What factors contributed to changes in employment during and after the Great Recession?

Farooq, Ammar; Kugler, Adriana D. Article What factors contributed to changes in employment during and after the Great Recession? IZA Journal of Labor Policy Provided in Cooperation with: IZA – Institute of Labor Economics Suggested Citation: Farooq, Ammar; Kugler, Adriana D. (2015) : What factors contributed to changes in employment during and after the Great Recession?, IZA Journal of Labor Policy, ISSN 2193-9004, Springer, Heidelberg, Vol. 4, pp. 1-28, https://doi.org/10.1186/s40173-014-0029-y This Version is available at: https://hdl.handle.net/10419/154702 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ ORIGINAL ARTICLE Open Access What factors contributed to changes in employment during and after the Great Recession? Ammar Farooq 1 and Adriana D Kugler 2* * Correspondence: [email protected] 2 Georgetown University, NBER, CEPR and IZA, Old North 311, 37th and O Streets, NW, Washington, DC 20057-1036, USA Full list of author information is available at the end of the article Abstract Unemployment increased drastically over the course of the Great Recession from 4.5 percent prior to the recession to 10 percent at its peak in October 2009. Since then, the unemployment rate has come down steadily, and it stood at 5.8 percent in November 2014. Based on existing analyses and some new evidence, this paper establishes that much of the change in unemployment during the Great Recession and during the recovery can be attributed to cyclical factors rather than structural factors. The paper then presents new suggestive evidence to quantify the employment impacts of various counter-cyclical policies introduced during this time. We conduct a counter-factual and find that employment would have been between 4.2 percent and 4.5 percent lower had it not been because of the spending in Medicaid injected in local economies by the Recovery Act. In addition, we conduct a differences-in-differences and triple difference analysis, which suggests that the Work Opportunity Tax Credits increased the likelihood of employment by about 4.7 percent for disconnected youth but had no effect on disabled and unemployed veterans. Finally, we also find evidence that suggests that the Hiring Incentive to Restore Employment (HIRE) Act increased employment of the unemployed by 2.6 percent and that the reemployment reforms introduced in 2012 as part of the UI extensions increased employment by 6 percent for the long-term unemployed. JEL codes: JE24, J23, J63, J64, J65, J68 Keywords: Employment; Labor demand; Unemployment incidence; Job search; Unemployment insurance; Tax credits; Hiring subsidies 1 Introduction In December 2007, the US economy entered into the deepest recession since the Great Depression. Like other recessions precipitated by financial crises, the Great Recession was accompanied by a substantial contraction in aggregate demand. Following the crisis, there were sharp drops in consumer spending and to a lesser extent in investment. This translated into a drastic fall in output, which hit a low point in the fourth quarter of 2008 with a contraction of real GDP of 8.3 percent. Figure 1 shows that by the beginning of 2010, the economy had turned around in terms of GDP, consumption and investment, though these remain low by historical standards. Just as the goods, services, financial, credit, and housing markets were all affected by the Great Recession, so was the labor market. The sharp contraction in demand © 2015 Farooq and Kugler; licensee Springer. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. Farooq and Kugler IZA Journal of Labor Policy (2015) 4:3 DOI 10.1186/s40173-014-0029-y generated massive layoffs, a sharp drop in employment and a rise in unemployment. While total employment and private sector employment have been growing steadily by 178,000 jobs per month and 188,000 jobs per month on average, respectively, since February 2010, 1 the loss of jobs during the Great Recession was so great that the economy only recently (in May 2014) recovered the 8.7 million jobs it lost. Unemployment reached a peak of 10 percent in October 2010 and has been declining steadily since. Indeed, the share of the short-term unemployed has returned to pre-recession levels, so the continued high unemployment reflects the larger share of long-term unemployed during the recovery. As explained below, the large share of long-term unemployed has important implications in terms of the speed at which the unemployment rate can continue to fall. In Section 2, we discuss the severity of the labor market downturn during the Great Recession. In Section 3, we review past evidence and present some new evidence on the extent to which unemployment during the Great Recession and the recovery can be attributed to cyclical or structural factors. In Section 4, we explain the policy tools used during this period and present some evidence of their employment impacts during the recession and later during the recovery. We conclude in Section 5. 