Treatment Dummy Variable Interpretation

[Lydia Gad]

Dear Statalist,

I need your advice urgently regarding the following:

I have a panel data and I have xtset it with country and time.
My model is: xtset Y B₁Period of treatment * treatment B₂X_{vector of other controls} i.country i.time

Now, my Y is in % and the coefficient on my B₁ is sometimes very high like 27.
The interpretation of B₁ is then to multiply 27 by 100 and say that receiving the treatment increases Y by 2700%, which does not make any sense.

Can you please help me with the interpretation? I know it is an elementary question but I am now not sure whether my model make sense or not?

Thank you so much for your help.

Best regards,
Lydia Gad

[Lydia Gad]

Thinking about the interpretation again, shouldn't it be: Receiving the treatment increases Y by 27 percentage points?

Thank you so much for your help.

[Chris Boudreaux]

Hi Lydia,

As per the FAQ, it would be much more helpful if you could report your exact results in this thread. This will make it much easier for us to understand the variables and output. As it is currently, written, I'm not sure which variables are in the interaction, which will affect the interpretation.

All you need to do is wrap the Code tags (i.e, the # button in the Statalist window) around your Stata output and copy and paste it to here.

[Lydia Gad]

Hi Chris,

Thank you for your help. I hope this is clearer.
The coefficient of interest "Period-treat" is a dummy variable that takes the value of 0 or 1.
The coverage ratio is in percentage terms.

. xtreg I3631_Coverage_ratio_A Period_treat A1100_Loans_and_advances_A Total_Liabilities_A RO
> E_A  ROA_A  Debt_Securities_A  Total_Assets_A i.new_referencearea
note: 27.new_referencearea omitted because of collinearity

Random-effects GLS regression                   Number of obs     =        439
Group variable: new_refere~a                    Number of groups  =         27

R-sq:                                           Obs per group:
     within  = 0.2018                                         min =          1
     between = 1.0000                                         avg =       16.3
     overall = 0.8207                                         max =         19

                                                Wald chi2(31)     =          .
corr(u_i, X)   = 0 (assumed)                    Prob > chi2       =          .

