The U.S. unemployment insurance system needs an overhaul. Since 1935, unemployment insurance (UI) has operated as the primary safety net for workers who lose their jobs through no fault of their own. But changes in work organization, increases in the number of women in the workplace, and growing concern for work/family balance have not been matched by changes in the way workers qualify for UI benefits. Over time the system has become less effective, with fewer than half of the nation’s unemployed even applying for benefits.

Part of the problem stems from the fact that each state establishes its own rules governing UI policy, generosity, eligibility, and revenue. Consequently, there is great variation between the 51 different UI systems that exist throughout the country. While the development of these state-level programs is overseen by the federal government, a wide range of practices-some much better than others-have evolved over the years.

In this analysis we focus on both the policy changes that have occurred throughout the 1990s and the state-to-state differences in UI policy (details by state are presented in the appendix). Special attention is given to those factors that are likely to affect low-wage workers, part-time workers, and workers with interrupted or limited labor market experience. For the 1990-2000 period, we analyze changes in eligibility requirements, benefit amounts, and the percentage of unemployed workers applying for benefits, referred to as insured unemployment.

Our general findings are that:

• Variations in the state-level rules and policies governing the UI system lead to inequitable treatment of workers from one state to the next, despite similar work and earnings histories.
• Eligibility for UI benefits is often too stringent for the lowest-income workers, many of whom fail to qualify for benefits at all. States with UI eligibility based on earnings (rather than hours) are unfair to low-income workers.
• Benefit amounts, especially for low-income workers, are below the poverty level in many states. Policy makers should raise the benefit amounts for workers at the bottom of the earnings distribution by raising the percentage of their wages replaced by the UI system.
• The percentage of unemployed workers who apply for and receive UI benefits continued to decline over the 1990s, largely as a result of lower benefits.
• Raising the minimum wage helps workers qualify for benefits and raises their UI benefit amounts. Increases in the minimum wage are associated with higher UI take-up rates. In fact, a $1 increase in the minimum wage raises the percentage of workers filing unemployment insurance claims by 1.9%.

We also find that, while most states made it easier to qualify for benefits and increased their benefit maximums, a smaller percentage of unemployed workers applied for and received those benefits. This contradiction is explained, in part, by the reduction in union employment and the migration of manufacturing from high-benefit states to low-benefit states. Lower levels of unionization raise the likelihood of workers being unaware of their benefits, and manufacturing’s migration to states with less generous UI systems makes applying for benefits less remunerative.

State-to-state variation
A comparison of UI programs by state clearly illustrates their differences.¹ Perhaps the most telling variation is each state’s maximum allowable benefit level. Table 1 lists the five states with the lowest and highest maximum allowable benefits.

The difference in maximum weekly benefits between Massachusetts and Alabama is striking. But other factors play an important role, too, such as the disparities between eligibility requirements from state to state. All benefits are based on a worker’s employment during the base period, which is usually the first four of the last five completed quarters. To qualify for maximum benefits in Alabama, a worker had to earn a total of $9,120 in the base period, with $4,560 earned in a single quarter during that period. Additionally, the worker had to meet other requirements, including being employed in a covered industry, not being self-employed, and not having been fired for cause or have quit. In Massachusetts, on the other hand, to qualify for the maximum benefits, the worker must meet a minimum earnings requirement of $12,402 in one quarter and $14,310 in the total base period; meeting these criteria will qualify the worker to receive the maximum benefit base of $477. Massachusetts also has a provision for dependents (called a dependents allowance), which pays an additional $25 per week per child under the age of 18 up to a maximum of $215 per week.

Vroman’s (1991) research into the complexity of the unemployment insurance systems reveals that many workers find the system too daunting. The most common reason for not applying for UI benefits is that workers do not think they are eligible-nearly 53% of all non-applicants believed they were not eligible for benefits (Vroman 1991, 25). A very small 5.3% of non-applicants cited either “too much hassle” and “too much like charity” as the main reasons for not applying.

Few researchers use the differences in state-level UI policy to analyze changes in insured unemployment, perhaps because quantifying the differences between the states’ programs is so painstaking. Blank and Card (1991) and Baldwin and McHugh (1992), though, are two notable exceptions. Atkinson and Micklewright (1991) do a good job at pointing out many of the shortcomings in previous research on unemployment insurance, noting that researchers regularly make assumptions about the provision of unemployment insurance that they know to be false. For example, many researchers create analytical models which wrongly assume that benefits are paid regardless of the reason for unemployment. Other analysts assume that benefits are available for the full unemployment spell, without a waiting period. Some research does not consider that a job search is a requirement to receive benefits. Confusion about the duration and calculation of benefits abounds. Finally, most analysts assume that, if benefits were not received, then this played no part in decisions about employment or unemployment. In other words, only those who receive benefits are thought to be influenced by the provision of benefits. In every instance these assumptions are likely to be incorrect and, in most cases, provide analysts with spurious results and policy makers with distorted conclusions from which to work.

Taken in toto these assumptions draw an inaccurate portrait of the U.S. unemployment system. In the United States, benefits are not paid to those who quit or are fired with cause. Benefits are paid only after a waiting period (usually one to two weeks). Eligibility requires evidence of a current job search, and refusal of a job offer often results in ending benefit payments. Workers must have made minimal contributions to the unemployment insurance system. Benefits are paid on a sliding scale based on total earnings and are limited to a maximum of 26 weeks in most states. Because of the myriad variations between states in these areas, research that makes generalized assumptions about unemployment insurance may easily result in inappropriate policy prescriptions.

