- Email: [email protected]
Twin deficits and financial integration in EU member-states
Journal of Policy Modeling 28 (2006) 595–602
Twin deficits and financial integration in EU member-states Theodore Papadogonas a,∗ , Yannis Stournaras b b
a Bank of Greece, Greece University of Athens, Greece
Received 1 September 2005; received in revised form 1 January 2006; accepted 1 February 2006
Abstract In this paper, we find that changes in general government balances in the EU-15 member-states are matched to a large extent by opposite changes in the private savings–investment gap, implying that changes in public sector deficits have a rather small relationship with current account deficits. Also, using an empirical framework implied by a well-known, intertemporal model of the current account, we find that current account developments in Greece are explained by factors which are related to financial and economic integration, such as interest rate spreads and growth differentials, as well as to the general government balance. © 2006 Society for Policy Modeling. Published by Elsevier Inc. All rights reserved. JEL classification: F15; F32; F41 Keywords: Financial integration; Public sector and current account balances
1. Introduction One of the central issues in both economic policy and open-economy macroeconomics is the relationship between public sector (general government) deficits and current account deficits. It is often argued that the two deficits are related strongly and positively (“twin deficits”) and this is the way that this relationship is usually presented in the financial press. If this is true, national or world current account imbalances can be (easily?) tackled to the extent that public sector deficits are, more or less, under government control. However, the relationship between public sector and current account deficits is more complicated, depending on the behavior of the private sector savings/investment gap, since the ∗
Corresponding author at: 10 Kivelis street, 111 46 Athens, Greece. Tel.: +30 2103203601. E-mail address: [email protected] (T. Papadogonas).
0161-8938/$ – see front matter © 2006 Society for Policy Modeling. Published by Elsevier Inc. All rights reserved. doi:10.1016/j.jpolmod.2006.02.002
596
T. Papadogonas, Y. Stournaras / Journal of Policy Modeling 28 (2006) 595–602
well-known national account identity states that the current account deficit is equal to the sum of the public sector deficit (that is, the difference between government investment and government saving) and the private sector deficit (that is, the difference between private sector investment and saving). Almost every contribution to the literature on open-economy macroeconomics examines the macroeconomic implications of a higher public sector deficit and, in particular, its effects on the real exchange rate, output, private savings, private investment and the current account. The Mundell–Fleming–Dornbusch model might be said to remain the standard paradigm although many authors are also using the new, open-economy macroeconomics framework (for differences and similarities between the different models see, among others, Lane (2001), Obstfeld and Rogoff (1996), Vines (2003)). According to the standard paradigm, the effects of a higher public sector deficit are transmitted through two channels of influence, namely the goods market (via the real exchange rate) and the capital account (via the real interest rate). A higher public sector deficit is associated with an appreciation of the real exchange rate and higher output (as aggregate demand increases). As a consequence, it is also associated with a deterioration of the current account. In addition, a current account deficit results in net asset decumulation and higher foreign debt. The impact of this on expenditure, as well as long-term considerations regarding the need to raise taxes to repay the public sector debt, are additional transmission mechanisms through which public deficits might affect external deficits. Two particular cases deserve special attention for being at the two opposite extremes. Firstly, the debt neutrality (Ricardian) hypothesis: according to one version this hypothesis suggests that in a world with no imperfections and infinite horizons, changes in budget deficits cause offsetting, oneto-one changes in private savings through anticipations of changes in future taxation. Therefore, national savings and the current account remain unaffected (Barro, 1988). Secondly, complete crowding out of net exports: in a small open-economy where all goods are perfect substitutes and freely traded (that is, the law of one price holds) while domestic production is either at the full employment level or is fixed due to a rigid real product wage, a higher budget deficit causes a one-to-one increase in the current account deficit. It may be noted that this result remains valid in two other cases: (a) in the conventional Mundell–Fleming–Dornbusch model with perfect capital mobility and a floating nominal exchange rate, and (b) under the ‘New Cambridge’ assumption (Fetherston & Godley, 1978) that the private sector’s (households and corporations) net acquisition of financial assets is zero (that is, under the assumption that private disposable income is equal to private consumption and investment). In an interesting contribution, Blanchard and Giavazzi (2002) attempt to explain current account developments in certain European Union (EU) member-states emphasizing factors related to economic and financial integration. First, the reduction in interest rate spreads and in currency risk due to nominal convergence, which, for net borrowing countries, increase private investment and reduce national savings. Unless government net lending (that is, the general government surplus) moves sufficiently in the opposite direction, this channel implies an increase in the current account deficit to GDP ratio. Second, the increase in competition through economic integration, which is expected to increase total factor productivity, improving the home country’s growth prospects. Unless the growth rate of trade partners exceeds the home country’s growth rate, this channel also implies an increase in the current account deficit to GDP ratio. Blanchard and Giavazzi (2002) use an explicit utilitarian approach with households living for two periods and maximizing a logarithmic utility function under a two-period budget constraint. Their model is thus different than the more traditional Mundell–Fleming–Dornbusch models of
T. Papadogonas, Y. Stournaras / Journal of Policy Modeling 28 (2006) 595–602
597
