I was reading this journal article about the decrease in the relevance of the income statement:
Basically what the author does is this:
(1) For each year, he takes that year's (say, 2010) income statement information from a set of companies as the $X$ values (independent variable), and the 1-year return (2011 share price/2010 share price) of the companies' stocks as the $Y$ values (dependent variable). He generates a regression model and calculates the adjusted $R^2$ for this model.
(2) He does the same in part (1) for 30 years
(3) He plots a graph of the adjusted $R^2$ for each year over time, and generates a regression model for it. It turns out that adjusted $R^2$ has decreased by about 10% to 7%. Thus, he concludes that the income statement has lost relevance over time.
I've seen adjusted $R^2$ used to compare different models using the same sample data. But I've never seen it used in such a manner before because the data used to generate each model is different.
Any advice would be much appreciated.