Cooking Official Statistics Is Not Easy, for Now
After the Bureau of Labor Statistics announced a drop in the unemployment rate—from 14.7% in April to “only” 13.3% in May—a friend emailed me to share his suspicion that the unexpectedly low figure was a propagandist lie. The probability of that is not zero, I explained to him, but it is extremely low.
These data are gathered (through a monthly survey of 70,000 households), assembled, analyzed, and summarized by bureaucrats from the Census Bureau and the BLS, many of whom are professional statisticians. Bureaucrats could of course be co-opted or corrupted by political leaders, as they were in Argentina and Greece not so long ago. But there are reasons why this is less likely to happen in America.
Any attempt at political interference in official statistical data (which would probably be a crime under federal law) could be resisted or blocked at many points in the process. Successful conspiracies involving a large number of people are rare because, like in the Prisoner Dilemma game, anyone has an incentive to defect before anyone else does. A political manipulation at the last stage would be visible to many who have participated in the process. (The BLS Commissioner apparently only sees the report once it is completed.) Any success at political manipulation would quite certainly be followed by some resignations. High-level bureaucrats have an incentive to preserve the value of their personal brands. A professional statistician suspected of having acquiesced to data fraud may be unable to find another job in his field. Moreover, a political manipulation would be interpreted as meaning that US statistical agencies having become of the Greek or Argentine sort. The credibility of all federal statistical agencies would suffer—and may take decades to recover. Treasury yields would probably increase as creditors would suspect that the federal deficit and debt numbers, for example, are cooked too.
Another part of the difficulty would be to reconcile false unemployment statistics with other numbers calculated by other federal statistical agencies, such as the Bureau of Economic Analysis (at the Commerce Department), which will, at the end of July, provide a first estimate of second-quarter GDP. And note that cooking a number one month may require cooking it again the following month and so forth, increasing the probability of the fraud being discovered.
Think also of the Department of Labor’s Inspector General, who may investigate any suspicion of statistical manipulation. It is true that federal Inspectors General may now be scared of investigating political malversations after President Trump removed five of them in different agencies over the past few months. But who knows, the Department of Labor Inspector General may still investigate out of personal integrity or because his legal responsibilities require it.
Fortunately, then, lying is not always easy in a government limited by the rule of law and constrained by numerous centers of power. We could say that, like in Kipling’s delicious novel The Man Who Would Be King (1888), even the king cannot do everything he wants.
The intriguing error in employment data made by the BLS over the past three months does not change my opinion. As my co-blogger David Henderson explained, an error by interviewers led to misclassify the workers furloughed due to the coronavirus as employed instead of “unemployed on temporary layoff” as they should have been. Without this error, the correct unemployment rate would have been closer to 20% in April and to 16% in May, as opposed to the published figures of 14.7 and 13.3%. This big error blunts the impact of the pandemic and especially of government measures to combat it.
The notification of this error in the BLS’s June 5 report covering May (available at https://www.bls.gov/bls/news-release/empsit.htm#2020) reads as follows:
However, there was also a large number of workers who were classified as employed but absent from work. As was the case in March and April, household survey interviewers were instructed to classify employed persons absent from work due to coronavirus related business closures as unemployed on temporary layoff. However, it is apparent that not all such workers were so classified. BLS and the Census Bureau are investigating why this misclassification error continues to occur and are taking additional steps to address the issue.
If the workers who were recorded as employed but absent from work due to “other reasons” (over and above the number absent for other reasons in a typical May) had been classified as unemployed on temporary layoff, the overall unemployment rate would have been about 3 percentage points higher than reported (on a not seasonally adjusted basis). However, according to usual practice, the data from the household survey are accepted as recorded. To maintain data integrity, no ad hoc actions are taken to reclassify survey responses.
(The constraint of maintaining data integrity exists to prevent intentional manipulation.)
A notice similar to the one above appeared in the report for April (published May 8) as well as in the report for March (published April 3); see also https://www.bls.gov/bls/news-release/empsit.htm#2020 for these reports. The same data collection error was committed three months in a row.
Let’s hope the BLS and the Census Bureau continue investigating until they find how the error happened. And let’s hope that their Inspectors Generals are (still) ready to do their own investigations if necessary.