2 The labor market during the Great Recession and recovery As the Great Recession evolved, the unemployment rate rose sharply from a low of 4.5 percent the year prior to the recession in June 2007 to a peak of 10 percent in October 2009. As Table 1 shows, this rise of 5.5 percentage points in the unemployment rate between the 2007 low and the 2009 high is unprecedented for any post-WWII recession period. However, this table shows that the drop of 4.2 percent after the recession is higher than the average drop of 2.75 percent over the last nine previous recessions. While it is argued that this recovery has been particularly slow, this is not apparent from the drops in unemployment after previous recessions shown in Table 1 and in Figure 2. In fact, the fall in the unemployment rate has been greater during this most recent recovery than during the recovery of the early and late 1950s, the late 1960s, the early 1970s, the early 1990s, and the early 2000’s. What has been atypical relative to all other recoveries going back to the early 1970s is that government jobs have not contributed to this recovery. In fact, the fall in Figure 1 US GDP, Investment and personal consumption, 2006 (quarter 1)-2013 (quarter IV). Source: Bureau of Economic Analysis. Farooq and Kugler IZA Journal of Labor Policy (2015) 4:3 Page 2 of 28 government jobs has held back the economy from a healthier employment recovery. Figure 3 shows government employment changes during recoveries of the 1970s, 1980s, 1990s, and 2000s as well as during the most recent recovery. This figure shows that every other recovery has been accompanied by government job growth, with the exception of the recovery from the Great Recession. While the role of government jobs on employment during previous recoveries had always been positive, the role of Table 1 Changes in unemployment rates during each post-war recession and recovery Recession Lowest unemployment rate in the year before start of recession Highest unemployment rate during or after the official end of recession Difference from troughtopeak Nov-48 3.7 7 3.3 Jul-53 2.6 6 3.4 Aug-57 4 7.4 3.4 Apr-60 5.2 7 1.8 Dec-69 3.4 6 2.6 Nov-73 4.8 8.9 4.1 Jan-80 5.9 10.7 4.8 Jul-90 5.3 7.6 2.3 Mar-01 3.9 6.2 2.3 Dec-07 4.5 10.0 5.5 Recession Highest unemployment rate during or after the official end of recession Lowest unemployment rate in the recovery (up to 6 years after the recession) Difference from peak to trough Nov-48 7 2.6 4.4 Jul-53 6 4 2 Aug-57 7.4 5.1 2.3 Apr-60 7 3.7 3.3 Dec-69 6 4.8 1.2 Nov-73 8.9 5.7 3.2 Jan-80 10.7 6.6 4.1 Jul-90 7.6 5 2.6 Mar-01 6.2 4.5 1.7 Dec-07 10.0 5.8 4.2 Figure 2 U.S. Unemployment rate, 1965-2013 (Dec). Source: Bureau of Labor Statistics-Current Population Survey. Farooq and Kugler IZA Journal of Labor Policy (2015) 4:3 Page 3 of 28 government jobs has been declining over the past decades since the 1970s. By contrast, the recent recovery has seen a large decline in government jobs, with the exception of the few months surrounding the hiring for the 2010 Census. In fact, close to 600,000 government jobs have been shed during this recovery. Much of the rise and sustained unemployment can be explained by private sector job losses during the recession and public sector job losses during the recovery. In addition, a big part of the sustained unemployment can be explained by the inability of people to exit unemployment as private sector employers have not been creating enough jobs, and public sector employers have not created jobs on net during this period. Figure 4 shows the drastic increase in layoffs during this time, with layoffs increasing by over 50 percent during the Great Recession. The vast majority of these discharges were the result of mass layoffs of more than 50 individuals. At the height of the recession, all dismissals were due to mass layoffs. Yet, there were also many experiencing individual layoffs before the peak of the recession. Since then, employers have substantially reduced mass and individual layoffs, and these are currently below the pre-recession levels in 2006. As shown in Figure 5, because there were so many individuals losing their jobs and so few new jobs created, the ratio of the number of unemployed to job vacancies reached a high of close to 7 unemployed per vacancy in July 2009 –the highest ratio since JOLTS data has been collected. As fewer people entered into unemployment and more jobs have been added, this has now dropped to about 3 unemployed per vacancy. Figure 6 shows that the chances of finding a job within the last month for an individual who has been unemployed for less than 6 months is over 20 percent, and the chances of finding a job within the last month for an individual who has been unemployed for more than 6 months is a little over 10 percent. This is because while job-to-job turnover substantially declined during the recession, there are still many employed looking for work that enter new jobs directly from other employment. 2 For example, even today, there are still 6.9 million part-time workers who would rather be working full time and who continue to look for full-time work. In addition, there are Figure 3 Government job creation in current and previous recoveries, 1975-2009. Source: Bureau of Labor Statistics-Current Employment Statistics survey. Farooq and Kugler IZA Journal of Labor Policy (2015) 4:3 Page 4 of 28 many individuals among those classified as out of the labor force that would like a job and would enter directly into jobs. There are 2.1 million marginally attached workers who are willing and available for work but did not look for work during the past month, which means that many workers entering new jobs come directly from the pool of individuals out of the labor force. 