Policy Research WORKING ,PAPERS Macroecon^mic Adjustment and Growth Country Economics Department The World Bank November 1991 WPS 799 Growth in Open Economies Sergio Rebelo A simple modification of recent growth models eliminiates the implausible implicationi that growth rates should be equalized in the presence of free international capital mobility and is consis- tent with evidence that points to low rates of savings in low income countries. Policy Research Working P4pers dissemnwiiale t,hi cii,iS *g v. k ii prorgc' ndll %Src 1 h,rgCu di's a nor1 ar'k saf fand all others interested ui developmein issue-. 'I hep parcro, d;% rlhnt2ed h!h RcKRscairmh Ads M,,rs S,ffl ern tfc mLmnmes 1mf thc authmmrs, renle;c only iheirviews, and should he used and lmmed dC,,rdglh 'I'h feiding., Fnierpretammoes, and cncius,nns Jrc the ammmJoi.m-m Ihe% shnuld notbeattnbutcdioihc World 3ank.itsl hord mfl)mrmestir% itsmalagemnettn)ran) of ise mlwhm:r o.intLncs Policy Research Macroeconomic Adjustment and Growth WPS 799 This paper -a product of the Macroeconiomic Adjustment and Growth Division, Country Economics Department - is part of' a larger effort in the Bank to understanid why growth rates diff'cr. This research was funded by the World Bank's Research Support Budget under a preparattion grant for the since-approved research project "How National Policies A'ffect Long-Run Growth?" (RPO 676-66). Copies are available f'ree from the World Bank, 1818 H Street NW, Washington DC 20433. Please contact Rebecca Martin, room N11-043, exicitsioin 39065 (54 pages). November 199 1. Rebelo surveys recent growth models lhait try to GNP would quickly be equalized all over the explain the diversity amilon-g countries in rales of' world. economic growth. Rebelo describes a simple modification of He finds that these models catn generate standard prefeireices that eliminates this implau- differences in growth rates onl inl thc absence ol' sibie cqualization of growth and is consistent international capital markets. Under these with the flct tihat the savings rate is lower in poor models, if there were 1'ree initenaitionlal capital countries tiha in ricih countries. mobility, the growth rate ol'conisumiiptioni and Policy Rcsearch Workling Papers N di .'niiiate nldings of s ork undcr \a! in the Bank. An objecti\ c of the scrics is to get thcsc Findings out quickl\. e% en if preser,itions are le-: than fully polished Tht findings, inicrpretations. and conclusions in these papers do not necessarily reprte.ent official B1ank policy Produced hv the Iihoicv Research Disseniination Ceniter TABLE OF CONTENTS I. Introduction 1 II. Why Do Growth Rates Differ? 2 11.1 Government Policy 3 The Linear Growth Model 4 The Effects of Taxation 8 The Effects of Sectoral Distortions 9 The Role of Government Expenditures 11 11.2 Trends in the Rate of Growth 12 II.3 International Trade 15 II.4 Poverty Traps 18 III. Empirical Evidence 20 The Shadow Real Rate of Return "1 IV. The Role of International Capital Markets 23 More Corplex Forms of Production and Accumulation 24 Nontradable Goods 25 The Role of Uncertainty 26 The Taxation of Foreign Investment 2P V. Savings and Growth 29 IV. Conclusion 35 References 37 Tables 44 Figures 47 Most of this paper was developed while I was visiting the World Bank in the uwmmer of 1990. 1 am thankful to the World Bank for providing me with a stimulating research environment and to David Renelt, Plyabha Kongsamut, and Maria Cristina Almero for research assistance. I benefitted from frequent conversations with Bill Easterly and from the comments of Robert Barro, Marianne Baxter, Jack Glen, Robert King, Ross Levine, Masao Ogaki, Charles Plosser, David Renelt, and Guido Tabellini. I am also grateful for the suggestions of Darticipants in the Carnegie-Rochester Conference, in the European Science goundation Conference on Economic Growth, and in seminars at the World Bank, Wharton School, New University of Lisbon, University of Lausanne, and at the Enrico Mattei Foundation in Milan. I. INTRODUCTION During the past fIve years growth theory has become one of tb --mot active areas of research ln economics. This paper evaluates the progress made by this recent literature in. explaining differences In rates of growth across countries ardL across historical epochs. This evaluation points to an. important shortcoAing of new growth theories: they can only explain cross-country differences In rates of growth by assuming the absence of lnternational capital markets. Heat recent models Imply that low rates of growth are the result of low real rates of return to investment by private agents. If this Is the case the development problem has a simple solution: allow stagnant countries to Invest in the developed world where the real rate of return Is high. This simple, costless policy would lead real per capita consumption and real per capita GN? to grow at the same rate In the developed and underdeveloped worlds. Growth models point to this implausible solution to the development problem because of their specification of lntertemporal preferences. The standard lntertemporal utility function implies that the optimal savings rate is identical for two countries which have the same real rate of interest but different income levels. This property leads to the prediction that an underdeveloped country which can Invest in the US will choose the same rate of savings, and hence will expand its GNP (but not Its GDP) at the same rate as the US. This paper studies a simple extension of standard preferences In which momentary utility has a Stone-Geary form: utility is derived from the amount of consumption that is above the level of subsistence consumption. With this modification it Is possible to generate different rates of growth even in the presence of perfect international capital markets. These preferences also imply that the rate of savings should be lower in poorer countries as seems to be the case in the data. The paper Is organized as follows. Section II summarizes different theories that attempt to explain the observed diversity In growth rates. Section III discusses empirical evidenne related to these theories. This section also examines Lhe relation between the implications of various models for the real rate of return and an indicator often used to evaluate the growth performance of LWC's: the Incremental Output Capital Ratio (ICOR). Section IV discusses the role of international capital markets. Section V studies the properties of the rate of growth when preferences have a Stone-Geary form -md presents evidence that accords with those properties. Section VI provides some conclusions. II. WHY DO GROWTH RATES DIFFER? This section summarizes some of the theories that have been advanced to explain cross-country differences In rates of growth. All the theories described can be seen as extensions of the neoclassical growth model of Solow (1956), Cass (1965) and Koopmans (1965). 1 We first review the channels by which government policy can affect rates of growth. We then describe three models that predict the presence of trends in rates of growth: the neoclassical growth model, and the models proposed by Romer (1986) and Jones and Manuelli (1990). The interactions between trade and growth are examined next. Finally, we discuss economies wtth poverty traps. IThis emphasis on models that extend the neoclassical framework to explain differences in rates of growth means that a large number of interesting papers associated with the new literature will not be reviewed in this section. Examples include Greenwood and Jovanovic (1990), Bencivenga and Smith (1988), Marcet and tarimon (1991:, and Levine (1991) on financial Intermediation and growth: Stokey 41988,1991) on the process of Introduction of new goods; Aghion and Howitt (1990) and Jovanovic and MacDonald (1990) on inncvation; Krugman (1991) on economic geography; and Schmitz (1989) on the interaction between Industrial organization issues and growth. 2 In describing these theories we always impose -ae symmetry hetween the economy under study and the rest of the world to rule out explanations of differences in growth rates that are based solely on the existence of cross-country differences In preferences or in technology. This type of symetry is conventional in modern macroeconomic theory but Is foreign to the develcpment economics tradition. Most of the models used In development economics emphasize the unique features of underdeveloped countries. II.1 GOVENMENT POLICY Before discussing how government policy can affect economic growth It Is useful to lay out a simple linear growth model with government. This model can be obtained by aggregating physical and human capital Into a composite good In a Lucas (1988)-Uzawa (1965) economy that follows a steady state path, has no externalities, and has Identical technologies In the output and human capital accumulation sectors.2 While the linear model Is very useful as an expositional device and as a guide to back-of-the-envelope calculations, it has some drawoacks. As Romer (1990a) has emphasized, this model (as all others of the Lucas-Uzawa variety) cannot explain why private firms invest in R&D. The linear model also tends to exaggerate the effects of certain public policies by ignoring the role of population size and assuming that the ratio of physical to human capital is fixed. But the qualitative effects of government policy described below can be built into virtually any endogenous growth model. Jones and Manuelli (1990) and Rebelo (1991) discuss the properties of the linear model and of related models. King and Rebelo (1990) compare the effects of taxation on growth and on welfare in the linear model and In a version of the Lucas-Uzawa mcdel in which there are no externalities and physical capital is used in the production of human capital. 