individual theoretical model. Although useful, this strategy is model-specific and focuses narrowly on individual characteristics. We take a different approach here, as do Demirgiir-Kunt and Levine (1996) and Demirgiic-Kunt and Maksimovic (1996). We use a multifaceted measure of over- all stock market development that combines the different individual character- istics of the functioning of stock markets. Thus, we prcvide an empirical assess- ment of whether overall stock market development is strongly connected with long-run economic growth. Individual Stock Market Development Indicators We use individual indicators of size, liquidity, and risk diversification. We measure the size of the stock market using the ratio of market capitalization divided by GDP. Market capitalization equals the total value of all listed shares. The assumption underlying th^ use of this variable as an indicator of stock mar- ket development is that the size of the stock market is positively correlated with the ability to mobilize capital and diversify risk. We measure the liquidity of the stock market in two ways. First, we compute the ratio of total value of trades on the major stock exchanges to CDP. This ratio measures the value of equity transactions relative to the size of the economy. Levine and Zeors 329 This liquidity measure complements the measure of stock market size because markets may be large but inactive. Second, we compute the ratio of the total value of trades on the major stock exchanges divided by market capitalization. This ratio, frequently called the turnover ratio, measures the value of equity transactions relative to the size of the equity market. The turnover ratio also complements the measure of stock market size as well as the total value of equity transactions divided by GDP, because markets may be small (compared with the whole economy) but liquid. The liquidity indicators do not directly measure the ease with which agents can buy and sell securities at posted prices. They do, however, measure the degree of trading, compared with the size of both the economy and the market. Since liquidity may significantly influence growth by easing investment in large, long-term projects and by promoting the acquisition of information about firms and managers, we include these two li- quidity measures in our stock market development index. Theory suggests that the ability to diversify risk-by investing in an interna- tionally diversified portfolio of stocks-can influence investment decisions and long-run growth rates (Devereux and Smith 1994; Obstfeld 1994). Barriers to international capital flows-such as taxes, regulatory restrictions, information asymmetries, and sovereign risk-may impede the ability of investors to diver- sify risk internationally. Thus, international capital flow barriers will impede risk diversification, reduce capital market integration, and keep arbitrageurs from equalizing the price of risk internationally. To measure the ability of agents to diversify risk internationally, we use Korajczyk's (1996) estimate of the de- gree of international integration of national stock markets. Korajczyk (1996) uses a multifactor International Arbitrage Pricing Model (IAPM) to measure stock market integration. The IAPM implies that the ex- pected excess return on each asset is linearly related to a linear combination of benchmark portfolios. For the benchmark portfolios, P, Korajczyk (1996) estimates the common factors based on an international portfolio of equities using the asymptotic principal components procedures of Connor and Koraiczyk (1986). Given m assets and T periods, consider the following regression: (1) R;, = a + b,P, + e, i = 1, 2,. . ., m; t = 1, 2,.. .,T where Ri4, is the excess retLrn on asset i in period t above the return on a risk- free asset or zero-beta asse. In perfectly integrated stock markets, the intercept in a regression of any asset's excess return on P should be zero. Specifically, the uPM plus the assumption of perfect integration imply that (2) a2 a . Rejection of the restrictions defined by equation 2 can be defined as rejection of the underlying asset pricing model or rejection of the assumption of market integration. 330 THE WORLD BANK ECONOMIC REVILW, VOL. 10, NO.2 Koraiczyk (1996) refers to ca as the mispricing of asset i relative to the bench- mark portfolio. We interpret estimates of the absolute value of the intercept terms from equation 1 as measures of market integration and the ability of agents to diversify risk internationally. Larger absolute values imply less inte- grated stock markets. To compulte estimates of stock market integration for each national market, we compute the average of the absolute value of ;q across all assets in each country. Simple Indexes of Stock Market Development To measure overall stock market development, we construct an index called STocK by averaging the means-removed values of the market capitalization ra- tio, the total value traded ratio, the turnover ratio, and the IAPM pricing error measure of stock market integration.4 Note that we multiply the absolute value of Korajczyk's (1996) pricing error measure by -1 before constructing the in- dex, STC. Thus, larger (less negative) values imply better stock market devel- opment. The means-removed market capitalization ratio for country i equals the market capitalization ratio for country i minus the mean for all countries, divided by the mean for all countries. Then we take a simple average of -the means-removed market capitalization ratio, the total value traded ratio, the turnover ratio, and the IAPM integration measure to obtain an index of stock market development. More formally, let S(i. j) equal the average value (over the relevant period) of variable j for country i. Let S(j) equal the average value of variablej across all countries. Define the means-removed value of S(i, j) as s(i, I), where (3) S(, j) = [S(1, j) - S)] / S@(). Then sTocKi for country i is (4) STOCK1(i) = Zj s(i, J) where we take the average across all the variables for country i. Financial Depth Gclb (1989), Ghani (1992), King and Levine (1993a, 1993b), and De Gregorio and Guidotti (1995) identify a significant correlation between financial depth and long-run economic growth rates in broad cross-country samples. To mea- sure financial depth, these authors typically use a measure of broad money, such 3. For alternative ways of measuring the ability of agents to diversify risk internationally, see Bekaert and Harvey (199S). 4. sTocK equals the stock market development index, rnxZ of DemirgfiV-Kunt and Levine (1996). All of the indexes discussed by Demirguq-Kunt and Levine (1996) yield similar results in the growth regressions. Thus, we simply report the results using one index We cail this index STOCK instead of INDEXZ for expositional purposes. Also, the WM pricing errors arc available only for twenty-four countries. Since the indexes are means-removed averages of the available indicators, sroc has values for all forty- one countries. For twenty-four countries,srocK aggregates information on size, liquidity, and UPM pricing errors. For the remaining seventeen countries, srocx aggregates information only on size and liquidity. Levine and Zervs 331 as M2, divided by GDP. We use the King and Levine (1993a, 1993b) measure of financial depth, DEPTH, to evaluate whether stock market development is signifi- cantly correlated with growth even after controlling for financial depth. DEPTH is defined as the ratio of liquid liabilities of the financial system to GDP. Liquid liabilities consist of currency held outside the banking system plus demand and interest-bearing liabilities of banks and nonbank financial interme- diaries. Ilf. STOCK MARKET DEVELOPMENT AND LONG-RUN EcONoMic GROWTH This section describes the framework and presents the results for the cross- country growth regressions to analyze the impact of stock market development on long-run economic growth. Cross-Country Growth Regression This section empirically evaluates whetlher the index of stock market devel- opment, STOCK, is strongly linked to long-run economic growth. To conduct this analysis, we use pooled cross-country, time-series growth regressions. We have data on forty-one countries during the period from 1976 to 1993. Each country has two observations, data permitting. The first observation for each country uses data from 1976 to 1985. The second observation uses data from 1986 to 1993. Thus, the dependent variable, GROWTH, is the real per capita growth rate averaged over the relevant period. The structure of our fzgression equation is the following: (5) GROWTH = X + (STOCK) + U where X is a set of control variables, ca is a vector of coefficients on the variables in X, P is the estimated coefficient on s-rocx, and u is an error term.5 The goal of the empirical analysis is to assess the strength of the independent partial correlation between stock market development and economic growth. Consequently, we use a large set of control variables, X, to control for a variety of factors that may be associated with economic growth. X includes initial in- come (the logarithm of initial real per capita GDP), initial education (the loga- rithm of the initial secondary school enrollment rate), a measure of political instability (the number of revolutions and coups), the ratio of government con- sumption expenditures to GDP, the inflation rate, and the black market exchange rate premium. S. Throughout the analysis we use heteroskedasticity-consistent standard errors as developed by White (1980). We also examine the statistical distribution of the error term and check for the importance of outliers. Bekaert and Harvey (1995) find that stock returns are often not normally distributed. Consequently, some readers may have concerns about the distribution of the error termL However, we do not use data on stock returns but an aggregate index of stock market size, liquidity, and integration. For a discussion of the properties of the error tern from cross-secton, time-series regressions involving asset pricing errors, see Bekaerr and Harvey (1995). 332 THE WORLD IRNK ECONOMIC REVIEW, VOL 10. NO.2 We include initial income and initial education because recent theoretical work suggests an important link between long-run growth and the initial per capita levels of physical and human capital (Lucas 1988; Mankiw, Romer, and Weil 1992). We follow Barro (1991), Barro and Sala-i-Martin (1992), and others in using the secondary school enrollment rate and initial income to proxy for the initial levels of per capita human and physical capital. We include political instability because it may be negatively associated with eco- nomic growth. We include a variety of macroeconomic indicators to evaluate the strength of the partial correlation between stock market development indexes and economic growth (Levine and Renelt 1992; Levine and Zervos 1993). We include the gov- ernment consumption ratio and the rate of inflation because the evidence sug- gests a strong connection between macroeconomic policy and economic activ- ity, as shown by Fischer (1993), Easterly and Rebelo (1993), and Bruno and Easterly (1995). We include the black market exchange rate premium because international price distortions may impede economic growth, as suggested by Dollar (1992) and Levine and Zervos (1994). We expect the government con- sumption ratio, the rate of inflation, and the black market exchange rate pre- mium to enter negatively. We use instrumental variables to estimate equation 5 for two reasons. First, instrumental variables will help us examine the relationship between growth and the predetermined component of stock market development. If the prede- termined component of stock market development (as identified by the instru- ments) is positively correlated with economic growth, this correlation will indi- cate that (a) stock market development does not simply follow economic development, and (b) contemporaneous shocks to both stock market develop- ment and economic growth are not the only factors that are driving the results. Thus, we use two-stage least squares to examine whether predetermined stock market development is closely associated with economic growth. In addition, because the LAPM pricing errors are generated regressors, which can lead to in- consistent standard errors, we use two-stage least squares to derive consistent standard errors as suggested by Pagan (1984).' For the instrumental variables, we use the logarithm of initial real per capita GDP, the logarithm of the initial secondary school enrollment rate, political in- stability (the number of revolutions and coups), the initial black market ex- change rate premium, the initial inflation rate, the initial ratio of government 6. The two-stage least squares estimator is consistent. Furthermore, we use White's (1980) heteroskedasticity-consistent standard errors. Although the generalized method of moments (GMM) estimator is sometimes used in pooled cross-section, time-series samples to obtain more effident estimators in the presence of heteroskedasticity and serial correlation in larg samples, rmm does not offer much value added in the current context. As we show, the results are already highly significant, so that a potentially more efficient estimator will only make the t-aistics Larger. Furthermore, we are working with a small sample of only seventy-nine observations, and we do nor have a true rime-series dimension to the data. We consider only two periods because we average the data over long time periods in order to focus on the relationship between stock market development and long&mn growth. Levine and Zervoa 333 spending to GDP, the initial ratio of exports plus imports to GDP, the initial ratio of market capitalization to GDP, the initial ratio of total value traded to GDP, and the initial turnover ratio.7These instruments, except for political instability, are predeterminea. We use these instrumental variables to extract the predetermined component of the government consumption ratio, the rate of inflation, the black market exchange rate premium, and STOCK. We obtained the stock market data from the International Finance Corporation's (lFc's) Emerging Markets Data Base and IMP (various issues). Data on real per capita GDP growth, secondary school enrollment rate, and govcrn- ment consumption ratio, and information on exports and imports are from the World Bank's National Accounts Data Base. We obtained the number of revo- lutions and coups from Barro (1991) and computed per capita GDP from Sum- mers and Heston (1988). Data on the black market exchange rate premium are from Picks Currency Yearbook (various issues) and International Currency Analy- sis, Inc. (various issues). Results Table I summarizes the results on the links between stock market develop- ment and economic growth. Regression 1 presents the regression results when we only include a constant, per capita GDP, the secondary school enrolhment rate, and political instability. Regression 2 includes also the government con- sumption ratio, the rate of inflation, and the black market exchang. rate pre- mium. All of the variables enter with the anticipated signs, but only initial in- come and political instability are consistently significant at the 0.05 level. Regressions 3 and 4 include the index STOCK. There is a significant, posi- tive correlation between the predetermined component of stock market de- velopment and real per capita GDP growth. The relationship between STOCK and growth remains significant at the 0.05 level whether or not we control for the government consumption ratio, rate of inflation, and black market exchange rate premium. Thus, stock market development is positively corre- lated with economic growth even after controlling for other factors associ- ated with long-run growth. As shown by Demirgiiu-Kunt and Levine (1996), stock market development is positively correlated with measures of financial intermediary development. Consequently, to assess the independent empirical link between stock market development and growth, we include the measure of financial depth, DEPTH, in the growth regression. As shown in regression 5, the predetermined component of DEPTH is positively and significantly correlated with long-run growth at the 0.05 significance level when SrocK is excluded. When all of the variables are included together in regression 6, the predetermined component of srocK re- mains positively and significantly correlated with growth. DEPTH, however, 7. The term "initial" refers to variables measured at the start of the estimation period. Since we use pooled cross-country time-series data for the period from 1976 to 1985 and from 1986 to 1993, 'initial" refers to 1976 and 1986 measures. 334 THE WORLD RANK ECONOMIC REVIEW, VOL 10. NO. 2 Table 1. Stock Market Development and Economic Growth, 1976-93 Regression Independent variable 1 2 3 4 S 6 Constant 0.023 0.061 0.041 0.050 0.042 0.047 (1.367) (2.147) (2.772) (1.726) (2.451) (1540) [0.1761 [0.035] [0.0071 (0.089] [0.0171 [0.128] Initial real per -0.011 -0.007 -0.012 -0.010 -0.015 -0.007 capita GDP" (2.256) (1.378) (2.62S) (1.83S) (2.637) (1.430) [0.027] [0.172] [0.011] [0.071] [0.010] [0.157] Secondary 0.023 0.013 0.020 0.019 0.018 0.022 school (2.044) (0.838) (2.095) (1.218) (1.468) (1.431) enrollment rate [0.045] [0.405] [0.040] 10.2271 10.146] 10.1571 Number of -0.019 -0.018 -0.015 -0.016 -0.011 -0.021 revolutions (2.710) (2.362) (23514) (2.337) (13.18) (2.241) and coups [O.008] [0.0211 [0.014] [0.022] [0.133] [0.0281 Ratio of -0.128 -0.090 -0.121 government (1.911) (1.805) (1.671) consumption [0.0601 [0.075] 10.099] expenditures to GDP Inflation rate -0.022 -0.020 -0.03S (2.001) (2.354) (1.844) [0.049] [0.021] [0.069] Black market -0.0002 -0.00001 0.0001 exchange rate (1.179) (0.973) (0.234) premium [0.242] [0.8121 [0.816] DErYHb 0.026 -0.021 (4.349) (0.965) [0.001] [0.338] srTaCc 0.015 0.012 0.020 (5.513) (43503) (2.205) [0.000) [0.000] [0.031] Note: Regression results are from pooled, crot ountry instrumental variables estimation. The dependent variable is average annual growth rate of per capita GD?. Instruments include the constant term; logarithm of initial r:al per capita GDP; logarithm of initial secondary school enrollment rate; the number of revolutions and coups; and inirial valuesof theratio ofgovernmezirconsumption expenditures to GDP, the rate of inflation, the black market exchange rate premium, marker capitalization, total value traded, the tumrnover ratios, and the ratio of international trade to GDP. Standard errors are in parentheses; P-values are in brackets. Data are for forty-one economies for two periods, 1976-85 and 1986-93. Each regression has seventy-nine observations (data are available for only one period for New Zealand, Pakistan, and Turkey). The forty-one economies are Argenina, Australia, Austria, Begiwum, Brazil, Canada, Chile, Colombia, Denmark, Filand, France, Germany, Greece, Hong Kong, India, ndonesia, Israel, Italy, Japan, Jordan, the Republic of Korea, Luxembourg, Malaysia, Mexico, the Netherlands, New Zealand, Nigeria, Norway, Pakistan, the Philippines, PortugaL Singapore, Spain, Sweden, Taiwan (China), Thailand, Turkey, the United Kingdom, the United States, Venezuela, and Zimbabwe. a. Logarithm of initial value 1976 for the 197685 period and 1986 for the 1986-93 period. LeWne and Zervos 33S Table 1. (continued) b. DEPrH is a measure of financial depth, the ratio of liquid liabilides of the financial intermfediaries to GDP. c. STOCK is the stock market development index, the average of means-removed values of the market capitalization, total value traded, turnover ratios, and asset pricing theory (APrr) mispricing indicator. sroac includes data on stock market size and liquidity for all forty-one economies. For the lAIm pricing error indicator, data are included for twenty-four economies: Argentina, Australia, Brazil, Chile, Colombia, Greece, India, Indonesia,Japan, Jordan, the Republic of Korea, Malaysia, Mexico, Nigeria, Pakistn, the Philippines, Portugal, Taiwan (China), Thailand, Turkey, the United Kingdom, the United States, Venezuca, and Zimbabwe. Source: Authors' calculations based on stock market data from the IFC's Emerging Markets Data Base and imF (various issues); data on real per capita GDP growth, the secondary school enrollment rate, and the government consumpion ratio and information on exports and imports from the World Bank's National Accounts Data Base; number of revolutions and coups from Barro (1991); data used to compute per capita GD? from Summers and Heston (1988); and data on the black market exchange rate premium from Picks Curreny Yearbook (various issues) and International Currency Analysis, Inc. (various issues). becomes insignificant.8 The instrumental variable results show that the prede- termined component of stock market development as extracted by the first- stage regression is strongly, positively correlated with growth. The empirical relationship between stock market development and long-ran growth remains strong even after controlling for initial conditions, inflation, the size of the government, the black market exchange rate premium, and the pre- determined component of financial depth. Moreover, the results hold after check- ing for outliers and removing individual countries. As discussed in the introduc- don, measurement, statistical, and conceptual problems plague cross-country growth regressions. Nonetheless, the results suggest a comparatively strong link between the functioning of stock markets and economic growth. IV. SUMMARY This aricle empirically evaluated the relationship between stock market devel- opment and long-run growh. The data suggest that stock market development is positively associated with economic growth. Moreover, the instumental variables procedures indicate a strong connection between the predetermined component of stock market development and long-run economic growth. Although these cross- country growth regressions imply a strong link between stock market development 8. Note that these results do not necessarily conflict with the findings of Gelb (1989), Ghani (1992), King and Levine (1993a, 1993b), and De Gregorio and Guidotti (1995). Fist, these studies of financial depth typically cover eighty countries over thirty years. This article, because of limited data availability on stock market development, covers only forty-one countries over eighteen years. Second, financial depth is a widely available indicator of overal financial sector development. In contrast, the stock market development index measures the functioning of only one part of the financial system. Clearly, researchers should attempt to build models of and develop data on the links between growth and the different components of the financial system: banks (private and public), nonbanks (mutual funds, private pension funds, insurance companies, and others), stock markets, bond markets, and dervatives markets. By adding stock markets to the study of the ties between finance and growth, we see this article as a small building block toward this longer-term objective. 336 THE WORLD BARK ECONOMIC REVIEW. VOL 10. NO.1I and economic growth, the results should be viewed as suggestive partial correla- tions that stimulate additional research rather than as conclusive findings. Much work remains to better understand the relationship between stock market development and economic growth. Careful case studies might better identify the causal interactions between the two. Future research also needs to identify the policies that will ease sound securities market development. REFERENCES The word "processed" describes informally reproduced works that may not be com- monly available through library systems. Atje, Raymond, and Boyan Jovanovic. 