Announcement

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

  • xtpmg vs. xtdcce2 (difference in results)

    Dear all,

    I am struggling to understand the differences between the Stata commands xtpmg and xtdcce2 (written by Jan Ditzen). If I understand correctly, the following two commands should be equivalent (both not controlling for cross-sectional dependence):

    HTML Code:
    xtpmg d.log_ResProd_short d.ResTaxRate_Infl_PPP d.log_PopDens d.log_GDP_pc, lr(l.log_ResProd_short ResTaxRate_Infl_PPP log_PopDens log_GDP_pc) ec(ec) replace pmg
    HTML Code:
    xtdcce2 d.log_ResProd_short d.ResTaxRate_Infl_PPP d.log_PopDens d.log_GDP_pc, lr(l.log_ResProd_short ResTaxRate_Infl_PPP log_PopDens log_GDP_pc) pooled(l.log_ResProd_short ResTaxRate_Infl_PPP log_PopDens log_GDP_pc) nocrosssectional
    However, I get very different results (at least regarding the p-values). Can someone explain the difference? Which model is superior in which case?

    HTML Code:
    Pooled Mean Group Regression
    (Estimate results saved as pmg)
    
    Panel Variable (i): Country_Code                Number of obs      =       645
    Time Variable (t): Year                         Number of groups   =        31
                                                    Obs per group: min =        11
                                                                   avg =      20.8
                                                                   max =        26
    
                                                    Log Likelihood     =  911.8572
    -------------------------------------------------------------------------------------
    D.log_ResProd_short | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
    --------------------+----------------------------------------------------------------
    ec                  |
    ResTaxRate_Infl_PPP |   .0643741   .0154175     4.18   0.000     .0341563    .0945919
            log_PopDens |   2.612486   .1629119    16.04   0.000     2.293185    2.931788
             log_GDP_pc |   .8725355   .0349202    24.99   0.000     .8040931    .9409778
    --------------------+----------------------------------------------------------------
    SR                  |
                     ec |  -.3689291   .0535052    -6.90   0.000    -.4737974   -.2640608
                        |
    ResTaxRate_Infl_PPP |
                    D1. |   1.428748   1.130847     1.26   0.206    -.7876705    3.645167
                        |
            log_PopDens |
                    D1. |  -2.042252   1.804974    -1.13   0.258    -5.579936    1.495432
                        |
             log_GDP_pc |
                    D1. |   .6399973   .0700586     9.14   0.000      .502685    .7773096
                        |
                  _cons |   -7.34668   1.069888    -6.87   0.000    -9.443622   -5.249738
    -------------------------------------------------------------------------------------
    HTML Code:
    (Dynamic) Common Correlated Effects Estimator - Pooled Mean Group (CS-ECM)
    
    Panel Variable (i): Country_Code                        Number of obs     =        645
    Time Variable (t): Year                                 Number of groups  =         31
    
    Degrees of freedom per group:                           Obs per group:    
     without cross-sectional avg. min   = 3                               min =         11
                                  max   = 18                              avg =         22
     with cross-sectional avg.    min   = 3                               max =         26
                                  max   = 18
    Number of                                               F(128, 517)       =       5.47
     cross-sectional lags               none                Prob > F          =       0.00
     variables in mean group regression = 97                R-squared         =       0.42
     variables partialled out           = 31                Adj. R-squared    =       0.28
                                                            Root MSE          =       0.10
                                                            CD Statistic      =      20.59
                                                               p-value        =     0.0000
    --------------------------------------------------------------------------------------
       D.log_ResProd_short|     Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
    ----------------------+---------------------------------------------------------------
     Short Run Est.      |
    ----------------------+---------------------------------------------------------------
       Mean Group:        |
     D.ResTaxRate_Infl_PPP|  1.144401   .8314861    1.38    0.169     -.4852817   2.774084
             D.log_PopDens| -1.663779   1.683031   -0.99    0.323     -4.962458   1.634901
              D.log_GDP_pc|  .7696895   .0380582   20.22    0.000      .6950968   .8442821
    ----------------------+---------------------------------------------------------------
     Adjust. Term         |
    ----------------------+---------------------------------------------------------------
       Pooled:            |
       L.log_ResProd_short| -.2480367   .7273092   -0.34    0.733     -1.673537   1.177463
    ----------------------+---------------------------------------------------------------
     Long Run Est.        |
    ----------------------+---------------------------------------------------------------
       Pooled:            |
       ResTaxRate_Infl_PPP|  .0519683   .7605495    0.07    0.946     -1.438681   1.542618
               log_PopDens|  2.546216   2.789462    0.91    0.361     -2.921029   8.013462
                log_GDP_pc|  .7305529   .8004832    0.91    0.361     -.8383653   2.299471
    --------------------------------------------------------------------------------------
    Pooled Variables: L.log_ResProd_short ResTaxRate_Infl_PPP log_PopDens log_GDP_pc
    Mean Group Variables: D.ResTaxRate_Infl_PPP D.log_PopDens D.log_GDP_pc
    Long Run Variables: ResTaxRate_Infl_PPP log_PopDens log_GDP_pc 
    Cointegration variable(s): L.log_ResProd_short
    Heterogenous constant partialled out. 
    Warning:
    Collinearities detected. One or more variables are dropped and set to zero.
    Use estat ebistructure to display more details.
    Also, if I want to control for cross-sectional dependence, my xtdcce2 command would look like this (if I understand correctly):

    HTML Code:
    xtdcce2 d.log_ResProd_short d.ResTaxRate_Infl_PPP d.log_PopDens d.log_GDP_pc, lr(l.log_ResProd_short ResTaxRate_Infl_PPP log_PopDens log_GDP_pc) pooled(l.log_ResProd_short ResTaxRate_Infl_PPP log_PopDens log_GDP_pc) crosssectional(d.log_ResProd_short d.ResTaxRate_Infl_PPP d.log_PopDens d.log_GDP_pc)
    I understand that one can control for cross-sectional dependence using xtpmg by including averages. Would I include those in the SR or LR equation? And again: which model would be better?

    Thank you very much in advance!

Working...
X