Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Interpreting coefficient of an independent variable that is a proportion

    Hi, my dissertation looks at a panel dataset of N=966, n=69 firms across T=14 years.
    Independent variable: WOB (proportion of women on board: in decimals)
    2 dependent variables: ROA (return on asset: in decimals), TQ (Tobin's Q measure: range between <1 to >1)

    I am unsure on how to interpret the effect of WOB on each of the dependent variables:-
    (i) ROA - Regression table 1
    ~ Am I right to infer: A 1% point (unit) change in WOB leads to a 7.34% point (unit) increase in ROA?

    (ii) TQ - Regression table 2
    ~Having difficulty with this as the coefficient is very big at 7.78, compared to ROA. Am I perhaps doing anything wrong?
    ~How can I interpret this coefficient using the idea of a 1% point (unit) change in WOB?



    For your information, the formulas of each variable is:
    WOB = number of women directors / total number of directors
    ROA = net income (loss) / book value of total assets
    TQ = (market value of stock + book value of debt) / book value of total assets

    Control variables: BS = board size, FS = firm size (natural log), FA = firm age (natural log)



    Code:
    xtset COMPANY YEAR, yearly
    Code:
    . xtsum WOB ROA TQ BS FS FA
    
    Variable         |      Mean   Std. Dev.       Min        Max |    Observations
    -----------------+--------------------------------------------+----------------
    WOB      overall |  .1495199   .1073334          0         .6 |     N =     966
             between |             .0772878          0   .3298424 |     n =      69
             within  |             .0750168  -.0692114   .4505403 |     T =      14
                     |                                            |
    ROA      overall |  .0673593   .1032169  -1.104439   .6541691 |     N =     966
             between |                .0603  -.0711783   .2470477 |     n =      69
             within  |             .0840633  -1.048663   .6061372 |     T =      14
                     |                                            |
    TQ       overall |  3.073908   2.173385   .5880593   26.99235 |     N =     966
             between |             1.494371   1.214629   8.949094 |     n =      69
             within  |             1.587621  -3.830402   21.11716 |     T =      14
                     |                                            |
    BS       overall |  9.257764   2.182079          3         18 |     N =     966
             between |             1.880398   4.714286   15.14286 |     n =      69
             within  |             1.128364   4.114907   14.40062 |     T =      14
                     |                                            |
    FS       overall |  8.414718   1.936693   3.070422   13.58957 |     N =     966
             between |             1.862711   5.032185   13.16478 |     n =      69
             within  |             .5725632   6.335153   10.44355 |     T =      14
                     |                                            |
    FA       overall |  3.138765   .5719432   .6931472    4.85203 |     N =     966
             between |             .5307901   1.992805   4.799363 |     n =      69
             within  |             .2217574   1.839107    3.85401 |     T =      14



    Code:
    reg ROA WOB BS FS FA, robust
    
    Linear regression                               Number of obs     =        966
                                                    F(4, 961)         =       7.36
                                                    Prob > F          =     0.0000
                                                    R-squared         =     0.0216
                                                    Root MSE          =     .10231
    
    ------------------------------------------------------------------------------
                 |               Robust
             ROA |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
    -------------+----------------------------------------------------------------
             WOB |   .0733581   .0377618     1.94   0.052    -.0007471    .1474632
              BS |  -.0079655   .0019387    -4.11   0.000    -.0117701    -.004161
              FS |    .007457   .0025467     2.93   0.003     .0024593    .0124546
              FA |   .0031653    .004069     0.78   0.437    -.0048199    .0111506
           _cons |   .0574504   .0201767     2.85   0.005     .0178549    .0970459
    ------------------------------------------------------------------------------
    Code:
     reg TQ WOB BS FS FA, robust
    
    Linear regression                               Number of obs     =        966
                                                    F(4, 961)         =      37.66
                                                    Prob > F          =     0.0000
                                                    R-squared         =     0.1679
                                                    Root MSE          =     1.9867
    
    ------------------------------------------------------------------------------
                 |               Robust
              TQ |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
    -------------+----------------------------------------------------------------
             WOB |    7.78169   .8359386     9.31   0.000     6.141214    9.422165
              BS |  -.2209036   .0284446    -7.77   0.000    -.2767243   -.1650829
              FS |  -.2787995   .0502885    -5.54   0.000    -.3774874   -.1801115
              FA |   -.026936   .0825185    -0.33   0.744    -.1888733    .1350014
           _cons |   6.386028   .4221714    15.13   0.000     5.557544    7.214512
    ------------------------------------------------------------------------------
    Thank you.