3 The evidence above shows that the shortand long-term unemployed are not equally likely to find jobs, and the unemployed, and in particular the long-term unemployed, may have to compete for jobs with currently employed workers and even with workers who have been out of the labor force. Indeed, there is evidence of duration dependence, which implies that it is harder to exit unemployment the longer an individual has been unemployed. Kroft et al. (2014) indeed find that duration dependence can account for much of the increase in long-term unemployment. The share of the long-term unemployed increased sharply during the Great Recession. While the long-term unemployed comprised less than 15 percent of all the unemployed prior to the recession, Figure 4 Layoffs and discharges and mass layoffs, 2006–2013. Source: Bureau of Labor StatisticsJob Openings and Labor Turnover Survey & Mass Layoff Statistics. Figure 5 Ratio of unemployed-to-job openings, 2006–2013. Source: Bureau of Labor StatisticsJob Openings and Labor Turnover Survey & Current Population Survey. Farooq and Kugler IZA Journal of Labor Policy (2015) 4:3 Page 5 of 28 by April 2010, 44 percent of the unemployed had been unemployed for more than 6 months. Figure 7 does show a slow decline in the share of the long-term unemployed in the past two years, with this share now standing at 30.7 percent –still high by prerecession standards. There are a number of reasons why the long-term unemployed may find it harder to find jobs than the short-term unemployed and contribute to keeping the share of the long-term unemployed and overall unemployment high. First, employers may simply take employment status or duration of unemployment as a signal of worker quality. Even though the extent of mass layoffs suggests that employment status and unemployment duration were probably relatively bad signals of quality during the Great Recession, 4 in their audit study, Kroft et al. (2012) find evidence of substantial statistical discrimination against the long-term unemployed. Second, the unemployed, and in particular the long-term unemployed, may find it much more difficult to find a job because individuals may lose their skills and motivation as they remain longer in Short-Term Unemployed Long-Term Unemployed Figure 6 Flows from unemployment to employment for shortand long-term unemployed, 1995–2012. Source: Bureau of Labor Statistics-Current Population Survey. Figure 7 Incidence of long-term unemployment, 2006–2013. Source: Bureau of Labor Statistics-Current Population Survey. Farooq and Kugler IZA Journal of Labor Policy (2015) 4:3 Page 6 of 28 unemployment. This may be because they are either less skilled or less desirable to hire as their spells of unemployment prolong. Third, the unemployed, may have less access to information about jobs because those in their networks may also be unemployed, and many jobs are filled through informal channels. There is anecdotal evidence that many employers turned to using informal channels during the recovery as a way to save on recruiting costs given the abundance of potential applicants. In fact, the share of individuals searching for work through family and friends grew from 19.7 percent in 2006 to 26.4 percent in 2007, then to 28 percent in 2008, and has stayed at over 32 percent since 2009. 5 This increased reliance on family and friends as a way to find jobs has occurred even though networks are likely less effective in generating job offers on average given the higher unemployment today. 6 Finally, as the financial assets of the unemployed deplete the longer they have been in unemployment, it becomes harder to pay for transportation, to move to take a job and to pay for other costs associated with looking for jobs. This section shows that the Great Recession was atypical in terms of the drastic impact it had on the labor market. All measures show large losses for workers, including extensive mass layoffs, widespread drops in employment, a steep rise in unemployment, and greater difficulty in finding employment. In addition, this section shows that the recovery after the Great Recession was different in that government jobs have delayed rather than sped up the recovery of the labor market, as in past recessions. Finally, even though the labor market is back to where it was in terms of most indicators, the Great Recession has been different in that it has had a long-lasting impact in the labor market because of its effect on long-term unemployment. The share of long-term unemployed is twice the pre-recession share, and the long-term unemployed face numerous difficulties finding employment even as the economy has continued to recover. 3 Cyclical vs. structural unemployment during the great recession and beyond In this section, we explore how much of the unemployment during the Great Recession and beyond was cyclical and how much reflected mismatches and other structural challenges faced by the economy. Below, we review the large body of past evidence and present some new evidence pointing to the unemployment problem being largely driven by cyclical factors during this period. Yet, the evidence also indicates that some new problems have emerged in the labor market which could potentially turn into structural factors if they persist over the years to come. 3.1 Evidence from Okun’s law A number of previous studies have re-estimated Okun’s Law during the Great Recession and recovery. These studies show that much of the change in unemployment is due to changes in GDP and that the persistent high unemployment is due to a slow recovery in GDP. Arteta et al. (2011) show that the majority of movements in unemployment since the Great Recession were due to changes in output. According to this study, the drop in GDP can explain 63 percent of the rise in unemployment during the recession. Similarly, about 57 percent of the drop in unemployment during the recovery can be explained by the rise in GDP. This indicates that both during the recession and during the recovery, cyclical factors were key in explaining changes in unemployment and that cyclical factors have remained important during the recovery. Importantly, Ball et al. (2012) find that the relation between the change in Farooq and Kugler IZA Journal of Labor Policy (2015) 4:3 Page 7 of 28 unemployment and the change in GDP has been stable over the periods 1990–91, 2001, and 2007–09. Rather, they conclude that the strength of economic growth relative to the trend is what has been different during these periods. In Okun’s law, the rest of the changes in unemployment are considered as unexplained or explained by factors not included in the regression. These could be factors such as skill, sectoral, or regional mismatches. 