To circumvent the types of problems described by Atkinson and Micklewright (1991), we take into consideration the specifics of each state’s program, providing a more complete picture of the labor market effects of unemployment insurance.

The purpose of eligibility requirements
Unemployment insurance eligibility is built around a single premise: labor force attachment. Both monetary and non-monetary eligibility requirements focus on the worker’s attachment to the labor force. If the worker is “casually” attached to the labor force, then he or she is likely to be denied benefits. Casual attachment is determined by several factors, including minimum earnings requirements (either in a base year or in a particular quarter) and continuous employment over several consecutive quarters. Additionally there are non-monetary eligibility requirements that further restrict an employee’s ability to collect UI compensation. The three most common reasons for ineligibility are: the worker is unemployed as a result of a labor dispute (strike, walkout, etc.); the worker left the job without “good cause”; or the worker was discharged due to “misconduct.”

Although it has been argued that the complexity of the states’ current UI systems are the result of institutional or historic legacies, the eligibility rules nevertheless serve an important purpose: to separate the “deserving” from the “undeserving.”

Those seen as deserving UI benefits are thought to be those who have “paid” for them. That is, once significant contributions to the UI system have been made, the worker is eligible for benefits. In most states, the basic eligibility rules have the following components: workers must be employed for a prescribed number of weeks during the year; they must earn a prescribed amount; they must have separated from employment through no fault of their own; and they must be actively looking for work rather than leaving the labor force. These combined factors are thought to measure a worker’s labor force attachment.

Unfortunately, many of these eligibility requirements bias UI policy against low-income workers, women, and contingent laborers. Workers must earn both the minimum earnings in the base year and the minimum in the peak quarter, criteria often difficult for low-income workers to meet. For example, full-time, full-year workers earning the minimum wage ($5.15) in 2000 would qualify for UI in all 50 states and D.C., but half-time, half-year workers with the same hourly wage would fail to qualify in eight states.² Additionally, employees who quit work due to employer-initiated schedule changes are ineligible for UI in 13 states; depending on the circumstances, they may be ineligible in another 25 states. For women, these unanticipated changes in schedule can be especially problematic if they are the ones primarily responsible for child care. Workers exclusively pursuing part-time work are deemed ineligible for UI in at least 30 states, since they are considered “unavailable” for full-time work.³ Finally, temporary workers are considered ineligible in 20 states if they refuse subsequent assignments.

As previously discussed, workers who are strongly attached to the labor force are likely to be eligible for unemployment insurance. Early framers of UI law believed that workers with a strong labor force attachment would re-enter the labor force with a minimal delay, and so they wrote eligibility rules to favor these workers. And although research has consistently found that UI benefits prolong spells of unemployment (Stigler 1962; McCall 1970; Mortenson 1970), they do so only by a small amount. Moffitt and Nicholson (1982) find that an increase in the benefit replacement rate of 10 percentage points (e.g., replacing 50% of lost income instead of 40%) would result in only one additional week of job search. Some researchers question even this finding (Atkinson et al. 1984; Atkinson and Micklewright 1985). Finally, more recent research by Acemoglu and Shimer (1999) finds that a “decrease in the generosity of UI from its current level would not only decrease welfare but also reduce the level of output.” Consequently, the negative impact of UI benefits is small and questionable, especially when considering their importance as a social safety net for workers.

Apart from providing necessary income support during a time of hardship, another benefit of unemployment insurance is that it may actually improve the quality of an unemployed worker’s job search. To the extent that unemployment insurance subsidizes a job search, it may influence the quality of the job match. Hence, improvements in unemployment insurance availability and benefits levels may result in better employer/employee matches and thereby reduce subsequent employment turnover and spells of insured unemployment.

And as the number of workers in contingent work arrangements grows, unemployment insurance is the best policy available to cover the increasing flow of workers into and out of these jobs. The system is already in place, has successfully provided a safety net for other displaced workers, and requires potentially minor changes to make it more amenable to alternative work forms or contingent work. Other researchers have proposed more dramatic policy changes4 that are not likely to be enacted given their sheer scope and magnitude. The more modest proposals presented in this report address Blank’s (1998) concern that “the unavailability of unemployment insurance to self-employed contractors and to many part-time or temporary workers who work limited hours or switch jobs frequently may exacerbate the economic uncertainty associated with contingent work.” Regardless of the employment relationship-low income, temporary, part time-all workers should be equally protected by employment policies such as UI.

Variations in UI eligibility requirements
Base periods. Determining eligibility for a recently laid-off worker begins with an examination of previous employment. In most states, only the most recent labor market experience is counted in making this determination. The period used to determine eligibility is often referred to as the base period. Most states have a base period that is the first four of five completed calendar quarters. In these states only wages and hours accrued in the first four out of five completed quarters are counted toward eligibility. This means that the most recently completed quarter is not counted nor is any portion of the current quarter. For a worker who filed an unemployment claim on April 1, 2001, the base period would be January 1 through December 31, 2000. Wages earned and hours worked between January 1 and March 31, 2001 would not be used to determine eligibility. The worst case occurs when a worker is unemployed a week before the end of a quarter. In this case 25 weeks worth of earnings would be disregarded for the purposes of eligibility.⁵ For workers living in states that use a “standard” base period, between 13 and 25 weeks’ worth of earnings are not counted toward eligibility. As of 2000, 39 states use this base period system.

