An Unemployment Insurance Scheme for the Euro Area? A Comparison of Different Alternatives Using Micro Data

We analyze different alternatives how a common unemployment insurance system for the euro area (EA) could be designed and assess their effectiveness to act as an insurance device in the presence of asymmetric macroeconomic shocks. Running counterfactual simulations based on micro data for the period 2000-13, we highlight and quantify the trade-off between automatic stabilization effects and the degree of cross-country transfers. In the baseline, we focus on a non-contingent scheme covering short-term unemployment and find that it would have absorbed a significant fraction of the unemployment shock in the recent crisis. However, 5 member states of the EA18 would have been either a permanent net contributor or net recipient. Our results suggest that claw-back mechanisms and contingent benefits could limit the degree of cross-country redistribution, but might reduce desired insurance effects. We also discuss moral hazard issues at the level of individuals, the administration and economic policy.


Introduction
The Great Recession and the resulting European debt crisis have revived the debate about deeper …scal integration in the European Monetary Union (EMU). The EMU is an atypical monetary union because monetary policy is decided at the central (European) level while …scal policy is carried out at the sub-central (member-state) level (Bordo et al. 2013). 1 Some observers argue that national automatic stabilizers provided insu¢ cient income insurance during the crisis as some EMU member states lost access to private capital markets and conclude that common …scal stabilization mechanisms are necessary to make EMU more sustainable (Bertola 2013, IMF 2013. While the main argument in favor of integrated …scal mechanisms in the euro area is that they should act as insurance devices in the presence of asymmetric macroeconomic shocks, the main concerns in the debate relate to the issues of permanent transfer ‡ows within the currency union and moral hazard, in particular with regard to negative incentive e¤ects inducing national governments to neglect structural reforms or …scal consolidation. How could a …scal risk sharing mechanism in the euro area be designed? The former President of the European Council, Herman van Rompuy, has suggested the following: "An EMU …scal capacity with a limited asymmetric shock absorption function could take the form of an insurance-type system between euro area countries. [...] The spe-ci…c design of such a function could follow two broad approaches. The …rst would be a macroeconomic approach, where contributions and disbursements would be based on ‡uctuations in cyclical revenue and expenditure items [...]. The second could be based on a microeconomic approach, and be more directly linked to a speci…c public function sensitive to the economic cycle, such as unemployment insurance."(Van Rompuy 2012). The European Commission built upon this initiative when launching its o¢ cial report entitled "A blueprint for a deep and genuine Economic and Monetary Union -Launching a European Debate"(European Commission 2012).
Since then, the perspectives of a European …scal union and di¤erent reform proposals along the lines of the Van Rompuy report have been analyzed in various studies. 2 For the 'macroeconomic approach'suggestions include a cyclical shock absorber based 1 In the following we equivalently use "EA", "EMU" and "Eurozone" to refer to the current 18 member states of the European Currency Union and thus, only to those EMU members who have already introduced the euro.
2 First analyses of potential insurance e¤ects if the EMU were more …scally integrated date back to the introduction of the euro (Fatás 1998 andForni andReichlin 1999), adding to the vast literature on insurance e¤ects in existing …scal federations such as the US (see e.g. Bayoumi andMasson 1995 andAsdrubali et al. 1996). More recent contributions include Bargain et al. (2013) who analyze the economic implications of a fully integrated European tax and transfer system and a …scal equalization mechanism based on taxing capacity and expenditure needs for 11 founding members of the euro area, and Feyrer and Sacerdote (2013) who ask to what extent economic shocks would be absorbed by the center if the EU were as …scally integrated as the US. The question of how to optimally design insurance mechanisms and the political economy of …scal unions has also gained renewed interest in the more theoretical literature (cf. Evers 2012, Farhi and Werning 2014, Luque et al. 2014). on output-gaps (Enderlein et al. 2013) and a stabilization fund for the euro area (Furceri and Zdzienicka 2013). For the 'microeconomic approach', the discussion has focused on the idea of a common EMU-wide unemployment insurance system (henceforth EMU-UI) as proposed among others by Deinzer (2004), Dullien (2014a) and Andor (2014). 3 Previous studies on the economic e¤ects of an EMU-UI system are based on aggregate macro-level data and focus on overall net contributions across euro area member states. This is the …rst paper based on household micro data that provides a comprehensive and systematic analysis of a wide range of design options for an EMU-UI system. 4 Our counterfactual simulation experiment includes the EA18 member states and covers the whole period since the start of the euro in 1999. Besides net contributions per member state, our analysis includes coverage ratios and automatic …scal stabilization e¤ects of a basic EMU-UI scheme that (partly) replaces national unemployment insurance systems. In addition, we explore the e¤ects of experience rating and claw-back mechanisms that are supposed to limit the amount of redistribution across member states. We also compare the basic EMU-UI scheme to a variant with 'contingent', i.e., trigger-based bene…t payments that provide income insurance only if the labor market situation deteriorates signi…cantly in a given member state. Moreover, we run several sensitivity checks regarding coverage and generosity levels of the scheme. We also discuss various concerns and potential adverse e¤ects of an EMU-UI system, in particular the view that such a system would lead to a transfer union in Europe and moral hazard issues. Importantly, the aim of the paper is not to serve as a policy proposal but rather as a conceptual experiment, providing general insights in the e¤ects of various design options for a basic EMU-UI.
