Policy, Research, and External Affairs WORKING PAPERS | Macroeconomic Adjustment and rlrowth Co intry Economics Department The World Bank October 1991 WPS 795 Economic Stagnation, Fixed Factors, and Policy Thresholds William Easterly Economic policies, not initial conditions, det,.rmine whether countries stagnate. The black market premium oni foreign exchange is an important factor in stagnation. The I'ohcl , Rescarch, anid Extrmal Affairs Complex dirLnbhuIes PRI: Working ldpcrs to disserinate the finding% of work in progross and to cncourage the exchange of ideas anmong l3ank staff and all othcrs iiterested in dcvelopment issues 'I lese papers canf) the names of the authors, reflect only their viCws, and shoiuld hb used and cited accordingly T'he findings, interpretations, and conclusions arc the authors' iwn T'hey should not he atinbuted to the World 13dnk, its B3oard of Direettors, its mandgemcnt, or an) of iws member counrncs. Policy, Research, and External Affairs Macroeconomic Adjustment and Growth WPS 795 This paper -a product of the Macroeconomic Adjustment and Growth Division, Country Economics Department -is part of a larger effort in PRE to assess the effect 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 Affect 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 N1 1-053, extension 39065 (39 pages). October1991. Many developing countries have experienced labor input and a broad concept of capital. economic stagnation. Africa had negative per Easterly extends the model to consider multiple capita growth in the 1970s and 1980s, and Latin capital goods and public capital. America in the 1980s. Per capita growth was significantly greater than zero only in 41 of 87 He finds that stagnation because of fixed developing countries in 1950-85, but it was factors is consistent with an array of statistical significantly positive in all OECD countries. evidence. Economic policies - not initial conditions - determine whether countries Analysis of decade-long growth rates in all stagnate. The black market premium on foreign countries shows a striking regularity: Episodes exchange is particularly helpful in explaining of rapid growth are limited largely to a middle stagnation. range of initial income; neither very poor nor very ricih countries experience rapid growth. Empiric, ' results show that growth first Episodes of negative growth are limited to low accelerates and then falls as income rises. and middle-income countries. Results confirm that initial income and policy variables have a different effect on whether a Easterly develops a simple model that sheds country stagnates than they do on the rate of light on this historical experience. The model growth once it starts growing, as expected from has two familiar elements from the growth the distinction between steady-state and transi- litcrature: (I) a Stone-Geary utility function tional effects. (saving is low at low incomes), and (2) fixed factors with the marginal product of capital These results suggest that cross-section bounded away from zero. The second property growth regressions may be misspecified because is derived by assuming an elasticity of substitu- of the nonlinearity inherent in the possibility of tion greater than one between an exogenous steady-state stagnation. Thc l'RE Working P'aper Series disseminatcs the findings of work under way in the Bank's Policy, Research, and Extemal Affairs Complex. An ohjective oftlhe scries is to get these findings out quickly, even ifprcsentations arc less than fully polished. The finding:, interprctat...ns, and conclusions in ihese papers do not necessarily represent official Bank policy. I'roduced by the P'RE Dissemination Center TABLE OF CONTENTS I. Introduction . .................................................1 IL Evidence on output stagnation and growth .................................... 3 TH. A model of policy-induced stagnation . ...................................... 11 IV. Empirical evidence ............................................... 23 V. Conclusion ...............................................30. 3 Bibliography ....... 34 I have benefitted from comments of Robert Barro, Jose de Gregorio, Stanley Fischer, Robert King, Michael Kremer, Ross Levine, Lant Pritchett, Sergio Rebelo, Dani Rodrik, Aiwn Young and Heng-Fu Zou, as well as of participants in a seminar at-University of Maryland and in the Northwestern University Summer Workshop. I am grateful for research assistance from Piyabha Kongsamut and Maria Cristina Siochi. 1. Introduction Stagnation due to fixed factors bulks large in both the old and the new literature on growth. The diminishing returns to endogenous factors with othi_r factors fixed exogenously is at the heart of classical and neoclassical theories of growth from Malthus to Solow. Ma!thus postulated a model with population growth constrained at zero because of the dimininishing marginal product of labor with land held constant. Ricardo put forward his model of rising rents and landlord enrichment based on a fixed supply of land. Mill recognized that other economic forces could offset diminishing returns to frxed factors, but the outcome was far from certain: Whether, at the present or any other time, the produce of industry proportionally to the labor employed, is increasing or diminishing ... depends upon whether population is advancing faster than improvement, or improvement thain population.' Mill seems to be a precursor to the broad notion of capital in the current growth literature in that his "improvement" includes inventions, institutional change, and education and trai,,i..g. Solow could afford to ignore land as a factor of production; in his model it is diminishing returns to capital with exogenous labor growth that prevents sustained per capita' growth. Thus, his famous conclusion that exogenous technological change, the "residual", was the force behind per capita growth. However, in retrospect, the predictions of the model seem to have had decidedly mixed success in describing postwar economic development. The prediction of the model that poor countries would grow faster than rich countries seems to have been confirmed among the subset of advanced countries (or among regions of one rich country) but not among all countries (Baumol and Wolff (1988), Banro and Sala-I-Martin (1989)). Some empirical studies have found that growth is inversely related to per capita income when policy variables are included (Barro (1991), Romer (1989), Levine and Renelt (1990)). The new literature on growth makes growth endogenous by postulating externalities to human or physical capital that overwhelm diminishing returns to fixed factors (Romer (1986, 'quoted in Abramovitz (1989), p.6 2 1990), Lucas (1988)). Another strand of the literature simply omnits fixed factors (Rebelo (1991), King and Rebelo (1990), Barro (1990)), arguing that even labor can be increased endogenously through investment in human capital, so that "everything is capital." This literature bears a resemblance to the development literature of the 1940's and 50's, which also argued that production depended only on capital, albeit for much different reasons -- in the famous Lewis surplus labor model, for example, an infinitely elastic supply of labor makes (physical) capital the only constraint on output.' The new literature on growth has also begun to address the apparent predicament of the poorest countries. Models with multiple equilibria are of particular interest here. Azariadis and Drazen (1990) show how a threshold requirement in the externality generated by human capital accumulation can yield multiple steady states in per capita growth, some characterized by low growth and no human capital investment, others by high growth and high investment in human capital. Similarly, using a model of endogenous fertility, Becker, Murphy, and Tamura (1990) postulate an increasing marginal product of human capital over low ircome ranges to gene:ate alternative steady states of high fertility and zero per capita growth and low fertility and high per capita growth. Murphy, Shleifer, and Vishny (1988) present a model with coordination externalities in which a "big push" may be needed to start development in a low income economy. In all of these models, initial conditions can play a critical role in whether a country develops. Again, these models echo earlier strands of the development literature -- e.g. the "low-level development trap" of Nelson (1956), and the "big push" theory of Rosenstein-Rodan (1947). Other endogenous growth models supply other elements useful to understand the apparent stagnation of the poor countries without reference to increasing returns or initial conditions. Rebelo (1991b) and Easterly (1990a) postulate modelsmi which the rate of saving 2The models cwtinue to be influential up to the piun. For a recent uample, wee Taylor (1989). 