1993. "Stock Markets and Development." Eu- ropean Economic Review 37(2/3, April):632-40. Bagehot, Walter. 1962. Lombard Street. Homewood, Ill.: Richard D. Irwin. Barro, Robert J. 1991. "Economic Growth in a Cross Section of Countries." Quarterly Journal of Economics S6(2, May):407-43. Barro, Robert J., and Xavier Sala-i-Martin. 1992. "Convergence." Journal of Poitical Economy 100:223-51. Bekaert, Geert, and Campbell R. Harvey. 1995. "rume-Varying World Marker Integra- tion." Journal of Finance 50(2, June)A403-44. Bencivenga, Valerie R, and Bruce D. Smith. 1991. "Financial Intermediation and En- dogenous Growth." Review of Economics Studies 58(2, April):195-209. Bencivenga, Valerie R., Bruce D. Smith, and Ross M. Starr. 1996. "Equity Markets, Transactions Costs, and Capital Accumulation: An Illustration." The World Bank Economic Review 10(2):241-65. Bhide, Amar. 1993. "The Hidden Costs of Stock Market Liquidity." Journal of Finan- cial Economics 34:31-51. Bruno, Michael, and William Easterly. 1995. "Inflation Crise and Long-Run Growth." World Bank, Policy Research Department, Washington, D.C. Processed. Cameron, Rondo, Olga Crisp, Hugh T. Patrick, and Richard Tilly. 1967. Banking in the Early Stages of Industrialization: A Study in Comparative Economric History. New York: Oxford University Press. Chandavarkar, Anand. 1992. "Of Finance and Development: Neglected and Unsetded Questions." World Development 20(1, January):133-42. Connor, Gregory, and Robert A. Koraiczyk. 1986. "Performance Measurement with the Arbitrage Pricing Theory: A New Framework for Analysis." Journal of Financia Economics 15:373-94. De Gregorio, Jose, and Pablo E Guidotti. 1995. "Fmancial Development and Eco- nomic Growth." World Development 23(3):433-48. Demirgiiq-Kunt, Ash, and Ross Levine. 1996. Stock Market Development and Fman- cial Intermediaries: Stylized Facts." The World Bank Economic Review 10(2):291- 321. Demirgiiq-Kunt, Ash, and Vojislav Maksimovic. 1996. "Stock Market Development and Financing Choices of Firms." The World Bank Economic Review 10(2):341- 69. Levine and Zervosa 337 Devereux, Michael B., and Gregor W. Smith. 1994. 'International-Risk Sharing and Economic Growth." International Economic Review 35(4, August):535-SO. Diamond, Douglas W., and Robert E. Verrecchia. 1982. "Optimal Managerial Con- tracts and Equilibrium Security Prices." Journal of Finance 37(May):275-87. Dollar, David. 1992. "Outward-Oriented Developing Economies Really Do Grow More Rapidly: Evidence from 95 wDcs, 1976-1985.' Economic Development and Cul- tural Cbange 40(April):S23-44. Easterly, William, and Sergio T. Rebelo. 1993. "Fiscal Policy and Economic Growth: An Empirical Investigation.' Journal of Monetary Economics 32(3, Decem- ber):417-S7. Fischer, Stanley. 1993. "The Role of Macroeconomic Factors in Growth." Journal of Monetary Economics 32(3, December)-.485-511. Gedb, Alan H. 1989. "Financial Policies, Growth, and Efficiency." wis 202. World Bank, Country Economics Department, Washington, D.C. Processed. Ghani, Ejaz. 1992. 'How Financial Markets Affect Long-Run Growth: A Cross- Country Study." wis 843. World Bank, Operations Evaluation Department, Wash- ington, D.C. Processed. Goldsmith, Raymond W. 1969. Financial Structure and Development. New Haven, Conn.: Yale University Press. Greenwood, Jeremy, and Bruce Smith. Forthcoming. "Fnancial Markets in Develop- ment and the Development of Fmancial Markets." Journal of Economic Dynamics and Control. Grossman, Sanford, andJoseph E. Stiglitz. 1980. -On the Impossibility of Informationaily Efficient Markets." American Economic Review 70:393-408. Holmstrom, Bengt, and Jean Tirole. 1993. 'Market Liquidity and Performance Moni- woring." Journal of Political Economy 101(4, August):678-709. International Currency Analysis, Inc. Various issues. World Currency Yearbook. Brook- lyn, N.Y. MF (International Monetay Fund). Various issues. International Financial Statistics. Washington, D.C. Jensen, Michael C., and Kevin J. Murphy. 1990. "Performance Pay and Top-Manage- ment Incentives." Journal of Political Economy 98(April):225-64. King, Robert G., and Ross Levine. 1993a. "Fmance, Entrepreneurship, and Growtdt Theory and Evidence." Journal of Monetary Economics 32(3, December): 513-42. 1993b. "Finance and Growtl: Schumpeter Might Be Right." Quarterly Journal of Economics 108(3, August):717-38. Koraiczyk, Robert A. 1996. "A Measure of Stock Market Integration for Developed and Emerging Markets." The World Bank Economic Revietw 10(2):267-89. Kyle, Albert S. 1984. "Market Structure, Information, Futures Markets, and Price For- mation." In Gary G. Storey, Andrew Schmitz, and Alexander H. Sarris, eds., Inter- national Agricultural Trade: Advanced Readings in Price Formation, Market Struc- ture, and Price Instability. Boulder, Colo.: Westview. Laffont, Jean-Jacques, and Jean Tirole. 1988. "Repeated Auctions of Incentive Con- tracts, Investment, and Bidding Parity with an Application to Takeovers." RAND Journal of Economics 19(Winter):516-37. 338 THE WORLD BANK CONOMIC REVIEW, VOL. 10. NO.I Levine, Ross. 1991. "Stock Markets, Growth, and Tax Policy. Journal of Finance 46(4, September):1445-65. Levine, Ross, and David Renel. 1992. 'A Sensitivity Analysis of Cross-Country Growth Regressions." American Economic Review 82(4, Seprember):942-63. Levine, Ross, and Sara Zervos. 1993. "What We Have Learned About Policy and Growth from Cross-Country Regressions." American Economic Review 83(May).426-30. . 1994. 'Looking at the Facts: What We Know about Policy and Growth from Cross-Sectional Analysis." In Luigi Pasinetti and Robert Solow, eds., Economic Growth and the Structure of Long-Term Development. New York: St. Martin's Press. Lucas, Robert E. 1988. "On the Mechanics of Economic Development." Journal of Monetary Economics 22(1, July):3-42. Mankiw, N. Gregory, David Romer, and David Weil. 1992. "A Contribution to the Empirics of Economic Growth." Quarterly Journal of Economics 107:407-37. Mayer, Colin. 1988. "New Issues in Corporate Finance." European Economic Review 32:1167-88. McKinnon, Ronald L 1973. Money and Capital in Economic Development. Washing- ton, D.C.: Brookings Institution. Meier, Gerald M., and Dudley Seers. 1984. Pioneers in Development. New York: Ox- ford University Press. Morck, Randal, Andrei Shleifer, and Robert W. Vishny. 1990a. "Do Managerial Ob- jectives Drive Bad Acquisitions?" Journal of Finance 45(1, March):31-48. 1 1990b. "The Stock Market and Investment: Is the Market a Sideshow?" Brookings Papers on Economic Actvity 2:157-215. Obstfeld, Maurice. 1994. "Risk-Taking, Global Diversification, and Growth." Ameri- can Economic Review 84(5, December):1310-29. Pagan, Adrian. 1984. "Econometric Issues in the Analysis of Regressions with Gener- ated Regressors." International Economic Review 25(1, February).221-47. Picks Currency Yearbook. Various issues. New York: Picks Publishing Corporation. Robinson, Joan. 1952. "The Generalization of the General Theory." In The Rate of Interest and Other Essays. London: Macmillan Press. Saint-Paul, Gilles. 1992. "Technological Choice, Financial Markets and Economic De- velopment." European Economic Reuiew 36(4, May):76341. Scharfstein, David. 1988. "The Disciplinary Role of Takeovers." Review of Economic Studies 55(April):185-99. Schumpeter, Joseph A. 1932. The Tbeory of Economic Deuelopment. Translated by Redvers Opie. Cambridge, Mass.: Harvard University Press. Shleifer, Andrei, and Lawrence Summers. 1988. "Breach of Trust in Hostile Takeovers." In Alan Auerbach, ed., Corporate Takeovers: Causes and Consequences. Chicago: University of Chicago Press. Shleifer, Andrei, and Robert W. Vishny. 1986. "Large Shareholders and Corporate ControL" Journal of Political Economy 94:461-88. Stern, Nicholas. 1989. "The Economics of Development A Survey." Economics Jour- nal 99(September):597-85. Stiglitz, Joseph E. 1985. "Credit Markets and the Control of CapitaLJ.Journal of Money, Credit and Banking 17(2, May):133-52. Levine and Zervos 339 . 1994. 'The Role of the State in Financial Markets." In Proceedings of the World Bank Annual Conference on Development Economics 1993. Washington, D.C.: World Bank. Summers, Robert, and Alan Heston. 1988. "The Penn World Table (Mvark 5): An Ex- panded Set of International Comparisons, 1950-1988." Quarterly Journal of Eco- nomics 106(2):327-68. White, Halbert. 1980. 'A Heteroskedasticity-Consistent Covariance Matrix and Direct Tests for Heteroskedasticity." Econometrica 48:817-38. THE WORLD BANK MCONOMIC REVIEW, VOL. 10, NO. Zt 341-369 Stock Market Development and Financing Choices of Firms Aslh Demirgiiu-Kunt and Vojislav Maksimovic In many developing countries with emerging stock markets, banks are fearful of stock market development because they think that stock markets will reduce the volute of their business. This artide empirically analyzes the effects of stock market development on firms' financing choices using data from thirty developing and in- dustrial countries from 1980 to 1991. The results imply that initial improvements in the functioning of a developing stock market produce a higher debt-equity ratio for firms and thus more business for banks. In stock markets that are already developed, further development leads to a substitution of equity for debt fImancing. By contrast, in developing stock markets, large finns become more levered as the stock market develops, wbereas small finns do not appear to be significantly affected by stock market development. Recent research in corporate finance has identified how asymmetries of infor- mation and imperfections in capital markets affect the firm's ability to raise funds and invest. Although empirical evidence suggests that specific imperfec- tions may significantly affect the firm's financial and investment policies, there has been little work on the effect of the level of development of the financial markets on the firm's policies. In this article we explore this relationship by providing empirical evidence on the association between the financing choices of the firm and the level of development of financial markets in thirty industrial and developing economies from 1980 to 1991. The finance literature suggests that stock markets serve important functions even in those economies in which a well-developed banking sector already exists, the reason being that equity and debt financing are in general not perfect substi- tutes. Equity financing has a key role in the management of the conflicts of inter- est that may arise between different stakeholders in the firm. Stock markets also provide entrepreneurs with liquidity and with opportunities to diversify their portfolios. Stock trading transmits information about the finr's prospects to Aslh Demirgiiq-Kunt is with the Policy Research Department at the World Bank. Vojislav Maksimovic is with the Coilege of Business and Management at the University of Maryland. This article was originally prepared for the World Bank conference on Stock Markets, Corporate Fmance, and Economic Growth, held in Washington, D.C., February 16-17,1995. The authors thank Gerard Caprio, Douglas Diamond, Robert Koraiczyk, Ross Levine, Sheridan Titnan, Dimitri Vittas, and two anonymous referees for useful comments and Qing-Hua Zhao for assistance with the data. 0 1996 The International Bank for Reconstruction and Development /Tm WORLD BANK 341 342 THE WORLD BANK ECONOMIC REVIEW. VOL 10. NO. 2 potential investors and creditors. For example, Allen (1993) contrasts the com- parative advantages of stock markets and financial institutions in processing in- formation about investmnent projects. As a result of the different attributes of debt and equity, the development of markets that facilitate the issuance and trade of equity should be reflected in the financing decisions of individual firms. Although differences in financial systems have been noted in the literature, there have been few attempts to formally model the effects of financial market development on firms' financing choices or on their investment decisions. No- table exceptions are Pagano's (1993) model of the effect of opportunities for diversification on entrepreneurs' portfolio choices; Bencivenga, Smith, and Starr's (1994) analysis of the effect of financial liquidity on technology choice; and Boyd and Smith's (1996) framework analyzing complementarities of debt and equity financing for capital investnents. The empirical work in this area is also sparse. There are empirical studies of firm debt-equity ratios by Titman and Wessels (1988) for the United States, Rajan and Zingales (1994) for a sample of industrial countries, and Demirgiii-Kunt and Maksimovic (1994) for a sample of developing countries. In addition, Mayer (1989) and Singh and others (1992) have looked at corporate financing patterns in industrial and developing coun- tries, respectively. This article empirically explores the effect of financial market development, particularly stock market development, on the financing choices of firms. We compare the relationship between the choice of capital structure and financial market development in our sample. We investigate the extent to which the varia- tion in the aggregate debt-equity ratios within these countries can be explained by (a) the level of development of the country's financial markets, (b) macro- economic factors, (c) the differences between the tax treatment of debt and equity securities, and (d) the firm-specific factors that have been identified in the corporate finance literature as determining financial structure. We find that in general there is a significant positive relationship between bank development and leverage and a negative but insignificant relationship between stock market development and leverage. However, when we break the full sample down into subsamples and control for the other determinants of firm financing, an interesting relationship between leverage and stock market development emerges. In stock markets that are already developed, further de- velopment leads to a substitution of equity for debt financing. By contrast, in developing stock markets, large firms become more levered as the stock market develops, whereas small firms do not appear to be significantly affected by stock market development. Our results have important implications. In many developing countries with emerging stock markets, banks are fearful of stock marker development be- cause they think that stock markets will reduce the volume of their business. Instead, our results imply that initial improvements in the functioning of a de- veloping stock market produce a higher debt-equity ratio for firms and thus more business for banks. These results also suggest that in countries with devel- Demirgfi-Kuntand?Maksimovic 343 oping financial systems, stock markets and banks play different yet complemen- tary roles. Thus policies undertaken to develop stock markets need not affect existing banking systems adversely. Our results are also consistent with the con- clusion of Demirgiiu-Kunt and Levine (1996) that stock market and financial intermediary development proceed simultaneously. The predicted relationship between financial market development and debt- equity ratios is discussed in section I. The sample of countries and the sources of the data are described in section IL The statistical model is described in section Im, and the results are reported in section IV. The conclusions are stated in sec- tion V. 1. FRAMEWORK FOR ANALYSIS Corporate finance theory suggests that corporations optimally structure fi- nancing packages to reduce the economic costs that result from taxes and from imperfections in financial markets. As financial markets develop, the compara- tive significance of different imperfections is likely to change. As a consequence, the issuance of specific securities may become more or less advantageous for certain categories of firms. Thus, there may be a relationship between financial market development and financing choices. In this section we focus on equity market development, in part because it has been most evident during our period of analysis (for a discussion, see Demirgfii-Kunt and Levine 1996). However, there are important spillovers between development of the equity market and development of the banking system. Inadequately Developed Financial Markets Here we consider three classes of imperfections that may result from inad- equately developed financial markets. In the first class, investors and entrepre- neurs face insufficient opportunities for diversification of portfolios. In the sec- ond, the firm lacks financing contracts appropriate for its investment projects. In the third, asymmetries of information between investors and the firm occur because stock markets do not efficiently aggregate information. For each of these imperfections, we identify the effect of financial market development on the firm's financing choices. DIVERSIFICATION BY ENTRERENEUS AND STOCK MARKET LIQUIDITy. In an economy in which equity markets are imperfect, entrepreneurs face the costs of diversifying their portfolios. Outside investors may require a premium to acquire the stock of a firm that is traded on an illiquid market. Moreover, as Pagano (1993) has emphasized, the benefits to the entrepreneur of exchanging the ownership of a stake in a firm for a portfolio of finmcial assets may be limited if the financial market on which these assets are traded does not provide opportunities for diversification. The costs of diversification may induce the entrepreneur to avoid the use of financial markets and, instead, to alter the firm's 344 THE WORLD BANK ECONOMIC REVIEW, VOL. 10. NO. 2 investment and product decisions so as to optimally balance the entrepreneures personal portfolio. This argument parallels the more familiar argument in finance that the firm's financial policies are chosen so as to take advantage of tax shields that the owners cannot exploit on their personal accounts. Here, the argument is that the firm's investment policy may be chosen to achieve a risk-return tradeoff that owners cannot obtain by altering their portfolio investments. The firm's investment policies may be affected in several ways by the owners' inability to diversify optimally in financial markets. First, the firm may diversify into areas in which it does not have a comparative advantage. Second, the firm may invest less than it would if its shares were widely held. Third, it may choose less capital- intensive production technologies that are subject to less long-term risk. OPTIMAL CONTRACTING AND FINANCIAL MARKETS. Conflicts of interest characterize relations between the firm and its customers and suppliers and between different classes of investors in the firm. These conflicts may induce the firm's owners, or managers who represent them, to harm the interests of the other parties. Because such opportunistic behavior can be anticipated, it may make it more difficult for the firm to obtain financing. However, by optimal structuring of the contracts between the firm and outside investors, the owners' incentives to engage in opportunistic behavior can be mitigated. There is a large literature on conflicts of interest between different classes of investors (see Barnea, Haugen, and Senbet 1985; Harris and Raviv 1991; Jensen and Meckling 1976; Myers 1977; and Myers and Majluf 1984). The corporate finance literature has identified several cases in which reliance on outside debt financing increases the incentives of the firm's owners to act opportunistically or to otherwise harm the creditors, customers, and suppliers. Jensen and Meckling (1976) argue that highly levered firns may have an incentive to take on projects that have negative expected net present values and are risky, thereby harming creditors. Myers (1977) shows that firms with significant risky growth opportunities may forgo profitable projects if the resulting increases in value are captured by the firms' creditors. Titman (1984) argues that as high leverage increases the probability of financial distress sufficiently, the firm will enter into contracts that it may be unable to execute. Maksimovic (1988) and Maksimovic and Titman (1991) argue that leverage increases the firm's incentive to renege on value-enhancing implicit contracts with rival firms or with customers. Because debt financing creates incentives to act opportunistically, a highly levered firm may not be able to obtain credit or to exploit fully opportunities for mutually beneficial contracting with customers, rivals, or suppliers. In these cases, issuance of equity would mitigate the incentive problems created by debt fi- nancing. EQurTY MARKr AND INFORMATON AGGREGATION. In addition to their primary role of supplying capital to the economy, equity markets have an important informational role. Equity markets aggregate information about the prospects Demirgfi-Kunt and Maksimovic 34S of the firms whose shares are traded (Grossman 1976). This aggregated information becomes publicly observable by the firm's creditors and investors. Equity markets thereby facilitate the monitoring of the firm by making it more profitable for investors to contribute capital to the firm. This role of financial markets is sufficiently important that many investment funds and mutual funds are prohibited from investing in companies whose stock does not trade on a recognized exchange. In addition to aggregating information, financial markets provide incentives for the investors' acquisition of information. As markets for publicly traded equity increase in size, it becomes profitable for analysts to invest in the acquisition of information about firms (for a formal model, see Grossman and Stiglitz 1980). The resulting increase in the quality of information further facilitates monitoring by creditors. The Effect of Developing an Equity Market Consider an entrepreneurial firm operating in an environment without a func- tioning equity market. The firm is financed by inside equity, trade credit, and bank borrowing. Because we are assuming that an effective equity market does not exist, the firm's initial debt-equity ratio will not be an economic optimum. Hence, once the market is opened we would expect the firm's owners to move away from the initial debt-equity ratio. The initially limited access to equity markets suggests that such a firm is likely to have a suboptimally high debt-equity ratio for its scale of operations. A possible secondary implication of limited access is that the firm may be suboptimally small: it may pass up growth opportunities that would be exploited if a functioning equity market existed. This may occur for the reasons identified above. First, if expansion can only be financed using the entrepreneur's own capital or debt, investment in risky growth opportunities may increase the risks borne by the undiversified entrepreneur enough to make it unattractive. Second, certain projects are optimally financed with equity capital; such projects may not be profitable if financed by debt. Third, in the absence of a public market aggregating information, informational asymmetries may make it too costly to raise capital from outside investors. Now allow an equity market to begin functioning. There will be three direct effects on the firm's debt-equity ratio. First, a substitution effect occurs as out- side equity is substituted for outside debt by firms that had previously been constrained to issue only outside debt. This effect will decrease the