Our main results are as follows. We …nd that a basic euro area unemployment insurance scheme with a replacement rate of 50 per cent, a maximum duration of ben-e…t receipt of 12 months and a broad coverage of all new unemployed with previous employment income could be implemented with a relatively small annual budget. Over the period 2000-13, average bene…ts would have amounted to roughly 49 billion euro per year …nanced by a uniform contribution rate across member states of 1.57 per cent on employment income. While the scheme does not lead to permanent redistribution per se as only short-term (rather than structural) unemployment is insured at the central level, our simulations show that a small number of member states would have 3 See also IAB (2013), Centre for European Policy Studies (2014), Dullien et al. (2014) and Lellouch and Sode (2014). Claeys et al. (2014) provide an overview of policy challenges associated with an EMU-UI system. 4 Jara and Sutherland (2014) also use micro data to analyze to what extent an EMU-unemployment insurance system would top-up national unemployment insurance systems in 10 euro area member states to guarantee a minimum level of income protection. Their analysis is conceptually di¤erent from ours as they compare stabilization gaps of existing national systems which would be …lled by the centralized unemployment insurance scheme while we focus on the economic e¤ects of the latter ignoring potential top-ups of national unemployment insurance systems. Both studies are thus complementary to each other. been net contributor or net recipient in each year of our simulation period. Largest net contributors are Austria, Germany and the Netherlands with average yearly net contributions of 0.2-0.42 per cent of GDP, while Latvia and Spain are the largest net recipients (average yearly net bene…ts of 0.33 and 0.53 per cent of GDP). We …nd that household incomes would have been stabilized in particular at the beginning of the recent economic crisis. Our measure for automatic stabilization, the income stabilization coe¢ cient, indicates that 36 per cent of the unemployment shock in 2009, measured as the decline in income due to the surge in unemployment, would have been absorbed by the scheme in the Eurozone. However, this e¤ect would have diminished the longer the crisis lasted as the share of (non-eligible) long-term unemployed was rising in the majority of member states. Schemes with lower coverage ratios and generosity levels generate smaller cross-country transfers, but also reduce desired insurance e¤ects. Country-speci…c contribution rates that would have balanced the budget in each member state over the period 2000-13 range from 0.75 per cent in the Netherlands to 3.3 per cent in Spain. The spread becomes larger if budgets are required to be balanced in each single year and range from 0.46 per cent in Luxembourg to 5.8 per cent in Latvia. However, revenue-neutrality can be imposed only ex-post when accumulated net bene…t payments and changes in the tax base are known. Therefore, we explore to what extent cross-country transfers can be restricted ex-ante by pre-speci…ed rules and analyze two di¤erent claw-back mechanisms, i.e., ex-post adjustments of country-speci…c contribution rates. We …nd that they would have led to smaller accumulated surpluses/de…cits relative to the benchmark of a uniform and time-invariant contribution rate in some, but not in all member states. Finally, we consider a contingent bene…t scheme which is activated if the unemployment rate in a given member state is 1 percentage point higher than in one of the previous three years. Under this system no member state would have been in a permanent net contributing / receiving position. With 21 billion euro per year, the overall budget and thus the amount of cross-country redistribution would have been less than half as large as under the non-contingent scheme in the baseline.
The paper is structured as follows. In section 2, we discuss di¤erent alternatives how a common EMU-UI system could be designed. In addition, we present key features of the simulated EMU-UI schemes. Section 3 describes the framework of the analysis. Baseline results are presented in section 4. Alternative EMU-UI schemes with clawback mechanisms based on experience rating and contingent bene…ts are analyzed in section 5. Section 6 concludes.
2 Possible characteristics of an EMU-UI system Design options. A common unemployment insurance system for the euro area could be designed in various ways. Three key options have been discussed in the literature and in the policy debate so far. A …rst option would be a common EMU-UI system that provides a basic level of insurance by partly replacing national unemployment insurance systems. Bene…ts from the euro area system could be topped up by additional payments from national unemployment insurance systems. Hence, there would be room for diversity across member states so that existing di¤erences with regard to replacement rates and bene…t duration could be preserved. The EMU-UI system would be …nanced by social insurance contributions with a contribution rate that could be uniform across Eurozone member states or country-speci…c and time-variant to restrict cross-country transfers. 5 An important feature of such a scheme is that it would provide income insurance to the unemployed (under certain eligibility conditions) irrespective of the size of the unemployment shock in a given member state. As an alternative, a common scheme could provide income stabilization only in the event of large (unemployment) shocks. Such contingent unemployment bene…ts would be triggered if the level and/or change in overall unemployment has reached a pre-determined threshold in a given period. 6 National unemployment insurance systems would still be in place in normal times. As a third option, the euro area unemployment insurance scheme could complement national systems by providing additional transfers which would either top up national bene…ts or kick-in if national bene…ts expire. The payout rules of this scheme could be trigger-based as well. Such a system would be comparable to the US unemployment insurance system where regular state bene…ts can be complemented by two types of bene…ts extension programs which are at least partly provided by the federal government, the Extended Bene…t program (EB) and emergency bene…ts. 7 Concerns. A major concern with an EMU-UI system is that it would lead to permanent transfers across euro area member states. How do the three variants for an EMU-UI system di¤er with regard to the risk of permanent redistribution? A basic EMU-UI scheme would not lead to permanent redistribution per se given that such a scheme conditions on changes in employment status rather than on unemployment levels. Di¤erences in unemployment rates alone do not (necessarily) lead to permanent redistribution because bene…ts would be targeted to cyclical (short-time) unemploy-5 Cf. Dolls et al. (2014) and Dullien (2014b). 6 Cf. Epaulard (2014) and Gros (2014). Other triggers could be the short-term unemployment rate or the insured unemployment rate which is used in the US unemployment insurance system (besides the total unemployment rate) as a trigger for bene…t extension programs (Nicholson et al. 2014). 7 Cf. Congressional Budget O¢ ce (2012) and Nicholson et al. (2014). Note that in the US regular state bene…ts are paid for a period which usually lasts not longer than 6 months. The large extensions of unemployment insurance provided by the US federal government in the 2009-12 period increased the bene…t duration to 99 weeks in many US states. Unemployment bene…ts in the EMU are usually granted much longer than regular state bene…ts in the US (Esser et al. 2013). ment and expire after a certain time span. It may nevertheless happen that (net) transfers are unevenly distributed across member states if ‡ows into unemployment diverge permanently or if there are permanent di¤erences in the level of short-term unemployment. 8 This risk could be reduced by claw-back mechanisms based on experience rating or if transfers were trigger-based as under the contingent bene…t scheme. Clearly, redistributive e¤ects of the former (latter) scheme would depend on the exact claw-back mechanism (choice of the trigger). The risk of permanent transfers would be high with an EMU-UI scheme that provides extended bene…ts after national unemployment bene…ts expire because such a scheme would be likely to cover not only cyclical, but also structural unemployment. Moreover, it could incentivize governments to cut national unemployment insurance bene…ts as the EMU-UI system would step in.
