Panel data: Static or dynamic?

[UA Munira]

Hello everyone,

I am working on a panel data where my dependent variable is claim rates and independent variable is sentencing (sentencing is a categorical variable with 0 or 1). I am trying to determine if sentencing affects the subsequent years' claim rates or not (after one year, after two years and after three years of sentencing). The question is does past sentencing affect the future claim rates of a industry. I was using static model (fixed effect) but did not get the significant result. Few days ago I just learned about the dynamic modelling. I am wondering which model should I use here static or dynamic?

Thank you for you help in advance.

[Phil Bromiley]

You didn't get a quick answer. As you may have read, you'll increase your chances of a helpful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex. I don't understand what claim rates and prior sentencing means. If claim rates is limited (e.g., between 0 and 1), you may need to take that into account.

Dynamic vs nondynamic is not a simple question. In a dynamic model, you look at the effects of variables holding constant the past level of the dv. There is a literature in political science I think saying that this can give you very different results than non-dynamic models. For example, holding past level of democracy constant throws out the immense stable differences in level of democracy and only looks at the minor variations over time. If you include a lagged dv, you must use an appropriate control for endogeneity.

The issue is really about your research question and the structure of the world you're modeling - we're not really qualified to make judgments on those issues.