A plan to reform the unemployment insurance system in the United States

 

Abstract

The unemployment insurance (UI) system has played an important role in delivering relief during the current pandemic. At the same time, this experience has highlighted the important challenges facing the UI system due to poor and underfunded administrative capacities, too few unemployed workers qualifying for UI benefits, inadequate levels of regular UI benefits, lack of effective triggers to tie benefit duration to economic conditions, and meager utilization of work sharing programs. In this proposal, I suggest remedies for each of these problems, and argue that these remedies are best achieved through converting the UI system to a fully federal program.

Introduction

Unemployment insurance (UI) in the United States was established in 1935 as a federal-state system. The last major changes were put in place in 1976, making the system in great need of major reform. Following the Great Recession, there was widespread agreement among experts that the UI system in the United States required substantial changes. There were numerous reports with detailed recommendations for reform (e.g., O’Leary and Wandner 2018), but unfortunately these recommendations were not implemented at the federal level. The Covid crisis has led to an unprecedented reliance on the UI system to deliver relief, income support, and fiscal stimulus in modern times. Importantly, the crisis has highlighted the serious strains on the UI system—with its patchwork of state rules and often impoverished administrative capacity from chronic underfunding. At the same time, it also demonstrated how the federal government can provide generous relief during a crisis through the UI apparatus.

Between the second week of March and the second week of May 2020, the number of individuals on UI jumped from 1.8 million to 24.9 million, as many employers (especially in hospitality, travel, and other service sectors) put their workers on temporary layoffs and most state governments enacted stay-at-home orders. Moreover, the federal government extended the potential benefit duration (PBD) and provided generous wage replacement over the course of the crisis. This included an unprecedented (federally funded) $600/week boost to UI benefits (Federal Pandemic Unemployment Compensation, or FPUC) that lasted through July 2020, followed by five weeks of $300/week boost (“Lost Wages Assistance”), then a further eight months of $300/week boost starting in January 2021. In addition, there was a dramatic increase in eligibility through the Pandemic Unemployment Assistance (PUA) program that was designed to extend benefits to those who may not qualify for UI due to limited work history or hours, or because they are self-employed. In the week ending January 30 of this year, for instance, 7.7 million of the total 18.3 million individuals receiving benefits received them through the PUA, demonstrating the importance of the program.

The fact that we were able to expand UI eligibility and benefits and provide relief to tens of millions of families during the crisis is a success story and demonstrates the potential to do more than we had in past downturns. At the same time, the implementation of the pandemic UI programs showed the limitations of the state-based UI system. One example of such limitations was the serious delay in processing UI claims in many states, which led to substantial drops in consumption (Farrell et al. 2020). In some states, those delays were created by major backlogs (Goodman, Cohen, and Chandler 2020). Additionally, archaic software infrastructure in many states made it impossible to change the benefit replacement rate or maximum benefit levels easily, leading Congress to use a flat dollar boost. The minimum and maximum benefit amounts vary greatly between states, creating difficulty in ensuring sufficient benefit amounts. There were numerous fraud attempts, which led some states (e.g., California) to halt all payments for some time. Short-time compensation (STC; also called work sharing) was not available in around half the states; even where they were available, lack of information and administrative hurdles kept the number of workers covered by work sharing at an extremely small number. These and many other limitations exposed the challenges for our current federal-state system.

The last major reforms in the UI system were enacted in 1976. Much has changed in the labor market since then, and we have learned much more about the possibilities and limitations of our current UI system. This especially includes the experience from the Great Recession as well as the current Covid crisis. This policy paper builds on other thoughtful and important proposals for reforming UI, including Chodorow-Reich and Coglianese (2019), O’Leary and Wandner (2018), West et al. (2016), Bennet (2020), and others. In this policy paper I have tried to synthesize some key aspects of the existing proposals, while updating and adding some elements in light of the experiences with UI during the Covid crisis.
There are five key components of my proposal:

• Make UI a fully federal program, similar to Social Security.
• Expand eligibility by reducing eligibility earnings thresholds, and increase the types of workers who qualify for benefits.
• Tie the PBD to state and national triggers for total unemployment rates (TURs).
• Make the replacement rates depend on the level of earnings, raise the rates substantially for those at the bottom of the pay scale, and add a recession boost.
• Revamp STC.

In this policy paper I begin with a background on UI, including the goals of the policy and how those goals are currently administered in the United States. Next, I discuss the challenges facing the current program. Following this, I lay out my proposed reforms to the UI system. Finally, I address possible questions

Background on UI in the United States

How UI is Currently Structured in the United States
UI provides partial income support to workers who lose their jobs for a period of time. The maximum length of time an individual can collect UI is known as the potential benefit duration, or PBD, while the share of income that is replaced by UI benefits is known as the replacement rate.

Under the 1935 Social Security Act, the UI system is a federal-state partnership, administered by each state. Because administration of the program is left to the states, there is tremendous variation in eligibility and replacement rates across states, and some variation in the PBD as well. In general, in all states eligibility is based on an earnings test, so only workers with sufficient earnings over a base period are eligible. The base periods vary across states. In some states, eligibility determination is based only on the standard base period, which is typically the first four of the previous five completed quarters prior to a UI claim. This ignores most-recent work history and puts weight on older earnings. Other states allow for an alternative base period where qualification is based on the immediately preceding four quarters in some states, which tends to increase eligibility rates.

The determination of the benefit also varies greatly across states. While most states typically start with replacing around 50 percent of income, there are wide variations in maximum benefit levels: currently the maximum weekly benefits range from $240 in Arizona to $855 in Massachusetts. Some states also have minimum benefit levels, though many do not; in Washington State, for example, the 2021 minimum benefit amount is $201. Factoring in the differences in rules, and averaged over all recipients, the replacement rate in Louisiana in 2019 was 35 percent, while it was 50 percent in the state of Washington. Moreover, the choice of quarters in the base period can also affect the determination of benefits and can adversely affect workers who have a more limited work history.

In normal times, most states typically provide 26 weeks of benefits, though as of 2021, seven states provide benefits for a shorter period (ranging between 16 and 22 weeks). During recessions, the PBD is typically longer. The federal government’s Extended Benefits (EB) program uses triggers to extend the PBD during downturns. In particular, there are two tiers to the EB program: Tier 1 extends PBD by 13 weeks and Tier 2 extends it by an additional 7 weeks. In addition, Tier 2 is optional and depends on the state’s decision to participate. The EB program is based on thresholds on a combination of the insured unemployment rate (IUR, based on UI filings) and the TUR (based on a household survey). However, as a practical matter, nearly all of the triggers onto EB during the Great Recession and its aftermath occurred via TUR triggers. These triggers are as follows:

• Tier 1: TUR of 6.5 percent, and 110 percent of the minimum rate during the last two years of the reporting period
• Tier 2: TUR of 8 percent, and 110 percent of the minimum rate during the last two years of the reporting period

Under current law, the federal government typically covers half of the cost of the EB program in Tiers 1 and 2. During downturns, some parts of EB have fully federally financed. In addition, during the two most-recent recessions, the federal government has introduced emergency programs to further extend benefit durations. I will discuss the use of these emergency programs and the EB programs in greater detail in the section titled Challenges with the Current System.

