An examination of the risk-return relation in the Australian housing market

Abstract

Purpose Extensive studies have investigated the relation between risk and return in the stock and major asset markets, whereas little studies have been done for housing, particularly the Australian housing market. This study aims to determine the relationship between housing risk and housing return in Australia. Design/methodology/approach The analysis of this study involves two stages. The first stage is to estimate the presence of volatility clustering effects. Thereafter, the relation between risk and return in the Australian housing market is assessed by using a component generalised autoregressive conditional heteroscedasticity-in-mean (C-CARCH-M) model. Findings The empirical results show that there is a strong positive risk-return relationship in all Australian housing markets. Specifically, comparable results are also evident in all housing markets in various Australian capital cities, reflecting that Australian home buyers, in general, are risk reverse and require a premium for higher risk level. This could be attributed the unique characteristics of the Australian housing market. In addition, there is evidence to suggest that a stronger volatility clustering effect than previously documented in the daily case. Practical implications The findings enable more informed and practical investment decision-making regarding the relation between housing return and housing risk. Originality/value This paper is the first study to offer empirical evidence of the risk-return relationship in the Australian housing market. Besides, this is the first housing price volatility study that utilizes daily data.

An empirical analysis of the Scottish housing market by property type

“…The estimation results are reported in Table . We found a negative and significant at the 5% level estimate of λ , indicating that investors would require a lower return to compensate for higher risk (Miles, ; Karoglou et al ., ; Lee, ). However, the values of all the three information criteria were higher and the log‐likelihood value was lower for the CGARCH‐M model compared to the values given for the CGARCH model, suggesting that the CGARCH model is preferred to the CGARCH‐M model.…”

“…For the selected model, we also estimate its GARCH‐in‐mean counterpart, which includes the conditional standard deviation as a regressor in the mean equation as follows The incorporation of the conditional standard deviation, which measures the risk, in the mean equation of the model is required for the identification and measurement of any risk–return relationship (Morley and Thomas, ; Karoglou et al ., ). Positive values of λ suggest that investors, especially risk‐averse ones, would demand a higher risk premium in return of increased risk, while negative values of λ indicate that investors would require lower risk premium during periods of high risk (Karoglou et al ., ; Lee, ).…”

“…Previous studies of risk–return relationships in real estate markets include those of Karoglou et al . (), Morley and Thomas () and Lee (). On the other hand, earlier studies of the determinants of housing prices include those of Case and Shiller (), Abelson et al .…”

“…A negative leverage EGARCH (d) parameter estimate and a positive asymmetry GJR-GARCH (c) parameter estimate suggest that negative return shocks have a greater impact on the conditional volatility than positive shocks of equal magnitude. Moreover, while the conditional variance of the GARCH model shows mean reversion to x, which is a constant for all time, the CGARCH model (Engle and Lee, 1999) decomposes the aggregate volatility into a short-term (transitory) component, h 2 t À q t , and a long-run (permanent) component, q t , (Lee, 2017) which is time-varying and slowly mean-reverting. The CGARCH model can thus be expressed as…”

“…The transitory volatility component, h 2 t À q t , converges to 0 with powers of a + b (Lee, 2017), while ϱ measures the speed of convergence to the long-run level of volatility (Karoglou et al, 2013). Christoffersen et al (2008) showed that including both volatility components allows the CGARCH model to outperform the GARCH model.…”