Faced with what I suspect is the biggest-ever forecasting error for a single data release — the U.S. employment report for May on Friday — the community of applied economists and Wall Street analysts is trying to determine what went wrong in their approach and how this applies to other key projections and policy recommendations. Their work is being complicated by questions about the robustness of some of the data compilation and classification practices used by the Bureau of Labor Statistics and Census Bureau, especially during a pandemic, as well as the high degree of political polarization of numbers. While the work is continuing, some elements are already clear, at least to me.
What makes this forecasting error so historic is more than the magnitude of the miss on the two widely reported and followed headline numbers: net jobs created, where the consensus expectation was off by an astounding 10 million, and the unemployment rate, off by some 6 percentage points, or what Wall Streeters like to call “six big figures.” In addition, economists and Wall Street analysts were uniformly wrong on the direction of the move, forecasting deterioration when the vast majority of the reported numbers pointed to a uniform improvement. The homogeneity of the miss was also notable as not one of the 78 economists surveyed by Bloomberg came close to the actual number. For example, the most optimistic projection for payrolls was a decline of 800,000; the actual number was an increase of 2.5 million.
With the Covid-19 disruptions posing challenges for data gathers, particularly for the compilation and classification of micro numbers, data revisions are likely in the future. The BLS has already pointed to one category in particular — the one meant to account for those “absent from work for other reasons” than being unemployed, such as jury duty. This category has jumped markedly in magnitude during the coronavirus-induced economic shock and probably now includes people that are unlikely to be employed again soon.
The BLS indicated that adjusting for this likelihood would result in an unemployment rate some 3 percentage points higher than the 13.3% reported on Friday. But, if applied consistently to the previous months data, too, this adjustment would do little, if anything, to alleviate the enormity of the data miss. For example, it would still have shown a drop in the unemployment rate, this time to 16.3% in May from 19.7% in April.
What is particularly puzzling for economists like me, and for Wall Street analysts, is that the BLS report seemingly missed a stage that the vast majority of all other data are going through, or expected to do so: that of getting bad at a slower rate before improving. This aspect will undoubtedly be an area of significant research, including the impact and implications of government policies such as the PPP (Paycheck Protection Program). After all, it is hard to find any higher frequency data that suggest that, as of mid-May (the effective snapshot date for the BLS jobs report), the economy had already turned sharply.
As embarrassing to the profession as this is — remember, only 12 years ago at a briefing on the global financial crisis, Queen Elizabeth put the profession on the spot by asking “why did nobody notice it?” — it’s more dangerous to appropriate policy making. A significant positive data surprise, especially one embraced eagerly by financial markets, which led to a record high for the Nasdaq and strong gains elsewhere, can easily be interpreted by some politicians that the economy is on the mend and that they can step back to the sideline and let the endogenous forces of recovery do the heavy lifting.
While welcoming the improvement in the rate of change, we should not lose sight of the overall levels given that Covid-19 has inflicted the biggest shock on the economy since the Great Depression. Whichever of the two unemployment rates you pick (13% or 16%), it is still well above the 10% peak reached during the Great Recession. The 2.5 million jobs reportedly created in May constitute only a small recovery of the 20.5 million lost in April. And pain and suffering are still exhibited every day in long line at many food banks around the country.