AN ASSESSMENT OF THE SUSTAINABILITY OF CURRENT ACCOUNT IMBALANCES IN OECD COUNTRIES Autores: Mariam Camarero(1)(*) Josep Lluís Carrion-i-Silvestre(2) Cecilio Tamarit(3) P. T. N.o 16/09 (1) Department of Economics. Jaume I University. (2) Department of Econometrics, Statistics and Spanish Economy. University of Barcelona. (3) Department of Applied Economics II. University of Valencia. (*) Corresponding author: Department of Economics, Jaume I University, Campus Riu Sec, E­ 12071 Castellon (Spain). Phone: 34+964728595. Fax: 34+964728591. e-mail: [email protected]; http://www3.uji.es/~camarero N.B.: Las opiniones expresadas en este trabajo son de la exclusiva responsabilidad de los autores, pudiendo no coincidir con las del Instituto de Estudios Fiscales. Desde el año 1998, la colección de Papeles de Trabajo del Instituto de Estudios Fiscales está disponible en versión electrónica, en la dirección: >http://www.minhac.es/ief/principal.htm. Edita: Instituto de Estudios Fiscales N.I.P.O.: 602-09-006-9 I.S.S.N.: 1578-0252 Depósito Legal: M-23772-2001 INDEX 1. INTRODUCTION 2. SOME STYLIZED FACTS ABOUT TH CURRENT ACCOUNT 3. BRIEF EMPIRICAL LITERATURE REVIEW 4. 1. 1. 1. 1. 1. 1. THEORETICAL FRAMEWORK 4.1. The classical flow equilibrium approach: sustainability of the current 1.1. account and the intertemporal budget constraint 4.2. The stock approach: the arithmetic of intertemporal solvency (Net 1.1. international debt to GDP ratio) 4.3. The unified approach of Gourinchas and Rey (2007): foreign debt and 1.1. the current account 5. ECONOMETRIC METHODOLOGY AND RESULTS 1. 1. 1. 1. 1. 1. 1. 5.1. 5.1. 1.1. 1.1. 1.1. 1.1. 1.1. Testing for current account sustainability and external debt solvency: panel analysis 5.1.1. Testing for the presence of multiple structural breaks 5.1.2. Testing 1(0) stationary on individual time series 5.1.3. The issue of cross-section independence 5.1.4. Panel data tests with cross-section dependence and structural 5.1.4. breaks 1. 5.2. Testing for the unified approach 1. 1.1. 5.2.1. Bai-Perron estimation results 1. 1.1. 5.2.2. VAR and impulse-response results: the case of US and Spain 6. CONCLUSIONS REFERENCES APPENDIX SÍNTESIS. Principales implicaciones de política económica —3— ABSTRACT In this paper we analyze the solvency and the sustainability of the current account for a group of twenty OECD countries using panel data methods. We test for the main hypotheses that have been formulated following both the flow and stock approaches in an intertemporal setting. Moreover, we also formulate a unified testable model based on Gourinchas and Rey(2007). The main results of the models can be tested using stationarity tests. For this purpose, we apply panel stationarity tests allowing for structural breaks and cross-section dependence. The evidence points to the solvency of the external accounts; however, the two variables evolve around a shifting deterministic component implying, hence, the non sustainability of the external position in most of the countries considered. We estimate, also allowing for structural breaks, the reduced-form parameters linking the two variables, following Gourinchas and Rey. This relationships is estimated for the sub-periods defined by the breaks. For the majority of the countries and time-periods the parameters are positive, smaller than one and significant, as expected. Concerning the dynamics, we find that solvency is reconvered after major shocks affecting the countries' external accounts. Keywords: Current account, panel data, structural breaks, cross-section dependence. JEL codes: F32, F41, C23. —5— Instituto de Estudios Fiscales 1. INTRODUCTION Since the beginning of the 1990s, current account (CA) imbalances have been widening considerably in the world economy. Economic globalization has meant an increase in international trade and capital mobility facilitating the financing of larger and more persistent current account imbalances. Among the OECD countries there is a clear trend toward larger imbalances, i.e. by 2007, the current account imbalances, whether surplus or deficit, of the OECD countries were more than twice as large as in 1988. However, the trend towards large imbalances is not confined to the OECD countries. These imbalances have been more acute between China and the oil exporter countries, on the one hand, and the US, on the other. Many emerging economies now show larger surpluses in their current accounts, although it may be necessary to distinguish between those who are enjoying a temporary surplus due to a favorable movement in the prices of their exports (as in the recent run-up in commodity prices) and those whose surpluses are the result of the pursuit of a particular development strategy. According to the World Trade Report 2008 (WTO, 2008), emerging East Asia has followed an export-led development strategy which was supported by exchange rate policies that anchored domestic currencies to the US dollar. It has been a successful development strategy resulting in the rapid mobilization and employment of tens of millions of workers. The means to bring this about is the cross-border transfer of goods and services to the centre country in exchange for financing its deficits (Dooley et al., 2007). The flow of savings to developed countries has also been encouraged by the lack of financial and capital market development in emerging Asian economies. The underdeveloped nature of the domestic financial or capital markets has become a bottleneck preventing the effective channeling of domestic savings into worthwhile investment projects at home. But the size of the imbalances has raised the key question of their sustainability and the nature of the adjustment process. Therefore, there has been a renewed interest in the study of the determinants of the dynamic adjustment of external imbalances. In part, larger current account imbalances reflect the impact of greater capital and financial market integration. A current account deficit reflects dissaving by domestic residents, an excess of absorption over income. The fact that it is occurring reflects a willingness by foreigners to finance that excess absorption by accumulating future claims on the earnings of domestic residents. As a consequence, net foreign liabilities have also been growing, generating concern that policy measures may be required if costly and destabilizing shifts in market sentiment are to be avoided. —7— The weight of experts' opinion suggests that these imbalances will ultimately decline although there is no consensus on when or on the manner, whether smoothly or abruptly, in which it would occur (Clarida, 2007). But there seems to be broad agreement that some combination of exchange rate and asset price changes would play a role during the process of adjustment. Studies of past adjustments in industrial countries point to the challenges ahead. Larger deficits take longer to adjust and are associated with significantly slower income growth during the current account recovery (Freund and Warnock, 2007). Consumption-driven current account deficits involve significantly larger depreciations than deficits financing investment. Obstfeld and Rogoff (2006) suggest that a large depreciation of the US dollar, something in the order of 30 per cent, could accompany the process. While temporary current account deficits may simply reflect the reallocation of capital to countries where capital is more productive, persistent deficits may be regarded as more serious. Deficits may lead to increased domestic interest rates to attract foreign capital. However, the accumulation of external debt due to persistent deficits will imply increasing interest payments that impose an excess burden on future generations. Now, adjustments to large current account imbalances are complex processes. The speed and economic effects depend on many factors. How much of the adjustment takes place through changes in asset valuation? How much through a reduction in absorption? How much in the form of expenditure switching? It will also matter how much international coordination among financial and central bank authorities takes place to ensure a supportive policy environment. Thus, the discussion above should not be seen as simplifying the challenges that are involved. If one can take a specific example, the soft-landing scenario requires that the acceleration of US export growth be matched by increased demand for US goods from the rest of the world. This would need to be triggered by just the right kinds of movements in exchange rates, asset and goods prices. Mann (2002) considers that sustainability should be viewed both from the domestic and international finance point of view. A sustainable current account is one that does not trigger feedback effects on domestic variables (investment and savings) or does not lead to significant international portfolio reallocations leading to changes in interest rates. We can distinguish three approaches in the theoretical literature that analyzes the current account balance. First, the conventional non-optimizing models, that comprises the Keynesian and monetary views, generally using reduced-form solutions and examining aggregated macroeconomic aspects. Although these models à la Mundell­ Fleming-Dornbusch provide a useful policy framework, the main drawback is that they are not based on microeconomic foundations and optimizing behavior of the economic agents. A second approach is the micro-founded intertemporal optimizing models developed in the 1980's that use the intertemporal budget —8— Instituto de Estudios Fiscales constraint. The major advantage of these models is that they deal with current and capital account behavior simultaneously through direct and portfolio investment flows across border along with trade in goods and services. The use of these models has facilitated the analysis of the sustainability of current account deficits. The intertemporal models developed until the late 1980's generally assumed perfectly flexible domestic prices and ignored the short-term price rigidities in product and factor markets. Finally, a third theoretical approach is the extension of the intertemporal models developed during the 1990's that introduced nominal rigidities and market imperfections into the dynamic general equilibrium models, being the Obstfeld-Rogoff Redux model the major milestone in the intertemporal approach to open-economy macroeconomics. These models provide a sound micro-theoretical framework, although they lack a matching empirical validation of the theoretical propositions. The empirical content in some of the models remains restricted to only calibrated simulations. The policy formulations at the central banks, government organizations, International Monetary Fund and the World Bank require an empirically tractable and econometrically estimable model to verify the theoretical propositions. More recently, some studies have extended the modern portfolio optimization theory to the current account and suggest that the marginal unit of wealth arising from a positive productivity shock is allocated according to the existing portfolio choices, and that changes in saving lead to changes in current account proportional to the share of foreign assets in total assets. Kraay and Ventura (2003) suggest that, in the long run, countries invest a marginal unit of saving in domestic and foreign assets in the same proportions as in their initial portfolios. In the short run, countries invest a marginal unit of saving mostly in foreign assets, and only gradually do they rebalance their portfolio back to its original composition. Countries not only try to smooth consumption, but also domestic investment, and they use foreign assets as a buffer stock. Lane and Milesi-Ferretti (2001, 2002) have examined the relationship between current account and changes in net foreign asset position at market value, and showed that the correlation between them is low or even negative. Lane and Milesi-Ferretti (2004) suggest that currency fluctuations influence the rates of return on inherited stocks of foreign assets and liabilities, in addition to operating through the traditional trade adjustment channel. The large gross cross-holdings of foreign assets and liabilities suggest that the valuation channel of exchange rate adjustment has grown in importance, relative to the traditional trade balance channel. More recently, Gourinchas and Rey (2007) have decomposed the external adjustment into a financial (valuation) channel and a trade (net export) channel and show that the deteriorations in net exports or —9— net foreign asset position of a country have to be matched either by future net export growth (trade adjustment channel) or by future increases in the returns of net foreign asset portfolio (financial adjustment channel). The valuation channel is important in the medium-term and the net export channel is important in a long-time horizon. The aim of this research is to test for sustainability following the framework defined in Milessi-Ferretti and Razin (1996) and Taylor (2002). According to this stream of the literature, it is possible to define two key concepts regarding the stochastic properties of the current account. First, the current account is said to be solvent if it is I(0) stationary. Second, the current account is sustainable if the economy is able to satisfy its long-run intertemporal budget constraint without a drastic change in private sector behavior or policy shifts. This is a more general concept and does not depend on any particular model. At the same time this concept of sustainability is a sufficient condition for other concepts to hold, with the advantage of its easy testability. According to Trehan and Walsh (1991), current account stationarity is a sufficient condition for the intertemporal budget constraint to hold. To this aim we first test for stationarity of two variables: the Current Account (CA) to GDP ratio and the Net Foreign Assets (NFA) position to GDP ratio. The first variable is representative of the traditional flow approach to the intertemporal budget constraint, while the second, provided that we use the stocks build by Lane and Milesi-Ferretti (2007), that consider the valuation effects in the financial markets, is already a methodological improvement compared to previous empirical work. Finally, as a second step we analyze a key relationship for the long run stability of the unified approach model of the current account adjustment as defined in Gourinchas and Rey (2007). For this purpose we use a panel data unit root test that allows for the presence of structural breaks and cross-section dependence. From an econometric point of view the contribution of this paper is twofold. First, we test for the presence of structural breaks affecting the CA time series, considering as a particular case the situation with no structural breaks. Once the presence of structural breaks has been investigated, then individual stationarity test statistics are computed. Second, such individual tests can be pooled to define panel data based test statistics, which permit an assessment of the CA stochastic properties using more powerful statistical tools. The statistical inference is conducted taking into account the presence of cross-section dependence through the computation of the bootstrap distribution and the use of approximate common factor models. As for the third variable analyzed, that is, the relationships between CA and NFA we first, use the Bai and Perron (1988) methodology in order to ascertain the possible structural breaks in the relationship on a country by country basis. It is worth mentioning that the application of the Bai-Perron methodology to estimate the number and position of the structural breaks requires the variables under analysis to be stationary in — 10 — Instituto de Estudios Fiscales variance, which is consistent with the null hypothesis and previous findings of the univariate analysis of the two ratios. Secondly, we analyze the relationship between both variables accounting for possible co-breaking relations in a dynamic heterogeneous panel setting. The remainder of the paper is organized as follows. Section 2 develops the present global imbalances situation in the economic relations at the world level, describing the main stylized facts. Section 3 displays a revision of the previous empirical literature, emphasizing the main issues related to the relationship between increasing economic integration and the external imbalances. In Section 4 we discuss the theoretical framework that guides our empirical investigation on the mechanisms of international financial adjustment. Section 5 presents our econometric methodology and describes the construction of our annual database for the OECD countries. The empirical results are presented in Section 6 and, finally, Section 7 concludes. 2. SOME STYLIZED FACTS ABOUT THE CURRENT ACCOUNT Prior to the empirical analysis developed in the next sections we will study the stylized facts associated with the current account and the external position of the developed countries. The first stylized fact that emerges is the intense degree of financial globalization that has occurred during the last decade, despite the financial crises and the reversal in global stock markets values in 2001-2002. During the last decade, the amount of financial wealth has increased steady in the world. The intensity of the process can be understood if we compare Figures 1 and Figures 2 and 3. Figure 1 shows the sum of exports and imports as a percentage of GDP for the Euro area countries, the UK, the US and Japan. Trade openness has increased steadily during the period considered. Even if the US and Japan are less open than the UK or the EMU countries, all of them have doubled, approximately, their level. This is in sharp contrast with Figures 2 and 3, where financial integration is measured as the sum of foreign assets and liabilities as a percentage of GDP, as proposed by Lane and Milesi-Ferretti (2008). We have presented this index selecting the countries according to criteria linked to their size and characteristics. The Southern and peripheral European countries (Spain, Ireland, Italy, Greece, and Portugal) are shown in the left graph of Figure 2. With the exception of Ireland (an outlier in comparison with the rest, due to its exceptionally high level), these countries have increased their financial integration around nine or ten times. The evolution is different in the graph on the right: larger Anglo-Saxon countries, such as the US, Canada, Australia and New Zealand, are four or five times more financially integrated than at the — 11 — beginning of the seventies. The exception is the UK in this case, a major financial center, especially after the advent of the euro. The behavior of the countries depicted in Figure 3 turns out to be similar to the other European countries mentioned above: financial integration has augmented around ten times. Thus, even if international trade has increased substantially, on the financial side international capital flows have expanded even more rapidly, and the financial linkages, as well as the real ones, have tightened across countries. A second stylized fact is that this process has caused important external imbalances in the current account. Figures 4 and 5 show the current account balance as a percentage of GDP for the same country groups as above1. In Figure 4, on the left graph, the peripheral Southern European countries have experienced significant deficits. The most dangerous positions, with deficits above 10% of GDP, are those of Portugal and Spain (and to a lesser extent, Greece); however, Italy and Ireland also worsened their balance after 2000. According to Blanchard (2007), these very large deficits in rich countries reflect mostly private saving and investment decisions. The question is why these countries have been able to experience such deficits without having suffered a reversal and, therefore, an adjustment. A possible explanation is that in a monetary union, the broad external balance of the European economy hides significant differences in external positions across individual European countries. Some of them, such as the Southern peripheral European countries are converging towards the core EMU countries. The external constraint that individual countries may face is not longer working in a monetary union. However, according to Lane and Milesi-Ferretti (2007), the exposures across Europe are very heterogeneous (differences in trade patterns, financial exposures, and net external positions) so that the process of adjustment may constitute an asymmetric shock. This implies bilateral real exchange rate adjustments between creditor and debtor countries as members of the Euro area. This heterogeneity can be observed in Figure 5: with the exception of France, after the euro, the “core” euro-area countries and the Nordic economies maintained a surplus in their current account balance. Therefore, some members of the EMU may experience significant and probably painful adjustments in their external position when the credit conditions become tighter. However, it should be emphasized that the EMU countries are not the only ones to have been affected, in recent years, by external imbalances. The case of the US has been discussed abundantly due to its magnitude and the persistence of the creditor position. Other OECD countries, such as New Zealand and Australia are also experiencing similar imbalances. 