3 &-Llnear Growth Modol In this economy population grows at the exogenous gross rate- r- and is composed by Identical agents. To simplify the. expositlon all variables are expressed in per capita terms. V'- will represent the gross growth rate of variable X by 7x- There is a single type of capital good (Zt) which Is a composite of physical and human capital. Production is a linear function of te.h stock of capital and can be used In consumption (Ct) or investment (It) t t Yt =AZt( Yt It +Ct' (2) The law of motion for the per capita stock of capital is the standard one: 7NZt1 =It t (1-8)Zto (3) where 8 represents the rate of depreciation. Since Zt Is * composite good that includes human capital It Is natural to assume that Investment Is irreversible, I t 0. To describe the operation of this economy it is convenient to think of a decentralization scheme in which there are two types of markets: spot factor markets and one-period credit markets. Households own the capital stock which they rent to firms for Rt. Profit (or value) maximization by firms and equilibrium In the factor market Implies that the rental price of capital must be equal to the marginal product of capital: Rt = A. Equilibrium In the credit market requires that the real rate of interest must be equal to rt = Rt -6 - A - a. The growth rate of this economy is determined by the savings 4 decision of private agents. Defining the savings rate as the fraction of Investment on net income. st1 (It-aZ/(Yt-6Zt), we can express the rate of growth of output as: =y (s r + I)/7N (4) Equation '4) shows that changes in r have a direct effect on the growth rate and an Indirect effect, associated with the response of the savings rate to changes in the real rate of return. To explore this indirect effect we need to make the savings rate endogenous. This will also clarify the effects of population growth on the rate of growth of per capita output. Thus we assume that each agent in the economy has perfect foresight and makes his savings decision so as to maximize his llfe-time utility: t-J = v (3 )t u(Ct (5 t=0 The momentary utility function u(Ct) Is assumed to be Isoelastic so as to be consistent with steady state growth:3 t u(Ct) = _ *> (6) These preferences Imply that households expand their consumption at a constant rate whenever the real Interest rate Is constant. The parameter iq represents the nature of the links between present and future generations. It is standard to assume that current generations care for the total utility of future generations, which corresponds to the case of ii=1 (e.g. Lucas (1988)). However, Barro and Becker (1989) choose O<1<1 in their analysis of endogenous fertility decisions. As will be clear 3The choice of preferences and production technologies consistent with steady state growth is usually justified by appealing to the Kaldor (1961) stylized facts of eronomic growth. See Romer (1989) for a discussion of these facts. S below, in equation (6), the value of 7n is crucial to determtne the Influence of exogenous population growth on the rate of expansion of the economy. In order to ensure that llfe-tlme-utility is filnite, so that (3) can be used to rank all feasible growth paths, we need to impose that: t$7w) ((A + (161W}< t. (7) N Finite utility conditions analogous to this one will be assumed to hold in all other models that we describe but will not be stated explicitly. The optimal rate of consumption growth from the standpoint of households is: V= ~STN (l+r)] (8) This implies that the optimal savings rate Is: s = 1t3 (l.r)1W /a/r (9) N It is well known that the response of savings to changes In the real interest rate depends on the relative strength of income and substitution effects. But the usual property that with o-1 these two effects cancel each other and ds/dr - 0, while ds/dr > O with oI, does not apply here. In order for these relations to hold we would have to define the savings rate as: se = UI + (1-8)Z / *Yt + Zt(1-6)1. The savings rate s will generally respond to changes In r even In the case of c-1. The growth rate of output can be obtained by replacing s in equation (8). Alternatively, since output grows at the same rate as consuzm;tion, Y can be obtained by replacing r In equation (8). One notable property of this economy Is that It has no transitional dynamics, It always grows at the following rate: 6 7y is{ 7Nhl [A + (1-6)1) (10) N This equation shows that an increase in the real rate of Interest r - A - 8 always leads to an incroase in the rate of growth. Expression (10) ignores the possibility of the non-negativity constraint on Investment being bincing. When this Is the case investment is zero and the gross rate of growth is (1-8)/. Equation (10) also shows that population growth has a negative impact on the rate of growth of per capita output whenever qO, C> (15) (Ct t where e(i-@) < I and <(i-o) < o to ensure strict concavity. Suppose that the government follows a balanced budget policy by choosing a combination of income tax rate r and government expenditures such that Gt = T AZt. In this case the rate of growth of this economy is, Ignoring the corner solution with zero Investment: J.= - { C' (A(1-T) + (1-8)1) (16) N If the government does not maintain a constant share of government expenditures 1A Income this economy will not follow a steady state path. In this case the effects of taxation are harder to characterize but the property that higher taxes imply lower growth is generally preserved. Evidence for developing countries suggests that certain 8The optimal path for government expenditure, abetractlng from the presence of distorting taxes, Is such that Ct/Gt G. 11 types of government expenditures may play an Important role in the development process. The existence of Infrastructures such as roads and an efficient legal system are often viewed as important preconditions for economic development. Barro (1990) has discussed these Issues in another extension of the linear model that views public expenditures as having a p.oductive role: the production function has the form Y t -- F(Zt#Gt). For simplicity we will consider the particular case of a Cobb-Douglas production function: t A Zt Gt (17) t t t Government expenditures are financed by income taxation: Gt T Yt, so the tax rate T coincides with the government share in output. It Is easy to see that the marginal product of capital for a given value of Tr is a A( T ' The rental price of capital will then be equal to Rt = (1-r) a A T(1/ while the real Interest rate Is rt = (1-T) a A (1/oa) 6. The rate of growth of this economy can be obtained by substituting rt in equation (8). Unless all governments choose the optimal value of T (which is (1-a)) rates of growth will be different but the response of growth to taxation Is now more complex: the growth effect of a marginal increase in T is positive if r < (l-a) and negative if if r > (1-X). II.2 TRENDS IN THE RATE OF GROWTH There are three growth paradigms that predict that If preferences, technology r4 public policy were identical across countries we should detect time trends in the rate of growth of per capita output: the neoclassical model. the Jones-Manuelll (1990) model and the Romer (1986) model. The neoclassical growth model predicts that poor countries should grow faster than rich countries as a result of 12 transitional dynamics. poor countries have a lower stock of capital and hence a higher real rate- of return which Induces a faster rate of growth. Using a battery of versions of the neoclassical growth model, King and Rebolo (1989) show that La order for these transitional dynamics to be Important the real rate, of interest has to take implausibly high. values In low income- countries. For instance, in order for transitional dynamics to explain the Japanese post-war growth. the real Interest rate In Japan in the end of World War It should have been close to 500%! These implausible predictions about the real rate of return can be traced to the presence of decreasing returns to scale to physical capital. When the production function Is Cobb-Douglas there Is a simple expression that relates the value of the real