The impact of UK household overconfidence in public information on house prices

ABSTRACT

We investigate if house prices are affected by the overconfidence of households who predict house prices using imperfect public information about economic outlook. For this purpose, we develop a new measure of household overconfidence in the Bayesian framework. For the three variables we test – changes in consumption, stock returns, and changes in human capital, we find that UK households were overconfident about the signals of consumption regardless of regions. However, households in London were overconfident about the signals of stock markets whereas those remote from London were overconfident about the signals of human capital. The results of household overconfidence appear positive in the UK housing market for our sample period from 1980 to 2018, in particular, 0.5% per quarter in London.

KEYWORDS:

• Overconfidence
• public signals
• households
• house Prices
• Bayesian updating

Acknowledgments

We would like to thank participants at the 52nd AREUEA Annual Conference and the AREUEA International Conference for helpful comments.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. http://www.imf.org/external/np/speeches/2014/060514.htm.

2. Overconfidence has been interpreted differently in the literature. Moore and Healy (2008) define overconfidence in three different ways: as over-placement, over-estimation, and over-precision. Over-placement (better-than-average) tends to be found in easy tasks, whereas overestimation is closely related to optimism with respect to expectations.

3. See the Appendix for the pricing model we drive from the Epstein-Zin utility. The popular hyperbolic absolute risk aversion class of utility functions is limited, as it does not properly distinguish between intertemporal substitution and risk aversion, and thus cannot explain a large shift in consumption over time caused by a small increase in the interest rate (or discount factor) when the elasticity of intertemporal substitution in consumption is large.

4. If random variables $X$ and $Y$ follow a jointly normal distribution and their standard deviations and covariance are ${\sigma }_X^2$, ${\sigma }_Y^2$ and ${\sigma }_{XY}$, the conditional expected value of $X$ given $Y$ is $E\left(X|Y\right)=E\left(X\right)+\frac{{\sigma }_{XY}}{{\sigma }_Y^2}\left\{Y-E\left(Y\right)\right\}.$

5. For a recent survey on behavioural real estate, see Salzman and Zwinkels (2017).

6. As in most studies for irrational price movements in housing markets (Salzman & Zwinkels, 2017), households in the resident property market are not just price takers. When households buy or sell their houses according to their expectation, ex post realised residential property returns reflect their posterior expectation.

7. Later in the empirical tests, we compare the ex post property return $r_{it+1}^b$ with the biased posterior expectation $E_{\hspace{0.17em}}^b\left(r_{it+1}|S_{kt}s\right)$ as well as with the rational posterior expectation $E\left(r_{it+1}|S_{kt}s\right)$. The empirical results show that $r_{it+1}^b$ is indeed closer to $E_{\hspace{0.17em}}^b\left(r_{it+1}|S_{kt}s\right)$ than to $E\left(r_{it+1}|S_{kt}s\right)$, indicating that overconfident households follow their biased posterior expectation for buying and selling their residential properties.

8. The ratio $\frac{{\gamma }_{ik}}{{\gamma }_{ik}^{b_p}}$ is not affected by cross-correlation between noisy signals ($S_{kt}$s) because the effects of multicollinearity, which appear in the denominators of regression coefficients (both ${\gamma }_{ik}$ and ${\gamma }_{ik}^{b_p}$), are cancelled out when ratios are calculated.

9. This is one reason why empirical studies on the effects of private information on asset pricing are rare. Most studies focus on theoretical consequences of overconfidence about private information (Daniel et al., 1998; Gervais & Odean, 2001; Odean, 1998).

10. Most macroeconomic variables are autocorrelated and thus their past values can be used to predict their future.

11. The least squares estimator ${\pi }_{kj}$ in the regression equation in Step 3 $x_{kt+1}=\mu +{\sum }_{k=1}^K{\pi }_{kj}\left(x_{kt+1}+{ϵ}_{kjt}^{\ast }\right)+{ϵ}_{kt+1}$ is ${\pi }_{kj}=\frac{Var\left(x_{kt+1}\right)}{Var\left(x_{kt+1}+{ϵ}_{kjt}^{\ast }\right)}=\frac{1}{1+\frac{Var\left({ϵ}_{kjt}^{\ast }\right)}{Var\left(x_{kt+1}\right)}}$ which increases as $var\left({ϵ}_{kjt}^{\ast }\right)$ decreases.

12. The Nationwide House Price Index is a seasonally adjusted house price index calculated using Nationwide’s lending data for residential properties at the post-survey approval stage. Nationwide is a large provider of household savings and mortgages in the UK and is the largest building society in the world.

13. These are the 12 first level Nomenclature of Territorial Units for Statistics regions (NUTS1 regions) that are used for statistical and administrative purposes.

