Mortgage Costs as a Share of Housing Costs—Placing the Cost of Credit in Broader Context

Abstract

Housing affordability is a key policy concern and an important component of sustainable homeownership. It follows that reducing housing costs without increasing the risk of mortgage default is a critical approach to sustaining homeownership for current and future generations. In this paper, we break down the different elements of housing costs, specifically focusing on the nuances of mortgage costs. We use internal Fannie Mae data to establish a pro forma of housing costs for different owner-occupant borrower profiles over a typical ownership period (all homebuyers, first-time homebuyers [FTHBs], and low-income first-time homebuyers [LI FTHBs]). We find that the biggest contributors to overall housing costs are transactions costs, ongoing utility expenses, property taxes, home improvement costs, and the component of the mortgage interest rate that compensates investors for the time value of money, with utilities and home improvement costs particularly conspicuous for FTHBs and LI FTHBs. The guaranty fees charged by the government-sponsored enterprises and private mortgage insurance are estimated to be less than 6% of the cost of homeownership. These general patterns hold across racial and ethnic groups, although mortgage insurance alone is roughly 6% of total costs for Black and Hispanic FTHBs and LI FTHBs compared to 2% for white FTHBs and LI FTHBs. Overall, our findings suggest that nonmortgage housing costs are key areas that policymakers should focus on to reduce housing costs and foster sustained homeownership rates.

Keywords:

• mortgages
• government-sponsored enterprises
• housing costs
• sustainable homeownership

Acknowledgments

We thank Ricky Goyette and Rebecca Meeker for their data assistance and feedback. Michael Gibson, Miguel Lanza, Kyle Mollica, Nuno Mota, Michael Rapetti, Nick Sapirie, and Nathan Sleeper all provided key data, feedback, and insights. We also received helpful feedback and suggestions from William Doerner, Douglas Duncan, Sarah Edelman, George Galster, Michael LaCour-Little, Lauren Lambie-Hanson, Jenny Schuetz, participants in the AEI Housing Research Collaborative, and three anonymous referees.

Disclaimer

The views expressed are those of the authors and not those of Fannie Mae nor the Federal Housing Finance Agency.

Notes

1 Throughout this paper, government-sponsored enterprise (GSE) mortgages refer to those that are purchased from lenders by Fannie Mae and Freddie Mac and converted into mortgage-backed securities (MBS). “Conventional” refers to loans that are not part of a government-insured mortgage program (e.g., the Federal Housing Administration [FHA] or the Department of Veterans Affairs [VA]), whereas “conforming” refers to loans that fall under the maximum loan dollar amounts set by the government and are eligible to be purchased by the GSEs. The GSEs set underwriting standards that provide guidance for lenders who plan to sell their mortgages to them, but these standards provide a minimum guideline for lenders and actual underwriting rules may be more stringent.

2 As stated in the Federal Housing Enterprise Safety and Soundness Act of 1992, and enforced annually through its Housing Goals and other programs.

3 For more information, see: https://www.nahb.org/News%20and%20Economics/Housing%20Economics/Indices/Housing%20Opportunity%20Index, https://www.nar.realtor/research-and-statistics/housing-statistics/housing-affordability-index, and https://www.hudexchange.info/programs/location-affordability-index/

4 Prepayment risk is a well-known and -modeled component of the time value of money for mortgage investors and is reflected in the GSE MBS rates in our data.

5 There is a long line of literature here; examples include Begley et al. (2021), Caplin et al. (2015), Faber (2018), Fout et al. (2020), Gerardi et al. (2023), Haupert (2022), and Hembre et al. (2021) among many others.

6 The small differences in credit scores across the three samples are not economically meaningful for GSE credit pricing. This is consistent with Beer et al. (2018), who demonstrate that there is only a moderate correlation between income and credit scores.

7 We use Fannie Mae averages, as noted above, for illustrative purposes. Because we are relying on averages, we face differing distributions within each variable. Therefore, using the average purchase price, LTV, closing costs, and mortgage amounts may not align with an individual borrower.

8 Although the ownership period for individual borrowers varies based on a number of factors (age, income, household composition, local housing market, and overall housing market cycle), the average owner duration has fluctuated between 7 and 8 years in recent years:

https://www.attomdata.com/wp-content/uploads/2021/04/Average-U.S.-Homeownership-Tenure-Q1-2021.png. Lengthening or reducing our assumed ownership period for different borrower profiles does not change our results or conclusions meaningfully.

9 Unfortunately, we have fewer data points for sales costs, so we cannot use precise cuts by borrower profile; instead, we use the average broker fee (6.7%) and other sales costs (0.7%) for the whole sample. Given this average is higher than the typically negotiated fee of 6.0% (Schnare et al., 2022), we also try using the median sales broker fee in our data of 5.0%, but our results are similar.

10 We exclude other potential fees such as condominium and homeowner association (HOA) fees, because only 5% of owner-occupied units are condominiums, and only roughly one quarter of owner households pay an HOA fee. However, their inclusion in our analysis would only increase the share of nonmortgage ongoing costs relative to other owner costs.

11 A recent JCHS (2021) report on home improvement expenditures shows that lower income homeowners are less likely to spend money on home upgrades. Similarly, Melzer (2017) finds that homeowners at higher risk of default spend less on both basic repairs and larger home improvements.

