Announcement

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

    fixed effect in policy adoption data

    I really hope you feel safe and healthy.

    I am coding policy adoption data in 48 U.S. states, and the dependent variable is whether or not states have adopted a policy. In this sense, after the adoption year, the observations are dropped. I am not sure whether I have to collect and code data after the adoption year but I think this does not make sense. This is because the likelihood after the adoption year is influenced by previous data, so there is an intertemporal correlation before and after the adoption year.

    As a result, when I used "xtlogit" with fixed-effect, it shows "outcome does not vary in any group." Does it mean there is no within variation in groups? I think this is because there is only one adoption in the dataset, so STATA cannot compare the likelihood, but I am not sure.
    I think just the "logit" command makes sense, though the adoption of the policy comes from the motivations and backgrounds within the states, not the comparison between the states (And I controlled the diffusion effect). Furthermore, if I use the logit model, It also has the same problem above, which means how I care and code data after the adoption year. Is only an alternative to use the event history model?



    Tags: None
  • #2
    Sukyae:
    it means that the same value (1 or 0, depending on the way you coded the levels of the dichotomous regressand) is repeated within panels.
    Neither -xtlogit, fe-, nor - logit- can work if regressand does not vary.
    As an aside, pleaee follow the FAQ and post what you typed and what Stata gave you back. Thanks.
    Kind regards,
    Carlo
    (StataNow 18.5)

    Comment

    • #3
      Thank you for your response.

      Code:
      clear
input str14 state int year byte adoptrps

"Arizona"    1990 0
"Arizona"    1991 0
"Arizona"    1992 0
"Arizona"    1993 0
"Arizona"    1994 0
"Arizona"    1995 0
"Arizona"    1996 0
"Arizona"    1997 0
"Arizona"    1998 0
"Arizona"    1999 0
"Arizona"    2000 0
"Arizona"    2001 0
"Arizona"    2002 0
"Arizona"    2003 0
"Arizona"    2004 0
"Arizona"    2005 0
"Arizona"    2006 1
"Arkansas"   1990 0
"Arkansas"   1991 0
"Arkansas"   1992 0
"Arkansas"   1993 0
"Arkansas"   1994 0
"Arkansas"   1995 0
"Arkansas"   1996 0
"Arkansas"   1997 0
"Arkansas"   1998 0
"Arkansas"   1999 0
"Arkansas"   2000 0
"Arkansas"   2001 0
"Arkansas"   2002 0
"Arkansas"   2003 0
"Arkansas"   2004 0
"Arkansas"   2005 0
"Arkansas"   2006 0
"Arkansas"   2007 0
"Arkansas"   2008 0
"Arkansas"   2009 0
"Arkansas"   2010 0
"Arkansas"   2011 0
"Arkansas"   2012 0
"Arkansas"   2013 0
"Arkansas"   2014 0
"Arkansas"   2015 0
end
      This is the coding I am using. Since Arizona adopted the policy in 2006, it is dropped. Arkansas does not adopt the policy, so it remains until 2015.

      1. The normal logit model

      Code:
      logit adoptrps c.repmean##c.peopleideo c.fossilper c.emission c.gdp c.unemploylag c.govideolag c.ideodiff c.price i.dem c.profess, nolog cluster(state)

Logistic regression                                     Number of obs =    596
                                                        Wald chi2(12) =  41.15
                                                        Prob > chi2   = 0.0000
Log pseudolikelihood = -85.747202                       Pseudo R2     = 0.2605

                                           (Std. err. adjusted for 48 clusters in state)
----------------------------------------------------------------------------------------
                       |               Robust
              adoptrps | Coefficient  std. err.      z    P>|z|     [95% conf. interval]
-----------------------+----------------------------------------------------------------
               repmean |  -.5666919     .21571    -2.63   0.009    -.9894757   -.1439081
            peopleideo |  -.4091162    .156218    -2.62   0.009    -.7152978   -.1029346
                       |
c.repmean#c.peopleideo |   .0109743   .0038899     2.82   0.005     .0033502    .0185985
                       |
             fossilper |   .0030436   .0086548     0.35   0.725    -.0139195    .0200066
              emission |   5.96e-09   1.04e-08     0.58   0.565    -1.43e-08    2.63e-08
                   gdp |   1.71e-06   8.30e-07     2.06   0.040     8.06e-08    3.33e-06
           unemploylag |  -.2343428   .1428422    -1.64   0.101    -.5143084    .0456227
            govideolag |  -.0106918   .0239021    -0.45   0.655    -.0575391    .0361554
              ideodiff |  -.0338794    .023423    -1.45   0.148    -.0797876    .0120288
                 price |   .0667522   .0916212     0.73   0.466     -.112822    .2463265
                 1.dem |    1.29745   .6208605     2.09   0.037      .080586    2.514314
               profess |    -.53709   .2134886    -2.52   0.012      -.95552   -.1186599
                 _cons |   18.39559   8.863633     2.08   0.038     1.023189    35.76799
----------------------------------------------------------------------------------------
Note: 1 failure and 0 successes completely determined.

