Do Arms Exports Stimulate Economic Growth?

Do Arms Exports Stimulate Economic Growth? Pavel Yakovlev Department of Economics College of Business and Economics West Virginia University Morgantown, WV 26505 Pavel.Yakovlev@mail.wvu.edu Draft Date: October 15, 2004 ABSTRACT Previous empirical research has yielded mixed results regarding how defense spending affects economic growth. Because defense spending can simultaneously affect growth through several channels and in opposite directions, I would argue that one should not expect to see a consistent overall relationship between these two variables. More conclusive evidence may come from estimating how defense spending affects growth through different channels. In an attempt to examine one of these channels, I use balanced panel data to estimate the relationship between economic growth and net arms exports for 62 countries (and some sub-samples) from 1990 to 1999. Using different econometric techniques, I find that net arms exports have a significant positive effect on economic growth in the entire 62 country sample, non-oil country sub-sample, and OECD sub-sample. Keywords: Growth, Defense, Military, Arms, Trade, Exports, Imports JEL Classification: O30, O38, H50, H56

Yakovlev 1 Do Arms Exports Stimulate Economic Growth? Escalating terrorist threats, regional conflicts and growing instability around the world in the past few years have called for greater domestic security spending and defense outlays in many developed as well as developing countries. Defense spending as a share of GDP can amount to double digits in countries like Israel and Jordan. Hence, the implications of defense spending for economic growth and development are far from trivial. Despite extensive empirical research on this topic in the last two decades, the overall empirical relationship between growth and defense spending remains vague. Because defense spending can simultaneously affect growth in opposite directions through several channels, I argue that one should not expect to see a consistent overall relationship between growth and defense spending. Perhaps a different approach that examines how defense spending may affect growth through various channels is warranted. I argue that one of these channels is arms exports. Although defense spending diverts valuable resources from other productive uses and puts a tax burden on a country s citizens, it may also create externalities and other growth incentives. What if years of strong investment in military R&D and capital can make a country into one of the leading arms and military technology exporters? Then, do arms exports reflect a de-facto return for providing national defense? Or do arms exports result in exporting a part of the tax burden associated with military R&D financing? If they do, then one should expect a positive relationship between economic growth and net arms exports. I find that net arms exports have a significant positive effect on per capita GDP in 62 country sample, non-oil country sub-sample, and OECD sub-sample from 1990 to1999. I. Literature Review 1

Yakovlev 2 Economic growth and defense spending relationship is a controversial research topic with serious political and socio-economic implications. Since Benoit s (1973) early influential work suggesting growth enhancing effect of military spending, many researchers have analyzed the empirical relationship between economic growth and defense spending, but found conflicting and inconsistent results. The advocates of military spending argue that military spending positively affects economic growth by stimulating aggregate demand, some infrastructure development, and technological externalities (spillovers) from military research and development (R&D). For instance, Benoit (1973, 1978) argues that an increase in a defense component of aggregate demand may increase the utilization of capital stock (especially in developing countries), which may lead to increased purchasing power, profit rate, and finally investment that, in turn, will generate both short run multiplier effects as well as higher long-term rates of economic growth. Moreover, military R&D expenditures may yield technological advances that could enhance productivity and growth in the civilian sector of the economy. For example, radars and jet engines developed by the military have found useful and productive economic applications as many other original military inventions. In addition, military training and education may increase the level of human capital that can later be used in the civilian economy for example, retired military working in the civilian airlines. In fact, military training in developing countries, especially, might increase the accumulation of human capital. Generally speaking, military expenditures may provide a variety of positive externalities. Hall (1988) finds that the growth of military spending stimulates services but retards non durable goods and construction. Hall observes that the covariance of military spending growth and the Solow residuals is positive and that military spending increases measured productivity in services and lowers it for non 2