  • #2
    Nur:
    welcome to this forum.
    You state you have a panel dataset and you -xtset- it.
    Then you went -regress-, but -regress- does not need -xtset- to work properly.
    More substantively:
    1) -regress- is not the first choice tool to consider for panel data regression with continuous regerssand (see -xtreg- instead);
    2) at the top of that, your -regress- code ignores the panel data structure of your dataset, as it considers all the observations as they were independent, whereas they are not due to their panel data structure.

    As an aside, I take the liberty to point you to any decent textbook on panel data econometrics.
    Kind regards,
    Carlo
    (Stata 18.0 SE)

    Comment


    • #3
      Hi Carlo,
      Thank you for the warm welcome.

      To share with you, I decided to take up 3 models for my dataset:
      1) Simple OLS: for the purpose to show it contains potential issues & isn't suitable for panel data (via xttest0), which motivates me to my next model
      2) FE (having tested for Hausman Wu test)
      3) System GMM (inspired by literature papers of the similar topic & also after correcting for serial correlation)

      The results for each model are as following:
      (i) Dependent Variable: ROA
      Code:
      reg ROA WOB BS FS FA, robust
      
      Linear regression                               Number of obs     =        966
                                                      F(4, 961)         =       7.36
                                                      Prob > F          =     0.0000
                                                      R-squared         =     0.0216
                                                      Root MSE          =     .10231
      
      ------------------------------------------------------------------------------
                   |               Robust
               ROA |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
      -------------+----------------------------------------------------------------
               WOB |   .0733581   .0377618     1.94   0.052    -.0007471    .1474632
                BS |  -.0079655   .0019387    -4.11   0.000    -.0117701    -.004161
                FS |    .007457   .0025467     2.93   0.003     .0024593    .0124546
                FA |   .0031653    .004069     0.78   0.437    -.0048199    .0111506
             _cons |   .0574504   .0201767     2.85   0.005     .0178549    .0970459
      ------------------------------------------------------------------------------
      Code:
       xtset COMPANY YEAR, yearly
             panel variable:  COMPANY (strongly balanced)
              time variable:  YEAR, 2007 to 2020
                      delta:  1 year
      
      . 
      end of do-file
      
      . do "/var/folders/4q/ss0sttk90dg6_5jph2lzjxj00000gn/T//SD00373.000000"
      
      . xtreg ROA WOB BS FS FA, fe vce(robust)
      
      Fixed-effects (within) regression               Number of obs     =        966
      Group variable: COMPANY                         Number of groups  =         69
      
      R-sq:                                           Obs per group:
           within  = 0.0616                                         min =         14
           between = 0.0265                                         avg =       14.0
           overall = 0.0018                                         max =         14
      
                                                      F(4,68)           =       7.11
      corr(u_i, Xb)  = -0.8156                        Prob > F          =     0.0001
      
                                     (Std. Err. adjusted for 69 clusters in COMPANY)
      ------------------------------------------------------------------------------
                   |               Robust
               ROA |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
      -------------+----------------------------------------------------------------
               WOB |   .1037013   .0547898     1.89   0.063      -.00563    .2130327
                BS |  -.0036958    .003398    -1.09   0.281    -.0104764    .0030849
                FS |  -.0443331   .0113535    -3.90   0.000    -.0669886   -.0216776
                FA |   .0675684   .0424617     1.59   0.116    -.0171626    .1522994
             _cons |   .2470374   .0829475     2.98   0.004     .0815182    .4125565
      -------------+----------------------------------------------------------------
           sigma_u |  .11009321
           sigma_e |   .0846534
               rho |  .62843864   (fraction of variance due to u_i)
      ------------------------------------------------------------------------------
      
      .
      Code:
       . xtdpdsys ROA WOB BS FS FA, lags(1) twostep vce(robust) artests(2)
      
      System dynamic panel-data estimation            Number of obs     =        897
      Group variable: COMPANY                         Number of groups  =         69
      Time variable: YEAR
                                                      Obs per group:
                                                                    min =         13
                                                                    avg =         13
                                                                    max =         13
      
      Number of instruments =     95                  Wald chi2(5)      =      15.99
                                                      Prob > chi2       =     0.0069
      Two-step results
      ------------------------------------------------------------------------------
                   |              WC-Robust
               ROA |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
      -------------+----------------------------------------------------------------
               ROA |
               L1. |   .2201171   .0818828     2.69   0.007     .0596298    .3806045
                   |
               WOB |  -.0269527   .0727193    -0.37   0.711      -.16948    .1155746
                BS |  -.0041597   .0044098    -0.94   0.346    -.0128027    .0044833
                FS |  -.0084459   .0165288    -0.51   0.609    -.0408418    .0239501
                FA |   .0236815   .0436667     0.54   0.588    -.0619036    .1092667
             _cons |   .0919001   .0534662     1.72   0.086    -.0128917    .1966919
      ------------------------------------------------------------------------------
      Instruments for differenced equation
              GMM-type: L(2/.).ROA
              Standard: D.WOB D.BS D.FS D.FA
      Instruments for level equation
              GMM-type: LD.ROA
              Standard: _cons
      (ii) Dependent variable: TQ
      Code:
      . reg TQ WOB BS FS FA, robust
      
