In recent years Australia has experienced high rates of immigration. We investigate the effect that immigration has had on home prices at the postcode level. The endogenity of immigrant inflows is accounted for using the Bartik shift-share approach. Using data from the censuses in 2006, 2011 and 2016 we find that immigrants raised home prices by around 0.6 percentage points per year compared with if there had been no immigration. This effect is not large—average annual home price appreciation was around 5% per annum—but neither is trivial. The effects of immigration on home prices were larger in the more recent part of the period examined and strongest in the state of New South Wales (excluding Sydney), Victoria, and the city of Canberra. Chinese immigrants were shown to have a stronger positive influence on prices than other immigrant groups.
Keywords: Immigration; Housing Prices; Australia.
Australia is a nation which has been heavily influenced by immigration over many years. Over the past decade the population has grown by around one percent per year on the basis of the net overseas inflow to Australia. The net flow of immigrants peaked at almost 1.5% in 2008 (Figure 1). Australia’s immigration rates are relatively high by international standards. This is reflected in data from the UN which shows that, as of the most recent data in 2015, almost 30% of Australia’s population was foreign-born. Other comparable countries have lower rates of foreign-born residents. Just over 20% of the populations of New Zealand and Canada are foreign-born while the figure is around 15% for the UK, US and Germany. Moreover, it is clear from Figure 2 that the proportion of foreign-born residents has increased more rapidly in Australia than in the other countries charted over the past decade.
Immigration has a range of effects on an economy. Perhaps most notably, it is a significant source of economic growth. Indeed, the high immigration rates in Australia are arguably one of the reasons why the Australian economy has performed so strongly—the country has been recession-free since the early 1990s. The strong growth effects of immigration reflect the fact that Australia’s migration intake is selective. Hence new immigrants usually bring with them large amounts of human and/or financial capital. The economic effects of immigrants in Australia on output, and in particular the labour market, have been well documented (Harrison 1984, Peter & Verikios 1996, Hugo 2006, Boubtane et al. 2016). In this paper our focus is on a less well research area. We investigate the impact that immigration has had on Australia’s housing market—more specifically on housing prices. This is an important policy question. It goes to whether there higher levels of immigration may make housing less affordable for locals—housing affordability has been a key policy focus in recent years as home prices in a number of Australian cities have risen significantly. Indeed there has been much media interest on this aspect of immigration (Daley et al. 2018, Macaulay 2019). The prevailing presumption of much of this discussion is that Australia’s large inflow of immigrants has bid up housing prices. While this view is not unreasonable we note that the effect of immigrants could also be in the other direction. That is, the inflow of immigrants to certain areas could lead to locals leaving the area (Aydede 2017). This in turn could lower prices. As we note below, some of the international literature on this topic has found just such an effect.
In Australia, there has been limited research on the question of whether immigration has impacted house prices. However, there some related work. Abelson et al. (2005) investigated the housing market of Australia between 1970 and 2003 and noted that migration rates positively affect housing prices. They argue that this could be due to changes in the age distribution of the population thus affecting the rate of home-buying. Furthermore, Kohler & van der Merwe (2015) studied the long-run trends in Australia’s house price growth across states. They found that migration inflows influence housing markets by impacting demographics, by changing fertility rates, raising incomes and changing average household size.
Perhaps the Australian research most related to our own is that by Wokker & Swieringa (2016) of the Australian government’s Federal Treasury—the department which provide economic advice to the government. They used confidential data to look at the relationship between foreign investment review board approvals and home prices at the postcode level over the period 2010-15. In Australia non-residents are required to obtain approval to purchase new or existing dwellings. It is generally the case that foreigners are given approval, without conditions, to purchase new dwellings. But it is very unlikely, though not impossible, that approval would be given to purchase an existing dwelling. They find a positive relationship between foreign investment approvals and home price appreciation. Though they note that the effects are quite small. They estimate that only around 0.6-1.0% of the total price rise was due to foreigners.
While the work of Wokker & Swieringa (2016) is useful there are a few caveats around it. First, as the authors note, there is a potential endogeneity problem with home prices, immigrant location choices and hence housing investment approvals. This is an issue that is widely appreciated in the literature and which we discuss at length below, but is not addressed by the authors. Second, the focus only on foreign investment review board decisions is perhaps too narrow. Many immigrants may have permanent residency, or obtain it reasonably soon after arriving. In this case they do not have to apply to purchase housing but can still of course impact home prices.
There is a large international literature looking at the impact of immigrants on housing markets which we draw upon. Perhaps the seminal work was Saiz (2007) who studied the impact of immigrant-inflows on the US housing market. He ran a regression of house rents (and prices) on the immigrant inflow relative to existing population with data for a number of US cities. The key contribution of Saiz (2007) was in outlining an approach to deal with the endogeneity problem using instrumental variables. The problem arises because there are likely to be many unobserved factors which influence home prices and the location choice of immigrants. In this case a simple regression of prices (or rents) on immigrant inflows will yield biased and inconsistent results. Saiz (2007) proposed using an instrument for endogenous variable, immigrant inflow, constructed using the shift-share or Bartik approach. This approach is now taken by much of the literature on this topic. It is also the approach that we take. It is discussed in more detail in the following section. The basic idea is to construct an instrument using previous migration patterns, which are exogenous to current price developments, to predict current migration patterns. Using his approach, Saiz (2007) found that an increase in the immigrant inflow per annum equal to 1 percentage point of the initial population shifted rental and housing prices by 1 and 2 percentage points per annum respectively.
