propensity score matching for merger data

[siz249]

Hi there, I am working on a merger data sample trying to reveal the merger gain, if there is any. I am using propensity score matching to construct my control group. However unlike standard Propensity score matching scenario, merger sample involved 2 parties at the same time. So in addition to match acquirer, I am also trying to match target. Therefore for every merger pair (an acquirer and its target), I need a matched pair (control acquirer and target). My dependent varible therefore would the weighted average of some measurement of the pair instead of one party only. I am stuck with this problem for a while and I know some people have done the modification. So I am wondering if anyone can help me to figure this out. Really appreciate!

[Joe Canner]

I think to answer this question we need more insight into what you are trying to do.

• Are the potential controls firms that have also been an acquirer or a target in a merger? If so...
  • What distinguishes cases from controls?
  • Are you interested in matching acquirers and targets separately or as pairs? In other words...
    • Is each index acquirer matched with a control acquirer based on certain characteristics and each index target matched with a control acquirer based on certain characteristics? If so, are the characteristics the same or different depending on whether you are matching acquirers or targets?
    • Is each index acquirer-target pair matched with a control acquirer-target pair based on a certain set of characteristics that are specific to that pair? (This assumes, of course, that potential control have also been involved in mergers.)

My initial assumption is that the answer to all of these questions is "no"; i.e., controls were not involved in mergers, and acquirers and targets are both matched to all possible controls. This can be done very simply, by having acquirers and targets both be labeled as cases ("treated"), matching without replacement (to avoid an acquirer and target pair both matching the same control), and then reshaping the data to pair up the acquirer-target pairs (one case pair and one control pair). You may not be able to do all of the usual canned statistics that propensity score matching programs give you, but you should be able to use the resulting data to do your own matched analysis. I may be missing something though...

[siz249]

Thanks for sharing! Here is my answer to your questions and hope it can give you a clearer picture:

1. first of all, I have a group of bank acquired their targets during some period (this is my treatment group). Then my control group involved banks had no M&As at all (neither be an acquirer or a target).

2. what I want to do is that I have performance measure (ROA, etc.) for the merged bank 1yr, 2yr and 3 yr later. I calculated the difference as compared to its pre-merger performance (pre-merger performance is calculated by weighted average performance of acquirer and its target based on asset size). This difference would be my performance changes after merger.

3. then from my control group, I want to use propensity score match to generate an close match for acquirer and target (charactoristics for both macthing are using same set of variables). Using the same way I calculated my pre-merger performance in step 2, I will calculate the difference before post- and pre-merger performance for this pair assuming simple combination (weighted average in this case). Because control group are close pair for actual acquier and target, I could assume what I got here would be the performance if this actual M&A transaction did not happend.

4. Therefore, comparing my treatment group and control group for performance changes, I should have so called Synergy if M&As creates value, or destroyed value if the other side turns out to be sure. This is what I want to find out, whether M&As have an economic rational.

Hope this help to clear the wonders and please feel free to share what you can think of. I really need help to get my research move forward.

Thanks!