      Linear regression                               Number of obs     =        966
                                                      F(4, 961)         =      37.66
                                                      Prob > F          =     0.0000
                                                      R-squared         =     0.1679
                                                      Root MSE          =     1.9867
      
      ------------------------------------------------------------------------------
                   |               Robust
                TQ |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
      -------------+----------------------------------------------------------------
               WOB |    7.78169   .8359386     9.31   0.000     6.141214    9.422165
                BS |  -.2209036   .0284446    -7.77   0.000    -.2767243   -.1650829
                FS |  -.2787995   .0502885    -5.54   0.000    -.3774874   -.1801115
                FA |   -.026936   .0825185    -0.33   0.744    -.1888733    .1350014
             _cons |   6.386028   .4221714    15.13   0.000     5.557544    7.214512
      ------------------------------------------------------------------------------
      Code:
      . xtset COMPANY YEAR, yearly
             panel variable:  COMPANY (strongly balanced)
              time variable:  YEAR, 2007 to 2020
                      delta:  1 year
      
      . xtreg TQ WOB BS FS FA, fe vce(robust)
      
      Fixed-effects (within) regression               Number of obs     =        966
      Group variable: COMPANY                         Number of groups  =         69
      
      R-sq:                                           Obs per group:
           within  = 0.1545                                         min =         14
           between = 0.0207                                         avg =       14.0
           overall = 0.0355                                         max =         14
      
                                                      F(4,68)           =       6.53
      corr(u_i, Xb)  = -0.6843                        Prob > F          =     0.0002
      
                                     (Std. Err. adjusted for 69 clusters in COMPANY)
      ------------------------------------------------------------------------------
                   |               Robust
                TQ |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
      -------------+----------------------------------------------------------------
               WOB |   4.715363   1.333387     3.54   0.001      2.05463    7.376095
                BS |  -.0355468   .0504514    -0.70   0.483     -.136221    .0651275
                FS |  -.8881094    .346856    -2.56   0.013     -1.58025    -.195969
                FA |   2.615657    1.17764     2.22   0.030     .2657124    4.965602
             _cons |   1.961209   1.629345     1.20   0.233    -1.290098    5.212517
      -------------+----------------------------------------------------------------
           sigma_u |  2.1499092
           sigma_e |  1.5175036
               rho |  .66745971   (fraction of variance due to u_i)
      ------------------------------------------------------------------------------
      Code:
      . xtdpdsys TQ WOB BS FS FA, lags(1) twostep vce(robust) artests(2)
      
      System dynamic panel-data estimation            Number of obs     =        897
      Group variable: COMPANY                         Number of groups  =         69
      Time variable: YEAR
                                                      Obs per group:
                                                                    min =         13
                                                                    avg =         13
                                                                    max =         13
      
      Number of instruments =     95                  Wald chi2(5)      =     189.05
                                                      Prob > chi2       =     0.0000
      Two-step results
      ------------------------------------------------------------------------------
                   |              WC-Robust
                TQ |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
      -------------+----------------------------------------------------------------
                TQ |
               L1. |    .537309   .0753259     7.13   0.000     .3896731     .684945
                   |
               WOB |   4.836092   4.473973     1.08   0.280    -3.932734    13.60492
                BS |  -.1939967   .1392793    -1.39   0.164    -.4669791    .0789858
                FS |  -.7659641    .668216    -1.15   0.252    -2.075643    .5437152
                FA |   2.199644    1.25319     1.76   0.079    -.2565621    4.655851
             _cons |   2.022582   2.592772     0.78   0.435    -3.059159    7.104322
      ------------------------------------------------------------------------------
      Instruments for differenced equation
              GMM-type: L(2/.).TQ
              Standard: D.WOB D.BS D.FS D.FA
      Instruments for level equation
              GMM-type: LD.TQ
              Standard: _cons

      Questions:
      ~Would you advise I stick with 2 models: FE & System GMM or can I proceed with OLS as the basic model?
      ~How can I interpret my results?


      Thank you.

      Comment


      • #4
        Kindly ignore this reply and the next. I posted the same reply twice as it had trouble uploading earlier.
        Last edited by Nur Batrisya; 25 Jan 2022, 07:43.

        Comment


        • #5
          -
          Last edited by Nur Batrisya; 25 Jan 2022, 07:38.