--------------------------------------------------------------------------------------------
    I3631_Coverage_ratio_A |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
---------------------------+----------------------------------------------------------------
              Period_treat |   26.67147   2.253275    11.84   0.000     22.25513     31.0878
A1100_Loans_and_advances_A |  -1.52e-09   3.55e-09    -0.43   0.669    -8.49e-09    5.45e-09
       Total_Liabilities_A |   1.49e-08   3.75e-09     3.98   0.000     7.56e-09    2.23e-08
                     ROE_A |  -.6494135   .2313493    -2.81   0.005     -1.10285   -.1959773
                     ROA_A |    8.34393   2.401678     3.47   0.001     3.636727    13.05113
         Debt_Securities_A |  -2.16e-07   7.96e-07    -0.27   0.786    -1.78e-06    1.34e-06
            Total_Assets_A |  -1.32e-08   2.88e-09    -4.57   0.000    -1.88e-08   -7.52e-09
                           |
         new_referencearea |
             BE (Belgium)  |  -7.255016   1.699754    -4.27   0.000    -10.58647   -3.923559
            BG (Bulgaria)  |   14.40076   3.649801     3.95   0.000      7.24728    21.55424
              CY (Cyprus)  |  -13.88762   2.200405    -6.31   0.000    -18.20034   -9.574906
      CZ (Czech Republic)  |   .5309052   3.570901     0.15   0.882    -6.467933    7.529743
             DE (Germany)  |   17.10475   12.28732     1.39   0.164    -6.977957    41.18746
             DK (Denmark)  |    5.47142   2.121775     2.58   0.010     1.312817    9.630023
             EE (Estonia)  |  -17.23595    2.43246    -7.09   0.000    -22.00349   -12.46842
               ES (Spain)  |  -8.517472   5.699551    -1.49   0.135    -19.68839    2.653443
             FI (Finland)  |  -25.08714   2.040973   -12.29   0.000    -29.08738   -21.08691
              FR (France)  |  -1.896817   11.90187    -0.16   0.873    -25.22405    21.43041
      GB (United Kingdom)  |   6.744207   9.994034     0.67   0.500    -12.84374    26.33215
              GR (Greece)  |  -6.575003   1.980087    -3.32   0.001     -10.4559   -2.694103
             HR (Croatia)  |   29.58684   3.622033     8.17   0.000     22.48779     36.6859
             HU (Hungary)  |   33.61253   5.856989     5.74   0.000     22.13304    45.09201
             IE (Ireland)  |  -18.68175   2.129423    -8.77   0.000    -22.85534   -14.50815
               IT (Italy)  |  -4.409037   3.805661    -1.16   0.247      -11.868    3.049921
          LU (Luxembourg)  |  -21.45035   2.153704    -9.96   0.000    -25.67153   -17.22916
              LV (Latvia)  |  -22.94123   2.274545   -10.09   0.000    -27.39926   -18.48321
               MT (Malta)  |  -17.33036   2.888248    -6.00   0.000    -22.99123    -11.6695
         NL (Netherlands)  |  -21.50378   4.176171    -5.15   0.000    -29.68892   -13.31863
              PL (Poland)  |   22.02575   3.239917     6.80   0.000     15.67563    28.37587
            PT (Portugal)  |  -8.162201   2.062478    -3.96   0.000    -12.20458   -4.119818
             RO (Romania)  |   17.11211   3.666844     4.67   0.000     9.925223    24.29899
              SE (Sweden)  |          0  (omitted)
            SI (Slovenia)  |   2.221463   2.358367     0.94   0.346    -2.400852    6.843777
            SK (Slovakia)  |   3.122234   2.479558     1.26   0.208     -1.73761    7.982079
                           |
                     _cons |   28.47181   3.480396     8.18   0.000     21.65036    35.29326
---------------------------+----------------------------------------------------------------
                   sigma_u |          0
                   sigma_e |  4.7123562
                       rho |          0   (fraction of variance due to u_i)
--------------------------------------------------------------------------------------------

[Chris Boudreaux]

I believe the interpretation is that period_treat is associated with a 26.67 percentage point increase in the coverage ratio. I have three questions though:

1) You mentioned an interaction previously. I do not see one here. That would affect the interpretation.
2) You have specified both random effects and fixed effects in one model. Which one do you want? You can estimate a fixed effects model by 'xtreg, fe' or you can do it by including the country dummies. You included the country dummies and specified the random effects model, 'xtreg' which does random effects by default.
3) Is your dependent variable--the coverage ratio--bounded between 0 and 1? If so, you might want to look into fractional response models or beta regressions. Here is an introduction to the method from Stata (https://www.stata.com/stata14/fracti...re%20excluded.)

[Lydia Gad]

Hi Chris,

1) The interaction term is the Period_treat dummy variable.

This is the first difference, creating control and treatment group:

