teffects ipwra to test self-selective effects

[Lynn Zhang]

Hi! I have ran a regression analysis, see below. It shows that people who do Economics, Mathematics and Computer Sciences earn significantly higher than others. It also shows that Asian Indian and Pakistani earn significantly more than others. I want to test whether Indian and Pakistani students are self-selecting into the higher paying subjects with "teffects ipwra" and conducted a test.

 reg ln_real_income year_2018_19 year_2019_20 Female Mature first_class lower_second_class third_class un
> classified London arts_humanities social_sciences medicine Economics Law Mathematics Computer_sciences P
> arents_no_degree imd_40pc disability polar1_2 asian_high_paying black_high_paying mixed_high_paying Arab
>  Asian_Bangladeshi Asian_Indian Asian_Pakistani black_African black_Caribbean Chinese no_info mixed_whit
> e_asian mixed_white_african mixed_white_caribeean

      Source |       SS           df       MS      Number of obs   =     2,062
-------------+----------------------------------   F(34, 2027)     =     31.26
       Model |  91.0613268        34  2.67827432   Prob > F        =    0.0000
    Residual |  173.677701     2,027  .085682141   R-squared       =    0.3440
-------------+----------------------------------   Adj R-squared   =    0.3330
       Total |  264.739027     2,061  .128451736   Root MSE        =    .29272

---------------------------------------------------------------------------------------
       ln_real_income | Coefficient  Std. err.      t    P>|t|     [95% conf. interval]
----------------------+----------------------------------------------------------------
         year_2018_19 |  -.0092687    .015784    -0.59   0.557    -.0402233     .021686
         year_2019_20 |  -.0586619   .0161195    -3.64   0.000    -.0902744   -.0270494
               Female |  -.0826012   .0139492    -5.92   0.000    -.1099575   -.0552449
               Mature |   .0382448   .0370626     1.03   0.302    -.0344398    .1109295
          first_class |   .0645985   .0166097     3.89   0.000     .0320247    .0971723
   lower_second_class |  -.0891009   .0381644    -2.33   0.020    -.1639464   -.0142554
          third_class |   -.058365   .0899435    -0.65   0.516    -.2347564    .1180263
         unclassified |   .1204162   .0230371     5.23   0.000     .0752373     .165595
               London |   .1986692   .0141862    14.00   0.000     .1708481    .2264902
      arts_humanities |  -.2094648    .020395   -10.27   0.000    -.2494621   -.1694675
      social_sciences |  -.1152332   .0226748    -5.08   0.000    -.1597014   -.0707649
             medicine |   .0317705   .0263738     1.20   0.228    -.0199521    .0834931
            Economics |    .247034   .0343026     7.20   0.000     .1797621     .314306
                  Law |  -.0295825   .0362345    -0.82   0.414    -.1006432    .0414783
          Mathematics |   .2362116   .0347456     6.80   0.000     .1680708    .3043525
    Computer_sciences |   .3802876   .0369847    10.28   0.000     .3077555    .4528196
    Parents_no_degree |   .0247399   .0197299     1.25   0.210    -.0139531     .063433
             imd_40pc |  -.0324318     .02108    -1.54   0.124    -.0737724    .0089089
           disability |   -.040524   .0205868    -1.97   0.049    -.0808975   -.0001505
             polar1_2 |  -.0528427   .0230941    -2.29   0.022    -.0981334   -.0075521
    asian_high_paying |  -.0357674   .0383723    -0.93   0.351    -.1110206    .0394858
    black_high_paying |   .0939296   .0981908     0.96   0.339    -.0986358     .286495
    mixed_high_paying |  -.0064315   .0584076    -0.11   0.912    -.1209768    .1081137
                 Arab |  -.0762608   .0990565    -0.77   0.441     -.270524    .1180023
    Asian_Bangladeshi |  -.0335714   .0946529    -0.35   0.723    -.2191985    .1520557
         Asian_Indian |   .1352684   .0332033     4.07   0.000     .0701522    .2003846
      Asian_Pakistani |   .2403407   .0715806     3.36   0.001     .0999616    .3807199
        black_African |  -.0290208   .0718641    -0.40   0.686    -.1699561    .1119145
      black_Caribbean |   .0679038   .1301019     0.52   0.602    -.1872437    .3230513
              Chinese |   .0672874   .0416096     1.62   0.106    -.0143146    .1488894
              no_info |   .0953633   .0606149     1.57   0.116    -.0235108    .2142374
    mixed_white_asian |   .0328668   .0405814     0.81   0.418    -.0467188    .1124524
  mixed_white_african |  -.0501021   .0914115    -0.55   0.584    -.2293724    .1291682
mixed_white_caribeean |   .0842579   .0897269     0.94   0.348    -.0917086    .2602245
                _cons |   10.28573   .0208051   494.38   0.000     10.24493    10.32654
---------------------------------------------------------------------------------------

I am not sure whether I have understoof teffects ipwra correctly: I assumed that being female, being the 40% most impoverished in the country (IMD_40pc), being in areas with the lowest higher education participation (POLAR1_2) and having disability affects one's choice to do higher paying subjects at university. In the results below, does it indicate that being an Asian means that one is 19.98% more likely to choose higher paying subjects?

teffects ipwra ( high_paying_subject Female imd_40pc polar1_2 disability) ( Asian imd_40pc p
> olar1_2 disability)

Iteration 0:   EE criterion =  5.742e-24  
Iteration 1:   EE criterion =  1.099e-33  

Treatment-effects estimation                    Number of obs     =      2,062
Estimator      : IPW regression adjustment
Outcome model  : linear
Treatment model: logit
------------------------------------------------------------------------------
             |               Robust
high_payin~t | Coefficient  std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
ATE          |
       Asian |
   (1 vs 0)  |   .1998188   .0362915     5.51   0.000     .1286888    .2709489
-------------+----------------------------------------------------------------
POmean       |
       Asian |
          0  |   .3066177   .0108662    28.22   0.000     .2853202    .3279151
------------------------------------------------------------------------------

My questions are: how to I correctly use teffects ipwra to test self-selective effects, and, how would I interpret the teffects ipwra results? Thank you in advance! Any help would be very appreciated!