          Comment


          • #6
            Nur:
            1) your -regress- model is not correctly coded, as you used -robust- vs -vce(cluster clusterid)- standard errors.
            Please note that, unlike -xtreg-, -regress- options for non-default standard errors do different jobs: -robust- takes heteroskedastcity into account, whereas -vce(cluster clusterid)- takes serial correlation into acoount (when you have both these nuisances, you can go -newey- or, for simpler error modelization, -vce(cluster clusterid)-);
            2) -xtreg,fe- is the first step to take when dealing with a panel dataset with a continuous regressand. From the high negative correlations (-0.8156 and -0.6843, respectively) the evidence of a panel-wise effect seems to be confirmed.
            However, looking at the limited within-Rsq (the R coefficient to be considered with -fe- specification) I would test if the functional form of the regressand is correctly specified (the procedure is similar to the one reported in the -linktest- entry, Stata .pdf manual);
            3) -gmm- estimator is extremely flexible but requires a very good command of econometrics to be applied correctly. I would advise you to compare the -fe- vs the -re- specification via the community-contributed module -xtoverid-.
            Kind regards,
            Carlo
            (Stata 18.0 SE)

            Comment


            • #7
              Carlo:
              Thank you for the suggestions.
              I will look into them in detail and reach out again if there's further clarification/advice needed.

              Comment


              • #8
                Carlo:
                1) Kindly check if this is correct for robust-cluster standard error:

                (i) Dependent variable: ROA
                Code:
                 xtreg ROA WOB BS FS FA, fe vce(cluster COMPANY)
                
                Fixed-effects (within) regression               Number of obs     =        966
                Group variable: COMPANY                         Number of groups  =         69
                
                R-sq:                                           Obs per group:
                     within  = 0.0616                                         min =         14
                     between = 0.0265                                         avg =       14.0
                     overall = 0.0018                                         max =         14
                
                                                                F(4,68)           =       7.11
                corr(u_i, Xb)  = -0.8156                        Prob > F          =     0.0001
                
                                               (Std. Err. adjusted for 69 clusters in COMPANY)
                ------------------------------------------------------------------------------
                             |               Robust
                         ROA |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
                -------------+----------------------------------------------------------------
                         WOB |   .1037013   .0547898     1.89   0.063      -.00563    .2130327
                          BS |  -.0036958    .003398    -1.09   0.281    -.0104764    .0030849
                          FS |  -.0443331   .0113535    -3.90   0.000    -.0669886   -.0216776
                          FA |   .0675684   .0424617     1.59   0.116    -.0171626    .1522994
                       _cons |   .2470374   .0829475     2.98   0.004     .0815182    .4125565
                -------------+----------------------------------------------------------------
                     sigma_u |  .11009321
                     sigma_e |   .0846534
                         rho |  .62843864   (fraction of variance due to u_i)
                ------------------------------------------------------------------------------


                Similarly, for my other outcome variable
                (ii) Dependent variable: TQ

                Code:
                xtreg TQ WOB BS FS FA, fe vce(cluster COMPANY)
                
                Fixed-effects (within) regression               Number of obs     =        966
                Group variable: COMPANY                         Number of groups  =         69
                
                R-sq:                                           Obs per group:
                     within  = 0.1545                                         min =         14
                     between = 0.0207                                         avg =       14.0
                     overall = 0.0355                                         max =         14
                
                                                                F(4,68)           =       6.53
                corr(u_i, Xb)  = -0.6843                        Prob > F          =     0.0002
                
                                               (Std. Err. adjusted for 69 clusters in COMPANY)
                ------------------------------------------------------------------------------
                             |               Robust
                          TQ |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
                -------------+----------------------------------------------------------------
                         WOB |   4.715363   1.333387     3.54   0.001      2.05463    7.376095
                          BS |  -.0355468   .0504514    -0.70   0.483     -.136221    .0651275
                          FS |  -.8881094    .346856    -2.56   0.013     -1.58025    -.195969
                          FA |   2.615657    1.17764     2.22   0.030     .2657124    4.965602
                       _cons |   1.961209   1.629345     1.20   0.233    -1.290098    5.212517
                -------------+----------------------------------------------------------------
                     sigma_u |  2.1499092
                     sigma_e |  1.5175036
                         rho |  .66745971   (fraction of variance due to u_i)
                ------------------------------------------------------------------------------

                Comment


                • #9
                  Nur: Usually results are not considered reliable unless you account for systematic changes over time. The so-called two-way fixed effects estimator includes dummy variables for the different years, so add i.year to both specifications. If WOB increases by 0.10 (a 10 percentage point increase) then ROA increases by about 0.01, so one percentage point. If WOB increases by 0.10 the TQ increases by about 0.47, and you can compare this with its standard deviation to get an idea of how large the effect is. N = 69 and T = 14 can be a bit dicey for clustering but it's probably okay.

