i ~ ~ ~ ~ ' /'>1 t' 'l / Policy, Research, and External Affairs WORKING PAPERS Macroeconomic Adjustment and Growth Country Economics Department The World Bank October 1991 WPS 794 How Do National Policies Affect Long-Run Growth? A Research Agenda William Easterly Robert King Ross Levine and Sergio Rebelo Which policies strongly affect long-run growth? Do policies explain why some poorcoLintries have stagnated and others have advanced? Do policies explain successive periods of rapid growth and stagnation in the sanie country? To what extent do national policies - rather than external influences - explain the stagnation of mainy countries in Africa, Latin America. and Asia in the 1980X()s I he Vi!ic%. Re,cirk Ii. a,nd I:xteniai! lair ('o:np!ry di,inh,:e, PR I. VN kirkirg !Pa;r' to, d',erl:litc tic I ii:digg of iiork in pri gre-\ did Li C1,l0aUdge the exclidnge of idead ariieog lank stIif and all iithers mnteremied n devclopmient rissuc% IThec pij[nr. canry thc n4me' of the a.ath)r, r,flcnit iinh their %ICewS and 'hold he used anid encd accordingy 'I he findings. interpretida ns, arid conchlsilon are LirC author' own Ihec shojlid not he dttributed to the World ltfnk t%s lBoard of D)irectors. its manageinent. or an if ts mcnhcr conints Policy, Research, and External Affairs Macroeconomic Adjustment and Growth WPS 794 This paper -a product of the Macroeconomic Adjustment and Growth Division, Country Economics Department -is paln ol a larger effort in PRE to assess the effecL of national policies on long-run growth. This research was funded by the World Bank's Research Support Budget under research project "Do National Policies AlITetc Long-Run Growth?" (RPO 676-66). Copies are available free from the World Bank, 1818 H Street NW, Washington DC 20433. Please contact Rebecca Martin, room Ni 1-053, extension 390(65 (60 pages).October 1991. The broad lorces behind economic growth- stagnation of many countries in Africa, Latin accumulation of produced factors, specialization, America, and Asia in the 1980s? They discuss economics of scale, and externalities - were five national policies: sketched out by the classical economists long ago. These same forces have been used in the * Fiscal policy. What are the growth effects of development literature to study various aspects different types of taxes and public spending (for of economic growtih. By building on the insights example, how do those that affect consumption ot growth and development economists, the compare with those that affect investment)? "new" theoretical literature on growth is also Does fiscal mismanagement create uncertainty contributing models that identify specific chan- that reduces growth? nels through which national policies may affect . MonetarY policy. Do countries with high long-run growth rates. But the empirical work on inflation tend to grow more slowly? Does the linking national policies to growth is still evolv- variance of monetary growth and inflation matter ing, and many basic issues about the long-run for growth? relationshiip betwecni policy and growth remain * Trade intervention. Do distortionary inter- unresolved. Among these issues: the effects on ventions (tariffs and quotas) in foreign trade growth of goverment size, trade policy, intema- affect growth or do they have only one-time tional capital flows, the allocation of public level effects? L es instability in trade and spending, an d the linancing of public spending. exchange rate policy affect growth'? F Financial policies. Do penalties on domestic Easterly, King, Levine, and Rebelo suggest financial intermediation affect growth? How that there are important opportunities to empiri- strong are the effects on growth through lower cally evaluate the theoretically predicted chan- investment and those through inefficient alloca- nels from policy to growth. They contepd that tion of investment? researchers should improve the design of cross- * Openness toforeign capital. How do country studies of growth and should conduct restrictions on direct foreign investment affect more detailed longitudinal case studies. growth'? Easterly, King, Levine, and Rebelo propose Their analytical framework is based on the a research agenda based on the endogenous simple idea that all factors of production can be growth literature, designed to address the ques- increased by investing in human or physical tions: How do national policies affect long-run capital. Economic growlh will be related to growth'? Which policies strongly affect long-run policies that affect the incentive to invest and the growth'? Do policies explain why some poor efficient use of capital and intermediate inputs. countries have stagnated and others have ad- Such a framework can be used to consider which vanced'? Do policies explain successive periods policies affect the long-run growtih rate rather of rapid growth and stagnation in the same than affecting simply the level of income oncc country? To what exctent do national policies - and for all. ralner than cxternal iniluences - explain the The I'RE Working Paper Scries disseminales the find(lings of work under way in the Bank's Policy, Rcscarch, and Extcrnal Affai7sG)mplex. An ohje tie .ofthe erie.s is to get thesc findings out quickly, cs'cn ifpresentations are lcss than fully polished. The findings. interpretations. and conclusions in thcsc papers do not necessarily represent official Bank policy. Produced by dhc PRE Dissemination Center TABLE OF CONTENT Executive Summary . ...................................................... 1 I. Objectives and strategy ................................................... 3 A. Growth experience of developing countries ............................ . 4 1. Experience over 1965-88 ...................................... 4 2. Long-run experience ......................................... 4 B. Contribution to knowledge of policy-related issues ................ 6 C. Empirical methodology ................... 9 D. Relevance of the agenda to policy .................................... 10 II. Design of the research agenda .............................................. 12 A. Analytical framework . ............................................. 12 1. Policies that affect incentives to capital accumulation ................. 12 2. Policief that affect efficiency of resource allocation .................. 17 B. Synthetic issues .................................................. 27 1. Growth and welfare .................. ....................... 27 2. Persistence of growth rates ........... ......................... 32 C. Empirical methodology ............................................. 34 1Il. Conclusion . .................................................. 42 Appendix I .................................................. 43 Bibliography . ..................................................... 49 Box 1: Measurement of growth . .............................................. 11 Box 2: How does growth begin? Why does growth stop.? .......... ................. 16 Figure 1: Average growth rates, 1870-1988; Growth rates, developing countries, 1870-1988; Growth rates, developed countries, 1870-1988 ............................. 7 Figure 2: Growth rates of output per capita, 1965 to 1989 .......................... 20 Table 1: Cross-country averages 1960-89 ........................................ 3 Table 2: Growth rates of output per capita, 1965 to 1989 ........................... 5 Table 3: Inequality and growth ......................... 29 Table 4: Income and social indicators ......................... 31 Table 5: Persistence of policies and growth performance ........... ................. 33 Table 6: Summary list of hypotheses to be tested . ................................ 36 Table 7: Data for analysis of policy and growth . ................................. 37 Table 8: Internationally available data sources . .................................. 38 Table 9: Matrix of correlations across policies . ................................... 41 EXECUTIVE SUMMARY Which policies, if any, strongly affect long run growth? Do policies explain why some poor countries have stagnated and others have advanced? Do policies explain successive periods of rapid growth and stagnation in the same country? To what extent do national policies--rather than external influences-explain the stagnation of many countries in Africa, Latin America, a.id Asia in the 1980's? Ths paper sets out a research agenda based on the endogenous growth literature designed to address the question "Do National Policies Affect Long-Run Growth?". Policy Issues This paper discusses five national policies: (1) Fiscal liy. What are the growth effects of different types of taxes and public spending, such as those that affect consumption versus those that affect investment? What is the growth impact of public investment? Does fiscal mismanagement create uncertainty that lowers growth? (2) Monetarv 2olicv. Do countries with high infQtion tend to grow more slowly? Does the variance of monetary growth and inflation matter for growth? (3) Trade intervention. Does distortionary intervention in foreign trade with tariffs and quotas have growth effects or only one-time level effects? Does instability in trade and exchange rate policy affect growth? (4) Financial policies. Do penalties on domestic tmancial intermediation affect growth? How strong are the effects on growth through lower investment and those through inefficient allocation of investment? (5) Openness to foreign capital. How do restrictions on direct foreign investment affect growth? Importance of the Issues Many of the developing countries have been undergoing adjustment for several years. Progress has been made, but the recovery of growth has been slow in coming. The critical issue in the 1990's will be designing policies to enhance per capita growth. This is necessary both to improve the standard of living and to service the loans they have received in support of adjustment efforts. Identifying the critical policies for restoration of growth will help frame policy recommendations in adjustment loans. Summar of analytical framework This paper's analytical framework is based on the simple idea that all factors of production can be increased through investment in human or physical capital. Economic growth win be related to policies that affect the incentive to invest and that affect the efficient use of capital and intermediate inputs. Suchi a framework can be used to consider which policies affect the long-run growth rate, as opposed to affecting the level of income once and for all. 2 The framework can also be used to analyze complex interactions among policies and initial conditions. In a highly distorted economy, a minimum degree of policy reform may be needed to have anv growth effect. The way policies interact will determine their net effect. For example, raising taxes may raise or lower growth depending on whether the taxes are used for government consumption or investment, whether the taxes penalize private consumption or investment, and whether the tax is easily evaded such that raising a given amount of revenue is highly distortionary. Empirkal methodoloqy The analytical framework derives testable predictions regarding the relationship between the national policies and long-run growth. We outline a variety of econometric and qualitative techniques to examine the accuracy of these predictions and determine which policies are most important in promoting growth. We argue that researchers should use fairly aggregate indicators of national policies. We recommend this approach for two reasons. First, in examining a broad collection of policies, it is not feasible within the context of one study to construct detailed cou"ntr-by-country indicators of every aspect of fiscal, monetary, trade, international capital, and domestic financial policies. But one should account for potentially important interactions among national policies, so that it is important that one include aggregate measures of all of these national policies in a study. Second, it is of interest whether one can predict the growth performance of countries by observing commonly used measures of national policies. If we can use aggregate measures of national policies to predict growth performance, then these measures will be useful target indicators in formalizing policy reform packages. The empirical methodology described in this paper consists of cross-country and pooled tests of the theoretical predictions. Although there is a large literature that regresses average growth rates on various explanatory variables, this approach could extend this literature by (1) using an analytical framework to consider the broad range of national policies listed above and interpret the results, (2) conducting sensitivity analyses to gauge the robustness of the results, and (3) examining theoretical predictions that have not been previously tested. The research could also examine the relationship between growth and broader measures of welfare such as social indicators and environmental measures. Preliminary evidence suggests high correlation between growth and other welfare measures, but the research could look at what factors could cause them to diverge. 