A range of researchers have applied these methods to other countries. In a recent study on Canada, Akbari & Aydede (2012) found a small but significant effect of immigrants on housing prices. Their model is built using census data from 1996-2006. They find an increase of about 0.1 percentage points in home prices for a 1 percentage point increase in the immigrant-to-population inflow. Gonzalez & Ortega (2013) find larger effects in Spain. They examine the period from 2000 to 2010 and find that immigration led to a 2 percentage point increase in the annual appreciation rate of housing prices. A similar effect is also found for Switzerland by Degen & Fischer (2017). They studied housing markets in 85 regions of Switzerland from 2001-2006 states and found that a 1 percentage point increase in the immigration inflow leads to home prices increasing by about 2.7 percentage points. This was about two-thirds of their annual increase—clearly a sizable impact.
In the UK, however, the story is quite different. Sa (2015) model the housing market of the UK and find that immigration led to falls in home prices. An increase in immigrants equal to 1% of the initial population decreases housing prices in that region by about 1.7 percentage points. They investigate the mechanism by which this effect occurs. They find that the arrival of the immigrants tends to lead to the exit of the locals while it also lowers the average income level. More recent research on the UK by Braakmann (2019) reaches similar conclusions. The author finds that prices for homes below the median tend to fall as a result of immigrant arrivals, though only modestly, with homes above the median being unaffected.
Stillman & Mar (2008) examine the impact of immigrants on the New Zealand housing market. They use census data from 1986-2006 and find a positive correlation between immigrant inflows and housing prices. In particular, they examine the impact of immigrant inflows on the population growths and find that international migration inflows could indirectly be impacting on housing prices appreciation in New Zealand via its effect on population. Coleman et al. (2007) takes a different approach and uses a structural vector auto-regression model on two data time windows from 1962-1982 and 1991-2006. They find that a net immigrant inflow equal to 1% of population led to a 10 percentage point increase in house prices. This is clearly a massive effect. They argue that it may result from the changes in natives housing demands and their expectations about future price growth occurring as a result of the migration waves.
In this paper, we examine the effects of immigrants on housing markets prices by comparing house price changes and immigrant inflows based on the 5-yearly censuses in 2006, 2011 and 2016. We use variation across Australia’s postcodes in immigrant arrivals to identify the effect of interest. Importantly, we adopt the instrumental variables approach of Saiz (2007). We also control for other determinants of home price appreciation such as the age profile of the postcode, its distance from the CBD, population density, housing structure type and level of education. We explore a large number of models for various regions across Australia and the major state and capital cities. In extending existing research we also look at the impact of immigrants from different countries such as China and India—the two countries which provide the largest number of immigrants to Australia. This enables us to explore the variation in the immigrant effect on home prices across an important dimension. This has not previously been addressed in the literature.
Our results indicate a small, but non-trivial, effect of immigrants on home prices. We find that home prices would have been around 0.6 percentage points per year lower had there been no immigration over the period examined. This compares with average home price rises over the period examined of around 5% per year. Our results are broadly consistent with the international literature. Most of the literature finds a positive effect—as have we—though our estimates are at the lower end of the spectrum in that immigrants had a relatively modest impact on home prices in Australia. However, in saying that Australia has relatively high home prices relative to its international peers. The median home price in 2016 in the urban postcodes that we examine was around $730,000 (see Table 1). Thus
our estimate of 0.6 percentage points amounts to around $4,400 per year in 2016 dollars. Importantly, we find there is variation in the effect of immigrants across time, property types, states, cities and by source country of the immigrants. Interestingly, the largest effects appear to be outside the main cities. Though Melbourne and Canberra have sizable effects. Larger effects of immigration are also observed in the more recent half of the data and for houses, compared with units (i.e. apartments/condos). Turning to the impact of Chinese and Indian immigrants, compared with all other immigrant groups, both these groups tend to impact in a more strongly positive way on price. Chinese immigrants had an impact almost twice as large as the effect of Indian and other source-country groups. Though, notably, this impact was larger in the first half of the sample. Indian immigrants had an impact which was more commensurate with that of other immigrant groups.
In section 2 we outline the methodology that we adopt. In particular we discuss the instrumental variables approach to estimation. Section 3 discusses the data used and outlines how home prices have evolved across Australia’s cities and states. It also provides some background on the other variables used in the model. Our results are discussed in section 4. We then conclude in section 5.
We adopt an approach which is fairly consistently used in the literature (Saiz 2007, Gonzalez & Ortega 2013). We estimate a model with the log change in home prices, ∆ ln hpt = ln hpt − ln hpt−1, as the dependent variable and immigrant inflows relative to population in the prior period—along with other variables—as the explanatory variables. Our primary interest is in the coefficient α in the model below—α represents the response of home prices to an increase in foreign inflows.
\[\begin{align} \Delta \ln h_{pt} = \alpha \left( \frac{\text{Immigrants}_{pt}}{\text{Population}_{pt-1}} \right) + {\bf x}_{pt-1}^T \boldsymbol{\beta} + \delta_{t} + \gamma_{c(p)} + \epsilon_{pt} \end{align}\]
The time periods (\(t\)) in our estimation will be the census years, \(t = 2006, 2011, 2016\). The model will be estimated across postcodes, denoted \(p\). So the dependent variable is the census-to-census postcode-level change in home prices. Identification of the effect of interest comes from variation across postcodes in the proportion of foreign immigrants and the extent to which this generates differential changes in home prices. The precise data used, and the lagged controls (\({\bf x}_{pt-1}\)), are discussed in detail in the following section. Note that we include dummy variables for time, \(\delta_t\) and for the state in which the postcode is located, \(\gamma_{c(p)}\). This is to control for any state- and time-specific fixed effects that may be driving price change.