Calculating the base period in this way clearly hurts workers who have limited or interrupted work histories. Additionally, low-wage workers who may need all of their earnings to qualify are more likely to be ineligible for benefits. Twelve states recognize that this system penalizes workers with low earnings and have adopted alternate base periods. In these states,6 if a worker fails to qualify for unemployment insurance benefits using the standard base period then an alternate base period is used. In the alternate base period, the last four completed calendar quarters are used to determine monetary eligibility, not the first four of five completed quarters. This effectively reduces the number of weeks of earnings that is excluded. Under an alternate base period, a minimum of zero weeks to a maximum of 11 weeks are not counted toward eligibility.

Earnings requirements. To qualify for unemployment insurance benefits, workers must have earned a minimum dollar amount during their state’s base period (this is true in 49 states). Additionally many states have quarterly earnings requirements. From 1990 to 2000, 10 states and the District of Columbia raised their inflation-adjusted minimum earnings requirements.⁷ By raising their minimum real earnings requirements, these states made it harder for low-wage workers to qualify for UI benefits.

To further understand the effects of these earnings requirements, we compare changes in median wages to changes in earnings requirements for each state8 over the course of two time periods, 1990-95 and 1995-2000. In the first half of the 1990s, median real wage growth across the states was negative, while the latter half of the decade saw real wages rise (see Mishel et al. 2000 for a discussion of median wage trends). If we examine median wage changes by state,9 we see that median wages fell by 2.2% from 1990 to 1995. Fortunately for many workers, UI earnings requirements declined by 8.8%. Thus, the median earner in many states had an easier time qualifying for UI benefits in the last half of the 1990s.

There were, however, nine states that actually raised their minimum earnings requirements faster than median wages increased in the early 1990s. These states’ percentage changes, UI earnings requirements, and median wages are listed in Table 2.

Since median wages are indicative of the wage distribution,¹⁰ many workers likely had increased difficulty qualifying for UI benefits in these states. Particularly daunting were the changes in Massachusetts, Ohio, and Maine, where earnings requirements increased considerably during a time when median wages were falling.

The period from 1995 to 2000 tells a somewhat different story. During this time state median wages increased by 7.3% while UI earnings requirements declined by 1.0%. Although this is likely to make UI eligibility easier in most states, six states raised their earnings eligibility requirements faster than the median wage during the last half of the 1990s. Interestingly, none of the states that raised their earnings requirements faster than median wage growth in the 1990-95 period did so again in the 1995-2000 period.

In addition to base period requirements, another significant eligibility factor involves a worker’s quarterly earnings. Most states (33 states in 2000) have quarterly earnings criteria that require workers to earn a state-determined minimum amount in at least one quarter of the base period. For example, Florida requires total earnings of $3,400 in the base period to qualify for minimum UI benefits, with earnings of at least $2,266 in one of those quarters. In other words, a worker who qualifies for minimum benefits must earn two-thirds of his or her minimum required earnings in a single quarter. This requirement is typically known as the “high-quarter” earnings requirement.

Our analysis indicates that, in the 1990-95 period, five states raised their high-quarter earnings requirements faster than their median wage grew. From 1995 to 2000, this occurred in three states. Overall, while the high-quarter requirements present a more difficult eligibility hurdle, fewer states raised high-quarter requirements (than total earnings requirements) relative to earnings in the 1990s. Nevertheless, seven states11 and the District of Columbia had real quarterly earnings requirements that were higher in 2000 than in 1990.

Weeks and hours. While most states have minimum earnings requirements, a few states require that an individual work a specified number of weeks. Typically, only those weeks in which a minimum weekly earnings threshold is exceeded are counted as weeks worked. This system has the same intrinsic inequities as state systems that use earnings as requirements for eligibility-workers who earn less per hour must work more in order to qualify for benefits. Over time, the number of states with this requirement has declined. In 1990, nine states had minimum weeks requirements; by 2000 that number had declined to three. In essence, these weeks requirements are really earnings requirements that double as distribution requirements. For example, a worker must work 20 weeks to qualify for UI in New Jersey in 2000, but only those weeks in which a worker earned 20% of the state’s average weekly wage are counted toward fulfilling this requirement.

The only state that does not explicitly tie UI eligibility to prior earnings is Washington. In Washington, workers qualify for UI once they work 680 hours in the base period. Washington also has an alternate base period, allowing the worker to count hours of work in the current calendar quarter. While Washington does not tie earnings to eligibility requirements, the de facto earnings requirement is 680 hours at the minimum wage. With Washington state’s minimum wage currently at $6.50, to qualify for UI most workers12 must have a minimum earnings of $4,420. This is $2,020 more than Florida, the next highest state. While Washington’s eligibility requirements are simple and easy to understand, the de facto earnings requirements are the highest in the country.

