7 Policy Res|e .. WfHRKING PAPERS Public Economics Policy Research Department The World Bank September 1993 WPS 11167 Determinants of Inflation among Franc Zone Countries in Africa Bruno Boccara and Shantayanan Devarajan Despite belonging to a monetary union with a common currency and pooled foreign reserves, the countries of Africa's franc zone (CFA) experience substantially different inflation rates, espe- cially in the short run. This is partly explained by the fact that for primary exporters in general, and for CFA zone members in particular, the volatility of commodity prices implies a high variance in government revenues. PoicyR ahWoding Papas dininat thefindg eof wCIs in p and zwagthedB hpdidma ama,Dr tffad u _td irndeydos md tiuu Thmepapam disbued by X Ruch Adviry Suff X tAnwaof thhmadhi yvsnu dboldbouedaWdeidaoy.Mwfind,inupu4d6adheuhe 'ow&Tb"sbeuld ambe audhmd to tn Wadd BDa. its Dod dDIW af its manaLmiu or m y of iw naw0bim PORGYt R_tob PuSb EoonmW WPS 1197 This paper-a productof the Public Economics Divis.on, Policy F.1search. Department-is partofa larger effort in the departnent to study structural adjustment in Sub-Saharan Africa. Copies of the paper are available free from the World Bank. 1818 H Street NW, Washington, DC 20433. Please contact Carlina Jones, room NIO-063, extension 37699 (September 1993, 45 pages). Despite belonging to a monetary union with a The fiscal relationship is the key behavioral common currency and pooled foreign rescrves, equation of the model, as the other r-lationships the countries of Africa's franc zone (CFA) are essentially derived from accounting identi- experience substantially different inflation rates, ties. especially in the short run. Boccara and Devarajan empirically test the Boccara and Devarajan develop a model of model for C6te d'lvoire. It tracks quite well the inflation differentials for the franc zone countries inflationary r thAt Cote d'Ivoire experienced based on behavioral differences in fiscel policy after the boom in coffee prices in 1975-76. responses to fluctuations in the price of the main export commodity. The model is based on the Since the countries are in a monetary union, fact that for primary exporters in general, and for if some countries have expansionary fiscal policy CFA zone members in particular, the volatility of (and thus inflation) the others must take a more commodity prices implies a high variance in contractionary fiscal stance. One issue for future govemment revenues. research is what determines whether a member country has the freedom to follow an expansion- The model identifies two effects: a monetary ary fiscal policy or whether it must contmct effect (commodity booms imply a surge in because other members have already expanded. foreign reserves which, if unsterilized, is infla- A discussion of potential "games" played by tionary) and a fiscal effect (higher govemment countries in a monetary union may shed light on revenues are, to varying degrees, accompanied these issues. by a marked increase in the level of spending, which is again inflationary). ThePolkcy ResearchpWorking P SaperSemitssdisseminates thendinpofworkunderwayinhfakAnobjeiv the fdsiuerie is to got these findings out quickly, even if presentations wre less than fully polished. The fvidings. interocaios, and conclusio in these ppes do not necesarily represent official Bank policy. Podwed by die Policy Resn h Dueiniian Center DETERMINANTS OF INFLATION AMONG FRANC ZONE COUNTRIES IN AFRICA by Bmno Boccam World Bank and Shantayanan Devarajan World Bank and Harvard University We are grateful to Stanley Fischer for conunents on an earlier draft and to Ajay Chhibber for comments and partial financial support. The Franc zone in Africa is a unique institution in the developing world. Thirteen sovereign countries are engaged in two monetary unions with France. Membership in the union involves pooled foreign reserves, a common currency (whose convertibility is guaranteed by France) and a fixed exc'hange rate with the French Franc. There are no foreign exchange implications for currency flows among members of the Franc zone. Membership in the Franc zone has provided countries with a substantial amount of monetary discipline. In particular, the inflation rates in these countries are much lower than in other developing countries (see Table 1). Table 1: Average Annual Inflation 1973-81 1982-89 CFA 12.0 4.3 Other: Sub-Saharan African 24.3 29.7 Low-income LDC's 18.4 33.3 Primary Producers LDC's 24.4 44.9 Note: Unweighted averages Source: Devarajan and de Melo [1991] In addition to lower overall inflation, one would expect little difference among inflation rates within the zone. Perfect capital mobility between the countries implies convergence of real interest rates, and thus of inflation rates if nominal interest rates are aligned (in the absence of country-specific risk premia). In practice, however, capital movements within the CFA zone are restricted and the interbank market is not sufficiently developed to enable perfect 1 arbitrage among various financial instruments'. Thus, and in spite of the fact that labor mobility is very high between the CPA countries, there can still be inflation differentials among member countries of the CFA zone (see Table 2). Table 2: Inflation within the CFA Zone 1983 1984 1985 1986 1987 Burkina 8.4 4.9 6.9 -2.6 -2.6 Cote d'Ivoire 5.9 4.3 1.9 6.6 5.3 Mali 9.8 12.4 7.8 -3.9 -16.7 Niger 2.5 8.5 -1.0 -3.2 -6.8 Senegal 11.7 11.8 13.0 6.0 -4.3 Togo 8.2 -3.5 -1.8 3.2 -0.6 Source: Direction dc la Statistiques des Etats Banque de France This paper analyzes inflation differentials in the Franc zone. In Section I, the existence of inflation differentials is verified by analyzing and conducting tests on the data. Section II de'elops a model of inflation in the CPA countries which incorporates important institutional aspects of the zone. In Section III, we test the model empirically and conduct a simulation. The main conclusion is that inflation differentials can be explained by the choice of 'For example, transferring funds to France is subject to fees and often involves substantial delays. Also banks themselves do not act as intermediaries between regions where there is a shortage of funds and regior.s where there is a surplus. 