We extend the model shown in equation (1) by breaking up the immigrant inflow term into inflows from certain countries. In particular from China and India—the two largest source countries for immigrants in Australia over the period. As well we include a third term which reflects the effects of immigrants from all other countries. This provides a basis for exploring the differential impact of different groups on home prices. Our models are estimated over a number of different geographic dimensions and for both house prices and unit (i.e. apartment/condo) prices.
However, a key issue with equation (1) is that the proportion of immigrants is likely to be an endogenous variable. That is, it is likely to be correlated with the error term. The problem of endogeneity in this equation arises from the likelihood that there are many variables which are omitted from the model but determine house price growth and are also related to immigrants’ location choice. There are many such omitted variables but to make the problem concrete consider the case of an amenity shock. Data on the amenities of a postcode are not available. Hence any amenity shocks will end up reflected in the error term. A positive amenity shock—such as a new school or train line or public park—will tend to raise prices but it will also likely attract more immigrants. In such a case, if we simply estimated model (1), we would get biased and inconsistent estimates of the effect of immigrants on home prices. This problem has been widely appreciated in the literature. The solution is to use instrumental variables and we now turn to this issue.
In order to consistently estimate equation (1) we need an instrumental variable for the ratio of a postcode’s immigrants in period \(t\) to the postcode population in the prior period. The approach we take, and that widely used in the literature, is to construct an instrument using the Bartik method (Bartik 1991, Saiz 2007).
The idea of the Bartik approach in this context is to use aggregate immigration to Australia from various countries to predict immigration at the postcode level. Here we use a postcode’s national share of existing residents who previously migrated from a particular country to determine its share of total new immigrants from that country. This is based on the argument of Altonji & Card (1989) that new immigrants are inclined to reside in areas where other immigrants with the same nationality have already settled. Recent evidence from Peeters & Chasco (2016) provides a more nuanced discussion of the way in which immigrants choose locations and the role played by existing immigrant groups. They find that certain immigrant groups, such as Chinese, do tend to cluster into areas where there are established groups from a similar background. While for a few ethnic groups the effects of established groups is weaker or even negative. If new immigrants do not follow the paths of previous immigrants then this could weaken the Bartik instrument approach. Though we provide evidence for our data, which deals predominantly with Asian immigration to Australia, that this Bartik-type instrument is strong.
The mechanics of the process of constructing the Bartik instrument are illustrated in the equations below:
\[\begin{align} \mathrm{Immigrants}_{pct}^* = \left( \frac{ \mathrm{Population}_{pct-1} }{ \sum_{p=1}^P \mathrm{Population}_{pct-1} } \right) \times \mathrm{Immigrants}_{ct}^{A} \end{align}\]
\[\begin{align} \mathrm{Immigrants}_{pt}^* = \sum_{c=1}^{C} \mathrm{Immigrants}_{pct}^* \end{align}\]
First, note that for each country we have a set of immigrants to Australia over a span of time. This is denoted \(\mathrm{Immigrants}_{ct}^{A}\). This is a period \(t\) value. However, it can reasonably be regarded as exogenous to anything that is happening in a particular postcode. That is, shocks to individual postcodes are not likely to have any effect on the aggregate immigration rate as there are many thousands of postcodes in Australia so each has a very small impact on the nation.
In equation (2) we share out the aggregate immigration from country \(c\) by postcode. This is done using the share of a postcode’s residents who hail from that country, denoted \(\mathrm{Population}_{pct-1}\), divided by the sum of all Australian residents who originally hail from that country. Because it is based on the population shares in period \(t − 1\) (in our case this is 5 years prior to period \(t\)) it is plausibly exogenous from what happens with home prices in period \(t\). This calculation gives, \(\mathrm{Immigrants}_{pct}^*\)—the number of immigrants which are expected to arrive in postcode \(p\) from country \(c\) in time \(t\).
Finally, in equation (3) we sum these across the various countries, \(c=1,2,\ldots,C\). This gives, \(\mathrm{Immigrants}_{pt}^*\), which is what can be used as an instrument for \(\mathrm{Immigrants}_{pt}\) in equation (1). In the case of constructing immigrant inflows from particular countries— China and India—we follow the same approach. The only difference is that we do not sum across countries as in equation (3).
Our source for postcode level housing price data over time is CoreLogic (formerly RPData). It is the leading real estate property data, information, analytics and services provider in Australia and New Zealand. CoreLogic provides data on the median price of houses and units (i.e. apartments/condos) sold in each postcode at a monthly frequency. The average prices are constructed by taking the median over property sales for the past year. Hence, in order to be consistent with the census data—our other source of data—we use the median prices for August of 2006, 2011 and 2016. This is because it is August when the ABS runs the 5-yearly Australian census.
We use the housing price data in a variety of ways. We use both the prices for houses and units as the dependent variable in different models. However, in our main models we use the average of the prices of houses and units as the dependent variable. In constructing the average price for a postcode we weight by the relative number of existing houses and units in each postcode.
Note that postcodes—the geographic identifier used in our modeling—are closely related to suburbs or groups of suburbs. Hence they provide a meaningful geographic segmentation of the housing market in Australia. Australian postcodes are broadly similar to US zip codes—though they tend to be slightly larger. Across our data the average postcode contains around 5000 dwellings but there is significant variability with some postcodes in the major cities containing up to 30,000 homes.
