After this post I decided to take a look at: A note on pitfalls of credit crunch regressions
The credit crunch regression:
Δy(t) = β0 + β1 × Δb(t) + ε(t),
where Δy(t) and Δb(t) denote growth rate of output and loans from banks. ε(t) is an i.i.d. error term. The regression should detect statistically significant effect of Δbt with a positive estimate
Using simulations in a Kiyotaki-Moore (KM) economy, the author notes:
In KM model, the credit crunch regression only works fine if the technology shocks are expected to be long-lived.
I.e. Neither the model by the author nor the regressions are as general as they appear to be.