u.eS I 2Li POLICY RESEARCH WORKING PAPER 1221 Does Research and Across all cotntries, more research and develoment Development Contribute (R& iviydosno to E,conomic Growth coetibUte to highler incomef. to Economic Growth RHigher income, t.owever, in Developing Countries? does lead to moreR&D activity, suggesting that R&D becomes important only after Nancy Birdsall a country reaches a certain Changyong Rhee stage of development. For developing countries, it appears that countries that are behind grow by catching up technologically, not by advancing the technological frontier. The World Bank Policy Research Departnent November 1993 POLICY RESEARCH WORKING PAPER 1221 Summary findings Using UNESCO Aata for research and developnient raither than vice versa. First, thtre is no evidence that (R&lD) expendirures and personnel, Birdsall and Rhee R&D in the OFCD in the early years of 1970-85 document international differer.ces in R&l) a-:tivities contributed to growth durmng the whiiole period. and assess the determinants of these differences and Second, the anralysis of deteriniiianits of R&D activities the link between R&I) and ecotiomic growth. suggests that level of income affects R&D activities; For a group of OECD cotiuitries, R&D activity and appairently R&D becomes important only after a economic growthi are correlated for one of their two coointry reaches a certain stage ot development. proxies. Contrary to the findings of Romer and For developing countries, the authors' results are Lichtenberg, however, they are not corielated acros, consistent with the widespread view, first proposed by all (including developing) countries. (Gershenkron. that countries ti.it are b hind grow by Moreover, even for OEC( countries, it appears catching up technologic,ally, not by ad-ancing the likely thit economic activity affects R&D activity technological frontier. This paper - a product of rhe Office of the Director, Policv Research Department - is part of a larger effort in the department to assess the sources of growth across counitries. Copies of tthe paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Sushma Rajan, room NI 1-045, extension 3,3747 (25 pages). November 1993. The Policy Research Working Paper Series disseminates the findings of uwork in progress to e:courage the exchange ol ideas about development issues. An objective of the series is to get the findings out quickly, even i. the presentations are less than fully polished. The papers carry the names of the authors and should be used and cited accordingly. The findings, interpretations. and conclusions are the authors' oum and should not be attributed to the World Bank, its Executive Board of Directors, or any' of its member countries. J Produced by the Policy Research Dissemination Center Coniments Welcome DOES R&D CONTRIBUTE TO ECONOMIC GROWTH IN DEVELOPING COUNTRIES? Nancy Birdsall and Chan3yong Rhee* * World Bank and University of Rochester. This paper was completed while the first author, now at Inter-American Development Bank, was Director, Policy Research Department, World Bank and the second author was visiting the World Bank as a Visiting Research Fellow. 1. Introduction R&D (research and development) activity is generally acknowledged as a likely source of economic growth, given it is a critical input to technological progress, and given evidence from firm and industry level studies of healthy rates of private and social returns (Mansfield [1977], Jaffe [1986]). Yet except for Romer[1989] and Lichtenberg [1992], we know of virtually no other cross-country studies of economic growth that incorporate country-level data on R&D activity -- despite a recent resurgence in the growth literature, incorporating an increasing array of economic, social and political variables.' The paucity of studies reflects in part the relatively poor state of country-level R&D data that is sufficiently comparable for international comparisons. For this paper, we developed such daLa by undertaking country-by-contry adjustments to information made available by UNESCO on R&D expenditures and personnel at the country level.2 On the basis of our adjusted data, we assess empirically the link between R&D activity and economic growth, and we analyze the determinants of R&D differentials across countries. We find that growth in R&D activity and economic growth are positively correlated, but only across OECD countries (conttrary to the findings in Romer [1989] and Lichtenberg [1992]). And even among OECD countries there is no evidence that R&D activity causes economic growth. Our cross country analysis does show a robust empirical link between R&D activities and the level (not growth rate) of income, suggesting that R&D activities become prominent only after a country reaches a certain stage of development. 'See, for example, Easterly (1993), Levine and Renelt (1992), and Barro (1991). 