Our other main source of data are the Australian censuses for the years; 2006, 2011 and 2016. From the census we obtain data on the population of each postcode, the country of birth and year of arrival of the postcode’s residents. Using this data we can calculate the immigration level of each postcode. Moreover, we can use data on the country and year of arrival to determine the share of the existing resident population which originates from each country. We can also calculate the national immigration rate. This is used to construct the Bartik instrument.
The census also provides a number of other variables which we use as controls in equation (1)—these are the variables denoted \({\bf x}_{pt-1}\). They are factors which may be associated with home price growth. The first such variable is the median age of persons in the postcode.
Age is likely to be related to income which in turn is likely to influence home prices. We also include the ratio of apartments as a share of all homes and the university-educated ratio—the proportion of persons in the postcode qualified to the level of a Bachelors degree or higher. This could be an important predictor of future population and income growth (Glaeser & Shapiro 2001, Glaeser & Saiz 2004). Distance to the CBD determines the access to the employment centre and may be predictive of price trends. Finally, we include the population density—the number of persons per square kilometre. This may be reflective of the extent to which there is capacity to expand housing supply in the postcode and hence be related to how prices respond to shocks. We restrict our estimation to postcodes located in major urban areas. That is, we focus on significant urban areas as defined by the Australian Bureau of Statistics. These are urban areas which have a population of at least 100,000 persons—there are 17 of them.2
This restriction is in order that the data is suitably reliable with enough persons and housing transactions to robustly estimate the parameters of interest. However, the regions included in our analysis make up at least three-quarters of Australia’s population and house a much higher proportion of the country’s foreign born and immigrant population. Hence, the results are likely to provide a good guide as to the effect of immigration on the Australian housing market more generally.
| Home Price Growth (over 5 years, %) | ||||||
| _ All Dwellings | 1826 | 26.00 | 15.27 | 14.50 | 24.65 | 37.37 |
| _ Houses | 1805 | 28.10 | 16.29 | 15.95 | 26.69 | 39.52 |
| _ Units | 1485 | 23.10 | 17.88 | 10.25 | 23.19 | 36.57 |
| Home Price Level (in 2016, $1000) | 795 | 731.01 | 371.79 | 477.97 | 619.34 | 871.44 |
| Immigrant Inflow (%) | 1826 | 6.32 | 6.98 | 2.16 | 4.58 | 7.86 |
| Median Age (years) | 1826 | 36.88 | 4.37 | 34.00 | 37.00 | 39.00 |
| University Educated (%) | 1826 | 17.28 | 10.06 | 8.97 | 14.90 | 24.18 |
| Apartment Ratio (%) | 1826 | 17.06 | 20.47 | 2.66 | 8.64 | 23.72 |
| Population Density (10,000 per km\(^2\)) | 1826 | 0.18 | 0.16 | 0.06 | 0.15 | 0.23 |
| Distance to CBD (`00s of km) | 1826 | 0.18 | 0.16 | 0.08 | 0.14 | 0.24 |
| Note: | ||||||
| Immigrant Inflow is the percentage of population which is newly arrived from overseas since the prior census. University Educated is ratio of those with the university education to the postcode population. Apartment Ratio is the ratio of apartments to total dwellings. |
| NSW | 276 | 32.80 | 16.86 | 19.60 | 32.46 | 47.38 | 5.98 | 6.92 | 1.59 | 4.00 | 7.38 |
| _ Sydney | 218 | 34.98 | 16.52 | 21.28 | 37.30 | 48.96 | 7.08 | 7.32 | 2.76 | 5.00 | 8.70 |
| VIC | 230 | 32.70 | 11.44 | 25.48 | 34.47 | 40.66 | 6.71 | 8.58 | 1.96 | 4.79 | 8.35 |
| _ Melbourne | 222 | 32.98 | 11.38 | 25.69 | 34.82 | 40.70 | 6.87 | 8.69 | 2.03 | 4.93 | 8.53 |
| QLD | 163 | 18.05 | 9.37 | 12.29 | 18.70 | 24.32 | 5.65 | 4.89 | 2.58 | 4.26 | 6.92 |
| _ Brisbane | 114 | 20.67 | 7.89 | 15.10 | 21.45 | 26.13 | 6.36 | 5.44 | 2.82 | 4.78 | 8.02 |
| Adelaide | 100 | 21.67 | 12.24 | 10.15 | 21.41 | 31.79 | 5.26 | 4.34 | 2.24 | 4.34 | 6.98 |
| Perth | 99 | 12.26 | 9.98 | 7.26 | 11.65 | 16.18 | 8.67 | 6.44 | 4.49 | 6.81 | 10.70 |
| Canberra | 22 | 19.37 | 11.49 | 11.18 | 17.51 | 28.90 | 8.35 | 11.41 | 1.97 | 5.17 | 10.25 |
| Note: | |||||||||||
| NSW denotes the state of New South Wales, VIC denotes the state of Victoria and QLD denotes the state of Queensland. |
Tables 1 and 2 provide a summary of the data. First to Table 1. In total, we have 1,826 postcodes in our estimation dataset. The mean percentage change in home prices (the average of houses and units) between the 5-years census time spans is 26% (around 5% per annum). The mean immigrant inflow, based on the population at the prior census, was 6.3% or about 1.3% per annum. The table provides summary statistics for the range of other variables used in the model.
