Digitized by the Internet Archive in 2011 with funding from Boston Library Consortium Member Libraries http://www.archive.org/details/optimaldesignofuOOblan / 1- Massachusetts Institute of Technology Department of Economics Working Paper Series THE OPTIMAL DESIGN OF UNEMPLOYMENT INSURANCE AND EMPLOYMENT PROTECTION. A FIRST PASS. Olivier Blanchard Working Paper 04-15 April 1,2004 RoomE52-251 50 Memorial Drive Cambridge, MA 02142 This paper can be downloaded without charge from the Social Science Research Network Paper Collection at http://ssrn.com/abstract=527882 i MASSACHUSETTS INSTITUTE OP TECHNOLOGY The optimal design of unemployment insurance and employment protection. A first pass.* Olivier Blanchardt April 1, and Jean Tirole * 2004 We * are grateful to Daron Acemoglu, Suman Basu, Javier Ortega, John Hassler, and to the participants to the third Toulouse Macroeconomics workshop for helpful comments. t MIT and NBER. IDEI and MIT. t Visiting Scholar, Russell Sage Foundation GREMAQ (UMR 5604 CNRS), Toulouse, CERAS (URA 2036 CNRS), Paris, and Abstract Much of the policy discussion of labor market institutions has been at the margin, with proposals to tighten unemployment benefits, reduce employment protection, and so on. There has been architecture should be. little discussion however of The paper what the ultimate goal and focuses on characterizing this ultimate goal, the optimal architecture of labor market institutions. We start our analysis with a simple benchmark, with risk averse workers, risk neutral firms and random shocks to productivity. In this benchmark, we show that optimality requires both unemployment insurance and employment protection the form of layoff taxes; it also requires that layoff taxes — in be equal to unemployment benefits. We then explore the implications of four broad categories of deviations: limits on insurance, limits on layoff taxes, ex-post firms or workers. The scope We wage bargaining, and heterogeneity show how the architecture must be modified for insurance may be more of in each case. limited than in the benchmark; so may the scope for employment protection. The general principle remains however, namely the need to look at unemployment insurance and employment protection together, rather than in isolation. Introduction Much of the policy discussion of labor market institutions has been at the margin, with proposals to tighten unemployment benefits, reduce employment protection, and so There has been on. architecture should be. provided, how should little What should be the level of insurance, Our What focus in this paper is make payments political economy on characterizing this ultimate goal, the We make match be optimal archi- no attempt to look at transition These are obviously highly relevant to policy makers, issues. start our analysis in Section 1 with a simple averse; entrepreneurs it to the state, or directly to the but they should come second, once the ultimate goal We should should be the size of these payments? tecture of labor market institutions. and how be financed? Should there be some form of employment it protection? If so, should firms workers they lay off? what the ultimate goal and discussion however of run firms and are is clear. benchmark. Workers are risk neutral. Productivity of risk any worker- random. In this benchmark, the proper institutions can achieve both production efficiency and full insurance. This can be achieved in a firm is number of ways. For example, insurance can be provided through the payment of unemployment benefits by an unemployment agency. And efficiency layoff taxes, leading firms to internalize the cost of can be achieved by using unemployment. In this benchmark, efficiency requires that layoff taxes be equal to unemployment benefits. This implies that layoff taxes to use payroll taxes. to unemployment The is —defined as We The presence equal to one. layoff taxes of the first Unemployment so, and there is benefits and employ- —are essential components of the optimal requires the presence of the other. then consider deviations from the benchmark. There ways of doing no need contribution rate, defined as the ratio of layoff taxes benefits, ment protection architecture: fully finance benefits, so there is is potentially a thousand no agreement among labor economists as to which imperfections are most relevant. Thus, our strategy is to explore the implications of four broad categories of deviations: Limits on insurance. Such either to the limits unemployed to search may come from for the need to give incentives another job, or to the employed workers not to shirk on the job. We show in Section 2 that, in this case, it is optimal to partly compensate for the limits on insurance by decreasing the incidence of un- employment, so by decreasing layoffs below the This efficient level. is achieved by choosing layoff taxes higher than unemployment benefits, thus by choosing a contribution rate higher than one. As layoff tax revenues benefits, the state then balances its Under this deviation, optimality calls for higher tighter the limits now exceed unemployment budget through a negative payroll tax. employment protection: The on unemployment benefits, the higher the optimal employment protection, but at an increasing cost in terms of efficiency. Reforms of the un- employment insurance system which allow search effort for higher insurance while thus have both direct insurance effects, maintaining but also indirect ones, as they allow to decrease employment protection and improve efficiency. Limits on layoff taxes. Shallow pockets may make firms to pay layoff taxes. feasible on and optimal to layoff taxes We show, fully insure in Section 3, that, even in this case, now be remains financed partly in turn implies that and this leads to too also consider, in this section, the case of multi-activity firms, and the question firms only partially internalize the cost of unemployment, many layoffs relative to the efficient level. of whether taxing layoffs We it workers against unemployment. But the limits imply that unemployment benefits must through layoff taxes and partly through payroll taxes. This We or impossible for it difficult show that, may more under shallow pockets, this optimal layoff tax schedule tax on the lead to first layoff for layoffs may —creating "snowball effects" indeed be the case, and derive the a firm composed of two jobs. Intuitively, the layoff should not be so high relative to the layoff tax on the second job as to lead the firm to want to close the second job. Ex-post wage bargaining. Our benchmark assumes that wages are set ex-ante. Thus, better insurance reduces the wage that risk averse workers are willing to accept and reduces expected labor costs. In fact, wages are often partly renegotiated ex-post. In that case, rather than leading to lower wages, social protection, be it more generous through unemployment benefits or layoff taxes, strengthens the bargaining position of workers, and thus is likely to lead to an increase rather than to a decrease wage is now workers. We Ex-post wage determination also implies that the depend on productivity, introducing wage risk for take up the case of ex-post wage bargaining in Section that, in this case, should be in wages. likely to less 4, employed and show workers have any bargaining power, the contribution rate if than one. The higher the bargaining power of workers, the lower the feasible level of insurance, and the lower the contribution rate. Heterogeneity. Our benchmark treats tical. This first consider heterogeneity of firms. is erogeneity stochastically is and workers as ex-ante iden- all consider various dimensions of het- 5. in the distribution of productivity. function firms We clearly not the case in fact. is erogeneity in Section We all We Weak assume two types of firms are more dominated by that of strong policy, the solution benefits partly is still from Their distribution risky: firms. not observable by the state, and the state firms, differing We relies show that if het- on a uniform tax to provide full insurance to workers, but to finance layoff taxes and partly from payroll taxes —to have a contri- bution rate below one. This leads to a transfer of resources from strong to weak firms, and, because a transfer self select, is weak firms are the relevant firms at the creation margin, such indeed desirable. If the state can instead offer tax menus and firms only the contribution rate for weak firms needs to be lower than one; the destruction margin of strong firms servable let and the is left state can condition taxes undistorted. If heterogeneity on observable characteristics, is ob- then even better can be done. Job subsidies can be used to transfer resources from strong to weak firms, while leaving destruction margins for both strong and weak firms unaffected. This reasoning suggests the desirability of treating different types of firms, for example new versus mature firms, differently. We then consider heterogeneity of workers. In parallel to the treatment of firms, we assume two types of workers, "low-ability" and the distribution of their productivity. "high-ability" workers, differing in We show that if heterogeneity is not observ- able by the state and the state relies on a uniform tax policy, then, to the extent that the state cares about inequality, full insurance and a contribution rate below one are again optimal. Lower layoff tax rates lead to a transfer from high-ability to low-ability workers, a desirable outcome as it reduces income inequality. Again, if feasible, offering menus or conditioning taxes on observable characteristics lead to a better outcome, allowing for transfers from high-ability to low-ability workers at a reduced efficiency cost. This reasoning again suggests the desirability of treating different workers, for example new entrants versus workers with higher seniority, differently. we Before is start, we want to emphasize