Yakovlev 3 durable goods and construction. Fredericksen and Looney (1983) discoveres that defense spending helps economic growth in resource-rich LDCs but not in resource constrained LDCs from 1960 to 1978, while Kennedy (1974) and Whynes (1979) using cross-country data estimate that defense spending has a positive effect on economic growth. The opponents of defense spending argue that these expenditures have a negative impact on growth because they result in a reallocation of resources from the more productive market ventures to the less productive military ventures financed with taxes, which create welfare losses and reduce labor supply. Others argue that higher defense outlays come at the expense of lower savings, which decrease investment and output growth. Moreover, increasing defense budgets come at the expense of other budgets such as education and health, which reflect investment in human capital a significant factor in long term economic growth as Mankiw, Romer, and Weil (1992) (hereafter MRW) have shown. Speaking of human capital, the military R&D sector employs a large number of skilled workers and talented scientists who could make productive contributions elsewhere in the market economy this represents the opportunity cost of employing human capital in the military sector at the expense of the civilian market economy. In addition, Heo (1998) points out that the military sector owns a large portion of capital goods, which can be used for a more productive investment. Furthermore, concludes Heo, since future productive capacities depend on saving and investment, an increase in defense spending can decrease planned investment and lower future output. Moreover, Perkins et. al. (2001) doubt that military training can significantly increase human capital because many of the skills learned in the military may not easily apply in the civilian economy. Perkins et. al. (2001) also question question 3

Yakovlev 4 whether or not military spending can significantly increase domestic aggregate demand because of the military s high import component, especially in developing nations. Finally, not all of the technology developed by the military sector can be applicable (or spilled over) to the civilian market economy or at least not in a timely manner due to military s secrecy. Despite some recent leaning towards viewing the growth-defense relationship as negative, the overall literature provides strong empirical support for either of the two arguments. For instance, Maizels and Nissanke (1986) discover a significantly negative relationship between economic growth and defense spending. On the other hand, Chowdhury (1991) shows that defense spending increases economic growth. Fredericksen and Looney (1983) and Deger and Smith (1983) report a growth retarding effect of defense spending in at least a majority of the developing countries in their sample. In addition, Deger and Sen (1983), Leontief and Dutchin (1983), and Faini, Annez, and Taylor (1984) reject the claim that defense spending stimulates economic growth, while Lim (1983) concludes that the negative impact of defense spending on economic growth is more pronounced among the poorer countries in Africa. In their summary of the defensegrowth literature, Perkins et. al. (2001) conclude that military expenditures are likely, after all, to have a negative overall effect on economic growth. They argue that military expenditures drain the Third World economies of scarce resources needed to finance long-term development efforts. Meanwhile, Chan (1995) concludes that the links between defense spending and economic growth are not yet fully understood. 4

Yakovlev 5 II. Theoretical Model Because the overall empirical relationship between growth and defense spending remains inconclusive, I suggest estimating the relationship between growth and various components of defense spending instead. I choose to look at one significant aspect of defense spending around the world arms trade. Generally, developed countries tend to be the net exporters of weapons systems and military technology, while developing countries tend to be the arms recipients. 1 I attribute this tendency to the high capital and technological requirements necessary for maintaining advanced defense industries, which only developed countries can afford. So how could arms exports benefit the exporting country? I hypothesize that arms exports can be associated with several positive effects on growth. First, arms exports can proxy for the level of technology in the exporting country and should, therefore, correlate with the Solow residual. Second, it has been noted that the capital intensive defense industry may be characterized as a decreasing cost industry. Hence, an export led increase in the output of a decreasing cost industry will have a positive productivity effect on the economy. Third, the revenue from the arms exports can stimulate consumption and investment in the exporting country. While it may not be possible to separate all the suggested effects of arms exports empirically, they should have a positive total effect on growth. To test for this total effect, I add the net arms exports variable (in an ad-hoc way) to the augmented Solow growth model. In 1956, Robert Solow developed a model that revolutionized the study of economic growth. He assumed an economy with a standard Cobb-Douglass production function with decreasing marginal returns to capital and fixed level of 1 See Figure 1 on arms exports and imports in developed countries in appendix. 5