Minimum wages and eligibility. There is a direct link between minimum wages and UI eligibility: increases in statutory minimum wages reduce the number of states where these workers are ineligible to receive UI benefits.

In 1989 the federal minimum wage was $3.35 (unadjusted for inflation). On four separate occasions in the 1990s the federal government increased the minimum wage until it reached $5.15 by 1997. Furthermore, throughout the 1990s, many states raised their minimum wage above the federally mandated minimum. By 2000, 10 states (including the District of Columbia) had minimum wages above the federal requirement.

In our state-level analysis we consider each state’s statutory minimum wage in determining UI eligibility (Table 3). In each case we examine the eligibility status of workers employed for 26 weeks at 20 hours per week during the base period.

In 1990, minimum wage workers working half-time for half the year failed to meet the monetary eligibility requirements in 16 states. The minimum wage increase of 1991 meant that in six more states, these low-wage workers qualified for UI benefits. Other minimum wage increases had similar effects. For example, had the minimum wage not been raised in 1996, the number of states in which these low-wage workers were ineligible to collect UI would have increased from 10 to 13. Similarly, the minimum wage increase late in 1997 reduced the number of states determining ineligibility for low-wage workers from 10 to eight in 1998.

Further evidence of the minimum wage’s effect on eligibility can be found by examining time periods when the minimum wage was not increased. From 1991 through 1996 there were no federally mandated increases in the minimum wage. During this period the number of states denying benefits to half-year, half-time minimum wage workers gradually increased. In 1991, 10 states had systems that denied these workers UI benefits. Over the next five years, the number of states increased to 15 before slowly dropping back down to 10 again, leaving half-time minimum wage workers no more eligible for UI benefits than in 1991. This trend suggests that states may respond to increases in the minimum wage by raising their UI earnings requirements.

It is worth noting that the estimates of eligibility in Table 3 may be overly generous. We assume that all of the previous earnings are “counted” when determining eligibility. In most states this is not the case; these states disallow a portion of a worker’s most current earnings. For example, in 2000, 40 states have base periods that disallow as much as three months of recent earnings. This implies that a recent labor market entrant who worked 20 weeks pr
ior to being laid-off would have 13 weeks of work not used in calculating benefits. At current minimum wage levels, even a full-time worker would fail to qualify for benefits under these circumstances in 19 states.

Eligibility and TANF recipients. Recent research indicates that previous welfare recipients are among those most likely to be affected by the minimum wage policies and alternative base periods in determining UI eligibility (Holzer 2000; Um’rani & Lovell 2000; Gustafson & Levine 1998). Many previous welfare recipients have a limited work history, making UI eligibility difficult to achieve. Even for those who work consistently the base periods adopted by many states make qualifying for UI difficult. In many cases these base periods disallow current earnings, penalizing workers whose employment history is recent, limited, or sporadic. Recipients of TANF (Temporary Assistance for Needy Families) benefits are very likely to fall into this category. Analysis of the 1993-96 Survey of Income and Program Participation (SIPP) indicates that about 5% of previous welfare recipients report receiving UI benefits. Most AFDC (Aid to Families With Dependent Children) recipients report receiving UI benefits before going on welfare. For those workers who received both welfare and UI, more than two-thirds report receiving UI prior to receiving AFDC. It is possible that these workers exhausted their unemployment insurance benefits and were then forced to apply for welfare. By contrast, only one-third of welfare recipients who received both UI and welfare report receiving unemployment benefits after receiving welfare. It is difficult to make any generalizations from such small percentages, but nevertheless, these findings suggest that qualifying for UI benefits after a spell of welfare receipt is considerably more difficult than getting UI benefits beforehand.

Recent analysis by Vroman (1998) concludes that no more than 20% of unemployed welfare recipients would be eligible for UI in a recession. This share is far below the national average of 37.4% of the unemployed applying for UI benefits. Additionally, we can find no evidence that states have made an effort to reform their UI systems in light of the new work requirements set out by the Personal Responsibility Work Opportunity Reconciliation Act in 1996. Between 1996 and 1997, 11 states increased their minimum earnings requirements necessary to qualify for UI, raising them, on average, by nearly 14%. Some states, such as Alabama, more than doubled the minimum earnings requirements necessary to qualify for UI benefits between 1996 and 1997. These increases create significant barriers to eligibility for previous welfare recipients.

Unemployment insurance benefits

Benefit levels. Once eligible for benefits, the next concern is the adequacy of the benefit amount. Typically this is measured in terms of wage replacement. Previous research on UI generosity typically examines the ratio of previous wages to UI benefits, which Vroman (1980) indicates is sensitive to the business cycle. During periods of economic recession, the pool of unemployed workers contains more high-wage workers than during times of economic expansion. As a consequence, wage replacement ratios are driven down during times of recession. Rather than examine individual wage replacement ratios, we first examine the real value of unemployment insurance benefits. We then examine the relationship between median wage growth in the state and maximum benefit amounts.

In 12 states, real maximum benefit amounts were lower in 2000 than in 1990.13 On average, maximum benefits amounts in these states fell by $23.40 per week, or by 8%. In D.C. and California, real maximum allowable benefits fell by $46.68 and $77.03, respectively.