      2. The xtlogit model. Data in each state is dropped after the adoption year.

      Code:
      xtlogit adoptrps c.repmean##c.peopleideo c.fossilper c.emission c.gdp c.unemploylag c.govideolag c.ideodiff c.price i.dem c.profess, nolog fe

note: 22 groups (364 obs) omitted because of all positive or
      all negative outcomes.
convergence not achieved

Conditional fixed-effects logistic regression        Number of obs    =    232
Group variable: id                                   Number of groups =     26

                                                     Obs per group:
                                                                  min =      2
                                                                  avg =    8.9
                                                                  max =     19

                                                     LR chi2(1)       = 108.00
Log likelihood = 0                                   Prob > chi2      = 0.0000

----------------------------------------------------------------------------------------
              adoptrps | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
-----------------------+----------------------------------------------------------------
               repmean |  -68.78283          .        .       .            .           .
            peopleideo |  -33.83849          .        .       .            .           .
                       |
c.repmean#c.peopleideo |   .8346062          .        .       .            .           .
                       |
             fossilper |  -35.22397          .        .       .            .           .
              emission |  -.0000533   473194.7    -0.00   1.000    -927444.6    927444.6
                   gdp |   .0133302          .        .       .            .           .
           unemploylag |  -75.70677          .        .       .            .           .
            govideolag |   7.996532          .        .       .            .           .
              ideodiff |   1.834573          .        .       .            .           .
                 price |   169.7478          .        .       .            .           .
                 1.dem |   205.6297          .        .       .            .           .
               profess |    46.8682          .        .       .            .           .
----------------------------------------------------------------------------------------
      3. The xtlogit model. Data in each state is not dropped and coded as "0"

      Code:
      . xtlogit adoptrps c.repmean##c.peopleideo c.fossilper c.emission c.gdp c.unemploylag c.govideolag c.ideodiff c.price i.dem c.profess, nolog fe
note: 19 groups (361 obs) omitted because of all positive or
      all negative outcomes.

Conditional fixed-effects logistic regression        Number of obs    =    551
Group variable: id                                   Number of groups =     29

                                                     Obs per group:
                                                                  min =     19
                                                                  avg =   19.0
                                                                  max =     19

                                                     LR chi2(12)      =  38.76
Log likelihood = -66.007936                          Prob > chi2      = 0.0001

----------------------------------------------------------------------------------------
              adoptrps | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
-----------------------+----------------------------------------------------------------
               repmean |  -.4125622   .2522502    -1.64   0.102    -.9069636    .0818391
            peopleideo |  -.2413848   .1792815    -1.35   0.178    -.5927701    .1100004
                       |
c.repmean#c.peopleideo |    .007829    .003824     2.05   0.041     .0003341    .0153239
                       |
             fossilper |    -.02222   .0481061    -0.46   0.644    -.1165061    .0720662
              emission |   1.26e-07   7.37e-08     1.72   0.086    -1.81e-08    2.71e-07
                   gdp |  -2.26e-06   3.21e-06    -0.70   0.481    -8.55e-06    4.03e-06
           unemploylag |  -.1852732   .1933301    -0.96   0.338    -.5641932    .1936468
            govideolag |  -.0306272   .0259252    -1.18   0.237    -.0814397    .0201853
              ideodiff |  -.0056636   .0266978    -0.21   0.832    -.0579904    .0466632
                 price |   -.225059   .1565869    -1.44   0.151    -.5319637    .0818457
                 1.dem |   1.401893   .8119022     1.73   0.084    -.1894065    2.993192
               profess |  -.9835737   .5055017    -1.95   0.052    -1.974339    .0071914
----------------------------------------------------------------------------------------
      Last edited by Sukjae Lee; 15 Mar 2022, 22:20.

      Comment

      • #4
        Code:
        "Arizona"  1990 0
"Arizona"  1991 0
"Arizona"  1992 0
"Arizona"  1993 0
"Arizona"  1994 0
"Arizona"  1995 0
"Arizona"  1996 0
"Arizona"  1997 0
"Arizona"  1998 0
"Arizona"  1999 0
"Arizona"  2000 0
"Arizona"  2001 0
"Arizona"  2002 0
"Arizona"  2003 0
"Arizona"  2004 0
"Arizona"  2005 0
"Arizona"  2006 1
"Arizona"  2007 0
"Arizona"  2008 0
"Arizona"  2009 0
"Arizona"  2010 0
"Arizona"  2011 0
"Arizona"  2012 0
"Arizona"  2013 0
"Arizona"  2014 0
"Arizona"  2015 0
"Arkansas" 1990 0
"Arkansas" 1991 0
"Arkansas" 1992 0
"Arkansas" 1993 0
"Arkansas" 1994 0
"Arkansas" 1995 0
"Arkansas" 1996 0
"Arkansas" 1997 0
"Arkansas" 1998 0
"Arkansas" 1999 0
"Arkansas" 2000 0
"Arkansas" 2001 0
"Arkansas" 2002 0
"Arkansas" 2003 0
"Arkansas" 2004 0
"Arkansas" 2005 0
"Arkansas" 2006 0
"Arkansas" 2007 0
"Arkansas" 2008 0
"Arkansas" 2009 0
"Arkansas" 2010 0
"Arkansas" 2011 0
end

        Comment

        • #5
          Sukyae:
          your code #3 looks more convincing than the previous ones.
          Check it with the literature in your research field, though.
          Last edited by Carlo Lazzaro; 16 Mar 2022, 02:39.
          Kind regards,
          Carlo
          (StataNow 18.5)

          Comment

          Previous Next
          Working...
          X