the current account, which emphasize competitiveness and national incomes of the home and “foreign” country as the main determinants of the current account. However, differences are more apparent than real: as we will explain later, competitiveness in the Blanchard and Giavazzi (2002) model does not appear in their current account equation because it is an endogenous variable, which depends on relative output. Actually, models with aggregate demand–aggregate supply equations of the Mundell–Fleming–Dornbusch tradition emphasize variables (as determinants of the current account), which are closer to those of Blanchard and Giavazzi (2002), see Stournaras (2004). The present paper addresses the following two empirical questions: (a) What is the empirical relationship between changes in the net lending of general government (that is, in the surplus of the public sector) and changes in the private sector savings–investment gap and, therefore, between changes in the net lending of general government and the current account balance in EU memberstates? (b) Can the Greek current account balance be explained by factors related to economic and financial integration a la Blanchard and Giavazzi (2002) as well as the general government deficit? 2. Government savings and investment and the private sector savings–investment gap It is well known that we can express the difference between national savings and national investment (that is, the current account surplus) as the sum of the net lending of general government (NLG)1 and the private sector savings–investment gap (Sp –Ip ). To the extent that NLG can be considered, more or less, as an exogenous variable under government control, it is important to examine the contribution of the net lending of general government to the overall balance between national savings and national investment and, thus, to the evolution of the current account balance. Incidentally it is interesting that many authors (see among others, Gourinchas (2002)) have stressed that the widening of current account deficits due to financial integration might cause international financial markets to take fright over the sustainability of the net foreign asset position of countries with widening current account deficits. This might require some insurance, or buffer. Smaller government deficits (or larger government surpluses) provide such an insurance. Hence it might be argued that the limits imposed by the Stability and Growth Pact on government deficits are justified not only on the grounds of fiscal and monetary stability, but also on grounds related to current account adjustment as a result of financial integration. In any case, general government net lending cannot be ignored in models analyzing the effects of financial integration. Since financial integration and, in particular, the convergence of interest rates, affect the balance between private savings and private investment, general government net lending could either reinforce this effect or move the balance between national saving and national investment (and, therefore, the current account balance) to the opposite direction. An important question that is usually asked in this context is the following: To what extent is a change in the net lending of general government matched by an opposite change in the private savings–investment gap (and through which transmission mechanisms) and, therefore, to what extent does a change in the net lending of general government affect the current account balance? This leads us to test the following equation: Sp − Ip NLG + ui = a 0 + a1 (1) GDP GDP i
1 NLG is equal to government savings minus government investment. Government savings is equal to current government revenue minus current expenditure.
598
T. Papadogonas, Y. Stournaras / Journal of Policy Modeling 28 (2006) 595–602
Table 1 Estimation results for Eq. (1)
Austria Belgium Denmark Finland France Greece Ireland Italy Netherlands Portugal United Kingdom
a0
a1
Adj. R2
a1 = −1a
0.04 (0.88) 0.02 (0.09) 0.16 (0.09) 0.53 (1.40) 0.08 (0.41) −0.08 (0.26) 0.11 (0.24) 0.10 (0.42) 0.08 (0.31) 0.03 (0.06) 0.08 (0.38)
−1.19*** (4.73) −0.71*** (6.87) −0.12 (0.44) −0.82*** (5.31) −1.01*** (5.04) −0.75*** (4.95) −0.74** (2.87) −0.89*** (5.72) −0.41*** (3.26) −0.93*** (3.44) −1.01*** (6.94)
0.45 0.59 0.02 0.50 0.52 0.42 0.31 0.59 0.24 0.31 0.60
0.76 2.81** 5.82*** 1.15 0.07 1.66 1.02 0.73 4.64*** 0.28 0.05
t ratios are in parentheses. ** Significant at the 5% level, *** significant at the 1% level (two-tailed tests). Results for Germany, Luxembourg, Spain and Sweden are not reliable due to insufficient observations. a t-tests for the hypothesis H : a = −1. ** Denotes rejection at the 5% level, and *** denotes rejection at the 1% level. 0 1
Table 1 presents results for 11 EU member-states for this equation.2 These imply a negative and statistically significant coefficient a1 for all member-states except Denmark with a value close to −1. Actually the hypothesis a1 = −1 cannot be rejected for most of these 11 member-states. Also, the constant a0 appears to be statistically insignificant in all member-states. These results imply that changes in the net lending of general government are matched to a large extent by opposite changes in the private sector savings/investment gap implying a negligible association between public sector and current accounts balances. This surprisingly strong result might be explained in many ways and has strong policy implications. We will not attempt to present all relevant theories here, since this is outside the scope of the present study, nor the full transmission mechanism. However, we will indicate certain possible explanations: (A) It might be argued very strongly that these results are consistent with the debt neutrality or Ricardian hypothesis presented earlier, since a stylized fact that would emerge from a Ricardian model is precisely the fact that the coefficient a1 in the above equation would be −1. An objection to this explanation could be the following: The precise Ricardian position does not refer to the relationship between government net lending and the private sector savings–investment gap. It refers to the relationship between private savings and government savings in a frictionless and fully rational world, where the private sector, in full anticipation of the future implications of increases in government current expenditure, increases its current savings. Moreover, earlier tests by McCallum (1993) for OECD countries reject this Ricardian relationship between private savings and government savings. (B) Mundell–Fleming–Dornbusch models suggest that a reduction in government net lending (that is, an increase in general government deficit) leads to higher output, an increase in real interest rates and an appreciation of the real exchange rate. These imply higher private savings and, perhaps, lower private investment (depending on interest rate and income
2
All data used in this study come from OECD, National Accounts and cover the period 1970–2003.
