The Political Economy of Flexicurity by Tito Boeri J. Ignacio Conde-Ruiz Vincenzo Galasso DOCUMENTO DE TRABAJO 2006-15

May 2006

The Political Economy of Flexicurity

Tito Boeri∗ Universit`a Bocconi, IGIER, and Fondazione Rodolfo Debenedetti

J. Ignacio Conde-Ruiz Prime Minister’s Economic Bureau and FEDEA

Vincenzo Galasso IGIER, Universit`a Bocconi and CEPR

February 2006

Abstract

We document the presence of a trade-of between unemployment benefits (UB) and employment protection legislation (EPL) in the provision of insurance against labor market risk. Diferent countries’ locations along this trade-of represent sta ble, hard to modify, politico-economic equilibria. We develop a model in which individuals determine the strictness of EPL and the size of a redistributive UB system in two distinct political environments. Agents are heterogeneous along two dimensions: employment status — insiders and outsiders — and skills — low and high. Unlike previous work on EPL, we model employment protection as an institution redistributing also among insiders, notably in favour of the low-skill workers. A key implication of the model is that flexicurity configurations with low EPL and high UB should emerge in presence of dispersed wage structures and progressive UB systems. Micro data on wage dispersion display correlations consistent with our results. The analysis of the experience of EPL reformers yields results which are in line with the relation between EPL and progressiveness of the UB system implied by our model.

Keywords: employment protection, unemployment insurance, political equi libria. JEL Classification: J68, J65, D72.

∗Tito Boeri, Universit`a Bocconi, IGIER and Fondazione Rodolfo Debenedetti, via Salasco 3-5, Milano, 20136, tito.boeri@unibocconi.it. Jose Ignacio Conde-Ruiz, Oficina Económica del Presidente (Presidencia del Gobierno), Complejo de la Moncloa, 28071 Madrid, jiconde@presidencia.gob.es, Vincenzo Galasso, IGIER, Universit`a Bocconi, via Salasco 5, 20136, Milano, vincenzo.galasso@unibocconi.it. We thank Alberto Alesina, Lars Calmfors, Pietro Garibaldi and Guido Tabellini for useful comments and Domenico Tabasso for skilful research assistance. All remaining errors are ours. The views expressed herein are those of the authors and not necessarily those of the Prime Minister’s Economic Bureau.

1. Introduction

Unemployment benefits (UBs) and employment protection legislation (EPL) protect individuals against the risks of job loss. Yet, while EPL protects those who already have a job, and does not tax the worker, UBs transfer income to the unemployed and are funded by taxes imposed on labor income. Across OECD countries there is a large variation in the use of EPL and UBs. Plotted against each other, various measures of the two institutions point to a trade-of between EPL and UBs: countries adopting strict EPL display smaller UB programs, and viceversa. “Flexicurity” countries adopt instead a combination of flexible labor market and large income security for the unemployed, by choosing low EPL and generous UBs.

Countries’ locations along this trade-of correspond to stable political-economic equilibria. Reforms of EPL are mostly marginal: they are confined to introducing “at the margin” more flexible contractual types, rather than modifying rules for workers who already have a permanent contract. Saving on fiscal cost of flexicurity is also proving very dificult: UB systems are adjusted by modifying enforcement rules (e.g., via activation schemes) rather than statutory replacement rates or the maximum duration of benefits. All this points to strong political opposition to reforms.

Why do diferent countries resort to alternative combinations of employment protection and unemployment insurance to protect individuals against the risk of being unemployed? Why is it so uncommon to reform these institutional configurations?

This paper provides a political-economic explanation of the observed trade-of between EPL and UBs, and of the cross-country variation in the use of the two policy instruments applying for the first time (to our knowledge) a multidimensional voting approach to endogenous labor market institutions theory. Our model bridges the gap between two streams of literature in the political economy of labor markets. On the one hand, our environment is similar to that proposed by Wright (1986) to examine UBs. On the other hand, it draws on Saint-Paul (1996, 1999a, 1999b and 2000) in modeling

choices over EPL.

Unlike Wright and Saint-Paul, we model EPL and UBs as multidimensional institutions operating redistributions not only between insiders to outsiders, but also between high and low-skill types. Hence, we introduce two conflicts of interest in our model. The first conflict arises in the transition between employment and unemployment: unemployment inflow and outflow rates are afected by the strictness of EPL. The second conflict occurs because both EPL and UBs operate — possibly at diferent degrees — some redistribution across skills.

In our political economy model, individuals decide the strictness of EPL and the size of UBs. Because of the multidimensionality of the issue space, the existence of a Condorcet winner of the majority voting game is not guaranteed. To solve this shortcoming, we analyze two alternative political systems. First, in a direct democracy environment, we concentrate on political equilibria induced by institutional restrictions, or structureinduced equilibria (see Shepsle,1979 and Persson and Tabellini, 2000). In this political system, the entire electorate votes simultaneously over the payroll tax financing unemployment benefits (hence over the size of UBs), and over the strictness of EPL; policy decisions are taken issue-by-issue. Although the median voter over each policy is typically — under reasonable specifications — a low-skill individual, high-skill types can still “vote with their feet”, by adjusting their labor supply in response to the tax financing the UB system. Hence, changes in the distribution of the population by skill level afect the UB/EPL policy mix via economic channels even when the identity of the pivotal political player does not change. We also analyze these redistributive issues in a representative democracy, in which voters choose between two parties or coalitions according to their policy platforms. Each party appeals to its own electorate, and within party decisions over the economic policy require unanimity, according to the party unanimity Nash equilibrium defined by Roemer (1999).

We show that flexicurity configurations with relatively low EPL and high UB emerge when wage diferentials between high and low-skill types are suficiently large or when unemployment benefits are strongly progressive, so that large redistributive transfers may take place via a generous UB system. Our model thus implies that countries with compressed wage structures have stricter EPL and that configurations with less EPL should be associated with stronger progressiveness in the design of UBs. The latter implication may represent a “possibility theorem” for Governments wishing to adopt flexicurity configurations: more flexibility may be introduced in the labor market, if UB systems are made more progressive. Also the fiscal costs of flexicurity can be reduced by concentrating UB generosity on the low-skill types.

Our empirical strategy ofers tests of these implications of the model. We document that wage structures allowing for larger premia on education (and also higher unconditional wage dispersion) are associated with flexicurity configurations, i.e., more UBs and less EPL. We also look at the policy experiments carried out in OECD countries in the 1990s and show that countries reforming EPL have also reformed UBs in the direction implied by our model.

The paper proceeds as follows: Section 2 documents the trade-of, characterizes the multidimensional conflicts involved by UBs and EPL and reviews the related literature. Section 3 presents the model and the economic environment. Section 4 develops the political system, and introduces the equilibrium concept. In section 5, we bring the model’s main assumptions and results to the data, and we conclude.

2. The UB/EPL trade-of

The theoretical literature acknowledging a welfare-enhancing role to labor market institutions suggests that UBs may be a close substitute for EPL. Both EPL and UB protect workers against uninsurable labor market risk. When severance payments and notice periods in case of dismissals are chosen optimally to maximize welfare of riskadverse agents, there is no role for unemployment insurance (Pissarides, 2001). These two institutions have also important design features in common. An experience-rated unemployment insurance scheme involves the same type (and possibly amount) of transfers from the employer to the employee than a severance pay or a statutory notice period in the event of a dismissal. The only diference is that EPL would be paid in one installment, while UBs are generally provided throughout the unemployment spell up to their maximum duration. The reform of the French unemployment benefit system recently advocated by Blanchard and Tirole (2003) exploits this substitutability between EPL and UBs: it involves an increase in the degree of experience-rating of the UB system, which confines EPL to a one-of monetary compensation for the job loss.

