Announcement

Collapse
No announcement yet.
X
Collapse
  • Page of 1
  • Filter
  • Time
  • Show
Clear All
new posts
  • #1

    Indicator Variable dropped from fixed effects regression results

    Good evening everybody,

    I have posted this question elsewhere in stata forum but I did not get any response: (https://www.statalist.org/forums/for...85#post1636485)

    Therefore I'm posting my issue again. I apologize in advance if this construed as a spam.

    I am trying to estimate a fixed-effects model using xtreg command in stata 17.0.

    My econometric model is as under:

    ki,t - ki,t-1 (1 - ΛhatZi,t-1 ) = β [(ΛhatZi,t-1 ) Xi,t-1] + ηi,t

    where     ΛhatZi,t-1 are estimated values from a pooled regression.
    I generate the LHS variable using the generate command and then try to use the following command:

    xtreg LHS_thirdstep c.soa_tier2_nocons1#( i.l.state1nonstate0 c.l.size c.l.profitability_w c.l.provforNPA_to_net_advances_w ///
    c.l.net_advances_to_assets_w c.l.market_funding_w i.l.Listeddummy1iflisted)] Y5-Y20, fe vce(robust)


    However, there seems to be an issue with my results. My indicator variables (two indicator variables taking value either 0 or 1) are not exhibiting results as they should be ideally doing (i.e. for the reference category 1). Also depending on how i order the indicator variables one of the category gets dropped. I reproduce the results from xtreg command with fe option.

    note: 1L.state1nonstate0#c.soa_tier2_nocons1 omitted because of collinearity
    Robust
    LHS_thirdstep Coef. Std. Err. t P>t [95% Conf. Interval]
    c.soa_tier2_nocons1#cL.size -1.204929 .8239942 1.46 0.149 -2.853162 .4433053
    c.soa_tier2_nocons1#cL.profitability_w .8311909 .4674114 1.78 0.080 -.1037711 1.766153
    c.soa_tier2_nocons1#cL.provforNPA_to_net_advances_ w .4286864 .3364938 1.27 0.208 -.2444014 1.101774
    c.soa_tier2_nocons1#cL.net_advances_to_assets_w -.0698509 .0623602 1.12 0.267 -.1945898 .054888
    c.soa_tier2_nocons1#cL.market_funding_w .0651142 .0322979 2.02 0.048 .0005087 .1297196
    L.Listeddummy1iflisted#c.soa_tier2_nocons1
    0 23.0813 14.32946 1.61 0.112 -5.581884 51.74448
    1 22.723 13.72806 1.66 0.103 -4.737198 50.1832
    L.state1nonstate0#c.soa_tier2_nocons1
    0 -.4240838 1.735835 0.24 0.808 -3.89627 3.048102
    1 0 (omitted)
    Y5 .2792379 .5636733 0.50 0.622 -.8482766 1.406752
    Y6 -.637602 .6677788 0.95 0.344 -1.973358 .6981545
    Y7 .7873123 .4665592 1.69 0.097 -.1459451 1.72057
    Y8 .5926671 .6799688 0.87 0.387 -.7674731 1.952807
    Y9 1.562704 .8756274 1.78 0.079 -.1888116 3.31422
    Y10 2.134863 .874211 2.44 0.018 .3861808 3.883546
    Y11 1.988466 1.172886 1.70 0.095 -.3576549 4.334587
    Y12 2.454433 1.119152 2.19 0.032 .2157971 4.69307
    Y13 2.115506 1.135905 1.86 0.067 -.1566421 4.387655
    Y14 1.494296 1.297193 1.15 0.254 -1.100476 4.089068
    Y15 1.31945 1.45346 0.91 0.368 -1.587903 4.226803
    Y16 1.908256 1.379692 1.38 0.172 -.8515392 4.668051
    Y17 2.02002 1.334612 1.51 0.135 -.6496013 4.689642
    Y18 2.59642 1.417224 1.83 0.072 -.2384511 5.43129
    Y19 2.7659 1.513867 1.83 0.073 -.2622841 5.794085
    Y20 2.62843 1.773836 1.48 0.144 -.9197715 6.176631
    _cons -.7346817 2.278107 0.32 0.748 -5.291574 3.822211
    sigma_u 2.4584079
    sigma_e 1.777728
    rho .65663981 (fraction of variance due to u_i)
    This is happening even when I estimate without inclusion of year FE:
    (Std. Err. adjusted for 61 clusters in Bankcode)
    Robust
    LHS_thirdstep Coef. Std. Err. t P>t [95% Conf. Interval]
    c.soa_tier2_nocons1#cL.size .0172878 .3759817 0.05 0.963 -.7347875 .7693631
    c.soa_tier2_nocons1#cL.profitability_w .5583617 .2504609 2.23 0.030 .0573654 1.059358
    c.soa_tier2_nocons1#cL.provforNPA_to_net_advances_ w .4084313 .1672226 2.44 0.018 .0739362 .7429264
    c.soa_tier2_nocons1#cL.net_advances_to_assets_w .0417223 .0391258 1.07 0.291 -.0365409 .1199856
    c.soa_tier2_nocons1#cL.market_funding_w .0287254 .032788 0.88 0.384 -.0368603 .0943111
    L.Listeddummy1iflisted#c.soa_tier2_nocons1
    0 4.354254 4.168328 1.04 0.300 -3.983643 12.69215
    1 3.457041 3.648269 0.95 0.347 -3.840583 10.75467
    L.state1nonstate0#c.soa_tier2_nocons1
    0 3.063894 1.689645 1.81 0.075 -.3158989 6.443686
    1 0 (omitted)
    _cons .4912929 .6968191 0.71 0.484 -.9025529 1.885139
    sigma_u 3.1152293
    sigma_e 1.8472187
    rho .73986044 (fraction of variance due to u_i)
    Any help would be much appreciated.