3 The belief that economic policy is a major determinant of economic growth has been expressed in the writings of economists for over 200 years. Much empirical work in the development literature has demonstrated such a Unk.' For example, Table 1 shows how over the past 30 years, fast-growing countries have had less government consumption, lower inflation, kss of a black market premium, and more trade than slow-growing countries. TABLE 1 Cross-Country Averoges: 1960-89 FYatgrow.rs Slovgrovers Share of investment in CDP 0.27 0.17 Secondary school enrollment rates 0.27 0.07 Primry school enrollmnt rates 0.90 0.52 Governsentl GDP 0.14 0.13 Coare nt consumption/ GDP 0.08 0.12 Inflation rat- 8.42 16.51 Standard deviation of inflation 8.75 19.38 Black mirkat exchag rate prmium 4.S5 75.03 Standard de tion of prem io 6.53 103.69 Shareof exports to GDP 0.44 O.29 Note: mea, per c pits-growth rate*** -L92 Fsstgrez a One- standard deviation greater than or equal to the-men growth rate. (cutoff" 4.0S n-ia) Slovgroverst Ona-standard deviation luea than. or equal. to the mean growth Tate. (cutoff -.0.2; n-I5) However, the nature of the relationship between policy and growth is far from settled. Controversy continues on which policies, if any, have growth effects as opposed to one-time effects on the level of income. For example, some authors have argued that policies that induce distortions are relatively unimportant because they have only one-time effects on income that seldom amount to more than a few percentage points (e.g. Rodrik (1990)). Others question the causality of the relationship between the policy variables and growth of the type shown in Table 1. Substantial controversy remains as to which policies explain relative successes and failures, and to what extent external factors play a role relative to national policies. Other doubts exist whether the relationship between policy variables and growth is the simple linear one usually assumed, or is more complex, with thresholds for effective minimum reforms. 1'Appdz I pnxd a Mbdet survey ofthe itertume. A moen mtecmisuvey an be found in Radt (1990). 4 A. Growth experience of developing countries The issue of which policies affect long-run growth has become especially critical as many developing countries seek to reestablish (or in some cases, establish for the first time) conditions for growth in the 1990's, after the poor record of the 1980's. 1. Experience over 1965288 Table 2 presents growth rates for 1965-89 and 1980-89 for different country classifications. Three important facts stand out: (1) growth rates over the last 25 years vary widely across countries and regions; (2) developing countries have not in general grown more rapidly than developed countries over the last 25 years; (3) the 1980's was a disastrous decade for many developing countries. Sub-Saharan Africa stands out with a poor growth record on all counts, East Asia with an outstanding growth record. However, wide divergences exist among countries in the same region. This record suggests that large variation in growth rates across countries needs to be explained; it is critical to understand the extent to which national policies can do so. Table 2 also presents some ranges of growth rates for growth in socialist countries, illustrating the uncertainty surrounding socialist economic performance. The high valt:zs of estimates imply respectable performance, which has been interpreted by some to support the conclusion that high distortions (such as those induced by planning) do not have growth effects.2 The low estimates imply a strong effect of distortions, since investment rates for most of these countries were very high. Preliminary analysis of developing country data indicates some suggestive regularities. We find that growth over time is not very stable-the correlation for growth in countries acroae successive 5-year periods is low.3 Moreover, poorer countries have more unstable growth rates, which may be related to greater policy instability in the poorest countries. The nigh variance of growth rates is at least suggestive that relatively frequent changes in policy may help explain changes in growth rates over successive periods (and possibly that uncertainty itself depresses growth). We will discuss the analysis of the lack of persistence in growth rates below. 2. Lone-run emerience While this project will focus mainly on the recent period for which most of the data is available, a longer-run perspective is also helpful. Figure 1 shows some estimates of long-run growth (since 1870) in developing and developed countries for which long time-series are available. The striking fact that emerges from this graph is that growth is much more unstable for 2Luca (1988). However, opinion seems to be swingiing towards lower estimates. IECSE is conducting research on the estimation of growth in socialist countria in tansition. 3The crs-peiod estikme ar genealy below .4 and in many cas below .2 (these results arm iable in Levine and Relt (1990b)). 5 Table 2 Growth rates of output per capita, 1965 to 1989 GDP per capita Annual averages Country group 1965-89 198n-89 Low- and middle-Incomn economies 1.4% 0.1% Low-income economies 0.5% -0.2% Middle-income economies 2.1% 0.3% Sub-Saharan Afrlca 0.6% -0.5% Highest - Botswana 9.4% 7.8% Lowest - Niger -3.2% -4.8% East Asia 4.1% 3.6% Highest - Korea 7.7% 8.2% Lerwest - Philippines 1.1% -1.8% South Asia 1.7% 2.3% Highest - Pakistan 2.3% 3.0% Lowest * Bangladesh 0.1% 0.8% Latin America anr the Caribbean 1.0W -1.2% Highest - 60-89 Brazil, 80-89 Colombia 4.2% 1.3% Lowest - Nicaragua -1.8% -3.9o OECD 2.6% 2.0% H:ghest - Japan 4.5% 3.4% Lowest - 65-89 New Zealand, 80-89 Netherlands 1.2% 1.4% Socialist economies (low range, high range) Low High China (1965-88) 5.4% Algeria (1965-88) 2.7% Yugoslavia (196548) 3.4% Czechoslovakia (1948488) 0.8% 3.5% Hungary (1948-88) 1.7% 5.1% Poland (1948488) -0.1% 3.6% Bulgaria (1948.80,1948-88) 2.8% 5.5% Soviet Union (1960-85) 1.6% 4.5% Average of last 5 socialist economies 1.4% 4.4% All averages are unweighted. Oil-dominated countries have been excluded. Sources: Non-socialist economies taken from WDR 1990 and upc,.ed 1989 with World Bank data Socialist economies as follows: Algena - from Summers and Heston (1988) China - from Statistical Yearbook of China 1984. Yugoslavia - from Statistical Yearbook of the SFRY, vareous years Czechoslovakia, Hungary, Poland, and Bulgaria - from Fischer and Gelb (1990), UN (1948), and WOR 1990. Soviet Union - High rate is from official numbers, and low rate is from Selyunin & Khanin (1987). 6 developing countries than for developed countries.4 Most of the developing countries show distinctive starts and stops in their pattern of development. The mc-nt famous case, Argentina, grew slightly above the OECD average over 1870-1913, but has lagg, d well behind ever since (with a particularly alarming decline after 1973). Ghana had respectable growth in 1913-50, then lagged badly in 1950-73, and finally went into catastrophic negative growth after 1973 (from which it is now recovering). Of the 1I developing countries shown, only 2 show reasonably steady growth -- Brazil and Colombia. Aside from the disruption of wars and other exogenous shocks, it is plausible that changes in policy regimes had much to do with shifts in growth performance. By contrast, the greater steadiness of growth in developed countries may reflect a more stable policy regime. For othet developing countries, there is some doubt whether sustained per capita growth has ever taken place. The incisive study of Reynolds (1985) concluded that per capita growth had noi begun in 7 out of the 40) countries he was analyzing.' The 1990 Woirld Development Report defines the level of 'extreme poverty" as US$275 per capita consumption in 1985 PPP prices, while USS370 per capita defines simple "poverty". According to data from Summers and Heston (1988), there are 8 countries below the extreme poverty line, while there are another 14 countries below the upper po%erty line.' If we assume that the po"^rty line (or the extreme poverty line) approximates the range of minimum subsistence income, this would imply that income in these countries today is not much different from what it was in the distant past.' Government policies may help to explain why countries have failed to grow. The analysis of the growth-poverty r.elation will be discussed in a later section. B. Contribution to knowledge of policy-related issues If specific policies help explain a significant portion of the substantial differences in long- run growvth shown here (and thus huge differences in per capita income), this should help convince even th - most reluctant statesman to adopt reforms. As Robert Lucas said, "The consequences for human welfare involved in questions like these are simply staggering: Once one starts to think about them, it is hard to think about anything else."8 Whether policy has growth effects has long been a controversial issue. In the traditional neoclassical framework, based on the work of Solow, policies have an effect only in the transition after a policy change, not in the long run. Recent theoretical models have extended the Solow model by endogenizing technological change, making it either another form of capital or 4The only eceptions to the steady growth pattemn of the develped countties are Japan and Spain. both of whom made the tansition from developing to deveped countries over the perod shown. 5 The countries wer Afghanisan, Banglade. Ethopia, Moambique, Nepal, Sudn. and Zaimes 6The "extremely pooe countries are Zaire, Chad, Ethiopia, Somalia, Malaswi, Tanzania, Ghana, and Rwanda, while the "pooe countries arm Zambia, Burundi Uberia, Niger, Burkina Faso, Guinea, Uganda, Angola, Mal, Sierra Leone, Central African Republic, Togo, Kenya, and Nigeria. It is significant that all of these countries are in Afrka, suggesting there are region-wide factor that need to be consudcred, as wil be done in one of the proposed case studies dcsu ibed below. 7 This argument was suggeted by Lant Pritchett 8Lucas (1988), p. S. 7 Figure 1 Average growth rates, 1870-1988 8.0 6.0 4.0 2.0 .R _ __ 0.0 -2.0 1870-191 3 1913-1950 1950-1973 1973-1988 Developed countries Developing countries Growth rates, developing countries, 1870-1988 8.0 6.0 4.0 -- -- -- -2.0 1870-1913 1913-1950 1950-1973 1973-1988 ARG BRA CHN COL EGY GHA 8.0 6.0 4.0 _ 2.0 88 -2.0 1870-1913 1913-1950 1950-1973 1973-1988 IND MEX PER PHL KOR - _.+e .. . . ... . _ - x._ 8 Figure 1 (continued) Growth rates, developed countries, 1870-1988 8sL 6.0t 4.0~~~~~~~~~~~~~~~- nn~~~~ ~ . ._ 0.0 -2.0 1870-1913 1913-1950 1950-1973 1973-1988 AUS AUT BEL CAN DNK FIN . . ..... - * 8.oH$ 6.0 4.0 2.0 0.0 -2.0 . 1870-1913 1913-1950 1950-1973 1973-1988 FRA DEU ITA JPN NLD NOR 8.0 6.0 4.0 ' . -. 