Applied Econometrics and International Development Vol. 141 (2014) ON THE IMPACT OF PUBLIC DEBT ON ECONOMIC GROWTH DAR, Atul A. * AMIRKHALKHALI, Sal ABSTRACT It is generally agreed that the rapid rise of public debt in most of the developed countries can be traced to the recent financial crisis, and the accompanying slump contributed to this problem. This has sparked debate about the likely adverse macroeconomic impacts of persistent large debts especially on longterm economic growth. In this paper, we attempt to contribute to this debate by examining the impact of public debt on economic growth in 23 OECD countries classified into four groups in terms of their average debttogdp ratio over the 19962007 period. We use a general empirical methodology, which is also likely to be better able to represent the law relating economic growth to its determinants. Our empirical results indicate that the marginal impact of debt is negative but very small and statistically insignificant in almost all cases. 1. Introduction The rapid rise in public debt in many developed countries over the past several years has sparked debate about the likely macroeconomic impacts, especially on longterm economic growth. While it is generally agreed that the recent problem of debt can be traced to the financial crisis of recent years, the accompanying slump contributed to this problem; and, there is fear that large debts will persist because they reflect underlying structural factors. As a consequence, adverse economic consequences are much more likely to occur. In general, theoretical considerations point to a negative impact of debt on growth and some of the empirical literature appears to support that view (Schclarek, 2004). However, more commonly, the empirical literature suggests that debt impacts adversely on growth only after a threshold is reached see for instance, Reinhart and Rogoff (2010, 2011), Caner et al. (2010), Kumar and Woo (2010), Checherita and Rother (2010), Patillo et al. (2002), and Clements et al. (2003). In general, the estimated threshold in the empirical studies cited above varies mostly because of different methodologies. For instance, Caner et al. (2010) considered 101 developing and developed countries over almost 30 years and found a threshold of 77 percent public debttogdp ratio, while Reinhart and Rogoff (2010), in their study of 44 advanced and emerging countries over a 200year period, found that public debt has adverse growth consequences only beyond a debttogdp ratio of 90 percent for these countries. (It should be noted that despite of acknowledging some errors in their 2010 paper, in a May 2013 errata, Reinhart and Rogoff have left their basic findings unchanged.) Much of the empirical growth literature uses some variant of the basic aggregate growthaccounting model developed by Solow (1956), with more generalized approaches incorporating theoretical insights offered by endogenous growth theory see, for instance, Barro (1991), Barro and SalaiiMartin (1992), Jones (1995), and Mankiw, Romer and Weil (1992), Dar and AmirKhalkhali (2003). An important feature in such studies is that * Atul A. Dar & Sal AmirKhalkhali, Saint Mary s University, Canada. Emails: atul.dar@smu.ca sal.amirkhalkhali@smu.ca
Applied Econometrics and International Development Vol. 141 (2014) economic growth is seen as depending not just on traditional variables like factor accumulation, but also upon various policy and/or institutional factors. Typically, in such models, various variables reflecting public policy, trade orientation, and institutional characteristics, are included as affecting growth via their impact on total factor productivity the Solow residual. In reality, much of the vast intercountry differences reflected in economic, political and other institutions are unobserved, but nevertheless are potentially important for growth. Our empirical methodology outlined below, is ideally suited to address this and other problems such as endogeneity. Our study modifies the growthaccounting model used by AmirKhalkhali and Dar (2012) to examine whether public debt plays a prominent role in explaining differences in growth rates in 23 OECD countries over the 19962007 interval. Unlike previous studies, which estimate their models using fixed and random effects methodology, we accommodate unobserved intercountry differences via random coefficients. Such