interest rates in two countries which are identical except for their level of output: r8 = (rA + 6) (Y /Y-)X/ 6 (18) In this expression a represents the labor share and Y A/Y0 the ratio of outputs in the two countries, A and B. If we assume that o - 2/3, 6 = 10 (which are standard values used In growth accounting studies, see Maddison (1987)), rA = 6.5% and that country B has half of the output of country A, we obtain a value for r3 = 56%. This is an extremely implausible value in light of historical evidence on rates of return (see Homer (1963), Siegel (1991). and Neal (1989)). Lucas (1990) discusses how differences in the levels of human capital and capital market imperfections can moderate these rate of return Implications. The effects of differences in human capital can be explained using the production function proposed by Mankiw, Romer and Weil (1990): Y = AKHt" " " where H is human capital and L raw labor. The relation between the real Interest rate of two countries, A and B is, In this case: rB = (Y /Y ) (r + 8) (H /H )M2 6 (19) 13 A A B A This expression shows clearly how differences In the levels of per capital human capital (HA and Ha) can offset the tendency for 13 the real rate of return to be low in poor countries. The Mankiw, Romer and Weil (1990) model preserves the Implications of the neoclassical model that, absent differences in technology, preferences and policy, poor countries should grow faster than rich countries. The same Is true about the Jones-Manuelli (1990) model which relies on a production function of the type: Y AZt + BZOT-x, where T represents non-reproducible factors such as land. The presence of T In the production function implies that the real rate of return Is high for low levels of Z. In contrast with these convergence Implications, the theory proposed by Romer C1986) predicts that growth should accelerate over time. This prediction is based on an Increasing returns to scale function such as the following: Yt = A Zt lxT ?t, 3l , M3 > 0 (20) t t al + M2 i The factors of production in this economy are Zt, which is a composite of physical capital, human capital and disembodied knowledge, and T, which represents non-reproducible factors of production. The variable Z represents the per capita capital stock in the economy. The positive effect of 7t on production Is meant to capture an external effect: the accumulation of capital Increases the stock of knowledge in the economy benefiting all production units.9 This externality implies that the competitive equilibrium is not Pareto optimal and hence there is scope for government intervention. In the borderline case In which al + oa - 1 this economy displays a behavior that Is identical to that of the linear model: there are .o transitional dynamics and the rate of growth is constant. If a+&3>1 It Is not feasible for this economy to 9See Benhabib and Jovanovic (1989) and Caballero and Lyons (1989) for attempts to Isolate empirically this type of externality. 14 grow at a constant rate and the competitive equilibrium displays In general growth rates that accelerate. 10 11.3 INTERNATIONAL TRADE The belief that international trade can be- one of the driving forces of economic growth is shared by many economists and policy makers. The model developed by Romer (1990a) has been used by Grossman and Helpman (1989a.1989b) and by Rivera-Batiz and Romer (1991) to explore the links between trade and growth. This model has the advantage of being consistent with the observation that profit maximizing firms Invest In R&D activities. To describe the essential features of Romer's model It Is useful to abstract from population growth and to view all variables as representing aggregate quantities. The production function for the output sector Is: Yt = 8 Lrt t x(l)1"di (21) The variable Lv denotes the number of quality-adjusted units of labor employed In the production of output. A continuum of capital goods Indexed by I e (O At I and represented by x(l) Is also employed in production.I& Each capital good x(l) costs u units of output to produce and does not depreciate over time. For this reason the aggregate stock of capital In this economy is Kt K VUAt x(i) di. To produce capital good i It Is necessary to 0 . Romer's (1986) model does not have the same Implausible Implications for the time series behavior of the real rate of return as the neoclassical model. Suppose for instance that o=+oa - 1.10. In the benchmark case of King and Robelo (1989) (which involves a real rate of return of 6.5% In the beginning of the century and a 6 fold Increase output from 1900 to 1980) tho model predicts that the real interest rate should have increased from 6.S% in 1900 to 9.7% In 1980. 1lThe assumption that there is a continuum of capital goods is convenient because it avoids dealing with integer constraints. 15 have; its design, that is. It is necessary that 1 e (O At]. New. designs are produced with labor according to the followiag technology: At+i = (L-Ly) At + At where L Is the total number of available efficiency units of time. The market for intermediate products is competitive so the rental price associated with x(i) is equal to the Its marginal product: R(i) = (1-a) B La x(i) (23) The inventor of a new type of capital has perpetual monopoly power in the market for that good as a result of a perfect patent system. Assuming that the economy is at the steady state, the quantity of each capital good produced is constant and the value of an additional patent is: PA =- v x(i) + (1+r)R(i)x(i)/r (24) where r is the real interest rate, vx(i) the cost of production of the capital good (since there is no depreciation production takes place only once along the steady state path) and (I+r)R(i)x(i)/r the present value of revenue from renting the capital good to firms in the output sector. Using equations (23) and (24) it is easy to see that the rental price that maximizes patent value is R(i) = v r/[(l+r)(l-a)j and that the optimal quantity of x(i) that should be produced is: x(i) - L ((-aM) B (l+r)/(rv)]1". The value of a patent is equal to: PA= 4XVX(i)/(1-a) (25) The symmetric role of the capital goods In production implies that In the steady state x(i) _ for all I S At. 16 Profit maximizing in the output sector and equilibrium in the labor market ensure- that: w - a. B LM-x- A (26) v Slace there Is free entry in the research sector the value of a patent has to be- in equilibriur-identical to its cost: PA8A/(l+r) = w (27) Using equations (25)-(27) it is possible to determine the quantity of labor employed in the output sector as being equal to: Ly = r/t(-a)61. This, in turn. determines the "Innovation rate": 7A 1 + 6L - r/(l-m) (28) Along the steady state path output Is given by Yt = B Ly x1 At, so Yt grows at rate 7A' Output is used in consumption and in the production of new capital goods: Yt = Ct + v x (At+ - At). Both consumption and production costs grow at rate VA. Equating ic n equation (24) to 7 A In (28) yields the equillbrium value of the real Interest rate and the equilibrium growth rate. This growth rate is suboptimal as a result of the presence of monopoly pricing In the intermediate goods market. Since equation (S) holds for this economy the property that higher real Interest rates lead to higher growth is preserved. Equation (28) shows that there is a strong scale fictor at work in this model, which may be troublesome in the presence of population growth. Integrating two economies identical to this one will increase the rate of growth, since L would double In the integrated area. Rivera-Batiz and Roner (1991) show that liberalizing the trade of goods and of Ideas (represented by At) has the same effects as economic Integration. The effects of tariffs and other trade barriers are theoretically ambiguous. In models similar to the one described Helpman and Grossman (1989a) show that tariffs increase the rate 17 of growth while Rivera-Batiz (1989) discusses a case in which they slow down economic growth. This variety of theoretical possibilities is not surprising given the presence of increasing returns and monopolistic competition. It is well known that second best results are more complex and less Intuitive than their first best counterparts. II.4 POVERTY TRAPS A striking feature of the growth phenonmeon in this century is that a large number of countries have failed to Improve "tir standard of living. Easterly (1991) shows that the ra '- f growth of per capita GDP Is not significantly different from zero in 46 out of 87 developing countries included in the Summers and Heston (1988) data set. This evidence suggests that some countries mav have been caught in a "poverty trap": a stable steady state which involves stagnation at low levels of Income. Models which display poverty traps have been proposed In the endogenous growth literature by Azariadis and Drazen (1990), Tamura (1989), and Becker, Murphy and Tamura (1990). 