Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Asymmetric International Risk Sharing in the Euro Area Holger Zemanek* University of Leipzig Institute for Economic Policy Grimmaische Straße 12 D-04109 Leipzig, Germany [email protected] Tel: +49-341-9733566 Fax: +49-341-973569 Previous version: 06.10.2010 Abstract In this paper we argue that persistent intra-euro area capital flows to southern Europe have created an asymmetric international risk sharing in the euro area. Theoretical results suggest that asymmetric international risk sharing reduces overall consumption smoothing as countries participate differently in international risk sharing. However, net borrowing countries are in particular negatively affected. An empirical panel analysis for the euro area confirms these propositions. Thus, beside rigid labour markets and only marginal fiscal risk sharing in the euro area, economic adjustment via the capital market is limited, too. This is an alarming result for the euro area. JEL-Codes: E32, F32, F36 Keywords: capital flows, international risk sharing, euro area, asymmetries * I thank Jan Fidrmuc, Bent Sørensen, Geoffrey Wood, and participants of the 3rd Economics and Finance PhD conference at Brunel University for helpful comments as well as the Institute of Economic Affairs, London, for support. 1 1. Introduction Main internal economic adjustment mechanisms in a monetary union are flexible labour markets, fiscal risk sharing or international risk sharing via capital markets. All three mechanisms mitigate income effects of asymmetric shocks or adverse business cycles. In the euro area, labour markets are quite rigid (EC 2008, Zemanek 2010) and fiscal risk sharing almost does not exist as competence on fiscal policies has remained at national level (EC 1993). Moreover, national fiscal policies are constrained by high deficits and debt levels after the financial crisis. Thus, internal adjustment within the euro area seems to rely basically on the functioning of the international risk sharing mechanism (EC 2008). However, persistent intra-euro area capital flows during the last decade from northern European countries to southern European countries have spurred an accumulation of foreign assets in the capital providing countries, in particular in Germany, while Greece, Portugal and Spain accumulated foreign liabilities. This asymmetric foreign capital asset distribution have created an asymmetric international risk sharing, which limits international risk sharing in the euro area. This becomes a crucial issue in the light of the necessary real adjustment in the euro area (EC 2010) and for the stability of the euro area. The aim of this paper is to shed light on asymmetric international risk sharing in theory and empirically for the euro area. The international risk sharing approach as adjustment mechanism in a monetary union was first proposed by Mundell (1973). Mundell argued that a monetary union would accelerate capital market integration, thus increased cross-border capital asset holdings. This portfolio diversification would provide a risk sharing mechanism between countries as income and consumption effects of an asymmetric shock are compensated by varying capital income and capital asset valuation. However, this approach assumes implicitly a relatively symmetric distribution of cross-border capital assets, which will be abandoned in our analysis. Empirical literature on international risk sharing among countries is substantial, as the question was deeply analysed prior to EMU and afterwards. For instance, Atkeson 2 and Bayoumi (1993) as well as Melitz and Zumer (1999) test risk sharing within the US and among countries to draw implications for EMU. Later research scrutinizes international risk sharing more in detail and looks for determinants, such as financial globalisation, that may have influenced the risk sharing mechanism, (Sørensen et al. 2006, Kose et al. 2007, Stavrev 2007, Demyanyk et al. 2008). The paper is structured as follows. In Section 2, we show first evidence for asymmetric foreign capital distribution as a source of asymmetric international risk sharing. Section 3 analyses asymmetric international risk sharing in a basic model. We test our hypotheses on asymmetric international risk sharing in a panel econometric framework in section 4. Section 5 concludes and gives implications. 