3 rises with income, in the tradition of the Stone-Geary consumption function. Rebelo (1991b) presents strong evidence for this hypothesis with analysis of cross-country saving rates. A country can then be stuck in a zero growth equilibrium with "subsistence" income and zero saving. Jones and Manuelli (1990) present an endogenous growth model in which the production function exhibits constant returns and diminishing marginal products of all factors, but the marginal product of capital is bounded away from zero.3 This model has two attractive features: (1) endogenous growth can be explained without any reference to market failure or externalities; (2) the model can generate either stagnant per capita income or sustained growth depending on the parameters. In this paper, we will combine the elements of Stone-Geary consumption behavior with a Jones-Manuelli production function to analyze possible causes of growth and stagnation. The paper is organized as follows. In section II, we present some descriptive statistics on the phenomenon of income stagnation. In section In, we present a model that explains stagnation and growth by policies such as income taxes. Some variations of the model to consider distortionary policies and government investment are also presented. Section IV presents some empirical results which relate the probability of stagnation to policy variables. Section V concludes. II. Evidence on output stagnation and growth Although the euphemism 'developing countries" is uiniversally used to describe poor countries, it is far from clear that sustained per capita growth is underway in all countries. Determining long-run growth tendencies is problematic because of the short time-series available for most countries. Reynolds (1985) concluded that 7 of 40 developing countries whose long-run The eariU grub htteature had aso considued this typc of mode (Gak and Sutherland (1968), Kurz (1968)). 4 experience he analyzed had not begun sustained per capita growth. The 1991 World Development Report of the World Bank shows negative or zero per capita growth for 19 devekc ing countries from 1965-89.4 All developed countries had per capita growth rates well above zero over this period. Income levels at or near subsistence in some low-income countries could also be taken as prima facie evidence that those countries have never grown.' Even for those countries that display positive per capita growth, i. is unclear whether this represents an underlying trend or merely random variation around a stationary income level. To test this for individual countries, the log change in real per capita GDP was regressed on a constant and then the significance of this constant was assessed. The results are shown in table 1.' Only 41 out of 87 develoiping covintries had significant positive per capita growth in the postwar period. In other words, growth is so low and/or the variation in output in 46 of the 87 countries is so great relative to the trend that it is impossible to discern whether the countries are growing or not.7 By contrast, all OECD countries had significant growth rates (not show- in the table). *Te countries arm Ethiopia, Chad, Tanzania, Zair, Madagascar, Uganda, Zambia. Niger, Togo, Benin, Central African Republic, Ghana, Mauritatia, Bolivia, Senegal, Peru, El Salvador, Jaraica and Argentina. Venezuela, Libya, and Kuwait also had negative growth but are excluded because their economies are dominated by oiL Many other countries that probably had negative growth are eexluded because of unavailablity of data: Afghanistan, Bhutan, Kampuchea, Liberia, Myanmar, Sudan, Vietnam, Lebanon. Mongolia, Nicauragua, Iraq, and Romtania. 5 Fe 1990 World Development ReCort defines USS375 per capita consumption as the poverty line in 1985 PPP prices. 10 counries wer below this level in 1988 according to Summen and Heston (1988). This argument was suggested by Lant Pritdhett. 6Countnes dominated by oil are ecluded. An earlier version of these reults is contained in Easterly (1990a). 7To discriminate between insignificance due to low growth and that due to high variation, we calculate the power of the test, as suggested by Andrts (1989). If the absolute value of true growth is less than the coeffident value under "egion of low power", the probability of failing to reject is greater than 50 pecent. A high value of thbis coeffoient implies a weak tes. For example, 13 of 28 countries with insignircnt positive growth have a region of low power spannmg more than (4,11, which mean that even if the true groth rate were above I peren t (or les than -1 percent), ther would still be a S0 percent1chance the test would fail to reect zem growtb. For these countriks, there is little chance of detecting whether growth is occuning - the test is indeed vwy wealk For the other countrie, the region of low power is within 1-1.11. This implies for thowe countries that there is a high probability of failing to rject zero gmwth only if grwtb is in fact close to zero. This technique was suggeted by Lant PritchetLt. Deiled results are available upon request. 5 Table 1 Per capita £owth performance of develoeing countries. 1950-85 Neative growth Positive but Postive and insignificant growth significant gr%Lh Afghlnutan Argentina Algeria Angola Bangladsb Barbados Benin Chile Botswana Bolivia Congo Brazil Burundi Cote d'lvoire Burkina Faso Centrai African Rep. El Salvador Cameroon Chad Ethiopia China Ghana Fiji Colombia Guinea Guttemala Costa Ria Guyana Haiti CyptUs Madagascar Honduras Dominican Republic Mali Jamaica Ecuador Mozambique Kenya Egypt Senegal Uberia Gabon Somalia Mauritania Hong Kong Sudan Mauritius India Zaire Nepal Indonesia Zambia Nicaragua Jordan Nigeria Korea Papua New Guinea Lesotho Peru Malawi Philippines Malaysia Rwanda Malta Sierra Leone Macico The Gambia Morocco Togo Myanmar Uganda Pakistan Uruguay Panama Paraguay Singapore South Africa Sri I anka Suriname Swaziland Syria Taiwan Tanzania Thailand Tunisia Turkey Zimbabwe 18 28 41 Sourc Summers and Heston data seL 6 Although there is doubt about the long-term trend of many countries, there is no doubt about recent stagnation in most developing countries. Table 2 shows ;rrowth rates by decade f r regional groups of developing countries. The African countries (almost all low-income economies) stagnated in both the 1970's and 1980's. Latin America stagnated in the 1980's in the aftermath of the external debt crisis. South Asia has done better than Africa and Latin America, but only East Asian countries (almost all middle-income economies) compare favorably to OECD countries. Figure I shows decade-long per capita growth rates graphed against initial per capita income level. Two striking facts are evident. One is that the phenomenon of negative growth is limited to developing countries. The second is that the t;pper boundary to the distribution displays a bell shape -- the most rapidly growing countries are at middle income levels. (This is more evident in figure lb which displavs a logarithmic scale). Contrary to the predictions of the Solow model, even the 'best" poor countries grow less rapidly than the "best" middle-income countries. However, beginning with Aiddle-income levels, the "best" growth rates decline with income level. The rapid growth of middle-income countries mirrors the earlier experience of "catch-up" of late industrializers such as Japan and Russia, as famously noted by Gerschenkron (1962).8 To see whether this pattern is due to the scarcity of observations in the tails of a bivariate normal distribution, Figure Ic graphs the observations from the sample stratified into equai groups. We still see a strong tendency for the upper boundary of the graph to show a benl shape. Thle "catch-up" pheomenon was attributed by Geshenkron to, among other things, the advantage that latecome* have in botrwing technoogy whicb they do not need to develoq themselve. For P rwt discusions of the dynamicx of technological diffuson and adoption, see Jovanovic, Lach (1990), Wan (1990), and Patento and rattt (1991). Table 2 Growth rates of output per capita, 1960 to 1989 GDP per capita growth Annual averages Country group 1960-70 1970-80 1980-89 Low and middle-income economies 2.2 1.7 0.1 Low-income economies 1.2 0.6 -0 2 Middle-income economies 3.0 2.7 0.3 Sub-Saharan Africa 1.4 -0.2 -0.5 East Asia 3.6 4.6 3.6 South Asia 1.4 1.4 2.3 Latin America and the Caribbean 2.4 2.0 -1.2 OECD 4.1 2.3 2.0 All averages are unweighted. Gil-dominated countries have been excluded. Regional aggregates include