the limits of our analysis. a static, one-period, model. enough already; but its Our only excuse is Our framework that the framework gets complex limitations prevent us from taking up a series of questions, such as the design of experience rating systems as a way of implementing layoff taxes, or the potential role of individual we mandatory unemployment accounts. Also, are aware that our broad categories of deviations from the not fit benchmark may each specific imperfection. Nevertheless, we believe that the conclusions we draw from our model, about the articulation of unemployment insurance and employment protection, about the relation of the optimal contribution rate and the relative use of layoff and payroll taxes to various imperfections, are quite we draw, in Section of workers, a continuum general and can help design better systems. In that context, 6, what we see as our 1 A 1.1 Assumptions main conclusions. benchmark Tastes and technology are as follows: • The economy of • mass is composed of a continuum of mass (at least) 1 of entrepreneurs, and the 1 state. Entrepreneurs are risk neutral. Each entrepreneur can start and run a firm. There is a fixed cost of creating a firm: I, which is the same for all en- trepreneurs. If a firm is created, a worker then revealed. Productivity is is hired, and the productivity of the match is given by y from cd G(y), with density g(y) on [0, 1]. The lay the worker who then becomes unemployed. off, Realizations are • and produce, or firm can either keep the worker Workers are ii across firms; there with risk averse, benefits, utility if unemployed no aggregate is risk. utility function U(.). Absent unemployment given by U(b) (b the wage equivalent of is is being unemployed). • The state observes only the employment employed or is status (whether the worker remains laid off). Accordingly, the state has three tools at its disposal, a payroll tax rate a layoff tax /, and unemployment benefits, t, /i. (Note the absence of job creation subsidies or taxes; these are straightforward combinations of layoff and payroll The taxes, and are redundant here). state runs a balanced budget. Decisions are taken in three stages: • • 1. The Stage 2. Entrepreneurs decide whether to start firms and pay the fixed cost. They hire workers, Stage state chooses t, and offer / and them /x, subject to budget balance. contracts, characterized threshold productivity level below which the worker is by w and y*, a laid off (the threshold be assumed to be interior in what follows.) will The assumption of a constant wage w rather than a schedule w(y) follows straightforwardly from workers' risk aversion. Note the absence of severance payments, to the worker. As • all all Stage 3. y > and r If y < will firms face the rium, If As y* workers are The , be clear, same cost i.e. direct payments from the firm they would be redundant here. and distribution of productivity, in equilib- initially hired. productivity of each job is revealed. then the firm keeps the worker, produces y, pays w to the worker, to the state. y* , state pays the firm lays the worker fx to the off; the firm pays / to the state, and the unemployed worker. Given these assumptions, expected utility is given by: Vw =G{y*)U{h+u) + {l-G{y*))U{w) The (1) free entry condition (or participation constraint) for entrepreneurs is given by: VF = -G(y*)f + f y dG(y) - (1 - G(y*))(w + r)>I (2) Jy" The government budget constraint VG = 1.2 is -G(j/*)(/z given by: - /) + (1 - G(y*))r > (3) The optimization problem The problem of the firm Before turning to the problem of the state, we need to look at the problem of the firm. At the time of hiring, a firm chooses the labor contract that maximizes the value of the firm, for a given level of expected utility for the worker: maxVjp Prom the first-order conditions, y* subject to is Vw = Vw then given by: ,.(, +T - fl -( gabffi±i!i and w is determined in ) ( 4) turn by the condition that expected utility be equal to Vw The first term in the expression for the threshold productivity producing. Were output was less term which says level will this the only term, the firm than the net cost of producing. But there that, unless the worker is is the net cost of would simply lay the worker is in off when general a second fully insured, the threshold productivity be lower than the net cost of producing. This insurance-like term may be familiar from the old later on; we line of research shall discuss it more length at The threshold defined by equation Ex post, layoff i.e. on implicit contracts, and (4) is will reappear then. not however in general time-consistent. once the firm has hired a worker, given the wage w, the payroll and tax rates r and /, the threshold productivity below which the firm will want to lay a worker off given by: is = (w+-T-f) y* The expression no longer In the benchmark, it (5) includes the insurance term. will be the case that workers are the two fully insured so expressions for the threshold will coincide. Later on, this will no longer necessarily be the case, and so the question will arise of what assumption to make, the choice of the ex-ante or the ex-post threshold. If y is observable and contractible, then the threshold will be given by the ex-ante value. If y is may not contractible however, the firm ex-ante value. What it needs to do is still be able to commit to the to put aside funds in amount {U(w) — U(b + n))/U'(w), to be paid to a third party (not the worker) in case of layoff. This will clearly lead the firm again to choose the ex-ante value for the threshold. Our maintained assumption will be that y* is not contractible, that firms do not put such funds aside, so firms choose the ex-post value of the threshold value given by equation assumption when it The problem of the The (5). matters in Section 2. state. state chooses tax rates expected value of workers' in this We —the shall discuss the implications of the alternative and unemployment benefits so utility, equation (1), as to maximize the (expected profits are equal to zero economy, so we can ignore entrepreneurs' utility), subject to the free entry condition, equation (2), the government budget constraint, equation (3), threshold chosen by firms, equation Note that we can consolidate the and the (5). free entry condition and the government budget constraint to get: VGF = -G(y*)n + [ y dG(y) - (1 - G(y*))w > I (6) Jy* Neither the payroll tax rate nor the layoff tax rate appear in the equation. the point of view of expected profit, given the balanced budget constraint, it From does not matter whether unemployment benefits are financed by one or the other tax. Thus, the consolidated on / and free entry condition depends on [i, and w, not y* directly t. This suggests the following formalization strategy: To set up the maximization problem as a mechanism design problem, finding the optimal allocation subject to condition (6). by the appropriate choice of / and so much in the (io, fi, y*) Then, to find out how the solution can be implemented t. This approach turns out to be useful, not benchmark but when we introduce deviations later. The optimization problem In the first step, we characterize the solution to the following problem. max Vw Vfg > subject to I w,n,y" or, equivalently max {G(y*)U{b + /i) + (1 - G(y*))U(w)} w,ii,y' subject to: -G{y*)pi + f y dG(y) - (1 - G(y*))w > I Jy" Note that /j., ditions, follows that: it w, y* fully characterize the real allocation. w = b = b V* and the levels of w and /i are determined From +n by condition the first-order con- (7) (8) (6) holding with equality. 10 The of condition first utility, a is full insurance condition: Workers achieve the same level whether employed or laid-off The second condition output, efficient for firms to it is is an and unemployed. efficiency condition: From the point of view of total produce so long as productivity exceeds the wage equivalent of being unemployed. Thus, in the optimal allocation, insurance does not stand in the way of We now efficiency. turn to the characterization of taxes which support the optimum. Implementation through payroll and layoff taxes Payroll dition and layoff taxes are determined by the combination of the threshold con- and the government budget constraint. Return to the threshold condition, equation (5). Replace both y* and w by their expressions from above, to get: f-r = The (9) ii net tax associated with laying off rather than keeping a worker, which to the layoff tax minus the payroll tax, is equal must be made equal to the unemployment benefits paid by the state to the worker. This implies a positive relation between the two tax rates: The higher the induce the firm to take the The second equation relation payroll tax, the higher the layoff tax needed to efficient decision. between the two tax rates follows from the budget constraint, (3): G(y*)f Unemployment benefits + (1 - G{y*))r = must be financed either G(y> by layoff or implies a negative relation between the two tax rates: (10) by payroll The higher taxes. This the payroll tax, the lower the required layoff tax. Combining equations (9) and (10 implies: / The layoff /*, t = tax should be equal to unemployment benefits. The payroll tax should be equal to ployment = zero. Define the contribution rate as the ratio of layoff taxes to benefits: r = (f/n)- An equivalent way unem- of stating the previous result is that the contribution rate should be equal to one. 11 We summarize these Proposition 1. results in Proposition In the benchmark, workers are fully insured (b threshold productivity The b). allocation is benefits (f = u.), is = /z w). The implemented through unemployment layoff taxes equal to or equivalently, through a unit contribution rate (r = = = 1). Payroll 0). Implications and discussion Our benchmark yields a Unemployment • + equal to the wage equivalent of being unemployed (y* taxes are equal to zero (r 1.3 1. number benefits of conclusions: and employment