an approach is a better description of the underlying law relating one variable to others in this context, the law relating growth to its determinants, including debt. We address the growthdebt issue in two ways. First, we estimate our model over the entire sample using random coefficients estimation, with debttogdp ratios as additional variables. Since the debt effects are countryspecific, we can use these to assess the nature of the relationship between growth effects and the size of debttogdp ratios. In the second approach, we classify countries into four groups according to the size of debttogdp ratios and estimate random coefficients models that are group specific. This would allow us to assess whether the effect of rising debt varies according to its size. A large debttogdp ratio can adversely impact on growth via distortions resulting from higher future taxes to pay interest payments, or via the buildup of more debt, which would have similar distortionary implications. It would seem therefore, the effects of debt or fiscal flow variables like government size are similar in terms of their impacts. The impact of both suggest adverse threshold affects. Thus, rising government size and debt may both have a positive (or zero) effect on growth at low levels, but a negative impact beyond some threshold. The rest of the paper is organized as follows. In Section II, we discuss the model, the data and the estimation techniques. Section III contains a discussion of our empirical findings, while Section IV concludes with a summary of the findings. 2. The empirical model, data & estimation As noted above, the theoretical literature generally points to an adverse impact on growth of public debt. The arguments in support of this are similar to those for the long run impact of fiscal variables such as government size. Essentially, an increase in the size debttogdp ratio, ceteris paribus, implies higher future taxes (on labour and capital). This can adversely impact on growth via their distortionary static and intertemporal effects on capital and labour markets. We have already noted that the empirical results on the effect of public debt point to the existence of threshold effects on growth; that is, the adverse effect on growth operate only beyond some high level of debt. The public debt threshold is generally found to exist at about 90 percent of GDP, but is likely lower for other types debt (e.g. external debt) and indeed, as Checherita and Rother (2010) find, it could be as low as 70 percent of GDP. 22
Dar, A.A., Amirkhalkhali, S. On The Impact of Public Debt on Economic Growth The existence of a threshold implies that debt would have a positive or neutral impact below that level. A positive impact can occur up to a point if public debt finances public capital, as noted by Checerita and Rother (2010), or because of complementarities between private and some types of public investment (e.g. infrastructure). As well, public debt can have a positive or neutral effect via a number of other channels. For instance, public debt can also loosen credit constraints on firms and households, and can thus crowdin private investment and hence promote growth (Woodford, 1990). However, there is considerable agreement that when debt reaches a high level and is persistent, as appears to be the case with many developed countries today, it would harm growth. This nonlinear impact of debt could be characteristic of large governments, since rising and persistently high levels of government spending drive government budget deficits and hence debt. Nevertheless, high budget deficits could also be the result of some significant untimely tax cuts. In light of this, the question of interest is whether the impact of public debt varies by its size. Specifically, does a rising debt have a positive impact on growth at low levels of its size and a negative impact at high levels? It is worth noting the rationale for a permanent growth impact of debt and other fiscal variables is provided by endogenous growth models. These models imply that long run economic growth can be permanently altered by policy action that permanently alters the levels of variables (such as investment rates in physical and human capital), or which affects the country s rate of technological advance. We turn next to the data used in this study. Our sample consists of data for twenty