Overall, 39 states raised their real maximum weekly benefit from 1990 to 2000. In these states the average maximum benefit increased by 15%. Leading the way were Indiana, Washington, and Massachusetts, which increased their maximum allowable benefits by $161.52, $116.89, and $105.46, respectively. Thus, most state UI systems in the 1990s were generous to middle-income workers. In fact, the median-wage worker employed full time was below the 50% wage replacement rate in only six states. This means that, in most states, workers earning the median wage had not yet reached the maximum benefit cap.

Analyzing maximum benefit amounts is a straightforward process. As benefits increase, fewer workers run into the cap set by the maximum allowable benefit. Minimum benefit amounts, though, are not as easy to analyze. Minimum benefits reflect two aspects of each state’s UI system: eligibility and generosity. For many states the minimum benefit amount is based on the minimum earnings requirement. This means that ease of eligibility is related to lower benefits. So although the benefits may be easier to get, unfortunately these low minimum benefits provide inadequate temporary income during unemployment. From 1990 to 2000, 32 states lowered their minimum benefit amounts. While this may sound problematic, it may, in fact, benefit low-income workers. Lower minimum benefit amounts usually imply easier eligibility. Since the minimum benefit amount is calculated from the minimum earnings requirements, lower weekly benefit minimums imply reduced barriers to eligibility. (Weekly benefit amounts for minimum and median wage workers in each state can be found in Appendix Table 3.)

Benefit generosity and ease of eligibility. As mentioned above, determining state generosity is a difficult task. Some states may have relatively high maximum benefit levels but make it difficult for workers to satisfy earnings requirements. Other states have relatively low earnings requirements, making it easy for an unemployed worker to qualify, but provide inadequate benefit levels. Ultimately, if a state has high benefits but difficult eligibility requirements, then the generosity of that state’s UI programs is mitigated. Conversely, states with lower eligibility requirements may be less generous. To begin to untangle this relationship, we analyze the minimum earnings requirements for eligibility and the maximum weekly benefits in the 1990s.

A nationwide analysis of the relationship between benefit levels and ease of eligibility in the U.S. indicates that there are small negative correlations between these two measures. This implies that states with lower earnings requirements tend to have higher maximum benefit levels. But such national analysis masks strong regional differences. In general, the negative relationship that we saw nationally was driven by large negative correlations in New England, the East North Central region, and the Pacific region (see Table 4).

Most of the other regional correlations are small and positive (with the exception of the Mountain region), indicating that states with lower eligibility requirements also have lower benefit levels, and vise versa. It is likely that the Middle Atlantic states have positive correlations due to generous benefits being associated with higher earnings thresholds, while the East South Central region is more likely to have lower minimum thresholds associated with lower maximum payments. A full-fledged multivariate analysis is necessary to determine which factors are driving these differences.

Declines in insured unemployment
In the 1980s, the unemployment system received considerable attention from academic and government researchers. In 1980, insured unemployment was 50%, but by the end of the 1980s, insured unemployment had fallen to 33% (see Figure A). Research by Burtless and Saks (1984), Vroman (1991), and Baldwin and McHugh (1992) suggests that a substantial portion of the decline was driven by policy changes. In particular, Baldwin and McHugh find that policy changes account for 55% of the decline in UI recipiency. By
contrast, Blank and Card (1991) find that UI policy changes had little effect on UI recipiency, and that population shifts from historically generous states to less-generous states had driven the national decline. By 1992, insured unemployment rates were on the upswing again, surpassing the 1980 level, only to drop again by another 16% by 1995. This trend raises an important question about the cyclical nature of the UI system.

These volatile national averages mask a considerable amount of state-to-state variation in insured unemployment as well. While the national average for insured unemployment was 38% in 2000, South Dakota and Georgia had rates of 18% and 20%, respectively. By contrast, Rhode Island, Massachusetts, and Alaska had insured unemployment above 62%.

Changes in national policy. One possible explanation for the volatility in insured unemployment rates is changes in UI policies at the national level. Most of the national policy changes occurred in the 1980s and served to limit the value of benefits and raise the bar on eligibility. In 1979, UI benefits became partially taxed, and in 1986 all UI benefits were subject to taxation. In 1982, the federal government also ceased making zero-interest loans to states with insolvent UI trust funds. This policy raised the costs of borrowing and effectively forced the states to reduce spending on UI.

But the 1990s saw the beginnings of UI policy changes that were designed to strengthen the program. Currently there is an effort underway to enact a Birth and Adoption Unemployment Compensation policy that would enable parents to receive UI benefits during a stint of unemployment that arises from these events. Efforts also are being made to expand UI to those workers employed fewer than 35 hours per week. Workers who search for part-time jobs exclusively currently are not eligible for UI benefits in most states. According to the National Employment Law Project, nine state have statutes that limit UI eligibility for part-time workers, nine more states have agency rules or regulations limiting UI access for these workers, and another 19 states have conflicting rules limiting the UI eligibility for part-time workers. The remaining 13 states have no statutory, agency, or regulatory policies regarding the UI eligibility of part-time workers. These results are similar to (although less encouraging than) the General Accounting Office (2000) finding that 30 states limit UI eligibility for part-time workers searching for part-time work. While nearly 20% of the U.S. labor force hold part-time jobs, it is unlikely that these workers have had much of an effect on changes in insured unemployment. This is largely due to the stability of part-time employment during the 1990s. As the share of the labor force holding part-time positions increases-as it has for several years now-the share of insured unemployment will likely decrease.