Figure 1 documents the aggregate trade-of between UB and EPL over OECD countries, for which we had comparable data on both institutional features. It displays, on the horizontal axis, an index of the strictness of employment protection for regular contracts¹ compiled by the OECD (OECD, 1999) on the basis of an assessment of national legislations. The vertical axis displays a summary measure of the generosity of unemployment benefits, namely the coverage of UBs (the fraction of unemployed receiving UBs) multiplied by the average gross replacement rate in the first-year of receipt of benefits.

Both EPL and UB measures are normalized in the 0-1 range. Higher values denote stricter EPL and more generous UBs. Most countries are located in the first and in the third quadrant of Figure 1, pointing to an inverse relationship between UB and EPL. The two relevant exceptions are the US and the UK, which display less UBs and EPL than a typical OECD country. The pairwise correlation of the two institutional features is -.39, which is statistically significant at 95 percent confidence levels².

¹The OECD overall EPL index combines information on “regular” and “temporary” contracts. Less EPL on the latter contracts (fixed-term contracts, temporary work agency, etc.) provides greater “flexibility at the margin” (more leverage in hiring via temporary contracts), which can actually insulate regular workers from labor market risk, by creating a “bufer stock” of flexible workers, who can be dismissed at low-cost in case of adverse shocks (Bentolila and Dolado, 1994; Boeri and Garibaldi, 2006).

²The US is an outlier in this context as it displays low both EPL and UBs by international standards. OECD statistics on UBs however fail to capture many state programmes which are targeted to the poor,

As first time job-seekers typically do not qualify for UBs, a low coverage of UBs may be associated with high youth unemployment rates, which tend to be positively correlated with EPL. However, the negative correlation between UB and EPL is stronger when concentrating on central age groups (Table 1), whose unemployment rate is uncorrelated with EPL (OECD, 1999). The negative correlation between UBs and EPL holds also when measuring UB generosity only in terms of replacement rates ³ (second row of Table 1).

This trade-of holds also at the micro level. Boeri, B¨orsch-Supan and Tabellini (2001) found that individuals, who feel protected by EPL, are less willing to purchase state-provided unemployment insurance; Ichino et al. (2003) found that judges are more favorable to workers (efectively making EPL stricter) when unemployment benefits are lower.

Moving up along the EPL-UB trade-of toward flexicurity configurations is proving extremely dificult for Governments. An inventory of reforms in this field (available at www.frdb.org) suggests that 93 out of 112 reforms of EPL carried out in the period 1986-2002 in Europe have been parametric, involving mainly the introduction of new contractual types “at the margin”, that is, limited to new hires. This is confirmed by the updating of the OECD index of the strictness of employment protection for regular workers: only four countries (Korea, Finland, Spain and Greece) out of 28 reduced the strictness of EPL for regular workers throughout the whole period. It is also proving dificult for countries to reduce the generosity of UB system towards configurations with lower fiscal costs. The location of the diferent countries along the EPL-UB trade-of thus seems to represent a stable politico-economic equilibrium.

such as the TANF (Temoporary Assistance for Needy Families), and for the extension of the maxiumum duration of benefits in states hit by regional shocks. The generosity and the progressiveness of the US UB system would be larger when accounting for these state programmes. In a companion study (Boeri, Conde-Ruiz and Galasso, 2004) we also show, either theoretically and empirically, that larger capital markets can reduce the demand for EPL and UBs, by providing to workers alternative insurance against shocks to their labour incomes.

³Buti, Pench and Sestito (1998) also found a negative pairwise correlation between UB replacement rates and EPL strictness.

2.1. Vertical redistribution operated by UBs and EPL

Political-economic models of EPL and UB (Saint-Paul, 1999a, 1999b and 2000; Wright, 1986) describe a one-dimensional distributional conflict between insiders and outsiders. However, both EPL and UBs may involve multidimensional conflicts. In fact, in addition to redistributing across the employment margins, EPL and UBs redistribute also across earning levels.

EPL unavoidably involves some fixed-cost, such as those associated with procedural obligations of employers vis-a-vis labor inspectors and unions. These fixed costs explain most (up to 85%) of the cross-sectional variation in the aggregate EPL OECD index, compared with roughly 40% explained by the pure transfer and earning-related (statutory severance pay and advance notice) component of EPL. More importantly, it is just these fixed (or progressive) components of EPL which are found to have a stronger impact on unemployment inflows (OECD, 1999; Bertola, Boeri and Cazes, 2000). This is consistent with economic theory: as suggested by Lazear (1990), EPL regulations involving just transfers from employers to employees can be undone by bonding wage contracts, provided that wages are suficiently flexible and workers are risk-neutral. Judges are also typically more protective of unskilled workers with low re-employment probabilities (Ichino et al., 2003; Marinescu, 2005), as if they were maximizing the joint welfare of the dismissed worker and the firm, tailoring firing costs to individual economic circumstances.

These asymmetric efects of EPL across earnings-skill levels are documented in Figure 2, which displays proxy quarterly job loss rates (defined as persons who are currently unemployed and who have been dismissed by their employer in the previous 3 months, as a proportion of dependent employment) by level of education, drawing on longitudinal data from the European Commission Household Panel (ECHP), plotted against the OECD EPL index for regular workers. As shown by the second-degree polynomial fitted into the scatterplot, unemployment inflows are less responsive to changes in the strictness of EPL in the case of the high-skill types.

Net replacement rates (NRR) of UBs are almost everywhere decreasing with earnings, as a result of benefit floors and ceilings or explicit rules reducing the indexation of benefits to wages at higher earning levels. As a measure of progressiveness in the design of UBs, table 2 displays the ratio of the replacement rates at the two earning levels, respectively and 100 per cent of the wage of the average production worker, during the first and third year of unemployment. For most countries the ratio is significantly greater than one pointing to strong progressiveness in the design of UBs, especially three years after job loss. Moreover, UBs also redistribute through the higher incidence and duration of unemployment at low skill levels.

Overall, vertical redistribution operated by EPL and UB is rather substantial.

3. A Political-Economic Model

3.1. The environment

In our economy, agents are infinitely long lived. In every period, they consume their current income, since, as in Wright (1986), we assume that no saving technology is available⁴. Preferences are defined over the infinite stream of consumption, and leisure, 1 l, where l represents labor, through a utility function, 2 with representing the subjective discount factor. The instant utility function is assumed to be log-linear , and is a parameter measuring the relative importance of leisure.

Agents may be of low or high skill type, l or h. The fraction of the type-j workers in the population is indicated by . Clearly, , and we consider that there are more low than high skill types, . Low and high skill agents are assumed to difer along several dimensions. When employed, low skill workers earn a pre-tax real wage equal to , whereas high skill workers earn , with representing the wage diferential; moreover the labor supply of the unskilled is rigid and normalized to . High skilled instead determine their labor supply, , by maximizing their utility when employed, that is, , where τ represents a proportional tax on labor income levied to finance an unemployment benefit system. Labor supply of the high-skill types is then: , which is non negative for

⁴This assumption greatly simplifies the analysis, without afecting our conclusions. The fact that agents are not allowed to self-insure against negative labor market shocks through private savings does not afect the redistributive motive driving our results.

