Significant difference between two regression coefficients (panel data)

[Ellen Veelema]

Hi Statalist-members,

I hope that you can help me out! I have a panel dataset containing 8803 observations in a timeframe 2015Q1 to 2017Q4, for which I want to test whether a certain new rule would change more for men than for women. Rule=1 when observations are from 2017Q2 to 2017Q4.
The following regressions are done using the xtreg, re vce (robust) command:

xtreg y alpha1*(Rule*Male) Ln_Assets MB Loss Issue restate Big4 CFO Specialist Litigation1 AnalystCoverage, re vce (rob
> ust)

xtreg y alpha2*(Rule*Women) Ln_Assets MB Loss Issue restate Big4 CFO Specialist Litigation1 AnalystCoverage, re vce (rob
> ust)

now I would like to test if there is a significant difference in alpha1 and alpha2. How should I do this? The suest command came to my attention, but I am not sure how to use it in my situation.
Can someone help me?

[Carlo Lazzaro]

Ellen:
I'm not sure I got your question right, but I would say that you can collapse the two codes in one only via interacting rule with gender (see -help fvvarlist-).
I assume that alpha1 and alpha2 are the coefficients you're interested in; if that were not the case, please provide more details. Thanks:

xtreg y i.Rule##i.Gender Ln_Assets MB Loss Issue restate Big4 CFO Specialist Litigation1 AnalystCoverage, re vce (robust)

[Ellen Veelema]

Hi Carlo,

My hypothesis is that the Men react more strongly to the new rule in terms of DACC than women.
This is the first regression that I want to do. I am interested here in -.0346849.

. xtreg DACC Disclosure_Male Ln_Assets MB Loss Issue restate Big4 CFO Specialist Litigation1 AnalystCoverage, re vce (robust)

Random-effects GLS regression                   Number of obs     =      7,160
Group variable: FormFilingID                    Number of groups  =        820

R-sq:                                           Obs per group:
     within  = 0.0480                                         min =          1
     between = 0.3406                                         avg =        8.7
     overall = 0.1908                                         max =          9

                                                Wald chi2(11)     =     116.62
corr(u_i, X)   = 0 (assumed)                    Prob > chi2       =     0.0000

                            (Std. Err. adjusted for 820 clusters in FormFilingID)
---------------------------------------------------------------------------------
                |               Robust
           DACC |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
----------------+----------------------------------------------------------------
Disclosure_Male |  -.0346849   .0176294    -1.97   0.049    -.0692379   -.0001319
      Ln_Assets |   .1291115   .0173606     7.44   0.000     .0950852    .1631377
             MB |   .0027905   .0015723     1.77   0.076    -.0002911    .0058721
           Loss |  -.1151629   .0183531    -6.27   0.000    -.1511343   -.0791915
          Issue |    .016028   .0088136     1.82   0.069    -.0012464    .0333025
        restate |   .0039874   .0148041     0.27   0.788    -.0250282     .033003
           Big4 |   .0094025   .0318952     0.29   0.768    -.0531109    .0719159
            CFO |  -.0002056   .0000257    -8.01   0.000    -.0002559   -.0001553
     Specialist |  -.0662087   .0460539    -1.44   0.151    -.1564728    .0240553
    Litigation1 |   .0162057   .0335644     0.48   0.629    -.0495793    .0819908
AnalystCoverage |    -.01718   .0029845    -5.76   0.000    -.0230295   -.0113305
          _cons |   -.642352   .0917915    -7.00   0.000    -.8222601   -.4624439
----------------+----------------------------------------------------------------
        sigma_u |  .40250027
        sigma_e |  .50163203
            rho |  .39165944   (fraction of variance due to u_i)
---------------------------------------------------------------------------------

Next I ran the following regression. I am interested in the value -.015349.

Random-effects GLS regression                   Number of obs     =      7,160
Group variable: FormFilingID                    Number of groups  =        820

R-sq:                                           Obs per group:
     within  = 0.0472                                         min =          1
     between = 0.3403                                         avg =        8.7
     overall = 0.1903                                         max =          9

                                                Wald chi2(11)     =     114.98
corr(u_i, X)   = 0 (assumed)                    Prob > chi2       =     0.0000

                              (Std. Err. adjusted for 820 clusters in FormFilingID)
-----------------------------------------------------------------------------------
                  |               Robust
             DACC |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
------------------+----------------------------------------------------------------
Female_Disclosure |   -.015349   .0330835    -0.46   0.643    -.0801915    .0494936
        Ln_Assets |   .1287988   .0172856     7.45   0.000     .0949196     .162678
               MB |   .0027903   .0015735     1.77   0.076    -.0002938    .0058744
             Loss |  -.1156454   .0183595    -6.30   0.000    -.1516294   -.0796614
            Issue |    .015224    .008936     1.70   0.088    -.0022902    .0327382
          restate |   .0166743   .0131684     1.27   0.205    -.0091352    .0424838
             Big4 |   .0093426   .0319078     0.29   0.770    -.0531955    .0718807
              CFO |   -.000208   .0000261    -7.97   0.000    -.0002592   -.0001569
       Specialist |  -.0655195    .046075    -1.42   0.155    -.1558248    .0247859
      Litigation1 |   .0160254     .03353     0.48   0.633    -.0496922    .0817429
  AnalystCoverage |  -.0171044     .00297    -5.76   0.000    -.0229255   -.0112833
            _cons |  -.6507357   .0935074    -6.96   0.000    -.8340068   -.4674645
------------------+----------------------------------------------------------------
          sigma_u |  .40250655
          sigma_e |  .50190492
              rho |  .39140775   (fraction of variance due to u_i)
-----------------------------------------------------------------------------------

Now I want to test whether the difference between men and women towards DACC (Y) after implementation of the new rule, is significantly different.

Female	Male	Difference (significant at 0.05 level)
-.015349	-.0346849	xxx

Some sort of table I want to make..

I hope that you can help me out!

[Carlo Lazzaro]

Ellen:
if you rule out interaction between -gender- and -rule-, see: https://www.stata.com/support/faqs/s...-coefficients/