2. Asymmetric foreign capital asset distribution in the euro area While Germany has experienced rising trade surpluses against euro area countries since EMU up to the recent crisis, other countries like Greece, Spain, Italy and Portugal have perpetuated large current account deficits (EC 2010, Zemanek et al. 2010). These opposite current account balances within the euro area reflect private intra-euro area capital flows from surplus countries to deficit countries. Blanchard and Giavazzi (2002) called these intra-euro area capital flows the end of the Horioka-Feldstein puzzle (Horioka and Feldstein 1980). Instead of domestic savings being invested domestically as found by Horioka and Feldstein (1980), savings are invested abroad in countries with the largest marginal return on capital. Therefore, correlation of domestic saving and domestic investment is observed low for the euro area (Blanchard and Giavazzi 2002). Why did the Horioka-Feldstein puzzle disappear for the euro area? Although, the literature gives many explanations, the main driver might be the European integration process. Spain, Italy, Greece and Portugal especially have taken advantage of improved access to international financial markets following the creation of EMU. The expected rate of return increased in these countries and long as well as short term 3 interest rates have converged strongly towards low German rates since the mid 1990s (Fagan and Gaspar 2007, Mendoza et al. 2007). Moreover, the abolition of exchange rate risk for lenders and borrowers, improved macroeconomic conditions, and lower borrowing constraints as a result of financial deepening accelerated capital flows within the euro area (De Santis and Lührmann 2006). After almost a decade of one-way capital flows from saving countries (mainly in the north of Europe) to southern Europe, the results are not only the widely mentioned large current account deficits (EC 2009) but also an asymmetric distribution of foreign capital assets in the euro area. Capital exporting countries, have lent large amounts to southern European economies, building up their foreign assets and lifting their net international investment position. In contrast, Greece, Portugal, Spain, Italy, Ireland, and France have substantial debt held by foreign lenders but relatively few assets abroad. This becomes evident in Table 1 which shows the net international investment positions (IIP) in percent of GDP and how they changed for EMU 12 countries. Germany, the Netherlands and Finland are those countries, which have seen a substantial rise of their net IIP since 1998. Table 1: Net international investment position in percent of GDP for EMU 12. Austria Belgium Finland France Germany Greece Ireland Italy Luxembourg Netherlands Portugal Spain 1998 -19.6 38.7 -76.9 9.0 0.4 -26.9 -1.6 -3.8 -25.0 -21.6 2002 -21.1 41.3 -40.6 3.0 5.6 -58.8 -20.0 -5.9 79.1 -27.0 -62.5 -46.8 2008 -14.4 31.4 -4.1 -18.1 25.3 -69.7 -54.1 -20.2 75.0 10.5 -91.9 -75.7 Δ 1998-2008* 5.2 -7.3 72.7 -27.1 24.9 -42.8 -34.1 -18.6 -4.1 14.3 -67.0 -54.1 Source: DataStream, 2010. * Percentage points, difference for Ireland and Luxembourg 2002-2008. Although data on IIP also include investment in and from countries other than EMU 12 countries, they allow us a first sight into asymmetric foreign capital asset 4 distribution.1 Data on foreign bank claims provided by the Bank for International Settlement2 partly confirm asymmetric capital asset allocation. Figure 1 shows net outstanding claims of German banks in Spain in relation to Spanish GDP. Until 2004, net claims remained relatively stable. However, since then, German banks have accumulated massive claims in Spain, outrunning claims of Spanish Banks in Germany. The outbreak of the financial crisis in 2007 stopped the development, however, on a high level. This evidence for asymmetric foreign capital asset distribution between euro area countries raises the question of their impact on international risk sharing and, thus, on the internal adjustment capability of the euro area. Figure 1: Net outstanding claims of German banks in Spain. 20 18 Percent of Spanish GDP 16 14 12 10 8 6 4 2 0 Jun.1999 Dec.2000 Mar.2002 Jun.2003 Sep.2004 Dec.2005 Mar.2007 Jun.2008 Source: Bank for International Settlement, 2010. Note: Net figures are calculated by claims of German banks in Spain minus claims of Spanish banks in Germany. GDP Figures: IMF World Economic Outlook 2010. 1 Detailed figures on IIP against single countries are not available. BIS Quarterly Review, Table 9B: Consolidated foreign claims of reporting banks - immediate borrower basis, December 2009. 