2 adjustmnent policies made by member countries of the monetary union. More specifically, given the countries' reliance on primary exports, the differences in inflation rates reflect different fiscal policy responses to changes in the international price of the main export conmmodity. We show that the changes in the export commodity prices have important effects on the domestic price level and therefore on the real exchange rate. Thus, a commodity boom will translate into inflation through a monetary effect (due to the unsterilized increase in foreign reserves) and through a fiscal effect (due to increased government spending, especially investment). I. INFLATION DIFFERENTIALS IN THE FRANC ZONE In this section we evaluate whether there are inflation differentials in the CFA zone. We test whether there are long-run relationships in price levels and inflation rates between the two unions, France and in the CFA countries and among countries in the Western (BCEAO) and Central (BEAC) monetary unions. First, unit root tests are performed for each price level2. We use an augmented Dickey- Fuller test (with two lags) which is based on least squares estimation of the following 2Data consist of yearly observations (all betw:en 1960 and 1988) ranging from 29 observations for France an Burkina Faso to 23 for Senegal and 21 for Cameroon. 3 regression equation Jup A X 1 X,= + E y1AX + 2 where the null hypothesis, Ho (B = 0) is no stationarity. The regression equation above is run for the price levels and the logarithm of the price levels and for the first two differences of these variables. As doc .mented in Engle and Granger (1987), the distribution of the t- statistic for the coefficient B is non-standard. The 99 (respectively 95) percent confidence level for rejection of non-stationarity is 3.77 (respectively 3.!7). Tests for higher order statioriarity are only relevant if all the lower order stationarity tests fail. The results show all price levels and their logarithms to be 1(2) (second-order stationarity) except for Burkina Faso for which we have first-order stationarity (and therefore, a stationary inflation rate). Next, cointegration tests are performned for price levels which have the same order of integration. The test, following Engle and Granger (1987), is an augmented Dickey-Fuller 4 test on the residual of the cointegration equation y, = P xi + C + el where x, and yt are the variables for which cointegration is tested and c is a constant. Table 3 shows the results of the cointegration tests which arc conducted between France and the two mo6. important countries of each of the unions, CMte d'Ivoire (which accounts for 38 percent of BCEAO GDP) and Cameroon (which accounts for 60 percent of BEAC GDP), between CMte d'Ivoire and Cameroon, between BCEAO countries (excluding Burkina Faso) and CMte c'Ivoire and, between BEAC countries and Camer2On. Cointegration is always rejected except between France and CMte d'lvoire and betwe n Senegal and Cote d'Ivoire. In sum, there is evidence of inflation differentials among the Franc zone countries in Africa. This can be seen in Figures 1 - 4, which plot inflation rates among various combinations of CFA countries and France. Although Honohan (1990) rejected cointegration between the French and CFA -one price levels, he nevertheless argued for inflation convergence based on his principal components analysis. This analysis showed that half of the variance of the zone members' inflation rates could be explained by a common factor (the principal component). Honohan (1990) then assumed this factor was tFrench inflation. Since the principal component analysis only accounted for half the variance, it is worth investigating what might explain the other half. Furthermore, the convergence of inflation rates described by Honohan (1990) is a long-run phenomenon. We are interested in explaining inflation differentials in the short-run. We turn therefore to a model which 5 interprets these differences in terms of fiscal polic,' responses to commodity price boom^. II. A MODEL OF INFLATION DIFERENTIALS IN TIE FRANC ZONE The starting point of our analysis i) th definition of the pric: level in a CFA country as: P - P,,P (oX, where PM is the domestic price of imports, PN is the domestic price of all other goods consumed in the ;-- zonomy, and a the share of imports in consumption. Thus, the inflation rate is a weighted average of the infladon rates of nontradables and imported goods. The domestic price of imports, in turn, is equal to the nominal exchange rate multiplied by the world price (absent any trade taxes.) Since the nominal exchange rate in a CFA country is fixed, domestic policymakers have little control over the price of imports. Furthermore, import prices are likely to mole in tandem across CFA countries, so that this component will not explain inflatdon differentials. Hence, the fobac of our investigation will be movements in PN. Recall that we classify as "nontradables" all goods in the consumption basket which are not imported. We therefore include import substitutes. This is jusified by the fact that, in most of these countries, domestically produced good are only imperfect substitutes for imported goods. Corsequently, domesdc prices of substitute goods are determined by domestic demand and supply conditions, rather than by the world prices of imported goods. In what follows, we present a model of how F, 6 is detemnined in a fixed-exchange rate regime that is subject to periodic shocks in its external terns of trade. Most of the member countries of the zone are dependent upon exports of primary commodities for foreign exchange. These commodities are also the main source of government revenues, as the governmemt sets the producer price below the world market pri.ce, with the difference accruing to a marketing board. A surge in the world price of this comrnodity has two effects. The first can be called the fiscal effect: the commodity boom is accompanied by a surge in government expenditures which often continues in spite of the si'bsequent decline in the price of primary commodities. The second can be called the monetary effect: the pooling of reserves and the maintenance of a fixed exchange rate result in incomplete sterilization of the foreign exchange. The domestic money supply rises foilowing the boom. Fluctuations in commodity prices, through these effects, ultimately cause changes in the domestic price level. 2.1 Specifying the Model It is assumed tha. there are three goods in the economy: a primary commodity C which is not consumed domestically, imports M and nontradables N. The notation used is the following: At = productivity variable; 7 H, - nominal money; Ri - nominal not foreign assets in CPA francs; Cr, - total domestic credit; Zi, = nominal net foreign transfers in CFA frar.es; Z2, = nominal capital account in CFA francs; pc", = international price of main export commodity; p'is = foreign price or imports; pM1 = domestic price of imports; = nominal government revxnues; l, = nominal government evpenditures (including hivestment); Gl, = nominal govermnent expenditures (excluding investment); 7t .