No, they do not sit around changing the numbers to serve the interests of Democratic presidents, or to harm Republican ones. The system has too many different steps, too many checks and balances, and too many people who do not want to do the wrong thing. In a sense, you could say that the BLS is too bureaucratic to do that. They are better thought of as an agency which maximizes process, and the successful execution of process, success being defined in heavily process-intensive terms.
Their ideology, if that is even the right word, is to maximize adherence to the process. And “defensibility of the estimate” is important there.
You might argue they are not very good at seeing “the big picture,” but that same emphasis makes it difficult for them to deviate much from established procedures.
If there were important reasons why we should be creating new, useful, but highly speculative estimates (how about “the number of jobs that were not created because of AI”?), the BLS would not be good at doing that. They would not do it at all. Such estimates would open them up to too much criticism, and the speculative nature of the enterprise would clash with their desire to be managing controllable and defensible processes.
Over the last twenty years, a lot of their innovations have come in the form of disaggregated, sector-specific or region-specific data, which is fine. Or more emphasis on “work from home” issues. Which is fine.
So they estimate “that which they can,” rather than producing unreliable estimates that might be highly interesting.
That is the sense in which the BLS — and many other parts of the government in fact — is biased. It can matter, but it is a mistake to be looking for partisan bias that skews the numbers.
The post In which ways is the BLS biased? appeared first on Marginal REVOLUTION.