                  Comment


                  • #10
                    Carlo:
                    2)
                    ~ What is panel-wise effect? Why does high negative correlations (-0.8156 and -0.6843) in the FE models indicate there is panel-wise effect? Is there any literature I can read to further enhance my understanding?
                    ~ What about the within-R^2 from the FE models (0.0616 and 0.1545) that made you want to check on the specification / functional form? Is it because they are small? Again, any literature I can read that discusses this?
                    ~ Nevertheless, I ran a RESET test manually for each dep. variable (below). As seen, both are misspecified. What does this imply & how can I solve it?

                    (i) Dependent variable: ROA
                    Code:
                    xtreg ROA WOB BS FS FA, fe vce(cluster COMPANY)
                    
                    Fixed-effects (within) regression               Number of obs     =        966
                    Group variable: COMPANY                         Number of groups  =         69
                    
                    R-sq:                                           Obs per group:
                         within  = 0.0616                                         min =         14
                         between = 0.0265                                         avg =       14.0
                         overall = 0.0018                                         max =         14
                    
                                                                    F(4,68)           =       7.11
                    corr(u_i, Xb)  = -0.8156                        Prob > F          =     0.0001
                    
                                                   (Std. Err. adjusted for 69 clusters in COMPANY)
                    ------------------------------------------------------------------------------
                                 |               Robust
                             ROA |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
                    -------------+----------------------------------------------------------------
                             WOB |   .1037013   .0547898     1.89   0.063      -.00563    .2130327
                              BS |  -.0036958    .003398    -1.09   0.281    -.0104764    .0030849
                              FS |  -.0443331   .0113535    -3.90   0.000    -.0669886   -.0216776
                              FA |   .0675684   .0424617     1.59   0.116    -.0171626    .1522994
                           _cons |   .2470374   .0829475     2.98   0.004     .0815182    .4125565
                    -------------+----------------------------------------------------------------
                         sigma_u |  .11009321
                         sigma_e |   .0846534
                             rho |  .62843864   (fraction of variance due to u_i)
                    ------------------------------------------------------------------------------
                    
                    . predict yhat1 if e(sample), xb
                    
                    . gen yhat2 = yhat1^2
                    
                    . gen yhat3 = yhat1^3
                    
                    . xtreg ROA yhat1 yhat2 yhat3, fe vce(cluster COMPANY)
                    
                    Fixed-effects (within) regression               Number of obs     =        966
                    Group variable: COMPANY                         Number of groups  =         69
                    
                    R-sq:                                           Obs per group:
                         within  = 0.0860                                         min =         14
                         between = 0.0214                                         avg =       14.0
                         overall = 0.0004                                         max =         14
                    
                                                                    F(3,68)           =      17.23
                    corr(u_i, Xb)  = -0.8354                        Prob > F          =     0.0000
                    
                                                   (Std. Err. adjusted for 69 clusters in COMPANY)
                    ------------------------------------------------------------------------------
                                 |               Robust
                             ROA |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
                    -------------+----------------------------------------------------------------
                           yhat1 |   .7047492   .2068873     3.41   0.001     .2919119    1.117586
                           yhat2 |   4.496776   1.691386     2.66   0.010     1.121669    7.871884
                           yhat3 |  -6.184528   5.120772    -1.21   0.231    -16.40287    4.033813
                           _cons |  -.0224359   .0214977    -1.04   0.300     -.065334    .0204622
                    -------------+----------------------------------------------------------------
                         sigma_u |  .11858146
                         sigma_e |  .08349592
                             rho |  .66854346   (fraction of variance due to u_i)
                    ------------------------------------------------------------------------------
                    
                    . test yhat2 yhat3
                    
                     ( 1)  yhat2 = 0
                     ( 2)  yhat3 = 0
                    
                           F(  2,    68) =    9.03
                                Prob > F =    0.0003
                    (ii) Dependent variable: TQ
                    Code:
                    xtreg TQ WOB BS FS FA, fe vce(cluster COMPANY)
                    
                    Fixed-effects (within) regression               Number of obs     =        966
                    Group variable: COMPANY                         Number of groups  =         69
                    
                    R-sq:                                           Obs per group:
                         within  = 0.1545                                         min =         14
                         between = 0.0207                                         avg =       14.0
                         overall = 0.0355                                         max =         14
                    
                                                                    F(4,68)           =       6.53
                    corr(u_i, Xb)  = -0.6843                        Prob > F          =     0.0002
                    
