RDP 2020-02: The Distributional Effects of Monetary Policy: Evidence from Local Housing Markets 3. Data

To measure local area housing prices we use data from a private company, CoreLogic. The geographic unit is Statistical Area Level 3 (SA3) (there are 358 SA3s in Australia, with populations varying between 30,000 and 130,000 people). SA3s often align to a regional city (for regional areas) or local government areas like city councils and transport or commercial hubs (for metropolitan areas).

The key variable is a hedonic index for real dwelling prices (this includes prices for detached houses and apartments). The hedonic index captures the change in housing prices controlling for observables such as land size, type of dwelling structure and the number of rooms. It can be thought of as a pure measure of housing prices. The housing price index is deflated by the (trimmed mean) consumer price index to arrive at a real measure.

In the baseline models, the stance of monetary policy is measured by the cash rate target, as published in RBA statistical table F1 (Interest Rates and Yields – Money Market). For the analysis, the sample period spans several housing price cycles covering the three decades between 1990 and 2019. The sample period roughly aligns with the introduction of inflation targeting in Australia, so our analysis should not be affected by changes in monetary policy regimes. For robustness, we also consider estimates of monetary policy shocks, which account for systematic changes in monetary policy by controlling for information that is contained in the Reserve Bank's forecasts (Bishop and Tulip 2017). The monetary policy shock series runs from 1992 to 2018.

The controls for macroeconomic variables such as GDP growth, the terms of trade and the unemployment rate are all sourced from the Australian Bureau of Statistics.^[11]

Footnote

We also have estimates of current rents by SA3 and quarter so that current rental growth can be included as an explanatory variable in the model. However, the available time series of these data only cover the period since 2005. So these data are mainly used for robustness. [11]