2 5 3. Asymmetric international risk sharing Capital markets are a third internal adjustment channel within a monetary union. They allow an international risk sharing between countries of a monetary union, to smooth consumption in the case of asymmetric shocks and adverse business cycles. This economic explanation was given by Mundell (1973) in his late research on optimum currency areas. Mundell argued that cross-country capital asset holdings3, and thus capital market integration, will increase as the devaluation risk of nominal exchanges rates disappears in a monetary union (McKinnon 2000). If each country is holding claims on output of all other members of a monetary union, asymmetric shocks or adverse business cycles are shared by varying capital income and capital valuation, which smoothes consumption over time in all countries of a monetary union. In general, the risk sharing mechanism by Mundell (1973) has two channels through which income is distributed. First, investment income is a direct channel. Dividends or profits surge in a boom, while they are supposed to shrink in a recession. In a boom (recession), a country has to pay more (less) dividends on its foreign liabilities, but receives less (more) investment income from its investments in the recession (boom) country. Second, capital valuations are an indirect income distribution channel. In a boom (recession), equities and real estates may rise (fall) in their value. However, it is unclear, to which extent these valuation gains or losses affect income. It depends on, whether valuation gains or losses are realized or valuation losses affect income via write-offs. The textbook case of international risk sharing by Mundell (1973) implicitly assumes a symmetric cross-country holding of capital assets. Each country holds capital assets of all other member countries relatively to their economic size. However, as we have shown in section 2.2, intra-euro area capital flows have led to a substantial asymmetry in foreign capital asset distribution during the last decade. International risk sharing becomes asymmetrical. 3 Capital assets are bonds, equities as well as direct credits by the banking system. 6 To compare the effects of symmetric vs. asymmetric international risk sharing, we set up a basic model of consumption smoothing for a two-country monetary union and assume fully adverse business cycles. Consumption in both countries (indicated by superscript i and j) depends on income (GDP) and net investment income. During a boom (denoted by h), GDP will increase by Yh − Y0 above the trend GDP (denoted by 0) and falls in a recession (denoted by l) by Y0 − Yl . However, the international risk sharing mechanism distributes some of GDP between both countries. First, each country has capital income from its foreign assets in the other country. That income depends on yield r and the size of its assets, measured as share a of foreign GDP, thus a i r Y j . Secondly, countries have to pay dividends on foreign liabilities, which would be a j rY i . ( Therefore, consumption ) Chi = Yhi − Y0i + a i rYl j − a j rYhi in country i ( can be expressed as ) in a boom or Cli = Y0i − Yl i + a i rYhj − a j rYl i in a recession. By assuming r equal over the business cycle, we only account for capital valuation effects. That simplifies the model without changing the interpretation. To analyse consumption smoothing effects, we calculate the consumption variation over the business cycle, relative to trend consumption ΔCh− l . Inserting consumption C0 equations yields for country i: [ ] ΔChi − l Yhi − Y0i ) + a i rYl j − a j rYhi ) − ((Y0i − Yl i ) + a i rYhj − a j rYl i ) ( ( = C0i Y0i + a i rY0 j − a j rY0i (1) If we assume for simplicity Yhi = Yhj , Yl i = Yl j , Y0i = Y0 j , Yhi − Yl i = 2 x and r equal in both countries, then the equation (1) reduces to: ( ( ΔChi −l 2 x − 2 xr a i + a j = i C0i Y0 + Y0i r a i − a j ) ) (2) In equation (2) the first part of the nominator ( 2 x ) is simply the amount that GDP varies during the business cycle. The second part − 2 xr (a i + a j ) shows the redistribution of GDP between both countries. The difference of both, the full 7 nominator, is the amount of undistributed GDP, thus the absolute consumption variation over the business cycle. Relative consumption variation, as stated in the full equation (2) depends positively on GDP variation x, negatively on GDP variation relative to GDP x/Y0, and negatively on the asset yield r. The shares of investment a i and a j determine relative consumption variation negatively. Rising investment shares increase the distributed GDP, thus increasing consumption smoothing. However, the effects of a i and a j are different. Both derivations of ΔChi − l with respect to a i and C0i a j are negative, but with: ⎛ ΔChi −l i ⎝ C0 δa i δ ⎜⎜ ⎞ ⎛ ΔC i ⎟⎟ δ ⎜⎜ hi −l ⎠ < ⎝ C0 δa j ⎞ ⎟⎟ ⎠ (3) This implies, that a variation of either a i or a j away from symmetric international risk sharing, thus any case of asymmetric risk sharing, affects consumption smoothing in both countries differently. If for example, either a i or a j falls below