= tax rate on commodity exports, 7", = tax rate on imports; = nominal exchange rate defined as CFA francs per dollar; pN = domestic price of nontradables; Pt = domestic price level. At, Z,i, Z2t, Pc*1, Pmt, 'rm, and Et are exogenous while the remaining 10 other variables are endogenous. All Greek letters represent parameters of the model. For any 8 variable x, dt .dt x A dependent economy model for the CFA countries is proposed and empirically tested in Devarajan and de Melo, (1987). Their model focuses on the adjustnen. needed in real variables to correct an external deficit. It ignores mc,netary aspects and treats government spei,ding as exogenous. It is shown that altervative assumptions about the labor market (full employment or real wage rigidity) do not alter the qualitative nature of the results. Since our model aims at capturing the effects of fluctuations in export prices on domestic inflation (and thus, on the real exchange rate), labor market and wage behavior issues are also ignored and the nominal wage is assumed to be constant3. The monetary side of the model is represented by equations (1) through (3) below. H, = (1--)R, + wCr, (1) 3We assume that the production of the exogenous level o' the main export commodity requires a fixed amount of labor LC. Denoting the labor force in the nontradables sector Lt,N Lk + LN is the employment level. A model which analyzes the impact of the fluctuations of the international price of coffee for Colombia can be found in Edwards (1986). A common feature of that model and ours is that terms of trade changes affect the money supply. 9 Cr,=R, (2) A PR,AI -1'Z114 + )2Z22, + 1(C + () - 42(, + A) (3) Equation (1) comes from the balance sheet of the monetary authorities, namely, the identity that base money equals net foreign assets plus domestic credit. Thus, the percentage change in money is equal to a weighted share of the percentage change in net foreign assets and of the percentage change in domestic credit. Equation (2) is the rule for domestic credit creation. The percentage change in domestic credit is assumed to be equal to the percentage change in net foreign assets since the balance of payments determines money in the absence of sterilization. In theory, this relationship only holds for the F.anc zone as a whole since it corresponds to the monetary rules of the BCEAO and of the BEAC. The solvency rule for each of the two central banks states that the ratio of foreign reserves to all assets be limited from below or A R + Cr ° Assuming the constraint to be binding, (i.e. the central bank issues as much credit as it can), R = ro(R+Cr) always holds and we have R proportional to Cr, which is equation (2). In fact, some countries have obtained a proportionally greater share of credit while others had their allocation of central bank credit kept more in check. By including the zone- 10 wide rule in a country specific model of inflation, we are imposing an additional restriction which in turn enables us to identify more clearly the relationship between fiscal-policy responses to changes in commodities price and inflation. Equation (3) comes from the log-differentiation of the equation defuiing nominal net foreign assets since = _I+ E1(Z+Z2,) + E,(P,CC, - pM M) where C, represents exports of the primary commodity and M, represents imports. We have: p -RI, 41 Zl 14~ =Z2PICCm *2 P" *M P R=-y ' - R ' - R 't1= R- '*2= RT The fiscal side of the model is described as follows. Receipts from commodity exports are the main source of government revenues in these countries. As pc* increases, the government starts spending (e.g, new projects, increases in public employment or in public sector wages). The longer the commodity boom lasts, the more difficult it will be for the government to reverse its spending pattern since it faces policy reversal costs (which increase with the length of the commodity boom) such as disinvestment costs for ongoing projects and firing costs for the public sector labor force. Once there is a commodity bust, the government must decide whether or not to reverse its spending policy. If the policy reversal costs are high, the government will prefer to let the public debt build up. Once the debt 11 build-up costs exceed the policy reversal costs, the government will start reversing its spending policy. (See, Boccara, 1991 for a description of policy reversal costs and of other effects which characterize the environment in which fiscal policy decisions must be made.) This discussion suggests the following equation to characterize fiscal policy: d = ytc + cG (4) The coefficients on p, and its lagged values characterize the extent of the government response to changes in commodity prices while the coefficient on GQ., chararterizes the inertia in the government's spending pattern'. The percentage change in government revenues is equal to: C. .C A ~~~.M. ~M(5 PI = 'E, + 11(fir+ , IC ,) + (1- 0)(P,' + M,)(S Equation (5) comes from the definition of trade taxes, T = E (TCPC¢C +rMPM*M) 4Equation (4) does not differentiate between positive export commodity price shocks and negative export commodity price shocks. In practice, the reluctance to reverse a spending pattern would be obrervnd only for the transidon from a commodity boom to a commodity bust while the government would immediately ieadjust its spending pattern if there is a commodity boom. It can be argued that the observed behavior is consistent with a high prior for ?rob(commodity boom at t I commodity bust at t-1) and a low prior for Prob(commodity oust at tI commodity boom at t-i). 12 and we have C P, Cc ~CPCs+tA Ms cp,c C + 'r pM M The percentage change in the export tax for the export commodity is given by: = +C (6) tA = %i, + XE, Equation (6) corresponds to the fixed producer price of the primary commodity export, so that all of the windfall gains from the price surge accrue to the government. From EPc'(1-,c) = Pc =constant, we obtain equation (6) with ;:=X i _c rm is assumed to be exogenous. On the supply side, the production function for nontradables is assumed to be of 13 the form: N, = A, Min fk,tLN); '). where f is a fi,nction (e.g. Cobb-Douglas) of k, and of the employment level 4. This assumes that a fixed amount of imported intermediate goods (3 for each unit of nontradables) is necessary to the production process which otherwise uses domestic factors of production. Profit maximization can be written as: Max [(P, - P, fp kI,LN,) - w1L,,,- rk,] with the capital stock k, fixed in the short run. The price that matters for production is P = pN - M and we have with