The Fed’s data snafus
Most everyone knows that the Federal Reserve Board is responsible for making monetary policy, handling prudential oversight of many of the nation’s banks and keeping the clearing and payment system flowing. But the Fed has another fundamental function that often goes unnoticed: collecting financial and economic data.
Good policymaking flows from having fresh and accurate data. From my little experience with the Fed they are not doing very well at this task.
Reuters is reporting that Fed Governor Elizabeth Duke believes that household debt has declined since the financial crisis of 2008 and that this reduction in household balance sheets will position families to participate in the recovery when conditions tick up. From Reuters:
Households’ caution about taking on debt and spending will stand them in good stead when the economic recovery becomes more robust, a top Federal Reserve official said on Saturday.
Household debt-to-income ratios skyrocketed during 2001-2007, but households cut debt and spending significantly during the financial crisis that began in 2007, [Fed Governor Elizabeth] Duke said.
Governor Duke is making some large assertions about household behavior. It’s hard to confirm her viewpoint because there has been a lot of confusion surrounding the data sets that the Fed is using to model household balance sheets.
The Federal Reserve publishes two major public datasets for household finances: the Survey of Consumer Finance (SCF) and the more recent “Consumer Credit Panel” (CCP). The SCF is conducted as a survey every three years by interviewing approximately 4,400 households who self report their debt and assets. The results of this data are then extrapolated to the nation’s 130 million households. In contrast the CCP is constructed by compiling a dataset of approximately 38 million credit reports and projecting those findings onto the general population. The CCP began publishing data in 2010 and updates its numbers every quarter.
My fellow Reuters blogger Felix Salmon has been wrestling with the student loan numbers published by the Federal Reserve and after some digging found them to be grossly underreported. This was data from the credit bureau reporting that the Fed extrapolates to the bigger population (CCP). From Felix’s blog (emphasis mine):
When it became obvious that the student-loan totals were too low, the NY Fed and
ExperianEquifax started looking at the data again. And eventually the culprit was found. There was a bucket of random obligations called “Miscellaneous”, which included things like utility bills, child support, and alimony. And it turns out that if you went burrowing in that miscellaneous debt, there was actually a pile of weirdly-categorized student loans in there.
When the NY Fed restates the Q2 figures, and from Q3 onwards, that pile of student loans will get included in the bigger student-loan figure, where it belongs. And the number will rise, probably by a couple of hundred billion dollars. Not enough to bring it to $1 trillion, but enough to make it bigger than total credit-card debt.
So the numbers the Fed had been using for student loans were undercounted by several hundred billion dollars. This mirrors a recent correction that the Fed made to credit card debt in the Survey of Consumer Finance. Earlier in the week the Federal Reserve Bank of New York published a paper describing how they uncovered the underreporting of credit card debt by households.
The paper, “Do We Know What We Owe? A Comparison of Borrower- and Lender-Reported Consumer Debt,” points out that credit card debt held by households is underreported by half.
Binyamin Appelbaum, a reporter for the New York Times, wrote a blog post summarizing the NY Fed paper. It claims that although credit card debt is alledgedly underreported, households are accurately reporting mortgage, auto and school loans when cross-referenced against private datasets. Appelbaum writes about the possible causes of under reporting that the research staff proposed staff: shame, lack of communication between spouses, and even ignorance.
The research staff of the NY Fed could be right in blaming consumers but given recent experience with data there may be flaws in their survey method, or maybe there are errors on both sides of the study from those who conduct it and those who participate in it.
There is another recent example of the Fed mistating the numbers. In May my Thomson Reuters colleague, Daniel Berger, began an investigation into the size of the municipal bond market, which the Federal Reserve estimated was $2.925 trillion. Daniel used a series of privately maintained databases to establish that the true size of the municipal bond market is closer to $3.7 trillion. The Fed concurred with this analysis and now Reuters and other media outlets use the revised number to state the size of the market.
I wrote in June this was a big miss for the Federal Reserve and raised the issue of the reliability of other data sets that the Federal Reserve had collected, especially the “Survey of Consumer Finance.” Including the municipal market restatement there have been three documented cases of errors in the Fed’s data that number in the hundreds of billions of dollars. These errors diminish the credibility of the Fed and call into question the data used in policymaking.
It’s important that these public data sets be as widely vetted and open to scrutiny as possible. Many U.S. households are ailing financially and this data is like taking their temperature. We need the best thermometer we can get, I don’t think the Fed’s data could currently claim that title.
Federal Reserve Bank of New York: Consumer Credit Panel data for Q2 2011
Chart source: Federal Reserve Bank of New York: Have Consumers Been Deleveraging?