North Carolina’s Employment Record: What role did Unemployment Insurance Reform play?

Patrick Conway Department of Economics University of North Carolina at Chapel Hill 22 December 2015 North Carolina’s labor force has experienced a turbulent past decade. The turbulence began with the financial crisis of 2007/2008: unemployment and the unemployment rate surged in that year and the years to follow to a level last seen during the Great Depression. In 2011, with the election of a Republican majority in both houses of the state legislature, further changes were put in motion. In February 2013, the legislature passed a sweeping reform of the unemployment insurance program. That reform reduced the maximum monthly insurance payment and reduced the maximum number of weeks for which the unemployed can appeal for insurance payment. The reform became law on 1 July 2013. It was so sweeping that the US Department of Labor was compelled to cancel extended Federal unemployment insurance payments for residents receiving them on that date; it could not continue disbursements in states that unilaterally rewrote the insurance contract as radically as North Carolina did. Conservative policy-makers pointed to the rapid fall in the unemployment rate in North Carolina in the months after the reform as evidence that reducing the payments for unemployment insurance led more individuals to seek work. Progressive policy-makers attributed the fall in the unemployment rate to another factor – a decision by a larger share of the working-age population to cease participation in the labor force. This distinction has important real-world consequences. If removing unemployment insurance leads to greater job search and job creation, then the reform has a positive effect on labor-force outcomes. By contrast, if removing unemployment insurance simply leads to discouraged workers no longer participating in the job market, there is no positive effect – the reduction in the unemployment rate simply masks the loss of work for the state’s residents. This is an empirical question. If the conservative position is correct, then we will observe two phenomena with unemployment insurance reform: labor-force participation will remain constant and employment will rise. If the progressive view is correct, then employment will be unaffected and labor-force participation will fall (due to the “discouraged worker” effect). In this note I investigate the empirical record. I use two databases maintained by the Federal government. The first is the Local Area Unemployment Survey (LAUS): it reports monthly totals of those employed, unemployed and in the labor force. It also provides indirectly a measure of the state working-age population. The second is the Current Population Survey (CPS); as part of its survey it provides household-level information about the decisions taken by

North Carolina’s Employment Record - 2

those of working age, and thus opens a window on the record of job creation and participation choice at the center of this debate. 1 A. Labor market evolution in North Carolina The empirical record for North Carolina demonstrates the features of its labor-market evolution since January 2000. For these illustrations I use the data of the Local Area Unemployment Statistics (LAUS) database of the Bureau of Labor Statistics (BLS), in seasonally adjusted form. I consider all residents 18 and over as the working-age population. The US has faced two labor-market downturns since 2000. The first, associated with the “dot com” bubble in the stock market, occurred in 2001-2002. The second, associated with the financial crisis, began in 2007. The dot-com downturn was relatively minor: the number employed began growing again in January 2002. The financial-crisis downturn, by contrast, was severe. It began in 2007 but did not reach its low point until December 2009. Since the beginning of 2010, the number of employed US residents aged 18 or older has been growing. The black line in Figure 1 illustrates these developments for the US excluding North Carolina (abbreviated as US in what follows).

1

In using the CPS database to create a panel of respondent behavioral choices in this way I follow the lead of Kroft et al. (2014). Thanks to those authors, and especially Matt Notowidigdo, for sharing their expertise in this technique.

North Carolina’s Employment Record - 3

•

North Carolina has gone through a very similar pattern of employment growth, as is evidenced by the red line in Figure 1. North Carolina’s reduction in employment was more pronounced than that for the US during the dot-com downturn, but by 2006 the state had reversed that relative drop. From the beginning of the financial-crisis downturn, North Carolina has had a less pronounced reduction and that has translated into faster employment growth over the 2000-2014 period.

Figure 2 illustrates this relationship in a novel way. The dashed blue line in the graph illustrates the ratio of NC employment to US employment. If the two grew at the same rate over time, then the ratio would remain 100. If it rises above 100, it indicates that NC employment grew more rapidly during that period.