Particular interest focuses on house prices and immigration. Table 2 reports summary statistics for these variables by some of the main cities and states. Among the areas included in the research, Sydney followed closely by Melbourne, has the highest price growth in Australia, with about 7.0% and 6.6% respectively on an annual basis across the postcodes. Considering the immigrant inflow proportions, the region with the highest immigration inflow is Perth. This averaged 8.67% over the two 5-year periods considered or 1.73% per
annum. This is likely due to the migration inflows toward this region during the mining boom between 2005 and 2013—Perth is a major mining state. Canberra, Darwin, Sydney, Melbourne and Brisbane also experienced quite high immigration rates. The smaller and more regional urban centres have lower rates.
We undertake some descriptive analysis to investigate how home prices and immigration have been related. In Figure 3 we construct a scatter plot of how home prices and the proportion of immigrants have been related. It provides a clear illustration of the diversity of experiences both in terms of price appreciation and the change in the population due to immigrants. There is a weak positive correlation between the two factors shown in the figure.
As noted above, we focus on the impact on prices of immigrants from certain source countries. In particular, we look at the effect of Chinese and Indian immigrants and combine other immigrants into another group. Figure 4 illustrates our reasons for focusing just on the two major immigrant groups to Australia—Chinese and Indians. They stand out as the two most significant immigrant groups by a large margin. There is a considerably smaller number of immigrants from the other countries meaning there is not sufficient data to accurately estimate further country-of-origin effects.
We estimate a number of different regression models for various housing markets. As a baseline model we simply estimate equation (1) using ordinary least squares (OLS). In all other models we use the Bartik instrumental variables approach to dealing with the problem of endogeneity outlined above. In addition we include a number of control variables in the model. These are factors which may help predict the future appreciation in home prices. All these variables are lagged—i.e. they are from the prior census. Our results across all immigrants are shown in Table 3.
| Immigrant Inflow | 0.091 | 0.478*** | 0.357*** | 0.548*** | 0.729*** | 0.290** | 0.563*** | 0.175 | 0.468*** | 0.346*** | -0.073 | -0.038 | 0.197 | -0.094 | 0.408* |
| (-0.061) | (-0.108) | (-0.113) | (-0.105) | (-0.121) | (-0.134) | (-0.135) | (-0.133) | (-0.113) | (-0.109) | (-0.160) | (-0.161) | (-0.217) | (-0.266) | (-0.210) | |
| Median Age | 0.001 | 0.003*** | 0.0004 | 0.004*** | 0.003** | 0.002 | -0.002 | -0.003 | 0.007*** | 0.007*** | -0.001 | 0.0003 | 0.0001 | -0.003 | 0.010** |
| (-0.001 ) | (-0.001 ) | (-0.001 ) | (-0.001 ) | (-0.001 ) | (-0.001 ) | (-0.002 ) | (-0.002) | (-0.002) | (-0.001) | (-0.001) | (-0.002) | (-0.002) | (-0.003) | (-0.004) | |
| University Educated | 0.122** | 0.113** | -0.070 | 0.189*** | 0.158*** | 0.109* | 0.031 | -0.135*** | 0.091 | -0.028 | 0.300*** | 0.104 | 0.124 | -0.064 | -0.590** |
| (-0.045 ) | (-0.050 ) | (-0.050 ) | (-0.048 ) | (-0.053 ) | (-0.063 ) | (-0.059 ) | (-0.051) | (-0.075) | (-0.074) | (-0.067) | (-0.076) | (-0.091) | (-0.124) | (-0.27) | |
| Apartment Ratio | -0.147*** | -0.217*** | -0.207*** | -0.265*** | -0.140*** | -0.152*** | -0.202*** | -0.148*** | -0.376*** | -0.318*** | -0.265*** | -0.128*** | -0.059 | -0.067 | 0.050 |
| (-0.026 ) | (-0.035 ) | (-0.032 ) | (-0.036 ) | (-0.034 ) | (-0.042 ) | (-0.038 ) | (-0.033) | (-0.05) | (-0.049) | (-0.036) | (-0.035) | (-0.078) | (-0.081) | (-0.090) | |
| Population Density | 0.204*** | 0.204*** | 0.259*** | 0.102*** | 0.207*** | 0.185*** | 0.147*** | 0.100*** | 0.210*** | 0.176** | 0.352*** | 0.130* | 0.131 | -0.178 | 0.744*** |
| (-0.029 ) | (-0.036 ) | (-0.035 ) | (-0.038 ) | (-0.037 ) | (-0.041 ) | (-0.031 ) | (-0.028) | (-0.080) | (-0.079) | (-0.068) | (-0.076) | (-0.095) | (-0.112) | (-0.212) | |
| Distance to CBD | -0.001 | 0.004 | 0.013 | -0.019 | 0.016 | 0.002 | -0.074* | -0.233*** | -0.152*** | -0.229*** | -0.071 | -0.223*** | -0.053 | -0.118** | -0.890*** |
| (-0.023 ) | (-0.043 ) | (-0.034 ) | (-0.025 ) | (-0.047 ) | (-0.048 ) | (-0.043 ) | (-0.046) | (-0.051) | (-0.054) | (-0.048) | (-0.070) | (-0.096) | (-0.059) | (-0.222) | |
| AR(1) | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| Year Fixed Effect | Yes | Yes | — | — | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| State Fixed Effect | Yes | Yes | Yes | Yes | Yes | Yes | — | — | — | — | — | — | — | — | — |
| No. Obs. | 1826 | 1826 | 911 | 915 | 1805 | 1485 | 552 | 436 | 459 | 443 | 325 | 227 | 199 | 195 | 44 |
| R2 | 0.82 | 0.816 | 0.919 | 0.91 | 0.841 | 0.707 | 0.935 | 0.961 | 0.927 | 0.933 | 0.862 | 0.931 | 0.952 | 0.75 | 0.938 |
| Adjusted-R2 | 0.818 | 0.814 | 0.917 | 0.909 | 0.839 | 0.703 | 0.934 | 0.961 | 0.925 | 0.931 | 0.858 | 0.928 | 0.949 | 0.738 | 0.922 |
| Std. Err. | 0.128 | 0.130 | 0.084 | 0.094 | 0.130 | 0.159 | 0.095 | 0.077 | 0.095 | 0.091 | 0.077 | 0.059 | 0.056 | 0.081 | 0.063 |
Our main focus is on the coefficient for the variable labeled ‘Immigrant Inflow’. We present a range of models and controls. Let us first turn to our leading set of results—these are for Australia as a whole. We also explore the sensitivity of our results to the region, the time examined and the property type.