A= M 14 where X is the share of value added in production. Percentage changes in the supply of nontradables are a function of percentage changes in the price P' and in the capital stock Ik and therefore, we have: N, = function (A, ,i,') We also have =k k-i s- K where I is investment and K the nominal capital stock. Kt is related to past investment (i.e., lagged once). Any positive contribution to savings, either from foreign sources (Z2*,t) or from domestic sources -- essentially public sector savings (i.e, primary surplus T,.,-G,.,.,) -- contributes to capital formation. Incorporating the identity between savings and investment gives us the following equation: I I Z2 T -G --1 N, = At + v(-2-) + v( v ) it + v' P K, Kt' and finally, N, = A, + v( ,: ') + v( ' K-') - VP, + vIlXpN + v (l-)p,U (7) K, 1s where v is the supply elasticity of capital for the production of nontradables and v' is the price elasticity of supply. Note that the coefficients for Z2.,,1/K, and for (T,.,-G1,,.1)/K, are equal since we do not differentiate among sources of savings when looking at capital formation. On the demand side, we assume fixed spending shares of imports and nontradables. We have V H, = PtN Nt + PIMM where V designates the constant velocity of money and we have substituted for absorption HV, using the quantity theory equation. Let a designate the fixed spending share on imports. Then, a VH, = P, MM, (I -a) V H, = P,NN, Therefore, we have Mg = Ht - iM (8) Nt = Ht Pt i(9) We assume the quantity of commodity exports to be exogenously determined. 16 The domestic price level and the domestic price of imports are defined by p = (p Ma (Pt N) -6 M U NMs P, =El (1 +t,) P Exports, other than the primary commodity for which domestic consumption is marginal, are excluded. Log-differentiating the equations above, we fi .d'y have a= ap + (l-a)N (10) .M ~~ M M. Pt' = Et + (a) rM + pt' (11) with I +TM 2.2 Estimating the Model In what follows, we ignore changes in productivity. Combining equations (1), (2), (3) with equations (8) and (11), we can see that the growth rate of money at time t can be expressed as a function of the growth rate of money at time t-1 and of the percentage 17 changes of exogenous variables at time t. We obtain the following equation for the monetary side of the model: 1+2 1 (P 2 -1 + 2t -*t + ¢2 2 + *1 5 + +*2 t + *2 (t )M Equation (4), the first equation of the fiscal side of the model, is repeated here for convenience. It expresses the percentage change in government expenditures at time t as a function of the percentage change in government expenditures as of time t-1 and of the percentage change at time t (including lagged values) of the international price of the main export commodity. G, = Eyi3tC + (II) Combining equations (5) and (6) with equations (8) and (11), we obtain the second equation of the fiscal side of the model: P = i(1+x) g + IJ(l+X) C +(1)(1) M + (-) This equation shows that the percentage change in government revenues is a 18 function of the percentage change in exogenous variables at time t and of the growth rate money at time t. Combiring equations (9), (7) and (10), we obtain an equation for the percentage change in the price of nontradables: 1+9'l-M(l-a, (I - Yg +(vU-V,-X)) A Pt= I 1+vl;X-v(1-U) K(jI This corresponds to the reduced form for the price of nontradables which clears the market. Combining the equation above with the price equations (10) and (11). we obtain the inflation equation of the model: Pnt = [ ] 1) (At - v I' ) A(t tt (IVt) 1+V'X-v(1-a) Kt where the function F, is equal to: [a + (1-a) (Vc-V'(1-X)) (4 + M+ 1 +v'l-v(1 -a) We now have defined a system of four equations. Equation (I) corresponds to the 19 monetary side of the model; equations (II) and (III) correspond to the fiscal side; equation (IV) corresponds to th inflation side of the model. This is a system of equation with 4 endogenous and 6 exogenous variables. A change in pc at time t affects nominal money at time t and government spending for a number of periods equal to the number of lags in equation (II). Government revenues are in turn affected for only one period at time t by the change in Pc' and by the change in nominal money. Inflation is a function of the change in nominal money at time t and of lagged investment, itself a function of lagged government revenues and government expenditures. The equations above capture the link between the change in the international price of the main export commodity and the inflation rate. Since the Franc zone countries are in a monetary union with a common currency pegged to the French franc, the only adjustmnent in the real exchange rate (ignoring changes in import taxes) is through the domestic price level. The difference in the adjustment experience of the CFA countries is naturally related to the choices of fiscai and monetary policies. The countries share a common central bank and the overall money supply is largely an exogenous variable. However, the distribution of the money supply, within the union, is left to the discretion of the central bhn, and is related to the fiscal policy adopted by each member country. Thus, the key behavioral difference between the member countries is their fiscal policy and more precisely the government's response to the fluctuations in the international price of its main commodity 20 export. For this reason, the model was specified with the same monetary rule for all the countries. In theory, as seen earlier, this monetary rule is only valid for the zone as a whole. Whether in a commodity boom or bust, it is the fiscal side that governs the macroeconomics since the price of nontradables pI, the only variable that can freely adjust in the Franc zone countries, is affected by government's demand for nontradables which in turn is a function of the fiscal response to the fluctuations in the international price of the main export commnodity. The essence of the model is that the behavioral diiferences in fiscal policy between the CFA countries are reflected in differences in inflation rate between these countries. III. ESTIMATION AND SIMULATION RESULTS In this section we test empirically the model of inflation differentials by estimating it for one country, C6te d'Ivoire, and by conducting a simulation with the estimated parameters. Section 3.1 presents estimation results for a slightly modified version of the model presented in equations (I), (II), (III) and (IV). Section 3.2 describes the simulation. 