0.0F 1870-1913 1913-1950 - 1950-1973 1973-1988 SWE CHE GBR USA ESP , _~.D* --._ - -**_-e._ 9 hypothesizing spillovers from physical and human capital investments. In the simple model that we use in this paper. all factors of production can be accumula-.ed. Such a model implies long-run growth effects both from the overall incentive to accumulate capital (related to overall taxes, macroeconomic stability, etc.) and from distortion of resource allocation. In this model, fiscal, monetary. trade, and financial policies can affect the level and efficiency if factor accumulation and thus steady-state growth rates.9 C. Empirical methodokly An important and influential strand of empirical research on economic growth has focused on "growth accounting": estimating the proportion of growth attributable to changes in labor and capital inputs, with the residual assumed to represent total factor productivity growth. This paper proposes a different approach. Instead of analyzing the "factor sources of growth", this agenda suggests analyzing the "policy sources of growth". The framework we set out yields testable implications regarding each policy and also illustrate the implications for growth of interactions among policies. Research could investigate the predictions from tLeory in an effort to better understand the relationship between national policies and long-run gro- h. There is also a substantial literature using cross-country and time-ser;es analyses to search for empirnal links between long-run growth rates and economic, political, and institutional variables."0 It is difficult to compare effects of policies from this literature, since authors study different sets of countries, over different years, use different explanatory variables, measure policy indicators differently, and employ different data sets. In addition, most investigators consider only 6 small number of explanatory variables in attempting to establish a statistically significant relationship between growth and a particular variable of interest. For example, many authors who examine the relationship between fiscal policy and growth omit variables measuring trade and monetary policies, while authors who study the importance of trade policy commonly omit fiscal and international frnancial policy variables. Thus, we suggest a number of extensions. First, test a set of hypotheses emerging from a common frameworl. Second, by compiling a comprehensive data base, compare new findings to past findings anJ discover the sources of important discrepancies. This will help extract the most reliable infercrnces. Third, consider a broad range of national policies. This will allow one to quantify which national policies are most important in determining growth and to address important interactions among policies. Fourth, conduct detailed sensitivity analyses to gauge the confidence one should have in the findings and to uncover new areas for economic research." 9However, the coequenc of the level effets in the neclasicl model approach the growth effectr of the "new" models as the capital share becomes large '0Appendt I oontan a survey of the theortical and empirical gwth literature. See also the revies by Chenezy, Robinson, and Syrquin (1986), Chapter I and Lerine and Renelt (1990b). tt See the orinal wor by Learer (198s) and Its appication to growb by Levine and Rendt (1990a). 10 D. Relevance of the agenda to policy The present model used by most World Bank economists assumes a linear relationship between investment and growth (the incremental capital-output ratio--ICOR). The ICOR is often assumed to improve (i.e. decline) in response to reforms such as trade or financial liberalization. Although often criticized for the lack of a clear theoretical foundation, the ICOR model is a useful first approximation to capture the link between investment and growth. However, it gives little guidance to evaluate what kind of reforms are likely to raise growth for a given investment rate, or what the magnitude of such growth effects might be. It also omits the effect of changes in the level and efficient allocation of human capital. Research could make a uwful contribution by providing a framework to explicitly connect policies to efficiency and growth. 11 Box 1:. MUMw M olgrg! We generally assume that GDP is the appropriate measure of national output, and. that it is measwured correctly in the available data. However, we should. take note of serious measurement issues that have been raised inthe literature, and use exsting.work- in the fie!4to avoid. drawing conclusions based:on'spurious growth in output. One difficulty that has been mentioned in the literature is that GDP is uncorrected: for depreciation of capital assets and depletion of 'natural resources (International.: Economics Department (1989), Ahmad et.al. (1989)) Data on depreciation and depletion:. are generally scant for developing countries, ..and definitional issues are: formidable. Nevertheless, a few rough corrections could be made. For example, it has been noted that output growth in oil producing countries is biased. since extraction of. oil is essentially . conversion of an asset into cash rather than true production. Takdng this and environmental degradation into account, Repetto et. al. (1989) lowered the estimate of: growth in Indonesia from 7 percent to 4 percent over 1971-84. In cases where countries with large mineral sectors are being studied, the-project will examine growth in the non- mineral sector to see how much it differs from the usual growth estimates. Another problem with measurement of both levels and growth rates of GDP is. changes in relative sectoral prices across countnes and over time. Drastically different relative prices or misaligned exchange. rates: makes dollar GDP incomparable. across countries -.this problem has been-addressed.in:the..International Compatisons Project. (Summers and Heston (1988)). A similar problem is that constant-price growth rates in a single country are not robust to changes in the base year, since the weighting of different . sectors can change drastically with a change in-the. base .year (Azam, Guillamont, and.. Guillamont (1988)). In the countries we.analyze,we will.look at the robustness of growth: rates and income levels across different methodologies. The results from a research agenda like this should be helpful in evaluating long-run growth effects of adjustment packages. Although the importance of long-run growth is universally acknowledged, little analytical effort is spent on evaluating the growth consequences of policies relative to their short-run macroeconomic impacts. It is weil known that policies to reduce macroeconomic imbalances could reduce long-run growth. For example, many debtor countries followed the combination of raising import taxes and reducing public infrastructure investment in response to the cutoff of external financing after 1982 even though it may well have reduced growth potential ((Corbo et al. (1987), Easterly (1989), Sachs (1990)). This research would also contribute to the academic debate on the size and sign of the long-run effects of macroeconomic adjustment anu structural reforms (e.g. Rodrik (1990), Sachs 12 (1987)). The bringing together of a diverse group of academic economists and Bank staff should generate fresh insights into the relationship between policy and growth. U1. Desigg of the research aeenda The strategy followed in the agenda is to set out and test the simplest possible framework under which policies have growth effects. This framework generally has unambiguous predictions as to which policies have growth effects, and the sign of those effects. We recognize that the framework leaves out numerous interactions that may complicate the relationships set out here. Indeed, whole branches of macroeconomics, public finance, international trade, and finance are devoted to anablzing the magnitude and signs of most of the effects predicted here. Nevertheless, we argue that the broad brush treatment given by the framework of this agenda is a useful starting point. The policymaker and country economist do not have the luxury of analyzing effects in isolation but must evaluate a set of policies together. For this, a simple framework is needed if the analysis is not to quickly become intractable. A. Analytical framework In this section, we set out a theoretical framework relating specific policies to growth. We present first policies that affect growth by affecting the level of investment, then policies that affect the efficiency of investment. Some of the policies enumerated above affect both the level and efficiency of investment, and will be discussed in both sections. 1. Policies that affect incentives to capital accumulation We first present a model in which the level of investment is the only economic variable that affects long-run growth -- we abstract from any investment allocation issues. Policies in this framework alter growth by affecting the level of investment. We assume that output (Y) is proportional to "capital" (K), which is broadly defined to include physical and human capital:12 (1) Y=AK The coefficient A is assumed not to vary either over time or across countries. Labor implicitly enters the model tbrough K, since human capital is utilized only to the extent that people are in the labor force and are employed. That is, if k is human capital per person and N is number of employed persons, then the component of K due to human capital is equal to kN. We assume an economy closed to inflows or outflows of capital, so the rate of saving and 12flb production functon was suggested by Rebelo (1991), and used In the work of Barm (1990, 1991) and King and Rebelo (1990).It bean a resemblance to the linear output-capital modds long used in the development Iitentumc 13 the rate of investment are equal.' If the economywide rate of accumulation of new capital (and saving) is a fixed proportion of olwput, i, then growth of output g will be given bvy (2) g = iA - aS where 6 is the rate of depreciation of capital. We should stress again that investment, i, is broader than the conventional defnition since we include also human capital accumulation."4 We see that growth depends only on the economywide level of investment, i, and the technologically fixed parameters, A and 6. We are omitting an explicit description of how population growth enters the model, although it will be included in the empirical implementation.' Since population growth is considered to be exogenously fixed for the purposes of this proposal, it does not affect the policy- growth relationships we set out whether we interpret growth as per capita or gross output growth.'6 Since growth depends only on investment, we need to analyze how policy affects investment We assume that economic agents maximize the present value of their future welfare extending indefinitely into the future. This is a useful first approximation to the plausible notion that investors must be rewarded with adequate returns to their capital in order to willingly postpone consumption."7 With investment responding in this way, the common growth rate of output, consumption, and capital will then be given by: L3ThIs asumption will be mainutined throughout the analysis cpt for the consideration of direct foreign lnvestmenL Tbe umptiMon of a closed capital market doea not rule out capital flows such as official credits, aid, or rationed commercial loans which can be eas accomodated in this framework. Few developing economies arm integrated with international capital markem so the simplifying asumption of a closed capital market seems reasonable. For those that are open, the model would need to incorporate complications such as adjustment costs of invemenLt Similar cffects of poliy obtain in such an open capital market model (details available upon request. 14lhi creates a potential measurement problem for "investment," to which we propose several imperfect but workable solutionL One is to try to directy measure human capital investment (such as education and health spending); this wiil generally only be possible in resticted smpkl Another is to asume that in the long run, the polices we are consideing do not affect the composition of invtment between human and physial capital this is restrictive, but permits the use of the conventional invetment rate as a prmsy for all investment. Fully, other proxies for human capital investment can be used, such as enrollment rates and health indicatos We will _teimeat with all three approaches and teat applicable restrictions In cas where the measurement problems prove insuperable, we can still use the reduced form relation between growth and poicies to be presented next. t51f labor supply is ewogenous. then it will grow with population To see the effect of population growth on per capita output goth, we noed to make an assumption about the degree of splllwer of capital acos generations If there is complete spilover (so that each new peron is endo with the cisting average, k), then population growth is neutra - there is no effect on per capita growth of higber population grwth. If there is zero or kla than complete spiLkawer. then higher population growth lowem per capita output growth. 