12 The mechanism at work in these models can be explained within a stylized example inspired by Azariadis and Drazen (1990). Consider a Lucas-Uzawa type model In which the equations that pertain to output production and to the accumulation of the capital stock are: Yt = AKtc (N H) O < cc <1 (Z9) t t tt(0 yt =~ * +It (30) 7NKt1 = It + (l- )Kt 0 < a <1 (31) The production function for human capital takes the following special form: 12The first poverty trap model was proposed by Nelson (1956). See Neves (1990) for a through discussion and critical evaluation of poverty trap models. 18 Ht+- Bit-HNt)Ht 1Ht If Ht> (32) Ht+ Ht if Ht H the economy is identical to tne one described In Lucas (1988) (abstracting from the production externality considered by Lucas). It will converge to a steady state where it will grow at rate Ty = [$AT (1+8)1 (1/0 The human capital technology described in (32) Is an extreme example since it Implies that economies with human capital below H cannot accumulate human capital. But it captures the essential idea explored by Azariadis and Drazen (1990) that economies In which human capital is low are less efficient at accumulating human capital. This generates a poverty trap In which the steady state real interest rate is low and so is the growth rate. Tamura (1989) and Becker, Murphy and Tamura (1990) explore a similar mechanism that works through the endogeneity of the fertillity rate: since it is relatively more costly for poor families to educate their children they tend to have a larger number of children with less education than those raised by rich families.13 This mechanism means that countries with low levels 13Erlich and Lui (1989a,1989b) explore the role of lmplicit familial contracts in economies that have this type of poverty trap. 19 of hu2an capital can converge to a low-level steady state in which there Is high population growth, no human capital accumulation and no growth in per capita output. IIr. EMPIRICAL EVIDENCE The data that ls currently available to test the theories described in section II are extremely scarce. There are some estimates of GDP. population, exports, Imports and government expenditures for developed countries and for some developing countries (mostly in Asia and Latin America) that go back to the 19th century (see Maddison (1982,1983)). But, for most countries, the data that is available covers the period from 1950 to the present. For this reason most growth studies have focused on cross-country regressions. These cross-country regressions are hard to interpret because most of the variables considered can be viewed as endogenous. These regressions results are also affected by measurement error problems and lack of robustness. The robustness problem has been examined In detail by Levine and Renelt (1990). These two authors show that virtually all policy variables (government share in GDP. proxies for property rights protection, measures of trade intervention, etc.) are related to growth but that the signs and statistical significance of these relations depend on the other variables included In the right hand side of the regression. This is partly a result of the fact that policy variables are highly correlated: countries that pursue highly distorting trade policies, also tend to employ highly distorting tax systems, fail to guarantee the protection of property rights, etc. In an exhaustive study that uses one of Barro's (1991) basic regression equations as point of departure, Levine and Renelt (1990) Identify two robust statistical relations: average rates of growth are positively related to the share of investment In GDP and the latter variable is positively related to the share of trade In GDP. Their basic regression equation, which employs 20 observations for 101 countries for the period 1960-1989 is (standard errors are reported in parenthesis): GYPm -0.83- O.3SxRGDP60 - 0.38xGP0 + 3.17xSEC + 17.SxINV (33) (0.85) (0.14) (0.22) (1.29) (2.68) R a 0.46 where GYP is the rate of growth of per capita GDP in the period from 1960 to 1989, RGDP60 Is the level of GDP in 1960, GPO is the growth rate of population. SEC Is the enrollment in secondary education In 1960, and INV the investment share in GDP. This regression accords with Barro's (1991) finding that growth is negatively related to the Initial level of per capita Income.14 The investment share is clearly the most Important variable In the regression described above. A regression of the rate of growth on the investment share yields the following results: GYP = -0.005 + 0.137xINV (34) 2 (0.004) (0.022) R' = 30.6% Below we explore one way of assessing the extent to which we can explain growth as a function of the investment share that Is, perhaps, more revealing than lookirg at the R from this regression. This involves computing some model-implied real rates of return as a function of the Investment share and the growth rate of output. Th Shadow eal Rte of Return The linear model described in section II implies that there should be a simple relation between growth. the real rate of 4This result, which continues to hold If we exclude the variable INV from regression (33), might be taken as Indicating the presence of cross-country convergence. Barro and Sala-i-Martin (1989) and Quah (1990) discuss a different notion of convergence that Involves a decline over time in the cross-country variability of per capita income. 21 return and .he investment share in GDP. Defining the investment share as It = It/(AZt), the equation of evolution for the per capita stock can be written as: VtNt t+1 l tAZ t +(1 6)Z t (35) It is easy to see that the- real interest rate rt -A - 6 can be written as a function of the investment share and of the rata of growth of aggregate output, tY which, in this model always coincides with the growth rate of capital. t NttYt - (1-6)1/1t - 6. (36) This formula Is almost identical to the expression for one of the key indicators used in practice to evaluate the performance of programs designed to help developing countries: the Incremental Capital-Output Ratio (ICOR): ICOR = It/(Yt+1 - (t = t/3(7Ntt37) where YV represents aggregate output. Equation (36) holds for the Easterly (1990) model when the economy Is at the steady state (in Easterly's model when investment Is reversible the transitional dynamics last only for one period). It also holds for Barro's (1990) model with a slight modification (a represents the share of capital In production):15 rt= MNttrYt -1-5)]/i - (38) When 6=0, the ICOR is the Inverse of the rate of return described in (36). Thus the linear model provides a theoretical foundation for an Indicator that has a long tradition in Formulas similar to this one can also be obtained for the neoclassical growth model, and for the models proposed by Romer (1986), Lucas (1988) and King and Rebelo (1990). 22 development economics. 16 Figure 2 depicts the real rates of retuurn associated with the-linear model. These were computed assumirg that B - .10 and using growth rates for non-overlapping 10-year periodz and average Investment rates for the countries in the-World Bank data set. 17 The range of values obtalned Is too wide to be believable, demonstrating the incomplete nature of our understanding of the growth phenomenon on the basis of rates of accumulation. IV. THE ROLE OF INTERNATIONAL CAPITAL MAPEKTS All the models reviewed in section II explain differences in rates of growth as the result of differences in real rates of return. But in order for these differences in rates of return to survive international capital markets must have a primitive form. WLth perfect capital markets rates of return would tend to be equalized across countries leading to uniform growth all over the wol.18 world.18 The only exception to this equalization result Involves 16The Bruno and Chenery (1962) model can also be seen as rationalizing the use of the ICOR. In their model the production function Is Leontleff and there is a significant fraction of the active population tnat Is unemployed. For this reason, the production function can be seen as being linear up to the point where all active population becomes employed. After that point the economy can no longer grow and hence no longer resembles the linear model. The levels of these rates of return are generally very high when compared with the 6.5% average real rate of return associated with common stock in the US ln the 1928-1988 period (see Ibbotson and Sinquefield (1988)), or even to the 30% return on equipment Investment estimated by De Long and Summers (1991). This Is partly a result of the narrow concept of Investment used which excluded, among other things, Investment In education. This is the reason why Grossman and Helpman (1989a), who assume 23 the case of taxation under a worldwide syste- considered by Klng and Rebelo (1990) and Rebelo (1991) and further examined below. 