three industrialized countries which are members of OECD: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Italy, Japan, Luxembourg, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, the United Kingdom, and the United States. The data span the 19962007 interval and were obtained from various issues of Economic Outlook published by OECD and International Financial Statistics published by International Monetary Fund (IMF). Table 1 presents averages of annual growth rates of real GDP (GY), employment (GE), real gross fixed capital formation (GI), and real exports (GX) as well as the ratios of investment (IY), government spending (GSY), and public debt (PDY) to GDP for these countries over the 19962007 period. The table classifies the twenty three OECD countries into four groups in terms of their average size of debt to GDP ratios. As can be seen from the table, the average public debt ratio varies from a low of 17.8% for Group I countries (Luxembourg, Australia, Norway, Switzerland, and New Zealand), to a high of 109.6% for Group IV economies (Belgium, Italy, Greece, and Japan). Group II (Germany, Ireland, UK, Spain, Iceland, and Finland) and Group III (Portugal, Sweden, Netherlands, Austria, Denmark, France, Canada, and USA) comprise the two intermediate groups of countries with averages of about 39.9% and 57.6%, respectively. While Group I has also the smallest size of government, it is Group II that enjoys the highest average growth rate of GDP, investment and employment. Nevertheless, Group IV with the highest public debt to GDP ratio has the lowest average growth rate of real GDP during this period. Table 2 provides a groupwise bivariate correlation analysis for these macroeconomic aggregates. The results are somewhat mixed in particular in the case of the public debt ratio, PDY, which is seen to be negatively correlated with GY for Groups I 23
Applied Econometrics and International Development Vol. 141 (2014) and IV, but this is significant only in the case of Group I. There is a positive correlation between PDY and GSY except in the case Group IV which is negative and significant. PDY has also a negative correlation with GE and GI except in the case of Group II. The negative correlation of PDY and GE is only significant for Group I. The correlation between PDY and GX is positive but it is only significant in the case of Group II. The correlation between PDY and IY is negative and significant except for Group IV. It is also interesting to note the reported negative correlations between GSY and IY in all cases, which is consistent with the crowding out effect. However, this could also imply the fiscal policy response to sluggish investment during this period. In order to go beyond simple correlations, we employ a model that is a generalisation of the commonlyused growthaccounting model based on the concept of an aggregate production function. In this model, the rate of economic growth is a function of capital and labour accumulation and total factor productivity. The standard growth accounting model involves the following aggregate production function in growth form: GY it = 1 + 2 GK it + 3 GL it + X it γ + u it (1) where GY is the rate of growth of real GDP, GK is the rate of capital accumulation, GL is the rate of growth of labour, X is a set of other variables, γ the associated vector of coefficients, while u is the disturbance term. The subscripts i (i=1,2,...,n) and t (t=1,2,...,t) index the countries and time periods in the sample, respectively. The X variables can be seen as those observed variables that impact on growth through total factor productivity. Generalizations of the model are usually generated by identifying measurable variables that capture the economic and/or political structure of a country, and which affect growth via total factor productivity. In this study, our generalization models total factor productivity as depending upon the rate of export growth (GX) and the public debt to GDP ratio (PDY), as well as upon other unmeasured differences across countries. In the absence of capital stock data, we use the growth rate of investment (GI) as a valid proxy variable for GK. In fact, GI closely captures the wide fluctuations in investment activity compared to investment rate which changes only slowly and shows relatively less variation across countries. In this study, we also use the growth rate of