In every period, agents may be either employed (insider) or unemployed (outsider). The transition between the two states is regulated by a Markov process, with skill specific transition probabilities. In particular, ) is the probability that a type-j employed worker becomes unemployed (the unemployment inflow rate); and is the probability that a type-j unemployed worker finds a job (the unemployment outflow rate). Our analysis concentrates on steady states. Thus, for each group of agents the unemployment rate is , while the total unemployment rate is . Clearly, we have that and . Moreover, stability conditions for the unemployment rate require that

3.2. Labor Market Institutions

We consider two types of labor market institutions: i) an unemployment benefit system, which in every period taxes the labor income of the workers and provides a transfer to the unemployed; and ii) an employment protection legislation, which afects unemployment inflow (and outflow) rates.

Unemployment Benefits (τ) Our insurance program imposes a proportional tax, on labor income and awards to any unemployed agent a transfer, . We allow for the UB scheme to entail some redistribution, from high to low skills individuals, and parametrize the degree of redistribution with

For every type of agent, the UB depends on the tax rate, which defines the amount of resources channeled to the UB scheme, on the unemployment rate, on the relative share of each type in the population, on their labor income, and on the degree of redistribution between the two types of unemployed individuals, according to the following expressions:

\[\begin{array}{r c l} b ^ {l} & = & \tau \frac {w ^ {l} l ^ {l} (1 - u ^ {l})}{u ^ {l}} + \tau \phi \frac {l ^ {h} w ^ {h} \rho_ {h} (1 - u ^ {h})}{u ^ {l} \rho_ {l}} \\ b ^ {h} & = & \tau (1 - \phi) \frac {l ^ {h} w ^ {h} (1 - u ^ {h})}{u ^ {h}} \end{array}\tag{3.1}\]

For , the UBs for the two types are completely isolated, and the generosity of each system depends on the skill-specific unemployment rate, as in a pure insurance scheme, which takes into account the diferent probabilities of being unemployed of high and low-skill types; for , some redistribution takes place from high to low skilled individuals.

Finally, the system is assumed to be budget balanced and thus the total amount of transfers to the unemployed equals total contributions:

\[b ^ {l} u ^ {l} \rho^ {l} + b ^ {h} u ^ {h} \rho^ {h} = \tau \left[ l ^ {l} w ^ {l} \rho^ {l} (1 - u ^ {l}) + l ^ {h} w ^ {h} \rho^ {h} (1 - u ^ {h}) \right].\]

Employment Protection Legislation (s) Labor markets may be regulated by norms protecting workers against the risk of job loss. As discussed above, economic theory and empirical evidence suggest that it is mainly red-tape and procedural costs which afect labor market flows⁵. These costs are fixed (hence protect more low-skill workers) and deadweight from the standpoint of the employment relationship (hence cannot be replaced by experience-rated UBs). Accordingly, we model EPL as protect ing only the low-skill workers, and disregard the existence of pure transfers such as mandatory severance payments.

⁵When EPL is confined to severance pay regimes, it can be replicated by any experience-rated unemployment benefit. This makes the UB-EPL substitutability rather uninteresting in that case. In the model below we indeed focus on the red tape component of EPL.

In our model, the degree of EPL is measured by a parameter , where s = 0 means no protection and s = 1 denotes maximum protection. As in Saint-Paul (1996 and 2000), we concentrate on the efects of EPL on unemployment inflow and outflow rates, a relationship on which there is little ambiguity in the empirical⁶ and theoretical literature. Consider the low skill types. A higher degree of EPL decreases the unemployment inflow rate, . Consistently with empirical evidence reviewed in section 2, we assume that this efect is larger when the labor market is flexible (s 0) than under strict EPL⁷, i.e.,

Also the unemployment outflow rate is negatively afected by the strictness of EPL, , in accordance with empirical evidence (OECD, 1999) and with economic theory (e.g., Bentolila and Bertola, 1990) predicting that in rigid labor markets employers hire less workers in upturns in order to reduce costs of dismissals during downturns. Figure 3 summarizes the behavior of the low skill inflow and outflow rates as a function of the strictness of EPL. Notice that a trade-of arises since more EPL decreases the unemployment inflow of low skill types, while reducing their outflow. The overall efect on the unemployment rate is therefore ambiguous, as in standard equilibrium models of the labor market (Mortensen and Pissarides, 2001). Provided that unemployment inflows are negative and convex in EPL, while unemployment outflows are linear (declining) in EPL, we expect unemployment to be decreasing for low levels of employment protection (as the efect on the inflow side dominates) and increasing for larger values of s (as the efect on the outflow side becomes relatively more important).

⁶See OECD (1999 and 2004) and Boeri (1999).

⁷It can be shown (results can be provided upon request by the authors) that Mortensen and Pissarides’ (2001) equilibrium search model also yields a convexity of the reservation productivity (hence unemployment inflows) in EPL, provided that the matching function is specialised as a Cobb-Douglas. This model also implies a negative efect of EPL on unemployment inflows and outflows. In the case of outflows, however, it is not possible to establish a priori the sign of the second derivative.

This assumption, which is standard in literature (see Persson and Tabellini, 2000), is consistent with empirical evidence and delivers an interior minimum at .

Finally, EPL leaves the high-skill types unafected, i.e., and b_are constant8_. Moreover, consistently with a large body of empirical evidence on hazards across employment and unemployment margins, we assume that the unemployment inflow rate is always higher for the low than for the high skill workers⁹, i.e. and that, for any degree of EPL, the unemployment outflow rate of the high skill workers is higher than the outflow rate of the low skill ones

3.3. Individual Preferences

As in Wright (1986) and Pissarides (2001), in our model individuals cannot save to insure against the unemployment risk. Thus, in every period, consumption is entirely determined by the employment status of the individuals: if employed, they consume ; if unemployed, they consume . Denote by the diference in utility between the two labor market stata for the low skill agents, i.e., . Define as the tax rate that makes the low-skill type indiferent between being employed or unemployed. The labor incentive constraint then requires

We can now characterize the indirect utility function with respect to EPL and UB. The expected lifetime utility of a type-j agent who is currently in state i, is given by:

\[V _ {i} ^ {j} (s, \tau) = \frac {\left(1 - \theta_ {i} ^ {j} (s , \tau)\right) ((1 - \tau) w ^ {j} l ^ {j} + \gamma \ln (1 - l ^ {j})) + \theta_ {i} ^ {j} (s , \tau) b ^ {j}}{(1 - \beta)}\tag{3.2}\]

⁸In a companion paper (see Boeri, Conde-Ruiz and Galasso, 2003), we show that our results hold also in an environment in which high skill agents do become unemployed, and EPL afects their inflow and outflow rates, provided that unemployment flows are less responsive for the high-skill types than for the low-skill individuals and that low-income insiders constitute a majority of the voters.