firm's debt- equity ratio. Second, outside equity will be substituted for inside equity. This will not affect the firm's debt-equity ratio. Third, the entrepreneur's ability to diversify risks may make expansion more attractive. The effect of such expan- sion on the firm's debt-equity ratio is ambiguous and will depend on the optimal financial structure of the firm. The development of an equity market may also have an indirect effect on the firm's leverage. Equity markets aggregate information that investors possess about firms. This makes it less costly for investors and financial intermediaries 346 l{E WORLD DANK ECONOMIC RLVIEW, VOL. 10, NO.2 to monitor firms. Thus, external equity and debt should become less risky. We would therefore expect to see an increasc in external financing. It is unclear, however, whether external equity or debt would benefit more. To the extent that debt is provided by the product market and by banks, who are probably already well informed, we would expect to see a decrease in leverage as finan- cial markets reduce the costs of monitoring to investors. The net effect of these considerations is that the effect of equity market devel- opment on the debt-equity ratio is ambiguous. All of the above arguments arc conditioned on the hypothesis that equity markets develop relative to the mar- ket for debt. If the debt market develops faster, then the effects may be reversed. II. DESCRPTION OF SAMPLE AND FINANCUUL MARKEr INDICATORS Our sample consists of thirty industrial and developing economies from 1980 to 1991. We have selected these economies because they have a developed or emerging stock market and because data on individual firms' financial struc- tures are available for a sufficiently large number of firms. Table 1 lists all the economies in the sample, together with several indicators of their economic development As an inspection of the table reveals, the sample represents a wide range of economic development: the gross domestic product (GDP) per capita for 1991 ranges from $27,492 for Switzerland to $359 for Pa- kistan. With the exception of Jordan and South Africa, all the economies expe- rienced growth in per capita income during the sample period. Some economies, especially Brazil, Mexico, and Turkey, experienced high rates of inflation in this period. In the absence of a theoryof financial market development, we use empirical indicators to measure the level of development of the equity market and finan- cial intermediaries in each country for each year of the sample. Our first three stock market indicators are the ratio of stock market capitalization to GDP, the ratio of total value of shares traded to GDP, and the ratio of the total value of shares traded to market capitalization. In our sample, data for these three indi- cators are drawn from the International Finance Corporation's (ic's) Emerging Markets Data Base. Our indicators of stock market development have been used in previous stud- ies (for example, Pagano 1993 and Demirguic-Kunt and Levine 1996) and pro- vide intuitive summary statistics for the level- of activity of the stock market and the significance of that activity for each of the economies in the sample. The ratio of stock market capitalization to GDP is a measure of both the stock market's ability to allocate capital to investment projects and its ability to provide signifi- cant opportunities for risk diversification for investors. The ratio of total value of shares traded to GDP and the ratio of total value of shares traded to market capitalization are indicators of market liquidity. The former measures the abil- ity to trade economically significant positions on the stock market, and the lat- ter is an indicator of liquidity of assets traded on the market, not adjusted for Demirgt;-Kunt andMaksimovic 347 Table 1. Economic Development Indicators, Selected Economies Average annual Average annual Life GDP per growth in GDP inflation, expectancy at capita, 1991 per capita, 1980-91 1980-91 birtb, 1991 Economy (U.S. dollars) (percent) (percent) (years) Switzerland 27,492 1.7 3.8 78 Japan 23,584 3.9 1.S 79 Norway 19,664 1.7 S.2 77 Sweden 19,649 1.6 7.4 78 United States 18,972 1.9 4.2 76 Finland 18,046 1.6 6.6 76 France 17,365 1.8 5.7 77 Austria 17,288 2.2 3.6 76 Netherlands 16,479 2.3 1.8 77 Germany 16,439 1.8 2.8 76 Canada 16,098 2.0 4.3 77 Belgium 16,051 2.2 4.2 76 Italy 14,570 2.5 9.5 77 Australia 13,095 1.6 7.0 77 United Kingdom 12,585 2.3 5.8 7S New Zealand 10,643 1.0 10.3 76 Singapore 10,294 4.9 1.9 74 Hong Kong 9,820 5.8 7.5 78 Spain 8,752 3.3 8.9 77 Korea, Rep. of 4,2S9 6.8 S.6 70 Malaysia 2,46S 3.6 1.7 71 South Africa 2,198 -1.0 14A 63 Brazil 2,073 2.1 327.6 66 Mexico 1,801 1.0 66.s 70 Turkey 1,375 3.1 44.7 67 Jordan 1,372 -2.1 1.6 69 Thailand 1,362 7.0 3.7 69 Zimbabwe 630 1.7 12.5 60 India 375 3.3 8.2 60 Pakistan 359 3.9 7.0 59 Note: Economies are in order of 1991 GDP per capita, from highest to lowest. Source: IMF (various issues). the size of the market relative to the economy. We also combine the three indi- cators in an equally weighted index of market development, lNrEx1. Table 2 lists the annual averages from 1980 to 1991 for the indicators of stock market development for each economy. We also report mnrEx1 augmented by Koraiczyk's indicator of securities mispricing (Korajczyk 1995). This indicator measures the extent of mispricing of securities relative to a domestic capital asset pricing model (cApm) for each country.? The augmented index is reported as INDEX2 in table 2. 1. The indicator is sinilar to the indicator estimated in Koraiczyk and Viallet (1989) and is described in that paper and Korajczyk (199S). We also used mispricing indicators obtained from international CAPM and asset pricing theory (Ar) models; these, however, are not reported since the results are not significantly differenL In addition, we have experimented with other indexes discussed in Demirgiu-Kunt and Levine (1996). Again, results were not significantly different. 348 nHU WORLD iMNK lCbNOMIC RhVIEW, VOL, IN, NO.2 Table 2. Indicators of Stock Market Development, Selected Economies, 1980-91 (annual average) Total value of Stock maarket Total value of traded shares capitalization traded shares to market Economy to cGD to GDP capitalization INDE.XP' 1ND21' Hong Kong 1.26 0.51 0.41 0.73 Japan 0.98 O.S3 0.51 0.67 Germany 0.24 0.29 1.23 0.59 United Kingdom 0.86 0.35 0.39 0.53 United States 0.61 0.36 0.58 0.52 Singapore 0.95 0.31 0.31 0.52 Switzerland 0.75 0.31 0.39 0.49 South Africa 1.35 0.07 0.05 0.49 Malaysia 0.88 0.16 0.16 0.40 -0.07 Korea, Rep. of 0.22 0.17 0.69 0.36 -0.21 Thailand 0.21 0.18 0.67 0.35 -0.22 Netherlands 0.46 0.19 0.39 0.35 Australia 0.49 0.15 0.30 0.31 Canada 0.46 0.13 0.29 0.29 Sweden 0.43 0.10 0.2S 0.26 Mexico 0.10 0.05 0.69 0.26 -0.63 Jordan 0.48 0.07 0.14 0.23 -0.24 India 0.07 0.04 0.59 0.23 -0.26 Norway 0.18 0.08 0.42 0.23 Austria 0.08 0.05 0.51 0.22 Brazil 0.11 0.05 0.48 0x21 -0.97 France 0.23 0.08 0.32 0.21 Spain 0.21 0.07 0.31 0.20 New Zealand 0.38 0.06 0.16 0.20 Belgium 0.31 0.04 0.12 0.15 Italy 0.15 0.04 0.23 0.14 Finland 0.17 0.04 0.18 0.13 Zimbabwe 0.10 0.01 0.08 0.06 -0.71 Pakistan 0.04 0.01 0.11 0.06 -0.20 Turkey 0.05 0.01 0.08 0.05 -0.31 Note: Economies are in order of the value for INDEX1, from highest to lowest. a. INDEXl is the average of the ratio of stock market capitalization to CDP, the ratio of total value of traded shares to GDP, and the ratio of total value of traded shares to market capitalization. b. rwmEx2 is the average of the indicators in INDExI and a pricing indicator estimated using a domestic cArm for developing countries. Source: Authors' calculations. In ten of the economies the financia' markets are classified as 'emerging" by IFc (various issues): Brazil, India, Jordan, the Republic of Korea, Malaysia, Mexico, Pakistan, Thailand, Turkey, and Zimbabwe. Interestingly, several emerg- ing markets, such as Korea, Malaysia, and Thailand, have ratios of market capi- talization to GDP that are higher than or comparable to those in some industrial economies, such as France and Germany. The correlation between the ratio of market capitalization to GDP and GDP per capita is only 0.23. Similarly, the ratios DemirgfiJKuntandMaksimovic 349 of total value of shares traded to GDP and total value traded to market capitali- zation are only weakly correlated with per capita CDP (correlation coefficients of 0.23 and 0.34, respectively). The principal indicators we use are measures of activity, rather than mea- sures of the institutional determinants of conditions under which securities are traded. This choice is in part due to the difficulty of quantifying differences in, for example, the regulatory environment that may affect firms' decisions to is- sue equity or debt in the United States and the United Kingdom. However, dif- ferences in the institutions among the ten emerging markets are large enough to be quantified. Table 3 details several important institutional indicators in the emerging markets. As shown in the table, by the end of our sample period the institutions in Brazil, Mexico, Malaysia, and Korea were more developed than those of, for example, Zimbabwe. The principal differences resulted from lower restrictions on dividend and capital repatriation and from higher quality of firm disclosures in the former group. An arithmetic average of the institutional indicators for emerging markets is listed in the fourth column of table 2. We use three empirical indicators to measure the significance of the banking sector in each of the economies in our sample (table 4). Each indicator quanti- fies different components of banks' provision of funds to the private sector in each of the economies. First, the ratio of banks' liquid liabilities (the M3 money supply) to GDP is an indicator of the size of the banking sector in relation to the economy as a whole. This indicator has been used in several studies of the effect of the financial sector on the growth of the economy (King and Levine 1993; Levine and Renelt 1992; Levine and Zervos 1996). Second, the ratio of domes- tic credit to the private sector to GDP measures the role of banks in the provision of longer-term financing to private corporations. A third indicator is the ratio of deposit bank domestic assets to GDP. The first and second indicators are aver- aged to yield an index we call FNDDEx1. The data used to construct the indicators are drawn from uvF (various issues). Although in many developing economies banks are the only significant financial intermediaries, in irdustrial economies significant insurance com- panies, pension funds, and other intermediaries also exist. To gauge the im- portance of financial intermediaries in general on provision of credit, we use the average of the ratio of assets held by deposit banks, nonbank private financial assets, and assets of private insurance and pension companies to GDP (nNEx2). Data on individual firms in Korea, lndia, Mexico, Jordan, Brazil, Turkey, Pakistan, and Zimbabwe come from the IFc's Corporate Finance Data Base. It consists of financial data on the hundred largest firms trading on the stock ex- changes of these countries. For some markets the data are only available for a subperiod, as noted in table A-1. Data on firms in the remaining countries in table 1 come from the Global Vantage Data Base. The number of firms avail- able from the Global Vantage sample is also noted in table A-1. Table 3. Institutional Indicators, Selected Economies, 1992 Restrictionis otn Restrictions on Price. Quiality of foreign foreign Restrictions on earninigs investor Secuirities ibwestors' investors' foreigners' Average informotationt Accountitgf protectioar exchange repatriation of repatriation of domestic institutional Economy publisneda standards lawsh comnnission' dividendsd capital$ investmentsd indicator' Brazil 1 2 2 1 2 2 2 1.71 Mexico 1 2 2 1 2 2 2 1.71 Malaysia 1 2 2 1 2 2 2 1.71 Korea, Rep. of 1 2 2 1 2 2 2 1.71 Thailand 1 1 1 1 2 2 2 1.43 Turkey I 1 1 1 2 2 2 1.43 Pakistan 0 1 1 1 2 2 2 1.29 India 1 2 2 1 1 1 1 1.29 Jordan 0 1 1 1 2 2 2 1.29 Zimbabwe 0 1 1 1 0 0 1 0.57 Note: All data are as of end-1992. Economies are in order of the average institutional indicator (eighth column), from highestto lowest. a. The value I indicates that comprehensive price-earnings information is published regularly and internationally; the value 0 indicates otherwise. b. The values 0, 1, or 2 indicate poor, adequate, or good, respectively. c. The value 1 indicates that the country has a securities exchange commission; the value 0 indicates otherwise. d. The values 0, 1, or 2 indicate that capital flows are restricted, somewhat restricted, or free, respectively. e. Average value of the indicators in the first seven columns. Souirce: Authors' calculations based on data from IFC (various issues). Demirgi4-Kunt and Maksimouic 3S1 Table 4. Indicators of Financial Intermediary Development, Selected Economies, 1980-91 (annual average) Ratio o,' Ratfo of Ratio of banks' liquid domestic credit deposit bank liabilities to private domestic Economy to GDP seeor to GD? assets to GDP FINDEXI' RNDEX2b HongKong 3.64 - - - Switzerland 2.82 3.01 3.12 2.91 3.12 Japan 3.41 2.13 2.45 2.77 1.42 Singapore 2.14 1.67 1.88 1.91 0.95 Jordan 2.00 1.13 1.34 1.36 1.34 Malaysia 1.87 1.24 1.54 1.56 0.67 Netherlands 1.63 1A7 1.89 1.5S 1.89 France 1.40 1.67 1.91 1.53 0.77 Germany 1.30 1.72 2.07 1.51 1.19 Austria 1.66 1.34 2.26 1.50 2.26 United Kingdom 1.31 1.62 1.62 1.47 092 United States 1.32 IAI 0.96 1.37 0.67 Spain 1.37 1.27 1.80 1.32 0.70 Finland 1.02 1.42 1.41 1.22 0.77 Norway 1.20 1.16 1.50 1.18 1.50 Thailand 1.26 0.96 1.19 1.11 0.54 Italy 1.48 0.67 1.05 1.07 0.68 Canada 1.26 0.87 0.95 1.07 0.56 Australia 1.10 0.89 1.01 0.99 0.80 Sweden 0.97 0.92 1.37 0.94 093 South Africa 1.06 0.72 0.76 0.89 0.76 New Zealand 0.97 0.71 0.88 0.84 0.88 Korea, Rep. of 0.77 0.88 0.92 0.83 0.45 Belgium 0.92 0.58 1.14 0.75 1.14 Pakistan 0.79 0.52 0.66 0.66 0.24 India 0.81 0.48 0.63 0.65 0.32 Turkey 0.57 0.36 0.49 0.46 0.25 Zimbabwe 0.77 0.14 0.33 0.46 0.22 Mexico 0.42 0.23 OA 0.32 0.16 Brazil 0.31 0.27 0.45 0.29 0.2 - Nor available. Note: Economies are in order of the value for nNrExl, from highest to lowest. a. FINDEx averages the ratio of banks' liquid liabilitics to GDP and the ratio of domestic credit to the private sector to GDI. b. FIEC2 averages the ratio of deposit bank domestic assets to GDP, the ratio of assets of private nonbanl financial corporations w CDP, and the ratio of assets of private insurance and pension compaaies to GDP. The last two terms are omitted when not available. Source: Authorse calculations based on data from nF (various issues). Research in the United States shows that financial policies are in part deter- mined by firm size. There are economies of scale in issuing securities (Ritter 1987). Larger firms may have more access to financial markets and be followed by a larger number of analysts. Table 5 provides information on the size distri- bution of firms. In each market, firms were ranked according to average size, measured by total assets, over the sample period. The average of each quartile of firm size is reported in the table. 352 THE WORLD BANK ECONOMIC PVIEW, VOL 10, NO.2 Table 5. Average Firm Size, Selected Economies (thousands of U.S. dollars) Smallest Second-smallest Second-largest Largest Economy quartile quartile quartile quartile Sweden 192,704 518,652 1,178,085 3,094,S30 Japan 116,234 256,922 556,993 4,160,906 Italy 85,290 255,479 697,713 4,476,867 Finland 69428 257,053 682,230 1,848,150 Korea, Rep. of 63,000 121,000 178,000 527,000 Spain 38,405 107,061 234,046 965,833 India 28,300 57,200 89,800 286,000 Norway 19,788 65,377 202,275 946,660 France 19,731 74,939 284,118 2,402,375 Switzerland 18,732 76,266 214,584 2,146,239 New Zealand 17,932 55,886 126,671 648,212 Germany 16,899 77,S79 266,325 2,779,747 Netherlands 14,596 69,812 216,311 1,958,973 Hong Kong 13,449 39,891 83,068 607,075 United States 13,484 50,751 137,437 1,220,276 Austria 11,884 40,867 149,432 1,039,347 Brazil 9,900 17,800 30,800 93,900 United Kingdom 9,448 35,739 110,966 1,180,701 Turkey 7,800 17,600 29,200 81,400 Singapore 7,541 26,065 68,452 206,160 South Africa 6.530 40,300 140,793 827,443 Mexico S,900 18,000 44,300 210,600 Zimbabwe S,900 11,600 21,000 64,400 Pakistan 5,700 11,800 17,600 76,500 Canada 5,519 32,984 106,909 629,526 Malaysia 4,886 14,092 29,770 148,555 Jordan 4,100 9,600 17,300 177,800 Belgium 4,092 31,236 144,011 1,242,865 Australia 2,961 18,059 59,657 509,707 Thailand 2,532 7,744 16,841 65,730 Note: The values are average total amsen, for each quartile of firns classified by total assets, during the countWs sample period. Economies are lisred in order of average value of assets in the smallest quarile Source: Global Vantage Data Base and zrcs Corporate Finance Data Basc. Table 5 reveals that firIn sizes differ materially across economies. The aver- age asset value of the largest quartile of Italian firms is approximately $4,500 million, but it is approximately $65 million in Thailand and Zimbabwe. The differences are equally marked in the smallest quartile: the average firm in the lowest quartile in Sweden is seventy-seven times larger than the average firm in the same quartile in Thailand. In fourteen of the markets the average firm in the smallest quartile has assets under $10 million. In some countries the difference in size between the largest and smallest firms is very large. Thus in Belgium the average firm in the largest quardle is approximately 300 times larger than the average firm in the smallest quartile. The large differences in firm sizes in the sample suggest that the devel- Demirgi-Kunt and Maksimouic 353 opment of markets may have different effects on large and small firms in the same market. M. DEERMINANAS oF FINANCIAL STRUCn1UE In order to isolate the contribution of financial market development to the firm's choice of financial structure, we control for other variables that may af- fect the firm's financing choices. We control for three categories of variables: individual firm characteristics, the tax levels in each of the economies in our sample, and macroeconomic factors. As discussed above, the firm's optimal financing mix will depend on the owners' ability to engage in opportunistic behavior at the expense of credi- tors and other parties. This, in turn, will in part depend on the composition of the firm's assets. We control for asset composition by measuring the firm's net fixed assets to total assets and net sales to net fixed assets. Firms with high ratios of fixed assets to total assets and low ratios of sales to fixed assets are predicted to have high long-term and short-term leverage, respec- tively (see Demirgiig-Kunt and Maksimovic 1994). We use two variables as proxies for the firm's requirement for debt fi- nancing: the ratio of earnings to total assets and the ratio of dividends to total assets. The ratio of earnings to total assets is included because several studies, such as Titman and Wessels (1986), have found an inverse relation- ship between profitability and leverage. The ratio of dividends to total assets is included because cash-constrained firms are unlikely to pay out large divi- dends. Our last two firm characteristics measure the firm's nondebt tax shields and its size relative to the economy (total assets to GDP). All other factors being equal, a firm with significant nondebt tax shields is less able to exploit tax shields obtained from debt financing than a firm with smaller, insignificant nondebt tax shields. Size relative to the economy is included as a measure of the firm's access to the financial markets. The sample means for each of these variables for each country are reported in table 6. The firm's choice of debt level will depend in part on the tax treatment of interest income relative to income derived from dividends and capital gains. For each economy and each year, we have calculated the relative tax advantage of debt and equity using data drawn from Coopers & Lybrand (various years) during our sample period (see Demirgiag-Kunt and Maksimovic 1995). Finally, we also control for two macroeconomic variables: the inflation rate and the growth rate of GDP (see table 1). Because debt contracts are typically written in nominal dollars, the rate of inflation may affect the riskiness, in real terms, of debt financing. GDP. growth is included as a measure of the growth opportunities available to firms in the economy. Fmance theory suggests that growth options should not be financed by debt. Thus, we would expect debt financing to be inversely related to per capita GDP growth. Table 6. Firml Characteristics by Economy, Selected Econotmies Net sales Long-tern: Sbort-tern: Total debt Net fixed Dividends to net Total debt to debt to total to total assets to Depreciation to total Earnintgs to fixed Nondebt assets to Economy total eqsuity equity equity total assets to total assets assets total assets' assets tax shieldb GDP Australia 0.563 0.653 1.248 0.385 0.033 0.025 0.064 4.509 -0.008 0.0024 Austria 1.201 1.495 2.696 0.293 0.051 0.017 0.075 3.477 0.012 0.0046 Belgium 0.764 1.259 2.023 0,221 0.039 0.022 0.092 6.153 0.030 0.0087 Brazil 0.139 0.421 0.560 0.640 - 0.002 0.057 1.166 0.017 0.0033 Canada 0.990 0.539 1.600 0.479 0.045 0.014 0.064 3.674 -0.031 0.0018 Finland 3.094 1.856 4.920 0.341 0.042 0.007 0.077 3.977 0.010 0.0154 France 1.417 2.108 3.613 0.234 0.043 0.014 0.094 7.727 0.010 0.0019 Germany 1.479 1.188 2.732 0.321 0.070 0.013 0.087 7.209 -0.007 0.0018 Hong Kong 0.309 0.967 1.322 0.344 0.017 0.057 0.121 6.676 0.020 0.0094 India 0.763 1.937 2.700 0.405 0.038 0.019 0.132 5.614 0.027 0.0006 Italy 1.114 1.954 3.068 0.327 0.041 0.014 0.080 3.287 0.000 0.0049 Japan 0.938 2.726 3.688 0.245 0.026 0.007 0.067 8.373 -0.016 0.0008 Jordan 0.266 0.915 1.181 0.459 - 0.033 0.073 2.979 - 0.0089 Korea, Rep. of 1.057 2.390 3.662 0.371 0.053 0.008 0.100 4.340 0.002 0.0023 Malaysia 0.284 0.639 0.935 0.405 0.021 0.026 0.087 3.264 0.010 0.0032 Mexico 0.375 0.442 0.817 0.579 - - 0.076 1.445 0.013 0.0021 Netherlands 0.710 1.297 2.156 0.334 0.043 0.020 0.094 7.500 0.018 0.0089 New Zealand 0.752 0.776 1.527 0.401 0.030 0.025 0.106 5.067 0.022 0.0224 Norway 3.495 1.880 .5.375 0.433 0.049 0.009 0.092 2.703 -0.005 0.0143 Pakistan 0.595 2.358 2.953 0.384 0.038 0.028 0.115 11.155 0.055 0.0012 Singapore 0.491 0.718 1.232 0.363 0.022 0.018 0.077 5.152 -0.004 0.0104 SouthAfrica 0.597 0.518 1.115 0I535 0.013 0.062 0.206 4.036 0.066 0.0120 Spain 1.086 1.649 2.746 0.446 0.040 0.016 0.095 3.613 0.017 0.0070 Sweden 2.879 2.321 5.552 0.342 0.036 0.011 0.100 4.398 0.021 0.0146 Switzerland 0.878 0.872 1.750 0.304 0.043 0.016 0.073 5.463 -0.081 0.0090 Thailand 0.51B 1.769 2.215 0.380 0.030 0.029 0.129 5.710 0.007 0.0007 Turkey 0.485 1.511 1.996 0,414 - 0.068 0.239 4.240 0.011 0.0011 United Kingdom 0.387 1.065 1.480 0.336 0.032 0.025 0.108 6.447 0.009 0.0010 United States 1.054 0.679 1.791 0.370 0.045 0.016 0.091 6.943 -0.0IS 0.0003 Zimbabwe 0.187 0.615 0.801 - 0.031 0.028 0.131 - 0.033 0.0063 - Not available. Note: The value of each item is calculated as the average of all firms for each economy's sample period. a. Income before interesr and taxes divided by total assets. b. Earnings before taxes minus the ratio of corporate taxes paid to corporate tax rate, deflated by total assets. Soudrcea IFC's Corporate Finance Data Base and Global Vantage Data Base. Demirgi4-Knnt and Maksimovic 35S Our control model for financial structure is yt = a, + , + Symi, + SB ti, + nd,t where y is a nmeasure of leverage, x are the firm-specific characteristics (the ratio of net fixed assets to total