As is evident from Figure 2 as well, the beginning of the dot-com downturn in 2001 brought larger downturns in employment in NC than in the US as a whole through 2005. At the beginning of 2006 the ratio returned to 100, indicating that NC had returned to employment growth parity with the US. In 2009 there is a further upturn in the ratio. This shows that employment reduction was less pronounced in North Carolina than in the US. By July 2010 the employment ratio reached 102, and then stayed roughly constant throughout the following four years before rising in 2015.

North Carolina’s Employment Record - 4

•

A cumulative four percent higher employment growth for NC than in the US as a whole since January 2000 (or February 2006) is a poor employment creation performance, because the population of North Carolina has grown at a more rapid rate than that of the US as a whole. With more people of working age, there should be more jobs – however, job growth has not kept up with population growth. This is also illustrated in Figure 2. The solid purple line illustrates the relative populations of NC and US. The populations grew at roughly the same rate from 2000 to 2004, but after that time the North Carolina population grew more rapidly. By December 2014, the NC population had grown cumulatively eight percent more than the US. (The sharp rise in the ratio in 2010 and the sharp fall in the ratio in 2012 are due to anomalies in the North Carolina population statistics.) While employment in North Carolina grew more quickly than in the US as a whole, that growth was not sufficient to offset the even-more-rapid growth in population.

•

This is not the story told by the unemployment rates in North Carolina and the US as a whole. In Figure 3 I illustrate the evolution of the unemployment rate since 2000. The NC unemployment rate (red line) rose above the US rate (black line) during both downturns: slightly, during the dot-com downturn, and more significantly during the financial-crisis downturn. From November 2008 to August 2013 the NC rate was above that of the US; since August 2013, however, the two have been quite similar.

North Carolina’s Employment Record - 5

Thus, while the employment record seems more anemic in North Carolina than in the US, the unemployment rate is nearly identical with that of the US as a whole. •

One explanation for this contradiction is found in the labor force participation rate. This is the percent of people of working-age who are either employed or actively seeking work. This percent has been declining steadily over time in the US as a whole, in part because of demographic shifts toward a greater percentage of residents entering retirement age. Figure 4 provides the labor force participation rate for the US as a whole and for North Carolina separately. While the two ratios differed from month to month, they followed a similar trajectory through December 2006: each ratio fell by about one percentage point from January 2000 through that time. Beginning with January 2007, the two series diverged: North Carolina’s ratio fell at a more rapid rate. Between January 2007 and December 2014, the participation rate in North Carolina fell by 6.6 percentage points, while the US participation rate fell by 3.5 percentage points. During 2015 the North Carolina rate has risen by about 1 percentage point while the US rate has stayed steady.

A fall in the labor force participation rate is an indication that individuals of working age have chosen not to seek out work. One explanation is that of the “discouraged worker”: someone who sought work for some time, had no success, and decided to stop looking. This helps to explain the evolution of the unemployment rate. The unemployment rate is calculated as the number of those actively seeking work divided by the number of individuals in the labor force. Since the discouraged worker is excluded from the

North Carolina’s Employment Record - 6

numerator, the unemployment rate is reduced by decreases in the labor force participation rate due to discouraged workers. •

One way to control for this “discouraged worker” complication is to examine the employment-population ratio for the US and for North Carolina. The employmentpopulation rate is the ratio of all those employed divided by the population over the age of 18 in the two geographic areas. Figure 5 illustrates this ratio for the US (black line) and North Carolina (red line). As is evident there, the experience in North Carolina essentially mirrored that of the US as a whole until December 2006. After that time, the share of residents working fell for both groups, but fell more rapidly in North Carolina. The ratios became roughly constant after December 2009. With the beginning of 2014 there is evidence of a recovery in the US as a whole; there is a slight rise in the ratio. In North Carolina, the ratio stayed roughly constant from 2010 to the end of 2014, even though the unemployment rate has fallen from 11.3 percent to 5.4 percent during the same period. The ratio then began to rise sharply in 2015, while the unemployment rate has been rising slightly.

We can make the same point in a slightly different way by comparing the outcomes of the unemployment rate and the employment/population ratio in North Carolina as illustrated in Figures 3 and 5. The unemployment rate in North Carolina was 6 percent at three points this century: in January 2002, in October 2003, and in May 2008. In these months, the employment/population ratio was very different: 62.5 percent, 60 percent, and 56.5 percent respectively. In recent years, the employment/population ratio consistent with 6 percent unemployment is much lower – and this indicates the massive shift of working-age population of the state towards non-participation in the labor force.