16rfhem is a rich literature on economic determinants of fertility nd population growb (see, for eacmple, Birdsal (1989)). We omit the potential feedback from policy to population growth in order to Ulmit the scope of the proposal. assuming that such feedback effects are small in comparion to the direct policy effec on growth that we analye 17EvAe= on the response of saving to rates of return is inconclusive (see Gerovitz (1988) and Schmidt-Hebbed. Webb and Corei (1990)). 14 (3) g = (A - 6 -p)/o where p is the rate of discount and l/a is the intertemporal elasticity of substitution.8 Growth is given by the difference between the net rate of return on capital (A-3) and the rate of discount p, times the intertemporal elasticity of substitution. The higher is the rate of return to capital, the higher is growth. Since many policies affect how much of the return to capital investors can retain, this will be the channel through which such policies affect growth. a. Fiscal policy We consider first a universal tax on income, r. The after-tax rate of return to capital becomes A(1-r), so growth will be given by: (4) g = (A(1-r) - 6 -p)/o The model makes a simple, testable prediction: growth is negatively and linearly related to the income tax rate, with coefficient -A/a. This assumes that the revenues from the tax are not used for anything that itself increases the economy's capital, such as government investment. It also assumes that the tax cannot be evaded. Since government capital and tax evasion both affect the efficiency of investment, the implications of these will be explored in the next section.' We next note that a tax on output, such as a sales tax of rate t, will have the same effect as an income tax in this frameworlk An output tax applied to investment purchases will imply a net rate of return of Ai(1+t). Thus, an output tax shifts down the net rate of return the same way an income tax does.' However, if the sales tax on output applies only to consumption, then there will be no growth effect. Since no tax is paid on investment, there is no effect on the rate of return, and t1We arc auuming that aU individuals have identical preferenc, so we are abstracting from the savings and investment effecta of distributional shifts strcued by, e.g, Taylor (1983). "9Note that if the tax mrvenues are implicitly or aplicity transfenred back to consunmrs, the tax would not affect the overall nte of saving and investment if private vng were a fixed ratio to private income. However, with endogenous investment and saving, the growth effecs of higher taxes are the same readle of whether the tmvenucs from the tax are conumed by the govenment, transferd bk to cosumers or otherwise wastd . 2AAn output tax of rate t is equivalt to an income tax of rate t/(l +t). 1S thus no effect on growth. This model makes the strong prediction that taxing investment (through a general income or output tax) lowers growth, while taxing consumption does not." The implication that investment taxes harm growth while consumption taxes do not could also be examined with cross-country data on the price of investment goods and consumption goods from Summers and Heston (1988), since prices can be assumed to reflect implicit taxes. De Long and Summers (1990) have put the relative price of producers' durables into a cross-section growth regression. Their finding that a low price of producers' durables is associated with high growth is suggestive. We suggest exploring these results further with disaggregated data on relative prices of different types of investment goods, and by separating out the effects of investment prices and consumption prices. b. Monetary policy Monetary policy in developing countries is often driven by the financing needs created by the public sector deficit.' The inflation created by monetary growth operates as a tax on holdings of money. It is reasonable to suppose that money is used to purchase investment goods, and money must be held for some minimum period to make the transaction. Inflation will thus act like an output tax on investment purchases, and the effects on growth will be essentially equivalent to those described for the output tax in the previous section: (5) g = (A/(1 +Br)-6-p)/a where a is the inflation rate and B is a parameter reflecting the average length of time that money must be held in advance of investment transactions. Thus we have the strong prediction that there is a negative relationship between growth and inflation.' This representation of the long-run growth effects of monetary policy as equivalent to a marginal tax is no doubt too simple. As we will examine in the next section, inflation may cause resources to shift between sectors. In addition, the effect of monetary policy will depend on fiscal policy decisions, so that it is impossible to treat them separately. With both income taxes and the inflation rate, the growth relationship becomes (6) g = A(1-r)/(l+&) P p - a 217he rltive vetit ot comumption and income tae been etensively debated in the pubik finance Literature wvthin the eontest of the trditional neoclasicl model (see Atkinso and Stiglitz (1980) for a discunion). Note that the strong condusion of no roth effea of consumption tam would be overturned if labor supply were highly responsive to real wages and the revenues from the consumption tax were completely transfered back to taxpaye Our hypotbesis depends on the joint probability of thene two cmnditi being low. VTbe consisten rltion betwen government revenue, spending bonWng, g h and the inlati u is stressed by and and van Wijnbergen (1969) and van Wljnbergen (1989). 23l is anakogou to resuts obtained for the neoclaca model by Stocknan (1981). See alo Mino (1990). 16 ar .2 Hm doe awh bek? WhY does aWh d? It ao useful to conside model In which growth may not tak, pace a all becaus.of bad policies. mentioned In th Intoduction, not aft deveoplng oounbtes show ovidonco ofu sutaned per capita growth.. Oter economies that previosly had experienced growth seem to have ground to a halt In the sat decades (for example weee zero percopita growthfor Argentina 196 . Nigeria 1960w47, nd the Philippines 197546). We presnh some basic models in which policy detwminee not only the vevsl of growth, but whethr growth akes plabe ta, (Se o N*eson (156). Azariadis and Drazen(199g), Becker, Murphy, andTzurm (1000, Murphy, Slhbro andVihn (198), RPbslo (1U90M and EastEdy j19U0 a. Economies with povert tra In, otlon HlAI, we asumed thsC othr things equal, all countis should have the sao saving rats. h reality poor countries, In paiular those wth Inoonme level nar th poverty lne, tnd to have very low rates of aving. This phnme cn besaily capturd by introducing Into th utili function the eubtn* evedof consumption, such that swngs goe to zeo as income approaches subsistence. This formulation makes th rats of sangs vwy. inelastic when the level of Income is nsa *ubsitence The growthrat of Irncom for an eonorm witih 01_he prefereno is: (1) 9=A (1-) wherex is the capital stock tht Is oonsient wih the subsistnce consumption level. In developed oonombss the term ,/K Is does to zero and so th rats of growth Is basically that given by equation (4). Policie that increae r in such a way that A(t-r) < p+6 will make the econony convergo towad th 'poverty trp' x. Onco the capital stock is near r a policy change that mako. A (1-r) > p+ will lad the eoonomy to expnd. Butth recoveryp roces can be veryslow. it can tak aong time forthe raofexpin to Increase signIfantly and tho brush with poverty will ave permanent tears. A period In which l is temporarily high will make ft Income level In the economy permanently lowe. b. Model with fixed factor So far we have asumed that fixed factors do not enter into productio. However a* model In which fixed factors bcome unimportant only at hIgher Income level, which seems broadly plusble in light of modem induWtrial development, would giv similar ruts. Jons and Manuelil (1990yhav proposed a production function of thsilype, ThI model ha the interetng property that policy can caus an economy to be tuck in zero perOapa growth, where fixedfactorsproventgrowthatheydonth SolowrmodeL Thiswould happenIfthtaxrats oran equivalepnt poenalt on capital accumulatlon is hIgh. A low tendency to save and/or high rate of population growth would make the zero growth equilibrium more likely. The Intuiion behind thertaturo of the-zero growth-trap nth.JoneeManuelli modal la stralghtfoward. A the captal-labor ratio rim, tho after-tax marginal product of captai will gradually decline to a consant which depens on the technology and on tho level of the tax rate. if this consta (e minimum after-tax rate of rturm to capit Is greater than the sum of the populatIon growth rate and tho rate of discount, then consumersc wHI find ft worthwhlls to keep inreasing t captalbor ratio, and growth will ue. if th conditon dor not hold (becaus otaxes hiGh, for exampe, thn consumer.wil Ot hv sufficnt lncne to indefinitely raise capiaWabor ratio, and output per workerwil stagnate; Thits a simiar result to th model of section a. although hers stagnation could lake phaoc. at any Income evel, not lus at subsistence Income. Both of th mods d this ectoon imply inequaliWee involving the t rate, which sugg99t an important poli0y lon. A mInlmumnthrfthold for reformmav be noeded to haw anv elfoct on arowth. 17 This indicates that the growth effects of inflation are smaller, the larger is the fiscal tax rate, and vice versa. Given these complexities, individual researchers will be encouraged to look also at interaction terms between inflation and other policy indicators, as well as at alternative measures of monetary policy. 2. Policies that affect efficiency of resource allocation In this section, we consider policies that affect the efficiency of investment. In other words, the policies considered here have an effect on growth even if investment is unchanged. These results are particularly interesting because they challenge the common presumption that efficiency matters only for the level of income, not for the rate of growth in the long run. If we consider the reduced form relation between growth and policy for the policies in this section, the relationships will generally be nonlinear, unlike the simple linear relationships implied by the policies in the previous section. We need to generalize the model of section 1 to consider two types of capital K, and K2: (7) Y = A F(KI,K2) where A is a technologically fixed parameter and F is a function capturing how the two types of capital can be combined to produce output. We assume that this function displays all the usual properties: constant returns to scale and a diminishing marginal product of each input. Note that any policy that affects the two types of capital the same way will have the same effects as those described in the previous section. However, we want to consider policies that penalize one of the types of capital relative to another. The two types of capital will be given various interpretations depending on the policy described. We will assume that the relative price of the two types of capital is fixed (at unity), either because they represent two alternative uses of the same good, or because they are traded internationally. In the absence of any policy intervention, producers in competitive markets would equate the marginal products of the two types of capital: (8) F, = F2 Using (8), we can solve for the ratio of K, to K2 and rewrite (7) as follows: (9) Y = AOK where K is the total value of the two types of capital (=K,+K.), and 0 is a function of the parameters of the function F, reflecting the efficient allocation of capital. In the fixed rate of investment case, the rate of growth will be given by: (10) g = iAo ^a is where i is the sum of the investments made in the two types of capital goods.' Note that growth now depends on the effi:iency parameter 0, even though A is still rixed. If the rate of investment is endogenous, then growth will be given by the following: ( 11) 8g = (F2 - a - p)lor Growth will respond to the net marginal product of capital (we could put the marginal product of either type of capital, since they are equal according to (8)), as in the previous model. These growth equations are the same as before, with the exception of the efficiency parameter 4. Policies that penalize one type of capital relative to another will lower this parameter and result in lower growth, even if investment is unchanged. Growth would also fail with endogenous investment, since the social rate of return to capital will fall with distortionary policies. a. Fiscal policy i. Differenial taxes We consider now a tax that applies to some types of capital goods but not others. One example would be given where K, is interpreted to be human capital and K2 physical capital, and T is a tax that applies to labor income (income from human capital) but not to profits (income from physical capital). Another example would be where income from capital in the formal sector (K,) is visible to the tax authorities and subject to tax, whereas income from capital in the informal sector (K2) can be hidden from the authorities and evades the tax. In any case of this type, the after-tax marginal products of the two types of capital will be equated: (12) (1-r) F1 = F2 The tax introduces inefficiency, because it induces too much of the second type of capital to be held, and too little of the first type. Thus, higher differential taxes will induce lower growth both for a given rate of investment, and in a reduced form relation where investment is endogenous. In terms of the equations above, the efficiency parameter i will be a negative (generally nonlinear) function of the tax rate: (13) g= A0(r) i -c 50<0 This model would suggest an equation considerably different from that usually tested in the literature: there would be an interaction term between the investment rate and a nonlinear function of the tax rate. The after-tax rate of return to capital (either (1-r)F, or F2) will fall, inducing lower growth also when investment is endogenous as in (11): 24We oontinue to uasune a cloed capital miet so that Imvstment equab saving. If the capita market we fully open and there Pee adjautmt csa to invment in the two goods we would get similar policy effecs to those discussd beklw. 