Monday, October 29, 2007

Economist's Notebook: Correlation is not Causation

Yesterday (Sunday) the Oregonian ran an op-ed piece by two OSU economists who have studied land values in areas with and without stringent land-use restrictions and found that land use restrictions have not depressed land values. From this one might be led to conclude that land use restrictions do not depress land values (as the headline writer seemed to do). But the evidence they produced doesn't actually tell us much about that.

NB: The authors are very careful to refrain from any causal statement - I am commenting on how this research will likely be presented, not on any error on their part.

To find a correlation (or lack of one) between land-use restrictions and property values does not tell us that land-use restrictions have had no affect on property values. In fact, I can imagine that places that are experiencing or are anticipating high growth rates are more prone to enact land-use restrictions. These restrictions might serve to limit growth and thus, after-the-fact, the growth rates in property values might look the same in low-growth no restriction areas. But the contra-positive (which is unobservable because it never happens) could very well be that without restrictions land values would appreciate a great deal faster than 'average.'

It is interesting that the authors point out themselves the problem with drawing causal conclusions from correlations, but in the opposite direction (they fail to point out the reverse argument that would reduce the value of their data to proponents of M49):

Because of the complex ways that land markets work, it is easy to confuse the effects of land-use regulations with "after-the-fact" opportunities created by those same regulations. For example, regulations that limit the supply of developable lands may give the impression that lucrative development opportunities abound for property owners. However, as we have demonstrated elsewhere, this perceived increase in property value is not the same thing as a reduction in value because of regulation.

To give an example supporting the quote above: suppose you have a property just outside an urban growth boundary and within the boundary property values are skyrocketing. What can you say about the affect of this land-use restriction on the value of your property? You cannot say that without the land-use restriction on your property it would have appreciated as fast as those within the boundary because in the absence of the land-use restriction, those properties might not have appreciated at nearly the rate that they did.

Still the message is the same, these are correlations - any conclusion about the affects of land-use restrictions on property values drawn from such data (anecdotal or survey data) would be inaccurate - including any claim about how land-use restrictions lower the values of private property. So, in the end, any real attempt to measure the true causal link is probably a fool's errand.

The reason I bring this up is not because I do or don't like land-use restrictions. Simply that I think the more sophisticated people become when confronted with statistics, the better the public policy that will result because you reduce the effectiveness of populist appeals.

What I would love to see discussed with M49 is the social cost / private right division. Like in my previous post (yes, 'I like to hear myself talk' is the obvious conclusion).


Jeff Alworth said...

Have you had a chance to read the full report? Is it available?

Patrick Emerson said...

Click on the link to the op-ed piece and there you will find the url (not clickable) to the full report. I have scanned it. It appeatrs to be precisely what they describe: a survey of selected properties in selected counties and a comparison of more and less restrictive land use regulations and the corresponding property values.

Causation is not asserted - but that is the conclusion supporters of M49 will want you to draw. But I would also point out that opponents of M49 are also crazy if they think they can say with any accuracy that M49 is costly for the same reasons. BTW, the fact that they 'select' certain counties and properties would not pass the sniff test for an academic paper.