December 06, 2014
Okay boys and girls, today we learn about the erratic pattern of wage data. Ideally the Bureau of Labor Statistics (BLS) would tell us exactly how much hourly wages rose each month. Unfortunately, BLS doesn’t have that ability. It has a very good survey of establishments that gives a reasonably close estimates of current hourly and weekly wages, but these numbers are not exact. And, since each month’s wage estimate includes a component of error, the changes from one month can contain a very large component of error.
To see the logic, imagine that the 95% confidence interval is +/- 0.1 percent. (I haven’t checked this, but 0.1 percent would be pretty good.) Suppose that one month it underestimates the average wage by 0.1 percent. Suppose the next month it overestimates the average wage by 0.1 percent. This would lead to a wage growth number from one month to the next that was 0.2 percentage points above the true number. In a context where monthly wage growth has been averaging less than 0.2 percent, this would be a very large error. That is why it is always advisable to take a longer period than a single month to assess wage growth. (My preferred measure is taking the rate of change for the most recent three months compared with the prior three months.)
Many foolish comments about the November employment report could have been avoided if reporters recognized the erratic nature of the monthly data. The 9 cent gain (0.4 percent) reported in the average hourly wage for November was widely touted. Unfortunately, reporters did not bother to note that BLS reported a gain of just 0.1 percent in October and 0.0 percent in September. As a result of the weak wage growth the prior two months, the average wage for these three months grew at just a 1.8 percent annual rate compared with the average of the prior three months. That is somewhat below the 2.1 percent increase over the last year.
When we look at these numbers we have two choices. One is to take the monthly data at face value, as almost all the reports on the November report did, and believe that wage growth virtually stopped in September and October and then surged in November. Alternatively, we can believe that the slowdown in September and October and the surge in November were both driven by measurement error.
Before deciding which route you prefer, let’s look at some additional data points. First, this is not the first time in the recovery that we have seen wages jump by 0.4 percent in a month. Wages also rose by 0.4 percent in January, July, and October of 2011 and June of 2013. Here is the sequence of wages in each of those cases:
January 2011: 0.1, 0.4, 0.0
July 2011: 0.1, 0.4., -0.1
October 2011: 0.2, 0.4, -0.1
June 2013: 0.1, 0.4, 0.0
Going further back before the recession, we can see that in April of 2006 the average hourly wage rose by 0.6 percent. Unfortunately it then fell by 0.1 percent in May.
Still undecided about whether we should think that wages surged in November? Let’s look at the some of the industries that were driving wage growth in the month.
There was an impressive 0.6 percent increase in the average hourly wage in the wholesale trade sector. Of course this followed a month of 0.0 percent wage growth in October and a drop of 0.4 percent in September. The average hourly wage in the information sector increased by 1.1 percent in November. This followed a decline of 0.6 percent in October and a rise of 0.1 percent in September. Wages in finance rose by 0.8 percent in November, following a drop of 0.1 percent in October and a rise of 0.3 percent in September.
If we want a bit more insight into the path of wage growth we can look at wages for production and non-supervisory workers, a group that includes more than 80 percent of employees. There is not much of wage surge in these data. The average hourly wage for these workers increased by 4 cents (0.2 percent) in November. That follows a gain of 0.1 percent in October, and 0.0 percent in September.
So, there you have it. We can think that after two months of very weak growth wages suddenly surged in November, and only for the minority of workers in supervisory positions, or we can think that we are looking at measurement error. The call is yours.
Comments