RDP 2015-08: Housing Wealth Effects: Cross-sectional Evidence from New Vehicle Registrations 7. Aggregate Implications and Relation to the Literature
August 2015 – ISSN 1448-5109 (Online)
- Download the Paper 754KB
Our findings are most comparable to Mian et al (2013) and Mian and Sufi (2014), who also use geographic variation to identify housing wealth effects. During the 2006 to 2009 collapse in US house prices, Mian et al (2013) estimate an MPC per dollar change in housing wealth for new motor vehicle consumption alone of 2.3 cents. Mian and Sufi (2014) estimate a smaller but still large MPC out of housing wealth for motor vehicles of 1.6 cents during the 2002 to 2006 boom in US house prices. In contrast, our directly estimated average MPC for new passenger vehicle consumption is 0.06 cents per dollar change in gross housing wealth, and the MPC implied by our elasticity estimate is 0.17 cents. Features specific to the 2002 to 2009 US house price cycle may account for the much larger US estimates. The collapse in house prices over the 2006 to 2009 period was unusually large, and the United States entered a deep and prolonged recession in 2008, both of which may have amplified the usual effect of housing wealth on consumption. Mian and Sufi (2014) show that, during the boom phase, households with low credit scores aggressively liquefied housing wealth. They also find that spending out of housing wealth was concentrated in low-income postcodes.
We can impute an aggregate MPC out of housing wealth by making an assumption on the relative size of the MPCs for new passenger vehicles and total consumption. If the MPCs out of housing wealth are the same for new vehicle and other consumption, we can simply scale our estimated MPC by the ratio of aggregate consumption to new passenger vehicle consumption. Doing so implies an MPC of 2 cents per dollar change in gross housing wealth. (New passenger vehicle consumption is 2.9 per cent of household final consumption expenditure.[17]) We view this as an upper bound on the MPC for total consumption. For the United States, Mian et al (2013) estimate that spending on motor vehicles accounts for 43 per cent of the overall MPC out of housing wealth, despite new vehicles being a small share of total consumption. The relatively large MPC out of housing wealth for motor vehicles is consistent with the importance of access to credit for the purchase of durable goods.[18] The MPC estimates by consumption category in Mian et al (2013) imply an aggregate MPC in Australia of only 0.14 cents per dollar increase in gross housing equity (multiply the MPC of 0.06 cents for new passenger vehicle consumption by 1/0.43). Given that Mian et al (2013) provide the only available information on the sensitivity of different components of consumption to a change in housing wealth, this is our best estimate of the MPC for total consumption. Accordingly, we view an MPC for total consumption of 2 cents as an upper bound but, based on the evidence from Mian et al (2013) that new vehicle consumption is particularly sensitive to housing wealth, our results suggest that the MPC for total consumption is likely be less than 0.25 cents per dollar change in gross housing wealth.
The MPC for total consumption implied by our work is smaller than most prior Australian research. Using the HILDA Survey, Windsor et al (2013) estimate an MPC of 3–4 cents per dollar change in house prices for younger home owners in Australia, which they interpret as evidence of a collateral constraints channel. In contrast to Mian et al (2013), they estimate similar MPCs out of housing wealth for durable and non-durable consumption. Their MPC estimates for non-durable consumption are surprisingly large, and may reflect the use of self-reported consumption data. Based on a panel of Australian states, Dvornak and Kohler (2007) estimate a similar MPC to Windsor et al (2013), reporting that a permanent one dollar increase in housing wealth raises annual consumption by around 2.5 cents. In contrast to both Windsor et al (2013) and Dvornak and Kohler (2007), we have followed Mian et al (2013) in estimating the relationship between housing wealth and consumption in differences rather than levels form. We suggest that estimates in levels form are likely to overestimate the magnitude of the relationship between house prices and consumption because they do not difference out or adequately control for economy-wide trends in current and expected incomes that independently affect both house prices and consumption. Thus our different methodology may explain why our estimates are small in the context of the existing Australian literature.
Fisher, Otto and Voss (2010) find more mixed evidence. They estimate that a transitory component accounts for a large share of the variation in housing wealth, using the cointegration technique developed by Lettau and Ludvigson (2003), and in the pre-2004 period find little evidence of a relationship between transitory variation in housing wealth and consumption. They find some evidence of a relationship between housing wealth and consumption in the post-2004 period, but caution that evidence of cointegration is weak.