Yakovlev 6 technology. The textbook Solow growth model treats the rate of saving, population growth, and technological progress as given exogenously. However, MRW improved upon the original Solow growth model. In their influential paper, MRW augment the standard Solow (1956) growth model with human capital and show that it can explain as much as eighty percent of cross-country variation in output per worker and can approximately predict cross-country convergence in the standards of living. The empirical success of MRW s model encourages its application for estimating the cross-country relationship between growth and net arms exports. MRW extend the textbook Solow growth model by augmenting the Cobb- Douglas production function with human capital in addition to the traditional factors such as physical capital and labor. The production function takes the Harrod-neutral form: (1) Y(t) = K(t) α H(t) β (A(t)L(t)) 1-α-β α + β < 1 Where Y(t) is output, K(t) is the stock of physical capital, H(t) is the stock of human capital, L(t) is the stock of labor, and A(t) is the effectiveness of labor. The effective units of labor, A(t)L(t), are assumed to grow exogenously at rate (n + g), where n is population growth and g is knowledge growth. The steady-state capital to labor ratio is related positively to the rate of saving and negatively to the rate of population growth. The α + β < 1 assumption allows for decreasing returns in production and income convergence. A constant share of output is assumed to be reinvested back into production. Then, MRW make the necessary assumptions and derivations, and propose the cross-country empirical model that relates output per capita to the factors of production: (2) ln(y/l) = A + α / (1-α-β) ln(i/y) - α+β / (1-α-β) ln(n + g + δ) + β / (1-α-β) ln(school) + ε 6

Yakovlev 7 Where Y/L is GDP per working-age person, I/Y is investment in physical capital as a share of GDP, (n + g + δ) is population growth plus knowledge growth plus depreciation rate δ (where g + δ = 0.05 by an educated guess), SCHOOL is the percentage of working-age population that is in secondary school. The labor participation rates are assumed constant. The assumption that the factors of production receive their marginal products allows to hypothesize that the coefficients for I/Y and SCHOOL each equal to 1, and the coefficient for (n + g + δ) equals to -2. Thus, all coefficients sum up to zero. This empirical model predicts that investment (or saving) and human capital increase the output per worker, while population growth decreases the output per worker. I adjust the empirical model in equation (2) for panel data and add to it in an ad-hoc way the net arms exports variable. Then, the empirical model becomes: (3) ln(y/l) it = A + bn it + α / (1-α-β) ln(i/y) it - α+β / (1-α-β) ln(n + g + δ) it + β / (1-α-β) ln(school) it + ε it Where N it is the net arms exports and the other variables are the same as before. I hypothesize that the net arms exports should have a positive coefficient for the reasons addressed previously. N it is not being expressed in the log form because many countries in the sample tend to be net arms importers (Resulting in negative net arms exports values for these countries). If I specified the model correctly, then my estimates should reflect the long run steady-state relationship between output per capita and its determinants. III. Data The data set comes from the World Development Indicators (2002). The data set is a balanced cross-sectional time series panel covering 62 countries from 1990 to 7