2Our adjustments were made on the basis of details on definitions and exclusions that are included in the footnotes to the published UNESCO data. We believe our findings are fully consistent with the view that growth in developing countries is largely a matter of technological catch-up, as first proposed by Gerschenkron, i.e. based on learning from and adapting technologies developed elsewhere rather than on advancing the technological frontier.3 This probably requires eytensive investments in education and training, including in engineering, but not extensive R&D activity. The paper is organized as follows. Section II describes our data. Section III sets out cross country analysis of the growth-R&i) ink, and Section IV sets out our analysis of R&D differentials across countries. Section V concludes. (I. Data The first international attempt to standardize the definition of R&D and the measurement of R&D inputs was made by OECD member countries at a conference in Frascati, Italy, in 1963. The resulting "Frascati Manual" has provided the basis for OECD's collecting and publishing R&D data on a regular basis. Since 1969, UNESCO has published standardized R&D statistics for some developing countries, based on the Frascati concepts. We use the published UNESCO sources, plus a supplementary UNESCO tape, for our country R&D data.4 3Page and Pack (1993) distinguish between technological catch-up and increases in "technical efficiency" and show that the most successful economies of East Asia have grown primarily on the basis of the former. Their finding is consistent with the view of Romer (1989) that growth is affected by the transfer of ideas, and the empirical finding of Kim and Lau (1993) that there has been little or no increase in TFP growth in East Asia, when TFP growth is defined narrowly as increases in technical efficiency. 'We are grateful to Mr. Morton M. Brown of the statistical office in the UNESCO, Paris for providing a supplementarv tape. 2 We use measures of R&D inputs, rather than outputs. Within a country the difficulty of measuring R&D inputs is well known. The concept itself is bard to define; according to the OECD definition, the main criterion to distinguish R&D inputs from related scientific activities is "the presence in R&D of an appreciable element of novelty" (Frascati Manual (1981]). There are no adequate measures of R&D output.' Much of R&D output is in the form of newly improved conmnodities whose quality-adjusted price indices are rarelv available. A significant portion of R&D output is produced in industiies such as health, deferse, and space, where extemalities are large and are difficult to measure and capture statistically. To measure R&D inputs, we construct a stock variable and a flow variable: (i) Personnel: As a r%roxy for the accumulated stock of R&D capital, we use the number of scientists and engineers employed in R&D activities. (ii) Expenditures: As a proxy for the change in R&D stock, we use gross domestic expenditure on R&D. Following commnop --actice, we ignore the depreciation and lagged effects of past R&D investments.6 5There is no internationally agreed system of output measurement. TWO indicators, the number of patent ap;lications and the technical balance of payments, are available for OECD member countries for a limited samp!e peiiod. 6In the UNES ) yearbooks, these measures are broken down by specialization (natural science, social science, etc.), sector of performance (private industries, universities, the governments), source of funds (government funds, productive ernterprise and special funds, etc.), and type of R&D activity (fundamental research, applied research, etc.). 1The more specific the measure, the larger are the differences in definitions across countries and tle shorter are available sample periods. 3 Even for the basic R&D input proxies, some national specifications vary from the intemational norms set out in the Frascati Manual. Details of differences between national and in. rnational specifications are set out in footnotes of the UNESCO Statistical Yearbook and the supplementary tape. On the basis of individual footnotes', we exclude countries (or observations) if their specification is too different from the irternational norm. For example, we exclude the expenditure and manpower data for Greece and the manpower data for Norway since they do not include R&D activities in private productive sectors. We exclude Fiji, French Polynesia and Tonga since their data relate to one research institute only. At the same time, to avoid excluding too many countries, we permit some variation in the R&D input definitions. The most severe example of this is the manpower variable for several countries, including Japan, which is based on the number of physical persons, rather than the international standard of full time equivalent persons. The R&D personnel for Japan are as a result estimated to be overstated by about 30 percent, according to other studies.8 Our data set consists of manpower and expenditure data for 21 OECD countries and 19 developing countries, whose sample periods differ widely across countries. Table I shows the sample averages and the growth rates of our R&D input proxies for each :ountry. The first two columns list the average and the annual growth rates of R&D expenditures as a percentage of GDP between 1970 and 1985. The third column shows the average of the number of scientists and engineers (employed in R&D activities) per million population 'The proportion of footnoted observations is large, indicating the primitive state of intemnational standardization of R&D measurement. 'OECD, Proposed Standard Pi.ctice for Surveys of Research and Experimental Development: "Frascati Manual 1980", Paris, 1981. 