Jack Welch, former CEO of General Electric, set off a firestorm last Friday with a single tweet:
Unbelievable jobs numbers..these Chicago guys will do anything..can’t debate so change numbers
Then U. S. Rep. Allen West chimed in: “I agree with former GE CEO Jack Welch, Chicago style politics is at work here. Somehow by manipulation of data we are all of a sudden below 8 percent unemployment, a month from the Presidential election.”
This is crazy talk. Here’s why.
The Bureau of Labor Statistics (BLS) collects and processes the data that underlie the monthly jobs report (most recent release is here). BLS staff members are civil servants who serve without regard to party and are insulated from political manipulation.
This National Public Radio piece and this Washington Post article make clear how the BLS goes to great lengths to protect the data after it is collected and while it is processed. It’s critical that they do so because early access to the jobs numbers could, for example, affect financial markets.
Similar procedures are in place for data such as Gross Domestic Product, which is collected and analyzed by the Bureau of Economic Analysis.
As I discussed in an earlier post, the BLS collects its data through two surveys, one of households and one of businesses. The data rarely match up — the definitions of employment used in the two surveys differ — but they tend to show the same trends.
The September jobs report was pretty typical. Both surveys indicated private-sector job growth and public-sector job losses. The 114,000 jobs gained according to the business survey were in line with most economists’ expectations, especially given the second quarter’s anemic GDP growth rate of 1.3 percent.
So, what was Jack Welch’s problem? The controversy focused on the household survey: It showed employment increasing by 873,000 during September and the unemployment rate falling from 8.1 percent in August to 7.8 percent.
These are big changes, but they are not at all strange or inexplicable.
First, the increase in employment is probably driven by the fact the data are seasonally adjusted by the BLS. My colleague John Olson has a nice description of this process. We know, for instance, that high school and college students leave their jobs when school is approaching. Typically, they leave at the end of August and beginning of September, so the data are adjusted to take this into account.
However, it looks like this pattern is starting to change and students are leaving their jobs earlier due to more schools staring in late August or because of students’ desire to have more vacation time. It turns out that when this change is combined with seasonal adjustment, the August numbers will be under-reported and the September numbers will be over-reported.
This is just one example of a more general problem: Our seasonal adjustment procedures might be messing up the numbers. This seems likely given that employment was virtually unchanged in August and jumped in September.
Second, the unemployment rate is derived from survey data, which means that there is a margin of error. For example, the August unemployment rate was 8.1 percent, with a margin of error of plus or minus 0.2 percentage points. That is, the BLS economists are quite confident that the unemployment rate was between 7.9 and 8.3 percent.
Now, let’s look at the September rate. The 7.8 percent figure could represent a “true” rate as high as 8 percent or as low as 7.6 percent. The result is that the true rate might have been 7.9 percent in August and 7.8 percent in September, which isn’t much of a change. We just can’t tell given our survey data.
Welch came back in The Wall Street Journal on Wednesday with a column headlined “Why I Was Right About That Strange Jobs Report.” He makes himself out to be a victim, writing, “Imagine a country where challenging the ruling authorities — questioning, say, a piece of data released by central headquarters — would result in mobs of administration sympathizers claiming you should feel ‘embarrassed’ and labeling you a fool, or worse.”
Economists question data all the time. That’s why the BLS releases regular revisions to the figures; they collect additional information, perform more tests to make sure that their procedures are correct, and generally try to improve the numbers. We implicitly carry around a “handle with care” sign in our heads whenever we work with this stuff.
That’s not what Jack Welch did. He accused professional economists of either conspiring with or bowing to pressure from political operatives to manipulate fundamental data about the U.S. economy. He provided no evidence to back up his charge but simply smeared professional economists for no reason.
I’m sorry, Mr. Welch, it’s not a conspiracy. It’s just statistics.