assets, the ratio of earnings to total assets, the ratio of net sales to net fixed assets, nondebt tax shields, the ratio of dividends to total assets, and the ratio of total assets to firm size), m are the macroeconomic fac- tors (the growth rate of GDP per capita and the inflation rate), t are the tax variables, d are the time and country dummies, and i and t denote country and time, respectively. In the regressions reported in section IV, this equation is aug- mented by indicators of financial institution and stock market development. Using this model, it is possible to estimate the impact of stock market devel- opment on the financing choices of firms. A negative coefficient estimate for the stock market variable indicates that firm leverage decreases with a marginal development in stock markets, leading the firms to substitute equity for debt. If the coefficient estimate is positive, however, this implies complementarities be- tween debt and equity finance. If the coefficient estimate is not significant, we can conclude that stock market development does not affect firm leverage. IV. RESULTS In this section we present our analysis of the effects of financial market devel- opment on the financing choices of firms. First, we discuss the correlations be- tween firms' capital structures and indicators of financial market development. Second, we discuss financial market development as a determinant of the capi- tal structure of all firms. Third, we characterize the relationship betwee - stock market development and firms' capital structure in developing and developed financial markets. Fourth, we analyze the different effects of stock market de- velopment on large and small firms. Simple Correlations Our primary focus is on the effect of financial market development on the use of equity and debt financing in each of the economies under consider- ation. To this end, we use the average ratio of debt to equity as the depen- dent variable. Specifically for each economy, for each year, and for the firms in our sample, we calculate the average ratios of short-term debt to total equity, long-term debt to total equity, and total debt to total equity. The simple correlations of these three ratios with each other and with indicators of stock market and financial institution development are shown in table 7. By using aggregated data we avoid problems posed for empirical testing by the observed heterogeneity of capital structures adopted by seemingly iden- tical firms (Myers 1984). Such heterogeneity is predicted by theories that focus on macroeconomic (Miller 1977) or industry-level (Maksimovic and Zechner 1991) determinants of financial structure. Table 7. Correlations of Leverage atnd Financial Inidicators Short- Total Total valute Deposit termt Total value of of shares Banks' bank GDP debt to debt to Stock uarket shares traded to liquid dotmestic per Measure total total capitalization traded to market liabilities assets to capita, or index equity equity to GDP GDP capitalizatiotn INDEXI' to GDP CDP FINDEXlb FINDEX2c 1991 Long-term debt 0.531 0.890 -0.191 -0.094 0.OS4 -0.120 -0.106 0.194 0.066 0.162 0.471 to total equity (0.000] (0.0001 (0.002] (0.1411 10.3981 [0.060] [0.0881 (0.0021 [0.2941 [0.0101 [0.0001 Short-term debt 0.846 -0.261 -0.007 0.076 -0.106 0.008 0.130 0.066 -0.036 0.087 to total equity 10.0001 [0.000] [0.9101 (0.2311 [0.097] [0.9021 10.038] [0.295] (0.5711 [0.153] Total debt to -0.246 -0.051 0.079 -0.117 -0.065 0.191 0.074 0.083 0.344 total equity [0.000] [0.42 1 (0.2151 [0.066] [0.2931 [0.0021 10.2391 [0.1881 [0.000] Stock market 0.664 0.051 0.782 0.555 0.365 0.500 0.268 0.228 capitalization [0.000] (0.277] [0.000] [0.000] (0.0001 [0.000] (0.0001 [0.0001 tO GDP Total value of 0.523 0.894 0.592 0.470 0.594 0.311 0.334 shares traded [0.0001 (0.000] [0.0001 [0.0001 (0.000] 10.000] [0.000] tO GDP Total value of 0.648 0.178 0.249 0.270 0.239 0.198 shares traded [0,000] (0.0001 [0.0001 [0.0001] [0.0001 (0.000 to market capitalization INDEXI 0.530 0.462 0.529 0.315 0.292 (0.000 [(0.0001 [0.000] [0.0001 (0.0001 Banks' liquid 0.816 0.951 0.707 0.451 liabilities to GDP °(0.000 [0.000] (0.0001 [0.0001 Deposit bank 0.90S 0.868 0.645 domestic assets [0.000] [0.0001 10.0001 to GDP FINDEX jb 0.742 0.631 [0.0001 10.0001 FINDEX2c 0,578 10.0001 Note: P-values are in brackets. a. INDEX I is the average of the ratio of stock market capitalization to GDP, the ratio of total value of traded shares to GDP, and the ratio of total value of traded shares to market capitalization. b. FINDEX I averages the ratio of banks' liquid liabilities to cDP and the ratio of domestic credit to the private sector to GDP. c. FINDEX2 averages the ratio of deposit bank domestic assets to GDP, the ratio of assets of privatc nonbank financial corporations to GDP, and the ratio of assets of private insurance and pension companies to CDP. The last two terms are omitted when not available. Source: Authors' calculations based on IFC'S Corporate Finance Data Base and Global Vantage Data Base. Demirgiic-Kunt and Maksimovic 3S7 As revealed in table 7, the firms' use of short-term and long-term debt in an economy is positively correlated. The ratio of long-term debt to total equity is negatively correlated with the size of the stock market, positively correlated with the size of the banking sector, and positively correlated with real per capita income. The results for short-term debt to total equity are similar. Thus, a large stock market is associated with reductions in both long-term and short-term debt financing. Interestingly, the level of activity of the stock market (as mea- sured by the ratio of total value of shares traded to market capitalization or the ratio of total value of shares traded to GDP) is not significantly correlated with the long- or short-term debt-to-equity ratio. Similarly, the ratio of banks' liquid liabilities to GDP, which has been used as a measure of the size of the banking sector, is not correlated with the financing choices of firms. Financial Market Development as a Determinant of a Firm's Capital Structure Although the simple correlations between debt and the level of development of the stock market and the banking sector suggest that equity is a substitute for both short-term and long-term debt financing, they do not take into account the other determinants of firms' financing choices. Thus, for example, the observed correlations may result from differences in industry composition, in tax regimes, or in growth rates and macroeconomic factors. To investigate these issues fur- ther, we perform an ordinary least squares (ois) regression of each of the firms' financing variables on the firm characteristics, macroeconomic factors, tax vari- ables, and time and country dummies as well as the financial intermediary and stock market indicators. By controlling for the determinants identified in the literature, this regression is a more conservative test of the relationship between financing choices and indicators of stock market development than the simple correlations reported in table 7. Furthermore, it is likely that some of the time and country dummies may pick up unmeasured differences in financial markets between countries and over time. We have explored this specification in unreported regressions. Our indepen- dent variables explain approximately 80 percent of the variation in the ratios of short-term, long-term, and total debt to total equity. Among the newly added control variables, firm characteristics and country dummies have highly signifi- cant explanatory power. This is consistent with the results of firm-level regres- sions in Demirgiiq-Kunt and Maksimovic (1994), where these variables are dis- cussed in detail. Consistent with corporate finance theory, tax variables are significant in the regressions using the long-term debt ratio. The macroeconomic variables, growth and inflation, jointly have a negative and significant effect on short-term debt and total debt, but not on long-term debt. The growth variable itself has a negative and significant sign in all three equations, indicating that debt financing is indeed inversely related to growth, as theory predicted. Turning to the variables of primary interest, the ratio of deposit bank domes- tic assets to GDP is positively related to firms' debt levels. This relationship is generally significant at the S percent level in the case of long-term debt and at 3S8 THE WORLD DANK ECONOMIC REVIEW, VOL 10, NO.2 the 10 percent level in the case of short-term debt. A stronger association with long-term debt is expected because financial intermediaries are likely to have a comparative advantage in making long-term, loans, whereas short-term financ- ing may be available through trade credit. The stock market indicator INDEX1, or alternative stock market indicators- such as the ratio of stock market capitalization to GDP, the ratio of total value of shares traded to GDP, and the ratio of total value of shares traded to market capitalization-consistently yield negative, but insignificant, coefficients in the regressions explaining the short-term, long-term, and total debt-to-equity ra- tios. This pattern suggests that a relationship between the financing choices of firms and stock market development does exist, but that it may not be captured with the simple linear specification. Next we explore this finding further, and attempt to characterize more fully the interactions between stock market devel- opment and financing choices. Stock Market Development and a Firm's Capital Structure in Developed and Developing Markets Pagano (1993) and others argue that stock markets may play different roles in financing enterprises in economies with small stock markets com- pared with those with well-developed markets. To investigate the possibility that stock markets may have different effects on firms' financing choices as the level of market development varies, we split the sample into subsamples and estimate the effect of stock market development separately in each. We use INDEX1 scores to split the sample into those markets that are developed and those that are developing. The fifteen markets with the largest values for INDEXl in table 2 are classified as developed, and the remaining markets are classified as developing.2 Consistent with the findings of Demirgiig-Kunt and Levine (1996), the well- developed stock markets in developing countries such as Korea, Malaysia, and Thailand belong to the developed group, whereas the relatively underdeveloped markets in some European countries, such as Austria, Finland, and Italy, fall into the developing category. This grouping is superior to a split based on indus- trial compared with developing countries because it takes into account the fact that some markets classified as emerging may already have a significant role in financing the national private corporate sector. The average ratios of stock market capitalization to GDP in the two subsamples over the sample period are shown in figure 1. The difference between the two groups is evident and appears to be constant through time. 2. Results are not sensitive to minor changes in the cutoff. Including Mexico in the developed category and Sweden in the developing category, or dropping both from the sample, yields very similar results. Splits based on other indexes calculated in Deniigiic-Kunt and Levine (1996) are not signicantly different because all those indexes are 99 percent correlated with mDExl. Another intering split could be based on capital market integration. HowevCr, as Bekaert and Harvey (1995) discuss, capital market integration is dosely correlated with market size and liquidity, measures that compose INDExl. Demirgfi-Kunt and Maksimouic 359 Figure 1. Average Ratios of Stock Market Capitalization to GDP in Deeloping and Developed Stock Markets, 1980-91 Average annual stock market capitalization to GDP (percent) 90 _ ~~~~~Developed stodc ma rics 70- 60- 50- 40- 30- 10- ._. - o Developing stock markers O-~~ 1 I I IIII 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 Note The data are annual averges for ffteen economes with developing stock markets (dexco, Jordan, Indla, Norway, Austria, Brazi, Prance, Spain, New Zealar4 Belgium, Italy. Finland, Zimbabwn Pakistan, and Turkey) and fifteen economies vwth developed stock markets (Hong Kong, Japan, Germany, the United Kingdan, the United States, Singapore, Switzerland. South Africa, Malaysia, Korea, Thailand, the Netelnds, Australia, Canada, and Sweden). The split between developing and developed stock markets was based on each economy's value for an Index of stock mrwket development, INDEXI (see table 2). Source Data fom iFs Emerglng Mrkets Datn Base We examine the effect of stock market development on firm financing in the developing and developed market subsamples separately. Table 8 shows the coefficients of the stock market development indicator in our equation explain- ing firms' choice of short-term, long-tern, and total debt to equity in the two subsamples. The basic equation with firm characteristics, macroeconomic fac- tors, and tax variables was estimated separately on each subsample with one indicator of stock market development (the ratio of stock market capitalization to GDP, the ratio of total value of shares traded to GDP, the ratio of total value of shares traded to market capitalization, niDExl, or nEDx2) and one indicator of financial intermediary sector development (the ratio of banks' liquid liabilities to GDP, the ratio of deposit bank domestic assets to GDP, FINEX1, or INDEX2). Table 8 shows an interesting contrast between the developed market and developing market subsamples. The coefficients of the stock market indica- 360 THE WORLD ANK ECONOMIC REVIEW. VOL 10, NO.2 Table 8. Capital Structure and Stock Market Development in Developed and Developing Stock Markets Dependent variable Total Total value of and financial Stock market value of shares traded intermediary sector capitalization shares traded to market explanatory variable to GDP to GDP capitaliZation INDEXl' INDEX2b Developed stock marketr Short-term debt to total equity Banks' liquid liabilities to GDP -0.23 -0.36 -0.04 -0.27 -D.14 Deposit bank domestic assets to GDP -0.07 -0.48 -0.06 -0.24 -0.16 FENDK1d -0.04 -0.47 -0.07 -0.23 -0.16 FINDEC2' -0.06 -0A8 -0.07 -0.23 -0.17 Long-term debt to total equity Banks' liquid liabilities to GDP -0.43 -0.62* -0.34* -0.78** -0.43 Deposit bank domestic assets to GDP -0.S0* -.81** -037* -0.90*** -453* F1DExKId -.S4* 4.82** -0.37* -0.94*** -0.52* FINDEx2t -0A9* -0.80** -0.37 -0.90*** -053* Total debt to total equity Banks' liquid liabilities to GDP -0.49 -0.93' -0.35 -0.95' -0.52 Deposit bank domestic assets to GDP -0.47 -1.18** -0.38 -1.04*" -0.60 iNDEXld -0.42 -1.21*" -0AO -1.07** -0.62 nINDEx2e -0.44 -1.15** -0.38 -1.02' -0.61 Deeloping stock maetsf Short-term debt to total equity Banks' liquid liabilities to GDP 0.10 0.97 0.06 0.20 0.74"*' Deposit bank domestic assets to GDP 0.16 0.84 0.05 0.17 0.69*** FINDEXld 0.15 0.93 0.07 0.22 0.74*** FINDEXZe 029 1.21 0.08 0.31 0.76*** Long-term debt to total equity Bankes' liquid liabilities to GDP 1.06 2.26 0.03 050 0.99*** Deposit bank domestic asset to GDP 1.04 1.94 -0.01 0.39 0.90** FINDEXId 1.03 2.10 0.02 0.47 0.97*** FINDEK2F 1.27 252 0.04 0.59 0.98** DernirgiJ;-Kutnt and Makslmovic 361 Table 8. (continued) Dependent variable Total Total value of and financial Srock market value of shares traded intermediary sector capitalization sbares traded to market explanatory variable to GDP to GDP capitalization INDEx1 INDEX2h Total debt to total equity Banks' liquid liabilities to GDP 1.40 3.61 0.16 0.93 1.80*** Deposit bank domestic assets to GDP 1.39 3.09 0.09 0.76 1.65*** FINDEXid 1.39 3.39 0.16 0.90 1.77*** FINDEX2' 1.79 4.06 0.18 1.11 1.81*** * Significant at the 10 percent level. Significant at the S percent level. Significant at the 1 percent level. Note: Values in the table are for coefficients on the indicators of stock market development from regressions of short-term, long-term, and total debt-to-equity ratios on finn characteristics (net fixed assets to total assets, earnings to total assets, net sales to net fixed assets, nondebt tax shields, dividends to total assets, and total assets to GDP), macroeconomic factors (inflation and growth in GDP per capita), tax variables, the stock market variables listed in the table, the financial intermediary variables listed in the table, and time and country dummy variables. Only the indicated stock market and financial intermediary variables are included in each regression. The split between developed and developing stock markets is based on INDExl. a. INDEXl is the average of the ratio of stock market capitalization to GDP, the ratio of total value of traded shares to GDP, and the ratio of total value of traded shares to market capitalization. b. INDEx! is the average of the indicators in INDExl and a pricing indicatorestimated using a domestic cALPM for developing countries. c. Hong Kong, Japan, Germany, the United Kingdom, the United States, Singapore, Switzerland, South Africa, Malaysia, Korea, Thailand, the Netherlands, Australia, Canada, and Sweden. Se table 2 for values for NDExl. d. FINDEX1 averages the ratio of banks' liquid liabilities to GDP and the ratio of domestic credit to the private sector to GDP. e. FiNix2 averages the ratio of deposit bank domestic assets to GDP, the ratio of assets of private nonbank financial corporations to GDP, and the ratio of assets of private insurance and pension companies to cDp. The last two terms are omitted when not available. L Mexicojordan, India, Norway, Austria, Brazil, France, Spain, Ncw Zealand, Belgium, Italy, Finland, Zimbabwe, Pakistan, and Turkey. See table 2 for values for INDEX1. Source: Authors' calculations based on IFc's Corporate Finance Data Base and Global Vantage Data Base. tor in the developed market subsample are uniformly negative, bult th. coef- ficients in the developing market subsample are all positive with one excep- tion. These patterns suggest that in economies with more deveoped stock markets, further development of the market leads to a substitution of equity financing for debt financing. This is seen most clearly in the case of long- term debt, where the coefficients are predominantly statistically significant. By contrast, in those economies in which the stock market is developing, further development of the market leads to opportunities for risk sharing and for aggregation of information that allow firms to increase their bor- rowing. 362 HI' WORLD RANK .CONOMIC Itl!VlW, VOI 10, NO.I Differenrces between Large and Small Firms It is likely that the effect of stock market development may be different for large and small firms. In particular, the information aggregation role uf the market is likely to be more significant for large firms that trade often and are followed by many analysts. To test this hypothesis, we formed portfolios con- sisting of the largest and smallest quartiles of firms in each country on the basis of their asset size. (Since our data include mostly the largest firms in the coun- tries, we are really looking at the differences between large and very large firms within each country.) Our basic regression equation was then estimated on four subsamples: the largest quartile of firms in developed and developing stock markets and the smallest quartile of firms in developed and developing stock markets. Table 9 reports the results of the splits according to size. The coefficients of the stock market variable for large firms in the developed stock markets sample are uniformly negative. For the ratio of long-term debt to total equity the coef- ficients are statistically significant at the S percent level when the ratio of stock market capitalization to GDP is used as the indicator of market development. By contrast, the coefficients of the stock market variable for large firms in the de- veloping stock markets sample are uniformly positive and for the most part statistically significant at the 5 percent level for the short-term, long-term, and total debt-to-equity ratios. These findings suggest that for large firms in developed stock markets, fur- ther market development acts to enhance opportunities for substitution of eq- uity for debt financing. By contrast, large firms in developing stock markets take advantage of further stock market development to increase their borrow- ing. The coefficients of the stock market indicator for small firms in developed stock markets are negative. This accords with the results for large firms in the same markets and suggests that small firms are also taking advantage of stock market devefopment by substituting debt for equity financing. Interestingly, the coefficients of small firms in developing markets are also predominantly nega- tive, although not statistically significant. We follow Barro (1991) to obtain a visual representation of the interaction between the financing choices of large firms and stock market development. That is, we subtract from the dependent variables of the regressions (long- and short-term debt-to-equity ratios) for the two subsamples of developed and de- veloping markets all the independent variables multiplied by their estimated coefficients with the exception of the stock market indicator. We use nDEx1 as the stock market indicator and the ratio of deposit bank domestic assets to GDP as the financial intermediary variable, and we estimate the regression using slope and intercept dummies for the developed markets. Figure 2 shows the unex- plained residuals for the short- and long-term debt-to-equity ratios plotted against mDEx1 at the sample means of each variable during the sample period for each Demirgi-Kunt and Maksimovic 363 Table 9. Capital Structure and Stock Market Development: Developed and Developing Stock Markets, by Small and Large Firms Dependent variable and Smallest quartile of flrns Largest quartile of flrmos financial intermediary Stock market Stock market sector explanatory capitalization capitalization variable to GDP INDEXlb INDEX2c to GDP INDEX1b iNDEx2r Developed stock marketsd Short-term debt to total equity Banks' liquid liabilities to GDP -0.32 -0.64 -0A1 -0.39 -0.59 -0.44 Deposit bank domestic assets to GDP -0.43 -0.t!5 -0.51 -0.14 -0.28 -0.27 FINDEX1? -0.47 -0.74 -0.53 -0.25 -0.39 -0.31 PINDEX2' -0.44 -0.79* -0.56 -0.17 -0.26 -0.26 Long-term dcbt to total equity Banks' liquid liabilities tO GDP -0.68 -1.28 -0.84 -0.39** -0.29 -0.03 Deposit bank domestic assets to GDP -0.85 -1.72 -1.19 -0.39** -0.28 -0.03 w!SDEXle -1.22 -1.92' -1.25 -.45** -0.31 -0.04 FNDDC2F -0.85 -1.67 -1.15 -0.39** -0.28 -0.03 Total debt to total equity Banks".iquid liabilities to GDP -0.71 -2.17 -1.29 -0.68 -0.83 -0.48 Deposit bank domestic assets to GDP -1.03 -2.67* -1.71 -0.62 -0.66 -0.35 RNDEX1e -1.44 -2.99*> -1.85 -0.71 -0.76 -OA1 FiNDEx2l -1.02 -2.63* -1.71 -0.65 -0.63 -0.34 Developing stoc marketsg Short-tern debt to total equity Banks liquid liabilities LoGDP -1.11 -1.02 -0.22 0.47 1.64"w 1.23*** Deposit bank domestic assets to GDP -1.07 -1.05 -0.24 0.02 1A9** 1.00**> FINDEXl -1.13 -1.06 -0.24 0.26 1.58*' 1.18**> iNDEXf -1.17 -1.05 -0.21 0.74 1.77** 1.19*0* Long-term debt to total equity Banks' liquid liabilities to GDP -0.34 -0.26 023 2.85* 3.06** 2.17*** Deposit bank domestic asset to GDP -0.34 -0.30 0.19 1.80 2.70* 1.72** FINDEXle -0.39 -0.30 0.21 2.27 2.88* 2.0S*** FINDEX2f -0.43 -0.28 025 3.10** 3.23*- 2.16*** (Table continues on the following page.) 364 THE WORLD BANK ECONOMIC REVIEW. VOL iS, NO.2 Table 9. (continued) Dependent variable and Smallest quartile of firms, Largest quartile