xi: xtivreg . Too many dummies

[ana garcia gtz]

I am doing some work and have the following command for an instrumental varial regression with panel data. I have 1052 observations per year for 17 years. I am using Stata 17 on mac 12.1

tab year, gen(anio)

for varlist anio*: gen aptoX= porcentaje_apto*X

gen paptoyear= porcentaje_apto*year

xi: xtivreg mortalidad_porhab (hectareas_sembrada=papto) i.year i.fips, first

I have panel data where year goes from 2003 to 2019, porcentaje_apto is percentage of municipality where a certain tree can be grown (it does not change from year to year) , mortalidad_porhab is infant death rate for each municipality per year, hectareas_sembrada is how many hectares of the tree are grown for each municipality per year and fips is the ID for each municipality.

Problem is i have too many dummies which makes stata run very slow with this code.

Is there a way to get same results without the need of the dummies?
I've also tried

xtivreg mortalida_porhab (hectareas_sembrada= paptoyear), fe first

but i don't get same results so dont know if i am doing somenthing wrong.
thanks for any tips you can provide

[Andrew Musau]

xtivreg is from SSC (FAQ Advice #12). You probably have a large number of municipalities in the dataset. Install ivreghdfe from SSC and absorb the fixed effects. You also should cluster your standard errors at the municipality level.

ssc install ivreghdfe, replace
ivreghdfe mortalidad_porhab (hectareas_sembrada=papto), first absorb(fips year) cluster(fips)

[Nick Cox]

Cross-posted at https://www.reddit.com/r/stata/comme..._many_dummies/ Please note our policy on cross-posting, which is that you should tell us about it.

As you're using Stata 17, xi: should be unnecessary in most cases. See

help xi

on this point: it explains that factor variable notation is usually enough.

[Richard Williams]

FYI, xtivreg may have started life as a user-written command, but now it is part of official Stata. The PDF documentation for the command states "We thank Mead Over of the Center for Global Development, who wrote an early implementation of xtivreg." The documentation also says factor variable notation is supported.

It often strikes me how many times we see people using xi: A few commands, especially older user-written commands (but also sem), do not support factor variable notation, but when a command does support FV it is usually if not always better to use FV than xi: Output produced by xi: tends to look terrible and various post-estimation commands (especially margins) do not work correctly if you use xi:

[Andrew Musau]

FYI, xtivreg may have started life as a user-written command, but now it is part of official Stata.

Thanks Richard. What was confusing me was ivregress and user-written commands such as ivreg2 from SSC. Of course, there is no xtivregress (consistent with there not being xtregress).