I have panel data with municipalities in years being the unit of analysis. I want to examine the effect of industrial mining on interpersonal trust on the municipality level.

The social trust data stems from a household survey conducted in 3 waves (2012, 2014, 2016). I assigned respondents from those surveys to their respective municipality in which they live, which means I collapsed the individual level survey data of each wave to the municipality level in order to merge it with the mining information of that particular year.

My question is now on the research design: I'm planning on using matching techniques. However I'm not sure how to apply it to my case. On the one hand, I have panel data in the sense that each municipality shows up in each wave, which would indicate I should apply matching to panel data even though it is usually a cross-sectional method.

On the other hand, I feel like my final data set is not a true panel in the sense that the actual data for each municipality is from the survey which does not interview the same people over time. I just aggregated the data and kind of made it a "fake panel". Thus, should I treat all the municipalities as a cross-section, ignoring the time component? Also, it seems to me that the alleged panel structure does not convey a lot of useful information anyways because the treatmeant (mining) is largely time-invariant over the time horizon of my sample making fixed effects measures pointless.

Thus, is there any value of my data being in panel structure right now? Is it valid to treat each municipality as an independent observation? Or should I rather create a data set where each municipality is recorded once and the variables are the averages of the 3 given waves (which would have the merit that it would cut down my dataset to 1/3 of the number of cases I have right now)?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.