19 (14) g = (F2(r) - a p)la F2' 0 at lowT, 01 <0 at highr, 02 > ° The testable prediction is that growth for a fixed investment rate is higher, the higher the share of government revenue going to productive investment, and is related nonlinearly to the tax rate. With endogenous investment, the equation becomes: (17) g = (F(r,sd) - 6 * p)-o F2,>O at low r, F2,0 Growth is nonlinearly related to the tax rate (positively at low levels, negatively at high levels), and positively related to the share of government spending going to investment. ii. uncertainty and fiscal policy Uncertainty about marginal tax rates can also have growth effects. If the marginal tax rate on the taxed capital good is uncertain (because of macroeconomic instability, for example), risk averse investors will substitute out of the capital good with the uncertain private return and into the other capital good. This uncertainty has an additional negative effect on the economy's growth rate from the distortion induced by the tax itself. In particular, assuming the tax rate on the first capital good is random with an expected value of r, this will introduce a positive risk premium, P, into the equation that determines the allocation of resources: (18) E[(1-r)F1] = E[F21 + P(VAR(r); r) where E[ ] is the expected value operator. P increases when the variance of the tax rate, VAR(r), rises. Thus an increase in the variance of the tax induces a substitution into capital good 2, which lowers the marginal product of capital and the economy's growth rate.' The testable prediction emerging from this analysis is that increased uncertainty regarding the private returns to investment across capital goods alters the allocation of resources and lowers the economy's growth rate for a given investment rate.3 This uncertainty may reflect high and variable fiscal deficits, an "overhang' of external debt, possibilities of expropriation, uncertainty 27 Sa th taCtboo by lnao (1987) for a formal derivation. 28 here cam some dirmsuan when tax uncertaint will not alter savinp decisons. Iis oocus If tax revenues arecmpletly rebetd with certaintY, ta uncauftY is uncurlated with otber policies nd producAvity shod., and agents are identicaL 22 regarding tax enforcement, or uncertainty about the tax rate in specific sectors of the economy.' For example, individuals have to decide how much time and resources to invest in human capital and how much to invest in physical capital. If there is more uncertainty regarding the private returns to physical capital, the uncertainty will cause too much human capital investment and lower growth. Various empirical proxies for these risks will be used to measure the risk associated with tax uncertainty including the variability of the fiscal deficit, the external debt ratio, the variance of average tax rates, variations in the sectoral composition of tax revenues, and measures of political uncertainty that have been used to quantify the uncertain returns to easily appropriated investments (Barro (1991)]. It is worth emphasizing the interactions between tax uncertainty and other policies. The magnitude of the risk premium in equation (18) depends on the existence and level of policy distortions besides tax uncertainty. Thus, the growth effects of increased tax uncertainty depend on the marginal tax rate and other policies.' Similarly, in the case where tax revenues can be spent on activities that increase the return to private investment, the magnitude of the risk premium in equation (18) and the growth effects of increased tax uncertainty will depend on the manner in which government resources are spent. The potential importance of these interactions and interrelationships among policies will be examined empirically by including interaction terms in the cross-country and pooled studies. b. Monetary policy With more than one type of capital good, inflation may have an efficiency impact, in addition to the effect on growth through the level of investment described in the previous section. This would be true if some investment goods required money to be used to purchase them, while others did not. For example, if there is a subsistence or household sector that does not use currency for transactions, then inflation (r) will act as a tax on capital in the modern sector but not in the subsistence sector. Growth will be lowered for a given rate of investment the same way as described for a tax on one type of investment good in the previous section: (19) g = A0(X) i -a 0'<6 Uncertainty about inflation can also slow growth by distorting investment decisions if inflation affects sectors differentially. The inflation tax and uncertainty about inflation will cause investors to devote a smaller fraction of their investments to the sector (which we assume to be Kj) affected by the inflation tax, which reduces economic growth. Formally, (20) [(1-r)/(1 +#x)]E(F,) = E(F2) + P(VAR(r); xr, r) '"For the roic of the cternal debt overhang in creating uncertainty, see Saclu (1988). On uncertainty induced by unsble polida, it is interesting to reall the higher variance of growth ratc in low inme countrim (Section JA1). T7his mie the interesting poambility that policy uncertainty could contribute to a low.incotme trap' of the kind dicusaed in Box 2. " For cample, uncerainty about taxm has less neptive impact on growth the higher is the expected marginal tax rate. This is becaue a high tax rate ilf lowera growth so much that the additonal effect from uncertainty is smalL 23 where x = inflation, VAR(x) = variance of inflation, P( is the risk premium associated with uncertain inflation, and E( is the expected value operator. Empirically, the variance of inflation and of growth rates of monetary aggregates can be used to measure monetary policy uncertainty. Again, it is worth noting that interactions between inflation uncertainty, the inflation rate, and fiscal policy may be empirically important. The risk premium depends on policies other than inflation uncertainty, and the growth effects of inflation uncertainty will depend on the level of other policy distortions.3" Indeed, higher inflation alone could induce greater uncertainty about future marginal taxes because inflation can interfere with the value of collected taxes. Examination of these policy interactions will be an important component of the empirical inquiry. c. Financial pocy Policies toward domestic financial market activities can also h wve important growth effects in this model. Financial market policies affect growth by interfering with the ability of financial markets to manage risk, evaluate and monitor firms, gather information, and mobilize resources. As the discussion above has already indicated, differential uncertainty about the rate of return to investment can reduce growth by distorting the allocation of investment. Assume that sector one is composed of firms that receive productivity shocks.'2 Individuals can diversify away the risk of productivity shocks by investing in a financial intermediary (such as a bank) that owns or lends to many firms. However, if investors cannot diversi4y away this risk by investing in many sector one firms because policies interfere with the ability of financial markets to allocate risk, then the return to investing in sector one, F,, becomes random. This will alter the allocation of investment: (21) E(F1) = E(F2) + P(VAR(FI)), where P is a risk premnium that increases when the variance of returns to sector one firms rises, VAR(Fj). Uncertainty plays the same role in preventing the efficient allocation of capital (where F, = F2) that a tax does. The empirical prediction illustrated in equation (21) is that financial market policies that interfere with the ability of financial markets to help investors diversify risk can reduce growth for a given rate of investment by altering allocation decisions.3 The types of financial policies that can inhibit financial market activities include direct taxes on financial institutions, high reserve requirements, interest rate controls, direction of credit toward favored sectors and other less 31 For eample, in the model above, the risk pemium becomes smaller if the inflation ate is higher or marginal ux racr ar hier for a given vaiance of inflation because higher inflatin itself reducs Investment in the currency-using sector. Tbus, this predic that the higher the infation rate or the higher th fscal tax rate, the smaller is the negative effect of ineasd inflation uieanty on giwib 32 Seaor two firms can ahto reee shocs. The imporgant point is that one sector is riser than another, o that unceainty an alter alloation decisionL 33 Line (1990, 1991) demonates this formal; for specific finncal instutions 24 direct intrusions. Some finanial policies may also directly distort resources by favoring one type of capital relative to another (for example, a credit subsidy for certain types of investments) In the individual case studies, attempts will be made to gather direct measures of these policies. In the cross-country analyses, aggregate measures of the performance of the financial system will be used.34 Financial niarkets do more than manage risk. They evaluate firrs, monitor managers, collect and process information about the national and global economy, and mobilize capitaL These traits are captured by Greenwood and Jovanovic (1990), Greenwald and Stiglitz (1989, 1990), and Levine (1990, 1991). In the context of this proposal's model, these financial market traits imply that restrictive financial market policies can reduce growth for a given investment rate by worsening the allocation of resources. Since financial market policies can both influence the uncertainty faced by investors and the rate of return to private investment, the growth effects of financial market policies will depend on other policies and the growth affects of other national policies will depend on financial market policies.35 Again, these potentially important interactions among policies would be an important and novel part of an empirical inquiry. d. Trade intervention and exch4nge rate policy To consider the effects of trade policy, we now interpret the two capital goods as representing two different types of goods that are both traded intemationally (with price ratio fixed at one for convenience). The second type of capital good K2 is made up of the domestically produced good, used for domestic consumption, investment, and exports. The first type of capital good, K, is imported from abroad. An import tariff (or an equivalent quota) now operates as a differential tax on the first type of capital good. As in the case of a domestic tax differentially applied, a higher import tariff T lowers growth for a given rate of investment; the relationship is nonlinear: (22) g = A (T) i-d 6 O Also the tariff distortion will only lower growth if it applies to investment goods (and intermediate goods); tariffs that distort only consumption decisions have a negative welfare impact but no effect on long-run growth. The testable prediction is that the average tariff rate on investment goods and intermediate inputs should enter in a growth equation as an interaction term with investment. While this result was framed in terms of a single import good, it would also hold for differential tariffs (or quotas) on different types of imported capital (or intermediate) goods. For example, we could interpret the two types of capital goods as both imported, while the domestic oLtput is only used for consumption and exports. Then a higher tariff on one of the imported 34There arc problems with exsting empirical prcfies of financial market policie The prome will be disced below, but they include identificaton of seveely nepthe real inteut rates. ratios of vety broad money menurs to GDP, and the fraction of all financial A held by the central bank. 35 In this model, for cample fiscal and inflation taxes have larger growth effects in an economy with well-functioning fincial marvet, than in an econmy with highly resrctive financal market policies The growth effects of policy interactions ia demonrated forMal in Luine (1991). 