[Gretl-users] (NBER) recession shading in gretl/gnuplot

Allin Cottrell cottrell at wfu.edu
Tue Nov 17 09:26:19 EST 2015


On Wed, 11 Nov 2015, Sven Schreiber wrote:

> here's a little detail thingy about the built-in NBER recession shading. It's 
> not really a bug (I think), but perhaps a little bit misleading.
>
> Gretl uses the dates of the NBER peaks and troughs (I'm talking about monthly 
> data now) to determine the beginning and end of the shading. For example, one 
> NBER peak is 2001M3 (March), and so gretl's shading starts immediately after 
> February, covering all of the March interval in the plot. Therefore, the 
> observed March value of any plotted variable is a bit "later" (= to the 
> right), because the point is centered within the March interval. So the peak 
> value will always be "inside" the recession.
>
> It seems that that's not the conventional way of presenting the recessions. 
> If you look at a time series plot on the FRED webpage for example, the March 
> value will be shown right at the beginning of the shading. Also, FRED offers 
> an indicator series for the NBER recessions which explicitly says "starting 
> with the period after the peak". It's a matter of convention, but I think the 
> FRED shading makes more sense than the current gretl shading.
>
> So: I suggest that either the shading should start in the middle of the 
> period (middle of March in this example). But given that that may not be 
> feasible with gnuplot, I would suggest to adopt the FRED convention and start 
> the shading right after the peak dates.

Right now the documentation for a "plotbars" file says:

# Format: each non-comment line must contain two space-
# separated dates in the format YYYY:MM.  These represent
# the start and end of "episodes" of some sort.

I take your point, but if we're to abide by this description, maybe 
the appropriate fix would be to modify the example NBER file: make the 
"start" of each NBER recession the month after the peak rather than 
the peak itself. In this case no changes to the plotting code would be 
required.

The alternative would be to leave the NBER file as it is and reword 
the description, replacing "start" (of an episode) with something like 
"period immediately preceding, as in NBER business cycle peaks". This 
would have to be accompanied by a change to the plotting code.

Preferences/thoughts?

Allin



More information about the Gretl-users mailing list