Justin Wolfers, a University of Michigan economist and senior fellow at the Brookings Institution, wrote a piece for The New York Times a couple of weeks ago that was partly a response to a previous piece I wrote for The Wall Street Journal about North Carolina’s recent drop in unemployment. I don’t agree with all of his conclusions, as you’ll see in a moment, but Wolfers — a careful scholar and co-editor of the influential journal Brookings Papers on Economic Policy — did advance the national debate about unemployment insurance and the labor market in two important respects.
First, Wolfers makes the point that using the Current Population Survey to track month-to-month changes in the labor market is a practice fraught with peril. Because this “household survey” draws from relatively small monthly samples in each state, there’s a lot of statistical turbulence. Trying to tease out firm conclusions for subgroups of workers, such as those who’ve dropped out of the labor force, is just not possible when working only with the monthly CPS data. Moreover, looking at the most commonly tracked measure of unemployment, the U-3 rate, doesn’t tell you whether unemployment drops are due to employment gains or to workers dropping out of the labor force. Instead, Wolfers points to job counts from the monthly Current Employment Survey, often called the “establishment survey” of employers, and from the Quarterly Census of Employment and Wages, which isn’t a random sample at all but an actual count of most payrolls in the United States (all jobs covered by unemployment insurance, actually).
Second, Wolfers observes that both sides of the debate about UI extended benefits — those (like me) who thought they should go away as well as those who thought they should be re-authorized — have made testable claims about the economic effects of benefit extensions. Analysts skeptical of the wisdom of extended benefits, usually but not always free-market economists, argue that after a certain point, the availability of UI benefits becomes an impediment to reemployment. The benefits act as a disincentive for unemployed workers to take the steps necessary to improve their lot in the long run, such as accepting an available job even if it isn’t ideal, moving to another city or state where the job prospects in their field are more promising, returning to school to retrain for a new career, or starting their own businesses. These are not easy decisions. Many workers will, understandably, delay such a decision as long as possible. Indeed, one of the explicit purposes of UI is to give laid-off workers some time to find the right fit, rather than having immediately to make a life-altering decision to pay the bills. At best, however, that’s an argument for a UI benefit denominated in weeks or months, not one denominated in years.
Supporters of extended benefits, usually but not always Keynesians or left-leaning analysts, argue that UI isn’t really generous enough to have the disincentive effect the skeptics claim, and in fact that UI payments actually boost economic growth and job creation by augmenting aggregate demand. Unemployed UI recipients, they point out, are likely to spend virtually all of the benefits they receive. Those benefits are funded by payroll taxes, the burden of which is shared in some fashion among business owners and employed workers. Both groups more likely to save than the unemployed are. By keeping UI checks flowing during weak economic recoveries, supporters say, extended benefits boost total consumer spending, hence job creation, hence overall economic growth.
If the skeptics are right, then ending extended benefits will have a net positive effect on the labor market as well as broader economic measures such as gross domestic product. The employment rate will rise. The unemployment rate will fall, partly because of higher employment within a given jurisdiction but also because some previous recipients will move to or take jobs in other jurisdictions. On the other hand, if extended-benefits supporters are right, then there will be a net negative effect on the economy from ending the benefit extensions. Consumer spending and, thus, employment will fall. There will also be a rise in foreclosures, welfare participation, and destitution.
Here’s an important point to remember, however: in neither case is the net effect likely to be very large. That’s because most unemployed workers are not eligible for UI in the first place. They either left their jobs voluntarily, got fired for cause, or held jobs that weren’t covered by UI, including self-employment. Changes in the amounts or duration of UI benefits have no direct effect on them. In my original piece, which argued that extended benefits tend to do more harm than good, I wrote that “the labor-market effects of reforming unemployment insurance may not be massive. But they certainly don’t appear to be negative.”