Yakovlev 8 1999 (After excluding the outlier countries with the negative savings and suspiciously large arms import figures). 2 The dependent variable Y/L is GDP per capita measured in constant dollars, I/Y is gross domestic savings expressed as a share of GDP, (n + g + δ) is population growth n plus g + δ = 0.05 (the value suggested by MRW, 1992), SCHOOL is the ratio of students enrolled in secondary schools to the total population of that age group. The net arms exports variable equals to arms exports as a share of GDP minus arms imports as a share of GDP. The net arms variable indicates whether a country is a net arms exporter (positive) or importer (negative). Because the arms exports and imports are very volatile and do not always match arms payments in a timely manner, the final net arms exports variable is an average weighted according to this formula: [(1/10) (t-4) + (2/10) (t- 3) + (3/10) (t-2) + (4/10)(t)]. I argue that this average measure of net arms exports is more meaningful for estimating the long run steady-state relationship than annual net arms exports. Some countries had a small number of missing values for arms exports or imports, which I replaced with the values generated by the weighted average formula above in order to make the panel balanced. The data sample also contains non-oil, intermediate, and OECD dummies: one for non-oil producing countries, one for the so called intermediate countries with population less than 1 million, and one for the OECD countries. IV. Empirical Results As a first step, I check if my model can produce meaningful results by comparing MRW (1992) findings with my own estimates using the same sub-sample categories applied to a panel data set with different years and countries though. This exercise should tell me how applicable is the MRW model in explaining 2 The estimates before and after taking the outliers out do not differ significantly. 8

Yakovlev 9 variation in GDP per capita in my cross-country time-series data. Table 1 in the appendix shows the basic estimates from the MRW (1992) paper. [Table 1 about Here] According to these estimates, physical and human capital positively affect economic growth, while population growth affects it negatively. See Table 2 for a comparison of my estimates based on a different sample with the MRW s estimates in Table 1. The coefficients in Table 1 and 2 for the non-oil and OECD samples are identical in signs but are somewhat different in magnitudes and standard errors. In fact, my OECD sub-sample yields statistically significant coefficients for physical capital and population compared to the MRW s OECD sub-sample, indicating that the basic MRW model produces meaningful results for my non-oil and OECD panel subsamples. Regressions for the entire panel of 62 countries also yield similar results. However, my intermediate sub-sample yields an unexpected negative coefficient (though not significant) for physical capital. [Table 2 about Here] These results indicate that the basic MRW (1992) model works well for the general sample of 62 countries as well as non-oil and OECD sub-samples, but not so well for the intermediate sub-sample. Since my intermediate sub-sample estimate for physical capital is not in accordance with the MRW (1992) findings, I find it reasonable to discard my intermediate sub-sample from further analysis. Now is the time to include the net arms variable into my regressions. The pooled OLS and robust regression estimates in Table 3 and Table 4 show that the net arms exports variable is positive and significant for different samples and estimation methods, while the coefficients (though not necessarily significant in every regression) for saving, population, and schooling are signed in accordance with the theory. Table 3 shows that the coefficient for net arms exports increases in 9

Yakovlev 10 magnitude and t-statistic from 0.17 (2.75) in the entire sample to 0.77 (7.90) in the OECD sub-sample. This evidence appears to support my argument that defense spending in developed countries is likely to lead to substantial arms exports, which should positively affect growth. The inclusion of non-oil, intermediate, and OECD dummies in the pooled OLS regression for the entire sample (62 countries) helps to explain 74% of cross-country variation in income per capita. This is a very good R- squared value considering that I am using panel data with a very diverse sample of countries. [Table 3 about Here] Table 4 shows that the intermediate and OECD dummies are positively and significantly related to growth, while the non-oil dummy is negatively and significantly related to growth. It seems that adding the net arms variable lowers the t-static for saving, which suggests that the two variables correlate. In fact, the correlation matrix in the appendix shows a positive correlation of 0.30 between savings and net arms exports, which is not alarming. This correlation is not surprising though. As discussed previously, rich countries or countries with large capital stocks usually tend to exporter military goods in large numbers. Moreover, it has been argued that saving needs to be modeled as endogenous. In fact, the Solow growth model has been criticized for its ad-hoc saving specification. I address the endogenous savings problem with the instrumental variable estimation method (IV). I use the average of five lagged annual saving figures and non-oil, intermediate, and OECD dummies as instruments explaining the endogenous nature of saving. The resulting pooled IV estimates shown in Table 4 are very similar to the other estimations. All the coefficients are significant (except for saving) and have the expected signs. The first stage IV estimates show a positive (although not significant) relationship between savings and net arms 10