T. Papadogonas, Y. Stournaras / Journal of Policy Modeling 28 (2006) 595–602
599
elasticities). Therefore, they imply a higher private savings–investment gap. However, the correlation between the increase in general government deficit and the increase in the private savings–investment gap is generally less than unity. A correlation coefficient equal to unity might be derived in models with product real wage rigidity, capital controls and credit rationing (see among others, Gibson, Stournaras, & Tsakalotos, 1992, 1994). (C) The result might be due to a combination of conjunctural factors and financial integration: Maastricht Treaty fiscal criteria promoted fiscal contraction (an increase in government net lending) for member-states with large government deficits and, at the same time, financial integration affected the private savings–investment gap: for borrowing countries the private savings–investment gap shrank as a result of interest rate convergence, which reduced private savings and increased private investment. Although we do not examine empirically the transmission mechanism and the full implications of this surprisingly strong result (which definitely requires further investigation), we can draw three conclusions. Firstly, we should expect a rather small association between changes in the current account deficit and changes in the public sector deficit, as there is strong evidence that changes in the public sector deficit are matched by opposite changes in the private sector savings–investment gap.3 Secondly, empirical efforts to analyze the effects of financial integration, especially on the current account should not ignore the contribution of government surpluses or deficits. Thirdly, a reduction in public sector deficits is not sufficient to reduce external imbalances: additional measures or incentives to affect the private savings–investment gap are also required if it is deemed that a current account deficit is a cause for concern. 3. The determinants of the current account balance in Greece In this section we will use the model proposed by Blanchard and Giavazzi (2002) in order to explain the current account balance in one of the EU member-states, Greece, making use of factors related to economic and financial integration as well as the public sector deficit. The model includes n countries, trading goods and assets among themselves. Each country produces its own good, but households in each country consume the same composite good. Households live for two periods and maximize logarithmic utility under an intertemporal budget constraint. The optimization problem is: Max{log(Ct ) + log(Ct+1 )}
(2)
Subject to Ct + [(1 + X)R]−1 Ct+1 = Pt Yt + [(1 + X)R]−1 Pt+1 Yt+1 Where
n
1 (σ−1)/σ Ci C= n
(3)
σ/(σ−1) (4)
i=1
C is the composite consumption good defined in Eq. (4), which is consumed by all countries, σ the elasticity of substitution (σ > 1), Y the domestic output, P the price of domestic output in terms 3 In a model in the tradition of new open economy macroeconomics, Erceg, Guerrieri, and Gust (2005) derive the same result for the US.
600
T. Papadogonas, Y. Stournaras / Journal of Policy Modeling 28 (2006) 595–602
of consumption, R the world interest rate in terms of consumption, t the current period, t + 1 the future period, and X is the spread at which the representative borrowing country borrows above the world interest rate R (this spread is affected by factors such as risk premia). Solving the above optimisation problem, and under various assumptions Blanchard and Giavazzi (2002) obtain the current account balance as a percent of output (1 − C/Y) for a borrowing representative country as: 1 1 1 + g 1−(1/σ) CAB = (5) 1− 2 1 + X 1 + g∗ Where CAB is the current account surplus as a percent of output, g* the world rate of output growth and g is the rate of output growth for the country under consideration.4 Although Blanchard and Giavazzi (2002) have used Eq. (5) in a heuristic way to explain certain stylized facts, they did not actually estimate it in the empirical part of their paper. In the present paper, we estimate Eq. (6), which is derived directly from Eq. (5), in which we have added a fiscal variable, the net lending of general government, to capture the contribution of the general government to national savings:5 CAB = a0 + a1 NLG + a2 X + a3 G
(6)
where the variables are described as follows: CAB, current account surplus (% of GDP); NLG, net lending (surplus) of general government (% of GDP); X, interest rate spread, ideally the difference in interest rates between Greek and German long-term bonds;6 G, difference between the real GDP growth rate of Greece and that of OECD. According to the augmented Dickey–Fuller test, all variables in Eq. (7) appear to be integrated of order one or lower. We then used Johansen’s Maximum Likelihood procedure to test for the presence of cointegration among the variables CAB, NLG, X and G. The maximum likelihood and trace statistics (Lmax and Ltrace ) both reject the null hypothesis of no cointegration in favour of one cointegrating relationship at the 5% level. Further, a lag of order k = 3 was chosen for the VAR of the Johansen procedure by the application of the Hannan–Quinn and Akaike criteria. The results of the econometric analysis are (t-ratios are in parentheses):7 CAB = a0 + 0.33∗∗∗ NLG + 0.52∗∗∗ X − 0.38∗∗ G (6.35)
(11.06)
(2.36)
(7)
where ** denotes the significant at the 5% level (two-tailed test) and *** denotes significant at the 1% level (two-tailed test). In the above equation, the explanatory variables are statistically significant and their coefficients have the expected signs: (a) an increase in the net general government lending (surplus) to ∗ , where Y* is output aggregated over all n countries (“world output”). Also In the model (1 + g∗ )Yt∗ = Yt+1 (1 + g)Yt = Yt+1 , where Y is domestic output. In Eq. (6) the world interest rate R and the price of domestic output in terms of consumption, P (which may be called the real exchange rate) do not appear because they are eliminated in the ∗ = 1/(1 + g∗ ). solution process. Indeed, Pt = (Yt /Yt∗ )−1/σ and 1/R = Yt∗ /Yt+1 5 In Blanchard and Giavazzi (2002) there is no public sector. It is straightforward to include it in the model. 6 Due to lack of data on long-term bond interest rates in Greece for a sufficiently long period, we estimate X (the interest rate spread) by the inflation differential between Greece and Germany. 7 In the information set of the model world oil prices were included to achieve a more correct specification. This variable appeared in the short-run dynamics of the model. 4
