The Jobs Numbers are Bad

The jobs report was released today, October 2nd, with the preliminary estimate of the unemployment rate at 7.9%. Following last month’s better-than-expected jobs report, Senate Republicans unveiled a slimmed-down coronavirus relief bill. The more comprehensive House bill remains in committee, where it is unlikely to progress. After a quick floor vote in the Senate, both bills remain effectively blocked, resulting in a very low likelihood of any additional relief. The chances for further coronavirus relief have slipped since June due to reduced urgency for broad legislation for some policymakers, and this jobs report only further reduced the urgency. Following the most recent release of the jobs report, the President and members of the administration have been bullish about the economy. In fact, the day the jobs report was released in September, the president praised the “unprecedented economic recovery” during a press briefing, signaling that the administration viewed the recession as being largely solved. The problem is that the recession is ongoing, the jobs numbers are still bad, and almost certainly wrong—potentially by a lot.

Throughout the recession, the jobs report has consistently overestimated employment, forcing the Bureau of Labor Statistics to make drastic corrections to the data. Revisions are very normal, and overestimation in preliminary figures is somewhat common during a downturn. However, the magnitude of these corrections is unprecedented, with many preliminary estimates having undergone revisions of historic proportions. The scale of the corrections has mirrored the unprecedented joblessness claims. In addition, there remains economic debate regarding the underlying methodology in calculating unemployment, with many noted economists offering alternative estimates. Collectively, this has concerning implications, considering that policy is made with respect to these preliminary figures.

For some context, the Bureau of Labor Statistics generally releases a monthly Employment Situation Summary the first Friday of every month, commonly referred to as the jobs report. The top line statistic in the jobs report is the unemployment rate, which summarizes the employment level of the previous month. The October jobs report pegged the unemployment rate at 7.9%, down from an all-time high of 14.7% only a few months prior. The report also features other employment statistics that focus on various sectors and occupations, such as non-farm employment or alternative measures of labor utilization.

Importantly, the jobs report also contains revisions. These reports and the figures within them are instrumental in setting policy, but revisions are often overlooked. Even though revisions provide a more accurate depiction of the data and the added value comes from reevaluating previous decisions and potentially correcting for the policy direction.

With unemployment data, the Bureau of Labor Statistics will release a preliminary estimate, and then issue additional revisions, typically two if needed, in the following months. After that process, the numbers will be more or less settled. Again, data revisions are common, most data releases will update prior estimates. Every data release is referred to as a vintage, or a data vintage, and reflects the best estimates at that time.

There is an entire literature on data vintages within economics, that consists of researchers doubling back on empirical work with updated data, to test whether prior conclusions still hold. Using data vintages, researchers can catalog when earlier results will be significant. The direction of revisions is expected to be random in nature, meaning that there should not be a pattern to the correction. Errors should be equally likely to be either an overestimation or an underestimation. However, the recent revisions to employment figures have tended towards overestimation.

Massive revisions are not the only problem with the unemployment rate. There are also important definitions and methodological assumptions that are worth reassessing. By slightly modifying these underlying assumptions, two Harvard Kennedy School economic professors, Jason Furman and Wilson Powell III, were able to project an alternative estimate. Furman and Powell argue that two issues are driving the lower unemployment rate, namely 3.7 million workers were deemed to have left the labor force, and a misclassification of an extra 1.1 million who were not at work for other reasons. After making these adjustments, Furman and Powell conclude that recent estimates of the unemployment rate were probably closer to 10%.

Although there are multiple official unemployment rates, the most commonly cited unemployment rate is called the U-3 by the Bureau of Labor Statistics. The U-3 consists of the total unemployed, as a percent of the civilian labor force. The U-3 is a middle-estimate, on a scale of labor utilization with U-1 as the most restrictive measure and U-6 as the most comprehensive. The precise definitional contours surrounding each category can result in misclassification. The U-3 unemployment rate is very susceptible to changes to the denominator, or who is considered in the labor force. Furman and Powell’s estimates leans on this fact, by expanding who is considered in the labor market the unemployment estimate ticks upward.

Regardless of the metric, the unemployment rate is improving, but policymakers should not conclude that the recession is over. The recession is still very much ongoing. The unemployment numbers confirm this. The Sahm Rule, a real-time recession indicator named after economist Claudia Sahm, still shows record peaks. When the Sahm Rule indicator rises above 0.5, that signals the start of a recession. The indicator has risen to 5.3, remaining higher than any point during of the Great Recession. Importantly, the indicator relies on comparing unemployment data. Though the Sahm Rule is not designed to double as a metric that signals when recessions have ended, having a recession indicator near a record high should provide policymakers serious pause.

Setting policy under the assumption that the unemployment rate is sharply falling and is nearing normal levels will result in different legislative priorities than recognizing that these metrics are difficult to accurately assess in real-time, especially in the midst of a recession. Importantly, as economist Kate Bahn notes “The jobs being added aren’t the jobs that were lost, because we’re still in an uncontrolled pandemic — people are going back to worse-quality jobs without increased pay.” Additionally, there remain millions of workers filing initial joblessness claims every month, even as unemployment declines. Legislation has stalled under the pretense that unemployment is normalizing, but even accepting the low-end estimates the joblessness numbers are still very bad. In addition, there remains active academic debate over what is a normal level of unemployment. Some argue that normal unemployment, or slack in the labor market, could be much lower than previously considered.

Regardless of the debate around the natural unemployment level, the metric, indicator, or any underlying methodology, the economic outlook is bleak. GDP forecasts are fairly low and the unemployment rate continues to far exceeds peer countries, in part because of legislative inaction and poor policy choices. The corrections have tended to rely on preliminary overestimations, and other metrics and estimates suggest that unemployment is far higher. Policymakers should remain cautious of the most optimistic projections and take a fuller view of the jobs report and the unemployment rate. The economy is improving, but has not completely improved, and policy should reflect that.

---

Note: The views expressed by authors do not necessarily reflect the views of Chicago Policy Review.

Mark Sheppard is a research assistant at the University of Chicago, Harris School of Public Policy and the Director of Communications for the Chicago Policy Review. Visualizations created by Mark Sheppard.