The US government agency responsible for publishing labour market statistics is struggling to pin down the actual unemployment rate in the world’s largest economy.
The problem is due to “misclassification” of workers in a key survey, struggles with data collection during the pandemic, and massive flows in and out of jobs.
On Friday, the Bureau of Labor Statistics, a unit of the US labour department, released its latest report on the American jobs market, showing unemployment dropping from 14.7 per cent in April to 13.3 per cent in May, an unexpected improvement.
But the data came with a caveat: the agency acknowledged that some furloughed employees had been labelled as working but absent, when they should have been classified as temporarily laid off. If not for that mistake, the unemployment rate would have been 3 percentage points higher.
The revelation was not new and in fact the problem has been getting smaller: in April, closer to the start of the pandemic, the BLS had said the “misclassification error” was in the order of 5 percentage points, which meant the US unemployment rate might have neared 20 per cent.
But the agency’s own admission that a higher, alternative jobless rate might be more accurate than the official one has placed it in the spotlight at a time when economic data is crucial in assessing the depth of the recession, and more politically sensitive given the presidential election year.
It also revealed the difficulty of producing economic statistics at a time of massive, sudden economic transformation.
“It’s not that the questions have changed or the interviewers have changed, the circumstances have changed dramatically from anything that has been encountered before,” said Erica Groshen, a former BLS commissioner under president Barack Obama, now at Cornell University. “You have employers who have never laid anybody off before, you have workers who have never been laid off before, this happened essentially overnight, and both sides are intending to pick up where they left off as soon as they can,” she said.
The BLS’s main quandary is that in its judgment — and its instructions to field staff — people furloughed during the pandemic should be classified as unemployed on a temporary basis, because it is an involuntary absence unlike other forms of leave. But when asked about their status, many respondents have been answering that they remained employed, but absent for “other reasons”. That category is usually reserved for limited departures from the workplace due to medical leave or jury duty, rather than homebound people waiting for pandemic restrictions to be lifted.
“At a moment like this, the line is blurry over how furloughed / hiatus / temporarily closed workers should be classified. They gave guidance to the field personnel but clearly there was ambiguity and some reported it one way and some the other,” said Austan Goolsbee, an economics professor at the University of Chicago’s business school and former economic adviser to Mr Obama.
BLS declined a request for an interview but on Friday said it and the US Census Bureau were investigating why discrepancies continued to occur and was taking additional steps to address it, suggesting a ramped-up training effort for field staff to prompt respondents differently so they do not place themselves in the wrong category.
In the data release, BLS revealed other difficulties it was having with producing the unemployment statistics at this time, including the fact that the survey of 60,000 households they are based on are partly conducted through in-person interviews that had to be cancelled due to social distancing guidelines. That led the response rate to be 15 percentage points lower than pre-pandemic levels, at 67 per cent, the BLS said.
The trouble has been compounded by the epic nature of the shifts happening in the labour market at the moment. The separate establishment survey released by BLS at the same time as the unemployment data showed that employers created 2.5m jobs in May, which was also a surprise, marking a small rebound from the 20.5m positions shed in April. But that was a net figure: according to a JPMorgan analysis of the data, 7.7m workers moved from unemployment to employment last month, and another 5.5m moved from outside the labour force into employment. But at the same time, 4.9m people moved in the opposite direction, from employment to unemployment.
Inevitably, given the heated political atmosphere in Washington, the confusion over the unemployment rate triggered some accusations of political bias at the BLS, suggesting Donald Trump was trying to suppress the true state of the jobless crisis. Paul Krugman, the New York Times columnist, apologised after saying on Twitter one could not “completely discount” the possibility that the US president had interfered with it.
However, former US economic officials have dismissed the idea of any political influence, saying statistics are assembled by civil servants in a very quick timeframe. The commissioner, currently William Beach, a former Republican staffer on Capitol Hill is the sole political appointee at the agency, and reviews the reports only once they are completed.
“I think there would have been red flags [if it had been tampered with]. Either the release would not have come out in time, or it would have been missing some of the back-up information that normally accompanies it,” said Ms Groshen. “And I would have been really surprised if somebody hadn’t gotten an inkling of it at BLS, because they are so dedicated to the integrity of the product and the agency that word would have gotten out”, she said.