1 Caballero et al. (2008) divide the world into four groups: the United States (and similar economies such as Australia and the United Kingdom); the Euro Zone; Japan; and the rest of the world. This classification also emerges from our stylized facts analysis. — 12 — Instituto de Estudios Fiscales An alternative, although complementary, approach to the nature and dimension of external imbalances can be gathered by looking at the net foreign assets (NFA hereafter) position of this same group of countries. Figures 6 and 7 show quite clearly another stylized fact: the preeminence or the persistence of the net debtor positions among the developed countries. The only exceptions are Japan, Norway (the only oil exporting country in the sample), and a group of core EMU members (Germany, France and Belgium). The negative values of the NFA position (sometimes reaching 50% of GDP or even 75% as in Spain) reflect the cumulated effect of persistent current account deficits and therefore, the imbalance between foreign assets and liabilities. Many rich countries have benefited from the high degree of international financial globalization and have been able to finance their growing current account imbalances through foreign capital entries. However, the deterioration of the NFA position has been severe in many cases and calls for painful adjustments. 3. BRIEF EMPIRICAL LITERATURE REVIEW As the current account represents the rate at which a country accumulates or decumulates foreign assets, one approach to judging whether an external balance of a given size is a problem or not is to see whether it is consistent with the assumption that all external debts will ultimately be repaid. This is the notion of intertemporal solvency. This concept, however, is a relatively weak criterion as far as giving warning of an emerging problem. The reason is that solvency requires only that, in the very long run, all debts be repaid. Since this is equivalent to saying that large trade deficits today will be offset by equally (in present value terms) large trade surpluses in some future period, a country can remain technically solvent even while running large external deficits as long as policies are adjusted as needed in the future to bring about the required surpluses that enable debts to be repaid. Therefore, it can be argued that intertemporal solvency imposes too few restrictions on the evolution of the current account and external debt over the medium term to be of much operational value in telling us when a country's external position warrants attention from policy makers. A more demanding criterion is sustainability. This concept adds on to the notion of solvency the idea that policies remain constant for the indefinite future. Thus, an external position is sustainable if, under the assumption that policies do not change, the country does not violate its intertemporal solvency constraint. The problem with the sustainability concept is that what matters for the current account are people's expectations of future policies rather than the policies themselves. These expectations are notoriously difficult to — 13 — observe and measure, which makes the sustainability concept difficult to apply operationally. Economists do not agree on a precise definition of a sustainable current account. In general, sustainability refers to a stable state in which a current account deficit generates no economic forces of its own to change its trajectory. In this research, a country's current account deficit is defined as unsustainable when it triggers a sharp hike in domestic interest rates, a rapid depreciation or some other abrupt domestic or global economic disruption. Using this definition, a sustainable current account is one that changes in an orderly fashion through market forces without causing jarring movements in other economic variables, such as the exchange rate. The traditional Keynesian approach to the current account put the emphasis on international price competitiveness and relative demand in explaining current account movements. However, the intertemporal approach that appeared from the beginning of the 1980's has emphasized the role of forward-looking expectations in explaining current account patterns. The current account of a country is treated as a reflection of consumption and investment decisions that span over long-term horizons. Thus, the standard intertemporal model of the current account considers the current account from the saving-investment perspective and features an infinitely lived representative agent who smooths consumption over time by lending or borrowing abroad. As the global integration of the financial markets increased from mid 70's, there was a rapid expansion of two-way capital flows and gross external asset and liability positions that contributed to the creation and sustainability of current account imbalances. Therefore, the intertemporal approach became a more appropriate framework to analyze the dynamics of the current account. The intertemporal approach to the current account stresses that, since the current account is the difference between national saving and investment, external deficits or surpluses result from intertemporal investment and consumption decisions by firms, households and the government2. Thus, when international markets provide limited insurance opportunities, borrowing and lending enables economic agents to smooth consumption through intertemporal trade, enhancing economic efficiency. The empirical applications of this approach evolved along two main lines of research. The first strand of the literature applied the present value test, as developed by Campbell and Shiller (1987). Under some simplifying assumptions and using a methodology developed by these authors in a different context, one can estimate the current account series that would have been optimal from a consumption smoothing perspective. The standard model implication is that the 2 See Obstfeld and Rogoff (1995, 1996) for a survey of the literature. — 14 — Instituto de Estudios Fiscales current account balance equals the present value of expected future declines in net output (output less investment and government spending). The intertemporal approach to the current account was first popularized by Sachs (1981) and considers net accumulation of foreign assets as a way for domestic residents to smooth consumption intertemporally in the face of idiosyncratic income shocks. Namely, in response to positive temporary shocks to net output, domestic households can increase both current and future consumption by lending internationally, either directly or through financial institutions. Conversely, in response to permanent shocks that raise net output in the long-run by more than in the short run, domestic households can optimally smooth consumption by borrowing in the international financial markets. To the extent that the permanent increase in net output is driven by shocks to productivity, borrowing in international financial markets allow the domestic economy to sustain higher rates of domestic investment without cutting current consumption. For more that two decades, these basic propositions have been tested using variants of the present-value model originally conceived by Campbell (1987) and Campbell and Shiller (1987) with mixed results. Starting with Ahmed (1986) and Sheffrin and Woo (1990), economists have compared actual current account data with this optimal benchmark leading to the general result that while the model-predicted current account is positively correlated with the actual series, the latter is substantially more volatile, what implies a statistical rejection of the model. Although the positive correlation means that consumption-smoothing plays a role in the dynamics of the current account, the finding of excess current account volatility has been used to reject the proposition of limited international capital mobility, as stated by Feldstein and Horioka. The present value framework was then extended in several directions in more recent papers. These studies have tried to generate extra predicted volatility through real exchange rates and interest rates variability (Bergin and Sheffrin, 2000), by incorporating consumption habits (Gruber, 2004) or by adding an exogenous world real interest rate shock (Nason and Rogers, 2006). The extent to which the model performance is driven by the empirical failure of the auxiliary assumptions commonly adopted to make the model testable is unclear but has been claimed as the main reason for that. In addition, present-value tests do not distinguish between temporary or permanent shocks driving the dynamics of a country's net foreign liabilities. The second strand of the literature has applied standard econometric techniques to establish if there is a long-term relationship between the current account and macroeconomic fundamentals – i.e. relative GDP per capita, the demographic structure or fiscal policy3. Recent literature addressing these issues 3 See, for example, Debelle and Faruquee (1996), Chinn and Prasad (2003) or Bussiere et al. (2004). — 15 — has used DSGE models with non conclusive results4. Moreover, due to the lack of a precise definition, no universally accepted measure of sustainability exists. Many economists gauge sustainability by examining the value of a country's external obligations. In this context, two commonly used measures are the ratio of the country's current account deficit to GDP and the ratio of the country's net international debt to GDP. Insight into the causes of the deficit can be gained by looking at how the deficit is financed. Examining the ratio of net international debt to GDP provides an alternative method for assessing the sustainability of a country's current account deficit. Net international debt is the accumulation over time of current account deficits. If an economy runs a current account deficit consistently, net international debt may become so great that foreign investors lose confidence in the economy's ability to service its debt or, worse yet, repay the principal. Once this happens, interests rates must rise or the borrowing country's currency must depreciate to enable the country to continue financing its deficit. In this case, the current account deficit has generated economic forces of its own to change its trajectory, and the current account deficit and the associated debt have become unsustainable. In balance of payments terminology, net capital inflow is the financial counterpart of the current account deficit. Thus, current account positions which appear justified from such perspective can only materialize subject to the constraints implied by international capital flows. In other words, a country that is solvent may nevertheless not be able to finance a particular current account deficit if investors are not willing to provide the required funds, i.e. if the country is liquidity constrained. However, recent empirical literature trying to test for this approach still relies only on flows to assess the dynamics of the adjustment process5. From a theoretical perspective, the above flow approaches have a major drawback, as they ignore valuation effects of stocks of foreign assets and liabilities and assume that the current level of net foreign assets (NFA) is sustainable. Although this mechanism could help to a gradual rebalancing, these benefits could turn into a problem if policies are not consistent with a credible medium-term policy framework aimed at external and internal balances, as expectations may not be well anchored. In this case, investor preferences may quickly change and the fallout from disruptive financial market turbulence would likely be more elevated than it had been otherwise. Moreover, a country 4 See, for instance, Blanchard and Giavazzi (2002), Fagan and Gaspar (2007) or Bems and Schellekens (2007). 5 For example, Bussière et al. (2004) extend the standard intertemporal model by introducing habit formation and non-ricardian consumers to account for current account behavior in the OECD and in EU acceding countries. Similarly, Zanghieri (2004) extends this analysis by projecting the future level of debt using the forecasts of current account minus FDI flows. Depending on the assumed share of FDI in the current account deficit, CEECs' debt will be stabilized (high share of FDI) or will continue to grow (low share of FDI). — 16 — Instituto de Estudios Fiscales running persistent current account deficits might be at the same time improving its NFA position if capital gains on its foreign assets exceed those on its foreign liabilities (Lane and Milesi-Ferretti, 2006). Additionally, if the country is located away from its equilibrium level of NFA, the current account deficit can be sustained precisely because the economy is adjusting to a higher level of long­ term liabilities. Edwards (2001) shows that this adjustment process can lead to quite substantial current account deficits. In our view, a stock approach can cope successfully with this problem. Stocks are also less volatile and can provide long term relationships that are easier to estimate. The stock approach has recently been used by several authors thanks to the development of an external wealth database by Lane and Milesi-Ferretti (2006). They use their own estimates of external positions to study the determinants of NFA in developing and industrial countries and they find that public debt, GDP per capita and a set of demographic variables give a good account of the patterns of external holdings (Lane and Milesi-Ferretti, 2001). Calderon et al. (2000) use a dataset constructed by Kraay et al. (2005) to test a portfolio model on a set of developing and industrial countries. Gourinchas and Rey (2007) use monthly data and an intertemporal budget constraint view to measure external imbalances in the United States. External imbalances should be especially disruptive in developing markets. IMF (2005) uses a methodology close to Gourinchas and Rey (2007) to show the different roles played by valuation effects in emerging and industrial countries. The idea behind this is that valuation effects are destabilizing in developing countries because of liability dollarization (see for example Obstfeld, 2004). The more an economy is dollarized, the worse will be a reaction of its net foreign assets position to depreciation. And since the reaction of the exchange rate to excess external liabilities will be to depreciate, a dollarized indebted country should become even more indebted, unless it runs substantial trade surpluses. On the other hand, if these surpluses are not accompanied by a surge in productivity, it should take place thanks to a shift in demand from tradables to nontradables, and this is possible only through further real exchange rate depreciation. This mechanism initiates a vicious circle that badly affects firms' balance sheets and their capacity to invest, thus leading to an output collapse. Indeed, IMF (2005) finds that valuation effects play a stabilizing role in industrial countries but not in developing countries. Although stock imbalance measures proved useful at predicting future flows (see Lane and Milesi-Ferretti, 2001, IMF, 2005 and Gourinchas and Rey, 2007), no attempt has been made at using them to predict the particular phenomenon of sudden stops in capital flows, which can be defined as a sharp, disruptive reversal in the current account. The mean reversion property of current account has several implications for international macroeconomics. First, a stationary current account is consistent with sustainability of the external debts. In this case, there is no incentive for the — 17 — government to make drastic policy changes and default on its international debts in the near future. Second, stationarity of the current account validates the modern intertemporal model as, theoretically, the model combines the assumptions of perfect capital mobility and consumption smoothing behavior to postulate that the current account acts as a buffer to smoothing consumption in the event of shocks. From an empirical point of view, the stationarity and sustainability of OECD current account balances has been the focus of many researchers over a number of years6. The literature on the sustainability of the current account examines the question within two alternative frameworks. On the one hand, a time series perspective is employed where researchers investigate either the long-run relationship between exports and imports or the stationarity of the external debt process (see Chortareas et al., 2004)7. With the exception of Liu and Tanner (1996), who consider the impact of structural breaks, the above mentioned studies generally find that current accounts are non-stationary for OECD countries. Tests that rely on linear approximations are likely to be imprecise on short samples when the observed current account is persistent, as it typically is. A persistent current account does not necessarily mean a non-stationary one. A stationary current account will be considered persistent if its process of mean reversion is slow. The small sample problems can occur even if the current account is stationary but persistent. Therefore, panel data can improve the information in relatively short sample databases. On the other hand, panel unit root techniques have been employed since unit root tests applied to single series suffer from low power. In recent years a number of alternative procedures have been proposed to test for the presence of unit roots in panels that combine the information from the time series dimension with that from the cross-section dimension. Studies that employ panel data methods include Wu (2000), Wu et al (2001), Holmes (2006) using Im, Pesaran and Shin (2003) test (IPS) and cointegration tests. However, due to the heterogeneous nature of the alternative hypothesis in their test, one needs to be careful when interpreting the results, because the null hypothesis that there is a unit root in each cross section may be rejected when only a fraction of the series in the panel is stationary. Moreover, the hypothesis that the current account balances or the debt series adjust to long-run equilibrium in a continuous fashion is troublesome when we suspect that in many cases discontinuities may be present in the mean-reversion process. We can identify several sources of breaks rooted in policy or institutional investors' behavior. 6 See, inter alia, Trehan and Walsh (1991), Otto (1992), Wickens and Uctum (1993), Liu and Tanner (1996), Wu (2000), Wu et al. (2001), Holmes (2006) and Holmes et al. (2007). 7 The strand of this empirical literature using single equation unit root tests usually rejects the mean reverting behavior of the current account. See, among others, Husted (1992), Ghosh (1995), or Bergin and Sheffrin (2000). — 18 — Instituto de Estudios Fiscales The presence of high government debt may have repeatedly induced abrupt corrective actions requiring sudden adjustments. More specifically, the fiscal health of the government tends to affect international investors' perception of expected profitability and the investment climate in the country. This perception, in turn, may trigger abrupt reversals in capital flows and imbalances in the current account (i. e. the EMS crisis in the early 1990s). Another channel that may lead to discontinuities in the series is the level of a country's indebtedness, which reflects the willingness of foreign lenders to hold domestic assets. Investors may be unwilling to lend beyond a level of foreign debt they consider “normal” and withdraw large amounts of funds, creating majors imbalances in the balance of payments. 