⁹This assumption mainly captures the diference in the job-to-job reallocation between low and high ability types. In fact, high ability types tend to have more job-to-job mobility and a lower unemployment inflow rate than the low-ability types. Additionally, high-ability workers have more firm specific human capital, which reduces incentives of employers to fire them.

where is the discount factor, and

\[\theta_ {I} ^ {j} (s) = \frac {\beta F ^ {j}}{1 - \beta + \beta (F ^ {j} + H ^ {j})} \mathrm{and} \theta_ {O} ^ {j} (s) = \frac {1 - \beta + \beta F ^ {j}}{\beta F ^ {j}} \theta_ {I} ^ {j} (s)\tag{3.3}\]

represent the (discounted) proportion of time that respectively a current insider and outsider type-j will spend unemployed during their lifetime; clearly, Notice that the expected utility of high-skill types, as well as , do not depend on EPL. It is useful at this juncture to define the degree of EPL which minimizes the (discounted) time spent unemployed by a low-skill insider and outsider respectively¹⁰: and . It is easy to see that — where is the degree of EPL which minimizes the unemployment rate — since and take into account the current employment status of the agent. Figure 4 summarizes the behavior of , and with respect to s. Finally, notice that as approaches 1, current employment conditions lose their relevance and the indirect utilities of low skill insiders or outsiders coincide:

4. The Political Environment

The degree of EPL and the level of UBs are determined in the political arena, where the individual preferences — described by the indirect utility functions at equations 3.2 and 3.3 — are aggregated into a policy outcome. We analyze two diferent political environments, in which the policy outcomes are decided once-and-for-all. First, we concentrate on a direct democracy, in which — at every election — agents vote over the income tax¹¹ which finances UBs, , and the strictness of . Due to the multidimensionality of the issue space, a Nash equilibrium typically fails to exist in this simple majority voting game. Hence, we follow Shepsle (1979) in analyzing voting equilibria induced by institutional restrictions, i.e.,

θ_I (s)

u^l (s),

θ_O (s)

F^l (s)

H^l (s)

τ ≤

{TA, TB}

¹⁰Notice that, as for the unemployment rate, the assumptions on and stated in the text are suficient for and to have a minimum, albeit not to be convex.

11The condition that τ ≤ min {τA, τB} guarantees a positive labor supply by the high skill insiders and satisfies the labor incentive constraint for the low skill insiders.

Structure-Induced Equilibria. Second, we consider a case of a representative democracy with electoral competition between two parties or coalitions that appeal to diferent constituencies of electors. Individual voters choose between the two parties according to their multidimensional policy platform, which is defined over the degree of EPL and the level of UBs. In this second environment, we use the concept of Party Unanimity Nash Equilibrium (PUNE), as introduced by Roemer (1999).

4.1. Direct Democracy and Issue-by-issue Voting

In a direct democracy with majority voting, individuals express their preferences directly over the policy issues. Since the issue space is bi-dimensional, (τ, s), and thus a Nash equilibrium typically fails to arise, we impose on the voting game a set of institutional restrictions, which convert a multi-dimensional election into a simultaneous issue-byissue voting game, in which a structure induced equilibrium exists (see Conde-Ruiz and Galasso, 2003 and 2005).

The concept of structure induced equilibrium — or issue-by-issue voting — applied to our political game can be summarized as follows. For every value of the degree of s, each voter determines her most preferred value of the level of UB, τ; analogously, the most preferred level of s is chosen for every given τ. In other words, every agent votes two reaction functions: τ (s) and . A duple is an equilibrium of this voting game if represents the outcome of a majority voting over the jurisdiction the level of unemployment benefit — when the other dimension is fixed at , and likewise for . In the following analysis, we will concentrate on the decision by the low skill insiders, who coincides with the median voter over every single issue¹², provided that the low-skills unemployed are not a majority,

The political decision over the EPL, s (τ ), depends on the trade-of that EPL creates between the low skill insiders’ inflow and outflow rates. An increase in the degree of

¹²A more detailed analysis of the political game is in the appendix.

EPL has two efects on their indirect utility (see eq. 3.2). First, it has an impact on the (discounted) proportion of time that current insiders spend unemployed during their lifetime, . Since the utility is larger when employed, this efect is positive for — where represents the degree of EPL that minimizes — and weakly negative thereafter, see Figure 4. Second, an increase in EPL modifies their future unemployment benefit (see eq. 3.1), through changes in the unemployment rate. This efect is positive for represents the degree of EPL that minimizes the unemployment rate for the low skill types — and weakly negative thereafter. Therefore, a low-skill insider chooses a degree of EPL between and , as she trades of the current positive efect of a decrease in the unemployment inflow rate, with the future negative impact on the average unemployment rate, and thus on the level of UBs, see Figure 4.

How does this degree of , depend on the UB level? It is easy to see that for the relevant range of the tax rate, there exists a negative relation between EPL and UB (see proposition A.1 in the Appendix), since a higher level of unemployment insurance for the low-skills reduces the cost, in terms of consumption, of being unemployed; thus leading a low-skill insider to require a lower degree of EPL. An example of the reaction function of s with respect to is at Figure 5.

In determining the tax rate that finances the UB, for any given level of EPL, τ (s), the low skill insiders trade of the cost represented by the current contribution and the expected future benefit from a UB system that redistributes in their favor. Their most preferred tax rate will thus be increasing in the progressiveness of the UB system and in the wage diferential between high and low skill workers, but decreasing in the strictness of EPL (see proposition A.2 in the Appendix). The intuition is straightforward. More earning inequality or a more progressive UB system involve more cross-skill redistribution operated by the UB system, thereby making unemployment benefits more appealing to low skill types. More EPL — especially, for , i.e., within the relevant range chosen by the low-skill insiders — reduces the low-skill insiders’ probability of being unemployed and hence to cash in the transfer. At the same time as large EPL increases unemployment, it also reduces the level of UBs via the budget constraint. The reaction function of with respect to s (see Figure 5) is thus negatively sloped.

To fully characterize the political equilibria of this issue-by-issue voting game over the strictness of and the UB level, one needs to obtain the duple at the intersection of the two reaction functions of the low skill insider. This is described in the next proposition and characterized geometrically in Figure 5, where the reaction functions, and , are portrayed.

Proposition 4.1. If , there exists a SIE of the voting game , such that and

Labor market policies, composed of EPL and UB, are decided by the low skill insiders. The above Proposition suggests that, if the progressiveness of the UB system and the wage diferential between high and low skill workers are suficiently large, the UB system can successfully be used to redistribute towards the low skilled, and thus these pivotal players will thus endorse flexicurity configurations with a large redistributive UB scheme, but also with some labor market regulations — EPL — in order to reduce the risk of becoming unemployed.

Figure 5 displays this equilibrium outcome. For low earning inequality (small A) and in case of a less progressive UB system (small φ), instead, the low skill insiders will support configurations with a strict EPL.