T. Papadogonas, Y. Stournaras / Journal of Policy Modeling 28 (2006) 595–602
601
GDP ratio of one percentage point increases the current account surplus to GDP ratio (or reduces the current account deficit to GDP ratio) by one-third of a percentage point; (b) a reduction in the interest rate spread (difference between the interest rate at which the domestic country can borrow and the world interest rate) of 100 basis points reduces the current account surplus to GDP ratio (or, increases the current account deficit) by one-half of a percentage point; (c) a widening of the growth differential (in favour of the home country) of one percentage point reduces the current account surplus (or increases the current account deficit) to GDP ratio by over a third (0.38) of a percentage point. As indicated in Section 1, there are many similarities between the findings of this model and the findings of more traditional ones (Mundell–Fleming–Dornbusch models) regarding the current account balance. First, output growth differentials between the home country and the world economy affect exports and imports in more traditional models: a high domestic output growth and a low output growth of the world economy affect imports of the home country more than its exports, depending, of course, on the relative income elasticities. Second, a reduction in interest rate spreads in more traditional models shift the aggregate demand schedule to the right. With a positively sloped aggregate supply curve, output increases and the real exchange rate appreciates. As a result, the current account balance deteriorates (see, among others, Stournaras, 2004). Third, an increase in government net lending (that is, a reduction in the government deficit) shifts the aggregate demand schedule to the left, domestic output falls and the real exchange rate depreciates. As a result the current account balance improves. Blanchard and Giavazzi (2002) treat output as exogenous (their model does not specify a supply side). This prevents them from considering important effects on the current account balance emanating from shifts in total factor productivity, the product and labour market structure, or from exogenous shocks affecting the cost of domestic output (e.g. an oil price shock). In Stournaras (2004) a positive total factor productivity shock shifts the aggregate supply schedule to the right, domestic output increases and the real exchange rate depreciates. Hence the effect on the current account balance is ambiguous, depending on relative elasticities. 4. Summary and conclusions The two main findings of this paper are (a) changes in the net lending of general government are matched to a large extent by opposite changes in the private sector savings–investment gap, in almost all EU member-states and (b) current account developments in Greece are explained, as suggested by intertemporal models of the current account using a utilitarian framework, by factors related to economic and financial integration, such as the growth differential between Greece and OECD and interest rate spreads (as approximated by inflation differentials between Greece and Germany) as well as by the public sector (general government) deficit. The findings imply that an improvement in the general government balance has a positive impact on the current account balance, although the effect is rather small as changes in general government balances are found to be strongly associated with opposite changes in the private sector savings–investment gap. Further research is needed, however, to investigate in more depth the causal links between these factors. The above findings also imply that we should treat changes in the current account balance with care. It is by no means the case that an increase in the current account deficit should always be a source for concern as this could simply reflect the forces of financial integration and real convergence. Finally, the findings suggest that the public sector deficit is not a sufficient instrument to affect the current account deficit; if it is deemed that the
602
T. Papadogonas, Y. Stournaras / Journal of Policy Modeling 28 (2006) 595–602
current account is a source for concern, measures or incentives are also needed in order to affect the private sector savings/investment gap. Acknowledgements The authors gratefully acknowledge helpful discussions and comments with Vassilis Droucopoulos, Heather Gibson and Ioanna Mpardaka. The usual disclaimer applies; remaining errors are, of course, our own. References Barro, R. (1988). The Ricardian Approach to Budget Deficits, NBER Working Paper, no. 2685. Blanchard, O., & Giavazzi, F. (2002). Current account deficits in the Euro area: The end of the Feldstein–Horioka puzzle? Brookings Papers on Economic Activity, 2, 147–209. Erceg, C., Guerrieri, L., & Gust, C. (2005). Expansionary fiscal shocks and the trade deficit, Board of Governors of the Federal Reserve System, International Finance Discussion Paper, 25. Fetherston, M., & Godley, W. (1978). New Cambridge macroeconomics and global monetarism. In A. Brunner & A. Meltzer (Eds.), Public policies in open-economies. North Holland. Gibson, H., Stournaras, Y., & Tsakalotos, E. (1992). Twin deficits in credit rationed economies, Studies in Economics, no. 92/9, University of Kent. Gibson, H., Stournaras, Y., & Tsakalotos, E. (1994). Investment and credit-rationing in four European economies. Greek Economic Review, 16(2), 65–81. Gourinchas, P. O. (2002). Current account deficits in the Euro area: The end of the Feldstein–Horioka puzzle? Brookings Papers on Economic Activity, 2, 147–209. Lane, P. (2001). The new, open-economy macroeconomics: A survey. Journal of International Economics, 54, 235–266. McCallum, J. (1993). Government spending and national saving. In M. Baldassari, R. Mundell, & J. McCallum (Eds.), Debt, deficit and economic performance. MacMillan/St. Martin’s. Obstfeld, M., & Rogoff, K. (1996). Foundations of international macroeconomics. MIT Press. Stournaras, Y. (2004). Aggregate supply and demand, the real exchange rate and the denomination of oil prices, Working Paper 55, University of Athens. Vines, D. (2003). John Maynard Keynes 1937–1946: The creation of international macroeconomics. The Economic Journal, 113, F338–F361.