25 goods will introduce a differential between their before-tax marginal products, lowering growth for a given amount of investment. This suggests that measures of dispersion of tariffs should also enter the growth equation in interaction with the investment rate. We should also note that uncertainty as to tariff rates will have the same denressing effect on growth (for given investment) as tax uncertainty, suggesting that a measure of variance of tariff rates (or import quotas) over time should be added to the equation. This implies that temRorary import liberalization will not help growth.' Fmally, the effect of exchange rate controls could be examined in this framework. Rationing of foreign exchange at the official exchange rate often leads to the emergence of a black market, where foreign currency is sold at a large premium over the official rate. If allocation of foreign exchange at the official rate is made for some types of imported inputs but not for others, then the black market exchange rate premium acts like a differential tax on those inputs for which no foreign exchange is allocated.37 Again, the relationship between the black market premium p and growth will be nonlinear, albeit always negative: (23) g =A+(p)i-6 a'<0 e. Policies on foreign direct investment The discussion of foreign direct investment often refers to the special benefits thought to be embodied in foreign capital--technological and commercial know-how and other human capital attributes of foreign specialists, the advanced technology embodied in physical capital investments made by foreign firms, etc)' With this in mind, a natural way to treat foreign direct investment is as a separate factor of production that complements domestic capital. In terms of the model that we are using in this section, we can think of capital type 1 as the stock of cumulative foreign direct investment, while capital type 2 is the capital stock owned and operated by nationals. Since capital type 1 contains unique features associated with foreign investment, nationals do not have the option of investing in it; conversely, foreigners cannot invest in type 2 capital. We also continue to maintain the assumption that nationals do not have access to international capital markets. We will study only the case of endogenous investment and saving by nationals. Since foreign nationals have the option of investing their money at the international interest rate, the equilibrium condition for the stock of foreign investment will be that its marginal 36A htceture on unenainty and investment supports this conclusion (e.g. van Wijnbergen (1985) and Rodrik (1989)). 37his assume that the authorities can enforce that the inputs imported at the official rate are used in production and not simply resold on the black market. If inputs ar freely traded on the black market, then the allocation of foreign echange at the offidal rate simply generts pure rents, with no effciency impications. However, generaly resourc are used to evade the controls of the authoritie, or usd to lobby to recdve the rmts, in which case the eficency effecs described continue to hold A large hterature on rent4ed ad smuggling makes these points (e.g. Krueger (1974)). 38See Hediner (1989) and lnotai (1990) for a survey. See also the dicusion in Appendix 1. 26 product be equal to the international interest rate.' We consider policies that levy a differential income tax on the income from foreign capital, which implies that it is the after-tax marginal product that is equated to the international interest rate: (24) (1-r) A F1 = r The marginal product of type 1 capital (foreign investment) will be higher the lower is the ratio of foreign to domestic capital. A higher tax on foreign investment will require a higher marginal product from foreign capital, and thus will imply a lower ratio of foreign to domestic capital in the long-run. With endogenous investment, the growth rate of the economy will be given by: (25) g = (A F2 (r) -a - p)/A F2,< 0 Citizens in the economy will accumulate type 2 capital at a rate that depends on the marginal product of that capital. More foreign investment (because of lower taxes, for example) implies a higher marginal product of domestic capital, so it will imply a higher growth rate. (The ratio of foreign to domestic capital will be fixed in the long run by (24) and thus the two types of capital will grow at the same rate.) The theory thus makes a strong prediction: taxation of foreign capital will decrease foreign investment and lower growth.'0 f. Sectoral policies The model of distortions can be applied naturally to the consideration of sectoral policies. Equation (7) can be used to think of policies that affect the relative prices of different types of investment goods, such as physical versus human capital, or equipment versus structures. Disaggregated data from Summers and Heston (1988) can be used to examine the relative prices of different types of physical investment goods as measures of price distortions. In testing the predictions of the model, one could exploit also the disaggregated data on quantities of investment by type available from this data set. Equation (7) is also a useful short-cut to think of the sectoral composition of inputs :o aggregate production.4" For example, type 1 capital could be interpreted as being made up of manufacturing goods and type 2 capital as being made up of agricultural goods (although we 390foue this determination of foreign direct investment is ovenimplified for the purposs of clarit. Iizondo (1990) contains a survey of the theory. flris argument could be taken to an artme, implying that subsidizins forign capital would actually raise growth. ere are two caves that should be mae to this condusin. First, subidis of foreign capital must be financed-such as by a tax on domeic capitaL With Cobb-Dougla production, for -ampe, growth would dedine with an icreae in the subsidy rate fianced by an increae in the tax rate on domestic capital Secood an inrese in the growth rate achieved by a ubsidyon forign intment actually worawe nce it induces invesment whose return d not coer the oppottunity cost to the economy of the capital whc is given by the inteanstonl intest rate, r. Ther is a long trdition in the literatur cognizing this kind of Immisertzing growth", eg. Bhagwati and Brecher (1980), and Brcher and Dza-Aleandro (1977). The same cooclusion holds true for any other disortion (such as tariffs) that acts as a subsidy to forign Investment. 4tAlthough we speak of all inputs as "capital, this frumwork can alo be used to analyze distortions of pric of intermediate inputs. Such dituorin have analogus effects on growth to those of ditortins of capital goocd pnces. 27 continue to think of output as one composite good). The wedge between marginal products of the two types can be interpreted as a differential tax that distorts the relative price of the two sectors. The tax may be implicit rather than explicit, as could arise from quantity rationing, price controls, differences in effective protection rates, etc. Relative price data from Summers and Heston (1988) could be used also to assess sectoral relative price distortions."2 World Bank data on the sectoral composition of output like that used by Syrquin and Chenery (1989) can be used to examine whether the structure of output (relative to a "normal' structure) is related to such sectoral distortions.' Other direct measures of sectoral distortions could also be used, such as the measures of differential taxation of agriculture developed by Krueger, Schiff, and Valdes (1988). The disaggregated data will be used in the examination of each of the policy areas listed above. The relative price and quantity data will be related to broad measures of the various policies to assess the sectoral implications of policies. For example, one would test whether the existence of financial repression is associated with relatively expensive capital goods, relatively cheap industrial goods, etc. The resulting effects on growth can then be examined in cross-country regressions. g. Interactions among resource allocation policies For simplicity, we have so far presented interaction effects as involving at most two policies together. Of course, in real world applications, there are many policies that affect many types of capital goods in different ways. In particular, some policies may have offsetting effects on growth. For example, one policy may penalize one type of capital while another policy subsidizes it, with zero net effect. While such an exact offset is unlikely, the total effect of a set of distortionary policies on growth will be generallY less than the sum of the individual policy effects. In the empirical implementation, it is important to include interaction terms among the policies to capture these offsets. B. Synthetic issues Some issues cut across policy areas: 1. Growth and welfare A major preoccupation of the development literature is the extent to which per capita income growth translates into welfare improvements for the poor and for the population as a whole. There are three elements that will be considered: (1) how the additional income arising from growth is distributed, (2) how much income is correlated with social indicators, and (3) how 42An anaogous execise is that of Dollar (1990). who looked at the price of the same consumption basket relative to a benchmark country from Summes and Heson (1988) as a measre of trade poicq distortion (including a corection for vatiation of nontradable pic acrms countrie). Dc Long and Summes (1990) is also relevant here ,OA ve-kown regularity discused by Chenery and Syrquin (1989) is the tendency of the share of industry to rise and agriculture to fall as per capita income rs. The approach of the "patterns" literature associated with these authors is to associate deviations from this path with policy repmes and other factos We would folow this approach to eamine specifically the effect of sectoral distoniof. We may ao eramine the predictions of differnt models for the trend in the agricultural shamr For -ampe, Rebelo (1991) presents a model in whi some ectors use Ibred factors (like agriculture using land), wbile other sectots use only reproducible inputs Sustained grth is still poible. and the hare of the fixod factor sectors will fall under some pammeter configurations. 