North Carolina’s Employment Record - 7

B. The Unemployment Insurance Reform in North Carolina. Once the financial-crisis downturn began in 2007, North Carolina rapidly used up the balance in its Unemployment Insurance Trust Fund. It borrowed from the Federal Unemployment Insurance Trust Fund to continue making unemployment insurance payments. The Republican majorities in both legislative houses passed a sweeping reform to the unemployment insurance law in North Carolina in early 2013, and NC governor Pat McCrory signed the bill into law on 19 February 2013. 2 The reform became law on 1 July 2013. There were two salient changes to the unemployment insurance policy. First, maximum weekly benefits were cut from $530 to $350. Second, the number of weeks for which recipients are eligible for benefits fell from 26 (or 52, with Federal extensions) to 12. Its reforms were so sweeping that the US Department of Labor ruled that the unemployment insurance system in North Carolina had been qualitatively altered; it cut off the access of North Carolina residents to extended Federal benefits. While nine states reformed their unemployment-insurance programs during the financial-crisis downturn, North Carolina’s reform was the most extreme: it was the only one of the nine to lose extended Federal benefits for its residents (GAO, 2015). Those wishing to investigate the effect of the policy then use a comparison of North Carolina’s employment performance to that of other states to determine the impact of the policy. Hagedorn et al. (2014), for example, used the 1 July 2013 reform as a “treatment” event, and looked for differences from behavior in “control” states such as South Carolina and Georgia. (These were also states that had reformed their unemployment-insurance policy, although not in as extreme a manner as in North Carolina.) Returning to Figures 1 through 5, there is little evidence that the reform on 1 July 2013 (or even, if we believe that workers anticipated the effect of the reform in earlier months, looking back to 1 January 2013) had any differential effect on employment in North Carolina. Figures 1 and 5 describe the evolution of employment (or employment/population) in North Carolina and in the US – there is no evidence of a differential effect in North Carolina. Figure 2 looks at North Carolina employment as a percent of US employment – that too varies little during the reform window. Figure 3 illustrates an unemployment rate in North Carolina falling more rapidly than in the US as a whole, but that more rapid decline is a continuation of a trend that dates back to early 2010. In Figure 4, the labor participation rate does exhibit a decline more rapid than that observed in the rest of the US. A cursory view of these data series thus suggests that the progressive argument has the greater basis in the facts: employment is unchanged but the labor force participation is substantially different. However, it is difficult to draw conclusions based on aggregate data such as these. To evaluate more precisely, I turn to the evidence found in the Consumer Population Survey.

2

Jen Wilson: “NC Gov. Pat McCrory signs Unemployment Insurance Reform Bill”, Charlotte Business Journal, 19 February 2013.

North Carolina’s Employment Record - 8

C. Labor-force choices as highlighted in the Consumer Population Survey. The question at hand is really one of worker choice and opportunity: do workers choose to participate in the labor force, and if so do they have the opportunity to take a job? To address this, I have collected employment information from household respondents to the Current Population Survey (CPS) by month from January 2000 to August 2015. The CPS queries household respondents about the labor-market behavior of each working-age adult in the household for four consecutive months; after a break of eight months, the CPS queries the household for an additional four months. A panel of data is created by matching observations across successive months by household identifier, line number (in the survey questionnaire), age, gender and race. This data panel is used to estimate the historical transition probabilities into and out of employment, unemployment and non-participation status for each worker. 3 For comparability to the aggregate LAUS data, the sample is limited to those of working age (i.e., aged 18 and above). Table 1 reports the number of panel observations for the rest of the US (top) and for North Carolina (bottom) in the CPS-based data. The number reported in each case is the average number of monthly observations for that year. There is a drawback in using this database to focus upon North Carolina, especially the unemployed of North Carolina, since the number of observations each month is relatively small. This leads to relatively greater volatility in North Carolina’s monthly average transition probabilities. In the figures that follow, a locally weighted least-squares regression of a univariate polynomial process is used to smooth the series over time. 4 I also aggregate the observations to quarterly cells to increase the number of observations in each cell. The individual observations remain month-to-month transitions.

