ppmlhdfe Pseudo R2 0.99+

[Amos Machado]

Here is the model to estimate the impact of free trade agreements (fta) and preferential trade agreements (pta) on exports (exports) for annual data from 1990-2018 with 4 year intervals:

ppmlhdfe exports fta pta , a(im#year ex#year im#ex) cluster(im#ex)

The value of pseudo R2 remains as high as 0.99+, and it stays there for various model specifications. Should I doubt something for this? Is there some alternative measure I may additionally calculate?

[Phil Bromiley]

Welcome to Stata list. You will increase your chances of useful answer by following the FAQ on asking questions-provide Stata code in code delimiters, readable Stata output, and sample data using dataex.

With a user written procedure, help with this sort of issue often depends on someone active on the list happening to use that procedure. I would certainly worry a little bit about such high R squares. While there's not a lot to do without the data, I would wonder a lot about all of the variables in your a() statement. How many free parameters is this statement creating relative to the actual sample size? And it may be that the effective sample size is not exactly what you think it is if you have a lot of zeros.

[Amos Machado]

Thanks, dear Phil,
I will follow the valuable comments.

[Carlo Lazzaro]

Amos:
have you detected quasi-extreme multicolinearity issues in your predictors?
That said, posting what you typed and what Stata gave you back (as recommended by the FAQ and wisely echoed in Phil's reply) will increase your chances of getting helpful replies. Thanks.

[Amos Machado]

Hello,
Here is the relevant example given in the ppmlhdfe help file. The pseudo R2 in this case is also 0.99+. The data I have is very similar except that I have around 40% zero values for trade, the dependent variable.
And Phil, it is pseudo R2, so may not be interpreted as it would be an R2 with a value of 0.99+, I think.

use "http://fmwww.bc.edu/RePEc/bocode/e/EXAMPLE_TRADE_FTA_DATA" if category=="TOTAL", clear
egen imp = group(isoimp)
egen exp = group(isoexp)
ppmlhdfe trade fta, a(imp#year exp#year imp#exp) cluster(imp#exp)

[Chinmay Sharma]

Its not surprising. You have importer year fixed effects, exporter-year effects, as well as importer-exporter effects. You are capturing a lot of variation this way, and its not surprising that the Pseudo R2 is this high.