                                                   (Std. Err. adjusted for 69 clusters in COMPANY)
                    ------------------------------------------------------------------------------
                                 |               Robust
                              TQ |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
                    -------------+----------------------------------------------------------------
                             WOB |   4.715363   1.333387     3.54   0.001      2.05463    7.376095
                              BS |  -.0355468   .0504514    -0.70   0.483     -.136221    .0651275
                              FS |  -.8881094    .346856    -2.56   0.013     -1.58025    -.195969
                              FA |   2.615657    1.17764     2.22   0.030     .2657124    4.965602
                           _cons |   1.961209   1.629345     1.20   0.233    -1.290098    5.212517
                    -------------+----------------------------------------------------------------
                         sigma_u |  2.1499092
                         sigma_e |  1.5175036
                             rho |  .66745971   (fraction of variance due to u_i)
                    ------------------------------------------------------------------------------
                    
                    . predict yhat11 if e(sample), xb
                    
                    . gen yhat21 = yhat11^2
                    
                    . gen yhat31 = yhat11^3
                    
                    . xtreg TQ yhat11 yhat21 yhat31, fe vce(cluster COMPANY)
                    
                    Fixed-effects (within) regression               Number of obs     =        966
                    Group variable: COMPANY                         Number of groups  =         69
                    
                    R-sq:                                           Obs per group:
                         within  = 0.2104                                         min =         14
                         between = 0.0157                                         avg =       14.0
                         overall = 0.0361                                         max =         14
                    
                                                                    F(3,68)           =       7.02
                    corr(u_i, Xb)  = -0.7522                        Prob > F          =     0.0004
                    
                                                   (Std. Err. adjusted for 69 clusters in COMPANY)
                    ------------------------------------------------------------------------------
                                 |               Robust
                              TQ |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
                    -------------+----------------------------------------------------------------
                          yhat11 |   .2926819    .189139     1.55   0.126    -.0847391    .6701028
                          yhat21 |   .2644516   .0894592     2.96   0.004     .0859386    .4429647
                          yhat31 |   -.018888   .0100898    -1.87   0.066    -.0390219    .0012458
                           _cons |  -.1143682   .7025185    -0.16   0.871    -1.516222    1.287486
                    -------------+----------------------------------------------------------------
                         sigma_u |  2.4460105
                         sigma_e |  1.4656575
                             rho |  .73581104   (fraction of variance due to u_i)
                    ------------------------------------------------------------------------------
                    
                    . test yhat21 yhat31
                    
                     ( 1)  yhat21 = 0
                     ( 2)  yhat31 = 0
                    
                           F(  2,    68) =    5.14
                                Prob > F =    0.0083

                    Comment


                    • #11
                      Carlo:
                      3)
                      Step 1: I've run the Hausman test at the start (to decide between FE & RE), do I still need to run -xtoverid- ?
                      Step 2: Breusch-Pagan test for Heteroscedasticity (reject)
                      Step 3: -xtserial- for Serial Correlation (reject)
                      Step 4: Introduce robust-cluster s.e. in FE model --> solve heterosc. + serial correlation
                      Step 5: RESET test for Misspecification /Functional form (reject)
                      Step 6: What can I do?
                      Step 7: Should I test for panel unit root test?


                      Hausman test below:
                      Code:
                      . xtreg ROA WOB BS FS FA, fe
                      
                      Fixed-effects (within) regression               Number of obs     =        966
                      Group variable: COMPANY                         Number of groups  =         69
                      
                      R-sq:                                           Obs per group:
                           within  = 0.0616                                         min =         14
                           between = 0.0265                                         avg =       14.0
                           overall = 0.0018                                         max =         14
                      
                                                                      F(4,893)          =      14.65
                      corr(u_i, Xb)  = -0.8156                        Prob > F          =     0.0000
                      
                      ------------------------------------------------------------------------------
                               ROA |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
                      -------------+----------------------------------------------------------------
                               WOB |   .1037013   .0426983     2.43   0.015     .0199006    .1875021
                                BS |  -.0036958   .0026303    -1.41   0.160     -.008858    .0014665
                                FS |  -.0443331   .0067035    -6.61   0.000    -.0574895   -.0311766
                                FA |   .0675684   .0175433     3.85   0.000     .0331374    .1019994
                             _cons |   .2470374    .048278     5.12   0.000     .1522858    .3417889
                      -------------+----------------------------------------------------------------
                           sigma_u |  .11009321
                           sigma_e |   .0846534
                               rho |  .62843864   (fraction of variance due to u_i)
                      ------------------------------------------------------------------------------
                      F test that all u_i=0: F(68, 893) = 7.51                     Prob > F = 0.0000
                      
                      . estimate store fe
                      
                      . xtreg ROA WOB BS FS FA, re
                      
                      Random-effects GLS regression                   Number of obs     =        966
                      Group variable: COMPANY                         Number of groups  =         69
                      
