ppmlhdfe: Should NTM notifications with no reports be coded as zero or missing?

[Ebru Aricioglu]

Dear all,
I am estimating a model using the ppmlhdfe estimator with robust standard errors.
One of my explanatory variables is the number of non-tariff measure (NTM). The model is like this: ppmlhdfe import lngdp lnrer ntm, absorb(country year) vce(robust) For some country–year observations, there are no notifications reported. In this case, should these observations be coded as zero notifications, or treated as missing values?
Thanks in advance.
Ebru

[Clyde Schechter]

It depends. And it may be different for different country-year observations. If you have no notifications for a country year because there actually were no NTMs for that country in that year, that should be coded as a zero. But if you have no notifications because the data for that country and year are, for whatever reason, simply unavailable, and some NTMs may have occurred (or not), but we don't know, then that is coded as missing value.

In short: code 0 when there are actually no NTMs, code missing value when the number of NTMs is not known (and might or might not be zero) due to some failure of the reporting process.

Added: This has nothing to do with -ppmle-; you would do the same for any command.