1

Government Size, Openness, and Output Growth: Industrial and Developing Economies under Globalization André Varella Mollick Corresponding Author Department of Economics and Finance University of Texas - Pan American 1201 W. University Dr. Edinburg, TX 78539-2999, USA E-mail: [email protected] Tel.: +1-956-316-7135 and fax: +1-956-381-2687.

René Cabral Escuela de Graduados en Administración Pública y Política Pública Tecnológico de Monterrey, Campus Monterrey Ave. Rufino Tamayo, Garza García, NL, CP. 66269, Mexico. E-mail: [email protected] Tel.: +52-81-8625-8347 and fax: +52-81-8625-8385

Abstract: We examine in this paper the effects of government size (G/Y) on industrial and emerging economies’ output growth over the “globalization years” of 1986-2004. Controlling for standard determinants of the Solow growth model and three openness measures, we find under fixed effects and dynamic panel data models strong negative long-run effects of government size on output growth in industrial economies. In emerging markets, the effect of G/Y on output growth is ambiguous. We interpret this finding as capital inflows leading to higher savings under savings complementing capital inflows in emerging markets. Overall, openness has positive effects on output growth. Keywords: Capital Inflows, Economic Growth, Globalization, Government Size, Dynamic Panel Data Methods. JEL Classification Numbers: F32, H11, H50, O47.

2 1. Introduction The relationship between foreign capital and economic growth has been explored from different angles. An important question is whether foreign capital complements domestic capital and is able to make output grow steadily in the long-run. Also, the role of government on output growth is very controversial and empirical evidence is mixed at best; see Kneller et al. (1999). Standard economic reasoning suggests that positive government expenditures may either spur growth under idle resources and a weak state of aggregate demand or may crowd-out private consumption and investment substantially. This paper reconsiders these issues using panels of industrial and emerging market economies for the “globalization years from 1986 to 2004”. With the Solow (1956) model as benchmark, the rate of population growth and the ratio of investment to output (I/Y) are the key control variables. Capital flows from abroad and government expenditures provide additional channels to the original model, through their effects to the stock of capital and to the savings rate of the economy, respectively. This paper conducts a reconsideration of openness effects on economic growth, initially without government and then allowing for long-run government effects on growth. In doing so, we allow for “flows of capital in a global world” as an alternative measure of openness. We argue that our research strategy is important for several reasons. First, Mishkin (2009) contrasts the opening of financial markets to opening to trade in goods and services. He points out that opening of financial markets to foreign capital directly increases access to capital, lowering its cost. Henry (2007) documents the cost of capital falling, investment booming, and growth rate of GDP rising after capital account liberalizations. In addition, financial opening promotes reforms to the financial

3 system which will improve its functioning. At the same time, opening domestic markets to foreign goods can spur financial development as long as it can weaken the political power of entrenched business interests that might block institutional reforms. Rajan and Zingales (2003) find that a deeper financial sector is positively associated with greater trade openness. Free trade may also promote financial deepening by reducing corruption. It is an empirical matter, however, to determine which of the two types of openness current account effects through trade or capital account effects through capital flows - has a stronger impact on long-run economic growth. While emerging market economies have abolished capital controls and experienced a significant amount in inflows, they have been unable to fully enjoy the risk sharing benefits of financial globalization as documented by Kose et al. (2009). Second, the optimal size of government is far from being an established issue and its positive effect on output depends on the relative efficiency of the public sector. Acemoglu (2005) constructs a theory of “Politics and Economics in Weak and Strong States” and shows that when the state is excessively strong the ruler imposes such high taxes that jeopardize economic growth. When the state is excessively weak, the ruler anticipates that he/she will not be able to extract rents in the future and under-invests in public goods. Both weak and strong states create distortions. Barro (1990) extends endogenous-growth models to include tax-financed government services on production and assumes exogenous government actions; in his model variations in the share of productive government expenditures in GDP affect the growth and saving rates. Barro (1990) goes on to conclude that the ratio of government spending to output (G/Y) and per capita GDP would show little correlation across

4 countries because each government goes to the point at which the marginal effect of G/Y on growth is close to zero. Kneller et al. (1999) find strong support for Barro (1990)’s endogenous growth model, in which taxation and public expenditure can affect the steady-state growth rate. Exploring a panel of 22 OECD countries from 1970 to 1995, they find that: i) distortionary taxation reduces growth; and ii) productive government expenditure enhances growth. Growth regressions in Mueller and Stratmann (2003) from 1960 to 1990 show that the relationship between government size and growth is negative and significant for the full sample as well as for subsamples. They summarize their findings on the links between income level and government size as follows: there is a positive association between government size and growth in low-income countries, where the government sectors tend to be small, and a negative relationship in the high-income countries, where government sector has perhaps grown too large. Third, the reexamination of Solow’s model by Mankiw et al. (1992) suggests a much lower fit for the human capital version across the OECD subsample. Also noteworthy is Kraay and Raddatz (2007) application of Solow (1956) to poverty traps: at low levels of income, investment is so low that it can only sustain very small levels of capital stock per capita. For openness measures, distinctive patterns emerge as well between emerging markets and industrial economies. Studies suggest that developed and developing countries may have different causal mechanisms when exploring economic growth under both external flows and the government sector.1

1

De Gregorio (1992) examined 12 Latin American countries over 1950-1985 and found no strong output effects of the share of trade to GDP. Focusing on developing countries from 1970-1997, Yanikkaya (2003) shows that there is no straightforward relationship between openness and economic growth, which depends on the type of openness measure used. More recently, Astorga (2009) examines the six largest Latin American economies and finds that physical and human capital are the key growth determinants over a century of data.

5 Fourth, most of the existing empirical estimates have assumed that government expenditure or openness measures are entirely exogenous to economic growth. This paper explores the class of estimators developed by Arellano and Bond (1991), which are suitable to handle endogeneity of the regressors. We also perform careful analysis of empirical tests in order to verify the adequacy of the empirical fits of the various specifications. Endogeneity has plagued past research, especially with respect to openness and growth. 2 Given this state of knowledge, not many papers have examined the effect of government size on economic growth allowing for financial measures of globalization. One previous such paper is Edison et al. (2002), who report estimations of international financial integration effects on economic growth per capita for 57 countries. Although they usually find positive growth effects on the capital flow measure, they show mixed results for government balances. Without international financial measures, OLS and IV cross-country regressions find that government balance has no impact on growth. When a dynamic panel GMM estimator is used, however, increases in government surplus have a negative impact on growth in the long-run. Another study in Kneller (2007) compares the rate of growth following trade liberalization from growth in other developing countries that did not liberalize. He finds that countries that incur trade liberalization (measured by

2

For instance, Lee et al. (2004) document a two-way relationship between openness and growth using fixed-effects and differenced GMM estimators. Combes and Saadi-Sedik (2006) use system-GMM estimators and decompose the standard trade openness measure into natural openness (estimating the level of trade openness a country should have based on structural factors) and trade policy (difference between trade openness and natural openness). Their empirical work focuses on 66 developing countries over 19741998 and finds that trade openness increases a country’s exposure to external shocks.

6 dummy variables) increase their spending on welfare but not other forms of expenditure. There is thus a link between government expenditure and openness as in Rodrik (1998).3 Using the standard controls of the Solow growth model of rate of population growth and investment/output ratios, we find systematic positive and significant effects of openness on real output growth over the “globalization years” of 1986-2004. We also observe strong negative long-run effects of G/Y on output growth in industrial economies. In emerging markets, the effect of government size on output is ambiguous. Our most likely explanation is through the link between savings and foreign capital as in extensions of Solow (1956) to the open-economy by Barro et al. (1995) and Verdier (2008). In industrial economies capital inflows lead to lower savings under savings substituting capital inflows. In contrast, in emerging markets capital inflows lead to higher savings under savings complementing capital inflows. This paper has five more sections. Section 2 summarizes the Solow model of economic growth; section 3 introduces the data and major features for both sets of countries; section 4 contains the empirical methodology under both fixed effects and dynamic panel data models; section 5 discusses the results; and section 6 concludes.