generate treat=1 if ((new_referencearea==1 | new_referencearea==3 | new_referencearea==5 | new_referencearea==7 | new_referencearea==9 | new_referencearea==10 |  new_referencearea==11 | new_referencearea==12 | new_referencearea==14 | new_referencearea==17 | new_referencearea==18 | new_referencearea==19 | new_referencearea==20 | new_referencearea==21 | new_referencearea==22 | new_referencearea==23 | new_referencearea==25 | new_referencearea==28 | new_referencearea==29) & (new_timeperiodorrange==9 | new_timeperiodorrange==10 | new_timeperiodorrange==11 | new_timeperiodorrange==12 | new_timeperiodorrange==20 | new_timeperiodorrange==21 | new_timeperiodorrange==22 | new_timeperiodorrange==23 | new_timeperiodorrange==24 | new_timeperiodorrange==32 | new_timeperiodorrange==33 | new_timeperiodorrange==34 | new_timeperiodorrange==35 | new_timeperiodorrange==36 | new_timeperiodorrange==44 | new_timeperiodorrange==45 | new_timeperiodorrange==46 | new_timeperiodorrange==47 | new_timeperiodorrange==48))
replace treat=0 if ((new_referencearea==4 | new_referencearea==6 | new_referencearea==8 | new_referencearea==13 | new_referencearea==15 | new_referencearea==16 | new_referencearea==24 | new_referencearea==26 | new_referencearea==27) & (new_timeperiodorrange==9 | new_timeperiodorrange==10 | new_timeperiodorrange==11 | new_timeperiodorrange==12 | new_timeperiodorrange==20 | new_timeperiodorrange==21 | new_timeperiodorrange==22 | new_timeperiodorrange==23 | new_timeperiodorrange==24 | new_timeperiodorrange==32 | new_timeperiodorrange==33 | new_timeperiodorrange==34 | new_timeperiodorrange==35 | new_timeperiodorrange==36 | new_timeperiodorrange==44 | new_timeperiodorrange==45 | new_timeperiodorrange==46 | new_timeperiodorrange==47 | new_timeperiodorrange==48))

This is the second difference, a difference in time for the treatment:

generate Period = 0 if (new_timeperiodorrange==1 | new_timeperiodorrange==2 | new_timeperiodorrange==3 | new_timeperiodorrange==4 | new_timeperiodorrange==5 | new_timeperiodorrange==6 | new_timeperiodorrange==7 | new_timeperiodorrange==8 | new_timeperiodorrange==13 | new_timeperiodorrange==14 | new_timeperiodorrange==15 | new_timeperiodorrange==16 | new_timeperiodorrange==17 | new_timeperiodorrange==18 | new_timeperiodorrange==19 | new_timeperiodorrange==25 | new_timeperiodorrange==26 | new_timeperiodorrange==27 | new_timeperiodorrange==28 | new_timeperiodorrange==29 | new_timeperiodorrange==30 | new_timeperiodorrange==31 | new_timeperiodorrange==37 | new_timeperiodorrange==38 | new_timeperiodorrange==39 | new_timeperiodorrange==40 | new_timeperiodorrange==41 | new_timeperiodorrange==42 | new_timeperiodorrange==43)
replace Period = 1 if (new_timeperiodorrange==9 | new_timeperiodorrange==10 | new_timeperiodorrange==11 | new_timeperiodorrange==12 | new_timeperiodorrange==20 | new_timeperiodorrange==21 | new_timeperiodorrange==22 | new_timeperiodorrange==23 | new_timeperiodorrange==24 | new_timeperiodorrange==32 | new_timeperiodorrange==33 | new_timeperiodorrange==34 | new_timeperiodorrange==35 | new_timeperiodorrange==36 | new_timeperiodorrange==44 | new_timeperiodorrange==45 | new_timeperiodorrange==46 | new_timeperiodorrange==47 | new_timeperiodorrange==48)

And the Period_treat is then:

generate Period_treat = Period*treat

2) I tried xtreg,fe but it omits Period_treat, my coefficient of interest.
Therefore I included xtreg..... i.new_referencearea, I think statistically that is fine, or? I saw many Statalist forms saying that xtreg ....i.country is fine.


3) summarizing the coverage ratio is:

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
I3631_Cove~A |        551    43.22249     11.0617   6.871287   71.42498

which means it is between 0 and 100 as it is in percentage terms.

Thank you chris for your help, I really appreciate it.

[Carlo Lazzaro]

Lydia:
just one step aside from Chris' helpful advice.
Your R-sq between=1.000 is obviously the warning signal that something is wrong with your model specification and this suspect is confimed by sigma_u=0 (ie, no evidence of a group-wise effect).
The usual recipe would be to consider a pooled OLS, instead.

[Lydia Gad]

Hi Carlo,

Thank you so much for your response.