7 Thus, an extension of Okun’sLaw would include measures of mismatches in the regression. Estevão and Tsounta (2011) estimate a relation of changes in unemployment at the state level on changes in gross state product (GSP), a measure of skill mismatches in the state, and a measure of geographical immobility. This study finds that much of the change in state unemployment can be explained by GSP and that only about 0.5 percentage points of the increase in the NAIRU can be explained by skill gaps. 3.2 Evidence from the Beveridge curve The Beveridge Curve, which establishes the relation between the job openings rate and the unemployment rate, is yet another way to disentangle how much of the change in unemployment is due to cyclical factors and how much is due to other factors. The Beveridge Curve, estimated by the Bureau of Labor Statistics over the last decade, 8 shows that during the early 2000s, unemployment increased and the job openings declined starting in November 2001 until around November 2007. These movements along the Beveridge Curve are consistent with cyclical factors driving the changes in unemployment during this period. In December 2007, the unemployment rate increased rapidly together with much slower drops in the job openings rate. These movements since 2007 are also consistent with cyclical factors driving these changes. In October 2009, the unemployment rate started to drop and the job openings rate started to rise. The backward movement to the Northwest points to a decline in cyclical unemployment. Yet, the fall in unemployment has not been fast enough to match the rise in the job openings rate. This indicates that for a given job openings rate, the unemployment rate is higher than it used to be, pointing potentially to the riseintheimportanceofstructuralfactors. A number of studies have examined and explained this shift in detail. Diamond (2013) explains that the Beveridge Curve may be using proxies rather than the correct measures of those searching for work and the right measure of job openings. Indeed, the unemployment rate may not capture everyone looking for work. Diamond (2013) does a thorough analysis of flows from and to employment, unemployment, and out of the labor force and finds transitions of similar magnitudes from non-employment to employment and back as from unemployment to employment and back. This is indicative that many of those classified as out of the labor force may be actively looking for work. In addition, Davis et al. (2012) have found that the speed of filling vacancies and the proportion of hiring varies by industry and over the business cycle. For instance, the industry composition of job openings has been changing over time, with many fewer job openings in construction, which have short durations, and many more job openings in health and education, which have longer durations. Given that long durations imply low closing rates of vacancies or high job opening rates, then one may worry that changes in the composition of vacancies may be accounting for the high opening rate associated with the same unemployment rate as before. Farooq and Kugler IZA Journal of Labor Policy (2015) 4:3 Page 8 of 28 Table 2 Descriptive statistics for different WOTC eligible groups Full sample Veterans Disabled veterans Disabled Veterans Before WOTC Eligibility (Pre-2008) Disabled veterans during WOTC eligibility (Post-2008) Unemployed veterans Unemployed Veterans Before WOTC eligibility (Pre2010) Unemployed veterans during WOTC eligibility (Post-2010) Youths (16–24 year olds) Disconnected youth Disconnected youth before WOTC eligibility (Pre2010) Disconnected youth after WOTC eligibility (Post2010) Employed 0.729 (0.444) 0.730 (0.444) 0.549 (0.498) 0.561 (0.496) 0.545 (0.498) 0.627 (0.484) 0.643 (0.479) 0.604 (0.489) 0.485 (0.500) 0.340 (0.474) 0.350 (0.477) 0.326 (0.469) Weeks looking for work 1.405 (6.162) 1.364 (6.070) 1.244 (5.803) 1.057 (5.203) 1.310 (6.012) 20.90 (12.66) 19.68 (12.27) 22.72 (13.02) 1.597 (6.550) 4.706 (12.05) 4.151 (11.32) 5.557 (13.06) Male 0.480 (0.500) 0.903 (0.296) 0.896 (0.306) 0.905 (0.293) 0.891 (0.312) 0.894 (0.308) 0.896 (0.306) 0.892 (0.311) 0.508 (0.500) 0.472 (0.499) 0.452 (0.498) 0.502 (0.500) Less than H.S. 0.117 (0.321) 0.0373 (0.189) 0.0326 (0.178) 0.0427 (0.202) 0.0276 (0.164) 0.0390 (0.194) 0.0435 (0.204) 0.0322 (0.176) 0.431 (0.495) 0.308 (0.462) 0.341 (0.474) 0.259 (0.438) H.S. diploma 0.299 (0.458) 0.329 (0.470) 0.264 (0.441) 0.268 (0.443) 0.260 (0.439) 0.370 (0.483) 0.374 (0.484) 0.364 (0.481) 0.210 (0.407) 0.432 (0.495) 0.426 (0.495) 0.442 (0.497) Some college 0.294 (0.456) 0.380 (0.485) 0.444 (0.497) 0.448 (0.497) 0.442 (0.497) 0.418 (0.493) 0.415 (0.493) 0.423 (0.494) 0.299 (0.458) 0.180 (0.384) 0.162 (0.369) 0.206 (0.404) Bachelor’s and higher 0.290 (0.454) 0.253 (0.435) 0.260 (0.439) 0.241 (0.428) 0.271 (0.444) 0.173 (0.378) 0.167 (0.373) 0.181 (0.385) 0.0600 (0.237) 0.0799 (0.271) 0.0709 (0.257) 0.0937 (0.291) White 0.795 (0.404) 0.809 (0.393) 0.793 (0.405) 0.808 (0.394) 0.787 (0.409) 0.793 (0.405) 0.801 (0.399) 0.782 (0.413) 0.796 (0.403) 