12 There is a wider literature discussing the distinction between smaller maintenance expenditures and more substantial improvements, as well as disparities in housing upkeep and repair needs by age, geography, and socioeconomic status; see Begley and Lambie-Hanson (2015), Bendimerad (2005), Davidoff (2004), Divringi et al. (2019), Gyourko and Tracy (2006), Holupka and Newman (2011), and Mayer and Lee (1981), among others.

13 For more information, see 2019 AHS Definitions.pdf (census.gov). Given the well-documented disruptions to various housing data series during the COVID-19 pandemic, we have used data from 2019 or earlier for the costs in this table.

14 Here, we focus on mortgage costs as relevant to the borrower. For a detailed overview of the primary–secondary spread and components of lender profits over time, see Fuster et al. (2013, 2021).

15 https://servicing-guide.fanniemae.com/THE-SERVICING-GUIDE/Part-A-Doing-Business-with-Fannie-Mae/Subpart-A2-Getting-Started-with-Fannie-Mae/Chapter-A2-3-Servicer-Compensation/A2-3-02-Servicing-Fees-for-Portfolio-and-MBS-Mortgage/1581707841/A2-3-02-Servicing-Fees-for-Portfolio-and-MBS-Mortgage-Loans-09-09-2020.htm

16 FHFA puts out an annual summary report on g-fee characteristics and trends, which includes summary information on g-fee dynamics using GSE data (FHFA 2020).

17 For example, Fannie Mae’s LLPA matrix is available here: https://singlefamily.fanniemae.com/media/9391/display

18 Note that 25 basis points is the average LLPA, which we assume is fully included in our calculated zero-point rate. For LLPAs above 25 basis points, the note rate may increase to take into account the additional LLPA. For borrower profiles that face an additional LLPA, we divide the amount above 25 by five and add the additional basis points to the stated note rate.

19 HomeReady^® is also subject to an LLPA cap. Begley et al. (2021) provide an overview of borrower characteristics and early performance outcomes for the HomeReady^® affordable mortgage, along with a view of the borrower cost and performance tradeoffs between FHA and HomeReady across the credit score–LTV matrix. Hembre et al. (2021) similarly provide a comparative analysis of HFA loan performance using GSE data.

20 Although there is also the difficult-to-predict issue of interest rate fluctuations and hedging by lenders between loan closing and sale to the secondary market, this is close to the MBA estimate of an average of 355 basis points in lender net secondary marketing income for 2020, although they also show that this number varies widely based on lender characteristics and the average was smaller in 2018 and 2019 (282 basis points) (Mortgage Bankers Association [MBA], 2021, Tables B2, C2, and D2).

21 For example, mortgage brokers typically charge more for their services than direct lenders do. Lenders may also have their own internal risk-based loan charges and/or lenders may also “buy up” or “buy down” the g-fee in their transaction with the GSEs, both of which will also change their specific profits. For a detailed explanation of lender tradeoffs in the secondary market, see Fuster et al. (2013). Additionally, a discussion of market frictions, increases in lender profits and GOS, and credit supply constraints during COVID is covered in Fuster et al. (2021).

22 https://www.mgic.com/-/media/mi/rates/rate-cards/71-61284-rate-card-pdf-bpmi-monthly-july-2018.pdf?la=en

23 B-8.1-04: Termination of Conventional Mortgage Insurance (05/15/2019) (fanniemae.com)

24 Of course, price appreciation will vary greatly by location, price tier, and market cycle. We believe that a 3.75% nominal growth rate is conservative given we also use a 2.0% discount rate in our analysis, implying real price growth of 1.75% over time. Additionally, the three borrower profiles include housing values that are solidly within the middle tier of typical national housing values for 2020 according to the Zillow ZHVI ($268,418 as of December 2020).

25 We also tried a range of discount rates to reflect different risk premiums and borrower opportunity costs, but higher rates do not substantially change our results. https://www.federalreserve.gov/faqs/economy_14400.htme

26 When we use the median broker fee in our sample of 5.0%, broker fees as a share of total costs fall to 10.3%, and the other costs increase slightly so that utility costs now overtake broker fees. However, the overall results, including the key components of housing costs, remain the same.

27 Papers exploring programs and policy solutions to assist with energy efficiency, weatherization and other home repair costs in detail include: Acquaye (2011), Begley and Lambie-Hanson (2015), Divringi et al. (2019), Fuller et al. (2010), Moulton (2022), Murray and Mills (2014), Perl (2015), Rohe et al. (2010), and Van Zandt and Rohe (2011).

28 Mortgage research overwhelmingly finds that lower income and nonwhite borrowers are less likely to refinance and that lower balance loans are less likely to benefit from refinancing (Agarwal et al., 2017; Gerardi et al., 2023, 2021; Keys et al., 2016).

29 Indeed, innovation is already occurring in this area, with new models of online brokerages offering rebates and flat fees. One such example is Yoreevo: https://yoreevo.com/how-it-works.

30 Only roughly 14% of all taxpayers itemize deductions, making them eligible for such deductions (Eastman & Tyger, 2019).

Additional information

Notes on contributors

Jaclene Begley

Jaclene Begley, PhD, is a senior economist and research director in the Economic & Strategic Research Group at Fannie Mae.

Mark Palim

Mark Palim, PhD, is the deputy chief economist and vice president in the Economic & Strategic Research Group at Fannie Mae.