protection ments by firms to the state at the time of layoff —defined — are as the pay- both central compo- nents of the optimal set of labor market institutions. They are complements higher need for in that higher unemployment unemployment insurance from firms to the benefits, (a lower b), require higher benefits. The reason is simple: This makes firms We have The outcome is both described the equilibrium as one where the state provides unemployment and firms pay layoff taxes to the state. But an equivalent allocation can clearly be achieved without the state: All which case, is insurance of laid off workers, and production efficiency. benefits to workers, firms to unemployment fully internalize the cost of insurance provided by the state to the unemployed. full payments state. Indeed, in this benchmark, layoff taxes must be equal to • coming from a make severance payments in amount \i is needed is for to the workers directly. In this payments to workers and payments by firms are automatically equal; there no need for third party intervention; and, indeed, firms are eager to provide such insurance to the risk averse workers as this allows them to reduce expected labor costs. In fact, going beyond the confines of this benchmark, there are good reasons to conclude that unemployment insurance should be provided by a separate agency, financed by taxes paid by firms: 12 • As of the time of the layoff, the loss from The outcome able: how long he is of search is unemployment a is random going to be unemployed. severance payment to offset that If the firm were to make a one-time one-time payment would do a loss, this poor job of insuring the laid-off worker. the firm decided instead to pay If the laid-off worker over time, contingent on his unemployment status, other issues would arise: The difficulty for worker, and determine whether he is still many the firm to actually track the unemployed or has found another job; the difficulty in monitoring his search effort is vari- know uncertain, and the worker does not and making sure that he indeed looking for another job. Rather obviously, individual firms cannot monitor laid ance. The off workers well enough to provide them with adequate insur- role of monitoring unemployment status and search be therefore delegated to an agency, private or public. The existing administrative structure, is likely to intensity must state, given its be in the best position to do the monitoring, and to administer the payment of unemployment benefits, either alone or in conjunction with the private sector. There are good reasons to dissociate benefits (which provide insurance to • workers) from layoff taxes (which provide the right incentives for firms). The benchmark, this two serve will different purposes. While the two are equal in the no longer be the case when we relax some of the benchmark assumptions. This dissociation could not be achieved if one limited oneself to the use of severance payments. The benchmark model about the issues. the labor market. But is it useful in setting up the clearly ignores a We now introduce those stage, number we and as a way of thinking of important imperfections in see as most important, and derive their implications. 2 Limits on insurance In our benchmark, workers were fully insured. In practice, incentive reasons com- mand that workers not be perfectly insured. This may be because workers must have incentives to search while unemployed, and because they must have incentives not to shirk when employed. 13 2.1 A With this in generic constraint mind, we introduce the following additional constraint in our opti- mization problem: {l-G(y*))(U(w)-U(b + fJ.))>B The expected utility gain (11) from being employed relative to being unemployed must exceed some value B. Call this constraint the incentive compatibility condition for the worker, or A ICW condition. simple shirking interpretation is Modify the benchmark to allow the as follows. worker to shirk or not shirk. Once hired, but before productivity is revealed, the worker has to decide whether to shirk or not shirk. Shirking brings private benefits B but results in zero productivity and thus a layoff. Shirking is unobservable. Thus, to prevent shirking, the condition above Under the search effort interpretation, the constraint that the from being employed must be constraint in a must hold. sufficient to justify search effort dynamic model. Even in expected utility gain would be the natural our static model, a simple modification of the benchmark leads to a constraint close to equation (11). Suppose that, if laid off, the unemployed can, by exercising search effort, get another job in an alternative sector same (home production, job abroad) as equation (11), with w at wage w. Then, the instead of w. The ICW constraint is the basic implications, in particular with respect to the contribution rate, are the same as those derived below. The optimization problem 2.2 The state solves the ICW same problem as in the benchmark, subject to the additional constraint. So: max {G(y*)U(b + fi) + (1 - G(y*))U(w)} subject to: -G{y*)n + f V Jy" dG(y) - (1 - G(y*)) w > I 14 and (l-G(y*))(U(w)-U{b The + (i)) > B solution to this problem can be characterized as follows. Unemployment straint, benefits, /z, can be no higher than allowed by the incentive con- thus providing incomplete insurance: U{b _* + „) = U{w)1 The threshold level of productivity, y* . 9 Or = , ^ ( (12) ) satisfies: ^_Mp> reorganizing: » For any concave i utility, r ,1. U(w)-U(b + n), n the term in brackets is negative; the stronger the curva- ture (the stronger the degree of absolute risk aversion), the if workers are risk averse, the optimal threshold is unemployed. The lower more negative. Thus, lower than the efficient Firms keep some workers although their productivity alent of being .„,,. is level. lower than the wage equiv- layoff rate serves as a partial substitute for unemployment insurance. The level of w itself is given by condition (6), holding with equality. Implementation through payroll and layoff taxes Replacing y* from (13) in the expression for the threshold decision of firms, gives us the first relation between / and f- T = »+ (5) r: U(w)-U(b + ri JTv) or, equivalently: /-t =p+ (l-G(y*))U'(w) 15 Together with the budget constraint, this condition yields: If B is the layoff tax rate, the more negative the payroll tax rate. In other words, positive, the equilibrium has is the contribution rate, that unemployment / > fi and r < 0. The higher B, the higher the ratio of the layoff tax paid by the firm to the is benefit received by the worker, must now exceed one. The reason is the need to decrease layoffs, a need coming from the limits on the unemployment insurance that can be provided to the workers. Proposition We summarize these results in 2. Proposition 2. In the presence ity is lower than the of limits on insurance, the threshold productiv- wage equivalent of being unemployed (y* < b): The lower incidence of unemployment partly compensates for the limits on insurance unemployed. The allocation employment benefits (f In other words, it is > is u.), when implemented through layoff taxes that exceed un- and negative payroll taxes to balance the budget. implemented through a contribution rate greater than one. 2.3 Implications and discussion • Despite our appeal to either shirking or search effort as motivations for equation (11), there is a sense in which the results carry more generality under the search interpretation. In the shirking interpretation, equation (11) was derived under the implicit assumption that, tivity was equal to zero. shifts the distribution nates H(.), less Our is the worker shirked, produc- Under the more general assumption that shirking from G(.) to H(.), where G(.) stochastically domi- built examples than one. Under a search the firm • we have if in which the optimal contribution rate effort interpretation is however, productivity in unaffected by search effort, and thus the problem does not arise. results in this extension bear a close relation to the results obtained in the "implicit contract literature" (in particular Baily (1974), Azariadis (1975), Akerlof and Miyazaki (1980)). 16 That literature looked risk averse workers. at the optimal contract between risk neutral firms and Under the assumption that there were neither severance payments nor unemployment benefits, one of the conclusions was that there would be overdevelopment, that firms would to the efficient outcome. One lay too few workers off relative of the criticisms addressed to those papers was the question of why firms did not offer unemployment insurance or severance payments. In the discussion here, the limits come from monitoring problems, and the solution takes the form of a But the logic is very much Indeed, the connection is layoff tax rate imposed by the state. the same. even closer than The the choice of the threshold by firms. looks. Recall our discussion of it results above were derived under the assumption that firms choose the threshold according to equation (5), that they choose it ex post and do not to workers at the time of hiring. are bound by that commitment, In this case, = and t 0. it is We feel bound by any earlier can ask what happens so that y* is if, commitment in fact, firms given instead by equation easy to check that the layoff tax rate is given by: / (4). = pi In other words, because firms voluntarily distort the destruction margin, the state does not need to introduce any other distortion, and thus can • still With set the contribution rate equal to one. respect to variations in B, unemployment protection (defined as layoff taxes) are sis is now purely normative, the same trade-off model of labor market institutions. is benefits substitutes. likely to and employment While our analy- emerge in a positive To the extent that the unemployment insurance system provides limited insurance (for any reason), there will be a