Conference – 2012 General Discussion

Participants commented on the paper's recommendation for central banks to begin conducting regular surveys of housing price expectations. One participant said the Bank of Japan has been conducting such land price surveys for over 10 years. However, they noted this period in Japan has been characterised by price declines and was, therefore, a poor case study for analysing the usefulness of these data in assessing the degree of extrapolative expectations during periods of rapid housing price appreciation. The same participant also picked up on the issue of a ‘four-year memory’, where housing price expectations are estimated using the lagged four-year average rate of housing price growth. The participant said while a ‘four-year memory’ was a useful rule of thumb for policymakers, what is needed is a theoretical foundation. In response the paper's author, John Muellbauer, said the ‘four-year memory’ was just a well-established artefact of the data and therefore seemed to be a serious candidate for doing exploratory econometrics. Another participant echoed the paper's plea for public institutions to collect survey data on housing price expectations. The participant said requests for such data from a central statistics bureau in the past had been rejected and found it perplexing and interesting that there was very little motivation to collect these data.

In regard to the proposed ‘risk-adjusted’ user cost approach, one participant sought clarification on two problems they had encountered when considering the approach using Chinese data. First, the data exhibit too much noise, posing a problem for identification, and second, this approach is problematic because it is possible for the user cost to be negative, and the log cannot be calculated. In responding, Prof Muellbauer referred to the application of the risk-adjusted user cost approach to French data. In this case, transaction costs combined with a risk premium are large enough to ensure the user cost was positive throughout the sample. He also said non-linearities are so important in these data that they solve the identification problem, noting a time-varying approach adjusting for risk – where risk increases in response to recently experienced volatility – outperforms alternatives, which include not risk-adjusting or taking a linear version of the model.

The feedback channels through which housing price declines could affect the real economy were the subject of much discussion. One participant reaffirmed the importance of the local government channel in China given local governments' reliance on land sales for revenue. However, the participant wanted more evidence for the proposition that lower housing prices could stimulate consumption in the Chinese economy. Using Japan as an example, the participant said very low housing prices have not boosted consumption in the Japanese economy. The participant also said while Japan does not exhibit housing wealth effects on consumption, it does have a very strong property market investment channel particularly for small and medium-sized firms that, in the Japanese bubble economy, speculated in the land market. Regarding this investment channel, Prof Muellbauer agreed that credit liberalisation had affected companies' responses to real estate price fluctuations rather than households. However, he disagreed with the notion that there were no housing wealth effects on consumption in Japan. Citing previous research, he said a stable consumption function had been estimated for Japan covering the last 50 years and, with appropriate controls, there was a slightly negative wealth effect. Prof Muellbauer indicated that this result was robust across the age distribution. Another participant shared their insights regarding the housing wealth effect on consumption. For Hong Kong, the participant said empirical estimates from micro-data suggest that wealth effects are predominately caused by a precautionary saving motive rather than credit constraints. They said the dominance of the precautionary saving motive suggests consumption feedback loops from the housing market to the real economy are broadly based and not only transmitted through credit constrained households.

On the topic of instruments for macroprudential policy, such as maximum loan-to-value ratios, one participant expressed concern about the effectiveness of these policies for mitigating risks in both the residential and commercial real estate markets. The participant said if observers and policymakers are intent on squeezing risk out of the residential market with such instruments, they cannot ignore the possibility that risk-taking behaviour would go elsewhere, such as loans to other parts of the property development sector, including commercial real estate. Moreover, it was noted that it is within these sectors that the risks are more correlated. Spain and Ireland were proffered as examples of where banks' lending books became untenable primarily due to bad loans to the property development and commercial real estate sector. In sum, therefore, the participant said policymakers needed to balance any focus on macroprudential tools for mitigating risks in the household sector with the realisation that these tools may not be effective in other areas of the property market and may indeed increase risks in these sectors. Prof Muellbauer agreed, but noted the performance of residential mortgages in, say, Ireland, is still very important for banks' balance sheets.

Tangential to these comments, another participant asked whether the policy implications borne out of the models were robust enough to make specific recommendations to regulators. Prof Muellbauer said he did not have a complete model of the system, and as such any recommendations or forecasts would be piecemeal. He said work completed for the United Kingdom on mortgage defaults is robust; this model has been running for a couple of years and has recently been revised for the government. It outperforms other forecasts for arrears and possessions and the risks that it flags are correct.