    Thanks and regards.
    Tags: None
  • #2
    Gagandeep:
    the warning message note: 1L.state1nonstate0#c.soa_tier2_nocons1 omitted because of collinearity explains the reason of this issue (and Stata decides what to drop in tese instances).
    Probably the substantive issue is the your model should be a bit more parsimonious: I fail to get how you can easily disseminate such a bunch of interactions to your audience.
    Kind regards,
    Carlo
    (Stata 19.0)

    Comment

    • #3
      Originally posted by gagandeep sharma View Post
      Good evening everybody,

      I have posted this question elsewhere in stata forum but I did not get any response: (https://www.statalist.org/forums/for...85#post1636485)

      Therefore I'm posting my issue again. I apologize in advance if this construed as a spam.

      I am trying to estimate a fixed-effects model using xtreg command in stata 17.0.

      My econometric model is as under:

      ki,t - ki,t-1 (1 - ΛhatZi,t-1 ) = β [(ΛhatZi,t-1 ) Xi,t-1] + ηi,t

      where       ΛhatZi,t-1 are estimated values from a pooled regression.
      I generate the LHS variable using the generate command and then try to use the following command:

      xtreg LHS_thirdstep c.soa_tier2_nocons1#( i.l.state1nonstate0 c.l.size c.l.profitability_w c.l.provforNPA_to_net_advances_w ///
      c.l.net_advances_to_assets_w c.l.market_funding_w i.l.Listeddummy1iflisted)] Y5-Y20, fe vce(robust)


      However, there seems to be an issue with my results. My indicator variables (two indicator variables taking value either 0 or 1) are not exhibiting results as they should be ideally doing (i.e. for the reference category 1). Also depending on how i order the indicator variables one of the category gets dropped. I reproduce the results from xtreg command with fe option.

