Read on Twitter
As one of the authors of this paper let me take this occasion to add a few points on composite indicators of financial conditions (most of these points are valid also for "financing" conditions)
#Econtwitter
@fwred
1/N https://twitter.com/fwred/status/1352636177933881346
#Econtwitter
@fwred
1/N https://twitter.com/fwred/status/1352636177933881346
FCIs have become very popular after the GFC. Economists love simple indicators, the way all humans like simple answers to complicated questions. Econometricians (I am also guilty of this) love complicated things, like kids that like to signal that they can do tricky stuff.
2/N
2/N
As a result, FCIs with complicated methods have become widespread. In this paper we show that even a simple average of individual indicators performs as well, if not better, than complicated ones. They are also more transparent.
Yet, how should one use these indices?
3/N
Yet, how should one use these indices?
3/N
Best use is to capture in a parsimonious setting the cross sectional/time series behaviour of a number of financial variables for a number of regions/countries. As dimensions (variables/countries) can become quickly large, data reduction helps analysis. But care is needed.
4/N
4/N
Most of these indices are standardized (0 mean, 1 standard deviation), hence conclusions like: financial conditions "are tigher" in country X than in Y make no sense. One can say: financial conditions have tightened in country X more than in Y (relative to historical norm)
5/N
5/N
Also, shocks matter. Interest rates can rise for good reasons (growth) or for bad reasons (risk of default). An FCI does not distinguish this. Also important: FX depreciation may or may not raise financing costs (depending on whether you have debt in foreign currency).
6/N
6/N
Conclusion: FCIs can be useful (better if they are simple and transparent) when they are used appropriately, they can be useless (even dangerous) if used inappropriately. Just like all the tools that economists and policy makers have in their toolbox.
N/N
N/N