2. The Theoretical Framework The details of the Solow (1956) model are reproduced in any intermediate level Macroeconomics textbook such as Abel et al. (2008, Chapter 6) and are only briefly 3

Several papers have dealt with the link between government size and openness per se. While Rodrik (1998) documented a positive covariation between trade openness and the size of government, Alesina and Wacziarg (1998) noted a negative covariation of country size with trade openness and with share of public consumption in GDP. A methodological modification in Ram (2009) adds to the debate with a 41-year panel data covering the period 1960-2000 for 154 countries. His fixed-effects estimates suggest more support for the direct relationship thesis in Rodrik (1998). Causality tests in Benarroch and Pandey (2008) support the notion that larger government size leads to lower openness, the opposite to Rodrik (1998). For causation running in both directions, see Svaleryd and Vlachos (2002).

7 summarized here. The economy is closed and there is no government. The starting point is a constant returns to scale production function of y = f (k) with k = K/L. The usual concavity on f is assumed and an increase in k raises the amount of output produced per worker. The rate of population growth is represented by n and the depreciation rate of capital stock is given by d. The other important equation is the steady-state gross investment, a straight-line equation that depends on the stock of capital. Total investment comprises net investment (nKt) and capital depreciation (dKt): It = nKt + dKt = (n + d) Kt. Making national savings (S) a function of income: St = sYt, where s is the constant savings rate, and equating S to I yields: sYt = (n+d) Kt. The steady-state capitallabor ratio k* is determined by this equilibrium condition. The fundamental determinants of long-run living standards include: i) the savings rate (an increase leads to higher investment and then capital stock); ii) the rate of population growth (an increase implies more output must be used to equip new workers with capital, leaving less output for consumption or to increase K/L); and iii) productivity changes (higher productivity raises incomes). We consider the first two factors in this paper and allow for two (empirical) modifications of this basic framework.4 One important modification has been performed

4

Previous works have reconsidered extensions of Solow (1956) growth model. Mankiw et al. (1992) showed that the augmented (including human capital) Solow model provides a good description of crosscountry data, with the exception of the OECD subsample. For the latter, the textbook Solow model explain less than 6% of the variation in per capita income and the performance of the augmented model is somewhat better but explaining still less than 30% of income variations for these countries. Nonneman and Vanhoudt (1996) suggest a further augmentation of the Solow model (to three types of capital) by including the endogenous accumulation of technological know-how. Bajo-Rubio (2000) generalizes theoretically Solow growth model to government, including government-providing inputs into the production function and finds a non-monotonic relationship between the rate of growth of per capita output and the size of the public sector. This leads to an inverted U-shaped relationship between the two variables.

8 by Barro (1990).5 With positive tax-financed government expenditures, the savings function shifts down and the production function will move downwards.6 If the steadystate investment per worker does not change, then equilibrium output per worker will fall. In terms of the typical diagram, government expenditures imply a shift downwards of the savings function. Output and capital-labor ratio would fall correspondingly. If the economy is open, capital flows from abroad can be added to the existing level of capital stock and could in principle ameliorate this effect if external capital complements domestic capital. The precise effects of both shocks (government budget and external capital) are, however, subject to empirical analysis and may be more important in one particular case (such as with higher levels of income with higher degree of openness) than in others. Extensions of this economy to the open-economy setting have been put forward by Barro et al. (1995). While in the closed economy debt is zero for the household and the interest rate r is determined by S = I, in an open economy r is determined abroad and

5

Barro (1990, p. S121) mentions that “aside from problems of measuring public services and the rates of growth and saving, the empirical implementation of the model is complicated by the endogeneity of government. Within the theoretical model (and with a Cobb-Douglas production function), the government sets the share of productive expenditures, G/Y, to ensure productive efficiency. Therefore, … each government would operate at the same point, G/Y = α. Within this framework of optimizing governments, cross-sectional variations in G/Y arise only if α differs from country to country. The parameter α, which measures the productivity of public services relative to private services, could vary across countries for a number of reasons. These include geography, the share of agricultural production, urban density, and so on.” 6 We stick to the traditional size of government treatment of G/Y. One should acknowledge, however, that G/Y can be decomposed into several components. Garen and Trask (2005), for example, consider nonbudgetary measures such as government ownership of enterprises, price controls, and asset expropriation. While each of these may have little impact on government expenditures, they can make the role of government sizable. They demonstrate that the scope of government is much larger in less open economies when considering non-budgetary measures and that higher levels of non-budgetary government are positively correlated with trade barriers. Rayp and Van De Sijpe (2007) implement a methodology based on data envelopment methods to measure and explain government efficiency in 52 developing countries. They find that government expenditure efficiency is primarily determined by structural country variables and government indicators, while economic policy determinants count less. Bose et al. (2007) examine the growth effects of government expenditures for a panel of 30 developing countries over the 1970s and 1980s. They find that current expenditures are insignificant in growth regressions.

9 foreign debt per worker can be positive or negative. They show that in the closed economy the ratio between physical and human capital stays constant, whereas in the open economy the ratio falls during the transition. The fall in the ratio causes diminishing returns to human capital to set in faster than otherwise and the speed of convergence is greater in the open economy than in Solow (1956). A recent contribution by Verdier (2008) introduces domestic (K) and foreign capital (Z), along with labor input, in a CES production function, and derives the equilibrium condition, in which S = I in domestic capital minus net factor payments on debt. Using the same mechanism as in Barro et al. (1995), k/z (in per worker terms) falls during the transition: k is relatively high initially but becomes less important with openness. Verdier (2008) shows that k does not jump immediately to its steady-state since domestic capital accumulation is constrained and k and z are complementary in production. The relative importance of the two types of capital determines the degree to which countries are constrained. In this setting, domestic savings and foreign investment are complements. In other words, “the country’s ability to attract capital flows may be directly linked to its savings rate.” Verdier (2008, p. 139). Since G affects the savings rate, this paper investigates the effects of government policies on output per worker when precisely both types of capital are present.

3. The Data We have a total of 55 economies in our sample: 22 are industrial and 33 emerging economies. The main sources of our data are the IMF’s International Financial Statistics database and the multi-country dataset on foreign assets compiled by Lane and Milesi-

10 Ferreti (2007). Data for GDP per capita (our dependent variable), investment, government spending and trade are obtained from the former source. Financial globalization indexes were constructed based on the database in Lane and Milesi-Ferreti (2007). Table 1 lists the economies in the sample and provides some descriptive statistics. Over the period from 1986 to 2004, output growth per capita stood at 2.23% on average for the 22 industrial economies and at 2.50% for the 33 emerging markets. With a close to 21% level for I/Y in all countries, industrial economies had a much larger rate of government size (19.4%) than in emerging markets (13%) and a lower rate of population growth (0.59% versus 1.59%). However, openness measures varied considerably across countries: they are much higher for industrial economies if measured by capital flows and are slightly higher for emerging markets if measured by trade considerations. [Table 1 here] Our first measure of globalization is the traditional trade openness (TO), which is calculated as total trade, the sum of exports and imports, over GDP. Following Lane and Milesi-Ferretti (2007), we employ two alternative measures of financial globalization. First, a measure of international financial integration (IFI) with respect to GDP: IFIit = (FAit + FLit)/GDPit, where: FA (FL) denotes the stock of external assets (liabilities). Second, a financial integration measure also with respect to GDP and based on portfolio equity and FDI stocks: GEQit = (PEQAit + FDIAit + PEQLit +FDILit)/GDPit, where: PEQA (PEQL) denotes the stock of portfolio equity assets and FDIA (FDIL) denotes the stock of their direct investment assets (liabilities). In Table 1 the countries with the largest averages rates of per capita growth across emerging and industrial countries are, respectively, China (8.41%) and Ireland (5.47%).

