How should job displacement wage losses be insured? Updated

Severance pay is the most common approach

Severance benefits are mandated in most countries and supplied voluntarily by many firms in countries that do not mandate them. Severance pay offers scheduled benefits based on expected losses and its efficiency depends crucially on the quality of the earnings loss forecast. Both mandated and voluntary severance benefits are often linked to job tenure, reflecting a common finding in the literature that job displacement earnings losses increase with job tenure, indeed almost linearly. The welfare value of a common severance benefit schedule of "x weeks of pay for each year of service" is transparent [4].

Unfortunately, a benefit algorithm that calculates benefits based solely on pre-displacement wages and years of service generally results in a modest fit to earnings losses at the individual level. Benefits that match losses only on average will undercompensate some displaced workers (those with unusually large wage losses) while overcompensating others. However, a benefit algorithm so good that it perfectly matches individual earnings losses would be equivalent to actual-loss insurance with no moral hazard distortions.

The limits of fixed benefits are apparent in the wage data; wages are noisy. That is apparent in the distribution of wage changes across types of separation (or no separation at all), as shown in Figure 1. Many workers who quit voluntarily accept wage losses, presumably because of the offsetting attributes of their new job. The distribution of wage changes over the course of one year for young workers in the US is quite large for displaced workers and voluntary leavers alike, and even among job stayers. From 1981 to 1982, almost 20% of stayers suffered a year-to-year real wage loss of 10% or more, which is certainly lower than the 34% of layoffs who suffered a similar wage loss, but less dramatically different than might have been imagined. The percentage of young quitters who suffer real wage declines of at least 10% is only slightly lower than that of young re-employed laid off workers, 29% versus 34%, despite the fact that these job separations are voluntary.

Improvement in loss forecasting and benefit schedules is conceptually possible, if difficult to implement. A review of the wage loss literature identifies covariates other than job tenure that might improve the efficiency of scheduled payments. It is well known, for example, that wage losses are on average higher in a deep recession [5], [6]. Indeed, in good times, average losses may actually be negative (a wage gain)-average re-employment wages of displaced workers may exceed pre-displacement wages, as shown in Figure 2. However, even in those years, large numbers of displaced workers experience large earnings losses.

Almost all workers are covered in mandated severance programs, but coverage is a problem in voluntary severance plans [3]. In the US, for example, only about one-quarter of the workforce is covered by severance plans. Incorporating severance into the unemployment insurance (UI) system would seem an inexpensive way to expand coverage if policymakers accepted the importance of this insurance instrument.

The financing of severance benefits may explain the reluctance to vary benefits with business conditions. Both mandated and voluntary programs are employer funded, usually on a pay-as-you-go basis. The last-in, first-out (LIFO) principle of many layoff protocols causes the value of payouts to increase at a growing rate with the severity of the cutback. Clearly, additional payouts in bad times may strain firm finances.

Although severance costs may be substantial to a firm when it must make a major reduction in its workforce, the expected cost of severance plans at the time of hire is relatively modest. Many workers never qualify for severance benefits. For example, workers who leave voluntarily (quit) to seek better opportunities elsewhere are normally not eligible for benefits. Similarly, older workers who retire do not generally qualify for severance benefits. Moreover, many workers who do receive benefits receive only small payments or none at all if there is a minimum service requirement. Under the usual LIFO rule, long-tenured workers do not often experience permanent job loss. Large severance payouts require decades of relatively stable employment punctuated by a precipitous drop.

There are a host of potential moral hazards to consider in any wage-loss insurance scheme. In the case of severance pay schemes, firing cost problems, perhaps more usefully labeled layoff moral hazard, may arise, because payouts at the time of displacement are substantial. The firm is the only fully reliable source of information on its own economic distress, and, if separation benefits are firm-financed it may choose to conceal low demand, retaining workers when it would actually be more efficient to release them. The practical importance of this concern remains under investigation, but the empirical record appears to reject claims of broader distortions of employment and unemployment [7].

One study argues that attention must be paid to the form of the severance benefit payout [3]. Displaced workers appear to consume benefits quickly, which argues for extended periodic payouts, not a single lump sum, if government interest in consumption smoothing is to be met.

Actual-wage insurance proposals

Actual-wage insurance is primarily known through proposals for programs. The primary proposal, which has been labeled the Brookings proposals [8], involves benefits linked to 50% of wage losses for fully employed workers for a relatively short period of time: two years [9]. This encourages a rapid return to work.

The absence of actual-loss wage insurance, despite its policy appeal, suggests that moral hazard and other concerns are large for this insurance instrument. Such concerns are not hard to imagine. Search moral hazard is an obvious problem. Finding a well-paid job likely involves more demanding search activities than finding a low-paying one, so the insured displaced worker is likely to engage in less vigorous search. Indeed, if fully insured, the worker would simply accept the first job offer, whatever its quality (wage).

Less familiar moral hazard problems emerge in actual-loss plans, including distortions in job choices by workers and in the provisions of jobs of various types by firms. For example, wages are only one dimension of a job; other dimensions include required effort, occupational safety, training opportunities, and fringe benefits. An extreme example might be labeled the "volunteer problem." Individuals often work long hours for almost no pay for causes that they care about. If the wage insurance plan compensated displaced workers for half their wage losses, token wage payments would qualify volunteers for half pay from the wage insurance systems. Less extreme examples are more common.

