WASHINGTON, March 6 (Reuters) - Bolstering government economic reports with private data could improve Federal Reserve policymaking by helping officials better anticipate changes in jobs and inflation, a team of top economists has concluded.
Following a year in which official data was interrupted by the record-long federal government shutdown and the head of the Bureau of Labor Statistics was fired by President Donald Trump after a run of historically large revisions to data, the new research concluded
the Fed would likely have cut interest rates sooner when initial BLS reports on job growth showed continued strength while private data from sources like payroll processor ADP had begun to flag weakness.
"Signals from...private data could potentially reduce the noise that policymakers face," compared to relying on government reports alone, the researchers found after studying how employment records from ADP, 401(k) account information from investment giant Vanguard, and checking account data from JPMorgan could complement the monthly government employment and inflation reports.
They found that combining public and private data produced better forecasts of the change in private nonfarm payrolls reported by BLS each month as well as the final revisions made to that number two months later, and also improved a six-month-ahead forecast of job growth. Since the anonymized data also included information on income and wage growth, it helped anticipate consumer inflation reports.
The research team included top economists from the three organizations that contributed data, along with University of California at Berkeley professor Yuriy Gorodnichenko and University of Chicago Booth School of Business professor Anil Kashyap. The findings were presented at a Booth monetary policy conference in New York.
It's common for Fed officials to use private data in assessing the economy, with reports on credit card spending a staple in analysis of consumer trends, and bank and association data on mortgages and real estate transactions also central data points. But the pandemic, with its historic economic swings and a decline in the response rates to government surveys, stoked interest in alternative and real-time methods to understand hiring and inflation, the two issues at the center of the Fed's policy debate.
Large revisions in jobs data last year, Trump's subsequent firing of the BLS commissioner, and the lack of any reporting during the government shutdown added fuel to the discussion.
One challenge, however, has been winnowing through the massive amount of raw information available in a world of cellphone tracking and ubiquitous barcode scanners for data series that add value to forecasts, have a long enough track record to be studied over time, and can be benchmarked as revisions to government data home in on final results. In the case of jobs data that "real" number takes over a year to emerge, with initial government surveys revised twice as responses trickle in from businesses, then again as the survey-based series is benchmarked against a more universal quarterly census of businesses that almost all firms complete.
The Fed, however, can't wait a year, and has to watch the economy in something close to real time so it can respond as problems develop.
The research team found that an index created using private data improved forecasts, and highlighted the Fed's policy debate last year as a moment when ADP data in particular indicated "that the labor market was losing steam" beginning that spring, even as the initial reports from BLS in May and June showed strong job gains. Those were eventually washed away in subsequent revisions, prompting the Fed to move ahead with rate cuts in September and Trump to fire the head of the BLS.
Federal Reserve Governor Christopher Waller, citing ADP data, had warned about the likely revisions at the time, urging his colleagues to move ahead with rate cuts in the summer because of the labor market weakness he thought was developing, and dissenting at the July meeting when the Federal Open Market Committee voted to hold rates steady.
The researchers were careful to note private data could only serve to bolster official reports, not replace them. But they also said that their hybrid approach seemed to work because private data may add information on smaller businesses and lower-income households that is not as well captured in government surveys despite the effort put into developing representative samples, proper weighting, and understanding the birth and death of small firms over time.
"Private sector data contribute a signal precisely in the areas of the economy, such as smaller firms and lower-income workers, where government surveys - constrained by sample sizes and response rates - may be weaker," the group wrote, suggesting that even better prediction could come with the use of an even broader set of private data sources than those used in the study.
(Reporting by Howard Schneider; Editing by Dan Burns and Andrea Ricci)