only developing countries. Sources: WDR 1981, 1982, and 1990. Ft9ufo t(a) Fguro I t(b) Per capita income and growti I Per capita Income and growth Per capita incore and growPti (Loa arincltoeascading r (Lidnear scaling) Per capka ouixd gfowth Loija(ithmic scating}) Pumpb vdpra ourP-ic d"sw Per capia Luart growthn Pet capta ouZ,0ul1cowth Pe cp SdmpIe SrowtU 10.0 F 0 I 8.0 8a0 00.0 00 0 00 _ 0 0, 9 0 0 02 601 0 0 0 0 0 6.0 00 0 0 o§ 0 .n A., ,&.0 0 (PO ° '° Oooog ,oE a@t O os; S°8aoo e.0 "V!4 |. ° D 0 &D , 0 0 0 CD 0 ' 0 e 0 0 tO0 0 0 4.0 ~~~~~~~~ ~ ~~~ ~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0 oo090 ia? 0 0000 ~~~~~~~~ 40 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~0 00 % o~~ 0 -5.0 ~ *04~0 .0 ° 0 O ° 0t r 20 eO 08 1 o 2000 400 t ooo 0010 80000 120 0 t 0.0 __ 0 o 0 oeg08 00 0 0 0 Oo oo 0 ~~~0 8 0 090 0 o 0 8 0 o4r 0 0 Q604 0 0 2.00 Iniia 2er capia 0 cord a (1985 p Ics) Sal pt cailaIncotza 1 c000 0 0 0 00 0 o ¶ % 0 0 ~ ~ ~ ~ ~ ~ -20 0 000" *o 0-D o 0 080 a 8 0 0 000 0 0 0 -2.0 ~~ ~ ~~~~~0 0 00 0 4.0[ 0 80 l o 0 0 0~~~~~~~~~*4 A ~~~ ~ ~~~~200 600 2000 4000 15000 200 400 600 1200 2000 4000 6600 15000 Inilial ~ ~ ~ ~ ~ ~ ~ ~~~~~~~~~~~taper capila kicrne(1985 prIces) Isriial per capita Incm 18 rcs tIi c ia Icme (1985 piices) kia per capila bxofm (I 98i p*as) 9 To assess whether the patterns presented are statistically significant, Table 3 shows a contingency table and tests for the irlependence of income and growth under various classifications. With a 3-way classification (low, niiddle, and high income, and high positive, mediun positive, and negative growth) the independence of growth and income is decisively rejected. This is an interesting contrast to the well-known lack of significance of the simple linear corTelation between growth and income (in this sample, ;he correlation coefficient is .06).9 Ihe pattern that high growth rates are disproportionately r ,presented among middle income countries is confirmed statistically . t the 5 percent level, as is the relative absence of negative growth at high incomes.'" III. A model of policy-induced stagnation In this section we present a model where policy and model parameters determines whether a country is in one of 3 possible long-run equilibria: (1) zero per capita growth with income at subsistence; (2) zero per capita growth with income above subsistence; or (3) positive per capita growth. Only one equilibrium at a time exhibits saddle point stability, so the outcome is well-defined. We show how the model displays the alternative equilibria depending on the overall rate of incomn. tax. We then consider some extensions of the model to the case of 9This lack of simple correlation between per capita iwcome and growth was noted by, among others, Summers and Heston (1988) and Barro (1991). 1'Alternative breakdowns of growth and income were tested to assess the robust. ens of these results. With high growth defined alternatively as greater than 4 percent and greater than 3 percent, disproportionate rpresentation of high growth at middle ir,comes is confitmed even more strongly. With the income breakpoints at 700 and 7000, a tendency toward high growth at middle noomes is confirmed if high growth is defined as greater than 3 or 4 percent, but not S percent. With income breakpoints at 800 and 6200 (chosen as the 1980 per capita incomes corresponding to the borderline low and high income countriea in the WDR), high middle income growth is again confimed with the 3 and 4 percent definitions, but not the 5 percent. We conclude the result of greater frequency of high growth at middle incomes is reasonably robust. The general independence of growth and initial income (i.e. also induding the lack of negative growth at high income) is rejected at the 1 pecent level with all of these breakdowns, to Table 3: Contingency table of per capita income and per capita growth, decade averages initial per capita income Y < 600 6000 > Y > 600 Y > 6000 totals per capita growth #observations: g>5 1 23 2 26 5>g>O 24 109 51 184 g5 2.9% 13.1% 3.7% 9.8% 5>g>O 68.G% 62.3% 94.4% 69.7% g<0 28.6% 24.6% 1.9% 20.5% sum 100.0% 100.0% 100.0% 100.0% proportions of growth g>5 3.8% 88.5% 7.7% 100.0% 5>g>o 13.0% 59.2% 27.7% 100.0% g<0 18.5% 79.6% 1.9% 100.0% sum 13.3% 66.3% 20.5% 100.0% Chi-squared statistics for rejecting independence of growth and income: for entire table (4 d.f.) 23.58 * for growth > < 5% (2 d.f.) 6.36 * for growth > < 0% (2 d.f.) 14.73 * ccrrelation coefficient of growth rates and per capita income: 0.06 (t-statistic) (.98) Sources: growth, World Bank data;per capita income (85 prices), Summers and Heston (1988) significant at 1 % ** significant at 5% 11 multiple types of capital goods, which is relevant for the analysis of policies that affect resource allocation. Two such policies that are considered are differential taxes on investment goods, and government investment in infrastructure. 1. The model The production function for the single good is a conventional CES function for capital K and labor L: l (1) Y - A (7K6 + (l1-)L6)E The elasticity of substitution between capital and lakor is 1/(e-1). The only difference from a conventional neoclassical specification is that capital is defined more broadly than just fixed physical assets. As in Rebelo (1991b) and Barro (1990), we have in mind a broad concept of capital that includes human capital, "knowledge" capital, "organizational" capital, etc." However, unlike Rebelo and Barro, but like Jones and Manuelli (1990), a fLxed factor like "raw" labor still has a role in production.'2 With such a broad concept of capital in mind, it is assumed e>O, i.e. the elasticity of substitution between capital and labor is greater than one. It is true that a great deal of econometric evidence suggests that this elasticity is less than or equal to one. However, if (1) is the true relation where K is defined to be "broad" capital, the estimation of (1) with only physical capital and labor included would result in biased coefficient estimates, because of the omission of I'Since some or all of these nontraditional types of capital are embodied in people, K should be thought of as including an element hL where h is embodied capital per person. We ignore this complication to simplify the prsentation. 12Another production function tht satisfic the Jones-Manuelli property is Y - AK + BK7L1' (dubbed the 'Sobelow function" by Sala4-Martin (1990) because it is a liner combination of the Rebelo and Solow models). 12 other non-physical types of capital. A large substitution elasticity is plausible if we think of labor- saving innovation (traditionally considered exogenous) as a way of substituting physical capital, human capital, and "knowledge" for labor.' With an elasticity greater than one, this production function obeys the Jones-Manuelli property that the marginal product of capital approaches a nonzero limit as the capital-labor ratio goes to infinity. Specifically, if s>O, then: (2) 1im - k-#aoA- where k is the capital-labor ratio, and y is per capita income. It is assumed that infinitely-lived producer-consumer dynasties maximize the per capita welfare of themselves and their descendants: co (c - c 1-u (3) m x-pt (c- C9) -1Ift 3) max f e P L d 01a Utility is an isoelastic function of per capita consumption in excess of a "subsistence" level of cr The labor term in the intertemporal utility function reflects the weight placed on numbers of descendants vis-a-vis the per capita utility of those descendants, as in Rebelo (1991b) an'd Becker and Barro (1988). If ,B is equal to zero, then only the per capita welfare of future dekscendants is considered. If f= 1, then the aggregate welfare of descendants is considered -- one l-'31s function also has the apparently counter-intuitive property that neither input is strictly essential. i.e. there could still be poitive production vith zero labor. Howwert keeping the broad definition of capital in mind, this does not imply some 21st century fantusy of madhine doing all the work Capital indudes human capitd embodied in persons. 