Maybe "economic policy uncertainty" is just firms disliking regulation
Why did economic policy uncertainty rise strongly in 2016, but the stock market is doing well?
But that result [from the model in their earlier paper which says that uncertainty is bad for stocks] assumes that the precision of political signals is constant over time. In contrast, we argue here that political signals have become less precise in recent months, especially after the November 2016 election.
They state that Trump on one day says this and on another day says that. And that therefore firms get more noisy signals about the future course of economic policy and that the lack of a signal isn’t the same as uncertainty over outcomes. This even got coverage in the respected German daily FAZ (in German).
I’m not completely convinced. Maybe much of what we refer to as “economic policy uncertainty” is just firms being annoyed at regulation. Regulation, justified or not, is likely not great for corporate profits. Baker, Bloom and Davis (2016) think (especially of their industry-specific) indices as measuring “regulatory policy uncertainty” (p.1621). But what if it’s more a proxy for “regulatory policy”?
It’s like when people say “risk has gone up”, they often only refer to downside risk. With Trump, I think, actual “uncertainty” (or what Pastor and Veronesi call “the precision of political signals”) is up, but the expected value of how much regulation there will be is far down. So expected profits rise and thus stocks benefit. But at the same time the newspapers are full of the words “uncertainty”, because there really is uncertainty about the future course of regulatory policy.
If you look at the time series of the EPU index, the fact that it jumps up around wars and elections I find convincing that it measures a significant amount of “uncertainty”. My conclusion is that it’s a mixture of both the subjective expected future level of regulation and the uncertainty around it.
“Uncertainty” may also have become a fashionable buzz word in the last couple of years and this would mechanically push the Baker et al. indicator up. They don’t correct for long-run changes in word use and it’s arguably tricky and a bit arbitrary to do so.
Don’t get me wrong. I think the Baker et al. paper is great and the indicator is carefully prepared and tested. It actually inspired me to be writing a similar paper at the moment. I’m measuring the use of language indicating financial stress in several newspapers since the 19th century.
But still, counting words in newspapers will obviously yield a noisy indicator and interpreting word frequencies as proxies for the unobserved variable of interest requires strong assumptions.