4 between 1970 and 1975, and the fourth column shows annual growth rates between 1973 and 1985. The last two coiBinns are foi the number of scientists and engineers not normalized by population.' Following the OECD convention, Table 1 puts OECD countries into three groups (major, medium, and small R&D performner) on the basis of the amount of R&D expenditures. The groups based on R&D expenditures are not necessarily the same as the groups based on R&D expenditures as a percentage of CiDP. Table 2 breaks down R&D expenditures by source of funds and b,- type of R&D activities. The first two columns in Table 2 report the average ratios of government funded R&D to the sum of government aind privately funded R&D expenditurcs.'" The last two columns report the awerag. ratios of basic R&D expenditures to the sum of basic and applied (including experimental) R&D expenditures. Data for these refined categories are available for a smaller set of countries. Some characteristics emerge from Tables I and 2. First, R&D activities in developing countries are negligible. The average R&D expenditure as a percentage of GDP between 1970 aa,d 1985 is only 0.36 percent for developing countries, just 25 percent of the average for OECD countries of 1.42 percent. Between 1970 and 1975, the average number of scientists and engineers employed in R&D activities is 172 per million population for developing 91n calculating sample averages and annual growth rates, the number of observations included differs across countries, and we use our discretion to minimize bias due to the differences in sample periods. We tried to avoid the concentration of samples around certain periods; if the values for 1985 are not available, either 1984 or 1986 figures are used. 10The source of funds code in the UNESCO data set includes other categories such as foreign funds and other funds. Since their proportion is usually minimal and we are not sure whether they should be treated as part of privately-funded expenditures, we exclude them in calculating the ratios in Table 2. 5 countries, just 14 percent of the average for OECD countries of 1,248 per million population. Moreover, given that the developing countries in our sample are almost certainly those with more R&D activity than those not in the sample, the average R&D activity for all developing countries is almost certainly lower than in our sample. Second, manpower proxies indicate that both the average growth rates and the variance of R&D activities is higher in developing countries than in the OECD countr.es. Between 1970 and 1985, the average growth rates of the number of scientists and engineers per rnillion populaticn are 3.84 percent for 18 OECD countries and 6.04 percent for 11 developing countries. S..andard deviations are 1.97 and 4.86, respectively. During that period, the growth rates vf R&D expenditures as a percentage of GDP are not very different between OECD countries and developing countries. They are 2.27 percent for 18 OECD countries and 2.42 percent for 11 developing countries. The standard deviations are 1.96 and 5.29, respectively. Among developing countries, the average R&D expenditures as a percentage of GDP of India and Korea are similar in size to those for low income OECD countries such as Spain, Portugal, and Iceland. The growth rates of the former are greater. Thirdly, though high R&D investment is often regarded as a major factor explaining Japan's high economic growth, Table I indicates that Japan s R&D activities, in terms of R&D expenditures or the growth rates of manpower indices, do not seem exceptional by OECD standards. (Japan's manpower indices are overstimated as noted above.) Surprisingly, even Japan's ratio of applied R&D is not higher than the OECD average. Its ratio of government funded R&D is lower. 6 Though not reported in Table 1, the time series of our data shows that R&D expenditures have grown as rapidly a. or more rapidly than G!)P in the majority of countries; the R&D elasticity with resnect to GDP seermls to be greater than 1. For OECD countries, this series data also reveal that R&D expenditures started to accelerate from the late 1970s, after faltering in the middle of the decade. III. R&D and Economic Growth To analyze the relationship between R&D activities and economic growth, we use a simple cross-country regression model as in Dowrick and Nguyen [1989]: (1) qi =ao + alk + a2Ri + a3l1 - a4 InYi* + E, where qi, ki, RI, and II are the average annual growth rates of real GDP, capital stock, R&D capital stock, and employment, respectively, of country i. Y* denotes initial per capita CDP of country i relative to a technological leader, the United States, in our analysis. The termn is included to capture the convergence effect, including the likelihood of total factor productivity catch-up effects. The reduced form (1) can be easily derived by introducing R&D capital stock in Dowrick and Nguyen's [1989] production function. Equation (1) can be rewritten to incorporate the irnvestmrrnt-GDP ratio by adopting the cormmon practice of proxying capital stock growth by the average share of investment in output (I/Q)": "This approximation is exact if capital-output ratios are constant across countries and over time. With the same logic, if R&D capital-output ratios are constant, we can proxy R in equation (2) by the average R&D expenditures as a percentage of GDP. 7 (2) q! = ao + a,(I/Q); + a2R, + a31i - a4 InY1j* + e; Table 3 presents the OLS regression results for equation (2). Definitions and sources of variables are summarized in the appendix. For R1, we use three measures: the growth rates of the number of scientists and engineers employed in R&D activities per milllon pcpulation (GSCP708S), the growtil rates of the number of scientists and engineers not denominated by population (GSCi7085), and the average R&D expenditures as a pe ,entage of GDP (RNDi_Y). Since our R&D proxies are available