propensity score matching for a unbalanced panel

[abhilash nair]

Hi,

I am analysing dividend behaviour of firms which give interim, special dividends or bonus shares (together i call them as "Quasi dividends") vis a vis firms that mostly pay regular dividends. Firms pay both regular and quasi dividends Hence, in my sample I have firms that pay regular dividends and/or quasi dividends over years. I have 8,000 firm year observations around a event (Dividend declaration), the treatment being if they have declared quasi dividends or not and some explanatory variables. The outcome to be tested is the cumulative abnormal return around the event.

I seek help in the following queries:
(i) Can I use propensity score matching (teffects psmatch) in this case where I have cross sectional time series data
(ii) If yes then the same firm can be there in control and treatment groups. Is this ok?
(iii) If I cant use teffects psmatch, is there any other way to conduct propensity score matching between the two groups.

Thanks
Abhilash

[David Radwin]

Regarding (iii), yes, there are several user-written programs for matching you can find using search matching, net. The most popular of these programs (it seems) is psmatch2, perhaps because it offers a dialog box and many options for other types of matching, but there are also the programs pscore,optmatch, etc. And of course you could use the built-in program teffects nnmatch which is similar to teffects psmatch.

I'm afraid I can't help with (i) or (ii).

[abhilash nair]

Thanks David.