Dr. Steven N. Weisbart, CLU, Triple-I Senior Vice President and Chief Economist
COVID-19 pandemic has not only disrupted our economy – it has complicated the data we routinely use to understand economic developments. This is a bit like finding out the thermometer you use to tell if you have a fever is unreliable.
Here are two examples of why it’s hard to know what’s happening.
What is the correct unemployment rate?
The April 2020 Bureau of Labor Statistics (BLS) employment report said the U-3 rate – just one of six unemployment measures BLS reports – was 14.75 percent. This number is derived by dividing the number of people counted as unemployed (23.078 million) by the civilian labor force (156.481 million), which is everyone who is either working or unemployed and looking for work.
But when the virus was recognized as a major public health threat in mid-March and April and many businesses and organizations were shut down, throwing many millions out of work, some who were affected decided to retire. This means they were no longer counted as part of the civilian labor force. This is most vividly seen by comparing the civilian labor force in February (164.6 million) with its count in April (156.5 million)—a drop of 8.1 million.
The large number of retirees affected the unemployment rate: if they had not retired, most would likely have been counted as unemployed. To keep the math in our example simple, let’s say 7 million of the retirees had remained in the labor force and been counted as unemployed (maybe the other 1 million would have retired then anyway—virus or no virus). The unemployment count would have been 30 million (23 million counted plus 7 million un-retirees) and the civilian labor force would have been 163.5 million (156.5 counted plus 7 million un-retirees).
The unemployment rate would have been announced as 30 million divided by 163.5 million, or 18.35 percent, instead of 14.75 percent.
So, which one is correct?
Are seasonal adjustments still correct?
Macroeconomists have long recognized that many economic data have seasonal patterns. For example, retail sales often spike in the last quarter of the year because of the holidays. Sales for some items, such as those bought for “back to school,” spike at other times. So, to see what’s really happening, economic data are often adjusted to account for the seasonal effects and reported after these adjustments are made.
To see the effect of seasonal adjustments, look at the following two graphs. The first is employment in the construction industry that is not seasonally adjusted. The second is the same industry and time; the only difference is that its data are seasonally adjusted.
Construction employment obviously dips in the cold months, and the drop shown in the first graph doesn’t represent any significant economic change, so the seasonal adjustment in the lower graph lets us see only changes beyond the seasonal adjustment, such as what happened in 2020.
The problem, from an economic analysis viewpoint, is that the amount of seasonal adjusting to apply is a judgment call, and it is often based on a historical period in which conditions were much as they are now. But what’s happening now has no satisfactory historical precedent.
So should we keep using the seasonal adjustment factors from before, or do they not apply to the current economic situation?
These are just two examples of datasets or analytical approaches whose relevance can be called into question in light of COVID-19 – further complicating the already complex and nuanced endeavor of attempting to understand and anticipate economic developments.