3

When considering unemployment status, I treat an observation for a given working-age adult of unemployment followed by employment followed by unemployment as a string of three months of unemployment. 4 I use the Stata command lpoly to create these average transition probabilities, with epanechnikov kernel, degree zero, and bandwidth chosen optimally by the program. The standard errors are derived from the conditional variance calculated for the locally weighted regression for each quarter in the sample.

North Carolina’s Employment Record - 9

Table 1: Average Number of Observations per Month in the CPS-based panel

Year 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

Rest of US Employed 10806 11539 12531 12384 12158 12061 12038 12049 11923 11734 11472 11386 11355 11206 11148

Unemployed 408 516 670 698 621 568 523 533 643 1069 1091 994 879 788 676

Non-participants 5273 5574 6197 6476 6345 6254 6200 6217 6243 6480 6392 6522 6553 6611 6682

Unemployed 10 14 17 16 14 12 11 10 16 26 23 22 20 20 14

Non-participants 138 128 134 141 141 136 128 127 137 142 138 139 142 150 151

NC Year 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

Employed 279 265 254 261 251 239 232 229 226 224 219 206 218 216 223

North Carolina’s Employment Record - 10

Figure 7 illustrates the average probability of transitioning from being employed to being unemployed for North Carolina (NC) and for the rest of the US (ROUS) (that is, the US excluding NC) in a given month. Both the ROUS and NC began with an average transition percentage of just above 1 percent in early 2000: this indicates that an average worker faced a 1 percent chance of becoming unemployed in the next month. •

•

•

For the ROUS (the orange line) that transition probability rose to just under 1.4 percent in the dot-com downturn, and fell back to 1.2 percent in late-2006. The financial-crisis downturn caused the transition probability to rise above 1.6 percent in early 2009, and then to fall to just over 1.2 percent again by early 2015. In North Carolina (the solid blue line), by contrast, the transition probability from employment to unemployment was higher on average throughout the dot-com downturn. It fell to below the ROUS average in 2005-2006. The probability of transitioning to unemployment then rose above the ROUS average through mid-2013, and at that time fell below that of the ROUS. The dotted lines in Figure 7 define a 90 percent confidence interval for the NC transition probability. The ROUS transition probability lies within that confidence interval for

North Carolina’s Employment Record - 11

nearly every period: while the two probabilities diverge, the difference is not a significant one. 5

Figure 8 reports the smoothed transition probability from employed to labor-force nonparticipant. This is most likely an indicator of the probability of retirement, although transition to disability status or returning to school is also included. •

•

•

5

For the ROUS, this probability begins just above 2.8 percent in 2000, rises slightly during 2006 and 2007, and then returns to its earlier level in 2010. From 2010 to 2015 there is an upward trend, with the probability rising above 3 percent by 2015. In North Carolina the upward trend is larger and observed throughout the 2000-2015 period. It began at 2.5 percent in 2000 and rose rapidly to over 3 percent at the onset of the financial crisis in 2007. It then stabilized for the period 2007-2010, and since 2010 it has been rising again, to above 3.2 percent by 2015. The extended spell of higher-thanROUS transition probabilities from employment to non-participation since 2007 reflect the cumulatively lower labor-force participation rate in North Carolina. The confidence intervals for the ROUS transition probabilities indicate that in the beginning of the period the North Carolina probability fell significantly below that of the ROUS. Beginning in the financial-crisis downturn, North Carolina’s probability rose above that of the ROUS, but not significantly so.

There is a confidence interval for the ROUS transition probability as well, but it is a much tighter band around the estimated transition probability. It is excluded to clarify the figures.

North Carolina’s Employment Record - 12

Figures 7 and 8 report on two transition probabilities from employment. The third probability is that the individual remains employed in the following period. We can infer that probability from the other two: for example, in early 2000, the probability that an employed worker stayed employed in the ROUS in the following month was 96 percent (i.e., 1 – 0.01 – 0.03). Those employed in North Carolina, then, have transition probabilities from employment similar to those in the rest of the US. The transition to unemployment has more volatility than in the rest of US, while the transition to non-participation in the labor market is monotonically increasing over time at a greater rate than in the rest of the US. In both cases, however, the differences are not statistically significant.