A further concern related to moral hazard is that a common EMU-UI system could undermine incentives for national governments to address structural weaknesses of the labor market. One argument against this claim is that national governments would still bear the cost of long-term unemployment under a basic, contingent or non-contingent EMU-UI system. This argument is much weaker, however, with an extended bene…t program which is likely to cover also structural unemployment. Moreover, incentives to pursue active labor market policies such as short-time work could be adversely a¤ected by an EMU-UI system given that the cost of short-term unemployment would be borne by the common pool.
Additional concerns relate to other moral hazard issues including administrative manipulation and adverse incentive e¤ects at the individual level with regard to job search and labor supply. National administrations would have incentives to use their discretion to increase the number of bene…t recipients. Incentives to manipulate would depend on the characteristics of the system, e.g. the required employment period or a waiting period for EMU-UI bene…ts. The longer both periods are, the more costly would administrative manipulation be, but longer periods would also reduce desired insurance e¤ects. Distortions at the individual level depend on the overall bene…t level (EMU plus national bene…ts) relative to the status quo and in case of the extended bene…t scheme also on the bene…t duration. The e¤ect of a common EMU-UI system on migration responses in case of unemployment is ambiguous. The portability of unemployment bene…t claims might increase the willingness to migrate and to search for a job in a member state with better labor market conditions, but could potentially also reduce incentives for active job search if EMU-UI bene…s are more generous than national bene…ts.
Key features of the simulated EMU-UI schemes. The current debate focuses on a basic EMU-UI system (contingent and non-contingent) as the risk of permanent transfers and moral hazard issues are perceived to be less severe compared to an ex-8 Economies where seasonal employment like in tourism plays an important role would be likely to have larger ‡ows into and out of unemployment. tended bene…t system. In the baseline scenario, we therefore focus on a basic, noncontingent EMU-UI scheme with a replacement rate of 50 per cent of previous gross earnings and a broad coverage of the short-term unemployed. 9 Eligible to EMU-UI bene…ts are all newly unemployed with previous employment income for a period of up to 12 months. The scheme is …nanced by social insurance contributions with a uniform contribution rate across member states and calibrated to be revenue-neutral at the Eurozone-level (but not the member-state level) over the simulation period. This scheme is labeled as variant A henceforth. In addition, we explore how our results change if we vary some key parameters of the baseline scheme in terms of coverage rates and generosity levels. With regard to the former, we introduce a waiting period of 2 months after job loss before eligibility to EMU-UI bene…ts begins in order to diminish the e¤ect of seasonal unemployment (variant B). Moreover, while the coverage rate of the newly unemployed is assumed to be 100 per cent in the baseline (upper bound estimate) 10 , we assume as a lower bound estimate that only the share of short-term unemployed covered by national unemployment insurance systems receives bene…ts from the euro area scheme (variant C). Variants B and C are otherwise identical to variant A, i.e., the replacement rate is 50 per cent of previous gross earnings and bene…ts are only paid up to the 12th month after job loss. In terms of generosity, we consider a scheme with a maximum bene…t amount of 50 per cent of median gross income in a member state (variant D), a replacement rate of 35 per cent of gross earnings (variant E) 11 , and a scheme where variants D and E are combined (variant F). Note that in variants D-F, the coverage rules of variant A are applied. Additionally, we compare the baseline EMU-UI scheme (variant A) to three alternative variants in which we impose revenue-neutrality at the member-state level (experience rating), adjust contribution rates based on past net balances in each member-state (claw-back mechanisms) and make the basic EMU-UI scheme trigger-based (contingent bene…ts). The analysis of redistributive and stabilizing properties of these additional variants is an important extension to the previous literature because they are often assumed to alleviate the risk of permanent redistribution and moral hazard issues.

