Sacking over hiring numbers: How Trump has undermined US data collection

McEntarfer’s firing has been condemned by most outside the White House. The National Association for Business Economics, for instance, said the accusations against the BLS were “unfounded” and “threatens the long-standing credibility of our economic data infrastructure”.

Staffing shortages

Collecting quality data is difficult at the best of times. Surveyors all over the world, including in India, are facing the issue of lower response rates. According to the BLS, the household response rate in the Current Population Survey – which helps provide data on the labour force – has fallen from 88.3 per cent in April 2015 to 68.1 per cent in April 2025.

But rising non-response rates is an issue that plagues all countries. In India, the Ministry of Statistics and Programme Implementation (MoSPI) has, for some time, been concerned by the increasing apathy of the upper class to government surveys, particularly households living in gated societies. In the US, there has been another challenge.

On January 20, the Trump administration froze federal hiring, with the Department of Government Efficiency (DOGE) aggressively cutting government jobs. This has affected data collection, with the Wall Street Journal reporting in June the BLS had told some economists that staffing shortage had led to Consumer Price Index (CPI) data being collected from fewer shops starting April. “These procedures will be kept in place until the hiring freeze is lifted, and additional staff can be hired and trained,” the BLS said.

As it turns out, the BLS suspended CPI data collection in the cities of Lincoln, Nebraska and Provo, Utah in April “to align survey workload with resource levels”. In June, data collection in Buffalo, New York was suspended, with the BLS saying on July 29 that “roughly 15 per cent of the sample in the other 72 areas also was suspended from collection, on average”.

Meanwhile, the Trump administration’s spending cuts for the current fiscal would lower the budget and headcount of the BLS by around 8 per cent. According to the Council for Community and Economic Research, this would limit the ability of the BLS’ to meet growing demand for sub-state and real-time labour market data.

“The agencies, BLS included, are in survival mode,” news website Politico reported in June, quoting Michael Horrigan, a former BLS Associate Commissioner. “You can’t save that money simply by cutting sampling or cutting indexes. The only way to save that kind of money is by cutting survey programs.”

‘Imputation’ in US too

To get around fewer CPI data points, the BLS has been using a technique known as ‘imputation’. When a product’s price is missing, it is derived from that of an item in the same category and same place – or ‘home cell imputation’.

For example: if the BLS does not have the price of a loaf of 100 per cent whole wheat bread in a selected store in Washington, DC, then the agency estimates the price from the price change for all types of bread in the same area. And if prices for all types of bread in the area are unavailable, then the BLS would take the price from a broader region – or ‘different cell imputation’.

India has seen its own cases of imputation, but only in extenuating circumstances. One such instance was the Covid pandemic, when prices of several items were not available due to restrictions on movement of people and lack of transactions. As a result, MoSPI used imputation to calculate CPI inflation for April 2020 and May 2020. But from a policy perspective, the Reserve Bank of India’s Monetary Policy Committee treated these two inflation numbers as a “break” in the CPI series.

The problem for watchers of US CPI data is that the BLS has been increasingly using more and more imputation over the last one year – and of the worse variety. ‘Different cell imputation’ – in which price data from a wider region is used to guess missing numbers – made up 35 per cent of all imputed prices in June, up from just 8 per cent in June 2024. Nothing can make up for actual data collection.