11 Meanwhile, those with the lowest growth rates are Venezuela (0.005%) and Switzerland (0.89%). On average, the group of emerging markets presents slightly higher investmentto-output ratio (21.36% versus 20.99%) and per capita GDP growth rate (2.50% versus 2.23%) than the group of industrial economies. Interestingly, the investment-to-output ratios are positively related to the growth rate of the output per capita in emerging markets (correlation coefficient of 0.75) but not among industrial economies (correlation coefficient of -0.03). The correlations between our three measures of globalization and the output per worker are always positive and on average larger for industrial economies. In Table 1, trade openness is the only measure of globalization that shows a higher average index among emerging economies (0.73) than across industrial countries (0.66). Financial globalization measures display positive correlation coefficients with output per capita growth and observe in general a stronger link in industrial economies: 0.35 against 0.01 for assets related globalization (IFI); 0.29 against 0.16 for equity related globalization (GEQ); and 0.41 against 0.25 for trade openness (TO). On government size, there are substantial differences in size (13% on average for emerging markets versus 19.4% for industrial economies) but there are weak negative correlations between G/Y and output growth, varying from -0.10 in emerging markets to -0.16 in industrial economies. Correlation between G/Y and I/Y is negative in industrial economies (-0.43) supportive of a large crowding-out and very weak (0.07) for emerging markets. Finally, in the same vein correlation between G/Y and OPEN is negative in industrial economies (except for positive 0.21 with TO) and close to zero throughout for emerging markets. The latter suggests no association between openness and G/Y in emerging markets.

12

4. Methodology The basic model employed here builds on Mankiw et al. (1992) empirical treatment of Solow (1956). Defined in logs, the benchmark empirical equation is:

Yit = β0 + β1i + β2 Zit + εit

(1),

where: Yit is the real income per capita in country i at time t, β0 is an intercept term, β1i represents a vector of country specific factors, Zit is a vector of traditional explanatory variables, such as the rate of population growth (n=∆Lt/Lt-1) and the investment-to-output ratio (It/Yt); and εit is the stochastic error term. From the traditional Solow (1956) model, we expect the two control variables to have different impacts on output per capita: β2 < 0 for population growth or β2 > 0 for investment-output ratio. Kose et al. (2006, 2009) refer to the “globalization period” from 1986:3 to 2003:4 or from 1987 to 2004 as the one in which there were dramatic increases in the volume of cross-border trade in both goods and assets. Henry (2007) lists eighteen developing countries that liberalized their stock markets between 1986 and 1993, including Argentina, Brazil, Chile, Mexico, Philippines, South Korea, Thailand, and Turkey. Several authors, including Bekaert et al. (2005) and Quinn and Toyoda (2008) have documented that these increasing capital flows have resulted on higher output growth. Kose et al. (2006) argue that using the sum of gross capital inflows and outflows as a ratio to national GDP (financial openness) yields a nice symmetry with the widely-used measure of trade openness, which is the sum of imports and exports as a ratio to GDP.

13 Since trade openness and financial openness capture different aspects of trade and financial flows in a global world, we explore them in turn in the following augmented model, accounting for globalization:

Yit = β0 + β1i + β2 Zit + β3 OPENit + εit

(2),

where: OPEN is a vector containing any of our three measures of globalization: IFI, GEQ or TO.7 This empirical equation considers the influence that trade and financial globalization might have had on per capita output growth. We expect β3 > 0 on the basis of more circulation of trade or financial flows helping the productive sector. Since our sample focuses on the post-liberalization years, it is natural to assume that import substitution policies conducted by developing countries are part of the past.8 The link between (2) and (1) can be grounded on recent theoretical models which have explored the relationship between an index of capital/output ratio and economic growth. One such example is Beaudry and Collard (2006), who show that during a period of globalization (limited to trade openness in their paper) we should observe an increase in the social returns to capital accumulation. Regressing change in financial openness on changes in trade openness and changes in GDP per capita, Aizenman (2008) finds a highly significant positive

7

Interactive variables can be constructed in order to capture the interactive measures between trade and financial flows: IFI*TO, and GEQ*TO, respectively as in Aizenman (2008). These coefficients turned out to be very small and are thus omitted in this version. 8 While Astorga (2009) acknowledges technological spillovers and the international transmission of knowledge as a source of growth for open economies, he mentions that protection might also boost higher longer-term growth in those cases where it encourages learning by doing by the infant industry argument and/or investment in research-intensive sectors, which might justify a priori an ambiguous sign for β3, at least for Latin American countries.

14 association between changes in financial and trade openness in developing countries.9 Greater TO increases the effective financial openness, by providing more opportunities for capital flight. However, authorities need to spend resources on monitoring and enforcing existing capital controls in order to curb illicit capital flows. This view has been exposed by Rodrik (1998) with government spending playing a risk-reducing role in economies exposed to a significant amount of external risk. The implication is that the relationship between openness and government size is strongest when external risk (terms of trade shocks, capital flight) is highest. A reexamination on 96 countries for the 19702000 period by Benarroch and Pandey (2008) casts doubt on this positive relationship and suggest reverse causation problems since Granger causality tests show that higher lagged government size leads to lower trade openness. One way to capture the link between openness and growth, allowing for the role of administrative costs in monitoring trade and financial transactions, is to interact government size with measures of openness, such as: Gov*IFI, Gov*GEQ, and Gov*TO. A study adopting a similar motivation on the complementary role of economic and institutional reforms on openness in Chang et al. (2008) interacts the openness measure with proxies of, respectively, educational investment, financial depth, inflation stabilization, public infrastructure, governance, labor market flexibility, and ease of firm entry and exit. In our case, to the extent that the hypothesis in Rodrik (1998) is true, the joint information of the increase in government activities along with higher trade and financial operations would have a positive effect on economic growth. On the other hand, 9

Specifically, Aizenman (2008) reports that a 10% increase in TO is associated with an average 2.6% increase in financial openness for all the developing countries, and with a 0.7% increase in financial openness in the sample of developing countries excluding countries in which financial openness exceeds 100% and trade openness exceeds 60%, and small island economies. The size of this effect in the OECD countries is 2%, but at a substantially lower significance level.

15 a negative sign on the government-openness interactive terms would imply that the bigger size of the government crowds-out private sector activities and contributes to lower economic growth. Given the unit root of the series commonly found in levels and the endogeneity of regressors such as G/Y and OPEN it is important to allow for a dynamic specification that contains lagged real GDP per capita as independent variable, as in:

Yit = β0 + β1 Yit-1 + β2 Zit + β3 OPENit + εit

(3),

This dynamic specification can be estimated using the dynamic panel methods developed by Arellano and Bond (1991). The first differentiation eliminates country specific effects and time-invariant explanatory variables from equation (2). The GMM procedure for dynamic models of panel data addresses endogeneity and controls for unobserved country-specific factors. The consistency of the GMM estimators depends on whether lagged values of the explanatory variables are valid instruments in the growth regression. We expect output persistence to be fundamentally different between industrial economies and emerging markets. The expected signs are as in the static model and a value of β1 close to 1 would indicate a high degree of persistence.