EU-SILC and EUROMOD
Di¤erent methodological approaches for an analysis of the economic e¤ects of an EMU-UI system are possible. While previous research has mainly used aggregate macro level data, we rely on representative household micro data for the EA18 and use EURO-MOD, a static tax-bene…t calculator for the European Union countries, for counterfactual simulations. The key advantage of using a micro data approach in the present context is that it enables us to account for heterogeneity in various characteristics of the populations in di¤erent countries which macro data approaches cannot capture. EUROMOD input-data are mainly based on the European Union Statistics on Income and Living Conditions (EU-SILC) released by Eurostat (Eurostat 2012). The simulated components include most direct taxes (especially income taxes on all sources of income including tax credits, payroll taxes and social insurance contributions) and bene…ts (e.g. welfare bene…ts, social assistance and some transfers based on previous contributions, e.g. unemployment bene…ts). 12

Simulation experiment
An important feature of EUROMOD is that it allows for counterfactual ex-ante simulations. In our empirical analysis, we introduce an unemployment insurance scheme for the current 18 member states of the euro area and ask what would have happened if such a scheme had been introduced from the start of the euro in 1999. 13 In a …rst step, we reweight our base year household micro data from 2008 such that labor market conditions (unemployment rate, earnings and size of labor force) correspond to the levels observed in the starting year of our simulation period. 14 In the next step, we simulate a sample of repeated cross-sections for each euro area member state for the period 2000-13 by reweighting our cross-country micro data such that total unemployment, short-term unemployment, earnings and the size of the labor force follow observed trends in each member state during the simulation period. 15 An increase (a 12 Sutherland and Figari (2013) provide more detailed information on EUROMOD and the underlying input data. 13 We assume that the current EA18 would have existed from 1999 onwards. In fact, it would complicate the interpretation of our results if we included new member states only after adoption of the euro. 14 Note that EUROMOD is based on cross-sectional data and not panel data. The …rst data year available in the current version is 2008. Hence, we need to reweight the data (as commonly done in microsimulation studies) using macro aggregates to re ‡ect the potentially changing structure of the economy during the simulation period. This allows us to construct a series of reweighted cross-sections for the period of analysis. 15 Earnings growth along the intensive margin and changes in the size of the labor force are modeled in order to account for changes in the tax base of the euro area unemployment insurance system. Growth rates in nominal compensation per employee, unemployment rates and the size of the labor force are obtained from the AMECO database, information on short-term unemployment from decrease) of the unemployment rate is modeled by increasing the weights of the unemployed (employed) while the weights of the employed (unemployed) are decreased correspondingly, i.e., in e¤ect a fraction of employed (unemployed) households is made unemployed (employed). 16 Our analysis is based on the following simplifying assumptions. First, we do not take into account general equilibrium e¤ects of an EMU-UI system, i.e., our analysis remains in a partial equilibrium context. This implies that we abstract both from potential moral hazard of national governments and administrations which could have adverse labor market e¤ects as well as from potential growth-enhancing e¤ects of an EMU-UI scheme. Accounting for these macroeconomic feedback e¤ects would require to link our micro data to a macro-econometric simulation model. However, past validations of micro-macro linkages point to a considerable degree of uncertainty of macroeconomic projections (Peichl 2009). Second, we do not simulate individual behavioral responses, e.g. potential migration responses, changes in hours worked or di¤erent patterns of entries and exits to the labor force which could follow the introduction of an EMU-UI. 17 In the light of these assumptions, our results should be interpreted as '…rst-round' e¤ects of an EMU-UI system.

Coverage rates
Before analyzing coverage rates of various EMU-UI schemes, we …rst provide descriptive statistics on the unemployment rates of EMU member states. Figure 1 shows that there are signi…cant di¤erences in both levels and trends in unemployment rates and the share of short-term unemployed for the period 2000-13 across euro area member states. Di¤erences between Germany on the one hand and Greece, Ireland and Spain on the other hand are particularly remarkable. In Germany, the unemployment rate increased from 2001 onwards, peaked at 11.3 per cent in 2005 being the second highest rate in the euro area in that year, but constantly fell afterwards. Contrary, unemployment rates increased tremendously in Greece, Ireland and Spain from 2008/2009onwards, up to 14.7 per cent in Ireland in 20123) per cent in Spain (Greece) in 2013. Other member states such as Cyprus, Estonia, Italy and Portugal were also hit by Eurostat. 16 See Immvervoll et al. (2006), Bargain et al. (2012) and Dolls et al. (2012) for similar applications of the reweighting approach. When modeling (un)employment shocks, the new (un)employed are evenly chosen from the full cross-section. An alternative approach would be to assume that the sociodemographic characteristics of the (un)employed remain constant which seems less realistic given the length of our simulation period. 17 Bargain et al. (2013) account for labor supply behavior after the introduction of a European tax and transfer system. They …nd that labor supply responses are marginal and do not alter their main results. large unemployment shocks during the crisis. This would have led to increasing shares of bene…t recipients of the euro area scheme relative to the labor force, in particular in those member states most a¤ected by rising unemployment rates. However, the share of short-term unemployed (relative to total unemployment) was falling the longer the crisis lasted. Hence, coverage rates of the baseline EMU-UI scheme (which are equal to the share of short-term unemployed) would have declined as well in spite of rising unemployment rates in recent years. Figure 2 summarizes average coverage rates of EMU-UI over the period 2000-13 for the baseline scenario of full coverage of all new unemployed (variant A), a scenario with a waiting period of 2 months at the beginning of the unemployment spell (variant B) and a scheme covering only the share of short-term unemployed which receives national unemployment insurance bene…ts (variant C). 18 Figure 2 shows that di¤erences in average coverage rates are substantial ranging from 34 per cent in Slovakia to 78 per cent in Finland in the baseline. A waiting period would to some extent exclude seasonal unemployment (like in tourism) from coverage. Our results indicate that coverage rates would indeed decline signi…cantly as a considerable fraction of unemployment spells in euro area member states lasts for not more than 2 months. Finally, coverage rates are much lower than in the baseline if we apply national coverage rates of the short-term unemployed. Lowest coverage rates of roughly 10 per cent are found for Greece, Italy and Slovakia, whereas more than 40 (50) per cent of the short-term unemployed are covered by national unemployment insurance systems in Austria (Finland). FR 18 Information about the share of unemployed whose unemployment spell does not exceed 2 months (waiting period) and the share of unemployed receiving national unemployment insurance bene…ts is obtained from Eurostat. Coverage rates for each year of our simulation period are shown in Table 2 in the Appendix.  Waiting period: no bene…ts paid in the …rst 2 months of the unemployment spell (variant B). National coverage: only the share of short-term unemployed covered by national unemployment insurance systems eligible to bene…ts from euro area scheme (variant C). Sources: EUROSTAT and own calculations based on EUROMOD.