0.753 (0.432) 0.756 (0.430) 0.747 (0.435) Black 0.115 (0.319) 0.128 (0.334) 0.140 (0.347) 0.127 (0.333) 0.144 (0.351) 0.136 (0.342) 0.132 (0.338) 0.142 (0.349) 0.108 (0.310) 0.143 (0.350) 0.145 (0.352) 0.141 (0.348) No. Observations 141,3820 100,591 8,223 3,140 5,733 6,517 3,906 2,611 260,842 32,656 19,766 12,890 Notes: Mean coefficients; SD in parentheses. Unemployed Veterans are defined as veterans who had searched for more than 4 weeks for work during the last year. Disconnected Youth are defined as 16–24 year olds who are not enrolled in school and who worked less than 26 weeks in the previous year. Data is from Annual Social and Economic (ASEC) supplement of the Current Population Survey spanning the years 2003–2013. Farooq and Kugler IZA Journal of Labor Policy (2015) 4:3 Page 15 of 28 10 20 30 40 50 Percent 2003 2004 2005 2006 2007 Year Disabled Veterans Disabled Non-Veterans Figure 9 Employment rates of disabled veterans and non-veterans. 50 55 60 65 70 Percent 2002 2004 2006 2008 2010 Year Non-Veterans Veterans Figure 10 Employment rates of unemployed veterans & non-veterans. Farooq and Kugler IZA Journal of Labor Policy (2015) 4:3 Page 16 of 28 Table 3 Difference-in-difference (DD) effects of work opportunity tax credits (WOTC) on disabled veterans and unemployed veterans All veterans Recent veterans only (1) (2) (3) (4) (5) (6) (7) (8) Disabled veteran X WOTC eligibility period for disabled veterans 0.0491*** (5.57) 0.0471*** (4.83) 0.0499*** (5.69) 0.0501*** (5.72) −0.0127 (−0.30) −0.0115 (−0.30) −0.0043 (−0.10) 0.0003 (0.01) Disabled veteran dummy −0.237*** (−24.95) −0.212*** (−27.84) −0.239*** (−25.13) −0.242*** (−25.31) −0.0641 (−1.48) −0.0635* (−1.72) −0.0706 (−1.62) −0.0801* (−1.81) WOTC eligibility period for disabled veterans −0.0416*** (−13.99) −0.0393*** (−5.23) −0.0204*** (−5.00) −0.0129* (−1.89) −0.0634*** (−5.53) −0.0397** (−1.97) −0.0182 (−1.18) −0.0105 (−0.49) Unemployed veterans X WOTC eligibility period for unemployed veterans 0.0203** (1.97) 0.0181* (1.75) 0.0208 (0.76) 0.0188 (0.68) Unemployed veterans dummy −0.167*** (−20.06) −0.163*** (−19.58) −0.128*** (−4.32) −0.121*** (−4.11) WOTC eligibility period for unemployed veterans −0.0308*** (−7.15) −0.0295*** (−3.55) −0.0609*** (−5.36) −0.0484*** (−2.65) State fixed effects No Yes No Yes No Yes No Yes Time fixed effects No Yes No Yes No Yes No Yes Demographic controls Yes Yes Yes Yes Yes Yes Yes Yes N100,584 100,591 100,584 100,584 9,505 9,514 9,505 9,505 Notes: The table above reports marginal effects from probit models. Demographic controls include age, a quadratic in age, dummies for education attainment, and gender and race dummies. Unemployed Veterans are defined as veterans who had searched for more than 4 weeks for work during the last year. Data for this table was restricted to veterans only. Data is from Annual Social and Economic (ASEC) supplement of the Current Population Survey spanning the years 2003–2013. t statistics in parenthesis. *p < 0.10, **p < 0.05, ***p < 0.01. Farooq and Kugler IZA Journal of Labor Policy (2015) 4:3 Page 17 of 28 Table 4 Difference-in-difference-in-difference (DDD) effects of work opportunity tax credits (WOTC) on disabled veterans and unemployed veterans All veterans eligible for WOTC based on employment history Only recent veterans eligible for WOTC based on employment history (1) (2) (3) (4) (5) (6) (7) (8) Disabled X veteran X WOTC eligibility period for disabled veterans 0.0472*** (4.06) 0.0485*** (4.20) 0.0471*** (4.06) 0.0480*** (4.16) −0.0193 (−0.45) −0.0180 (−0.43) −0.00910 (−0.22) −0.0084 (−0.20) Disabled X WOTC eligibility period for disabled veterans 0.00847 (0.86) 0.00766 (0.78) 0.00952 (0.97) 0.00870 (0.89) 0.0169*** (2.71) 0.0164*** (2.63) 0.0166*** (2.66) 0.0162*** (2.59) Veteran X WOTC eligibility period for disabled veterans −0.0182*** (−5.71) −0.0185*** (−5.81) −0.0139*** (−3.26) −0.0144*** (−3.36) −0.0402*** (−2.65) −0.0310** (−2.08) −0.00807 (−0.47) −0.0005 (−0.03) Disabled X veteran 0.154*** (33.90) 0.152*** (32.99) 0.154*** (34.11) 0.152*** (33.23) 0.174*** (13.00) 0.172*** (12.71) 0.171*** (12.40) 0.169*** (12.13) Disabled dummy −0.560*** (−77.57) −0.559*** (−76.79) −0.563*** (−78.36) −0.562*** (−77.58) −0.442*** (−76.49) −0.442*** (−76.36) −0.444*** (−77.05) −0.445*** (−76.91) Veteran dummy −0.0137*** (−5.85) −0.0145*** (−6.17) −0.0114*** (−4.77) −0.0121*** (−5.05) 0.00567 (0.45) −0.00208 (−0.16) 0.00948 (0.74) 0.00235 (0.18) WOTC eligibility period for disabled veterans −0.0168*** (−21.24) −0.0147*** (−7.12) −0.0100*** (−9.07) −0.0266*** (−14.66) −0.0174*** (−22.69) −0.0151*** (−7.39) −0.0107*** (−9.97) −0.0271*** (−15.14) Unemployed X veterans X WOTC eligibility period for unemployed veterans 0.0118 (0.95) 0.0105 (0.84) 0.0155 (0.48) 0.0130 (0.40) Unemployed X veterans −0.0295*** (−3.86) −0.0296*** (−3.88) −0.0199 (−0.72) −0.0175 (−0.64) Veterans X WOTC eligibility period for unemployed veterans −0.00888** (−2.04) −0.00824* (−1.89) −0.0447*** (−3.22) −0.0447*** (−3.22) Unemployed X WOTC eligibility period for unemployed veterans 0.00983*** (3.00) 0.0093*** (2.82) 0.0109*** (3.43) 0.0103*** (3.24) WOTC eligibility period for unemployed veterans −0.0102*** (−8.82) 0.0111*** (5.41) −0.0103*** (−9.21) 0.0110*** (5.41) Unemployed dummy −0.119*** (−48.30) −0.117*** (−47.60) −0.122*** (−51.18) −0.120*** (−50.46) Farooq and Kugler IZA Journal of Labor Policy (2015) 4:3 Page 18 of 28 Table 4 Difference-in-difference-in-difference (DDD) effects of work opportunity tax credits (WOTC) on disabled veterans and unemployed veterans (Continued) State fixed effects No Yes No Yes No Yes No Yes Time fixed effects No Yes No Yes No Yes No Yes Demographic