Effects of immigration on house prices in Canada

Abstract

From the turn of the present century until late 2008, house prices in some developed countries, including Canada, rose sharply compared to the increases in their per capita incomes. Some in the public circles of these countries argue that immigration fueled this rise. Each year, Canada admits about 225 000 immigrants, but information on the effect of immigration on house prices in this country is lacking. Our extensive econometric analysis based on panel data at census division levels obtained from the 1996, 2001 and 2006 population censuses indicates a statistically significant but small effect of immigration on prices of privately owned dwellings in Canada. An out migration of the native born from the areas where new immigrants settle, or an increased supply of housing due to expectations of higher demand in those areas may have caused this result.

Keywords:

• immigration
• immigrant integration
• housing policy for immigrants
• housing markets
• housing bubble
• social policy

Acknowledgements

We thank Dr D. DeVoretz (Simon Fraser University), Dr Charles Beach (Queens University), Dr E. Falaris (University of Delaware) and participants of a session at the Canadian Economics Association meeting, Toronto (28–31 May 2009), for their valuable comments on an initial draft. Responsibility for any shortcomings remains with us.

Notes

¹ Gradual market clearing can be incorporated into the basic stock-flow model with the addition of a price (P) adjustment mechanism ΔP = σ(P* − P), where σ is the annual percentage rate at which actual prices converge to the equilibrium price. However, we believe a 5-year gap between censuses provides a sufficient time for prices to clear housing markets completely.

² Recent studies also incorporate hedonic price determination methods (Bover and Velilla, 2002) in their empirical analyses.

³ This displacement effect can also occur through labour markets: depending on the level of substitution between immigrant and local workers, an inflow of newcomers can reduce local real wages and force native born workers to move out from the region.

⁴ We thank an anonymous referee for bringing these two points (population growths and household sizes) to our attention.

⁵ All dollar values (for price, rent and income) have been deflated to 1996 prices using Gross Domestic Product (GDP) deflators.

⁶ We will test the assumption that E[є_itє_is ] = 0 for all s ≠ t for contemporaneous correlations across cross-sectional units, as well as for serial correlation within cross-sectional units.

⁷ We have applied the Hausman test to see the appropriateness of fixed-effect models. The test results strongly reject the null hypothesis that the random effect estimators are consistent, which is also supported by a high correlation between X and u, as presented in Table 2.

⁸ Fixed-year effects take care of macro factors such as changes in the rate of GDP growth, tax regulations, interest rates and economic and political conditions over the period between 1996 and 2006.

⁹ Our econometric specifications in this article have benefited from Mayo (1981) and Saiz (2007).

¹⁰ In 2007, while Ontario still received the largest number of immigrants, its per capita inflow declined from 11 to 9 in 1996.

¹¹ The relevance of cross-sectional data in our analysis also arises from the fact that house prices are unlikely to experience the arbitrage of tradable divisible commodities, so it is unlikely that the law of one price holds.

¹² Our dependent variable, y_it is the average value in a CD of private dwelling as provided by the census data. This value is reported by the census respondent. We do not expect there to be a systematic bias in this reporting since the average values are based on a large number of observations (each of the census data that we used is based on 20% sample of Canadian population). We have also controlled the effect of housing quality variations across CDs.

¹³ We also used the density of people who were a different CD a year ago. Although the sign changes, the results are not significantly different than zero.

¹⁴ This argument also refers to whether or not immigrants are perfect substitutes of nonimmigrants in labour markets, which is one of the long-standing arguments in the labour economics literature (Borjas, 1994).

¹⁵ For example, a better quality of education and health care that may attract more immigrants may also increase housing prices.

¹⁶ Saiz (2007) uses the inflow of recent immigrants as an IV for immigration based on the fact that the role of networking is crucial for recent immigrants to pick their destination cities.

¹⁷ A wide range of studies addresses this problem in the immigration literature. Some selected ones are Duffy et al. (2005), Saiz (2007) and Card and DiNardo (2000).

¹⁸ Many studies (see Saiz (2007) for the list of studies) report that the local networks of immigrants, family, kin and friendship for newcomers tend to be more important in their destination decisions.

¹⁹ Because of the lagged immigration densities, we had to use the 1991 census for finding the number of immigrants in each of 253 CDs. We removed all CDs from the 1991 census that are not in our panel data pooled from 1996, 2001 and 2006 censuses. However, since some CDs are new and not among 253 CDs the number reduced to 235 CDs in our IV estimations.

²⁰ The correlation coefficient between recentimm/pop and immratio is 0.68.

²¹ We also estimated the same regressions in Table 6 using 2001 and 1996 censuses individually. The coefficients of immigration ratios are not significantly different from those given in Table 6.

²² During most of the period 1971 to 1996, Canada was experiencing a shift in its immigrant inflows from western Europe to Asia and Africa. During the period 1996 to 2006, most immigrants came to Canada from Asia and Africa.