Bad jobs report, good wage report?

4/03/2015 03:27:00 PM
Tweetable
The BLS jobs report comes out on the first Friday of every month, and today's was a disappointment, showing a lower than expected increase of 128,000 jobs in March. In addition, January and February figures were revised down, making for a disappointing first quarter to 2015. First quarter was also staggeringly weak in 2014--GDP actually shrank, making it the only decline in GDP on record that didn't turn into a recession.

On the other hand, wages rose. As Danny Vinik notes:
"What’s even more interesting in this report is that wage growth actually beat expectations, rising 0.3 percent in March versus the expected 0.2 percent. In the past year, wages have grown 2.0 percent. That’s only a slight beat and can be attributed to the noisy report. Yet it’s still tough to square a slight increase in wages with the ugly jobs numbers. It’s a complete reversal of the narrative that we’ve seen over the past year where the jobs numbers repeatedly came in above 200,000 yet wage growth was muted. Simply put, it doesn’t make any sense!"
According to Vinik, wage growth contradicts the bad jobs growth.

Theory predicts that wages growth should rise as labor markets tighten, and decline in weak labor markets, and I don't dispute that. However, I'm on record criticizing people who made a big deal out of wage growth in the past monthly reports because, in my view, wage growth seems to be too persistent for that kind of analysis--any month-over-month change in wages is far more likely to be noise than signal.

So, here's a few simplistic regressions to help think about the correlations:
This the unemployment rate, the percent change in employment, and the change in the unemployment rate, regressed on wages. The p-values give a sense of the strengths of the correlations between these variables--contemporaneous wage and employment numbers are hardly correlated at all. In fact, wages and employment actually have a slightly negative correlation of -0.04, which I'd take to say that wage growth is about as likely to rise as to fall in the same month as a strong jobs report. If you're concerned about multicolinearity, here's the regression with just wages and employment:
In fact, the relationship looks even weaker, if that's possible. (Note: I did the regressions with both the absolute and percent change in employment just to be sure--they are essentially the same results)

The unemployment rate, on the other hand, is actually predictive of the changes in wages.
So we would expect to start seeing more wage growth in low unemployment environments. But my goodness that is a low R-squared. What this says is that low unemployment months have higher wage growth on average, but in any particular month we are still only slightly more likely to see the two go in the same direction as opposite directions.

And just to complete the thought, here's what we get if we look at the change in unemployment instead of the level:
Same thing, just weaker correlation than with the levels.

I'm not saying that strong labor markets don't lead to wage increases. But clearly we won't be able to see this relationship in real-time monthly datapoints. All of that commentary is just tilting at statistical noise, not actual economic indicators, in my opinion.