Another important and often overlooked national policy contributing to UI eligibility is the minimum wage. As discussed earlier, increases in the federal and state minimum wages resulted in increased eligibility. Future increases in the minimum wage likely will have a similar positive effect on insured unemployment.

Changes in state-level policies. Among the most important state policy changes in the 1990s are those in minimum earnings requirements and alternate base periods. The minimum earnings requirements and alternate base periods measure ease of eligibility. If eligible for UI, an unemployed worker must decide if applying for benefits is worthwhile. This decision is influenced by the expected unemployment duration and the value of UI benefits. To better understand these relationships, we examine several factors, including exhaustion rates (i.e., the percentage of UI recipients who exhaust benefits), maximum weekly benefit levels, and the percentage of wages replaced by benefits.

State policies may also influence insured unemployment in ways that go beyond statutes and regulations. In particular, a state administration can choose the level of enforcement it wishes to pursue. Enforcement varies considerably, and we suspect that enforcement stringency is loosely tied to trust fund solvency. States with larger trust funds (per worker) may be likely to have higher insured unemployment rates. Of course, states with more stringent enforcement regimes may have larger trust funds. Since this relationship is at best ambiguous and causality is difficult to determine, we do not include any measure of trust fund adequacy in the analysis.

Migration, manufacturing, and unionization. While manufacturing employment has been on the wane nationally throughout the 1990s, 24 states actually saw manufacturing employment increase from 1990 to 2000. These states were located predominately in the South, Midwest, and the West.¹⁴ These shifts in manufacturing enable us to determine the extent to which employment shifts across the country have been from states with more generous UI systems to those with less generous systems.

As for unionization rates, the percent of workers who are members of a union has fallen considerably since the 1950s. The 1990s were no different-between 1990 and 2000 unionization rates fell from 13.3% to 11.5 %. Unions typically provide important information to workers regarding their workplace rights. Because union members are more likely to be more aware of these rights and benefits, it is likely that states with higher unionization rates will have higher insured unemployment rates.

What drives insured unemployment levels? The results of our analysis,15 shown in Table 5, lend support to most of our guesses about the effects of national- and state-level policies and the changes in the manufacturing sector and unionization rates. Particularly interesting is the effect of the minimum wage-a $1 increase in the minimum wage results in a 1.6 to 1.9 percentage point increase in insured unemployment. But state policies, on the whole, yield substantively small and statistically insignificant impacts on insured unemployment levels. An increase of $1,000 in minimum earnings requirement reduced insured unemployment by only 0.18%-0.25%. The benefit replacement rate, however, has a substantively important impact, as there is a strong positive relationship between the percentage of weekly earnings replaced by UI benefits and UI uptake rates. A 1% increase in the benefit replacement ratio yields a 0.69%-0.74% increase in insured unemployment. The percent of the state’s workforce that is unionized also had the expected impact-higher unionization rates result in a higher percentage of unemployed workers filing for benefits.

Additionally, two important demographic results are revealed by the regression analysis. As the labor force became less white, the percentage of workers receiving UI benefits declined. It is difficult to explain why this is the case, but the model clearly finds that, during the 1990s, states with a lower percentage of white workers had fewer UI recipients. Interestingly, a similar phenomena occurred for women. During the 1990s, increases in the percentage of women in the labor force are related to decreased UI recipiency. Combining these effects-a 1% decline in the share of the labor force that is white and a 1% increase in the share that is female-resulted in a 1.0% decline in UI recipiency.

Our analysis also reveals two unexpected relationships (although neither is statistically significant at the 10% level). First, manufacturing seems to have a negative effect on UI recipiency rates. It is likely that Blank and Card’s (1991) mobility result explains this negative coefficient associated with manufacturing in our results. As we mentioned before, this may be explained by the migration of manufacturing jobs to stat
es with historically lower UI application rates. The second unexpected relationship was that between UI uptake rates and the alternate base periods used to determine eligibility. Easing eligibility for lower-income workers, especially recent entrants and those with intermittent employment histories, should increase the application rates for these types of workers. But the negative effect of alternate base periods on insured unemployment is not statistically significant, indicating that the effect is indeterminate.

The effects from the preferred specification (Table 5, column 2), are summarized in Table 6. These results indicate that there would be a 1.9 percentage point increase in the number of unemployed workers claiming benefits if the minimum wage were increased by $1. Additionally, we are 95% certain that this effect is at least 0.15%, and could be as large as 3.6%. Again, since minimum wage increases act to both raise eligibility and benefit levels, this effect does not seem inordinate. Similarly, the mean effect of raising the benefit-to-wage ratio by 1% is to increase the percentage of unemployed workers claiming UI benefits by 0.74%. Again, we are 95% certain that the true effect lies between 0.39% and 1.1%.