Costs and Benefits of Providing UI
With the current structure of UI benefits in mind, it is worth thinking about the costs and benefits when designing an ideal UI program. Economists typically approach the design of UI as balancing two factors. On the benefit side is the goal of smoothing consumption of workers and family members by preventing large drops in consumption after job loss. Too low an unemployment benefit would lead to inadequate consumption smoothing if individuals do not have sufficient other sources of insurance or savings, and are not able to easily borrow against future income. On the cost side is the possibility of moral hazard from providing a transfer tied to someone not working. Too generous a benefit may be costly because it could lead workers to reduce their effort to find a job since they are receiving a benefit without having to work. This can reduce the rate at which workers exit out of unemployment, thereby prolonging spells of joblessness and costing taxpayers additional resources.

A key consideration here is that if consumption drops are larger—say because more workers are consuming what they receive in income—the value of insurance is greater for a given level of moral hazard. One implication of this reasoning is that poorer workers—who typically have fewer disposable assets and are closer to subsistence level—are particularly benefited from a more-generous unemployment benefit. In general, distributional concerns should push us toward providing more insurance for low-income earners. So, if a 50 percent benefit replacement rate during normal times is reasonable for the typical unemployed worker, it may be inadequate for those at the bottom.

One important aspect of the moral hazard costs is that they are smaller during downturns. Why? Because during downturns jobs are rationed, and there are many more workers searching for jobs than there are vacancies. As a result, if a generous UI benefit leads one worker to forgo taking a particular vacancy, it is very likely to be filled by a different worker; there is a positive search externality from a more generous UI during downturns. The musical chairs aspect of job finding during downturns means that, whatever we think is the right level of generosity in normal times, it can (and should) be higher during downturns (Landais, Michaillat, and Saez 2018a). In addition, the insurance value of UI is likely longer in downturns, since unemployment spells are longer in those periods. It is easier to draw on savings or to borrow to keep up consumption when facing a 5-week-long spell as opposed to a 20-week-long one.

Finally, besides the microeconomic objective of consumption smoothing, UI also has a macroeconomic role in stimulating consumption and serving as an automatic stabilizer (Chodorow-Reich and Coglianese 2019). In particular, those who are unemployed have a high marginal propensity to consume (MPC), which makes transfers to UI recipients a very effective form of stimulus (Ganong and Noel 2019). Factoring in this aggregate demand externality further supports the idea of increasing UI generosity during downturns.

There is a large body of literature that has used quasi-experimental methods to evaluate the impact of benefit generosity (both PBD and replacement rate) on job-finding rates as well as on consumption smoothing. In appendix A I both lay out the theoretical framework that is used in public finance to analyze optimal UI policies and review in detail the microeconomic and macroeconomic employment impact of UI policies. Prior to the current crisis, the micro evidence (tracking individual unemployed workers and their exit out of UI recipiency) suggests a moderate degree of responsiveness of UI spells to PBD and replacement rates. Summarizing across studies, a 10 percent increase in PBD typically raised the average UI spell length by 0.1 to 6 percent, with a midpoint around 3 percent. For a typical UI spell length of around 10 weeks, this would translate to two additional days, which is a fairly modest sized impact. The evidence on replacement rate is both thinner and somewhat more mixed. Across studies, a 10 percent increase in benefit level typically raised the average UI spells by 1 to 9 percent, with a midpoint around 5 percent.

However, these micro elasticities may substantially overstate how responsive macrolevel employment is to UI benefit generosity, especially in downturns, since they do not factor in search and aggregate demand externalities. In recent years we have seen a number of studies provide a macro employment impact of the extension of PBD to 99 weeks during the Great Recession. On the whole, most (though not all) macro estimates from the Great Recession suggested smaller employment losses than micro studies. Overall, this pattern is consistent with the importance of search externality and stimulus effects during downturns.

There is also evidence on the benefits side of UI. First, a range of studies have documented consumption declines following job loss, with food or nondurable consumption dropping between 6 and 9 percent. More-recent work using high-quality bank account transaction data has also found that nondurable consumption drops by 12 percent when UI benefits are exhausted (Ganong and Noel 2019). Importantly, Ganong and Noel find sharp drops in spending for necessities like groceries, goods from drugstores, and medical copayments. Those authors also document that, when the state-level replacement rate is lower, the drop in consumption from unemployment is larger, highlighting the importance of an adequate benefit replacement amount. Moreover, the reduction in spending they find for necessities has implications for the possible distributional heterogeneity in the insurance value: for those families close to survival constraints, the reduction in these necessities is likely to come at a very high cost in welfare. Consistent with this point, the consumption drop is found to be much larger for households with low liquid assets. Since lower-income families tend to have lower assets, this is another rationale for providing relatively more-generous benefit levels to those with lower incomes.

Another important finding is how sensitive spending is to UI benefits. Overall, Ganong and Noel’s findings suggest an MPC out of UI spending of around 0.8, meaning that $0.80 of each dollar of UI benefits is spent. This is quite high, showing substantial barriers in access to credit among those who are unemployed. This also suggests that UI is likely to be well targeted for the purpose of providing stimulus during downturns.

What We Have Learned About UI During the COVID Crisis
The UI system played a very important role during the Covid crisis in delivering relief to those who were not able to work during the lockdown period, and to both short- and longer-term unemployed workers subsequently. FPUC extended the PBD initially to 39 weeks, though this was subsequently extended further by another 24 weeks in December 2020.

Importantly, the FPUC also expanded the benefit level by an unprecedented $600/week beyond the usual benefit level, fully funded by the federal government between March and July of 2020. As shown by Ganong, Noel, and Vavra (2020; henceforth GNV), the FPUC increased the benefit level by an unprecedented amount. Averaged across states, the median replacement rate rose from 48 to 145 percent. The goal of this increase in benefits was to allow most workers to receive at least close to their usual earnings during the period when individuals were laid off due to the onset of the pandemic and (initially) due to the lockdowns that mandated business closures in most states. This $600/week boost expired at the end of July and was not renewed via legislation at that time due to lack of political consensus on the terms of renewal. However, the Trump administration created a temporary Lost Wages Assistance program that provided a $300/week boost to those who were unemployed via agreements with states typically for the five weeks following the expiration of FPUC, covering most but not all UI recipients. While the exact timing of receipt of the payments varied across states, it covered payments for the period of unemployment during August and early September, so by the middle of September the median replacement rate had fallen back sharply to pre-pandemic levels, averaging around 48 percent across states. Finally, under legislation passed in December 2020, the FPUC benefit boost was renewed, albeit at $300/week instead of $600/week.

