In lieu of an abstract, here is a brief excerpt of the content:

Brookings-Wharton Papers on Urban Affairs 2005 (2005) 89-95



[Access article in PDF]

Comments

[Return to Article]
[References]

Steven M. Sheffrin: Wasi and White use an econometric approach to quantify the lock-in effect that stems from the assessment provisions of Proposition 13. These provisions limit increases in property taxes to 2 percent a year as long as the owner of the property does not change. The authors' econometric method is based on analyzing the differences in the behavior of homeowners and renters in California versus Texas and Florida, two other states that experienced rapid growth and immigration over the last two decades but did not have the assessment features of California. The underlying data indicate sharp increases in tenure for both homeowners and renters in California relative to Texas and Florida. The authors find that these differences remain after they introduce a wide range of controls, although the magnitude of the effects does decrease.

As background, it is important to point out that the tax benefits to homeowners under Proposition 13 depend on when their property was purchased. The research that O'Sullivan, Sexton, and I conducted indicates that the largest beneficiaries of Proposition 13 were the owners of property before the passage of the proposition in 1978 who had their assessments rolled back to the values that prevailed in 1975.1 By 1990, many of these owners were elderly. For example, in Los Angeles County, 82 percent of the homeowners older than sixty-five years had owned their homes since 1975. Purchasers of homes after that time had various experiences. In the early 1990s, for example, housing prices fell in Los Angeles County and recent purchasers did not benefit from the assessment provisions of Proposition 13. However, in the late 1990s purchasers throughout California did gain substantially.

Comparing California with Texas and Florida, Wasi and White find that Proposition 13 increases tenure overall by 1.67 years for homeowners and 0.98 [End Page 89] years for renters. With a systematic set of econometric controls—including measures for rent control and an affordability index for housing—these estimates fall to 0.66 years for owners and 0.44 years for renters. These effects are estimated over the entire range of homeowners and renters in California, not just those with largest gains from Proposition 13. In the second part of their paper, the authors look at the lock-in effects relative to the magnitude of the tax benefits under Proposition 13. For a subsidy level of $2,600 per year, as the authors find for San Jose, California, their estimates predict an increase in tenure for homeowners between 2.6 and 3.3 years.

These effects are substantially larger than in prior research. In our own work, which used simulation methods, we estimated a lock-in effect of approximately a year for the subsidy levels that prevailed in San Jose. Quigley's (1987) econometric results based on a mortgage lock-in effect were also of similar magnitude. Wasi and White's estimates are approximately three times as large as prior estimates, even with a large set of controls in the regression.

As the authors recognize, the tax subsidy of Proposition 13 increases with tenure length. Thus any factor (observed or unobserved) that increases tenure length will also increase the measured Proposition 13 subsidy. Any regression method that tries to estimate the causal effects of the tax subsidy on tenure length must cope with this important confounding effect. The authors do try to control for the endogeneity of the Proposition 13 subsidy through their econometric methods, but they may not have eliminated all of the endogeneity in their approach. This may account for the size of their estimate.

One important data note: the key tenure data in their study are based on rather broad reporting intervals, and the authors use the midpoint of the intervals in their empirical work. This raises some econometric issues that could be addressed: Does the use of this procedure cause any potential bias in the results? Should the midpoint be used as a point estimate or should the points in the interval be...

pdf

Share