What happens to those who are unemployed? Figure 9 illustrates that transition both for the ROUS and for NC. •

•

In the ROUS, the percent of workers transitioning from unemployment to employment in the next month was above 25 in early 2000. This percentage fell to 20 during the dotcom recession, and rose again to 24 at the end of 2007. During the Great Recession, this percentage fell to 14 in 2010 and rising to 18 in 2015. For North Carolina, this transition probability to employment has been very close to that of the ROUS. The re-employment probability in North Carolina was less than that in the ROUS both during the dot-com recession and in the period 2010-2015, but the difference between the two was rarely significant.

North Carolina’s Employment Record - 13

Workers also transition from unemployment to labor-force non-participation. Figure 10 illustrates the historical percent of those unemployed who transitioned to non-participation in the following quarter.

•

•

•

For the ROUS, the transition from unemployment to non-participation occurred about 25 percent of the time in early 2000. This percent fell during the dot-com recession, and then rose to roughly the same percent in mid-2006. The percent turning from unemployment to non-participation fell to about 20 percent in early 2009, and then has risen again to just over 25 percent in early 2015. In NC, the transition percent was on average well below the ROUS through 2000-2003: it remained below 20 percent most of that time. In 2004-2005 the NC percent rose above that of ROUS, and then nearly coincides with that of ROUS for 2006-2009. In 20102013 the probability of transitioning to non-participation fell below that of ROUS, and beginning in late 2012 the NC transition probability rose well above that of the ROUS, reaching 33 percent in 2015. The confidence intervals for the NC transition probability indicate that the North Carolina probability was not significantly different from that of ROUS for the initial 12 years of the sample. Beginning in 2013, those unemployed in NC were significantly more likely to transition to non-participation in the labor force. This greater non-participation in recent years has contributed to the significantly smaller labor-force participation rate in North Carolina currently.

The final transition probability is that the unemployed individual remains unemployed in the next month. That percent is implied by those observed in Figures 9 and 10. For the ROUS in

North Carolina’s Employment Record - 14

early 2000 that was 50 percent (i.e., 1 – 0.25 – 0.25); for NC that was 54 percent (i.e, 1 – 0.28 – 0.18). Comparing the results for transition from employment to non-participation and from unemployment to non-participation provides an estimate of the share of the transition due to “discouraged workers”. About three percent of those employed transition to non-participation in a given month. However, about 20 percent of those unemployed transition to non-participation in a given month. While there are some reasons why we might expect a higher percentage from the unemployed (e.g., the transition from unemployment to disability), that percentage differential is in large part due to those who become discouraged. The evidence on the behavior of the unemployed provides insight into the impact of labor-force policy reform in North Carolina. During the most recent years (i.e., since 2011), there is no evidence that reform led to greater transitions of the unemployed to employment. In fact, that transition probability for NC was less than the comparable probability for the rest of the US. We do see, however, that in NC there is a steadily increasing percentage of the unemployed transitioning to non-participation in the labor market since 2011, and that percentage is significantly higher than for the rest of the US since the beginning of 2013. The final piece of the transition puzzle is found in the transitions from labor-force nonparticipation. Figure 11 illustrates the percent of those not participating in the labor force who in the next period are employed (because they’ve left school and taken a job, for example). As is evident, this is a low-probability event – only between 2 and 3 percent of the non-participants begin employment in the next quarter. •

•

•

For the ROUS, there is a slight downward trend throughout the period considered. The downturn was slight from 2000 to 2007, became pronounced in 2007-2010, and then rose slightly in the period 2010-2015. From a high of 3.3 percent in early 2000, the percentage fell to 2.8 percent in early 2015. In NC, the probability is less than in ROUS in almost all periods. While it is roughly constant during 2000-2007, there is a more pronounced downturn during 2007-2010. The percentage stays low through 2013, and then begins to rise. The confidence intervals for the NC transition probability illustrate that the differences in transition probabilities from non-participation to employment between North Carolina and ROUS are significant during nearly all the periods considered.