4.2. Representative Democracy and Party Unanimity Nash Equilibrium (PUNE)

In this section, we consider a diferent political institution to aggregate individual preferences into a policy outcome. In a representative democracy, all voters — low and high skill, insiders and outsiders — choose between two parties or coalitions. Every individual determines her electoral decision according to the policy platform presented by each party. The policy space is , and voters have an exogenous probability of turning out to elections: for O and

We call the two parties: Right (R) and Left (L). Each party appeals to a constituency: the left party seeks the support of low skill insiders and outsiders, while the right party seeks the support of all high skill agents and of low skill insiders. Parties are assumed to be uncertain about the distribution of voter types, so that, if the Left and Right parties propose policies and respectively, the Left wins by majority vote¹³ with probability, . Under the pair of proposals, , the expected utility of the Left constituency is

\[\begin{array}{r c l} \Pi^ {L} (y _ {L}, y _ {R}) & = & \pi (y _ {L}, y _ {R}) [ \rho^ {l} (1 - u ^ {l}) V _ {I} ^ {l} (y _ {L}) + \rho^ {l} u ^ {l} V _ {O} ^ {l} (y _ {L}) ] + \\ & & (1 - \pi (y _ {L}, y _ {R})) [ \rho^ {l} (1 - u ^ {l}) V _ {I} ^ {l} (y _ {H}) + \rho^ {l} u ^ {l} V _ {O} ^ {l} (y _ {H}) ] \end{array}\]

while the expected utility of the Right constituency is

\[\begin{array}{r l r} \Pi^ {R} (y _ {L}, y _ {R}) & = & \pi (y _ {L}, y _ {R}) [ \rho^ {l} (1 - u ^ {l}) V _ {I} ^ {l} (y _ {L}) + \rho^ {h} (1 - u ^ {h}) V _ {I} ^ {h} (y _ {L}) + \rho^ {h} u ^ {h} V _ {O} ^ {h} (y _ {L}) ] + \\ & & (1 - \pi (y _ {L}, y _ {R})) [ \rho^ {l} (1 - u ^ {l}) V _ {I} ^ {l} (y _ {H}) + \rho^ {h} (1 - u ^ {h}) V _ {I} ^ {h} (y _ {H}) + \rho^ {h} u ^ {h} V _ {O} ^ {h} (y _ {H}) ] \end{array}\]

Every party’s partizan ideology corresponds to the bliss point of the party’s militants, who are assumed to be low skill outsiders in the Left party and high skill insiders in the Right party. Each party is composed of three groups of actors: i) militants, who want the party to adhere as closely as possible to its principles, i.e., to its partizan ideology; ii) opportunists, who only care about winning the elections; and iii) reformists who wish to maximize the expected utility of the party’s constituency. Given a proposal by the opposition, in every party the final decision about the policy requires inner-party unanimity, which coincides with unanimity between opportunists and militants, since the agreement of the reformists would automatically follow. More precisely, the Left party would deviate form to only if

(y_L, y_R),

L,

¹³Given a pair of policies if all voters are indiferent between the two parties, we assume that high skill types vote for party R , low skill outsiders vote for party and low skill insiders split their votes equally between the two parties.

\[(\pi (y _ {L} ^ {\prime}, y _ {R}), V _ {o} ^ {l} (y _ {L} ^ {\prime})) \geq (\pi (y _ {L}, y _ {R}), V _ {o} ^ {l} (y _ {L}))\tag{4.1}\]

and similarly, the Right party would switch from to only if

\[(1 - \pi (y _ {L}, y _ {R} ^ {\prime}), V _ {I} ^ {h} (y _ {H} ^ {\prime})) \geq (1 - \pi (y _ {L}, y _ {R}), V _ {I} ^ {h} (y _ {H}))\tag{4.2}\]

which allows us to define a party unanimity Nash equilibrium as follows.

Definition 4.2. A policy pair is a party unanimity Nash equilibrium (PUNE) if and only if there is no at which condition 4.1 holds and there is no at which condition 4.2 holds.

We can now state the following proposition, according to which the policy outcome associated with the issue-by-issue voting game in the previous political environment is also a party unanimity Nash equilibrium of the game described in this section.

Proposition 4.3. is a PUNE equilibrium, where and

A formal proof is in the appendix, but the intuition of this result is straightforward. The policy platform chosen by both parties targets the low skill insiders — which also in the previous political game coincided with the pivotal (median) voters (see section 4.1). A deviation in the policy platform by a party towards more extreme positions, such as its partizan ideology — given the other party’s platform — would be welcome by the militants, but be opposed by the opportunists, since it would reduce the party’s probability of winning the election.

4.3. The Trade-of

In both political environments, low-skill insiders represent the pivotal political players targeted by the policies. The incentives faced by these low-skills insiders — and thus the resulting labor market policies — may be afected by some crucial characteristics of the economy. More specifically, the next proposition analyzes how a change in the degree of progressiveness of the UB system, and in the wage diferential between high and low-skills workers, A, (recall that modifies the outcome of our politico-economic equilibrium in both political games.

Proposition 4.4. An increase in the wage diferential between high and low-skills workers, A, or an increase in the progressiveness of the UB system, , induce a change in an equilibrium outcome from to with and

This proposition contains the crucial theoretical result of the paper and provides a political economic explanation for the observed trade-of between EPL and UB. In countries with large wage diferentials and progressive UB systems, the low skill insiders, which typically constitute the pivotal political players, favor flexicurity configurations with large UB schemes, in order to appropriate more resources — when unemployed — from the high skill individuals, and low EPL. In the jargon of the issue-by-issue voting described in section 4.1, this involves an upward shift in the reaction function τ (s), as displayed in Figure 6. Because of the larger UB transfer, due to the wider wage diferential or to a more progressive design, the diference in utility between the good state — employment — and the bad state — unemployment — shrinks, and hence low skill insiders vote for less EPL. The reaction function s (τ ) shifts downwards, and a new equilibrium¹⁴ with more UB and less EPL is reached, as displayed at point B in Figure 6.

¹⁴Notice that also an increase in the discount factor tilts the equilibrium towards more UB and less EPL, since it reduces the relevance to the insiders of their current employment status, thereby inducing them to accept less EPL in exchange for more UB (for a formal treatment of this aspect, see the companion paper, Boeri, Conde-Ruiz and Galasso, 2003).

5. Empirical Relevance

Our model implies i) a negative cross-sectional correlation between the progressiveness of UBs and the strictness of EPL, and ii) a positive cross-sectional correlation between generosity of UBs and wage dispersion¹⁵.

Figure 7 plots the OECD index of strictness of EPL for regular workers and the measure of the progressiveness of UB systems discussed in section 2. We combine timeseries and cross-sectional variation for the ECHP countries to maximize observations (104). The correlation coeficient is -.78 and the t-statistics is 12.6, which is significant at 99 per cent. Consistently with the predictions of our model, EPL for regular workers is lower in countries where UB systems are more progressive.

Another key implication of our model is that countries with less compressed wage structures should exhibit institutional configurations with more UB and less EPL (flexi curity). Clearly, wage dispersion may be afected by UBs as well as by other institutions (such as minimum wages) providing floors to wage setting. In order to control for these efects, we measure wage dispersion as the Gini coeficient over the distribution of predicted wages out of a standard Mincer-type earning equation corrected for self-selection. In particular, we predict wages based on the following equation:

\[\log (w _ {i}) = \alpha + \beta E D U _ {i} + \gamma_ {1} T E N _ {i} + \gamma_ {2} T E N _ {i} ^ {2} + \lambda_ {i} + \varepsilon_ {i}\]

where EDU represents years of education and TEN denotes tenure in the current job, while λ is a Heckman correction term. As can be seen from Figure 8, countries with more UB and less EPL (the ratio between the two measures is displayed in the vertical axis) exhibit more dispersed earning structures over observables (and controlling for self-selection). In this case, the correlation coeficient is .34 and the t-statistics is 3.6.