any symmetric case, consumption variation in country i increases compared to symmetric risk sharing, thus consumption smoothing is reduced. However, a reduction of a i , the country i’s own share of foreign assets, will increase consumption variation relatively more than if a j reduces. Moreover, any asymmetric international risk sharing makes both countries participate differently in international risk sharing. Consumption smoothing will be different in both counties. In the light of the net IIP figures of Table 1, a reduction of a i would be the southern European perspective and the reduction of a j the perspective of Germany. Thus, Germany is expected to benefit more from international risk sharing with more smoothed consumption than southern European countries with highly negative net IIP figures. That relationship is displayed in Figure 2, which shows the graphical solution of equation (2) by using the same values as used for Table 2: Y=100, r=0.1 and x=2. While, the dashed line shows relative consumptions variation for country i dependent on a i given a j =2 (net IIP of country j is 200 percent), the solid line shows relative 8 consumption variation for country i dependent on a j , given a i =2 (net IIP of country i is 200 percent). In the case of symmetric international risk sharing, as both countries have an equal share of assets held abroad of 200 percent of foreign GDP. Consumption varies by 2.4 percent in both countries over the business cycle, although GDP varies by 4 percent (point A). Consumption is equally smoothed in both countries. However, if country j (say Spain) has a share of foreign assets of only 50 percent, consumption in country i (say Germany) will slightly more fluctuate by 2.6 percent (point B) over the business cycle. In contrast, from Spain’s perspective a i reduces to 0.5 (50 percent), increasing consumption fluctuation for Spain to 3.6 percent (point C), which is close to GDP variation. Overall, asymmetric international risk sharing reduces consumptions smoothing in all countries of a monetary union compared to a symmetric case. This result is true for both types of countries, net foreign asset and net foreign liability countries, but in particular for the latter. Figure 2 – Graphical solution of asymmetric international risk sharing Relative consumption variation in country i 0.050 0.045 0.040 C 0.035 ai(aj=2) 0.030 B 0.025 0.020 0.0 A j i a (a =2) 0.5 1.0 9 1.5 2.0 a i, a j 4. Empirical analysis Taking our analysis in sections 2 and 3 as a starting point, we test the empirical significance of asymmetric international risk sharing and its effect on consumption smoothing in a panel econometric framework for the euro area. In this context, we test the following two hypotheses: 1. Participation in international risk sharing differs significantly between euro area countries, revealing an asymmetric risk sharing. 2. The asymmetric foreign capital asset distribution in the euro area is a source for asymmetric international risk sharing, and reduces overall consumption smoothing within the euro area. Model and data The empirical analysis follows closely Sørensen et al. (2006), Kose et al. (2007) and Stavrev (2007). International risk sharing is measured by the correlation of GDP growth and consumption growth relative to a country group, which is the euro area in this analysis. The idea is to analyse how a deviation of a country’s GDP growth from the average GDP growth of the euro area systematically affects the deviation of consumption growth relative to the euro area. This can be shown in the following baseline estimation relationship: (ΔC i ,t ) ( − ΔCtEMU = β ΔYi ,t − ΔYt EMU ) (4) ΔCi ,t is the consumption per capita growth rate, ΔYi ,t is GDP per capita growth rate for a specific country i in time t.4 The subscript EMU indicates euro area consumption per capita growth or euro area GDP per capita growth. Coefficient β is the percentage of uninsured risk in the euro area. If β is zero then relative deviations of per capita GDP growth in country i ( ΔYi ,t − ΔYt EMU ) will not systematically affect relative per capita consumption growth ( ΔCi ,t − ΔCtEMU ). That is the case for perfect risk sharing within the euro area. In reverse, β will be 1 for non risk sharing as no 4 ( ) ( ) We calculate ΔCi , t = ln Ci , t − ln Ci , t −1 and ΔYi , t respectively. 