19 To describe the impact that international capital markets can have let us return to the linear model of section II and consider a country in whi.ch the real rate of return Is r - A(1-c) - a as a result of transactions costs represented by c. In developed countries these transaction costs do not exist and the real rate of return Is r - A - 6. What would happen if the private sector were allowed to invest In the international . capital market at rate r ? In the extreme case in %hich there are no lrreversibilities or adjustment costs associated with the canital Installed, production in the LDC would be driven to zero and all capital would be moved abroad. GDP would be zero but GNP and consumption would grow at the same rate as in the developed world. This represents a solution to the development problem that In our view is markedly utopian. It suggests that the World Bank should be setting up mutual funds in developing countries that allow their population to Invest their savings In the developed world. This simple, costless policy would lead to an equalization of the growth rate of GNP and of consumption. Below we discuss some modifications of this scenario that might mitigate the growth equalization implication. More comolex forms of production and accumulation The first modification has to do with the reversibility of Investment. If investment were Irreversible all new Investment the presence of international capital markets, focus on explaining the world growth rate. instead of cross-country differences in rates of expansion. 19The effects of liberalization of international capital markets are particularly dramatic in models with Increasing returns, see Correia (1990). 24 would take place abroad and per capita GDP would tend to decline. In this case liberalizing capital markets leads to high growth rates in per capita GNP which quickly decline toward the growth rate of developed countries.20 Introducing sectors that use factors of production that cannot be moved abroad (c.g. land and labor) elimLnates thet. prediction that GDP would embark on a declining path after the opening of international capital markets. Consider for instance the poverty trap model of section II.4. If an economy that has Ht1. E(Mt) t R 7 , E(i+r ) I () exp [(va/2) (o-1)l (42) As one would expect, the variance of the rate of growth, which is given by: var(vt) = (E(1+rt) 1l 7 Q]t MM exp (rv2) El - exp(-v2)1 (43) t t ~~~~N always increases with a mean preserving spread. ne taxation of foreign investment Whether tax-driven differences in rates of growth can survive In the presence of international capital markets depends on how income from foreign investment is taxed. There are essentially two systems regarding the tax treatment of foreign investment: the worldwide system and the territorial system. Under the worldwide tax system an investor pays domestic taxes on Income from foreign Investment but receives credit for any taxes paid abroad on the same Income. Since this credit cannot exceed the amount of domestic taxes associated with foreign income, the relevant tax rate Is In this * 0 case maxUr, ), where T is the foreign tax rate. Under the territorial system foreign income is exempted from taxation.23 23This discussion ignores other relevant features of the tax system such as the role of transfer prices, the possibility of tax deferral, and the existence of a tax credit by country versus an overall tax credit. See Slemrod (1988) and Swenson (1989) for more detailed discussions of the tax treatment of foreign 28 The worldwide system is used by the US, the UK, Japan and Canada. while the territorial system is followed by France and by the Netherlands. It is easy to see that under the worldwide system taxation can still generate different rates of growth in the presence of lnternational capital markets. If v >T no foreign Investment will be undertaken. If tr sT there is still no reason to move Investment abroad since the relevant tax rate continues to be r. For this reason the effects of taxation are the same as In a closed economy. 24 Under the territorial system an increase in domestic taxes 0 rate above r has the dramatic Implications already discussed: all new investment Is undertaken abroad. In summary, with the exception of taxation under the worldwide system, the mechanisms described In section II do not survive as sources of growth differentials in the presence of international capital markets. These mechanisms can nevertheless be important to explain differences in rates of growth across historical periods. Goodfriend and McDermott (1990) Is an example of a model that can be viewed as explaining how the world as a whole moves from primitive forms of production to modern, specializatior-based, forms of organization. V. SAVINGS AND GROWTH In the absence of international capital markets differences in real rates of return that reflect aspects of the technology or of government policy translate into differences In rates of lnvestment. 240ne problem with the worldwide system is that it Is only optimal under very restrictive assumptions; see Feldstein and Hartman (1979). 29 growth. In the opposite polar case of perfect International capital markets the real rate of return Is the same all over the world and. as a consequence, technology and government policy cannot Influence the rate of growth of consumption and have generally a small impact on the rate of expansion of GNP. This section considers a simple extension of the standard preference specification that is consistent with differences in rates of growth with perfect international capital markets. These preferences have the time-separable form described in (5) but have a different momentary utility function: u(Ct) (44) The only difference between this Stone-Geary function and the one described (6) involves the consumption subsistence term C. With this utility function the elasticity of intertemporal substitution, which is (1-C/C t)(1/). is no longer constant: It Is equal to zero when C = C and converges to (1/o) when t - consumption grows at a sustained rate.2 Before we discuss the role of International capital markets it is useful to describe the competitive equilibrium for a closed economy when the technology is linear and the government levies a proportional tax on income as in section II.1. In this case, it is possible to solve the model in closed form. Given the level of per capita capital stock. Zt' the optimal level of consumption associated with these preferences is given by: t a [Zt -I+ C, (45) 25King and Rebelo (1989) discuss the transitional dynamics of a version of the neoclassical model that has these preferences. Christiano (1989) uses a neoclassical model with Stone-Geary preferences In which a trend is introduced in subsistence consumption to interpret the growth of Japan In the post-war period. 30 In this expression C s Q/(1+r-v) is the minimum level of capital that can support the subsistence level of consumption. The real interest rate Is given by rt a A(1-) - 6. The constant a is equal to a = (i+r) - E' y" (ler))]" . The optimal rate of savings Is given by the following hyperbola: St - (r - a)/r + (a3 - g)/(rZt) (46) The rate of savings converges to (TN-l)/r when the level of capital approaches g. When 1cr > 71 /1, the country pursues N unbounded growth. When l+r < 1 / I It converges toward the N subsistence level Z. To have a better feel for the implications of this model it is useful to look at two examples. The first. depicted in Figure 3, pertains to country A where the private rate of return is low, for one of the reasons suggested in section II, so that I+r < g /O. The level of income in this economy converges toward the subsistence level. Suppose that agents in this country are allowed to invest in the developed world and that currently the level of Income in the economy is twice the subsistence income level. As we have seen in section III, if C were zero the growth rate would suddenly Increase to the same level as that of the developed world. But since C > 0 and the economy is close to subsistence consumption. there is a very long period of slow growth. The reason for this is that locally around the subsistence steady state the elasticity of Intertemporal substitution is close to zero. As consumption increases so does the elasticity of lntertemporal substitution, eventually converging to 1/o, and leading households to choose steeper paths for their consumption profiles. Figure 4 shows that the pattern of evolution that would follow a liberalization of financial flows. This transition process can be extremely slow, It can take a long time for the economy to grow at a "healthy rate" and the brush with poverty 31 will leave permanent scars: there is no tendency for this country to catch up with the level of Income of the rest of the world. After the liberalization of capital flows all new investment is undertaken abroad but Initially these flows are very small and the effects on growth almost negligible. One way to see that the- short term effec1.s of financial liberalization are likely to be disappointing is to look at the relation between the savings rate and the rate of growth associated with equation (4): Ty (s r + 1 )/T. When the level of consumption is elose to C the response of savings to changes In r is very small and s is close to zero. As a consequence, eliminating internal distortions or allowing investment abroad to raise r will have almost no short term impact on the rate of growth. The idea that poor countries have lower savings rates is not new. It was one of the empirical regularities discovered by Kuznets (1966, pp. 426-7) ano it is a prominent feature of the famous Kaldor (1956) model. This idea is, however, often associated with the time-honored question concerning the relation between the Income distribution and growth. In the Kaldor (1956) model there are two groups of agents that have different marginal savings rates. Thus, redistributing Income raises the aggregate savings rate and the rate of growth. The Stone-Geary preferences described in (44) do not Imply any relation between Income distribution and the rate of growth.26 The reason for this is that the consumption decision rule (45) Is linear Implying that although different agents have different average savings rates, they all have the same marginal rate of savings. There are several sources of evidence consistent with the Idea that poor countries have lower savings rates. Glovannini (1985) has estimated the elasticity of intertemporal substitution in LDC's is extremely low and often insignificantly 26See Williamson (1989, Lecture 3) for an historical discussion that supports the idea that Inequality and growth are unrelated. 32 different from zero. This is exactly what one would expect from our Stone-Geary formulation. Recall that the elasticity of intertemporal substitution is equal to (1--C/Ct)(1/W). so countries with consumption levels close to C should be expected to have low values for the rate of intertemporal substitution.27 Atkeson and Ogaki (1991) show that two types of utility functions rationalize simultaneously the fact that rich people have higher savings rates and lower food expenditure shares than poor people: a two-good extension of (44) and a 2-good addilog utility function They discuss a large tody of evidence that supports both of these empirical regularities and describe simulations in which their models reproduce the S-shaped pattern of savings as a function of income suggested by Kuznets. They also prove aggregation results for these utility functions thu.t guarantee that the relevant elasticity parameters can be estimated using aggregate data. In a separate paper (Ogak.i and Atkeson (1991)) they report estimation results focusing only the addilog utility function. This narrower focus is justlfiei by the fact that their 2-good extension of (44) has the counterfactual implication that the elasticity of the demand for food Is highest for poor people. Figure 5 shows the relation between the share of gross domestic savings in GNP and the level of real per capita GNP. 270ne possible problem with Glovannini's estimates is that they rely on data for the real return to time deposits. It is un:lear whether these real rates of return, which are often negative, are a good measure of the real rate of return to capital In LDCq'. In developing countries the financial Intermediary system is often primitive and retained earning are used to finance Investment projects whose rates of return are much higher than the time deposit rate. The fact that the change in time-deposits as a fraction of savings is small can be seen as suggesting that direct financing of investment is, In fact, an Important phenomenon In LDC's. Gelb (1989) reports that the change in M3 (which includes currency plus deposits) on savings during the period 1965-73 was 18.7% for countries with a positive real Interest rate, 12.7% for countries that had small negative real Interest rates and 6.4X for countries that had large negative real interest rates. 33 Each point corresponds to the average savings rate and Income level for a given country during a 10-year period (only non-overlapping perioda were used; the Figure has the same features If we take averages over a different time horizon). The solid line in this Flgure corresponds an estimate of the hyperbola described in (46). Figure 6 presents the same evidence but concentrating on countries with incomes less than 15000 1987 US dollars. The dotted line in Figure 7 depicts an estimate of (46) while the solid line was constructed by ordering the countries by income, dividing them into 10 classes and computing the simple average of the savings rate for these various classes. This set of Figures shows a relation between the rate of savings and the level of Income that accords remarkably with the predictions of equation (46). Table 1 shows that the relation between the inverse of the level of Income and the rate of savings holds up when we move from a bivariate comparison to a multivariate analysis. The data used in these regression is that of the Barro-Wolf data set extended to include World Bank measures of the savings rate and of the share of exports In GDP. Standard errors, reported in parenthesis, are based on White's (1980) heteroskedasticity-consistent covariance matrix. The first regression describes the relation between the savings rate and the inverse of the level of income 1960 (INVGDP60), depicted in Figures S through 7. The second regression includes the variables in the Barro-Wolf data set that are not highly correlated with the initial income level. These variables are the share of government consumption expenditures in GDP (GOV), number of revolutions per year (REVOL), number of assassinations per million population per year (ASSASS) and the magnitude of the deviation of the Investment deflator from the sample average (PPI60DEV). In this regression the coefficient on INVGDP60 continues to be negative and significant. The same result holds when we Include, in the third regression, the share of exports In GDP (XGDP) . The motivation to Include this variable is Levine and Renelt's (1990) finding that this is the only variable that 34 is robustly correlated with lnvestment.28 Regressions (6) and (7) on Table 2 Include the variables of the Barro-Wolf data set that are highly correlated with the initial level of income- ('able 2): the enrollment ratios for primary and secondary education in 1960 (PR1IH60 and SEC60, respectively) and the dumy variables for- Africa and Latin America. Not surprisingly all the variables in the regression become Insignificant, with the exception of the export share, as a result of multicolinearity. In contrast with what one might expect on the basis of life-cycle theories of savings, population-related variables (the growth rate of population, measures of the age structure, mortality rates) are insignificant when included in regression (3). Table 3 reports similar results for a set of regressions in which the share of investment Is used as a proxy for the savings rate. These regressions are presented to show that the results that we reported are not likely to be an artifact of measurement error in the savings rate series. VI. CONCLUSION The new growth literature has Identified several mechanisms which can potentially explain differences In rates of growth across countries. Unfortunately most of these mechanisms fall to generate different rates of expansion in the presence of International capital markets. Existing models predict that a libe-alization of capital flows would be followed by "capital flight" from stagnated countries, where the rate of return Is low, to fast growing countries. As a result, the growth rates of consumption and of GNP would be equalized around the world. A significant part of the new growth literature has been 28Romer (1990b) reports results similar to those of this equation for the investment share. He found that per capita real Income In 1960 is the only variable that has explanatory power for the Investment share after one controls for the influence of the share of exports in GDP. 35 devoted to studying alternative specifications of the technology that makes sustained growth feasible without exogenous technical progress. The thought experiment of liberalizing capital flows suggests that properties of the technology alone cannot rat.onalize the diversity of growth experiences that we observe. When we combine those properties with the standard time-separable, lsoelastic preferences we obtain the implication that allowing stagnant countries to Invest their savings In growing economies would solve the development problem. This suggests that the standard preference specification Is inadequate to think about growth in open economies. This paper explored perhaps the simplest modification of standard preferences that has two properties: It is consistent with evidence that poor countries save less than rich countries and Implies that a liberalization of capital flows would have negligible short term effects on the rate of growth. Empirical and theoretical work that Improves on this preference specification can potentially enhance significantly our understanding of the development problem. 