Twin deficits and financial integration in EU member-states Theodore Papadogonas a,∗ , Yannis Stournaras b b
a Bank of Greece, Greece University of Athens, Greece
Received 1 September 2005; received in revised form 1 January 2006; accepted 1 February 2006
Abstract In this paper, we find that changes in general government balances in the EU-15 member-states are matched to a large extent by opposite changes in the private savings–investment gap, implying that changes in public sector deficits have a rather small relationship with current account deficits. Also, using an empirical framework implied by a well-known, intertemporal model of the current account, we find that current account developments in Greece are explained by factors which are related to financial and economic integration, such as interest rate spreads and growth differentials, as well as to the general government balance. © 2006 Society for Policy Modeling. Published by Elsevier Inc. All rights reserved. JEL classification: F15; F32; F41 Keywords: Financial integration; Public sector and current account balances
1. Introduction One of the central issues in both economic policy and open-economy macroeconomics is the relationship between public sector (general government) deficits and current account deficits. It is often argued that the two deficits are related strongly and positively (“twin deficits”) and this is the way that this relationship is usually presented in the financial press. If this is true, national or world current account imbalances can be (easily?) tackled to the extent that public sector deficits are, more or less, under government control. However, the relationship between public sector and current account deficits is more complicated, depending on the behavior of the private sector savings/investment gap, since the ∗
Corresponding author at: 10 Kivelis street, 111 46 Athens, Greece. Tel.: +30 2103203601. E-mail address: [email protected] (T. Papadogonas).
0161-8938/$ – see front matter © 2006 Society for Policy Modeling. Published by Elsevier Inc. All rights reserved. doi:10.1016/j.jpolmod.2006.02.002
596
T. Papadogonas, Y. Stournaras / Journal of Policy Modeling 28 (2006) 595–602
well-known national account identity states that the current account deficit is equal to the sum of the public sector deficit (that is, the difference between government investment and government saving) and the private sector deficit (that is, the difference between private sector investment and saving). Almost every contribution to the literature on open-economy macroeconomics examines the macroeconomic implications of a higher public sector deficit and, in particular, its effects on the real exchange rate, output, private savings, private investment and the current account. The Mundell–Fleming–Dornbusch model might be said to remain the standard paradigm although many authors are also using the new, open-economy macroeconomics framework (for differences and similarities between the different models see, among others, Lane (2001), Obstfeld and Rogoff (1996), Vines (2003)). According to the standard paradigm, the effects of a higher public sector deficit are transmitted through two channels of influence, namely the goods market (via the real exchange rate) and the capital account (via the real interest rate). A higher public sector deficit is associated with an appreciation of the real exchange rate and higher output (as aggregate demand increases). As a consequence, it is also associated with a deterioration of the current account. In addition, a current account deficit results in net asset decumulation and higher foreign debt. The impact of this on expenditure, as well as long-term considerations regarding the need to raise taxes to repay the public sector debt, are additional transmission mechanisms through which public deficits might affect external deficits. Two particular cases deserve special attention for being at the two opposite extremes. Firstly, the debt neutrality (Ricardian) hypothesis: according to one version this hypothesis suggests that in a world with no imperfections and infinite horizons, changes in budget deficits cause offsetting, oneto-one changes in private savings through anticipations of changes in future taxation. Therefore, national savings and the current account remain unaffected (Barro, 1988). Secondly, complete crowding out of net exports: in a small open-economy where all goods are perfect substitutes and freely traded (that is, the law of one price holds) while domestic production is either at the full employment level or is fixed due to a rigid real product wage, a higher budget deficit causes a one-to-one increase in the current account deficit. It may be noted that this result remains valid in two other cases: (a) in the conventional Mundell–Fleming–Dornbusch model with perfect capital mobility and a floating nominal exchange rate, and (b) under the ‘New Cambridge’ assumption (Fetherston & Godley, 1978) that the private sector’s (households and corporations) net acquisition of financial assets is zero (that is, under the assumption that private disposable income is equal to private consumption and investment). In an interesting contribution, Blanchard and Giavazzi (2002) attempt to explain current account developments in certain European Union (EU) member-states emphasizing factors related to economic and financial integration. First, the reduction in interest rate spreads and in currency risk due to nominal convergence, which, for net borrowing countries, increase private investment and reduce national savings. Unless government net lending (that is, the general government surplus) moves sufficiently in the opposite direction, this channel implies an increase in the current account deficit to GDP ratio. Second, the increase in competition through economic integration, which is expected to increase total factor productivity, improving the home country’s growth prospects. Unless the growth rate of trade partners exceeds the home country’s growth rate, this channel also implies an increase in the current account deficit to GDP ratio. Blanchard and Giavazzi (2002) use an explicit utilitarian approach with households living for two periods and maximizing a logarithmic utility function under a two-period budget constraint. Their model is thus different than the more traditional Mundell–Fleming–Dornbusch models of
T. Papadogonas, Y. Stournaras / Journal of Policy Modeling 28 (2006) 595–602
597
