Same observations different year propensity score matching

[Britta Schijff]

Dear reader,

Currently I am doing a research for my masterthesis, whereby I use propensity score matching.

However, I have firms that in year 1 are not treated and in year 2 or 3 are treated.
If I perform propensity score matching, I think they will match the same firms that are not treated in year 1 with themselves in the years they are treated.
I've added fiscal year as independent variables, but still the region of common support is .99896195 which is pretty high in my opinion.

Does anyone has a solution for this problem?

Thank you in advance,

Britta Schijff

[Mike Lacy]

It would help if you explained how you would *like* the matching to proceed, that is, what rules do you *want* the matching to follow. Further, do you want to allow a firm to be a treatment subject in one year but a control in another? That is a possibility that can be reasonable. Also, how do you want "year" to enter into the match. For example, could a treatment firm in year 1 be matched to a control firm in year 3, as well as the reverse? Is there a particular Stata procedure you are wanting to use for this matching and analysis? Finally, as always is true here, showing a sample of your data using -dataex- (see the FAQ) is likely to be helpful?

[Britta Schijff]

Thank you Mike for your response, here is an example of my dataset:
input float year byte clawback float ERC
2012 1 -12.027376
2013 1 -.5941014
2010 0 3.655981
2011 1 5.722945
2013 0 3.839935
2010 0 .40065545
2012 0 1.054705
2016 1 .4786063
2011 0 -.7525854
2014 1 -1.7295245
2016 0 3.4072754
2010 0 -1.7473538
2015 0 1.1098427
2017 0 2.769554
2011 0 -5.518787
2014 1 .3320716
2016 1 -4.096886
2011 0 -1.3896435
2015 1 -.072744586
2013 1 .07528323
2010 0 -.9182339
2012 0 .08513422
2017 1 4.789506
2014 1 3.97917
2018 1 2.4470615
2010 1 -2.105691
2017 1 -.04314759
2016 1 .66527
2012 1 -2.5367224
2011 1 -12.405378
2013 1 -4.1389837
2014 1 -10.814457
2013 1 -9.319745
2015 1 4.998299
2012 1 .4108148
2015 1 1.186667
2010 0 3.436631
2011 1 -.52800524
2016 1 -.6646025
2012 1 -.5740604
2017 1 5.228627
2013 1 2.0792744
2014 1 -177.73994
2013 1 -1.483731
2017 1 .6701498
2012 1 -7.177568
2011 1 4.3022776
2014 1 8.751979
2011 1 .2954478
2017 1 -.029353876
2012 0 .8696177
2014 1 .7433214
2015 1 -34.590534
2013 1 5.972861
2014 0 .6484414
2015 1 -4.607118
2013 0 3.3004715
2017 1 1.0664916
2010 0 -3.013066
2016 1 12.525826
2010 0 2.0055554
2011 1 12.583147
2013 1 -1.223335
2017 1 6.060036
2014 1 -7.138487
2013 0 1.3249896
2015 0 -2.5531244
2011 1 11.09817
2010 0 -4.6756163
2011 0 3.1816576
2010 0 6.95943
2017 1 .5657401
2014 1 -.29041505
2016 0 -.54941344
2012 0 .13418517
2013 1 -.1566632
2015 1 .03448614
2010 0 -.7707962
2016 1 2.1189196
2017 1 2.481526
2016 0 2.937855
2011 1 -1.669809
2017 0 -9.02121
2010 1 -12.79592
2012 1 -.2172724
2013 1 3.087588
2011 1 237.6083
2010 1 14.911936
2012 1 -.01574306
2015 1 1.0012686
2016 1 -1.4361477
2013 1 -1.259526
2016 0 .17524956
2017 0 -.09338693
2014 0 2.629949
2013 1 -1.747321
2010 0 -7.020304
2012 1 -.13142298
2015 1 -.15453404
2011 0 -14.8819
end

I would like to match a '0' for clawback in year x with a '1' for clawback in the same year (and thus a different firm with the same characteristics). And then look at the change in ERC. I've already run the propensity score matching with this code:

. . global ylist ERC

.
. . global xlist year roa size industry leverage mtb ICMWD

.
. . global treatment clawback

pscore $treatment $xlist, pscore (myscore) blockid(myblock) comsup

I'm just not sure if this is the way to match the firms as I would like to..

Thank you in advance!

[Mike Lacy]

I have not happened to use the user-contributed command -pscore-. Looking at the help for it, though, I can't find any way to force the matches to be within-year. Perhaps this can be done with some feature of the built-in -teffects- commands.