5. Results 5.1. Bivariate Regressions Figure 1 illustrates the bivariate relationships across samples. For both samples of emerging and industrial economies we plot average per capita GDP growth rates against

16 average I/Y in the first row. Although there is a rather flat pattern found for industrial countries, there is a strong positive relationship between I/Y and output growth for emerging markets. In the latter, the variance of output growth is widely explained (56.6%) by movements in I/Y. There is also a weak negative relationship between G/Y and output growth in both samples as depicted in the second row. In contrast, the behavior of openness is remarkably different across samples: there seems to be a positive relationship for openness and economic growth in industrial countries, no matter which indicator is used. In particular, for industrial economies there is a higher than one coefficient (1.247) on TO when explaining about 17% of output variance. There are also positive relationships between openness and economic growth for emerging markets but the R2 is smaller, varying from 0% in IFI to about 6% in TO. In all but one of the six scatterplots, the fitted regressions confirm the positive association between average GDP growth rates and trade and financial globalization. [Figure 1 here]

5.2. Static Panel Data Methods Fixed effects-models suggest, for the full sample of countries in Table 2, the following set of results. First, the rate of population growth has a negative impact on economic growth per capita but this result is not robust to all specifications. Second, the I/Y ratio has a positive and statistically significant coefficient varying from 0.15 in column (5) to 0.29 in column (6) across models. Only in the standard Solow (1956) benchmark, I/Y has no impact on the rate of economic growth. Third, openness has strong and positive effects and are higher when TO is used (0.470 and 0.480), which

17 suggests that the trade channel is more important than the external financing for economic growth. Fourth, in columns (5) to (8), G/Y has no effect on the steady-state per capita growth, which is in line with ambiguous effects reported by previous research. Fifth, the empirical fit of the standard Solow model (measured by the R2 within) of 3.8% is improved to 6.7% when G/Y is included and to between 32% and 55% when each of the series in OPEN is included as regressor. [Table 2 here] Despite these major findings being broadly consistent with our expectations, there are substantial differences in the coefficients depending on the level of economic development. In Tables 3 and 4, the rate of population growth turns out to be strongly positive in industrial economies and strongly negative for emerging markets. The former result would be against the expected sign by the Solow (1956) model and would suggest that output growth of more mature economies would benefit from increases in population growth. Whenever significant, the coefficients on I/Y are invariably positive, ranging from 0.17 to 0.29 in emerging markets, to more stable values of around 0.22 or 0.26 for industrial countries. Again the specifications (1) and (7) in Table 3 and (1) in Table 4 do not yield any long-run effect of investment/output on economic growth. The coefficients on TO are very high for industrial countries (0.754 and 0.786) and less so for emerging markets (0.315 and 0.323), which are, however, much larger than the impact from the measures of financial globalization. The coefficients on IFI and GEQ are smaller but positive and statistically significant. It is possible to conclude from the full sample and for both sets of countries that openness measured in the traditional sense of trade of goods and services provides a larger effect on economic growth in the steady-state.

18 Previous research has documented mixed effects for some control variables when explaining output growth. In Kneller et al. (1999), for example, investment and population growth have negative but not statistically significant effects for the 22 OECD countries. Omitting initial income in their cross-country regressions, investment becomes positive but still not statistically significant in Kneller et al. (1999)10. Greenaway et al. (2002) find that the value of I/Y is positive for developing countries but very sensitive to the type of liberalization index used. Our estimates show that I/Y has a positive effect on output per-capita growth regardless of the globalization measure we employ. In Bekaert et al. (2005) there is no I/Y in the regressions but the rate of population growth has a strongly negative coefficient on economic growth. [Tables 3 and 4 here]

5.3. Dynamic Panel Data Methods One possibility is that the estimations in Tables 2 to 4 are biased due to the omission of dynamic effects. Allowing for dynamic effects through lagged output terms and the choice of instruments, Tables 5 to 8 report the first-step estimations of the Arellano and Bond (1991) model.11 We consider in Tables 5 to 7 endogenous regressors and allow for exogeneity of regressions in Table 8. For the full sample of countries in Table 5, the rate of population growth loses statistic significance and the I/Y ratio has a positive and statistically significant 10

Note that introducing the initial I/Y in our panel data set up would be the same than introducing fixed effects. Later, in the dynamic specification I/Y or fixed effects would be removed from the regression as the model is estimated in differences. 11 We conduct extensive search on the specification for the dynamic models. Following the literature on dynamic panels, we look for acceptable values of the Sargan test and also of the AB (2) test which tests for second-order serial correlation under heteroskedasticity robust specifications. Our basic criteria were that both tests were satisfied, which in some cases implied that the econometric model for one sample would have a different lag length than for the other sample. See more on this below.

19 coefficient varying from 0.10 to 0.20 across models, with the exception of column (1) for the Solow (1956) model of the closed economy and without government. Openness measures have strong and positive effects and, as before, are higher when TO is used (0.454 in column 4 and 0.238 in column 8). This suggests that the trade channel is more important than the external financing for economic growth. Also, for the combined full sample G/Y has no effect on the steady-state per capita growth, which is in line with ambiguous effects reported in previous research. The Sargan test of over-identifying restrictions (for testing orthogonality between the instruments and the residuals) does not reject the null hypothesis. The AB (2) tests based on a second-order serial correlation test of the residuals from the difference equation indicate that the use of endogenous t-3 dated variables is valid since there is no serial correlation of order 212. [Table 5 here] As in the static case, there are substantial different findings when we split the full sample according to the level of economic development in Tables 6 and 7. In line with what was a priori expected, the rate of population growth becomes strongly negative in both sets of countries, with usually statistically significant effects for industrial economies (between -1.36 and -2.39 without G/Y) in Table 6. The coefficients on I/Y, however, differ drastically: they are small and positive only in some cases for industrial economies (columns 3 and 9 in Table 6) and invariably positive varying from 0.17 to 0.27 in emerging markets. As suggested by Figure 1, there is a robust relationship between I/Y and output growth for emerging market economies in Table 7, which can not be verified for industrial economies in Table 6. Among the series in the OPEN vector, the 12

We employ 3 as the maximum number of lags of the dependent variable. Borwsher (2002) find that this is the number of lags that maximizes the power of the Sargan test using a Monte Carlo experiment with different sample sizes.

20 coefficients on TO are the highest for industrial countries (0.058 when G/Y is included) and also the highest for emerging markets: 0.171 statistically significant at 10% and 0.372 statistically significant at 5%) when G/Y is not included. The coefficients on IFI and GEQ are smaller but positive and statistically significant in general. For both sets of countries, openness measured in the traditional sense of (X+M)/Y provides a larger effect on economic growth in the steady-state. However, between IFI and GEQ measures, GEQ effects are much higher and significant for emerging markets, which can be explained by the role of portfolio equity and FDI stocks for these countries. This can be consistent with Manova (2009), who shows that the effects of liberalizations are more pronounced in economies with initially less active stock markets. Foreign equity flows may therefore substitute for an underdeveloped domestic financial system. Also, Kose et al. (2009) found that, in general, FDI and portfolio equity stocks seem to improve risk sharing outcomes while debt stocks have the opposite effect. Focusing on Tables 6 and 7, there is one remarkable change in the estimates of dynamic models. Increases in government size bring about a negative effect on economic growth in industrial economies, consistent with the Solow (1956) model in which the savings rate makes the production function shift downwards for a given stock of capital. Kneller et al. (1999) find important effects of certain government expenditures on growth for the 22 OECD countries. Under GMM estimations, Bekaert et al. (2005) find G/Y has mixed effects: not statistically significant for a full sample of 95 countries but negative and statistically significant (from -0.031 to -0.038) for another sample of 76 countries with a different indicator of capital account openness.

21 In this paper the magnitude of the β4-coefficient on government size is very close across specifications but always negative for industrial countries: at -0.098 when there are no capital flows and varying from -0.073 to -0.104 depending on each OPEN series is included. No such negative effect is found for emerging markets, which support Adam and Bevan (2005), who pointed out that for an economy not on its steady-state growth path there is a range over which deficit-financing may be growth enhancing.13 [Tables 6 and 7 here] On robustness issues, we allowed for several interaction effects between financial globalization measures themselves and also between them or trade together with G/Y. Except for some very minor statistically significant values in a few cases, the interactions turnout out not to be statistically significant and are omitted in this paper. A further modification is to allow a different stochastic process for the regressors. In particular, G/Y is now assumed to be exogenous in Table 8. The negative effect of G/Y on output growth becomes amplified for industrial economies, varying from -0.232 without openness to between -0.187 for GEQ in column (3) to -0.205 for trade openness in column (4). With this modification, the previous no-result on the G/Y coefficient for emerging markets is maintained at the right of Table 8: the β4-coefficients are always not statistically significant for emerging markets. In other words, the negative effect of G/Y on output growth in fact becomes larger if we relax the endogeneity assumption. Allowing for exogenous stochastic processes in Table 8, we continue to find negative effects only for industrial countries. 13

Adam and Bevan (2005) allow for the marginal effect of the fiscal deficit on growth to vary around a threshold value of the deficit for a panel of 45 developing countries over 1970-1999. They also find that fiscal deficits may be growth enhancing if financed by limited seigniorage; they are likely to be growthinhibiting if financed by domestic debt; and to have opposite flow and stock effects if financed by external loans at market rates.