10 consumption smoothing via the capital market occurs. Then relative deviations of GDP growth are perfectly correlated with relative deviations in consumption growth. Hence, β is expected to be in the range of 0 and 1. We base our empirical analysis on a panel of quarterly data for twelve euro area countries.5 Our data set covers the period from Q1/1996 to Q3/2009. As we use year on year growth data, the seasonal pattern of quarterly data is not relevant. We leave Cyprus, Malta, Slovakia and Slovenia out of the sample because they have been euro area members for a too short period relative to other countries and relative to the full sample period. As data are not available for the full sample length for all EMU 12 countries, the number of observations reaches at maximum 632. Appendix 1 shows the data availability for each country. Data sources for consumption6 and GDP are the Eurostat Database on Quarterly National Accounts and for population figures the IMF World Economic Database. The latter are only available in an annual frequency and are assumed to be stable during the year. To evaluate hypothesis one, that participation in international risk sharing differs between euro area countries, we refer to the following standard panel estimation equation, with country-specific effects ρ i , a constant α and μ i,t indicating the white noise error term: (ΔC i ,t − ΔCtEMU ) = α + β EMU (ΔYi ,t − ΔYt EMU ) + ρi + μi ,t (5) However, to discriminate between country differences in consumption smoothing, we will include a threshold dummy variable. The estimation equation changes according to Sørensen et al. (2006). Starting from equation (5), we split uninsured risk β EMU into a country specific effect and a remaining uninsured effect: β EMU = β1 + β T ,i Di (6) 5 The respective countries are Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal and Spain. 6 We use final consumption, which includes public consumption. Therefore, we implicitly account for varying public budgets policy. 11 By doing so, the overall euro area uninsured risk β EMU consists of a country specific uninsured risk βT ,i with T indicating our threshold dummy approach. Di is a country specific dummy for each country i, which is 1 for the specific country i and 0 for all other 11 countries.7 β1 is the remaining uninsured risk. Inserting equation (6) into (5) yields: (ΔC i ,t ) ( ) ( ) − ΔCtEMU = α + β1 ΔYi ,t − ΔYt EMU + β T ,i ΔYi ,t − ΔYt EMU Di + ρ i + μ i ,t (7) Equation (7) includes an interaction term. Therefore, the interpretation of the coefficients changes (Hardy 1993: 81-82), which allows us to differentiate between single countries while using the explanatory power of the full data set. With respect to our estimation equation, the coefficients can be interpreted as follows: - β1 gives the value of uninsured risk if Di is zero, hence, yielding the average uninsured risk for all countries but country i. - βT ,i gives the number of units that β 1 changes if Di becomes one. - The sum β 1 + β T ,i is the uninsured risk for country i. Given that interpretation, we cannot reject our first hypothesis if there is a significant coefficient βT ,i , and thus, a significant variation in consumption smoothing between euro area countries. In a similar way, we test the validity of hypothesis two that an asymmetric foreign capital asset distribution has affected consumption smoothing in the euro area. For this specification, we use the net international investment position (IIP) in percent of GDP as a proxy for foreign capital asset distribution within the euro area (compiled from DataStream). We assume that overall uninsured risk β EMU is determined by the net IIP. Therefore, we again split β EMU in one part which is affected by the net IIP β IIP and the remaining uninsured risk β 2 : β EMU = β 2 + β IIP IIPi ,t (8) 7 Di is basically equal to the dummy variables related to country specific effects ρi. However, we use different notations, to separate different meanings. While ρi captures the country specific effect related to the constant α, Di indicates the country specific effect on uninsured risk β. 12 Inserting equation (8) into (5), the estimation equation becomes: (ΔC i ,t ) ( ) ( ) − ΔCtEMU = α + β 2 ΔYi ,t − ΔYt EMU + β IIP ΔYi ,t − ΔYt EMU IIPi ,t + ρ i + μ i ,t (9) In this specification, we interact relative GDP growth with the respective net IIP. The interpretation of coefficients changes in a similar way as above (Jaccard and Turrisi 2003). In so doing, β 2 indicates the value of uninsured risk if the net IIP is zero. This is the level of risk sharing for a hypothetic symmetric capital distribution. In contrast, β IIP shows whether risk sharing will be affected by a net IIP unequal zero. According to our analysis in section 3, we would expect a negative coefficient β IIP . Net foreign liabilities increase the share of uninsured risk, thus, they reduce a country’s consumption smoothing within the euro area. We also expect a lower uninsured risk β 2 for a hypothetic symmetric risk sharing compared to β EMU where we