Budgetary e¤ects and …nancial ‡ows
For the baseline scheme (variant A), a uniform contribution rate across member states of 1.57 per cent on employment income would have led to revenue-neutrality at the euro area level over the period 2000-13. 19 Note that the scheme can run de…cits and surpluses in single years which has important implications for its automatic stabilization e¤ects as discussed in the next section. Figure 3 shows the evolution of contributions and bene…ts for the EA18. While contributions would have almost constantly grown over the period due to growth in nominal earnings, bene…t payments would have ‡uctuated to a much larger extent. On average, bene…ts and contributions amount to 49 billion euro per year. The scheme would have run surpluses from 2000-03 and from 2006-08 and de…cits in the remaining years, in particular during the recent …nancial and economic crisis. Figure 4 shows average yearly net contributions as well as minimum and maximum payments for the baseline scenario. Relative to GDP, Austria, Germany and the Netherlands would have been the largest net contributors with average net contributions of 0.2 per cent in Germany, 0.25 per cent in Austria and 0.42 per cent in the Netherlands. Latvia (-0.33 per cent) and Spain (-0.53 per cent) would have been the largest net recipients. Interestingly, the majority of member states would have been net contributor in some years and net recipients in other years. Notable exceptions are Austria and the Netherlands (France, Latvia and Spain) which would have always been net contributors (recipients). Finally, we compare the baseline scheme to variants with lower coverage and generosity levels. In terms of coverage, we introduce a waiting period for the …rst two months of the unemployment spell (variant B) and assume that only the share of short-term unemployed covered by national unemployment insurance systems is eligible to the euro area scheme (variant C) as in the previous section. Moreover, we alter the generosity by capping the maximum bene…t amount at 50 per cent of median income in a given member state in a given year (variant D), by reducing the replacement rate to 35 per cent of gross income (variant E) and by combining the latter two scenarios (variant F). Results are presented in Figures 5 and 6 and in Tables 3-6 in the Appendix. Figure 5 shows that a waiting period would reduce net contributions considerably in most member states compared to the baseline scenario, in some cases by almost 50 per cent, indicating that a large share of the short-term unemployed was able to …nd a new job within a short time period. Seasonal (un)employment patterns are one factor explaining this …nding. When only the share of short-term unemployed covered by national unemployment insurance systems is eligible, net contributions shrink further and some member states which are a net contributor in the baseline become a net recipient (Belgium, Germany) or vice versa (Cyprus, Estonia, Greece, Portugal, Slovakia). This is due to large di¤erences in coverage rates of national unemployment insurance systems as discussed in section 4.1. Figure 6 shows that net contributions become smaller the less generous the euro area unemployment scheme is. In the least generous case of a scheme with a 35 per cent replacement rate and bene…ts capped at 50 per cent of median income, average net contributions shrink to 0.25 per cent of GDP in the Netherlands, the largest net contributor, and to -0.31 per cent in Spain, the largest net recipient. Interestingly, Estonia and Portugal become net contributors rather than net recipients if bene…ts are capped which is due to low median incomes in these member states.

Automatic …scal stabilization
Automatic …scal stabilization is associated with the ability of taxes and transfers to automatically stabilize disposable income and consequently consumption in the event of macroeconomic shocks. This relies on a simple mechanism: in the presence of a given negative shock to gross income, taxes decline and transfers increase, with the decline in disposable income being smaller than the shock to gross income (Auerbach and Feenberg 2000, Kniesner and Ziliak 2002, Dolls et al. 2012. Several components of government budgets are a¤ected by the macroeconomic situation in ways that operate to smooth the business cycle, with progressive income taxes and unemployment bene…ts being the most prominent examples. 20 A common measure for estimating automatic stabilization based on micro data is the "normalized tax change"used by Auerbach and Feenberg (2000) which can be interpreted as "the tax system's built-in ‡exibility" (Pechman 1973(Pechman , 1987. Based on this idea, Dolls et al. (2012) de…ne the "income stabilization coe¢ cient", I , that shows how changes in market income Y M (de…ned as the sum of all incomes from market activities such as (self)-employment, business and property income) translate into changes in disposable income Y D (market income minus taxes plus bene…ts) through changes in net tax payments T . They extend the concept of normalized tax change to include other taxes as well as SIC and transfers.