How did workers and family members respond to job loss, to the FPUC, and to its subsequent expiration? Farrell et al. (2020) find that spending among those losing jobs dropped by 20 percent prior to receiving UI. At the same time, following UI receipt recipient spending rose by 10 percent as compared to the pre-job-loss consumption level, in spite of the loss in earnings. The $600/week boost helped turn what would have been a very sharp and protracted reduction in spending into a gain in spending from the baseline. Relatedly, upon the expiration of the FPUC in July 2020 spending declined by 14 percent in August. (Importantly, the consumption drop was substantial even though recipients had built up savings from the $600 boost.)

The unprecedented expansion of the benefit level also created concerns about unintended consequences. In particular, the fact that many unemployed workers were receiving benefits in excess of their usual earnings was considered problematic by many observers. While such benefit levels were seen as tolerable during a very short-term (one to two months) period where many workers were unable to work during lockdowns and closures in the hospitality sector, many in Congress and statehouses argued that the benefit boost was too generous (Iacurci 2020), and slowed down the reopenings in the summer months. This was an important factor behind the lapse in the FPUC payment at the end of July, which was expected to aid the labor market recovery by incentivizing work.

The evidence so far, however, suggests a very limited impact of the UI policies on employment outcomes. Early in the Covid crisis a number of papers evaluated the impact of the $600 boost by comparing the trajectory of employment typically between March and June across groups with very different replacement rates (due to differences in earnings). These papers used a number of new high-frequency datasets, including Homebase and the Census Pulse Survey, but found little systematic evidence that the higher replacement rates were associated with lower employment recoveries (e.g., Altonji et al. 2020; Bartik et al. 2020; Dube 2020b).

The best evidence comes from three studies that leverage the sharp expiration of the FPUC at the end of July 2020 and the fact that the replacement rates fell differently for different groups from the expiration of the flat $600 boost. Finamor and Scott (2021) use personnel data from Homebase, a scheduling service whose clients tend to be small businesses, especially restaurants. They assess whether those who were not employed by their employer (or by another Homebase employer) return to employment following the FPUC expiration, and find no relationship between the change in replacement rates (based on pre-pandemic earnings) and change in reemployment. Ganong et al. (2021) use transaction-level data of individuals with a bank account at JP Morgan Chase. They consider unemployment duration of individuals receiving UI benefits, and find little change in the exit rate out of unemployment following the expiration, and that this change was only slightly larger in states where workers saw larger drops in the replacement rate. Their estimates suggest duration responses that are smaller than indicated in existing literature.

In contrast to these microlevel estimates, Dube (2021) uses the Household Pulse Survey to study the impact of the July expiration of the FPUC $600 boost on aggregate employment probability, using cross-state variations in the change in median replacement rates following the expiration. Dube does not find any evidence of increased employment in states where the benefit replacement rate fell more. Jobs actually rose slightly less in states where the replacement rate fell more, though this difference was not statistically significant.

Overall, the evidence from the pandemic shows a surprisingly small impact of benefit generosity on employment—both at micro and macro levels. Needless to say, the Covid crisis has had unique features—including a very high share of unemployed workers who were on temporary layoffs and then were recalled, especially in the earlier part of the crisis. At the same time, the fact that the employment effects were elusive in spite of the unprecedented level of benefit replacement suggests policymakers worried about negative incentives have more latitude to provide more-generous benefits, especially during downturns when the moral hazard cost is less due to other compensating factors.

What are some takeaways from the consumption response to UI during the Covid crisis, and what implications do those takeaways have for policy design? From a purely insurance perspective, the $600/week supplement not only smoothed consumption but actually boosted it, with a small to no efficiency loss in terms of employment. Even though some of the circumstances during this downturn are unusual, the evidence from experience nonetheless suggests we can likely raise benefit levels, especially for those at the bottom, to allow much lower consumption drops than have occurred in the past. Moreover, the strong consumption response supports the macro-stabilization goal of boosting UI benefits during downturns.

Challenges with the Current System

State Administration is Ineffective
The state administration of this program has created a patchwork of rules that have little benefits but many costs. At a basic level, a state-based system prevents insurance across states. For example, if a particular state is hit harder by a downturn, it will be in a deeper hole when it comes to rebuilding its trust fund and will therefore need to raise payroll taxes by a greater amount. Moreover, a state-based system does not allow easy adaptation to new circumstances, such as changing benefit levels—as demonstrated during the current crisis. And perhaps most importantly, state administration makes it more difficult to ensure that systems in all 50 states (and the District of Columbia) are aligned with program objectives, including adequate eligibility benefit standards, efficient processing of applications and timely benefit payments, and use of linked earnings and income data for benefit calculation and verification. As demonstrated during the current crisis, there was substantial variation in the timing by state; this variation was costly in terms of consumption loss (Farrell et al. 2020).

Along with benefit levels, the employer taxes vary tremendously across states. There is wide variation in the definition of the taxable wage base across states, and states have often been reluctant to raise taxes to sufficiently fund adequate benefits (O’Leary and Wandner 2018). As a result, taxes ranged between $50 per beneficiary in Florida to $631 in Oregon in 2020 (DOL 2021). The tax variation puts employers in very different positions in different states. A DOL report in 2020 found that around half of the states’ UI trust funds were underfunded (DOL 2020). At the same time, when tax rates are raised they happen right after downturns—but this can potentially hamper the recovery in the labor market. As Johnston (forthcoming) shows, these tax increases after the Great Recession reduced hiring and employment and prolonged the recovery. By insufficiently smoothing the taxation over the business cycle, states have also hampered effective recoveries. Finally, because employers pay the entire payroll tax, they have put pressure on state governments to not raise the tax rates, contributing to the underfunding.

In the case of UI, allowing state administration and a hybrid federal-state partnership for financing has not produced any clear gains. In contrast, the patchwork of rules and pressure to cut benefits for budgetary reasons has produced a system that is inadequate and arbitrary.

Too Few Unemployed Workers are Eligible for Benefits
One of the key failures of the current UI system is how few workers who are unemployed received assistance through the program prior to the Covid-19 recession. In 2019 only 28 percent of all unemployed workers in the country received UI benefits. Moreover, this percentage varied tremendously across states: in 2019 a mere 9 percent of unemployed workers in North Carolina received UI, while the share in New Jersey was 59 percent (DOL 2019).

What are the key factors that drive the eligibility problem? There are several important contributors. The most important one is that the thresholds of minimum earnings exclude many workers who lose jobs. The thresholds vary across states, which in part explains the recipiency variation. When the thresholds are too high (as they often are), many low-wage, part-time, seasonal, and volatile-schedule workers are excluded even when they lose their jobs. Relatedly, in some states, much greater weight is put on older earnings, even if more recent earnings are higher. A solution to this problem is through additionally allowing an alternative base period where qualification is based on earnings from the immediately preceding four quarters, as is done in the majority of states today. By taking the greater of earnings from the alternative base period or standard base period, allowing alternative base period helps increase eligibility.