controls Yes Yes Yes Yes Yes Yes Yes Yes N1,285,543 1,285,543 1285543 1,285,543 1,285,543 1285543 1,285,543 1,285,543 Notes: The table above reports marginal effects from a probit model. Demographic controls include age, a quadratic in age, dummies for educational attainment, and dummies for gender and race. Unemployed Veterans are defined as veterans who had searched for more than 4 weeks for work during the last year. Unemployed dummy refers to all individuals in the sample who had searched for more than 4 weeks during the last year. Data for this table leaves out individuals who could have qualified for WOTC, including SNAP and TANF recipients. Data is from Annual Social and Economic (ASEC) supplement of the Current Population Survey spanning the years 2003–2013. t statistics in parenthesis. *p < 0.10, **p < 0.05, ***p < 0.01. Farooq and Kugler IZA Journal of Labor Policy (2015) 4:3 Page 19 of 28 includes state and time fixed effects. The results are similar with or without state and time fixed effects and show that the likelihood of employment of disconnected youth increased by 0.015, or an increase of 4.7 percent, relative to the pre-treatment employment for this group. We find mixed results of targeted tax credits on employment. While we find a positive and non-trivial impact on the employment of disconnected youth, we do not find evidence of an impact on veterans. Previous evidence on tax credits is Table 5 Difference-in-difference (DD) effects of work opportunity tax credits (WOTC) on disconnected youth under the ARRA (1) (2) Disconnected youth X WOTC eligibility period 0.0152* (1.83) 0.0157* (1.88) Disconnected youth −0.289*** (−101.09) −0.285*** (−98.47) WOTC eligibility period (Under ARRA) −0.0688*** (−22.97) −0.0236*** (−4.00) State fixed effects No Yes Time fixed effects No Yes Demographic controls Yes Yes N260,841 260,841 Notes: The table above reports marginal effects from a probit model. Demographic controls include age, a quadratic in age, dummies for educational attainment, and gender and race dummies. Disconnected Youth are defined as 16–24 year olds who are not enrolled in school and who worked less than 26 weeks in the previous year. Data for this table is restricted to 16 to 24-year-olds and leaves out individuals who could have qualified for WOTC, including SNAP and TANF recipients. Data is from Annual Social and Economic (ASEC) supplement of the Current Population Survey spanning the years 2003–2013. t statistics in parenthesis. *p < 0.10,**p < 0.05, ***p < 0.01. 35 40 45 50 55 Percent 2002 2004 2006 2008 2010 Year Disconnected Youth Rest of the 16-24 yr olds Figure 11 Employment rates of disconnected youth & rest of 16-24 yr olds. Farooq and Kugler IZA Journal of Labor Policy (2015) 4:3 Page 20 of 28 also mixed. Katz (1998) finds that the Targeted Jobs Tax Credit, a major wage subsidy program for the economically disadvantaged introduced between 1979 and 1991 had modest but positive employment effects. Hamersma (2008) argues that the WOTC had minimal effects on the employment of targeted groups because of low take-up of the credits. Burtless’(1985) analysis of a randomizedtargetedwagesubsidy program in Dayton, Ohio suggests that vouchers may have even hurt the targeted groups by stigmatizing them. Our results are a little smaller than the ones reported by Katz (1998) for disadvantaged youth of a reduction in employment of 7.7 percent due to the discontinuation of the Targeted Jobs Tax Credit, though the TJTC applied to an older age group of 23 to 24-year-olds, who Katz (1998) argues are more attached to thelaborforceand,thus,morelikelytobenefitfromthecredits 13 . 4.2.4 Impact of the Hiring Incentives to Restore Employment (HIRE) Act An alternative to tax credits attached to individual groups are tax credits provided to employers hiring any workers, which would avoid the problem of stigmatization. In March 18, 2010 the Hiring Incentives to Restore Employment (HIRE) Act was passed, which instead gave a direct payroll tax exemption of 6.2 percent to employers hiring unemployed individuals who had been unemployed for at least 60 days or who worked less than 40 hours (part-time workers) in the last 60 days. The HIRE Act expired on December 31, 2010. While there has been no evaluation of this program, at the time, the Treasury Department indicated that there were 3.2 million jobs created which, in principle, qualified for these credits over the time period during which the credits were effective. Here we attempt to quantify the impact of the HIRE Act. Table 6 presents descriptive statistics for those who qualified for the HIRE Act. We define these as those who have been unemployed for at least 60 days in the past year. Those who qualify have lower employment and have searched more. They are also more likely to be male and African American, and they are more likely to be high school drop outs or to have only a high school degree. The first row in Table 7 reports the interaction between eligibility for the HIRE Act and an indicator for 2011, since the CPS reports employment in March for the previous past year. The specification controls for demographic characteristics, an indicator of whether the person is eligible, and state and time fixed effects. The results show that the HIRE increased employment by 1.6 percentage points or 2.6 percent relative to the employment for this group pre-HIRE Act. Grijalva and Neumark (2013) similarly find that state tax credits increased employment for the unemployed, but that the effects were not large. By contrast, evaluations of similar tax credits in other countries suggest that these credits have been effective in encouraging hiring. Kugler (2011) presents an extensive literature review with evidence on the effectiveness of payroll tax cuts for employers from a number of natural experiments around the world as well as from cross-country panel data studies. While there is no data on take-up of credits from the HIRE Act, there was a perception that take-up of the hiring credits was low, and this maybeonereasonwhytheimpactwasnotbigger. 