The US Bureau of Labor Statistics (BLS) recently published data which made the present US economy look lackluster. The graph below shows the month-on-month (not the year-on-year!) change in the number of US jobs. The July data were quite low, while the May and June data were revised downwards. The total number of jobs is, however, still increasing.
Despite this increase, the new data, including the revisions, led to the replacement of the head of the BLS by the US president. Or did the president of the US sense another, larger and darker, danger, which led him to act? More on this at the end. Officially the US president in charge perceived these revisions as a personal attack. Is this president right, and can these data be perceived as an intentional attack on his person and track-record? This question has to be discussed. Currently, university training of economists pays little attention to the whys and hows of gathering economic statistics. This includes labor market statistics. This regrettable situation enables gaslighting when it comes to stating opinions about the purpose of published data. So, why and how are these data gathered? Spoiler: the president is wrong. But if I were head of the BLS, I would skip the first data release to diminish revisions and to provide the administrative state as well as politicians with more robust data. There is always a trade-off between timely and precise data. And it may be good to be fast and roughly right. But it´s not good to be fast and roughly wrong.
First, the ´why´ question. Why should a country spend loads of money on frequent and timely statistics on the labor market? The answer is political. Data are needed for economic policy. Labor market data might show (like in the USA) lackluster job growth or (like in Spain) a falling labour market participation rate. Generally, ´we´ want the opposite. Economic policies might help to get there. Or, to quote the BLS (this is about data on job openings):
Remarkably, job openings are stable and relatively high in the USA, while ´new jobs´ increase less fast. This suggests that the dynamics of job search in the US may have shifted. ICE, anybody? The answer to this last question is quite important for economic policy.
Second, ´what´?. What´s a job? To be able to count jobs, one has to define them. This definition is based on the concepts in the national accounts guidelines. The job concept of the national accounts relates to ´gainful employment´, i.e. roughly wage labor plus the self-employed. Non-monetary work, like much housework, is not part of the accounts and not counted as jobs. The new 2025 national account guidelines contain the next overview:
Based on these guidelines, organisations like the BLS develop definitions (to be found on their website), and surveys (to be found on their website) and reach out to 21.000 US businesses (every month) to obtain labor market information. And, maybe to the abhorrence of some, many national but also international organisations, like the invaluable International Labour Organization (ILO), are involved:
The results from the surveys are aggregated and published. All economic data gathered are prone to revisions, meaning that published data are frequently changed. These changes are based on standard procedure and made to improve the data, not to make a president look bad. The BLS is transparent about these changes; you can see them here. I´ve examined these changes briefly. Publishing the data one month later would lead to smaller revisions on average. By now, we would have the June data but not the provisional July data.
Summarising: the statisticians are right, within a margin of error. But the error could be smaller. Trump is wrong, without margin. In my opinion, the data do not exclude the possibility that ICE actions have led to a less well-functioning US labour market. At this moment this is just a hypothesis. But the possibility, unlike the revisions, poses a real threat to Donald Trump and his government. He might sense this. The BLS measures it. Hence, the replacement of the head of the BLS.
Personal point: the spelling corrector indicates that labour and labor and lacklustre and lackluster are all wrong. TLLM: too large language models.