the current account, which emphasize competitiveness and national incomes of the home and “foreign” country as the main determinants of the current account. However, differences are more apparent than real: as we will explain later, competitiveness in the Blanchard and Giavazzi (2002) model does not appear in their current account equation because it is an endogenous variable, which depends on relative output. Actually, models with aggregate demand–aggregate supply equations of the Mundell–Fleming–Dornbusch tradition emphasize variables (as determinants of the current account), which are closer to those of Blanchard and Giavazzi (2002), see Stournaras (2004). The present paper addresses the following two empirical questions: (a) What is the empirical relationship between changes in the net lending of general government (that is, in the surplus of the public sector) and changes in the private sector savings–investment gap and, therefore, between changes in the net lending of general government and the current account balance in EU memberstates? (b) Can the Greek current account balance be explained by factors related to economic and financial integration a la Blanchard and Giavazzi (2002) as well as the general government deficit? 2. Government savings and investment and the private sector savings–investment gap It is well known that we can express the difference between national savings and national investment (that is, the current account surplus) as the sum of the net lending of general government (NLG)1 and the private sector savings–investment gap (Sp –Ip ). To the extent that NLG can be considered, more or less, as an exogenous variable under government control, it is important to examine the contribution of the net lending of general government to the overall balance between national savings and national investment and, thus, to the evolution of the current account balance. Incidentally it is interesting that many authors (see among others, Gourinchas (2002)) have stressed that the widening of current account deficits due to financial integration might cause international financial markets to take fright over the sustainability of the net foreign asset position of countries with widening current account deficits. This might require some insurance, or buffer. Smaller government deficits (or larger government surpluses) provide such an insurance. Hence it might be argued that the limits imposed by the Stability and Growth Pact on government deficits are justified not only on the grounds of fiscal and monetary stability, but also on grounds related to current account adjustment as a result of financial integration. In any case, general government net lending cannot be ignored in models analyzing the effects of financial integration. Since financial integration and, in particular, the convergence of interest rates, affect the balance between private savings and private investment, general government net lending could either reinforce this effect or move the balance between national saving and national investment (and, therefore, the current account balance) to the opposite direction. An important question that is usually asked in this context is the following: To what extent is a change in the net lending of general government matched by an opposite change in the private savings–investment gap (and through which transmission mechanisms) and, therefore, to what extent does a change in the net lending of general government affect the current account balance? This leads us to test the following equation: Sp − Ip NLG + ui = a 0 + a1 (1) GDP GDP i
1 NLG is equal to government savings minus government investment. Government savings is equal to current government revenue minus current expenditure.
598
T. Papadogonas, Y. Stournaras / Journal of Policy Modeling 28 (2006) 595–602
Table 1 Estimation results for Eq. (1)
Austria Belgium Denmark Finland France Greece Ireland Italy Netherlands Portugal United Kingdom
a0
a1
Adj. R2
a1 = −1a
0.04 (0.88) 0.02 (0.09) 0.16 (0.09) 0.53 (1.40) 0.08 (0.41) −0.08 (0.26) 0.11 (0.24) 0.10 (0.42) 0.08 (0.31) 0.03 (0.06) 0.08 (0.38)
−1.19*** (4.73) −0.71*** (6.87) −0.12 (0.44) −0.82*** (5.31) −1.01*** (5.04) −0.75*** (4.95) −0.74** (2.87) −0.89*** (5.72) −0.41*** (3.26) −0.93*** (3.44) −1.01*** (6.94)
0.45 0.59 0.02 0.50 0.52 0.42 0.31 0.59 0.24 0.31 0.60
0.76 2.81** 5.82*** 1.15 0.07 1.66 1.02 0.73 4.64*** 0.28 0.05
t ratios are in parentheses. ** Significant at the 5% level, *** significant at the 1% level (two-tailed tests). Results for Germany, Luxembourg, Spain and Sweden are not reliable due to insufficient observations. a t-tests for the hypothesis H : a = −1. ** Denotes rejection at the 5% level, and *** denotes rejection at the 1% level. 0 1
Table 1 presents results for 11 EU member-states for this equation.2 These imply a negative and statistically significant coefficient a1 for all member-states except Denmark with a value close to −1. Actually the hypothesis a1 = −1 cannot be rejected for most of these 11 member-states. Also, the constant a0 appears to be statistically insignificant in all member-states. These results imply that changes in the net lending of general government are matched to a large extent by opposite changes in the private sector savings/investment gap implying a negligible association between public sector and current accounts balances. This surprisingly strong result might be explained in many ways and has strong policy implications. We will not attempt to present all relevant theories here, since this is outside the scope of the present study, nor the full transmission mechanism. However, we will indicate certain possible explanations: (A) It might be argued very strongly that these results are consistent with the debt neutrality or Ricardian hypothesis presented earlier, since a stylized fact that would emerge from a Ricardian model is precisely the fact that the coefficient a1 in the above equation would be −1. An objection to this explanation could be the following: The precise Ricardian position does not refer to the relationship between government net lending and the private sector savings–investment gap. It refers to the relationship between private savings and government savings in a frictionless and fully rational world, where the private sector, in full anticipation of the future implications of increases in government current expenditure, increases its current savings. Moreover, earlier tests by McCallum (1993) for OECD countries reject this Ricardian relationship between private savings and government savings. (B) Mundell–Fleming–Dornbusch models suggest that a reduction in government net lending (that is, an increase in general government deficit) leads to higher output, an increase in real interest rates and an appreciation of the real exchange rate. These imply higher private savings and, perhaps, lower private investment (depending on interest rate and income
2
All data used in this study come from OECD, National Accounts and cover the period 1970–2003.
