Differences-in-differences mergers

[Robin Boender]

Hello all,

I am currently writing my master thesis about financial performanes of merging hospitals in comparison with non-merging hospitals (from 2007 till 2016) and i have some questions regarding this topic:
Firstly, i have to find an appropriate control group to compare my merging hospitals with. I collected ratios of the following variables: solvency, liquidity, current ratio, rentability, total staff, total admissions, nursing days, bed capacity. Based on these variables i want to find an appropriate control group (non-merging hospitals) (via paired matching??) to compare the merging hospitals (14 hospitals) with.

Secondly, i want to investigate whether reduction in bed capacity increases financial performanes (solvency, liquidity, current ratio, rentability) of hospitals (thus, merging hospitals and non-merging hospitals). In order to investigate that question, i am creating a delta variable, which indicates whether bed capacity reduces or increases over time (2007 till 2016). My question to you guys: how can i make STATA to link the detla variable of bed capacity to solvency, liquidity, current ratio, rentability in order to know whether this reduction or increase result in performing better on these financial indicators?

Thirdly, by conducting a logit model i want to investigate whether underperforming hospitals have a higher propensity to merge. My time window consists of 3 pre-merger years from the moment the merging hospitals engaged in a merger. But, how do i determine T_min3 for non-merging hospitals, since each merging hospital engaged in a merger in different years? I already know i have to create a binary variable (merger = 1 / non-merger = 0), which functions as dependent variable and solvency, liquidity, current ratio, rentability as explanatory variables. And how do i add this time indication to this logit model with the previously mentioned variables?

I would be great if you could help me out!

Kind regards.

[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, try to simplify your posting to the critical issues. You are also likely to increase your chances of a useful answer by asking about one analysis at a time particularly when they require different data structures.

You need to decide whether to use propensity score matching, another matching technique, or an ad hoc approach to matching (e.g., match by size and geographic area). I'd see what folks do in your discipline.

Without a full understanding of your data structure, no one can tell you how to "link" a delta variable. Normally, we'd treat such data as a panel where the panel variable is hospital and time is year. If you do this, then when you calculate the delta it will lie in the right observation. Now, you may want to lag the explanatory variables or such such, but that is easy once you xtset your data.

Third - if you're looking at a matched sample, then you probably should look at a conditional logit. The panel variable then groups the matched hospitals including year. But this can't be answered precisely without more on your data.

[Robin Boender]

Thanks for your feedback regarding posts and, of course, for your reply. I will specify it more according to the FAQ next time. For now, I really appreciate your answer!