only after 1970, the data are the sample averages or the growth rates between 1970 and 1985. LNDIFF70 is the log of the gap in GDP per capita in the year 1970, The first three columns of Table 3 report the regression results for the full sample. The next four columns report the results for OECD countries and developing countries. Consistent with the results in previous studies (Abramovitz (19861, Baumol and Wolff 11988], Barro [1991]), we find that high investment rates are strongly associated with high GDP growth, and there is evidence for catch-up in levels of total factor productivity. As for the growth-R&D nexus, our results are mixed. The average growth rates of the number of scientists and engineers, whether denominated by population or not (GSCP7085, GSC17085), are significant in explaining GDP growth rates for the full sample and O2CD countries, but not for the developing countries. I' e take this evidence at face value, it suggests that R&D activities become important only after a country reaches a certain degree of economic development. 8 However, the average R&D expenditure as a percentage of GDP (RND _Y) is insignificant in explaining economic growth even for OECD countries. Breaking R&D expenditures into sub-fields does not alter the rcsult. table 4 eeports regression results when government funded and privately funded expenditures enter the regressions separately.12 They do not provide evidence for a significant return from privately funded or government funded R&D .Ilestment. How can our mixed result be best interpreted? One might argue that the expenditure variable is a better proxy since manpower irr ices could be proxying the changes in human capital instead 3f R&ID capital. To check the validity of this argument, we include proxies for human capital accumulation used in Barro [1991], the primary and the secondary school enrollment rates, in the equation (2). The second column in Table 4 shows that both the secondary enrollnient rates and GSCP7085 are significant in explaining GDP growth.'3 The magnitude and significance of the coefficient of GSCP7085 do not change after the inclusion of the enrollment rates. Our measure of the number of scientists and engineers employed on R&D activities is thus probably not simply playing the role of (less educated) human capital. Of course our proxies may represent highly educated human capital, but it would be hard to distinguish the inputs of highly educated human capital from R&D capital anyway. Moreover, there is a positive case for choosing GSCP7085 or GSC17085 over RNDIY. '2Because of the limited number of non-OECD country observations, we report the results for OECD countries only. Neither are applied and fundamental R&D investments significant in growth equations, which is not shown since it covers only 9 countries. '3Differently from Barro [1991], the fact that the primary school enrollment rates are insignificant in our regression is because our samnples are mostly OECD countries and the sample periods of the regressions is from 1970. 9 As mentioned in Section 11, since our selection criterion ziows a certain degree of arbitrariness in national specifications of R&D measurement, international comparison of the levels of our proxies could be misleading. However, if national specifications have been time- invariant, comparing the growth rates of our proxies would be less problematic. This argument suggests that GSCP7085 or GSC17085 are better proxies than RNDIyY."4 At the same time, the results in Table 3 cannot be interpreted as demonstrating a causal relationship from R&D activities to economic growth. The positive partial correlation between GSCP7085 and economic growth may be due to simultaneous equation bias: Growth of R&D capital stock could simply follow economic growth. In order to control for this possible endogeneity, we estimated the relationship between initial R&D activities at the beginning of the sample period and subsequent economic growth, shown in Table 5. Following Barro [1991] and other cross country studies, we regressed the growth rates of real GDP per capita between 1970 and 1985 on initial level of real GDP (GDP70), initial primary and secondary school enrollment rates (PRIM, SEC), and the ratio of government spending to GDP (GOV), together with a proxy for initial R&D capital stock,'5 the average number of scientists and engineers per million population between 1970 and 1975 (SCI_P). We use the average value between 1970 and 1975 because some of our sample countries do '"We think this problem is more than a measurement error problem. First, international comparability of RND 1- Y is less problematic among OECD countries, but we still have the same problem for OECD countries only. Second, the partial correlation between RND I -Y and GSCP7085 (GSC17085) is negative. The correlation coefficient between them are -0.48, -0.38, and --0.46 for the full sample, for OECD countries, and for developing countries, respectively. It suggests that both variables cannot be used as a proxy for the change of R&D capital stock. Section IV further discusses this negative finding. 15As shown in Mankiw, Romer, and Weil [19921, this specification can represent the transitional dynamics of a neoclassical growth model with human and R&D capital. 