In our simulations, we follow their approach and calculate the income stabilization e¤ects of all variants of the EMU-UI system presented throughout this paper. is computed using arithmetic changes ( ) in bene…t and contribution payments as well as changes in employment income from year t to t+1 P which are aggregated across individuals i in each member state. Note that changes in employment income as well as in contribution and bene…t payments are calculated for employment changes along the extensive margin only in order to isolate the stabilizing e¤ect in the event of unemployment shocks from (intensive margin) income shocks. The income stabilization coe¢ cient for euro area unemployment insurance bene…ts is positive in a given member state if total bene…t payments in year t + 1 are higher than in year t and the total change in employment income following entries into employment/unemployment is negative, and zero otherwise: Accordingly, the income stabilization coe¢ cient for contribution payments is positive if total contributions in year t + 1 are lower than in year t and the total change in 20 Automatic stabilization might not only have e¤ects on disposable income and consumption but also on GDP itself (cf. Fatás and Mihov 2001). If fewer taxes are collected and more transfers are paid in a recession, this should support private incomes and dampen adverse movements in aggregate demand. employment income is negative, and zero otherwise: The individual components of can be a summed up to the total income stabilization coe¢ cient, I : Income stabilization coe¢ cients for the so-called GIIPS countries (Greece, Ireland, Italy, Portugal and Spain), which were hardest hit during the recent crisis period, are shown in Figures 7 and 8. Tables 7 and 8 in the Appendix provide the full set of results. Note that we focus on the total income stabilization coe¢ cient which is mainly driven by increased bene…t payments and only to a small extent by lower contribution payments. In fact, SIC is equal to the contribution rate of the scheme (see last row of Table 1). In the baseline scenario, the euro area unemployment bene…t scheme would have absorbed a considerable fraction of the overall unemployment shock in 2009 both at the euro area level (36 per cent) as well as in individual member states. The fact that all member states would have been stabilized in 2009 can be explained by the capacity of the scheme to build up de…cits in years with rising (short-time) unemployment. 21 For the GIIPS countries, we …nd income stabilization coe¢ cients in a range between 23-31 per cent at the beginning of the crisis, but lower stabilization e¤ects in the following years which is due to rising long-term unemployment (and hence lower coverage rates) in the more recent years of the crisis. 22 In line with our results presented in section 4.2, the euro area unemployment bene…t scheme is less e¤ective in stabilizing disposable incomes the lower the coverage rates and the less generous the scheme is. These …ndings indicate that there is a trade-o¤ between the amount of redistribution (ex-post) across member states on the one hand and the insurance and stabilization e¤ects on the other hand as both are positively correlated. Hence, it is interesting to compare the stabilizing e¤ects of a basic EMU-UI system covering only short-term unemployment (for a maximum period of one year, i.e., 52 weeks) to the extended and emergency unemployment bene…ts provided by the US federal government (with bene…ts extended from 26 up to 99 weeks in many US states in the 2009-12 period). Given the longer duration of UI payments in the US, one can conclude that a basic EMU-UI system (without further extensions) would only be an e¤ective automatic stabilizer in short recessions. In prolonged recessions with more and more long-term unemployed individuals, the coverage rates of the EMU-UI scheme decline (see Figure 1) and a program such as the federal extensions in the US unemployment insurance system would be more e¤ective as an automatic stabilizer in such a setting.  09 10 11 12 13 09 10 11 12 13 09 10 11 12 13 09 10 11 12 13 09 10 11 12 13 Baseline Waiting period Nat. coverage of new unempl.
Note: Total income stabiliziation coe¢ cient I for variants A (baseline), B (waiting period) and C (coverage of national unemployment insurance systems). Sources: EUROSTAT and own calculations based on EUROMOD.  5 Alternative scenarios

Experience rating and claw-back
The baseline scheme is calibrated to be revenue-neutral at the EMU-level over the simulation period with a uniform contribution rate (1.57 per cent) across member states. The analysis in the previous section has shown that a uniform contribution rate would have led to permanent transfers in the euro area with 5 member states of the EA18 being either net contributor (Austria and the Netherlands) or net recipient (France, Latvia and Spain) in each year of the 2000-13 period. Therefore, an interesting analytical exercise is to calculate country-speci…c contribution rates that balance the budget in each member state either over the whole simulation period or in each year (experience rating). The former are shown in Table 1 for the di¤erent speci…cations of the euro area unemployment insurance scheme presented throughout this paper. The last row of Table 1 shows the uniform contribution rates that where estimated in the previous section and balance the budget at the euro area, but not the member-state level. Given the large di¤erences in net contributions across member states presented in the previous section, it is not surprising that country-speci…c contribution rates di¤er signi…cantly ranging from 0.75 per cent in the Netherlands to 3.3 per cent in Spain for the baseline scenario. Less generous schemes (columns B-F in Table 1) require lower contribution rates for revenue-neutrality.  Figure 9 presents average country-speci…c contribution rates that balance national budgets in each year as well as maximum and minimum contribution rates over the period. In Austria and the Netherlands, the two member states that would have been permanent net contributors, revenue-neutral contribution rates would have always been below the uniform (Eurozone-wide) contribution rate of 1.57 per cent (dashed horizontal line), while the opposite is true for France, Latvia and Spain, the three net recipients throughout the simulation period in the baseline scenario (variant A). However, an important implementation problem with experience rating is that in practice, the revenue-neutral contribution rate is not known ex-ante (neither for the euro area as a whole nor for its member states). A potential way to reduce cross-country transfers could be to determine pre-speci…ed rules, i.e., claw-back mechanisms, that adjust (country-speci…c) contribution rates automatically after certain time intervals based on previous balances. In the US, for instance, each state places its unemployment insurance payroll taxes in a trust fund with the Treasury and state-speci…c tax rates are raised if trust funds become insolvent (Vroman and Woodbury 2014). 23 We consider two stylized types of claw-back mechanisms and examine their e¤ect on each member state's net contributions over time. Under the …rst claw-back mechanism (Claw-back A), member-state speci…c contribution rates are applied in the current period that would have balanced the (national) budget in the previous period. Under the second claw-back mechanism (Claw-back B), past accumulated de…cits or surpluses are taken into account as follows. In the initial period, we apply the Eurozone-wide revenueneutral contribution rate (1.57 per cent) which leads to an unbalanced budget at the member-state level. In the subsequent periods, country-speci…c contribution rates are applied that would have reduced the net balance of the previous period by 50 per cent. For both claw-back mechanisms, contribution rates are adjusted in 3-year intervals (2000-02, 2003-05, 2006-08, 2009-11, 2012-13).
