Elsevier

Journal of Monetary Economics

Volume 67, October 2014, Pages 62-77
Journal of Monetary Economics

Business cycle implications of mortgage spreads

https://doi.org/10.1016/j.jmoneco.2014.07.005Get rights and content

Highlights

  • We quantify the business cycle effects of the residential mortgage spread.

  • Shocks to this spread reduce house prices and real quantities.

  • This indicates the presence of financial frictions in residential mortgage markets.

  • Large-scale asset purchases of mortgages have sizable effects on the business cycle.

Abstract

How do aggregate quantities at the business cycle frequency respond to shocks to the spread between residential mortgage rates and government bonds? Using a structural VAR approach, we find that mortgage spread shocks impact the real economy by both economically and statistically significant magnitudes: a 100 basis point decline in the spread causes a peak increase in consumption, residential investment and GDP by 1.6 percent, 6.2 percent and 1.9 percent, respectively. Presumably, these effects are magnified when the policy rate is held fixed, as was the case in the US during the recent implementation of unconventional monetary policy.

Introduction

What are the quantitative business cycle effects of time variation in the residential mortgage interest rate spread? Surprisingly, this question is almost unexplored in the existing literature despite the substantial cyclical variation of this spread in the data. While Hubbard and Mayer (2009), Guerrieri and Lorenzoni (2011) and Hall, 2011a, Hall, 2011b all have referenced the issue, none have empirically documented the relationship between mortgage spreads and aggregate quantities. We define the mortgage spread as the difference between the average interest rate on newly issued mortgages at a given maturity and the government bond rate of the corresponding maturity. By using this definition, we separate the mortgage spread from the term premium. We restrict our analysis to the prime mortgage market. This is not because we think that subprime mortgages are unimportant, but rather because the two markets merit separate analysis.

Why might mortgage rates affect the macroeconomy? Theoretically, the mortgage rate, and thus the mortgage spread, potentially affects aggregate economic variables through several channels: (i) house prices and residential investment through the user cost of housing, (ii) as one relevant rate in the consumption/savings decision and, (iii) the post-interest disposable income of any household with a mortgage. If house prices are affected by mortgage spreads, then housing wealth and collateral values are also affected. In the presence of binding collateral constraints or, more generally, if credit extension is decreasing in household leverage, mortgage spreads will influence spending decisions through this collateral channel.

The motivation for exploring the business cycle effects of residential mortgage spread variation – and, more specifically, innovations to this spread – is threefold. First, this paper seeks to contribute to the general understanding of what drives business cycles and document the quantitative importance of mortgage spread innovations for aggregate variables.1 Second, if the mortgage spread affects aggregate variables, then monetary policy should take that into account. Accordingly, the analysis herein also explores how monetary policy historically has responded to mortgage spread innovations. This paper thereby complements Cúrdia and Woodford׳s (2009) analysis which addresses how monetary policy optimally is conducted in a stylized model with one lending spread that applies to all types of loans. Third, and perhaps most importantly, this paper׳s research question has bearings on unconventional monetary policy intended to affect the business cycle through the mortgage spread, such as the Federal Reserve׳s recent purchases of mortgage backed securities (MBS). To our knowledge, this paper is unique in that it empirically quantifies the business cycle effects of mortgage spread innovations without relying on a specific theoretical model.

The US is the primary country of study over the sample period 1983q1–2011q4. We start by documenting the substantial time variation in the mortgage spread and that the spread is countercyclical. Furthermore, the maximum absolute cross-correlation occurs when the mortgage spread leads GDP by 2–3 quarters. In other words, the spread is lowest immediately prior to GDP peaks and highest immediately prior to GDP troughs. A very similar pattern has been documented by Kydland et al. (2012) for nominal mortgage rates.

Our main exercise is inspired by Gilchrist et al.׳s (2009) and Gilchrist and Zakrajšek׳s (2012) work on the macroeconomic effects of corporate bond spreads. The role of innovations to mortgage spreads for business cycles is documented by estimating a structural vector autoregression (SVAR). The baseline SVAR includes the following seven variables in levels: consumption, residential investment, GDP, the consumer price index, the mortgage spread, the policy interest rate, and house prices. The identifying restriction is that mortgage spread shocks do not affect aggregate quantities or consumer prices on impact but are allowed to contemporaneously affect the policy rate and house prices.

The mortgage spread impulse responses obtained are consistent with the simple theoretical relationships mentioned above. They are also consistent with an interpretation of mortgage spread shocks as credit supply shocks: aggregate quantities and house prices all decrease following a positive innovation to the spread. A mortgage shock of 100 basis points (bps) yields a decrease of 1.6 percent in consumption, 6.2 percent in residential investment, and 1.9 percent in GDP. These responses are gradual and reach a trough after more than one year. House prices respond faster and decline by 2.6 percent. A 100 bps mortgage spread shock yields a fast and strong 184 bps offsetting response of the policy rate. From the point of view of unconventional monetary policy these results provide a lower bound. The reason is that in a setting where the policy rate is fixed and cannot accommodate the mortgage shock, such as during the recent zero-lower bound period in the US, the responses of all other variables will be stronger.

