Inattention and Inertia in Household Finance: Evidence from the Danish Mortgage Market

Size: px

Start display at page:

Download "Inattention and Inertia in Household Finance: Evidence from the Danish Mortgage Market"

Cory Horton
8 years ago
Views:

1 nattention and nertia in Household Finance: Evidence from the Danish Mortgage Market S. Andersen, J.Y. Campbell, K.M. Nielsen, and T. Ramadorai Copenhagen, Harvard, HKUST, Oxford March 20, 2014 Andersen et al (2014) nattention and nertia GWU/FRB / 20

Ramadorai Copenhagen, Harvard, HKUST, Oxford March 20, 2014

2 ntroduction nertia in Household Finance Households often respond slowly to changed circumstances, e.g. Participation, saving, and asset allocation in retirement savings plans (Agnew, Balduzzi, and Sunden 2003, Choi, Laibson, Madrian, and Metrick 2002, 2004, Madrian and Shea 2001) Portfolio rebalancing in risky asset markets (Bilias, Georgarakos, and 2010, Brunnermeier and Nagel 2008, Calvet, Campbell, and Sodini 2009). Mortgage re nancing is an important example of this Woodheads important in industry prepayment models for US FRMs Random time-variation in inertia generates prepayment risk, important for pricing MBS (Gabaix, Krishnamurthy, and Vigneron 2007) nertia leads to cross-subsidies from sluggish re nancers to prompt re nancers, and may inhibit constructive nancial innovation (Campbell 2006, Gabaix and Laibson 2006) Similar phenomena in the UK system with ARM teaser rates (Miles 2004). Andersen et al (2014) nattention and nertia GWU/FRB / 20

Participation, saving, and asset allocation in retirement savings plans (Agnew, Balduzzi, and Sunden 2003, Choi, Laibson, Madrian, and Metrick 2002, 2004, Madrian and Shea 2001) Portfolio rebalancing

3 ntroduction Understanding Mortgage Re nancing nertia Do prompt re nancers look di erent from sluggish re nancers? n the US, HMDA forms only tell us borrower characteristics at the time of mortgage origination, so we don t observe non-re nancers Survey data, e.g. American Housing Survey, very noisy (Schwartz 2006) s there evidence that some people also re nance too soon? Agarwal, Driscoll, and Laibson (2013) present an approximate closed-form solution to the real options problem of optimally re nancing with xed costs Agarwal, Rosen, and Yao (2012) nd evidence of both errors of commission (re nancing too soon) and errors of omission (re nancing too late), but they can only observe re nancers so the worst errors of omission are excluded To what extent is sluggish re nancing explained by constraints? n the US, re nancing requires positive home equity n the US, re nancing requires a su ciently high credit score Attempts to control for these constraints are inevitably imperfect (Archer, Ling, and McGill 1996, Campbell 2006, Schwartz 2006). Andersen et al (2014) nattention and nertia GWU/FRB / 20

Agarwal, Driscoll, and Laibson (2013) present an approximate closed-form solution to the real options problem of optimally re nancing with xed costs Agarwal, Rosen, and Yao (2012) nd evidence of both

4 Danish Data Mortgage Data from Denmark n this paper we use Danish data to surmount many of the problems of previous research Denmark has predominantly FRMs, like the US, but with important special features Funding with covered bonds, xed-rate maturity-matched bonds with integer coupons Re nancing does not require positive home equity or a credit check provided the mortgage principal balance does not increase Re nancing involves buying back the underlying mortgage bond, either at market value or face value When buying back at face value, the re nancing incentive is the bond s coupon rate less the current mortgage yield. Andersen et al (2014) nattention and nertia GWU/FRB / 20

credit check provided the mortgage principal balance does not increase Re nancing involves buying back the underlying mortgage bond, either at market value or face value

5 Danish Data Administrative Data from Denmark Denmark also has high-quality administrative data, which we match over the period All mortgages from 5 largest mortgage banks (out of 7) with a 94% market share Demographic information from Civil Registration System ncome and wealth from the tax authority Education from the Ministry of Education Medical treatments from the National Board of Health We start with 2.7 million households We match education and income for 2.5 million We look only at households with a stable number of adults and a single address, 2.0 million About 800,000 households have mortgages, 600,000 have a single mortgage, and 500,000 have a xed-rate mortgage in two successive years ( or ). Andersen et al (2014) nattention and nertia GWU/FRB / 20

Health We start with 2.7 million households We match education and income for 2.5 million We look only at households with a stable number of adults and a single address, 2.