Of firms financial intermediary Stock moarket Stock maxrket sector explanatory capitalization capitalization variable to GDP INDEX1I INDEX2' to GDP INVEXlb INDEX2' Total debt to total equity Banks' liquid liabilities toGDP -1.SO -1.26 0.00 3.98* 4.92** 357*** Deposit bank domestic assets to GDP -1.38 -1.31 -0.06 2.13 4.30** 2.77*0* FINDEX1 -1.52 -1.32 -0.04 3.10 4.65** 3.38** FINDEx2' -1.57 -1.28 0.05 4.48** 5.24** 3.54*** Significant at the 10 percent level. * Significant at the 1 percent level. * Significant at the 1 percent level. Note: Values in the table are for coefficients on the indicators of stock market developmnent from regressions of short-term, long-term, and total debt-to-equity ratios on firm characteristics (net fixed assets to total assets, earnings to total assets, net sales to net fixed assets, nondebt tax shields, dividends to total assets, and total assets to GDP), macroeconomic factors (inflation and growth in GDP per capita), tax variables, the stock market variables listed in the table, the financial intermediary variables listed in the table, and time and country dummy variables. Only the indicated stock market and financial intermediary variables are included in each regression. The split between developed and developing stock markets is based on tNDExl. a. Based on total assets; see table S. b. =ExI is the average of the ratio of stock market capitalization to GDP, the ratio of total value of traded shares to GDP, and the ratio of total value of traded shares to market capitalization. c. INDEX2 is the average of the indicators in NDExl and a pricing indicator estimated using a domestic CAPM for developing countries. d. Hong Kong, Japan, Germany, the United Kingdom, the United Srates, Singapore, Switzerland, South Africa, Malaysia, Korca, Thailand, the Netherlands, Australia, Canada, and Sweden. See table 2 for ralues for INDExl. e. FmDExi averages the ratio of banks' liquid liabilities to GDP and the ratio of domestic ccedit to the private sector to GDP. J. rmnc2 averages the ratio of deposit bank domestic assets to GDP, the ratio of assets of pnrvate nonbank financial corporations to GDP, and the ratio of assets f private insurance and pension companies to GDP. The last two terms are omitted when not available. g. Mexico,Jordan, India, Norway, Austria, BraziL France, Spain, New Zealand, Belgium, Italy, Finland, Zimbabwe, Pakistan, and Turkey. See table 2 for values horzNDEx1. Source: Authors' calculations based on IFC's Corporate Finance Data Base and Global Vantage Data Bas. economy. Moving to the right along the x-axis corresponds to greater stock market development. The visual evidence is striking. It suggests that for economies with developing stock markets, debt-equity ratios of large firms increase with the development of the stock market. For large firms in economies with more developed markets, firther development is associated with lower debt-to-equity ratios. Taken together, the results sug :st that further development of stock mar- kets may affect firms differently in economies where the markets already play a significant role than in those where they do not. If stock markets are already significant, further development leads to a substitution of equity financing for debt. However, in economies where stock markets are too small to have a sig- Demirgfi-KuntandMaksimoviic 36S Figure 2. Leverage and Stock Market Development Short-Term Debt Ratio of short-term debt to tomal equicy 0.8 - 0.6 - 0.4 0.2 0.0 - I I I -0.2 - -0.4 - 0 0.1 0.2 0.3 0.4 0.5 o.6 0.7 Long-Term Debt 0.6 0.4 0.2-_ -0.2 -0.4 - 0 0.1 0.2 0.3 0.4 0.5 o.6 0.7 Level of stock market development, iNDEXi Nohe Vatues are the unexplained residuals for short- and long-term debt-oequity mrtos plated against INDEXI at the sumple means of each variable during the sample period for each economy. MEX1 is the average of the ratio of stock market capimlization to GDP, the ratio of tal value of traded shares to GDP, and the ratio of total value of shares traded to makcet capitalization. See table 2 for the economies induded in the sample and for values for iNDEXI. Residuals are from regremssions of the debt-to-equity atios on INDEl, the raio of deposit bank domestic assets to GDP, firm charactertics, mcroeconomnic and tax variables, and slope and intercept dummy variables for the developed markes. Sobwue Authors! calculations. 366 TrHE WORLD DANK ECONOMIC REVIEW. VOL 10. NO. 2 nificant role in the economy, as measured by our indicators, development per- mits large firms to increase their leverage. One possible explanation of these results is that, at early stages of stock mar- ket development, improven ments in information quality, monitoring, and corpo- rate control may be large enough to induce creditors to lend more. For these firms, debt and equity finance are complementary. This information aggrega- tion role of the market may be especially important for large firms because their stocks are traded more often and are followed by many analysts. Small firms may not benefit as much from stock market development, at least initially, be- cause they may find their access limited by high issuance costs. Even those small firms that list may not have their stock traded as often because it may be more costly for traders to acquire information about their prospects. V. CONCLUSIONS This article has empirically explored the effect of financial market develop- ment, particularly stock market development, on the financing choices of firms. We used aggregated firm-level data for a sample of thirty countries from 1980 to 1991. We measured stock market development by the ratio of market capi- talization to GDP, the ratio of total value traded to GDP, and the ratio of total value of shares traded to market capitalization. Taking all the countries in the sample together, we find that there is a statis- tically significant negative correlation between stock market development, as measured by market capitalization to GDP, and the ratios of both long-term and short-term debt to total equity of firms. There is also a statistically significant positive relationship between the size of the banking sector and leverage. The negative linear relationship between leverage and stock market develop- ment loses statistical significance when we control for variables that have been identified in the corporate finance literature as determining firns' financial struc- tures. However, when we break the full sample down into subsamples, an inter- esting pattern emerges. In developed markets, funther development leads to a substitution of equity for debt financing, especially for long-term debt. In devel- oping markets, large firms become more levered as the stock market develops, but small firms do not appear to be significantly affected by market develop- mcItn. These findings suggest that the development of a stock market initially af- fects directly the financial policies of only the largest firms. This may be because diversification of ownership and the aggregation of information provided by the development of stock markets initially benefits the larger firms more because of the need to spread fixed issuance costs and traders' costs of information acqui- sition. Moreover, these firms increase leverage. Thus, initially at least, an im- portant role of the stock market is to aggregate information and thereby induce lenders to extend credit to firms whose stock is traded. Demirgfi4-KuntandMakdmouic 367 Table A-1. Number of Finns and Sample Period, Selected Economies Economy Number offirmu Sampe period Australia 401 1983-91 Austria 44 1983-91 Belgium 89 1983-91 Brazil 100 1985-91 Canada 494 1983-91 Finland 55 1983-91 France 544 1983-91 Germany 359 1983-91 Hong Kong 173 1983-91 India 100 1980-90 Italy 81 1983-91 Japan 1,104 1983-91 Jordan 38 1980-90 Korea, Rep. of 100 1980-90 Malaysia 143 1983-91 Me3ico 100 1984-91 Netherlands 165 1983-91 New Zealand 41 1983-93 Norway 52 1983-91 Pakistmn 100 1980-88 Singapore 213 1983-91 South Africa 67 1983-91 Spain 116 1983-91 Sweden 68 1983-91 Switzerland 150 1983-91 Thailand 137 1983-91 Turkey 45 1982-90 United Kingdom 1,27S 1983-91 United States 3,247 1983-91 Zimbabwe 48 1980-88 Source: iyc's Corporate Fmance Data Base for developing countries and Global Vantage Data Base for industi countries. 368 THE WNORLD BANK ECONOMIC REVIEW. VOL t0, NO. 2 REFERENCES The word "processed" describes informally reproduced works that may not be com- monly available through library systems. Allen, Franklin. 1993. "Stock Markets and Resource Allocation." In Colin P. Mayer and Xavier Vives, eds., Capital Markets and Financial Intermediation. Cambridge, U.K.: Cambridge University Press. Barmea, Amir, Robert Haugen, and Lemma Senbet. 1985. Agency Problems and Finan- cial Contracting. Englewood Cliffs, NJ.: Prentice-Hall. Barro, R.J. 1991. "Economic Growth in a Cross-Section of Countries." QuarterlyJournal of Economics 106:407-41. Bekaert, Geert, and Campbell Harvey. 1995. "rune-Varying World Market Integra- tion." Journal of Finance 50(2):403-44. Bencivenga, Valerie, B. D. Smith, and R. M. Starr. 1994. "Liquidity of Secondary Capi- tal Markers: Allocative Efficiency and the Maturity Composition of the Capital Stock." Cornell University, Department of Economics, Ithaca, N.Y. Processed. Boyd, John, and Bruce Smith. 1996. "The Coevolution of the Real and Financial Sec- tors in the Growth Process." The World Bank Economic Retiew 10(2):371-96. Coopers & Lybrand. Various years. International Tax Summaries. New York: John Wiley and Sons. Demirgiic-Kunt, Ash, and Ross Levine. 1996. "Stock Market Development and Finan- cial Intermediaries: Stylized Facts." The World Bank Economic Review 10(2)4291- 321. Demirgiir-Kunt, Ash, and Vojislav Maksimovic. 1994. "Capital Structures in Develop- ing Countries: Evidence from Ten Countries." wns 1320. World Bank, Policy Re- search Department, Washington, D.C. Processed. . 1995. "Stock-Market Development and Financial Intermediaries." wrs 1462. World Bank, Policy Research Department, Washington, D.C. Processed. Grossman, Sanford J. 1976. "On the Efficiency of Competitive Stock Markets Where Traders Have Diverse Information." Journal of Finance 31:573-85. Grossman, Sanford J., and Joseph E. Stiglitz. 1980. "On the Impossibility of Informationally Efficient Markets." American Economic Review 70:393-408. Harris, Milton, and Artur Raviv. 1991. "The Theory of Capital Struaure." Journal of Finance 46(1):297-355. IFc (International Finance Corporation). Various issues. Factbook. Washington, D.C. iMP (International Monetary Fund). Various issues. International Financial Statistics. Washington, D.C. Jensen, Michael C., and William Meckding. 1976. "Theory of the Firm: Managerial Behavior, Agency Costs, and Capital Structure." Journal of Financial Economics 3:305-60. King, Robert G., and Ross Levine. 1993. "Finance, Entrepreneurship and Growth: Theory and Evidence." Journal of Monetary Economics 32(3):513-42. Korajczyk, Robert A. 199S. "A Measure of Stock Market Integration." Northwestern University, School of Management, Evanston, Ill. Processed. Korajczyk, Robert A., and Claude Viallet. 1989. "An Empirical Investigation of Inter- national Asset Pricing." Revietw of Financial Studies 2:55345. DemiygiJS-Kunt and Maksimouic 369 Levine, Ross, and David Renelt. 1992. 'A Sensitivity Analysis of Cross-Country Growth Regressions." American Economic Review 82(4, September):942-63. Levine, Ross, and Sara Zervos. 1996. "Stock Market Development and Long-Run Growth." The World Bank Economic Review 10(2):323-39. Maksimovic, Vojislav. 1988. "Capital Structure in Repeated Oligopolies." RANDJour- nal of Economics 19:389407. Maksimovic, Vojislav, and Sheridan Titman. 1991. "Financial Policy and Reputation for Product Quality." Review of Financial Studies 4:175-200. Maksimovic, Vojislav, and Josef Zechner. 1991. "Debt, Agency Costs, and Industry Equilibrium." Journal of Finance 46:1619 43. Mayer, Colin P. 1989. "Myths of the West." wis 301. World Bank, Policy Research Department, Washington, D.C. Processed. Miller, M. H. 1977. "Debt and Taxes." Journal of Finance 32:261-75. Myers, S. C. 1977. 'Determinants of Corporate Borrowing." Journal of Financial Eco- nomics 5:147-75. - ~. 1984. 'The Capital Structure Puzzle." Journal of Finance 39:575-92. Myers, S. C., and N. S. Majluf. 1984. "Corporate Financing and Investment Decisions When Firms Have Information That Investors Do Not Have." Journal of Financial Economics 13:187-221. Pagano, Marco. 1993. "The Flotation of Companies on the Stock Market: A Coordina- don Failure Model." European Economic Review 37:1101-25. Rajan, Raghuram, and Luigi Zingales. 1994. "Is There an Optimal Capital Structure? Evidence from International Data." University of Chicago, Graduate School of Busi- ness, Chicago. Processed. Ritter, Jay. 1987. "The Costs of Going Public." Journal of Financial Economics 19(2):269-81. Singh, Ajit, Javed Hamid, Bahram Salimi, and Yoichi Nakano. 1992. "Corporate Fi- nancial Structures in Developing Countries." IFc Technical Paper 1. International Finance Corporation, Economics Department, Washington, D.C. Processed. Titman, Sheridan. 1984. "The Effect of Capital Structure on a Firm's Liquidation Deci- sion." Jounal of Financial Economics 13:137-51. Titman, Sheridan, and Roberto Wessels. 1988. "The Determinants of Capital Structure Choice." Journal of Finance 43:1-19. THE WORLD BANK ECONOMIC REVIEW. VOL. 10. NO. 2: 371-396 The Coevolution of the Real and Financial Sectors in the Growth Process John Boyd and Bruce Smith The role of debt and equity changes over time and with the level of development. What are these changes, and why should they systematically occur across different countries and time periods? This article characterizes financial innovation as a dynamic process that both influences and is influenced by the deuelopment of the real sector. It focuses on the emergence and development of equity markets, using a model that allows for growth and for capital accumulation that is financed externally throughl a combination of debt and equity. As an economy develops, the aggregate ratio of debt to equity will generally fall; yet, debt and equity remain complementary sources for the financing :f capital investments. The results suggest how various government policy actions migbt affect capital accumulation and equity market activity. Economists have long believed that financial markets and institutions are impor- tant factors in determining "the wealth of nations." In the last five years or so, there has been a veritabkc renaissance of interest in this topic. This growing, recent literature has produced several important new reasons to think that finan- cial institutions matter very much in the development process. Most of this re- search, however, considers economies with a very limited class of financial mar- kets, and often all investment is by assumption entirely financed either with debt or with equity. Debt and equity markets are therefore not active simultaneously. Moreover, with some exceptions, the level of activity in financial markets does not evolve along with the economy, and the status of various markets is often exogenously imposed. Examples of the kind of literature we have in mind include Bencivenga and Smith (1991), Cooley and Smith (1992), Green- wood and Jovanovic (1990), Levine (1991), and Bencivenga, Smith, and Starr (1994, 1995, 1996). In the first three papers, all investment is financed by bank lending; in the last four papers, all investment is financed either inter- John Boyd is with the Research Department atthe Federal Reserve Bank of Minneapolis and with the Fmance Department in the Carlson School of Managemnent at the University of Minnesota; Bruce Smith is with the Research Department at the Federal Reserve Bank of Minneapolis and with the Department of Economics at Cornell University. This ardcle was originally prepared for the World Bank conference on Stock Markets, Corporate Finance, and Econornic Growth, held in Washington, D.C., February 16-17, 1995. The authors thank Gerald Caprio, Doug Diamond, Ross Levine, and Asli Demirgii-Kunt for helpful comments, and Jason Schmidt for excellent research assistance. The views expressed in this article are those of the authors and not necessarily those of the Federal Reserve Bank of Minneapolis or the Federal Reserve System. D 1996 The International Bank for Reconstruction and Development /lTE wORLD BANK 371 372 THE WORLD ANK ECONOMIC REVw VOL 10, NO.2 nally or by issuing equity. Of this literature, only Greenwood and Jovanovic (1990) has a volume of financial market activity that explicitly evolves over time. The treatment of financial markets in the modem theoretical literature is in sharp contrast with standard accounts of the role of financial markets in the growth process, such as those given by Gurley and Shaw (1955, 1960). They describe financial innovation as a dynamic process that both influences and is influenced by the development of the real sector. In poor, primitive environ- ments, they observe, capital formation is accomplished primarily with entrepre- neurs' savings. As the economy grows, specialized lending institutions, such as banks, emerge and help finance additional capital investment. The loan claims produced in this process are held by the banks tlhemselves and rarely traded. At this stage of development, entrepreneurs remain the only source of equity capi- tal and are the sole residual claimants. With further increases in per capita in- come and wealth, markets for tradable securities, both debt and equity, emerge and complement (but do not replace) bank lending. Gurley and Shaw did not ignore the fact that technological innovation can affect financial arrangements. But technological change was not the centerpiece of their theory of finance and development; rather, the endogenous, dynamical interaction of the financial and real sectors was. Their account of events is well supported by the data. Michie (1987) and Gurley and Shaw (1967) show how equity market development was strongly associated with real development in the history of the United States and the United Kingdom. And Antje and Jovanovic (1992) and Demirgfic-Kunt and Levine (1993, 1994) demonstrate that measures of equity market activity are positively correlated with measures of real activity, across different countries (and that the association is particularly strong for developing economies). Such observations are supported by the data reported in table 1. The objective of this article is to provide a better understanding of the coevo- lution of the real and financial sectors of an economy as it develops. We are particularly interested in whether it is possible to produce a theoretical frame- work that stylistically reproduces the evolution of financial markets in the growth process. We focus here especially on the emergence and development of equity markets, and we put particular emphasis on three main questions: (a) Why does the development of equity markets occur relatively late in the development pro- cess? (b) When this development occurs, what social costs and what social benefits result? and (c) What kinds of govermment policy actions are likely to foster or inhibit the development of equity markets? These questions are clearly of cen- tral importance in the context of economic development as Levine and Zervos (1995) demonstrate, measures of equity market activity are very strongly corre- lated with measures of real economic performance. Since our focus here is on the interactions among real development, equity mar- ket development, and the financing of capital investment, we must pose these ques- tions in an environment that allows for growth and capital accumulation. In addi- BoydandSmith 373 Table 1. Average Market Capitalization, Selected Economies, 1980-91 Economy Percentage of GDP Australia 0.49 Austria 0.08 Belgium 0.31 Brazil 0.11 Canada 0.46 Finland 0.17 France 0.23 Germany 0.24 Hong Kong 1.26 India 0.07 Italy 0.15 Japan 0.98 Jordan 0.48 Korea, Republic of 0.22 Malaysia 0.88 Mexico 0.10 Netherlands 0.46 New Zealand 0.38 Norway 0.18 Pakistan 0.04 Singapore 0.95 Spain 0.21 South Africa 1.35 Sweden 0.43 Switzerland 0.75 Thailand 0.21 Turkey 0.05 United Kingdom 0.86 United States 0.61 Zimbabwe 0.10 Source: Demiriq-Kunt and Maksimovic (1995). tion, we need a model in which capital accumulation is financed externally through a combination of debt and equity, and in which the level of development of debt and equity markets itself affects investment behavior. Such a model is developed and formally analyzed in Boyd and Smith (1994c, 199S); here we sketch the main compo- nents and results of their analysis. We then use that analysis as a basis for a discus- sion of how a variety of different government policy activities might affect the devel- opment of equity markets. Section I presents an overview, section II describes the model we use; section m analyzes the contractual relation between borrowers and lenders. Sections IV and V describe, respectively, the nature of the contracts and the evolution of the real and financial sectors; section VI focuses on possible policy implications; and section VII offers conclusions. I. OVERVIEW The composition of the external finance that investors-and firms more generally-obtain represents the solution to a constrained minimization prob- 374 THe WORLD DANK ECONOMIC REVIEW, VOL 1O, NO.2 lem. Given the quantity of external finance required, and given constraints on the availability of certain kinds of finance, entrepreneurs will raise external fund- ing in the lowest-cost way. Of course in practice the cost of external funds depends on a number of factors, including the level of market interest rates, the existence of various taxes, and the existence of various subsidies implicit in things like goverrnent credit or loan guarantee programs. It is these factors that suggest why government policy can be expected to affect the level of activity in equity markets. The issue of what is the lowest-cost way of raising external funding is of necessity a somewhat subtle one, however. Modigliani and Miller, for example, described circumstances under which the composition of a firm's liabilities is completely irrelevant to the cost of obtaining external funding. Obviously, any theory that can successfully address why equity market development and real development are related must find a way of evading this implication of the justly celebrated Modigliani-Miller Theorem. A common formulation that does evade the Modigliani-Miller Theorem, and often delivers a determinate liability structure, is the introduction of so-called bankruptcy costs, which are incurred when entrepreneurs cannot make contrac- tual payments to creditors. Since the contractual payments called for by debt, unlike those called for by equity, are not contingent on firm performance, the composition of a firm's debt and equity affects the probability of bankruptcy and hence expected bankruptcy costs. The composition of a risk-neutral firm's external finance will be chosen to minimize the expected costs of raising funds externally, inclusive of these banluktptcy costs. Under a theory of this type, if the role of debt and equity is changing ovaer time and with the level of development, it must be the case that entrepreneurs will be perceiving corresponding changes in bankruptcy costs. But what are these changes, and why should they occur in a systematic way across different coun- tries and time periods with the process of economic development? An answer to this question requires that we take a stand on the nature of these bankruptcy costs. Fortunately, a