4. THEORETICAL FRAMEWORK 4.1. The classical flow equilibrium approach: sustainability of the current account and the intertemporal budget constraint According to Taylor (2002) sustainability of the current account can be defined as the ability of an economy to satisfy its long-run intertemporal budget constraint without a drastic change in private sector behavior or policy shifts. It views the current account as the equilibrium outcome of forward-looking saving and investment decisions taken by rational individuals and driven by expectations of productivity growth, government expending, interest rates and other factors. This view emphasizes the role of the current account as a buffer against transitory shocks in productivity or demand in order to smooth the intertemporally-optimal consumption path. As we previously claimed, this is a rather general concept and does not depend on any particular model, with the advantage of its easy testability. According to Trehan and Walsh (1991), current account stationarity is a sufficient condition for the intertemporal budget constraint to hold. Consider a stochastic model with zero growth. The one period budget constraint is, C t + It + Gt + NFA t = Yt + (1+ ii )NFA t −1, (1) where Ct ,It ,Gt,NFA t and Yt are consumption, investment, government consumption, net stock of debt and income respectively. it is the world interest rate. Rearranging (1) and from national accounts identities we have that, NFA t = (1+ i i )NFA t −1 + NX t , — 19 — (2) where NX t is the net exports. Iterating (2) forward and assuming that the expected value E(i t | ϕ t−1 ) = i , with ϕ t−1 being the information set available in t − 1, we get ∞ j T ⎛ 1 ⎞ ⎛ 1 ⎞ NFA = ⎟ E(NFA t +T | ϕt−1). ⎜ ⎟ E(NX t + j | ϕt −1) + lim ⎜ T→∞⎝ 1+ r ⎠ 1+ r ⎝ ⎠ j=0 ∑ (3) Equation (3) simply states that international agents are able to lend to an economy if they expect that the present value of the future stream of next exports surpluses equals the current stock of foreign debt. Hence, the sustainability hypothesis, or long run budget constraint implies that: T ⎛ 1 ⎞ lim ⎜ ⎟ E(NFA t+T | ϕ t−1 ) = 0 t→∞⎝ 1+ i ⎠ (4) This transversality condition means that the present value of the expected stock of debt when t tends to infinity must equal zero, that is, a no-Ponzi game condition. Following Trehan and Walsh (1991), given that the current account CA t = NFA t − NFA t−1 , a sufficient condition for (4) to hold is that the current account is an I(0) stationary process. In the more realistic case of an economy with a positive rate of growth of output, we have that the sustainability condition t holds if the ratio ca t = CA is I(0) stationary. This means that sustainability is Y t possible with current account deficits as far as they do not grow faster than output in expected value. An obvious test of sustainability is hence a unit root test on ca t . This is what most of the literature has previously used as a test of sustainability. However, note that we are dealing here with expected values of future events. Changes in the agents' perceptions about risk, portfolio allocation decisions, future policy changes, transaction costs in international financial flows, among others, can lead to changes in the dynamics of current account mean reversion and, hence, equilibrium values of the current account. As previously mentioned, Taylor (2002) sees the speed of convergence towards equilibrium as a summary statistic of the degree of capital mobility. This is because it reflects how agents are prepared to allow for periods of current account deficits (surpluses) above the perceived equilibrium value. If, given the international financial environment, agent's perceptions about, for instance, the relative risk of US denominated assets changes due to large observed current account deficits, the speed of mean reversion and the mean of the current account itself would also change. That is, changes in the current account affecting the agent's perception can trigger adjustment dynamics leading to discontinuities in the time series. In this sense, it may be the case that tests that do not consider the existence of breaks are misspecified and reach wrong conclusions about the sustainability of the current account or arrive at too simplistic descriptions — 20 — Instituto de Estudios Fiscales of the current account dynamics. Moreover, the special nature of the financial markets, characterized by contagion effects may give rise to sudden stops or even reversals in the asset holdings leading again to breaks in the time series and to the existence of cross-section dependence. This fact may again lead to misleading conclusions. In this research we overcome these two problems through a new panel unit root test that considers the existence of multiple breaks and cross-section dependence. Although the saving-investment equilibrium approach does provide an analytical basis for the evaluation of external positions, its almost exclusive concern with flows limits its ability to assess the viability and adequacy of external indebtedness, a stock problem by nature. 4.2. The stock approach: the arithmetic of intertemporal solvency (Net international debt to GDP ratio) The arithmetic of solvency starts from the notion that an economy is intertemporally solvent if its (net) foreign indebtedness is no larger than the present discounted value of the stream of its future non-interest surpluses. The practical difficulty with this approach is that in principle any level of external debt is consistent with solvency provided that sufficient trade surpluses are generated in the indefinite future (Milesi-Ferretti and Razin, 1996). Thus, to make this approach operational, researchers typically assume that the economy targets a given debt-to-GDP ratio, and consider the particular case in which current policy would remained unchanged into the indefinite future (Corsetti and Roubini, 1991). The arithmetic of solvency is primarily concerned with the question of whether net external liabilities grow less rapidly than their (marginal) rate of return so that the present discounted value of net liabilities converges to some finite quantity. In practical terms, the arithmetic of solvency examines whether the net debt/GDP ratio grows more or less rapidly than the difference between the real interest rate and the economy's growth rate. Following Chortareas et al (2004) let us start with a stylized version of the nominal balance of payments identity defined in domestic currency: NFA t = ER t ∆L t − ∆A t ≈ NX t + i∗t ER t L t−1 − i t A t−1 (5) where NFA t is the net foreign assets position, NX t stands for trade balance, A(L) are domestic (foreign) assets held by foreigners (domestic residents), i (i∗ ) are the nominal rates of return on domestic (foreign) assets, and ER is the domestic price of the foreign exchange rate. Deflating by nominal GDP, and regrouping terms, the former identity can be rewritten as: ∆nfa t ≈ c t + ~ rt nfa t−1 — 21 — (6) where c t = τ t + (i t − i∗t − e t )b ∗t−1 is the primary current account deficit, b t − b ∗t is net foreign indebtedness, τ t is net exports, and ~rt = it − p t − y t is the growth-adjusted real return on net foreign debt. Further, e = ∆ logER t , p = ∆ logPt , and y t = ∆ log Yt , all other lower case letters denote variables as a ratio to nominal GDP. If (6) is deflated by a price index, f and c are real foreign debt and current account, and ~r is the real interest rate. Assuming ~r > 0, solving (6) forward, and imposing the no-Ponzi game condition, the Intertemporal Budget Constraint (IBC) is: nfa t = n nfa t = − ∑ ρ t c t+i (7) i=1 (1 + ~rt+s )−1 If this conditions holds, current and future discounted with ρ t = Π ns=1 primary current account surpluses are sufficient to pay off initial indebtedness. The traditional sustainability approach tests for the stationarity on nfa t . In the present exercise we take account of the valuation effects of stocks of foreign assets and liabilities using the new External Wealth of Nations Mark II (EWN II) database provided by Milesi-Ferretti and Lane (2007). According to them the size of countries' external portfolios is now such that fluctuations in exchange rates and asset prices cause very significant reallocations of wealth across countries, playing the exchange rate a dual role influencing both net capital flows and net capital gains on external holdings. 4.3. The unified approach of Gourinchas and Rey (2007): foreign debt and the current account In this subsection we summarize the model developed by Gourinchas and Rey (2007). This model follows an intertemporal approach and is based on two elements: an intertemporal budget constraint and a long run stability condition. They start from a country's intertemporal budget constraint and derive two implications. The first one is a link between the net foreign asset position and the future dynamics of the current account. If total returns on NFA are expected to be constant, today's net foreign liabilities must be offset by future trade surpluses (the so called “trade channel”). However, in the presence of stochastic asset returns, the expected capital gains and losses on gross external positions constitute a complementary adjustment tool called the “valuation channel”. The external constraint implies that today's imbalances must predict either future changes in the trade balance (flow adjustment), future movements in the returns of the NFA portfolio (changes in the stock of foreign assets), or both. In the short and medium term, most of the adjustment goes through asset returns, whereas at longer horizons it occurs via the trade balance. — 22 — Instituto de Estudios Fiscales The value of assets owned by domestic residents held abroad (A) minus the value of domestic liabilities to the rest of the world (L) is called the national net foreign asset position (NFA) . If its net foreign asset position is positive (NFA > 0) , the country is a net creditor to the rest of the world. Conversely, if NFA is negative (NFA < 0) then the country is a net debtor, because its outstanding liabilities to the rest of the world exceed its claims on the rest of the world. All nations are subject to a budget constraint that requires that the value of gross domestic expenditure, (GDE) or absorption, plus the change in the stock of foreign assets owned by domestic residents (A t − A t−1 ) equals the value of gross domestic product (GDP) plus the change in the stock of domestic debt owed to foreigners (L t − L t −1 ) . Combining this relationship with the definition of the current account, it follows that the change in the net foreign asset position is the same as the balance on the current account. (GDPt − GDE t ) + NFIt + UTt = (A t − L t ) − (A t−1 − L t−1 ) (8) Substituting in the definition of the net export balance (NX t = GDPt − GDE t ) and net foreign asset position (NFA t = A t − L t ) , this simplifies to: NX t + NFIt + UTt = CA t = NFA t − NFA t−1 (9) which says that the change in the net foreign asset position is the sum of net exports, net foreign income, and unilateral transfers or the balance on the current account. Therefore, if the current account is in deficit (CA < 0) , the change in the net foreign asset position is negative, indicating that the increase in foreign debt was greater than the increase in foreign assets over the year. A negative change in the net foreign asset position is referred to as a net capital inflow, since more capital flowed into the country through additions to the level of foreign debt than flowed out through purchases of foreign assets. Future current account and net foreign asset positions are related to the present current account and net foreign asset positions through future net foreign income flows8. The extent of these flows is influenced by the rates of return on foreign assets and foreign debt. Net foreign income is essentially the difference between interest earned on foreign assets and interest paid on foreign liabilities: NFIt = r A A t−1 − r L L t−1 (10) where r A is the rate of interest residents earn on their foreign assets and r L is the rate of interest that the country pays on its foreign liabilities. Theoretical 8 See Lane and Milesi-Ferretti (2002) and Gourinchas and Rey (2007) for a more complete discussion of the longer term relationship between the US net exports deficit and revaluations of the US net foreign asset position. — 23 — analyses typically assume that there is no differential between the interest rate on foreign assets and debt, and that the interest rate on foreign debt exceeds the growth rate of nominal GDP, which suggests that the economy must shift to a net export surplus to maintain its current negative net foreign asset position. In textbook examples there is no distinction between r A and r L , because they assume there is only one traded asset. However, this assumption is far from reality9, so it is important to allow for differences between r A and r L : NFIt = (r A − r L )A t−1 + r L (A t−1 − L t−1 ) = (r A − r L )A t−1 + r L NFA t−1 (11) Substituting expression (11) into (9) above, it follows that: NX t = NFA t − (1 + r L )NFA t−1 − (r A − r L )A t−1 − UTt (12) Dividing through by the level of GDP and imposing the foreign debt sustainability condition that the ratio of NFA to GDP be constant at nfa∗ , we find that the critical net exports to GDP ratio, nx ∗ at the current gross foreign asset to GDP ratio a∗ and typical unilateral transfer to GDP ratio ut ∗ is: nx ∗t = (g − r L )nfa ∗ − (r A − r L )a ∗ − ut ∗ (13) where g is the growth rate of nominal GDP. According to Kouparitsas (2005) when economists want to assess sustainability of the current account, they begin by calculating the net exports to GDP ratio that would be required to maintain the current net foreign assets to GDP ratio, nfa∗ . He refers to this as the critical net exports to GDP ratio, nx ∗ . Net exports to GDP ratios above nx ∗ will increase the nation's net foreign assets to GDP ratio above nfa∗ , while net exports to GDP ratios below nx ∗ will decrease it. For reasons explained below, negative net foreign asset positions are typically associated with a positive nx ∗ . Another way of stating this is that a country must give up a fraction of all future GDP equal to nx ∗ to maintain its current negative net foreign asset position. A country's current net foreign asset position is considered unsustainable if the associated nx ∗ is a relatively large fraction of GDP. Similarly, a current account deficit is considered unsustainable if it maintains or leads to an unsustainable net foreign asset position. According to this analysis, nx ∗ depends not only on nfa∗ , which is weighted by the difference between the growth rate of nominal GDP and the interest rate on foreign debt, 9 The experience is inconsistent with standard theoretical assumptions in many countries, like the US. First, the return the US earns on its private foreign assets exceeds the rate it pays on its private foreign debt. Second, on average, rates of return on most classes of US foreign debt have been roughly equal to the growth rate of nominal GDP. — 24 — Instituto de Estudios Fiscales but also the current ratio of domestic gross foreign assets to GDP, a∗ , which is weighted by the difference between the interest rates on foreign debt and foreign assets, and the typical ratio of unilateral transfers to GDP, ut ∗ . A by-product of this analysis is the current account to GDP ratio, ca ∗ , that would be required to maintain nfa∗ . ca ∗ only depends on nfa∗ , which is weighted by the growth rate of nominal GDP. Through a similar analysis, one can show that the critical current account to GDP ratio ca ∗ is: ca ∗ = g ⋅ nfa ∗ (14) Moreover, many statistic databases do not take into account the unrealized capital gains from both changes in local currency prices and exchange rate adjustments and this mechanism can be of increasing importance in a financially integrated world10. Although the theoretical relation (14) should hold between the current account and the net foreign assets, Gourinchas and Rey (2007) argue that they use trade balance instead of the current account to avoid possible discrepancies in the valuation of capital gains11. Consequently, they consider the accumulation identity for net foreign assets between t and t + 1 : NFA t+1 = R t+1(NFA t + NX t ), (15) where NX t are the net exports (difference between exports, X t , and imports, Mt ) and NFA t are net foreign assets (difference between gross foreign assets, A t , and gross foreign liabilities, L t , measured in the domestic currency). According to equation (15), the net foreign position would increase with net exports and with the total return on the net foreign asset portfolio R t+1 (see equation (2) above) . Next, the model is log-linearized. The following assumptions should hold: Assumption 1: (a) The ratios A t / W t , L t / W t , X t / W t and M t / W t are stationary, where Wt represents total household wealth. (b) The steady state values of the ratios, denoted µ aw , µ lw , µ xw and µ mw respectively, satisfy µ aw ≠ µ lw and µ xw ≠ µ mw . Assumption 2: The growth rate of household wealth Wt+1 / Wt is stationary with steady state value γ . 10 See Lane and Milesi-Ferretti, (2001) and Tille (2003) for detailed discussion of the size and history of valuation effects for the US and other nations. 11 However, in order to ease the comparison between the three theoretical approaches that we present in this research, in the empirical application we substitute CA t by NX t in equation (15). — 25 — Assumption 3: The return to the net foreign asset portfolio R t is stationary with a steady state value R that satisfies γ < R . Concerning Assumption 1, this is not particularly restrictive. The first part of the assumption would be verified in any model where exports, imports, external assets, liabilities and household wealth grow at the same rate along a balanced growth path. This will be the case in a wide variety of models, as long as assets and liabilities are not perfect substitutes. The second part of the assumption guarantees that some ratios are well defined. Assumption 2 is also an implication of the existence of a well-defined balanced growth path. Assumption 3 has an intuitive interpretation in this context: manipulating equation (15) if Assumption 3 holds, the steady state ratio of net exports to net foreign assets is stationary with an unconditional mean NX / NFA that satisfies: NX = ρ − 1< 0, NFA (16) where ρ = γ / R < 1 implies that the real growth rate of wealth is lower than the rate of return of the net foreign asset portfolio. Therefore, countries with steady state creditor positions (NFA > 0) should run trade deficits (NX < 0) , whereas countries with steady state debtor positions (NFA < 0) should run trade surpluses (NX > 0) . Lane and Milesi-Ferretti (2002) point out that the correlation between the change in the net foreign asset position at market value and the current account is low or even negative. They also note that rates of return on the net foreign asset position and the trade balance tend to commove negatively, suggesting that wealth transfers affect net exports. Moreover, Lane and Milesi-Ferretti (2004) show exchange rate effects on rates of return of foreign assets and liabilities. 5. ECONOMETRIC METHODOLOGY AND RESULTS In this section we describe the testing strategy we use to address the theoretical issues described above. The empirical application is based on a panel database that consists of 20 OECD countries, both European and from the rest of the world. The sample covers the period 1970-2006, and the data has been obtained from the World Bank and the new External Wealth of Nations Mark II (EWN II) database provided by Milesi-Ferretti and Lane (2007a). The two variables of interest are the current account balance as a percentage of GDP ( ca i, t ), and the net foreign assets stock also as a percentage of GDP ( nfa i,t ). We first test, using panel methods, both for the sustainability of the current account and the solvency of the net foreign assets position of our group of countries. Then, following what we call the “unified approach” by Gourinchas and Rey — 26 — Instituto de Estudios Fiscales (2007), we study the stability and the sign of the relationship linking the two variables. Concerning the two first hypotheses, we have applied panel data based test statistics following a two-step testing strategy that addresses the problems related to the issues of multiple structural breaks and cross-section dependence12. First, we have tested for the sustainability of the current account and for the external solvency by allowing for multiple structural changes in a panel setting that, to the best of our knowledge, has not been applied yet in this literature. Previous evidence has revealed that there might be some events that affect the current account and the net foreign asset position in a permanent way. It is well known that non accounting for structural breaks biases both unit root and stationarity tests towards concluding in favor of non-stationarity in variance13. Thus, this feature should be of special interest in our case, since this type of variables may be affected by major events such as currency crises or economic integration processes during the analyzed period. Second, we consider the existence of cross-section dependence amongst the individuals in the panel. Cross-section independence is hardly found in practice, especially when using macroeconomic time series that derive from globalized financial markets, as it is the present case. Moreover, it is worth mentioning that the existing literature has evidenced an increase in the degree of market integration, which should lead to higher correlation between financial and macroeconomic aggregates at the international level. As panel data unit root and stationarity tests are known to be biased towards concluding in favor of variance stationarity when individuals are cross-section dependent –see O'Connell (1998) and Banerjee, Marcellino and Osbat (2004, 2005)– the issue of cross-section dependence is of great importance. Therefore, we suggest computing the test statistic in Pesaran (2004) and Ng (2006) to assess whether the individuals in the panel are cross­ section independent. Furthermore, Ng's (2006) statistic is quite convenient since, in addition to testing for the null hypothesis of cross-section independence, it provides guidance about the best way to model cross-section dependence. The application of this statistic reveals that cross-section dependence is present in the panel data sets that we study. Then, our analysis considers two different ways to accommodate cross-section dependence. First, following the approach by Carrion-i-Silvestre et al. (2005) we compute the bootstrap critical values of the panel data stationarity test statistic, which allows us to consider a 12 We have applied as well classical panel unit root and stationarity tests without structural breaks finding mixed results. These results are available upon request from the authors. 