Currently, UI is available primarily to those workers who were involuntarily separated. While this is sensible in most cases, there are some workers who voluntarily separate for whom UI eligibility makes sense. The first set of such good cause voluntary separations include those who may see substantial changes in their work conditions due to reasons that are beyond the worker’s control (such as a large drop in hours, or erratic scheduling). In addition, a second set of voluntary separations involve extenuating family circumstances, include own or family member health reasons, or when child-care arrangements cannot be secured, or when a spouse or partner relocates. Many states have some form of these good cause exceptions, but many states do not, leading to reduced recipiency (see West et al. 2016). The importance of these factors (such as caring for family members, or inadequate child care) have been made even more prominent during the pandemic.

Finally, there are a set of unemployed individuals (like gig workers, self-employed individuals, or those entering the labor force for the first time) who would fall through the cracks of regular UI. The inadequacy of the current eligibility standards is also highlighted by how many unemployed workers have relied on the PUA program that allowed many otherwise ineligible workers to receive assistance. For example, for the week ending December 12, 2020, 8.5 million of the total 19.6 million individuals receiving some benefits did so through the PUA. While reforming the eligibility rules can certainly help with this, nonetheless there may be a role for a more permanent assistance program to help reach these individuals. There are some challenges in providing assistance to workers who have direct control over their hours, like the self-employed. It opens up the possibility of moral hazard, where any self-employed person who wants to stop working for a period can receive unemployment assistance during that time.

Lack of Effective, Automatic Triggers
As I discussed in the Background section, there are compelling reasons to expand UI PBDs during downturns. The economic costs stemming from reduced job finding are low when the labor market is slack, due to both search and aggregate demand externalities. While the federal government’s EB program is, in principle, equipped to do this, in practice some key program details prevent it from being effective. Unfortunately, the look-back provisions of requiring current TUR to exceed 110 percent of the minimum TUR over the past two years greatly diminishes the value of the EB program. As Chodorow-Reich and Coglianese (2019) show, most workers losing jobs during the early part of downturns have lower benefit durations, so the EB does not apply to them. The EB is really more relevant later in a downturn and during early recovery periods. However, precisely in this case, the look-back provisions lead the EB to trigger off prematurely. Moreover, even with the two tiers the EB program would allow PBD to rise to at most 46 weeks, which is insufficient for deeper downturns.

As a result of the inadequacy of the EB program, historically the federal government has introduced emergency programs to extend benefit durations. This included the Emergency Unemployment Compensation program in the Great Recession that extended PBD to as much as 99 weeks. During the Covid crisis, the Pandemic Emergency Unemployment Compensation extended the PBD initially by 13 weeks and subsequently by additional numbers of weeks. Moreover, while the federal government typically pays half of EB costs, the emergency extensions have been fully federally funded. The problem with this discretionary approach—as exemplified during the current crisis—is that each extension is often contentious, and decisions revolve around vagaries of politics at least as much as economic conditions. Moreover, unlike the state-level determination in the EB program, the emergency provisions are blunter, since they are uniform nationally, which prevents them from adequately addressing differential severity of downturns across states.

The haphazard way in which the FPUC expired and passed temporarily, and then was revived partially in December (and may be extended further under the Biden administration) highlights the problems with discretionary policy that is not tied to economic indicators. A successful reform proposal would produce a policy that would allow these extensions in PBD to occur in a more automatic fashion based on economic realities.

Inadequate Benefit Levels
Currently, a combination of factors makes the regular UI benefit levels inadequate for most unemployed workers. First, in most states the current benefit replacement rate averaged over all recipients is below 50 percent. Beyond that, states have widely varying maximum benefit levels, which lead to extremely low replacement rates in some states. For example, averaged over all recipients, the replacement rate in Louisiana in 2019 was only 36 percent, while it was 51 percent in the state of Washington. Moreover, the choice of quarters in the base period can affect the determination of benefits and can adversely affect workers who have a more limited work history.

At a broad level, I would argue that the 50 percent replacement rule is inadequate for many workers. Many optimal UI calculations from an insurance approach suggest that a 50 percent wage replacement may be reasonable as a general rule (Chetty 2008; O’Leary and Wandner 2018). To be sure, under different assumptions, such as workers placing a greater value on insurance, the optimal replacement rate is estimated to be higher. Still, a 50 percent replacement rate may work adequately for a representative worker in the middle of the wage distribution experiencing short or medium unemployment spells.

But if we approach social insurance from a more holistic perspective, we need to better address distributional concerns. Those concerns suggest greater replacement rates for workers at the lower end of the wage distribution. In particular, low-wage workers who are much closer to subsistence level have much greater difficulty over the unemployment spell; under a wide range of normative assumptions, a 10 percent reduction in consumption has very different welfare implications for low-wage workers who are close to their survival constraint than it does for those at the middle or the top. To give an example, a single mother of two children, working full time at $10/hour, would be at roughly 90 percent of the federal poverty level. If she loses her job and qualifies for UI, her annualized income would place her at below 50 percent of the federal poverty level, or in deep poverty. A moderately long unemployment spell can have devastating consequences for lower-income families.

As discussed in the Background section, the evidence suggests much larger drops in consumption among those with lower liquid assets; this is much more common among lower-wage workers. That evidence provides a prima facie argument for providing a greater replacement rate at the bottom of the distribution. Moreover, low-wage workers are likely closer to survival constraints, which may mean they reduce the consumption drops using very costly mitigation strategies—like reducing investments in health and education in order to pay for more-immediate necessities.

In addition, there are also strong, evidence-based, rationales for raising the optimal benefit levels in recessions. Despite that evidence, prior to the Covid crisis we had not boosted benefits sufficiently during downturns. As discussed in the Background section, there are three distinct rationales for raising benefits in recessions. First, transfers to UI recipients are well targeted to be spent and increase demand. The unemployed have much higher MPCs, making transfers to this group a very effective form of stimulus. Moreover, a boost to benefit levels provides a substantial boost to aggregate demand early in downturns by providing greater transfers to all unemployed individuals (not just to longer-term unemployed individuals, as is the case through PBD extensions) as soon as, say, the national TUR exceeds a threshold. This makes the automatic stabilizer aspect of UI much more meaningful than is typically implemented, since the boost in benefits would kick in much faster than the boost from the longer duration of benefits during a downturn (Chodorow-Reich and Coglianese 2019).

In addition, as I described above, the cost of providing more-generous benefits is lower during downturns because of search externalities. The moral hazard cost of workers not looking hard for a job is lower during downturns when jobs are rationed, since one worker finding a job is more likely to prevent someone else from finding a job during downturns. Finally, workers are likely to stay unemployed longer during downturns, which makes it harder to cover expenses by drawing down savings or borrowing more, which increases the insurance value of more-generous benefit levels during downturns.

