Immigration and regional housing markets: prices, rents, price-to-rent ratios and disequilibrium

ABSTRACT

The study examines the relationship between immigration and regional housing markets using wavelet coherence analysis in conjunction with Granger causality tests. We find that while there are some regional differences in the relationships, the impact of immigration on house rents is consistently more pronounced than the impact on house prices and, in the short to medium term, an increase in immigration leads to a decrease in price-to-rent ratios. The study uses data sets for different regional housing markets in New Zealand, a country with one of the highest immigration-to-population ratios in the world during the sample period of 1996–2017.

KEYWORDS:

• housing bubble
• rental yields
• user cost equilibrium
• migration
• time–frequency analysis
• wavelet coherence scalograms

JEL:

• J61
• R11
• R23
• R31

ACKNOWLEDGEMENTS

The authors gratefully acknowledge the Real Estate Institute of New Zealand (REINZ) and the Ministry of Business Innovation and Employment (MBIE), New Zealand, for making the data sets available, free of charge, for academic use.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Notes

1. The use of wavelet analysis in the area of urban and regional studies has been limited, with the exception of Flor and Klarl (2017) and Fan et al. (2019), who study housing cycle synchronization across regions in the United States and China, respectively. However, wavelet analysis has been used more extensively in some areas of economics, such as energy economics (e.g., Ben-Salha et al., 2018), financial economics (e.g., Ko & Lee, 2015) and macroeconomics (e.g., Aguiar-Conraria & Soares, 2011). For some excellent discussions of the application of wavelet analysis in economics and social sciences, see Crowley (2007), Ramsey (2002), Schleicher (2002) and Flor and Klarl (2017).

2. According to the Organisation for Economic Co-operation and Development (OECD), New Zealand’s average annual immigration-to-population ratio between 2003 and 2016 was 1.14%. This ratio for other top-ranked countries of the world for these years was: Australia (0.93%), Canada (0.76%), UK (0.55%), Italy (0.49%) and Germany (0.45%). The OECD also reports that the New Zealand housing market’s price-to-rent ratio was ranked fifth in the world in 2017 (among 37 comparison countries) and its price-to-income ratio was ranked fourth in the world in 2017 (among 32 comparison countries).

3. While regional analysis in the context of New Zealand is limited, see Fu et al. (2020), who investigate the relationship between New Zealand’s airport system and regional economic growth.

4. Since the REINZ median price and MBIE median rent indexes are constructed using simple medians, they do not adjust for quality changes of properties that take place from one period to the next and over the long term. This potentially overestimates the price and rent increases in markets, and finds the changes to be more volatile than they actually are (Hill & Syed, 2016).

5. The rental data are freely available for downloading from the MBIE website under a Creative Commons licence (see https://www.mbie.govt.nz/building-and-energy/tenancy-and-housing/rental-bond-data/).

6. The regional migration data can be obtained from the StatsNZ website (see http://infoshare.stats.govt.nz) and then under ‘Subject categories’ select ‘Tourism’, then from the Tourism drop-down menu select ‘International Travel and Migration – ITM’ and then click on ‘Permanent & long-term migration by ctry of residence, citizenship, and NZ area (Monthly)’. The monthly populations used in the paper are calculated by assuming a constant monthly growth over the year using the annual regional population data obtained from StatsNZ (see http://infoshare.stats.govt.nz). More specifically, if the population is P_t–1 and P_t in years t – 1 and t, respectively, then the constant monthly population growth is calculated as: (1/12) ln(P_t/P_t–1) – 1.

7. Using inflows may not measure the change in the population of a region as well as net immigration does. We have nevertheless conducted our Granger causality and wavelet coherence analysis using inflows, and when compared with our analysis using net immigration, we do not find any significant differences in the results (see Tables C.6–C.7 and Figures C.2–C.3 in Appendix C in the supplemental data online).

8. The growth of property prices in Canterbury has been relatively subdued since the Christchurch earthquake on 22 February 2011.

9. Saiz (2007), Shi et al. (2009), Hyslop et al. (2019) and Hearne (2020) have documented large differences in regional house prices and rents.

10. We conduct sensitivity analyses of user cost and market deviation with respect to each of the components of user cost. We find that user cost is most sensitive to expected capital gain, followed by its sensitivity to interest rates and then the other three components (see Figures B.3 and B.4 in Appendix B in the supplemental data online). These perturbations of the components changed the magnitude of market deviation from equilibrium in New Zealand as a whole and in our four regions; however, they do not change our finding that the market was overvalued in the sample period.

11. Regional GDP and unemployment data in New Zealand are available quarterly. We have calculated the monthly figures by assuming a constant monthly growth over the quarter using the quarterly data obtained from StatsNZ.

12. The Granger causality tests that use the number of lags selected by SIC and HQ criteria provide results similar to the test results shown in Table 1, which use the number of lags selected by AIC criteria. ADF, PP and KPSS unit root test results and SIC and HQ results are provided in Appendix C in the supplemental data online.

13. The coherence measures are not reliable in the zone below the cone in the scalogram. This is because these coherences are calculated using filled-in zero values due to the lack of sufficient data around the starting and end points of the sample (Schleicher, 2002).