                      R-sq:                                           Obs per group:
                           within  = 0.0321                                         min =         14
                           between = 0.0084                                         avg =       14.0
                           overall = 0.0006                                         max =         14
                      
                                                                      Wald chi2(4)      =      16.09
                      corr(u_i, X)   = 0 (assumed)                    Prob > chi2       =     0.0029
                      
                      ------------------------------------------------------------------------------
                               ROA |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
                      -------------+----------------------------------------------------------------
                               WOB |   .1031011   .0388552     2.65   0.008     .0269464    .1792559
                                BS |  -.0051465   .0024554    -2.10   0.036     -.009959    -.000334
                                FS |  -.0070367   .0039544    -1.78   0.075    -.0147871    .0007137
                                FA |    .009782   .0104881     0.93   0.351    -.0107743    .0303384
                             _cons |   .1280972   .0361077     3.55   0.000     .0573274    .1988671
                      -------------+----------------------------------------------------------------
                           sigma_u |  .05504877
                           sigma_e |   .0846534
                               rho |  .29719491   (fraction of variance due to u_i)
                      ------------------------------------------------------------------------------
                      
                      . estimate store re
                      
                      . hausman fe re
                      
                                       ---- Coefficients ----
                                   |      (b)          (B)            (b-B)     sqrt(diag(V_b-V_B))
                                   |       fe           re         Difference          S.E.
                      -------------+----------------------------------------------------------------
                               WOB |    .1037013     .1031011        .0006002        .0177037
                                BS |   -.0036958    -.0051465        .0014507        .0009432
                                FS |   -.0443331    -.0070367       -.0372964         .005413
                                FA |    .0675684      .009782        .0577864         .014063
                      ------------------------------------------------------------------------------
                                                 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(4) = (b-B)'[(V_b-V_B)^(-1)](b-B)
                                                =       48.97
                                      Prob>chi2 =      0.0000
                      Last edited by Nur Batrisya; 27 Jan 2022, 16:38.

                      Comment


                      • #12
                        Jeff:
                        1) Do you mean this?

                        Code:
                         xtreg ROA WOB BS FS FA i.YEAR, fe vce(cluster COMPANY)
                        
                        Fixed-effects (within) regression               Number of obs     =        966
                        Group variable: COMPANY                         Number of groups  =         69
                        
                        R-sq:                                           Obs per group:
                             within  = 0.0942                                         min =         14
                             between = 0.0344                                         avg =       14.0
                             overall = 0.0013                                         max =         14
                        
                                                                        F(17,68)          =       2.54
                        corr(u_i, Xb)  = -0.8372                        Prob > F          =     0.0035
                        
                                                       (Std. Err. adjusted for 69 clusters in COMPANY)
                        ------------------------------------------------------------------------------
                                     |               Robust
                                 ROA |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
                        -------------+----------------------------------------------------------------
                                 WOB |   .0672572   .0523924     1.28   0.204    -.0372902    .1718046
                                  BS |   -.002395   .0034457    -0.70   0.489    -.0092708    .0044808
                                  FS |  -.0505556   .0127517    -3.96   0.000    -.0760013     -.02511
                                  FA |   .0177616   .0532606     0.33   0.740    -.0885183    .1240415
                                     |
                                YEAR |
                               2008  |  -.0447429   .0217714    -2.06   0.044     -.088187   -.0012987
                               2009  |   -.011389   .0105424    -1.08   0.284     -.032426     .009648
                               2010  |   .0091338    .014177     0.64   0.522    -.0191561    .0374236
                               2011  |   .0234917   .0159843     1.47   0.146    -.0084044    .0553878
                               2012  |   .0152036   .0175619     0.87   0.390    -.0198407    .0502479
                               2013  |   .0270263   .0188135     1.44   0.155    -.0105154     .064568
                               2014  |   .0185476   .0199836     0.93   0.357     -.021329    .0584242
                               2015  |   .0095067   .0258717     0.37   0.714    -.0421196    .0611329
                               2016  |   .0193137   .0238466     0.81   0.421    -.0282715     .066899
                               2017  |   .0106246   .0279679     0.38   0.705    -.0451845    .0664336
                               2018  |   .0332663   .0310587     1.07   0.288    -.0287103     .095243
                               2019  |   .0440596   .0351293     1.25   0.214    -.0260398    .1141591
                               2020  |   .0338415   .0344703     0.98   0.330    -.0349428    .1026259
                                     |
                               _cons |    .435717   .1611177     2.70   0.009     .1142117    .7572223
                        -------------+----------------------------------------------------------------
                             sigma_u |  .11986156
                             sigma_e |  .08378281
                                 rho |  .67177388   (fraction of variance due to u_i)
                        ------------------------------------------------------------------------------
                        Code:
                        xtreg TQ WOB BS FS FA i.YEAR, fe vce(cluster COMPANY)
                        