The inadequacy of the regular benefit levels was laid bare during the Covid crisis, which led the federal government to provide a $600/week boost in March 2020, and a smaller $300/week boost in December of 2020. As discussed in the Background section, the consumption and employment responses suggest that this was a highly successful policy in boosting spending with minimal job loss. This was true even though the average median replacement rate under the $600 boost was much greater than 100 percent (and stood at 148 percent averaged across states.) At the same time, it is worth recognizing that the flat boosts were in part a reflection of the inability of the federal government to implement higher earnings replacement rates due to antiquated software and administrative hurdles. Ideally, a reformed UI program would provide higher replacement rates, especially at the bottom, and provide a recessionary boost, while preventing replacement rates from greatly exceeding 100 percent of usual wages in order to avoid elevated moral hazard problems.

Inadequate Utilization of Work Sharing
Economic downturns lead to inefficient dissolution of productive matches between workers and employers. Ideally, we want to encourage workers to remain linked to their workplaces during temporary downturns, returning to work when demand picks up. While temporary layoffs (during which workers collect UI benefits) can partly accomplish this objective, it is an imperfect solution. For example, many temporary layoffs may become permanent, and that uncertainty may lead workers to move on and possibly end in a job that is a worse match.

One way to avoid this eventuality is through STC. Under STC, instead of laying off workers, employers reduce their hours by as much as 40 percent to 60 percent, depending on the state. Then the UI system helps provide benefits to partly compensate for the lost hours through a pro-rated share of the workers’ weekly benefit amount. In this way, workers and employers can stay attached, and the reduced earnings are shared across workers more broadly. Moreover, by reducing the number of active jobseekers, STC can mitigate the reduction in labor market tightness during downturns, thereby helping unemployed workers as well. Especially when coupled with an adequate benefit replacement, the human costs of recessions can be greatly mitigated through STC. A particularly successful example is Germany, where use of STC has greatly mitigated the rise in unemployment during downturns.

Currently, 27 US states have STC programs. However, the participation rate in STC has been very low. For example, in the week ending January 30, 2021, about 0.5 percent of the 18.3 million continuing UI claims were through STC. The share was virtually the same prior to the pandemic. Those businesses that use STC are overwhelmingly satisfied with the program (Balducci et al. 2015). However, as von Wachter (2020) notes, the main hurdle to increased STC participation appears to be lack of employer awareness and administrative bottlenecks. Lack of awareness is a major problem, but there are simple solutions. For example, Houseman et al. (2017) conducted demonstration projects in two states and found that disseminating information about STC significantly raised use of that option.

Besides lack of awareness, administrative hurdles are also a barrier. Employers typically must fill out and mail a form with a list of affected workers and their information to the state agency, which then processes the forms and decides on approval. In most cases during the pandemic, approvals were slow, and the state agencies have not been provided adequate funding and administrative capacity to streamline the process. These barriers prevent greater use of STC.

Finally, as Abraham and Houseman (2014) point out, there are several other disincentives for use of STC. First, employers may not cut worker hours more than 40 to 60 percent (depending on the state), which may be too restrictive in some cases. Second, employers are required to pay for full health insurance benefits for workers using STC. This makes it more costly for employers to use STC instead of laying off workers.

Proposals for Reform

In light of past evidence on the costs and benefits of the UI system, and to address some of the key existing shortcomings in the current UI system, I propose five key changes, which I discuss in detail here.

Proposal 1: Make UI A Federally Financed and Administered Program
Under my proposal, UI will be made a fully federally administered program, like Social Security. Part of the costs of the program—reflecting regular benefits paid during normal times—would be paid using a federal payroll tax, while the balance would be paid using general federal revenue. I recommend that the payroll tax be levied on both employers and employees, similar to Social Security. Since the current system is based only on employer taxes, there will be some added costs on workers. However, note that even when employers nominally pay a payroll tax, it is partly paid for by workers through lower wages. In other words, if all of the added taxes needed to pay for the proposed benefits were nominally on employers, some of those taxes would be passed through to workers as lower wages. At the same time, the pass-through of employer taxes to wages is likely to be incomplete, which is why employers often lobby against increases in the tax rates. Importantly, as O’Leary and Wandner (2018) argue, moving away from an employer-only system could reduce some employer opposition to changes in benefit generosity.

Under my proposal, the payroll taxes would be set to pay for expenses during normal times, not downturns, and should be within a range that does not vary by the business cycle. (The range allows for some degree of experience rating, whereby taxes depend on the extent of use of benefits among workers laid off by a particular firm.) To be specific, the payroll tax rate schedules should be set to allow revenues to equal to outlays for benefits under Tiers 1 and 2 of benefit durations, as described in Proposal 2 below; in brief this means funding up to 33 weeks of UI spells. All of the cyclical financing should come from general funds. The lack of cyclical changes in tax rate schedule can help in the economic recoveries. In addition, the use of general funds from the federal budget to pay for cyclical UI helps provide greater stimulus during economic recoveries. Moreover, this automatically increases revenue collection during extended booms (without any pressure to cut the rates), thereby stabilizing activity.

Finally, the federal government should use all available earnings sources from UI employment and earnings data (ES-202 filings by employers) to automatically calculate the benefit levels. (The formula the federal government would use to automatically calculate those levels is described in Proposal 4.) Employers should also be required to report workers’ hours and not only earnings as they do currently in four states. When they file for UI, applicants should automatically be shown the default benefit level based on their earnings records on file.

Proposal 2: Expand Eligibility
In order to increase recipiency, I make the following recommendations.

Reduce the earnings eligibility threshold, setting the minimum earnings requirement at $1,000 in one quarter and $500 in a second quarter during the base period, allowing for both standard and alternative base periods. These thresholds are substantially lower than those used in most states. Having a more-realistic and more-consistent earnings eligibility threshold can go a long way toward raising eligibility. Both reducing the earnings thresholds and allowing for the alternative base period increase the odds that an unemployed worker qualifies for UI. These thresholds should be indexed to the median average weekly wage (AWW), similar to suggestions made by O’Leary and Wandner (2018).

Allow good cause voluntarily separating workers to be eligible for UI. This allowance would include workers who may see change in work circumstances (such as wage or hour cuts) that are beyond the worker’s control; and workers who have to quit due to extenuating family circumstances, such as own or family-member health reasons, or when child-care arrangements cannot be secured, or when a spouse or partner relocates. This recommendation is similar to suggestions in West et al. (2016) and Bennet (2020). Moreover, most of these conditions are covered by provisions that are already in place in at least some states, but under my proposal these provisions would be applied uniformly across the country.

In order to raise awareness and reduce the cost of application, the federal government should automatically send a letter to employees about potential eligibility when they separate from an employer.

Finally, the federal government should implement a Jobseeker’s Allowance (JA) program for those who are unemployed and searching for work, but who do not qualify for regular UI, closely following the proposal in West et al., 2016. This group would include those who are self-employed, those with inadequate earnings history, new entrants, and reentrants to the labor market.