demand for high employment protection. Continental Europe is An interesting characteristic of indeed the presence of such an inverse relation be- tween the generosity of the unemployment insurance system and the degree of • employment protection Governments in many (see for example Boeri [2002]). countries are exploring ways of providing generous insurance while making sure that the unemployed exert search effort and accept employment (for example by requiring them to accept "reasonable job offers"). In the context of our model, such reforms can be thought of as 17 decreasing B. welfare. The As one direct need to reduce and an such, they have both favorable direct layoffs better insurance. is below the The and indirect effects indirect one is on the decreased leading to better allocation efficient level, efficiency gain. Shallow pockets 3 Our benchmark model embodied the assumption that firms were risk neutral and had deep pockets. These assumptions are surely too strong. Even of aggregate risk, the owners of diversified, and thus are many in the absence firms, especially small ones, are not fully likely to act as if they were risk averse. And, even if entrepreneurs are risk neutral, information problems in financial markets are likely to lead to restrictions on the funds and focus on the implications 3.1 The I + A available to firms. We leave risk aversion aside, of limited funds. generic constraint simplest assumption to f where / > is make here is that each entrepreneur starts with assets therefore the free cash flow available to the firm after investment. In this context, we need to extend the set of tools available to the state. Until now, job subsidies or taxes were redundant. the case. So we taxes /, and job creation taxes The new It is no longer obvious that give the state four instruments, benefits t (job subsidies if t is constraint on the optimization problem is (i, this is payroll taxes r, layoff negative). therefore: f<f-t (16) Layoff taxes must be less than or equal to free cash flow minus job taxes (equivalently: plus job subsidies). We call it the "shallow pockets" constraint. This specification raises the issue of the potential role of outside, "deep pockets" investors in alleviating the constraint. We shall return to this point below. 18 3.2 The optimization problem With the introduction of job creation taxes, the government budget constraint becomes: GW(/-/i) + [l-G(w?)]T + t>0. Combining this equation with the shallow pocket constraint changed) threshold condition /, and (5) gives (16), and the (un- an equation which no longer depends on t, t: G(y*)n-[1-G(y*)](y*- W )<f The optimization problem of the state is max {G (y*) U(b + ») w,/j.,y~ (17) given therefore by: + [l-G (y*)} U (w)} subject to: G(y*)n-[l-G(y*)]{y*-w)<f -G (if)/i+ and / ydG(y)-[l-G(y*)]w>I Jy" where the second constraint the introduction of Prom t, is the consolidated budget constraint, which, despite turns out to be the same as before, equation (6). the first-order conditions, the solution can be characterized as follows: Workers are still fully insured: w = b+n and the threshold = lrj = (18) level is given by: y* where i.e. g(y*)y*/(l — I 7 U \ + 9J TV (19) \T) G(y*)) denotes the elasticity of employment with respect to the threshold (or, equivalently, with respect to the payroll tax, or with respect 19 to (minus) the layoff tax), ated with the constraint the efficient and the second constraints first > binding, v is and v and 6 are the non-negative shadow prices then y* 0, b: respectively. Provided that the first The optimal threshold is higher than level. Implementation through taxes straint > associ- is and the threshold equation imply that / we The government budget straightforward: is smaller than = equal to zero, f \x. The job con- creation The contribution rate tax t is irrelevant, less than one, and the difference between unemployment benefits and layoff taxes is made up by a To understand so if set it f- is positive payroll tax rate. these results, note first that the shallow pocket constraint does not affect the ability of the state to fully insure workers. To see why, suppose for example that w > b + Consider a decrease in fi. w and an increase in r by the same unit amount. This affects neither the threshold condition nor the firm's profit. Then, from the government budget constraint, the unemployment benefit can be increased by [(1— G(y*))/G(y*)] At. Together, these changes imply a change in utility of [-(1 will decrease Note w - G{y*))U'(w) + and increase fi (1 until - G{y*))U'(b + /i)]Ar > still plays no useful role here. why, consider the introduction of job subsidies in amount At (16), this allows the state to increase layoff taxes compensate Thus, the state workers are fully insured. also that the job creation tax or subsidy layoff taxes does not 0. < 0. To From equation by —At. But the increase for the increase in see in job subsidies, and to satisfy the government budget constraint, the state must increase payroll taxes by —At. The effect of similar increases in layoff and payroll taxes leaves the threshold chosen by firms unchanged, and thus has no effect on the allocation. In other words, the state has no way to move cash around so as to soften the shallow pocket constraint. Given full insurance, and the limit on layoff taxes, the state must therefore finance benefits partly from layoff taxes, shallow pocket constraint We summarize our is and partly from payroll taxes. This binding, the contribution rate is less is why, if the than one. results in the following proposition: Proposition 3 In the presence of shallow pockets, workers are still fully insured 20 — b+ (w Layoff taxes are however fj,). less than unemployment benefits, leading to a contribution rate less than one, and a threshold productivity level higher than the wage equivalent of being unemployed (y* 3.3 A > b). Multi-activity firms concern often expressed about layoff taxes (more generally, about payments imposed on firms in bad times) is that, even if the firm can pay the taxes, it can only do so at the expense of other activities in the firm. For example, in order to pay the layoff tax associated with the closing to close otherwise healthy activity In the extend i model we have it = 1, 2. as follows. The down of activity may have a firm 2, 1. just presented, firms are one-activity (one-job) firms. We assume firms to be composed of two joint distribution of productivity in the • With probability a, yi • With probability 1 - is two jobs drawn from distribution a, y\ is drawn from though, of course, their productivity may be G(.), as follows: is and distribution G(.), For convenience, we shall refer to firms for which y^ low), = We activities (two jobs), j/2 = and 2/i j/2 = 0. y\ as "strong firms" (al- and to firms for which 2/2 = as "frail firms." All firms have initial assets 2(1 On + /), sink investment 27, and hire two workers. the government side, the state levies payroll tax t per job, layoff taxes /i in case of a single layoff, 2/2 in case of two layoffs, and pays The assumption is benefits are the of the firm that laid that there is not restrictive. in a position to fully insure workers, unemployment is them no gain same argument benefits \x. that unemployment benefits and payroll taxes are not conditional on the number of jobs preserved the state unemployment off, same for all and given by By the same argument and so it does. This implies that unemployed, irrespective of the state \i = w — b. Similarly, it can be shown in allowing for different payroll taxes. Finally, as before, it as before, and by the can be shown that the job creation tax or subsidy plays no useful role here, and so, without loss of generality, we set it equal to zero. 21 The problem of the firms Ex-post strong firms face the choice of laying the workers, their profit both workers is off, is 4 2/2 by — 2/2 given by yi their profit is given . If —w — t). 2(y\ they keep If they lay Thus, their threshold productivity given by: y Ex-post worker frail off, their profit = * («, + r) - h (20) firms face the choice of laying off one or two workers. If they lay one their profit is is given by y\ — (w + t) — f\. If = w + T)-(2f2 -f1 ( This expression embodies the concern described tax /1 encourages layoffs by on the otherwise healthy priori, there strong or frail frail firms. It creates appears to be now two —that lay two workers frail is given by: (21) ) earlier: off, An increase in the layoff a "spillage" or "snowball effect" activity. constraints on layoff taxes. First, firms have to be able to pay layoff taxes: off, 2/2 Second, those they lay both workers given by —2/2. Thus, their threshold productivity y** A two workers. off zero or — 2(io + r) = < 2/. (22) firms that lay only one worker off must also be able to pay — (w + r)]. As the [j/i state does not observe j/i, it cannot levy more than 2/4- [y** — (w + t)], so /1 < 2/4 [y** - (w + t)}. Using condition (21), this condition reduces to 2/2 < 2/, the layoff tax. Their profit on the remaining activity so we only need is to impose condition (22). Tie optimization problem Using the result (which still holds here) that workers will be fully insured, the optimization problem can then be written max as: U(b 4 m) 22 subject to two conditions: The first is obtained by combining the government budget constraint, the insurance condition (w = b + (i), conditions (20) and (21) for the two thresholds, and the shallow pocket condition 2M The second - - 2a(l (22): - G(y*))(y* b) - (1 - a)(l - G(y**))(y** -b)< surance condition and the two threshold conditions, can be written 2a + (1 - a) f (y - b)dG(y) > Jy" {y- b)dG{y) Jy" Prom 2/ (23) the consolidated budget constraint which, using again the is full 27 full in- as: + 2M (24) the first-order conditions, the two thresholds are given by: v =y v where, as and v and earlier, 77 is the elasticity of 6 are the two non-negative + ej \r] employment with respect to the threshold shadow prices associated with the first and the second constraints respectively. Implementation through taxes olds, together is straightforward. The equality of the two thresh- with the two threshold conditions, implies that /1 firm has shallow pockets so v unemployment benefits the contribution rate is > 0, /1 = ji = / and layoff taxes less than one. Proposition 4 Limited is We . Whatever = /2, and difference exists made up through if the between positive payroll taxes; can summarize our results as follows: assets in a multi-activity firm raise the possibility of "snowball effects". In that case, the state in general does not want to collect the full collectable amount in case of partial layoffs. In the where a two-job firm has assets 2/ layoff, and 2/ for two layoffs (if f example we presented, after investment, the state levies tax is / for one not too large). 