I have looked up now Pooled OLS and run the following regression:

 reg I3631_Coverage_ratio_A Period_treat A1100_Loans_and_advances_A Total_Liabilities_A ROE_A  ROA_A  Debt_Securities_A  Total_Assets_A i.new_referencearea i.Time, vce (cluster new_referencearea)

The R-sqaured is reduced to 0.8242, is this considered better?

note: 27.new_referencearea omitted because of collinearity

Linear regression                               Number of obs     =        439
                                                F(23, 26)         =          .
                                                Prob > F          =          .
                                                R-squared         =     0.8242
                                                Root MSE          =     4.7736

                                   (Std. Err. adjusted for 27 clusters in new_referencearea)
--------------------------------------------------------------------------------------------
                           |               Robust
    I3631_Coverage_ratio_A |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
---------------------------+----------------------------------------------------------------
              Period_treat |   26.40972   2.225921    11.86   0.000     21.83428    30.98517
A1100_Loans_and_advances_A |  -3.50e-09   5.35e-09    -0.65   0.519    -1.45e-08    7.50e-09
       Total_Liabilities_A |   1.69e-08   5.15e-09     3.28   0.003     6.30e-09    2.75e-08
                     ROE_A |  -.5932689     .63963    -0.93   0.362    -1.908047    .7215094
                     ROA_A |   7.931393   7.675989     1.03   0.311    -7.846828    23.70961
         Debt_Securities_A |  -1.40e-07   9.69e-07    -0.14   0.887    -2.13e-06    1.85e-06
            Total_Assets_A |  -1.41e-08   3.31e-09    -4.25   0.000    -2.09e-08   -7.26e-09
                           |
         new_referencearea |
             BE (Belgium)  |  -7.546097   1.336614    -5.65   0.000    -10.29355   -4.798648
            BG (Bulgaria)  |   13.71833    4.52217     3.03   0.005     4.422876    23.01378
              CY (Cyprus)  |  -14.18068   1.913676    -7.41   0.000     -18.1143   -10.24706
      CZ (Czech Republic)  |  -.4291531   3.955256    -0.11   0.914    -8.559298    7.700992
             DE (Germany)  |   19.77372    14.2343     1.39   0.177    -9.485312    49.03275
             DK (Denmark)  |   5.143016   1.794059     2.87   0.008     1.455275    8.830757
             EE (Estonia)  |  -17.55299   3.769838    -4.66   0.000      -25.302   -9.803979
               ES (Spain)  |  -7.536704   6.426752    -1.17   0.252    -20.74708    5.673674
             FI (Finland)  |  -25.23638   1.295194   -19.48   0.000    -27.89869   -22.57407
              FR (France)  |  -.4810713   13.30886    -0.04   0.971    -27.83782    26.87568
      GB (United Kingdom)  |    7.91004   10.90463     0.73   0.475    -14.50475    30.32483
              GR (Greece)  |  -6.705464    1.67177    -4.01   0.000    -10.14184   -3.269091
             HR (Croatia)  |   28.96351   4.162181     6.96   0.000     20.40803      37.519
             HU (Hungary)  |   32.64777   5.624505     5.80   0.000     21.08643     44.2091
             IE (Ireland)  |  -18.67331    1.94267    -9.61   0.000    -22.66652   -14.68009
               IT (Italy)  |  -3.835743   4.010677    -0.96   0.348    -12.07981    4.408321
          LU (Luxembourg)  |  -21.76652    2.13112   -10.21   0.000     -26.1471   -17.38594
              LV (Latvia)  |  -23.35435   2.372446    -9.84   0.000    -28.23098   -18.47772
               MT (Malta)  |  -17.25479   2.264214    -7.62   0.000    -21.90895   -12.60063
         NL (Netherlands)  |  -20.52507   4.889536    -4.20   0.000    -30.57566   -10.47449
              PL (Poland)  |   21.50341   3.829923     5.61   0.000     13.63089    29.37593
            PT (Portugal)  |  -8.446582   1.725781    -4.89   0.000    -11.99397   -4.899189
             RO (Romania)  |   16.57896   4.650007     3.57   0.001     7.020738    26.13719
              SE (Sweden)  |          0  (omitted)
            SI (Slovenia)  |   1.911455    2.98924     0.64   0.528    -4.233016    8.055926
            SK (Slovakia)  |   2.865127   2.410635     1.19   0.245    -2.090004    7.820259
                           |
                      Time |
                    20269  |   .3252497   .5347412     0.61   0.548    -.7739265    1.424426
                    20361  |   .7219141   .8502811     0.85   0.404    -1.025864    2.469692
                    20453  |   1.384325   .9931123     1.39   0.175    -.6570464    3.425697
                    20544  |   1.352453   1.319414     1.03   0.315    -1.359641    4.064547
                    20635  |   1.510416   1.670778     0.90   0.374    -1.923918     4.94475
                    20727  |   1.122131   1.624416     0.69   0.496    -2.216904    4.461165
                    20819  |   1.214823   1.790773     0.68   0.504    -2.466163     4.89581
                    20909  |   1.783342   1.669986     1.07   0.295    -1.649363    5.216047
                    21000  |   1.864331   1.847858     1.01   0.322    -1.933996    5.662657
                    21092  |    2.21565   1.870559     1.18   0.247    -1.629339    6.060639
                    21184  |   .0140313   2.475975     0.01   0.996    -5.075408     5.10347
                    21274  |   2.175613   2.479799     0.88   0.388    -2.921686    7.272912
                    21365  |   .9468819   2.349925     0.40   0.690    -3.883458    5.777222
                    21457  |   1.095093   2.686052     0.41   0.687    -4.426166    6.616352
                    21549  |   1.154285   2.819219     0.41   0.686    -4.640704    6.949273
                    21639  |   2.131911   2.479525     0.86   0.398    -2.964825    7.228647
                    21730  |   1.833592   2.578084     0.71   0.483    -3.465735    7.132918
                    21822  |   1.580083   2.568156     0.62   0.544    -3.698837    6.859003
                           |
                     _cons |   27.77395    4.20707     6.60   0.000     19.12619     36.4217
--------------------------------------------------------------------------------------------