A key requirement would be that these individuals are actively seeking employment; to mitigate abuse the work search requirements should be more stringent for this group of workers than for regular UI recipients. Another way to mitigate concerns of abuse of the program is through making the assistance substantially less generous than regular UI, and by imposing a time limit of receipt (see West et al. 2016 for a detailed presentation of a JA plan). Under my proposal, the JA benefit level would be set at 20 percent of the AWW. The JA PBD in ordinary times would be 13 weeks (or half of usual length of regular UI benefits). The JA PBD would also respond to the same state and national triggers that would determine the regular UI PBD, as discussed in Proposal 3 below. In particular, the JA PBD should be set at UI PBD less 13 weeks. However, there would be additional limits on collecting the JA, with a 52-week total limit in any rolling five-year period. The JA program would be funded entirely out of regular federal funds.
Overall, how will these reforms affect recipiency rates? While it is difficult to answer this precisely, I estimate that combination of reforms could lead to a recipiency rate among the unemployed of around 55 percent, which was double the recipiency rate in 2019 nationally. For reference, in 2019 the three states with the highest recipiency were New Jersey (59 percent), Massachusetts (50 percent), and Connecticut (47 percent), averaging at 52 percent. I am assuming that the combination of reform will push the national recipiency rate to the current frontier across states. While it is difficult to project take-up of a new program, I estimate that JA would cover an additional 10 percent of the unemployed, based on the use of PUA during the pandemic.

Proposal 3: Tie PBD to State and National Triggers
The key recommendation is to reform the current EB program to accomplish three objectives: (1) remove the look-back period, (2) add more tiers to allow for adequate increases in the PBD during downturns, and (3) include both national and state triggers to allow for a robust response at the national level while allowing for heterogeneity across local areas. Since the IUR triggers have largely been irrelevant for EB triggers, following Chodorow-Reich and Coglianese (2019) I recommend focusing on TUR triggers.

In particular, I recommend seven tiers of PBD for the proposed federal UI program, where the PBD will depend on the maximum of state or national TUR:

• Tier 1: Under 5 percent (26 weeks)
• Tier 2: 5 percent (33 weeks)
• Tier 3: 6 percent (46 weeks)
• Tier 4: 7 percent (59 weeks)
• Tier 5: 8 percent 72 weeks)
• Tier 6: 9 percent (85 weeks)
• Tier 7: 10 percent (98 weeks)

To be concrete, if someone is in a state with a 10 percent unemployment rate while the national unemployment rate is 7 percent or higher, they would be in Tier 7 (with a PBD of 98 weeks). Similarly, if someone is in a state with 7 percent unemployment rate while the national rate is 10 percent, they would also be in Tier 7. What is the rationale for a national-level trigger beyond just a state-level one? If there is a clear signal that we are entering a national recession, it is useful to provide broad-based support and not wait until a particular state has officially crossed a threshold. State and national unemployment are both imperfect proxies for local labor market conditions, but state-level estimates are noisier signals, especially for smaller states. For this reason, if we see national level unemployment rate climb above 5 or 7 percent, it provides important information about the likely trajectories of the local labor market.

If the proposed triggers were in place, what would the PBD have looked like in the United States over the past two decades? To answer this, I simulate state-specific PBD based on state and national triggers. Figure 1 plots the average PBD across the United State over time, along with the national unemployment rate.

As figure 1 shows, during the extended Great Recession period (2008–12), the proposed policy would have raised the (average) PBD to levels that are broadly similar to what was discretionarily achieved via Emergency Unemployment Compensation over that period. For example, in 2011 the average PBD across states was 87 weeks, while under my proposal it would have been 85 weeks. Seventeen states would have had greater PBDs and 34 states would have had lower PBDs under my plan. At the same time, under my proposal the PBD would have increased more quickly during the Great Recession, making it a better automatic stabilizer. Under the proposed rules, the average PBD would have increased from around 30 weeks to 85 weeks between 2007 and 2009. In contrast, the PBD in reality increased from 26 to 68 weeks between those periods, and it took until 2011 for the PBD to catch up to my proposed rule. Another benefit is that by using triggers the phasing out of extended PBD would be smoother. For example, the Emergency Unemployment Compensation expired in 2014 rather abruptly, leading PBD to fall from as much as 99 weeks to 26 weeks in some states. In contrast, under my proposal PBD would have fallen more gradually based on both national- and state-level unemployment rates, falling to around 40 weeks by 2015 and to 26 weeks by 2017 when the labor market was considerably tighter.

Finally, in 2019—a year with a tight labor market—the average PBD was around 25 weeks while under my plan it would have been 27. In general, my proposed triggers would match the severity of the crises. For example, in the aftermath of the shallower 2001 recession, average PBD would have risen to a little under 50 weeks for a brief period, and stayed mostly under 40 weeks during the recovery period. In contrast, during the onset of the current crisis, PBD would have shot up to 98 weeks, but fallen as the labor market improved. However, it would have been close to 60 weeks in February 2021, which is broadly similar to where we were in actuality (average of 58 weeks).

Overall, while the proposed changes would alter the distribution of PBDs across place and time (to better match economic necessity), it would only modestly increase the overall number of weeks. The average PBD in the 2015–19 period would have been 30 weeks under my proposal, while in reality it was 25 weeks. During the 2010–14 period, the average PBD would have been 71 weeks under my proposal, as compared to 62 weeks in reality.

Proposal 4: Restructure and Increase the Benefit Replacement Rate
I propose the following benefit replacement determination. Regular UI benefits would be based on the AWW as follows (assuming that maximum and minimum benefits do not bind):

• First $400 of AWW: 80 percent marginal replacement
• rate
• Next $300 of AWW: 65 percent marginal replacement
• rate
• For additional AWW above $700: 50 percent marginal
• replacement rate

I propose setting the maximum benefit level at 80 percent of AWW, and a minimum benefit level at 20 percent of AWW in the United States. For example, in 2020 this would have meant a maximum benefit of $910 and a minimum benefit of $230. This is close to what states with more-generous benefits currently do; for example, in Washington State the minimum benefit level is $201 while the maximum benefit is $844. In addition, replacement rates would be capped at 100 percent of earnings for all workers. The benefits would be calculated based on the high-quarter method, as is done in 29 states currently, which protects workers with nonstandard and limited work histories. Specifically, this proposed change would take the highest quarter in the relevant base period for the purpose of benefit calculations. To show how the current and proposed replacement rates look for different workers, figure 2 shows the average replacement rates by earnings deciles of those who are unemployed. These data are based on pre-pandemic 2019 unemployed workforce in the Current Population Survey (CPS), based on the methodology developed in GNV. More details are provided in appendix A.

