In a year of almost unmitigated disaster, the May jobs report was a pleasant surprise. Instead of further heavy job losses and a rising unemployment rate, the report showed the economy adding 2.5 million payroll jobs, with the unemployment rate falling to 13.3% from 14.7% in April.
For those of us in the forecasting business, it is necessary, first, to understand why the numbers were better than expected. Second, we need to consider what these and other recent data points imply for the economic outlook. Finally, it is important to ask whether this altered forecast justifies either the very powerful recent rally in stocks or the still very low level of interest rates.
Most of all, though, it is crucial to be clear-eyed about the current state of the economy and speed of the recovery. The 2.5 million May jobs gain, while very welcome, followed a cumulative job loss of 22 million jobs in the prior two months. It’s a little like falling off a cliff onto a trampoline—the bounce may be impressive but the fall is more important.
For the record, U.S. non-farm payrolls rose by 2.5 million compared to our forecast of a 2.7 million decline and consensus expectations for an 8.0 million plunge. The unemployment rate fell from 14.7% in April to 13.3% in May compared to both our own and consensus forecasts of 19.8%. So why were the forecasts so wrong?
I believe the main problem was that many of the variables normally used to predict job gains and losses were unreliable, given the magnitude of the prior month changes and the unprecedented nature of the social-distancing recession. Consequently we, like many other forecasters, had to rely primarily on state unemployment claims data.
For background, the monthly household and payroll employment surveys always refer to the week containing the 12th of the month. Consequently, the two relevant weeks were the weeks ended Saturday, April 18th and Saturday, May 16th.
Between those two weeks, 14.2 million people filed initial claims for unemployment benefits and this was likely the source of many of the higher estimates of job losses. However, the change in continued unemployment benefits should have been a better measure of net employment changes, since they exclude claims filed but denied and net out those hired or rehired from the rolls of the unemployed. These continuing claims for unemployment benefits rose by 2.8 million between the two survey weeks.
This, of course, still implies a huge job loss between April and May. The most obvious explanation for why this didn’t occur is that continuing claims for the April survey week may have been artificially suppressed by the fact that state unemployment offices were simply overwhelmed in April and so couldn’t process claims and that this problem was much less serious in May. The undercount of those eligible for unemployment benefits in April then meant that the actual improvement in the labor market between the survey weeks for the monthly jobs report simply wasn’t visible in the claims data.
A similar argument can explain why employment, as measured by the household survey, rose by 3.8 million, helping the unemployment rate fall from 14.7% to 13.3%. Note that the measured unemployment rate still likely sharply underestimates the true unemployment rate (as it did in April) because the rolls of the unemployed exclude millions who are ready, willing and able to work but are not currently looking for a job because of the pandemic or who are being incorrectly classified as being employed but not at work when they are, in reality, unemployed.
Of course, explaining a missed forecast is one thing—understanding the underlying trend and projecting it forward is a separate but more difficult job. So where do we go from here?
The stronger-than-expected jobs report adds a little, but only a little, to estimates of second-quarter GDP. GDP is a measure of output, not income, and, to the extent that employers recalled workers to satisfy the requirements of the paycheck protection program, some of the higher employment numbers will not show up in higher output so much as in diminished productivity. Still, higher employment in the services sector suggest slightly higher consumer spending on services. In addition, other data last week on light vehicle sales and construction activity added a little to our estimates of second-quarter GDP.
Even with this, however, we still expect real GDP to fall by close to 40% annualized in the second quarter before rebounding by between 15% and 20% annualized in the third. Thereafter growth should slow before reaccelerating after the widespread distribution of a vaccine, hopefully in the first half of 2021. Even with this, however, we expect real GDP to be lower in the fourth quarter of 2021 than it was in the fourth quarter of 2019.