Economic data house of mirrors
admin | February 15, 2019
This document is intended for institutional investors and is not subject to all of the independence and disclosure standards applicable to debt research reports prepared for retail investors.
The unprecedented 35-day federal government shutdown led to significant delays in the flow of key economic data, as two of the three main federal agencies responsible for economic reports were shuttered. The Census Bureau and Bureau of Economic Analysis, both part of the Commerce Department, were closed, while the Bureau of Labor Statistics, part of the Labor Department, had funding and remained on schedule. Any hopes of a quick return to normal have been dashed on two fronts. The data agencies have been relatively slow to get back on track, and the delayed data that have been released so far have been unusually erratic. If this pattern continues, the next few months could be a bumpy ride.
The federal government had never been shut down for as long as it was in December and January, which severely impacted the monthly data cycle. The impacted data can be categorized into several buckets. First, there was some data for the month of November that was scheduled to be released just after the shutdown began. Presumably, the survey responses were all in and these data were in the process of being compiled and tabulated when government agencies were closed. Completing these releases would be the easiest task, and they were, of course, the first to be released after the shutdown ended.
For the December data reports, the underlying activity that they measure might have been somewhat compromised by the shutdown, as the federal government was partially closed for the last 10 days of the year. It is conceivable that some economic transactions could have been affected, for example, federal contractors may have been unable to operate normally. Though aside from the Christmas retail season, the last 10 days of the year are usually pretty quiet. The main impact on the data would be that the normal survey procedures were not followed. For example, the retail sales survey returns for December may not have been processed until early February, several weeks later than usual. The Census Bureau’s release argued that survey response rates were, if anything, higher than normal and there were no clear anomalies in the process, but the strange results do introduce doubt.
The January data reports will be the most severely impacted. Economic activity taking place during the month could suffer both directly and indirectly: closed agencies may have prevented or disrupted transactions in certain sectors; and lower business and consumer sentiment may have delayed spending. In addition, surveys may not have been sent out on a timely basis, and responses will certainly be processed and compiled with a noticeable delay.
Data for February and March will also be affected due to the bottlenecks and delays that developed during the shutdown. The latest Census Bureau calendar shows release dates have yet to be determined, or “TBD”, for all February and March data releases. Evidently the Census Bureau does not expect to get back on schedule until the April data cycle, which begins with retail sales due out on May 15.
The unprecedented length of the government shutdown puts us in uncharted territory, but so far the pace at which the Census Bureau and BEA are catching up is disappointing. By the end of February, over a month after the shutdown ended, the data calendar is going to still be about a full month behind. The preliminary estimate of Q4 GDP is now due out February 28, the day that the first revision would normally have been released, while the December construction spending report is currently scheduled for March 4, a month and three days later than the original date. Government agencies may begin to make up time more aggressively in March, which will cover the January data cycle, but requiring over three months to catch up after a one month shutdown is not a particularly impressive performance.
The delays are bad enough, but the early returns on affected data releases suggest that the reliability of the data may also be compromised. There have been three major releases so far that were delayed. Each of them has been shockingly far from the consensus expectation as compiled by Bloomberg (see Exhibit 1).
Exhibit 1: Economic data releases post-shutdown
There is no particular reason that November data releases would be distorted by the shutdown. To be fair, both new home sales and the trade balance tend to be extremely volatile from one month to the next. Both new home sales and the trade balance were much stronger than anticipated, with new home sales exceeding 62 out of the 63 forecasts, while the trade deficit was well below even the most optimistic estimate.
More troublesome, the December retail sales figures were shockingly weak, with one of the core measures posting its worst monthly performance since September 2001, including every month during the throes of the financial crisis. In the fall of 2008, and in the aftermath of the 9/11 terrorist attacks, it was evident to everyone that consumers were retrenching – in the latter case, commercial flights were grounded and most households were literally scared to venture out of their homes for the better part of a week. During the recent holiday period, there were a multitude of indications that the 2018 Christmas retail season was one of the strongest in recent memory. Various proxies, including MasterCard data on spending, Johnson Redbook surveys, corporate earnings reports on fourth quarter sales, and the Fed’s Beige Book, were unequivocally strong. If the consumer were truly as weak as the December retail sales report suggested, there would be other indications, but the evidence is almost universally lined up at the polar opposite extreme.
As with the other two releases, it is far from shocking to see month-to-month volatility in the retail sales report. Even by those standards the December release was extreme, missing the weakest of 74 economist forecasts by a full percentage point. There has been widespread speculation that there is a problem with the data. Perhaps in the rush to get the report out under unusual circumstances, a mistake was made – perhaps a state or two got left out or the seasonal adjustments were applied incorrectly. It’s unlikely something like that happened, but no better explanation has so far emerged. Retail sales figures should get back to “normal” in the coming month or two.
If the retail sales shocker is indicative of a pattern, where shutdown-affected data will be unusually volatile and thus not to be relied upon, then markets are going to be flying blind for considerably longer than previously imagined. In which case it is probably a good that the FOMC has declared itself in a wait-and-see mode for the next few months, because policymakers may not be able to see clearly until well into the spring.