23 The however that the result and due layoff tax schedule is linear is too strong, to the stochastic structure of our example. In our example, /i affects only the decision of "frail" firms to lay one or two workers if By off. contrast we know that, the relevant choice were between zero and one layoff as in the previous section, maximum then Unsurprisingly, layoff taxes if we add and y2 G(-), j/i whose relevant with probability a and trade-off that for the complementary fraction 1 — 2/2 from distribution H(y). (All firms are ex-ante = 1, h = h = /, 3.4 and for = 0, h drawn (= 2/2 ) J/i identical.) = > f2 To see this, - as above (y\ with probability of the firms, fi). between zero and one is structure: /i = < 2/ whenever 2/ of the firms have y\ and y 2 suppose that a fraction from firms = (/i we obtain a concave tax layoff to our example, = would be optimal = 1 and y2 is Then we know 2/. In general, A > f2 drawn - a), and 1 drawn that if . Implications and discussion Our benchmark looked at the case in which firms and concluded that / layoff tax = /x. By had enough assets to pay the we contrast, the shallow pocket firm looked at in this section had limited assets and could not pay more than /, leading to a lower contribution rate. This raises the question of We The how / is itself determined. have explored this question in two ways. first is that policy to assume, in contrast to the maintained assumption in this paper, is set after rather than before firms the state cannot commit and sets (t, /, invest. Suppose, for example, that n) after firms have invested and hired workers, but before they learn the productivity of the match. In this case, firms will obviously layoff, so /= choose to be "judgment proof", 0. The optimal threshold is i.e. to have no assets then given by: y* — 6/(1 — left in case of (1/t?)). (This condition directly follows from equation (19): Under ex-post policy setting, the entry condition becomes irrelevant to the decision problem of the state, so 8 and = 0, this equation obtains.) The second finance. is to assume that the firm has limited assets but has access to outside The answer, not surprisingly, firm and the investors. Suppose investors, first depends then on the relation between the that there are deep-pocket, risk-neutral, but that these investors are uninformed: Like the state, they cannot 24 observe productivity directly, just whether workers axe employed or unemployed. Then, it is clear that allowing for the presence of such investors we to the results do by itself. just derived. Whatever the investors For example, in the one-activity case, investors to the firm be equal to F, be D if is this is exactly the choice of t, /, t could not use (for t if it x to the investors does not. Then, the investors' break given by: G{y*)D 1 + But D difference the funds advanced by the and the payments of the firm production takes place, and even condition let makes no do, the state could already - G(y*))D > (1 same condition a given value of fj.) F as the condition faced earlier. And, for the by the state in its same reason the state to relax the shallow pocket constraint, nor can the presence of investors relax this constraint. Things are clearly different if investors are informed, of productivity y. In this case, they can ask give is more to the firm in clear that the firm bad and more states. Indeed, in the i.e. can assess the realization good of the firm in states, and absence of other constraints, it investors together are just like the deep-pocket firm of the benchmark. In that case / can be set equal to fi, independent of the free cash flow of the entrepreneur. Things become more interesting and so can ask for we assume that investors are indeed informed payments from the firm contingent on trepreneur enjoys a rent investors can get at if R > y, in case of continuation (so but that the en- if y is total value, most y — R). This assumption implies the standard corporate finance feature that the entrepreneur more eager than investors to continue is since he receives a private benefit even when the firm no longer makes profits. Setting up the model and characterizing the solution would take us too and we refer the reader to Blanchard and Tirole (2004). In that extension, we find that the threshold level of productivity is higher than the efficient level (y* leading to more layoffs than in the benchmark. But to set a unit contribution rate, in this case, far afield, and workers are it is still fully insured > b), optimal for the state / = /x = w — b. Thus, shallow pockets lead to inefficiently high layoffs, but do not justify a lower contribution rate. The reason for these results is that low entrepreneur's 25 solvency is the culprit and the state, constrained to balance the budget, can do nothing to boost this solvency. Ex—post wage 4 negotiation Our benchmark embodied the assumption that wages were set ex-ante, i.e. at the time of hiring. This had the implication that, by offering unemployment insurance to risk averse workers, a firm could not only offer a lower wage, but actually lower its unemployment To some firm offered the option of joining the layoff tax benefits system, extent however, there this is the case, a firm in A expected labor costs. would have voluntarily joined. is always some room which has to pay a for layoff tax ex-post bargaining. provision of benefits if laid off unemployment is in a stronger position. When lays a worker off if it a weaker bargaining position vis a vis that worker; a worker unemployment cum The who is will receive layoff tax and the benefits both lead to higher, not lower, wages, and thus to increase labor costs. In this section, we therefore modify our earlier assumption about wage setting, assume ex-post wage bargaining instead, and characterize the shape of optimal labor market institutions under ex-post wage bargaining. A 4.1 The formalization of wage bargaining on ex-post following formalization captures the effects of benefits and taxes wage determination Assume wage in a simple way: setting now takes place after productivity outcome of a two-stage game. In stage firm. The wage is 1, probability 1 — it down. maximize If by the worker with probability is the it /?, turns or it down, the by the firm with /3. Under the assumption that the firm chooses the threshold as to and realized, the worker makes a wage offer to the firm can either accept the offer or turn set in stage 2, either is profit ex-post, in stage 2, the highest and therefore the wage offered by the worker, is level of productivity so wage the firm equal to y +f— r. will accept, The highest 26 wage the worker b + fi. and therefore the wage will accept, Thus, the expected wage in stage 2 This implies that, in stage 1, the (risk neutral) firm, an i.e. = w(y) The higher the is offered given by 0(y the worker will make the by the firm, is equal to + f - t) + (1 - /?)(& + /x). highest offer acceptable by offer of + f-r) + (l-P)(b + (3(y layoff tax rate, or the higher i) (25) f unemployment benefits, the higher the wage. The threshold value for productivity, y* = y* , is w(y*) given by: +t-f and rearranging: Equivalently, using equation (25) y* = b + fi + r-f Just as in the benchmark, the combination of / efficient threshold, y* = But, as b. we (26) = //, r = would deliver the shall see, other considerations are now relevant. Expression (26) allows us to rewrite the wage schedule w(y) The wage paid {b + ti)+0{y-y*) (27) to the marginal worker, the worker in a job with productivity equal to the threshold level, plus = as: unemployment is equal to benefits. (6 + /z), the wage equivalent of being unemployed The wage then increases with (3 times the difference between productivity and threshold productivity. 4.2 The The optimization problem state solves the constraint that the same problem wage is as in the benchmark, subject to the additional no longer a decision variable, but is instead given by equation (27): 27 max G(y*) U{b + fi) + f U{w(y)) dG(y) subject to: -G(y*)fi+ f (y-w(y))dG(y)>I Jy" and The w(y) = b + /i + (5{y - y*) solution can be characterized as follows: The threshold level of V where E[U'\y > marginal y*} utility if = productivity =b+ 0(1 (J 1 , is implicitly defined by: ~ E[U'\y>y*} -G(y*)) W) EU r [ U'{w{y))dG(y))/{\ employed, and E U' = marginal ity, and level. is strictly positive, utility if so y* is The more employed greater than , the One way of getting G(y*)(b more + /J. + ri+ I ((b + U'{w(y)) dG(y) then the expected than the unconditional expected marginal The is utility. util- layoff rate exceeds the production efficient workers in bargaining, the higher the layoff rate. unemployment benefits, Vfg = |i, is then determined by I- intuition for the optimal threshold constraints, Jy' left level, feasible level of and the second fJ.) strictly risk averse, constraint above, the condition that first The b. + the expected value of is risk averse the workers, or the stronger the the higher the threshold Given y* the and workers is less G{y*)) G(y*)U'(b the unconditional expected value of marginal So long as - () ] is to combine the first and reorganize to read: i) f + P(y-y*))dG(y)<G(y*)b+ f Jy' side gives the value of total workers' y dG(y) - I income (the sum of wages, reservation 28 wages, and benefits) in the economy. in the The economy. The two are the same, as Suppose do not for the moment free entry implies zero profit Then, changes in y* change the shape of the income profile without affecting the total workers' income. more workers receive the same, but the income schedule point of insurance The assumption from y* this implies that, as why and increases, and just b so p, is greater than 6, 4.3 is flatter. where it = b, increase in y* income remains Indeed, the best value of y* from the workers receive the same income. all is however clearly decreases risk, at least a small increase in y* = 1 for example, then total income increasing in the degree of concavity of It is increasing in desirable; This implies that the optimal value of y* zero. than one. less It is is But, as y* increases, the efficiency loss b. income decreases. For y* but An overall workers' small changes in y* do not affect total income; must be equal to the utility function. /3, (fj, The b). the optimal y* exceeds total is The higher 1, + that changes in y* do not affect total income incorrect. Starting this explains = y* is income. that changes in y* do not affect total income, and so affect total workers' income. implies that income right side gives the value of total the share of the surplus going to workers: /3, the steeper the wage schedule, the larger the Implementation through payroll and demand for insurance. layoff taxes Replacing y* from (28) in the expression for the threshold decision of firms, gives us the first relation between / and f-T-p-pa The other relation is r: E[U'\y>y*\ (l-G(y*)) j^—] [i g{r) given, as before, is now negative, / < fi Together with the government budget constraint, this implies / is higher than the efficient from its benchmark The reason level. value, This namely (29) by the budget constraint, equation the second term on the right side of equation (29) contribution rate below one. is (26), is clear (3). —r < As /z. and so a from above: The optimal threshold achieved by lowering the contribution rate unity. By implication, payroll taxes positive, in order to finance the shortfall of the must be unemployment insurance system. 29 Note that = for 0, i.e. if workers have no bargaining power, then same characterization as in the benchmark: / As it (3 becomes positive, ' = and the wage schedule fi is and now =w— pL we obtain the b. increasing in productivity, can be shown that the equations above imply: # < is, 0. d/3 d/3 That d ± < both the unemployment benefit and the layoff tax decrease as the work- more bargaining power, and the ers acquire tax layoff falls faster, leading to a decreasing contribution rate. We summarize our Proposition 5. marginal worker results as follows: When wages is the same are set through ex-post bargaining, utility for the as for the unemployed workers, and workers with higher productivity receive a higher wage. These outcomes are independent of the state policy choices. The remaining choice for the state is that of y* optimal to choose a threshold higher than the efficient level: (y* decreases the uncertainty faced by workers at some is in 4.4 > It is . then This choice b). cost in efficiency. This choice turn implemented by a contribution rate lower than unity. Implications and discussion Under pure ex-post wage bargaining, the scope provide insurance to workers is extremely limited: benefits does not change the slope of the an amount equal to unemployment freedom is for An unemployment increase in unemployment wage schedule, increasing Put benefits. differently, the benefits to all wages by only degree of the location of the kink y* in the worker's income schedule. Indeed, in our framework, given the threshold level of productivity, the level of unemployment benefits entry condition —which implies that workers' income, the sum of wages, reserva- tion wages, is entirely determined and unemployment benefits, by the wage schedule and the must be equal to In that dimension however, our framework is too extreme. free total output. The static, one-period, nature of the model implies a fixed unemployment duration. In dynamic models, 30 such as Mortensen and Pissarides (1994) for example, the state free to is choose the level of unemployment benefits; but the higher the level of unemployment benefits, the longer the equilibrium result with the an increase same flavor obtains. In the Pissarides (2000) unemployment in unemployment duration. And an an increase benefits leads to exactly such that the "pain of unemployment" i.e. , model in irrelevance for example, expected duration the difference in the expected value of being employed or unemployed, remains constant. Heterogeneity 5 We have assumed so far that reality, more (or, they all they clearly are not. Firms face, workers and differ in firms are ex-ante identical. In generally, the distribution of productivity and in their initial assets. New productivity: Workers and relative demand shocks) also differ in the distribution of entrants have distributions characterized by higher uncertainty than workers with more labor market We all the distribution of productivity shocks seniority, and so on. study in this section the implications of heterogeneity, both on the firm and on the worker distributions, 5.1 side, both observed and unobserved. focus on differences in but discuss some other dimensions of heterogeneity later. Heterogeneity of firms Suppose there are two types of productivity distributions. The and "weak" which firms, "strong" , productivity of "strong firms" mulative distribution Gh{-) and The We is differ in their drawn from cu- that of "weak firms" from distribution Gl(-). distribution function of strong firms stochastically dominates that of firms: for all y in (0, 1), Gn(y) < Gl{v)- The fraction of strong firms is weak equal to P- We is start with the unable to tell assumption that this heterogeneity is the two types of firms apart. that the state sets a uniform policy strong and weak firms may (t, /, p); We we unobserved, so the state also start with the assumption shall look later at menus in which self select. 31 Note that given the distribution assumptions and uniform taxation, strong firms have higher expected profits than small firms. Thus, at the creation margin, the free entry condition is relevant for lay workers off, weak firms only. so the destruction margin is By contrast, both types may relevant for both types of firms. This in turn implies that we can no longer merge the free-entry condition and the government budget balance constraint: The former applies to a subset of firms (the weak firms) and the latter to the whole set of firms. We take the more "pedestrian" route of optimizing with respect to We therefore have to all the variables. assume that the state maximizes the welfare of workers. Allowing the state to also put some weight on the would not alter the results. max positive rents earned The [p(l problem is given by: + {l-p)G L {y*)]U(b + n) \pG H (v*) + by the owners of strong firms state's optimization - G H {y*)) + subject to the free entry condition for -G L {y*)f + / (1 - p){l - G L {y*))\ U(w) weak firms (y - w - T)dG L (y) > I the government budget constraint + (l-p)GL (y*)](n-f) +[p(l-GH (y*)) + (l-p)(l-GL (y*))]T > -[pGH (y*) and the threshold productivity condition, which is the same for weak and strong firms: y* The =w + t- f solution can then be characterized as follows: • The • The threshold state fully insures workers: w = b + \x. level of productivity is given by: 32 y *^ b P(G L (y*)-G H (y*)) , (P9H(y*) By p the definition of = 0, ployment is weak and strong the threshold level This solution is in turn benefits, so + (i-p)gL<y*)) firms, Gl(j/*) higher than the efficient is implemented by choosing > Gh(v*)- So, unless level. layoff taxes lower through a contribution rate lower than unity. than unem- The difference financed through payroll taxes. The intuition for types of firms firms. By why may the contribution rate lay workers off contrast, the creation than one is less as follows: Both and so the destruction margin applies to margin corresponds to the weak firms state can then improve workers' welfare strong to weak firms. Because the state by allowing is for The some cross-subsidy from Weak firms lay workers off more from a contribution so benefit only. all unable to assess the firms' strength, the cross-subsidy operates through the contribution rate: more than strong firms and is rate smaller than unity. Can the state do better by offering the state offers an option (tl, menus and targeted at /x,) strong firms, rather than the single option (thresholds, payroll and optimization problem is layoff taxes are identical to the letting firms self select? weak (r, /). Suppose firms and another (th,Jh) at With obvious notation change now indexed by the type of firm), the problem above, except that the last con- straint, y* is - [w + t - f] = replaced by the constraints: L - [w + rL - fL = H - + th - fH = y* y* [w ] ] and the incentive compatibility constraint that the strong firms do not want to masquerade as weak ones is given by: 33 -G H (vh) fn+ [ (v-w-t„) dG H (y) 'Vh , > -G H It is useful at this (y* L) h+ (y-w-rL dG H (y) ) [ point to define the net contribution rate as the ratio of the layoff tax minus the payroll tax to unemployment benefits, that, so long as the net contribution rate efficient: What matters here is is (fc — Ti)/fj,i, i = H,L. Note equal to one, the layoff decision not the layoff tax itself, is but the difference between the layoff tax and the payroll tax. The solution can then be characterized as follows: w = b+ • The • Strong firms face a net contribution rate equal to one, and so choose the state still fully insures workers: fi. efficient threshold: H— y* Weak firms • and b face a net contribution rate Jh-th = H below one, and so choose a threshold higher than the efficient level: In short, while offering a menu improves characteristics of the uniform policy, efficiency, the solution carries the main namely cross-subsidization of weak firms by strong firms, with a contribution rate for weak firms below one. If heterogeneity is observable, i.e. if the state is able to tell weak and strong firms apart, the state can do even better. Characterizing the optimal policy forward: by It is still optimal to subsidize weak firms. This can offering a job creation subsidy to the tion rate equal to 1 for from strong to weak weak is straight- now be done however firms, while setting the net contribu- both types of firms. This achieves the desired redistribution firms, while avoiding distortions at the destruction margin (Note that both layoff and payroll taxes must be higher than in the benchmark, as the extra revenue is needed to finance job creation subsidies to weak firms. The 34 difference between the layoff tax unemployment We and the payroll tax remains however equal to no distortion benefits, so there is at the destruction margin). have focused above on heterogeneity in productivity distributions. While we shall not present results here, we have also explored heterogeneity in assets, in- cluding the model of corporate finance described informally in Section 3 (with informed investors, and some non pledgeable income.) There, the same type of sults emerges, for the assets, have smaller same reason. profits, The weaker firms, i.e. those firms with the least and thus are the relevant firms at the creation margin. As, in order to satisfy investors, weaker firms also lay workers off more from a decrease firms, they benefit Thus, is if heterogeneity is in the layoff tax unobservable and the state re- relies more than other than stronger firms. on uniform taxation, it optimal for the state to decrease the contribution rate below one, transferring profits We from stronger to weaker firms, and favoring job creation. can summarize our results as follows: Proposition 6. Suppose that there are two types of firms: "strong" or "weak". Strong firms have productivity distribution Gh(-), weak firms G H {y) G l(.) , andG/,(y) > for ally in (0,1). If this heterogeneity the optimal pohcy higher than the is is unobserved, and if the state relies on a uniform policy, then to fully insure workers (w efficient level (y* > b), = b + fi), and implement it choose a threshold level through a contribution rate less than one. If the state offers a menu and has the following instead, then the optimal menu separates the firms properties: Workers are fully insured. Strong firms face a net contribution rate of unity, and choose the efficient threshold. Weak firms face a net contribution rate below one, and the threshold exceeds the efficient The underlying mechanism remains level. the same, cross-subsidization of weak firms through the use of a lower contribution rate. If heterogeneity is creation subsidy to implement it observed by the state, then the optimal policy weak is to offer a job firms, choose the efficient threshold for productivity, and through a unit net contribution rate for both types of firms. 35 Heterogeneity of workers 5.2 A similar set of issues arises on the workers' side. particular, some more uncertain and thus more are To think about the issue, we up a case very set Not all likely to workers are be In alike. laid off. similar to that of firms. We assume there are two types of workers, "high- ability" and "low-ability" workers. High- ability workers have a productivity distribution given by Gh(-), low- ability workers a distribution given by Gl(-), with Gl(v) fraction of workers with high ability We is assume that firms know the workers' earlier case of heterogeneity in firms, does not know the workers' We equal to abilities, we and that y € (0,1). The for p. abilities. start > Gn(y) Proceeding in parallel with the by assuming both that the state it chooses a uniform policy r, /, fi. can simplify the set-up of the optimization problem, by noting that work- ers of each type will more valuable to be fully insured, firms, they will and that, because high-ability workers are be paid more, both when employed and when unemployed. Suppose therefore that low-ability workers receive unemployed, and are receive to + A where fully insured, so A when employed, and fi w = b w when + p. \i when High-ability workers will then + A when unemployed (so w + A = b+p + A), the additional expected profit brought about to the firm by a high- is A ability worker. Noting that productivity the same for both types of workers, A= employed, and is [G L (y*) - GH (y*)] does not affect the layoff decision, so the threshold / + / [y - (b it + p) - follows that r] [dG H (y) A is - dG L given by: (y)] Jy' Payment of A can be achieved through higher wages when employed and severance payments from the firm to the high-ability workers when laid-off. Alternatively, it can be achieved by wage-indexed unemployment benefits, so a worker w+A receives p. +A is paid (and the firm pays correspondingly higher layoff taxes when laying a high-ability worker With who off.) this characterization of wage setting, and assuming that the state maximizes 36 a utilitarian social welfare function, the optimal policy is the solution to: max {(1- p)U(b + (j,)+pU{b + n+ A)} tJ,h,v" subject to the free entry condition -G L (y')f+ [ Jy' [y-(b + ri-T}dG L (y)>I and the government budget constraint [(1 - p) where The • G L (y*) + P G H A (y*)] (J - M)+[(l - p) [1 - GL (y*)] +p[l-GH (y*)]] r > and y* have been defined above. solution has the following form: If the proportion of low-ability workers for productivity is given by: pO-- „-_,, [(i - high enough, then the threshold is p)EW) - G H (y*)} p)9l{v*) + pg H {y*)) [(i + b)- u'{n + b + A)] p)U'(h + b) + P U'(p + b + A)] [U'jii - G L (y*) - G H (y*) is positive, and so So the fraction on the fight is positive, and the optimal threshold Note that than the U'(p, + b) - U'(n + is b + A). higher efficient level. This solution unity, is is in turn implemented by using a contribution rate below with the rest of unemployment benefits being financed by payroll taxes. • If however, the fraction of low-ability workers optimal to not facilitate their employability. and they receive unemployment productivity is benefit is No small enough, then job is created for it is them for sure. In this case, threshold given by: y* This solution \i is = b implemented by relying on a gross contribution rate greater 37 > than one: (f/p) The The 1; payroll tax rate and a net contribution rate equal is by r in turn given intuition for these results is as follows. = [(1 When — to one: / — r = p. p)/p] p- both types of workers are em- ployed, the use of a contribution rate below one leads to a transfer from high-ability to low-ability workers, and therefore reduces ciency. If the proportion of low-ability workers is very low, efits cost in effi- more efficient to efficient separations it is have them not work, and then to choose tax rates so as to have for high-ability workers. some inequality, but at Because low-ability workers receive unemployment ben- but no corresponding layoff taxes are collected, the solution requires having both a positive payroll tax and a correspondingly higher on high-ability layoff tax workers. This increases revenues while maintaining a net contribution rate equal to one (This result is related to the result in Cahuc and Jolivet (2003) where the need to finance a public good also leads to higher layoff and payroll tax Note that the Weak logic of the results workers (firms) are more under firm and worker heterogeneity likely to be laid off (to lay off), internalization of the externality of the layoff on the UI fund rates.) is similar: and an incomplete benefits the creation margin. What to the state can offer menus and firms then self-select? if show that the state then offers two menus, one aimed It is straightforward announce at firms that they have hired a high-ability workers, one aimed at firms that announce they have hired a low-ability worker. The to one. The second menu has a workers are fully insured. first menu has a net contribution rate equal net contribution rate below one. Both types of Thus, just as the uniform case, heterogeneity leads to lower layoff taxes, but in this case only for low-ability workers. What if differences between workers are observable? high-ability from low-ability workers disabled —then the optimal policy —for example is When if The tell apart workers are in bad health or to offer a job creation subsidy (or, equiva- lent^, uniformly lower payroll and layoff taxes) to firms worker. the state can if they hire a low-ability net contribution rate can then be set equal to one for both types, and the destruction margin is undistorted for both types. In general, when differences are only partly observable, the solution combines job creation subsidies and a net 38 contribution rate below one. We summarize our Proposition 7. results in the following proposition: Suppose that there are two types of workers: "high- ability" or "low-ability". High-ability workers have productivity distribution Gh{-), lowability workers Gl(-), and Gl(v) > Gh{v) If this heterogeneity is unobserved the optimal policy is and the state to fully insure workers (w higher than the efficient level (y* rate less than one. for all y in (0, 1). The lower > b), relies on a uniform = b + fi), and implement policy, then choose a threshold level it through a contribution layoff tax ieads in effect to a transfer from high- ability workers to low-ability workers. If the proportion of low-ability workers is small enough, then it is optimal to leave them unemployed, and use a unit contribution rate for the remaimng, high-ability workers. If the state offers a menu instead, then the optimal menu