Comment of the Day: The Wall Street Journal, the University of Chicago's Booth School of Business, and Stanford University's Hoover Institution Have Major, Major Quality_Control Problems: "And this, kids, is why it seems that you need to force economists to take a philosophy course...:
...before they can be allowed to practice.
What is Cochrane's mental model here? That social reality is forced (by god?) to match a particular function, which happens to be an exponential plus a whole lotta noise?
In physics, where we have some understanding of underlying reality, it is often (not always) legitimate to extrapolate a curve beyond measurements because we have good reasons to believe certain properties about the extrapolating function.
But none of that sort of reasoning holds in the social sciences. All we can say there is forms of summary statistics --- the DATA is well represented by...
To go beyond that to extrapolation makes very little sense. It might be justifiable IF you have some sort of (validated) theory of what's generating the distribution of data. This sort of thing can work (under limited circumstances) in finance, or even in macroeconomics. But Cochrane has no quantitative theory of the casual relationship between law and GDP beyond this data summary, and so there is UTTERLY no reason to perform his extrapolation.
Or to put it in terms he should understand:
Suppose I take data points from the left hand side of the Laffer curve and fit a nice little line through them (or hell, let's add some more functions --- a line plus a log, or a line plus a square root). Is Cochrane then happy with my confident prediction that if we tax at a rate of 90% we can raise massive amounts of revenue and, hell, let's ramp up the tax rate to 150% and raise revenue like there is no tomorrow? If we're going to be guided PURELY by curve-fitting, not by any sort of underlying theory, then what's wrong with this analysis?"