Federal Reserve officials have said countless times that they take a “data-driven approach” to policy decisions, including the current conundrum of when to cut interest rates. But what if the information is not as reliable as before?
That seems to be happening – and it’s making the job of central banks much more difficult.
“We have to make decisions in real time,” Fed Governor Christopher Waller said late last year. “Whatever information is released, I have to use it. The problem with data is that it is revised.”
This would not necessarily be a problem if the corrections, which could occur months after the initial reports were published, were relatively minor. However, many revisions have changed the game in the last few years.
For example, Waller noted that preliminary monthly employment numbers for 2021 led him to believe the job market is “good, but it’s not really great.” Even though inflation was at a 40-year high, he and other Fed officials were under the impression that they had to raise interest rates very carefully, fearing it could lead to job losses, Waller said.
But corrections have also come.
For example, in its preliminary jobs report for August 2021, the Bureau of Labor Statistics estimated that 235,000 new jobs were added.
As with any monthly job report, a job change should be reviewed three times. The first two are presented in the reports of the next two months. Those first two revisions to the August 2021 headline employment number put the number of jobs added at nearly 483,000, more than double the BLS’ first estimate.
The third and final revision comes in the calendar year after the original estimate is published, as the BLS reconciles the survey data it collects with actual data it later receives from state labor departments to arrive at the original estimate. This latest revision for August 2021 brought the monthly change in employment to 663,000, an increase of 428,000 from the first estimate.
Al Drago/Bloomberg/Getty Images
Federal Reserve Governor Christopher Waller said big changes in the data clouded his assessment of how the economy was faring.
That’s much higher than the adjusted month-over-month average change in the total number of jobs since 1973, the most recent revision data available, according to the BLS.
“When you look back, you say, ‘Oh my God, the labor market was a lot stronger then than the data releases show,'” Waller said in a lecture titled “Using Economic Data.” “Understanding the Economy” at the November conference.
In retrospect, it’s easy to say the central bank was too slow to raise interest rates, he said. “But in real time, maybe we’re not that late.”
Last year, the opposite effect showed itself. Revised employment gains for all but one month, July, indicated the labor market was weaker than initially reported. (Not all subsequent versions for November and December have been released).
With that in mind, Waller said last month, “there’s a good chance that December will be downgraded as well.”
“The frequency and volume of data revisions make the task of predicting how the economy will develop even more difficult,” Fed Governor Michelle Bowman said in November at an event hosted by the Ohio Bankers League. He highlighted recent changes to monthly job earnings as well as average hourly earnings.
The official summary of what Fed officials said and discussed during the September meeting – also known as the Fed minutes – noted: “Several participants observed that there are difficulties in assessing the state of the economy, as some data continue to be volatile and subject to large adjustments.” ”
Federal Reserve spokesmen declined to say which data Fed officials were referring to. It could have been any number of economic indicators reports.
For example, the Commerce Department initially estimated last April that the US economy would grow by 1.1% annually in the first quarter of 2023. That was well below the 2% rate that economists had predicted at the time. But the report may suggest that Fed officials, who want to curb inflation by raising interest rates to slow economic growth, are on the right track.
Then a month later, the Commerce Department revised its original estimate to 1.3%. A month later, in June, the Department released its next final revision. This time, he said, the economy is growing at the 2% annual rate that economists initially predicted.
It is perfectly normal for GDP data to change as it goes through a standard revision process to incorporate more comprehensive data that is available or revised. However, revisions to Q1 2023 GDP data were higher than usual.
Al Drago/Bloomberg/Getty Images
Governor Michelle Bowman said frequent and large revisions to economic data affect the Federal Reserve’s decision-making.
On average, the third revision is 0.6 percentage points higher or lower than the first estimate for a given quarter, according to a Commerce Department analysis of GDP data published from 1996 to 2022.
Consumer Price Index inflation data is also revised. However, unlike the other pieces of data, there is only a one-time revision that comes in February of the following calendar year as a result of updates to how seasonal adjustments are made.
Last year’s revision of CPI data from 2022 erased some of the progress thought to be made in bringing down inflation. Waller said last month he was concerned that CPI changes for 2023 later this month could also change the “inflation picture”.
“I hope the amendments confirm the progress we’ve made, but good policy is based on data, not hope,” he said.
For example, it is almost impossible to say with certainty how much prices are rising or how many people are being employed in an entire country’s economy at any given moment. Finding out how many new hires there are in a given month involves asking each employer how many people they have on their payroll. Therefore, government and other providers of economic data often rely on surveys to make complex estimates.
The BLS, the Census Bureau, and other government agencies that conduct surveys that inform economic reports work hard to make the best possible estimates with the data they collect. And more often than not, they do a great job.
But surveys are inherently imperfect.
Just as polls do not always predict the winning candidate, polls do not always reflect the exact truth. However, they can get pretty close to the truth.
Election surveys and government surveys have a sample size of respondents intended to be representative of the total group being studied. The larger and more diverse the sample, the closer the estimate will be to the true value.
Getting a large and diverse sample requires a lot of outreach to recruit people to be part of the group that the BLS and other agencies regularly recruit to answer a particular survey. For surveys used in many of the BLS’s monthly reports, including Employment, the Consumer Price Index, and Job Openings and Labor Turnover, the hiring rate fell sharply before the pandemic.
That raises the possibility that the survey sample could be more biased, said Erica Groshen, former commissioner of the BLS. The agency, he said, “tests for all the biases they know about, but there are more biases that may cause fewer people to respond.”
“You have to worry primarily that the employers who never agreed to participate in the survey differed to some degree from those who did,” said David Wilcox, a longtime Fed employee who is now an economist. Peterson Institute for International Economics and Bloomberg Economics.
For example, it may be that firms that agree to participate in the BLS survey are more financially stable than those that decline, he said.
“It shows that there is a potential for us to misjudge what’s going on in the economy,” he said, adding that this would apply mainly to preliminary estimates rather than final ones.
Laura Kelter, director of national estimates in the Current Employment Statistics division at the BLS, told CNN that declines in hiring rates, which help explain the drop in response rates, do not lead to larger changes between the initial and final employment estimates. “However, we continue to monitor the issue,” he said.
Kelter said the declines can be attributed to various challenges, such as the voluntary nature of survey participation, as well as survey fatigue, which is when people are bombarded with too many surveys.
Groshen agreed, adding that less civic responsibility, changes in technology like caller ID and spam filtering, and growing concerns about privacy and data security are also at play.
“Even in normal times, the Fed’s ability to perceive what’s going on in the economy is like an individual with vision out of both eyes, but by no means is that vision 20/20,” Wilcox said.
New York Fed President John Williams told CNN last year that he accepts the initial government data he gets at monetary policy meetings will change when officials meet again.
It therefore takes into account as much external data as possible, often from private sector data providers, to give it a better understanding of what is happening in the economy.
“We want to be data dependent, but not data point dependent,” Williams said. “I try not to get too excited about every data point we see.”