10 not have the data for the year 1970. Using average values has the additional advantage of avoiding a systematic bias in estimating the dependent variable and the initial values of independent variables from a common base year. The resuit in column 1, Table 5 shows that SCI_P is not significant in explaining the per capita GDP growth rat-. In order to check whether this result is due to a different specification from those in Table 3, the second column includes the average investment-GDP ratio and GSCP7085 in the regression. We confirm the previous result that the average investment ratio and GSCP7085 are significant explanatory variables, but still the initial R&D stock, SCI P, is insignificant. The third column reports the same conclusion for the OECD countries only.'6 The last three columns in Table 5 test parameter stability over time. Given a possible lag in the effect of R&D on economic growth, the fourth column regresses per capita GDP growth rates between 1975 and 1985 on the same set of independent variables, whose initial values are for the year 1970. Also, from previous cross country studies of growth, we know that the fit of the regression is much better if the sample periods are extended back to the 1960s. The fifth regression extends the samnple period; the dependent variable is the GDP growth rate for the period 1960 to 1985. and the initial values for independent variables (except SCI_P) are the flgures in the year 1960. Note that the estimated coefficient of SCI_P '6Barro and Lee [1993] report a similar finding for the data set including over 100 countries. They find that the initial values of educational attainment at higher (rather than secondary) level are insignificant for economic growth, even for OECD countries, and conclude that "human capital in the formn of higher education would be unimportant for most countries, which tend to adopt leading technologies rather than invent fundamentally new things" [p. 14-15]. The difference between their highly educated human capital stock variable and ours is that ours includes only those employed in R&D activities. 11 sould be biased upward since SCI_P is the average value in the middle of the sample period. In both cases, however, SCI P is not correlated with economic growth. The results for OECD countries and developing countries are not different (not reported in Table 5).17 The finding that our proxies for initial R&D activities are not correlated with economic growth contrasts sharply with Romer [1989] and Lichtenberg [1992] who use the same UNESCO data set. Romer [1989] finds a positive correlation between the initial number of scientists and engineers employed in R&D activities and subsequent economic growth. Lichtenberg [1992] finds that the ratio of privately funded R&D expenditures to GNP is highly correlated with economic growth in a sample of 59 countries, though government funded R&D expenditures are not. We believe their results should be treated cautiously because of probable data problems and possible resulting selection bias. Lichtenberg [1992] includes many developing countries that have just one annual observation on the R&D expenditure/GNP ratio, and uses the one observation as an average for the whole sample period. Available data for other countries are often too concentrated around certain periods to represent an average value for the period 1970 and 1985. Both authors include countries whose specifications diverge considerably from the international norms explained in Section II. The combination of results from Tables 3 and 5 suggests that the positive '7Our findings in Table S are not sensitive to White's heteroskedastic error correction, inclusion of other proxies for political stability and market distortions used in Barro [1991], and the choice of R&D proxies. Using the number of scientists and engineers (not denominated by population) or average R&D expenditure in percentage of GDP between 1970 and 1975, without a good theoretical reason, as a proxy for initial R&D capital does not alter the conclusions. 12 contemporaneous correlation between our proxies for R&D activities and economic growth (Table 3) is capturing the positive effect of economic growth on R&D activities -- tne reverse of the expected causation. Higher initial R&D activities do not necessarily lead to subsequent economic growth (Table 5). Is this result contradictory !o numerous micro-studies, which document high returns to R&D expenditures at the firm or industry level? (Griliches [1984], Jaffe [1986], Mansfield [1977]) At face value, our results suggest that high private returns to R&D expenditures in micro-studies do not reflect high social returns. They may result from enhanced monopoly power of a firm at the cost of its competitors. Alternatively, it may be that evidence of high returns fromr, micro studies of firms and industries in highly developed countries, mostly the United States, does reflect social gains, but such gains to R&D activities become important only after a country or a firm advances to a technologically leading group."8 IV. Determinants of R&D Expenditures This section analyzes what variables are important in explaining international differences in R&D expenditures as a percentage of GDP. The issue itself is interesting. In addition, this analysis provides an indirect test of the quality of our data. If large measurement errors or differences in national specifications are a major reason for our previous finding, it is less likely that we will find meaningful determinants of our R&D proxies. 18The other explanation may be the poor quality of the R&D proxies. International standardization in measuring R&D activities is still in the experimental stage. The UNESCO data, even with adjustments, may not provide an adequate basis for study of the effects of R&D on growth compared, for example, to other data, such as those on education enrollments, that routinely used in studies of growth. 