22 [Table 8 here] How to rationalize the contrasting patterns of government size on output growth reported in this research? One way to interpret this is as follows: for given I/Y and n, when capital inflows increase an expansionary fiscal policy causes G/Y to rise and then savings to decrease in industrial economies. This is the standard channel in which foreign financing replaces domestic savings. On the other hand, when capital inflows increase in emerging markets economies savings also increase because capital inflows complement domestic savings. This result is obtained in Verdier (2008) by modifying the neoclassical model with a credit constraint. In addition to the theoretical effect on the savings rate discussed in Section 2 along Solow (1956) and the extensions to the open-economy by Barro et al. (1995) and Verdier (2008), at least two additional possibilities are available to rationalize our contrasting output effects of G/Y. Rodrik (1998) argues that government spending plays a risk-reducing role in economies exposed to a significant amount of external risk. The implication is that the relationship between openness and government size is strongest when external risk (terms of trade shocks, capital flight) is highest. So, if government has any role in risk reduction when capital and goods and services flows are present, the evidence in this paper suggests that it does so in a dual way for the globalization years: very negatively for industrial economies and ambiguously for emerging markets. Another explanation is that emerging markets are not fully integrated into the world economy and therefore can not exploit all opportunities as do industrial countries. In this case, there could be a larger role for government services in help achieving a higher degree of risk sharing and, therefore, of economic growth in developing countries.

23 Kose et al. (2009) compile a dataset for 69 countries (21 industrial and 48 developing) over 1960 to 2004 and estimate the effect of changes in per capita consumption of a country with respect to changes in per capita world consumption on a regressor containing a difference between changes in country’s GDP with respect to changes in world GDP. The difference between the national and common world component captures the country-specific fluctuations. Kose et al. (2009) find that the risk-sharing coefficients vary substantially across countries, being the lowest for their emerging markets panel.14

6. Concluding Remarks One of the advantages of open economies is that they need not save in order to accumulate capital because they can borrow. In this setting, domestic savings and foreign investment can be complements as in the theoretical construction by Verdier (2008), based on the capital mobility model of Barro et al. (1995) and on the seminal Solow (1956) model. Since G affects the savings rate of the economy, this paper investigates the effects of government policies on output per worker when both types of capital (domestic and foreign) are present under the globalization years of 1986-2004. Sample correlations, bivariate charts, and a combination of fixed effects and dynamic panel data models suggests systematic positive and significant effects of

14

Kose et al. (2009) show that the extent of risk sharing appears to be higher in industrial countries than in developing countries. Moreover, interacting the difference in GDP with the measures of financial openness, they find for the globalization years that financial integration appears to have no significant impact in the globalization period since the interactive terms are not statistically significant. The risk sharing benefits of financial integration thus seem to have accrued mainly to industrial countries over the period. At a theoretical level, Devereux and Sutherland (2009) analyze two-way capital flows between the economies and find that a configuration where the emerging economy holds nominal bonds and issues claims on FDI can achieve a considerable degree of international risk-sharing. Under this setting, the evidence in this paper suggests that the role of government in coordinating these activities of enhancing risk-sharing has deleterious effects in industrial economies but ambiguous effects in emerging markets. A full exploration of this channel is left for further research.

24 openness on real output growth, with the current account-based openness channel turning out to be stronger than the capital-account one. We also observe strong negative long-run effects of G/Y on output growth in industrial economies and ambiguous effects of government size on output in emerging markets. The result for industrial economies is consistent with the standard channel in which foreign capital replaces domestic savings in industrial economies. On the other hand, when capital inflows increase savings may also increase in emerging markets because foreign capital complements domestic savings.

25 References Abel, Andrew, Ben Bernanke, and Dean Croushore (2008). Macroeconomics. 6th Edition. Pearson, Addison-Wesley: Boston, MA. Acemoglu, Daron (2005). Politics and Economics in Weak and Strong States. Journal of Monetary Economics 52: 1199-1226. Adam, Christopher, and David Bevan (2005). Fiscal Deficits and Growth in Developing Countries. Journal of Public Economics 89: 571-597. Aizenman, Joshua (2008). On the Hidden Links between Financial and Opening. Journal of International Money and Finance 27: 372-386. Alesina, Alberto, and Romain Wacziarg (1998). Openness, Country Size, and Government. Journal of Public Economics 69: 305-321. Arellano, Manuel and Stephen Bond (1991). Some Test of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations. Review of Economic Studies 58 (2): 277-297. Astorga, Pablo (2009). A Century of Economic Growth in Latin America. Journal of Development Economics, forthcoming. Bajo-Rubio, Oscar (2000). A Further Generalization of the Solow Growth Model: The Role of the Public Sector. Economics Letters 68: 79-84. Baltagi, Badi, Panicos Demetriades, and Siong Hook Law (2009). Financial Development and Openness: Evidence from Panel Data. Journal of Development Economics, forthcoming. Barro, Robert J., N. Gregory Mankiw, and Xavier Sala-I-Martin (1995). Capital Mobility in Neoclassical Models of Growth. American Economic Review 85 (1): 103-115. Barro, Robert J. (1990). Government Spending in as Simple Model of Endogenous Growth. Journal of Political Economy 98 (5): S103-S125. Beaudry, Paul, and Fabrice Collard (2006). Globalization, Returns to Accumulation and the World Distribution of Output. Journal of Monetary Economics 53: 879-909. Bekaert, Geert, Campbell Harvey, and Christian Lundblad (2005). Does Financial Liberalization Spur Growth? Journal of Financial Economics 77: 3-55. Benarroch, Michael, and Manish Pandey (2008). Trade Openness and Government Size. Economics Letters 101 (3): 157-159.

26 Bowsher, Clive (2002). “On Testing Overidentifying Restrictions in Dynamic Panel Data Models”, Economic Letters 77 (2): 211-220. Bose, Niloy, M. Emranul Haque, and Denise Osborn (2007). Public Expenditure and Economic Growth: A Disaggregated Analysis for Developing Countries. The Manchester School 75 (5): 533-556. Chang, Roberto, Linda Kaltani, and Norman Loayza (2008). Openness can be Good for Growth: The Role of Policy Complementarities. Journal of Development Economics, forthcoming. Combes, Jean-Louis, and Tahsin Saadi-Sedik (2006). How does Trade Openness Influence Budget Deficits in Developing Countries? Journal of Development Studies 42 (8): 1401-1416. De Gregorio, José (1992). Economic Growth in Latin America. Journal of Development Economics 39 (1): 59-84. Devereux, Michael, and Alan Sutherland (2009). “A Portfolo Model of Capital Flows to Emerging Markets”, Journal of Development Economics, forthcoming. Edison, Hali, Ross Levine, Luca Ricci, and Torsten Sløk (2002). International Financial Integration and Economic Growth. Journal of International Money and Finance 21: 749776. Garen, John, and Kathleen Trask (2005). Do More Open Economies Have Bigger Governments? Another Look. Journal of Development Economics 77: 533-551. Greenaway, David, Wyn Morgan, and Peter Wright (2002). Trade Liberalisation and Growth in Developing Countries. Journal of Development Economics 67: 229-244. Henry, Peter Blair (2007). Capital Account Liberalization: Theory, Evidence, and Speculation. Journal of Economic Literature 45 (4): 887-935. Kneller, Richard (2007). No Miracles Here: Trade Policy, Fiscal Policy and Economic Growth. Journal of Development Studies 43 (7): 1248-1269. Kneller, Richard, Michael Bleaney, and Norman Gemmell (1999). Fiscal Policy and Growth: Evidence from OECD Countries. Journal of Public Economics 74: 171-190. Kose, M. Ayhan, Eswar Prasad, and Marco Terrones (2009). “Does Financial Globalization Promote Risk Sharing?”, Journal of Development Economics, forthcoming. Kose, M. Ayhan, Eswar Prasad, Kenneth Rogoff, and Shang-Jin Wei (2006). “Financial Globalization: A Reappraisal”, IMF Working Paper 06/189.