As we can see, the average replacement rate for those in lower deciles is around 70–75 percent, declining to 40 percent at the top decile. In contrast, the current replacement rates (averaged across states) are around 50 percent for the bottom two-thirds of the workforce, declining to around 20 percent at the top. Moreover, the averages in the current replacement rates mask considerable heterogeneity by state. In contrast, under my proposal replacement rates would not vary by state for workers with the same earnings levels. Based on the simulations, the average weekly benefit levels in 2019 would have risen from $338 to $547 under my proposal, leading to an average replacement rate of 72 percent (as compared to the estimated current replacement rate of 44 percent).

In addition, I propose a boost to benefit levels during downturns to aid the automatic stabilizer role of UI policy, and because it is less costly (in terms of efficiency losses) to provide more-generous benefits during downturns. In particular, the recessionary boost would entail

• An additional $100/week boost above 6 percent national TUR; and
• An additional $200/week boost above 8 percent national TUR.

Figure 3 shows the same distribution, but now with a recessionary boost of $100/week. Here we see higher replacement rates throughout the distribution. A substantial portion of the beneficiaries in the bottom half of the distribution would be at the 100 percent replacement cap during downturns. The beneficial stimulus effects and the reduced cost of moral hazard (due to search externalities) allows a moregenerous benefit level during downturns.

To estimate the resulting increase in outlays on UI benefits, I simulate the stimulus spending that would have occurred had the recessionary boosts been in place since 2000. To clarify, these are just the added spending from the recessionary boost assuming the recipiency rates in this proposal; I will provide estimates for the overall added spending—including the change in regular benefit levels and recipiency rates—in the subsection on proposal costs below. In addition, note that the estimates for 2020 do not account for payments from FPUC. Figure 4 plots the simulated annual federal spending from the recessionary boost over this period.

Here are the key findings: During the extended Great Recession period (2008–14), the proposed recessionary boost would have provided an additional $320 billion or so in added stimulus—or roughly $45 billion per year. The proposed recessionary boost is proportionate to the size of the downturn: for example, it would have added a smaller boost of around $35 billion during the 2001 recession. Finally, during the current crisis, the automatic recessionary boost would have provided around $40 billion in added spending in 2020. In comparison, policymakers spent around $250 billion toward the boost in benefits in 2020. However, this comparison does not factor in the added spending that occurs under my plan from more generous baseline benefits. Factoring those in, as I do in the proposal costs subsection below, suggests a total of around $150 billion in added expenditures in 2020 from the proposed increase in benefit levels as compared to normal replacement rates (i.e., excluding FPUC or LWA).

It is also useful to keep in mind how UI income interacts with other safety-net programs to affect income after taxes and transfers. In particular, the Supplemental Nutrition Assistance Program (SNAP) is an important program that has important countercyclical properties. There is a strong distributional argument to exclude the UI recessionary boosts for the purpose of SNAP eligibility determinations in order to help those at the very bottom, to avoid SNAP eligibility reduction to crowd out the recessionary boost. Such an exclusion was adopted as part of the December 2020 relief bill, but only temporarily. Under my proposal the UI recessionary boost income would be excluded for SNAP eligibility determination at all times.

Proposal 5: Strengthen STC
Incorporate STC into UI at the federal level, and substantially increase administrative capacity and funding to process applications in a timely fashion. Streamline the employer application process by allowing online applications, using data from the regular employment and wage (ES-202) employer filings as a default, allowing employers to choose from the list of existing workers who they wish to enroll in STC. Information about hours should already be collected as part of the regular employer filings, making this process easier. Moreover, similar to what was suggested by von Wachter (2020), the federal government should allow employers to pay workers on STC directly, and should compensate employers for the cost through the STC program.

Increase awareness about the program through a major information
campaign using strategies that have been found to work (as in Houseman et al. 2017), and test the efficacy of other messaging campaigns. This echoes the proposal made by Abraham and Houseman (2014).

Allow employers to reduce hours by as much as 80 percent of enrolled workers, which will allow greater employer participation. This was proposed by then-candidate Biden in his 2020 position statement (Biden Harris 2020).

Provide financial incentives to employers to use STC by providing a refundable tax credit to reimburse employers for the added costs of providing full health benefits for workers during the period of reduced work hours. This also was proposed in the Biden position statement (Biden Harris 2020).

Proposal Costs

I provide two different cost estimates: one during expansionary periods, and another during downturns. I provide the incremental cost for each of the five proposals in the same order as the proposals discussed above. These are incremental costs and build on each other; added up, they provide the overall change in costs as compared to the status quo.

Costs During Expansionary Periods
Averaged over the 2015–19 period, the actual average UI outlays were around $30 billion/year. The federal administration of UI in Proposal 1 would not have any direct impact on program outlays. Under Proposal 2, I assume that recipiency rates among the unemployed would rise to around 55 percent, while 10 percent of the unemployed would be enrolled in the JA. While there is a lot of uncertainty around these estimates, when I average over the 2015–19 period I estimate that the proposed increase in recipiency rates would have led roughly to a $42 billion increase in annual outlays, combining both regular UI and the JA.

The tying of PBD to triggers in Proposal 3 has a modest positive impact on overall PBD. In the 2015–19 period the average PBD would have increased from 25 to 30 weeks under my proposal. Assuming that actual benefit durations increase by around 2 percent for a 10 percent increase in PBD (midpoint in the literature), I find that this would have led to a $3 billion increase in costs over the 2015–19 period.

To assess the impact of the benefit increases in Proposal 4, I simulate actual and proposed benefits in a non-recessionary year (2019), which were shown in figure 3 above. Averaged over the 2015–19 period, I estimate that outlays would have been around $42 billion higher due to the higher benefit levels in my proposal. To be clear, this is the incremental benefit of adding the changes to the replacement rate, factoring in the increases in recipiency rates from Proposals 2 and 3. Put differently, without assuming the increase in the recipiency rate to 55 percent (and another 10 percent through the JA), the incremental cost of higher replacement rates would have been about half as big.

The primary impact of Proposal 5 is changing the way in which workers experience reduced employment, and how they are compensated through the UI system. The added outlays would mainly be from the refundable tax credits to compensate for health benefits. It is difficult to estimate specific costs given uncertainties about how much STC will actually be taken up by employers. I estimate that as much as 10 percent of the UI recipients may shift to STC as a result of the policy. Assuming that half the covered workers are receiving employment-based health benefits, and an average cost of insurance (blending between single and family plans) of $10,000, this would imply around $2 billion additional annual outlays.

Overall, under my proposal, during expansionary periods, UI outlays would increase by around $89 billion, reaching $119 billion a year instead of the current $30 billion annual outlay. The major cost increases are roughly half due to increased eligibility, and half due to increased benefit levels. However, note that this decomposition depends on the order in which we account for the proposals. Had I first considered the benefit levels and then the eligibility expansion, the vast majority of increased outlays would have been due to eligibility expansion.