13 In Table 6, the dependent variable in the first two regressions is average R&D expenditures as a percentage of GDP between 1970 and 1985 (RNDI_Y). The results show that the initial level of per capita GDP (GDP70) and the initial number of scientists and engineer in R&D activities per million population (SCI_P) are major factors explaining cross country differentials in R&D expenditures. We find that GDP70 or SCI_P alone can explain 61 percent or 75 percent of the cross country variation. It is interesting to note that the average share of physical investment in GDP is not a significant factor once the initial level of GDP is controlled, though it is significant without GDP70. Other variables such as export share, government expenditure share, and proxies for initial human capital are not significant in explaining the variation of RN'D1_Y, and their inclusion does not alter the results. The results for the OECD countries are similar. The finding that the R&D expenditure share is strongly related to the level of income and the level of initial R&D stock is not surprising. A country with higher per capita GDP (larger physical capital stock) should have a lower marginal productivity of physical investment, and therefore, a larger incentive to invest in R&D, which can alleviate the diminishing marginal productivity of physical capital. In contrast, the positive effect of initial R&D stock suggests that the marginal productivity of R&D capital is not diminishing, i.e., there are constant or increasing returns to R&D investment. This is consistent with the view that huge fixed costs are involved in R&D investment. The third and the fourth columns report the results when privately funded and government funded R&D expenditures to GDP ratios are used as dependent variables, respectively. They show qualitatively the same results with those for total R&D expenditures. The finding is not 14 surprising, considering the high correlation coefficient (0.70) between privately funded and government funded R&D expenditures to GDP ratios.' The last two colurrns in Table 6 show typical regression results when the dependent variable is the growth rates of the number of scientists and engineers per million population (GSCP7085). As expected from our previous regressions in Table 3 and 5, the per capita GDP growth rate is a significant determninant of GSCP7085. In fact, we find other variables we used in this paper are not robustly related to GSCP7085 using Levine and Renelt's [1992] approach to assessing robustness. Note that GSCP7085 is negatively related to the initial level of income, GDP70, which is a major determinant of R&D expenditures. Ex post, this finding illustrates the reason for the negative correlation between our manpower and expenditure proxies for R&D activities (see footnote 14). We believe that R&D activities are income- elastic and led by high income countries, consistent with the results in columns 1-4. Therefore, at least for the OECD countries, expenditure data appears to provide a better proxy for growth in R&D activities than manpower data. At the same time, the finding in columns 5 and 6 confirms our earlier supposition of reverse causality, i.e., growth of income induces growth of R&D manpower indices in all countries. VI. Conclusions Using the UNESCO data for R&D expenditures and personnel, this paper documents international differences in R&D activities and assesses empirically the link between R&D and economic growth, and the deterninants of R&D differentials. Within our group of OECD 'The correlation coefficient between fundamental R&D and applied R&D expenditures to GDP rations is 0.89. 15 countries, R&D activity and economic growth are correlated for one of our two proxies. Contrary to the findings of Romer [1989] and Lichtenberg [1992], however, they are not correlated across all, including developing countries. Moreover, even for OECD countries, it is likely that economic growth affects R&D activity rather than vice versa. First, there is no evidence for OECD countries that R&D activity at the beginning of the period 1970-85 contributed to subsequent growth during the period. Second, our analysis of the determinants of R&D activities shows a robust link between these activities and the level of income, suggesting that R&D activities become important only after a country reaches a certain stage of development. For developing countries, our results are consistent with the widespread view, first proposed by Gerschenkron, that countries that are behind grow by catching up technologically, not by advancing the technological frontier. 