27 Kraay, Aart, and Claudio Raddatz (2007). Poverty Traps, Aid, and Growth. Journal of Development Economics 82: 315-347. Lane, Philip, and Gian Maria Milesi-Ferretti (2007). The External Wealth of Nations Mark II: Revised and Extended Estimates of Foreign Assets and Liabilities. Journal of International Economics 73: 223-250. Lee, Ha-Yan, Luca A. Ricci, and Roberto Rigobon (2004). Once Again, Is Openness Good for Growth? Journal of Development Economics 75: 451-472. Mankiw, Gregory, David Romer, and David Weil (1992). A Contribution to the Empirics of Economic Growth. Quarterly Journal of Economics 107 (2): 407-437. Manova, Kalina (2008). Credit Constraints, Equity Market Liberalizations and International Trade. Journal of International Economics 76 (1): 33-47. Mishkin, Frederic (2009). Globalization and Economic Development. Journal of Development Economics, forthcoming. Mueller, Dennis, and Thomas Stratmann (2003). The Economic Effects of Democratic Participation. Journal of Public Economics 87: 2129-2155. Nonneman, Walter, and Patrick Vanhoudt (1996). A Further Augmentation of the Solow Model and the Empirics of Economic Growth for OECD Countries. Quarterly Journal of Economics 111 (3): 943-953. Quinn, Dennis, and A. Maria Toyoda (2008). Does Capital Account Liberalization Lead to Growth? Review of Financial Studies 21 (3): 1403-1449. Rajan, Raghuram, and Luigi Zingales (2003). The Great Reversals: The Politics of Financial Development in the 20th Century. Journal of Financial Economics 69 (1): 5-50. Ram, Rati (2009). Openness, Country Size, and Government Size: Additional Evidence from a Large Cross-Country Panel, Journal of Public Economics 93 (1-2): 213-218. Rayp, Glenn, and Nicolas Van De Sijpe (2007). Measuring and Explaining Governments Efficiency in Developing Countries. Journal of Development Studies 43 (2): 360-381. Rodrik, Dani (1998). Why do More Open Economies Have Bigger Governments? Journal of Political Economy 106 (5): 997-1032. Solow, Robert M. (1956). A Contribution to the Theory of Economic Growth. Quarterly Journal of Economics 70(1): 65-94. Svaleryd, Helena, and Jonas Vlachos (2002). Markets for Risk and Openness to Trade: How are they Related? Journal of Development Economics 57: 369-395.

28

Verdier, Geneviève (2008). What Drives Long-Term Capital Flows? A Theoretical and Empirical Investigation. Journal of International Economics 74: 120-142. Yanikkaya, Halit (2003). Trade Openness and Economic Growth: A Cross-Country Empirical Investigation. Journal of Development Economics 72 (1): 57-89.

29 Figure 1. Per capita GDP Growth and its Determinants

30

Table 1. Descriptive Statistics Industrial Countries ∆(Y/L) I/Y G/Y

Country

N

AUSTRALIA AUSTRIA BELGIUM CANADA DENMARK FINLAND FRANCE GERMANY GREECE ICELAND IRELAND ITALY JAPAN NETHERLANDS NEW ZEALAND NORWAY PORTUGAL SPAIN SWEDEN SWITZERLAND UNITED KINGDOM UNITED STATES

19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19

2.05 1.97 2.07 1.70 1.56 2.03 1.74 2.19 1.97 2.04 5.47 1.74 1.69 2.32 1.99 2.28 4.28 2.87 1.96 0.89 2.37 1.96

24.33 22.12 19.94 19.69 19.49 20.87 19.31 20.69 19.92 20.16 19.24 20.19 27.56 21.40 20.66 21.34 25.00 21.12 18.12 24.13 17.76 18.75

418

2.23

20.99 -0.03 -0.43

Total/Average Correl (∆Y/L,X) Correl (G/Y,X)

IFI

GEQ

TO

Country

Emerging Market Economies N ∆(Y/L) I/Y G/Y

IFI

GEQ

TO

18.37 18.99 21.64 20.16 25.66 22.04 22.99 19.30 14.51 21.77 14.75 19.11 15.22 23.61 18.17 21.33 18.30 17.23 27.17 11.41 20.01 15.77

1.40 2.02 4.95 1.69 2.36 2.00 2.23 1.75 1.17 1.24 7.21 1.32 1.02 4.10 1.72 1.65 1.94 1.43 2.59 6.22 4.74 1.19

0.70 0.31 1.38 0.82 0.64 0.78 0.85 0.46 0.18 0.22 2.54 0.33 0.18 1.74 0.76 0.50 0.40 0.51 1.22 2.23 1.21 0.56

0.37 0.78 1.43 0.66 0.74 0.61 0.47 0.57 0.45 0.70 1.39 0.44 0.19 1.13 0.58 0.71 0.64 0.46 0.71 0.75 0.53 0.22

19.43 -0.16

2.54 0.35 -0.16

0.84 0.29 -0.11

0.66 0.41 0.21

ARGENTINA BANGLADESH BRAZIL BULGARIA CHILE CHINA COLOMBIA COSTA RICA CZECH REPUBLIC DOMINICAN REPUBLIC ECUADOR EGYPT EL SALVADOR HUNGARY INDIA INDONESIA ISRAEL KOREA MALAYSIA MEXICO MOROCCO NIGERIA PAKISTAN PANAMA PERU PHILIPPINES POLAND SINGAPORE SOUTH AFRICA THAILAND TUNISIA TURKEY URUGUAY VENEZUELA

19 19 19 14 19 19 19 19 12 19 19 19 15 19 19 19 19 19 19 19 19 18 19 19 19 19 19 19 19 19 19 18 19 19

1.04 2.55 1.07 1.91 4.40 8.41 1.47 2.25 3.26 2.82 0.97 2.45 3.34 2.11 3.39 3.16 1.53 5.90 3.91 1.08 1.85 1.04 1.90 1.54 0.75 1.48 3.01 4.39 0.22 5.09 2.57 2.15 2.10 -0.00

17.13 18.07 18.60 15.79 22.40 32.58 18.24 19.11 28.49 20.26 20.00 21.27 16.53 21.89 22.85 24.83 20.55 32.92 30.62 19.09 22.19 8.64 16.65 17.10 19.48 20.14 20.12 32.53 16.71 31.39 24.34 22.63 12.84 20.19

12.80 5.16 17.90 16.99 11.16 14.49 15.75 14.23 21.38 5.61 10.72 11.76 9.58 10.77 11.67 8.15 29.00 11.56 12.83 10.28 17.67 6.31 11.93 15.38 9.30 10.86 17.54 10.42 19.10 10.53 16.05 12.07 12.46 10.61

1.13 0.48 0.71 1.71 1.50 0.60 0.79 1.03 1.26 0.78 1.21 1.02 0.85 1.14 0.39 1.03 1.39 0.65 1.69 0.82 1.23 1.77 0.67 4.69 1.04 1.19 0.83 5.74 0.88 1.07 1.38 0.74 1.52 1.33

0.26 0.03 0.25 0.15 0.60 0.21 0.17 0.27 0.38 0.24 0.24 0.23 0.16 0.32 0.06 0.11 0.33 0.15 0.71 0.24 0.19 0.52 0.08 0.50 0.22 0.20 0.12 2.45 0.55 0.28 0.61 0.07 0.12 0.27

0.22 0.26 0.19 1.03 0.61 0.42 0.36 0.82 1.17 0.75 0.55 0.48 0.60 0.97 0.22 0.57 0.77 0.66 1.75 0.49 0.52 0.61 0.34 1.52 0.31 0.82 0.51 3.54 0.49 0.92 0.89 0.46 0.42 0.50

628

2.50

21.36 0.75 0.07

13.00 -0.10

1.30 0.01 0.03

0.33 0.16 0.00

0.73 0.25 0.03

Total/Average Correl (∆Y/L,X) Correl (G/Y,X)

Notice: For the correlation estimates X represents each of the explanatory varibles listed on each respective column: (I/Y), IFI, GEQ and TO. Bulgaria’s GDP is only available since 1991. The Czech Republic, El Salvador, Nigeria and Turley also present some missing data.