4.3 Employment impacts of policies to get the long-term unemployed back to work As shown in Figure 7, those unemployed for more than six months are about half as likely to find a job as those who have been unemployed for less than six months. Yet, Farooq and Kugler IZA Journal of Labor Policy (2015) 4:3 Page 21 of 28 the share of long-term unemployed increased sharply during the Great Recession and remains twice as high as before the recession. Yet, under most states’unemployment systems, individuals are entitled to unemployment benefits for up to 26 weeks, and the replacement rate is close to 50%. Given that long-term unemployment rises during recessions, over the past several decades, emergency unemployment compensation has been extended 8 times to provide additional unemployment benefits to the long-term unemployed. During this last recession, emergency unemployment compensation (EUC) was first introduced in June 2008, then extended in February and November of 2009, and again in December 2010, February 2012, and January 2013. The initial program introduced two ‘tiers’of additional weeks of benefits. In November of 2009, the program was expanded to include two additional tiers. While the exact weeks and qualification for each tier has changed with each new extension, the current four tiers have been in place since 2009. The latest extension of EUC in January 2013 provided 14 additional weeks of benefits in the first tier to all states. The second tier provided 14 additional weeks for states with unemployment rates over 6 percent. Tier 3 provided 9 additional weeks if the unemployment rate is above 7 percent, and tier 4 provided 10 additional weeks if the unemployment rate is above 9 percent. The rationale in providing more weeks of benefits in those states with higher unemployment is that those are precisely the places where the long-term unemployed will be facing the biggest hurdles in getting jobs. In addition to emergency unemployment compensation, extended benefits (EB) trigger in for up to 20 weeks in states where the unemployment rate is above 6% and remains above what it was in the past three years. Table 6 Descriptive statistics for different unemployed groups Full sample HIRE act eligible HIRE act eligible pre-2011 HIRE act eligible post-2012 Long term unemployed Long term unemployed pre-2013 Long term unemployed 2013 & 2014 Employed 0.729 (0.444) 0.609 (0.488) 0.605 (0.489) 0.623 (0.485) 0.540 (0.498) 0.536 (0.499) 0.563 (0.496) Weeks looking for work 1.405 (6.162) 23.62 (11.77) 22.96 (11.57) 24.62 (12.00) 34.08 (7.984) 34.04 (7.961) 34.30 (8.100) Male 0.480 (0.500) 0.544 (0.498) 0.541 (0.498) 0.540 (0.498) 0.533 (0.499) 0.534 (0.499) 0.529 (0.499) Less than H.S. 0.117 (0.321) 0.173 (0.378) 0.181 (0.385) 0.156 (0.363) 0.184 (0.388) 0.188 (0.391) 0.164 (0.370) H.S. degree 0.299 (0.458) 0.349 (0.477) 0.356 (0.479) 0.334 (0.472) 0.360 (0.480) 0.362 (0.481) 0.346 (0.476) Some college 0.294 (0.456) 0.287 (0.452) 0.278 (0.448) 0.308 (0.462) 0.281 (0.450) 0.277 (0.447) 0.304 (0.460) Bachelor’s and higher 0.290 (0.454) 0.191 (0.393) 0.185 (0.388) 0.202 (0.402) 0.175 (0.380) 0.173 (0.378) 0.186 (0.389) White 0.795 (0.404) 0.761 (0.427) 0.763 (0.425) 0.750 (0.433) 0.729 (0.444) 0.729 (0.444) 0.727 (0.446) Black 0.115 (0.319) 0.145 (0.352) 0.142 (0.349) 0.155 (0.362) 0.168 (0.374) 0.167 (0.373) 0.174 (0.380) No. Observations 1,413,820 74,985 48,004 18,839 35,368 29,687 5,681 Notes: Mean coefficients; SD in parentheses. HIRE eligible are defined as workers who searched for at least 60 days in the past year. Long Term Unemployed (LTU) are defined as unemployed who were unemployed for six month or more in the last year. Data is from Annual Social and Economic (ASEC) supplement of the Current Population Survey spanning the years 2003–2013. Farooq and Kugler IZA Journal of Labor Policy (2015) 4:3 Page 22 of 28 These extensions have provided income support to close to 25 million workers and their families since the beginning of the recession and have helped many of these families from falling into poverty. However, one concern with unemployment benefits extensions is that they may generate a moral hazard and cause people to search less and to lower their acceptance of jobs. Yet, there are a number of reasons why extending unemployment benefits may be beneficial on economic grounds. First, unemployment benefits are an automatic stabilizer and avoid big consumption drops by households facing unemployment. Gruber (1997) finds that consumption drops by 22% for those without unemployment benefits while only dropping by 7% among those receiving unemployment benefits. Also, Vroman (2010) finds a multiplier of 2 for unemployment insurance. This means that the economies of entire regions and states where the longterm unemployed received benefits grew by twice as much as the benefits received in those states. 