You’re looking too closely. The rule just needs to trigger “close enough” to the beginning relative to the duration to serve as a good tool. For comparison, the inverted yield curve rule has a range that varies from 6-12 months for predicting a recession (depending on the economist and pundit). Recall that the NBER doesn’t even declare a recession until after its over. So the reason why 2003 as a false positive is that no recession ever occurred “near” the trigger of the rule. So my question would be to what degree of accuracy do you require for a rule to be sufficiently useful? There is none better than the Sahm Rule in terms of fitting history (I say “fitting” in deference to both Sahm and Powell who insist it is not a predictive tool). I am surprised you want to give the Sahm Rule credit for the recession that ended in 2022 – we’re looking for an indicator that triggers in time to make economic decisions (as Sahm was doing).

For quibbles, you can use Powell’s term, the Sahm Rule has “statistical regularity”, aka, it is highly correlated with the onset of a recession.

The chart seems to indicate that the Sahm rule confirms recessions rather than predicts them; by the time the metric reaches 0.5 the recession is already in progress. Given that, there was a recession around 2002, so how was the nearby Sahm spike a false positive?