UI policy recommendations
Because so many components of a state’s UI system are intertwined, it is extremely difficult to propose an ideal program. The recommendations that follow adhere to three basic tenants: fairness, simplicity, and adequacy. Unfortunately, most states fail on these criteria. Most of the state UI systems are unfair due to the structured relationship between earnings and eligibility. In these states, workers who earn higher hourly wages may work fewer hours and still qualify for UI benefits. For example, a minimum wage ($5.15/hour) worker in Florida must work 660 hours in order to qualify for the minimum UI benefit of $32 per week. By contrast, a worker employed at $10 per hour would have to work only 340 hours to qualify for the minimum benefit. And the differences between states in this regard can be dramatic. According to the 2000 Green Book,16 a worker employed at the federal minimum wage for 40 hours per week for 52 weeks would be eligible for $216 per week in UI benefits in Connecticut but only $81 per week in California.

So the first recommendation would be to de-couple the earnings and eligibility relationship. This would improve fairness within and between the states. Washington state currently has the best system, with UI eligibility based on working 680 hours in the base period, regardless of earnings. We also would strongly encourage more states to adopt base periods that include a worker’s most recent earnings in determining eligibility. With the advent of computers, a quarter-long lag in reporting is unnecessary and unfairly disadvantages those with limited labor market experience.

As much of this report demonstrates, most states’ UI systems are far too complicated. Workers cannot make an accurate assessment of eligibility, and trying to figure potential UI benefits is an even more daunting task. States should alter their benefit formulas so that those who are unemployed have a reasonable expectation of their benefits.

Although variations in UI benefits are partly due to regional differences in the cost of living, states should still adopt average benefit amounts that exceed the federal poverty level. Average weekly benefits in 32 states are below the poverty level for a single-parent household with one child.¹⁷ Average weekly benefit amounts should be structured so that the unemployed don’t live in poverty while looking for a new job.

Based on the above criteria, the basic structure of Washington state’s UI system is the best in the country. By using only an hours requirement for eligibility, Washington eliminates the inequity between lower- and higher-paid workers. It is also one of 12 states that has adopted a base period that counts workers’ most recent earnings. The simplicity of these rules is easier for workers to understand and serves to de-mystify the UI system. Given Vroman’s (1991) finding that the majority of workers do not apply because they don’t believe they are eligible, clarity and simplicity are obviously important criteria for shaping any UI policy. That 45.7% of unemployed workers apply for benefits in Washington, a share considerably higher than the 37.4% national figure, attests to the need for greater simplicity. It should be noted that insured unemployment is higher despite the state’s steep hours-of-work requirement. Washington State’s 680 hour requirement is among the most stringent in the United States. Creating a system that has straightforward rules and transparent policies could substantially improve the uptake of UI benefits.

Conclusion
To understand the effectiveness of the UI system, researchers and policy analysts must consider the complexities resulting from the variations in policy between states. While national averages are certainly indicative of general trends, such aggregates often mask considerable differences between UI systems from state to state. In general most states eased eligibility requirements and raised both maximum benefit amounts and average weekly benefits over the 1990s. But these averages hide the fact that 13 states bucked the national trend and had declining benefit amounts, and that 10 other states raised the minimum earnings requirements necessary to qualify for UI.

Fortunately, an increasing number of states have revised the way in which they calculate base periods so that a worker’s most recent completed quarter is factored in when determining eligibility. Finally, we find that increases in the minimum wage both on the state and federal level had an important impact on both eligibility and on raising the percentage of unemployed workers applying for benefits.

-August 2001

Endnotes
1. All data taken from Comparison of State Unemployment Insurance Laws (U.S. Department of Labor 2000).

2. Unless otherwise noted, we use the state-specified minimum wage or the federal minimum wage in each state, whichever is higher.

3. Estimate from the GAO (2000). The National Employment Law Project estimates that 39 states would make an ineligible determination if a worker sought part-time work exclusively.

4. Other proposed policies have included “adopt[ing] measures that attempt to limit the creation of temporary jobs to a level that accommodates the worker’s need for flexible annual work schedules” (Golden and Appelbaum 1992) and reducing the costs unions currently face in organizing contingent workers by allowing sectoral bargaining for unions, or making legislative provisions for comparable pay and benefits schedules for regular full-time workers (duRivage, Carré, and Tilly 1999, 273-78).

5. In this case we have 13 weeks for the quarter most recently completed and another 12 weeks for the nearly completed quarter.

6. The 12 states with alternate base periods in 2000 are New Jersey, New York, Ohio, Maine, Massachusetts, New Hampshire, Vermont, Rhode Island, Michigan, North Carolina, Wisconsin, and Washington.

7. These states include Alabama, Colorado, Florida, Maine, Massachusetts, Michigan, Nebraska, New Mexico, New York, Ohio, and Utah.

8. All amount are 2000 dollars, inflated by the CPI-U.

9. Note that the unit of analysis is the state (unweighted), not the individual.

10. An analysis of the 20th wage percentile provides the exact same results. It appears that nine states in the 1990-95 period raised their earnings requirements faster than 20th pe
rcentile wages, while eight states raised their earnings requirements faster than the growth in the 20th percentile wages during the 1995-2000 period.

11. Alabama, Florida, Oregon, Nebraska, New Mexico, New York, and Utah.

12. Training wages and minimum wage exemptions will lower the earnings requirements.

13. These states include Alabama, Alaska, Arizona, California, District of Columbia, Florida, Maryland, Michigan, Minnesota, Mississippi, Texas, and West Virginia.