For Australia we estimate a model using the data from all urban postcodes and home prices which are the stock-weighted average of house and unit prices. This is shown in column (B) of the table. In this model we included the lagged home price change to control for momentum. In addition we included time and state fixed effects, to control for any regional differences in price trends, and the other variables listed. There were 1,826 observations used in the model and the \(R^2\) was 0.816. We find an estimated coefficient of 0.478. This is statistically significantly different from zero at the 1% level. It is important to note that this coefficient is very different from that by simply estimating OLS—this is shown in the column (A) of the table. In this case the estimation by instrumental variables leads to important changes in the estimates.
The coefficient α reflects the log change in home prices over a 5-year period for a one unit change in the Immigrant Inflow variable. This latter variable is measured as the 5-yearly rate. From Table 1 we can see that the immigrant inflow averaged 6.32% (or around 1.25% per year). This means that if there had been no immigration over a 5-year period examined then this would have lowered home price growth by around 0.6 percentage points per year.3 While this figure is not particularly large neither is it trivial. Table 1 shows that the average home price growth rate was 26% over a 5-year period or around 5% per annum. If immigration were to have been eliminated then it would have led to a relatively mild change in overall home price growth. But this is not to say the effect of immigrants is negligible either. Over a number of years six-tenths of a percentage point of extra growth accumulates to a meaningful impact on home prices. Indeed, at the average home price of $730,000 per year in 2016 this effected amounts to $4,400 per year.
The remaining models explore the extent to which the effect of immigrants has varied across time, space and over property types. Using data for the whole of Australia, we estimate a similar model for data just from 2006 and 2011, column (C), and then another model using data only from 2011 and 2016, column (D). The effect of immigrants is statistically significantly positive in both cases but larger at 0.548 in the later period. This implies that if immigration had gone from the average rate to zero over this period then prices would have grown by around 0.7 percentage points less per year than they did.
Again using data for Australia we look at the price impact of immigrants on houses and units separately—columns (E) and (F) respectively. The results indicate that immigrants have a significantly stronger effect on houses than units—the coefficient is 0.729 compared with 0.290 for units and both are statistically significantly different from zero. Thus house prices would have been around 0.9 percentage points lower, and unit prices 0.4 percentage points lower per annum, had there been no immigration.
Broadly, these results for Australia as a whole indicate a fairly robust finding of a causal effect of recent immigrants on home prices. This effect is of the order of around 0.6 percentage points per annum—though this varies somewhat over time and across property types. The remaining results shown in Table 3 explore geographic variation in the immigrant effect on home prices. We estimate our models using data for all three census years and combine prices for both houses and units. Our focus is on the main states—New South Wales (NSW), Victoria (VIC) and Queensland (QLD)—and the main state capital cities— Sydney, Melbourne, Brisbane, Adelaide, Perth and Canberra. We find a large effect of 0.563 for NSW. This implies that prices would have been 0.7 percentage points per annum lower had there been no immigration. But interestingly this is driven by urban areas outside the main city of Sydney where the coefficient is lower and not statistically different from zero. In Victoria we see a similar pattern with the state recording a statistically significant coefficient of 0.468 (0.6 percentage points per annum). This is higher than the coefficient for Melbourne of 0.346 (0.4 percentage points per annum), though this is still quite large and statistically different from zero. Interestingly in the other cities, barring Canberra, we find negligible effects for immigrants on home prices. Canberra records an effect around the size of that in Australia as a whole. Though it is only marginally statistically significant given the small sample size.
In the prior regressions a number of controls were included for price growth in the succeeding 5 years. These were important and in a number of cases had significant coefficients. Most notably, the Apartment Ratio is negatively correlated with home price appreciation. This was also the case for Distance to the CBD in the major states and capital cities. Higher population density was predictive of positive home price growth. These latter two factors indicate an increasing premium for proximity in some of Australia’s larger cities over this period.
