fixed effect interaction with many values of FE

[Adrienne Wold]

Hi,

I am trying to run a specification using reghdfe, where I interact a district indicator (uniqueid) with another variable using the specification below. The district variable has ~2500 values, and I have set the maxvar size using stata MP and set empty cells drop, but Stata stops responding every time I run the specification. Is there anyway to estimate this interaction in a way that gets around this?

 reghdfe y x##i.uniqueid w z , cluster(uniqueid) a(uniqueid year)

[Andrew Musau]

reghdfe is from SSC (as you are asked to explain). Do you need the dummy coefficients of the interaction term? If not, the solution is simple - absorb them.

reghdfe y w z , cluster(uniqueid) a(uniqueid year c.x#i.uniqueid )

Note that there is no need to include the variable "x" in the regression as it will be collinear with the interaction dummies, i.e., you cannot identify its effect.

[Phil Bromiley]

With just year and uniqueid as panel variables, some would use xtreg:

xtset uniqueid year
xtreg y c.x c.x#i.uniqueid i.year , vce(cluster, uniquid) fe

i.uniqueid drops one of the values (allowing an estimate on x which is essentially the value for the dropped panel).

[Fahad Mirza]

Andrew Musau just a quick question. what if I needed coefficients as well even after absorbing interaction

[Andrew Musau]

reghdfe y w z , cluster(uniqueid) a(uniqueid year c.x#i.uniqueid, savefe)

will save the estimated fixed effects as variables. Here is an example:

webuse grunfeld, clear
reghdfe invest mvalue kstock i.year, a(company)
qui reghdfe invest mvalue kstock, a(company year, savefe)
sort year
gen yearcoef= __hdfe2__-__hdfe2__[1]
tabdisp year, cell(yearcoef)

Res.:

. reghdfe invest mvalue kstock i.year, a(company)
(MWFE estimator converged in 1 iterations)

HDFE Linear regression                            Number of obs   =        200
Absorbing 1 HDFE group                            F(  21,    169) =      31.90
                                                  Prob > F        =     0.0000
                                                  R-squared       =     0.9517
                                                  Adj R-squared   =     0.9431
                                                  Within R-sq.    =     0.7985
                                                  Root MSE        =    51.7245

------------------------------------------------------------------------------
      invest |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
      mvalue |   .1177158   .0137513     8.56   0.000     .0905694    .1448623
      kstock |   .3579163    .022719    15.75   0.000     .3130667    .4027659
             |
        year |
       1936  |  -19.19741   23.67586    -0.81   0.419    -65.93593    27.54112
       1937  |  -40.69001   24.69541    -1.65   0.101    -89.44122    8.061213
       1938  |   -39.2264   23.23594    -1.69   0.093    -85.09647    6.643667
       1939  |  -69.47029   23.65607    -2.94   0.004    -116.1698   -22.77083
       1940  |  -44.23507   23.80979    -1.86   0.065      -91.238     2.76785
       1941  |  -18.80446     23.694    -0.79   0.429     -65.5788    27.96987
       1942  |  -21.13979   23.38163    -0.90   0.367    -67.29748    25.01789
       1943  |  -42.97762   23.55287    -1.82   0.070    -89.47334    3.518104
       1944  |  -43.09876    23.6102    -1.83   0.070    -89.70766    3.510134
       1945  |  -55.68303   23.89561    -2.33   0.021    -102.8554   -8.510689
       1946  |  -31.16928   24.11598    -1.29   0.198    -78.77665    16.43809
       1947  |  -39.39223   23.78368    -1.66   0.100    -86.34361    7.559141
       1948  |  -43.71651   23.96965    -1.82   0.070    -91.03501    3.601991
       1949  |   -73.4951   24.18292    -3.04   0.003    -121.2346   -25.75559
       1950  |  -75.89611   24.34553    -3.12   0.002    -123.9566    -27.8356
       1951  |   -62.4809   24.86425    -2.51   0.013    -111.5654   -13.39637
       1952  |  -64.63233    25.3495    -2.55   0.012    -114.6748   -14.58987
       1953  |  -67.71796   26.61108    -2.54   0.012    -120.2509   -15.18501
       1954  |  -93.52622   27.10786    -3.45   0.001    -147.0399   -40.01257
             |
       _cons |  -32.83631   18.87533    -1.74   0.084     -70.0981    4.425483
------------------------------------------------------------------------------

Absorbed degrees of freedom:
-----------------------------------------------------+
 Absorbed FE | Categories  - Redundant  = Num. Coefs |
-------------+---------------------------------------|
     company |        10           0          10     |
-----------------------------------------------------+

. 
. qui reghdfe invest mvalue kstock, a(company year, savefe)

. 
. sort year

. 
. gen yearcoef= __hdfe2__-__hdfe2__[1]

. 
. tabdisp year, cell(yearcoef)

----------------------
     year |   yearcoef
----------+-----------
     1935 |          0
     1936 |  -19.19741
     1937 |  -40.69001
     1938 |   -39.2264
     1939 |  -69.47029
     1940 |  -44.23507
     1941 |  -18.80446
     1942 |  -21.13979
     1943 |  -42.97762
     1944 |  -43.09877
     1945 |  -55.68303
     1946 |  -31.16928
     1947 |  -39.39223
     1948 |  -43.71651
     1949 |  -73.49509
     1950 |  -75.89611
     1951 |   -62.4809
     1952 |  -64.63232
     1953 |  -67.71796
     1954 |  -93.52622
----------------------

.

[S. M. Woahid Murad]

Hi Andrew Musau,

I am using the following command for my panel data estimation:

reghdfe y x1 x2 x3 x4, absorb(bank year bank*year) vce(cluster bank); where I declare my dataset as xtset bank year, yearly.

Does it make sense if I consider bank, year and their interaction together as fixed effects? Many previous studies drop bank and year individual fixed effects before including their interaction.

Many thanks for considering my request.


Kind Regards,
Woahid

[Andrew Musau]

I declare my dataset as xtset bank year, yearly.

If your panel identifier is bank and your time variable is year, then a bank-year is an observation in that dataset. You cannot have fixed effects at the observation level. So you probably want:

reghdfe y x1 x2 x3 x4, absorb(bank year) vce(cluster bank)

Additionally, no need to create interactions manually. Use factor variables.

help fvvarlist