Using the Payroll Data
Now let’s get to the evidence. During the six-month period in which North Carolina was the only state without extended benefits, July to December 2013, the establishment survey shows a 1.5 percent gain in payroll jobs in North Carolina. That’s much higher than the national average of 0.8 percent growth during the same period. However, Wolfers points out that North Carolina wasn’t the only state in our neighborhood to best the national average in job creation. Payroll jobs grew 1.6 percent in South Carolina and 1.1 percent in both Georgia and Tennessee. Virginia underperformed the nation at only 0.2 percent.
Wolfers then provide further evidence from the Quarterly Census of Employment and Wages, which showed that not only did South Carolina best North Carolina in payroll-job growth during the period but so did Tennessee and Georgia. However, the QCEW data for 2013 are raw and preliminary. Wolfers adjusted them for seasonal variation, understandably, but this does not appear to be standard practice. Unlike the CES, the QCEW is not intended for use as a time series. For periods of less than a year, the Bureau of Labor Statistics recommends that analysts use the payroll data from CES. A BLS official even told my colleague Don Carrington that economists had found some differences in seasonal patterns between the two datasets in the short run (in the long run the bureau uses the actual QCEW count to adjust the CES trends).
I know these are technical points. But that’s the nature of the discussion. If we use the dataset BLS recommends, from the establishment survey, then North Carolina’s 1.5 percent job growth from June 2013 to December 2013 outperformed any regional average one might choose, including our immediate neighbors (1.1 percent), the nine states of the Census Bureau’s South Atlantic region (1 percent), or the 12 states that make up the Southeast region used by the Bureau of Economic Analysis (also 1 percent).
During a subsequent radio program in which Wolfers and I participated, he argued that the best comparison isn’t to a regional average or a set of peer states but simply to South Carolina, given that the two states are contiguous, have a similar industrial mix, and often experience similar economic trends. That excludes too much relevant data, it seems to me. Since the Great Recession, for example, the variations in Tennessee’s employment have been about as close to North Carolina’s as South Carolina’s have, while GDP growth in Tennessee has actually been more comparable to North Carolina’s than South Carolina’s GDP growth has been. Manufacturing makes up 21 percent of North Carolina’s economy, 17 percent of South Carolina’s, and 16 percent of Tennessee’s. Surely it makes sense to compare North Carolina to several regional peers, not just to the one with which it happens to rhyme.
Using Other Economic Data
The more fundamental disagreement I have with Wolfers, however, has nothing to do with a choice of payroll datasets or regional comparisons. It is that by discounting the application of the Current Population Survey to this discussion, he tosses out the only set of data capable of providing answers to all the relevant questions. You see, the payroll data — be it from CES or QCEW — cover only about 90 percent of the labor market. Those who are self-employed or work for employers not covered by UI aren’t included. Moreover, payroll data tells us the number of covered jobs, not the number of employed people or where they reside.
Imagine that you are a long-time unemployed worker living in Gastonia. If your UI checks run out, you have several choices. One would be to accept a job in Gastonia. That will show up in a CES or QCEW count of payroll jobs. But what about other choices? If you keep your residence in Gastonia but take a job over the border in South Carolina, that won’t show up in a count of NC payroll jobs. Nor would your decision to move to Georgia to take a job. Nor would a familiar choice for displaced workers: starting your own business.
There is no doubt that the monthly household statistics from the CPS are less than ideal for addressing these questions. The sample sizes for states are too small. Fortunately, the Bureau of Labor Statistics supplements them with other measures of the labor market. They are based on 12 months of data in each state, thus addressing the sample-size problem, and are updated every quarter.
They consist of six different rates, U-1 through U-6. The U-3 rate is the standard one, consisting of the number of people unemployed and looking for work divided by the civilian labor force (which is the sum of all employed workers and the unemployed looking for work). More useful are the broader measures: U-4 (which adds in discouraged workers who’ve stopped looking for jobs), U-5 (which adds in all workers who have dropped out of the labor force for any stated reason), and U-6 (which adds in workers with part-time jobs who’d rather be working full-time).
