Federal Tax Receipts Show A Record Plunge In May, Raising More Doubts About Employment Data Accuracy

Submitted by Joseph Carson, former chief economist at Alliance Bernstein

The Monthly Treasury Statement for May showed federal withheld income tax receipts falling a record 33% from the comparable period one year ago. The decline in May tax receipts exceeds the 30% decline in April. Monthly tax (gross) receipts have been reported since 1973 and April and May declines are the largest on record.   

Federal withheld tax receipts are directly related to workers paychecks. The scale of the decline in tax receipts is nearly three times the decline in reported household and payroll employment. The unprecedented gap raises questions about the accuracy of the April and May employment reports.

Tax Receipts vs. Employment

Federal withheld (gross) income tax receipts are highly correlated with employment levels and wage growth since taxes are withheld from workers paychecks. Monthly receipts can be noisy, often influenced by the number of workdays. Nonetheless, back-to-back monthly declines are rare and have only occurred during periods of exceptionally large declines in employment or when there have been legislative changes that lower peoples’ withheld tax payments.

There are no legislative changes that would result in dramatically lower withheld gross federal income tax receipts. So the logical conclusion is that the sharp drop in withheld income tax receipts is directly related to a plunge in wage and salary income.

Without question, the tax data raises doubts over the scale of reported job loss as well as industries that experienced the largest declines.  Tax receipts are off over 30%, while employment levels are off roughly 13%. How can tax receipts fall three times more than employment?  As puzzling as that appears to be what is equally puzzling is that the vast majority of job loss was concentrated in lower-wage industries, such as leisure and hospitality and retail trade.  If job loss was concentrated in low wage industries one would not expect tax receipts to fall three times as fast as overall employment.

In my recent article, BLS Fails Its Mandate: “Fearless Publication of the Facts” published on June 7, I made the argument that the Bureau of Labor Statistics (BLS) statistical methodology failed to ensure an accurate account of the employment situation. The tax data for April and May offers strong evidence that the employment data is inaccurate.

Mr. William W. Beach, BLS Commissioner, is quoted as saying there was no political bias in the published reports and the unemployment rate reported below its actual rate was “accidental”.  I agree with Mr. Beach on the first point.  BLS government statisticians operate with the highest integrity. I know that to be the case because I started my career as a government economist/statistician at the Department of Commerce.

However, BLS statistical methodology did fail; not once, but twice in reporting grossly inaccurate employment statistics.  The US statistical system is the “gold standard” of the world, producing the most accurate, and always operating with the mandate, “Fearless Publication of the Facts”.  But the employment reports  for April and May show that no statistical methodology is perfect, and its the responsibility of BLS to ensure the accuracy of the data.  The scale of the error in April was so large it should have set alarm bells so to avoid another “accidental” report of bad data.

The sharp drop in withheld income tax receipts strongly suggests that the “error” term of the household employment could even be larger than what BLS has stated.  BLS said that the number of households entering in their survey for the first and second time was 30 percent below the average of the past 12 months. As a result, BLS was compelled to use a historically low number of responses to estimate household employment for April and May.  So its highly possible that the number of people misclassified as employed instead of temporarily unemployed could be far larger than the 8.4 million for April and the 5.4 million in May.

Household employment data is based on a sample of 60,000 households out of a total household population of 125 million.   Federal tax receipts are unambiguous. They reflect withheld income taxes taken directly from 30 million business establishments employing over 150 million workers before the pandemic.  Which data series–reported household employment or withheld taxes—do you think offers a more accurate picture of the current employment situation?