                        Fixed-effects (within) regression               Number of obs     =        966
                        Group variable: COMPANY                         Number of groups  =         69
                        
                        R-sq:                                           Obs per group:
                             within  = 0.2210                                         min =         14
                             between = 0.0950                                         avg =       14.0
                             overall = 0.1026                                         max =         14
                        
                                                                        F(17,68)          =       6.17
                        corr(u_i, Xb)  = -0.6725                        Prob > F          =     0.0000
                        
                                                       (Std. Err. adjusted for 69 clusters in COMPANY)
                        ------------------------------------------------------------------------------
                                     |               Robust
                                  TQ |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
                        -------------+----------------------------------------------------------------
                                 WOB |   2.463905   1.271564     1.94   0.057    -.0734615    5.001271
                                  BS |  -.0146124   .0600168    -0.24   0.808    -.1343741    .1051492
                                  FS |  -1.115915   .3754191    -2.97   0.004    -1.865052   -.3667775
                                  FA |   .8742513   1.124884     0.78   0.440     -1.37042    3.118923
                                     |
                                YEAR |
                               2008  |  -1.224219   .1794651    -6.82   0.000    -1.582336   -.8661024
                               2009  |  -.7324359   .1951532    -3.75   0.000    -1.121858   -.3430137
                               2010  |  -.2648907   .2415088    -1.10   0.277     -.746814    .2170325
                               2011  |  -.4661647   .2588768    -1.80   0.076    -.9827453    .0504159
                               2012  |  -.2641439   .2968133    -0.89   0.377    -.8564256    .3281377
                               2013  |   .2204134   .3448908     0.64   0.525    -.4678056    .9086324
                               2014  |   .1504903   .3805593     0.40   0.694     -.608904    .9098846
                               2015  |   .2690584   .4306915     0.62   0.534     -.590373     1.12849
                               2016  |   .2798808   .4438906     0.63   0.530    -.6058891    1.165651
                               2017  |   .7286053    .528214     1.38   0.172    -.3254294     1.78264
                               2018  |   .9848381   .6265339     1.57   0.121    -.2653909    2.235067
                               2019  |   1.001124   .6225822     1.61   0.112    -.2412194    2.243467
                               2020  |   1.512466   .6739646     2.24   0.028      .167591    2.857342
                                     |
                               _cons |   9.330034   2.943174     3.17   0.002     3.457021    15.20305
                        -------------+----------------------------------------------------------------
                             sigma_u |  2.0520993
                             sigma_e |  1.4673429
                                 rho |  .66168679   (fraction of variance due to u_i)
                        ------------------------------------------------------------------------------
                        2) Could you explain why you transformed WOB from 0.01 to 0.10 (to 10% point), and down-transformed ROA / TQ by one decimal place ( about 0.01 / 0.47) when interpreting the results? Also, about making comparisons, do you mean I can compare the effects of WOB on ROA vs TQ in terms of the size of their s.e.?

                        3) Do you recommend I stick to robust s.e. or can still use robust-cluster s.e.? If the former, how can I solve for serial correlation?
                        Last edited by Nur Batrisya; 27 Jan 2022, 17:00.

                        Comment


                        • #13
                          Nur:
                          1) your -fe- codes are correct. However, as Jeff helpfully reminded us, -i.timevar- should be included in the right-hand side of your regression equation;
                          2) as you know, the -fe- estimator allows the panel-wise effect (ui) to be correlated with the vector of coefficients. Your pretty high negative correlations between them and -ui- seem to support the evidence of a panel-wise effect;
                          3) you do not have to run Ramsey's test manually if you test the possible misspecification of the functional form of the regeressands via fitted vales and their powers. For both your model a misspecification issue arises; this means that you should change the metric of your regerssand (ie, ln it) and/or add predictors/interactions in the right hand side of your regression equation. This issue is also supported by the low within R_sqs;
                          4) if you impose non-default standard errors due to heteroskedasticity and/or serial correlation, you should skip -hausman- and go -xtoverid-;
                          5) I fail to get how you could run -estat hettest-, as it is not supported by -xtreg-. Do you mean you used the community-contributed module -xttest3- instead?

                          That said, I stronglyrecommend you to consult a textbook on panel data econometrics (Jeff's one is obviously a -must have-; Statalisters dealing with this stuff also like
                          https://www.stata.com/bookstore/microeconometrics-stata).
                          Please note that folowing the "trials and errors" strategy in this (as in many other) research field(s) is surely time-consuming and possibly misleading.
                          Kind regards,
                          Carlo
                          (Stata 18.0 SE)

                          Comment

                          Working...
                          X