Comparing regression coefficients of different independent variables in separate regressions

[Vinh Ng]

Dear all,

Suppose I have the following dataset

input dv iv1 iv2 firmid year
1 8 1 155 2001
3 9 1 155 2002
5 10 1 155 2003
8 15 1 155 2004
9 18 1 166 2001
12 16 1 166 2002
11 14 1 166 2003
15 12 1 166 2004
14 5 0 177 2001
14 7 0 177 2002
15 3 0 177 2003
17 5 0 177 2004
18 2 0 177 2005
19 4 0 177 2006
12 8 0 177 2007
3 6 1 188 2001
5 8 1 188 2002
2 9 1 188 2003
1 8 1 188 2004
6 12 1 199 2001
7 9 1 199 2002
4 10 1 199 2003
56 11 1 199 2004
23 8 1 199 2005
10 3 0 199 2006
11 3 0 200 2001
12 4 0 200 2002
54 1 0 200 2003
6 2 0 200 2004
5 5 0 200 2005
end

I'd like to compare regression coefficients of two different independent variables (iv1 and iv2) in two separate regressions. For example:

reghdfe dv iv1, absorb(firmid year) vce(cluster firmid)
reghdfe dv iv2, absorb(firmid year) vce(cluster firmid)

I'd like to use reghdfe command instead of reg as I need to control for firm fixed effects. I'm aware that I can't use suest in this case as it doesn't capture reghdfe regression output.

Could anyone please help me find a way to go around this problem?

Thank you very much for your help,

Vinh

[Andrew Musau]

reghdfe is from SSC, as you are asked to explain (see FAQ #12). For two dependent variables and the same independent variable, see
https://www.statalist.org/forums/for...erent-outcomes

input dv iv1 iv2 firmid year
1 8 1 155 2001
3 9 1 155 2002
5 10 1 155 2003
8 15 1 155 2004
9 18 1 166 2001
12 16 1 166 2002
11 14 1 166 2003
15 12 1 166 2004
14 5 0 177 2001
14 7 0 177 2002
15 3 0 177 2003
17 5 0 177 2004
18 2 0 177 2005
19 4 0 177 2006
12 8 0 177 2007
3 6 1 188 2001
5 8 1 188 2002
2 9 1 188 2003
1 8 1 188 2004
6 12 1 199 2001
7 9 1 199 2002
4 10 1 199 2003
56 11 1 199 2004
23 8 1 199 2005
10 3 0 199 2006
11 3 0 200 2001
12 4 0 200 2002
54 1 0 200 2003
6 2 0 200 2004
5 5 0 200 2005
end

reghdfe dv iv1, absorb(firmid year) vce(cluster firmid)
reghdfe dv iv2, absorb(firmid year) vce(cluster firmid)
reshape long iv, i(firmid year) j(which)
gen group= 1.which
reghdfe dv group#(c.iv), a(firmid#group year#group) cluster(firmid)

Res.:

. reghdfe dv iv1, absorb(firmid year) vce(cluster firmid)
(dropped 1 singleton observations)
(MWFE estimator converged in 4 iterations)

HDFE Linear regression                            Number of obs   =         29
Absorbing 2 HDFE groups                           F(   1,      5) =       0.21
Statistics robust to heteroskedasticity           Prob > F        =     0.6650
                                                  R-squared       =     0.3091
                                                  Adj R-squared   =    -0.2091
                                                  Within R-sq.    =     0.0039
Number of clusters (firmid)  =          6         Root MSE        =    14.3803

                                 (Std. Err. adjusted for 6 clusters in firmid)
------------------------------------------------------------------------------
             |               Robust
          dv |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         iv1 |  -.3603447   .7838261    -0.46   0.665    -2.375234    1.654544
       _cons |   15.44132   6.135466     2.52   0.053    -.3303987    31.21304
------------------------------------------------------------------------------

Absorbed degrees of freedom:
-----------------------------------------------------+
 Absorbed FE | Categories  - Redundant  = Num. Coefs |
-------------+---------------------------------------|
      firmid |         6           6           0    *|
        year |         6           0           6     |
-----------------------------------------------------+
* = FE nested within cluster; treated as redundant for DoF computation

. 
. reghdfe dv iv2, absorb(firmid year) vce(cluster firmid)
(dropped 1 singleton observations)
(MWFE estimator converged in 4 iterations)
warning: missing F statistic; dropped variables due to collinearity or too few clusters

HDFE Linear regression                            Number of obs   =         29
Absorbing 2 HDFE groups                           F(   1,      5) =          .
Statistics robust to heteroskedasticity           Prob > F        =          .
                                                  R-squared       =     0.3202
                                                  Adj R-squared   =    -0.1897
                                                  Within R-sq.    =     0.0199
Number of clusters (firmid)  =          6         Root MSE        =    14.2644

                                 (Std. Err. adjusted for 6 clusters in firmid)
------------------------------------------------------------------------------
             |               Robust
          dv |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         iv2 |       12.6   3.47e-15  3.6e+15   0.000         12.6        12.6
       _cons |   5.234483   2.09e-15  2.5e+15   0.000     5.234483    5.234483
------------------------------------------------------------------------------

Absorbed degrees of freedom:
-----------------------------------------------------+
 Absorbed FE | Categories  - Redundant  = Num. Coefs |
-------------+---------------------------------------|
      firmid |         6           6           0    *|
        year |         6           0           6     |
-----------------------------------------------------+
* = FE nested within cluster; treated as redundant for DoF computation

. 

. 
. reghdfe dv group#(c.iv), a(firmid#group year#group) cluster(firmid)
(dropped 2 singleton observations)
(MWFE estimator converged in 4 iterations)
warning: missing F statistic; dropped variables due to collinearity or too few clusters

HDFE Linear regression                            Number of obs   =         58
Absorbing 2 HDFE groups                           F(   2,      5) =          .
Statistics robust to heteroskedasticity           Prob > F        =          .
                                                  R-squared       =     0.3146
                                                  Adj R-squared   =    -0.2208
                                                  Within R-sq.    =     0.0119
Number of clusters (firmid)  =          6         Root MSE        =    14.3225

                                 (Std. Err. adjusted for 6 clusters in firmid)
------------------------------------------------------------------------------
             |               Robust
          dv |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  group#c.iv |
          0  |       12.6   6.80e-15  1.9e+15   0.000         12.6        12.6
          1  |  -.3603447   .7907935    -0.46   0.668    -2.393144    1.672455
             |
       _cons |    10.3379   3.095002     3.34   0.021     2.381944    18.29386
------------------------------------------------------------------------------

Absorbed degrees of freedom:
--------------------------------------------------------+
    Absorbed FE | Categories  - Redundant  = Num. Coefs |
----------------+---------------------------------------|
   firmid#group |        12          12           0    *|
     year#group |        12           0          12     |
--------------------------------------------------------+
* = FE nested within cluster; treated as redundant for DoF computation

[Vinh Ng]

Many thanks, Andrew. Your code works perfectly.

Cheers,

Vinh