13 See Perron (1989) for univariate statistics, or Carrion-i-Silvestre, del Barrio and LópezBazo (2001) for panel data statistics. — 27 — wide form of cross-section dependence. Second, we compute the panel data unit root and stationarity test statistics proposed in Harris et al. (2005) and Bai and Carrion-i-Silvestre (2009), which model the presence of cross-section dependence through the estimation of approximate common factor models as in Bai and Ng (2004). In both cases, the analysis considers the existence of multiple structural breaks. In addition, the approach that is adopted here is general enough to consider the non-break situation as a particular case embedded in the testing procedure. Therefore, our analysis does not impose the existence of structural breaks, but accounts for the possibility that they are present in the data. Finally, note that proceeding in this fashion accounts for the existence of a tension or trade-off between cross-section dependence and misspecification concerning the presence of structural breaks: the former introduces a bias towards stationarity in variance while the bias due to the latter goes in the opposite direction. This feature implies that the empirical analysis of the current account balances should be addressed carefully to avoid the effects of this tension. Then, we analyze the relationship between the current account balance (cai, t ) and the stock of net foreign assets (nfai, t ). Once the panel methods have concluded that the two variables are stationary, we can use the Bai-Perron procedure to assess the stability and the sign of the relation linking them. This will constitute a test of the unified approach of Gourinchas and Rey (2007) as described above. Its empirical implementation on a large cross-country time series sample poses two main issues. First, the model defines a long-run relationship between the ratio of CA and NFA. However, given the imperfections in international financial and factor markets, stock equilibrium does not hold in every point in time but is achieved gradually in the long-run. Therefore, in the empirical analysis, this relation might also present instabilities. Second, it seems reasonable to assume that countries can differ in the market imperfections and barriers to portfolio reallocation that govern the short term dynamics, and perhaps even in the parameters characterizing the long-run equilibrium. Thus, we must take into account the very likely possibility of parameter heterogeneity across countries. In the third part of this paper we first propose to apply in a single-country context the methodology developed by Bai and Perron (1998) to ascertain possible breaks in the relationship. Additionally, we use standard VAR methods of estimation to capture the dynamics of the adjustment. A salient feature of our analysis is that we do not use detrended variables as in Gourinchas and Rey (2007). We think that this would neglect relevant information that is embedded in the trend of the variables. As Gourinchas and Rey (2007) claim, trends might be representing structural changes in the world economy, such as financial and trade globalization. Therefore, we have considered an econometric strategy — 28 — Instituto de Estudios Fiscales that allows identifying and including the possible structural breaks in the estimation of the relationships between the variables. 5.1. Testing for current account sustainability and external debt solvency: panel analysis We use panel data methods in the empirical analysis of the theoretical hypotheses described in subsections 4.1. and 4.2. above. For simplicity, we present the testing strategy and the results simultaneously, as the same tests and procedures are applied in the two cases. However, in the discussion, we will explain separately the main conclusions. 5.1.1. Testing for the presence of multiple structural breaks The first stage of our analysis consists of assessing the presence of structural breaks affecting the ca i, t and nfa i,t time series using the following specification: y i,t = α i + mi ∑ θi,k DUi,k,t + e i,t , (17) k =1 where y it is the variable of interest, whereas t = 1,K,T , i = 1,K,N , with DUi,k, t = 1 for t > Tbi,k and 0 elsewhere – Tbi,k denotes the k th break point for the { } i th individual, k = 1,K,m i – and where e i,t are assumed to be a stationary process satisfying the strong-mixing conditions given in Phillips (1987) and Phillips and Perron (1988). This specification permits a high degree of heterogeneity assuming that the structural breaks may have different effects on each individual time series. For this purpose, the break points are located at different dates for each individual, and the individuals may have different number of structural breaks. Under these conditions, the estimation of the number and position of the structural breaks, if any, can be carried out using the sequential testing procedure proposed by Bai and Perron (1998). When computing the statistic we have to specify a maximum number of structural breaks, which in this case has been set equal to m i = 5 ∀i . The number of structural breaks is estimated using critical values at the 5% level of significance. It is worth mentioning that the application of the Bai-Perron methodology to estimate the number and position of the structural breaks requires the variables under analysis to be stationary in variance, which is consistent with the null hypothesis that we have specified, i.e., that the solvency hypothesis holds. Furthermore, the test statistic that is used is consistent against the alternative hypothesis of non-stationarity in variance, even when structural breaks are present in the analysis – see Lee, Huang and Shin (1997), Kurozumi (2002) and, Carrion-i-Silvestre (2003), among others. — 29 — Panel A in Table 1 reports the estimated number and position of the structural breaks for each individual in the current account panel data set. We can see that, except for Italy and New Zealand, the procedure detects at least one structural break for each time series, which indicates that previous analyses in the literature that do not account for the presence of structural breaks may have led to misleading conclusions. It should be stressed that the estimated number of structural breaks does not attain the maximum that has been defined. Concerning nfa i,t , we present the breaks and their position in Panel A of 4. With the exception of Germany, we find at least one structural break in all the countries in our group. We should note that the two variables we consider, current account balance and nfa i,t , are very different in nature: whereas the first one is a net flow, the second one, nfa i,t is a cumulated series (a stock). Therefore, we do not expect to find that the breaks are placed at the same dates for the two variables. As the literature on current account reversals describes (see Freund (2005) and Debelle and Galati (2007), for example), the adjustment usually takes place once the current account imbalance reaches a certain threshold. The effect on the net foreign asset position would critically depend on the relative position of assets and liabilities when the event takes place. Figure 8 depicts the CA time series for all the countries involved in our analysis along with the estimated deterministic component. The countries have been divided according to their condition of EU members during the studied period. This presentation allows us to establish a comparison of the break dates and the direction of the changes that have been estimated. In Table 2 we present an approximation to the main events explaining the structural breaks found in the data. We have ordered the countries following two criteria: (i) their EMU (or EMS) membership and (ii) their external position in terms of the current account. Two countries, Ireland and France could not be clearly placed in a current account category and, hence, they are considered separately. In the Table we have limited ourselves to the main milestones in European integration and international economic events. Other issues, however, may explain a particular structural break. We next analyze the countries individually. At the beginning of the 70's, the first oil shock triggered the collapse of the Bretton Woods system inducing effects on different countries. Belgium and Austria, decided to link its currency to the Deutsche Mark at the end of Bretton Woods – therefore, a policy change may have happened in 1974 and 1975 for Belgium and Austria, respectively. Two non-EMU countries suffered structural changes at the beginning of the eighties. Australia in 1980, when the dollar experienced a depreciation linked to a terms of trade worsening – in 1979 the Australian financial market — 30 — Instituto de Estudios Fiscales experienced a process of deregulation, and the dollar freely floated in 1983. The break in Norway in 1979 is possibly linked to the increase in oil prices. A large group of countries have a structural break in the mid-eighties. Both Belgium and Germany followed recovery programs. For example, president Martens in Belgium devalued the frank in 1982 and started an export-led policy. Ireland also devalued in 1983 in an answer to a twin deficits problem, followed by a tight fiscal policy14. Austria in 1980 started a system of cooperative arrangement for its exchange rate. Finally, Portugal suffered a deep recession, with terms of trade losses, fiscal deficits and increase in foreign debt service. Concerning non-EMU countries, the Reagan administration started a program at the beginning of the eighties that reduced policy intervention and allowed the free floating of the dollar. In early 1981, the new Reagan Administration decided to move away from what it judged to have been the heavy intervention inherited from the previous administration. From 1981 through early 1985, the dollar continued to strengthen, for several reasons. US monetary conditions were restrictive in the context of a robust recovery, and prospects for continued large US fiscal deficits exerted upward pressure on real interest rates. Meanwhile, monetary authorities abroad initially were reluctant to raise interest rates because their recoveries appeared more fragile. Investment, including foreign investment, boomed in the United States, attracted by the increasingly favorable business climate. In addition, dollar­ denominated assets were sought as a safe haven following the onset of the international debt crisis and amid apprehensions about the political situations in some European countries. Another large group of structural changes is found during the first half of the nineties. Most of the breaks are linked to the free capital movements in Europe and the German Unification in 1990, together with the EMS crises in 1992 and 1993. Portugal and France suffered a slowdown in economic activity in an effort to fulfill the Maastricht criteria. In the case of Austria, EU membership occurred in 1995, together with Sweden and Finland. The only structural break that Finland suffered occurred in 1994, the year of the referendum for EU accession. Sweden presents two structural breaks: the first one (in 1994) can be related to inflation targeting policy that started in 1993, whereas the second one (in 2001) is placed at the peak of an economic expansion. Finally, the end of the nineties and the beginning of 2000 accumulates another group of structural changes. Those in EMU countries and the US are 14 Membership of the EMS always posed problems for Ireland by virtue of the fact that the UK, the country's major trading partner, is not a member of the system. Such problems became most acute when a depreciation in Sterling put pressure on Irish companies in traditional industrial sectors. Such considerations prompted a devaluation of the Irish pound at the March 1983 re-alignment. — 31 — linked to the creation of the monetary union in 1999, the launching of the euro in 2001 and its effects on the dollar15. At the same time, Norway established an inflation targeting strategy, whereas Sweden, also outside the EMU, experienced an economic expansion. In contrast, the Asian crisis affected the demand of commodities and deteriorated Canadian dollar (and its terms of trade, suffering an adverse current account shock). Beginning in the summer of 1997, Malaysia, Indonesia, Thailand and South Korea (and some other Asian countries) fell into a serious recession, sparked by the collapse of their pegged­ exchange-rate regimes. As these countries are large users of raw materials, their recessions led to a significant fall in the world's demand for raw materials, and thus a large decline in raw materials prices. In the next year or so, the average prices of raw materials fell by about 30 per cent. All countries that export raw materials experienced a sudden decline in demand for their currencies, which lost value as a result – Canada, New Zealand, and Australia. This type of shock is a negative current-account shock, because it reflects a reduction in the demand for Canadian goods or services, the transactions of which are recorded in the country's current account of the balance of payments. In Japan, the real estate bubble burst and the current account was declared to be a monetary policy target. From the joint analysis of Table 4 and Figure 9, a similar pattern can be found in the structural breaks of nfa i,t . In Table 4 the breaks are classified according to the type of country (EMU member or third country) and its NFA position (positive or negative). In addition, an arrow indicates the direction of the break (where ↑ stands for an improvement in the position and ↓ for a worsening). With two exceptions (Australia in 1978 and Belgium in 1977), the rest of the breaks are accumulated in the mid-eighties and the end of the nineties-2000. The economic and political events described above are also valid for the NFA variable and describe the reasons for the occurrence of the breaks. We have compared the structural breaks found in the current account balance over GDP with the results of a strand of literature that studies the current account reversals16. The main conclusion that can be derived is that the structural changes detected in the variable coincide with reversals or 15 The dollar broadly strengthened against other currencies after the mid-1990s because market participants expected to receive higher rates of return on their investments in the US than abroad. For example, consider for a moment the fate of the euro versus the dollar since the euro's launch on january 1, 1999. The dollar strengthened by 30% against the euro primarily because market participants anticipated brighter prospects and higher rates of return in the US than in Euroland, and capital flowed out of euro-denominated assets into equities, bonds and other US investments. 16 See, for example, Freund (2005), Debelle and Galati (2007) and De Hann et al (2008) for a detailed list of these episodes. — 32 — Instituto de Estudios Fiscales adjustments17 in the majority of the countries. This is, for example, the case of Australia in 1980, of Canada, Finland and Sweden in the mid-nineties, and Portugal, the UK and the US in the eighties. However, we have identified slightly different dates for the structural breaks in nfa i,t . This result is not surprising, as this variable is a stock and a process of adjustment takes several years. Therefore, the bulk of the structural changes are found in the eighties and the nineties, but also around 2000. For example, whereas the Finish current account has a break in 1994, its net foreign assets have two breaks: one in 1996 and another one in 2001. In the US, the dates are 1982 for the current account and 1984 in nfa i,t Later, a second break appears in 1999 in the current account, coincident with a break in net foreign assets. The literature dates a reversal in 2000. 5.1.2. Testing I(0) stationarity on individual time series Once the break points have been dated, we proceed to analyze the order of integration of the yt time series. The estimation of the model in (17) with the break points that have been obtained above can be used to compute the individual stationarity test in Kwiatkowski et al. (1992) –henceforth, KPSS statistics– given by η̂i (λ i ) = ω̂i−2 T−2 T ∑ Ŝi,t2 , (18) t=1 where Ŝ i,t = ∑tj=1 ê i, j is the partial sum process that is obtained using the estimated OLS residuals of (17), ω̂i2 denotes a consistent estimate of the long­ run variance of the error term e i,t , which, based on the evidence reported in Carrion-i-Silvestre and Sansó (2006), has been estimated following the procedure described by Sul et al. (2005), using the Quadratic spectral kernel. In ′ i i / T,...,Tb,m / T ⎟⎞ , which (18), λ i is defined as the vector λ i = (λ i,1,...,λ i,mi )′ = ⎜⎛ Tb,1 ⎝ i, j ⎠ indicates the relative position of the dates of the breaks on the entire time period T for each individual. Thus, the computation of the individual KPSS statistic permits to get a first analysis of the stochastic properties current account and the net foreign asset position (see Panel A in Tables 1 and 4, respectively). The statistics in Panel A offer the computation of the individual KPSS along with the corresponding simulated critical values at the 5 and 10% level of significance. Focusing on the individual statistics of ca i, t , we can see that 17 A current account adjustment is defined by three conditions: (i) the current account should exceed 2% of GDP prior to the adjustment; (ii) the average deficit should decline by at least 2% of GDP over three years and be reduced by at least a third; (iii) the largest deficit during the five years after the peak should not be wider than the smallest deficit during the three years before the peak (Debelle and Galati, 2007). — 33 — the null hypothesis of I(0) cannot be rejected at the 5% level of significance for fifteen out of twenty countries – the exceptions are Ireland, Japan, Netherlands, Portugal, and Sweden. Therefore, the constraint is met for the majority of the countries in the panel data set, although the fact that for these countries ca i, t is found to be I(0) evolving around a broken deterministic component implies that the current account is not sustainable. The results are similar in the case of nfa it : the null hypothesis of stationarity is rejected at 5% in eight cases (Austria, Belgium, Italy, Japan, Spain, Sweden, UK and US). This individual based inference can be improved if we combine the individual statistics through the definition of panel data statistics. Thus, the literature on non-stationary panel data statistics argues that a better characterization of the stochastic properties of the time series can be obtained if we increase the amount of information when performing the inference. However, some cautions have to be taken when computing these panel-data-based statistics, since some of them rely on the critical assumption of cross-section independence. This assumption is investigated in the next section for our panel data sets. 5.1.3. The issue of cross-section independence The independence assumption imposed in the so-called first generation panel data statistics has been widely criticized in the recent literature, since it has been shown that non accounting for cross-section dependence amongst the individuals might bias the statistical inference in favor of variance stationarity – see Banerjee et al. (2004, 2005). Although it is now common practice to apply panel data unit root and stationarity tests that take into account cross-section dependence, few really test whether the individuals are cross-section dependent. In this subsection we test the null hypothesis of non correlation against the alternative hypothesis of correlation using the approach suggested in Pesaran (2004) and Ng (2006). Besides, this framework allows us to gain some insight on the kind of cross-section dependence in terms of how pervasive and strong is the cross-section correlation. We can allow for the presence of the structural breaks when testing the null hypothesis of non correlation among the individuals in the panel. We will then estimate an autoregressive model to isolate cross­ section dependence from the autocorrelation that might be driving the individual time series. In addition, the estimation of the autoregressive model includes dummy variables to capture the level shifts that have been detected using Bai and Perron (1998) in the previous section, which aims at isolating cross-section dependence from both autocorrelation and structural breaks in the individual time series. Pesaran (2004) designs a test statistic based on the average of pair-wise Pearson's correlation coefficients p̂ j , j = 1,2,K,n , n = N(N − 1) / 2 , of the residuals — 34 — Instituto de Estudios Fiscales obtained from an autoregressive (AR) model that includes dummy variables to capture the structural breaks. The CD statistic of Pesaran (2004) is given by CD = 2T n n ∑ p̂ j → N(0,1.) j =1 This statistic tests the null hypothesis of cross-section independence against the alternative of dependence. The procedure proposed by Ng (2006) works