¹⁵A link between wage compression and EPL, yet not UB, was also in Bertola and Rogerson (1997), who viewed them as complementary policies.

5.1. Lessons from the Reformers

The above pairwise correlations are not informative as to the order of causality of the correlation between countries’ locations along the UB/EPL tradeof and earning dispersion. Better insights in this respect may come by contrasting the experience of the few countries that reduced EPL for regular workers in the period covered by data with that of countries with similar initial institutional configurations that did not reform their labor market. As recalled in section 2, the only major reformers were Korea, Finland, Spain and Greece. The radical reforms of EPL that occurred in Finland and Spain were split into a number of milder liberalization measures. In particular, in Finland, there were three waves of reforms: in 1991, 1996 and 2001, while in Spain reductions of EPL were enacted in 1994 and 1997. In Korea, reforms were carried out in 1998. In Greece, EPL was reformed in the late 1980s (a period not covered by ECHP data) and was much milder than in Finland and Spain. Hence, we concentrate on Finland, Spain and Korea.

Table 3 compares the experience of Spain with that of Greece, the experience of Finland with that of Denmark, and the experience of Korea with that of Japan. These “matches” are chosen by drawing on a taxonomy of labour market and social policy institutions in the EU (Bertola et al., 2001), pulling together, on the one hand, Nordic and, on the other hand, Southern European countries, and taking Japan as the closest match to Korea in the 1990s. Importantly, the changes to the dispersion fo earnings before the reforms are likely to be associated to exogenous factors, such as a spread of ICT technologies (e.g., the percentage of households PC users increased by 34 base points in Spain compared with only 15 per cent in Greece) or women participation in the labour market (2.7 for Finland compared with only 0.8 for Denmark) . In the period covered by data, there was a spread of ICT technologies and an increase in female participation to the labour market. Both factors are likely to induce exogenous changes in earning distributions.

As shown in Table 3, Spain reduced EPL after that in a period when its earning distribution had became more dispersed: with an increase of the Gini coeficient by 6 base points. The reduction in EPL was also associated with an increase in generosity and progressiveness of UBs, while the opposite happened in Greece. Moving to another match, Finland reduced EPL for regular workers without having experienced an increase in earning dispersion, but — unlike Denmark — it adjusted the design of its UB system by making it more progressive. Turning to the last match, table 3 shows that also in Korea EPL reduction was associated with an increase in earning dispersion, unlike developments in Japan. UB adjustments were also made in Korea in line with the predictions of our model.

Overall, there is some indication that countries reducing EPL for regular workers experienced an increase in earning inequality or in the progressiveness of UBs compared with countries with initial similar institutional configurations, which did not reform EPL for regular workers.

5.2. Conclusions

OECD countries provide diferent types of insurance to workers against labor market risks, by combining diferent degrees of employment protection and unemployment insurance. A heated debate has taken place over the need to reform some of the existing labor market institutions, and some form of consensus has emerged even among academics and international organizations that countries should adopt more “mobility-friendly” institutional configurations assigning a greater weight to UB and less importance to EPL in protecting workers against labor market risk. However, reforming institutions along these lines has proved dificult and politically costly.

Unlike previous literature, this paper characterizes EPL and UBs as schemes redistributing not only between insiders and outsiders, but also across skill groups. Our theoretical model suggests that “flexicurity” configurations, characterized by less EPL and more UB, should emerge in countries with less compressed wage structures. Empirical findings are in line with this theoretical implication. Moreover, we document that the few countries reforming EPL for regular contracts also featured increasing progressiveness of UBs and/or widening earning diferentials.

Our analysis may inspire a political feasibility theorem: reforms of employment protection need to trade labor market flexibility with state-provided unemployment insurance which redistributes in favor of the low-skill segments of the workforce. The trade-of is likely to become steeper when there is a lower degree of redistribution across skill groups embedded in the design of UBs. This implication of our model difers from the prescriptions in Blanchard and Tirole (2003) that configurations with more UB and less EPL could be obtained by introducing experience-rating in the UBs, that is, internalizing the costs of dismissals to the employers “responsible” for the redundancies. This paper implies that political flexibility of reforms does not require that UBs mimic EPL. UBs can still pool risk across employers, but either wage diferentials grow larger or UBs should become more progressive in order to win consensus to reforms.

References

1. [1] Acemoglu, D. and R. Shimer (2000) ”Productivity Gains from Unemployment Insurance”, European Economics Review, 44, 1195-1224.
2. [2] Acemoglu, D., P. Aghion and G. Violante (2001) “Deunionization, Technical Change and Inequality”, Carnegie-Rochester Conference Series on Public Policy.
3. [3] Bentolila, S. and G. Bertola (1990), “Firing cost and labor demand: how bad is eurosclerosis?” Review of Economic Studies 57: 381-402.
4. [4] Bentolila, S. and Dolado, J. (1994): Labour Flexibility and Wages: Lessons from Spain, Economic Policy, n.18, 55-99.
5. [5] Bertola, G. (1990): Job Security Employment and Wages, European Economic Review, 851-866.
6. [6] Bertola, G., Boeri, T. and Cazes, S. (2000): Employment Protection in Industrialized Countries: the Case for New Indicators, International Labour Review, n.1.
7. [7] Bertola, G. and Boeri, T. (2002) EMU Labor Markets Two Years On: Microeconomics Tensions and Institutional Evolution, in Buti, M. and Sapir, A. (eds.) EMU and Economic Policy in Europe, Edward Elgar.
8. [8] Bertola, G. and Rogerson, R. (1997): Institutions and Labour Reallocation, European Economic Review, 41, 1147-71.
9. [9] Blanchard, O. and Tirole, J. (2003) Contours of Employment Protection Reform, MIT Dept of Economics Working Paper Series, n. 03-35.
10. [10] Boeri, T. (1999) Enforcement of Employment Security Regulations, On-the-job Search and Unemployment Duration, European Economic Review, 43, 65-89.
11. [11] Boeri, T., A. B¨orsch-Supan and G. Tabellini (2001),“Would you like to shrink the welfare state? A survey of European citizens” Economic Policy April, 9-50.
12. [12] Boeri, T., Conde-Ruiz J.I. and V. Galasso (2003) “Protecting against Labor Market Risk: Firing Costs or Unemployment Insurance?”, CEPR Discussion Paper N. 3990.
13. [13] Boeri, T. and Macis, M. (2005)
14. [14] Boeri, T. and P. Garibaldi (2006) ”Two Tier Reforms of Employment Protection: a Honeymoon Efect?”, mimeo.
15. [15] Buti, M. , L. Pench ad P. Sestito (1998), “European Unemployment: contending theories and institutional complexities” Policy Paper 98/1, The Robert Schuman Centre, European University Institute.
16. [16] Cazes, S., T. Boeri and . Bertola (1999), “Employment Protection and Labor Market Adjustment in OECD Countries: Evolving Institutions and Variable Enforcement” ILO Employment and Training Papers 48.
17. [17] Clark, A. and Postel-Vinay, E. (2004) Job Security and Job Protection, mimeo, Paris.
18. [18] Conde-Ruiz, J.I. and V. Galasso (2005), “Positive Arithmetic of the Welfare State”, Journal of Public Economics.
19. [19] Conde-Ruiz J.I. and V. Galasso (2003) “Early Retirement” Review of Economic Dynamics 6: 12-36.
20. [20] Foerster, M. and M.Pellizzari (2000) ”Trends and Driving Factors in Income Distribution and Poverty in the OECD area”, OECD Labour Markets and Social Policy Occasional Papers, n.42.
21. [21] Ichino, A., M.Polo and E.Rettore (2003) Are Judges Biased by Labor Market Conditions?, European Economic Review, 47(5), 913-944.
22. [22] Lazear, E. (1990), “Job Security Provisions and Employment”, Quarterly Journal of Economics 105: 699-726.
23. [23] Ljungqvist, L. and T. Sargent (2002) ”The European Employment Experience,” CEPR Discussion paper 3543 September.
24. [24] Marinescu, I. (2005) Are Judges Sensitive to Economic Conditions? Evidence from UK Employment Tribunals, mimeo Ioana, LSE, Harvard, and NBER
25. [25] Mortensen, D. and Pissarides, C. (2001) Unemployment Responses to ’Skill-Biased’ Shocks: The Role of Labor Market Policy, Economic Journal, 109 (242-265).
26. [26] OECD (1999 and 2004) Employment Outlook, OECD, Paris.
27. [27] Persson, T. and G. Tabellini (2000) Political Economics. Explaining Economic Policy. The MIT Press, Cambridge, Massachusetts, London, England.
28. [28] Pissarides, C. (2001) Employment Protection, Labor Economics, 8, 131-59.
29. [29] Roemer, J. E. (1999) The Democratic Political Economy of Progressive Taxation. Econometrica, 67: 1-19.
30. [30] Saint-Paul, G. (1996), “Exploring the Political Economy of Labor Market Institutions” Economic Policy 23: 265-315.
31. [31] Saint-Paul, G. (1999a), “ The Political Economy of Employment Protection”, CEPR Discussion Paper, No. 2109
32. [32] Saint-Paul, G. (1999b), “ Assessing the Political viability of Labor Market Reform: the case of Employment Protection”, CEPR Discussion Paper, No. 2136
33. [33] Saint-Paul, G. (2000), “The Political Economy of Labor Market Institutions” Oxford University Press.
34. [34] Shepsle, K.A. (1979), “Institutional Arrangements and Equilibrium in Multidimensional Voting Models”, American Journal of Political Science, 23(1)
35. [35] Wright, R (1996), “The Redistributive roles of Unemployment Insurance and the Dynamics of Voting”, Journal of Public Economics 31: 377-399.

