Comment by andrewla
17 hours ago
It is an interesting discussion, but I don't see anything in there about how to assess the accuracy of any of the methods against some sort of objective truth. I would have to read the underlying papers I think to get at this, but I don't feel a strong need because I feel like the larger epistemic point is unaddressed in any of the summaries.
The closest I saw was one study that compared the model to a human generated data set, which is just kicking the can down the road. The article semi-concludes
> I think it's quite plausible but not a slam dunk. That said, if the objection is, "valuing land separately from improvements is fundamentally impossible, and we can never get better at it, so we shouldn't try," I think that's plainly ruled out.
I do not agree with this assessment -- you can create a bunch of models and show that the models have good intra-model agreement, but the fundamental point has not been touched.
Maybe but at some point, isn't your problem the same valuing a single dollar? We can compare the USD against other fiat currency, but they're all valued against the USD or some other bucket of fiat currencies. Or, you could try to tie the value of the dollar against a bucket of goods, but that's also just moving the goalposts around.
Now we're wandering a bit far afield, but estimates and measurements like this are fine if they are designed for an operational purpose. If I want to know the unimproved value of my land so that I can evaluate whether to buy a similar property and build a similar house on it, then it's fine for me to use whatever estimation protocol is going to help me make a decision. If my protocol is bad then my estimates will be bad then <shrug emoji> it's my problem.
Similarly trying to measure the value of the dollar -- what is the operational purpose of that measurement? This is a real problem in any sort of macroeconomic analysis, and Goodhart's law makes it far far far worse when trying to apply it for practical purposes. Mostly you have to accept that there is not going to be a quantitative metric that captures the underlying squishy concept so better not to think about the problem of, say, inflation, in purely quantitative terms.