.

Thank you so much for your help.

Best regards,
Lydia Gad

[Carlo Lazzaro]

Lydia:
-xtreg- R-sq between and pooled OLS are different beasts.
I'm not clear with the measure of time in your regression; that said you can test its joint statistical significance via:

testparm(i.Time)

[Chris Boudreaux]

Lydia,

Bear in mind what I said previously. You still have fixed effects (i.e., the new_referencearea dummies) in your model so that is actually a fixed effects model and not a pooled one. If you want pooled, you have to omit the new_referencearea dummies. Your parameter estimates are almost identical as before, with the exception that you now include year fixed effects in your second model.

Regarding your interaction, it would be better if you allowed Stata to do the interaction for you--this will specify all relevant terms. So instead of manually creating an interaction, you could do the following:

i.Period##i.treat

This syntax will create two dummies plus their interaction. This is what you want in most cases.

[Chris Boudreaux]

I don't think I was entirely clear. You can specify the interaction in the regression--not to generate an interaction by itself. Using your variables, it would look like this:

 
 xtreg I3631_Coverage_ratio_A i.Period##i.treat A1100_Loans_and_advances_A Total_Liabilities_A RO > E_A  ROA_A  Debt_Securities_A  Total_Assets_A, fe

[Lydia Gad]

Hi Carlo,
Thank you for your help.

I receive this error:

no such variables;
the specified varlist does not identify any testable coefficients
r(111);

I have included i.Time since my dependent variables might be affected by yearly shocks and I would like to control for that. It is also implemented in the literarture.


Hi Chris,
Thank you so much for your help.