North Carolina’s Employment Record - 15

Figure 12 indicates the percent of non-participants in each period who report themselves unemployed in the following period. These are individuals who have not been looking for work, but in the next period begin (unsuccessfully) to search for jobs. These probabilities are also very small: in the range of 1.5 to 4 percent per quarter. •

•

•

For the ROUS, this rate began at just under 2 percent in early 2000. It rose slightly to 2.3 percent during the dot-com recession, but returned to just under 2 percent in early 2007. It rose rapidly to 3.3 percent with the onset of the financial-crisis recession, and from 2010 to the present has been falling. The value in early 2015 is 2.4 percent. In NC, this transition probability followed the same general path. The probability that someone out of the labor force will enter and be unemployed began at 1.5 percent in early 2000, but rose during the dot-com recession to 2.6 percent in late 2002. It then fell to 2 percent in 2004-2006, and rose slightly during the financial-crisis recession to 2.8 percent in 2010. The percent then fell from 2010 on, returning to 2 percent in early 2015. Consideration of the confidence intervals around the NC transition probabilities suggests that the NC record in this regard was insignificantly different from that of the ROUS until 2008. During the financial-crisis downturn and thereafter, the NC probability remained significantly below that of ROUS: those who choose not to participate in the labor force in NC are significantly less likely to come back and search (unsuccessfully) for jobs.

North Carolina’s Employment Record - 16

The remaining transition probability is for a non-participant in the labor force to remain a nonparticipant in the following month. Not surprisingly (since this dataset includes those of retirement age), the likelihood of remaining a non-participant is roughly 95 percent both in NC and in the ROUS in early 2000. Here also we find significant differences between those in NC and those in ROUS who are not participating in the labor force. Since the beginning of the financial-crisis downturn in 2008, non-participants in NC have been significantly less likely than those in the rest of the US to either be employed in the next month or to be unemployed in the next month. In other words, non-participation has been a more likely final destination for those who drop out of the labor market. D. Are these results driven by the growing retired and school-age population in North Carolina? The attractiveness of North Carolina as a retirement destination is given as one explanation for the falling labor-force participation rate in North Carolina relative to that in the rest of the US; the increased age of college students is given as another. 6 To ensure that the results of the preceding section are not biased by responses by school attendees and retired individuals, I redo the analysis using the sample of working-age individuals between 25 and 60 years of age. This should include only those who are prime candidates for returning to employment if unemployed and for returning to labor-force participation if currently out of the labor force.

6

These are themes, for example, in the NC Commerce Department LEAD FEED by Andrew Berger-Gross entitled “Back to School”, 1 December 2014.

North Carolina’s Employment Record - 17

I won’t present the analytic details here, but the results of non-parametric regression are available on demand. The results are very similar to those of the larger working-age population. In particular: •

•

Those unemployed in North Carolina have a significantly greater likelihood to be out of the labor force in the next month than those unemployed in ROUS. The likelihood of remaining unemployed next month or of being employed next month are insignificantly different when comparing NC and ROUS. The behavior of those out of the labor force in North Carolina was insignificantly different from those out of the labor force in the ROUS during the period since the Financial Crisis began. They were no more likely to be employed or unemployed in the following month than their counterparts in ROUS.

In summary, the evidence for this sample of working-age individuals also suggests that the significant impact of North Carolina labor-force policy was to increase the number of workers who left the labor force, not to encourage those without jobs to become employed. E. Conclusions and suggestions for further research. There are two ways to interpret the unemployment insurance reform in North Carolina and the subsequent decline in the unemployment rate. The conservative position is that restricting access to unemployment insurance compelled the unemployed to search harder for jobs, leading to an upturn in employment and no decline in labor force participation. The progressive position is that the unemployment insurance reform led to a fall in labor force participation and no differential increase in employment opportunities. Using the other 49 states of the US as the control group, I find no evidence of a positive differential in employment opportunities in North Carolina in the period around the unemployment insurance reform, especially once controlling for growth of the working-age population. By contrast, there is evidence of significant differential in labor force participation rates between North Carolina and the rest of the US, with North Carolina’s labor force participation shrinking at a more rapid rate than in the rest of the US. This is consistent with the progressive critique of the unemployment insurance reform. Both the LAUS database and the transitions derived from the CPS database tell a similar story. The LAUS trends in employment growth are quite similar when North Carolina and the US as a whole are compared. The CPS-derived transition to employment from unemployment (or for that matter, from unemployment to employment) is similar in North Carolina to in the rest of the US. There is no significant evidence that the unemployment-insurance reform in North Carolina had an effect on the creation of jobs or the decision of the unemployed individual to accept a job. North Carolina does differ significantly from the rest of the US, however, in its residents’ decisions about labor force participation. The LAUS data illustrate that the fall in labor-force participation rate has been more rapid in North Carolina than in the US as a whole. The CPSderived transitions from unemployment to labor force non-participation became significantly larger at the time of the unemployment-insurance reform. We also note that the transition from non-participation to either employed or unemployed status is significantly lower in North

