...Then to Both Housing and Consumption

John was amongst the very first economists to analyse how housing booms and busts could influence consumption and to consider wider interactions between housing and the economy, for example, labour markets and migra­tion.

A comprehensive survey paper on consumption by Muellbauer and Lattimore (1995) proposed consumption functions that allowed for sluggish adjustment and controlled for real interest rates, uncertainty, expected future income growth, wealth and housing wealth interacted with credit conditions. By including liquidity, credit constraints, uncertainty and expectations effects, they incorporated insights not only from Hall (1978), but also from Deaton (1991) and Carroll (1992) on buffer-stock saving, and the literature on credit constraints associated with Akerlof, Jaffee, Stiglitz and Weiss. This survey paper influenced the consumption function adopted in the large Federal Reserve FRB-US model introduced in 1996,¹4 although the model unfortu­nately did not build in the recommended credit channel influences or wealth disaggregation.

The common practice in the empirical literature on consumption is still to aggregate all forms of household financial and housing assets minus all debt into one wealth variable, namely net worth. Muellbauer and Lattimore (1995), as did Muellbauer and Murphy (1989), noted that using total net

¹⁴ For example, FRB-US used the recommended wealth to income form of wealth effects, which gives a far better approximation than the commonly used log wealth formulations and a high discount rate to formulate permanent income given the perceived riskiness of income. worth ignores substantial costs to households if they fail to meet debt obliga­tions, a risk leading many households in practice to put greater (negative) weight on a unit of debt than on a unit of gross assets. Disaggregating net worth at least into liquid assets minus debt, illiquid assets (such as pension wealth) and housing wealth reflects differences in the marginal propensities to consume of these different components. Another argument for separating out housing wealth is that housing is both a consumption good and an asset, and intertemporal choice theory then implies that it should be treated differently from financial wealth.

The related question of what determines house prices was studied by Muellbauer and Murphy (1997). They inverted the demand for housing ser­vices, as in Poterba's (1984) classic house price model, and found that easier mortgage credit conditions—in the presence of controls for real interest rates and expected income—bolstered house prices. The first time the inverse demand approach was applied to spatial house price determination was in Cameron et al. (2006a), which explained variations in regional house prices with a system of inverted demand equations. Prices in each region depend not just on same-region incomes and housing stocks, but also on those of other regions. Consequently, there are cross-regional spillover or “ripple” effects of house price changes. An implication of this spatial analysis is that where addi­tional housing is constructed, it may have a substantial effect on national house prices. Muellbauer (2019c) applied these ideas to a model of Parisian house prices: as in London, prices in Paris proved to be more sensitive to interest rates and credit conditions than elsewhere in the country.

An innovative feature of John's research has been his emphasis on the important double role for credit conditions in influencing both house prices and consumption, partly via the collateral role of housing wealth (e.g. bor­rowing against housing equity can contribute to swings in consumption across housing price booms and busts). In much of his research on housing and consumption, John has tried account for changing credit conditions by improving measures of difficult-to-observe credit availability. Beginning with a 2000 working paper, published in revised and updated form as Aron and Muellbauer (2013a), a technique of extracting a measure of credit conditions as a latent variable common to several equations was developed.

Fernandez-Corugedo and Muellbauer (2006) used the latent variable method in a 10-equation system for aggregate debt and proportions of mort­gage loans in high loan-to-value and high loan-to-income tranches to extract a mortgage credit conditions index for the UK. Duca and Muellbauer (2014) christened the technique, “latent interactive variable equation system” (LIVES), since the interaction of credit conditions with housing wealth or the house price-to-income ratio is a crucial feature of the credit channel. Applying the method to the US, they showed that the interaction of the fall in house prices and the contraction of credit after 2007 explained the sharp fall in con­sumption relative to income. More generally, their paper emphasised the importance of household balance sheets. Consistent with Brainard and Tobin's (1968) vision of integrating finance and monetary policy into macro-models, and to incorporate a more modern Bernanke-Gertler-Gilchrist type of house­hold financial accelerator, they explained methods for endogenising house­hold balance sheets in plausible and tractable ways.

Cognizant of the heterogeneity in financial regulations and development across economies, Muellbauer and his collaborators have applied the LIVES method to several other countries, including Australia (Muellbauer and Williams 2012), Canada (Muellbauer et al. 2015), Germany (Geiger et al. 2016) and France (Chauvin and Muellbauer 2018). For most countries, Muellbauer and his co-authors used a spline approach^[211] to estimate the time variation in the effect of non-price terms of both mortgage and consumer (non-real estate-secured) credit on consumption. The inclusion of credit con­ditions through the LIVES methodology in aggregate, time-series models of consumption addresses an omitted variable bias found in conventional mod­els of consumption. For most countries, this notably improves model fit and raises the speed of adjustment of actual to equilibrium consumption levels from implausibly slow to more plausible.

In his 2007 Jackson Hole Symposium article, John warned of the downside risk that the housing-fuelled US consumption boom of the mid-2000s would unwind in a bust. In a study of the US, the UK and Japan, Aron et al. (2012), Muellbauer and co-authors show how credit-augmented, generalised con­sumption models outperform the “barebones” LCH/PIH models in those countries experiencing substantial financial liberalisation or substantial inno­vations in lending, by controlling for critical time-variation in credit con­straints, uncertainty and the composition of disaggregated household wealth and debt. The mortgage-equity channel is not uniform, ranging from large impacts in the financially liberalised UK and US, to small, negative impacts in Japan, with onerous down-payment requirements for a mortgage and with­out the possibility of borrowing against housing equity. The impact of a hous­ing bust for the macroeconomy and for financial stability thus critically depends on the architecture of credit markets and the type of shocks experi­enced. Moreover, the effect of housing wealth on consumption is largely a collateral effect, consistent with much micro-evidence.

Differences in the structure of household balance sheets and in the inter­temporal elasticity of substitution can affect the magnitude and even the sign of the direct impact of real interest rates on consumption. In Japan, house­hold bank and saving deposits were huge, while debt levels were relatively low and relatively few households owned equities. A fall in real interest rates then reduced the consumption of pensioners and those saving for retirement by more than it increased the consumption of debtors. Thus, monetary policy in Japan is likely to be much less effective than in the UK or the US. This paper was awarded the Kendrick Prize for the best macroeconomic paper in the Review of Income and Wealth in 2012 and 2013.

In countries with more complete data, Muellbauer and his co-authors were able to develop measures specific to mortgage availability. For the UK and the US, loan-to-value ratios were constructed for first-time homebuyers—the marginal homebuyers in most countries—whose housing wealth, unlike that of previous owners, is not directly affected by earlier capital gains or losses on housing. Incorporating such measures greatly improves house price models for countries where mortgage constraints have notably varied, as in the UK and the US (see Duca et al. 2011, 2016).

3.8