| Other Immigrants Inflow | -0.068 | 0.387 | -0.480** | 0.917*** | 0.393 | 0.268 |
| (0.111) | (0.249) | (0.221) | (0.236) | (0.259) | (0.322) | |
| Chinese Immigrants Inflow | 0.006 | 0.682*** | 1.559*** | -0.137 | 1.553*** | 0.349 |
| (0.17) | (0.264) | (0.386) | (0.222) | (0.321) | (0.297) | |
| Indian Immigrants Inflow | 0.904*** | 0.367 | 0.588** | 0.847*** | 0.273 | 0.242 |
| (0.25) | (0.319) | (0.266) | (0.317) | (0.42) | (0.388) | |
| Median Age | 0.001 | 0.003** | -0.002 | 0.005*** | 0.002 | 0.002 |
| (0.001) | (0.001) | (0.001) | (0.001) | (0.001) | (0.001) | |
| University Eduated | 0.132*** | 0.106** | -0.057 | 0.218*** | 0.126** | 0.106 |
| (0.045) | (0.052) | (0.052) | (0.050) | (0.054) | (0.066) | |
| Apartment Ratio | -0.147*** | -0.209*** | -0.135*** | -0.293*** | -0.107*** | -0.149*** |
| (0.026) | (0.041) | (0.038) | (0.042) | (0.04) | (0.052) | |
| Population Density | 0.215*** | 0.201*** | 0.253*** | 0.109*** | 0.197*** | 0.184*** |
| (0.029) | (0.036) | (0.037) | (0.041) | (0.038) | (0.041) | |
| Distance to CBD | 0.003 | 0.002 | 0.016 | -0.014 | 0.012 | 0.001 |
| (0.023) | (0.043) | (0.032) | (0.026) | (0.046) | (0.049) | |
| AR(1) | Yes | Yes | Yes | Yes | Yes | Yes |
| Year Fixed Effect | Yes | Yes | Yes | Yes | Yes | Yes |
| State Fixed Effect | Yes | Yes | Yes | Yes | Yes | Yes |
| No. Obs. | 1826 | 1826 | 911 | 915 | 1805 | 1485 |
| R2 | 0.821 | 0.816 | 0.917 | 0.907 | 0.841 | 0.706 |
| Adjusted-R2 | 0.819 | 0.814 | 0.915 | 0.906 | 0.839 | 0.703 |
| Std. Err. | 0.128 | 0.130 | 0.085 | 0.096 | 0.13 | 0.159 |
In Table 4 we report results for two significant immigrant groups in Australia—Chinese and Indians. As well, we include an Other group reflect all remaining nationalities. We estimate these models at the national level across various dimensions. A base model, shown in column (P), is also shown which simply uses OLS. The other models use instrumental variables.
The results indicate significant effects across immigrant groups. In model (Q) for all of Australia and both property types the Chinese coefficient is 0.682 and statistically significant at the 1% level. This is higher than the coefficient for both Indian immigrants and Other immigrants. The large coefficient for Chinese immigrants reflects a bigger coefficient in the earlier period, from 2006-2011, and a bigger coefficient for houses than units. These results point towards a degree of diversity in the impact that immigration can have on home prices. The nature and source-country of the migrant clearly plays a non-trivial role in the effect on home prices in the destination country.
The purpose of this research was to investigate the impact of immigrants on prices in the housing market at postcode level in Australia from 2006 to 2016. We make use of data from the Australian censuses in 2006, 2011 and 2016 to relate changes in home prices to the change in the proportion of newly arrived immigrants between these years. Because there are likely to be many unobserved factors which both drive home price growth in a certain area and also influence immigrants choice of destination we use instrumental variables. The shift-share approach of Bartik (1991) is used to construct an instrument for the immigrant inflow.
We find that had there been no immigration over the period examined then home prices would have been around 0.6 percentage points lower per year. We note that this effect is not particularly large—the average annual growth in home prices over the period has been about 5%—but neither is it trivial. At average home prices of $730,000 per year in 2016 this amounts to $4,400 per year. Our estimates of the effect of immigrants are broadly consistent with the effect found in other countries. Most of the international literature has found a positive effect—as we did—however, our estimates are on the smaller end of the spectrum in absolute value. This may reflect the fact that Australian home prices are high by international standards so a smaller percentage effect is required for the same dollar effect. Moreover, as we noted in the introduction, Australia has experienced relatively high rates of immigration by comparison with other countries. This means the overall effect of immigration may be be similar in Australia to that in other countries.
Importantly, our research illustrated a degree of variation in the effects of new immigrants over time, across property types, regions and by immigrant source country. Immigrants had larger effects on house prices than unit prices—raising them by around 0.729 percentage points per annum for a one percentage point increase in the immigrant inflow. This compares to 0.290 percentage points for units. The effects were largest in more recent times and in the states of NSW (excluding Sydney) and Victoria and in the city of Canberra. Interestingly, we found that Chinese immigrants had larger effects on home prices than Indian immigrants or immigrants from all other countries. Chinese arrivals had a particularly strong impact on house prices, as opposed to units, and over the period 2006-2011.
Abelson, P., Joyeux, R., Milunovich, G. & Chung, D. (2005), ‘House prices in Australia - 1970 to 2003 - facts and explanations’, IDEAS Working Paper Series from RePEc .
Akbari, A. H. & Aydede, Y. (2012), ‘Effects of immigration on house prices in Canada’, Applied Economics 44(13), 1645–1658.
Altonji, J. G. & Card, D. (1989), The effects of immigration on the labor market outcomes of natives, Working Paper 3123, National Bureau of Economic Research.
Aydede, Y. (2017), ‘Immigration and location choices of native-born workers in Canada’, Papers in Regional Science 96, 113–134.
Bartik, T. (1991), Who Benefits from State and Local Economic Development Policies?, W.E. Upjohn Institute.
Boubtane, E., Dumont, J.-C. & Rault, C. (2016), ‘Immigration and economic growth in the OECD countries 19862006’, Oxford Economic Papers 68(2), 340–360.
Braakmann, N. (2019), ‘Immigration and the property market: Evidence from England and Wales’, Real Estate Economics 47. Coleman, A., Landon-Lane, J. et al. (2007), Housing markets and migration in New Zealand, 1962-2006, Technical report, Reserve Bank of New Zealand.