13 is indifferent between an increase in aggregate dynasty "income" because of more descendants and an increase due to higher per capita "income" of an unchanged number of descendants. We assume that income is taxed by the government at rate r. Per capita consumption is constrained by: (4) c - (1-r)A (-ykE + 1-7) - i where i is investment per capita. The evolution of the capital-labor ratio is given by: (5) - i - (6+n)k where q is the rate of exogenous labor growth. The first-order conditions yield the following equation for the growth of consumption: a (1.ilr)A-(-y + (1-7)k ) 6) S= F The first expression in brackets is the familiar condition that growth of per capita consumption is given by the net marginal product of capital less the discount rate and labor growth rate (adjusted by ,B), times the intertemporal elasticity of substitution. The second expression in brackets is the ratio of "excess" (i.e. above subsistence) consumption to total consumption. This term will be close to zero with low consumption and close to one with high consumption. Equation (6) displays two possible zero-growth equilibria. One is the 'modified golden rule" equilibrium where the net marginal product of capital (the marginal product less depreciation) is equal to labor growth (adjusted by f) plus the discount rate. The other is the 14 subsistence equilibrium where consumption is equal to subsistence consumption c,. We will see that at most one of these can be stable, and that the tax rate will determine which one is stable, if either. Unlike nonconvex models, initial conditions do not affect the outcome.14 The value of the capital-labor ratio at subsistence will be given by the k, that satisfies the condition that subsistence consumption is just equal to after-tax income less the investment required to replace depreciated capital and keep up with labor growth. 1 (7) c8 - (1-r)A (-ks + 1-]6 - (6+n7)k This equation could have two solutions for k,: one less than the "golden rule" consumption- maximizing k, and one greater. The lesser one, where the derivative of consumption with respect to k is positive, is the relevant one (the higher one will be dynamically inefficient and unstable). It is also conceivable that (7) would have no solution -- i.e. subsistence is not feasible. There will always be some r that implies infeasibility of subsistence -- this range of r is ignored here. From (7), we can show that the subsistence capital stock will be positively related to c,, r, n, and 6. A higher subsistence requirement, higher taxes, higher labor growth, and higher depreciation all force the consumer to accumulate more capital to satisfy her subsistence requirements. '4Models with multiplicity of equilibria and dependence on initial conditions include Murphy, Shleifer and Vishny (1989), Becker, Murphy, and Tamum (1990) and Azariadi and Drazen (1990). See also tbe discussion in Sa-i-Martin (1990). With nonconwesties in the prcsent modeL whether the economy grows or stagates would depend on initial conditions, since the marginal product cue would inteowt the time preference line in more than one place. However. note that polides could be such as to avoid the dependence on history. A policy change could shift the after-tax marginal pmduct cunve entirely above the sum of labor growth, depreciation. and discount rates, leading to susained growth regarles of initial conditions. 15 The subsistence equilibrium will be stable if the first term in brackets in (6) is negative, i.e. if the net marginal product of capital is less than the discount rate plus the adjusted labor growth rate, evaluated at the subsistence capital stock k,: 1- (8) (1-r)Ay(7+ (1-7)lJE) k - ((l-5)-+6+p) < 0 If (8) holds, then at subsistence the consumer will not find it worthwhile to accumulate more capital (point A in figure 2).' A higher tax rate will make it more likely that (8) holds, both because it lowers the first term directly, and because it increases the subsistence capital stock k,. Similarly, higher labor growth, higher depreciation of capital, and a higher discount rate will all make it more likely that (8) holds, i.e. that the economy will be stuck at subsistence. If (8) is violated at the subsistence level k,, then the consumer will want to accumulate more capital until the net marginal product falls to equality with the discount rate plus the rate of labor growth (i.e. till (8) holds with equality such as point B for a capital level above k, in figure 2). This will be the modified golden rule equilibrium of the Solow-Cass model. Note that this equilibrium will only be feasible if it yields a value of consumption above subsistence. Thus, another interpretation of (8) is that it gives the condition for the subsistence capital stock to lie above the modified golden rule capital stock. Whichever capital stock is greater will be the stable equilibrium. '-"is rWit mifmw that of Rebelo (1991b). 1(' Figure 2 Net marginal product of capital and alternative steady states Net Marginal Subsistence Product Capltal of Capital \~~~~ - - (l-T)dY 6 B 0 g n(1- g) p k 17 However, it is not assured that there is any capital stock sufficiently large to equate the net marginal product to the discount rate plus labor growth. Recall from (2) that the marginal product of capital approaches a positive minimum as the capital labor ratio goes to infinity. If this minimum, net of depreciation, lies above the discount rate plus labor growth, then no stable fixed income equilibria will exist: (9) ~ (1-r)Aye - ((l-jB)r7+6+p) (9) a>0 If (9) holds, then the consumer will find it worthwhile to increase capital indefinitely (see rT curve in figure 2). In the limit, per capita growth will approach the expression given in (9). From (9), we see that stagnation is more likely with higher taAes, higher labor growth, higher depreciation, and a higher discount rate."6 If there is growth, the same factors make growth lower. (However, note that if P = 1, labor growth will have no effect on whether per capita growth takes place.) Combining this result with the previous one, we see policy continuity -- as the tax rate rises, it lowers the rate of growth, until finally growth stops all together. Further increases in the tax rate lower the fixed level of income until income falls to subsistence. A range of tax rates will be consistent with subsistence."7 Figure 3 shows the conventional Cass-Koopmans phase diagrams for the subsistence and modified golden rule equilibrium. Multiple steady states exist, but only one steady state at a time exhibits saddle-point stability. 16A similar result is noted in Jones and Manuelli (1990). 17Fwm (8) and (9) it is apparent that changes in A are equivalent to changes in r of opposite sign. T7his implies that a permanent one-time aogenous technological shift can be sufficient to escape a subsistence income trap, or to move from zero per capita growth to positive growth. This gives an interesting contrast to the traditional neoclassical model in which continuous technological progress is required for per capita growth. Similarty, a negative shock (like a civil war) could induce stagnation in a previously growing economy. 18 The transitional properties of this model are also interesting. From (6), we can see that two offsetting factors will be at work in determining the speed of growth during a transition from stagnation to growth. A country with an initially low per capita capital stock will have a high before-tax marginal product of capital and would grow rapidly, just as in the Solow model. However, this is offset by the low saving propensity at low income levels, as reflected in the second term in (6). Simulations of the saddle path of the model reveal a "hump-shaped" pattem of accelerating then decelerating growth as shown in figure 4."8 This property of the -nodel offers a possible explanation of the rapid growth of middle-income countries compa- .o very poor and very rich countries. 2. Extensions of the Model A model with multiple inputs is relevant to analyze policies which distort the allocation of resources among different activities, policies that are common in developing countries. This extension will also be useful to examine the role played by public sector capitaL I'Accelerating growth (but not decerating) during the transition is noted in the Rebefo'(1991) application of a Stone-Geary utility function. A "hump-shaped" relationship between transitional growth and per capita income is observed in the analysis of King and Rebelo (1990) of the tansitional dynamics of the Solow model with Stone-Geary utility, for eaactly the same reason as in this paper. 