21 3.1 Estimation Results The data used for C6te d'Ivoire came from various sources, International Financial Statistics (IMF), African Economic and Financial Data (World Bank), and World Bank country reports. Variables for which we had several data sources were systematically checked for consistency. Excluding missing observations, the final data consists of yearly observations from 1970 until 1987. Thrnc variables had to be constructed, capital K, an index of commodity export price PCOM and government revenues T. The capital variable is needed since it appears in equation (M). Relying on investment data, a time series for K was constructed starting in 19605 using an initial capital output ratio of 5 and a depreciation rate of 5 percent. Taylor, (1979) suggests as a rule of thumb a depreciation rate of 4% and a capital output ratio of 3. Since he was referring to developing c3untries as a whole, we use slightly higher figures for African countries. The two main export commodities for Cote d'Ivoire are coffee and cocoa. PCOM is constructed as a weighted sum of the international price of coffee and of cocoa (whose movements are incidentally strongly positively correlated), the weights being the 5K is built recursively as K, - K, (1-6) + 1, with Kto - S Ydo. Although estimation starts at 1970, we started the series in 1960 so that our final result would be less sensit5ve to the selected value for the inidal capital output rado. Several values of & were tried (3%, 5% and 7%) but did not modify the estimadon results greatly. 22 quantity shares (Q /cocQ. + QOff.) and (Qc.f/Q,. + Q fe.).6 The quantity shares were obtained from African Ec.onomic and Financial Data. Data on governmemn revenues were constructed using data on government (current) expenditures and on the government's (primary) deficit available from World Bank documents. Difficulty in obtaining reliable information on government expenditures and revenues prevented us frrm testing th model for a country other than C6te d'Ivoire. As one of the more advanced diversified coastal countries, Cote d'lvoire's experience may deviate from that of a poorer, undi' ersified, landlocked country like Niger. The modifications made to the original system given by equations (I) through (IV) are explained below. First, it is assumed that there are no changes in the tax rate 7 'I and equations (I), (III) and (IV) are modified accordingly by removing the corresponding terms. Second, in equation (IV), the percentage changte in nominal money at time t has 6We use the fact that export revenues PQ are equal to P,Q, plus P2Q2 which implies that P - Pi (Ql1(Ql + Q2)) + P2 (Q2/(Q1 + Q2)) 7Data Yas unavailable for 9?. 23 been replaced by its one period lagged value since equation (IV) gave a better fit when the variable was replaced by its one period lagged value. Recall that the percentage change in money in equation (IV) comes from the log-differentiation of the demand side equations. These equations, in turn, are based on fixed spending shares and on the quantity theory of money. Therefore, specifying that total spending (between time t-1 and time t) on nontradables PN,N, is equal to (l-a)VH, or to (1-a)VH,., are equal approximations, since what is really required under our assumptions is an average of the money stock over the period between t-1 and t. Finally, since investment data are available, equation 'IV), can also be estimated using the investment figures (instead of computing investment as in equation (7)), directly. In that case, an equation defining government revenues is no longer necessary and equation (III) can be deleted from the system'. In sum, the system of equation which is estimated is the following: H,= 1+ (P Ht-1 + * Zlt + 42 Z2 + 1 Pt + +2 E,) (I) $Note that in both cases, keeping or deleting the equation from the system did not (as we would expect under the assu:nption that (III) is well-specified) modihy in any significant way the coefficients from the remaining equation. 24 1.2 *C* G, = 2y tit, + cG, X(II f.0 or i-2 2,t= i0 ytPA; + CG2,..1 (II') which is the same equation as above except that govermnent expenditures G are replaced by government capital expenditures G2, l+v[ -a l - V )+ F, (E, ,P") (I 1+V'X-V(1-cc) K7)+F ElP'(V with F, equal to (1-a) (Va-v(l-1))] + pMs [a + I+'-(-)Ict, Pt The results obtained from the estimation by three-stage least squares are shown in Table 4 and in Table 4'. The results correspond to a final estimation for which insignificant variables that did not contribute to a better fit were deleted. In Table 4', the fiscal relationship is estimated with G2 instead of G (i.e equation (II') instead of (II)). In both cases, there is only one lagged term (in addition to the current term) for the percentage change in the index price of export commodities. The results confirm the hypothesis that fiscal policy is driven by fluctuations in 25 the price of the main export commodity. The coefficient corresponding to percentage change (lagged once) in PCOM is significant. Without a lag, the percentage change in PCOM was found insignificant and this whether or not the variable lagged once is also in the equation. A two period lag structure did not improve the results and showed the percentage change in PCOM lagged twice to be insignificant (at least when the same variable lagged only once was also present in the equation). Estimation with G2 instead of G gave similar results in terms of fit and significance (i.e., the one period lagged change in PCOM is significant while it is not without a lag or with a two period lag). Furthermore, the change in government expenditures lagged once is no longer significant. The fact that lagged percentage change in G2 is no longer significant reinforces strongly our hypothesis since usually a one period lagged value of the independent variable is expected to be significant (unless there is lack of correlation in policy orientation between one time period and the next).9 The significance of the lagged percentage change of G (which is equal to G, plus G2) is explained by the fact that G, (government consumption) responds much less to a change in the price of the export commodity than G2 (government investment) so that G and its one period lagged value are more correlated than G2 and its one period lagged value. Therefore, the imputed fiscal response to fluctuations in export commodity prices is explained mostly by the public sector investment response. The results show that both lagged money growth and current percentage change 9Note that we are not talking about G2 in level (i.e the policy variable itself) but about the percentage change in G2 (i.e the policy orientation from one period to the next). 