T. Papadogonas, Y. Stournaras / Journal of Policy Modeling 28 (2006) 595–602
599
elasticities). Therefore, they imply a higher private savings–investment gap. However, the correlation between the increase in general government deficit and the increase in the private savings–investment gap is generally less than unity. A correlation coefficient equal to unity might be derived in models with product real wage rigidity, capital controls and credit rationing (see among others, Gibson, Stournaras, & Tsakalotos, 1992, 1994). (C) The result might be due to a combination of conjunctural factors and financial integration: Maastricht Treaty fiscal criteria promoted fiscal contraction (an increase in government net lending) for member-states with large government deficits and, at the same time, financial integration affected the private savings–investment gap: for borrowing countries the private savings–investment gap shrank as a result of interest rate convergence, which reduced private savings and increased private investment. Although we do not examine empirically the transmission mechanism and the full implications of this surprisingly strong result (which definitely requires further investigation), we can draw three conclusions. Firstly, we should expect a rather small association between changes in the current account deficit and changes in the public sector deficit, as there is strong evidence that changes in the public sector deficit are matched by opposite changes in the private sector savings–investment gap.3 Secondly, empirical efforts to analyze the effects of financial integration, especially on the current account should not ignore the contribution of government surpluses or deficits. Thirdly, a reduction in public sector deficits is not sufficient to reduce external imbalances: additional measures or incentives to affect the private savings–investment gap are also required if it is deemed that a current account deficit is a cause for concern. 3. The determinants of the current account balance in Greece In this section we will use the model proposed by Blanchard and Giavazzi (2002) in order to explain the current account balance in one of the EU member-states, Greece, making use of factors related to economic and financial integration as well as the public sector deficit. The model includes n countries, trading goods and assets among themselves. Each country produces its own good, but households in each country consume the same composite good. Households live for two periods and maximize logarithmic utility under an intertemporal budget constraint. The optimization problem is: Max{log(Ct ) + log(Ct+1 )}
(2)
Subject to Ct + [(1 + X)R]−1 Ct+1 = Pt Yt + [(1 + X)R]−1 Pt+1 Yt+1 Where
n
1 (σ−1)/σ Ci C= n
(3)
σ/(σ−1) (4)
i=1
C is the composite consumption good defined in Eq. (4), which is consumed by all countries, σ the elasticity of substitution (σ > 1), Y the domestic output, P the price of domestic output in terms 3 In a model in the tradition of new open economy macroeconomics, Erceg, Guerrieri, and Gust (2005) derive the same result for the US.
600
T. Papadogonas, Y. Stournaras / Journal of Policy Modeling 28 (2006) 595–602
of consumption, R the world interest rate in terms of consumption, t the current period, t + 1 the future period, and X is the spread at which the representative borrowing country borrows above the world interest rate R (this spread is affected by factors such as risk premia). Solving the above optimisation problem, and under various assumptions Blanchard and Giavazzi (2002) obtain the current account balance as a percent of output (1 − C/Y) for a borrowing representative country as: 1 1 1 + g 1−(1/σ) CAB = (5) 1− 2 1 + X 1 + g∗ Where CAB is the current account surplus as a percent of output, g* the world rate of output growth and g is the rate of output growth for the country under consideration.4 Although Blanchard and Giavazzi (2002) have used Eq. (5) in a heuristic way to explain certain stylized facts, they did not actually estimate it in the empirical part of their paper. In the present paper, we estimate Eq. (6), which is derived directly from Eq. (5), in which we have added a fiscal variable, the net lending of general government, to capture the contribution of the general government to national savings:5 CAB = a0 + a1 NLG + a2 X + a3 G
(6)
where the variables are described as follows: CAB, current account surplus (% of GDP); NLG, net lending (surplus) of general government (% of GDP); X, interest rate spread, ideally the difference in interest rates between Greek and German long-term bonds;6 G, difference between the real GDP growth rate of Greece and that of OECD. According to the augmented Dickey–Fuller test, all variables in Eq. (7) appear to be integrated of order one or lower. We then used Johansen’s Maximum Likelihood procedure to test for the presence of cointegration among the variables CAB, NLG, X and G. The maximum likelihood and trace statistics (Lmax and Ltrace ) both reject the null hypothesis of no cointegration in favour of one cointegrating relationship at the 5% level. Further, a lag of order k = 3 was chosen for the VAR of the Johansen procedure by the application of the Hannan–Quinn and Akaike criteria. The results of the econometric analysis are (t-ratios are in parentheses):7 CAB = a0 + 0.33∗∗∗ NLG + 0.52∗∗∗ X − 0.38∗∗ G (6.35)
(11.06)
(2.36)
(7)
where ** denotes the significant at the 5% level (two-tailed test) and *** denotes significant at the 1% level (two-tailed test). In the above equation, the explanatory variables are statistically significant and their coefficients have the expected signs: (a) an increase in the net general government lending (surplus) to ∗ , where Y* is output aggregated over all n countries (“world output”). Also In the model (1 + g∗ )Yt∗ = Yt+1 (1 + g)Yt = Yt+1 , where Y is domestic output. In Eq. (6) the world interest rate R and the price of domestic output in terms of consumption, P (which may be called the real exchange rate) do not appear because they are eliminated in the ∗ = 1/(1 + g∗ ). solution process. Indeed, Pt = (Yt /Yt∗ )−1/σ and 1/R = Yt∗ /Yt+1 5 In Blanchard and Giavazzi (2002) there is no public sector. It is straightforward to include it in the model. 6 Due to lack of data on long-term bond interest rates in Greece for a sufficiently long period, we estimate X (the interest rate spread) by the inflation differential between Greece and Germany. 7 In the information set of the model world oil prices were included to achieve a more correct specification. This variable appeared in the short-run dynamics of the model. 4
