RDP 2020-02: The Distributional Effects of Monetary Policy: Evidence from Local Housing Markets 1. Introduction

‘… it is pretty clear that there is no such thing as the Australian housing market. What we have is a series of separate, but interconnected, markets.’ (Lowe 2019)

It is well known that the price of housing is sensitive to changes in interest rates (e.g. Himmelberg, Mayer and Sinai 2005; Jorda, Schularick and Taylor 2015; Williams 2016; Gibbs, Hambur and Nodari 2018; Saunders and Tulip 2019). It is also well known that the price of housing varies by location (e.g. Glaeser, Gyourko and Saks 2006; Saiz 2010). And yet we know little about how these findings fit together. Do interest rates contribute to housing price variation across regions? Are housing prices in more expensive areas more sensitive to interest rates than in cheaper areas? If so, by how much? And what does this tell us about the transmission of monetary policy?

The variation in housing price cycles is clear in Australia, both between and within states (Figure 1). For example, the two most populous states of New South Wales and Victoria experienced strong run-ups in housing prices between 2013 and 2017, followed by relatively steep falls. In comparison, the Northern Territory experienced a steep decline in housing prices over the same period. Looking within states, the more expensive areas of New South Wales and Victoria saw much larger increases in housing prices than cheaper areas during the recent upswing.

Figure 1: Distribution of Housing Prices by State and Territory
2006 = 100, selected price quintiles
Figure 1: Distribution of Housing Prices by State and Territory

Note: Price quintiles are calculated within state or territory

Sources: Authors' calculations; CoreLogic data

In this paper we explore the distributional effects of monetary policy on housing prices across local housing markets in Australia. In studying the effects of monetary policy, the varying housing market experiences both between and within states and territories is important for a couple of reasons.

First, the distribution of housing prices helps to identify the causal effects of monetary policy. The relationship between housing prices and monetary policy is typically identified using time series models that are subject to endogeneity problems. For instance, the central bank may raise interest rates if it expects the economy to be stronger in the future. At the same time, asset prices, including housing prices, may rise if households also expect higher future economic growth. This will confound any results and likely attenuate estimates; clouding the true effect of monetary policy on the housing market. In this example, it may appear that increases in monetary policy lead to increases in housing prices, when in reality it is stronger expected economic growth that drives both. By exploiting the fact that the central bank is unlikely to systematically respond to conditions in any specific local housing market we can better identify the effects of monetary policy.

Second, the distribution can tell us why monetary policy works. Any regional variation in the sensitivity to monetary policy may tell us why monetary policy matters to the housing market. For example, expensive areas may be more supply constrained than cheaper areas given they are more likely to be in coastal areas. If a demand shock due to a change in interest rates has a greater effect on housing prices in these areas, then this may tell us that supply constraints are binding and important to the transmission of monetary policy to housing markets. In other words, the economy may be more sensitive to monetary policy if more areas are supply constrained (holding all else constant). This information could be important to take into account when setting monetary policy.

To study the distributional effects of monetary policy, we estimate a model in which each local housing market is allowed to respond differently to monetary policy. The resulting distribution of price sensitivities across local markets provides information about the factors that link monetary policy and the housing market. When interest rates fall, areas with less elastic housing supply should experience larger increases in housing prices than areas with more elastic supply (Aastveit and Anundsen 2016). Consistent with this, we find that supply-related factors, such as land availability, explain some of the cross-sectional variation in local housing price sensitivities. But other factors are relevant too. For example, housing prices are more sensitive to monetary policy in regions with higher shares of housing investors and higher levels of income, on average. These findings are consistent with research that emphasise the role of expectations in driving housing price dynamics (e.g. Kaplan, Mitman and Violante 2019).

We also find that the local housing markets with relatively high levels of mortgage debt are those in which local housing prices are more sensitive to changes in monetary policy. This result is consistent with research that suggests that monetary policy has state-dependent effects, wherein the state of the mortgage market affects the potency of monetary policy (Beraja et al 2019).