'Hump-shaped" transition paths also follow from technology adoption models because of the well4akown logistic cue for new product output (Jovanovic and LaXh (1991), Wan (1990)). Figure 3 Subsistence and modified golden rule equilibria C-O [0 =o] C0 't1] Cs --- c=o-t2 C j i , l ~~~~~~~k _0 T T1% CS ----- ---- -- --- -- -- - -- - C = /I k =O[ t = k0 to] , 1 /t ,~~~~~ 1Ž 21> 2 FIGURE 4 Growth during the transition to the steady state 0.07 0 .05 -~ 0.0l4 -~ ~ ~ ,.03 0.024 00 - ' 2 4 6 8 Per capita income (log) 20 a. Multiple inputs with distortionary policies We extend the production function to include two generic types of capital, K, and K2, with elasticity of substitution 1/(19-1). Capital and labor continue to have elasticity of substitution l/(e-1) (which is still assumed to be greater than one in absolute value): (10) Y - A OK + (1-OMKI l + (1-,)L] c We assume the consumer-producer still maximizes (3). The equations of accumulation of the two types of capital in per capita terms are: (11) 'i - (6+q)kl (12) t2 - 2 (6+n)k2 The type of distortionary policy that we consider will be a sales tax that falls on investment purchases of type 1, with type 2 investment exempted or able to evade the tax. Easterly, King, Levine, and Rebelo (1990) show how this type of structure can be applied to many types of distortionary policies in developing countries, including sales taxes that are evaded by the underground economy, import tariffs and quotas, administrative credit allocation, black market premia in dual foreign exchange markets, and inflation taxes that fall on the monetized sector but are avoided by the non-monetized sector.'9 '9A sales tax on investment type i is also equivalent to a tax on the income from capital tvpe 1. The income tax equivalent to a sales tax t is t/(I+t). The proceeds of the tax are assumed to be nonproductivelv dissipated. 21 Per capita consumption must obey the household's per capita budget constraint: (13) c - A [ E k( + (1.la)k ] 6 + 1-1 1 (1+r) - i where r is the rate of sales tax on type 1 investment. The first order conditions for maximizing (3) imply that the ratio of marginal products of type 1 to type 2 capital is equal to I +T, which implies the following ratio of the type 2 to type 1 of capital, denoted 4): (14) G = [lJ --a) (1+tj i The growth in per capita consumption along the optimal path will be given by: (15) [2 - (C1-p)t7+6+P] [ ] where r. is the derivative of per capita output with respect to the per capita stock of type 2 capital. This in turn will be given by: (16) r2 Ay( 1-a) Lao + 1-a) + (1-y)k2j (aZ + 1-a) As before, if ef>O (elasticity of substitution between capital and labor greater than one in absolute value), the marginal product of capital will go to a nonzero limit as both capital-labor ratios go to infinity (recall the ratio 4D is given by (14)). Specifically, we have: 1l- v17) lim r Ay (I-a) (a- + 1-a) k2 22 If this limit is greater than the sum of the discount rate, depreciation rate, and labor growth, then positive per capita growth will ensue at the asymptotic rate: 1 1-0 (18) g Aye (1-a) (ax @ + 1-ca) - (S+(G-,8)t+p) In other words, per capita growth will take place if the right-hand side of (18) is positive. If (18) is negative, per capita output will stagnate and the capital-labor ratio will be such as to satisfy the modified golden rule. From (14) and (18), it can be seen that distortionary policies tend to make stagnation more likely.' An increase in the distortionary tax r will increase the ratio of type 2 to type 1 capital (14) above the socially optimal level. This could lower the asymptotic marginal product of type 2 capital (17) sufficiently that (18) becomes negative and growth stops. Further tax increases can cause a regression toward subsistence income, just as in the previous section. b. Public capital and growth The model of the previous section can also be used to discuss the effect of public capital on growth. It is plausible that there are capital inputs that can only be provided by the public sector. We consider public capital inputs that will not be forthcoming in a competitive market system (say because of technological difficulties in charging per unit of use), but otherwise satisfy the usual properties of private goods (rivalry in consumption, perfect divisibility, diminishing marginal product, etc.). Equation (10) can then be used to cover the case of public and private capitaL Government 'capital investment" includes all activities that contribute to human or physical 2Thm is in contrst to the argument that distortionary policies ony have level effects, as argued by Lucas (1988) and Young (1991). A micw-buad modd with effet of distortionazy polices on growth is Murphy, Shilifer and Vishny (1991. 23 capital, such as education, highways, basic health measures, or electrical distribution. K1 is interpreted as a government capital input, and K2 is interpreted as private capitaL. We assume that the government finances the construction of public capital with lump-sum taxes (taxes which do not affect growth). The government is assumed to follow a policy rule where the ratio of public to private capital is maintained constant over time. r in equation (14) can be interpreted as a measure of the ex-post distortion induced by supplying too little (positive r) or too much (negative r) public capital. Equation (18) now gives the asymptotic growth rate determined by private sector investment in type 2 capital. If (18) is negative, output will stagnate. We see that output is mor_ likely to stagnate the lower is the ratio of public to private capital (i.e. the higher is the ratio of type 2 to type 1 capital cD). The reason is simple: lower public capital lowers the asymptotic rate of return to private capital, possibly below the critical value given by the sum of the depreciation, population growth, and discount rates. Since growth is determined by the private return to capital, higher public capital always increase, the likelihood of growth, even if it is suboptimal from the standpoint of total welfare.2" IV. Empirical evidence The model makes several predictions: (1) countries that penalize capital or distort its allocation are more likely to stagnate (and such policies will cause lower growth if a country is growing); (2) initial income does not affect whether countries stagnate; (3) countries that do grow will follow a hump-shaped transition path where growth rises and then falls with rising income. 21An obvious etenson is to consider public capital spending financed by a tax that affects growth. This was considered for more general growth models in Barro (1990) Bam and Sa1ai4Martin (1990), and Eterl (1990b), and is not diectly considered hem 24 These predictions differ from those of other endogenous growth models underlying recent work on growth (e.g. Barro (1991)), in that right-hand side variables do not affect growth continuously. The model suggests that counties can be in one of two regimes -- either sustained growth, where right-hand side variables have growth effects, or stagnation, when a function of right-hand side variables passes a threshold level. The determination of stagnation involves steady- state factors -- forward-looking consumer-producers decide on the basis of preference, production, and policy parameters whether steady-state growth is worthwhile.' The growth of growing countries. on the other hand, includes transitional dynam.s such as the aforementioned hump- shaped relationship between growth and initial income. Policy variables also could have transitional effects on growth that differ from their effects on the stagnation/growth outcome. This formulation suggests the use of limited dependent variable methods to take into account the truncation of growth rates induced by stagnation. A probit equation will be specified to predict whether countries stagnate. A truncated regression will predict the growth rate of growing countries. Under a null hypothesis of the conventional continuous model, both methods would still yield consistent estimators of the effect of right-hand side variables on growth. The continuous model in effect imposes the restriction that the coefficients are the same in the two equations.' This implication can be tested by nesting the probit and truncated regressions within a tobit equation (which also yields consistent estimators of the continuous model under the null hypothesis) and constructing a likelihood ratio statistic for equality of coefficients between the probit and truncated equations.24 If equality of coefficients were rejected, the continuous model 22Positive (or negative) growth could still result from transitional dynamics from one stagnant equilibrium Wo another, such as that due to a favorble (or unfavorable) policy change. There is no a priori reason to epect such effects to be large for stagnating countries. 23Adjusted for the standard error, since the probit is based on the standard normal distribution. 