RDP 2014-04: Home Price Beliefs in Australia 1. Introduction

As the real estate agent said, ‘Location, location, location’, and we're right next door to the airport. It will be very convenient if we ever have to fly one day. Dale Kerrigan, The Castle (1997)

Housing is the largest component of household wealth in Australia. Variation in housing prices has been shown to be important for household leverage, portfolio allocation decisions and consumption (Ellis, Lawson and Roberts-Thomson 2003; Kohler and Smith 2005; Berger-Thomson, Chung and McKibbin 2009; Windsor, Jääskelä and Finlay 2013). However, timely data on the prices of individual homes are not readily available. For this reason, households are typically required to infer or form a belief about the value of their home when making these economic decisions. As the quote above from the Australian film The Castle suggests, these beliefs can be very subjective.

This paper explores homeowners' beliefs about housing prices in Australia. Our goal is to understand differences between homeowners' beliefs and market-inferred home sale prices and whether these differences are important for economic decision-making. To this end, we estimate the average homeowner belief at the postcode level, and compare the distribution of these beliefs with the distribution of average prices inferred from transactions data. We refer to these differences between homeowner beliefs and market-inferred home sale prices as ‘home valuation differences’. We also explore whether these valuation differences are correlated with certain demographic and economic variables and whether valuation differences are significant in explaining household consumption, leverage and portfolio allocation decisions across postcodes.

Our paper makes three contributions:

1. We estimate the difference between beliefs and prices (hereafter ‘home valuation differences’) in a way that is free of recollection bias. That is, unlike the previous literature, we construct our comparisons of beliefs to prices using a method that does not rely on the ability of surveyed households to recall the purchase price of their homes. To do this, we use separate hedonic regressions on household survey data and on unit-record sale price data.
2. We regress our measure of home valuation differences on various household characteristics (e.g. age, income and education), the local area unemployment rate and a proxy for housing market information (the tenure of ownership).
3. We investigate whether the size of home valuation differences across postcodes is correlated with household spending, leverage and the share of risky assets held in households' financial portfolios.

Our approach allows us to focus not just on average beliefs about home prices, but also higher moments of the distribution of beliefs, and to relate any differences between beliefs and market values to households' economic decisions.

Understanding how well Australian homeowners assess the value of their homes is important for a number of reasons. First, self-assessed home values sourced from household surveys are the main source of data used to measure the distribution of household wealth (and related financial indicators, such as leverage) in Australia. If homeowners do not accurately value their homes, then survey measures of household wealth may be biased. For example, if home valuation differences vary systematically with age then the estimated age profile of household wealth using self-assessed home values will be biased, giving a misleading picture of the actual distribution of wealth by age.

Second, by focusing on the distribution of average differences in beliefs and prices across postcodes, our approach provides insight into alternative theories of homeowner belief formation. In particular, we consider whether beliefs are unbiased on average (rational) or whether there is skewness in beliefs that could reflect optimism or pessimism. Some models of decision-making under uncertainty that focus on factors such as robust control (Hansen and Sargent 2008; Bidder and Smith 2012) and ambiguity (Epstein and Schneider 2008) predict that some households may hold pessimistic beliefs and therefore undervalue their homes.

In contrast, Genesove and Mayer (2001) show that loss aversion may cause some homeowners to hold optimistic beliefs relative to market-inferred prices when prices are declining. Likewise, the recent literature on optimism and other rational biases (see, for example, Van den Steen (2004) and Brunnermeier and Parker (2005)) predicts that some households may hold optimistic beliefs and hence overvalue their homes. In particular, households may trade off the utility gains from optimism with any costs from making distorted decisions because of overvaluation. An appealing feature of our paper is that we can provide empirical evidence on these alternative theories of belief formation.

To begin with, we document some new stylised facts about Australian homeowners' ability to value their own homes. Home valuation differences, measured at the postcode level, are defined as the difference between the average of homeowners' beliefs about the value of their homes and the average price in the same postcode based on transactions data. Both measures are constructed by controlling for the differing characteristics of properties over which beliefs are formed, or that are sold, in any given period. This is done in two steps. First, we estimate average homeowner housing price beliefs across postcodes using hedonic regressions on household survey data. Second, we estimate average market values across postcodes using separate hedonic regressions on home sales data.

We find that homeowners' home price beliefs are unbiased on average across postcodes. In terms of the absolute differences, we find that half of the average home valuations fall within 11 per cent of the average market value across postcodes. Although beliefs are unbiased on average, we do find statistically significant differences between average beliefs and average sale prices for many postcodes. In particular, a relatively large share of postcodes are undervalued (have a significant negative valuation difference) and a relatively large share of postcodes are overvalued (have a significant positive valuation difference).

Certain average household characteristics are correlated with valuation differences. In particular, postcodes with older homeowners are more prone to overvalue their homes, on average. In contrast, postcodes in which homeowners have lived in their homes for a relatively long time or in regions with relatively high unemployment are more likely to undervalue their homes, on average.

We also explore how home valuation differences are associated with households' consumption and financial decisions. We find evidence that valuation differences are positively associated with spending, leverage and the allocation of wealth to ‘risky’ assets, such as equities, after controlling for a number of other factors, including average income and the average sale price of homes in the postcode.

Importantly, we show that our results are unlikely to be due to omitted characteristics in the hedonic regression. Our key findings also hold under an alternative approach to estimating home valuation differences using repeat sales.

In Section 2 we discuss some of the existing literature and the motivation for our research. In Section 3 we discuss the data and in Section 4 we outline the hedonic regression modelling. In Section 5 we document the key stylised facts about the distribution of home valuation differences before turning to the determinants of these differences in Section 6. We explore the correlation between valuation differences and household decision-making in Section 7 before we consider the robustness of our findings in Section 8. We then draw conclusions in Section 9.