well-developed model of the microeconomic foundations of bankruptcy costs exists, and we will exploit that model here. Specifically, we employ a standard costly state verification (csv) model of the type originally developed by Townsend (1979), and subsequently extended by Diamond (1984), Gale and Hellwig (1985), and Williamson (1986). In this model, external credi- tors can observe some component of a firm's returs only by bearing a (fixed) cost. So long as the firm honors its contractual commitments, there is no need to monitor the firm's returns, and the external creditors do not incur monitoring costs. By contrast, when the firm cannot honor its commitments, extenal credi- tors must verify the firm's returns, and therefore incur these costs. In the most standard csv environments, the entire set of firm returns is unob- servable (Diamond 1984; Gale and Hellwig 1985; Wiliamson 1986). This fact has a strong implication. Since equity claims promise payments based on firm BoydandSmith 375 performance, and since firm performance is costly to observe, the use of equity would require that excessive verification costs be incurred. It is therefore opti- mal for firms to be 100 percent debt financed. Indeed, debt claims make con- tractual payments contingent on firm performance only in the event of a bank- ruptcy, and therefore the use of debt minimizes expected verification costs. The csv model cannot be adequate for our purposes because it predicts that equity markets will never be active. However, the csv environment does provide a simple and tractable explicit model of bankruptcy costs. Therefore, we pursue the implications of a csv model altered in one basic way. Whereas the conven- tional csv literature gives investors access to only a single investment technol- ogy with unobservable returns, we give investors access to two technologies. In particular, we assume that physical capital can be produced using either of two technologies. One, the unobservable-return technology, yields a return that is freely observable only by the initiating investor, and hence is subject to the csv problem. The other, the observable-return technology, yields a return that is freely observable to all agents. Under the assumption that the expected amount of capital produced (exclusive of verification costs) by the former technology exceeds that produced by the latter, agents undertaking capital investments face a tradeoff. The technology with the unobservable return is intrinsically more productive, but it is also associated with larger bankruptcy costs.' Our analysis allows capital producers to make a choice as to how heavily they will use each technology; their choice will depend on the relative expected returns on the two technologies and on the perceived costs of state verification. When the perceived costs of verification are low, the unobservable-return technology will be par- ticularly attractive, and its higher expected return implies that-when costs are perceived to be low enough-it will be used exclusively. In this event our model mimnics the conventional csv environment, and expected verification costs are minimized by having firms be 100 percent debt financed. Equity markets will not be active. As the perceived costs of state verification rise, it will eventually become economical for capital producers to take actions to reduce the expected costs of state verification. The action they can take is to use more heavily the observ- able-return investment technology. The higher perceived verification costs are, the more heavily the observable-return technology will be used. Because the return on this technology is observable, it is not costly to issue claims that bear payments contingent on some aspects of firm performance, and hence some use of equity can be optimal. At the same time, so long as the unobservable-return technology is in use at all, the expected verification costs associated with it must be minimized. To do so, firms will continue to issue some debt. Indeed, as we demonstrate formally later, the minimization of expected verification costs will dictate that firms issue a determinate amount of debt and equity. Moreover, the 1. If the expected amount of capital produced by the observable-return technology is higher than that produced by the unobservable-return technology, then the latter technology is dominated and will never be used. This would leave us with a model where the Modigliani-Miller Theorem applies. 376 THE WORLD ANK ECONOMIC REVIEW. VOL. 10, NO.2 issue of some equity-when equity is used-reduces the cost of issuing debt. In this important sense, equity finance is not just a substitute for debt finance, but rather a complement of debt finance. This theoretical result is consistent with many empirical findings, such as those of Demirgii;-Kunt and Levine (1994), Rojas-Suarez and Weisbrod (1994), and Demirgiiq-Kunt and Maksimovic (1995), suggesting the complementary nature of debt and equity market activity in the financing decisions of a firm, and in the general process of development. According to the scenario we have described, the volume of equity market activity mnust increase as an economy develops because capital producers are more actively using the observable-return technology in an effort to reduce the expected costs of state verification. Why should capital producers be expected to use this technology more intensively as an economy becomes more devel- oped? They will do so if and only if the perceived costs of state verification rise along with the development process. In the model we describe later, entrepreneurs perceive relative verification costs that rise during the development process because entrepreneurs are en- gaged in the production of physical capital. As an economy develops, we typi- caily expect the relative price of capital to fall, a proposition that is supported by a wealth of empirical evidence (see, for example, Greenwood, Hercowitz, and Krussell 1995). Suppose that state verification technologies use some com- bination of inputs, possibly capital and labor; the relative price of labor will rise in the development process, and certainly will rise relative to the price of capi- tal. As a consequence, the costs of state verification must rise relative to the value of what entrepreneurs produce as an economy develops. This rise in perceived costs will, as we have argued, induce firms to use the observable- return teclmology more intensively and to raise more of their funds in equity markets. There is good evidence both that developing economies have higher per unit costs of bankruptcy (World Bank 1989) and that their firms rely less heavily on equity markets (Demirgiiu-Kunt and Maksimovic 1995; Levine and Zervos 1995). Although we think that this scenario is itself highly plausible, the intuition underlying it depends on few details of the model we present. Indeed, an intu- itively simpler (but formally more complex) scenario is that development is asso- ciated with the use of increasingly more specialized and complex technologies. Thus as economies develop, external monitoring becomes more difficult, other tiings being equal; as a result, firms take more actions to economize on monitor- ing costs. This scenario will generate the same conclusions we have just described. Our analysis also suggests that financial market frictions will be less severe in industrial economies than in developing economies. This appears to be consis- tent with observation (World Bank 1989), and indeed it is often argued that developing economies are less developed because their financial market fric- tions are more severe than those of their more developed counterparts (McKinnon 1973; Shaw 1973). In all our examples, economies that are more developed lose fewer resources (per unit of funding) because of the presence of intermediation Boyd and Smitb 377 costs than do less mature economies (see section V and Boyd and Smith 1995), a social benefit yielded by the development of equity markets and an endogenous outcome. Financial market frictions become less severe over time as a natural consequence of development. In this sense, the cvolution of finiancial markets in the development process does tend to provide an cconomy with a more efficiently functioning set of capital markets. Finally, our results provide several suggestions about the consequences of various government policies, both for the level of rcal activity and for the level of activity in equity markets. After pursuing the formal analysis, we offer an informal discussion of how various policy actions might affect capital accumu- lation and equity market activity. We focus particularly on policies that affect the opportunity costs of external finance, as many government policies do in practice. The analysis of how such policies affect both the financial system and the level of real activity has not previously been formally undertaken. Our re- sults indicate that government policies which lower the opportunity cost of ex- ternal funds should be expected to attenuate equity market development. A par- ticular finding is that high inflation-which acts to reduce real interest rates-will interfere with the development of equity market activity. Again, such a finding is well supported empirically (Choi, Smith, and Boyd 1995). 1I. THE MODEL In this section we lay out a model that formalizes the intuition described in section I. Specifically, we consider an economy populated by a sequence of two- period-lived, overlapping generationis, plus an initial old generation. Each genera- tion has the same large population, and is identical in its composition. In par- ticular, agents in each young generation are divided into two types, which we term borrowers and lenders. All borrowers are identical, ex ante, and borrowers constitute a fraction a e (0,1) of the population. All lenders are also identical, ex ante, and they constitute a fraction 1 - c of the population. Lenders are en- dowed with one unit of labor when young, which they supply inelastically. They are retired when old. Borrowers are cndowed with no labor, but are endowed with access to high-return investment projects, which are described later. All agents, both borrowers and lenders, are risk neutral and care only about old- period consumption; thus, all young-period income is saved. There is a single consumption good at each date, which is produced according to a standard, commonly available constant-returns-to-scale production func- tion with capital and labor as inputs. In particular, if K, is the capital stock at time t, and L, is the labor input at time t, then the production of final goods and services is given by F(K,, Lt). In addition, if kt - KIIL, is the capital-labor ratio at t, then f%k) _ F(k,, 1) is the intensive production function. We maintain standard assumptions on f, that is, f(O) 0 O, f(k) > 0 > f(k) for all k Ž 0, and fsatisfies the usual Inada conditions. For simplicity, we also assume that capital is used in production and then depreciates completely. 378 11i4 WORLD BANK rCONOMIC REVIEW. VOL 10, NO.2 Capital Production Technologies Our assumptions imply that agents who can produce capital (borrowers or entrepreneurs) require external funding in order to do so. Our primary focus is on whether tney raise this funding through debt or equity markets, and on the extent to which each set of markets is used. As argued in section 1, an interesting analysis of this question requires that some type of bankruptcy cost be present. We therefore assume that capital can be produced at each date using one or more of the following three technologies: a. There is a commonly available, nonstochastic linear technology whereby one unit of current output invested at t yields r > 0 units of capital at t + 1. In addition, there are two stochastic linear technologies that convert current output into future capital. b. Technology o (for observable return) produces y units of capital at t + 1 per unit invested at t, where y is an independent and identically distributed (iid) (across agents and across time periods) random variable, realized at t + 1. We assume that ye(y1, Y2, . . ., yNl, and we let pn - prob(y = yn). Obviously U < pn 5 1 for all n, and Ep, = 1. Finally, we assume that for any investor the amount of capital yielded by investments in technology o is publicly observable at zero cost. c. Technology u (for unobservable return) is assumed to produce w units of capital at t + 1 per unit invested at t, w is a continuous, iid (across investors and periods) random variable with cumulative density function G and probability density furctiong; g is continuously differentiable with support [0,ii]. In addition, the return or. investments in technology u can be observed (by any agent other than the initiating investor) only by bearing a fixed cost of y > 0 units of the current consumption good. Thus, a csv problem arises for investments in technology u. We assume that only borrowers are endowed with access to the investment technologies o and u, and ownership of these investment opportunities cannot be traded. Moreover, we impose an upper bound on the scale at which any borrower can operate the investment technologies. Thus, we let 4 be the invest- ment in technology o, and it be the investment in technology u, by a representa- tive borrower at t, and we let i4 - it + i4. Then each borrower faces the maximum scale of operation constraint i < q, where q is an exogenously given parameter. The assumption of linear capital production technologies, along with the exist- ence of an upper bound on their operation, will make it easy to determine how much external finance each entrepreneur desires. Finally, we define Q, - i4i4 to be the fraction of total investment done in technology o by a representative borrower. Define 9 Z lp,,y, to be the expected gross return (in units of capital) on in- vestments in technology o, and w =- J 'wg(w)dw to be the expected gross return (in units of capital), not inclusive of verification costs, on investments in technol- Boyd and Smith 379 ogy u. We assume that w > 9 > r. Thus, the commonly available technology is rclatively unproductive. It should be clear that, unless w > y holds, the unob- sarvable-return technology will never be employed. Finally, we assume that the initial old agents are endowed with Ko > 0 units of capital per capita. No agents thereafter are endowed with either capital or the consumption good. Trade and Finance Two kinds of trade take place in this economy: capital and labor are rented in competitive factor markets, and funds are transferred from lenders to borrow- ers. The inherited capital stock (dle proceeds of the previous period's invest- ment) and labor are both supplied inelastically. Both factors are demanded by competitive producers and hence are paid their marginal products. Thus, if co, is the real wage rate at time t, and p, is the rental rate for capital at time t, we have (1) Pt = f(k,) (2) co, = fk,) - kf'(k,) =- o(kj). Clearly, ol(kj) > 0 holds, for all k,. We assume throughout that the potential supply of funds by lenders is at least as great as the demand for funds by borrowers. The supply of funds by lenders is (1 - a)o), at t, since lenders save their entire young-period income. The maxi- mum demand for funds by borrowers is aq. Hence we assume that (3) (1 - ccx)(k) > aq for all relevant values of kr When equation 3 holds, any marginal savings must be invested in the commonly available technology, yielding r units of capital per unit invested. Thus, the opportunity cost of funds faced by lenders at t must equal rp,i , since the rental value of the capital obtained at t + 1 is Ptl. m. FUNDING CONTRACTS BErWEEN BORROWERS AND LmlDNtEs A funding contract specifies a quantity of resources that will be transferred to a particular borrower (i,), as well as how these resources will be allocated among technologies o and u. We assume that external investors can observe both i -id the allocation of investments among the two investment technologies. In addi- tion, the contract specifies a set of payments that are contingent on firm perfor- mance. Since the return y is observable, repayments can always be made con- tingent on it. However, since the return w can be observed externally only if monitoring (or verification) occurs, repayments can only meaningfully be made contingent on w if state verification occurs. Hence a contract must specify a set 380 THE WORLD BANK ECONOMIC REVIEW1 VOL 10, NO.2 of states A,(y), in which monitoring will occur, and B,(y), in which monitoring will not occur at t + 1. This set can obviously be conditioned ony. In monitoring states only, repayment can also be made contingent on w. We let R,(w, y) be the promised repayment at t + 1, in units of consumption and per unit borrowed at time t + 1, if monitoring occurs at that date. In addition, we let x,(y) be the promised repayment at t + 1, in units of consumption and per unit borrowed at time t + 1 if monitoring does not occur. Funding contracts are assumed to be announced by borrowers and either ac- cepted or rejected by lenders. In order to avoid rejection, such contracts must satisfy the following three constraints. a. Contracts must be feasible, that is, specify nonnegative consumption levels for borrowers. Thus, a borrower's repayment can never exceed the borrower's total available resources, or (4a) Rt(w, y) S Pt+i [Oty + (1 - Ot)w] (4b) xt(y) < pt.[4Oy + (1 - H)wJ, for all w E B,(y) must hold. Note that Pt.i appears as it does in equation 4a since investment returns are in units of capital, while repayments are in units of consumption. b. Contracts must be incentive compatible, so that borrowers have an incentive to announce truthfully when a monitoring state has occurred. They will do so if and only if repayments are lower in monitoring than in nonmonitoring states, or if and only if (5) R,(w, y) < xj(y), for all w e A,(y). c. Since lenders can always invest in the commonly available technology, the expected repayment-per unit borrowed-must at least equal the opportunity cost of funds (rp1tB) plus expected monitoring costs (which, without loss of generality, we assume are born by lenders). Note that we abstract from stochastic monitoring (see Boyd and Smith 1994b for a defense). Thus, contractual repayments must satisfy the expected return constraint (6) Pn Z,, A ,, Rt(w, yJg(w)dw + M.P. J.EB,(n) Xt(YJg(W)dWJ -Y 4PM JweAty,) g(w)dw > rp+li, Borrowers announce funding contract terms to maximize their own expected utility, subject to the constraints given by equations 4, 5, and 6. The expected utility of a borrower is simply the expected return on the borrower's investments, less the expected repayment inplied by the contract. Thus, borrowers choose contract terms to maximize Boydand Smitb 381 (7) it4{pt,[Ot + (1 - e3t) ;J - Z,p, JWeA#, )Rt(w, y)g(w)dw - IPp 1 Bt(,,,) x,(ynJg(w)dw) subject to equations 4, 5, and 6. This completes the specification of the model environment and the descrip- tion of how agents behave. We now describe the solution to the contracting problem of borrowers as a prelude to an analysis of the evolution of the capital stock, real activity, and the volume of equity market transactions. IV. TIE NATuUR OF EQUILIBRIUM CONTRAcTs Boyd and Smith (1994c) show that equilibrium contracts between borrowers and lenders have the folluwing properties. First, borrowers operate their invest- ment projects at the largest possible scale, so that i, = q. There is no reason for them not to do so, since the fact that 9 > r implies that unused capacity can always be profitably exploited. Second, if the borrower's ex post return [%y + (1 - 06)w]p,+. is no less than the promised payment x,ty), it is feasible for bor- rowers to fully meet their external commitments and to avoid state verification, thereby minimizing verification costs. Third, if the ex post return is less than xt(y), it is not feasible for borrowers to meet their contractual commitments. In this case, providers of external funding monitor the firm, learn the true value of w, and retain the entire value of the firm's output. All of these results are direct analogs of findings in conventional csv models (Diamond 1984; Gale and Heliwig 1985; Williamson 1986). It wiJl now be useful to define a new variable. Let (8) Zm2{ [X((YjIP'+1 - G'yJAI(1 - O). Formally, z,t is the smallest possible realization of w, conditional on y = y,, that enables an entrepreneur to meet contractual commitments and to avoid verifi- cation. In other words, if y = y,, at t, then monitoring occurs if and only if the return on technology u is less than z, In effect, then, the variables z,,, govern the conditional probability (and expected cost) of state verification. If y = yn at t, the probability of monitoring is just G(z,). In addition, since w and y are indepen- dent, if the borrower's optimization problem displays enough concavity, bor- rowers will wish to smooth monitoring costs across realizations of y. Hence Zit = z2 =.. * = zNt = zt holds, and the probability of state verification, G(z,), will be independent of the realization of y. Thus, the variable ; indexes the amount of state verification that is associated with a particular contract. The optimal value of Zt depends on two endogenous factors. First, it depends on the composition of a borrower's investment between technologies o and u; that is, on e=- io/it. Note that this composition is chosen by the borrower. The higher is e, the more investment is done in the observable-return technology, 382 THE WORLD BANK ECONOMIC REVtEW. VOL 10, NO.2 permitting less monitoring to occur, so that higher values of El are associated with lower values of z,. Second, z, depends on the relative cost of monitoring, t1+1 - y/qp,,1. Monitoring uses y units of final goods and services; ylq is this cost relative to the size of individual investment projects. Dividing this quantity by the relative price of capital gives monitoring costs measured in units of capital. Since investment projects yield capital, this puts monitoring costs and invest- ment returns in comparable units. The cost of monitoring is taken as parametric by the borrower, but it is endogenous to the economy. To denote the depen- dence of z, on 9, and on t,t,+ we write ; = z(O9d tt+). z(OA tt,1) must be chosen, given 0, and ttel so that providers of external funding receive the market- expected return (rp,+1) on their investments at t. As we have argued, the probability of monitoring is simply G(z,) = G[z(O,; t4+.)]. Entrepreneurs then want to choose 0, to maximize the expression (9) qpt+1f03' + (1 - O,)i' - r - In particular, qp,+1[0,9 + (1- 9,) wv] is the expected amount of capital produced, valued at the market price of capital; qp,+1r is the expected return that must be offered to obtain q units of funds, and qp,+1 i,+1 G[z(9,; t,+1)] represents the ex- pected costs of state verification. Each borrower chooses 0, to maximize the expression in equation 9. That expression clearly reflects the foliowing tradeoffi higher values of 0, reduce the expected amount of capital produced, but they also reduce the probability of monitoring, G[z(EA,; t,+1)]. The higher is 1, the greater is the value of a reduction in monitoring costs; hence we expect the optimal choice of O, to rise with tt+1. To denote the relation between the choice of 9 and the scaled monitoring cost t, let 88(t) denote the borrower's optimal choice of 9. Boyd and Smith (1994c) describes formal conditions under which the intuition above is valid and, in particular, under which (10) d0*/dt 2 0. The solution to the borroweres problem is completely summarized by the composition of investment [9= 9(t,+], by the amount of monitoring called for [Zt = z(9lt,+1)], and by the repayment schedules x,(y) and R,(w, y). Once the borrower's problem has been solved, however, we are left with the further prob- lem of how the optimal contract can be supported in the marketplace. We now describe how an optimum can be implemented by having borrowers issue an appropriate mix of debt and equity. An optimal funding contract calls for a repayment rate of xA(y) if state verifi- cation does not occur at t + 1, where (11) x,(y)= p,+J(1 - 0z(*t, 1))+b'J]. Notice that this promised repayment consists of the sum of two terms; Boyd and Smith 383 pt.