Calling B.S. On the BLS & Incumbency Isn’t What It Used to Be – Top 3 Takeaways, August 5th, 2025
Takeaway #1: The year was...
Hold on by Wilson Phillips was the top song, Milli Vanilli was exposed as a fraud. The Hubble Space Telescope was launched; Elon Musk was 19. The World Wide Web was created. The Berlin Wall came down. Operation Desert Shield took place. You know what else took place in 1990 that’s particularly relevant to current news? It’s the last time that the Bureau of Labor Statistics meaningfully changed the way they collect information for their monthly jobs report and overall, the general methodology has been essentially the same since 1984. That’s right the Bureau of Labor Statistics (BLS) has been using call-out data as part of its data collection methods for the Current Employment Statistics survey, which produces the monthly jobs report, since 1984 with the introduction of computer-assisted telephone interviewing in 1990, which involves calling businesses to collect employment, hours, and earnings data. Now, if you were trying to do the job you do today with the methods and technology of 1990, how effective would you be? Btw, I did a little research on how many jobs are effectively the same today as in 1990 and the high-end estimate was 3%. Fun fact(s): Do you know what two professions are the least likely to be impacted by technology over the past 35 years? Local live music and janitorial work. Speaking of flushing... That’s the reason for my top takeaway today. The recent decision by the Trump administration to flush the head statistician at the Bureau of Labor Statistics. As was noted by team Trump, when accounting for revisions, the average error rate in the BLS’ initial reporting has been 40%. As I was asking yesterday – if you were wrong or failed at 40% of your work how long would you remain employed? But because it’s the GSS news media and it’s President Trump, headlines like this from CNBC persist: Trump stokes conspiracies about jobs data, as White House defends firing BLS chief. So...
Takeaway #2: Is the idea that a 40% failure rate is supposed to be an acceptable thing?
The reality is this is nothing new because the BLS hasn’t been doing anything new. As I addressed in a Q&A from over three years ago in addressing significant differences in reporting between the monthly ADP Report and the BLS’ jobs report: The BLS figures include government jobs that aren't factored into the ADP Report. If all other factors were equal there’d always be a natural variance to a certain extent due to the government report factoring in government job changes not factored into ADP’s reporting. Public sector employment accounts for approximately 15% of all US employment and 13.5% of overall employment in Florida specifically. Therefore, variances in the neighborhood of 15% or so between the two reports could be explainable without additional factors coming into play but as for the methodology... ADP processes the HR and payroll data for over 14% of Americans in the workforce. They’re far and away the leader in payroll processing which is why they produce monthly reporting that’s used by economists to gauge real-time conditions in employment. The starting point for the ADP Report is actual data covering a 6th of the entire US employment pool. Moody's then uses the data, taking into account sector and regional variances, to project the overall job changes monthly. The BLS uses call out surveys that's then used to project the result. It’s for this reason that the initial reporting by the BLS has commonly proved unreliable independent of other factors. Rather than starting with actual data, they’re using estimates provided by HR professionals about changes within their organization. If the HR professionals provide incorrect estimates to the government – you get the garbage in-garbage out effect. The BLS’s methodology is analogous to political polling. It’s a sample, and done well can be highly accurate, however there’s inherently a margin of error, prior to projections, that’s not present within the ADP Report. And that in a nutshell is what goes on here. The federal government has still been doing things as though it was 1990. The biggest issue with the previous BLS chief was a lack of modernizing the data collected and projected by the agency. That’s not a conspiracy but it is typical of the way government often does things.
Takeaway #3: Incumbency isn’t What it Used to Be
One thing is certain with the current mess in Texas. What does or doesn’t happen there won’t being staying there. That’s true of the Democrats who fled to the sanctuary city of Chicago to attempt to prevent a special session to take up redistricting. That’s true if Governor Greg Abbott and Republicans in Texas are successful in redistricting that could result in potentially five more GOP seats. But while we wait and watch what will or won’t be which with the prospect of the governor potentially removing 51 elected democrats from office and perhaps arresting them too – which would be quite the thing. Anyway, while digging into the implications of Texas’ redistricting thing – which does hold the potential to impact which party has control of the House in the next Congress, I found this. Incumbency isn’t what it used to be. On average in the 2024 election cycle how big of a benefit do you think incumbency proved to be? The answer is...only 1.1%. The lowest in U.S. history. Earlier in this century that advantage peaked at 7.7% and it’s been on the descent ever sense. What this means is that even a little redistricting could net some big changes. That’s true in Texas and in Florida if we happen to go there too as Governor DeSantis seemingly wants to do. Stay tuned...