I run the last command with the interaction term you specified and the output:

RO ambiguous abbreviation
r(111);

when I run xtreg, fe, it omits Period_treat, which is the interesting coefficient:

xtreg I3631_Coverage_ratio_A Period_treat A1100_Loans_and_advances_A Total_Liabilities_A ROE_A  ROA_A  Debt_Securities_A  Total_Assets_A, fe

note: Period_treat omitted because of collinearity

Fixed-effects (within) regression               Number of obs     =        439
Group variable: new_refere~a                    Number of groups  =         27

R-sq:                                           Obs per group:
     within  = 0.2018                                         min =          1
     between = 0.0105                                         avg =       16.3
     overall = 0.0095                                         max =         19

                                                F(6,406)          =      17.11
corr(u_i, Xb)  = -0.4651                        Prob > F          =     0.0000

--------------------------------------------------------------------------------------------
    I3631_Coverage_ratio_A |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
---------------------------+----------------------------------------------------------------
              Period_treat |          0  (omitted)
A1100_Loans_and_advances_A |  -1.52e-09   3.55e-09    -0.43   0.669    -8.51e-09    5.47e-09
       Total_Liabilities_A |   1.49e-08   3.75e-09     3.98   0.000     7.54e-09    2.23e-08
                     ROE_A |  -.6494135   .2313493    -2.81   0.005    -1.104205   -.1946216
                     ROA_A |    8.34393   2.401678     3.47   0.001     3.622653    13.06521
         Debt_Securities_A |  -2.16e-07   7.96e-07    -0.27   0.786    -1.78e-06    1.35e-06
            Total_Assets_A |  -1.32e-08   2.88e-09    -4.57   0.000    -1.88e-08   -7.50e-09
                     _cons |   44.55126   2.798249    15.92   0.000     39.05039    50.05213
---------------------------+----------------------------------------------------------------
                   sigma_u |  11.396829
                   sigma_e |  4.7123562
                       rho |  .85399623   (fraction of variance due to u_i)
--------------------------------------------------------------------------------------------
F test that all u_i=0: F(26, 406) = 58.27                    Prob > F = 0.0000

I have run a hausman to make sure that FE model is the way to go, and also in the literature at hand FE model has been implemented:

. xtreg I3631_Coverage_ratio_A A1100_Loans_and_advances_A Total_Liabilities_A ROE_A  ROA_A  D
> ebt_Securities_A  Total_Assets_A, fe

Fixed-effects (within) regression               Number of obs     =        463
Group variable: new_refere~a                    Number of groups  =         27

R-sq:                                           Obs per group:
     within  = 0.1979                                         min =          1
     between = 0.0082                                         avg =       17.1
     overall = 0.0086                                         max =         20

                                                F(6,430)          =      17.68
corr(u_i, Xb)  = -0.4727                        Prob > F          =     0.0000

--------------------------------------------------------------------------------------------
    I3631_Coverage_ratio_A |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
---------------------------+----------------------------------------------------------------
A1100_Loans_and_advances_A |  -2.46e-09   3.24e-09    -0.76   0.448    -8.84e-09    3.91e-09
       Total_Liabilities_A |   1.62e-08   3.48e-09     4.64   0.000     9.32e-09    2.30e-08
                     ROE_A |  -.4944818   .2064842    -2.39   0.017    -.9003256   -.0886379
                     ROA_A |    7.17511   2.207866     3.25   0.001     2.835558    11.51466
         Debt_Securities_A |   8.65e-08   7.36e-07     0.12   0.907    -1.36e-06    1.53e-06
            Total_Assets_A |  -1.36e-08   2.74e-09    -4.95   0.000    -1.89e-08   -8.18e-09
                     _cons |   44.02287   2.662234    16.54   0.000     38.79026    49.25548
---------------------------+----------------------------------------------------------------
                   sigma_u |   11.28489
                   sigma_e |  4.7635099
                       rho |  .84876659   (fraction of variance due to u_i)
--------------------------------------------------------------------------------------------
F test that all u_i=0: F(26, 430) = 58.47                    Prob > F = 0.0000

