Hand waving, model making, and variable lags

I would suggest that anyone who is interested in economic models reads this post, namely because I agree with it wholeheartedly

While it might seem cool to run some regressions and get a result that you believe will tell you the future it is important to realise what your implicit assumptions are.  If you add a variable to you model and it is significant but you don’t know what it has to do with anything you should be careful.  This is how using a lag functions.

You need to ask yourself why is this lag significant? what is the process behind it? Data can only really be used when you can frame it with a model of how the world works.

Personally, I hate using too many lags unless I have an understanding of why the lag matters.  When you do an empirical model it isn’t just the “significance of the variables” that matters – it is the believability of the implicit model that it represents.