28 income growth relates to the environment.4 While recognizing the large literature on these issues, the project will examine fresh evidence for the relationship between welfare and growth. This is important both to assess whether raising income growth is a sensible policy objective, and to identify any feedback effects on growth from the social consequences of the growth process a. Inequality and growth Table 3 shows a comparison of per capita incomes of the poorest 20 percent of the population and the average per capita income for those countries that have data on income distribution. The table shows a high degree of correlation (.77) between the per capita income and the income of the poorest part of the population -- differences in inequality do not dominate differences in per capita income across countries. If we think of per capita income as the sum of past growth, this suggest that growth dominates redistribution as a factor in the income of the poor. But the outliers give pause to the notion that cross-sectional growth differences are all that matters -- the income of the poorest fifth in Brazil is only sixty percent of that in Morocco, despite Brazilian per capita income being two and a half times larger. Similarly, the poorest in Botswana are slightly worse off than in India, even though Botswana's per capita income is three times larger. However, even in these outliers, the relationship between growth and inequality is unclear -- would Botswana's poor be better off if the country had not grown? A large literature has found similar patterns to that displayed in table 1, both intertemporally and cross-sectionally. Fields' (1989) survey of the literature indicates strong evidence of reduction in absolute poverty being associated with growth, with poverty "more apt to decrease the more rapid is economic growth".4 The notion that absolute poverty tends to increase with economic growth has been decisively refuted. The 1990 WDR found that income of the poorest tenth grew more rapidly than per capita income in 9 out of 11 developing country growth episodes. But outliers still exist -- in Costa Rica, the poor suffered an absolute decline in income despite average per capita growth of 3.5 percent over 1971-86.' The famous Kuznets hypothesis that inequality first rises and then falls with income has also been examined in the literature. Cross-section studies have tended to confirm it, while intertemporal studies --which seem more appropriate -- have found little evidence for it (Fields (1989), WDR 1990). Other studies have looked at how inequality itself may lower growth because it tends to lead to policies that harm growth. Alesina and Rodrik (1991) found that higher inequality tends to lower growth in the subsample of democracies but is insignificant in nondemocratic countries. Almost identical results were found by Persson and Tabellini (1991), alt-ough the significance of 4"We do not consider anotheW imu In this sction that was biefly mentioned elewhere in this propwa growth could be immilulzing becuse policy distortions may eist that cause too much ivsmcet. '5FeI (1989), p. 174. "6WDR 1990, p. 4P. 29 Table 3 Inequality and growth Percent share of Per capita income Per capita income of 1988 income lowest 20 percent lowest 20 (various years, 1980's) percent Bangladesh 9.3 170 79 India 8.1 340 138 Pakistan 7.8 350 137 Ghana 6.5 400 130 Sri Lanka 4.8 420 101 Indonesia 8.8 440 194 Philippines 5.5 630 173 Cote d'Ivoire 5.0 770 193 Morocco 9.8 830 407 Guatemala 5.5 900 248 Botswana 2.5 1010 126 Jamaica 5.4 1070 289 Colombia 4.0 1180 236 Peru 4.4 1300 286 Costa Rica 3.3 1690 279 Poland 9.7 1860 902 Malaysia 4.6 1940 446 Brazil 2.4 2160 259 Hungary 10.9 2460 1341 Yugoslavia 6.1 2520 769 Venezuela 4.7 3250 764 Correlation Coefficien., per capita income and income of lowest quintile: 0.77 Source: World Development Report .990 inequality was rather marginal. Lindert and Williamson (1984) present evidence against the Kaldor hypothesis that, because it is the rich that save, inequality is necessary for higb saving and ghwtL The project would consider inequality in two ways. Frst, measures of inequality (e.g. the ratio of the ircome share of the top to the bottom quintile) will themselves be tried as dependent variables as part of each policy task, to examine whether the same policies that affect growth also affect inequality. Second, inequality measures will be used as independent variables in growth equations to test whether higher inequality can make a given policy more or less damaging to growth. The synthesis task wil explore further the empirical regularities between income grow and inequality to address the effectiveness of per capita growth as an instrument to reduce poverty. 30 b. Growth and other welfare indicators Per capita income shows a high degree of correlation with social indicators such as life expectancy, daily calorie supply, low birth weight, secondary enroUment, infant mortality, maternal mortality, and access to safe drinking water and sanitation services (table 4). The correlation is less strong between income and the crude death rate, population per physician, primary enroUment, and literacy ratios. There is also a correlation between percent changes in calorie supply, population per physician, and infant mortality and income growth, while percent changes in primary and secondary enrollment, literacy, life expectancy, and the crude death rate are essentially uncorrelated with growth.'7 Simple correlations may understate the relation between social indicators and income because the relationship may be nonlinear, as suggested by Ingram (1989). He confirmed a statistical relationship that implies that life expectancy, primary enrollment, and daily calorie intake improve rapidly as income rises from low levels, then level off as income rises beyond a certain middle income threshhold. This is plausible since these indicators are inherently bounded. This helps to explain Ingram's finding of strong unconditional convergence across countries in social indicators, even though he finds no evidence of unconditional convergence in incomes. However, there are siz.zable outliers to the income-social indicator relationship. The WDR 1990 noted a striking phenomenon -- a major improvement in life expectancy and primary enroUlment in Africa over 1965-85 despite negative growth in per capita consumption.4 Conversely, Pakistan registered little improvement in net enrollment over 1965-85 (it was still only 43 percent in 1985) despite per capita growth of 2.5 percent over the period.' The relationship between growth and improvements in social indicators supports the study of growth as a generally useful proxy for broader notions of welfare improvement. However, the existence of large outliers suggests an examination of factors that can cause social improvement and growth to diverge. Social indicators to some extent reflect conscious choices made by the government and by private individuals. For example, the allocation of public expenditure has a major impact on many indicators. In each of the policy studies, we will develop one or more social indicators to be tested as dependent variables along with per capita growth as functions of policies. Some of the social indicators may also be proxies for the level or rate of accumulation of human capital, which suggests their inclusion on the right hand side of those equations that also include physical capital investment. 47Work for the 1991 WDR by Suiit Bhalla found reasonable correlations between changes in per capita income and chae in educational attainment, infant mortality, and political liberty, although the ast correlation is rather weak However, he cites otber studies that claim a high corrlation between per capita income and political and civil liberty (Dasgupta and Weale (1990), Scully (1988)). Looking at infant morality, the 1991 WDR found evidence that both income growth and government health apenditure explain its decIe in developing countries (King and Rowazweig (1991), Bhalba and Gilt (1991)). Correlations between changes in indicators can be seen as a stronger test of association given the posibility of spurious correlations between variables with trendas 4OWDR 1990, p. 40 Another ocial indicator, per capita daily calorie supply, did fall in 16 African counties over 1965-86. (WDR 1990, table 28). *ibild., p. 43. 31 Table 4 Income and Social Indicators 1985 Corrlation Correlation bet.196-85 betweeen per capita per capita growth and income and level of: percent change 1965-85 in Life expectancy .67 .02 Crude death rate -.31 .01 Daily calorie supply per capita .70 .33 Babies with low birth weight (%) -.52 Primary enrollment .28 -.15 Secondary enrollment .76 -.09 Population per physician -.35 .33 Infant mortality rate -.64 -.40 Maternal mortality rate -.43 Percent of population with access to safe drinling water .66 to sanitation services .60 Female literacy ratio .14 Male literacy ratio .18 Deforestation -.11 c. Growth and the enviromnent Another important welfare indicator is the state of the enironment. The poor quality and great scarcity of data has inhibited quantitative work relating growth to environmental degradation. Table 4 shows that one of the few quantitative measures available, the percent rate of deforestation, shows little correlation with aggregate growth (and in fact is the "wrong" sign compared to the popular perception that growth causes deforestation). However, this is just as likely to reflect the unreliability of the data as any lack of relationship between growth and deforestation. 32 Some recent research on industrial countries has suggested that, with proper policies, reasonable economic growth can be compatible with preserving the environment.' For example, the rapid fall in energy use in the OECD countries (23 percent fail in energy requirements per unit of GDP over 1970-87) suggests that other inputs can be elastically substituted for exhaustible resources if price incentives are sufficiently strong (Pearce (1990)). However, this process should not be seen as costless -- some have argued that environmental regulations lowered productivity growth in the U.S. in the 70's and 80's (Jorgenson and Wilcoxen (1989)). In the developing wor.d, deforestation, soil erosion, and pollution have imposed significant economic costs, both measured and unmeasured. It is conceptually unclear whether higher economic growth worsens or improves the environment. The tradeoff in welfare between conventional income and environmental quality is also unclear (and may differ across countries or across income levels). While higher growth implies higher growth in demand for exhaustible resources, it also may imply a shift away from use of exhaustible resources. For example, deforestation in the Sahel and the Philippines is associated with consumption of fuelwood by the poorest segment of the population -- increased income could lead to substitution of less environmentally costly forms of energy for fuelwood. Rising income could also lead to greater demand for environmental preservation.5' 2. Persistence of growth rates A surprising fact is the low persistence of countries' growth performance across periods, as mentioned in the first section. Table 5 shows the cross-section correlation between growth in subsequent decades to be around .3. This captures the persistence of relative performance, since the correlation is with respect to deviations from the average for ail countries. In other words, this correlation measures the degree to which being an above-average performer one period is a good predictor of being an above-average performer the following period. Since this correlation is with respect to the period average, it removes any common global trends in growth rates. Several hypotheses could explain the low persistence of relative performance. Policies. While policy regimes are commonly thought to be highly persistent, this is not true for all policies, as shown in table 5. The government current expenditure variables -- consumption and education spending -- are the most persistent, far more so than growth. 