16 Appendix: Definitions of Variables in Table 1-6 AVG EX70: Average ratio of exports to GDP for the years 1960-85; from World Bank BESD. AVG17085: Average ratio of public plus private investment over GDP for the years 1970-85; from Barro [1991]. GOV7085: average ratio of real government "consumption" expenditure to real GDP for the years 1960-85; from Barro [1991] GRNDI_Y: government funded expenditures on R&D as a percentage of GDP; average for the years 1970-85; calculated from UNESCO data set. GR7085: growth rates of GDP per capita, in 198W) $US, for the years 1970-85; from World Bank BESD. GrP70: GDP per capita in 1980 $US, for the year 1970; from World Bank BESD. GPOP7085: growth rates of population for the years 1970-85; from World Bank BESD. GRSCP7085: growth rates of the number of R&D scientists and engineers per million population for the years 1970-85; calculated from UNESCO data set. GRSC17085: growth rates of the number of R&D scientists and engineers for the years 1970-85; calculated from UNESCO data set. LNDIFF70: log difference between GDP70 of each country relative to U.S. GDP70. PRND1_Y: privately funded expenditures on R&D as a percentage of GDP; average for the years 1970-85; 'calculated from UNESCO data set. PRIM: primary school enrollment rate; from Barro [1991] RND_Y: Expenditure on R&D as a percent of GDP; average during the period 1970- 75; calculated from UNESCO data set. RND1 Y: same as above for the years 1970-85. SCI_P: average number of R&D scientists and engineers per million population for the years 1970-75; calculated from UNESCO data set. SEC: secondary school enrollment rate; from Barro [1991] TGDP7085: growth rates of GDP, in 1980 $US, for the years 1970-85; from World Bank BESD. 17 References Abramovitz, Moses, "Catching Up, Forging Ahead, and Falling Behind," Journal of Economic History, XLVI (June 1986), 385-406 Barro, Robert, " Economic Growth in a Cross Section of Countries," Quarterly Journal of Economics, CVI (1991), 40743 Barro, Robert and Jongwha Lee, "Sources of Economic Growth," miineo, Harvard University, April 1993. Baumol, William and Edward Wolff, "Productivity, Growth, Convergence, and Welfare: Reply," American Economic Review 78, December 1988, 1155-59. Dowrick, Steven, and Duc-Tho Nguyen, "OECD Comparative Growth," American Economic Review, LXXIX (December 1989), 1010-30. Easterly, William, "Explaining Miracles: Growth Regressions Meet the Gang of Four," presented at the NBER's Fourth Annual East Asian Seminar on Economics, San Francisco, June 17-19, 1993. Griliches, Zvi, R&D, Patents and Productivity, Chicago: University of Chicago Press, 1984. Jaffe, Adam B., "Technological Opportunity and Spillovers of R&D: Evidence from Firms' Patent, Profits, and Market Value," American Economic Review 76, (December 1986), 984-1001. Kim, Jong-II, and Lawrence Lau, "Sources of Economic Growth of the Newly- Industrialized Countries on the Pacific Rim," presented at the "Conference on the Economic Development of the Republic of China and the Pacific Rim in 1990 and Beyond," Taiwan, China, 1993. Levine, Ross, and David Renelt, "A Sensitivity Analysis of Cross-Country Growth Regressions," American Economic Review 1(4) (September, 1992), 942-963. Lichtenberg, Frank R., "R&D Investment and International Productivity Differences," NBER Working Paper #4161, September 1992. Mankiw, Gregory, David Romer, ani David Weil, "A Contribution to the Empirics of Economic Growth," Quarterly Journal of Economics 107, (May 1992), 407-38. Mansfield, Edwin, The Production and Application of New Industrial Technology, New York: N.W. Norton, 1977. OECD, Proposed Standard Practice for Surveys of Research and Experimental Development: "Frascati Manual 1980", Paris, 1981. 1o Page, John and Howard Pack, "Accumulation, Exports, and Growth in the High Performing Asian Economies," presented at the Carnegie-Rochester Conference on Public Policy, 1993. Romer, Paul, "What Determines the Rate of Growth and 1'echnological Change?," working paper #279, The World Bank, September 1989. UNESCO. *Statistical Yearbook, Paris, 1963-present. 19 TABLE : R&D INPUTS [1] Expenditure in [2] Researchers per 3] Number of 0 of GDP mill. population Researcheru (1) avg(%) (2)growth() (3)avg (4)growth () (5)Avg (6)growth (1970-85) (1970-85) (1970-75) (1970-85) (1970-75) (1970-85) OECD Countries Major R&D Performer U.S. 2.44 0.79 2523.6 1.30 531516 2.27 Japan 2.16 2.70 3230.9 3.06 349323 4.04 Germany 2.29 2.32 1550.2 4.96 95379 3.73 France 1.89 1.33 1207.5 3.14 62678 3.69 U.K. 2.15 0.71 1400.3 n.a. 78193 n.a. Medium R&D Performer Italy 0.84 1.85 605.3 4.45 33054 4.83 Canada 1.15 0.66 1006.1 3.00 22073 4.11 1.97 0.07 1730.1 1.92 23117 2.62 Netherlands Sweden 2.01 4.89 1497.9 5.24 12207 5.46 _.28 2.74 1481.5 3.64 9496 3.76 Switzerland Australia 1.05 -0.66 1838.6 0.05 24600 1.43 Belgium 1.48 1.1S 1328.9 1.55 12925 1.81 Small R&D performer Austria 0.99 4.93 617.4 4.67 4640 4.78 Norway 1.33 2.57 n.a. n.a. n.a. n.a. Denmark 1.05 1.75 959.8 3.97 4804 4.21 Finlend 1.08 3.96 1247.6 6,30 5828 6.72 New n.a. n.a. 852.7 n.a. 2549 n.a. Zealand Ireland 0.75 0.41 687.2 3.75 2129 5.00 Spain 0.35 6.13 219.3 5.93 7570 6.80 Portugal 0.35 1.16 223.7 3.93 2050 4.77 Iceland 0.61 5.94 756.8 8.29 161 9.34 LDCs Argencina 0.38 3.23 289.7 1.84 7300 3.38 Ecuador n.a. n.a. 92.1 n.a. 569 n.a. Egypt n.a. n.a. 299.4 2.62 10665 5.17 Ghana n.a. n.a. 359.9 n.a. 3260 n.a. Guatemala 0.24 8.78 47.6 n.a. 269 n.a. India 0.51 5.95 76.2 7.64 44412 7.50 Iran n.a. n.a. 128.9 -3.01 3829 0.40 Jordan 00.32 -3.64 78.9 3.95 207 6.65 20 tl] Expenditure in (2] Researchers per (33 Number of % of GDP mill. population Researchers (1) avg(%) (2)growth(%) (3)avg (4)growth(V) 'S)Avg W6growth (1970-85) (1970-85) (1970-75) (1970-85) (1970-75) (1970-85) Korea 0.57 8.79 192.4 11.74 6533 13.31 Mauritius 0.40 -6.53 127.7 6.86 107 8.20 Mexico 0.32 6.33 98.8 7.70 5417 10.67 Nigeria n.a. n.a. 35.0 n.a. 2066 n.a. Peru 0.34 -3.07 131.4 11.00 188S 13.52 0.19 0.23 n.a. n.a. n.t. n.a. Philippines Singapore n.a. n.a. 228.2 13.01 502 14.38 Sri Lanka 0.17 0.34 n.a. n.a. n.a. n.a. Sudan n.a. n.a. 153.3 n.a. 2311 n.a. Trinidad & 0.53 6.21 n.a. n.a. n.a. n.a. Tobacco Uruguay n.a. n.a. 419.7 3.09 1150 2.74 Total Avg 1.04 2.32 770.13 4.68 38188 5.70 OECD Avg 1.41 2.27 1248.27 3.84 64214 4.41 LDC Avg 0.36 2.42 172.45 6.04 5655 7.81 21 TABLE 3: R&D and Economic Growth (1) Dependent variable: TGDP7085 Indep (1) (2) (3) (4) (5) (6) (7) variables ALL ALL ALL OECD OECD LDCS LDCS Constant -0.022 -0.022 -0.001 -0.0007 0.011 -0.054 -0.030 (-2.14) (-2.16) (-0412) (-0.103) (1.307) (-1.80) (-0.53) AVG17085 0.153 0,148 0.133 0.052 0.075 0.236 0.161 (3.84) (3.87) (3.22) (1.99) (2.62) (2.57) (0.92) GPOP7085 0.461 0.273 0.126 1.343 0.846 0.487 0.175 (1.36) (0.87) (0.33) (4.21) (2.35) (0.64) (0.14) LNDIFF70 -0.008 -0.008 -0.007 -0.004 -0.0022 -0.015 -0.010 (-2.46) (-2.66) (-2.46) (-1.62) (0.55) (-1.96) (-1.39) GSCP7085 0.')016 0.0017 0.0012 (2.C7) (2.41) (0.79) GSC17085 0.0018 (2.49) RND I Y -0.004 -0.0049 0.0414 _ 0.67 0.69 (-0.947 (-1.76) (0.66) R2 0.67 0.69 0.47 0.