      note: 1L.state1nonstate0#c.soa_tier2_nocons1 omitted because of collinearity
      Robust
      LHS_thirdstep Coef. Std. Err. t P>t [95% Conf. Interval]
      c.soa_tier2_nocons1#cL.size -1.204929 .8239942 1.46 0.149 -2.853162 .4433053
      c.soa_tier2_nocons1#cL.profitability_w .8311909 .4674114 1.78 0.080 -.1037711 1.766153
      c.soa_tier2_nocons1#cL.provforNPA_to_net_advances_ w .4286864 .3364938 1.27 0.208 -.2444014 1.101774
      c.soa_tier2_nocons1#cL.net_advances_to_assets_w -.0698509 .0623602 1.12 0.267 -.1945898 .054888
      c.soa_tier2_nocons1#cL.market_funding_w .0651142 .0322979 2.02 0.048 .0005087 .1297196
      L.Listeddummy1iflisted#c.soa_tier2_nocons1
      0 23.0813 14.32946 1.61 0.112 -5.581884 51.74448
      1 22.723 13.72806 1.66 0.103 -4.737198 50.1832
      L.state1nonstate0#c.soa_tier2_nocons1
      0 -.4240838 1.735835 0.24 0.808 -3.89627 3.048102
      1 0 (omitted)
      Y5 .2792379 .5636733 0.50 0.622 -.8482766 1.406752
      Y6 -.637602 .6677788 0.95 0.344 -1.973358 .6981545
      Y7 .7873123 .4665592 1.69 0.097 -.1459451 1.72057
      Y8 .5926671 .6799688 0.87 0.387 -.7674731 1.952807
      Y9 1.562704 .8756274 1.78 0.079 -.1888116 3.31422
      Y10 2.134863 .874211 2.44 0.018 .3861808 3.883546
      Y11 1.988466 1.172886 1.70 0.095 -.3576549 4.334587
      Y12 2.454433 1.119152 2.19 0.032 .2157971 4.69307
      Y13 2.115506 1.135905 1.86 0.067 -.1566421 4.387655
      Y14 1.494296 1.297193 1.15 0.254 -1.100476 4.089068
      Y15 1.31945 1.45346 0.91 0.368 -1.587903 4.226803
      Y16 1.908256 1.379692 1.38 0.172 -.8515392 4.668051
      Y17 2.02002 1.334612 1.51 0.135 -.6496013 4.689642
      Y18 2.59642 1.417224 1.83 0.072 -.2384511 5.43129
      Y19 2.7659 1.513867 1.83 0.073 -.2622841 5.794085
      Y20 2.62843 1.773836 1.48 0.144 -.9197715 6.176631
      _cons -.7346817 2.278107 0.32 0.748 -5.291574 3.822211
      sigma_u 2.4584079
      sigma_e 1.777728
      rho .65663981 (fraction of variance due to u_i)
      This is happening even when I estimate without inclusion of year FE:
      (Std. Err. adjusted for 61 clusters in Bankcode)
      Robust
      LHS_thirdstep Coef. Std. Err. t P>t [95% Conf. Interval]
      c.soa_tier2_nocons1#cL.size .0172878 .3759817 0.05 0.963 -.7347875 .7693631
      c.soa_tier2_nocons1#cL.profitability_w .5583617 .2504609 2.23 0.030 .0573654 1.059358
      c.soa_tier2_nocons1#cL.provforNPA_to_net_advances_ w .4084313 .1672226 2.44 0.018 .0739362 .7429264
      c.soa_tier2_nocons1#cL.net_advances_to_assets_w .0417223 .0391258 1.07 0.291 -.0365409 .1199856
      c.soa_tier2_nocons1#cL.market_funding_w .0287254 .032788 0.88 0.384 -.0368603 .0943111
      L.Listeddummy1iflisted#c.soa_tier2_nocons1
      0 4.354254 4.168328 1.04 0.300 -3.983643 12.69215
      1 3.457041 3.648269 0.95 0.347 -3.840583 10.75467
      L.state1nonstate0#c.soa_tier2_nocons1
      0 3.063894 1.689645 1.81 0.075 -.3158989 6.443686
      1 0 (omitted)
      _cons .4912929 .6968191 0.71 0.484 -.9025529 1.885139
      sigma_u 3.1152293
      sigma_e 1.8472187
      rho .73986044 (fraction of variance due to u_i)
      Any help would be much appreciated.

      Thanks and regards.
      Thanks Carlo for your reply,

      Unfortunately I cannot tinker with the model as it is well established in the literature.It comes from: Jiang, C., Liu, H., & Molyneux, P. (2019). Do different forms of government ownership matter for bank capital behavior? Evidence from China. Journal of Financial Stability, 40, 38–49. https://doi.org/10.1016/j.jfs.2018.11.005

      I want to understand the reasons behind dropping of the indicator variables. I understand that time-invariant variables may be dropped since they tend to be collinear with fixed effects. In my case two variables: state1nonstate0 and Listeddummy1iflisted are both time-invariant, they do not change values for a bank within the sample period. However, in the present model both the variables are being interacted with another variable. Thus, whenever the indicator variable is one, the interactions will not be same for a particular bank over the sample period. Then why are these variables still being dropped?

      I manually create the interactions of the independent variables and then run the regression, which runs fine. But i don't find that approach productive. Any pointers would be much appreciated

      Thanks and regards.

      Comment

      Previous Next
      Working...
      X