Last Friday’s relatively positive national employment report from the U.S. Bureau of Labor Statistics (BLS) elicited cries of foul play from the far right – started by none other that Jack Welch, the former head of GE who tweeted that the books were being cooked to help President Obama following his disastrous debate with Mitt Romney. Numerous other right wing pundits quickly jumped on board and started building on the story.
The idea that there is a left-wing conspiracy behind the numbers is false—as the large majority of reasonable folks on both sides of the aisle quickly acknowledged. But this hasn’t deterred Mr. Welch, who decided to double down on his claims that someone is cooking the books with an editorial in last week’s Wall Street Journal titled, of all things, “I Was Right About That Strange Jobs Report.”
No, Mr. Welch, you were not.
Does the economy appear to be growing at a pace that would bring the unemployment rate down by half a percentage point in two months? No, of course not. The U.S. economy has experienced slower than normal growth in both economic output and overall employment. That does not tend to lead to dramatic drops in the unemployment rate.
But if you understand the BLS data, you realize that this is not a relevant question for two months of household numbers. The U.S. unemployment rate is not measured well (as Mr. Welch himself makes abundantly clear). It is based on a tiny sample of 60,000 out of 110 million households. And yes, many of the survey questions can be a bit confusing, and on occasion wrong answers are given.
What this means is that the month-to-month fluctuations in the unemployment rate are what we in economics call ‘noisy’—sharp movements that happen in the short run and are not based on any true underlying economic force, positive or negative. It is simply randomness. To accurately assess the unemployment rate, you have to consider what the underlying mid-term trend suggests.
This graphs lays out that trend. Over the past two years the U.S. unemployment rate has been falling modestly—at slightly less than one percent per year. Sometimes the rate has fallen a bit faster—such as in late 2010/early 2011, and in the past two months. Sometimes it has flattened out or even risen a tiny bit—as happened in May though July of this year, and March through June of 2011.
So the relevant question to ask is does the current pace of economic growth suggest that the unemployment rate has dropped by slightly less than 2 percentage points over the last two years? And the answer is yes, that makes perfect sense.
And was the sharp drop in unemployment over two months that surprising anyway? No. After all, unemployment had risen over the three months that preceded the drop despite the fact that the U.S. economy was adding jobs and growing. Another short term data oddity that eventually works its way out of the system. We were due for a little revision to the downward trend.
But what is either an example of extreme cynicism or poorly developed critical thinking skills is that Mr. Welch writes: “Bottom line: To suggest that the input to the BLS data-collection system is precise and bias-free is—well, let's just say, overstated.” As stated, I agree about the precise part, no one ever said it was precise. But not being precise is completely different than being biased and to put the two in the same sentence is an absolute non-sequitor. Being biased presumes the results are being pushed in one direction or another either because of problems with the design of the measure, or on purpose by the investigator. ‘Noise’ does not equal bias by any means.
On the other hand, Mr. Welch’s frantic comments about the two-month decline are about as perfect an example of bias as one will ever see. Why doesn’t Mr. Welch comment on the odd increase in the unemployment rate over the 3 months prior to the drop? Is he only able to believe economic estimates that favor his candidate’s chance of winning the White House?
Ironically, for all his complaining about the noise in the nation’s employment statistics, Mr. Welch never mentions that it is his own Republican Party that has consistently pushed to reduce the budgets of the U.S. Census and BLS. They recently took aim at the Census’s American Community Survey—a hugely useful annual data source that has been a boon to social researchers not to mention businesses looking for better market data. I guess its better to have no information at all than better information with less noise?
What’s most humorous is that if any group should have a reason to quibble with the BLS, it’s the Democrats. Why? Because it turns out that the BLS has actually been undercounting new payroll jobs—to the order of 385,000 jobs overall and 450,000 private sector jobs as of the first quarter of the year. That’s about .4% of all payroll jobs in the nation, and a number that only applies to the first quarter of 2012. This takes the slow pace of job growth, and makes it look… not too bad. And definitely something that would tend to bring that unemployment rate down.
The BLS has reported the undercounting on their website. See: http://www.bls.gov/ces/cesprelbmk.htm.
But before you go writing editorials to the Wall Street Journal about how it’s the Republican controlled House that has infiltrated the BLS to keep those payroll numbers down to help win the White House, understand that this undercounting is yet another example of noise.
The BLS ties its Current Employment Statistics (CES) payroll employment estimates to the unemployment insurance data that is roughly lagged by 9 months. But the system has some known issues because of something called the birth-death model. Without getting technical, the CES tends to overestimate job growth when the economy is starting to slow, and underestimate it when the economy is starting to improve. The latter is likely the reason why jobs have been under-counted recently. When the BLS revises their 2012 estimates of the number of payroll jobs in the country, these issues will be ironed out of the data. Another example of why it is critical to look at longer term trends to get an accurate read of employment data.
And to quote my hero Bill Murray – that’s a fact, Jack.