3
$\begingroup$

I have a panel data set of housing prices from different neighborhoods in a city - and wish to test whether or not a policy has had an impact on the housing prices. The policy (treatment) is simply a sign signaling whether it is a good or bad neighborhood (based on parameters which are already known).

Initially I figured I could mirror the paper by Card and Krueger (1994) with a difference-in-differences estimation however what complicates my issue is that the policy enables neighborhoods to go from good-bad and bad-good over a span of 10 years. So they could potentially get a treatment multiple times.

Any suggestions into papers which has studied policies in a similar fashion or at methods which could be applied is greatly appreciated.

$\endgroup$
5
$\begingroup$

I think a DiD will not work here as the parallel trend assumption probably doesn't hold in your case.

However, given that the treatment is based on a number of parameters that are (I assume) observed, you could potentially use some type of discontinuity design to estimate the effect of the policy around the threshold values of these parameters.

To be more precise, your treatment neighbourhoods would be the the neighbourhoods that have the sign but are very close to loosing it, while the control neighbourhoods would be the neighbourhoods that are just below the threshold for obtaining the sign. Then if prices increases are a continuous function of all other covariates, the difference in treatment vs control would measure the effect of the signs for neighbourhoods around the threshold.

Of course there are potential endogeneity issues that you'll need to take into account as you can expect neighbourhoods to do some effort to get above the threshold.

$\endgroup$
1
  • $\begingroup$ This has given me some thought - thank you for the feedback. I was struggling with whether it was the correct method or not - I will look into a discontinuity design, thank you! $\endgroup$
    – Sirmimer
    May 22 at 19:47

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.