Under my proposal, employee-based payroll taxes would pay for regular benefits that are part of Tiers 1 and 2; payroll taxes would not pay for any cost increases from the JA, or from the the health-care subsidy costs of STC, all of which would be paid for by general revenue. As a result, the annual outlays that need to be paid for using payroll taxes would be less than the $119 billion, and are likely closer to $100 billion. During the 2015–19 period, average employer payroll taxes were around 0.6 percent of total wages. I estimate that the payroll taxes needed to cover the outlays under my proposal would amount around 2 percent of total wages.

Costs During Severe Downturns
To estimate the outlays during a severe downturn or slack labor market, I follow a similar strategy as above, but now average over the 2008–12 period. The actual average UI outlays over this period were around $102 billion/year, including the ad hoc changes to the programs that policymakers made during those years. The expansion of eligibility from Proposal 2 would add $30 billion annually to the outlays during that downturn. The change in PBD from my Proposal 3, factoring in both mechanical and behavioral responses, would be around $3  billion. To calculate the average benefit levels under my proposal during that period, I use the 2011 CPS. Incorporating the recessionary benefit boosts as well as regular UI benefit changes, I find that reforming the A Plan to Reform the Unemployment Insurance System in the United States 17 benefit structure in Proposal 4 would cost $132 billion more during that downturn than the cost of benefits in place in those years. The increased outlays from Proposal 4 are substantially greater during downturns, since the recessionary boosts will provide stimulus to the economy. Finally, I estimate STC health subsidies would cost around $4  billion annually during downturns, using the same assumption on increased take-up as before. Overall, the average outlays would increase by $170  billion/year relative to the outlays in those years, leading to total average outlays of $272 billion under my current proposal during the 2008–12 severe downturn period. Of that $272 billion, approximately $152 billion of those benefits would be financed out of the overall federal budget and $120 billion would be financed by the payroll tax revenue. Since all of the additional costs beyond regular UI benefits in Tiers 1 and 2 in my proposal are to be financed by general federal revenue, the combination of increased eligibility and benefit generosity would provide substantial amount of stimulus during downturns.

Questions and Concerns

Are any parts of the proposal feasible without federalizing UI?

Moving to a fully federally funded and administered system is a key plank of my proposal: there are few benefits and many costs of our current hybrid system. At the same time, transitioning to a fully federal system has costs and requires overhauling the administrative structure of UI. There are many institutional arrangements that can help facilitate the transition process. For example, state agencies and personnel can continue to play a role in administering benefits while UI becomes a federal program.

It is important to emphasize that much of my other proposals can be implemented under the current system as long as the federal government is able to provide financial incentives and support for the states to adopt these reforms. This includes adopting a uniform eligibility and benefit determination, STC, as well as the PBD tiers system as described above. This would require major funding to overhaul antiquated software, hire additional personnel, and adopt national standards. The federal government in return would fully fund the reformed EB system for (say) Tiers 3 and above, as well for the recessionary boosts and for costs associated with STC. Overall, the funding balance would strongly move toward the federal government, giving a substantial incentive for states to adopt the necessary standards. In this sense, the fully federal administration of the UI program—while greatly desirable—is also severable from the other key parts of this proposal.

What is the role of evidence and evaluations?

To evaluate the impact of more-ambitious policies, we have to first experiment with more-ambitious policies. We did that during the Covid crisis, which allowed us to expand our thinking about UI generosity. However, we will need to further evaluate any enacted changes, and use evidence to guide any course correction.

As part of any legislation, we should incorporate rigorous program evaluations over the course of one to two years following implementation. Researchers can use quasi-experimental approaches to study the impact of higher benefits on both micro and macro employment effects, as well as other outcomes during unemployment spells, including changes in consumption, savings, self-reported well-being, health, and food security. The federal government can make high quality matched administrative data (linked Longitudinal Employer-Household Dynamics [LEHD] data to UI recipiency records and other outcome measures as feasible) available to researchers for this purpose. In addition, as I recommended in this proposal, all employers should be required to report employee hours and not just earnings; this is done currently in only four states. This requirement will aid the evaluation objectives as well. We should use these findings to finetune policy details, including possibly altering the benefit levels or PBD to achieve a better balance, should changes be needed.

Why only use TUR triggers and not IUR triggers?

In my main proposal, I suggest reforming the EB program using only TUR triggers. In practice, IUR triggers had rarely been activated for the EB program; moreover, TUR provides a more-accurate overall state of the labor market. For example, IUR will depend on take-up and eligibility of UI recipients, which varies greatly across time and space. It is difficult to know how to set effective IUR triggers without knowing what the impact of the proposed changes will be on actual recipiency rates. At the same time, it is true that the IUR is calculated at a higher-frequency level, which is helpful for faster response. This was true during the Covid crisis, but is unlikely to be relevant in most downturns. As an alternative, we can further modify the proposal by allowing a national IUR trigger for rapid response.

If reforming the replacement rates is difficult and takes time, is it possible to implement a simpler reform to the benefit levels, especially in the short term?

In Proposal 4, I propose marginal replacement rates based on earnings brackets, which allows greater control over how the average replacement rate changes by the level of earnings. If this were infeasible, especially in the short run due to administrative constraints, an alternative would be to add a simple $100 boost to all recipients in normal times, instead of the earnings bracket–based determination. The additional recessionary boosts of $100 and $200 would remain as is. In appendix A I show that the average replacement rates by earnings level from this simpler plan are broadly similar to what I suggest in the proposal.

Does the proposal target the right level of benefit generosity?

While 50 percent replacement may be sufficient for middle and high-income workers with sufficient assets and access to credit, it is insufficient for those who are at the bottom of the pay scale. Those with low earnings may already have had difficulty meeting needs if they had low hours; in this case 50 percent of average earnings can be far too low for meeting basic needs. Moreover, while the replacement rates are higher during recessions due to the recessionary boost, the evidence suggests that even smaller downsides in terms of moral hazard, and more upside in terms of stimulus, from greater generosity during downturns. And finally, capping benefits at 100 percent of earnings (as long it is above the minimum benefit level) guards against overly dis-incentivizing work.

Will the increased benefit levels hold back recovery and growth?

Based on the evidence, the overall increase in the benefit level would likely have a moderate impact on the exit out of unemployment benefits and recovery, especially during normal times. During downturns, the employment effects would be even smaller since jobs are rationed, and would possibly be positive due to the stimulus effects.

Why use earnings and not hours-based threshold?

In this proposal, I have suggested using realistic earnings based thresholds. This can help with expanding eligibility. At the same time, there are compelling arguments for eventually moving away from earnings toward an hours-based determination of eligibility. The hours-based thresholds can be 120 hours in one quarter and 60 hours in a second quarter. By collecting hours information from employers as part of UI tax reporting, as is done now in four states, we can make this easier.

Would having a single national standard for benefit determination lead to worsened standards in some states?

The proposed benefit levels would be more generous than those in any state. However, the federal program can accommodate states’ desire to go above the federal standard. In that case, the payroll taxes for employers/employees in those states would need to be adjusted to cover the additional expenses.