Yakovlev 11 exports, which goes along with one of my arguments that sufficiently large arms exports may increase capital inflow and investment in developed countries. Another pooled IV estimation, but with the generalized method of moments (GMM) estimator, yields the expected coefficients, although neither population nor saving appear significant. On the other hand, net arms exports, schooling, and nonoil, intermediate, and OECD dummies are highly significant. In addition to these estimates, I employ panel FGLS estimator. Although the panel FGLS estimator shown in Table 4 lowers the estimates magnitudes compared to the other estimation methods, it produces highly significant coefficients with the expected signs. This FGLS estimator does not assume the IID error structure. Instead, it assumes heteroskedastic error structure with cross-sectional correlation and panelspecific AR (1) autocorrelation. This is a reasonable assumption for the crosscountry time series data that tends to be very heterogeneous and non-stationary. The heteroskedasticity tests show significant heteroskedasticity presence, which I address by utilizing robust standard errors in pooled OLS and IV regressions. Under these assumptions, the FGLS results are very convincing and agree with the OLS and IV estimates. Perhaps these successful error assumptions for the FGLS estimator could explain why the two-way fixed effects estimator yields unexpected estimates: significant and negative coefficients for net arms exports and schooling. In fact, several panel data growth studies have pointed out that their data is plagued with non-stationarity, autocorrelation, and heteroskedasticity problems. This information further compromises the possibility of meaningful inference from the two-way fixed effects estimator that does not control for these error structures. Therefore, these bizarre two-way fixed effects estimates are very likely to be unfit for inference. 11

Yakovlev 12 Another empirical evidence for the positive relationship between economic growth and net arms exports comes from regressing the Solow residual on net arms exports, which yields a significantly positive relationship between these two variables. This evidence supports of my propositions that the level of net arms exports should be positively related to the level of productivity or technology in that country. This means that developed countries that can afford an advanced industrial complex and that tend to be the net arms exporters are also likely to be more productive than developing net arms importing countries. However, it is difficult to say whether this relationship implies that the military complex produces significant positive externalities or that net arms exports proxy for productivity in developed countries. Perhaps, future research can investigate the direction of causality in this relationship and reexamine it using more sophisticated econometric methods. V. Conclusion Several different estimation methods consistently show that net arms exports have a significantly positive effect on economic growth in different samples, which suggests that this relationship is robust across different estimation methods and countries. Saving, population growth, and schooling have the expected signs and usually significant coefficients. These results suggest that even if the total empirical effect of defense spending on growth turns out to be negative, an advanced industrial defense sector may appear beneficial for economic growth if it works through the arms exports channel. Nevertheless, it remains inconclusive whether net arms exports simply proxy for the level of technology rather than tax burden exporting or de-facto return to providing national defense. A further empirical examination of the 12

Yakovlev 13 relationship between the Solow residual and net arms exports may yield more insightful evidence. Perhaps, future theoretical and empirical research should also explore how per worker output relates to various factor inputs in civilian and defense sectors of the economy. I think that a two-sectored growth model with solid micro foundations may provide the basis for more conclusive empirical estimates of how different defense spending components can affect economic growth. 13