14 Second, Krueger and Mueller (in progress) find that unemployment benefits help the unemployed stay attached to the labor force rather than going into disability insurance. Both Rothstein (2011) and Farber and Valletta (2013) find that the unemployment insurance extensions reduced exits from unemployment but that this was largely due to reductions in exits from the labor force rather than a decrease in exits to employment. This is particularly important given the decline in labor force participation which started in 2000 and which has continued during the recession and also given the rise in disability insurance enrollments since the 1980s. These studies suggest that while the UI extensions may prolong unemployment, this is not because the unemployed are turning down job offers but because they are staying attached to the labor force rather than going into disability or stopping their job searches. 15 Also, Table 7 Difference-in-difference (DD) effects of unemployment assistance programs on the employment of the long-term unemployed (1) (2) HIRE eligible X 2011 0.0184*** (3.57) 0.0159*** (3.07) HIRE eligible −0.0894*** (−33.62) −0.0875*** (−32.98) 2011 Year dummy −0.0174*** (−11.49) LTU X LTU assistance period 0.0327*** (5.39) 0.0304*** (4.97) Long Term Unemployed (LTU) −0.106*** (−26.47) −0.102*** (−25.53) LTU assistance period (2013,2014) −0.00895*** (−7.66) 0.0106*** (5.02) State fixed effects No Yes Time fixed effects No Yes Demographic controls Yes Yes N1,199,947 1,199,947 Notes: The table above reports marginal effects from probit models. Demographic controls include age, a quadratic of age, dummies for educational attainment, and gender and race dummies. HIRE eligible are defined as workers who searched for at least 60 days in the past year. Long Term Unemployed (LTU) are defined as unemployed who were unemployed for six months or more in the last year. Data for this table leaves out any individuals eligible for WOTC, including SNAP and TANF recipients. Data is from Annual Social and Economic (ASEC) supplement of the Current Population Survey spanning the years 2003–2013. t statistics in parenthesis. *p < 0.10,**p < 0.05, ***p < 0.01. Farooq and Kugler IZA Journal of Labor Policy (2015) 4:3 Page 23 of 28 Chetty (2008) shows evidence that unemployment benefits may also provide liquidity to individuals during periods of unemployment, which may improve the quality of the jobs they get. Finally, Farooq and Kugler (2013) find that public insurance programs increase labor mobility by increasing occupational and industry mobility but also mobility into self-employment and wage employment. In addition to extending the period of time for which individuals can receive unemployment benefits, the Middle Class Tax Relief and Job Creation Act of 2012 introduced important reforms to help the long-term unemployed get back to work. First, the 2012 act introduced reemployment assistance and eligibility assessments (REAs) for those getting additional unemployment insurance. The REAs required in-person checkins in UI offices, skill assessments, and job search counseling during those visits for the long-term unemployed. This proposal was based on a number of studies of randomized trials in Nevada, Minnesota, Illinois and Florida showing substantial reemployment effects. Michaelides et al. (2012) showed that UI recipients who were randomly assigned to REA’s were 15 percent less likely to exhaust benefits, reduced the period for which they received benefits by 3 weeks, and increased their earnings by 18% in the 6 quarters following participation in REAs. Second, the reform introduced a selfemployment assistance (SEA) program which allows the long-term unemployed to continue using UI benefits while setting up their own business. Benus (2009) evaluated a similar program, the Growing America through Entrepreneurship (GATE) program introduced in Pennsylvania, Minnesota, and Maine, which randomly assigned half of the people to training and business counseling and provided assistance in applying for business financing, and found that those assigned to the program were 6 percent more likely to own a business, were likely to start their business sooner, and their businesses had greater longevity. They also found that the program was most effective among those receiving UI. Benus et al. (1994) and Benus et al. (1995) also show positive employment impacts from the Self Employment Enterprise Demonstration (SEED) and the Massachusetts Enterprise Program, which allowed the unemployed to continue claiming benefits while receiving entrepreneurial training and setting up a new business. Table 6 shows descriptive statistics for the long-term unemployed and other groups. The long-term unemployed are less likely to be employed and to have searched longer for work. They are also more likely to be male and African American and to have dropped out of high school or to have only a high school degree. Figure 12 shows the pre-treatment employment for the long-term unemployed and those unemployed for less than six months. While the short-term unemployed have higher employment throughout, the trends are similar, except that the long-term unemployed experienced much sharper drops in employment both in the 2001 and 2009 recessions. Table 7 shows the results from a regression that interacts the long-term unemployment dummy for those unemployed for more than 6 months with the 2012 and 2013 dummies, after the WOTC and HIRE Act were no longer in effect, but when the REAs where already in effect. The results are reported in row 4 of Columns (1) and (2), and they suggest that the introduction of REAs increased employment by 3 percentage points or 6 percent for the long-term unemployed. While the effect is not directly comparable with the studies cited above, the effect seems somewhat weaker compared to the randomized trial states which spent similar amounts on eligibility assessments and reemployment ($85 vs. $200 Farooq and Kugler IZA Journal of Labor Policy (2015) 4:3 Page 24 of 28