I am researching firm level data with Stata 13. My endogenous variable is (unfortunately only) of binary nature and indicates whether a firm engaged in R&D activities or not (0;1). I use a panel for EU based firms that among others comprises the years 2006,2007,2008 and 2009. I moreover am lucky enough to have a set of moderators (such as firm size, age, industry dummies and quite a few more).

Given the turbulent period, I figured that I might run the risk of foregoing to check whether there might be a structural break following the financial crisis.

In this first attempt, I rely on a random effects discrete choice model. I ran the same regression once before and once after the crisis. In fact, the coefficients of the moderators used change quite a bit in size, significance, and in two cases even in sign. I also incorporate time dummies, and again I find significant coefficients in the years depicted.

I discussed my findings in a small group and - among others - I got the suggestion to plot the residuals over time. I do not quite get how this could indicate a structural break. Moreover I do not know how to do so in Stata 13 and - more specifically - which kind of presentation might be suitable.

What do you think? Is plotting the residuals the way forward here? And if so, how? Moreover, what do you think I could moreover do to make sure I "explain the financial crisis" away (in terms of variance)?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.