24Greene (1990) has a lucid description of this procedure Note the 'equality of coefficient" must be evaluated with probit coefficients (baud on the sandard normal) adjusted for the size of the standard aror. 25 would be shown to be inappropriate and the prediction of different regimes for growth and stagnation would be confirmed. The empirical problem is to define stagnation. The approach taken here is to define a country as stagnating when its growth rate is below 0.1 percent. AU countries with negative growth are presumed to be in a transition towards a lower fixed-income equilibrium. In the tobit equation, for example, the dependent variable will be defined as zero for ali observations that satisfy this definition of stagnation, while the actual per capita growth will form the dependent variable in the non-stagnating cases.' The set of right-hand side variables indicated by the model include (1) implicit or explicit taxes on capital (section 11.1), (2) variables reflecting policy distortions of resource allocation (section 1I.2.a), and (3) variables that reflect government physical and human capital spending (section II.2.b). Labor growth and per capita income will also enter as explained earlier. The equations are estimated alternatively with all variables defined as 10-year averages (table 4) and 30-year averages (table 5), except for per capita income, which is given as income in the first year of the period. We also show in the table the probit coefficients adjusted by the standard error to be comparable to the tobit and truncated coefficients. One set of variables common to other empirical growth work that is not used here are investment ratios and measures of human capital like primary and secondary enrollment ratios. Since the model relies on a definition of capital that includes unobservable components like training and knowledge, it seems best to estimate reduced forms that do not rely on measuring 2lThis proedure is equivalent to discarding the information contained in differences among negative grwth observations. As argued before, this does not bias the estimates even if the true model is continuous. The sample is endogenously truncated, and the limited dependent variable methods then correct for the truncation. This roundabout prodecure allows us to test the implication of repwme change with separate pn*it and tmncated regressns. 26 capital accumulation. Only exogenous variables (policy and other) will be included in these regressions.' The results obtained are as follows:7 Lkelihood ratio tests. The restriction that coefficients are the same across probit and truncated regressions is rejected in five out of the six regressions using decade averages (Table 4). It is notable that the coefficients on per capita income (and income squared) are insignificant in the probit equation but significant in the truncated equation. These results support the prediction that initial income does not affect whether a country stagnates or not but does affect the growth rate if it grows.8 In the regressions with 30-year averages (Table 5), the restriction is not rejected. It makes sense that the restriction is more likely to be rejected with 10-year averages but not with 30-year averages, since one would expect transitional effects to be stronger with the former. However, one would have expected transitional effects to still be important with 30-year growth rates. Initial per capita income. The hump-shaped relation between income and growth is confirmed by 5 out of the 6 regressions using decade averages (Table 4), as both income and income squared are significant with the predicted signs in the truncated regressions.' The maximum of the hump varies between $900 and $1800 in 1985 prices. This contrasts to the negative linear effect of per capita income on growth found by, among others, Balassa (1985), 26Some regrsios ere tun with the total investment ratio on the right-hand side to cumine whether it cruciatly affeds the remltL Tbe interpretation of such equato would be that station would oocr ether beause inetment was too low or bemuse other vanabls lowefed the efficiency of inVmSment. Investment was genally significant and the results wse otherwise similar to those tpoted hber However, thes regrks are problematic bceause investmet is pmably endogenous, which is difcult to addrs in the limited dependent variable contect '?For the citation of previous mesult, I was muited greatly by the survey of Renedt (1991). mWhen the sample is restricted to developing countrie equality of coeffidents is reJected in 3 out of the 6 regresuioni. The weer esult is not surprisng in view of the narrower range of the per capita income variable in this case 29Apin, the results ar weaser if the sampe is limited to deveoping countries, with only I out of the 6 truncated regresion showing significnt cffidents on income and inoome squared. 27 Barro (1991), FLscher (1991), Grier and Tullock (1989), Landau (1986), and Murphy, Shleifer, and Vishny (1991).3 A quadratic term was found to be marginally significant by Barro (1991), but with the opposite sign from that found here. No effect of per capita income on growth is detected in the regressions with 30-year averages for all variables (except initial income) in Table 5. Again, it is not surprising that transitional effects are weaker with 30-year averages." The black market exchange rate premium. This is a widely available measure of price distortion. reflecting an implicit tax on producers of traded goods that are priced according to the official exchange rate. For example, it is a tax on exporters that are forced to deliver foreign exchange at the official rate, rather than the black market rate. We assume the proceeds of the "tax' are dissipated. An increase in the black market premium should than make stagnation more likely. Levine and Renelt (1990) and Easterly (1990b) found this variable to be insignificant in cross-section regressions." By contrast, the black market premium is found to be consistently significant here, no matter what other right-hand side variables are included. The significance of the probit and truncated coefficients varies -- with 30-year averages, it is the probit that is consistently significant, while with 10-year averages, both are generally significant. We conclude that the black market 3OHowever. Michael Kremer reports finding the "hump-shaped' pattern predicted here in unpublished results. 3UAlthough again the lack of significance with 30.year averages is disconcerting, since simulations seem to indicate that transitional dynamics in the Jones-Manuelli model can be quite prolonged. 32However, other measures of price distortions have been found to be significant in the literature. Barro (1991) reports that the absolute value of deviations of the relative price of investment goods is significantly negative in a cross-country growth regression. De Long and Summaer (1991) found that a high relative price of equipment investment goods has a negative effect on growth. De Long and Summers (1991) and Easterly (1990a) found a dummy variable measuring outward trade orientation from the 1987 World Development Repoin to have a positive and significant effect on growth. Dollar (1990) found a measure of general overvaluation of real erchange rates, based on Summers-Heston relative price data, to lower growth. 28 riremium is a good predictor both of whether countries stagnate, and how fast they grow if they do not stagnate. Public investment as a share of GDP. The theory predicts that higher public investment (measured conventionally as physical investment only) makes stagnation less likely. This variable is only available for the 1970's and 80's, and only for a reduced sample of countries. (For this reason, this variable was omitted from the regressions with 30-year averages). Other studies, such as Barro (1991) and Khan and Reinhart (1990), have generally found this variable to be insignificant in growth regressions. The regressions with decennial averages show some evidence that higher public investment makes stagnation less likely."3 Public investment is positive and significant at 5 percent in one probit regression and at 10 percent in another. H{owever, we also find the puzzling and significant result that higher public investment causes growth to be lower in the truncated regressions. This surprising finding is inconsistent with theoretical predictions and merits further investigation. Govemment consumption. The share of government consumption in GDP is found to have a significant effect on growth in studies such as Barro (1990, 1991), Romer (1989a), and Easterly (1990b). In terms of this model, government consumption can be seen a proxy for the part of the tax burden not offset by productive spending. Thus, an increase in government consumption should make stagnation more likely. We fmd no evidence for such an effect, however, as govemment consumption is only significant at the 10 percent level in one regression, and of the wrong sign. 33No regrssions were run induding public investment for the 30-ycar averages, because the sample size was too small for the use of nonlinear e_nowetnic techniques. 