This line from the latest Case-Shiller house price report caught my eye:
Las Vegas and Miami share the dubious title of the weakest markets in January, reporting double-digit annual declines of 19.3%.
Now, what do Las Vegas and Miami have in common? Yup, they are locations for two of the three CSI franchises.
Even if we allow for the fact that prices in the third CSI area - New York - are holding up relatively well, the CSI effect seems big. Prices in the three CSI cities have fallen by an average of 14.8% in the last 12 months, compared to an average fall of 8.7% in the other 17 metropolitan areas covered by the Case-Shiller indices.
Coincidence? No. There's an obvious reason for this - the salience effect. CSI causes people to exaggerate the extent of violent crime in its cities, causing people, at the margin, to avoid them.
Granted, the CSI effect doesn't seem significant in conventional statistical terms. Regressing annual price changes upon a CSI dummy (1 for Las Vegas, Miami and New York and zero for the other 17 cities) yields a p-value of 15.1%, against the null hypothesis of a zero effect. But remember:
1. Why should we be classical statisticians about this? If your Bayesian prior is that there is a CSI effect - as theory predicts - the evidence provides some corroboration.
2. Don't confuse statistical and economic (in)significance. As we've seen, the effect is economically significant - worth over $15,000 for a house priced at $250,000 this time last year.
3. CSI is not the only show affecting perceptions of crime and therefore prices. If we add a Numb3rs effect (giving LA a dummy value of 1 as well), then the p-value for a combined CSI+Numb3rs effect drops to 5.9% - significant at the 10% level, as they say. Papers have been published with less.
Pedants will moan that this is just a nonsensical excuse to carry a picture of Emily Procter. But what do they know?