separates workers and has the following properties: Workers are fully insured. Firms hiring high-ability workers face a net contribution rate of unity, and choose the efficient threshold. Firms hiring low-ability workers face a net contribution rate below one, and the threshold exceeds the efficient level. The underlying mechanism remains the same, cross-subsidization of low-ability workers through the use of a lower contribution rate. If heterogeneity is observed by the state, then the optimal policy job creation subsidy to low-ability workers, and choose an productivity, it to offer a threshold for through a unit contribution rate. Conclusion 6 We and implement efficient is see our analysis as suggesting a few general principles: The use of layoff taxes employment transfers same benefits. is the natural counterpart to the state provision of un- The second requires the from firms to workers, can be seen principle. In that sense, tion are basic components first. Severance payments, direct as a crude way of implementing the unemployment insurance and employment protec- of the optimal set of labor market institutions. In the 39 benchmark, they allow the state to achieve both insurance and an efficient full allocation. The principle is important. But it must be adjusted to take into account the imperfections present in the labor market. Some imperfections lead to higher layoff taxes than in the benchmark. This case for constraints on the affecting search or work amount effort. The amount the higher the optimal of insurance the lower the limit on the provision of insurance, of employment protection, thus the higher the optimal layoff tax. This however comes at some cost in Most others is which can be provided without efficiency. lead to lower layoff taxes than in the benchmark: Constraints on the assets available to firms in case of layoffs obviously put a on limit layoff taxes. In this case, insurance, but some it it is still feasible must be financed through both efficiency cost. The and optimal to provide layoff and payroll full taxes, again at asset constraint also affects the layoff tax schedule, the number relation of the layoff tax to the of layoffs. Ex-post wage bargaining reduces the scope for insurance. And it is optimal in this case to have a lower layoff tax rate than in the benchmark. This provides income insurance, but at some some efficiency cost. Finally, heterogeneity of firms or of workers also typically implies a lower layoff tax rate than in the benchmark. To the extent that weak firms face higher layoff risk, a reduced layoff tax rate, together with a corresponding increase in the payroll tax, transfers resources argument holds from strong to weak for workers. A firms, improving efficiency. A parallel reduced layoff tax rate transfers resources from strong to weak workers, improving aggregate welfare. To the extent that the state can offer rates for menus and let firms weak on observable a firms, or are very our results suggest offering lower layoff tax weak workers. To the extent that the characteristics, job subsidies for less distortionary solution We self-select, than lower much aware however one-period model, there is weak firms or state can condition weak workers provide layoff tax rates. of the limits of our analysis. a number of issues still Even within our to be explored. Let us mention two. 40 An important issue not taken up here layoffs as triggered reservation is that of quits versus layoffs. by productivity shocks (shocks to wage shocks (shocks to b, and quits y), or to the disutility of work If we think of by as triggered —which we do not have explicitly in our model) and we think of the layoff tax as applying only in , The case of layoffs, this raises two sets of issues. workers they would The actions by firms is actions by firms to induce (harassment), and actions by like to lay off to quit instead workers to induce firms they would The second first is like to quit to lay them off instead (shirking). and workers together to mislabel quits and incentives to harass, shirk, or cooperatively misreport, each case on the contribution rate. We layoffs. depend very much in have informally explored these issues in Blanchard and Tirole (2003), but a formal treatment remains to be given. Another role, issue the role of judges, who, in is and are often ultimately is that this countries, play a central whether in charge of deciding argument not. Clearly, the logic of our many European layoffs were justified or better accomplished through a is combination of layoff taxes and severance payments, with the decision then being left to the firm. But our look at the implications of imperfections, from shallow pockets to heterogeneity, also suggests the desirability of adapting layoff taxes to particular situations. This can in principle be done through offering menus, or allowing taxes to be conditional on observable characteristics of firms, or by leaving some in fact have the information, the discretion to judges. It remains to more informed ability, be shown however that judges do and the incentives, to take better and decisions. Then, and obviously so, there are dynamic issues we could not consider at all in our one—period model. Dynamic models of the labor market with risk aversion and imperfections are notoriously hard to derives optimal institutions — solve. The only model we know which defined as the optimal combination of payroll taxes, layoff taxes, job creation subsidies or taxes, and unemployment benefits developed by Mortensen and Pissarides (2001). However, it assumes — was risk neutrality and so cannot deal in a convincing way with the interaction between insurance and efficiency. self- insurance (A model by Alvarez and Veracierto [1998] has risk averse workers, as well as state-provided insurance, payroll severance payments. While it shows (numerically) the and effects of layoff taxes, and changes in some 41 of these instruments, We benefits.) • The it does not give a characterization of optimal taxes and see two extensions of our first is model as crucial. the role and the implications of self insurance by workers (in terms of the model here, the role and implications of the endogeneity of In this context, an important question vidual unemployment accounts such is the role, if any, of mandatory b). indi- as are being considered or introduced a number of Latin American countries (Three papers provide a useful in starting point here. All three allow for self-insurance, and look at the role of state-provided insurance in the presence of other imperfections. In Hansen and Imrohoroglu (1992), moral hazard in search limits the scope for state- provided insurance. In Acemoglu and Shimer (1999, 2000), state-provided insurance affects search, which in turn affects match quality.) • The second is the role of experience rating systems, such as the US system, as ways of implementing the collection of layoff taxes. This requires a careful look not only at the dynamic problem of the firm, but at the exact nature of the financial constraints that There is a final set of issues for it faces. which our analysis extended. These issues indeed provided the initial is useful but must also be motivation for our project. In most European countries, governments are considering reforms of their labor market institutions. In what labor market many emerging countries, governments are considering institutions to put in place. The challenge practical implications of our conclusions for such purposes. this in We is to draw the have tried to do Blanchard and Tirole (2003), focusing on France, characterizing existing institutions, and sketching a tentative path for reforms. But, here again, much more can and should be done. 42 References [1] Acemoglu, D., and R. Shimer, (1999), "Efficient Unemployment Insurance", Journal of Political Economy, 107-5, October, 893-928 [2] Acemoglu, D., and R. Shimer, (2000), "Productivity Gains from Unemployment Insurance", European Economic Review, 44-7, June, 1195-1224 [3] Akerlof, G., and H. Miyazaki (1980) "The Implicit Contract Theory of Unemploy- ment Meets the Wage [4] Bill M. Veracierto Alvarez, F., and provisions," working paper [5] Argument," Review of Economic Studies, 48: 321-338. (1998), "Search, Self-Insurance and Job Security WP98-2, Federal Reserve Bank Azariadis, C. (1975) "Implicit Contracts and of Chicago Underemployment Equilibria," Jour- nal of Political Economy, 83: 1183-1202. [6] Baily, M.(1974) Economic [7] Blanchard, O., and MIT [8] "Wages and Employment under Uncertain Demand," Review of Studies, 41: 37—50. Tirole (2003), "Contours of J. Employment Protection Reform" working paper 03-35, November Blanchard, O. and J. Tirole (2004), "Notes on Corporate Finance and Layoffs", mimeo, January [9] [10] [11] Boeri, T. (2002) "Making Social Europe Compete," mimeo, Universita Bocconi. Cahuc, P. (2003), France, May Cahuc, P. and G. "Pour une Meilleure Protection de l'Emploi", mimeo Paris Jolivet, (2003), "Do mimeo I. tection Legislation?" [12] , Paris We I, Need More Stringent Employment Pro- Hansen, G. and A. Imrohoroglu, (1992) "The Role of Unemployment Insurance in an Economy with Liquidity Constraints and Moral Hazard", Journal of Economy, Political 100-1, February, 118-142 43 [13] Ichino, A., fort," [14] mimeo, EUI and University of Basel. Mortensen, D., and C. Pissarides (1994) "Job Creation and Job Destruction Theory [15] and R. Riphahn (2003) "The Effect of Employment Protection on Ef- of Unemployment," Review of Economic in the Studies, 61-3(208): 397-116. Mortensen, D., and C. Pissarides (2001) "Taxes, Subsidies and Equilibrium Labor Market Outcomes", mimeo, April [16] Pissarides, C. (2000) Equilibrium Unemployment Theory. Second edition, MIT Press. [17] Pissarides, C. (2002) cations for Optimal 3837 "Consumption and Savings with Unemployment Risk: Impli- Employment Contracts," mimeo, March :28 44 Date Due Lib-26-67 MIT LIBRARIES 3 9080 02617 8365