employment (GE) rather than that of total labour force (GL) because, given the existence of persistent episodes of unemployment in these countries over the sample period, employment more accurately captures the extent of labour utilization. As a result, the above model can be rewritten as: GY it = 1 + 2 GI it + 3 GE it + α 4 GX it + α 5 PDY it + u it (2) In recent years, this type of model has become a popular one for studying the determinants of economic growth rate. However, a significant problem with a structural model such as (2) is the endogeneity of the explanatory variables. For instance, it is hard to ignore the endogeneity arising from the bidirectional relationship between GY and PDY. This means that a standard interpretation of the estimated parameters is problematic because of the correlation between explanatory variables and the error term. Further, in the above model, intercountry differences are assumed away by virtue of the 24
Dar, A.A., Amirkhalkhali, S. On The Impact of Public Debt on Economic Growth assumption that all coefficients are the same across countries. This is a questionable assumption a priori; at least one that should be treated as a testable proposition. We overcome both problems by adopting the more general random coefficients model which permits us to treat the fixedcoefficients assumption as a testable proposition. In addition, the random coefficients model can be seen as a refinement of the stochastic law relating economic growth to its main determinants [see Pratt and Schlaifer (1984, 1988)]. For this purpose, we postulate that GY it = 1 + 2 GI it + 3 GE it + α 4 GX it + α 5 PDY it + W it ' (3) here W is the set of excluded variables that along with those that are included are sufficient to determine GY. However, in the linear, deterministic law stated by (5), neither the slope coefficients nor W are unique in that they are sensitive to the parameterization chosen. To ensure uniqueness, we also postulate W it = 1i + 2i GI it + 3i GE it + 4i GX it + 5i PDY it + e it (4) Substituting (6) into (5) yields GY it = 1i + 2i GI it + 3i GE it + 4i GX it + 5i PDY it + v it (5) With 1i ' + 2i ' 3i ' 4i ' 5i ', and v it =e it '. Note that (5) is a random coefficients model, and that the disturbance is not the joint effect of excluded variables; instead, it is the joint effect of the remainder of the excluded variables after the effect of included variables has been factored out. Note also that although the included variables cannot be uncorrelated with every variable that affects GY, they can be uncorrelated with the remainder of every such variable (see Pratt and Schlaifer,1988). Thus, each of our explanatory variables can be uncorrelated with u, and this regression model can be taken to represent the law relating GY to its determinants. The varying coefficients model represented by (5) also accommodates intercountry heterogeneity. In this study, these regression models are estimated using Swamy and SwamyMehta random generalized least squares (RGLS) estimators. For more details of the RGLS estimation methods, see Swamy (1970), Swamy and Mehta (1975), and Swamy and Tavlas (1995, 2002). We first estimate the random coefficients model for the pooled sample. This allows debt to affect growth via its impact on total factor productivity. We then classify countries into four groups according to the relative size of public debt and estimate the model for each group separately. This allows the impact of debt and other determinants of growth to vary by the debt level; alternatively, this is equivalent to allowing public debt to affect growth via its impact on individual factor productivity and total factor productivity. 25
Applied Econometrics and International Development Vol. 141 (2014) 3. Empirical results Table 3 contains the results based on the pooled sample that is, 23 countries over the 19962007 period, estimated using the Swamy RGLS technique. The validity of the random coefficients model can be tested using the Swamy s gstatistic which follows a distribution, see Swamy (1970) for more details. Note first that the gstatistic is statistically significant at the 5% level, thereby vindicating the random coefficients model. The investment, employment, and export growth rates coefficients have the expected positive sign and are each statistically significant at the 5 percent level. The public debt coefficient