Estimating fundamental and affordable housing price trends: a study based on Singapore

ABSTRACT

Policy-makers often impose some cooling measures on the housing market when housing prices rise fast. Such policies yield limited success if housing prices are driven up by fundamentals. Estimating a fundamental price trend from observed price data is a challenge. We present an empirical methodology to separate housing price trends into fundamental and affordable components. Deviating from the common practice, we replace current income by a long-run income measure constructed from household incomes at different quantiles. This income measure provides a more suitable basis for constructing affordable house price levels. It also serves as a better fundamental variable, especially for segmented housing markets like that of Singapore. These price trends provide policy-makers with useful information to intervene into property markets to achieve desirable outcomes. Analysing Singapore data using this methodology shows the magnitudes of the price gaps between actual and fundamental prices and how housing affordability fluctuates over price cycles.

KEYWORDS:

• Fundamental housing price
• affordable housing price
• lifetime income
• counter-factual simulations

JEL CLASSIFICATION:

• R21
• C51

Acknowledgements

Authors would like to thank two referees for their extensive and constructive comments on the paper. Thanks are also due to the discussants of the paper at the Seventh Joint Economics Symposium of Fudan University, HKU, Keio University, NUS and Yonsei University held at NUS, for their valuable comments on this exercise.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Subletting, downsizing and lease buyback of public housing by the government are a few options available to monetize the housing wealth. But none of these options is that popular yet.

² To cite a few: Poterba (1984), Case and Shiller (1987), Topel and Rosen (1988), Mankiw and Weil (1989), Dipasquale and Wheaton (1994), Mayer and Somerville (2000), Hwang and Quigley (2006). Leung (2004) provides an extensive literature review.

³ Singapore has a thriving rental market because of the presence of a large foreigner population in Singapore.

⁴ See Gu (2008) and Lum (2011) for a review of Singapore government policies in the housing market.

⁵ In the United States, housing affordability is monitored by three main real estate institutes and government agencies: National Association of Realtors (NAR), US Department of Housing and Urban Development (HUD) and the National Association of Home Builders (NAHB) all of which focus on short-run affordability measures. See detailed discussions in Quigley and Raphael (2004).

⁶ See Cochrane (2005) for a textbook treatment and Campbell (2008) and Engsted, Pedersen, and Tanggaard (2012) for further assessments of the transformation.

⁷ Usually these variables are expressed in real terms. The inflation effect, however, cancels out in the formulation in Equation 1. Apart from this, we prefer modelling in nominal terms because the consumer price index (CPI) tends to increase when house prices increase and the real house price (P/CPI) tends to be over–deflated, especially during housing price bubbles.

⁸ See also Abeysinghe and Jayawickrama (2013).

⁹ The opposite will happen if negative bubbles (troughs) are dominant.

¹⁰ If income (y) is lognormally distributed or log of income is normally distributed with mean $\mu$ and variance ${\sigma }^2$, then $E(y)=e^{\mu +0.5{\sigma }^2}$ and Median(y) = Md(y) =$e^{\mu }$. This gives, $E(y)/Md(y)=e^{0.5{\sigma }^2}$ and as the variance of log-income distribution increases the mean departs from the median at the rate given above.

¹¹ Unless otherwise specified, the main source of data used in the study is the STS database of the Department of Statistics, Government of Singapore.

¹² These are transaction-price based indices worked out from a representative basket of properties. The weights are based on a moving average of transaction values. These indices do not control for quality explicitly. Deng, McMillen, and Tien (2012) have proposed a matching approach to construct constant-quality private housing price indices for Singapore over 1995–2010. The major turning points and cycles captured by their indices match closely with the government indices. Therefore, the URA and HDB price indices are reasonably good indicators for the work in this exercise.

¹³ Any type of data interpolation may result in measurement errors. However, cointegrating regressions still produce consistent parameter estimates if the measurement errors are stationary.

¹⁴ This has been roughly the average prime lending rate.

¹⁵ Note that the series presented in Figure 4 of Abeysinghe and Gu (2011) are in real terms whereas those in Figure 3 in this paper are in nominal terms. Moreover, the income series of this study includes employer’s CPF contributions that were not available to Abeysinghe and Gu (2011).

¹⁶ We do not include Johansen and Hansen test results for brevity. Note that Hansen test supports cointegration in the full sample when it is used with Dynamic OLS. It seems that lags in Johansen and DOLS compensate for parameter instability observed with OLS and FMOLS estimates.

¹⁷ This measure subsumes housing accessibility, which is the ability to pay upfront payments such as a down payment. In this exercise, we are only interested in long-run affordability. We provide information on housing accessibility and affordability separately at http://www.fas.nus.edu.sg/ecs/scape/housing.html.

¹⁸ Results can be obtained from the authors.