Yakovlev 14 References Ball, Nicole. 1983. Defense and Development: A Critique of the Benoit Study, Economic Development and Cultural Change 31, 507-24. Barro, R. 1991. Economic Growth in a Cross Section of Countries, The Quarterly Journal of Economics. May Benoit, E. 1973. Defense and Economic Growth in Developing Countries Boston: D.C. Heath & Co. -----. 1978. Growth and Defense in Developing Countries, Economic Development and Cultural Change 26, 271-80. Biswas, B., and R. Ram. 1986. Military Expenditures and Economic Growth in Less Developed Countries: An Augmented Model and Further Evidence, Economic Development and Cultural Change 34, 361-72. Chan, S. 1985. The Impact of Defense Spending on Economic Performance: A Survey of Evidence and Problems ORBIS (Summer) 403-34. -----. 1986. Military Expenditures and Economic Performance, World Military Expenditures & Arms Transfers 1986, 29-37. Chowdhury, A.R. 1991. A Causal Analysis of Defense Spending and Economic Growth, Journal of Conflict Resolution, March, Vol. 35 Issue 1, p80. Deger, S., and S. Sen. 1983. Military Expenditure, Spin-Off and Economic Development, Journal of Development Economics 13, 67-83. Deger, S., and R. Smith. 1983. Military Expenditure and Growth in the Less Developed Countries, Journal of Conflict Resolution 27 (June), 335-53. Faini, R., P. Annez, and L. Taylor. 1984. Defense Spending, Economic Structure and Growth: Evidence Among Countries and Over Time, Economic Development and Cultural Change 32, 487-98. Fredericksen, P.C., and R.E. Looney. 1982. Defense Expenditures and Economic Growth in Developing Countries: Some Further Empirical Evidence, Journal of Economic Development 7, 113-24. -----. 1983. Defense Expenditures and Economic Growth in Developing Countries, Armed Forces and Society 9, 633-45. Hall, R.E. 1988. The Relation between Price and Marginal Cost in U.S. Industry The Journal of Political Economy, Vol. 96, No. 5. (Oct., 1988), pp. 921-947. Heo, Uk. 1998. Modeling the Defense-Growth Relationship Around The Globe, Journal of Conflict Resolution, October, Vol. 42 Issue 5, p637, 21p Kennedy, G. 1974. The Military in the Third World New York: Scribner. Leontief, W., and F. Dutchin. 1983. Military Spending: Facts and Figures New 14

Yakovlev 15 York: Oxford University Press. Lim, D. 1983. Another Look at Growth and Defense in Less Developed Countries, Economic Development and Cultural Change 31, 377-84. Maizels, A., and M. Nissanke. 1986. The Determinants of Military Expenditures in Developing Countries, World Development 14 (September): 1125-1140. Mankiw, G., Romer, D., and Weil, D. 1992. A Contribution to the Empirics of Economic Growth, The Quarterly Journal of Economics, May Perkins, Dwight, et. al. 2001. Economics of Development, 5th edition. W.W. Norton. SIPRI Yearbook, 1970-1985. http://www.sipri.org Smith, D., and R. Smith. 1980. Military Expenditure, Resources and Development Birbeck College, London. Mimeo. Temple, J. 1999 The New Growth Evidence, Journal of Economic Literature, Vol. XXXVII (March), pp. 112-156 US Statistical Abstracts. 1967-1977 Whynes, D. K. 1979. The Economics of Third World Military Expenditure London: Macmillan. 15

Yakovlev 16 Appendix Figure 1. Arms Imports and Exports in Developed Nations Arms Imports and Exports in Developed Nations: 1970-99 70000 60000 R 2 = 0.5101 Constant dollars, millions 50000 40000 30000 20000 10000 0 1970 1972 1974 1976 1978 1980 1982 1984 Year R 2 = 0.8742 1986 1988 1990 1992 1994 1996 1998 Source: World Military Expenditures and Arms Transfers, US Arms Control and Disarmament Agency. Arms Imports Arms Exports Arms Exports Trend Arms Imports Trend Table 1. Mankiw, Romer, and Weil s (1992) Estimates: 1960-1985 Sample: Non-oil Intermediate OECD Countries: 98 75 22 Constant 6.89 (5.89) 7.81 (6.56) 8.63 (3.94) ln(i/y) 0.69 (5.31) 0.70 (4.67).28 (0.72) ln(n + g + δ) -1.73 (4.22) -1.50 (3.75) -1.07 (1.43) ln(school) 0.66 (9.43) 0.73 (7.30) 0.76 (2.62) R-squared 0.78 0.77 0.24 Dependent variable: log GDP per worker in 1985 16

