The accuracy of homeowners’ valuations in the twenty-first century

Abstract

We study the accuracy of homeowners’ self-reported estimates using data from the 2001–2019 American Housing Survey. By comparing sellers’ estimates with deflated transaction prices for the same properties, we find evidence that American sellers overestimate the value of their homes by 1.3% on average. The mean absolute error is equal to 15.6%. Correcting for selection bias, we find that the (absolute) valuation error is strongly related to several household characteristics, most importantly marriage and education. Moreover, we find that housing market conditions such as recent local price growth and volatility are important determinants for the size and direction of the error. In particular, volatility in house prices decreases the accuracy of homeowners in estimating their property’s value. Our results imply that transaction prices are preferred over self-assessed values in several applications such as wealth estimations and hedonic price models.

Appendices

Appendix

Theory

In this section, we derive a simple model of sellers’ valuations to see why a relationship between the valuation error and household, house and market characteristics may exist. We assume that the sales price is equal to the true market value, which the seller does not observe before the sale. The seller, however, did observe the previous purchase price $P_0$ which was the market value when he/she purchased the house at t=0. After t=0, it is costly to learn and the seller only partially observes the true market price. Therefore, the seller’s estimate of the market value is equal to a weighted average of the true market value and the previous purchase price as a reference point. Genesove and Mayer (2001) and others show that the purchase price is indeed an important reference point for households. Therefore, the seller’s estimate of the market value $V_t$ is equal to:

$$\begin{aligned} V_t= \lambda (X)P_0+ (1-\lambda (X))P_t \end{aligned}$$

(4)

The weight $\lambda (X)$ can be seen as the cost to learn the true market price. If the cost to learn the true market price is higher, more weight will be given to the previous purchase price. The weight $\lambda (X)$ is a function of seller, house and market characteristics X. Indeed, for some households (lower education, low income,...) the cost to learn the true market price may be higher due to a lack of knowledge or connections with knowledge about the true market value. The ability to learn the true market price also depends on the number of comparable sales in the neighborhood. In markets with more comparables (such as urban areas), sellers’ valuation will be more accurate, in line with our empirical findings. For atypical houses with fewer comparables or markets with volatile price evolutions, it will be more difficult to obtain the true market price. The valuation error as a percentage from the true market price is then equal to:

$$\begin{aligned} \frac{V_t-P_t}{P_t}=\lambda (X)\left( \frac{P_0}{P_t}-1 \right) \end{aligned}$$

(5)

Assuming that prices grow at a constant annual growth rate of g such that $P_t=(1+g)^tP_0$, we can rewrite as:

$$\begin{aligned} \frac{V_t-P_t}{P_t}=\lambda (X)\left( \frac{1}{(1+g)^t}-1 \right) \end{aligned}$$

(6)

Therefore, the valuation error depends on the seller, house and market characteristics as these characteristics affect the cost of learning the true market price. In line with our empirical findings, higher price growth g and the longer a homeowner is in possession of the property t have a negative effect on the valuation error as it will increase the difference between the current market price and the reference point (original purchase price).

In our empirical analysis, we estimate the following reduced-form model to study the relationship between the valuation error and the different characteristics:

$$\begin{aligned} \frac{V_t-P_t}{P_t}=\beta X + \gamma g + \varepsilon \end{aligned}$$

(7)