29 Government exgenditure on education. This measures one form of productive government investment Higher education spending should make stagnation less likely. This variable was found to have an insignificant effect on growth in Diamond (1990). In contrast, we find some evidence here that government spending on education influences stagnation and growth, as it is significant in the tobit and truncated regressions for 30-year averages, and in one of the tobit and probit equations for decennial averages. The significance is not robust, as it vanishes when variables like government consumption are added. Labor force growth. The model predicts that population or labor growth has a zero or negative effect on growth, depending on whether the parameter fi is equal to or less than one, respectively. If it is one, then consumers place a value on the number of their descendants that exactly offsets the negative effect of having to spread future capital around more people. Thus, higher labor growth will either make stagnation more likely or will have no effect. Mixed results for the effects of labor growth on per capita growth have been reported in the literature: Barro found it to be significantly negative, Grier and Tullock (1989) positive, and Balassa (1985), De Long and Summers (1990), Landau (1986) and Mankiw, Romer and Weil (1990) insignificant. The results here include some weak evidence for high labor growth making stagnation more likely. The coefficient on labor in probit equations is significant at 5 percent in one instance, but the coefficient reverses sign in other specifications. Inflation Inflation represents a tax on investment to the extent that cash must be held in advance of investment transactions. Higher inflation will make stagnation more likely if these cash-in-advance requirements are significant. Inflation was found to be significantly negative in growth regressions in Grier and Tullock (1989) and Fischer (1991). The results here show some evidence that inflation makes stagnation more likely, as-the coefficient is significant and negative in one of the truncated regressions (regression v in table 4). The coefficient on 30 inflation in the probit equation with the same set of variables is not significant. 'ven the significance of the inflation coefficient in the probit regression vanishes in other specifications. Financial reRression dummy. Controls on interest rates such that real interest rates in the financial system are highly negative will lead to allocation of credit by administrative fat. This imposes a tax on investment by those who do not have access to subsidized credits; we presume the subsidized crec&its themselves to flow to nonproductive uses. This variable is defined as 1 if the average real deposit interest rate over the period is less than -5 percent and 0 if it is greater. A value of 1 makes stagnation more likely. This variable was found to be significant in Gelb (1990), Easterly (1990b), and Roubini and Sala-i-Martin (1991). However, we fail to find any evidence for financial repression affecting growth or stagnation in these results. The main effect of including this variable is to render insignificant per capita income and the likelihood ratio test statistic. Time and continent dummies. The model considers only national policies as affecting growth, but it is plausible that there are also global influences (the well-known slowdown in world growth in the 80's for example). The regressions control for these influences by putting one dummy variable each for the decades of the 60's and 70's. We also consider continent dummies that other studies have found to be significant. The dummies for Latin America and Africa are generally significant in these results, as they have been in most other studies (e.g. Barro (1991)). This suggests there are other factors influencing growth and stagnation that have not been captured here. The time dummies are also generally significant in both probit and truncated regressions, which could be indicating some exogenous worldwide productivity trends.34 34Te model of section 11 could be modified to incorporate rngenous productivity growth. Stagnation would then be defined as gowth equal to the cogenous productivity trend. The result here indicates that stagnation (defined as zero growth and below) became more likely in the 1980m which could be interpreted as a decrease in aogeMous productivity grMwtb. 31 V. Conclusion Stagnation due to the presence of fixed factors is consistent with an array of statistical evidence. Economic policies, and not initial conditions, determine whether countries stagnate. The black market premium on foreign exchange is particularly helpful in expiaining stagnation. Empirical results show that growth frst accelerates and then falls as income rises. Results confirm that initial income and policy variables have a different effect on whether a country stagnates than they do on the rate of growth once it starts growing, as expected from the distinction between steady state and transitional effects. These results suggest that cross-section growth regressions may be misspecified because of the nonlinearity inherent in the possibility of steady-state stagnation. 32 \ Tmhk 4 Tobi. Prebi. end Truaed sqguk rmults regmaion by decades Cr-swht in paradsa) I)pmdt vdiabIe PMr capta growtb Consant Exdu gc ia- lMM Govt Intil Squarc of Govt labor focsc Finan- lime lsUN Asiu Africn 3Mmt Sampbi LJkcIlbo.l Mcx- . Raft tin LIvecit expaad. kvel Iial isv cmOnp. growth cit dummy r ofODP 8oab shae of dum ym per -Pkt per capot GDP L TobM 11042 402 -103 40$ 0.003 -0.003 -QOD3 402Qr" 60 256 (1.36) (-231) (49) (49) (1.16) (425) (-2.8) (-2.5) Probh 391 -1.782r 123 -I0.SI 0.43 0.38 -5.62 -5.84 60 (1101) (-2.) (0.5) (.I.57 (1.28) (0) (401) (401) Adjused Problt 1104 40r- 0.002 4106 0o04 4004 406 406 TNun0td Q0S 402 4.009 4.006 0.002 -0.004 403-- 4.0r- 48 (1.2) (-L55) (1.9 (413) (Q66) (436) (-237 (-1W) 2 TobAt 401 402 11OOO 401 OOD6 60 417 (441) (-186) (02) (428) (t.49) Prob -.2.17 -1.47P 0124 -5.44 0.47 60 (-407) (.04) (0192) (.1.21) (1.47) Adjuted Probit 4Q3 4022" 1104 40S1 O07 Tnrcat.d Q02 4007 4005 1105 1.006 48 (0.55) (4149) (476) (O15) (1.16) 3 Tobkt 0.o4 40r2 4004 Q2S- 1O1OI -Ur, 40r 54 9.95 (19) (-243) (-L33) (2.4) (0.15) (-2.4S) (-1L93) Probit 403 -1.$9 1o2 7.53 1143 -5.11 472 $4 (QOt) (-1.91) (1O1) (0148) (1100) (-40) (401) Aduted PrLblt 103 401 11002 0.065 OD0037 404 404 TnrJated 1106 402 40DS 0 29W" 0.000 402r 42-" 43 (146) (.1.U) (4.24) (232) (1107) (-231) (-2.3) 4 Toblt 4005 402 " 4001 0.39' 54 499 (419) (-2.3) (43) (1.76) Probit *1.84 1.41"- 01196 10.46 54 (493) (-1.97) (172) (0.68) AdjUs Probet 402 402r 11002 1113 Tnmcr e 1103 401 4004 Q14r 43 (am7) (1LOI) (-97) (1.93) 34 Table 5 (continued) Tcbiu, Prol,i and Trun4ated r.gression reahu bor 30year-avege penod Cr-atstics in parcothe) Dq.sdaet varablie per capita gmlh Constm Echnge Infatic Inlitial Square of Govenment Ljbor Govt ep Ast Afrnc LAtin Sample Ueliod Rote weel initbl kvd ojPti force on eduation dummy dumm Aneia size ratio Premium Of Income or inotme share of GDP rcthb sbare of dummy test per cata per capita CDP S. Tobit 0.04 402- Q007 40DS 029- Q002 402-- 402" 52 1106 (134) (-268) (Q.81) (-1.46) (238) (023) (-2.47) (-2.09) Probst 3.72 -Z24 * L67 404 1213 0.8 $529 -5.0S 52 (QOI() (-2 03) (L1) (41) (0.73) Coaool) (401l) (410l) Adjusted Probit O0t3 402er L014 4-O03 olto aGOO.s -005 -004 Trctted 0.06 -0.02 Q.003 400S 025$ 400004 -002 402-- 41 (152) (-21) (0.53) (-L34) (1.95) (4005) (-z23) (-z28) 6 Tobit 404 402"* Q02 4002 028, 0001 4027o a0oz 54 10.05 (432) (-2.26) (047) (456) (Z34) (all) (-229) (-187) Probit -5.36 -1.46 27 419 6,75 0.3 -5.00 -4.7 54 (401) (-176) (0.61) (46) (0.43) (0) ( 1) (401) Adjusted Probit 405.0 40r 0023 40016 006 0.0026 404 404 Trnaned J.0. -0.02 0.006 40009 0 30tr OO004 4002 40r. 43 (0.04) (-116) (L0.) (414) (217) (0.06) (-1.93) (-223) - Sadficane at 10% levld -- Significant at 5% ksed - Signifa at 1% levd 35 Var,able Definition Dependent variable: Per capita income growth = log (1 + growth); SOURCE: World Bank Database. Per cap income growth, PROBIT = dummy of per capita income growth; Equal to 0 if growth < 0.1; 1 otherwise. SOURCE: World Bank Database. Per cap income growth, TOBIT = Equal to 0 if growth < 0.1; 'g' otherwise, where 'g' is the growth and TRUNCATED rate. In the truncated regressions, the g <0.1 % observations are excluded. SOURCE: World Bank Database. Independent variables: Exchange rate premium = log (1 + black market exchange rate premium); SOURCES: 1960-83: World Currency Yearbook; 1984: Pick's Currency Yearbook (various years), 1984-89: Financial Times International Reports' Statistical Market Letter, 1989: Africa Analysis for certain countries. Inflation = log (1 + inflation rate); SOURCE: World Bar. Database. Public Investment, share of GDP = log (1 + real public investment as a share of GDP); SOURCE: Pfefferman, G. and A. Madarassy. 