do not check for net IIP using equation (5). To check for changes over time and robustness we estimate for equations (9) and (5) additional to the full sample (Q1/1996-Q3/2009) several sub-samples. These are a “pre-EMU” sub-sample (Q1/1996-Q4/1999), an EMU period sub-sample (Q1/1999-Q3/2009) and two “late-EMU” sub-samples (Q1/2002-Q3/2009) and (Q1/2004-Q2/2009). Additionally, we recursively estimate β 2 and β EMU over time and display their development. For all specifications, we estimate a standard least square dummy variable (LSDV) estimator with country fixed effects, as suggested by the Hausman-Test (Hausman 1978), and with robust standard errors. Panel unit-root tests based on Levin et al. (2002), Maddala and Wu (1999) and Choi (2001) reject non-stationarity of our time series: 13 Table 2: Panel Unit Root test results (Q1/1996-Q3/2009). Consumption (ΔC-C EMU ) Panel Unit Root Test Levin/Lin/ ADF-Fisher PP-Fisher chiChu t* chi-square square 0.000 0.000 0.000 Output (ΔY-Y EMU ) 0.000 0.000 0.000 Interaction term IPP*(ΔY-YEMU) 0.000 0.002 0.000 Note: Lag selection has been conducted using the modified Hannan-Quinn criterion, an individual intercept is not allowed.8 Entries are p-values. Estimation Results Results concerning our hypothesis one are shown in Table 2. Eight of twelve country specific threshold dummies β T ,i reach the common levels of significance. Spain, France, Italy, the Netherlands and Austria have a positive coefficient. This suggests that these countries share on average less risk via the capital market than the respective remaining eleven EMU countries. In particular, Spain and Italy perform badly in our test on consumption smoothing. On the other hand, Luxembourg, Finland and Greece have a negative coefficient indicating a relatively better risk sharing within the euro area for these countries. The performance of Greece is surprising, but has to be interpreted carefully as only 35 observations are available for Greece. Overall, we cannot reject our hypothesis two. Results suggest significant differences in international risk sharing participation among euro area countries, which might be evidence of asymmetric international risk sharing. 8 A variation of the lag selection criteria as well as estimating the regression equation without intercept and trend or without trend does not change the results significantly. 14 Table 2: LSDV estimation results, discriminating for country specific risk sharing Sample # Country β1 β T,i Constant R-square Observations Q1/1996-Q3/2009 1 2 Belgium Germany 3 Ireland 4 Greece 5 Spain 6 France 0.526*** (0.130) 0.206 (0.131) 0.003** (0.001) 0.527*** (0.131) 0.202 (0.131) 0.003** (0.001) 0.487** (0.178) 0.131 (0.178) 0.003** (0.001) 0.544*** (0.131) -0.533*** (0.131) 0.004*** (0.000) 0.518*** (0.130) 0.424*** (0.130) 0.002** (0.000) 0.530*** (0.128) 0.273* (0.128) 0.003** (0.001) 0.41 632 0.42 632 0.42 632 0.42 632 0.42 632 0.41 632 10 Austria 11 Portugal 12 Finland Sample # Country β1 β T,i Constant R-square Observations 7 Italy 8 9 Luxembourg Netherlands 0.469*** (0.126) 0.477*** (0.126) 0.003*** (0.001) 0.663*** (0.073) -0.562*** (0.073) 0.002*** (0.000) 0.524*** (0.130) 0.313** (0.130) 0.003** (0.001) 0.524*** (0.131) 0.235* (0.131) 0.003** (0.001) 0.522*** (0.133) 0.204 (0.133) 0.002** (0.001) 0.554*** (0.138) -0.283* (0.138) 0.003** (0.001) 0.45 632 0.50 632 0.42 632 0.42 632 0.42 632 0.42 632 Robust standard errors are reported in parentheses. *, ** and *** indicate significance at 10%, 5% and 1% level. Related to hypothesis two, the impact of net IIP on consumption smoothing in the euro area is shown in Table 3. The respective coefficient β IIP becomes significant in almost all samples, at least at the 10 percent level. The net IIP has a positive impact prior to EMU. Nevertheless, the coefficients are negative in the EMU samples and rise in their absolute values since 1999 indicating a rising importance of the net IIP for international risk sharing within the euro area. In the sample Q1/2004-Q3/2009, international risk sharing increases (decreases) on average by 0.005 units per each percentage of net IIP assets (liabilities) per GDP. That figure seems to be small. However, net IIP positions have reached quite high levels (negative as well as positive) for single countries (see Table 1). Moreover, the levels of uninsured risk assuming a symmetric international risk sharing β 2 are significantly lower than β EMU in Table 4 where we do not check for 15 net IIP, referring to equation (5). Based on a t-test with H0: βEMU= β2, the t-statistic for sample Q1/1999-Q3/2009 is 2.68, for sample Q1/2002-Q3/2009 3.34 and for sample Q1/2004-Q3/2009 3.37. The critical value for 5 percent