as follows. First, we get rid of the autocorrelation pattern in the individual time series through the estimation of an AR model. This allows us to isolate the cross-section regression from serial correlation. Taking the estimated residuals from the AR regression equations as individual series, we compute the absolute value of Pearson's correlation coefficients (p j = p̂ j ) for all possible pairs of individuals, j = 1,2,K,n , where n = N(N − 1) / 2 , and sort them in ascending order. As a result, we obtain the sequence of ordered statistics given by {p[1 : n], p[2 : n],K, p[n : n] }. Under the null hypothesis that p j = 0 and assuming that individual time series are Normally distributed, p j is half-normally distributed. Furthermore, let us define φ j as ( ), where Φ denotes the cdf of the standard Normal distribution, so that φ = (φ1,K, φn ) . Finally, let us define the spacings as ∆φ j = φ j − φ j−1 , j = 1,K,n . Φ Tp[j : n] Second, Ng (2006) proposes splitting the sample of (ordered) spacings at arbitrary ϑ ∈ (0,1) , so that we can define the group of small (S ) correlation coefficients and the group of large (L ) correlation coefficients. The definition of the partition is carried out by minimizing the sum of squared residuals Q n (ϑ) = [ϑn] n j=1 j=[ϑn]+1 2 2 ∑ (∆φ j − ∆ S (ϑ)) + ∑ (∆φ j − ∆L (ϑ)) , where ∆ S (ϑ) and ∆ L (ϑ) denotes the mean of the spacings for each group respectively. A consistent estimate of the break point is obtained as ϑ̂ = argmin ϑ∈(0,1) Q n (ϑ) , where some trimming is required. Following Ng (2006) the trimming is set at 0.10. Once the sample has been split, we can proceed to test the null hypothesis of non correlation in both sub-samples. Obviously, the rejection of the null hypothesis for the small correlations sample will imply also rejection for the large correlations sample as the statistics are sorted in ascending order. Therefore, the null hypothesis can be tested for the small, large and the whole sample using the Spacing Variance Ratio (SVR ) in Ng (2006), which under the null hypothesis converges to the standard normal distribution. The results in Tables 5 show that for ca i, t the null hypothesis of independence is rejected for the whole sample of spacings, while it is not — 35 — rejected for the L and S samples at the 5% level of significance. Since the proportion of non significant correlations in the L and S group is similar, this leads us to conclude that cross-section dependence is not pervasive. In this case, the factor models suggested by Bai and Ng (2004) might not be a suitable approximation to account for the cross-section dependence that appears in the panel data set. Besides, Pesaran's (2004) CD statistic strongly rejects the null hypothesis of independence. As for the variable nfa i,t , we can see that the SVR statistic does not reject the null hypothesis of cross-section independence at the 5% level of significance neither of the whole, small and large sample of spacings. On the contrary, Pesaran's (2004) CD statistic strongly rejects the null hypothesis of cross-section independence. Although the two statistics lead to contradictory results, we prefer to proceed in a conservative way and consider that the time series of nfa& i,t in the panel data set are cross-section dependent – further note that the p-value of the SVR test is 0.124, not far away from the 10% level of significance. In all, the evidence that is obtained in this section indicates that cross-section dependence has to be considered when computing the panel data statistics if misleading conclusions are to be avoided. 5.1.4. Panel data tests with cross-section dependence and structural breaks The specification estimated above permits the computation of two different panel data stationarity statistics. First, we have applied the approach suggested in Carrion-i-Silvestre et al. (2005) to test the null hypothesis of panel variance stationarity allowing for multiple level shifts. Thus, note that the specification given in (17) is one of the two models considered by these authors. The OLS estimated residuals from (17) are used to obtain the individual KPSS statistics computed in the previous sections, which in turn can be combined to define the panel stationarity test statistic: LM(λ ) = N −1 N ∑ η̂i (λ i ), i =1 with η̂i (λ i ) defined in (18). Note that η̂i (λ i ) has been defined such that the long­ run variance is heterogeneous across individuals. However, it would be possible to use a homogeneous estimate of the long run variance, i.e., ω̂ 2 = N −1 ∑Ni=1 ω̂i2 . Using these elements we can define the panel data statistic 2 −1 N 2 2 Z(λ ) = N (LM(λ ) − ξ ))ς , where ξ = N −1 ∑N i=1 ξ i and ς = N ∑i=1 ς i , with ξ i and ς i being the individual mean and variance of ηi (λ i ) respectively. Note that these two possibilities for the definition of the long-run variance estimate gives rise to two different statistics, i.e., the Z(λ ) when the long-run variance homogeneity is imposed and the Z(λ ) for heterogeneous long-run variance. — 36 — Instituto de Estudios Fiscales Under the null hypothesis of variance stationarity and assuming cross-section independence, the Z(λ ) panel data statistics are shown to converge to the standard normal distribution. However, this limiting result is not obtained when individuals are cross-section dependent, as it is in our case. In this situation, we can compute the bootstrap distribution of the Z(λ ) statistics to account for the presence of a general form of cross-section dependence. The computation of the bootstrap distribution follows the lines given in Maddala and Wu (1999). To be specific, we have defined the (T × N) -matrix of the OLS estimated residuals from (17) ê = (ê1,K,êN ) , and have resampled with replacement the rows of the ê matrix so that the first matrix of resampled residuals ê ∗(1) is obtained, where the superscript ∗ (1) indicates the first resampling. Conditional on the estimated parameters and structural breaks, we have computed the bootstrap variables ∗(1) y i,t = α̂ i + m̂i ∑ θ̂i,k DUi,k,t + ei,t∗(1), k=1 for each i , where α̂ i and θ̂i, k are the OLS estimates of the parameters in (17). ∗(1) This is repeated 20,000 times so that we define y i,t ,K, y ∗i,t(2,000 ) series for each individual, which can be used to approximate the empirical distribution of the Z(λ ) statistics. Table 1, panel B, presents the Z(λ ) statistics as well as the bootstrap critical values for ca it . According to these statistics, the null hypothesis of I(0) cannot be rejected for the current account imbalance, regardless of the assumption made about the long-run variance estimation. Concerning the net foreign assets, the results are presented in Table 4, where we reach the same conclusion: the variable is stationary. Although we have already obtained that the variables in the two panels are I(0) stationary processes, we have checked the robustness of our results computing panel data unit root and stationarity tests that control for the presence of cross-section dependence using approximate common factor models proposed in Bai and Ng (2004), Harris et al. (2005) and Bai and Carrion­ i-Silvestre (2009). The common factors approach decomposes the observable variables as follows y i, t = α i + mi ∑ θi, k DUi, k, t + Ft′ πi + ξi, t , (19) k =1 t = 1,K,T , i = 1, K,N , where Ft is a (r × 1) -vector that accounts for the common factors that are present in the panel, and ξi, t is the idiosyncratic disturbance term, which is assumed to be cross-section independent. Note that the specification given by (19) is similar to the one in (17), where the disturbance term e i,t in (17) has been expressed as e i, t = Ft′ π i + ξ i, t giving rise to the specification in (19). The unobserved common factors (Ft ) and idiosyncratic — 37 — disturbance terms (ξi, t ) are estimated using principal components on the first difference model. The estimation of the number of common factors is obtained using the panel BIC information criterion in Bai and Ng (2002), with a maximum of six common factors. Tables 1 and 4, respectively, report the results of applying this method. For both the current account and the net foreign assets the ADF statistic computed from the idiosyncratic disturbance terms rejects the null hypothesis of unit root, while the procedure detects at least one non­ stationary common factor (one in cai, t , and six in nfa i,t ) where r1 denotes the number of non-stationary common factors so that r = r0 + r1 , and r0 is the number of stationary common factors. This set-up allows us to compute two panel data test statistics that consider the presence of multiple structural breaks. First, we have the panel data stationarity test statistic in Harris et al. (2005), which is given by ˆ = T −1/ 2 ∑T ˆ + cˆ ) / ω ˆ {aˆ k,t }, being C S Fk = (C k k t =k +1 â k,t the autocovariance of order k , +r̂ â k,t = ∑N i=1 ẑ i,t ẑ i,t −k . We define ẑ i, t as the i th element of the (N + r̂ ) × 1 vector (F̂1, t ,K,F̂r̂, t ,ξ̂1, t ,K,ξ̂N, t )′ that contains the estimated common factors (F̂) and the idiosyncratic disturbance (ξˆ i ) , with ĉ = (T − k )−1/ 2 ∑Ni=1ĉ i , being ĉ i a correction term defined in Harris et al. (2005) and, ω̂2 {a t } is a consistent estimate of the long-run variance of {a t } . Under the null hypothesis of joint variance stationarity of the common and idiosyncratic components the statistic SFk → d N(0,1) . We follow Harris et al. (2005) and use k = (3T )1/ 2 . The value of S Fk = 2.546 statistic with p-value of 0.005 leads to the rejection of the null hypothesis of I(0), which contradicts the previous results that have been found using the Z (λ ) statistics. However, it should be borne in mind that the Ng's statistic has revealed that cross-section dependence is not pervasive, so that the use of the common factor model might not be correct. Second, we can compute the panel data unit root test in Bai and Carrion-iSilvestre (2009), which has been shown to be robust to the presence of multiple structural breaks affecting the level. These authors propose the computation of the panel data version of the modified Sargan-Bhargava (MSB) statistics using the estimated idiosyncratic disturbance term (ξ̂i ), with up to three different ways to pool the individual information. In this case and using the notation in Bai and Carrion-i-Silvestre (2009), we have the Z → d N(0,1) , Pm → d N(0,1) and P → d χ 22N panel data unit root test statistics. Panel B of 1 reports the results for the variable ca i, t , showing that the null hypothesis of I(1) can be rejected using the Pm and P statistics at the 5% level of significance, although it is not rejected when using the Z test. The finite sample analysis in Bai and Carrion-i-Silvestre (2009) shows that the Pm and P statistics are the ones with better finite sample [ ] — 38 — Instituto de Estudios Fiscales performance compared to the Z test, as the Z test suffers from mild size distortions problems (underrejection) while the Pm and P statistics have the correct size. Thus, we rely on the conclusions of the Pm and P statistics, that the null hypothesis of panel unit root for ca i, t is rejected. We have conducted the same analysis for the nfa variable, whose own results are reported in Panel B of Table 4. In this case we present the statistics for different number of common factors, given that the use of the panel BIC always estimates the number of common factors that equals the maximum that is allowed. First, the SFk statistic does not reject the null hypothesis of I(0) regardless of the number of common factors that is specified. Second, the conclusions drawn from the panel unit root tests depend on the number of common factors that are considered. In general, we can see that when we impose either one, three, four or five common factors, the null hypothesis of unit root is rejected at least at the 10% level of significance. The converse is obtained when the number of common factors is fixed at two or six. Taken together, the results that have been obtained indicate that nfa i,t is I(0) stationary. To sum up, our results show that there is evidence that both the current account and the net foreign assets variables can be characterized as I(0) stationary processes once structural breaks and cross-section dependence are allowed for. 5.2. Testing for the unified approach 5.2.1. Bai-Perron estimation results Following the discussion in the section devoted to the theoretical models, we now proceed to test for what we have called the “unified approach”. For this purpose we have estimated the following model: CA t NFA t = µj + βj + ut ; GDPt GDPt t = Tj−1 + 1,...,Tj j = 1,...,m + 1, where T0 = 0 and Tm+ q = T. Given that the variables that define this model have been characterized as I(0) stochastic processes, we can apply the Bai and Perron (1998) procedure to estimate the number and position of the structural breaks. The reason for the choice of this method is that, due to the presence of an important number of structural changes in the unit root analysis, we expect to find also breaks in the relationship linking ca t and nfa t . In fact, only some of the changes cancel-out and the majority remain. In Table 6 we present the results for all the countries in our sample. We show in the first four columns the dates where we have found the structural changes. In the following columns we offer the results for the mean and the slope parameters in the — 39 — different regimes. There are five cases where no structural change was found: France, Greece, New Zealand, Portugal and Spain. Moreover, the relationship is not significant in France and in New Zealand. As all of them had breaks individually, we assume that they cancel-out. A second group of countries has only one structural break: Australia (1976), Japan (1987) and the Netherlands (1993). Probably the changes in regime are related to exchange rate movements. In particular, the Dutch one is found in the 1993 EMS crisis. Two structural breaks appear in Finland (1988, 1994), Ireland (1982, 1996), Italy (1992, 1998), Sweden (1993, 2001) and the UK (1981, 1986). Whereas the EMU members changes are linked to the transition towards the new regime (Italy, Finland and Ireland), in the UK the oil prices are the main determinants of the changes in its external position. The rest of the countries have three structural breaks except for the US, which has four. As we have already described in the previous sections, the relation between the two measures of the external position of the countries should be positive. This is what we find: the structural changes are mainly capturing the activation of the adjustment mechanisms in the face of external imbalances. However, the parameters in the different regimes maintain the positive link between the two series. Take, for example, the case of Denmark in Table 6: all the β̂ are significant and positive. This is also true for the majority of the countries, no matter the number of breaks or its EMU membership. This is the first step of this part of the analysis, where we have estimated the reduced-form equation for all the countries in the sample. In the next subsection we will study the causality and the dynamics of the adjustment for two of the countries: the US and Spain. Note that the representation that we will use only concerns the dynamics or short-run analysis. We have restricted the last issue of the study to just two countries that are those (together with Portugal) in our sample with the larger current account deficits from 2000 onwards and where the current account adjustment seems to have been sluggish. 5.2.2. VAR and impulse-response results: the case of US and Spain We have now investigated the Granger causality between ca and nfa variables through the estimation of a VAR(p) model specification, allowing for structural breaks that affect both the constant and the coefficients of the lagged variables. Our analysis relies on the previous results where the two variables under investigation have been characterized as I(0) stationary processes. In this case we can apply the procedure in Qu and Perron (2007) to estimate a system of equations that allow for the presence of common structural breaks affecting the different equations. — 40 — Instituto de Estudios Fiscales Let us define the vector of variables y t = (ca t ,nfa t )′ . The model that has been estimated using Qu and Perron (2007) approach is given by: y t = µ j + β j,1y t −1 + β j,2 y t −2 + L + β j,p y t −p + u t ; t = Tj−1 + 1,K,Tj where β j,k denotes the (2 × 2) matrix of parameters that is associated to the k -th lag in the j -th regime, k = 1,K,p , j = 1,K,m . Given the limitation imposed by the small number of observations that we dispose of, we have restricted both the number of structural breaks and the maximum lag order for the VAR model up to two. Further, the covariance matrix of the disturbance terms is assumed not to change across regimes. The order of the VAR model is chosen using the Bayesian information criterion (BIC). Spain We should first note that in the Bai-Perron procedure applied to the reduced-form model (or structural representation) no breaks were found in the case of Spain. The dynamics differ: two structural changes (1985 and 1995) were found. This finding does not contradict the above-mentioned stability. It only calls for different types of adjustment in the three regimes. The first one, previous to EU accession; a second one between 1986 and the application of the Maastricht criteria; the third one after 1996 (roughly EMU membership). The estimates for the Spanish case using the whole period that goes from 1972 to 2006 are the following ones: ⎛ − 2.749 ⎞ ⎡0.592 ⎜⎜ ⎟⎟ + ⎢ ⎝ − 1.767 ⎠ ⎣0.702 ⎛ − 3.438 ⎞ ⎡ 0.661 ⎜⎜ ⎟⎟ + ⎢ ⎝ − 3.283 ⎠ ⎣0.402 ⎛ ca ⎞ ⎛ 0.562 ⎞ ⎡0.730 ⎜⎜ t ⎟⎟ = ⎜⎜ ⎟⎟ + ⎢ ⎝ nfa t ⎠ ⎝ − 3.081⎠ ⎣2.106 ⎛ ca t ⎞ ⎜⎜ ⎟⎟ = nfa t ⎠ ⎝ ⎛ ca t ⎞ ⎟⎟ = ⎜⎜ ⎝ nfa t ⎠ − 0.265 ⎤ ⎛ ca t −1 ⎞ ⎜ ⎟⎟ + û t 0.767 ⎦⎥ ⎝⎜ nfa t −1 ⎠ t = 1972,K,1985 − 0.188 ⎤ ⎛ ca t −1 ⎞ ⎟⎟ + uˆ t ⎜ 0.776 ⎦⎥ ⎝⎜ nfa t −1 ⎠ t = 1986,K,1995 0.075 ⎤ ⎛ ca t −1 ⎞ ⎜ ⎟ + uˆ t 0.824⎦⎥ ⎝⎜ nfa t −1 ⎠⎟ t = 1996,K,2006 These estimates satisfy the stationarity condition for the first and second regime, but not for the third regime. Although this might seem to contradict our previous results, we should bear in mind that the analysis is constrained to consider at most two structural breaks. Thus, the procedure that we have applied does not allow to include a third structural break in the model because of the small number of observations that we have, a situation that might be behind the non-stationarity of the VAR model in the third regime, specially, if we take into account that the euro was launched in 1999. In order to evaluate this suspicion, we have carried out the estimation of the VAR model using the observations up to 1999. The estimates of the model for this subperiod are: — 41 — ⎛ ca ⎞ ⎜⎜ t ⎟⎟ = ⎝ nfa t ⎠ ⎛ ca t ⎞ ⎜⎜ ⎟⎟ = ⎝ nfa t ⎠ ⎛ − 2.589 ⎞ ⎡0.569 − 0.229 ⎤ ⎛ ca t −1 ⎞ ⎟⎟ + û t t = 1972,K,1982 ⎜⎜ ⎟⎟ + ⎢ ⎥ ⎜⎜ ⎝ − n1.610 ⎠ ⎣0.797 0.769 ⎦ ⎝ nfa t −1 ⎠ ⎛ − 5.087 ⎞ ⎡ 0.911 − 0.371⎤ ⎛ ca t −1 ⎞ ⎜⎜ ⎟⎟ + ⎢ ⎥ ⎜⎜ ⎟⎟ + uˆ t t = 1983,K,1993 ⎝ − 3.739 ⎠ ⎣0.559 0.698 ⎦ ⎝ nfa t −1 ⎠ ⎛ ca ⎞ ⎜⎜ t ⎟⎟ = ⎝ nfa t ⎠ ⎛ − 3.037 ⎞ ⎡1.360 − 0.163 ⎤ ⎛ ca t −1 ⎞ ⎜⎜ − 26.844 ⎟⎟ + ⎢ 5.051 − 0.623 ⎥ ⎜⎜ nfa ⎟⎟ + uˆ t ⎝ ⎦⎝ ⎠ ⎣ t −1 ⎠ t = 1994,K,1999 Now the estimated VAR models for each subperiod satisfy the stationarity conditions. We can use these estimates to compute impulse response functions (IRF) to evaluate the effects of the shocks of each equation. Figures 11 to 13 present the IRF that are based on the Cholesky's orthogonal decomposition of the covariance matrix of the disturbance terms. As can be seen, the same qualitative picture is obtained for the first two regimes. Thus, we can see that the shocks that affect the current account, no matter from which equation they come, cause a decrease in the current account in the long-run. The opposite is observed for the effects on nfa t , for which the shocks cause an increase in the variable nfa t in the long-run. However, the main difference between the two regimes comes from the magnitude of these effects: while the magnitude of the effects on the current account has increased, whereas the one on nfa t has decreased. This situation changes for the third regime, where a shock (of magnitude equal to one standard deviation) coming from the disturbance that affects the CA equation produces a positive effect both on ca t and nfa t in the long-run –the total multiplier for ca t equals 6.915 and for nfa t is 22.274. On the other hand, when the shock comes from the disturbance that affects the nfa t equation it causes a negative effect on both ca t and nfa t variables in the long-run– the total multipliers are -0.925 and -2.050, respectively. US The estimated model for the US using the statistical information for the whole period gives the following results: ⎛ ca ⎞ ⎜⎜ t ⎟⎟ = ⎝ nfa t ⎠ ⎛ ca ⎞ ⎜⎜ t ⎟⎟ = ⎝ nfa t ⎠ ⎛ 1.348 ⎞ ⎡0.414 − 0.224 ⎤ ⎛ ca t −1 ⎞ ⎜⎜ ⎟⎟ + ⎢ ⎥ ⎜⎜ nfa ⎟⎟ + û t t = 1972,K,1982 3.130 0.053 0.545 ⎝ ⎠ ⎣ ⎦⎝ t −1 ⎠ ⎛ − 1.256 ⎞ ⎡0.577 − 0.115 ⎤ ⎛ ca t −1 ⎞ ⎟⎟ + û t t = 1983,K,1995 ⎜⎜ ⎟⎟ + ⎢ ⎥ ⎜⎜ ⎝ − 2.024 ⎠ ⎣0.070 0.694 ⎦ ⎝ nfa t −1 ⎠ ⎛ ca ⎞ ⎜⎜ t ⎟⎟ = ⎝ nfa t ⎠ ⎛ − 0.596 ⎞ ⎡0.854 0.028 ⎤ ⎛ ca t −1 ⎞ ⎜⎜ ⎟⎟ + ⎢ ⎥ ⎜⎜ ⎟⎟ + û t ⎝ − 4.807 ⎠ ⎣0.927 0.512⎦ ⎝ nfa t −1 ⎠ — 42 — t = 1996,K,2006 Instituto de Estudios Fiscales It can be verified that these estimates satisfies the stationarity conditions for each regime, whose IRF are reported in Figures 14 to 16. The structural breaks, although just two (note that the structural model in the previous section had four breaks) capture the same type of discontinuities than above. The changes occur in 1982 and 1995. As can be seen, the same qualitative conclusions are obtained for the first and second regimes. Thus, the shocks that come from the disturbance of the ca t equation affect positively the ca t in the long-run, whereas the effect is negative for the nfa t . Besides, the shocks coming from the disturbance of the nfa t equation affects positively nfa t itself in the long-run, whereas the effect is negative on the current account. An interesting feature is that the effects of the shocks seem to favor the net foreign asset position of the US when going from the first to the second regime: (i) a shock coming from the disturbance of the ca t equation reduces the nfa t by -0.109 in the first regime and by -0.048 in the second regime, while (ii) a shock coming from the disturbance of the nfa t equation increases the nfa t by 2.839 in the first regime and by 4.154 in the second regime. As for the third regime, the results indicate that the effect of the shocks is always positive for all variables. 