26 in PCOM are significant in the money growth equation. The other variables which appear in equation (I) were all insignificant. The results indicate that changes in commodity export prices can have (through the unsterilized accumulation of foreign reserves) important effects on the money supply. The results for the inflation equation show that investment, an indicator of fiscal policy, is significant. Other variables and in particular, the percentage changes in the exchange rate (in dollars per CFA) and in the international price of imports, are not significant. Although money creation (lagged once) is insignificant, it becomes significant (whether or not it is lagged), regardless of whether one uses G or G2 in the fiscal relationship, if the inflation equation is estimated with investmeiit proxied by Z2 + T - GI (as is done in the original model, see equation (7)). However, it is again insignificant if investment is proxied by G2 alone. The conditions under which different proxies can be used for investment are outlined in what follows. Using the savings equal investment identity and a decomposition of investment by sources and of savings by sources, it can be shown that using the primary surplus GI - T plus Z2 as a proxy for investment is a valid assumption if private saving is close to zero. Using G2 as a proxy for investment is a valid assumption if private saving is essentially equal to government private borrowing and if foreign borrowing by the private sector is close to zero. 27 It can also be shown that using G2 plus Z2 (where Z2 is the nominal capital account) as a proxy for investment is a valid assumption if again private saving is essentially equal to government private borrowing and if foreign borrowing by the public sector is close to zero. All these assumptions are not necessarily in contradiction with each other. For example, whether public or private foreign borrowing can be assumed to be close to zero, depends upon whether public enterprises are considered to be in the public or in the private sector. To summarize, the results conform well to the predictions of our model and are consistent with our explanation of inflation in a typical CFA country. They highlight the dependence of the economy of Cote d'Ivoire on its two export commodities: coffee and cocoa. 3.2 Simulation Results The results above have shown that changes in commodity export prices have important effects on the domestic price level (and hence, given the fixed exchange rate, on the real exchange rate). The purpose of the simulation exercise is to analyze the response of 2 typical CFA economy (or more generally a typical primary commodity exporter with a ixed exchange rate) to exogenous changes in the international price of its main export commodity. The simulation scenario is that of a commodity boom followed by a commodity 28 bust. The international price of the main export commodity is the only exogenous variable allowed to vary in the simulation and all the other exogenous variables of the model are assumed to be constant. The simulation scenario does not correspond exactly to the actual behavior of the price of coffee (or that of cocoa) since there were two coffee (and cocoa) booms. Figure 5 shows the simulation scenario while Figure 6 shows the actual behavior of PCOM for CMte d'Ivoire. As can be seen in Figures 7 and 8, the behavior of investment and of public capital expenditures are similar in CMte d'Ivoire'°. Thus, equation (IV) is modified with investment proxied by public capital expenditures G2. To be consistent, G2 is also used in the fiscal equation (i.e equation (II') instead of (II)). Using G2 in the inflation equation (instead of investment) simplifies the simulation and emphasizes the key behavioral aspects of the economy reviewed above. Withuut G2, an additional equation relating investment to fiscal policy would be needed as well as a specification of privatte and direct foreign investment (the latter can probably be considered as exogenous). l0The scales of the figures are different. Thus, public investment appears to be a (constant) fraction of total investment. This would suggest that the behavior of the p-ivate sector, as far as investment goes, is similar to the public sector's behavioral response to fluctuations in the price of the main export commodity. 29 Thus, the simulation model is the following: H, = [ X2 ] (p H,,1 + W1 Pt) G2, 5 EYCP1t + Co 2-I i*0 Pt l+vXl -v(l -a) Kt Note that the first two equations can be estimated (or simulated) independently since no other endogenous variable except the independent variable is included. This reflects the fact that a change in the price of the export commodity has, as has been explained earlier, a fiscal and a monetary effect (i.e., the first two equations of the simulation) which in turn translate into inflation. Again it should be emphasized that the key behavioral equation of the model is the specification of fiscal policy. The remaining equations are essentially accounting identities which are written in the context of a primary commodity exporter in a fixed exchange rate regime. To stress this point, the simulation uses only the estimated parameters 30 for the fiscal equation while all parnmeters for the other equations are assigned "realistic" values independently. The values and the rationale for selecting them are given below. In the monetary relationship the parameters selected are Al = 02 = 2 and p - 0.95. Since 4,1 (02) is the ratio of exports (imports) to reserves, the underlying assumptions are balanced trade (at least in the long run) and that the central bank has enough reserves to cover six months of imports. A decrease in reserves would increase AI and i2. p is simply the ratio of reserves from one period to the next. In the inflation relationship, the parameters selected are ot (share of imports including intermediate inputs in the consumption basket) = 0.25; v (supply elasticity of capital for the nontraded good) = 3; V (supply price elasticity for the nontradable good) =0.4 and A (value added in production") = 1.333. Finally, in the fiscal relationship, the parameters used are the estimated values of a fiscal equation (with G2 as in (II')) which is part of a model for which G2 is used as a proxy for investment in the inflation equation. The parameters are r = 0.198, Yo = 0.227 and y = 0.556. Figure 9 shows the predicted public investment. It corresponds to a simulation 'It was shown earlier that X is equal to PN/ (PN - #Pm). With PM - 1 (numeraire) and 6 - 1/4 (i.e. 0.25 unit of import needed for the production of each unit of domestic good), we get X - 4/3 or 1.333. 