The monthly nonfarm payrolls (“jobs”) report and the “advance” GDP reading put out by the U.S. Bureau of Economic Analysis every quarter are the two biggest market-moving economic data releases for both stocks and bonds.
Less widely-followed releases, like the monthly reports on regional manufacturing activity put out by the Federal Reserve Banks of Philadelphia and Chicago, have less of an impact on the market.
Ironically, the jobs and GDP releases contain the least amount of information about the course of future economic growth, whereas the regional data are much better predictors of such activity.
These are the findings of a new report by Goldman Sachs chief economist Jan Hatzius.
“Our results suggest that these investors will put too much weight on indicators that tend to get heavily revised and are not very informative in real time,” says Hatzius. “Moreover, many investors seem to look for a simple, high-profile summary of the performance of the economy, rather than piecing together a composite picture from many different reports. And the highest-profile indicators are the monthly employment report and the quarterly GDP report.”
The charts below show the impact on markets of various economic indicators. The nonfarm payrolls and advance GDP releases are clearly the biggest market movers, while the Chicago and Philly Fed reports are among the least likely to move markets.
The next chart shows each data release’s impact on Goldman’s proprietary “current activity indicator,” one of the firm’s favourites for measuring economic growth.
“Our most important finding is that the impact of both nonfarm payrolls and advance GDP is small and statistically insignificant,” says Hatzius. “In contrast, the Philly Fed index, the Chicago PMI, and initial jobless claims contain a statistically significant and economically meaningful amount of information for growth. For example, an increase in the first-release Philly Fed index equivalent to a [one standard deviation] surprise (about 8 points) is associated with an average growth pace of our CAI over a 6-month period that is almost 0.3 [percentage points] higher.”
So, if the nonfarm payrolls and advance GDP releases are the biggest market movers, yet contain the least relevant information for the future course of economic growth, then where’s the disconnect?
Hatzius identifies two reasons:
1. First-release data can look very different from fully revised data.
It is difficult to overemphasize the importance of using first-release data for the explanatory variables in our analysis … payrolls and GDP are subject to heavy revisions between the first release and the fully revised version. Some of these revisions occur in the next monthly or quarterly release, and some occur with much longer lags via the annual revisions. And they are very important; if we re-run our regressions using fully revised data, both payrolls and GDP become statistically significant as predictors of future growth. But revised data are obviously not available in real time.
2. Timelier indicators are much more valuable than less timely ones.
Nonfarm payrolls are always released three weeks after the survey week, which typically means the first Friday of the month. In contrast, jobless claims are released just five days after the week to which they refer and the Philly Fed is released near the middle of the month. This issue is very significant in the context of our analysis. We can see this by re-running the regressions reported in Exhibits 4 and 5 without controlling for timelier economic indicators for the current month, e.g. jobless claims and the Philly Fed. In this version of our analysis, payrolls now become highly statistically significant. (Advance GDP is only marginally significant even if we drop all of the timelier indicators.)
There is, of course, one other major factor tying the nonfarm payrolls report to big moves in the market: the Federal Reserve’s decision in December with the introduction of the “Evans Rule” to condition changes in monetary policy largely on developments in the labour market.
“Given the limited usefulness of first-release nonfarm payroll numbers, we believe that the distinction between an initial print of 162,000 and one of 180,000-200,000 is too small to determine whether Fed officials should make a significant policy change,” says Hatzius. “But if the markets have some reason to believe that this distinction will in fact drive Fed policy, the outsized response to relatively small and economically meaningless payroll surprises may not be so irrational.”