14. States that increased the number of manufacturing jobs include New Hampshire, Vermont, Indiana, Michigan, Wisconsin, Minnesota, Iowa, North Dakota, South Dakota, Nebraska, Kansas, Georgia, Kentucky, Arizona, Oklahoma, Texas, Montana, Idaho, Wyoming, Colorado, Arizona, Utah, Nevada, and Oregon.

15. Table 5 shows the results of a panel regression of the states’ insured unemployment. The dependent variable in all regressions is the percentage of unemployed workers currently filing for UI benefits and those continuing to receive benefits [(initial claims + continuing claims) / total unemployed]. Since we are examining annual data, the dependent variable is the state annual average. This is the variable most commonly analyzed in the literature. Some authors have highlighted problems with this dependent variable. In particular the IU/TU ratio is thought to overcount the number of UI recipients, since as many as 10%-15% of initial claimants will ultimately be denied benefits (Bassi & McMurrer 1997). We believe that including filers who are ultimately denied is a better measure of how individuals respond to changes in UI policy and have consequently included them in the analysis.

Table 5 includes two sets of regression results using a one-step panel data estimator developed by Arellano and Bond (1991). We estimate a difference model with lagged endogenous variables. This estimation strategy is robust to a number of error specifications including auto-correlated and heteroskedastic errors (Arellano and Bond 1991, 281-82). All standard errors are estimated using the robust variance-covariance estimator.

16. Committee on Ways and Means, U.S. House of Representatives, Overview of Entitlement Programs, Green Book (2000).

17. The U.S. Census poverty level for one adult, one child equals $11,869. This number is calculated based on 26 weeks of UI eligibility at the average weekly benefit amount.

References
Acemoglu, D., and R. Shimer. 1999. Productivity Gains From Unemployment Insurance. Journal of Political Economy. Vol. 107, No. 5. pp. 893-928.

Atkinson, A.B., and J. Micklewright. 1991. Unemployment Compensation and Labor Market Transitions: A Critical Review. Journal of Economic Literature. Vol. 29, No. 4, pp. 1679-1727.

Arellano, M., and S. Bond. 1991. Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Data. The Review of Economic Studies. Vol. 58, pp. 277-97.

Blank, Rebecca M., and David E. Card. 1991. Recent Trends in Insured and Uninsured Unemployment: Is There an Explanation? Quarterly Journal of Economics. Vol. 106, No. 4, pp. 1157-89.

Clark, Kim B., and Lawrence H. Summers. 1979. Labor Market Dynamics and Unemployment: A Reconsideration. Brookings Papers on Economic Activity. Vol. 1, pp. 13-72.

Clark, Kim B., and Lawrence H. Summers. 1982. “Unemployment and Labor Market Transitions.” In M.N. Baily, ed., Workers, Jobs, and Inflation. Washington, D.C.: Brookings Institution.

duRivage, Virginia L., Françoise J. Carré, and Chris Tilly. 1998. “Making Labor Law Work for Part-Time and Contingent Workers” in K. Barker and K. Christensen, eds., Contingent Work: American Employment Relations in Transition. Ithaca, N.Y.: ILR Press.

Golden, Lonnie, and Eileen Appelbaum. 1992. What Was Driving the 1982-1988 Boom in Temporary Help Employment? American Journal of Economics and Sociology. Vol. 51, No. 4, pp. 473-93.

General Accounting Office. 2000. Unemployment Insurance: Role as Safety Net for Low-Wage Workers Is Limited. Washington, D.C.: GAO.

Gustafson, Cynthia K., and Phillip B. Levine. 1998. Less-Skilled Workers, Welfare Reform, and the Unemployment Insurance System. Working Paper 6489. Cambridge, Mass.: National Bureau of Economic Research.

Holzer, Harry. 2000. Unemployment Insurance and Welfare Recipients: What Happens When the Recession Comes? Washington, D.C.: Urban Institute.

Hudson, Kenneth. 1999. No Shortage of ‘Nonstandard’ Jobs. Briefing Paper. Washington, D.C.: Economic Policy Institute.

Mortensen, Dale T. 1970. Job Search, the Duration of Unemployment and the Phillips Curve. The American Economic Review. Vol. 60, No. 5, pp. 847-62.

U.S. Department of Labor. Various years. Comparison of State Unemployment Insurance Laws. Washington, D.C.: DOL Employment and Training Division.

U.S. Department of Labor. Various years. UI Data Summary. Washington, D.C.: DOL.

Um’rani, A., and Vicki Lovell. 2000. Unemployment Insurance and Welfare Reform. Washington, D.C.: Institute for Women’s Policy Research.

Vroman, Wayne. 1980. State Replacement Rates in 1980. Unemployment Insurance Compensation: Studies and Research, Vol. 1. Washington, D.C.: National Commission on Unemployment Compensation.

Vroman, Wayne. 1991. The Decline in Unemployment Insurance Claims Activity in the 1980s. Unemployment Insurance Occasional Paper 91-2. Washington, D.C.: U.S. Department of Labor, Employment and Training Division.

The author would like to thank the Ford Foundation for support of this project. Research assistance was provided by Matthew Walters. The author also benefited from conversations with Eileen Appelbaum, Maurice Emsellem, Rick McHugh, Becki Smith, Mark Baldwin, John Schmitt, and Chirag Mehta.