is negative but not statistically significant. To assess whether and to what extent these aggregate results mask intercountry and/or groupspecific differences, we first look at the countryspecific and then groupwise estimates of the model, obtained using the SwamyMehta (1975) RGLS method. The countryspecific estimates are reported in Table 4. The investment growth rate has a statistically significant positive impact on the growth performance for all countries except Luxembourg. In the case of the employment growth rate, the positive impact is statistically significant for most countries in Group I but for only about half of countries in the other three groups. The export growth rate is also a significant contributor to growth performance for all countries except Australia, Norway (Group I), UK, Spain (Group II), Portugal (Group III), and Greece (Group IV). In the case of public debt ratio, the impact of PDY on growth is negative and statistically significant only in the cases of Luxembourg (Group I) and USA (Group III). For the rest of the countries, the size of the PDY coefficient is very small and its sign is positive (but insignificant) in almost half of the countries in each group. Table 5 presents the groupwise results using the random coefficients GLS estimator. The reported gstatistic is statistically significant at the 5% level for each group, thereby supporting the random coefficients model. The results show that the impacts of investment, employment and export growth rates are positive for all four groups. The contributions of investment and export growth rates are significant in all cases but the impact of employment growth rate is only statistically significant for Group III at the 10% level. The impact of public debt is negative for all groups but not statistically significant. Group I shows the largest negative impact of the public debt ratio. However, this group contains Luxembourg and also some relatively small countries, In light of their small size, the negative impact of debt for Group I is not surprising. The negative but small, insignificant impact of the debt ratio in Groups II and III is also not surprising because both groups are below or at par of the 60% limit (set by the European Union). What is interesting is the insignificance of the impact of the debt ratio in Group IV. In other words, these results appear not to lend support to the debt threshold hypothesis. As well, it is interesting to note that the marginal impact of debt is very small in all cases. 4. Summary and concluding remarks In this paper, we attempted to empirically investigate the impact of public debt on economic growth for a group of 23 industrialized countries within a production function framework in a much more general manner than previous studies. To this end, we classified the countries into four groups in terms of their public debttogdp ratio. We 26
Dar, A.A., Amirkhalkhali, S. On The Impact of Public Debt on Economic Growth also developed the estimating model so that it more accurately represented the law relating the rate of economic growth to its determinants. The random coefficients approach is particularly suited here since, as stressed by the political economy literature, institutional factors are likely to play a major role in explaining intercountry differences in growth rates. In other words, a random coefficients treatment appears to be a more reliable way of taking these factors into account than models that try to quantify them. We first estimated the random coefficients model for the pooled sample, allowing debt to affect growth via its impact on total factor productivity. We found that the public debt coefficient was negative but small and not statistically significant. In the case of the countryspecific estimates of the model, the impact of the debttogdp ratio on growth was mixed but still insignificant for all countries except Luxembourg and the USA. In these two cases, the debt coefficient was negative and statistically significant. We then estimated the groupwise model. Note that this allows the impact of debt and other determinants of growth to vary by the debt level; alternatively, this is equivalent to allowing public debt to affect growth via its impact on individual factor productivity and total factor productivity. Our groupspecific results were mixed. The impact of public debt was negative for