31

Table 2. Fixed Effects Model of GDP Per-capita: Post-1986 Years for the Full Sample Yit = β0 + β1 Yit-1 + β2 Zit + β3 Oit + β4 (G/Y)t + εit , O = IFI, GEQ, TO

∆L/L ln(I/Y)

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

-12.545*** (3.524) 0.046 (0.094)

-8.709*** (2.150) 0.281*** (0.034) 0.282*** (0.016)

-1.357 (1.765) 0.181*** (0.031)

-4.161* (2.454) 0.162*** (0.037)

-11.140*** (3.207) 0.152*** (0.044)

-8.596*** (2.158) 0.286*** (0.035) 0.284*** (0.016)

-1.559 (1.748) 0.175*** (0.032)

-3.904 (2.444) 0.168*** (0.038)

ln(IFI) ln(GEQ)

0.161*** (0.011)

ln(TO)

0.164*** (0.011) 0.470*** (0.041)

ln(G/Y) constant

R2 within

8.887*** (0.277)

8.176*** (0.108)

8.626*** (0.092)

9.191*** (0.130)

0.034 (0.042) 8.519*** (0.176)

0.038

0.471

0.544

0.317

0.067

0.040 (0.041) 8.035*** (0.176)

(0.058) (0.036) 8.764*** (0.142)

0.480*** (0.040) 0.054 (0.041) 9.012*** (0.190)

0.473

0.547

0.320

Notes: Logarithms are taken on output, globalization, investment/output and government spending/output series. Newey-West standard errors robust to autocorrelation and heteroskedasticity are reported in parentheses. The symbols *, ** and *** refer to levels of significance of 10%, 5% and 1%.

32 Table 3. Fixed Effects Model of GDP Per-capita: Post-1986 Years for Industrial Economies Yit = β0 + β1 Yit-1 + β2 Zit + β3 Oit + β4 (G/Y)t + εit , O = IFI, GEQ, TO

∆L/L ln(I/Y)

(1)

(2)

(3)

(4)

(7)

(8)

(9)

(10)

19.204*** (4.621) -0.019 (0.080)

5.274*** (1.274) 0.228*** (0.038) 0.266*** (0.009)

9.965*** (1.674) 0.244*** (0.041)

14.746*** (2.866) 0.210*** (0.068)

19.030*** (4.577) -0.038 (0.092)

5.261*** (1.263) 0.226*** (0.043) 0.266*** (0.009)

9.757*** (1.633) 0.223*** (0.048)

14.966*** (2.821) 0.263*** (0.071)

ln(IFI) ln(GEQ)

0.162*** (0.008)

ln(TO)

0.162*** (0.008) 0.754*** (0.038)

ln(G/Y) constant

R2 within

10.147*** (0.275)

9.465*** (0.119)

9.498*** (0.129)

10.224*** (0.198)

-0.109 (0.187) 10.528*** (0.737)

0.127

0.852

0.813

0.583

0.130

-0.011 (0.075) 9.505*** (0.310)

-0.127 (0.083) 9.939*** (0.348)

0.786*** (0.039) 0.256*** (0.095) 9.335*** (0.390)

0.852

0.817

0.597

Notes: Logarithms are taken on output, globalization, investment/output and government spending/output series. Newey-West standard errors robust to autocorrelation and heteroskedasticity are reported in parentheses. The symbols *, ** and *** refer to levels of significance of 10%, 5% and 1%.

33

Table 4. Fixed Effects Model of GDP Per-capita: Post-1986 Years for Emerging Markets Yit = β0 + β1 Yit-1 + β2 Zit + β3 Oit + β4 (G/Y)t + εit , O = IFI, GEQ, TO

∆L/L ln(I/Y)

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

-22.466*** (5.137) 0.073 (0.113)

-14.038*** (3.914) 0.282*** (0.043) 0.240*** (0.044)

-6.885** (3.095) 0.173*** (0.036)

-14.033*** (3.912) 0.187*** (0.039)

-20.759*** (4.618) 0.205*** (0.043)

-13.737*** (3.916) 0.287*** (0.044) 0.248*** (0.044)

-6.832** (3.107) 0.168*** (0.036)

-13.778*** (3.887) 0.189*** (0.040)

ln(IFI) ln(GEQ)

0.142*** (0.019)

ln(TO)

0.146*** (0.020) 0.315*** (0.046)

ln(G/Y) constant

R2 within

8.933*** (0.332)

8.241*** (0.139)

8.685*** (0.105)

9.001*** (0.132)

0.03 (0.041) 8.490*** (0.178)

0.101

0.348

0.435

0.34

0.226

0.036 (0.043) 8.108*** (0.207)

-0.046 (0.039) 8.794*** (0.156)

0.323*** (0.045) 0.035 (0.038) 8.893*** (0.186)

0.349

0.464

0.34

Notes: Logarithms are taken on output, globalization, investment/output and government spending/output series. Newey-West standard errors robust to autocorrelation and heteroskedasticity are reported in parentheses. The symbols *, ** and *** refer to levels of significance of 10%, 5% and 1%.

34

Table 5. Dynamic Model of GDP Per-capita: Post-1986 Years for the Full Sample Yit = β0 + β1 Yit-1 + β2 Zit + β3 Oit + β4 (G/Y)t + εit , O = IFI, GEQ, TO

ln(Yt-1) ∆L/L ln(I/Y)

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

0.693*** (0.141) -15.573 (11.457) 0.042 (0.033)

0.287 (0.197) -6.917 (5.174) 0.198*** (0.059) 0.268*** (0.084)

0.098 (0.097) -2.746 (2.492) 0.141*** (0.031)

0.344 (0.244) -10.641 (8.382) 0.097** (0.039)

0.626*** (0.224) -9.733 (6.223) 0.099** (0.049)

0.474** (0.233) -6.559* (3.796) 0.188** (0.080) 0.166** (0.080)

0.382 (0.246) -3.748 (3.127) 0.122*** (0.045)

0.538** (0.244) -4.873 (3.552) 0.089** (0.045)

ln(IFI) ln(GEQ)

0.166*** (0.025)

ln(TO)

0.103*** (0.040) 0.454** (0.192)

ln(G/Y) cons

3.542** (1.745)

7.480*** (2.069)

10.018*** (1.077)

7.516*** (2.806)

0.077 (0.079) 3.876* (2.306)

Sargan Test

55.55 [0.491] -1.30 (0.193) 0.24 (0.813)

55.11 [0.508] -1.63 (0.102) -0.54 (0.586)

55.37 [0.498] 0.37 (0.709) 1.14 0.254

54.90 [0.516] -2.12 (0.033) 0.71 (0.480)

55.5 [1] -2.61 (0.009) -1.19 (0.233)

AB(1) AB(2)

0.064 (0.063) 5.264** (2.315)

0.011 (0.060) 6.787** (2.698)

0.238* (0.127) 0.103 (0.085) 4.900* (2.579)

54.27 [1] -3.23 (0.001) -1.27 (0.205)

53.63 [1] -1.35 (0.176) 0.44 (0.653)

54.86 [1] -3.48 (0.001) -0.52 (0.603)

Notes: Logarithms are taken on output, globalization, investment/output and government spending/output series. The Table reports the first-step estimators of the Arellano and Bond (1992). The Sargan test reports that under the null the overidentified restrictions are valid. AB (1) and AB (2) correspond to the Arellano-Bond test for serial correlation, under the null of no autocorrelation. Robust standard errors are reported in parenthesis. The symbols *, **, and *** refer to levels of significance of 10%, 5%, and 1%, respectively.