6 A First Look at the Data Summary Statistics (Table 1) Andersen et al (2014) nattention and nertia GWU/FRB / 20

7 A First Look at the Data Do Re nancers and Non-Re nancers Look Di erent? (Table 3) Andersen et al (2014) nattention and nertia GWU/FRB / 20

8 A Model of nattention and nertia A Model of nattention and nertia Pr(Refinancing) = Pr(γ 0 b it + A(δ 0 s it ) (z it ) + ɛ it ) > 0, ɛ it N(0, Σ). Vectors b it and s it contain characteristics of household i at time t. b it determines baseline probability of re nancing s it shifts attention to incentives Vector z it contains characteristics of household i s mortgage at time t (.) is the re nancing incentive function A(.) is the attention function for household i at time t We use the exponential form A(δ 0 s it ) = exp(δ 0 s it ). As A(s it )! 0 the household ignores the incentive, and re nancing probability goes to the baseline. Andersen et al (2014) nattention and nertia GWU/FRB / 20

b it determines baseline probability of re nancing s it shifts attention to incentives Vector z it contains characteristics of household i s mortgage at time t (.

9 A Model of nattention and nertia The Re nancing ncentive (z it ) = C old it Y new it O(z it ). n the Danish system the interest saving is given by old bond coupon less new mortgage bond yield. O(z it ) is a threshold determined by re nancing costs. We use the Agarwal, Driscoll, and Laibson (2013) approximate closed-form solution: s q σκ it O(z it ) 2(ρ + λ it ). m it (1 τ) Andersen et al (2014) nattention and nertia GWU/FRB / 20

O(z it ) is a threshold determined by re nancing costs.

10 A Model of nattention and nertia The Agarwal-Driscoll-Laibson Threshold O(z it ) s q σκ it 2(ρ + λ it ). m it (1 τ) σ = is interest rate volatility, τ = 0.33 is tax rate (for mortgage interest deduction), ρ = 0.06 is discount rate m i,t is the size of the mortgage in DKK λ it is the base rate of principal reduction: λ it = µ it + Y old it exp(y old it T it ) 1 + π t µ it = µ = 0.1 (in this version of the paper) κ it = f + θm it is the re nancing cost in DKK f = 10, 000, θ = Andersen et al (2014) nattention and nertia GWU/FRB / 20

06 is discount rate m i,t is the size of the mortgage in DKK λ it is the base rate of principal reduction: λ it = µ it + Y old it exp(y

11 Re nancing ncentives and Mistakes Recent History of Danish Mortgage Rate (Figure 1) Andersen et al (2014) nattention and nertia GWU/FRB / 20

12 Re nancing ncentives and Mistakes Variation in Re nancing ncentives (Table 4) Andersen et al (2014) nattention and nertia GWU/FRB / 20

13 Re nancing ncentives and Mistakes Errors of Commission and Omission (Table 5) De ne commission as re nancing when (z it ) = Cit old Yit new O(z it ) < k. De ne omission as re nancing when (z it ) = C old it Y new it O(z it ) > k. Andersen et al (2014) nattention and nertia GWU/FRB / 20

14 Re nancing ncentives and Mistakes Who Makes These Errors? (Table 6) Andersen et al (2014) nattention and nertia GWU/FRB / 20

15 Estimating nattention and nertia Estimating the Full Linear Model (Table 9) Andersen et al (2014) nattention and nertia GWU/FRB / 20

16 Estimating nattention and nertia Nonlinear E ects (Table 9) Andersen et al (2014) nattention and nertia GWU/FRB / 20

17 Estimating nattention and nertia Nonlinear E ects (Figure 2) Andersen et al (2014) nattention and nertia GWU/FRB / 20

18 Estimating nattention and nertia Proportional nertia and nattention? (Table 10, Figure 3) Andersen et al (2014) nattention and nertia GWU/FRB / 20

19 Conclusion Conclusion Household characteristics generally move inertia (low baseline re nancing probability) and inattention (insensitivity to re nancing incentive) in the same direction The unconditional correlation of inertia and inattention is 0.82 But we can reject a deterministic linear relationship between these variables Middle-aged and older households have more inertia and inattention than younger households Education reduces inertia and inattention, particularly among more educated households ncome reduces inertia and inattention, particularly among poorer households Wealth increases inertia and inattention, particularly among wealthier households Could this result from a high cost of time, or less attention paid when mortgage is relatively less important? Andersen et al (2014) nattention and nertia GWU/FRB / 20

82 But we can reject a deterministic linear relationship between these variables Middle-aged and older households have more inertia and inattention than younger households Education reduces inertia

20 Conclusion Caveats and Next Steps This is a highly preliminary rst draft, with several methodological weaknesses that we hope to address in future versions We assume equal moving probability for all households Would be better to estimate moving probability as a function of household characteristics We focus on interest-rate reduction as the motivation for re nancing Maturity extension may be another relevant motive (Danish system imposes no barriers to this) We do not use information on the exact timing of re nancing within the year We do not use the fact that errors of commission must have occurred before the start of our sample if households have old mortgages with high interest rates. Andersen et al (2014) nattention and nertia GWU/FRB / 20

extension may be another relevant motive (Danish system imposes no barriers to this) We do not use information on the exact timing of re nancing within the year We do not use the fact that

Inattention and Inertia in Household Finance: Evidence from the Danish Mortgage Market

Inattention and Inertia in Household Finance: Evidence from the Danish Mortgage Market Ste en Andersen, John Y. Campbell, Kasper Meisner-Nielsen, and Tarun Ramadorai 1 This draft: July 2014 1 Andersen: