Department of Economics
Master Thesis in Economics
Advisor: Bertil Holmlund
Fall 2006
On Unemployment Insurance and Experience Rating
On Unemployment Insurance and Experience Rating
Survey Study
Katya Vasileva
January 2007
Abstract
There have been numerous studies, both theoretical and empirical, dealing with how the degree of experience rating affects employment. The results have been often contradicting, giving ground for ongoing debate among the economists. Here in this thesis I will present the most central studies done concerning the effects on employment induced by layoff costs and UI payroll tax in particular. A special interest will be put on the implicit-contract model, since it appears to be especially useful when studying the effects of experience rating. The theoretical approaches and the existing empirical evidence will be discussed. By doing this survey I will make an attempt in explaining and summarizing the crucial and sometimes contradicting conclusions made though the years, so that a reader unfamiliar with those concepts will be able to easily grasp.
Table of Content
Introduction………..3
The US System of Experience Rating……….4
Theoretical Issues………...9
Empirical Investigation………..20
Conclusion……….28
Introduction
The American way of financing the unemployment insurance by taxing the employers based on how often and how many people they have laid off in the past has been a topic of studies for many years. This financing practice of the UI is well known as
experience rating meaning that each employer’s UI tax rate is determined by his/her layoff experience. Is an imperfect experience rating, where the employer is obliged to reimburse the UI fund only for a fraction of the UI benefits his former workers received, encouraging more layoffs than perfect experience rating, when the firms repay in taxes the full value of the UI benefits paid to their employees? There have been numerous studies, both theoretical and empirical, dealing with how the degree of experience rating affects employment. The results have been often contradicting, giving ground for ongoing debate among the economists.
Here in this thesis I will present the most central studies done concerning the effects on employment induced by layoff costs and UI payroll tax in particular. A special interest will be put on the implicit-contract model, since it appears to be especially useful when studying the effects of experience rating. The theoretical approaches and the existing empirical evidence will be discussed.
By doing this survey I will make an attempt in explaining and summarizing the crucial and sometimes contradicting conclusions made though the years, so that a reader unfamiliar with those concepts will be able to easily grasp.
The UI system studied in this thesis is limited to the US unemployment insurance, since it is the only one financed via experience rating. Effects caused by other elements of the UI system different from the UI tax are not studied. I limit the paper only to studying effects on employment and no welfare aspects are considered. Finally, I present one particular theoretical model, since it is the most relevant and useful method for examining the role of experience rating by focusing on layoff costs implied by the structure of the unemployment insurance.
literature and presents the literature in question. Next section is a review of some empirical studies done on the topic. Finally, section five summarizes and concludes the survey and some general insights on the topic are made.
The US System of Experience Rating
The United States unemployment insurance program dates back to 1935, when the Congress launched it as a part of the Social Security Act. The unemployment insurance aimed two main problems. The first one was to support people’s purchasing power by assisting workers who have involuntary lost their jobs with UI benefits to compensate their lost income and the second was to act as a stabilizing factor during times of recession or seasonal layoffs. In 1939 the Federal Unemployment Tax Act (FUTA) was established to specify the way of financing the unemployment insurance (UI) by imposing state taxes in accordance to the firm’s past layoff behavior and uniform federal taxes on the employers. Today the UI tax mechanism works as follows. Both, the state and the federal taxes are collected and transferred to the Federal Unemployment Trust Fund (UTF), but while the state taxes are put into the state’s account and used to pay the regular UI benefits and 50 percent of the extended UI benefits to the unemployed in the particular state, the federal taxes are used for UI administrative costs nationwide. What is left from the federal unemployment administration account goes to cover 50 percent of the extended UI benefits in the state, and if there is some tax money left after that, they go to the loan fund financing other states, which have exhausted their resources. So, one can say that each state deposits its UI taxes in the Federal Unemployment Trust Fund and later withdraws money to cover the regular benefits and the half of the extended benefits in the state, while the federal taxes serve three additional accounts.1
1
The two main elements of the US unemployment insurance are the payments to the eligible unemployed workers, on one side, and the financing of these payments, on the other. The most striking difference between the financing of the American UI and of those in the rest of the world is the experience rating practice, where employer’s UI tax rate depends on his layoff history. This practice meets a lot of support from the U.S. federal government and is designed to encourage employers to stabilize employment. Under experience rating the higher the layoff rate of a firm for some specified time period, the higher the future tax rate on that firm subject to some range of minimum and maximum values. Note that the tax rate of a firm with no layoffs cannot go below some minimum level for the state just like the maximum tax rate is not high enough so that firms with high layoff rates are subject to perfect experience rating. The result of that is that firms with low layoff rates subsidize other firms with high layoff rates. All states’ UI programs must satisfy some minimum requirement set by the government as for example the federally mandated minimum annual tax base of $7000, and a maximum tax rate of at least 5.4 percent. Different states have different UI laws though and therefore different range of the tax rates, different tax bases, different methods of experience-rating accounting etc.
Experience rating can be categorized as perfect and imperfect (or incomplete). People refer to perfect experience rating when firm’s tax liabilities are equal to the full amount of UI benefits generated by workers laid off by the firm. Similarly, imperfect experience rating is when the firm is liable for less than the whole amount of benefits
Employer UI Taxes State Taxes Federal Taxes UTF State Account UTF Federal Account (Adm.) Extended Benefits Account Loan Fund
Figure 1 – Distribution of UI taxes
used by former employees and is subsidized for the rest. If firms are not liable for an increase in the UI benefits drawn by their former employees, then their marginal cost of an additional lay off is zero and one can conclude that they have an incentive to temporary lay off workers when product demand falls. In practice, nowhere in USA there is perfect experience rating. All states have been using the imperfect system, but the degree of experience rating among them can vary greatly. According to Topel (1984) there are three reasons for the experience rating to be incomplete. First, the range determined by the maximum and the minimum tax rate makes the rating non-effective for employers with very high and very low layoff rates, i.e. those with zero or low layoff rates are imposed the minimum state tax rate, which is positive for the majority of the states, and in this way subsidize those with very high layoff experience, who cannot be charged with more than the maximum tax rate; the so-called cross-subsidization. Second, the UI benefits aren’t taxed as regular income until 1986 and the third factor listed by Topel (1984) is the absence of interest rates combined with the fact that tax payments lag behind benefits charged. Rephrasing the last argument, the present values of tax increase is less than the present value of benefits since tax increase aren’t subject to interest, therefore the longer the time of benefit repayment, the lower the present value of the tax liability and the lower the degree of experience rating.
Topel claimed that there has been decline in the range of rated unemployment through the years, which made experience rating less effective and relevant for employment decisions. Each state’s reserves depend on the interest accrued on the existing reserves, the tax revenues and the benefit payments, shortly written in per capita units as rt =rt−1 +
τ
t −ρ
lt, where lt is the firm-specific layoff rate and ρis the ratio of benefits charged to the taxable wage base, called charge rate. Considering the reserves dynamics (rt =rt−1 +τt −ρlt) and by presenting data on how benefit levels grew fasterto non-experience rated systems, so the range of rated unemployment became narrower. The years after, wages and tax rates have been increased in luckily successful attempt to fix this damage.
The degree of the experience rating can be defined as the present value of the tax liabilities result of layoffs, relative to the present value of UI benefits drawn by the laid off workers2. Thus an employer who by laying off an additional worker uses benefits with present value greater than the increase in taxes is imperfectly experience rated. The four basic methods for experience-rating accounting are the reserve ratio method, the benefit ratio method, the benefit-to-wage ratio and the payroll variation methods. The reserve ratio and the benefit ratio are the most widespread and the reserve ratio method accounts for more than the half of the cases. Topel (1984) describes these two and here I present them briefly.
The reserve ratio,rt, is the ratio of the firm’s reserve account (rt =rt−1 +τt −ρlt) to its taxable payroll, i.e. the sum of its tax payment minus the sum of all benefits charged over some recent time period, all this divided by the sum of all workers’ taxable wages, which are the actual wages up to some maximum over the same period. When computing taxes under the reserve ratio method, the tax rate is a decreasing function of the reserve ratio in the form of discrete small steps (see figure 2). Those steps are calculated so that specified percentages of the taxable payroll fall in each tax bracket.
2
Topel and Welch (1980) p. 367
Reserve ratio 0 max τ τ min
τ
Figure 2 – A typical reserve ratio tax function
Source: Topel (1984)
In all US states the tax schedule is a step function applied only for tax rates in the range (
τ
min;τmax), which can be approximated as τ =η0 −η1rt, whereη
1 is the slope given by the change in the tax rate (the height of the step) divided by the change in the reserve ratio (the width of the step). Firms whose unemployment/layoff rates lie outside an interval determines by the maximum and the minimum tax rates are not subject to experience rating and their marginal tax costs are zero, while all other firms are applying the experience rate system and the fraction of UI benefits, which they have to repay,according to Topel (1984), is given by 1
1 1 < + = i e
η
η
, where i is the nominal interest rate usually around 0.1 and
η
1 =0.3 in a typical reserve ratio state. Then usually e = 0.75, which implies that for every dollar paid in UI benefits to former employees, the firm should pay $0.75 in UI tax. One should note that e need not be constant throughout the range and that if benefits were subject to interest rate, then there would be perfect experience rating regardless of the step sizeη
1.The benefit ratio, (BRt), the second most used method, is defined to be the
quotient of the total benefits charges to a firm during some time period and the total taxable wages paid over the same period. With this method the tax rate is an increasing step function of the benefit ratio. The beginning and the ending ratio for each step, which in turn marks the corresponding tax bracket, is calculated so that a specific fraction of the taxable payroll belongs to that tax bracket. The tax schedule used for calculating the tax rates between
τ
min and τmax is τt =η0 +η1BRt and, just like in the reserve ratio case, it is used only by those firms whose layoff rates are in a certain range. Those employers’ degree of experience rating is given by = 1[1−(1+ ) ]<1−
iT i e
T
η , where
η
1 is the slope ofthe schedule, i is the nominal interest rate and T is the time period used in the benefit ratio calculations, three or four years for the majority of the states.
Theoretical Issues
One of the most spread theoretical approaches in the UI literature is the job-search theory, which has helped economist to study individuals’ behavior during unemployment and unemployment duration in particular. The model assumes that each unemployed person chooses a search intensity, which combined with his/her preferences for work and reservation wage can influence the probability of finding job. There is evidence though (see Feldstein (1978)), that a large fraction of unemployed workers are often rehired by their previous employer in a sense that they might experience periods of “temporary” unemployment when temporary laid off due to low product demand. For those workers the theory of job-search is irrelevant since it is the employer, not the worker, who determines the end of the unemployment spell. The workers probably know and expect that they will be recalled to work and don’t engage themselves with job-searching. For studying that group of the unemployment population, which turns out to be significantly big (see Feldstein (1978)), the most useful model and the approach that appears to be particularly useful when studying the effects of experience rating is the implicit-contract
model.
employment schedule and the wage schedule based on the output prices which occur with certain probability. The firm has a production function f(.) with f'(.) > 0 and f"(.) < 0.The worker’s utility is a function of their income and hours of work, which is assumed to be differentiable, concave, strictly increasing in income and strictly decreasing in hours of work, so that U1(y,h)>0, U2(y,h)<0, U11(y,h)≤0, U22(y,h)≤0 and
0 ) , ( ) , ( ) , ( 2 12 22 11 y hU y h −U y h ≥
U 3. In the short-run the workers are assumes to have no
other employment opportunities, when unemployed, they receive unemployment benefits. The employee’s expected utility subject to given contract should not be lower than utility expected from alternative employment and the employer’s expected profit should be non-negative. An optimal contract then is this one for which no pareto improvement can be achieved, i.e. no increase in the worker’s expected utility would be possible without decreasing the firm’s expected profit and the other way around. The optimal contract is determined either by maximizing the expected profit while maintaining certain competitive level of workers’ expected utility or vice versa.
The first one to implement the implicit-contract model in a context of unemployment insurance was Feldstein (1976). He modifies the model by letting the employers to contribute to the financing of the UI benefits, which their laid-off workers receive. This adjustment in the model is totally relevant for the US unemployment insurance system due to the experience rating practice.
The theoretical model which Martin Feldstein develops in his paper from 1976, “Temporary Layoffs in the Theory of Unemployment” studies the change in the employment and hours of work when the annual product demand falls. The model in use ignores employees risk aversion, provided that they have access to the capital market and have sufficient assets so that their consumption will not depend on the current employment state. The assumption of risk-neutrality appears, later in other studies, to be of great significance for the outcomes, but I will discuss that in awhile. The employees are assumed to be identical, permanently attaches to the firm, they are paid the same hourly wage regardless of the demand state (the so-called wage stickiness by Baily (1974)), and they use their personal assets to smooth consumption when actual and
3
expected income differ due to variation in employment and working hours. Since the labor market is competitive, the contract which the firm offers to the employees should be so that it gives at least as high utility level as if they were employed by other firm and then the problem in Feldstein (1976) is formulated as to maximize the utility of the attached workers by offering them the optimal combination of employment, wages and working hours, taking into account the constraints introduced by demand, production, taxes and cost of capital.
The price-taking firm’s production function is defined as X=G(K, N, h), where X denotes the current annual rate of output, K is the capital stock, N is the number of employees in the firm and h is the average working hours per employee.
Feldstein (1976) lets λ denote the fraction of low demand period(s) during the year. For that part of the year the price falls from its normal value of p0 to p1. The total annual
revenue can be expressed as
) , , ( ) , , ( ) 1 ( ) 1 ( −λ p0X0 +λp1X1 = −λ p0G K N0 h0 +λp1G K N1 h1 . (1) The employees laid off as a result of the low product demand is the number difference of the initial workers (N0) and the labor force size during low demand periods (N1) and they
receive UI benefit at an annual rate b. Then the size of the total annual unemployment insurance benefits paid out is
B =λb (N0 - N1). (2)
The unemployment insurance tax paid by the firm is a sum of some fixed amount T and a fraction of the total amount UI benefits received by the temporary laid off workers, eB +
T, where e = 1 under perfect experience rating and e < 1 otherwise.
The firm’s revenue spreads over the aggregate wage bill (W), the UI tax and the depreciation and interest costs of capital (cK):
) , , ( ) , , ( ) 1 ( p0G K N0 h0 p1G K N1 h1 T eB W cK+ + + = −λ +λ . (3)
The aggregate annual income of all attached workers to the firm is viewed as the sum of the wage bill and the UI benefits. Wage income and benefits are taxed differently at a rate of ty and tb respectively. In the beginning of the period each worker has the same chance
of being laid off, that’s why the expected net income per employee is
The last step in the setting of the model before assessing the effects of demand change is to define the worker’s utility function which Feldstein (1976) takes as separable and depending only on the working hours and the consumption level, which is relatively constant due to smoothening. That is, the utility of consumption is unaffected by the number of working hours as well as the utility of leisure is independent of the consumption level. ) 0 ( ) ( ) ( ) 1 ( ) ( 0 1 0 1 0 1 0 v N N N h v N N h v y u U = + −
λ
+λ
+λ
− , (5)where u(y) is the total utility of consumption, v(h0) is the utility of working h0 hours in
times of normal product demand, v(h1) is the utility of working h1 hours of those who
were not temporary laid off when demand decreased and v(0) is the utility if the laid off workers, who work zero hours and receive unemployment insurance benefits, while expecting to be recalled.
The employer’s problem is then to maximize each worker’s utility, eq. (5), subject to the constraints eq. (1) to eq. (4) by choosing appropriate h0, h1 and N1. In this way the
constraint of the optimal expected profit of the firm is kept at zero. To estimate the optimal employment and working hours, the paper first states the three first-order conditions when maximizing U:
0 )] 0 ( ) ( [ ). ( ' 1 0 1 = − + ∂ ∂ v h v N N y y u λ , (6) 0 ) ( ' ) ( ' 1 0 1 1 = + ∂ ∂ h v N N h y y u
λ
, (7) 0 ) ( ' ) 1 ( ) ( ' 0 0 = − + ∂ ∂ h v h y y u λ . (8) The derivatives of expected income with respect to employment and working hours can be obtained by differentiating equation (4) after plugging the expressions for W and B from eq. (3) and eq. (2) respectively:0 1 1 0 1 ) 1 ( ] / [ ) 1 ( N b t eb N G p N t N y λ y − b λ − + ∂ ∂ − = ∂ ∂ , (9) 1 1 0 1 / ) 1 ( h G p N t h y y ∂ ∂ − = ∂ ∂ λ , (10) 0 0 0 0 / ) 1 )( 1 ( h G p N t h y y ∂ ∂ − − = ∂ ∂ λ . (11) If substituting each of these derivative expressions in its respective first-order condition, after some rearranging we obtain:
1 1 1 ) ( ' ) 1 ( ) ( ) 0 ( ) 1 ( ] ) 1 ( ) 1 [( 1 p y u t h v v t b e t t G y y y b N − − + − − − − = , (12) 1 1 1 1 ) ( ' ) 1 ( ) ( ' p y u t h v N G y h − − = , (13) 0 0 0 0 ) ( ' ) 1 ( ) ( ' p y u t h v N G y h − − = (14) , where 1 N G =∂G/ N∂ 1, Gh1= ∂G/ h∂ 1, and Gh0= ∂G/ h∂ 0.
When looking closely to each of the F.O.C. above, Feldstein makes number of useful conclusions. For simplicity he assumes constant marginal utility of income, (u´(y)> 0 and u´´(y) = 0).
When the number of working hours is held fixed, eq. (12) implies that the optimal marginal product of an additional worker in a low demand period,
1
N
G , varies inversely with the price, p1.Therefore, a decrease/increase in product demand leads to
decrease/increase in the size of the labor force. The magnitude of this change in N1
depends on the expression inside the brackets of eq.(12), that is, the unemployment insurance subsidy, ) 1 ( ] ) 1 ( ) 1 [( y y b t b e t t J − − − −
= , and the value of leisure,
) ( ' ) 1 ( ) ( ) 0 ( ) ( 1 1 y u t h v v h L y − −
= . The value of the UI subsidy is the remaining part of the UI benefits, )
1
on the other side is the difference between the utility of being laid off and the utility of working h1 hours under reduced product demand.
An observation made by totally differentiating eq. (12) with respect to N1 and the
subsidy J, is that a change in the UI subsidy size has an opposite effect on employment; 0 1 1 1 < = NN G p dJ dN
. With other words, an increase/decrease in the degree of experience rating leads to increase/decrease in employment. Note that here the working hours are still not allowed to vary.
When both labor force and working hours are allowed to vary, the effect of experience rating on employment can be evaluated by total differentiation of eq. (12) and eq.(13). Feldstein (1976) excludes eq.(14) from this particular analysis, since under the assumption of constant marginal utility of income, Gh0 (the optimal working hours during a period of normal product demand) is not affected by the degree of experience rating, while this is not true for Gh1(the optimal working hours during a period of reduced product demand), because that variable is indirectly affected by the UI subsidy,
J, via N1. The results of the differentiation are summarized by ∂N1/∂J <0and 0
/ 1 ∂ >
∂h J , which re-confirm the finding that an increase in the unemployment insurance subsidy decreases employment and show that higher subsidies increase the number of working hours per employee, implying that the imperfect experience rated system encourages firms to layoff workers rather than to reduce hours per worker.
A final remark made by the author is that the UI subsidy has the same negative effect on employment even when there is no change in the demand (p0 = p1),
since∂N1/∂J <0.
Feldstein’s theory became fast popular and was prevailing for several years in the labor market economy field. In a study by Burdett and Hool (1983) though, the results’ general validity was put under doubt. The corner stone was the assumption made by Feldstein (1976) about employees’ risk neutrality. The authors use analytical tools to approach the question. They view the implicit contract as a dual problem, that of the firm and the workers simultaneously, and with the help of isoprofit and indifference curves construct a set of optimal contracts (contract curve), one of which will be chosen out. The employees’ attitude towards risk and temporary layoffs (Z) appears to matter a lot. First by maximizing employees’ expected utility and by keeping Π*=0, as in Feldstein (1976), in the most common scenario, when working is preferred to being laid off (Z>0) and when letting workers to be risk averse, Burdett and Hool (1983) show that an increase in experience rating might actually lead to an increase layoffs:
e
N ∂
∂ 1/ > 0, if Z ≤ 0 or workers are risk neutral, when Π*=0 e
N ∂
∂ 1/ = ?, if Z > 0 and workers are risk averse, when Π*=0
When looking at the other case, maximizing employer’s expected profit and maintainingU*=0, ∂N1/∂e>0 regardless of the attitude towards risk. Combining the two cases to estimate the dual solution, Burdett and Hool (1983) conclude that greater experience rating increases employment.
Presenting the set of Burdett and Wright (1989) model, I’m keeping Feldstein’s notation to ease the reader. There are N identical attached workers and the contract specifies the employment of Nj workers, where Nj < N and j=0,1 depending on whether
the product demand is normal or low respectively. The production function f(.) satisfies f
´ > 0 and f ´´< 0. Revenue is denoted byRj(Nj)= pjf(Nj). The probability to be laid off is (N – Nj)/N. Each attached worker has a utility function u(yj,h), where yj is the
income he/she receives and works either one unit of time, h=1, or not at all, h=0. The employer’s tax bill is t =eb(N−Nj)+δN +T, where e is the marginal tax cost of laying off a worker i.e. the degree of experience rating, b is an UI benefit rate, Nδ is the payroll tax and T is a lump-sum tax.
_
U here is the reservation level of a contract C. Thus the firm’s expected profit function and the worker’s expected utility look like:
] ) ( ) ( [ ] ) ( ) ( )[ 1 ( 1 1 1 1 1 0 0 0 0 0 T N N N eb N y N R T N N N eb N y N R − − − − − + − − − − − − = Π δ λ δ λ (15) )] 0 , ( ) ( ) 1 , ( [ )] 0 , ( ) ( ) 1 , ( [ 1 1 1 1 0 0 0 N u y N N u b N b u N N y u N N U = −λ + − + λ + − . (16)
The maximization problem consists of finding the optimal firm size-contract pair solution by maximizing firm’s profit function subject to the constraints
_
U
U ≥ and Nj ≤ N. The
Lagrangian and the following first-order conditions become:
0 )] 0 , ( ) 1 , ( [ ) ( ' ) 1 ( 0 0 0 0 0 0 = − − + + − − = ∂ ∂ β µ λ N b u y u eb y N R N L , (19) 0 )] 0 , ( ) 1 , ( [ ) ( ' 1 1 1 1 1 1 = − − + + − = ∂ ∂ β µ λ N b u y u eb y N R N L , (20) 0 ) 1 ]( ) 1 , ( [ 0 0 0 = − − = ∂ ∂ N N N y u y L
µ
λ
, (21) 0 ] ) 1 , ( [ 1 1 1 = − = ∂ ∂ N N N y u y Lµ
λ
, (22) 0 / _ = − = ∂ ∂L µ U U , (23) 0 /∂ = − ≥ ∂L βj N Nj . (24) The authors show that when product demand is normal and in the case of an optimal ‘labor contract’ (when employees prefer work to temporary layoff, i.e. h=1), there is a full employment (N0 = N). β0 expressed from eq. (19) is plugged into eq. (18),1
β
is set to zero and y1 = y2 = y from (21) and (22). Solving the resulting system by totaldifferentiation yields the following outcomes:
N fixed Exogenously N=N0>N1 N determined endogenously N=N0>N1 N determined endogenously N>N0>N1 dN0/de 0 (N0 fixed) _ _ dN1/de + ? _ dl/de _ _ _
dM/de + - (if f´´ is increasing) - (if f´´ is increasing)
When letting firm size to adjust, higher degree of experience rating leads to decrease in the optimal firm size, since firms get more reluctant to hire people due to the potential layoff costs.
Greater experience rating does in fact reduce the number of layoffs. dl/de<0 in all three cases.
When product demand is low, Feldstein’s model predicts that higher experience rating (lower unemployment insurance subsidy) leads to increase in employment, whereas dN1/de is ambiguous in the “labor contract” case and negative in the “leisure
contract” case. This outcome puts a lot of doubt on Feldstein’s (1976) result concerning employment during low demand period.
The main prediction of positive relationship between the degree of experience rating and the average employment level, when firm size is fixed, reverses when Burdett and Wright (1989) relaxes the N fixed assumption. They need to assume though that f
´´(.) is increasing so that dM/de < 0 in all cases. Therefore, “under some reasonable conditions the model implies that greater experience rating .... will reduce average employment, because the desire to reduce layoffs is partially realized via fewer hires.”
The importance of the endogenous firm size model is studied also when work hours per person can vary so that the assumption of either working full-time or not at all is relaxed. In a case of working less than full-time, the UI pays partial benefits. Burdett and Wright (1989b)4 show that when short-time compensation is allowed, higher degree of experience rating reduces optimal firm size, increases hours per worker and decreases total man-hours, while N fixed models predict increase in total man-hours.
The authors point out that the unemployment caused by higher UI taxes is most sensible in the short-run, since the unattached workers might get soon employed by another firm. Another remark is that an increase in unattached unemployment is not necessarily bad, since it could be so that there was an excessive employment from the beginning.
4
Using the same framework as in Burdett and Wright (1989), but with the exception that the UI budget constraint is taken in consideration, Fath and Fuest (2002) re-examines the employment effects of experience rating in a standard implicit contract model with endogenous firm size. Excluding work sharing possibilities, setting the maximization problem in an identical way, therefore obtaining the same first-order conditions to derive optimal y1 = y2 = y, assuming that the employment constraint is
binding when normal product demand (N0 = N ) and constant reservation utility,
_
U , Fath and Fuest (2002) claim that by introducing tax on layoffs, the contributions to the UI system and its budget constraint should change and therefore the constraint needs to be taken into account. That is, given the level of UI benefits, b, an increase in the degree of experience rating may not imply higher adjustment costs in the future, since the revenue obtained could be used by the UI system to decrease firms’ payroll tax, for example. First, by neglecting the UI budget constraint and by letting the revenue, result of grater experience rating, be neutralized by lump sum taxes and transfers as in Burdett and Wright (1989), Fath and Fuest (2002) confirm their results that experience rated system reduces the number of temporary layoffs (d(N −N1)/de<0) and firms’ attaches labor force (dN/de<0). Afterwards, when introducing the UI constraint into the maximization problem (I am still using Feldstein (1976) notation),
e N N N N y b N N ) [(1 ) ] ( ) ( − 1 =
δ
−λ
+λ
1 +λ
− 1λ
eq.(25), and by assuming that the payroll tax adjusts so that the benefits paid to unemployed are always equal to the revenue of the unemployment insurance system, the authors find that
0 0 1 > = dR de dN , ? 0 = = dR de dN and ( ) ? 0 1 = − = dR de N N d
. Thus, a revenue neutral change in e, has positive effect on the employment during low demand and an ambiguous effect on the attached labor force and the number of temporary layoffs. Higher experience rating increases average employment when labor productivity during low demand is significantly different compared to productivity during normal demand period.
even increase employment in the long-run, which despite the slightly different approach, kind of supports Feldstein (1976) and opposes Burdett and Wright (1989) conclusions.
Empirical Research
Feldstein (1978) is an empirical study standing out as the first econometric research about the effects of UI on temporary layoff unemployment. Unlike the majority of preceding works, Feldstein (1978) is not dealing with the average unemployment duration or the total unemployment level, but with the amount of temporary layoffs. It uses data from the Current Population Survey (CPS, March 1971) on 24 545 individuals between 25 and 55 years old, including only those who are in the labor force, who aren’t out of risk being laid off and who are not working seasonally. Applying the ordinary least-square method, Feldstein (1978) runs several regressions relating the explanatory binary variable for temporary layoff unemployment to the rest of the variables: benefit replacement ratio(s) (the ratio of potential UI benefits to previous after-tax earnings), union membership, age, sex, marital status, race, wage rate and industry-occupation variable. Recognized weaknesses of the model are the lack of information on firms’ degree of experience rating and the availability of cash transfers or fringe benefits.
suggests that variations in BEN < 0.5 have small effect on the layoff rates, and significant effect for benefit replacement ratio larger than 0.5 .
Intuitively, temporary layoffs should be more probable for union members, since unionized firms’ workers believe that in case of a temporary layoff they will be shortly recalled, assured by the union contract, and employers are also more willing to lay off people knowing that they will return when recalled. After studying both, the whole sample and the union members separately, Feldstein (1978) proves that the temporary layoff rate of union members is twice as much the rate of the entire sample and three times as much the rate of nonunion workers.
This empirical study shows that the unemployment insurance system at the time, encourages temporary layoff unemployment, but wouldn’t that be the case of any UI? One can predict that with no UI, employers wouldn’t consider laying off their workers during periods of low demand, fearing they might find another job meanwhile. Not to offer unemployment insurance is not a solution to the temporary unemployment problem, since those who are changing job will loose their financial support during the search process which will probably cause inefficient outcomes. As Feldstein suggests, perfect experience rating will prevent firms from distortionary layoff behavior, but it will also distort hiring and permanent layoff decisions. A possible solution to the problem, which he suggests, is to have a “ one-month perfect experience rating” meaning full experience rating only for the UI benefits paid during some initial short period of the unemployment, so that the firms loose their incentives for temporary layoffs without influencing hiring and permanent layoff decisions.
In the same line of reasoning Topel (1983) claims that Feldstein’s results don’t infer anything about the effect of experience rating on unemployment. In a survey paper with Welch from 1980, they claim that the first one to do this by explaining firms’ layoff behavior via experience rating is Brechling (1979). The main results of that study are summarized as follows: an increase in the lower limit for experience rating increases flows into layoff; extending the range of rated unemployment rates reduces layoff flows and increasing the range of layoff rates with a lower payback ratio, increases the layoff rate. This early study, according to Topel and Welch (1980), showed the way towards further investigation of tax incentives in this context.
Some years later Topel (1983) conducts a study, where by measuring the extend of subsidization of UI across states, he estimates the effect of incomplete experience rating on the transitions from work to layoff and from temporary layoff to work. The sample in use consists of 8280 full-time full-year employed individuals between 20 and 65 years old, 6.7 percent of which are on temporary layoffs. The data represents 29 industries and 19 reserve-and-benefit ratio systems and is collected from the CPS report for March 1975.
Theory predicts that UI subsidy increases unemployment by changing the frequency and the duration of the layoff spells, and that the higher the subsidy, the more the layoffs. Topel’s approach to estimate the UI effect on transitions from and to unemployment uses information on employment status and duration of current unemployment spell. The main variables in the model are the benefit replacement ratio5 and the amount of subsidized benefits6, and their coefficients are separately studied in two different specifications of the model. The average benefit replacement ratio is computed to be 0.56 first by estimating potential benefits using available personal and state-law information, as a fraction of the after tax income. Regarding the calculation of 5 ) 1 ( i i i t w b ratio −
= , where b is the benefit level, w is the wage and t is the income tax
6 − − = e t w b subsidy i i i 1
subsidized benefits per person, the benefit level, the wage, the income tax rate and the individual degree of experience rating are used. The individual degree of experience rating is the only source of variation and it is a function of the equilibrium unemployment rate and can be estimated by using tax tables and the respective formulas described in a previous section. Apart from the subsidy and the replacement ratio variables the model controls also for wage, age, education, race, sex and industry effects. As mentioned before the model has two specifications, the first one uses the replacement ratio to measure the effects ignoring experience rating information, as in Feldstein (1978), while the second one focuses on the experience rating effects by using a measure of the UI subsidy as proportion of weekly earnings (on average 31 percent). Each of the specifications reports separate estimates on entering layoff, on leaving layoff and on the total effect, which is the difference of the first two.
layoffs, especially in industries with high unemployment, where the experience rating is more incomplete and the subsidy is larger. The results of Topel (1983) therefore imply that just by increasing the degree of experience rating, without changing UI benefit levels, a large reduction in temporary layoffs can be accomplished.
A consequent study done by the same author, Topel (1984), conducted in almost identical way and using data for the period 1973-1976, reaches the same conclusion: subsidies, in particular, have powerful effect on unemployment and just by reducing them with 33 percent and keeping benefits on the same level, layoff unemployment decreases from 3.4 to 3.1 percent.
Another empirical study on the topic of experience rating effect on unemployment is Anderson and Meyer (2000). That research takes advantage of the events in Washington State, where after 13 years of fixed UI tax rate for all employers, the state adopted an experience-rated system from the beginning of 1985. This change was a result of federal legislation requiring all states to have a tax rate of at least 5.4 percent. Such rate would have overtaxed most of the firms in Washington state, therefore a state law introduced the non-uniform tax rate. The authors conduct a natural experiment, expecting different taxes across firms in the same market to have effect on the employment level, since the tax differences cannot be shifted to changes in wages or prices.
The theoretical framework in Anderson and Meyer (2000) builds up on the adjustment cost and the compensation package approaches7, suggesting that an increase in the tax cost of additional layoff (higher degree of experience rating) leads to fewer annual amount of temporary layoffs and ambiguous overall effect on the employment level, since the cyclicality of the annual employment is expected to decrease. Using data for the 25-year period from 1972 to 1997 on all US states’ employment ratios collected from the Bureau of Labor Statistics and population data obtained from the Census Bureau for individuals of age between 18 and 64 years old, the authors conduct two different
7
The adjustment cost approach treats the unemployment insurance as an adjustment cost for the firm, since by paying fraction of the UI benefits for its ex-workers, the firm’s cost of adjusting the level of
econometric approaches, one of which is applying the difference-in-difference technique8. The forced introduction of experience rating in Washington State makes it easier to argue that the tax differences which occurred after 1985 are exogenous.
The constructed outcome variables are monthly employment/population ratio, annual range of monthly employment/population ratio and log of these two. All states apart from Washington State are used as control group. There are two control variables for which the Anderson and Meyer (2000) check separately: an interaction variable of state dummies with the log of the US unemployment rate and an interaction variable of state dummies with the log of the state unemployment rate. Three sets of results are produced, so that each state’s response to the overall business cycle is checked. The first model is not checking for any of the two control variables. The second uses the first control variable and the third one uses the second control variable.
Anderson and Meyer (2000) first perform a difference-in-difference analysis in two steps. The first one is a regression of one of the constructed outcome variables above on a dummy variable for state , on the interaction of dummy variables for state and the time period after the introduction of experience rating, 1985 to 1997, and on one or none of the two control variables depending on the specification. After this first step, 51 estimates are obtained, representing each state’s change in the particular dependent variable from before and after 1985. The second regression is a weighted least square9 of these estimates on a dummy variable for Washington State. The coefficient obtained is the corrected difference-in-difference estimate of the effect of experience rating on the outcome variables (monthly employment/population ratio, annual range of monthly employment/population ratio, log of monthly employment and annual range of the log of monthly employment).
The alternative econometric approach compares Washington state with the neighboring states of Oregon and Idaho by running regressions of each of the outcome
8
Difference in differences is a technique used in econometrics that measures the effect of a treatment at a given period in time. It is often used to measure the change induced by a particular treatment or event.
9
variables on the average marginal tax cost10 in the state at that time, on state dummies, a dummy variable for the time period after the introduction of experience rating, 1985 to 1997, and/or one of the control variables presented above. Due to the uniform tax rate in Washington State up to 1985, the marginal tax cost appeared to be zero for all firms in the state, while the average marginal tax costs in Oregon and Idaho were positive for the whole period. The advantages of this alternative method which Anderson and Meyer conduct are that both, the continuity of the average marginal tax cost variable and the similarities in the industry distribution between Oregon, Idaho and Washington, could give better estimates for the effect of experience rated UI tax on employment than the difference-in-difference estimators.