Cumulative net contributions under Claw-back A and B are presented in Figure 10, together with net contributions that would have accrued under uniform and country-speci…c contribution rates (column A of Table 1). The latter two cases can be interpreted as benchmark scenarios for our claw-back mechanisms as they are based on contribution rates which ex-post guarantee revenue neutrality at the euro area and member-state level. Figure 10 illustrates that ex-post adjustments of contribution rates reduce accumulated de…cits or surpluses, relative to the counterfactual of a uniform contribution rate, in some, but not all member states. In France, Germany, Greece, Latvia, Malta, and Spain, this holds for both claw-back mechanisms and in Austria, Belgium, Finland, Luxembourg, and the Netherlands for Claw-back B. In a few cases, the accumulated net balance at the end of the simulation period is ampli…ed (Cyprus, Ireland, Portugal). In other member states, claw-back mechanisms result in a net contributor (recipient) becoming a net recipient (contributor), namely Austria, Belgium, Finland, Luxembourg and the Netherlands under Claw-back A and Italy, Slovenia and Slovakia under Claw-back A and B. Note that we …nd similar results with other claw-back mechanisms based on shorter or longer adjustment periods. 24 These simulations show that claw-back mechanisms could be useful tools to reduce cross-country transfers, but that this does not hold in a systematic way in all member states. The trade-o¤ between the degree of cross-country redistribution and automatic stabilization e¤ects would become especially apparent if member states were forced to increase the contribution rate and thus the tax wedge on wages during a recession. Country-specific Claw-back A Claw-back B Note: Uniform and country-speci…c contribution rates as in Table 1,

Contingent transfers
As a further variant we simulate an EMU-UI scheme with contingent bene…ts which are activated once certain triggers are reached and analyze its stabilizing and redistributive properties, in particular if such a scheme reduces cross-country transfers. Our choice of the trigger is guided by the US Extended Bene…t (EB) program which permits states to use either the insured or the total unemployment rate to qualify for extended unemployment bene…ts (Nicholson et al. 2014). We choose the total unemployment rate as a trigger so that activation of contingent transfers is independent from eligibility conditions of national unemployment insurance systems. Precisely, bene…ts from the EMU-UI system are triggered if the unemployment rate in year t is at least 1 percentage point higher than the unemployment rate in i) year t 1, ii) years t 1 or t 2, iii) years t 1 or t 2 or t 3. Longer look-back periods ensure that EMU-UI bene…ts can remain activated in sustained periods of high unemployment rates. 25 In all other dimensions (payout rules, uniform contribution rate across member states), the contingent bene…t schemes i-iii are identical to the baseline scheme (variant A) which implies that by construction member states are net contributors in those years when contingent bene…ts are not triggered. Table 9 in the Appendix shows that while with a three-year look-back period, contingent bene…ts would have been triggered in all member states at least once, they would not have been activated in Malta (Belgium and Malta) in any year with a twoyear (one-year) look-back period. The divergent development of unemployment rates since the start of the euro in 1999 becomes evident by a comparison of activation periods in di¤erent member states. While the short-term unemployed in Germany or Luxembourg, for instance, would have been eligible to EMU-UI bene…ts only in the period 2003-05 (and in 2013 in Luxembourg under variant iii), transfers would have been activated in Greece, Ireland, Italy and Spain only from 2008 onwards (with the exception of Greece under variant iii in 2000). Not surprisingly, with average yearly bene…t and contribution payments of 13, 19 and 21 billion euro at the Eurozone-level, the overall budget of the contingent bene…t schemes i-iii would have been signi…cantly lower than in our baseline scenario with non-contingent bene…ts (49 billion per year). Consequently, revenue-neutral contribution rates would have been less than half as large as in the baseline (0.41, 0.61 and 0.68 rather than 1.57 per cent on employment income). Figure 11 compares cumulative net contributions under the contingent bene…t schemes to the baseline variant. A key …nding is that the redistributive e¤ects of the contingent bene…t schemes di¤er to the baseline in several instances as evidenced by the fact that a few member states change their net contributing position in terms of accumulated net contributions at the end of the simulation period (Finland, France, Italy, Slovenia). Austria and the Netherlands, the two member states which would have been net contributors in each year in the baseline, are now net receivers in some years. In the Netherlands, accumulated net contributions are reduced by more than 50 per cent by the end of the simulation period relative to the baseline. Spain, a net recipient in the baseline throughout the simulation period, is a net contributor until 2007 and a net recipient in the remaining years. These results show that an EMU-UI system with contingent bene…ts could indeed provide more targeted transfers to member states which see their labor market conditions signi…cantly deteriorating.
What are the automatic stabilization e¤ects of such a scheme? Given that the contingent bene…t schemes considered here correspond to the non-contingent baseline scheme in all dimensions besides the activation of the scheme, stabilization e¤ects are similar once EMU-UI bene…ts are triggered. However, it must be taken into account that countries that have not reached the trigger (but might well be in a recession) would be worse o¤ compared to the baseline EMU-UI system as the link between contribution and bene…t payments would be broken. The reason is that households in these member states would need to …nance both their national unemployment insurance system as well as the EMU-UI system because the former would not be (partly) replaced by the latter if the trigger is not reached. This potential destabilizing e¤ect could be prevented by suspending contribution payments to the EMU-UI system under certain circumstances such as rising unemployment rates.

