Behavioural biases and nonlinear adjustment: evidence from the housing market

ABSTRACT

Using the threshold vector error correction model, this study finds substantial evidence for asymmetric mean-aversion in the Korean housing market, arising from behavioural biases. In order to effectively capture behavioural biases from the prospect theory, special attention is paid to the extreme tails of price deviations from long-run rationality. The major findings are highly consistent with the prospect theory predicting the risk-aversion effect after prior losses and the house money effect after prior gains. The overall speed of adjustment at losses is assessed about 7 times as high as at gains. The evidence that price changes are serially dependent, even after controlling the behavioural biases, suggests that house prices are also driven by investors’ extrapolative expectation.

KEYWORDS:

• Housing market
• nonlinear adjustment
• prospect theory
• threshold vector error correction model

JEL CLASSIFICATION:

• L85
• R21
• C51
• E44

Acknowledgment

The author thanks the editor and anonymous referee for valuable comments. The earlier version, entitled “loss aversion and asymmetric price adjustment in the Korean housing market”, has been presented in the KFA section at the 20th annual meeting for the allied Korean business administration associations. The usual disclaimer applies.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

¹ See Ambrose and Kim (2003) and Kim (2013) for theoretical analyses of chonsei and Cho (2006) on the Korean housing price parity. See Kim (2004) for a comprehensive survey on the Korean housing market.

² This parity is equivalent to the case of ‘pure chonsei’ in Kim (2013). Arbitrageurs may refer to flippers in Leung and Tse (2017) and they are called by the mass media as ‘gap investors’.

³ Assuming that rent R is paid at t + 1, the Euler equation is written as$S_t=E_t(m_{t+1}(S_{t+1}+R_{t+1}))$, where m denotes a stochastic discount factor. Cochrane (2005) shows that the existence of discount factor m implies the law of one price (as a result of no arbitrage) holds.

⁴ Equation (1)(1) $\left(S_t-D_t\right)=(S_{t+1}^e-D_t)/(1+r_t)$(1) is derived from the perspective of investors (flippers), based on a survey by Case and Shiller (1988). In contrast, Cho (2006) writes the arbitrage-free housing price parity as $S_t=\left(S_{t+1}^e+r_t{D}_t\right)/\left(1+r_t\right)$, which is applicable to owner-occupants.

⁵ The random walk model has had an element of truth for speculative markets including housing markets (Shiller (1991)).

⁶ Leung and Tse (2017) explore an equilibrium model about how house prices are adjusted when mismatches are mediated by flippers. They claim that house prices are state-dependent by dynamic interactions among end-user-owners, renters, and flippers.

⁷ Barberis, Huang, and Santos (2001) assume that loss and gain are measured as a deviation from the risk-free return as a reference, and that investors rationally forecast changes in the benchmark reference. Without loss of notions, the benchmark price (S_t*) is the rational price when b₀ = 0. Loss is assessed with the ratio of benchmark rational price to actual price. The use of S* in the prospect function is partly justified by a finding by Case, Shiller, and Thompson (2012) that homebuyers were very much aware of trends of home prices.

⁸ Kahneman and Tversky (1979) claim that investors are willing to take risks in order to avoid a loss. In contrast, Thaler and Johnson (1990) suggest that if investors suffer an unpleasant loss from unsuccessful prior efforts, they will subsequently act in a more-risk-averse manner.

⁹ The current definition of prospect utility is slightly different from the specification by Barberis, Huang, and Santos (2001). However, notions are almost the same. Two kinds of losses are conceivable: fundamental losses (S-S*<0) and transient losses (ΔS<0).

¹⁰ Framing refers to the way a problem is posed for investors. The process by which investors formulate such problems for themselves is called mental accounting. Narrow framing is the tendency to treat individual gambles separately from other proportions of wealth. Thus, investors with narrow framing often evaluate it as if it is the only gamble so far, rather than accumulating it to previous bets to see if the new bet is worthwhile to take. This serves as a rationale for time-varying reference points as assumed in this article.

¹¹ Barberis et al. (2015) present an asset pricing model with heterogeneous agents where some investors form extrapolative expectation and others hold fully rational beliefs.

¹² The mispricing z_t is assumed to be net of effective transaction costs within the interval [${\gamma }_1,{\gamma }_2$], where the former is the lower threshold value above which opportunity cannot cover costs (losses are not perceived by investors), and the latter the upper threshold value below which costs are not justified by opportunity (gains are not perceived by investors).

¹³ Housing data are accessible at https://onland.kbstar.com/. Corporate bond rates are accessible at http://ecos.bok.or.kr/.

¹⁴ Case and Shiller (1989) find that there is significant serial dependence of housing returns and it is too strong to be justified by arbitrage barriers. They conjecture that a housing market is driven largely by extrapolative expectations in the sense that people form expectations on the basis of past price movements rather than any knowledge of fundamentals (Case and Shiller (1988)).

¹⁵ See Christiano, Eichenbaum, and Trabandt (2018) and Stiglitz (2018) about recent debates on DSGE models.