The estimates of the effect of experience rating on employment are mixed depending on which control variable has been used in the regression. The theoretical prediction that the range of cyclical employment variation over time is expected to decrease is confirmed. Using interaction of state dummies with the log of the state unemployment rate as a control variable, there is a negative effect on the annual range of monthly employment/population ratio of 11 and 29 percent with the difference-in-difference and the alternative approach respectively. One should note though that the results obtained by doing difference-in-difference are not statistically significant. Employers in Washington appear to have become less likely to lay off workers than were their counterparts in Oregon and Idaho after the reform on 1985. Overall, the results suggest that experience rating stabilizes employment and thus lowers unemployment.
As Anderson and Meyer conclude, employers have an interest of keeping more workers attaches than they need and to “cycle” them through spells of lower demand, as a result of the subsidized UI. Card and Levine (1994) argue further that the effects of imperfect experience rating vary with the state of demand, where more temporary layoffs are generated in a trough of the business cycle (recession), and fewer layoffs are generated at the peak of the cycle (boom). In their empirical work they measure the effect of imperfect experience rating during the cyclical downturn in the beginning of the 1980s
10
and the expansionary periods before and after, 1979 and 1987 respectively, as well as in different months of the year for seasonal effects, by combining CPS micro data for 1979 – 1987 and a database of US unemployment insurance tax rates. The sample comprises data for 180000 individuals in 36 states and 5 industries (construction, durable/nondurable manufacturing, trade and services). To solve potential endogenity problems, where high unemployment industries are not experience rated at all, the experience-rating effects are measured within these 5 main industry branches. When the degree of experience rating increases, Card and Levine anticipate lower layoff unemployment during state of low demand. Regarding the effect on unemployment during cyclical peak, the expectations were uncertain due to the varying number of firms experiencing a peak during the business cycle.
Conclusion
one of the reasons for the observed outward shift of the Phillips curve and higher average unemployment rate at that time. Economists have been proposing different ways to reduce the adverse effects while still assisting unemployed people. The most common and one of the modest proposals suggests widening the tax rate range by lowering the minimum tax rate and increasing the maximum tax rate, so that employers’ tax liabilities reflect their layoff experience even better. A more extreme proposal, made by Robert Topel, is to experience-rate individual workers so that workers with a history of long unemployment spells pay higher tax rates11. The federal government adopted the Tax Reform Act in 1986, according to which unemployment benefits are taxed just like regular income, which is a small but serious step towards more complete experience rated system.
Eveline Burns (1944) outlines three stages of social insurance evolution, where each consequent stage is characterized with higher level of social responsibility and lower degree of individual responsibility. Back in 1960 Joseph M. Becker calls the UI system controversial and places the United States of that period in stage two approaching stage three, but both Joseph Becker and Eveline Burns express doubt in that if the step towards the third stage could be ever done. Today, roughly half a century later, a huge scientific progress in the field of the unemployment insurance has been made, but based on my research, I believe that the nature of the unemployment insurance is still controversial and we are most probably still in the last phase of stage two.
11
References:
Anderson, P. and Meyer, B., “The Effects of Unemployment Insurance Payroll Tax on Wages, Employment, Claims and Denials,” Journal of Public Economics 2000, 78, 81-106.
Azariadis, C., “Implicit Contracts and Underemployment Equilibruia,” Journal of Public
Economics 1975, 83, 1183-1202.
Baily, M. N., “Wages and Employment under Uncertain Demand,” Review of Economic
Studies 1974, 41, 37-50.
Becker, Joseph M., “Twenty-Five Years of Unemployment Insurance: An Experiement in Competitive Collectivism,” Political Science Quaterly 1960, 75, 481-499.
Blough, S., “A Simple Exposition of Some Basic Results of Implicit Contract Theory,”
Economic Theory 1991, 1, 293-297.
Brechling, Frank, “Layoffs and Unemployment Insurance,” Unpublished Working Paper, Center for Naval Analyses, 1979.
Burdett, K. and Hool, B., “Layoffs, Wages and Unemployment Insurance,: Journal of
Public Economics 1983, 21, 325-357.
Burdett, K. and Wright, R.: Optimal Firm Size, Taxes and Unemployment. Journal of
Public Economics 1989, 39, 275-287.
Burdett, K. And Wright, R., “Unemployment Insurance and Short-time Compensation: the Effects on layoffs, hours per worker, and wages,” Journal of Political Economy 1989,
39, 275-287.
Burns, Eveline M., “Social Insurance in Evolution,” American Economic Review 1944,
34, 199-211.
Card, D. and Levine, Ph., “Unemployment Insurance Taxes and the Cyclical and Seasonal Properties of Unemployment,” Journal of Public Economics 1994, 53, 1-29. Fath, J. and Fuest, C., “Temporary Layoffs and Unemployment Insurance: Is Experience Rating Desirable?” Center for Economic Studies & Ifo Institute for Economic Research 2002, Working Paper No. 663(4).
Feldstein, M., “Temporary Layoffs in the Theory of Unemployment,” Journal of
Feldstein, M., “The Effect of Unemployment Insurance on Temporary Layoff Unemployment,” American Economic Review 1978, 68, 834-846.
Topel, R., “On Layoffs and Unemployment Insurance,” American Economic Review 1983, 73, 541-559.
Topel, R., “Experience Rating of Unemployment Insurance and the Incidence of Unemployment,” Journal of Law and Economics 1984, 27, 61-90.
Topel, R. and Welch, F., “Unemployment Insurance: Survey and Extensions,”