all groups but not statistically significant. The negative but insignificant impact of the debt ratio for Groups I, II and III do not seem that surprising because these groups are below or at about the somewhat reasonable debt ratio limit of 60%. However, the insignificance of the impact of debt ratio in Group IV seems quite interesting. The small marginal impact of the debt ratio on growth would appear not to support the debt threshold hypothesis. These results have important fiscal policy implications. While fiscal discipline could be an important long term objective of policy makers, the timing of its implementation seems to be crucial. Within this context, our empirical results lend strong support to the view that at a time of sluggish growth, the appropriate fiscal policy would appear to be a demandoriented growth strategy to deal with demand uncertainties, and this might require more deficit spending and borrowing. Nevertheless, a key factor in being able to implement this stimulative policy would depend upon governments will, as well as their ability to borrow at reasonably affordable interest rates. References AmirKhalkhali, S. and A. Dar (2012), On Explaining InterCountry Differences in Economic Growth Rates: Does Regulatory Quality Matter? Journal of Applied Econometrics and International Development, Vol. 122, 2012, 4150. Barro, R. J. (1991), Economic Growth in a Cross Section of Countries, Quarterly Journal of Economics,106: 40733. Barro, R and X. SalaiMartin (1992), Convergence, Journal of Political Economy, 100: 22351. Caner, M., Thomas Grennes, and Friederike KöhlerGeib (2010), Finding the Tipping Point When Sovereign Debt Turns Bad, Policy Research Working Paper, WPS5391 (July), The World Bank. Chari, V.V., and Patrick J. Kehoe (2006), Modern Macroeconomics in Practice: How Theory is Shaping Policy, Journal of Economic Perspectives, 20(4):328. 27
Applied Econometrics and International Development Vol. 141 (2014) Checchetti, Stephen G., M.S. Mohanty, and Fabrizio Zamplolli (2011), The Real Effects of Debt, BIS Working Papers, No. 52, Bank for International Settlements. Checherita, C. and P. Rother (2010), The Impact of High and Growing Government Debt on Economic Growth: An Empirical Investigation for the Euro Area, Working Paper Series, No 1237 (August), European Central Bank. Clements, B., R. Bhattacharya and T.Q Nguyen (2003), External Debt, Public Investment, and Growth in LowIncome Countries, IMF Working Paper, 03/249. Dar, A. and S. AmirKhalkhali (2003), On the Impact of Trade Openness on Growth: Further Evidence from OECD Countries, Applied Economics, 2003, 35: 176166. Jones, C.I. (1995), Time Series Tests of Endogenous Growth Models, Quarterly Journal of Economics, 110: 495525. Kumar, M. and J. Woo (2010), Public Debt and Growth, IMF Working Paper, 10/174. Mankiw, N.D., P. Romer, and D. Weil (1992), A Contribution to the Empirics of Economic Growth, Quarterly Journal of Economics, 107: 40737. Patillo, C., H. Poirson, and L. Ricci (2002), External Debt and Growth, IMF Working Paper, 02/69. Pratt J. W. and R. Schlaifer (1988), On the Interpretation and Observation of Laws, Journal of Econometrics 39, 2352. Reinhart, Carmen M. and Kenneth S. Rogoff, Growth in a Time of Debt, American Economic Review, 100(2): 57378. Reinhart, Carmen M. and Kenneth S. Rogoff, The Forgotten History of Domestic Debt, Economic Journal, 121(552: 31950. Solow, R.M. (1956), A Contribution to the Theory of Economic Growth," Quarterly Journal of Economics, 70: 6594. Swamy, P.A.V.B., (l970), Efficient Inference in a Random Coefficients Regression Model. Econometrica 38, 3ll23. Swamy, P.A.V.B. and Mehta, J.S., (1975), Bayesian and NonBayesian Analysis of Switching Regressions and of Random Coefficient Regression Models. Journal of the American Statistical Association 70, 593602. Swamy, P.A.V.B.and G.S. Tavlas, (1995), Random Coefficient Models: Theory and Applications. Journal of Economic Surveys, 165196. Swamy, P.A.V.B., and G.S. Tavlas, (2002), Random Coefficient Models in Companion to Theoretical Econometrics, edited by B.H. Baltagi. Basil Blackwell, 410428. Woodford,M. (1990), Public Debt as Private Liquidity, American Economic Review, 50(2): 38288. 28