Conclusion
The economic crisis in the Eurozone has revived the debate on deeper …scal integration and has brought this topic to the top of the European policy agenda. A common unemployment insurance system is one key reform proposal which could serve as a …scal risk sharing mechanism in the euro area. Supporters of this idea argue that a centralized EMU-UI system would dampen asymmetric shocks in the Eurozone and provide income insurance to those households which are most vulnerable. It would thus not only improve the economic resilience of EMU and make its institutional architecture more sustainable, but also strengthen the social dimension of European policy-making. However, main concerns include the risk of permanent transfer ‡ows across member states and moral hazard for national governments and administrations, which could lead to adverse labor market e¤ects. The aim of this paper has been to present di¤erent options for the design of a common unemployment insurance system and to assess their redistributive and stabilizing properties. Moreover, we have discussed how di¤erent design options would a¤ect moral hazard issues. In our empirical analysis, we have used counterfactual simulation techniques based on harmonized European micro data to examine the economic e¤ects of a hypothetical common EMU-UI system for the time period 2000-13. Our main results can be summarized as follows. A basic scheme, partly replacing national unemployment insurance systems, with a replacement rate of 50 per cent, a maximum duration of bene…t receipt of 12 months and a broad coverage of all new unemployed with previous employment income could be implemented with a relatively small annual budget. On average, it would have amounted to 49 billion euro per year at the Eurozone-level …nanced by a contribution rate of 1.57 per cent on employment income. The scheme would have provided signi…cant income stabilization at the beginning of the recent economic crisis absorbing 36 per cent of the unemployment shock in 2009 at the Eurozone-level, but due to its focus on short-term unemployment this e¤ect would have diminished, the longer the crisis lasted. We …nd, perhaps surprisingly given that the scheme does not lead to permanent redistribution per se, that 5 out of 18 member states would have been either net contributor or net recipient in each year of our simulation period. Running various sensitivity checks regarding the coverage and generosity of the scheme, we show that there is a trade-o¤ between the degree of cross-country redistribution and desired automatic stabilization e¤ects.
We therefore investigate whether claw-back mechanisms based on experience rating and contingent bene…t schemes lead to smaller accumulated net balances. In a …rst step, we calculate country-speci…c contribution rates that would have balanced national budgets either over the whole period or in every single year. These range from 0.75 per cent in the Netherlands to 3.3 per cent in Spain in the former and from 0.46 per cent in Luxembourg to 5.8 per cent in Latvia in the latter case. In the next step, analyzing two stylized claw-back mechanisms under which country-speci…c contribution rates are adjusted in 3-year intervals based on previous net balances, we …nd that they lead to smaller net contributions in some, but not all member states. Finally, our analysis shows that a common EMU-UI system with contingent bene…ts would lead to less cross-country redistribution as it would provide more targeted transfers to member states with deteriorating labor market conditions. However, claw-back mechanisms and contingent bene…ts can have undesirable side e¤ects such as pro-cyclical adjustments of contribution rates or a broken link between contribution and bene…t payments if bene…ts are not activated.
We should emphasize that our analysis has a number of limitations which should be taken into account in the interpretation of the results. Most importantly, it is not the objective of this paper to establish whether or not the introduction of an EMU-UI scheme is desirable in terms of economic welfare. Our analysis focuses on the …nancial ‡ows implied by di¤erent unemployment insurance schemes and the ability of these ‡ows to act as an automatic stabilizer. In so far our analysis is purely positive, rather than normative. In addition, we take economic behavior as given. If EMU-UI had the desired stabilizing e¤ects, the …nancial ‡ows in the system would di¤er from those calculated here; the redistributive e¤ects would probably be smaller. However, if the moral hazard e¤ects dominated, the …nancial ‡ows from contributors to recipients could also be larger. Adding behavioral e¤ects to the analysis is a promising subject for future research.  Immvervoll, H., Levy, H., Lietz, C., Mantovani, D. and Sutherland, H. (2006). The Notes: Coverage rates in per cent of all unemployed. A: Baseline, all new unemployed with previous employment income covered. B: Waiting period, no bene…ts paid in the …rst 2 months of the unemployment spell. C: National coverage, only the share of short-term unemployed covered by national unemployment insurance systems receives bene…ts. Sources: EUROSTAT and own calculations based on EUROMOD.      Notes: Income stabilization coe¢ cients. A: Baseline, all new unemployed with previous employment income covered. B: Waiting period, no bene…ts paid in the …rst 2 months of the unemployment spell. C: National coverage, only the share of short-term unemployed covered by national unemployment insurance systems receives bene…ts. Source: Own calculations based on EUROMOD.   Table 9: Trigger for contingent bene…ts   2000  2001  2002  2003  2004  2005  2006   i  ii  iii  i  ii  iii  i  ii  iii  i  ii  iii  i  ii  iii  i  ii  iii  i  in a given member state in year t is at least 1 percentage point higher than in t-1 (one-year look-back period).
Contingent scheme ii): 2-year look-back period, i.e., bene…ts are triggered if unemployment rate in year t is at least 1 percentage point higher than in t-1 OR t-2. Contingent scheme iii): 3-year look-back period, i.e., bene…ts are triggered if unemployment rate in year t is at least 1 percentage point higher than in t-1 OR t-2 OR t-3. Source: AMECO.