There are no guarantees when predicting Federal Reserve monetary policy decisions. But yesterday’s surprisingly strong increase in US private payrolls in February by ADP’s reckoning gives the hawks a potent new data point to make the case for squeezing monetary policy at next week’s FOMC meeting.
A quick review of Wednesday’s news: the crowd was looking for a gain of 183,000, based on Econoday.com’s consensus forecast. The actual increase was sharply higher at a sizzling 298,000. One month is hardly definitive proof of anything in economics, but as I noted yesterday the year-over-year change for payrolls has been ticking higher lately and February’s ADP data feeds into that trend.
The bigger issue is deciding if the latest ADP figures will translate into a similarly bullish picture in tomorrow’s official payrolls report for February? Econoday.com’s consensus forecast sees the Labor Dept.’s estimate of private employment rising 190,000 – a respectable gain, but moderately below January’s 237,000 advance. But in the wake of the red-hot increase via the ADP report, there’s a case for revising tomorrow’s forecast up, perhaps by a lot.
Regressing the monthly changes on the ADP and Labor Dept. data sets tells the story. Based on the relationship since 2001 (the start for the ADP figures), yesterday’s release implies that the government tomorrow will report a gain of 297,000 for private US payrolls in February – more than 100,000 above the consensus forecast.
Regression analysis isn’t fate, of course, in part because any one point forecast has a relatively wide confidence interval. At a 95% confidence level, the forecast variation that applies to the 297,000 point estimate for tomorrow’s report ranges from roughly 280,000 up to 313,000. At a 99% confidence level, the range expands to a 275,000-to-318,000 band. But that’s still encouraging and so the regression analysis implies that the worst-case scenario tomorrow will still deliver a solid improvement over January’s employment gain.
The question is whether the ADP is a reliable predictor for the Labor Dept. data that typically arrives a few days later? Let’s dig a bit deeper into the numbers for additional perspective. For instance, the next chart below compares the differences between the two data sets through time. Both estimates are attempting to estimate changes in private employment. If there was no difference in the two time series, the line in the graph below would consistently be zero, indicating no variance between the two estimates of monthly changes in payrolls. In reality, there are differences from month to month. Note, however, that the differences have narrowed in recent years. The ADP data tends to exceed the Labor Dept. numbers, but the spread has narrowed over the last five years or so vs. the previous period. Perhaps that’s a sign that the ADP estimates are becoming more accurate in terms of anticipating the government’s report each month.
For another perspective, here’s a chart of the rolling 12-month correlation between the monthly percentage changes for ADP and Labor Dept. figures. Here too there are clues for thinking that the differences between the two data sets has tightened in recent years, which suggests that the forecasting accuracy of the ADP estimates has improved. As the correlation reading approaches 1.0, the spread between the two time series narrows. With that in mind, the 0.86 correlation for January — the highest in four years — suggests that ADP’s predictive ability is higher these days.
None of this insures that tomorrow’s report will be stellar. Any one number can deliver big surprises, for good or ill, no matter how many econometric tests you run. But from a probability perspective, the analysis above speaks loud and clear and so it appears that the government is set to report that US companies hired more workers in February vs. January. In turn, the 89% probability of a rate hike next week (based on Fed funds futures as of Mar. 8 via CME data) is on track to tick even higher once Friday’s release hits the streets.