The importance of mortgage spread shocks is moderate in terms of variance decomposition at business cycle frequencies. Roughly 10 percent of consumption, GDP and house price variation is due to the spread shock at short horizons. In terms of variance decomposition the mortgage spread shock is as important for the business cycle as the excess (corporate) bond premium shock documented in Gilchrist and Zakrajšek (2012).

We find similar results for the UK and Sweden. Mortgage spread innovations also appear both statistically and economically important for these countries. Furthermore, they induce the same qualitative dynamics. However, the mortgage spread shock is more important for aggregate quantities and house prices in these countries compared to the US and its impact is faster. This difference may be due to the much shorter duration of the typical mortgage contract in the UK and Sweden compared to the US.

Our results are robust to several variations in both the SVAR specification and the sample period. Perhaps most importantly, the importance of mortgage spread innovations is not diminished when a corporate bond spread is controlled for in the VAR. We use an alternative measure of the mortgage spread that accounts for the prepayment option in US mortgages. We change the basic identification approach by using sign restrictions. For a given size of the mortgage spread shock, effects on aggregate quantities are larger for both of these alternative specifications.

To aid in the interpretation of the mortgage spread shock, we estimate a SVAR specification where the quantity of mortgage debt outstanding is added to our baseline VAR. Mortgage spread shocks drive the price and the quantity of credit in opposite directions. This corroborates other indications that mortgage spread shocks should be interpreted as credit supply disturbances.

The takeaways from this paper are the following: (i) business cycle fluctuations are affected by financial frictions in the residential mortgage market, (ii) innovations in the mortgage spread predominantly capture movements in credit supply and are moderately important for business cycle variation in aggregate quantities and house prices, (iii) in general, the policy rate appears to partially offset mortgage spread innovations, and (iv) if unconventional monetary policy in the form of asset purchases in mortgage markets succeeds in affecting the mortgage spread, then it has sizable effects on aggregate quantities and house prices.2 Similarly, macroprudential policies that affect the mortgage spread will have sizable business cycle effects.

The paper is organized as follows. The remaining part of this section describes the related literature. Section 2 characterizes the mortgage spread. Section 3 contains the quantitative exercises and results. Robustness exercises are documented in 4 Robustness, 5 Interpretation of mortgage shocks and conclusions concludes. Online appendices contain additional material.

Although their primary focus differs, three papers clearly relate to the present paper as they include empirical analysis of the effects of a household borrowing spread on the macroeconomy. Both Darracq Pariès et al. (2011) and Gerali et al. (2010) estimate a DSGE model using one measure of the retail bank loan rate and another for the retail deposit rate, in addition to standard macro variables. Musso et al. (2011) estimate structural VAR models to compare the US׳ and the Euro area׳s monetary transmission mechanisms, with a focus on housing. They include a mortgage rate in their structural VAR models. As detailed in Section 3.4, it is critical to use a mortgage spread against a government bond with a matching maturity in order to avoid confounding the short risk-free rate, the term premium and the mortgage spread.

This paper addresses time variation in aggregate mortgage conditions in terms of prices, measured as spreads. This complements the mainly theoretical existing literature that deals with time variation in the quantity dimension of mortgage credit conditions, often measured by the maximum loan-to-value ratio (LTV). Guerrieri and Lorenzoni (2011) explore changes in both these dimensions in a theoretical model with heterogeneous agents. An important finding in their model is that changes in the LTV ratio only mildly affect the aggregate variables while spread changes have a major impact.

The present paper also complements the literature addressing other non-price aspects of mortgage supply. Wilcox (2009) and Muellbauer and Williams (2011) empirically attempt to capture all non-price aspects of mortgage conditions using a latent variable approach.

Kydland et al. (2012) build a theoretical model showing that the cyclical properties of mortgage rates can explain the fact that residential investment leads business investment in the US. The same mechanism explains why housing starts lead business investment both in the US and in other OECD countries. In their model, the mortgage rate enters through the first order condition for residential investment.

The literature examining the macroeconomic effects of the Federal Reserve purchases of MBS is very thin. Chung et al. (2011) provide an estimate of the joint effect of all Large-Scale Asset Purchase (LSAP) programs by using the FRB/US model. Gambacorta et al. (2014) take a broader approach and estimate the effect of the size of central bank balance sheets on aggregate variables in a cross-country SVAR study. Gertler and Karadi (2013) provide a model of how LSAPs affect the macroeconomy. They model LSAPs as a form of financial intermediation and find that central bank purchases of private assets (corporate bonds) are more powerful than purchases of long-term government debt. For a model of LSAPs that distinguishes between corporate debt and mortgage debt, see Dai et al. (2013).