. estimate store fe

. xtreg I3631_Coverage_ratio_A  A1100_Loans_and_advances_A Total_Liabilities_A ROE_A  ROA_A  
> Debt_Securities_A  Total_Assets_A, re

Random-effects GLS regression                   Number of obs     =        463
Group variable: new_refere~a                    Number of groups  =         27

R-sq:                                           Obs per group:
     within  = 0.1966                                         min =          1
     between = 0.0086                                         avg =       17.1
     overall = 0.0121                                         max =         20

                                                Wald chi2(6)      =      99.71
corr(u_i, X)   = 0 (assumed)                    Prob > chi2       =     0.0000

--------------------------------------------------------------------------------------------
    I3631_Coverage_ratio_A |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
---------------------------+----------------------------------------------------------------
A1100_Loans_and_advances_A |  -1.07e-09   2.74e-09    -0.39   0.698    -6.44e-09    4.31e-09
       Total_Liabilities_A |   1.39e-08   3.04e-09     4.56   0.000     7.90e-09    1.98e-08
                     ROE_A |  -.5834656    .206057    -2.83   0.005    -.9873298   -.1796013
                     ROA_A |   8.096589   2.200684     3.68   0.000     3.783328    12.40985
         Debt_Securities_A |   6.55e-09   7.37e-07     0.01   0.993    -1.44e-06    1.45e-06
            Total_Assets_A |  -1.15e-08   1.74e-09    -6.59   0.000    -1.49e-08   -8.07e-09
                     _cons |    42.8582   2.088609    20.52   0.000     38.76461     46.9518
---------------------------+----------------------------------------------------------------
                   sigma_u |  9.1938164
                   sigma_e |  4.7635099
                       rho |   .7883641   (fraction of variance due to u_i)
--------------------------------------------------------------------------------------------

. estimate store re

. hausman fe re

Note: the rank of the differenced variance matrix (2) does not equal the number of
        coefficients being tested (6); be sure this is what you expect, or there may be
        problems computing the test.  Examine the output of your estimators for anything
        unexpected and possibly consider scaling your variables so that the coefficients are
        on a similar scale.

                 ---- Coefficients ----
             |      (b)          (B)            (b-B)     sqrt(diag(V_b-V_B))
             |       fe           re         Difference          S.E.
-------------+----------------------------------------------------------------
A1100_Loan~A |   -2.46e-09    -1.07e-09       -1.40e-09        1.73e-09
Total_Liab~A |    1.62e-08     1.39e-08        2.31e-09        1.70e-09
       ROE_A |   -.4944818    -.5834656        .0889838        .0132755
       ROA_A |     7.17511     8.096589       -.9214792        .1779405
Debt_Secur~A |    8.65e-08     6.55e-09        8.00e-08               .
Total_Asse~A |   -1.36e-08    -1.15e-08       -2.08e-09        2.12e-09
------------------------------------------------------------------------------
                           b = consistent under Ho and Ha; obtained from xtreg
            B = inconsistent under Ha, efficient under Ho; obtained from xtreg

    Test:  Ho:  difference in coefficients not systematic

                  chi2(2) = (b-B)'[(V_b-V_B)^(-1)](b-B)
                          =       22.03
                Prob>chi2 =      0.0000
                (V_b-V_B is not positive definite)

.

I am thinking of reg X Period_treat Ys..... i.new_referencearea i.Time and by this will be implementing a FE model and keeping the Period_treat variable. What do you think of this?

I am so grateful for your time and effort.


Best regards,
Lydia Gad

[Chris Boudreaux]

Sorry that was a typo on my end from when I copy and pasted your code. Use this instead:

 
  xtreg I3631_Coverage_ratio_A i.Period##i.treat A1100_Loans_and_advances_A Total_Liabilities_A ROE_A  ROA_A  Debt_Securities_A  Total_Assets_A, fe

It was saying RO is am ambiguous abbreviation due to the space between RO and E_A.