'Trends in Private Investment in Developing Countries". IFC Discussion Paper No. 11, World Bank (1989). Initial level of income = log (GDP per capita income) initial year for each decade, i.e., per capita 1960 for the 60's, 1970 for the 70's, 1980 for the 80's. SOURCE: Summers and Heston. Government consumption as a = log (1 + real government consumption as a share of GDP); share of GDP SOURCE: World Bank Database. Labor force growth = Average annual growth rate of population of working age (15-64). Based on data in World Development Indicators, World Bank (1987). Government expenditure on = log (1 + nominal government expenditure on education as a share education, share of GDP of GDP). SOURCE: World Bank Database. Dummy Variables: Financial Dummy = Dummy variable for financial policy distortions. I if real interest rates are less than -5%, 0 otherwise. Based on World Develonment Report 1989. Data from- 1965-85 from Financial Policy Division, World Bank. 36 BIBUOGRAPHY Abramowitz, Moses. 1989. Thincing about Growth. Cambridge: Cambridge University Press. Andrews, Donald W. K 1989. "Power in econometric applications." Econometrica 57 (Sep.): pp.1059-1090. Azariadis. C and A. Drazen. 1990. "Threshold Externalities in Economic Development." Quarterly Journal of Economics 105: pp. 501-26. Balassa, Bela. 1985. "Exports, Policy Choices, and Economic Growth in Developing Countries After the 1973 Oil Shock." Journal of Development Economics 18 p. 23-35. Barro, Robert. 1990. "Government Spending in a Simple Model of Endogenous Growth. "Journal of Political Economy 98: pp. S103-S125. 1991 "Economic Growth in a Cross Section of Countries."Quarterlv Journal of Economics. (same as De Long and Summers), pp. 407-444. and Xavier Sala-i-Martin. 1989. "Economic Growth and Convergence Across the United States." Harvard University. and . 1990. "Public Finance in Models of Economic Growth." NBER Working Paper No. 3362. Cambridge, MA. NBER. Baumol, William and E. Wolff 1988. "Productivity Growth, Convergence, and Welfare: Reply." American Economic Review. December. Becker, Gary S. and Robert Barro. 1988. "A reformulation of the economic theory of fertility." Quarterly Journal of Economics 103 (Feb.): pp. 1-25. Kevin Murphy, and Robert Tamura. 1990. "Human Capital,Fertility, and Economic Growth." Journal of Political Economy 98(5). De Long, J. B. and L H. Summers. 1991. "Equipment, Investment, Relative Prices, and Economic Growth." Quarterly Journal of Economics May 1991, Vol CVI, Issue 2, pp. 445- 502. Diamond, J. 1989. "Government Expenditure and Economic Growth: An Empirical Investigation." IMF Working Paper No. 89/45. Washington, DC: IMF. Dollar, David. 1990 "Outward-oriented countries do grow more rapidly". Mimeo, World Bank Easterly, William. 1990a. "How Does Growth Begin? Models of Endogenous Development." World Banl. 37 1990b. "Endogenous Growth in developing countries with government-induced distortions." World Bank . R. King, R. Levine, and S. Rebelo. 1990. "Do national policies affect long-run growth? A research agenda." World Banlk Fischer, Stanley. 1991. "Growth, Macroeconomics and Development," NBER Macroeconomics, Cambridge, MA (forthcoming). Gale, and Sutherland. 1968. "Analysis of a one good model of economic development." In Dantzig, George B. and Arthur F. Veinott, Jr. Mathematics of the decision sciences Part II. Providence: American Mathemstical Society. 1968. Gelb, Alan. H. 1989. "Financial policies, growth, and efficiency." PPR Working Paper No. 202. Washington, DC: World Bank. Gerschenkron, Alexander. 1962. Economic Backwardness in Historical Perspective.Cambridge MA, Harvard University Press. Greene, W.H. 1990 Econometric An .ysis. New Yorlc MacMillan. Grier, Kevin and Gordon Tullock 1989. "An Empirical Analysis of Cross-National Economic Growth, 1951-1980." Journal of Monetary Economics 24:p.259-276. Heston, Alan and Robert Summers. 1988. "What We Have Learned about Prices and Quantities from International Comparisons: 1987." American Economic Review 78(2): 467-473. Jones, Larry and Rodolfo Manuelli. 1990. "A Convex Model of Equilibrium Growth." Journal ofPolitical Economy. 98 (5): pp. 1008-38. Jorgenson, Dale W. 1961. "The development of a dual economy." Economic Journal 71 (Jun):pp. 309-334. Jovanovic, B. and S. Lach. 1990. "'he diffusion of technology and inequality among nations." Paper presented at a NBER conference on economic growth, November 1990. King, Robert, and Sergio Rebelo. 1990. "Public Policy and Economic Growth: Developing Neoclassical Implications." Journal of Political Economy 98: pp. S126-150. Kurz, M. 1968. "The general instability of a class of competitive growth processes." Review of Economic Studies 35 (Mar) : pp. 155-174. Landau, Daniel. 1986. "Government and Economic Growth in the Less Developed Countries: An Empirical Study for 1960-80." Economic Development and Cultural Change 35 (October) p.35-75. Levine, Ross, and David Renelt. 1990a. "A Sensitivity Analysis of Cross-Country Growth Regressions." World BankL 38 Lucas, Robert. 1988. "On the Mechanics of Economic Development." Journal of Monetary Economics 22 (July) p.3-42. Mankiw, N. Gregory, David Romer. and David Weil. 1990. "A Contribution to the Empirics of Econonuc Growth." Harvard University. Murphy, Kevin, Andrei Shleifer, and Robert Vishny. 1989. "Industrialization and the Big Push." Journal of Political Economy 97 (October) p.1003-1026. . 1991. "The allocation of talent: implications for growth," Quarterly Joumal of Economics May 1991, Vol CVTI, Issue 2, pp. 503-530. Nelson, Richard R. 1956. "A theory of the low-level equilibrium trap in underdeveloped economies." American Economic Review 41: pp. 894-908. Parente, Stephen L. and Edward G. Prescott. (1991) 'Technology Adoption and Growth," NBER WP#3733 Rebelo, Sergio. 1991a forthcoming. "Long run policy analysis and long run growth." Jourmal of Political Economv. 1991b. "Growth in open economies." World Bank. Renelt, David. 1991. "Economic Growth: A Review of the Theoretical and Empirical iterature" Word Bank PRE Working Paper 678. Reynolds, Lloyd G. 1985. Economic Growth in the Third World. 1850-1980. New Haven and London: Yale University Press. Romer. Paul. 1986. "Increasing Returns and Long-Run Growth." Journal of Political Economv.94 p.1002-1037. . 1989. "What determines the rate of growth of technological change?" PPR Working Paper No. 279. Washington, DC: World Bank. . 1990. "Endogenous technological change." Journal of Political Economy 98: pp. S71-S102. Rosenstein-Rodan, P.N. 1943. "Problems of Industrialisation of eastern and south-eastern Europe." Economic Journal 53:pp. 202-211. Roubini, Nouriel and Xavier Sala-I-Martin. 1991. "The Relation Between Trade Regime, Fmancial Development and Economic Growth." NBER Fourth Annual Interamerican Seminar on Economics Sala-I-Martin, Xavier. 1990. "Lecture notes on economic growth tI): Introduction to literature and neoclassical models." NBER Working Paner No. 3563. Cambridge, MA: NBER. Taylor, Lance. 1989. Stabilisation and Growth in Developing Countries: A Structuralist 39 Approach. London: Harwood Academic Publishers. Wan, Henry. 1990. 'Trade, development and inventions." World Bank seminar. Young, Alwvn. 1991. "Learning by Doing and the Dynamic Effects of International Trade," Quarterly Journal of Economics. May 1991, Vol CVI, Issue 2, pp. 369-406. Policy Research Working Paper Series Contact 2191 Author I2L for laaer 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 September 1991 S. Jonnakuty Shocks and Fiscal Reform: Klaus Schmidt-Hebbel 39074 Sim:jlations for Zimbabwe, 1988-95 WPS773 Are Cihana's Roads Paying Their Reuben Gronau September 1991 J. Francis Way" Assessing Road Use Cost and 35205 User Ci arqes 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 Devaralen October 1991 A. Bhalla 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 Ian 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 Octeber 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 Autho for paper WPS785 How Did the Asian Countries Avoid Ishrat Husain October 1991 S. King-Watson the Debt Crisis9 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. Cabana 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 William Easterly October 1991 R. Martin Affect Long-Run Growth? Robert King 39065 A Research Agenda Ross Levine Sergio Rebelo WPS795 Economic Stagnation, William Easterly October 1991 R. Martin Fixed Factors, 39065 and Policy Thresholds