Determinants of Residential Investment Growth Rate in the US Economy (1992–2019)

ABSTRACT

The leading role of residential investment in the business cycles is a robust stylized fact for the US economy. The housing bubble of the 2000s has increased interest in the macroeconomic relevance of this expenditure. However, there is still controversy surrounding its determinants. This article is an attempt to fulfill this gap. We propose to combine mortgage interest rate and house price inflation — two of the most relevant variables according to the literature — in a single index: houses’ own-rate of interest. This index represents the real cost of buying houses, so we expect a negative relationship with the growth rate of residential investment. Regarding the US economy from 1992 to 2019, we find a unidirectional negative correlation between houses’ own-rate of interest and residential investment growth rate in the long run. In the short-run adjustment process, we report no statistically significant effect of residential investment growth rate on houses’ own- rate of interest. Our results are robust to lag order specification and show that houses own-rate of interest explains more than half of the variability of residential investment rate of growth.

KEYWORDS:

• Residential investment
• asset price inflation
• mortgage interest rate
• vector-error correction model

JEL CODES:

• E39
• E44
• R39

Acknowledgments

The authors would like to thank but not implicate Franklin Serrano, Fabrício Pitombo Leite, Matias Vernengo, Carolina Baltar, and one anonymous referee for useful comments and suggestions, as well as comments by the participants of UFRJ Political Economy research seminars and the participants of the 4^th Demand-led growth Workshop for their valuable comments on a previous draft. Gabriel Petrini gratefully acknowledges the financial support in the form of research grant from the Brazilian National Council for scientific and Technological Development (CNPq) under Grant 130777/2018-8 and 140721/2020-7. The usual disclaimer applies.

Disclosure Statement

The authors reported no potential conflict of interest.

Notes

1 For a detailed chronology, see McCarthy and Peach (2002, Appendix B) and Green and Wachter (2005); for a discussion on the consequences for households’ access to consumption and mortgage credit, see Moysich (1997) and Wall (2010). Just to name a few examples, the Reform, Recovery, and Enforcement Act (FIRREA) in 1989 and Federal Deposit Insurance Corporation Improvement Act (FDICIA) in 1991 were important in increasing the credit volume to households in the following periods (Wall 2010).

2 We should note two important exceptions. In a letter to President Franklin D. Roosevelt, in the mid of the Great Depression, Keynes (1978, p. 436) presented the case of housing as the best aid to fight the crisis and to foster economic recovery. The other exception is Duesenberry (1958) who criticizes ordinary theories of aggregate investment because they put firms’ investment in equipment and structures together with residential investment. Thus, he proposes a model dedicated for the latter type of investment, emphasizing the role of house prices.

3 For the other countries in the sample, Kydland, Rupert, and Šustek (2016) find that residential investment is coincident (and not lagging) with GDP. Additionally, they conclude that long-term nominal mortgage rate is a relevant transmission channel from interest rates to housing costs.

4 For the other countries, they find inconclusive results regarding fluctuations due to institutional heterogeneity. However, Huang (2020) claim that for most G7 countries, residential investment at least amplifies the business cycle.

5 It worth noting that Figure 3 is just a visualization procedure. We are aware of the econometric problems regarding this type of filter. For details, see Hamilton (2017).

6 This adapted Tobin’s Q is defined as the ratio between the market price of new houses and its construction costs.

7 It could be argued that houses’ own-rate is relevant for the decision to purchase houses in general, whether they are newly built or not. Nevertheless, we restrain our analysis to macroeconomic data and relate it to the previously mentioned literature. In this case, we are only looking at residential investment (i.e., newly built houses) and not the whole housing market.

8 All data has been retrieved from FRED database. The codes are MORTGAGE30US, A011RL1Q225SBEA, and CSUSHPISA respectively. We have resampled all the non-quarterly data series by the end of the quarter and defined growth rate as percent change from the previous quarter.

9 As already discussed by the econometric literature (see Lütkepohl (2005)), we consider the Granger causality test to only indicate temporal precedence and not causal inference.

10 Although residential investment may not influence house construction costs, it may affect the price of land in particular regions since there are lands of different qualities that are scarce (Castilho 2020). While it is possible to build new houses to accommodate demand, creating land in prestigious places is impossible. In this sense, it is conceivable to have a feedback effect of residential investment on prices in locally specific housing markets. Whether this influences a national home price index, such as the Case-Shiller index, is a matter of empirical investigation we discuss further in this section. We expect this effect to be not very important, as stated in our hypotheses 2 and 7.

11 For the purpose of determining the order of the model, we use four information criteria. Two (BIC, HQIC) indicates no lag, the other two (AIC, FPE) indicate four lags (see Table 7 in appendix B available online).

12 We use data from 1999 to 2019 due to availability.