(5 0.66 0.74 0.43 D.W. 1.56 1.60 1.34 1.87 2.09 1.65 1.18 of obs. 28 28 30 18 |_20 10 10 Figures in parenthesis are t-statistics. For the definitions of variables, see appendix. 22 i.able 4: R&D and Economic Growth Dependent variable: TGDP7085 Independent Variab es (1) (2) l__________________ OECD OECD Constant 0.00 77 -0.0091 (0.965) (-0.646) A VG17085 0.0792 0.0433 (2.982) (1.743) GPOP7085 0.5783 1.2389 (1. 703) (4.056) LNDIFF70 -0.0039 -0.0063 l ________________________ (-1.083) (-2.22) Gov't funded R&D -0.0008 (-0.142) Privately Funded R&D 0.0034 _____ ____ _____ ____ ____ _ ___ ____ ____ (-0. 740) GSCP7085 0.0197 (2.93) SEC 0.0197 (1.97) PRIM -0.004 (-0.367) R2 0.0683 0.81Q D. W 1.63 1.305 L of observation 18 18 23 Table 5: Growth 4 Initial R&D Capital Dependent (1) (2) (3) (4) (5) Variable/ Independent GR7085 GR7085 GR70S5 GR7585 GR6085 variable All All OECD All All Constant -O 0093 -0.0241 0.0147 -0.026 -0.0049 (-0.46) (-0.80) (1.020) (-1.060) (-0.413) Initial GDP -0.0048 -0.002 -0 .0021 -0.0059 o .0074 ___ _ ______ _ (-2.26) (1.54) (-2.61) (-2.26) (->.967) SEC 0.0544 0.032 0,0215 0.056 0.0424 (1.992) (1.44) (1.88) (1.66) (;.133) PRIM 0.0305 -0.0065 0.0093 0.0373 0.0447 (1.429) (-0.272) (0.847) (1.438) (3.898) GOV -0.0467 0.0310 -0.0299 -0.0022 -0.0430 (-0.759) (0.506) (-1.211) (-0.029) (-0.935) SCI_P -0.000 0.000 0.000 0.000 0.000 (-0.200) (0.034) (0.202) (0.405) (0.254) AVGI7085 0. 1257 (2. 496) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ GSCP7085 0.0027 (2.810) R_ _ 0.363 0.601 0.485 0.327 0.530 D.W. 1.48 1.29 2.02 1.72 1.67 #of obs. 36 29 20 36 36 2 4 Table 6: Determinants of R&D Expenditures Dependent (1) (2) (3) (4) (5) (6) RND1 Y RND1_Y PRND1_Y GRN !. Y GSZP7085 GSCP'085 variab/e/ Alif ALL ALL .. ALL ALL Independent Variab,'e Constant -0.439 0.145 -0. 755 -0.014 10. 36 8.871- (-0.88) (0.36) (-1.42) > 0.04) (3.00) (2.3.) GDP70 0.228 0.097 0. 129 0.120 -0.335 -0.105 (5.31) (2.26) (3.06) (4.49) (-1.33) (-0.31) GR7085 8.625 3.908 8.152 i.695 94.52 101.71 (1.08) (0.62) (1.12) (0.36) (2.35j (2.50) AVGI7085 0.163 -2.545 1.301 -0.624 -14.72 -8.982 (0.C8) (-1.65) (0.77) (-0.5c' (-1 24) (-0.68! AVG EX70 -0.854 0.748 0.104 0.243 1.766 -0.039 (-1.03) (0.92) (O.10) (0.39) (0.29) (0.00C GOV7085 2.007 1.434 0.529 0.132 -199.64 -18.028 (1.00) (0.95) (0.24) (O 09) (-1 63) (-1.49) SCI_P 0.0006 -0,001 (4.72) (-1.02, R2 0.65 0.84 0.48 0.63 0.3 _ 0.42 D.W. 1.92 2.00 2.08 1.60 1.85 1.84 # of obs. 29 25 21 21 26 26 25 Policy Research Working Paper Series Contact Thtle Author Date for paper WPS1208 Primary School Achievement in Levi M. Nyagura October 1993 I. Conachy English and Mathematics in Abby Riddell 33669 Zimbabwe: A Mufti-Level Analysis WPS1209 Should East Asia Go Regional? Arvind Panagariya October 1993 D. Ballantyne No, No, and Maybe 37947 WPS1210 The Taxation of Natural Resources: Robin Boadway October 1993 C. Jones Principles and Policy Issues Frank Flatters 37699 WPS1211 Savings-Investment Correlations Nlandu Mamingi October 1993 R. Vo and Capital Mobility in Developing Countries 31047 WPS1212 The Links between Eoonomic Policy Ravi Kanbur October 1993 P. Attipoe and Research: Three Examples from 526-3003 Ghana and Some General Thoughts WPS1213 Japanese Foreign Direct Investmern: Kwang W. Jun November 1993 S. King-Watson Recent Trends, Determinants, and Frank Sader 33730 Prospects Haruo Horaguchi Hyuntai Kwak WPS1214 Trade, Aid, and Investment in Sub- Ishrat Husain November 1993 M. Youssef Saharan Africa 34637 WPS1215 How Much Do Distortions Affect William Easterly November 1993 R. Martin Growth? 39065 WPS1216 Regulation, Instautions, and Alice Hill November 1993 D. Evans Commitment: Privatization and Manuel Angel Abdala 38526 Regulation in the Argentine Telecommunications Sector WPS1217 Unitary versus Collective Models Pierre-Andre Chiappori November 1993 P. Attipoe of the Household: Time to Shift the Lawrence Haddad 526-3002 Burden of Proof? John Hoddinott Ravi Kanbur WPS1218 Implementation of Trade Reform in John Nash November 1993 D. Ballantyne Sub-Saharan Africa: How Much Heat 37947 and How Much Light? WPS1219 Decentralizing Water Resource K. William Easter November 1993 M. Wu Management: Economic Incentives, Robert R. Hearne 30480 Accountability, and Assurance WPS1220 Developing Countries and the Bernard Hoekman November 1993 L. O'Connor Uruguay Round: Negotiations on 37009 Services Policy Research Working Paper Series Contact Title Author Date for paper WPS1221 Does Research and Development Nancy Birdsall November 1993 S. Rajan Contribute to Economic Growth Changyong Rhee 33747 in Developing Countries? WPS1222 Trade Reform in Ten Sub-Saharan Faezeh Foroutan November 1993 S. Fallon Countries: Achievements and Failures 38009