Dar, A.A., Amirkhalkhali, S. On The Impact of Public Debt on Economic Growth TABLE 1. Average Annual Growth of Output (GY), Employment (GE), Investment (GI), Exports (GX); and Government Spending (GSY), Investment (IY), and Public Debt to GDP ratios (PDY), 19962007 Groups GY GE GI GX GSY IY PDY Countries Group I Luxembourg 4.67 3.79 6.26 2.03 39.78 21.96 7.20 Australia 3.53 0.66 5.50 2.50 35.19 21.75 9.40 Norway 2.85 1.51 5.86 3.39 44.9 20.59 19.0 Switzerland 1.98 1.00 1.99 0.69 34.79 21.79 25.9 New Zealand 2.97 1.90 4.33 0.30 39.84 23.05 27.9 Average 3.20 1.77 4.79 0.97 38.90 22.90 17.85 Group II Germany 1.53 0.51 1.40 0.72 46.98 18.27 33.6 Ireland 6.46 1.13 7.00 3.53 35.02 24.63 34.9 UK 2.76 1.09 4.42 1.96 41.83 17.26 38.5 Spain 3.48 3.69 5.35 0.08 39.75 28.38 42.9 Iceland 4.40 3.93 9.38 0.52 43.76 24.80 43.2 Finland 3.55 1.42 5.05 0.35 50.85 20.93 46.9 Average 3.70 1.96 5.43 0.34 43.03 21.87 39.86 Group III Portugal 2.16 1.29 2.49 1.91 44.83 24.72 54.7 Sweden 2.76 1.11 4.64 0.19 56.67 18.38 55.7 Netherlands 2.75 1.76 3.58 0.30 46.18 20.20 56.4 Austria 2.50 2.08 1.54 0.03 51.95 23.01 57.4 Denmark 1.87 0.65 3.87 1.22 54.05 21.19 58.4 France 2.09 1.10 3.52 0.35 52.87 20.40 61.4 Canada 3.01 1.96 5.72 3.18 41.64 21.21 62.0 USA 2.98 1.19 3.87 1.54 35.82 19.53 62.7 Average 2.52 1.39 3.65 0.99 48.01 21.11 57.59 Group IV Belgium 2.20 1.36 3.26 1.71 50.18 21.51 97.8 Italy 1.28 1.80 2.16 4.48 48.55 21.03 107.6 Greece 3.74 1.42 5.75 0.18 44.96 22.04 107.4 Japan 1.16 0.08 0.15 1.45 37.92 23.75 125.7 Average 2.09 1.12 2.83 1.95 45.41 22.03 109.60 Overall Average 2.90 1.57 4.22 0.81 44.28 21.76 53.4 29
Applied Econometrics and International Development Vol. 141 (2014) TABLE 2. GroupWise Correlation Analysis Groups Variables GY GE GI GX GSY IY PDY Group I Group II Group III GY 1 GE 0.479* 1 GI 0.534* 0.291* 1 GX 0.321* 0.342* 0.009 1 GS 0.017 0.041 0.078 0.073 1 IY 0.198 0.051 0.193 0.231 PDY 0.324* 0.314* 0.148 0.020 0.126 0.411* 1 0.403* 1 GY 1 GE 0.289* 1 GI 0.626* 0.395* 1 GX 0.375* 0.113 0.149 1 GS 0.471* 0.236* 0.240* 0.063 1 IY 0.314* 0.474* 0.187 0.054 0.390* 1 PDY 0.187 0.170 0.111 0.239* 0.361* 0.266* 1 GY 1 GE 0.538* 1 GI 0.746* 0.453* 1 GX 0.109 0.077 0.109 1 GS 0.178 0.189 0.083 0.293* 1 IY 0.062 0.343* 0.035 0.111 PDY 0.095 0.145 0.005 0.107 0.021 0.197* 1 0.310* 1 Group IV GY 1 GE 0.339* 1 GI 0.655* 0.350* 1 GX 0.356* 0.146 0.193 1 GS 0.049 0.343* 0.120 0.176 1 IY 0.050 0.205 0.087 0.092 PDY 0.084 0.195 0.090 0.089 * significant at the 5% level. 30 0.673* 1 0.461* 0.044 1
Dar, A.A., Amirkhalkhali, S. On The Impact of Public Debt on Economic Growth TABLE 3. Pooled Results: Random Coefficients GLS Model: GY it 2i (GI) it + 3i (GE) it + 4i (GX) it + 5i (PDY) it +u it Countries All 1.994* 0.916 0.186* 0.023 0.248* 0.119 0.127* 0.033 gstatistic = 314.3* * and ** indicate statistical significance at the 5% and 10% level, respectively. TABLE 4. Countryspecific Results: Random Coefficients GLS Model: GY it 2i (GI) it + 3i (GE) it + 4i (GX) it + 5i (PDY) it +u it Countries 0.006 0.038 Luxembourg 2.890* 0.053 0.678* 0.151* 0.487* Australia 2.015* 0.076* 0.117 0.008 0.106 Norway 0.576 0.083* 0.272* 0.042 0.084 Switzerland 1.431 0.244* 0.252* 0.204* 0.002 New Zealand 1.106 0.202* 0.258* 0.156* 0.021 Germany 1.673* 0.237* 0.046 0.232* 0.017 Ireland 1.110 0.286* 0.468* 0.224* 0.019 UK 3.119* 0.147* 0.204 0.022 0.031 Spain 2.105* 0.176* 0.218* 0.055 0.008 Iceland 3.230* 0.097* 0.312* 0.143* 0.023 Finland 1.838* 0.216* 0.226 0.103* 0.005 Portugal 3.440* 0.197* 0.092 0.071 0.032 Sweden 0.986 0.223* 0.091 0.136* 0.012 Netherlands 2.017 0.213* 0.328* 0.140* 0.010 Austria 0.465 0.239* 0.392* 0.120* 0.015 Denmark 0.111 0.203* 0.105 0.080** 0.021 France 3.295 0.191* 0.206 0.060** 0.034 Canada 0.531 0.197* 0.706* 0.146* 0.024 USA 3.152* 0.268* 0.032 0.162* 0.080* Belgium 2.008 0.161* 0.307* 0.145* 0.09 Italy 0.330 0.258* 0.051 0.199* 0.025 Greece 2.145 0.113* 0.005 0.022 0.02 Japan 2.744 0.193* 0.390** 0.177** 0.034 * and ** indicate statistical significance at the 5% and 10% level, respectively. 31
Applied Econometrics and International Development Vol. 141 (2014) TABLE 5 Groupwise Results: Random Coefficients GLS Model: GY it 2i (GI) it + 3i (GE) it + 4i (GX) it + 5i (PDY) it +u it Groups g statistics I 2.004 0.132* 0.315 0.112** 0.056 46.89* II 2.179* 0.193* 0.397 0.130* 0.009 104.5* III 2.117 0.216* 0.244** 0.114* 0.011 79.25* IV 2.807 0.181* 0.186 0.136* 0.020 30.28* * and ** indicate statistical significance at the 5% and 10% level, respectively. Journal published by the EAAEDS: http://www.usc.es/economet/eaat.htm 32