Panel Data Regression Analysis - Whether to Use Panel Structure?



Hi guys, a very conceptual question while planning a research paper.

GOAL: I am investigating the effects of enrollment in insurance programs on health inequality.

For insurance programs, there are three different programs rolled out in different years. I imagine I’d stratify by insurance accordingly.
For health inequity, we are looking at medical care utilization (binary) and catastrophic health expenditure (binary) between poverty and nonpoverty groups (will be categorized in diff ways).

For all three insurance schemes, the panel data ‘sandwiches’ the enrollment year.


  1. I’m wondering if I should even use the panel data structure. For example, insurance X was rolled out in 1998. Can I just used the data pre-1998 and post-1998 as pooled data and run logistic regressions? Or, should I just use data post 1998, pooled, and do the same analysis? Or, should I somehow account for the panel structure and set the data by year?
  2. Would it be beneficial to only compare within insurance groups? What about across enrollees in all three?

I’m at a bit of loss since my econometrics isn’t that good. Not looking for direct answers, but moreso how each of these analyses would change the research question I’m looking at.