31 of the estimated fiscal equation and replicates the "hump" in public capital expenditures associated with a commodity boom followed by a commodity bust. Figures 10 and 11 show actual and predicted inflation in CMte d'lvoire. The simulation shows the increase in inflation during the commodity boom followed by a decrease during the commodity bust and a stabilization (with a minor increase) once commodity prices stabilize again. The predicted behavior is again consistent with the data. The differences in levels in the later periods may be attributed to the fact that the simulation scenario does not correspond to the exact values of the exogenous variables in CMte d'lvoire. To summarize, the simulation performs well and could be used for other CFA countries (or more generally countries whose economic structure is similar). Naturally, for the simulation to be more than the tool we use here to enhance the explanation of our model, it would be necessary to have forecasts of export commodity prices'2 and to know the fiscal parameters of a particular country. To get these parameters, one needs only to estimate the fiscal equation since it is independent of the rest of thc system. It is alsn possib!e to use the simu:ati.n under different hypothetical fiscal responses. However, such an exercise is not meaningful since a plausible range for the magnitude of the fiscal equation parameters is not known. At the very least, we would need to estimate the fiscal equation for other CFA 12Although the simulation can be used as a predictive measure of real exchange rate behavior under different scenarios. 32 countries'3. IV. CONCLUSIONS In this paper, we first showed that, despite the monetary union and other extensive ties with France, a fixed exchange rate and substantial intra-zonal labor mobility, there are, at least in the short run, substantial inflation differentials across the countries of the CFA zone. We next proposed a model of inflation differentials for the CFA zone. The model is based on the fact that for primary exporters in general, and CFA zone members in particular, the volatility in commodity prices implies a high variance of government revenues. The model identifies two effects: a monetary effect (commodity booms imply a surge in foreign reserves which, if unsterilized, is inflationary) and a fiscal effect (higher government revenues are, to varying degrees, accompanied by a marked increase in the level of public spending which again is inflationary). Although some countries get a proportionally greater share of the pooled reserves, our model ignores this'4 (by imposIng on each country a constraint on the level of reserves which in fact is only valid for the zone as a whole) and attributes the ceifferences in inflation across countries to the fiscal response. The fiscal relationship is the key behavioral equation of the model since the other relationships 13Data on govermment expenditures (general and on capital) is unfortunately very unreliable and often inexistent altogether. "The rationale for doing this is explained in Section 3. 33 are essentially derived from accounting identities. The model is empirically tested for CMte d'Ivoire. The results show the key parameters to be statistically significant. The model is also simulated. The simulation exercise requires estimated coefficients from the fiscal relationship only since all the other parameters are selected independently. Therefore, once the fiscal behavioral response is known, the simulation provides a tool to analyze real exchange rate responses under different scenarios of fluctuations in export commodity prices. The simulation exercise shows that the path of inflation in Cote d'lvoire can, indeed, be tracked by a model like ours. The rationale for the fiscal behavior itself is left unexplained by the model (see, Boccara (1991) for an analysis of fiscal policy in primary commodities exporters). The typical fiscal response to commodity booms can hardly be considered as optimal. As stressed in Cuddington (1988), the expected response would be to limit the increase in expenditures to the perpetuity equivalent of the present value of the windfall gain. This would imply an increase in net savings in good times and reduction in net savings in bad times. This has, however, limited practical applications since it is often difficult to tell whether shocks are temporary or permanent. Furthermore, since the countries are in a monetary union in which foreign reserves are pooled, it is necessary that if some countries have expansionary fiscal policy (and thus, inflation) that others take a more contractionary fiscal stance. Another issue for 34 ftuture research is what determines whether a member country has the freedom to follow an ex- -nsionary fiscal policy or whether it must contract because other members have already expanded before. A discussion of potential "games" played by countries in a monetary union may help in understanding these issues. 35 REFERENCES Boccara, B. "Fiscal Policy in Primary Commodities Exporter LDC's: An Empirical Investigation of the Franc Zone in Africa", Unpublished Ph,D. Dissertation, Massachusetts Institute of Technology, 1991. Cuddington, J. "Fiscal Policy in Commodity-Exporting LDC's. Working Paper WPS 33, Policy, Planning and Research, World Bank, July 1988. Devarajan, S. and J. de Melo. "Adjustment with a Fixed Exchange Rate: Cameroon, Cote d'Ivoire and Senegal", World Bank Economic Review, Vol. 2, No. 2, pp. 447-487, 1987. Devarajan, S. and J. de Melo. "Membership in the CFA Zone: Odyssean Journey or Trojan Horse?" in Chhibber, A. and S. Fischer, eds., Economic Reform in Sub-Saharan Africa, World Bank, 1991. Edwards, S. " Commodity Export Prices and the Real Exchange Rate in Developing Countries: Coffee in Colombia" in S. Edwards and L. Ahamed (eds.) Economic Adiustment and Exchange Rates in DeveloDing Countries, University of Chicago Press, 1986. Engle, R. and C. Granger. "Cointegration and Error Correction: Representation, Estimation and Testing", Eonometrica, Vol. 55, No.2, pp. 251-276, March 1987. Honohan, P. "Price and Monetary Convergence in Currency Unions: the Franc and the Rand Zone", PRE Working Paper, World Bank, 1990. International Monetary Fund, "International Financial Statistics", various issues. Taylor, L. Macro Models for Developing Countries, Economics Handbook Series, MacGraw Hill, New-York, 1979. United .4ations Development Programme, World Bank. African Economic and Financial Data, 1989. 