Spatial Dynamic Panel Model with FE when N large and T small

[Ruth Nikijuluw]

Dear all,

I would like to run a spatial dynamic panel model with Fixed Effect (FE). As far my understanding, i cannot run using the xsmle command since my N is relatively larger (N=490) while T is really small (T=5). I read some articles and those suggest that i need to run using a GMM procedure to get a consistent and unbiased results. Any idea of how to this in STATA, or any user-written command that i can use?

Many thanks

[Phil Bromiley]

You'll increase your chances of a useful answer by following the faq on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex. Also, do not provide pictures.

I don't do spatial models. I know Stata 15 has a variety of such models. There are also several user-written routines. If findit did not find what you want, I'd try to google the issue.

[Ruth Nikijuluw]

Dear Phil,

Thanks for your suggestion.

The specific "xsmle" code that i used is below:

xsmle pc2 lnexpcap lnpop povtrate povtgap urban_rate lngrdpcap lnexpcapprov, dlag(1) wmat(W_2) model(sdm)fe type(ind) nolog

Using this code, i am able to produce this following results:

Dynamic SDM with spatial fixed-effects               Number of obs =      1744

Group variable: kode_2009                         Number of groups =       436
Time variable: year                                   Panel length =         4

R-sq:    within  = 0.4483
         between = 0.3628
         overall = 0.3605

Mean of fixed-effects = -20.5626

Log-likelihood =  -607.4401
------------------------------------------------------------------------------
         pc2 |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
Main         |
         pc2 |
         L1. |   .1518592   .0197937     7.67   0.000     .1130642    .1906541
             |
    lnexpcap |   -.017293   .0657726    -0.26   0.793    -.1462048    .1116189
       lnpop |  -.6265645   .2048895    -3.06   0.002    -1.028141   -.2249884
    povtrate |  -.0193569   .0136373    -1.42   0.156    -.0460854    .0073717
     povtgap |   .0337402   .0163293     2.07   0.039     .0017354    .0657449
  urban_rate |   .0249641   .0093417     2.67   0.008     .0066546    .0432736
   lngrdpcap |  -.1532616   .0807036    -1.90   0.058    -.3114378    .0049146
lnexpcapprov |  -.0750031   .0506973    -1.48   0.139    -.1743681    .0243618
-------------+----------------------------------------------------------------
Wx           |
    lnexpcap |   .2834804   .3215404     0.88   0.378    -.3467272    .9136879
       lnpop |   .9132967   2.462981     0.37   0.711    -3.914057     5.74065
    povtrate |   .0957065   .0641256     1.49   0.136    -.0299774    .2213904
     povtgap |   .1681391   .1408838     1.19   0.233    -.1079881    .4442662
  urban_rate |   .1614315   .1038291     1.55   0.120    -.0420699    .3649329
   lngrdpcap |   1.218524   .5473999     2.23   0.026     .1456401    2.291408
lnexpcapprov |     .21519   .1475523     1.46   0.145    -.0740072    .5043872
-------------+----------------------------------------------------------------
Spatial      |
         rho |   .4431088   .1161655     3.81   0.000     .2154287     .670789
-------------+----------------------------------------------------------------
Variance     |
    sigma2_e |   .1383996   .0037524    36.88   0.000     .1310451    .1457541
------------------------------------------------------------------------------ 

However, after read some references, it's been advised that because my N is much more larger than the number of T, running using a quasi MLE methodology- which is what assumed by xsmle- may not lead to consistent results. The references suggest to run using a GMM procedure hence therefore i am trying to find a procedure of how to run a spatial dynamic panel with GMM. I looked at STATA 15 spatial module, but i could not find any specific command or option that can handle this case.

Many thanks!

[Aleksey Oshchepkov]

Dear all,

I have the same question as Ruth posted a year ago. I have a panel with large n (80), small T (6) and would like to run a spatial dynamic panel model. The only command in Stata I found is -xsmle-, but it is using qMLE which is appropriate for large T and n (based on Lee , de Jong, & Yu, 2008. Quasi-maximum likelihood estimators for spatial dynamic panel data with fixed effects when both n and T are large Journal of Econometrics, 146, pp. 118-134). Are there any other commands in Stata that could work for the small-T case, e.g., some commands using GMM?

Many thanks!

[Mushtaq Zulqarnain]

Hello Dear Professor, I am trying to apply Dynamic Spatial panel model on a panel data. but i cant find the commands and package. can you please guide me how to estimate this specific model. thank you