A. Appendix

Proposition A.1. For , the degree ofEPL chosen by the median voter, displays the following features: i) is convex in τ and decreasing for , where s.t and iii) is decreasing in A and in

Proof of Proposition A.1: For , the median voter over s is a low ability insider, since high skill agents are unafected by s and thus may be assumed to abstain from voting on this issue. Hence, for given the most preferred level of 2 , is obtained by maximizing eq. 3.2 with respect to s. Denote and

i) To see that , we equates to zero the first order condition resulting from this maximization problem:

\[\underbrace {\frac {\beta^ {2} (F ^ {l} H _ {s} ^ {l} - F _ {s} ^ {l} H ^ {l}) \Delta^ {l}}{(1 - \beta + \beta (F ^ {l} + H ^ {l})) ^ {2}}} _ {I} - \underbrace {\frac {\beta (1 - \beta) F _ {s} ^ {l} \Delta^ {l}}{(1 - \beta + \beta (F ^ {l} + H ^ {l})) ^ {2}}} _ {I I} + \underbrace {\theta_ {I} ^ {l} \frac {\partial b ^ {l}}{\partial s}} _ {I I I}\tag{A.1}\]

where

\[\frac {\partial b ^ {l}}{\partial s} = - \frac {F _ {s} ^ {l} H ^ {l} - F ^ {l} H _ {s} ^ {l}}{(F ^ {l} + H ^ {l}) ^ {2}} \left(\frac {b ^ {l}}{u ^ {l}} + \frac {\tau w ^ {l} \overline {{l}}}{u ^ {l}}\right)\]

If we evaluate this FOC in , the first and the third terms, i.e., I and , are equal to zero, while the second term, and thus the entire FOC, is positive, since Therefore, . On the other hand, if we evaluate this FOC in , the first two terms, i.e., I and , are equal to zero, since for while the third term, and thus the entire FOC, is negative, since 0 for . Finally, simple algebra shows that the second order condition of this maximization problem evaluated at is negative, so that is a maximum.

ii) To prove that is convex in τ and decreasing for , where , we apply the implicit function theorem to the FOC at . Since

, we have that the sign of is equal to the sign of Notice that the FOC at eq.A.1 can be written as

\[- \Delta^ {l} \frac {\partial \theta_ {I} ^ {l} (s)}{\partial s} + \theta_ {I} ^ {l} (s) \frac {\partial b ^ {l}}{\partial s} = 0\tag{A.2}\]

and its derivative as

\[\frac {\partial F O C (s _ {I} ^ {l} (\tau))}{\partial \tau} = - \frac {\partial \theta_ {I} ^ {l} (s)}{\partial s} \frac {\partial \Delta^ {l}}{\partial \tau} + \theta_ {I} ^ {l} (s) \frac {\partial^ {2} b ^ {l}}{\partial s \partial \tau} = \frac {\partial \Delta^ {l}}{\partial \tau} \left(\frac {\theta_ {I} ^ {l} (s)}{u ^ {l}} \frac {\partial u ^ {l}}{\partial s} - \frac {\partial \theta_ {I} ^ {l} (s)}{\partial s}\right)\]

since . Recall that and for , so that the sign of is equal to the sign of

Since

\[\frac {\partial \Delta^ {l}}{\partial \tau} = - \left(w ^ {l} \overline {{l}} + \frac {\partial b ^ {l}}{\partial \tau}\right) = - \frac {1}{\rho^ {l} u ^ {l}} \left\{\rho^ {l} w ^ {l} \overline {{l}} + \rho^ {h} \phi (1 - u ^ {h}) \left(w ^ {h} - \frac {\gamma}{(1 - \tau) ^ {2}}\right) \right\}\]

and is decreasing for , where s.t , and is convex.

iii) To prove that is decreasing in A and in we apply again the implicit function theorem to the FOC at . Since SOC , we have that the sign of is equal to the sign of , and the sign of is equal to the sign of , where

\[\frac {\partial F O C (s _ {I} ^ {l} (\tau))}{\partial A} = - \frac {\partial \theta_ {I} ^ {l} (s)}{\partial s} \frac {\partial \Delta^ {l}}{\partial A} \leq 0 \mathrm{and} \frac {\partial F O C (s _ {I} ^ {l} (\tau))}{\partial \phi} = - \frac {\partial \theta_ {I} ^ {l} (s)}{\partial s} \frac {\partial \Delta^ {l}}{\partial \phi} \leq 0\]

since

\[\frac {\partial \Delta^ {l}}{\partial A} = - \frac {\tau (1 - u ^ {h}) \phi w ^ {l} \rho^ {h}}{u ^ {l} \rho^ {l}} < 0, \frac {\partial \Delta^ {l}}{\partial \phi} = - \frac {\tau (1 - u ^ {h}) l ^ {h} w ^ {h} \rho^ {h}}{u ^ {l} \rho^ {l}} < 0\tag{A.3}\]

\[\mathrm{and} \frac {\partial \theta_ {I} ^ {l} (s)}{\partial s} \leq 0 \mathrm{for} \widehat {s} ^ {l} \leq s _ {I} ^ {l} \leq \widetilde {s} _ {I}.\tag{A.4}\]

q.e.d.