Being annual averages, these statistics aren’t a perfect fit for evaluating a six-month period, admittedly. On the other hand, they actually take into consideration all the worker responses we’re interested in. North Carolinians who start their own businesses or take cross-border jobs are counted as employed North Carolinians. Those who leave North Carolina to find jobs elsewhere are no longer counted as unemployed North Carolinians.
If you start with average U-5 rates as of June 2013 and then compare them with average U-5 rates as of December 2013, you are replacing six months of data before the UI change with six months of data after it. During this period, North Carolina’s U-5 rate declined from 10.5 percent to 9.3 percent. That 1.2 point drop was the largest in the nation — larger than South Carolina’s drop of .9 and significantly larger than the average of our neighboring states (.6), the South Atlantic average (.6), the Southeast average (.5), and the national average (.4).
If you then advance to the average U-5 rates as of June 2014, you have an entire 12 months of data from after the end of extended benefits in North Carolina. If the liberals were correct, the data would paint a dismal picture. They don’t. Our state’s average U-5 rate through June 2014 was 8.1 percent, down from 10.5 percent a year earlier. (Again, these measures include those who’ve dropped out of the labor force.) That drop of 2.4 points far exceeds the regional and national averages. In fact, as of June 2014, North Carolina’s U-5 rate now matches the national U-5 rate for the first time since the onset of the Great Recession.
In my original Journal piece, I pointed to another piece of evidence: state GDP. If the liberals were correct about the macroeconomic effects of ending extended benefits, then surely there ought to be at least some adverse effect evident in the broadest macroeconomic indicator. But look at real (inflation-adjusted) GDP per state for 2013. North Carolina’s grew by 2.3 percent in 2013, compared to the national average of 1.8 percent, the Southeast average of 1.6 percent, and an average of just 1.2 percent for our neighboring states.
It would be foolish to argue that North Carolina’s impressive gains in economic performance during the period were wholly or substantially caused by ending the disincentive effect of extended UI benefits. There just aren’t enough affected workers to explain the trend. Also, the June 2013 to June 2014 comparison for U-5 includes six months in which the whole country, not just North Carolina, had ended extended benefits. Similarly, the GDP figures for 2013 include six months in which everyone had extended benefits and six months in which only North Carolina had exited them. Many other factors — policy and nonpolicy — must be in play here.
Alternative Conclusions
There are three possible interpretations of the available (imperfect) evidence. One of them, advanced by Justin Wolfers in his Times piece, is that we should look solely at the payroll data and primarily at a comparison between North Carolina and South Carolina. He concludes that the end of extended benefits had no significant effect on labor-market outcomes, positive or negative. I don’t agree with Wolfers, but his position is certainly defensible and capably argued.
Another interpretation, the one I find more persuasive, is that we should look at a range of relevant, valid statistics — payroll jobs, yes, but also U-5 and state GDP — and use a reference group larger than a single state. These decisions put North Carolina’s economic performance in a more favorable light, and suggest that unemployment-insurance reform likely played a positive role in the trends, although probably not a massive one.
The third possible interpretation, still favored by liberals and some conservatives, is that ending extended benefits had deleterious results for the labor market and larger economy, first in North Carolina in the last half of 2013 and then in the nation as a whole during 2014. I’m unaware of any persuasive evidence for this conclusion. Since the entire extended-benefits program expired at the end of 2013, the American labor market has clearly experienced substantial improvement. The U-3, U-4, U-5, and U-6 rates are all down substantially. The employment-population ratio is up. Fewer people are on food stamps. The gains appear to be particularly strong among the long-term unemployed.
In short, the supporters of UI extended benefits predicted dire economic consequences from the expiration of those benefits. The predicted consequences didn’t happen. On that, Wolfers and conservatives agree.
-30-
Hood is president of the John Locke Foundation.