The monthly nonfarm payrolls report put out by the U.S. Bureau of Labor Statistics is perhaps the single most important report in determining the Federal Reserve’s outlook for monetary policy at the moment. The fact that it moves markets the most of any regular economic data release reflects this.
However, a new paper by Johns Hopkins economist Jonathan Wright has sparked a lot of chatter about how the seasonal adjustments used by the BLS in the report have become seriously distorted in the wake of the financial crisis of 2008, causing the agency to systematically overestimate job creation from October to April every year since then, while at the same time underestimating job creation over the May-September months.
That’s why, every year since 2008, the economy has appeared to stage a “recovery” in the winter months, only to weaken again over the summer.
Wright says the effect of these distortions from inaccurate seasonal adjustments has weakened each year since the financial crisis — in other words, the problem is slowly going away on its own.
Unfortunately, it didn’t go away fast enough to create another big problem this year.
In May — right at the end of that October-April period that gets an artificial boost from seasonal adjustment errors — the Fed signaled its intent to taper back its quantitative easing program later in the year, based on improvement in the data.
By September, Wall Street was convinced that this monumental shift in policy was imminent, and had priced such an outcome into financial markets.
Then, the Fed shocked markets by refraining from tapering, citing a lack of substantial improvement in the underlying economic data.
“We have never heard investors as angry as the day after the Fed decided to not taper its asset buying program,” says BofA Merrill Lynch economist Ethan Harris. “Some investors argue that it was a deliberate attempt to introduce volatility into the market. Others said Fed credibility is shot.”
BAML analysts reproduced the work of Jonathan Wright, attempting to answer the question of whether the Fed was in turn misled by the seasonal adjustment errors in the nonfarm payrolls report.
“Our analysis suggests that might be the case,” says BAML interest rate strategist Ruslan Bikbov.
In a note to clients, Bikbov writes (emphasis added):
The Chart of the Day shows that although the six-month moving average of BLS [seasonally-adjusted] employment increased by 85k from September 2012, when the Fed initiated QE3, to April 2013, the adjusted moving average increased by only 47k over the same time period.
In fact, when we repeated our analysis with the data available to the Fed by May of this year, we found even a smaller improvement in employment statistics adjusted for biases. The adjusted six-month moving average increased by only 35k from September 2012 to April 2013, while official BLS statistics showed 78k growth in the moving average over the same time period.
Clearly, the 35k improvement in monthly payroll growth from the start of QE3 to April 2013 barely qualifies for “substantial improvement” in the outlook for the labor market. Nevertheless, the Fed started talking publicly about tapering asset purchases in May. This suggests the Fed might have overreacted to biases in official BLS statistics. Similarly, the drop in the moving average of employment growth since April, which could be one reason for the decision not to taper in September, might be smaller than suggested by published BLS data.
Although our analysis applies only to employment, it is plausible to assume other [seasonally-adjusted] data could be biased as well. This also could have influenced Fed communication this year.
Bikbov warns that the seasonal adjustment errors could continue to cause problems for markets headed into the end of the year, causing another sell-off in the Treasury market.
“Our analysis suggests official payroll statistics may be biased higher in the coming months,” he writes. “Given that the Fed has deemphasized the importance of the unemployment rate, the market could be especially sensitive to payroll data.”