6. CONCLUSIONS In this paper we have empirically revisited the debate of the external accounts sustainability in the OECD countries for the period 1970-2006. We try to reconcile the main theoretical approaches and formulate hypotheses that can be tested in a panel framework. Current account imbalances have steadily increased in rich countries over the last 20 years and there appears a widely shared worry that these deficits are too large, and government intervention is required. Using the concept of sustainability as the ability to meet the long run intertemporal budget constraint, we test for stationarity in the current account and the stock of net foreign assets of the OECD countries. In addition, we call for a unified approach were we relate both the flow and the stock approaches in line with Gourinchas and Rey (2007). Concerning the empirical methodology, we argue that there are several reasons to believe that these variables may suffer from discontinuities. Previous evidence has revealed that there might be some events that affect the current account and the external debt in a permanent way. If this is the case, it is well known that non accounting for structural breaks biases both unit root and stationarity tests towards concluding in favor of non-stationarity in variance. Moreover, the independence assumption imposed in the so-called first — 43 — generation panel data statistics has been widely criticized in the recent literature, since it has been shown that non accounting for cross-section dependence amongst the individuals might bias the statistical inference in favor of I(0) stationarity. In this research we aim at filling the gap in the literature on external sustainability in several respects. First, we improve previous empirical work on the intertemporal model by testing for the stationarity of the current account and the net foreign assets stock by applying panel tests. Second, we allow for multiple structural breaks and cross-section dependence. Finally, we relate the identification of the structural changes with the literature on current account reversals, trying to assess how the countries regain solvency through adjustment processes. Concerning the classical flow approach, based on the current account, our results point at just two cases of strict sustainability: only two countries have not experienced structural changes during the analyzed period, namely, Italy and New Zealand, showing stationarity, and therefore, external sustainability. The rest of the countries have experienced up to four breaks in their current account for the period considered. These discontinuities correspond to major institutional changes or policy measures that have induced a series of breaks in the path followed by the variables. Focusing on the individual statistics, we can see that the null hypothesis of I(0) cannot be rejected at the 5% level of significance for fifteen out of twenty countries – the exceptions are Ireland, Japan, Netherlands, Portugal, and Sweden. In general, the individual country results point to the fact that policy measures or, otherwise, abrupt readjustments, are still needed to keep the sustainability of the current accounts. This evidence would be against a smooth self regulating capacity of the markets, and therefore, against laissez-faire, the so-called Lawson doctrine. However, the increasing financial integration process among the OECD countries may be relaxing the external constraint. In fact, the evidence obtained indicates that cross-section dependence has to be considered when computing the panel data statistics if misleading conclusions are to be avoided. Finally, our results show that there is evidence of the current account being an I(0) stationary process once structural breaks and cross-section dependence are allowed for. The stock approach has been tested using the same empirical methodology. Again the results show that, even though the whole panel turns out to be stationary once we account for structural changes and dependence, only one country, namely Germany, does not suffer from structural changes. Once the two variables are found to be stationary, we analyze whether there is a relationship linking them, their sign, stability and dynamics. We first assess the presence of structural changes in the relation using the Bai-Perron — 44 — Instituto de Estudios Fiscales procedure and estimate the reduced-form parameters for the subperiods defined by the structural changes. For the majority of the countries and time­ periods the parameters are positive, smaller than one and significant, as expected. Concerning the dynamics, we have applied a recent method proposed by Qu and Perron (2007), which consist of the estimation of a VAR model also allowing for structural changes. We restrict this analysis to the cases of Spain and the US, which are those where the external imbalances are more severe. The results point to stationary adjustment for the majority of the sample, the exception being Spain after 1996. This implies that solvency is recovered after major shocks that affect the countries' external accounts. However, the high degree of financial globalization may have increased the persistence of the disequilibria for some peripheral EMU members, as in the case of Spain. — 45 — Instituto de Estudios Fiscales APPENDIX Figure 1 TRADE OVER GDP 80 T RAEMU T RAUK T RAUS 70 T RAJ AP 60 50 40 30 20 10 0 1 960 19 63 19 66 196 9 1 972 19 7 5 197 8 1 981 1 984 19 87 199 0 1 9 93 19 9 6 1 99 9 2 0 02 200 5 Figure 2 FINANCIAL INTEGRATION Fina ncial integration= asse ts + liabilities o ver GDP 2500 900 FI PORT UGA L FI ITA L Y FI GRE E C E FI SPA I N FI IRE L A ND FIU S FIC A NA DA FIU K FIN E WZE FIA U ST RA LI A 800 2000 700 600 1500 500 400 1000 300 200 500 100 0 0 1970 1975 1980 1985 1990 1995 2000 2005 1970 1975 1980 1985 1990 1995 2000 2005 2000 2005 Figure 3 FINANCIAL INTEGRATION Financi al in tegration= assets + liab ilitie s over GDP 60 0 1000 FI SW E D EN FI NOR W AY FI DE N MA R K FI FIN L AN D FI AU S TR IA 50 0 FI FR AN C E FI GE R FI NE TH FI BE LG IU M FI JA PA N 800 40 0 600 30 0 400 20 0 200 10 0 0 0 1970 197 5 1980 19 85 1990 1 995 2000 2005 1 970 — 47 — 1 975 1980 1985 1990 1995 Figure 4 CURRENT ACCOUNT CA ove r GD P 5.0 4 CA POR CA IT CA GR E CA ES P CA IR E 2.5 CA US CA CAN CA UK CA NZ C A A US L 2 0.0 0 -2.5 -2 -5.0 -4 -7.5 -6 -10.0 -8 -12.5 -10 -15.0 -12 -17.5 -14 1970 1975 1980 1985 1990 1995 2000 2005 1970 1975 1980 1985 1990 1995 2000 2005 1980 1985 1990 1995 2000 2005 1980 1985 1990 1995 2000 2005 Figure 5 CURRENT ACCOUNT CA over GDP 20 10.0 CA SW E CA NOR CA DK CA FIN CA AU S 15 CAFR CAGE R CANE T CAB E L CAJ A P 7.5 5.0 10 2.5 5 0.0 0 -2.5 -5 -5.0 -10 -7.5 -10.0 -15 1970 1975 1980 1985 1990 1995 2000 2005 1970 1975 Figure 6 NET FOREIGN ASSETS NFA over GD P 50 25 NFAP OR NFAI T NFAGR E N FAE S P NFAI R E 25 NFA US NFA CA N NFA UK NFA NZ NFA AU S L 0 0 -25 -25 -50 -50 -75 -75 -100 -100 -125 1970 1975 1980 1985 1990 1995 2000 2005 1970 — 48 — 1975 Instituto de Estudios Fiscales Figure 7 NET FOREIGN ASSETS NFA over GDP 100 75 NFAS W E NFAN OR NFAD K NFAFI N NFAA U S 50 NFAFR NFAGE R NFAN E T NFAB E L NFAJ A P 50 0 25 -50 0 -100 -25 -150 -200 -50 1970 1975 1980 1985 1990 1995 2000 2005 1970 1975 1980 1985 1990 1995 2000 2005 Table 1 RESULTS FOR THE MODEL WITH MULTIPLE BREAKS AFFECTING THE MEAN Panel A: Individual information Tests mi i Tb,1 Australia Austria Belgium Canada Denmark Finland 0.122 0.037 0.031 0.057 0.073 0.067 1 2 4 1 1 1 1980 1975 1974 1998 1989 1994 France Germany Greece Ireland 0.041 0.034 0.096 1 3 1 3 1992 1984 1998 1984 Italy Japan Netherlands New Zealand 0.181 ∗∗ 0.035 0 2 1982 0.256 ∗∗ 0.109 1 1992 0.053 1979 1984 1999 0.221 ∗∗ i Tb, 2 1981 1984 0.160 ∗∗ 2 2 Spain 0.269 ∗ 1 1992 1990 2001 1991 1998 2001 0 Norway Portugal i Tb, 3 1999 1995 i Tb, 4 2000 Critical values 10% 5% 0.197 0.179 0.060 0.230 0.156 0.186 0.256 0.232 0.068 0.302 0.190 0.237 0.169 0.091 0.231 0.085 0.212 0.110 0.299 0.102 0.354 0.128 0.463 0.158 0.168 0.209 0.350 0.453 0.127 0.100 0.161 0.118 0.244 0.319 (Follow) — 49 — (Continuation) Panel A: Individual information Tests mi i Tb,1 i Tb, 2 Sweden 0.428 ∗∗ 2 1994 United Kingdom United States 0.086 0.042 1 2 1986 1982 i Tb, 3 i Tb, 4 Critical values 10% 5% 2001 0.180 0.233 1999 0.158 0.114 0.191 0.138 Panel B: Panel data based unit root and stationarity test statistics Bootstrap distribution Test 90% 95% Z(λ) (Homog) -0.521 4.620 5.853 Z(λ ) (Heterog) -1.904 4.882 5.617 Test p-value Num. of factors SFk 12.546 0.005 1 Z Pm -0.735 11.921 0.231 0.027 5 5 P 57.178 0.038 5 Table 2 MAIN EVENTS AND BREAKS FOUND IN THE DATA EMU countries Main events CA Surplus Beginning 70s (Bretton Woods ends) AUS (75) First oil shock BEL (74) Both Non-EMU countries CA Deficit CA surplus CA deficit Beginning 80s Second oil shock AUS (81) NOR(79) IRE(84) Mid-80s lower oil prices BEL(84), GER(84) Beginning-mid 90s BEL(92) German unification, K mov, EMS crises GER(90), FIN(94) FR(92) POR(95) DK(89), SWE(94) End 90s, beginning 2000 Asian Crisis, EMU AUST(80) POR(84) IRE(91) GRE(98) CAN(98), NOR(99) BEL(00), GER(01) IRE(98) SPA(99) — 50 — SWE(01), JAP(01) US(99) Instituto de Estudios Fiscales Table 3 MAIN EVENTS AND BREAKS FOUND IN THE DATA Main events Non-EMU countries EMU countries Negative NFA Positive NFA Negative NFA Positive NFA Beginning 70s (Bretton Woods ends) NZ (76 ↓ ), CAN(77 ↓ ) First oil shock Beginning 80s AUS(76) BEL(80), FR(82) Second oil shock Mid-80s GRE(80) SWE(80 ↓ ) NZ(81), DK(83) IRE(80), SP(80) UK(80 ↑ ) AUST(78) JAP(84), US(84) ITA(84 ↓ ) lower oil prices Beginning-mid 90s DK(77) IRE(88) BEL(90,96) AUST(85) ESP(92),AUS(96) JAP(90,95) NOR(90 ↑ ),DK(91), NZ(91) UK(89,95 ↓ ) SWE(92), US(94 ↓ ) POR(00),ITA(01) SWE(97 ↑ ) DK(97), CAN(98 ↑ ), NZ(99) GRE(01),SPA(01) JAP(00), NOR(01) AUST(01), US(99 ↓ ) German unification, NET(93 ↓ ),FR(95),IRE(96 ↑ ),FIN(96 ↓ ) K mov, EMS crises End 90s, beginning 2000 FIN(01 ↑ ) Asian Crisis, EMU Table 4 RESULTS FOR NFA VARIABLE WITH MULTIPLE BREAKS AFFECTING THE MEAN Panel A: Individual information Tests mi i Tb,1 i Tb, 2 i Tb, 3 2001 i Tb, 4 Critical values 10% 5% Australia 0.098 3 1978 1985 0.099 0.122 Austria 0.214 ∗∗ 2 1976 1996 Belgium 0.129 ∗∗ 3 1980 1990 Canada 0.105 2 1977 1998 Denmark 0.046 4 1977 1983 Finland 0.097 2 1996 2001 0.196 0.255 France 0.048 2 1982 1995 0.098 0.113 0.142 0.179 1996 0.075 0.085 0.149 0.188 1991 1998 0.056 0.063 (Follow) — 51 — (Continuation) Panel A: Individual information Tests mi i Tb,1 i Tb, 2 i Tb, 3 i Tb, 4 Critical values 10% 5% Germany 0.092 0 0.356 0.459 Greece 0.054 2 1980 2001 Ireland 0.042 3 1980 1988 Italy 0.239 ∗∗ 2 1984 2001 Japan 0.517 ∗∗ 4 1984 1990 Netherlands 0.076 1 1993 New Zealand 0.052 4 1976 1981 Norway 0.025 2 1990 2001 Portugal 0.033 1 2000 Spain 0.439 ∗∗ 3 1980 1992 2001 0.079 0.092 Sweden 0.201 ∗∗ 3 1980 1992 1997 0.079 0.092 United Kingdom 0.123 ∗∗ 3 1980 1989 1995 0.075 0.086 United States 0.123 ∗∗ 3 1984 1994 1999 0.084 0.099 0.151 0.193 1996 0.071 0.081 0.123 0.149 1995 2000 0.073 0.088 0.169 0.209 1991 1999 0.061 0.071 0.136 0.169 0.252 0.330 Panel B: Panel data based unit root and stationarity test statistics Bootstrap distribution Test 90% 95% Z(λ) (Homog) -1.349 13.442 15.204 Z(λ ) (Heterog) -5.898 17.010 17.824 Number of factors SFk p-value Z p-value Pm p-value P p-value 1 -0.824 0.795 -1.300 0.097 4.194 0.000 77.512 0.000 2 -0.489 0.668 -1.271 0.102 0.664 0.253 45.942 0.240 3 -0.446 0.672 -2.092 0.018 2.431 0.008 61.747 0.015 4 -0.279 0.610 -1.591 0.056 1.645 0.050 54.712 0.061 5 -0.761 0.223 -1.229 0.110 2.569 0.005 62.976 0.012 6 -0.584 0.280 -1.049 0.147 0.533 0.297 44.766 0.279 — 52 — Instituto de Estudios Fiscales Table 5 SPACING VARIANCE RATIO AND CD STATISTICS FOR THE CA AND NFA VARIABLES. DETERMINISTIC FUNCTION GIVEN BY A CONSTANT TERM WITH LEVEL SHIFTS Ng’s spacing test Whole sample ca nfa Pesaran’s CD test Small group Large group svr (η) p-val svr (η) p-val η̂ svr (η) p-val Test p-val 2.355 1.158 0.009 0.124 -0.674 -0.490 0.250 0.688 120 194 -0.859 -0.659 0.195 0.745 6.825 3.907 0.000 0.000 Figure 8 CURRENT ACCOUNT OVER GDP AND ESTIMATED BREAK POINTS (Follow) — 53 — (Continuation) (Follow) — 54 — Instituto de Estudios Fiscales (Continuation) — 55 — Figure 9 NET FOREIGN ASSETS OVER GDP AND ESTIMATED BREAK POINTS (Follow) — 56 — Instituto de Estudios Fiscales (Continuation) (Follow) — 57 — (Continuation) — 58 — 1994 1988 Finland Ireland Greece Germany 1982 1996 2001 1993 1986 1976 Denmark 1990 1993 1987 1981 Canada 1985 1999 1992 1984 Belgium France 2001 1993 1981 Austria Tˆb, 3 1976 Tˆb, 2 Australia Tˆ b,1 Tˆb, 4 β̂ 1 µˆ 2 β̂ 2 -5.107 -0.307 -2.559 0.034 (-2.919) (-2.507) (-4.183) (2.660) -0.722 0.274 0.461 0.044 (-3.550) (11.332) (1.279) (1.156) -1.990 -0.129 2.337 0.021 (-9.232) (-5.842) (7.096) (0.412) -3.500 -0.008 24.978 0.689 (-4.837) (-0.369) (11.006) (11.632) 11.282 0.677 -2.303 0.032 (6.711) (7.985) (-4.549) (2.447) -6.491 -0.255 -10.324 -0.180 (-1.348) (-0.964) (-3.467) (-2.449) 0.215 0.059 (0.655) (1.124) -3.050 0.588 4.044 -0.024 (-3.982) (4.923) (3.384) (-0.328) -1.894 0.072 (-3.834) (4.310) -0.373 0.174 5.831 0.117 (-0.886) (16.478) (14.181) (16.715) µˆ 1 β̂ 3 µˆ 4 β̂ 4 -0.558 0.040 (-2.970) (4.795) -0.835 -0.018 1.812 0.144 (-2.666) (-0.492) (3.606) (4.435) -1.740 0.030 -5.626 -0.306 (-3.362) (1.083) (-4.945) (-5.891) 4.841 0.019 1.205 0.068 (5.766) (0.430) (1.124) (2.780) 0.196 0.096 3.137 0.117 (0.082) (1.573) (22.869 (20.314) 12.410 0.283 3.685 0.105 (11.703) (10.786) (17.772) (9.680) 5.845 -0.011 (6.004) (-0.875) µˆ 3 Table 6 ESTIMATION OF THE CA AND NFA RELATIONSHIP µˆ 5 (Follow) β̂ 5 1993 Netherlands 1993 1981 1976 Sweden UK US Spain Portugal Norway 1982 1986 2001 1985 1987 Japan 1977 1998 1992 Italy New Zealand Tˆb, 2 Tˆb,1 (Continuation) 1988 1999 Tˆb, 3 µˆ 1 β̂ 1 µˆ 2 β̂ 2 µˆ 3 β̂ 3 µˆ 4 β̂ 4 µˆ 5 β̂ 5 1.558 0.192 -0.211 0.144 4.841 0.347 (-1.513) (6.400) (6.572) (4.031) (3.134) (5.195) -0.710 0.461 1.740 0.040 (-3.294) (11.543) (6.140) (3.608) -1.901 0.280 6.675 0.116 (-1.001) (2.463) (9.895) (3.111) -4.791 0.012 (-3.122) (0.520) 2.228 0.214 14.232 0.003 2.325 0.375 8.802 0.259 (1.570) (6.359) (5.771) (5.149) (5.086) (6.474) (7.524) (0.081) -1.188 0.092 (-0.844) (2.656) 0.065 0.119 (0.132) (5.482) 5.161 -0.126 -0.607 0.049 2.581 0.025 (-1.211) (1.553) (1.427) (0.438) (8.662) (-2.410) -1.963 0.425 3.510 -0.147 -2.215 -0.034 (-3.610) (4.038) (1.619) (-1.308) (-8.734) (-1.303) -0.973 0.234 -3.884 0.511 -2.740 0.211 -0.183 0.212 -12.677 -0.381 2001 (-2.725) (3.703) (-4.615) (4.324) (-30.940) (8.590) (-1.410) (15.374) (-11.193) (-6.456) Tˆb, 4 Instituto de Estudios Fiscales Figure 10 CA AND NFA VARIABLES (Follow) — 61 — (Continuation) (Follow) — 62 — Instituto de Estudios Fiscales (Continuation) Figure 11 SPAIN. ACCUMULATED IMPULSE RESPONSES FOR THE FIRST REGIME (1972-1982) — 63 — Figure 12 SPAIN. ACCUMULATED IMPULSE RESPONSES FOR THE SECOND REGIME (1983-1993) Figure 13 SPAIN. ACCUMULATED IMPULSE RESPONSES FOR THE THIRD REGIME (1994-1999) — 64 — Instituto de Estudios Fiscales Figure 14 US. ACCUMULATED IMPULSE RESPONSES FOR THE .RST REGIME (1972-1982) Figure 15 US. ACCUMULATED IMPULSE RESPONSES FOR THE SECOND REGIME (1983-1995) — 65 — Figure 16 US. ACCUMULATED IMPULSE RESPONSES FOR THE THIRD REGIME (1996-2006) — 66 — REFERENCES AHMED, S. (1986): Temporary and permanent government spending in an open economy, Journal of Monetary Economics, n.º 17, pp. 197-224. BAI, J. and CARRION-I-SILVESTRE, J.L. (2009): Structural changes, common stochastic trends, and unit roots in panel data, Review of Economic Studies, forthcoming. BAI, J. and NG, S. (2002): Determining the Number of Factors in Approximate Factor Models, Econometrica, n.º 70, pp. 191-221. – (2004): A PANIC Attack on Unit Roots and Cointegration, Econometrica, 72, 4, pp. 1127-1177. BAI, J. and PERRON, P. (1998): Estimating and Testing Linear Models with Multiple Structural Changes, Econometrica, n.º 66, pp. 47-78. BALTAGI, B.H. (2005): Econometric Analysis of Panel Data. Third edition. John Wiley & Sons. BANERJEE, A. (1999): Panel Data Unit Roots and Cointegration: An Overview. Oxford Bulletin of Economics and Statistics, Special issue, pp. 607-629. BANERJEE, A.; MARCELLINO, M. and OSBAT, C. (2004): Some cautions on the use of panel methods for integrated series of macro-economic data, Econometrics Journal, n.º 7, pp. 322-340. – (2005): Testing for PPP: Should We Use Panel Methods?, Empirical Economics, n.º 30, pp. 77-91. BLANCHARD, O. (2007): “Current account deficits in rich countries”, IMF Staff Papers, vol. 54 (2), pp. 191-219. BLANCHARD, O. and GIAVAZZI, F. (2002): Current account deficits in the euro area: The end of the Feldstein-Horioka puzzle. Brookings Papers on Economic Activity 2, pp. 147-209. BEMS, R. and SCHELLEKENS, P. (2007): Finance and convergence: What's ahead for Emerging Europe? IMF Working Paper WP/07/244. BERGIN, P.R. and SHEFFRIN, S.M. (2000): Interest rates, exchange rates, present value models of the current account, Economic Journal, n.º 110, pp. 535-558. BREITUNG, J. and MEYER, W. (1994): Testing for Unit Roots in Panel Data: Are Wages on Different Bargaining Levels Cointegrated?. Applied Economics, n.º 26, pp. 353-361. BREITUNG, J. and PESARAN, M.H. (2007): Unit Roots and Cointegration in Panels, in Matyas, L. and Sevestre, P., The Econometrics of Panel Data (Third Edition), Kluwer Academic Publishers. — 67 — BUSSIÈRE, M.; FRATZSCHER, M. and MULLER, G.J. (2004): Current account dynamics in OECD and EU acceding countries - An intertemporal approach ECB working paper, n.º 2281. CABALLERO, R.J.; FARHI, E. and GOURINCHAS, P.O. (2008): “An Equilibrium Model of Global Imbalances and Low Interest Rates”, American Economic Review, vol. 98(1), pp. 358-393. CALDERÓN, C.; LOAYZA, N. and SERVEN, L. (2000): External sustainability. A stock equilibrium perspective, World Bank Policy Research Working Paper, n.º 2281. CAMPBELL, J.Y. (1987): Does saving anticipate declining labor income? An alternative test of the permanent income hypothesis, Econometrica, n.º 55, pp. 1249-1273. CAMPBELL, J. . and SHILLER, R. (1987): “Cointegration and tests of present value models”, Journal of Political Economy, n.º 93, pp. 1062-1088. CARRION-I-SILVESTRE, J.L. (2003): Breaking date misspecification error for the level shift KPSS test, Economics Letters, n.º 81, pp. 365-371. CARRION-I-SILVESTRE, J.L.; DEL BARRIO-CASTRO, T. and LÓPEZ-BAZO, E. (2001): Level Shifts in a Panel Data Based Unit Root Test. An Application to the Rate of Unemployment, Proceedings of the 2001 North American Econometric Society. – (2005): Breaking the panels: An application to the GDP per capita, Econometrics Journal, n.º 8, pp. 159-175. CARRION-I-SILVESTRE, J.L., and SANSÓ, A. (2006): A Guide to the Computation of Stationarity Tests, Empirical Economics, 31, pp. 433-448. CLARIDA, R.H. (2007): G7 Current Account Imbalances: Sustainability and Adjustment, Chicago, University of Chicago Press. CHINN, M. and PRASAD, E. (2003): Medium-term determinants of current accounts in industrial and developing countries: An empirical exploration. Journal of International Economics, n.º 59, pp. 47-76. CHORTAREAS, G.E.; KAPETANIOS, G. and UCTUM, M. (2004): An investigation of current account solvency in Latin America using non-linear nonstationary tests, Studies in Nonlinear Dynamics & Econometrics, Vol. 8, issue 1. pp. 1-17. CORSETTI, G. and ROUBINI, N. (1991): Fiscal deficits, public debt and government solvency: evidence form OECD countries, Journal of the Japanese and International Economies, n.º 5, pp. 354-80. DE HANN, L.; SCHOKKER, H. and TCHERNEVA, A. (2008): “What Do Current Account Reversals in OECD Countries Tell Us About the US Case?”, The World Economy, vol 31(2), pp. 286-311. DEBELLE, G. and FARUQEE, H. (1996): What determines the current account? A cross-sectional and panel approach. IMF Working Paper WP/96/58. DEBELLE, G. and GALATI, G. (2007): “Current account adjustment and capital flows”, Review of International Economics, n.