North Carolina’s Employment Record - 18

Carolina than in the rest of the US: once an individual leaves the labor force, she is significantly less likely to return. Accepting the progressive interpretation of labor-market evolution in North Carolina leads to strong policy implications. The growth in employment does not differ from that of the rest of the US, suggesting that the reform did not create new jobs (or fill new jobs that had been unfilled prior to the reform). The fall in the unemployment rate is due to the decision of discouraged workers to exit the labor force and should be seen as a challenge for policy, not a success. As the economy improves, the discouraged workers will return – as we’ve seen in recent months with the unemployment rate rising despite solid employment growth. This is the real Carolina Comeback, as working-age individuals return to the labor force, and policy should work to encourage that. There are a number of follow-on questions that arise from these results. While the unemployment-insurance reform is the event that is posited to have triggered change in laborforce behavior, the figures suggest that the difference in labor-force participation in North Carolina dates from earlier – perhaps the beginning of 2011. It will be useful to find the temporal turning point indicated by the data to see if it corresponds to the Republican takeover of both legislative houses in North Carolina. The behavior described here is the behavior of the individual worker, but that worker’s decisions will be influenced by the range of policy initiatives introduced by the North Carolina legislature. It is important to recognize as well that the conclusions in sections C and D are based on small samples of North Carolina residents. The Current Population Survey surveys a large number of US residents, but the number of North Carolina residents reporting themselves to be unemployed (for example) is relatively small in each month. The Labor and Economic Analysis Division of the North Carolina Commerce Department has access to more complete databases of the unemployed in North Carolina, and it will be important that they confirm these conclusions using the universe of those unemployed or out of the labor force. Finally, there are questions of measurement to address. As Hagedorn et al. (2014) point out, different data sources on the same phenomenon can lead to different conclusions. They found that their calculations of employment and unemployment for North Carolina based on the CPS and the LAUS database yielded different trends over time. The database that provided the most support for the conservative view, though, was the Current Employment Statistics survey. It will be important to investigate that data source as well to see if the divergences those authors identified remain when examined as above.

North Carolina’s Employment Record - 19

Bibliography: Berger-Gross, A.: “Back to School”, NC Commerce Department LEAD FEED, 1 December 2014. Accessed 22 December 2015 at http://www.nccommerce.com/lead/researchpublications/the-lead-feed/artmid/11056/articleid/25/back-to-school-the-decline-in-ncs-youthlabor-force-participation-rate. Farber, H., J. Rothstein and R. Valletta: “The Effect of Extended Unemployment Insurance Benefits: Evidence from the 2012-2013 Phase-out”, American Economic Review Papers and Proceedings, 105/5, 2015, pp. 171-176. General Accountability Office (GAO): “UNEMPLOYMENT INSURANCE: States' Reductions in Maximum Benefit Durations Have Implications for Federal Costs”, Report 15-281, April 2015. Hagedorn, M., F. Karahan, I. Manovskii and K. Mitman: “Case Study of Unemployment Insurance Reform in North Carolina”, mimeo, 25 March 2014. Kroft, K., F. Lange, M. Notowidigdo and L. Katz: “Long-term Unemployment and the Great Recession: The Role of Composition, Duration Dependence and Non-Participation”, NBER Working Paper 20273, June 2014. Wilson, J.: “NC Gov. McCrory signs Unemployment Insurance Reform Bill”, Charlotte Business Journal, 19 February 2013.