Proposition A.2. For , the UB level chosen by the median voter, displays the following features: i) , (ii) is decreasing in s and is increasing in A and in .

Proof of Proposition A.2: Consider the decision of agent (insider or outsider) over for given s. For and , the maximization of 3.2 with respect to yields the following first order condition, where the first term represents the marginal utility cost from paying higher taxes, while the second term denotes the marginal utility benefit from the higher UB, weighted by the discounted probability of being employed or unemployed:

\[- \left(1 - \theta_ {i} ^ {j}\right) w ^ {j} l ^ {j} + \theta_ {i} ^ {j} \frac {\partial b ^ {j}}{\partial \tau} = 0\tag{A.5}\]

with

\[\frac {\partial b ^ {l}}{\partial \tau} = \frac {1}{u ^ {l}} \left(w ^ {l} \bar {l} (1 - u ^ {l}) + \frac {\rho^ {h} w ^ {h} \phi (1 - u ^ {h})}{\rho^ {l}} - \frac {\gamma \rho^ {h} \phi (1 - u ^ {h})}{\rho^ {l} (1 - \tau) ^ {2}}\right)\]

and

\[\frac {\partial b ^ {h}}{\partial \tau} = \frac {(1 - \phi) (1 - u ^ {h})}{u ^ {h}} \left(w ^ {h} - \frac {\gamma}{(1 - \tau) ^ {2}}\right).\]

Clearly, current outsiders will choose a higher tax rate than current insiders, for , since . Moreover, it is easy to see that high ability insiders will prefer a zero tax rate. In fact, eq. A.5 can be written as

\[- \frac {1 - \theta_ {I} ^ {h}}{1 - u ^ {h}} \left(w ^ {h} - \frac {\gamma}{1 - \tau}\right) + \frac {\theta_ {I} ^ {h}}{u ^ {h}} \left(w ^ {h} - \frac {\gamma}{(1 - \tau) ^ {2}}\right) < 0\]

since . Finally, simple algebra shows that the most preferred tax rate by a low ability insider is

\[\tau_ {I} ^ {l} (s) = 1 - \left(\frac {w ^ {h}}{\gamma} - \frac {\rho^ {l} w ^ {l} \overline {{l}}}{\gamma \rho^ {h} \phi (1 - u ^ {h})} \left(\frac {u ^ {l}}{\theta_ {I}} - 1\right)\right) ^ {- 1 / 2}\tag{A.6}\]

which yields for

For , the previous results and imply that the median voter will belong to the low skill insider for any possible ranking: (1) ; and (3) , which proves (i) above since

Finally, using eq. A.6, simple algebra proves (ii) and (iii) above.

Proof of Proposition 4.1: We need to show that the reaction functions and cross — at least — once. Recall that, by Proposition is decreasing in s and by Proposition A.1, is decreasing in . Moreover, recall that for (see the prof of proposition A.1) and min . We thus need to consider three cases: (i) and (ii) and ; and (iii) and

Case (i): is the upper bound on the tax rate for any . By inspecting eq. A.1, it easy to see that the reaction function ranges between for (and hence and for (and hence e Consider the reaction function . Clearly, we have that, for . To show that the two reaction functions cross, we thus need to establish that for s close to the reaction function is above . Since , for , we need to show that . From the expression of at Proposition A.2, it is easy to see that if and only if

Case (ii): eis the upper bound on the tax rate for any . In this case, the reaction function ranges between for and for , since . Consider the reaction function . Clearly, we have that, for . The remaining of the prof follows from case (i).

Case (iii): is lower than the upper bound on the tax rate . In this case, the reaction function has a minimum at and the increases for τ between and or . As in cases (i) and (ii), to show that the reaction functions cross, we need to show that . Recall that is such that whereas is such that . It is easy to see that the last equation (the first order condition of the low skill insiders) evaluated at is negative, which shows that . The remaining of the proof follows case (i) above. q.e.d.

Proof of Proposition 4.3: A) First, we have to proof that if then is a best response over the space of inner-party unanimity proposal set, that is:

\[\begin{array}{r c l} \text {(i)} & \pi (y _ {L} & = y _ {R}, y _ {R}) > \pi (y _ {L} ^ {\prime}, y _ {R}) \forall y _ {L} ^ {\prime} \text {and} \\ \text {(ii)} & V _ {o} ^ {l} (y _ {L} & = y _ {R}) > V _ {o} ^ {l} (y _ {L} ^ {\prime}) \forall y _ {L} ^ {\prime} \end{array}\]

Condition (i) is clearly satisfied. In fact, if party L deviates from , it would lose half of the votes by the low ability insiders, but gain none, since it always has all the votes of the other group in its constituency (the low ability outsiders). Thus, will maximize the number of votes for party and is a best response in the PUNE sense. (Notice that by choosing a proposal the utility of L party militants (the low ability outsiders) may indeed increase; but this proposal would be blocked by the opportunists in the party.

B) The same reasoning can be used to show that, if , then is a best response.

Therefore, is a PUNE. q.e.d.

Proof of Proposition 4.4. To prove this proposition, we show that an increase in A or φ moves the reaction functions and as displayed in Figure 6. (i) An increase of or moves the reaction function to the left, since and , as shown in the proof of proposition A.1. (ii) It is straightforward to see that an increase of A or moves the reaction function upward, by just deriving at Proposition 4.1 w.r.t. A or q.e.d

Figure 1: The UB/EPL Tradeof

Table 1: Alternative measures of the trade-of (late 1990s)

Source: European Community Household Panel (ECHP)

Figure 2: Unemployment Inflows by skill level and EPL

Source: ECHP.

Table2: Progressiveness in the Design of Unemployment Benefit System

Note: Ratio of the replacement rate at 2/3 and 100 at the wage of an average production worker for a single 40-year-old worker who is considered a good approximation to the average situation of an unemployed person.

Table 3

Note: ∆EPL: change in the value of the OECD EPL index for regular workers, over reform period (1994-2001 for Spain and Greece; 1996-2001 for Finland and denmark; 1994-2001 korea and Japan).

∆Gini earnings:change in the Gini coeficient before the reform (1997 for Spain;

Figure 4: EPL: Low-skill Insiders and Outsiders

Figure 5: The UB-EPL Political Equilibrium

Figure 6: Wage Diferential and the EPL-UB Trade-Of

Figure 7: Progressiveness of UB systems and EPL for regular workers

Figure 8: The tradeof and earning dispersion

1996 for Finland; 1998 for Korea).

∆UB Generosity: change in the generosity, the coverage of UBs (the fraction of unemployed receiving UBs) multiplied by the average gross replacement rate in the first-year of receipt of benefits.

∆UB Progressiveness: the ratio of the replacement rates at the two earning levels 100 per cent or 2/3 of the wage of the average production workerratio of the replacement rates at the two earning levels 100 per cent or 2/3 of the wage of the average production worker.