A new issue of JQAS just came out, with this fun little piece. It uses structural break models to analyze exogenous shocks in Barry Bonds's (Bonds'?...I have the same issue when using my own name: what's the grammatically correct choice here?).
Now, let's begin by saying that the answer to whether there was a structural break here is pretty obvious. But since my entire dissertation was about structural breaks (an example plot here 3rd one down) in the 4 North American Leagues (and one of the authors of this paper designed the main statistical test I used for stationarity in the presence of breaks), it's always fun to see the method in action.***
I have not read through the entire paper; however, I can offer up a few notes about the method itself.
First, it can be very sensitive. While they do not use exactly the same method to model the breaks that I did, I can say that very small changes in the data have impacted my analyses rather unexpectedly (see the downward shift earlier in Bonds's career that seems to be a product of one huge outlier).
Secondly, defining an "exogenous" shock is not as straight forward as you would think. In the context of statistical modeling like this, an exogenous shock can be anything not included in the model. So, if you model attendance of the Miami Heat without a dummy variable for their NBA Championships, you'll probably find an "exogenous" shock in your model around there. But we all know where this came from.
With that said, these models can be very useful when we want to look at stationary segments and avoid issues with unit roots due to one time changes. They can also be useful when we don't have an easy explanation for what happened. And they're always fun in a descriptive sense (as shown in this paper).
***Add in that the first game I ever saw in the Big House was the Appy
State game, and Strazicich does his work there, makes for fun coincidence as well.