º 15(5), pp. 989-1013. — 68 — DOOLEY, M.P.; FOLKERTS-LANDAU, D. and GARBER, P. (2007): Direct investment, rising real wages and the absortion of excess labor in the periphery, in Clarida (2006), op. cit. EDWARDS, S. (2001): Does the current account matter? NBER Working Paper, n.º 8275. FAGAN, G. and GASPAR, V. (2007): Adjusting to the euro. ECB Working Papers, n.º 716. FREUND, C. (2005): Freund, Caroline, 2005. "Current account adjustment in industrial countries," Journal of International Money and Finance, Elsevier, vol. 24(8), pp. 1278-1298. FREUND, C. and WARNOCK, F. (2007): Current account deficits in industrial countries: the bigger they are, the harder thay fall?, in Clarida (2006), op. cit. GHOSH, A. (1995): International capital mobility amongst the major industrialized countries: too much or too little?, Economic Journal, n.º 105, pp. 107-128. GRUBER, J. (2004): A present value test of habits and the current account, Journal of Monetary Economics, n.º 51, pp. 1495-1507. GOURINCHAS, P.O. and REY, H. (2007): International Financial Adjustment, Journal of Political Economy, n.º 115, 4. HARRIS, D.; LEYBOURNE, S. and MCCABE, B. (2005): Panel stationarity tests for purchasing power parity with cross-sectional dependence, Journal of Business and Economics Statistics, n.º 23, pp. 395-409. HOLMES, M.J. (2006): How sustainable are OECD current account balances in the long-run?, Manchester School, n.º 74, pp. 626-43. HOLMES, M.J.; OTERO and PANAGIOTIDIS, T (2007): “On the sustainability of the EU’s current account deficits”, Department of Economics, WP 2007-06, Loughborough University. HUSTED, S. (1992): The emerging US current account deficit in the 1980's: a cointegration analysis, The Review of Economics and Statistics, n.º 74, pp. 159-166. IM, K.; PESARAN, M.H. and SHIN, Y. (2003): Testing for Unit Roots in Heterogeneous Panels, Journal of Econometrics, n.º 115, pp. 53-74. IMF (International Monetary Fund) (2005): Globalization and external imbalances, World Economic Outlook, Chapter III, Washington. KAO, C. (1999): Spurious Regression and Residual-based Tests for Cointegration in Panel Data, Journal of Econometrics, n.º 90, pp. 1-44. KRAAY, A.; LOAYZA, N.; SERVÉN, L. and VENTURA, J. (2005): “Country Portfolios,” Journal of the European Economic Association, MIT Press, vol. 3(4), pp. 914-945. KRAAY, A. and VENTURA, J. (2003): “Current accounts in the long and the short run” en M. Gertler and K. Rogoff (eds.), NBER Macroeconomics Annual 2002, The Mit Press, Cambridge, pp. 65-94. — 69 — KOUPARITSAS, M. (2005): Is the US current account sustainable?, Chicago Fed Letter, n.º 215, june. KUROZUMI, E. (2002): Testing for Stationarity with a Break, Journal of Econometrics, n.º 108, pp. 63-99. KWIATKOWSKI, D.; PHILLIPS, P.C.B.; SCHMIDT, P.J. and SHIN, Y. (1992): Testing the null hypothesis of stationarity against the alternative of a unit root: How sure are we that economic time series have a unit root, Journal of Econometrics, n.º 54, pp. 159-78. LANE, P.R. and MILESI-FERRETTI, G.M. (2001):The external wealth of nations: Measures of foreign assets and liabilities for industrial and developing countries, Journal of International Economics, n.º 55, pp. 263-294, – (2002): External wealth, the trade balance, and the real exchange rate, European Economic Review, n.º 42, pp. 1049-1071 – (2004): “The Transfer Problem Revisited: Real Exchange Rates and Net Foreign Assets" , Review of Economics and Statistics, n.º 86, pp. 841-857. – (2006): “A global perspective on external positions”, in R. Clarida (ed.), G7 Current Account imbalances: sustainability and adjustment, University of Chicago Press, Chicago. – (2007a): The external wealth of nations mark II: Revised and extended estimates of foreign assets and liabilities, 1970-2004, Journal of International Economics, 73, pp. 223-250. – (2007b):”Europe and global imbalances”, Economic Policy, July, 519-573. – (2008): “"International Investment Patterns", Review of Economics and Statistics, n.º 90(3), pp. 538-549. LEE, J.; HUANG, C. J. and SHIN, Y. (1997): On Stationary Tests in the Presence of Structural Breaks, Economics Letters, n.º 55, pp. 165-172. LIU, P. and TANNER, E. (1996): International intertemporal solvency in industrialized countries: evidence and implications, Southern Economic Journal, n.º 62, pp. 739-749. MANN, CL (2002): Perspectives on the US current account deficit and sustainability, Journal of Economic Perspectives, n.º 16, pp. 131-152. MILESI-FERRETTI, G.M. and RAZIN, A. (1996): Sustainability of persistent current account deficits, NBER working paper 5467. NASON, J. and ROGERS. J.H. (2006): The present value model of the current account has been rejected: round up the usual suspects, Journal of International Economics, n.º 68, pp. 159-187. NG, S. (2006): Testing cross-section correlation in panel data using spacings, Journal of Business and Economics Statistics, n.º 24, pp. 12-23. — 70 — OBSTFELD, M. (2004): “External adjustment” Review of World Economics, Vol. 140 (4), pp. 541-568. OBSTFELD, M. and ROGOFF, K. (1995): The intertemporal approach to the current account, in Handbook of International Economics, 3, Grossman, G., and K. Rogoff (eds.), North-Holland. Amsterdam. – (1996): Foundations of International Macroeconomics, MIT Press, Cambridge, MA. – (2006): The unsustainable U.S. current account position revisited, in Clarida (2006), op. cit. O'CONNELL, P.G.J. (1998): The Overvaluation of the Purchasing Power Parity, Journal of International Economics, n.º 44, pp. 1-19. OTTO, G. (1992): Testing a present-value model of the current account: evidence US and Canadian time series, Journal of International Money and Finance, n.º 11, pp. 414-430. PERRON, P. (1989): The Great Crash, the Oil Price Shock and the Unit Root Hypothesis, Econometrica, n.º 57, 6, pp. 1361-1401. PESARAN, M.H. (2004): General diagnostic tests for cross section dependence in panels, Cambridge Working Papers in Economics, n.º 435, University of Cambridge. – (2007): A simple panel unit root test in the presence of cross section dependence, Journal of Applied Econometrics, n.º 22, pp. 265-312. PHILLIPS, P.C.B. (1987): Time series regression with a unit root, Econometrica, vol. 55, n.º 2, pp. 277-301. PHILLIPS, P.C.B. and PERRON, P. (1988): Testing for a unit root in time series regression, Biometrika, n.º 75, pp. 335-346. QU, Z. and PERRON, P. (2007): Estimating and testing structural changes in multivariate regressions, Econometrica, n.º 75, pp. 459-502. SACHS, J. (1981): The current account and macroeconomic adjustment in the 1970's, Brookings Papers on Economic Activity, n.º 12, pp. 201-268. SHEFFRIN, S.M. and WOO, W.T. (1990): Present value tests of an intertemporal model of the current account, Journal of International Economics, n.º 29, pp. 237-253. SUL, D.; PHILLIPS, P.C.B. and CHOI, C.Y. (2005): Prewhitening Bias in HAC Estimation, Oxford Bulletin of Economics and Statistics, n.º 67, pp. 517-546. TAYLOR, A.M. (2002): A century of current account dynamics, Journal of International Money and Finance, n.º 21, pp. 725-748. TAYLOR, M.P. and SARNO, L. (1998): The behavior of real exchange rates during the post-Bretton Woods period, Journal of International Economics, n.º 46, pp. 281-312. — 71 — TILLE, C. (2003): The impact of exchange rate movements on U.S. foreign debt, New York Fed Current Issues in Economics and Finance, n.º 9, pp. 1-7, TREHAN, B. and WALSH, C. (1991): Testing intertemporal budget constraints: theory and applications to US federal budget deficits and current account deficits, Journal of Money, Credit and Banking, n.º 26, pp. 206-223. WICKENS and UCTUM (1993): Wickens, M. R. & Uctum, Merih, 1993. "The sustainability of current account deficits : A test of the US intertemporal budget constraint," Journal of Economic Dynamics and Control, Elsevier, vol. 17(3), pp. 423-441, may WTO (2008): World Trade Report 2008. Geneva. WU, J.L. (2000): Mean reversion of the current account: evidence from the panel data unit root test, Economics Letters, n.º 17, pp. 423-441. WU, J.L.; CHEN, S.L. and LEE, H.Y. (2001): Are current account deficits sustainable? Evidence from panel cointegration, Economics Letters, n.º 72, pp. 219-224. ZANGHIERI,P. (2004): "Current Accounts Dynamics in new EU members: Sustainability and Policy Issues," Working Papers 2004-07, CEPII research center. ZIVOT, E. and ANDREWS, D.W.K. (1992): Further Evidence on the Great Crash, the Oil Price Shock, and the Unit-Root Hypothesis, Journal of Business & Economic Statistics, n.º 10, 3, pp. 251-270. — 72 — SÍNTESIS PRINCIPALES IMPLICACIONES DE POLÍTICA ECONÓMICA En este trabajo se ha realizado una revisión empírica del debate sobre la solvencia y sostenibilidad externa de los países de la OCDE durante el período 1970-2006. Los desequilibrios por cuenta corriente se han incrementado de forma constante durante los últimos veinte años, existiendo una preocupación creciente sobre el tamaño que han alcanzado y la necesidad de tomar medidas de ajuste más allá de los mecanismos de mercado. Con este fin se contrasta la hipótesis de solvencia, entendida como la capacidad de un país para cumplir su restricción presupuestaria intertemporal, tanto desde el punto de vista del saldo acumulado de deuda exterior (enfoque stock) como por los flujos de la cuenta corriente en términos de valor presente (enfoque flujos). Asimismo, se analiza la relación entre las dos variables dentro del enfoque unificado de Gourinchas y Rey (2007). Desde un punto de vista metodológico, se contrasta la hipótesis de estacionariedad de las variables implicadas en un contexto de datos de panel no estacionarios, realizando un conjunto de mejoras respecto a los trabajos empíricos precedentes. En primer lugar, se amplia la información mediante el uso de datos de panel, lo que añade información cross-section a la contenida en la dimensión temporal de los datos. En segundo, se considera la existencia de cambios estructurales. Este punto tiene gran importancia, pues permite distinguir dentro del enfoque flujos entre los conceptos de solvencia y sostenibilidad. Esta última se conseguiría sólo cuando la estacionariedad se alcanza sin necesidad de ajustes bruscos por parte de los agentes privados en sus sendas temporales de consumo e inversión o bien, cuando no son necesarias medidas de ajuste por parte de las autoridades económicas. Por tanto, la existencia de cambios estructurales implicaría la no sostenibilidad de los desequilibrios exteriores, aún si existe solvencia a largo plazo. En tercer lugar, consideramos la existencia de relaciones de dependencia o correlación cruzada entre los países miembros del panel considerado. Esta situación es esperable en áreas integradas económicamente como es el caso de la OCDE. Su no consideración genera sesgos en los resultados que pueden haber conducido a error en investigaciones anteriores. Por último, se analiza la relación dinámica entre las dos variables de ajuste para los casos de Estados Unidos y de España, de especial relevancia por la evolución y magnitud de sus desequilibrios exteriores durante el período de análisis considerado. Los resultados obtenidos señalan que individualmente sólo tres países no han experimentado cambios estructurales durante el período analizado: Italia e Irlanda en el caso del saldo de balanza por cuenta corriente y Alemania para el caso del saldo acumulado de deuda exterior. En el resto de países se han identificado hasta un máximo de cuatro discontinuidades que se corresponden con cambios institucionales de importancia o con medidas de política económica (en su mayoría variaciones de — 73 — tipo de cambio). Estos resultados indican que las intervenciones institucionales o, alguna forma de ajuste brusco, se desencadena con el fin de poder mantener la sostenibilidad de las cuentas exteriores de los países de la OCDE estudiados. Esta evidencia empírica iría en contra de la capacidad autorreguladora del mercado para alcanzar un “ajuste suave” y, por tanto, también en contra de la llamada “doctrina Lawson” que aboga por un “laissez-faire”. Con todo, el proceso de integración financiera creciente entre los países de la OCDE puede estar relajando la restricción exterior de sus economías. De hecho, los resultados obtenidos indican la existencia de dependencia estadística en cada sección cruzada del panel analizado. Una vez que se consideran contrastes que engloban la posibilidad de cambios estructurales y dependencia “cross-section” los resultados indican estacionariedad conjunta del panel y, por tanto, la solvencia (si bien no la sostenibilidad) en el área en su conjunto. Este resultado se confirma al analizar la dinámica del ajuste para el caso de España y los Estados Unidos, países con severos desequilibrios exteriores pero que, una vez considerados los cambios estructurales, alcanzan un ajuste estacionario para el conjunto de periodos, excepto para España a partir de 1996. Este hecho estaría indicando que la solvencia exterior se recupera tras shocks de un tamaño importante sobre los saldos externos y que, sin embargo, el alto grado de globalización financiera puede haber aumentado la persistencia de los desequilibrios en los miembros periféricos de la Unión Monetaria Europea, como es el caso de España. — 74 — NORMAS DE PUBLICACIÓN DE PAPELES DE TRABAJO DEL INSTITUTO DE ESTUDIOS FISCALES Esta colección de Papeles de Trabajo tiene como objetivo ofrecer un vehículo de expresión a todas aquellas personas interasadas en los temas de Economía Pública. Las normas para la presentación y selección de originales son las siguientes: 1. Todos los originales que se presenten estarán sometidos a evaluación y podrán ser directamente aceptados para su publicación, aceptados sujetos a revisión, o rechazados. 2. Los trabajos deberán enviarse por duplicado a la Subdirección de Estudios Tributarios. Instituto de Estudios Fiscales. Avda. Cardenal Herrera Oria, 378. 28035 Madrid. 3. La extensión máxima de texto escrito, incluidos apéndices y referencias bibliográfícas será de 7000 palabras. 4. Los originales deberán presentarse mecanografiados a doble espacio. En la primera página deberá aparecer el título del trabajo, el nombre del autor(es) y la institución a la que pertenece, así como su dirección postal y electrónica. Además, en la primera página aparecerá también un abstract de no más de 125 palabras, los códigos JEL y las palabras clave. 5. Los epígrafes irán numerados secuencialmente siguiendo la numeración arábiga. Las notas al texto irán numeradas correlativamente y aparecerán al pie de la correspondiente página. Las fórmulas matemáticas se numerarán secuencialmente ajustadas al margen derecho de las mismas. La bibliografía aparecerá al final del trabajo, bajo la inscripción “Referencias” por orden alfabético de autores y, en cada una, ajustándose al siguiente orden: autor(es), año de publicación (distinguiendo a, b, c si hay varias correspondientes al mismo autor(es) y año), título del artículo o libro, título de la revista en cursiva, número de la revista y páginas. 6. En caso de que aparezcan tablas y gráficos, éstos podrán incorporarse directamente al texto o, alternativamente, presentarse todos juntos y debidamente numerados al final del trabajo, antes de la bibliografía. 7. En cualquier caso, se deberá adjuntar un disquete con el trabajo en formato word. Siempre que el documento presente tablas y/o gráficos, éstos deberán aparecer en ficheros independientes. Asimismo, en caso de que los gráficos procedan de tablas creadas en excel, estas deberán incorporarse en el disquete debidamente identificadas. Junto al original del Papel de Trabajo se entregará también un resumen de un máximo de dos folios que contenga las principales implicaciones de política económica que se deriven de la investigación realizada. — 75 — PUBLISHING GUIDELINES OF WORKING PAPERS AT THE INSTITUTE FOR FISCAL STUDIES This serie of Papeles de Trabajo (working papers) aims to provide those having an interest in Public Economics with a vehicle to publicize their ideas. The rules gover­ ning submission and selection of papers are the following: 1. The manuscripts submitted will all be assessed and may be directly accepted for publication, accepted with subjections for revision or rejected. 2. The papers shall be sent in duplicate to Subdirección General de Estudios Tributarios (The Deputy Direction of Tax Studies), Instituto de Estudios Fiscales (Institute for Fiscal Studies), Avenida del Cardenal Herrera Oria, nº 378, Madrid 28035. 3. The maximum length of the text including appendices and bibliography will be no more than 7000 words. 4. The originals should be double spaced. The first page of the manuscript should contain the following information: (1) the title; (2) the name and the institutional affi­ liation of the author(s); (3) an abstract of no more than 125 words; (4) JEL codes and keywords; (5) the postal and e-mail address of the corresponding author. 5. Sections will be numbered in sequence with arabic numerals. Footnotes will be numbered correlatively and will appear at the foot of the corresponding page. Mathematical formulae will be numbered on the right margin of the page in sequence. Bibliographical references will appear at the end of the paper under the heading “References” in alphabetical order of authors. Each reference will have to include in this order the following terms of references: author(s), publishing date (with an a, b or c in case there are several references to the same author(s) and year), title of the article or book, name of the journal in italics, number of the issue and pages. 6. If tables and graphs are necessary, they may be included directly in the text or alternatively presented altogether and duly numbered at the end of the paper, before the bibliography. 7. In any case, a floppy disk will be enclosed in Word format. Whenever the document provides tables and/or graphs, they must be contained in separate files. Furthermore, if graphs are drawn from tables within the Excell package, these must be included in the floppy disk and duly identified. Together with the original copy of the working paper a brief two-page summary highlighting the main policy implications derived from the research is also requested. — 77 — ÚLTIMOS PAPELES DE TRABAJO EDITADOS POR EL INSTITUTO DE ESTUDIOS FISCALES 2004 01/04 Una propuesta para la regulación de precios en el sector del agua: el caso español. Autores: M.a Ángeles García Valiñas y Manuel Antonio Muñiz Pérez. 02/04 Eficiencia en educación secundaria e inputs no controlables: sensibilidad de los resultados ante modelos alternativos. Autores: José Manuel Cordero Ferrera, Francisco Pedraja Chaparro y Javier Salinas Jiménez. 03/04 Los efectos de la política fiscal sobre el ahorro privado: evidencia para la OCDE. Autores: Montserrat Ferre Carracedo, Agustín García García y Julián Ramajo Hernández. 04/04 ¿Qué ha sucedido con la estabilidad del empleo en España? Un análisis desagregado con datos de la EPA: 1987-2003. Autores: José María Arranz y Carlos García-Serrano. 05/04 La seguridad del empleo en España: evidencia con datos de la EPA (1987-2003). Autores: José María Arranz y Carlos García-Serrano. 06/04 La ley de Wagner: un análisis sintético. Autor: Manuel Jaén García. 07/04 La vivienda y la reforma fiscal de 1998: un ejercicio de simulación. Autor: Miguel Ángel López García. 08/04 Modelo dual de IRPF y equidad: un nuevo enfoque teórico y su aplicación al caso español. Autor: Fidel Picos Sánchez. 09/04 Public expenditure dynamics in Spain: a simplified model of its determinants. Autores: Manuel Jaén García y Luis Palma Martos. 10/04 Simulación sobre los hogares españoles de la reforma del IRPF de 2003. Efectos sobre la oferta laboral, recaudación, distribución y bienestar. Autores: Juan Manuel Castañer Carrasco, Desiderio Romero Jordán y José Félix Sanz Sanz. 11/04 Financiación de las Haciendas regionales españolas y experiencia comparada. Autor: David Cantarero Prieto. 12/04 Multidimensional indices of housing deprivation with application to Spain. Autores: Luis Ayala y Carolina Navarro. 13/04 Multiple ocurrence of welfare recipiency: determinants and policy implications. Autores: Luis Ayala y Magdalena Rodríguez. 14/04 Imposición efectiva sobre las rentas laborales en la reforma del impuesto sobre la renta personal (IRPF) de 2003 en España. Autoras: María Pazos Morán y Teresa Pérez Barrasa. 15/04 Factores determinantes de la distribución personal de la renta: un estudio empírico a partir del PHOGUE. Autores: Marta Pascual y José María Sarabia. 16/04 Política familiar, imposición efectiva e incentivos al trabajo en la reforma de la imposición sobre la renta personal (IRPF) de 2003 en España. Autoras: María Pazos Morán y Teresa Pérez Barrasa. 17/04 Efectos del déficit público: evidencia empírica mediante un modelo de panel dinámico para los países de la Unión Europea. Autor: César Pérez López. — 79 — 18/04 Inequality, poverty and mobility: Choosing income or consumption as welfare indicators. Autores: Carlos Gradín, Olga Cantó y Coral del Río. 19/04 Tendencias internacionales en la financiación del gasto sanitario. Autora: Rosa María Urbanos Garrido. 20/04 El ejercicio de la capacidad normativa de las CCAA en los tributos cedidos: una primera evaluación a través de los tipos impositivos efectivos en el IRPF. Autores: José María Durán y Alejandro Esteller. 21/04 Explaining. budgetary indiscipline: evidence from spanish municipalities. Autores: Ignacio Lago-Peñas y Santiago Lago-Peñas. 