RDP 2014-04: Home Price Beliefs in Australia 2. Background and Existing Literature

The idea that households form subjective beliefs about the value of their own home is intuitive. Unlike financial assets such as equities, housing is an asset that is relatively hard and costly to value.^[1] The main reasons for this are:

1. Housing is a heterogenous asset that is not sold on a large centralised market. Instead, there are non-trivial search costs for buyers and sellers to successfully match either through bilateral negotiation or an auction.
2. Selling or valuation costs can be non-trivial.
3. Housing is traded infrequently. On average over the past decade, around 6 per cent of the total number of homes in Australia are sold each year;^[2] suggesting an average holding period of about 16 years.
4. Housing price data are generally only available with a lag.
5. Obtaining detailed price information about values in one's locality can be expensive, and especially so if one wants to obtain updates of this information regularly.

The fact that housing is an asset that is comparatively more difficult to price than financial assets, and is a large component of households' wealth, implies that households infer or form beliefs about the value of this asset when making economic decisions. In particular, standard theory would suggest that when making either a portfolio allocation decision or choosing a consumption path, beliefs about the value of the home will influence these decisions.

This raises the main questions of interest in this paper. First, how well do homeowners' beliefs match objective measures of home prices? This is addressed in Section 5. Second, are certain household characteristics correlated with beliefs? More specifically, what characteristics explain whether a household overvalues, undervalues or correctly values their home? This is addressed in Section 6. Third, does the self-reported belief or the market-inferred value of the home matter for household economic decisions? That is, do households that undervalue their homes choose different consumption, leverage and portfolio allocations to households that overvalue their homes? This is addressed in Section 7.

A novel feature of our paper is the data we use. We gauge the accuracy of self-assessed home values using a near-census of housing sale prices as the benchmark measure of valuation. The early research instead compared estimates of housing prices by homeowners and professional appraisals (see Kish and Lansing (1954) and Kain and Quigley (1972)). However, Robins and West (1977) showed that homeowners and professional appraisers assess the value of homes with the same degree of inaccuracy.

The literature has also compared self-assessed home values to recalled sale prices. In these studies, homeowners that have recently moved are surveyed and asked to make an assessment of the current value of their homes, as well as recall the original sale price of their homes (e.g. Goodman and Ittner 1992). Local housing price indices are typically used to control for the passage of time between the current estimate and the initial purchase price. For instance, Goodman and Ittner (1992) find an implied home valuation difference of around 6 per cent (for similar approaches see Ihlanfeldt and Martinez-Vazquez (1986) and Kiel and Zabel (1999)).

In a similar vein, Melser (2013) assesses home valuation differences for Australia using household panel data. He compares self-assessed current home prices to the initial purchase prices recalled by surveyed homeowners. He finds that Australian homeowners have a positive bias of around 4 per cent in estimating the value of their homes.

The limitation of these approaches are the small samples of sale prices (generally less than 1,000 observations); their inability to distinguish between valuation bias and recollection bias;^[3] and their reliance on external indices to update self-assessed home values.^[4]

Relative to previous studies, one advantage of our approach is that we can focus on valuation bias, abstracting from either recollection bias or the use of external indices to infer a belief. Another advantage is that we have a near-census of sale prices as the benchmark measure of valuation. Finally, the timing of these sale prices are matched to the timing of self-assessed home values, thereby avoiding the use of external benchmarks to ‘update’ self-assessed home values.

Footnotes

This is gradually changing over time with the advent of internet-based home valuation tools. [1]

This estimate does not include all newly built homes and transfers within families that are not sales. This measure also masks significant differences in turnover across regions. [2]

In Appendix B we directly estimate the degree of recollection bias and find that surveyed homeowners understate the purchase price of their homes by about 3.4 per cent, on average. [3]

A recent paper by Henriques (2013) is an exception. For a panel of non-moving US homeowners, Henriques compares the growth in self-assessed home values over the period 2007 to 2009 for each homeowner to the growth in regional house price indices. She finds that the median home valuation difference on the rate of change in housing prices is around 2.5 per cent. Agarwal (2007) is another exception. The benchmark housing price data used in this US study comes from homeowners' financial institutions, with the financial institution's estimate of the market value coming from the Case-Shiller repeat sales index. However, despite both of these studies having access to market values, they only have access to self-assessed home values for particular periods. Agarwal uses observations from households who engaged with a particular financial institution in 2002, while Henriques uses observations from homeowners who were surveyed in 2007 and 2009, in the midst of the US housing downturn. [4]