significance is 1.98. Thus, we have to reject the H0; β2 is in all three samples significantly lower than βEMU. Table 3: LSDV estimation results for β 2 , controlling for net IIP Sample # β2 β IIP Constant R-square Observations Q1/1996Q4/1999 13 Q1/1996Q3/2009 14 Q1/1999Q3/2009 15 Q1/2002Q3/2009 16 Q1/2004Q3/2009 17 0.851*** (0.051) 0.007*** (0.002) 0.000** (0.000) 0.510*** (0.140) -0.003 (0.003) 0.002*** (0.001) 0.270*** (0.067) -0.003* (0.002) 0.005*** (0.001) 0.221*** (0.046) -0.004*** (0.001) 0.006*** (0.000) 0.212*** (0.051) -0.005*** (0.000) 0.007*** (0.000) 0.83 105 0.43 551 0.23 446 0.23 333 0.38 240 Robust standard errors are reported in parentheses. *, ** and *** indicate significance at 10%, 5% and 1% level. Table 4: LSDV estimation results of β EMU Sample # β emu Constant R-square Observations Q1/1996Q4/1999 18 Q1/1996Q3/2009 19 Q1/1999Q3/2009 20 Q1/2002Q3/2009 21 Q1/2004Q3/209 22 0.669*** (0.103) -0.000 (0.000) 0.532*** (0.127) 0.003*** (0.001) 0.450** (0.164) 0.004** (0.001) 0.375* (0.175) 0.004*** (0.001) 0.384** (0.174) 0.004*** (0.001) 0.68 124 0.41 632 0.30 508 0.19 372 0.20 267 Robust standard errors are reported in parentheses. *, ** and *** indicate significance at 10%, 5% and 1% level. That result is confirmed by recursive estimations. We estimate equations (5) and (9) forty-two times. We start with the full sample and reduce the sample size in every step by starting one quarter ahead until Q4/2006. The sample end remains always Q3/2009. Therefore, we constantly increase the weight of more resent data. Figure 5 shows the development of β EMU and β 2 over time and the corresponding p-values below. It clearly shows, that we measure more uninsured risk in the case of 16 asymmetric international risk sharing (indicated by β EMU ) rather than assuming a hypothetic symmetric international risk sharing (indicated by β 2 ). Figure 3: Development of β EMU and β 2 over time based on recursive LSDV estimations (variable start quarter – end quarter Q3/2009). Coefficient 0.600 0.500 βEMU 0.400 0.300 β2 0.200 0.100 1996Q01 1997Q03 1999Q01 0.20 2000Q03 2002Q01 start quarter 2003Q03 2005Q01 2006Q03 2005Q01 2006Q03 p-value 0.15 βEMU 0.10 0.05 β2 0.00 1996Q01 1997Q03 1999Q01 2000Q03 2002Q01 2003Q03 start quarter Therefore, we cannot reject our second hypothesis. Persistent capital flows, leading to asymmetric capital asset distribution among euro area countries, seem to have affected the international risk sharing mechanism and thus consumption smoothing in the euro area. Results suggest that, overall consumption smoothing within the euro area seems to have been limited by asymmetric risk sharing. Furthermore, countries with net foreign assets are supposed to share risks much better than countries with net foreign liabilities, which is in line with our theoretical findings. We can summarize our empirical analysis for the euro area that (i) the participation in risk sharing differs significantly between countries. The divergence in net IIP seem to be one important determinant, for (ii) different country specific consumption 17 smoothing and (iii) for limiting overall consumption smoothing throughout the euro area. That leaves countries with high foreign liability less insured against adverse economic developments within the euro area and with more volatile consumption. 5. Conclusion In this paper we have analysed how an asymmetric foreign capital asset distribution changes international risk sharing and effects consumption smoothing in the euro area. That analysis has become necessary after almost a decade of one-way capital flows in particular from Germany to southern European euro area countries. That resulted in a divergence of current account balances as well as net international investment positions within the euro area. Substantial asymmetric foreign capital asset distribution creates an asymmetric international risk sharing mechanism, which reduces, in comparison to a symmetric case, consumption smoothing in the monetary union as a whole and in both types of countries – countries with net foreign assets countries as well as with net foreign liabilities. However, countries with high net foreign liabilities are in particular exposed to consumption volatility. These theoretical propositions are confirmed by our empirical analysis for the first 12 euro area countries. Hence, the international risk sharing mechanism does not work as properly and as equally in the euro area as assumed by the theoretical symmetric international risk sharing approach by Mundell (1973) and as argued by the European Commission (EC 2008). Without sufficient labour market flexibility, very limited room for fiscal policy and also poorly working international risk sharing, the adjustment capability of the euro area seems to be very limited. That is an alarming result for the euro area in general and in particular for currently suffering euro area countries facing real adjustment needs. 