The analogous literature addressing corporate interest rate spreads and how shocks to these affect business investment and the business cycle more generally is well developed. For structural VAR approaches, see, for example, Gertler and Lown (1999), Gilchrist et al. (2009), Gilchrist and Zakrajšek (2012), Meeks (2012) and Furlanetto et al. (2013). Helbling et al. (2011) consider international transmission of this type of shocks. Two examples of estimated DSGE models that allow for financial shocks and use corporate spread data are Christiano et al., 2011, Christiano et al., 2014. Jermann and Quadrini (2012) and Nolan and Thoenissen (2009) also show the importance of shocks to firm financing by taking models to the data, but do not specifically include spread data. Fornari and Stracca (2012) apply a structural VAR approach without spreads that instead uses sign restrictions for identification of financial shocks. More generally, the business cycle literature appears to be relaxing the previously prevalent assumption that one interest rate is sufficient to characterize the economy.

Section snippets

Data

This section documents the countries and sample periods used. In addition, the measurement of the mortgage spread is described and the spread is characterized.

Our main country of study is the US. The sample period is 1983q1–2011q4. The sample begins in 1983 in order to avoid the Regulation Q and the Volcker disinflation periods. The UK and Sweden are also studied. These two countries provide an international perspective and contrast the US in that mortgage contracts have a much shorter

Mortgage spread innovations and business cycles

This section documents this paper׳s main quantitative exercises and results. The aim is to quantify the effects of innovations to the mortgage spread on the rest of the economy. We do this by estimating a structural VAR.

Robustness

The VAR specification is altered in the following six ways to document the robustness of the results: (i) inclusion of a corporate bond spread, (ii) the use of sign and zero restrictions for identification, (iii) the use of a spread measure that accounts for the option value of prepaying the mortgage, (iv) the change of the ordering of variables so that mortgage shocks are allowed to affect fewer variables contemporaneously, (v) the sample period is shortened to 1983q1–2008q2, and (vi) the lag

Interpretation of mortgage shocks and conclusions

We have used a SVAR with aggregate quantities, consumer prices, the mortgage spread, the federal funds rate and house prices to extract exogenous innovations to the mortgage spread. Strong indicative evidence is found that these innovations should be interpreted as credit supply shocks. Nevertheless, it is not obvious what the concrete underlying factors are that generate these innovations. Recall that the baseline identification is set up such that only spread movements that are orthogonal to

Acknowledgments

I am grateful to Jan Alsterlind for starting this project off. Thanks to Kristopher Gerardi (the editor) and the referee for valuable suggestions. Thanks also for feedback at many presentations and in private conversations. Warm thanks to Mattias Villani for sharing his VAR code and Jonas Arias, Juan Rubio-Ramírez and Dan Waggoner for sharing their sign and zero restriction code. The views expressed in this paper are solely the responsibility of the author and should not be interpreted as

References (42)

  • L. Christiano et al.

    Risk shocks

    Am. Econ. Rev.

    (2014)
  • Chung, H., Laforte, J., Reifschneider, D., Williams, J., 2011. Estimating the Macroeconomic Effects of the Fed׳s Asset...
  • Cúrdia, V., Woodford, M., 2009. Credit Frictions and Optimal Monetary Policy....
  • M. Darracq Pariès et al.

    Macroeconomic propagation under different regulatory regimesevidence from an estimated DSGE model for the euro area

    Int. J. Cent. Bank.

    (2011)
  • Dai, M., Dufourt, F., Zhang, Q., 2013. Large Scale Asset Purchases with Segmented Mortgage and Corporate Loan Markets....
  • F. Fornari et al.

    What does a financial shock do? First international evidence

    Econ. Policy

    (2012)
  • R. Fry et al.

    Sign restrictions in structural vector autoregressionsa critical review

    J. Econ. Lit.

    (2011)
  • Furlanetto, F., Ravazzolo, F., Sarferaz, S., 2013. Identification of Financial Factors in Economic Fluctuations....
  • Fuster, A., Willen, P., 2010. $1.25 Trillion is Still Real Money: Some Facts About the Effects of the Federal Reserve׳s...
  • A. Fuster et al.

    The rising gap between primary and secondary mortgage rate

    FRB NY Econ. Policy Rev.

    (2013)
  • X. Gabaix et al.

    Limits of arbitragetheory and evidence from the mortgage-backed securities market

    J. Financ.

    (2007)
  • Cited by (59)

    • Mortgage loan and housing market

      2023, International Review of Economics and Finance
      Citation Excerpt :

      It is the mortgage length that positively affects the homeownership rate in the long run for a single mortgage contract. To help the readers understand the importance of the mortgage market for economic activity and macroeconomic volatility, one strand of literature that comes to mind is the one that documents the importance of US mortgage interest rates/spreads for business cycles in the US, Europe, and emerging markets; see, e.g., Cheng and Chiu (2020), Horvath and Rothman (2021), Musso et al. (2011), and Walentin (2014). More related studies of this strand are still emerging.

    • Financial shocks and inflation dynamics

      2021, Macroeconomic Dynamics
    • International Housing Markets and the U.S. Subprime Crisis

      2024, Journal of Money, Credit and Banking
    View all citing articles on Scopus
    View full text