36 Table 3. Cointegration Tests Results Cointegration Tests Results for Price Levels (Augmented Dickey Fuller Statistic, 2 lags) France and major country of each zone C6te d'Ivoire / France Yes -3.56 Cameroon / France No _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _- 1 .4 9 Major country of each zone C6te d'Ivoire / Cameroon | No ____________________________________________________ -1.65 BCEAO countries Niger / C6te d'Ivoire No __________________________________________________ f-1.41 Senegal / C6te d'Ivoire Yes -3.49 Togc / Cote d'Ivoire No _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _-2 .8 0 BEAC countries Gabon / Cameroon No Congo / Cameroon No -1.19 37 Table 4. Estimation Results (3SLS) Equation 1 Fiscal Relationship Dependent Variable HG 17 Observations from 1971:1 until 1987:1 coefficient t-statistic HG 1 .528 3.71 HPCOM .072 .65 HPCOM 1 .233 2.11 RBAR**2 .278 Equation 2 Monetary Relationship Dependent Variable HHl 17 Observations from 1971:1 until 1987:1l coefficient t-statistic _ HH 1 .678 5.32 HPCOM .281 3.14 RBAR**2 .216 Equatior 3 Inflation Depende.nt Variable HP 17 Observations from 1971:1 until 1987:1 coefficient t-statistic HHI1 .114 .95 100* I 1/K 1.09 3.50 RBAR**2 .188 I indicates a one period lag. Any variable whose label starts with an H (hat) is in percentage change. G is government expenditures, H is money, P the price level, PCOM the export commodity price index, I is investment and K capital. 38 Table 4'. Estimaton Results (3SLS) Equation 1 Fiscal Relatonship Dependent Variable HG2 17 Observations from 1971:1 until 1987:1 coefficient t-statistic HG2 1 .233 1.16 HPCOM .251 1.14 HPCOM 1 .241 2.50 RBAR**2 -_.20 Equation 2 Monetary Relationship Dependent Variable HH 17 Observations from 1971:1 until 1987:1 coefficient t-statistic HH_1 .671 5.53 HPCOM .282 1 3.15 RBAR**2 .218 Equation 3 Inflation Dependent Variable HP 17 Observations from 1971:1 until 1987:1 coefficient t-statistic HH 1 .120 1.02 100* 11/K 1.09 3.48 RBAR**2 .189 The notation is identical to that of Table 4. G2 is public capital expenditures. 39 Cow ceuntries 0.12 0.11_i 0.09 0.0U * 0.07 0.06 0.03 0 0.04 0:03 - -0.01 60O i62 I 64 1 iS SI 8 I 70 1 72 1 i4 I ~i I 'a I aiO I82 I 814 I if. 8 I1 63 63 67 69 71 73 73 77 79 a81183 S3 87 t france + Cote d'ivoir O cameron Figure 1 ICECO: Coeatal countries 0.13 0.06 0.07 0.06 0.05 S £L 0.04 0.03 0.02 0.01 01 -0.02 Ij ~j 60 62 64 6 668 7 72 74 7 78 80 8 4 8 61 63 65 67 69 71 73 75 77 79 81 33 65 87 time Figure 2 2nw +tg oedker 40 *CEA: LandloCkeg cUnwtr1*6 0.12I -0.04 _ 0.07F 0.050 6 0.04 0.02 0.01 0 -0.0`1 -0.02 -0.03 -0.05 6'01216'41616e I1 7'o 7I21 i4 17 I76 160 iS l e11 '2 14 1i1s " 61 63 65 67 69 71 73 75 77 79 61 U U5 87 time 0 bwrdno fmeo + niqer O cete, d1vIr Figure 3 UEAC countries 0.12 0.11 0.1 0.03 0.05 0.07 0.06 0.05 = 0.04 0.03 0.02 0.01 0 -0.01 -0.02 -0.03 -0.04 O~~~~~~~~~6 1 gI6 iO J 2I4I g1t 11 63 63 67 63 71 73 75 77 73 31 U 6 I7 0 Cngo gebe. * c Figure 4 41 Export Commodity rrice O ~ ~~~~~~~~~~/\ c.o aN o 2 6 8 10 I :c :3 i Figure 5 Simulated Export Commodity Price Cote D'lvoire Poreign Pr1ce of Expe* CmmdniL Pcom 110 go L r !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~I 70h IIi 8 0 40 1' 10 T I \ I 6 71 73 75 77 79 AlUUSS 70 72 74 76 76 30 32 84 3e Figure 6 42 Cote D'Ivoire Inv.gtmornt I 700 goo 600 e--5/ 500 / 400k 300 r 200 01 I ' X ' ' ' ' I -- 69 70 71 72 73 74 75 76 77 75 79 50 SI 82 53 54 55 86 87 Figure 7 >°' Cote D'Ivoire GovL Invostmont Expenditure G2 320 300 - E E 260 \* vW r ,!/0 '\i' X I 240 ~ 220j 200k [ / \ (10 |:ti /i' 120 / 1001- A ,oo , so- SoF 40 h_ _ _ _ _ _ 20 - - -- I 69 71 73 75 77 79 51 53 55 57 a6 Figure 8 ye 43 Govt. Inv. Spending ° 0 2 4 6 8 1 0 1 2 1 4 ; 6 1 8 20 periods Figure 9 Simulation Results for Government Investment Spending 44 Cote D'Ivoire X Chonge in P (Inclation) 0.12 0.11 0.1 0.02 _ 0.01 8 0.07 0.06 0.05 0.04- 0.03 0.02- 0.01 0 -0.01 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 15 36 87 85 Figure 10 Inf lItion co 0 0 o I 0 2 4 6 8 10 12 14 16 18 20 periods Figure 11 Simulation Results for Inflation 45 Policy Research Working Paper Series Contact Title Author Date for paper WPS1 178 Productivity of Public Spending, John Baffes September 1993 C. Jones Sectoral Allocation Choices, and Anwar Shah 37699 Economic Growth WPS1 179 How the Market Transition Affeced Bartlomlej Kaminsld September 1993 P. Kokila Export Performance In the Central 33716 European Economies WPS1 180 The Finandng and Taxation of U.S. Harry Hulzinga Septembsr 1993 R Vo Direct Investment Abroad 31047 WPS1 181 Reforming Heatth Care: A Case for Zeliko Bogetkc September 1993 F. Smith Stay-Well Health Insurance Dennis Heftley 36072 WPS1 182 Corporate Governance in Central Cheryl W. Gray September 1993 M. Berg and Eastern Europe: Lessons from Rebecca J. Hanson 31450 Advanced Market Economies WPS1183 Who Would Vote for Inflation in Cheikh Kane September 1993 T. Hollestelle Brazil? An Integrated Framework Jacques Morisett 30968 Approach to Inflation and Income Distribution WPS1 184 Providing Social Benefits in Russia: Simon Commander September 1993 0. del Cid Redefining the Roles of Firms and Richard Jackman 35195 and Government WPS1 185 Reforming Hungarian Agricultural Morris E. Morkre September 1993 N. Artis Trade Policy: A Quantitative David G. Tarr 37947 Evaluation WPS1 186 Recent Estimates of Capital Flight Stijn Claessens September 1993 R. Vo David Naud6 31047 WPS1 187 How Should Sovereign Debtors Andrew Warner September 1993 J. Queen Restructure Their Debts? Fixed 33740 Interest Rates, Flexible Interest Rates, or Inflation-indexed WPS1 188 Developmentaiism, Socialism, and Mario Marcel September 1993 S. Florez Free Market Reform: Three Decades Andr6s Solimano 39075 of Income Distribution in Chile WPS1189 Can Communist Economies Alan Gelb September 1993 PRDTM Transform Incrementally? China's Gary Jefferson 37471 Experience lnderjit Singh WPSI 190 The Government's Role in Japanese Yoon Je Cho September 1993 T. lshibe and Korean Credit Markets: A New Thomas Hellmann 37665 Institutional Economics Perspective Pollcy Research Working Paper Serie Contact Title Author Date for paper WPS1 191 Rent-Sharing in the Multi-Fibre Geoffrey J. Bannister September 1993 A. Daruwaia Arrangement: The Case of Mexico 33713 WPS1 192 Effects of Tax Reform on Argentina's Jacques Morisset September 1993 G. Carter Revenues Alejandro lzquierdo 30603 WPS1 193 The Armenian Labor Market in Milan Vodopivec September 1993 S. Florez Transition: Issues and Options Wayne Vroman 39075 WPSI 194 How Fast Has Chinese l..dustry Tom Rawski September 1993 E. Khine Grown? 37471 WPS1 195 The Enterprise Sector and Mark Schaffer September 1993 E. Khino Emergence of the Polish Fiscal 37471 Crisis, 1990-91 WPS1 196 Corporate Tax Structure and Jeffrey Bernstein September 1993 C. Jones Production Anwar Shah 37699 WPS1 197 Determinants of Inflation among Bruno Boccara September 1993 C.Jones Franc Zone Countries in Africa Shantayanan Devarajan 37699 'N,.:. 34 gj.,w,S.-gzfl. g.2s:k -e, S* : XiA U-A..S :44