29 An important by-product of this research might be the solution to another classical problem in development economics: the relation between the dynamics of lncome distribution and the growth process. 29Allen (1989) and Zervos (1991) are examples of studies of the consequences of departures from standard preferences for the implications of growth models. 36 References Aghion, P. and Howitt, P. (1990) A Model of Growth Through Creative Destruction. Working paper No. 3223, National Bureau of Economic Research. Allen, E. (1989) U.S. versus Japan Savings Rates: Habit Formation? Mimeo, University of Chicago. Atkeson, A. and Ogaki, M. (1991) Does the Intertemporal Elasticity of Substitution Vary with Wealth and Does That Matter for Aggregate Savings? Mimeo, University of Rochester. Azarladis, C. and Drazen, A. (1990) Threshold Externalities in Economic Development. Q0Uarerlv Journal of Economics, 105: 501-26. Barro, R. (1990) Government Spending In a Simple Model of Endogenous Growth. Journal of Political Economy, 98: S103-125. (1991) Economic Growth in a Cross Section of Countries. Ouarterly Journal of Economics, 100 407-444. Barro R. and Becker, G. (1989) Fertility Choice in a Model of Economic Growth. Econometrica, 57: 481-501. Barro, R. and Sala-i-Martin, X. (1989) Economic Growth and Convergence Across the United States. Mimeo, Harvard University. Barro. P.. and Wolf, S. (1989) Data Appendix for Economic Growth in A Cross Section of Countries. Mimeo, Harvard University. Becker, G.S., Murphy, K.M. and Tamura, R. (1990) Human Capital, Fertility, and Economic Growth. Journal of Political Economy, 98: S12:S37. Bencivenga V. and Smith, B. (1988) Financial Intermediation and Endogenous Growth. Working paper No. 159, Rochester Center for Economic Research. Benhabib, J. and Jovanovic, B. (1989) Externalities and Growth Accounting. Working paper No. 3190, National Bureau of Economic Research. 37 Bruno, H. and Chenery, H. (1962) Developing Alternatives in an Open Economy: The Case of Israel. Ecgonoi Journa, 72: 79-103. Butter, W.and Ketzier, K. (1991) Persistent Differences in National Productivity Growth Rates with Common Technology and Free- Capital Mobility: the Role of Private Thrift, Public Debt, Capital Taxation, and Policy Towards Human Capital Formation. Working paper No. 3637, National Bureau of Economic Research. Caballero, R. and Lyons, R (1989) The Role of External Economies In U.S. Manufacturing. Working paper No. 3033, National Bureau of Economic Research. Cass, D. (1965) Optimum Growth In an Aggregative Model of Capital Accumulation. Review of Economic Studies, 32: 233-240. ,Chamley, C. (1986) Optimal Taxation of Capital Income In General Equilibrium with Infinite Lives. Econometrica, 54: 607-22. Christiano, L. (1989) Understanding Jaoan's Savings Rate: the Reconstruction Hypothesis. Federal Reserve of MInnea2olis Quarterly Review, 2: 10-2S. Cohen. D. and Michel. P. (1991) Property Rights on Foreign Capital and Long Run Growth. Working paper No. 9106, CEPREMAP. Correia, r. (1991) External Debt and Economic Growth. Mimeo, Bank of Portugal. De Long, J. and Summers, L. (1991) Equipment Investment and Economic Growth. Quarterly Journal of Economics, 106: 445-502. Easterly, W. (1990) Endogenous Growth in Developing Countries with Government Induced Distortions. Mimeo, World Bank. (1991) Economic Stagnation, Fixed Factors, and Policy Thresholds. Mimeo, World Bank. Easterly, W., King, R.. Levine, R. and Rebelo, S. (1990) Do National Policies Affect Long Run Growth?: A research proposal. Mimeo, World Bank. Ehrlich, I. and Lul, F. 38 (1989a) Longevity, Human Capital Accumulation and Economic Growth. Discussion Paper No. 8907, SUNY at Buffalo. (1989b) Social Insurance, Self Insurance and Economic Growth. Mimeo, SUNY at Buffalo. Feldstein, M. and Hartman, D. (1979) The Optimal Taxation of Foreign Source Investment Income. Quarterl2 Journal oL Ec2n2oics 93: 613-629. Gelb, A. (1989) Financial Policies, Efficiency and Growth: An Analysis of Broad Cross-Section Relationships. Mimeo, World Bank. Glovannini, A. (1985) Saving and the Real Interest Rate in LDC's. Journal of Develo2ment gg2nomigs, 18: 197-217. Goodfriend, M. and McDermott, J. (1990) Early Development. Mimeo, University of Chicago, Greenwood, J. and Jovanovic, i. (1990) Financial Development, Growth, and the Distribution of Income. Journal ofL 22ltlcUal gcomogMv 98: 1076-1107. Grossman, G. and Helpman, H. (1989a) Comparative Advantage and Long-Run Growth. Working paper No. 2809, National Bureau of Economic Research. (1989b) Growth and Welfare In a Small Open Economy. Working Paper No. 2970, National Bureau of Economic Research. Homer, S. (1963) Hs2 Hso anterest Rat2s. New Brunswick: Rutgers University Press. Ibbotson, R. and Sinquefleld, R. (1988) $tggks. Bgadi, a&fla ind Inflak2n_. 1288 Yearbopk. Chicago: Ibbotson and Associates. International Finance Corporation (1990) ua2rsj.ggn StUKarkets Factbook. Washington D.C. Jones, L. and Manuelli, R. (1990) A Convex Model of Equilibrium Growth. Journal 2L PolE.a t gLQ . 98: 1008'38. Jones, L., Manuelli, R. and Rossi, P. 39 (1990) Optimal Taxation in Convex Models of Equilibrium Growth. Mimeo. Northwestern University. Jovanovic, S. and MacDonald, G. (1990) Competitive Difusion. Working paper No.253, Rochester Center for Economic Research. Kaidor, N. (1956) Alternative Theories of Distribution. Review of Economic Studies 23: 83-100. (1961) Capital Accumulation and Economic Growth. l Theory _f gatal. eds. F. Lutz and D. Hague. New York: St. Martin's Press. King, R.G. and Rebelo, S. (1989) Transitional Dynamics and Economic Growth in the Neoclassical Model. Working paper No. 3185, National Bureau of Economic Research. (1990) Public Policy and Economic Growth: Developing Neoclassical Implications. Journal of Political Economy, 98: 5126-SiSO. Koopmans, T. (1965) On the Concept of Optimal Economic Growth. The Econometric AyDroach to Development Eianning. Pontific Acad. Sci., Amsterdam: North-Holland. Krugman, P. (1991) Increasing Returns and Economic Geography. Journal of Political Economy, 99: 483-499. Kuznets, S. (1966) Modern Economic Growth: Rate, Structure amg Soread. New Haven: Yale University Press. Levhari D. and Srinivasan, T. (1969) Optimal Savings Under Uncertainty. Review oL ,Economic Studies, 36: 153-164. Levine, R. (1991) Stock Markets, Growth and Policy. Journal of Flnance, forthcoming. Levine, R. and Renelt. D. (1990) A Sensitivity Analysis of Cross-Country Growth Regressions. Mimeo. World Bank. Lucas, R. (1988) On the Mechanics of Economic Development. Journal 2L Monetary Economics, 22: 3-42. 40 (1990) Why Doesn't Capital Flow from* Rlch to Poor Countries?. Amersac £Economic R2vle 80: 92-96. (1991) Supply Side Economics: an Analytical Review. Oxford Economlc Paers, forthcoming. Maddison, A. (1982) Phaseg 2L Ca*ltalLLtic D#_veloRVt. Oxford: Oxford University Press. (1983) A Comparisc,n of Levels of GDP Per Capita in Developed and Developing Countries, 1700-1980. J2rniM gf Economic HlistrL 43: 27-41. (1987) Growth and Slowdown In Advanced Capitalistic Economies: Techniques of Quantitative Assessment. Journal of Economic Literature, 25: 649-698. Mankiw, G.. Romer, D., and Weil, D. (1990) A Contribution to the Empirics of Economic Growth. Mimeo, Harvard University. Marcet, A. and Marimon, R. (1991) Communication. Commitment and Growth. Mimeo, Universitat Pompeu Fabra. Mino, K. (1990) Money and Endogenous Growth In a Cash-in-Advance Economy. Mimeo, Hiroshima University. Neal. C. (1989) Macrofinancial Indicators for 117 Developing and Industrial Nations. Mimeo, World Bank. Nelson, R. (1956) A Theory of Low Level Equilibrium Trap In Underdeveloped Economies. American EgnomL2. Review, 46: 894-908. Neves, J. (1990) P& Validadg Clenliflca do Cgnceito 4& EamilibrLo de Pobreza. Lisboa: Centro de Estudos Fiscais. Direccao Geral de Contribuiq8es e Impostos. Minist6rio das Finanqas. North, D. (1987) Institutions, Transactions Costs and Economic Growth. Economic Eniry 25: 419-428. Ogaki. M. and Atkeson, A. (1991) Estimating a Model of Engle's Law, Saving, and Risk Searing with Indian Panel and U.S. Time Serles Data. Mimeo, University of Rochester. 41 Quah, D. (1990) International Patterns of Growth: I.Persistence in Cross-country Disparities. Mimeo, Massachusetts Institute of Technology. Rebelo, S. (1991) Long-Run Policy Analysis and Long-Run Growth. Janal of Political Economy, 99: 500-521. Renelt, D. (1991) Economic Growth: A Review of the Theoretical and Empirical Literature. Mimeo, World Bank. Rivera-Batiz, L. (1389) Growth, Technology and Trade between Asymmetric Countries. Mimeo, University of Chicago. Rivera-Batiz, L. and Romer, P. (1991) International Traos with Endogenous Technical Change. E=22240 FcongMi2 Review. forthcoming. Romer, P. (1986) Increasing Returns and Long Run Growth. Journal of Polltical Economy, 94: 1002-1037. (1989) Capital Accumulation In the Theory of Long Run Growth. Modern Buslness Cv_qe Theory, ed. R. 8arro. Cambridge, MA: Harvard 'Iniversity Press. (1990a) Endogenous Technological Change. LJrnal _o Political Economy, 98: S71-S102. (1990b) Capital, Labor, and Productivity. BrookirjS FARREr on Aconom£ctivity:L MIcrE2coomcs1, 337-420. Washington: Brookings Institution. Schmitz, J. (1989) Imitation, Entrepreneurship, and Long-Run Growth. 