35

Table 6. Dynamic Model of GDP Per-capita: Post-1986 Years for the Industrial Economies Yit = β0 + β1 Yit-1 + β2 Zit + β3 Oit + β4 (G/Y)t + εit , O = IFI, GEQ, TO (1) ln(Yt-1) ln(Yt-2) ∆L/L ln(I/Y)

1.312*** (0.058) -0.316*** (0.060) -2.392*** (0.459) 0.015 (0.025)

ln(IFI)

(2) 1.149*** (0.056) -0.273*** (0.055) -1.988*** (0.547) 0.034 (0.022) 0.038*** (0.008)

ln(GEQ)

(3) 0.985*** (0.067) -0.187*** (0.061) -1.364* (0.696) 0.051*** (0.020)

(4) 1.157*** (0.063) -0.205*** (0.061) -1.887*** (0.528) 0.028 (0.022)

(7) 1.229*** (0.068) -0.247*** (0.065) -0.901* (0.514) -0.006 (0.017)

(9) 1.095*** (0.062) -0.208*** (0.051) -0.140 (0.584) 0.025** (0.013)

(10) 1.153*** (0.057) -0.202*** (0.054) -0.531 (0.475) 0.016 (0.018)

0.021*** (0.003) 0.070*** (0.021)

ln(G/Y) constant

0.031 (0.168)

1.422*** (0.291)

2.391*** (0.329)

0.572*** (0.211)

-0.098*** (0.027) 0.520*** (0.171)

Sargan Test

21.331 [1] -2.654 (0.008) -1.226 (0.220)

20.554 [1] -2.488 (0.013) -0.515 (0.607)

20.616 [1] -2.246 (0.025) -0.157 (0.875)

20.954 [1] -2.027 (0.043) -1.323 (0.186)

20.099 [1] -3.179 (0.002) -1.375 (0.169)

AB(2)

1.157*** (0.061) -0.261*** (0.053) -0.573 (0.422) 0.020 (0.018) 0.030*** (0.007)

0.038*** (0.005)

ln(TO)

AB(1)

(8)

-0.080*** (0.020) 1.438*** (0.178)

-0.104*** (0.018) 1.630*** (0.201)

0.058*** (0.013) -0.073*** (0.026) 0.801*** (0.143)

20.660 [1] -3.040 (0.002) -0.697 (0.485)

19.870 [1] -3.063 (0.002) -0.696 (0.487)

19.417 [1] -2.839 (0.005) -1.338 (0.181)

Notes: Logarithms are taken on output, globalization, investment/output and government spending/output series. The Table reports the first-step estimators of the Arellano and Bond (1992). The Sargan test reports that under the null the overidentified restrictions are valid. AB(1) and AB(2) correspond to the Arellano-Bond test for serial correlation, under the null of no autocorrelation. Robust standard errors are reported in parenthesis. The symbols *, **, and *** refer to levels of significance of 10%, 5%, and 1%, respectively.

36

Table 7. Dynamic Model of GDP Per-capita: Post-1986 Years for Emerging Economies Yit = β0 + β1 Yit-1 + β2 Zit + β3 Oit + β4 (G/Y)t + εit , O = IFI, GEQ, TO

ln(Yt-1) ∆L/L ln(I/Y)

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

0.701*** (0.116) -7.584*** (3.885) 0.096*** (0.028)

0.267 (0.210) -22.025 (13.626) 0.273*** (0.088) 0.163 (0.106)

0.069 (0.077) -7.482 (5.764) 0.175*** (0.050)

0.190 (0.182) -17.517 (11.398) 0.196*** (0.061)

0.513** (0.245) -13.023* (7.523) 0.137** (0.061)

0.469** (0.237) -9.725* (5.649) 0.192** (0.088) 0.134 (0.088)

0.386 (0.249) -6.210 (4.419) 0.126*** (0.048)

0.467* (0.250) -9.331* (5.637) 0.120** (0.054)

ln(IFI) ln(GEQ)

0.159*** (0.037)

ln(TO)

0.086** (0.036) 0.372** (0.161)

ln(G/Y) Constant

3.027** (1.217)

7.328*** (2.195)

9.740*** (0.833)

8.521*** (1.920)

0.035 (0.054) 4.909** (2.477)

Sargan Test

32.73 [0.996] -1.24 (0.213) 0.66 (0.510)

32.42 [0.995] -1.42 (0.155) 0.14 (0.888)

31.72 [0.996] 0.26 (0.798) 0.91 (0.364)

33.12 [0.994] -1.66 (0.097) 1.00 (0.316)

32.49 [1] -2.16 (0.305) -0.67 (0.503)

AB(1) AB(2)

0.038 (0.055) 5.148** (2.275)

-0.015 (0.053) 6.444** (2.611)

0.171* (0.087) 0.049 (0.055) 5.445** (2.559)

30.25 [1] -2.76 (0.006) -1.15 (0.249)

28.93 [1] -1.58 (0.114) 0.25 (0.806)

31.4 [1] -2.98 (0.004) -0.42 (0.667)

Notes: Logarithms are taken on output, globalization, investment/output and government spending/output series. The Table reports the first-step estimators of the Arellano and Bond (1992). The Sargan test reports that under the null the overidentified restrictions are valid. AB (1) and AB (2) correspond to the Arellano-Bond test for serial correlation, under the null of no autocorrelation. Robust standard errors are reported in parenthesis. The symbols *, **, and *** refer to levels of significance of 10%, 5%, and 1%, respectively.

37 Table 8. Dynamic Model of GDP Per-capita: Post-1986 Years for the Industrial and Emerging Economies

(1) ln(Yt-1) ln(Yt-1) ∆L/L ln(I/Y)

1.153*** (0.061) -0.145** (0.062) -2.620*** (0.524) -0.021 (0.024)

ln(IFI)

IEs (2) 1.059*** (0.050) -0.138*** (0.052) -2.300*** (0.479) -0.002 (0.025) 0.027** (0.011)

ln(GEQ)

(3)

(4)

(1)

0.935*** (0.063) -0.082* (0.048) -1.790*** (0.550) 0.014 (0.021)

1.079*** (0.040) -0.099** (0.041) -2.283*** (0.430) -0.009 (0.024)

0.286 (0.242)

-28.201* (15.251) 0.220*** (0.070)

cons

Sargan Test AB(1) AB(2)

(3)

(4)

0.261 (0.206)

0.069 (0.077)

0.182 (0.175)

-21.448 (13.426) 0.265*** (0.085) 0.167 (0.104)

-7.669 (5.842) 0.170*** (0.048)

-16.085 (10.908) 0.191*** (0.060)

0.029*** (0.005)

ln(TO) ln(G/Y)

EMEs (2)

0.160*** (0.038)

-0.232*** (0.047) 0.639** (0.289)

-0.201*** (0.047) 1.547*** (0.339)

-0.187*** (0.043) 2.320*** (0.327)

0.042** (0.019) -0.205*** (0.050) 0.892*** (0.261)

21.75 [1]

20.48 [1]

20.6 [1]

20.85 [1]

0.007 (0.043) 7.388*** (2.592)

-0.012 (0.038) 7.444*** (2.155)

-0.047 (0.031) 9.898*** (0.829)

0.384** (0.157) 0.065 (0.066) 8.446*** (1.825)

30.1 [0.998] -1.08 (0.278) 1.22 (0.222)

30.85 [0.997] -1.44 (0.149) 0.23 (0.815)

30.31 [0.998] 0.25 (0.801) 1.12 (0.260)

31.20 [0.997] -1.75 (0.081) 1.04 (0.301)

Notes: The Sargan test reports that under the null the overidentified restrictions are valid. AB (1) and AB (2) correspond to the Arellano-Bond test for serial correlation, under the null of no autocorrelation. Robust standard errors are reported in parenthesis. The symbols *, **, and *** refer to levels of significance of 10%, 5%, and 1%, respectively.