18 References Atkeson, Andrew / Bayoumi, Tamim 1993: Do Private Capital Markets Insure Regional Risk? Evidence from the United States and Europe, Open Economies Review 4, 303-324. Blanchard, Oliver / Giavazzi, Francesco 2002: Current Account Deficits in the Euro Area: The End of the Feldstein-Horioka Puzzle? Brookings Papers of Economic Activity 33, 147-186, Brookings Institution, Washington, DC. Choi, In 2001: Unit Root Tests for Panel Data, Journal of International Money and Finance 20, 249-272. de Santis RA, Lührmann M 2006: On the Determinants of External Imbalances and Net International Portfolio Flows: A Global Perspective, ECB Working Paper 651. European Central Bank, Frankfurt/Main. Demyanyk, Yuliya / Ostergaard, Charlotte / Sørensen, Bent E. 2008: Risk Sharing and Portfolio Allocation in EMU, European Economy Economic Papers 334, European Commission, Brussels. European Commission (EC) 1993: European Commission 1993: Stable Money – Sound Finances, Community Public Finances in the Perspective of EMU, European Economy 53, Brussels. European Commission (EC) 2008: EMU@10: Success and Challenges after 10 Years of Economic and Monetary Union, European Economy 2/2008, Brussels. European Commission (EC) 2009: Quarterly Report on the Euro Area 8 (1). Brussels. European Commission (EC) 2010: Surveillance of Intra-Euro-Area Competitiveness and Imbalances. European Economy 1/2010. Brussels. Fagan, Gabriel / Gaspar, Vitor 2007: Adjusting to the Euro, ECB Working Paper 716, European Central Bank, Frankfurt/Main. Hardy, Melissa A. 1993: Regression with Dummy Variables, Sage, Newbury Park. Hausman, Jerry .A. 1978: Specification Tests in Econometrics, Econometrica 46 (6), 1251–1271. Horioka, Charles / Feldstein, Martin 1980: Domestic Saving and International Capital Flows, Economic Journal 90, 314-329. Jaccard J / Turrisi R 2003: Interaction Effects in Multiple Regression. Sage, Newbury Park. Kose, M. Ayhan / Prasad, Eswar / Terrones, Marco 2007: How Does Financial Globalization Affect Risk Sharing? Patterns and Channels, IMF Working Paper 238, International Monetary Fund, Washington, DC. Levin, Andrew / Lin, Chien-Fu / Chu, Chia-Shang James 2002: Unit root tests in panel data: asymptotic and finite-sample properties, Journal of Econometrics 108 (1), 1-24. Mendoza E / Quadrini V / Rios-Rull JV 2007: Financial Integration, Financial Deepness and Global Imbalances, NBER Working Paper 12909, National Bureau of Economic Research. Cambridge, M.A. 19 Maddala, G.S. / Shaowen Wu 1999: A Comparative Study of Unit Root Tests with Panel Data and a new Simple Test, Oxford Bulletin of Economics and Statistics 61, 631-652. Melitz, Jacques / Zumer, Frederic 1999: Interregional and International Risk Sharing and Lessons for EMU, Carnegie-Rochester Conference Series on Public Policy 51(1), 149-188. Mundell, Robert 1961: A Theory of Optimum Currency Areas, The American Economic Review 51, 657-665. Mundell, Robert 1973: Uncommon Arguments for Common Currencies, in: Johnson, Harry / Swoboda, Alexander (ed.): The Economics of Common Currencies, London, 114-132. Sørensen, Bent E. / Wu, Yi-Tsung / Yosha, Oved / Zhu, Yu 2007: Home Bias and International Risk Sharing: Twin Puzzles Separated at Birth, Journal of International Money and Finance 26, 587-605. Stavrev, Emil 2007: Growth and Inflation Dispersions in EMU: Reasons, the Role of Adjustment Channels, and Policy Implications, IMF Working Paper 167, International Monetary Fund, Washington, DC. Zemanek, Holger 2010: Competitiveness in the Euro Area: The Problem that still needs to be solved, forthcoming in Economic Affairs. Zemanek, Holger / Belke, Ansgar / Schnabl, Gunther 2010: Current Account Balances and Structural Adjustment in the Euro Area. International Economics and Economic Policy 7 (1), 83-127. 20 Appendix 1. Data availability and data sources Data: Nominal GDP, quarterly Source: Eurostat Availability:* Q1/1995 – Q3/2009 Ireland: Q1/1997-Q/2009 Greece: Q1/2000-Q/2009 Consumption, quarterly Eurostat Q1/1995 – Q3/2009 Ireland: Q1/1997-Q/2009 Greece: Q1/2000-Q/2009 Population, annually IMF World Economic Outlook 1995-2009 International Investment Position (IPP), annually, Assets-Liabilities IMF International Financial Statistics, via DataStream 1995-2008 (quarterly data were created by linear interpolation) Ireland: 2001-2008 Greece: 1998-2008 Luxembourg: 2002-2008 Portugal: 1996-2008 * y-o-y growth figures used in the empirical analysis start four quarters later. 21