(l - Oj)z(9t t,+1), which is not contingent in any way on firm performance, and p910y, which depends on the observable component of firm performance (y). Standard usages of the terms direct us to label the first component as the return to debt holders in nonmonitoring states, and the second component as the return to equity holders. In the csv literature, payments that are not con- tingent on firm performance in nonverification states are associated with debt. In addition, Boot and Thakor (1993) call securities that use information about firm performance (here about y) "equity,' while they call securities that do not use information about firm performance "debt." Both terminologies are consis- tent with our usage of the terms debt and equity. (It should be noted that no part of our analysis is inconsistent with the possibility that a firm's debt-or even equity-is held by intermediates.) In order to detennine the amount of debt and equity issued by a representative firm, it is necessary to determine how monitoring responsibilities are allocated between debt and equity holders. Following standard interpretations, we assume that debt holders are paid first. If it is infeasible to make the promised total payment, which is q(1 - 0e)z(O9 L, ,j)t,* to debt holders, debt holders monitor and become residual claimants on investment returns. If the returns on invest- ments are large enough to repay debt holders, but not large enough to fully repay both debt and equity holders-in the amount qxA(y)-debt holders are fully re- paid while equity holders receive less than promised. Equity holders then verify the return and retain all firm income (net of monitoring costs), less payments to debt holders. We assume that both debt and equity holders can coordinate their monitoring activity, thereby avoiding duplication of monitoring effort. This interpretation of monitoring-and of which agents perform it-is incon- sistent with the standard terminology employed in the existing csv literature in one significant way. Specifically, equity holders cannot literally be construed as forcing firms into bankruptcy proceedings, and therefore we should no longer refer to all states in which monitoring occurs as bankruptcy states. Nonetheless, the interpretation that some monitoring is done by equity holders does reflect what is often observed in practice when firms experience low profits but are still able to cover their payments of principle plus interest. Under these circumstances there is likely to be a conffict between the outside equity holders and the inside owner-managers. Outside equity holders cannot force bankruptcy, but they can undertake a variety of costly actions against inside owner-managers; such ac- tions, among other things, have the effect of uncovering information about the firm and include the hiring of outside auditors, various attempts to force changes in firm policies, or even attempts to replace the incumbent management These kinds of actions may be channeled through the board of directors, or even through formal class action suits, and hence are coordinated among external equity hold- ers. In the model all of these activities are, of necessity, represented by costly monitoring. Given these allocations of payments and monitoring responsibilities among debt and equity holders, the quantities of debt and equity issued by any bor- 384 THE WORLD BANK ECONOMIC REVIEW, VOL. 10, NO.I rower are determined by two considerations. First, the real value of debt (d,) and equity (e*) issued at t must raise the necessary external funds (that is, it = q = d, + e, must hold). Second, both debt and equity holders must obtain the market expected return (Tp,,1) between t and t + 1 on their assets. These two sets of conditions completely determine d, and e,. Loosely speaking, it is economical to issue equity only when technology o is used. Thus, if 8, = 0, then et = 0. By the same token, debt is employed to minimize the verification costs associated with technology u. Thus, if 0, = 1, no debt is required, and e, = q (the firn is 100 percent equity financed). For values of 0, between 0 and 1, higher values of Ot are generally associated with greater use of equity, for given values of the rela- tive monitoring cost parameter t.e,,. It is also true that, for given values of 0,, higher values of xttl (higher monitor- ing costs) tend to be associated with less use of equity. (See Boyd and Smith 1994c for a formal statement.) When t.+, rises, in general entrepreneurs will be induced to take actions to further economize on state verification, so that O* will rise as well. Such actions tend to increase the volume of equity issued. Thus, theoretically, the net effect of higher values of z,.+ on the volume of equity issued is ambiguous. Examples produced by Boyd and Smith (1994c, 1995) suggest that, when t,+1 rises, we should typically expect 8(t,41) to rise enough so that firms increase their reliance on equity finance. In other words, it is reasonable to expect that higher perceived monitoring costs lead to the increasingly heavy utilization of technology o. Since investment in this technology is associated with the use of equity, higher perceived monitoring costs are also typically asso- ciated with a greater volume of equity market activity. V. GENERAL EQUILIBRIUM: THE EvoLufoN OF THE REAL AND FINANCLc SECTORS In this section, we integrate the analysis of investment and financing choices of a firm into a conventional neoclassical growth model of the Diamond (1965) variety. The result is a standard one-sector growth model, with one exception. As in most one-sector growth models, starting from some initial capital stock, the economy will (under one technical condition) converge monotonicaily to a steady-state capital stock and output level. Again, as in most (one- and two- sector) growth models, the accumulation of capital will imply that the marginal product of capital, and hence its relative price, declines as an economy develops. In our model, this decline in the relative price of capital has three dosely related implications that are not present in more standard models. First, entre- preneurs are engaged in the production of capital, but state verification con- sumes final goods and services. The latter is not essential, as we discuss in more detail below. Thus, the price of what entrepreneurs produce falls relative to the cost of monitoring or, in other words, the perceived relative cost of monitoring rises in the development process. Second, as we have argued, the decline in the relative price of capital causes entrepreneurs to shift the composition of invest- BoydandSmitb 385 ment toward the observable-return technology, in order to economize on (in- creasingly expensive) state verification. This shift in the composition of invest- ment is typically associated with the increasingly heavy use of equity markets. And third, as the composition of investment changes, less monitoring occurs, and the quantity of resources consumed by the presence of the csv problem declines. This is the social benefit associated with the development of equity markets, and it conforms to the observation that the development of direct trans- actions in securities tends to reduce the costs of intermediation (Wavson and others 1986). Of course this benefit is gained only at some cost. The shift in the composi- tion of investment implies that less capital is being produced over time per unit invested. It is this cost-which is highest when the capital stock is relatively low and capital has a high relative price-that prevents large levels of equity market activity early in the development process. We now formalize this intuition. Capital Accumulation and Production If the capital-labor ratio at time t is k, then each young lender earns co(k), all of which is saved. Hence per capita savings is (1 - a)uw(k). Each young borrower seeks to obtain q units of funds, and hence the per capita demand for external funding is aq. In equilibrium, any savings in excess of aq must be invested in the publicly available capital production technology. Hence, the level of investment in this technology is-in per capita terms-(1 - aoo)(kt) - aq, and the quantity of capi- tal at t + 1 produced using this technology is r[(1 - cx)co(k.) - cxq]. In addition, of the axq units of funds obtained by borrowers, a fraction 0* is invested in technol- ogy o at t, yielding y units of capital per unit invested, while 1- 0* is invested in technology u at t, yielding wil units of capital per unit invested. The per capita capital stock at time t + 1 is given by the sum of these terms; (12) Kt.+ = a4q[8,y5 + (1 - 6*) - r] + r(1 -ca)(k,). Converting the left-hand side of equation 12 into a capital-labor ratio [K,+11(1 - a)] yields (13) kt.1 = [al(1 - a)]q[w' - r - (fit - ')On + awkd. Given the initial capital-labor ratio ko, equation 13 describes the evolution of the equilibrium sequence of capital stocks (kj,. If equation 10 holds, then it is easy to confirm that equation 13 gives k,+1 as an incresing function of kr This function is depicted for a particular set of parameter values in the upper right-hand quadrant of figure 1. In this figure there is a unique, asymptotically stable, steady-state equilibrium capital-labor ratio, denoted by k . Boyd and Smith (1995) shows that, if equation 10 holds and if yaq 2 k(1 - x), then there necessarily exists at least one asymptotically 386 THE WORLD BANK ECONOMIC REVIEW, VOL 10. NO.2 Figure 1. Evolution of tie Capital Stock, Equity Ratio, Composion of Inweshnent and Quantity ofMonitonng Relative to the Development of an Economy IV. Quantity of monitoring L. Capital stock Next-period capital stock, Kt I -3-20 #Zz -1.44/ State verification occurs if Wzt -542 515 .488 4 1.44 2.32 3.20 .119 .079 .039 k k Cuft- Compcsition of caprio investment, capital MI. Composition of H. Equity ratio investment Equity ratio, Et9Q Note The panTneter values are for the expected return on investments in the unobservable-return technology, w- 2.5; for the expected gross return on investments in the observable-retum technology, y -22805; for the rare of retum on estment, r- 05; for the production function, B - 2.0; fortie soale of investment projects, q - 1.0; for the resources used in monitoring, y - 0S; for borrowers as a percenmge of the population, cE - 0-5; and for the production funcdon, A - 0.5. stable steady-state equilibrium with (1 - ao)w(k) > aq. Any such steady-state equilibrium is approached monotonically. If k0 < k£ holds, the capital stock-as well as the level of real activity-will rise over time as the steady state is ap- proached. BoydandSmith 387 Relative Monitoring Costs Rise over Time As we have just described, under the appropriate technical assumptions [k,r is an increasing sequence. It follows that it+. = (t(k,41)I is an increasing sequence as well, so that borrowers perceive effective monitoring costs that are rising over time. This fact has an important implication. If equation 10 holds (which it does uniformly in the examples reported by Boyd and Smith 1994c, 1995), then (6*,] is also increasing over time. Thus, as an economy develops, an increasing fraction of investment will take place in the observable-retum technology. As a consequence, the development process will be accompanied by a dedining (gross of verification cost) return on investment. This shift in the composition of investment also allows for the possibility that total resources consumed in state verification decline as an economy grows. In particular, resources used in monitoring in real terms are given by a-yG[z(O',, t,1J, per capita. As k rises, the variable z, = z (0,; iL,1) that describes the amount of monitoring can, theoretically speaking, either rise or fall. However, it is intui- tive that it should fall; an intuition confirmed by all of the numerical examples reported by Boyd and Smith (1995). Thus, we expect more advanced economies to typically use fewer resources (and a smaller fraction of total resources) in dealing with financial intermediation costs. Because our model has a fixed quan- tity of total resources (caq) always being transferred to entrepreneurs, the analy- sis does not distinguish between a decline in the unit costs and a decline in the total costs of intermediation. A more general analysis would have more funds being transferred in financial markets as an economy develops. Such an analysis would then predict that the unit costs of intermediation fall with development, although the behavior of total costs would depend on the rate of growth of intermediary activity. In this sense, then, developing economies will appear to face financial market frictions that are more severe than those their industrial counterparts face. How- ever, in the model, this outcome is purely endogenous and does not depend on their financial systems being intrinsically more severely flawed. In the sense de- scribed above, the evolution of financial market activity provides an economy with an increasingly more efficient set of financial markets as that economy develops. Developing economies naturally tend to display relatively large costs associated with informational asymmetries. Activity in Equity Markets The equilibrium equity ratio e'7q may, theoretically, either increase or de- crease with the growth process. However, all of the numerical examples pro- duced by Boyd and Smith (1994c, 1995) have the property that elq is increasing in t. Thus, as an economy develops, capital is accumulated, and t increases, our results suggest that the typical observed pattern should be that the volume of equity market activity increases over time. This prediction is consistent with a wealth of empirical evidence documenting the strong positive correlation be- 388 THE WORLD i r.K ECONOMIC REVIEW, VOL. 1, NO.2 tween the level of activity in equity markets and real development (Antie and Jovanovic 1992; Demirgiiu-Kunt and Levine 1993, 1994; Demirguc-Kunt and Maksimovic 1995; Gurley and Shaw 1967; Levine and Zervos 1995; Michie 1987). In order to illustrate the theoretical coevolution of the real and the financial sectors in the growth process predicted by our model, we present a numerical example that is fully general equilibrium in nature. Given a specification of a probability distribution for w, equation 13 and the definition of i(k,,1) describe the equilibrium law of motion for k,, which we trace out and represent diagrammatically in figure 1. Once k,,1 is obtained for each value of k,, we can compute the corresponding values of the equilibrium relative monitoring cost parameter, t,¼ the equilibrium composition of investment, 9n the equilibrium quantity of monitoring, which is related to z, and the equilibrium equity ratio e*tq. Figure 1 depicts how each of these variables evolves as the economy moves along its growth path. In addition, the total resources consumed by monitoring (measured in units of currcnt consumption) at t are ayG(z4,). Thus, figure 1 illustrates how the resource loss implied by the existence of the credit market friction changes as an economy develops. In order to trace out the equilibrium law of motion for k0, it is necessary to specify a set of parameter values (iv, y, r, a, q, y), a production function Pk), and a probability distribution for w. Here we assume Cobb-Douglas production, so thatAik) =BkA.In addition, we assume thatwhas the followingtriangular distri- bution: 4wiW2; O s w wl2 (14) g(w) = 4(w^-W/2 lV/2 < W < W. Equation 14 yields an analytically tractable, symmetric, and unimodal distribu- tion of retums on technology u. Given our assumptions on the production function and the probability distri- bution of returns on technology u, a specification of the vector (ib, y, r, CZ, q, '7, B, 0) is sufficient to allow us to derive the values of k,,+, t,.i, 9, and Zt corre- sponding to each value of k, In addition, the determination of e 7q requires that the entire probability distribution for y be specified. An Example This example is drawn from Boyd and Smith (1995). We setN = 2, yV = 0.01, and yV = 4.551, with p= = = 0.5. Thus, y = 2.2805. Equation 14 implies that w, = w12. In addition, we set wb = 2.5, r = 0.5, a = 0.5, q = 1.0, y = 0.8, B = 2.0, and = 0.5. Figure 1 depicts various aspects of how the capital stock (or capital-labor ratio) and other equilibrium quantities evolve as an economy develops. The up- per right-hand quadrant of figure 1 represents the equilibrium law of motion for Boyd and Smitb 389 k, given these parameter values. As noted previously, k,,1 is a monotone increas- ing function of k,. There is a unique nontrivial steady-state equilibrium capital- labor ratio satisfying all of our hypotheses and, in addition, the steady state is asymptotically stable. The steady-state value of the capital-labor ratio is ap- proximately 2.83, and steady-state per capita output is about 1.68. The lower right-hand quadrant of figure 1 depicts what the equilibrium eq- uity ratio (eI/q) would be for each value of the current capital-labor ratio. For low-enough values of k, the relative cost of state verification is small enough that O' = 0. When this occurs, all investment by borrowers is done in technology u, all external finance takes the form of debt, and the equilibrium quantity of equity is zero. Thus, at low current capital stocks, there is no equity market activity. Over time, of course, lk,j will be an increasing sequence, and once kt, is large enough (here greater than or equal to about 1.2), equity market activity will begin to be observed. The steady-state equilibrium value of the equity ratio for this economy is approximately 0.1, so at the early stages of its development this economy can have no equity market activity. As it approaches its steady state, however, equity market activity will emerge endogenously, and ultimately eq- uity finances a significant fraction of capital investment. For this example, mar- ket capitalization is about 3 percent of gross domestic product (GDP) in the steady state. This approximates observed market capitalizations in Pakistan and Turkey (see table 1). The upper left-hand quadrant of figure 1 shows the value of z t (the return realization below which monitoring occurs) corresponding to each value of kf,2. If (kj is an increasing sequence, (z*4 is a decreasing sequence so that, as this economy develops, the quantity of resources expended in state verification de- clines. (This property is shared by all of the examples reported in Boyd and Smith 1995.) In this sense, the evolution of the debt and equity markets that goes on during the development process provides this economy with more-effi- cient capital markets. This reduction in monitoring costs is made possible by the fact that the com- position of investment (O*,) changes along with k. The behavior of e t is displayed in the lower left-hand quadrant of figure 1. As this figure suggests, the variation of H * over time is quite important in delivering the fairly intuitive result that financial market frictions consume fewer resources as an economy develops. Of course we do not intend to imply that the share of the intermediary sector in real activity typically declines in the development process. This is obviously not the case. What we do mean to imply is that the per unit costs of intermedia- tion typically decline as an economy develops. Such a prediction is, indeed, con- sistent with an array of evidence on transactions costs in developing countries compared with those in industrial countries (World Bank 1989). In our model, simplicity has dictated that a constant quantity of total invest- ment be intermediated over time. Thus, the declining unit costs of intermedia- tion are reflected in declining total resources consumed by this activity. A richer 390 THE WORLD BANK ECONOMIC REVIEW, VOL 10,4 NO.2 model would allow for the total volume of financial activity to grow over time; such growth would be facilitated by a decline in unit costs. The fact that the composition of investment is endogenous is important in our model. If 9, were exogenously fixed, as it is in more conventional csv envi- ronments, more resources would necessarily be employed in state verification as an economy developed. This would imply that financial market frictions loom larger as an economy develops, and it would also imply that equity market ac- tivity declines with increased real development. Both implications appear con- trary to observation. Thus, we believe that a full understanding of the coevolu- tion of the real and the financial sectors in the development process requires confronting firms with an endogenous decision regarding their investment com- position. Summary Over time, as an economy develops, its agents become wealthier and accu- mulate more capital. As capital becomes relatively more abundant, its price dedines. This change in relative prices has several effects. The one most central to this model is that, as the price of capital falls, the relative cost of monitor- ing-as perceived by borrowers-rises. This happens because we have assumed that the monitoring technology employs the final good as an input, a good whose price rises relative to capital in the growth process. However, the assumption that monitoring uses final goods is not essential to the results, although it does yield a substantial simplification. It deserves emphasis that the same forces would be at work if monitoring employed only labor, or if it employed some combina- tion of labor and capital. Under any of these specifications, borrowers will per- ceive relative monitoring costs that rise with the level of development. As a consequence, borrowers substitute away from investment technologies that are monitoring-intensive {technology u), and into technologies that economize on monitoring (technology o). As the allocation of investment changes, so too will the optimal mix of finan- cial claims used to finance that asset allocation. As in more standard csvr envi- ronments, debt is employed to minimize the amount of monitoring necessitated by financial contracts. As technology o is used more extensively, the amount of monitonng will fall of its own accord, and firms will therefore make increasing use of equity relative to debt as an economy develops. Our model necessarily abstracts from a number of important issues concern- ing the role of financial markets in economic development. Perhaps most impor- tant among these is market integration-meaning the extent to which assets with the same risk exhibit the same expected rates of return. As discussed in Bekaert and Harvey (1995), market integration varies considerably across devel- oping nations (and over time). And these differences have important implica- tions for the level and composition of capital investment, as well as for financial contracting. However, our model cannot deal with such issues since (for tracta- bility) we have assumed universal risk neutrality. BoydandSmitb 391 VI. TOWARD A SET OF POUCY IMPLICATONS We have yet to undertake a systematic welfare analysis of the competitive equilibrium allocations that arise in this economy. Nor havc we formally ana- lyzed the consequences of specific policy actions. Howc' !r, in this section we make a few observations about the likely consequences of various policy inter- ventions that might be contemplated in this context.2 We have displayed an example economy in which equity markets may be in- active early in the development process. This lack of equity market activity is not necessarily a signal of any allocative inefficiency, and at this phase attempts to stimulate equity market activity would, at best, simply benefit some agents at the expense of others. Indeed, it is easy to think of interventions intended to stimu- late the development of equity markets that would be harmful. By the same token, in an economy with active equity markets, there would be no obvious case for interfering with the level of activity in these markets. However, for develop- ing economies, it ik not likely that financial (or other) markets will be operating free from government interference. Many specific forms of government inter- vention can easily be analyzed in our framework, and we now offer some specu- lations about the consequences of certain kinds of policy interventions. Many government policies have the effect of altering the opportunity cost of funds perceived by borrowers. Policies that affect the real rate of return to sav- ings can be expected to do so, as can changes in the tax treatment of interest or dividend income, or various interest subsidy or loan guarantee programs. Examples constructed in Boyd and Smith (1995) suggest that a reduction in the oppornmity cost of funds tends to depress O, other things being equal, that is, tends to favor the increased use of technology u. Since technology u is more productive than technology o (gross of verification costs), this tends to result in more capital accumulation and in a corresponding upward shift in the law of motioni for the capital stock. It also tends to attenuate the level of equity market activity. Thus, policies that reduce the opportunity cost of funds to borrowers are likely to depress the level of activity in equity markets and to increase the reliance on debt finance. Inflation A particularly obvious macroeconomic factor that affects the opportunity cost of funds to borrowers is the rate of inflation. Higher rates of inflation act to reduce the real rate of return on real balances, or on any other savings instru- 2. A technical observition is that competitive equilibrium allocations in our economy are what Balasko and Shell (1980) term "weakly Pareto Optinal"; that is, it is not possible to achieve a Pareto improvement by transferring resources between a finite number of generations. However, this does not imply that competitive equilibrium allocations are fully Pareto optimal; capital overaccumulation is still possible, as it is in the standard Diamond (1965) modeL If our economy does not display capital overaccumulation- that is, if steady-state real interest rates are not too low-and if there are no other interventions, then competitive equilibria will be Parcto optimal, and any policy interventions that are not entirely negative in their consequences must necessarily be redistributive. In this situation, policies must be evaluated with respect to whether these redistributions are deemed socially desirable. 