14. There are significant restrictions in regional data for consumption and disposable household income, particularly for the 1980 s. Due to data unavailability, we use national consumption data instead of regional consumption data for the entire sample period. For the gross disposable household income, we use national data until the end of 1989 and then use regional data from 1990 to the end of the sample period.

15. Suppose that labour income grows at the rate of g and human capital grows at the rate of r. When these two rates are constant over time and r > g, human capital is calculated as discounted future labour income: $HumanCapita{l}_t=\frac{LabourIncom{e}_t}{r-g}$. Therefore, the change rate in the human capital is equivalent to the change in labour income: $\frac{HumanCapita{l}_t-HumanCapita{l}_{t-1}}{HumanCapita{l}_{t-1}}=\frac{LabourIncom{e}_t-LabourIncom{e}_{t-1}}{LabourIncom{e}_{t-1}}$.

16. The detailed properties of the variables can be obtained from the authors upon request.

17. The results are consistent with Havranek and Sokolova (2020) who show that the permanent income hypothesis is consistent with data after correction for publication bias using a comprehensive meta-analysis of 3000 tests of the permanent income hypothesis reported in 144 studies.

18. The results of the pooled regression with 12 regions are marked ‘All 12 Regions’ whereas those marked ‘UK’ present estimates of a univariate regression with the aggregate data. The difference between the two lies in how each region is weighted: the pooled regression treats each region equally whereas the ‘UK’ data are value-weighted.

19. As explained in Proposition 1, $\frac{{\delta }_{ik}}{{\delta }_{ik}^b}=\frac{{\gamma }_{ik}}{{\gamma }_{ik}^{b_p}}$ regardless of cross-correlation between the three variables. Thus we use ${\rho }_{ik}$ estimated with ${\delta }_{ik}$ and ${\delta }_{ik}^b$ to calculate ${\gamma }_{ik}$ in the multivariate regression.

20. Many behavioural studies that try to explain boom-bust price patterns in housing market have not been published partly due to the difficulties in estimating fundamental values of houses. In order to overcome this problem, we use a different approach which compares the responses of property returns to contemporaneous explanatory variables with those to the signals of these explanatory variables available to households.

21. London housing market is indeed distinct from other UK markets. For example, our results may be affected by concentration of foreign investment in London than any other regions.

22. These differences in London and Scotland are significant at the 1% level. The detailed results of Table 6 can be obtained from the authors upon request.

23. Stock returns are not affected by a different proxy for human capital changes.

A1. When a random variable is conditionally lognormally distributed, its expected return is $log{E}_t\left[X\right]=E_t\left[logX\right]+\frac{1}{2}va{r}_t\left[logX\right]$ (Campbell, Lo, and Mackinlay, 1997, pp. 306–307).

Additional information

Notes on contributors

Soosung Hwang

Soosung Hwang is a professor of Financial Economics at College of Economics, Sungkyunkwan University, Seoul. His main research area is asset pricing in different markets such as equities, real estate, and emerging markets. In particular, he is interested in behavioural aspects of decision makers and financial econometrics. Soosung has published more than 40 academic papers in leading finance journals and top real estate journals such as Real Estate Economics and Journal of Real Estate Economics and Finance, and served as referees or editor/associate editors for various finance/economics journals. He has extensive work experience in finance industry: he has worked as a fund manager/consultant for several hedge funds and investment banks in London and Seoul.

Youngha Cho

Youngha Cho Youngha is a senior lecturer at School of the Built Environment, Oxford Brookes University, UK. She hold a PhD from London School of Economic UK. Her research expertise lies in housing and land policy, macro and micro market anaysis, finance and mortgage market, housing need and affordability, Intermediate tenure and residential mobility, including statistical modelling and large data analysis. She has over 15 years of related research and consultancy experience for a range of research bodies including the Economic and Social Research Council (ESRC), Joseph Rowntree Foundation (JRF), Department and Community of Local Government (DCLG), University of Cambridge, Korea National Housing Corporation, So-Gang University, and Korea Research Institute of Human Settlement (KRIHS) in Korea. For past 10 years she has conducted several international comparative research with collaborators from US, Hong Kong, Singapore, South Korea and Southern African countries. She has published several research papers in leading journals such as Journal of Housing Economics, Journal of Real Estate Finance and Economics, Real Estate Economics, Journal of Housing and Built Environment and Construction Management and Economics journal. She is an active member of variety of academic research bodies, such as ENHR, APNHR, ERES, AsREA and AREUEA

Jinho Shin

Jinho Shin is a big data analytics at Kakao Bank which is one of global major internet banks, Seoul. His main research area is asset pricing in financial markets and real estate markets. He has published several papers in real estate journals and economics journals. He has extensive work experience in banking and finance industries, for example, he has worked at Standard Chartered Bank, Korea Credit Bureau and other financial institutions for over 18 years.