[Carlo Lazzaro]

Lydia:
works for me with this toy-example:

. use "https://www.stata-press.com/data/r16/nlswork.dta"
(National Longitudinal Survey.  Young Women 14-26 years of age in 1968)

. xtreg ln_wage i.race i.year

Random-effects GLS regression                   Number of obs     =     28,534
Group variable: idcode                          Number of groups  =      4,711

R-sq:                                           Obs per group:
     within  = 0.1058                                         min =          1
     between = 0.0975                                         avg =        6.1
     overall = 0.0907                                         max =         15

                                                Wald chi2(16)     =    3310.99
corr(u_i, X)   = 0 (assumed)                    Prob > chi2       =     0.0000

------------------------------------------------------------------------------
     ln_wage |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
        race |
      black  |  -.1279049   .0128852    -9.93   0.000    -.1531594   -.1026504
      other  |   .0900274   .0537654     1.67   0.094     -.015351    .1954057
             |
        year |
         69  |    .085839   .0123757     6.94   0.000      .061583     .110095
         70  |   .0702121   .0115599     6.07   0.000     .0475552     .092869
         71  |   .1200453   .0114213    10.51   0.000     .0976599    .1424307
         72  |   .1329282   .0117421    11.32   0.000     .1099142    .1559422
         73  |   .1480458   .0113851    13.00   0.000     .1257314    .1703603
         75  |   .1615023   .0112522    14.35   0.000     .1394483    .1835562
         77  |   .2215681   .0112623    19.67   0.000     .1994945    .2436418
         78  |   .2603374   .0115062    22.63   0.000     .2377857    .2828891
         80  |   .2685209    .011652    23.04   0.000     .2456834    .2913585
         82  |   .2858463   .0113927    25.09   0.000      .263517    .3081756
         83  |   .3132819    .011535    27.16   0.000     .2906736    .3358902
         85  |   .3656784    .011431    31.99   0.000      .343274    .3880827
         87  |   .3814745   .0113629    33.57   0.000     .3592036    .4037454
         88  |   .4370321   .0113003    38.67   0.000     .4148839    .4591802
             |
       _cons |     1.4612   .0109536   133.40   0.000     1.439732    1.482669
-------------+----------------------------------------------------------------
     sigma_u |  .36492114
     sigma_e |  .30294584
         rho |  .59200363   (fraction of variance due to u_i)
------------------------------------------------------------------------------

. testparm(i.year)

 ( 1)  69.year = 0
 ( 2)  70.year = 0
 ( 3)  71.year = 0
 ( 4)  72.year = 0
 ( 5)  73.year = 0
 ( 6)  75.year = 0
 ( 7)  77.year = 0
 ( 8)  78.year = 0
 ( 9)  80.year = 0
 (10)  82.year = 0
 (11)  83.year = 0
 (12)  85.year = 0
 (13)  87.year = 0
 (14)  88.year = 0

           chi2( 14) = 3202.52
         Prob > chi2 =    0.0000

.

[Carlo Lazzaro]

Lydia:
as your -hausman- outcome is limping, you can constrast it via the community-contributed programme -xtoverid- (just type -search xtoverid- from within Stata to spot and install it).
Please note the -xtoverid- needs the -re- specification only to work properly, as the null is, non technically speaking, that -re- is the way to go (conversely, if the -xtoverid- outcome reaches statistical significance, you should switch to -fe-; if there's no evidence of a group-wise effect, -xtoverid- gives back a warning message, that advises you to go pooled OLS).
In your case, it should be:

. xtreg I3631_Coverage_ratio_A  A1100_Loans_and_advances_A Total_Liabilities_A ROE_A  ROA_A  Debt_Securities_A  Total_Assets_A, re
xtoverid