50See the useful srvey of Pearce (1990) and the heuristic treatment of Anderson (1990). SlWe have worked wut an eaample in which natural ources (assumed to be in fued supply) not used in production enter the utility function. Sustained growth equires that the elasticity of substitution in production between reproducible inputs (capital) and nauanl raourcns be greater than one. With a Cobb-Dougi utility function, the use of natural reources in production gos asymptotikly to zero in the optimal plan, ie. all natural resou are preseved for their utility value. Whether the market can repliote the opal plan depends on appropriate pricing of natural resources 33 Table 5 Persistence of policies and growth performance Cross-section Correlation Cross-Section Correlation between 1960s and 1970s between 1970s and 1980s Per capita Growth .26 .32 Black market premium .62 .45 Inflation .40 .41 Labor force growth .70 .79 Government Consumption/GDP .76 .64 Government Education Spending/GDP .75 .74 Government Investment Spending/GDP .49 Real interest rates .48 .32 Trade orientation .56 Comparing 1973-85 with 1963-73, based on 1987 WDR List of four categories: 1. strongly outward oriented 2. moderately outward oriented 3. moderately inward oriented 4. strongly inward oriented However, government investment, trade orientation, inflation, the black marL et premium, and real interest rates are not very persistent (although still more persistent than growihi.52 We will examine the ability of policies to explain the low growth persistence by examining the standard errors of pooled time-series cross-section tegressions that use decennial averages for growth rates and policy variables. Random error. A stochastic growth modeL such as that of Rebelo (1991a), implies that growth will be given by a component dependent on policies and a stochastic term reflecting random shocks (such as those to the rate of return to capital). The policy component itself is less 52Most of theme "poUi" meures contain endosflous elementL However, it sems plausible that ca-country variation in, for emmpke, in&"ti blak madret pumi. and re interat rates a mainly due to poliy chois on money creation, exchange rate oatrals, and interet rae cotwIa, rCuecdvY. 34 than perfectly persistent, as noted above. A plausible variance level in the stochastic term could explain a significant portion of the low persistence of growth rates. We performed some Monte Carlo simulations that indicate that the observed low persistence of growth rates, combined with persistence of about .6 of policies, could be reproduced with a standard deviation of 1.7 percentage points on growth rates. The project will explore further the amount of random variation underlying growth rates. The random error could be interpreted as cyclical movements, random technological or natural shocks (such as a drought), or external shocks.53 External shock. External shocks represent one particular type of random error that could cause low persistence in growth rates. However, studies such as Mitra et al. (1990) and Balassa and McCarthy (1984) have found little evidence for strong effects of external shocks on growth rates.54 Transitional dynamics. Some models in this proposal would imply transitional dynamics that could cause growth to be unstable over time even with unchanged policies. For example, initially rapid and then decelerating growth could arise from transitional dynamics in an initial situation where there are imbalances between the quantities and rates of return of different types of capital. Transitional dynamics could be incorporated in the analysis by, among other things, including income level or initial capital stocks as an independent variable. C. Empirical Methodology Empirical analysis should use formal econometric techniques in evaluating the predictions of the analytical framework. This section (1) repeats the main hypotheses, (2) discusses measures of national policies and, (3) outlines the cross-country procedures. Since the empirical predictions have already been discussed in the Analytical Framework section above, Table 6 simply collects the major hypotheses that could be tested. The major themes that emerge from the analytical framework are (1) national policies or uncertainty about national policies affect long-run growth (and do not just have one-time level effects) by altering the level of investment; (2) policies also affect growth by distorting the allocation of resources; and (3) there are important linkages among policies. For example, for a given level of investment, distortionary taxes like inflation tend to slow growth but productive government expenditures can contribute to faster growth. Thus, the growth effe cts of more inflation may depend on how the resources from the inflation tax are spent. In addition, the model indicates that policies that affect the allocation of resources generally have non-linear effects on growth. Some indicators of national policies are provided in Table 7; Table 8 lists international 53tn the Monte Cuazb Lnulti we assumed tht the rndom enw wa aeriy unorrlated Seal conaatb of an et-Or in the leel of output (like a cydical eor term) couwld induce neptie autocowelatlon of growth rates, which would make obeving a bw crao piod corelation ielky. he simubati is also constained to rprduce the obherved crousctioaal variance. he hedp of LAt Prltdcett on this simulation i grtefuly acknwledged. 5See the Summary of thb kew in Bhalla (1991 WDR. chapter 2). 35 data sources. We recognize that there are conceptual and statistical problems with these measures. Research is needed to design and construct more accurate measures of fiscal, monetary, trade and financial policies. Also, since the listed indicators are commonly used, it is useful to determine whether these variables reliably predict growth. If these policy indicators are useful in explaining growth, then they may be useful target indicators in formalizing policy reform packages. Empirical implementation should update earlier datasets. For example, the compilation of census data on educational attainment by Psacharopoulos and Arriagada (1986) could be updated with 1990 census observations in some countries. From the WDR 1991 database, one could use the following variables: 1. Estimates of factor inputs (capital stock, labor force, arable land, and mean education of the worling population for 68 countries). Some of the specifications that will be tested include factor accumulation on the right hand side so as to test for the effect of policies on efficiency. The factor input data will be useful in this regard. 2. Welfare indicators (infant mortality, female/male ratios, mean index of civil and political liberties). These and other indicators will be useful to examine the correlation between growth and more general welfare improvement. 3. Data on project economic rates of return used in a recent paper by Kaufnann (1991). 4. Government expenditures for health and education (original source is IMF and UNESCO); private spending on health (available for 19 countries, original source is United Nations). The government spending data will be important for the analysis of fiscal policies; the private health spending will be useful in measuring factor inputs. 5. Trade policy indicators (average tariff rates, nontariff barriers (UNCIAD), index of trade iiberatlization - 1960-84 (Papageorgiou, Choksi, Michaely), index of trade liberalization - 1978-88 (Halevi, Thomas), price distortion index (modified version of Dollar (1991). 6. Direct foreign investment flows (55 countries, 1970-89). 7. Political indicators (irregular e%ecutive transfers). 8. Measures of inequality (35 countries). 36 .ummar t d h Hbo lo bw Teoted ..ml c it 6t*ocpstb ededtbe5t .pmuylidaesXoThnie erlbotWb.^p-fTW In bold 'ip giv e.eoo _no.dohinpbiciinethes wme.hyp o.wUlI be tesed lesformaltl In te qut . tdles. NoW^tomefthe enonwtrlo Implcos reAtotho rduoed form etion ben growthand pol . whl othe ratr Odeiusu .qUt@wt@ttor. th deedr vadble.e ohw Ineteted pOlc: vbZb h .; . ............................ .. ,~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.... . 1. F O:af : :h A . UnLicm kwLme.oe as andimiounr . s.e roenum are nOt : t onp ti_ humvend ad aa ween#Iow g4wth bV reducngh tIn eweti" . l T :- AbI, ri. m _w b: dpeiI g a wa gi *he lud 1'tg _* __eien qie __ hr ar beteg dc.: B. . Tm. (Ineudin he lnX o tax thdIesiel atetposeeosaogot ie ra*.6 hwoment by diaoing reoure allocaton T ul hIn tw msu mwO . mitllg.*m* negiv , -dh ur pelI*nom. rbsenu ra C. Conpon taes he no growth eoffct. Memaum of tw oonsunWdon U e wU nxt be. _ l11 - in Oggt - e n D. Shfing govewme rnd to pr _ pubio good c grow f g hweAsmt rfse axd for a given tax fe end txsructur. Thh d mn, hf pisa osph bwesne hm _gaeu sp e ig9 awg bealpdus.y pm remd.Sdt hudshthv kaehutrs F_- Taepwntton productilvnvtmetshavpositveor regtlv growtheffectsdonding ontire disortion Iduced byttaxes dwheth prviin ot publI goode I too bw or too hlgh. Mike , beoa pose vly-oloped: o r af Ire wowiegmir e:b _Id h. i w.itinr rais h:to I rr hl~_1 '* d B 4o_ ninn.e hrt. .. .S .. ... . F. Unrtait abouA fcal. nd monetay policy reduce.sgrowth by distorting the allon of InveatmnL Te bill e hiui n_w d hioi wl :be u-ds..a. Trade Pollels and Polkies on International Cala.l Mlows A. Dktodd;iofdo top by twrNuft w quote nnd ewcha ired to s ounoertalntyao .*: tiepoliis elw.ow Ah for a ghivn ut. on veim Wdlsbrn* resource alloctin prsdumd i_k.dus ove nnvi -be l eiy negw inurli c _an _ .: kweebnsethr a UvulJOa ftgrseslen.W B;.. R-ic or diteu nt taxes on oreibt n investment (or uneoanty -- outhes. poi birslo gcwIh~ AdSuIVvaiM or1 tuevlblu nasnubigope,ussIm atiw, op :: b slgnlls.* posile -n W* W r _o- 111. Flhrnoldl 8.er Policiz. -............................ Tme loett domesti 11nannohinwkst acvle (such as hinbrestrat.oonrd, direce ur .:: highreewwrequlirmsnn, and owrr tes strd ri ll.fo on elngsnd rmouroumeobIksl.l Ths.Aowsrsgrwd. both: by reduolkg..kweent::.nd distortIn:g.tt* Wallocaion of nesdment. ssvq eNu,d.adwaivulablIuowmUswUtAldg icvdhmetI Ituds A-*-oti II".qMinapo 37 Tible 7:: Dta for analysis of policy and growth MEASURES OPt OMT1TT EMPLOYMENT, AND POPULATION GROWrH : Gioss output growth (GNP and GDP) Rodbustess checks: 4ltrmative bas- years: excluding extraction of exhaustible resources Population growth: Labor force grwth Employment growth (man-hours where available,:which:will be seldom) MEASURES OF PRIVATE FACTOR ACCUMULATION Private investment in physical capital Private education spending Measures of enrollment (primary, secondary, higher-level) Measures of human capital stock from previous studies: Measures of healtb (such as infant mortality and.life. expectancy) MEASURES OF:POUCY Fiscal de¢.ts:.. consolidated nonfinancial public sector deficits . .structurl deficit: variance of overalln ad structural .deficit... Public revenue policy (levels and variances): Cbnsumption taxes-bases and rates:.: Domestic sales taxes.on production,inputsaninvestment goc s ncomeeaxes ..pua::m::: -corporate . . -household Import taxes -.-:::.: .. ,,, . . - . .. . ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.......... Reen.ue offcommodit marketig boards::::.. Public spending:.::: Physical capital formation... -infrsitructuref(utilities, coSmi.ns, transport) . -other. Education- -.e..her salaries::.. -primary/secondary/higher-level -other . 38 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~. . .:.,:.... . ... ........ Table 7 (continuation) . . a th. : .. . ............: -basic health. .-other. . ...... Goods-subsidies . . ~ ~ ~~~~~~~~~. .. :.'.:.'..:. Monehry polisv ~~~~~~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~. . . . ............. Mean rate f.mony grwt, inflation .. ..:. Variance.of monetary growth,.inflation: Trade intervention Percent of imports subject to licensing -consumption -intermediate inputs -capital goods Average tariff rate and measures of dispersion -consumption:. -inteimediate inputs, . .. ..... .-investment::.. Ratio domestic:producer:price/world price for maJor commodities;... Black market eiohange rate.premium -(mean and-variance - . .:. :.:: Trade policy indicators (see sou.rces below).. . ..... Financial sector policies- Differential domestic deposit. rate and international.interes2 rate.: Real domesticdedposit interest rate. Spread.betwee deposit. and .lending rate..-. .... . . .......... Ex-post reserve equirement: (bank reserves/deposits). Interest-rate on go nment bonds Subsidized credits (amounts:and interest rates) Share of totallfinancial assets held by central bank Openness to -forein caoital Direct foreign investment flow-s Other net capitalflo.vs.- Restrictions on diect foreign investment, forein bdrrowing.:and capit e.p.ort .(and ... variance of restrictions over time). ~~~~~~~~~~~~~~~. . . . .. : .. '.:. : . '::- . : , ........... . . 39 TABLE&AA pi 1d::t: c 1auesm . GM" ouPA Nrawth GDP) "., FIncF.),:IFWodm ,, ~~~~~~~da8ita,Smmr n:d:testorn(1988' "' Population roh IFS, Wod Bknk da, Su o (1988 Labor foroe grWWt- World Bank ata Employme ,growth hours: ILO.. Prvate tIeatmentrtk physica capital P a & Mladaras (W1969) :Honot prime so" (1"; - :9 Hbb~Wb,and Cru (tcar: ':'::g :: ,s,*",#w.. ~ ~~~~~~~~~~~~~ ...-. :..:. PrIat fnacia savin 1K0nolhan&Atys(99 Private educio spenn Lau Jamis & 'Low 1199(:': Measurs d .LoIms UNESCO. Wr wand WON 9W, World D::::u. Report