Daley, J., Coates, B. & Wiltshire, T. (2018), ‘How migration affects housing affordability’, The Conversation . URL: http://theconversation.com/how-migration-affects-housing-affordability-92502
Degen, K. & Fischer, A. M. (2017), ‘Immigration and Swiss house prices’, Swiss Journal of Economics and Statistics 153(1), 15–36.
Glaeser, E. L. & Saiz, A. (2004), ‘The rise of the skilled city’, Brookings-Wharton Papers on Urban Affairs pp. 47–105.
Glaeser, E. L. & Shapiro, J. (2001), Is there a new urbanism? The growth of U.S. cities in the 1990s, Working Paper 8357, National Bureau of Economic Research.
Gonzalez, L. & Ortega, F. (2013), ‘Immigration and housing booms: Evidence from Spain’, Journal of Regional Science 53(1), 37–59.
Harrison, D. S. (1984), ‘The impact of immigration on a depressed labour market: The South Australian experience’, Economic Record 60(1), 57–67.
Hugo, G. (2006), ‘Temporary migration and the labour market in Australia’, Australian Geographer 37(2), 211–231.
Kohler, M. & van der Merwe, M. (2015), Long-run trends in housing price growth, Technical report, Reserve Bank of Australia.
Macaulay, J. (2019), ‘Immigration: the house price elephant in the room’, The Spectator . URL: https://www.spectator.com.au/2019/02/immigration-the-house-price-elephant-in-the-room/
Peeters, L. & Chasco, C. (2016), ‘Identifying local determinants of destination choices of international immigrants to the Madrid metropolitan area’, Papers in Regional Science 95, 281–307.
Peter, M. W. & Verikios, G. (1996), ‘The effect of immigration on residents’ incomes in Australia: Some issues reconsidered’, Australian Economic Review 29(2), 171–188.
Sa, F. (2015), ‘Immigration and house prices in the UK’, Economic Journal 125(587), 1393– 1424.
Saiz, A. (2007), ‘Immigration and housing rents in American cities’, Journal of Urban Economics 61(2), 345 – 371.
Stillman, S. & Mar, D. (2008), ‘Housing markets and migration: Evidence from New Zealand’, IDEAS Working Paper Series from RePEc .
Wokker, C. & Swieringa, J. (2016), ‘Foreign investment and residential property price growth’, Treasury Working Paper 2016-03.
| Australia | |||||
| _ (B) All | Weak instruments | 1 | 1810 | 851.79 | 0.000 |
| Wu-Hausman | 1 | 1809 | 19.08 | 0.000 | |
| _ (C) 2006-11 | Weak instruments | 1 | 896 | 547.79 | 0.000 |
| Wu-Hausman | 1 | 895 | 0.84 | 0.361 | |
| _ (D) 2011-16 | Weak instruments | 1 | 900 | 430.99 | 0.000 |
| Wu-Hausman | 1 | 899 | 87.85 | 0.000 | |
| _ (E) Houses | Weak instruments | 1 | 1789 | 922.45 | 0.000 |
| Wu-Hausman | 1 | 1788 | 14.51 | 0.000 | |
| _ (F) Units | Weak instruments | 1 | 1469 | 924.71 | 0.000 |
| Wu-Hausman | 1 | 1468 | 3.10 | 0.079 | |
| NSW | |||||
| _ (G) All | Weak instruments | 1 | 543 | 322.98 | 0.000 |
| Wu-Hausman | 1 | 542 | 39.01 | 0.000 | |
| _ (H) Sydney | Weak instruments | 1 | 427 | 228.94 | 0.000 |
| Wu-Hausman | 1 | 426 | 11.50 | 0.001 | |
| VIC | |||||
| _ (I) All | Weak instruments | 1 | 450 | 268.29 | 0.000 |
| Wu-Hausman | 1 | 449 | 45.49 | 0.000 | |
| _ (J) Melbourne | Weak instruments | 1 | 434 | 260.40 | 0.000 |
| Wu-Hausman | 1 | 433 | 32.42 | 0.000 | |
| QLD | |||||
| _ (K) All | Weak instruments | 1 | 316 | 318.39 | 0.000 |
| Wu-Hausman | 1 | 315 | 0.00 | 0.974 | |
| _ (L) Brisbane | Weak instruments | 1 | 218 | 212.86 | 0.000 |
| Wu-Hausman | 1 | 217 | 0.76 | 0.385 | |
|
Weak instruments | 1 | 190 | 296.10 | 0.000 |
| Wu-Hausman | 1 | 189 | 0.22 | 0.637 | |
|
Weak instruments | 1 | 186 | 54.41 | 0.000 |
| Wu-Hausman | 1 | 185 | 4.77 | 0.030 | |
|
Weak instruments | 1 | 35 | 30.63 | 0.000 |
| Wu-Hausman | 1 | 34 | 2.69 | 0.111 |
This is the pre-peer reviewed version of the following article: “The impact of immigration on housing prices in Australia”, which has been published in final form at https://doi.org/10.1111/pirs.12497. This article may be used for non-commercial purposes in accordance with Wiley Terms and Conditions for Use of Self-Archived Versions.↩
The urban areas are; Adelaide, Brisbane, Cairns, Canberra, Central Coast, Darwin, Geelong, Gold Coast, Hobart, Melbourne, Newcastle, Perth, Sunshine Coast, Sydney, Toowoomba, Townsville, Wollongong.↩
This is calculated as: \(\left( \exp( 0.478 \times 0.0623) - 1 \right) / 5 \times 100\).↩