Insurance may also affect the attributes of jobs that firms offer. In particular, firms have an incentive to shift their compensation packages toward non-wage aspects of the job from direct wage payments if the displaced are a significant proportion of their new hire pools. Non-wage elements of the compensation package might include better health insurance, improved job training, and reduced work intensity. Any compensating differential wage adjustments would be partly subsidized by the wage insurance program.

To curb such moral hazard distortions, proposed actual-loss plans usually have narrow ambitions, sharply limiting program generosity, which of course reduces the insurance value of the plans. Common restrictions in proposed wage insurance plans [8], [9] include: permitting wage gainers to keep all their gain (full insurance among the displaced would call for pooling wage gains among winners and losers); and limiting the subsidy to 50% of wage losses.

A practical difficulty with actual-loss wage insurance is that the impact of a given displacement on wages grows more uncertain as wage shocks of all types hit the worker. It is not possible to imagine that the difference between the worker's actual wage and predicted wage ten years in the future reasonably measures the impact of that single event. As a consequence, most actual-wage loss proposals impose severe limits on the duration of wage losses, for example: limiting the period of loss recovery to two years.

Additional features often found in proposed plans include re-employment bonuses; greater program benefits if the displaced worker finds their next job more quickly. For example, the two-year limit on loss recovery raises the question, when does the two years begin? If the two years begins at the date of displacement, aggregate plan benefits fall with longer unemployment spells. A displaced worker who is immediately re-employed would receive benefits for the full two years, while a displaced worker who is re-employed after six months would recover only 18 months of lost wages. The re-employment literature gives little reason to believe that this refinement is of serious value [10].

Plan costs may also be limited by excluding high-wage displaced workers, where, of course, "high-wage" would be defined by the plan. For example, the program could be limited to low-earning individuals, say those making US$50,000 or less after displacement. However, there is no evidence that low-wage workers suffer higher proportional wage losses than higher-wage workers following job displacement, which makes the rationale for this feature unpersuasive. However, the possibility of fraud no doubt grows with the size of the individual payout.

A serious concern with implementing an actual-wage insurance plan is the lack of real-world experience with such plans [8]. Two small pieces of evidence have emerged from the Trade Adjustment Assistance (TAA) literature. TAA is essentially actual-wage insurance with the additional condition that job displacement must arise from import penetration [11]. The latter is hard to establish with any precision [12], but the practical operation following that determination is the same as for any displacement.

One large Canadian demonstration project assesses the distortions generated by the (limited generosity) actual-wage plan described previously [13], and the results are reassuring, if somewhat surprising. The investigators undertook a project that offered displaced workers wage insurance that was more generous than the basic actual-loss plan, covering 75% of wage losses for full-time workers. Although the study focuses on the assessment of the re-employment bonus feature of actual-wage plans (there were no effects), the results bear directly and importantly on a variety of potential distortions in actual-loss plans. The authors find few behavioral distortions of any sort, implying that moral hazard distortions may not be substantial. However, take-up rates were low, about 20% of the trade displaced. Those who qualified for small payments chose not to enroll and, more importantly, many displaced workers who otherwise qualified for benefits did not secure another full-time job in the required time (two years).

Actual-wage insurance was introduced into US TAA legislation in 2002, though the focus remained on job training and extended unemployment benefits during the training period. In 2009, greater emphasis was put on wage insurance, which was offered as an alternative to the training/extended UI option [14]. As with the Canadian study, the wage insurance take-up rate was low. Among those who did take up wage insurance, post displacement employment and earnings behavior of recipients varied only modestly from the behavior of those in the traditional track, and not at all in the long term. A demonstration project would presumably be required to determine the impact of a broader displacement program, which did not already have the TAA base support.

Alternative models of wage loss insurance

In addition to insurance plans, severance savings plans are relatively common (see [15] for specific examples). These are typically extensions of pension plans, which, if generous enough, can cover wage losses without threatening retirement well-being. The advantage of such saving plans is that workers have little incentive to game the system, because resources are the workers' own. Search moral hazard in particular might not be a problem under these schemes, although all savings schemes transfer losses only across time for the same individual, and not across individuals in various loss statuses.

The most common actual-wage insurance proposal [8], [9] has been discussed for 25 years without being adopted, which at least hints that there is a problem with the basic design of the proposed plan. For example, the greatest threat to workers' economic security comes from the persistence of wage losses among the displaced. Initial wage losses of voluntary job quitters or even job stayers may be almost as substantial, but presumably more ephemeral (Figure 1). Perhaps "catastrophic" wage loss insurance would be more valuable to the worker. Loss coverage under current proposals is front-loaded, often only covering the first two years of losses. If the primary objective is to compensate workers for large losses an alternative plan would be to cover wage losses only later, in say post-displacement years three through six. Under this type of back-loaded insurance, targeting would be improved as resources would be directed to those most harmed by the displacement.

There may be other gains to back-loaded benefits, such that moral hazard problems are likely to be eased. The adverse impact of the wage subsidy on a person's search for a better paying job may be reduced if displaced workers know they will face wage losses for one or more years before receiving wage loss compensation. The program could be integrated with the UI plan to limit especially large income losses between the end of the UI program and the start of the wage insurance program.