22/04 Local governmets' asymmetric reactions to grants: looking for the reasons. Autor: Santiago Lago-Peñas. 23/04 Un pacto de estabilidad para el control del endeudamiento autonómico. Autor: Roberto Fernández Llera 24/04 Una medida de la calidad del producto de la atención primaria aplicable a los análisis DEA de eficiencia. Autora: Mariola Pinillos García. 25/04 Distribución de la renta, crecimiento y política fiscal. Autor: Miguel Ángel Galindo Martín. 26/04 Políticas de inspección óptimas y cumplimiento fiscal. Autores: Inés Macho Stadler y David Pérez Castrillo. 27/04 ¿Por qué ahorra la gente en planes de pensiones individuales? Autores: Félix Domínguez Barrero y Julio López-Laborda. 28/04 L a reforma del Impuesto sobre Actividades Económicas: una valoración con microdatos de la ciudad de Zaragoza. Autores: Julio López-Laborda, M.ª Carmen Trueba Cortés y Anabel Zárate Marco. 29/04 Is an inequality-neutral flat tax reform really neutral? Autores: Juan Prieto-Rodríguez, Juan Gabriel Rodríguez y Rafael Salas. 30/04 El equilibrio presupuestario: las restricciones sobre el déficit. Autora: Belén Fernández Castro. 2005 01/05 Efectividad de la política de cooperación en innovación: evidencia empírica española. Autores: Joost Heijs, Liliana Herrera, Mikel Buesa, Javier Sáiz Briones y Patricia Valadez. 02/05 A probabilistic nonparametric estimator. Autores: Juan Gabriel Rodríguez y Rafael Salas. 03/05 Efectos redistributivos del sistema de pensiones de la seguridad social y factores determinantes de la elección de la edad de jubilación. Un análisis por comunidades autónomas. Autores: Alfonso Utrilla de la Hoz y Yolanda Ubago Martínez. 14/05 La relación entre los niveles de precios y los niveles de renta y productividad en los países de la zona euro: implicaciones de la convergencia real sobre los diferenciales de inflación. Autora: Ana R. Martínez Cañete. 05/05 La Reforma de la Regulación en el contexto autonómico. Autor: Jaime Vallés Giménez. — 80 — 06/05 Desigualdad y bienestar en la distribución intraterritorial de la renta, 1973-2000. Autores: Luis Ayala Cañón, Antonio Jurado Málaga y Francisco Pedraja Chaparro. 07/05 Precios inmobiliarios, renta y tipos de interés en España. Autor: Miguel Ángel López García. 08/05 Un análisis con microdatos de la normativa de control del endeudamiento local. Autores: Jaime Vallés Giménez, Pedro Pascual Arzoz y Fermín Cabasés Hita. 09/05 Macroeconomics effects of an indirect taxation reform under imperfect competition. Autor: Ramón J. Torregrosa. 10/05 Análisis de incidencia del gasto público en educación superior: nuevas aproximaciones. Autora: María Gil Izquierdo. 11/05 Feminización de la pobreza: un análisis dinámico. Autora: María Martínez Izquierdo. 12/05 Efectos del impuesto sobre las ventas minoristas de determinados hidrocarburos en la economía extremeña: un análisis mediante modelos de equilibrio general aplicado. Autores: Francisco Javier de Miguel Vélez, Manuel Alejandro Cardenete Flores y Jesús Pérez Mayo. 13/05 La tarifa lineal de Pareto en el contexto de la reforma del IRPF. Autores: Luis José Imedio Olmedo, Encarnación Macarena Parrado Gallardo y María Dolores Sarrión Gavilán. 14/05 Modelling tax decentralisation and regional growth. Autores: Ramiro Gil-Serrate y Julio López-Laborda. 15/05 Interactions inequality-polarization: characterization results. Autores: Juan Prieto-Rodríguez, Juan Gabriel Rodríguez y Rafael Salas. 16/05 Políticas de competencia impositiva y crecimiento: el caso irlandés. Autores: Santiago Díaz de Sarralde, Carlos Garcimartín y Luis Rivas. 17/05 Optimal provision of public inputs in a second-best scenario. Autores: Diego Martínez López y A. Jesús Sánchez Fuentes. 18/05 Nuevas estimaciones del pleno empleo de las regiones españolas. Autores: Javier Capó Parrilla y Francisco Gómez García. 19/05 US deficit sustainability revisited: a multiple structural change approach. Autores: Óscar Bajo-Rubio. Carmen Díaz-Roldán y Vicente Esteve. 20/05 Aproximación a los pesos de calidad de vida de los “Años de Vida Ajustados por Calidad” mediante el estado de salud autopercibido. Autores: Anna García-Altés, Jaime Pinilla y Salvador Peiró. 21/05 Redistribución y progresividad en el Impuesto sobre Sucesiones y Donaciones: una aplicación al caso de Aragón. Autor: Miguel Ángel Barberán Lahuerta. 22/05 Estimación de los rendimientos y la depreciación del capital humano para las regiones del sur de España. Autora: Inés P. Murillo. 23/05 El doble dividendo de la imposición ambiental. Una puesta al día. Autor: Miguel Enrique Rodríguez Méndez. 24/05 Testing for long-run purchasing power parity in the post bretton woods era: evidence from old and new tests. Autor: Julián Ramajo Hernández y Montserrat Ferré Cariacedo. — 81 — 25/05 Análisis de los factores determinantes de las desigualdades internacionales en las emisiones de CO2 per cápita aplicando el enfoque distributivo: una metodología de descomposición por factores de Kaya. Autores: Juan Antonio Duro Moreno y Emilio Padilla Rosa. 26/05 Planificación fiscal con el impuesto dual sobre la renta. Autores: Félix Domínguez Barrero y Julio López Laborda. 27/05 El coste recaudatorio de las reducciones por aportaciones a planes de pensiones y las deducciones por inversión en vivienda en el IRPF 2002. Autores: Carmen Marcos García, Alfredo Moreno Sáez, Teresa Pérez Barrasa y César Pérez López. 28/05 La muestra de declarantes IEF-AEAT 2002 y la simulación de reformas fiscales: descripción y aplicación práctica. Autores: Alfredo Moreno, Fidel Picos, Santiago Díaz de Sarralde, María Antiqueira y Lucía Torrejón. 2006 01/06 Capital gains taxation and progressivity. Autor: Julio López Laborda. 02/06 Pigou’s dividend versus Ramsey’s dividend in the double dividend literature. Autores: Eduardo L. Giménez y Miguel Rodríguez. 03/06 Assessing tax reforms. Critical comments and proposal: the level and distance effects. Autores: Santiago Díaz de Sarralde Míguez y Jesús Ruiz-Huerta Carbonell. 04/06 Incidencia y tipos efectivos del impuesto sobre el patrimonio e impuesto sobre sucesiones y donaciones. Autora: Laura de Pablos Escobar. 05/06 Descentralización fiscal y crecimiento económico en las regiones españolas. Autores: Patricio Pérez González y David Cantarero Prieto. 16/06 Efectos de la corrupción sobre la productividad: un estudio empírico para los países de la OCDE. Autores: Javier Salinas Jiménez y M.ª del Mar Salinas Jiménez. 07/06 Simulación de las implicaciones del equilibrio presupuestario sobre la política de inversión de las comunidades autónomas. Autores: Jaime Vallés Giménez y Anabel Zárate Marco. 18/06 The composition of public spending and the nationalization of party sistems in western Europe. Autores: Ignacio Lago-Peñas y Santiago Lago.Peñas. 09/06 Factores explicativos de la actividad reguladora de las Comunidades Autónomas (1989-2001). Autores: Julio López Laborda y Jaime Vallés Giménez. 10/06 Disciplina credititicia de las Comunidades Autónomas. Autor: Roberto Fernández Llera. 11/06 Are the tax mix and the fiscal pressure converging in the European Union?. Autor: Francisco J. Delgado Rivero. 12/06 Redistribución, inequidad vertical y horizontal en el impuesto sobre la renta de las personas físicas (1982-1998). Autora: Irene Perrote. — 82 — 13/06 Análisis económico del rendimiento en la prueba de conocimientos y destrezas imprescindibles de la Comunidad de Madrid. Autores: David Trillo del Pozo, Marta Pérez Garrido y José Marcos Crespo. 14/06 Análisis de los procesos privatizadores de empresas públicas en el ámbito internacional. Motivaciones: moda política versus necesidad económica. Autores: Almudena Guarnido Rueda, Manuel Jaén García e Ignacio Amate Fortes. 15/06 Privatización y liberalización del sector telefónico español. Autores: Almudena Guarnido Rueda, Manuel Jaén García e Ignacio Amate Fortes. 16/06 Un análisis taxonómico de las políticas para PYME en Europa: objetivos, instrumentos y empresas beneficiarias. Autor: Antonio Fonfría Mesa. 17/06 Modelo de red de cooperación en los parques tecnológicos: un estudio comparado. Autora: Beatriz González Vázquez. 18/06 Explorando la demanda de carburantes de los hogares españoles: un análisis de sensibilidad. Autores: Santiago Álvarez García, Marta Jorge García-Inés y Desiderio Romero Jordán. 19/06 Cross-country income mobility comparisons under panel attrition: the relevance of weighting schemes. Autores: Luis Ayala, Carolina Navarro y Mercedes Sastre. 20/06 Financiación Autonómica: algunos escenarios de reforma de los espacios fiscales. Autores: Ana Herrero Alcalde, Santiago Díaz de Sarralde, Javier Loscos Fernández, María Antiqueira y José Manuel Tránchez. 21/06 Child nutrition and multiple equilibria in the human capital transition function. Autores: Berta Rivera, Luis Currais y Paolo Rungo. 22/06 Actitudes de los españoles hacia la hacienda pública. Autor: José Luis Sáez Lozano. 23/06 Progresividad y redistribución a través del IRPF español: un análisis de bienestar social para el periodo 1982-1998. Autores: Jorge Onrubia Fernández, María del Carmen Rodado Ruiz, Santiago Díaz de Sarralde y César Pérez López. 24/06 Análisis descriptivo del gasto sanitario español: evolución, desglose, comparativa internacional y relación con la renta. Autor: Manuel García Goñi. 25/06 El tratamiento de las fuentes de renta en el IRPF y su influencia en la desigualdad y la redistribución. Autores: Luis Ayala Cañón, Jorge Onrubia Fernández y María del Carmen Rodado Ruiz. 26/06 La reforma del IRPF de 2007: una evaluación de sus efectos. Autores: Santiago Díaz de Sarralde Míguez, Fidel Picos Sánchez, Alfredo Moreno Sáez, Lucía Torrejón Sanz y María Antiqueira Pérez. 27/06 Proyección del cuadro macroeconómico y de las cuentas de los sectores institucionales mediante un modelo de equilibrio. Autores: Ana María Abad, Ángel Cuevas y Enrique M. Quilis. 28/06 Análisis de la propuesta del tesoro Británico “Fiscal Stabilisation and EMU” y de sus implicaciones para la política económica en la Unión Europea. Autor: Juan E. Castañeda Fernández. — 83 — 29/06 Choosing to be different (or not): personal income taxes at the subnational level in Canada and Spain. Autores: Violeta Ruiz Almendral y François Vaillancourt. 30/06 A projection model of the contributory pension expenditure of the Spanish social security system: 2004-2050. Autores: Joan Gil, Miguel Ángel Lopez-García, Jorge Onrubia, Concepció Patxot y Guadalupe Souto. 2007 11/07 Efectos macroeconómicos de las políticas fiscales en la UE. Autores: Oriol Roca Sagalés y Alfredo M. Pereira. 02/07 Deficit sustainability and inflation in EMU: an analysis from the fiscal theory of the price level. Autores: Óscar Bajo-Rubio, Carmen Díaz-Roldán y Vicente Esteve. 03/07 Contraste empírico del modelo monetario de tipos de cambio: cointegración y ajuste no lineal. Autor: Julián Ramajo Hernández. 04/07 An empirical analysis of capital taxation: equity vs. tax compiance. Autores: José M.a Durán Cabré y Alejandro Esteller Moré. 05/07 Education and health in the OECD: a macroeconomic approach. Autoras: Cecilia Albert y María A. Davia. 06/07 Understanding the effect of education on health across European countries. Autoras: Cecilia Albert y María A. Davia. 07/07 Polarization, fractionalization and conflict. Autores: Joan Esteban y Debraj Ray. 08/07 Immigration in a segmented labor market: the effects on welfare. Autor: Javier Vázquez Grenno. 09/07 On the role of public debt in an OLG Model with endogenous labor supply. Autor: Miguel Ángel López García. 10/07 Assessing profitability in rice cultivation using the Policy Matrix Analysis and profit­ efficient data. Autores: Andrés J. Picazo-Tadeo, Ernest Reig y Vicent Estruch. 11/07 Equidad y redistribución en el Impuesto sobre Sucesiones y Donaciones: análisis de los efectos de las reformas autonómicas. Autores: Miguel Ángel Barberán Lahuerta y Marta Melguizo Garde. 12/07 Valoración y determinantes del stock de capital salud en la Comunidad Canaria y Cataluña. Autores: Juan Oliva y Néboa Zozaya. 13/07 La nivelación en el marco de la financiación de las Comunidades Autónomas. Autores: Ana Herrero Alcalde y Jorge Martínez-Vázquez. 14/07 El gasto en defensa en los países desarrollados: evolución y factores explicativos. Autor: Antonio Fonfría Mesa. 15/07 Los costes del servicio de abastecimiento de agua. Un análisis necesario para la regulación de precios. Autores: Ramón Barberán Ortí, Alicia Costa Toda y Alfonso Alegre Val. 16/07 Precios, impuestos y compras transfronterizas de carburantes. Autores: Andrés Leal Marcos, Julio López Laborda y Fernando Rodrigo Sauco. — 84 — 17/07 Análisis de la distribución de las emisiones de CO2 a nivel internacional mediante la adaptación del concepto y las medidas de polarización. Autores: Juan Antonio Duro Moreno y Emilio Padilla Rosa. 18/07 Foreign direct investment and regional growth: an analysis of the Spanish case. Autores: Óscar Bajo Rubio, Carmen Díaz Mora y Carmen Díaz Roldán. 19/07 Convergence of fiscal pressure in the EU: a time series approach. Autores: Francisco J. Delgado y María José Presno. 20/07 Impuestos y protección medioambiental: preferencias y factores. Autores: María de los Ángeles García Valiñas y Benno Torgler. 21/07 Modelización paramétrica de la distribución personal de la renta en España. Una aproximación a partir de la distribución Beta generalizada de segunda especie. Autores: Mercedes Prieto Alaiz y Carmelo García Pérez. 22/07 Desigualdad y delincuencia: una aplicación para España. Autores: Rafael Muñoz de Bustillo, Fernando Martín Mayoral y Pablo de Pedraza. 23/07 Crecimiento económico, productividad y actividad normativa: el caso de las Comunidades Autónomas. Autor: Jaime Vallés Giménez. 24/07 Descentralización fiscal y tributación ambiental. El caso del agua en España. Autores: Anabel Zárate Marco, Jaime Vallés Giménez y Carmen Trueba Cortés. 25/07 Tributación ambiental en un contexto federal. Una aplicación empírica para los residuos industriales en España. Autores: Anabel Zárate Marco, Jaime Vallés Giménez y Carmen Trueba Cortés. 26/07 Permisos de maternidad, paternidad y parentales en Europa: algunos elementos para el análisis de la situación actual. Autoras: Carmen Castro García y María Pazos Morán. 27/07 ¿Quién soporta las cotizaciones sociales empresariales?. Una panorámica de la literatura empírica. Autor: Ángel Melguizo Esteso. 28/07 Una propuesta de financiación municipal. Autores: Manuel Esteban Cabrera y José Sánchez Maldonado. 29/07 Do R&D programs of different government levels overlap in the European Union. Autoras: Isabel Busom y Andrea Fernández-Ribas. 30/07 Proyecciones de tablas de mortalidad dinámicas de España y sus Comunidades Autónomas. Autores: Javier Alonso Meseguer y Simón Sosvilla Rivero. 2008 11/08 Estudio descriptivo del voto económico en España. Autores: José Luis Sáez Lozano y Antonio M. Jaime Castillo. 12/08 The determinants of tax morale in comparative perspective: evidence from a multilevel analysis. Autores: Ignacio Lago-Peñas y Santiago Lago-Peñas. 13/08 Fiscal decentralization and the quality of government: evidence from panel data. Autores: Andreas P. Kyriacou y Oriol Roca-Sagalés. 14/08 The effects of multinationals on host economies: A CGE approach. Autores: María C. Latorre, Oscar Bajo-Rubio y Antonio G. Gómez-Plana. — 85 — 15/08 Measuring the effect of spell recurrence on poverty dynamics. Autores: José María Arranz y Olga Cantó. 16/08 Aspectos distributivos de las diferencias salariales por razón de género en España: un análisis por subgrupos poblacionales. Autores: Carlos Gradín y Coral del Río. 17/08 Evaluating the regulator: winners and losers in the regulation of Spanish electricity distribution (1988-2002). Autores: Leticia Blázquez Gómez y Emili Grifell-Tatjé. 18/08 Interacción de la política monetaria y la política fiscal en la UEM: tipos de interés a corto plazo y déficit público. Autores: Jesús Manuel García Iglesias y Agustín García García. 19/08 A selection model of R&D intensity and market structure in Spanish forms. Autor: Joaquín Artés. 10/08 Outsourcing behaviour: the role of sunk costs and firm and industry characteristics. Autoras: Carmen Díaz Mora y Angela Triguero Cano. 11/08 How can the decommodified security ratio assess social protection systems?. Autor: Georges Menahem. 12/08 Pension policies and income security in retirement: a critical assessment of recent reforms in Portugal. Autora: Maria Clara Murteira. 13/08 Do unemployment benefit legislative changes affect job finding? Evidence from the Spanish 1992 UI reform act. Autores: José M. Arranz, Fernando Muñoz Bullón y Juan Muro. 14/08 Migraciones interregionales en España y su relación con algunas políticas públicas. Autora: María Martínez Torres. 15/08 Entradas y salidas de la pobreza en la Unión Europea: factores determinantes. Autores: Guillermina Martín Reyes, Elena Bárcena Martín, Antonio Fernández Morales y Antonio García Lizana. 16/08 Income mobility and economic inequality from a regional perspectiva. Autores: Juan Prieto Rodríguez, Juan Gabriel Rodríguez y Rafael Salas. 17/08 A note on the use of calendar regressors. Autor: Leandro Navarro Pablo. 18/08 Asimetrías y efectos desbordamiento en la transmisión de la política fiscal en la Unión Europea: evidencia a partir de un enfoque VAR estructural. Autor: Julián Ramajo. 19/08 Institutionalizing uncertainty: the choice of electoral formulas. Autores: Gonzalo Fernández de Córdoba y Alberto Penadés. 20/08 A field experiment to study sex and age discrimination in selection processes for staff recruitment in the Spanish labor market. Autores: Rocío Albert, Lorenzo Escot, y José A. Fernández-Cornejo. 21/08 Descentralización y tamaño del sector público regional en España. Autor: Patricio Pérez. 22/08 Multinationals and foreign direct investment: main theoretical strands and empirical effects. Autora: María C. Latorre. — 86 — 23/08 Una aproximación no lineal al análisis del impacto de las finanzas públicas en el crecimiento económico de los países de la UE-15, 1965-2007. Autor: Diego Romero Ávila. 24/08 Consolidación y reparto de la base imponible del Impuesto sobre Sociedades entre los Estados Miembros de la Unión Europea: consecuencias para España. Autores: Félix Domínguez Barrero y Julio López Laborda. 25/08 La suficiencia dinámica del modelo de financiación autonómica en España, 2002-2006. Autores: Catalina Barceló Maimó, María Marquès Caldentey y Joan Rosselló Villalonga. 26/08 Ayudas públicas en especie y en efectivo: justificaciones y aspectos metodológicos. Autores: Laura Piedra Muñoz y Manuel Jaén García. 27/08 Las ayudas públicas al alquiler de la vivienda. un análisis empírico para evaluar sus beneficios y costes. Autores: Laura Piedra Muñoz y Manuel Jaén García. 28/08 Decentralization and spatial distribution of regional ecomonic activity: does equalization matter?. Autores: Santiago Lago-Peñas y Diego Martínez-López. 29/08 Childcare costs and Spanish mothers’s labour force participation. Autora: Cristina Borra. 30/08 Pro-poor economic growth, inequality and fiscal policy: the case of Spanish regions. Autores: Luis Ayala y Antonio Jurado. 2009 01/09 Does the balance of payments constrain economic growth?. Some evidence for the new EU members. Autores: Oscar Bajo-Rubio y Carmen Díaz-Roldán. 02/09 Imputación a valor de mercado de los rendimientos de la vivienda en Propiedad del IRPF. Autores: Luis Ayala Cañón, Jorge Onrubia Fernández y María del Carmen Rodado Ruiz. 03/09 Income poverty and multidimensional deprivation: lessons from cross-regional analysis. Autores: Luis Ayala Cañón, Antonio Jurado y Jesús Perez-Mayo. 04/09 Reglas fiscales activas: el caso de España (1981-2007). Autor: Juan E. Castañeda Fernández. 05/09 Índices trimestrales de volumen encadenados, ajuste estacional y Bechmarking. Autores: Ana M.ª Abad, Ángel Cuevas y Enrique M. Quilis. 06/09 Fiscal decentralization and economic growth in OECD countries: matching spending wit revenue decentralization. Autores: Norman Gemmell, Richard Kneller e Ismael Sanz. 07/09 Una estimación del voto estratégico en las elecciones generales españolas, 2000-2008. Autores: Enrique García Viñuela y Joaquín Artés. 08/09 La tributación del transporte como instrumento frente al cambio climático. Autor: Miguel Buñuel González 09/09 The ins and outs of unemployment and the assimilation of recent immigrants in Spain. Autores: José I. Silva y Javier Vázquez. 10/09 Decomposing the determinants of health care expenditure: the case of Spain. Autores: David Cantarero Prieto y Santiago Lago-Peña. — 87 — 11/09 La clase beta de medidas de desigualdad. Autores: Luis José Imedio Olmedo, Elena Bárcena Martín y Encarnación M. Parrado Gallardo. 12/09 Right incentives to enhance efficiency in public expenditure. Autor: Tamón A. Takahashi Iturriaga. 13/09 Fiscal decentralization and public sector employment: a cross-country analysis. Autores: Jorge Martínez-Vázquez y Ming-Hung Yao. 14/09 Factores explicativos de los resultados de las Comunidades Autónomas Españolas en PISA 2006. Autores: José Manuel Cordero Ferrera, Eva Crespo Cebada y Daniel Santín González. 15/09 A proposal to empirically evaluate the sensitivity of the speed of convergence in the EU. Autoras: Sonia de Lucas Santos, Inmaculada Álvarez Ayuso y M.ª Jesús Delgado Rodríguez. 16/09 An assessment of the sustainability of current account imbalances in OECD countries. Autores: Mariam Camarero, Josep Lluís Carrion-i-Silvestre y Cecilio Tamarit. — 88 —