Table 2. Pooled OLS on Panel Data without Net Arms Exports, 1990-1999 Estimation: Pooled OLS* Pooled OLS* Pooled OLS* Sample: Non-oil Intermediate OECD Countries: 59 53 22 Constant 7.58 (6.34) 11.72 (10.67) 6.15 (4.81) ln(i/y) 0.18 (2.21) -0.05 (0.59) 1.18 (10.28) ln(n + g + δ) -2.22 (5.29) -3.84 (10.31) -1.63 (2.74) ln(school) 1.10 (9.87) 1.01 (8.07) 0.64 (3.11) R-squared 0.51 0.60 0.39 Dependent variable: ln(gdp per capita). * Regressions with robust standard errors and corrected for heteroskedasticity. Yakovlev 17 Table 3. Pooled OLS and Robust Regressions on Panel Data, 1990-1999 Estimation: Pooled OLS* Pooled OLS* Pooled OLS* Robust Regression Sample: Entire Sample Non-oil OECD Entire Sample Countries: 62 59 53 62 Constant 6.99 (5.92) 7.25 (6.12) 6.72 (6.22) 8.11 (10.63) Netarms 0.17 (2.75) 0.18 (2.66) 0.77 (7.90) 0.35 (9.05) ln(i/y) 0.14 (1.56) 0.09 (1.02) 1.16 (8.15) 0.15 (1.83) ln(n + g + δ) -1.89 (4.51) -1.92 (4.58) -0.83(1.96) -2.42 (8.71) ln(school) 1.15 (10.52) 1.13 (10.05) 0.25 (2.05) 1.14 (16.16) R-squared 0.51 0.52 0.61 - Dependent variable: ln(gdp per capita). * Regressions with robust standard errors and corrected for heteroskedasticity. 17

Yakovlev 18 Table 4. Pooled OLS, IV, and GLS on Panel Data, 1990-1999 Estimation: Pooled OLS* Pooled IV* IV-GMM* Panel FGLS Sample: Entire Sample Entire Sample Entire Sample Entire Sample Countries: 62 62 62 62 Constant 7.13 (10.53) 6.98 (5.93) 7.14 (10.60) 10.03 (43.02) Netarms 0.15 (3.11) 0.17 (2.75) 0.15 (3.18) 0.04 (13.28) ln(i/y) 0.11 (1.67) 0.14 (1.47) 0.10 (1.48) 0.07 (4.38) ln(n + g + δ) -0.20 (0.82) -1.89 (4.52) -0.20 (-0.84) -0.59 (8.73) ln(school) 0.57 (7.05) 1.15 (10.50) 0.57 (7.06) 0.17 (9.16) Non-oil -1.58 (6.62) - -1.58 (6.68) -2.38 (24.66) Intermediate 0.18 (2.29) -.18 (2.31) 0.39 (14.43) OECD 1.67 (18.12) - 1.67 (18.23) 2.04 (46.92) R-squared 0.74 0.51 - - Dependent variable: ln(gdp per capita). The IV-GMM and FGLS estimates contain z-statistic in parentheses; otherwise it s a t-statistic. * Regressions with robust standard errors and corrected for heteroskedasticity. Table 5. Correlation and Standard Deviation Matrix Correlation / lngdp Netarms ln(i/y) ln(n + g + δ) ln(school) St. Deviation lngdp 1.3045 - - - - Netarms 0.2669 0.9483 - - - ln(i/y) 0.2797 0.9483 0.4612 - - ln(n + g + δ) -0.5648-0.3125-0.3157 0.1551 - ln(school) 0.6616 0.1085 0.2358-0.5568 0.5790 Standard Deviations are on the diagonal. Table 6. Variance Inflation Factors for the Entire Sample of 62 Countries VIF 1/VIF Netarms 1.28 0.779797 ln(i/y) 1.21 0.828872 ln(n + g + δ) 2.16 0.462236 ln(school) 1.79 0.559343 18