392 THE WORLD BANK ECONOMIC REVIEW, VOL. 10. NO.2 ment bearing a fixed nominal return. In addition, particularly in many devel- oping countries, binding nominal interest rate ceilings limit the flexibility of nominal returns, and relatively high rescrve requirements force banks to hold large amounts of real balances. Both kinds of factors will tend to allow increases in the rate of inflation to put downward pressure on real returns to savings. Indeed, Choi, Smith, and Boyd (1995) document that in several countries the real rate of return on both safe savings instruments and on equity is very strongly negatively correlated with the rate of inflation. Thus, higher rates of inflation will tend to reduce the real returns perceived by savers and to lower the oppor- tunity cost of funds to borrowers. Our previous observations, then, suggest that higher rates of inflation are likely to reduce 6 t and hence to depress the volume of activity in equity markets. This conjecture that high inflation is detrimental to equity market activity is consistent with the finding of Tun Wai and Patrick (1973), Choi, Smith, and Boyd (1995), and Boyd, Levine, and Smith (1995) that higher rates of inflation do tend to be injurious to stock market operations. At this point, we withhold any conjectures about the effects of higher inflation on real activity. An analysis of this issue would require a monetary version of the present model. This is a sufficiently large modification that we do not currently want to speculate about the consequences of higher inflation for, as an example, the steady-state capital stock. If we are correct, higher rates of inflation lead to heavier use of the invest- ment technology that is subject to the csv problem, and to a greater volume of resources being consumed by state verification. In this sense, higher rates of inflation would appear to make an economy's financial markets function less efficiently. Thus, macroeconomic policies that are conducive to price stability will tend to favor the development of equity markets, and to foster the efficient operation of capital markets in general. For further theoretical results on this point, see Choi, Smith, and Boyd (1995) or Boyd and Smith (1994a). There are, of course, any number of other methods by which government interventions can be used to reduce the opportunity cost of funds to borrowers. For instance, a reduction in ceiling rates of interest paid on deposits would tend to do so, as would an increase in the effective tax rate on capital income. Our previous reasoning suggests that these actions would tend to shift the composi- tion of investment in favor of technology u. The shift in favor of technology u would typically favor increased capital formation, depending on whether sav- ings rates are sufficiently insensitive to variations in the real (after tax) rate of interest, an assumption that we make in our formal analysis. The shift in favor of technology u would also tend to reduce the volume of equity market activity; in addition, it would tend to lead to more resources being consumed in the state verification process. In this sense, high real returns to savers tend to be con- ducive to the development of equity markets and to economizing on the re- sources consumed by monitoring. Boyd and Smnith 393 The Relation between Debt and Equity Markets A final issue of government policy concerns the relationship between debt r.nd equity markets. It is often argued that the allocative importance of equity markets in developing economies is not very great, as debt markets constitute an effective substitute for them (see, for instance, the discussion in Rojas-Suarez and Weisbrod 1994). We believe that thi: irgument is called into serious ques- tion by our analysis. In particular, by using equity markets appropriately, firms in our model substantially reduce their costs of issuing debt. Indeed, Boyd and Smith (1994c) produce examples in which it is impossible for firms to issue debt without issuing some equity, simply because the costs of 100 percent debt fi- nance are too great for this to be feasible. This situation is most likely to obtain as economies become relatively developed, and it suggests that at some levels of development equity markets will be a necessary complement to debt markets. When this occurs, equity market activity will appear endogenously if it is not hampered by government intervention. VII. CONCLUSIONS We have developed a model in which capital is produced by investors who make use of two technologies. One yields a high expected return, but is subject to an informational friction. The other yields a lower expected return, but has the advantage of full public observability. Investors must make a decision regard- ing how heavily they will use each technology. This decision depends, among other things, on the relative price between capital and the resources used in state verification. As an economy develops, investors will perceive a relative cost of moni- toring that rises over time. As a result, under conditions that we typically expect to prevail, less use will be made of the unobservable-return, and more use will be made of the observable-return technology. Since investment in the unobservable-return technology is generally associated with the use of debt finance, while the use of the observable-return technology is associated with equity, we also typically expect the ratio of equity finance to rise as an economy develops. This intuition is confirmed by all of the examples in Boyd and Smith (1994c, 1995). Moreover, as we have seen, it is possible to produce parameter values such that-at low levels of development-there will be no use of equity markets. Equity market activity can be observed for such parameters only once the economy attains a critical level of real development. Such examples support the conclusion of Gurley and Shaw (1960, p. 92) that "the selection of financial assets evolves in the growth process," and that the variety of financial claims increases as well. It is also the case that, in all of our numerical examples, the quantity of re- sources consumed by monitoring declines as an economy develops. This pro- 394 THE WORLD BANK ECONOMIC REVIEW. VOL. 10, NO. 2 vides a sense in which the endogenous evolution of debt and equity markets in the development process provides an economy with a more efficient set of capi- tal markets. Finally, our analysis provides a sense in which debt and equity markets func- tion as complements rather than substitutes. A case against the importance of equity markets in financing real development is often made on the basis that existing credit markets are close substitutes for equity markets. Our analysis calls the validity of such arguments into serious question. RErERENCES The word "processed" describes informally reproduced works that may not be com- monly available through library systems. Antje, Raymond, and Boyan Jovanovic. 1992. "Stock Markets and Development." Draft. New York University, Department of Economics, New York. Processed. Balasko, Yves, and Karl Shell. 1980. "The Overlapping-Generations Model. I: The Case of Pure Exchange Without Money." Journal of Economic Theory 23(3):281-306. Bekaert, Geert, and Campbell R. Harvey. 1995. "Time-Varying World Market Integra- dion." Journal of Finance L(2):403-45. Bencivenga, Valerie R., and Bruce D. Smith. 1991. 'Financial Intermediation and En- dogenous Growth." Review of Economic Studies 58(2):195-209. Bencivenga, Valerie R., Bruce D. Smith, and Ross M. Starr. 1994. "Liquidity of Second- ary Capital Markets, Capital Accumulation, and the Term Structure of Asset Yields." Draft. Cornell University, Department of Economics, Ithaca, N.Y. Processed. 1995. "Transactions Costs, Technological Choice, and Endogenous Growth." journal of Economic Tbeory 67(1):153-77. - . -1996. 'Liquidity of Secondary Capital Markets: Allocative Efficiency and the Maturity Composition of the Capital Stock." Economic Theory 7(l):10-50. Boot; Arnoud W., and Anjan V. Thakor. 1993. "Security Design." Journal of Finance 48(4):1349-78. Boyd, John H., and Bruce D. Smith. 1994a. "Capital Market Imperfections in a Mon- etary Growth Model." Draft. Federal Reserve Bank of Minneapolis, Research De- partment, Minneapolis, Minn. Processed. - . 1994b. "How Good Are Standard Debt Contracts? Stochastic versus Non- stochastic Monitoring in a Cosdy State Verification Environment." Journal of Busi- ness 67(4):539-61. 1994c. "The Use of Debt and Equity in Optimal Financial Contracts." Draft. Federal Reserve Bank of Minneapolis, Research Department, Minneapolis, Minn. Processed. . 1995. "The Evolution of Debt and Equity Markets in Economic Development." Draft. Federal Reserve Bank of Minneapolis, Research Department, Minneapolis, Minn. Processed. Boyd, John H., Ross Levine, and Bruce D. Smith. 1995. "Inflation and Financial Mar- ket Performance." Draft. Federal Reserve Bank of Minneapolis, Research Depart- ment, Minneapolis, Minn. Processed. Boydand Smith 395 Choi, Sangmok, Bruce D. Smith, and John H. Boyd. 1995. "Inflation, Financial Mar- kets, and Capital Formation." Draft. Federal Reserve Bank of Minneapolis, Research Department, Minneapolis, Minn. Processed. Cooley, Thomas F., and Bruce D. Smith. 1992. 'Financial Markets, Specialization, and Learning by Doing." Draft. Cornell University, Department of Economics, Ithaca, N.Y. Processed. Demirgiuc-Kunt, Aslh, and Ross Levine. 1993. 'Stock Market Development and Finan- cial Intermediary Growth: A Research Agenda." Draft. World Bank, Policy Research Department, Washington, D.C. Processed. 1994. "Stock Markets and Financial Intermediaries: Stylized Facts." Draft. World Bank, Policy Research Department, Washington, D.C. Processed. Demirgiiu-Kunt, Aslh, and Vojislav Maksimovic. 1995. "Stock Market Development and Firm Financing Choices." Draft. World Bank, Policy Research Department, Washington, D.C. Processed. Diamond, Douglas W. 1984. "Financial Intermediation and Delegated Monitoring." Review of Economic Studies 51(3):393-414. Diamond, Peter A. 1965. "National Debt in a Neoclassical Growth Model." American Economic Review 55(5):1126-S0. Gale, Douglas, and Martin Hellwig. 1985. "Incentive-Compatible Debt Contracts: The One-Period Problem." Review of Economic Studies 52(4):647-f63. Greenwood, Jeremy, Zvi Hercowitz, and Per KrIussell. 1995. "Long-Run Implications of Investment-Specific Technological Change." Draft. University of Rochester, De- partment of Economics, Rochester, N.Y. Processed. Greenwood, Jeremy, and Boyan Jovanovic. 1990. 'Financial Development, Growth, and the Distribution of Income."Journal of Political Economy 98(5, October):1076- 1107. Gurley, John G., and Edward S. Shaw. 1955. "Financial Aspects of Economic Devel- opment." American Economic Review 45(4):515-38. 1960. Money in a Theory of Finance. Washington, D.C.: Brookings Institution. 1967. 'Financial Structure and Economic Development." Economic Devel- opment and Cultural Change 15:257-68. Levine, Ross. 1991. "Stock Markets, Growth, and Tax Policy." Journal of Finance 46(4):1445-65. Levine, Ross, and Sara Zervos. 1995. "Poicy, Stock Market Development, and Long- Run Growth." Draft. World Bank, Policy Research Department, Washington, D.C. Processed. McKinnon, Ronald L 1973. Money and Capital in Economic Development. Washing- ton, D.C.: Brookings Institution. Michie, R. C. 1987. The London and New York Stock Exchanges, 1850-1914. Lon- don: Allen and Unwin. Rojas-Suarez, Liliana, and Stephen Weisbrod. 1994. "Financial Market Fragilities in Latin America: From Banking Crisis Resolution to Current Policy Challenges." Draft. International Monetary Fund, Research Department, Washington, D.C. Processed. Shaw, Edward C. f973. Financial Deepening in Economic Development. New York: Oxford University Press. 396 THE WORLD BANK ECONOMIC REVIEW, VOL 20, NO.2 Townsend, Robert M. 1979. "Optimal Contracts and Competitive Markets with Costly State Verification." Journal of Economic Tbeory 21(2):265-93. Tun Wai, U., and Hugh T. Patrick. 1973. "Stock and Bond Issues and Capital Markets in Less Developed Countries.' International Monetary Fund Staff Papers 20:2S3- 317. Watson, C., G. Maxwell, R. Kincaid, C. Atkinson, E. Kalter, and D. Folkerts-Landau. 1986. "International Capital Markets: Detelopments and Markets." IMF World Economic and Financial Surveys. Washington, D.C.: International Monetary Fund. Williamson, Stephen D. 1986. "Costly Monitoring, Financial Intermediation, and Equi- librium Credit Rationing." Journal of Monetary Economics 18(2):159-79. World Bank. 1989. World Development Report 1989. New York: Oxford University Press. World Deelent Report 1996: From Pla to Market W - orld Deivelopment Reporl 1996: From Plan to u!iaa WULI, AtarLet steps back from the extraordinary 1| ' s array of recent events and policy changes in 28 for- mer centrally planned economies-those in Centria and Eastern Europe and the newly independent states of the rormner Soviet Union, along with China, Mongolia, and Vietnam-to ask what we have learned about the ingredients of a successfid transi- tion to a market economy and how they should be pursued. The Report addresses the initial chaUenges of transition and how different countries have responded. it also analyzes the longer-term agenda of consolidating the early reforms by developing the insdtutions and policies that will help new systems to develop and prosper. Publisbedfor the World Bank by Oxford Uxnf ty Pcssr English (Hardback) Stockno. 6110 ER(ISBN 0-19-521108-1) $45.95 English (Paperback) Stock no. 61107ER (ISBN 0-19-521107-3) $22.95 Frfbobming in jgflfaAq: $22395 awck: Arabic Stock no. 13269ER (ISBN 0-8213-32694) September 1996 Chinew Stock no. 13268ER (ISBN 04213-3268-6) September 1996 French: Stock no. 13264ER (ISBN 04213-3264-3) August 1996 Cermat Stock no. 13266ER (ISBN 0-8213-32664) September 1996 Juputese Stock no. 132671ER (ISBN 0-8213-3267-8) September 1996 Portuguese: Stock no. 13271ER (SBN 0-8213-3271-6) September 1996 Iussiani Stock no. 13270ER (ISBN 0-8213-3270-8) September 1996 Spanish: Stock no. 13265ER (ISBN 0-8213-3265- ) August 1996 World Bank Publications For pTenlimin local cuneny, conadyWadd hnskDi*or. ln fI U.S. cnITheWoddBnk. PA 7Z47-M9.PIib IpNa, PA 19170-0I. Phone:MR473 1155, FatCM522-26;. Spingard hmdIng USsM& For *dmIi defw ivieryoW&fUS.. add USlMOOloron lembplt AS6O for arhadWoa kmn P*etby USSdmanson a US, bankpyEablebDtleWbdoBakrby MaorbrniRpess~~~~~~~~~~~~~~~~~~~12 Distributors of World CHINA GERMANY Bank Publications ChianaiinclaEconomic UNO-Vedg Prces and awdt19,ms V 1m Publislhn House 8, Da Fo Si PoppeisdorferAlee55 aountto counw Consultyour DgJ 53115Bom ical distrbutrbe tortoplacingan B GREEC adeS h GREECE COLOMBIA Papasotrlou SA. ALBANIA InfoeriLtda. 35,Sourara Sir. Mdon Ltd. ApartadoAeeo3427 10662Athens PerlatRexhepiStr BogotA D.E. Pa. 9, Shk. 1,Ap. 4 HONG KONG, MACAO Trna COTE DIVOIRE Asia2000 Lt. Centre d'Ediflon et de Dfffusion Sabes& Ciculation Depatmnt ARGENTINA Afrcalnes (CEDA) Seabird House, OficnadelLbrolntemaclonel 04 Ba..-41 unItl0ll02 Av. Crdoba 1877 AbWijan 04 Plateau 22-28Wyndham Stre, Cental, 1120BuenosAirs HongKong CYPRUS AUSTRAUA, RJI, PAPUA CnterofApplitidResearch HUNGARY NEW GUINEA, SOLOMON CypusCoIg Foundation forlarketEionomy ISLANDS, VANUATU, MD 6. Dioges Stkeet Engomi Donbovad Ut 17-19 WESTEM SAMOA P.O. BoxX206 H1117Budapest DA Informtion Sevoes NMagia 648WhithorsoRoad INDIA Mitcham3132 CZECH REPUBLIC Aried Publishers Ltd. Visofia NationalInformationCenter 751 MountR prena, Konriktska 5 Madras-600002 AUSTRIA CS-11357 Praguel Gerold anddCo. INDONESIA Graben3l DENMARK Pt Indira Umited A-1011 Wien SamfundsLlteratur Jalan Bombudur2o RosenoemnsMAl6 P.O. Box 181 BANGLADESH DK-197OFredenksbergC J*kat1a032O Micro Indusies Developmet Assistance Society (MIDAS) EGYPT, ARAB REPUBUC OF IRAN HoLueS,eRoad16 AlAhrarn Kowkab Publishers DhanmondfRtAmea AlGalaStaeet SPO. Box19575-511 Dheaa1209 Cairo Tehun BELGIUM The Middle EastObserver Ketab Sara Co. Publishers Jean De Lannoy 41, Sherif Street Khaled EsiamboliAve.,6th St Av. du Roi202 lro Kusheh DelafroozNo.8 1060Brsels Teliran FINLAND BRAZIL Akeinn NoKik5Uppa IRELAND PuilicaoresTecnicas RO. Box 23 GovenwmentSupplesAgency IntnaioisLida FIN-0037 Heisirn 0g antSolilhair Rua Peixoto Gomids, 209 4-5 Haourt Road 01409SaoPaulo, SR FRANCE Dtbin2 World Bank Publications CANADA 66 avenued'l6na Renml Publishing Co. Ltd. 75116 Pais l294Atoma Road Oftawa Ontario KIB 3W8 ISRAEL NETHERLANDS PORTUGAL Yaomot Uteeture Ld. DeUndeboomlhOrPublikales UvrarlarPoriugal P.O. Box 56055 PO. Box 202 RuaDoCarno70-74 TelAviv61560 748oAEHaaWrgn 1200Llsbn R.Y. Intenational NEW ZEALAND ROMANIA PO. Box 13056 0SCO NZ Ud. Compani De Ubrailucuresi TelAviv61130 Pivate Mal Bag 99914 S.A. New Market,Auckland Str. Upscarn no. 26 PalestnlanAuthodtyMk1d1eEast sector3 IndexinomnadonSercs NIGERIA Bucharest PO.B. 19502Jerusalem UrIver*y Press UNited Thre Crnis Building Jericho RUSSIAN FEDERATION ITALY Private Mal Bag 50 Isdatelstvo es Mir> UcosaComrnssionariaSansoni badan 9a, Kolpachniy Pereulok SPA Moscow101831 Via Duca Di Calabria, 111 NORWAY Casela Postale 552 NarveseninformationCentr SAUDI ARABIA, QATAR 50125Rreze BookDepartment JanrBookStare P.O. Box 6125 Etterstad P.O. Box 3196 JAMAICA N-0602 Oslo 6 Rlyadh 11471 Ian Randle Pubtishrs Ud. 206 Old Hope Road. Klngston 6 PAKISTAN SINGAPORE, TAIWAN, MirzaBookAgency MYANMAR, BRUNEI JAPAN 65,Shahrah-e-Quaid-e-Azarn Asahgate PublishingAsia Eastem Book Sewice P.O. Box No. 729 Paciir Pt. Lld. Hwg3Chom A Lahore 540 41 KalangPudding Road #4-03 Bunkyo-ku 113 GoldenWheelBuling Toko Oxford Uniersit Press Singapore349316 S Bangabre Town KENYA Shame FaBisal SLOVAK REPUBLIC Afrca Book Service (EA) Ld. PO. Box 13033 Slovart G.T.G. Ltd. OuaranHouse Kaachi-75350 Krupnska4 MfanganoStret P.O. Box 152 P0. Box 45245. Nairobi PERU 85299 BratislavaS Editorial Desarolo SA KOREA, REPUBLIC OF Apartado3824 SOUTH AFRICA, BOTSWANA Daejon Tradng Co. Ltd. Unal Forsngle rites: PO. Box 34 Oxford Universiy Press Southemr Yeoeida, SoulI PHILIPPINES Afria intenatonal Booksource Center, P.O. Box 1141 MALAYSIA Inc, CapeTown 8000 UniWry of Malaya Sulte720, Cityland 10 CooeativeBookshop, Uniited, CondornlriiumTower2 Forsubsac on onrds: P.O Box 1127 H.V da Casta, comer Intemational Subscription Svc. Jalan Pantal Barn VaIeroSt, Makati, Metro Mania P.O. Box41095 59700 Kuala Lumpur Craihall POLAND Jdhannesburg2024 MEXaCO Internaoal Publishing Savim INFOTEC UI. Piekna 31/37 Apatdo Postal22-860 00-SllWarawa 1406OTMpen Mexico D.F. SPAIN SWITZERLAND UGANDA Mundi-PrensaU ibros,SA UbrairePayot GustroLtd. Castdol37 Service Insfflulionnel MadhVWIBucing 28001 id COts.de-Monbnon30 P.O. Box 9997 1002Lausanns Plot 164Jinja Rd. MundiPrensa Barcelona Kmpla Consell de Cmnt, 391 bbn Diermen EdidonsTechnfques OQOOgBamelana Ch.deLacuez4l UNITED IaNGDOM CHl807BIonay MbinfoOLtd. SRI LANKA, PO. Box 3 THE MALDIVES TANZANA Alon, Lake House Bookshop Oxford University Press HampshireGU342PG P.O. Box 244 Makaba Street iVand 100, SirChittampalamA. P.O. Box 5299 GarnerMawfla DaresSabam ZAMBIA Colonro2 University Bookshop THAILAND Gmeat East Road Campus SWEDEN Cental Books Distibution P.O. Box 32379 Fatzes Customer Service 306 Slom Road [usaks Regerigsgaton12 Bangkok S-10647 ZIMBABWE Stockholm TRINIDAD & TOBAGO, LongmanZbnbabwe JAMAICA (Pe.) ld. WVennergren-WillarnsAB Systemftcs Studies Unit Toure Road,Ardbennis R 0 Bx 1305 #9WattsStreet PO. Box ST125 S-17125 Cum"pe Southerton So-lna Tdrtdad Hame WestIndies Coming in the next issue of THE WORLD BANK ECONOMIC REVIEW Septemnber 1996 Volwme 10, Nuinber 3 * Why Do Govemments Initiate Public Enterprise Reform? knj Jose Edgardo Cnmpos and Hadi Sale/ii Esfahani * Roads, Land Use, and Deforestation: A Spatial Model Applied to Belize by Keiiinethi M. Clhomyiitz and David A. Gray v A New Database Measuring Income Inequality by Klaus Deiniinger aind Lyin Squire * From Plan to Market by Mar-tha de Melo, Cet'det Den izer, anid Alauu Gelb * Hungary's Bankruptcy Experience, 1992-93 by Cliery Gray, Sabine Schlorke, and Miklos Szanyi * Industrial Centralization in Indonesia by J. Venion Hen7derson and Ari Kuircoro * Guidelines on Searching for a Dalton-Improving Tax Reform: An Illustration with Data from Indonesia by Shlomo Yitzhaki and Jeffrey D. Lewis