12MMIJ of Polltltli L;gaomy, 97: 3- Slegel, J. (1991) The Real Rate of Interest from 1880-1990: A Study of the U.S. and the U.K.. Mimeo, Wharton School. Slemrod, J. (1988) Effects of Taxation with International Capital Mobility. Uneasv Com£r2mlsu : EStr2bgm gL n kIbrld I5ncMe-ConsumtitL2n L. eds H. Aaron, H. Galper, and J. Pechman. Washington D.C.: Brookings Institution. Solow, R. (1956) A Contribution to the Theory of Economic Growth. guortrrly no_ualn __ Eo nomS cs., 70: 65-94. 42 Soto, H. (1989) M.Othr Pa1th. New York: Harper and Row. Stokey, N. (1988) Learning by Doing and the Introduction of Now Goods. Journal, of Peoltical Econo , 96: 701-717. (1991) Human Capital. Product Quality and Growth. Ouar.ter.ly IQE g2L EgEono2ic, 106: 587-616. Summers, R. and Heston, A. (1988) A New Set of International Comparisons of Real Product and Price Level.. Estimates for 130 Countries, 1950-1985. Rgview 2L Incom2 Il Wealth, 34: 1-25. Swenson, D. (1989) The Impact of U.S. Tax Reform on Foreign Investment In the U.S.. Mimeo, Massachusetts Institute of Technology. Tabellini, G. and Persson, T. (1990) Is Inequality Harmful for Growth?: Theory and Evidence. Mimeo, Institute for International Studies, Stockholm, Sweden. Tamura, R. (1989) Fertility, Human Capital and the "Wealth of Nations". Mimeo, University of Iowa. Uzawa, H. (1965) Optimal Technical Change in an Aggregative Model of Economic Growth. l.atrnatiolal Em2micf2 Review, 6: 18-31. White, H. (1980) A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity. Efgnometria, 48: 817-838. Williamson. J. (1989) Inequality and Modern Economic Growth: What does History Tell Us?. The Kuznets Memorial Lectures, Yale University. Yuen, C. (1990) Taxation of Human Capital. Mimeo, University of Chicago. Zervos, D. (1991) Balanced Growth with Non-time Additively Separable Preferences. Mimeo, University of Rochester. 43 Table 1 Savings Rate Regressions (1) (2) (3) (4) (5) CONSTANT 0.240 0.310 0.278 0.128 0.132 (0.017) (0.OS0) (0.061) (0.072) (0.078) INVGDP60 -0.050 -0.419 -0.039 -0.019 -0.018 (0.009) (0.010) (0.010) (0.015) (0.020) GOV -0.377 -0.397 -0.230 -0.224 (0.311) (0.300) (0.302) (0.296) ASSASS -0.011 -0.006 -0.008 -0.009 (0.022) (0.024) (0.018) (0.019) REVOL -0.075 -0.046 -0.011 -0.012 (0.097) (0.096) (0.093) (0.093) PPI60DEV 0.006 0.007 0.004 0.006 (0.035) (0.034) (0.030) (0.033; XGDP 0.098 0.072 0.074 (0.068) (0.073) (0.079) SEC60 -0.042 -0.046 (0.092) (0.130) PRIM60 0.141 0.138 (0.053) (0.060) AFRICA -0.010 (0.037) LAAME -0. 001 (0.060) R2 0.14 0.17 0.19 0.24 0.24 44 Table 2 Variable Correlations with INVGDP6O GOV 0.35 ASSASS -0.17 REVOL 0.18 PPI60DEV 0.22 XGDP -0.16 SEC60 -0.65 PRIM60 -0.69 AFRICA 0.67 LAAME -0.25 4!5 Table 3 Investment Rate Regressions (1) (2) (3) (4) (5) CONSTANT 0.234 0.229 0.206 0.070 0.074 (0.010) (0.019) (0.024) (O..031) (0.036) INVGDP60 -0.041 -0.034 -0.034 -0.004 -0.009 (0.006) (0.005) (0.005) (0.007) (0.007) GOV 0.107 0.073 0.137 0.122 (0.099) (0.102) (0.093) (0.092) ASSASS -0.007 -0.005 -0.007 -O. 003 (0.014) (0.01S) (0.010) (0.010) REVOL -0. 125 -0.090 -0.027 -0.024 (0.021) (0.029) (0.027) (0.027) PPI60DEV -0.049 -0.048 -O.059 -0.062 (0.018) (0.088) (0.015) (0.015) XGDP 0.088 0.092 0.085 (0.023) (0.020) (0.018) SEC60 0.125 0.115 (0.044) (0.047) PRIM60 0.068 0.077 (0.025) (0.025) AFRICA 0.012 (0.016) LAAME -0.016 (0.013) R2 0.29 0.47 0.49 0.63 0.64 .46 Lti Grow Raw p p p 0 -C I I p 0 I I / / I I p _ I / / / I I 1 p I - I - n / C / / C = / C, ,1 n / -q I - / / II / / // / / / / / / / / / p - - / / p Shadow Rel Rate of Re= (LA ILAA c .. c C a ~0 00 0 0 0 0 a0 00 0 @0 0 0 0 aaa 09 000 0 a 00oo00 0 0 a00 0 b 0o0o o 00 0 0 0 0 o0o 0 0 0 ~ o 0 0 0 00co0 o o o a 0 0 00 crb00 0 0 0 00 0 4 cto 0 0 0 a00 o 000 000 a0 0 0m 0o- 0a 0 o 0% 0 01)0 o oo0 o o 0o 00 o 00 0 t 00 tio o°o o0 o 00 0 O° Oo 0v°c o 0 o CfD C* 0,*O00~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ oe o ooo a 0o o8008) b.b ~ ~ I FlCIIRE 3 4200D , Y(L) | 1401 , C(t) 4000 \ 120 3800 ~ ~ ~ ~ ~ ~~~~2 3800 \ 100 _C _ ___ ___10 3600 - 80 3400 . A I 0 50 100 150 0 50 100 150 time (years) time (years) s10-3 GroWth Rate Y(l) -x1O-3 Growtlh Rate C(t) 0 0 -2- -2- * ~~~~~~~~~-4 -4-- 0 5U) 100 150 0 50) 100 150 time (years) time (years) FHIIIRE, 4 X104 Y(t) 1000 - C( t) 3-7 2 .500/ 0 0 _ . a 0 50 100 15t 0 50 10( 150 time (years) Limiie (years) 0.02 Growth Rate Y(t) ___ o G tlrowli 1ate C(t) 0.015 - 0.015 0.01 - 0.01 0.005 . ---------- 0 50 100 150 0 5( 100 150 time (years) tiimke (years) FIGU(RE 5 J.2 _ 0.8 0.6 0.21 0 Q -0.2 . -0.4 4 5 6 7 8 9 10 II LAg of 1960 Rea;l 1er Capil; (GDP FIGgIIE 6 0.6 Q.5 Q.4 0.3 0.2 0.i -01 -0.1 -0.3 - 5 6 7 89 Pog 19601 Real lPtr Capita GD11 Saig Rat 5p - P p 'C~~U la N _'C. £~~~~~~~ I ( Policy Research Working Paper Series Contact vw AuthorDate for Dauer WPS771 Macroeconomic Structure and Ibrahim A. Elbadawi September 1991 S. Jonnakuty Policy in Zimbabwe: Analysis and Klaus Schmidt-Hebbel 39074 Empirical Model (1965-88) WPS772 Macroeconomic Adjustment to Oil Ibrahim A. Elbadawi Septernber 1991 S. Jonnakuty Shocks and Fiscal Reform: Klaus Schmidt-Hebbel 39074 Simu,lations for Zimbabwe, 1988-95 WPS773 Are Ghana's Roads Paying Their Reuben Gronau September 1991 J. Francis Way? Assessing Road Use Cost end 35205 User Charges in Ghana WPS774 Agricultural Pricing Systems and Mark Gersovitz October 1991 B. Gregory Transportation Policy in Africa 33744 WPS775 The Macroeconomics of Public William Easterly October 1991 R. Martin Sector Deficits: A Synthesis Klaus Schmidt-Hebbel 39065 WPS776 Enforcement of Canadian "Unfair" Mark A. Dutz October 1991 N. Artis Trade Laws: The Case for Competition 37947 Policies as an Antidote for Protection WPS777 Do the Benefits of Fixed Exchange Shantayanan Devarajan Octoter 1991 A. 3halla Rates Outweigh Their Costs? The Dani Rodrik 37699 Franc Zone in Africa WPS778 A Dynamic Bargaining Model of Eduardo Fernandez-Arias October 1991 S. King-Watson Sovereign Debt 31047 WPS779 Special Programme of Research. Janet Nassim October 1991 0. Nadora Development and Research Training 31019 in Human Reproduction WPS780 Optimal User Charges and Cost tan G. Heggie October 1991 P. Cook Recovery for Roads in Developing Vincy Fon 33462 Countries WPS781 The Korean Consumer Electronics Taeho Bark October 1991 N. Artis Industry: Reaction to Antidumping 37947 Actions WPS782 The Economic Effects of Widespread Patrick Conway October 1991 N. Artis Application of Antidumping Duties Sumana Dhar 37947 to Import Pricing WPS783 The Origins and Evolution of J. Michael Finger October 1991 N. Artis Antidumping Regulation 37947 WPS784 Chemicals from Poland: A Tempest Andrzej Olechowski October 1991 N. Artis in a Teacup 37947 Policy Research Working Paper Series Contact ClUe Author Dat fgr par WPS785 How Did the Asian Countries Avoid ishrat Husain October 1991 S. King-Watson the Debt Crisis? 31047 WPS786 Fiscal Policy for Managing Sadiq Ahmed October 1991 B. Prasertwaree Indonesia's Environment 82477 WPS787 Private Investment Under Macroeco- Klaus Schmidt-Hebbel October 1991 S. Jonnakuty nomic Adjustment in Morocco Tobias Muller 39074 WPS788 How Expectations Affect Reform Francesco Daveri October 1991 S. Jonnakuty Dynamics in Developing Countries 39074 WPS789 Intrahousehold Inequality and the Lawrence Haddad October 1991 J. Sweeney Theory of Targeting Ravi Kanbur 31021 WPS790 Reforming and Privatizing Hungary's Esra Bennathan October 1991 B. Gregory Road Haulage Jeffrey Gutman 33744 Louis Thompson WPS791 Measuring Real Exchange Rate Lant Pritchett October 1991 k. Gabana Instability in Developing Countries: 37947 Empirical Evidence and Implications WPS792 Reducing Labor Redundancy in Jan Svejnar October 1991 B. Gregory State-Owned Enterprises Katherine Terrell 33744 WPS793 Decollectivization and the Karen M. Brooks October 1991 C. Spooner Agricultural Transition in Eastern and 30464 Central Europe WPS794 How Do National Policies Affect William Easterly October 1991 R. Martin Long-Run Growth? A Research Robert King 39065 Agenda Ross Levine Sergio Rebelo WPS795 Economic Stagnation, Fixed Factors, William Easterly October 1991 R. Martin and Policy Thresholds 39065 WPS796 Excess Liquidity and Monetary Gerard Caprio, Jr. October 1991 W. Pitayatonakarn Overhangs Patr:^k Honohan 37666 WPS797 Using Field Visits to Improve the Richard Heaver October 1991 0. Nadora Quality of Famly Pianning, Health, 31091 and Nutrilion Programs- A Supervisor's Manual WPS798 Agriculture's Decline in Indonesia Will Marlin October 1991 M. Sanchez Supply or Demand Determined7 Peter G. Warr 33731 WPS799 Growth in Open Economies Sergio Rebelo November 1991 R. Martin 39065