T. Papadogonas, Y. Stournaras / Journal of Policy Modeling 28 (2006) 595–602
601
GDP ratio of one percentage point increases the current account surplus to GDP ratio (or reduces the current account deficit to GDP ratio) by one-third of a percentage point; (b) a reduction in the interest rate spread (difference between the interest rate at which the domestic country can borrow and the world interest rate) of 100 basis points reduces the current account surplus to GDP ratio (or, increases the current account deficit) by one-half of a percentage point; (c) a widening of the growth differential (in favour of the home country) of one percentage point reduces the current account surplus (or increases the current account deficit) to GDP ratio by over a third (0.38) of a percentage point. As indicated in Section 1, there are many similarities between the findings of this model and the findings of more traditional ones (Mundell–Fleming–Dornbusch models) regarding the current account balance. First, output growth differentials between the home country and the world economy affect exports and imports in more traditional models: a high domestic output growth and a low output growth of the world economy affect imports of the home country more than its exports, depending, of course, on the relative income elasticities. Second, a reduction in interest rate spreads in more traditional models shift the aggregate demand schedule to the right. With a positively sloped aggregate supply curve, output increases and the real exchange rate appreciates. As a result, the current account balance deteriorates (see, among others, Stournaras, 2004). Third, an increase in government net lending (that is, a reduction in the government deficit) shifts the aggregate demand schedule to the left, domestic output falls and the real exchange rate depreciates. As a result the current account balance improves. Blanchard and Giavazzi (2002) treat output as exogenous (their model does not specify a supply side). This prevents them from considering important effects on the current account balance emanating from shifts in total factor productivity, the product and labour market structure, or from exogenous shocks affecting the cost of domestic output (e.g. an oil price shock). In Stournaras (2004) a positive total factor productivity shock shifts the aggregate supply schedule to the right, domestic output increases and the real exchange rate depreciates. Hence the effect on the current account balance is ambiguous, depending on relative elasticities. 4. Summary and conclusions The two main findings of this paper are (a) changes in the net lending of general government are matched to a large extent by opposite changes in the private sector savings–investment gap, in almost all EU member-states and (b) current account developments in Greece are explained, as suggested by intertemporal models of the current account using a utilitarian framework, by factors related to economic and financial integration, such as the growth differential between Greece and OECD and interest rate spreads (as approximated by inflation differentials between Greece and Germany) as well as by the public sector (general government) deficit. The findings imply that an improvement in the general government balance has a positive impact on the current account balance, although the effect is rather small as changes in general government balances are found to be strongly associated with opposite changes in the private sector savings–investment gap. Further research is needed, however, to investigate in more depth the causal links between these factors. The above findings also imply that we should treat changes in the current account balance with care. It is by no means the case that an increase in the current account deficit should always be a source for concern as this could simply reflect the forces of financial integration and real convergence. Finally, the findings suggest that the public sector deficit is not a sufficient instrument to affect the current account deficit; if it is deemed that the
602
T. Papadogonas, Y. Stournaras / Journal of Policy Modeling 28 (2006) 595–602
current account is a source for concern, measures or incentives are also needed in order to affect the private sector savings/investment gap. Acknowledgements The authors gratefully acknowledge helpful discussions and comments with Vassilis Droucopoulos, Heather Gibson and Ioanna Mpardaka. The usual disclaimer applies; remaining errors are, of course, our own. References Barro, R. (1988). The Ricardian Approach to Budget Deficits, NBER Working Paper, no. 2685. Blanchard, O., & Giavazzi, F. (2002). Current account deficits in the Euro area: The end of the Feldstein–Horioka puzzle? Brookings Papers on Economic Activity, 2, 147–209. Erceg, C., Guerrieri, L., & Gust, C. (2005). Expansionary fiscal shocks and the trade deficit, Board of Governors of the Federal Reserve System, International Finance Discussion Paper, 25. Fetherston, M., & Godley, W. (1978). New Cambridge macroeconomics and global monetarism. In A. Brunner & A. Meltzer (Eds.), Public policies in open-economies. North Holland. Gibson, H., Stournaras, Y., & Tsakalotos, E. (1992). Twin deficits in credit rationed economies, Studies in Economics, no. 92/9, University of Kent. Gibson, H., Stournaras, Y., & Tsakalotos, E. (1994). Investment and credit-rationing in four European economies. Greek Economic Review, 16(2), 65–81. Gourinchas, P. O. (2002). Current account deficits in the Euro area: The end of the Feldstein–Horioka puzzle? Brookings Papers on Economic Activity, 2, 147–209. Lane, P. (2001). The new, open-economy macroeconomics: A survey. Journal of International Economics, 54, 235–266. McCallum, J. (1993). Government spending and national saving. In M. Baldassari, R. Mundell, & J. McCallum (Eds.), Debt, deficit and economic performance. MacMillan/St. Martin’s. Obstfeld, M., & Rogoff, K. (1996). Foundations of international macroeconomics. MIT Press. Stournaras, Y. (2004). Aggregate supply and demand, the real exchange rate and the denomination of oil prices, Working Paper 55, University of Athens. Vines, D. (2003). John Maynard Keynes 1937–1946: The creation of international macroeconomics. The Economic Journal, 113, F338–F361.