We then directly study the effects of monetary policy on housing wealth inequality. For this, a specific version of the model is estimated in which local housing markets are sorted by how expensive they are.[1] We find that more expensive regions are more sensitive to changes in monetary policy. The variation in price responses across regions also suggests that reductions in the cash rate increase housing wealth inequality. However, this effect appears temporary with the price differences between regions disappearing about two years after the interest rate change. We investigate if this differential is linked to land availability and supply constraints, and find some evidence to support this with the effect being more concentrated in areas that have higher land availability constraints. However, we are not able to rule out other channels by which monetary policy could generate changes in the housing wealth distribution.

This paper is closely related to several papers that study the effect of monetary policy on regional housing markets. Most of this research focuses on the United States.[2] Consistent with our findings for Australia, there is substantial variation across US local housing markets in the sensitivity of housing prices to monetary policy (Cooper, Luengo-Prado and Olivei 2016; Fischer et al 2018).[3] This heterogeneity is typically linked to variation in housing supply elasticities (Aastveit and Anundsen 2016) and, more specifically, to variation in local government regulations (Fischer et al 2018). There is also some evidence of asymmetric effects of monetary policy across US local housing markets (Aastveit and Anundsen 2016). However, some studies suggest that monetary policy has a limited role to play in explaining the variation across local housing markets (Del Negro and Otrok 2007).[4] Moreover, while many studies suggest that regional heterogeneity is due to supply-side factors, other potential explanations could be important too (e.g. Fischer et al 2018; Beraja et al 2019).[5]

This paper is also related to a broader literature that looks at the regional effects of monetary policy. The heterogeneity in the sensitivity of local economies to interest rates has been documented for the United States (Carlino and DeFina 1998), Europe (Rodriguez-Fuentes and Dow 2003) and Australia (Vespignani 2013). This research typically explores the mechanisms that explain why there are regional differences in sensitivity to monetary policy. The heterogeneity in local responses to monetary policy is typically linked to variation in industrial structure (e.g. share of output due to the manufacturing sector), household demographics (e.g. age structure of the population) or firm dynamics (e.g. firm size). Our results do not suggest these factors are the main drivers behind the regional differences in housing market responses to monetary policy.

Finally, our research is related to the literature that studies the distributional effects of monetary policy (e.g. Doepke, Schneider and Selezneva 2015; Coibion et al 2017). To the best of our knowledge, there has been little research that specifically links monetary policy to housing wealth inequality, which is surprising given the strong relationship between interest rates and housing prices.[6]

Footnotes

Otto (2007) examines the effect of changes in mortgage interest rates on housing prices across Australian capital cities. However, he does not examine the heterogeneity in responses within cities nor link it to monetary policy. [1]

The only Australian study of which we are aware is Lim and Tsiaplias (2018). They find that the stance of monetary policy has varying (nonlinear) effects across city-based housing markets in Australia. We extend their analysis by allowing the sensitivity of housing prices to interest rates to vary for each individual local housing market. We also more closely examine which economic variables matter to this sensitivity. [2]

Cooper et al (2016) is the study most similar to this paper, although they ask a slightly different research question. We are interested in the effect of an aggregate shock (national monetary policy changes) on regional housing prices. They study the effect of regional monetary shocks on regional housing prices, with their identification strategy implicitly abstracting from the aggregate effects of national monetary policy changes. The regional monetary policy shocks are estimated based on a counterfactual monetary policy rule for each US state. [3]

Del Negro and Otrok estimate a dynamic factor model on US state data and find that most of the variation in local housing prices is explained by regional shocks, with a limited role for monetary policy. However, their sample period excludes the national housing price cycle over the mid to late 2000s, which some have linked to the stance of monetary policy. For a summary of the debate on this, see Dokko et al (2011). [4]

Beraja et al (2019) indicates that the regional distribution of housing equity influences the aggregate effect of monetary policy, and that lower levels of equity are associated with regions being less responsive to interest rate cuts due to a reduced ability to refinance mortgages. [5]

For an overview of research on the links between monetary policy and inequality, see Colciago, Samarina and de Haan (2019). [6]