Commentators and economists alike were shocked last Friday when the government reported the U.S. added 2.5 million jobs in May, and that the unemployment rate dropped to 13.3 percent from 14.7 percent a month earlier. Forecasters had expected the economy to shed 7.5 million jobs in May, not to add millions to payrolls.
Instead of trying to figure out what the numbers meant, some of the president’s critics were quick to challenge the numbers themselves. Paul Krugman, for example, wrote on Twitter about 20 minutes after the numbers were released that “you can’t completely discount the possibility that” the president’s political team had “gotten to” the Bureau of Labor Statistics, the government agency that produces the monthly jobs report.
To his credit, Krugman quickly apologized. But some are still questioning the integrity of the numbers.
This is absurd. There isn’t a shred of evidence the BLS fudged the numbers, let alone violated its integrity. It’s important to learn from this controversy, because the U.S. will likely see more such surprises in economic data in 2020.
To calculate the unemployment rate, the government surveys 60,000 households each month. These households — which create a statistically valid picture of the U.S. population as a whole — are not asked directly if they are unemployed. Instead, they are asked a series of questions designed to obtain an accurate picture of their job-market activities.
People who reported that they didn’t work in May and expected to be recalled to their jobs should have been classified as unemployed on a temporary layoff. Many were, but several million workers who should have been counted as unemployed were instead listed as employed but absent from work. If these workers had been properly classified, the unemployment rate would have been 16.3 percent according to the BLS, not the 13.3 percent that was officially recorded.
If the BLS thought the official number was wrong or misleading, why didn’t it just change the number? The bureau’s longstanding practice is to accept data from the survey as recorded. This helps ensure the integrity of the data by keeping it free from exactly the type of ad hoc changes the Trump administration’s critics are worried about.
Challenges in properly classifying workers in the middle of an economy-shaking pandemic are a far cry from fudging the numbers for political purposes. Besides, the reason everyone knows this happened is that the BLS announced it in its monthly employment report. If officials were trying to cook the books, they wouldn’t have put the recipe on page six of the report. The same problem happened when calculating the unemployment rates for March and April, as well — and the BLS noted it in both reports.
If that doesn’t convince you, consider this: Reassigning the status of the workers in question from employed to unemployed would increase May’s unemployment rate by three percentage points. But making the same change for April would increase the unemployment rate in that month by 4.8 percentage points. So the drop in May would have been larger if the misclassification hadn’t occurred. The upshot is that, if anything, the president’s supporters have a stronger case that the BLS is cooking the books against them than his critics have that the numbers are being fudged to help President Trump.
Economic statistics are constructed for normal times. The U.S. economy, shaken by the pandemic, is not in a normal time, and that makes answering standard survey questions much more difficult. If you are absent from work at the time of the survey due to illness, then you should be counted as employed. But what if you aren’t at work because of fear of getting sick?
You should be counted as employed if you’re absent from work because of a vacation. But if you’re not at work but are getting paid through the Paycheck Protection Program, how should you answer the survey questions? If your employer shuts down, then you’re supposed to be counted as a temporary layoff. But if your employer tells you that he expects business to be back to normal soon and will want you back, you may not think of yourself as having lost your job.
Labor-market statistics are not the only ones that are harder than usual to forecast and interpret. The average monthly increase in personal income in the five years before the pandemic was 0.4 percent. In April, personal income increased by 10.5 percent. Forecasters expected a 6.5 percent drop.
In the five years before the pandemic, the average personal savings rate was 7.4 percent. In April, households saved 33 percent of their income. Is this because people were locked at home and couldn’t spend? Or has the pandemic changed people’s preferences or expectations about their future income, increasing their desire for a large rainy-day fund?
Economic numbers will continue their wild ride throughout this summer and into the fall. Typically, around two million workers transition from unemployed to employed every month. In May, 7.7 million did — by far the largest flow in the history of the data. Flows into unemployment were unusually large, as well. Forecasting net changes in this environment is a major challenge. Observers should expect to be surprised by economic data.
False accusations of foul play may not be confined to May’s employment report. The nonpartisan Congressional Budget Office expects the economy to grow this summer at a record 21.5 percent annual rate. In the fall, the budget office forecasts 10.4 percent growth.
Expect the president’s critics to claim that the books may be cooked. Their suspicions about corruption would be wildly misplaced, but their sense that the numbers are off will be well-grounded in the reality that the economy will be in terrible shape despite this impressive growth.
It’s important to smack down hard this nonsense about foul play. The data are only as good as the quality of survey responses. And the data are used for critically important decisions.
If the government’s measure of consumer prices is off by even a little bit, Social Security payments for retirees, determined in part by price inflation, will be off. Hundreds of billions of dollars of federal funds are distributed to states each year based on government survey results. The Federal Reserve’s monetary policy is heavily influenced by economic data. It’s important for the data to be as accurate as possible, and that requires public trust in the data and in the process.