Paying Too Much? Borrower Sophistication and Overpayment in the US Mortgage Market

ISSN: 1962-5361

Disclaimer: This Philadelphia Fed working paper represents preliminary research that is being

circulated for discussion purposes. The views expressed in these papers are solely those of

the authors and do not necessarily relect the views of the Federal Reserve Bank of Philadelphia

or the Federal Reserve System. Any errors or omissions are the responsibility of the authors.

Philadelphia Fed working papers are free to download at: https://philadelphiafed.org/research-

and-data/publications/working-papers.

Working Papers

RESEARCH DEPARTMENT

WP 24-11

PUBLISHED

June 2024

Paying Too Much?

Borrower Sophistication

and Overpayment in the

US Mortgage Market

Neil Bhutta

Federal Reserve Bank of Philadelphia

Consumer Finance Institute

Andreas Fuster

Swiss Finance Institute at EPFL

Aurel Hizmo

Federal Reserve Board

DOI: https://doi.org/10.21799/frbp.wp.2024.11

Paying Too Much? Borrower Sophistication and

Overpayment in the US Mortgage Market

∗

Neil Bhutta Andreas Fuster Aurel Hizmo

†

June 5, 2024

Abstract

Comparing mortgage rates that borrowers obtain to rates that lenders could oﬀer for the

same loan, we ﬁnd that many homeowners signiﬁcantly overpay for their mortgage, with

overpayment varying across borrower types and with market interest rates. Survey data

reveal that borrowers’ mortgage knowledge and shopping behavior strongly correlate

with the rates they secure. We also document substantial variation in how expensive

and proﬁtable lenders are, without any evidence that expensive loans are associated

with a better borrower experience. Despite many lenders operating in the US mortgage

market, limited borrower sophistication may provide lenders with market power.

∗

This paper was previously circulated as “Paying Too Much? Price Dispersion in the US Mortgage Mar-

ket.” Adithya Raajkumar and Jacqueline Blair provided excellent research assistance. We thank Jason Allen,

Robert Avery, John Campbell, Nick Embrey, Seraﬁn Grundl, Katherine Guthrie, Hajime Hadeishi, Michael

Haliassos, Chris Hansman, Gregor Matvos, Raven Molloy, John Mondragon, Christopher Palmer, Saty Pa-

trabansh, Chad Redmer, James Rowe, David Zhang, as well as seminar and conference participants at the

American University (Kodog), Arizona State University, Bank of England, Baruch College (Zicklin), EPFL,

Federal Reserve Bank of Atlanta, Federal Reserve Board, Freddie Mac, Norges Bank, NYU Stern, Oxford

ıd, University of Copenhagen, AFA Annual Meeting, AREUEA National Conference, CEPR European

Conference on Household Finance (Ortygia), Cherry Blossom Financial Education Institute, FCA-Imperial

Household Finance Conference, FDIC Consumer Research Conference, and the NBER Summer Institute

(Real Estate) for helpful comments. Special thanks to Jay Shultz, Saty Patrabansh, and Bob Avery for

helping us access and use data at FHFA. The views expressed are those of the authors and do not necessarily

reﬂect those of the Federal Reserve Board, the Federal Reserve System, or the Swiss National Bank.

†

Bhutta: Federal Reserve Bank of Philadelphia; Fuster: EPFL, Swiss Finance Institute, and

CEPR; Hizmo: Federal Reserve Board. Emails: neil.bhutta@phil.frb.org; andreas.fuster@epﬂ.ch; au-

[email protected]v.

1 Introduction

Survey data indicate that half of the borrowers taking out a mortgage in the US only seriously

considered one lender, and just 3 percent of the borrowers considered more than three lenders.

Ninety-ﬁve percent of the respondents reported that they were satisﬁed that they received the

lowest interest rate for which they could qualify.

Taking these facts at face value, one might

be led to conclude that there is little variation in mortgage pricing, or that borrowers are very

eﬃcient at ﬁnding the best rates. This might seem a reasonable conclusion, especially when

considering that the mortgage market appears highly competitive: the majority of mortgages

in the US are standardized and guaranteed by the government, and there are hundreds of

lenders oﬀering mortgages on any given day. However, in contrast to borrowers’ perceptions,

we ﬁnd that many borrowers substantially overpay for their mortgage, and that overpayment

varies systematically across borrower types and over time. We argue that this overpayment

reﬂects, at least in part, limited borrower sophistication, which provides lenders with market

power.

To assess overpayment, we use unique data on both available rates—the rates that lenders

could oﬀer for speciﬁc mortgages/borrowers in each market and each day—and data on the

mortgages locked, or obtained, by consumers. The available rates are inclusive of the fees

and markups that a borrower would pay if they chose a particular lender. Importantly, these

“oﬀer” rates are lenders’ private information, rather than publicly posted, and may not be

automatically oﬀered to prospective borrowers. The data on locked mortgages include key

variables for evaluating mortgage pricing, including several that are unavailable in any other

dataset, such as “discount points,” exact time of rate lock (as opposed to the closing date),

and the lock period (e.g., 30 or 60 days).

For a given borrower, we compare the rate they locked against the distribution of available

rates for the same type of loan and borrower (same loan-to-value [LTV] ratio, credit score

[FICO], points, etc.) on the same day in the same market. Motivated by a simple search

model (e.g., Carlson and McAfee, 1983), we construct a metric we refer to as the Expected

Gain from Additional Search (EGain): the expected amount by which a borrower could

reduce their rate if they were to obtain one additional rate quote.

We ﬁnd that EGain diﬀers substantially across borrower types. For example, borrowers

Statistics are based on the National Survey of Mortgage Originations (NSMO), which covers mortgage

originations from 2013 to 2020.

getting Federal Housing Administration (FHA) loans, who tend to have lower incomes and

credit scores, have an average EGain of 28 basis points (bp), and, remarkably, for a quarter

of these borrowers, EGain exceeds 40bp. In contrast, borrowers getting “jumbo” loans—who

tend to have high incomes—have an average EGain of just 4bp, implying such borrowers

are much more likely to obtain rates that come close to the best available. We interpret

these results as consistent with borrowers across diﬀerent market segments varying in their

ﬁnancial “sophistication,” which we think of as encompassing factors such as the recognition

that lenders may oﬀer diﬀerent rates, and the ability to search and negotiate for low rates.

We also ﬁnd that the expected gain from search varies over time, inversely with interest

rates. When market interest rates (and lenders’ oﬀer rates) are higher, the average EGain

is lower, implying that borrowers tend to ﬁnd better deals when rates are high. We argue

that this relationship is in part driven by “behavioral” factors: When the level of rates is

low, borrowers may feel less compelled to search for a good deal or negotiate hard than when

rates are higher, even though in dollar terms the consequences are the same. Indeed, we also

ﬁnd explicit evidence from survey data on mortgage borrowers that shopping eﬀort increases

with market rates.

We provide supporting evidence of ineﬀective shopping and negotiation by analyzing

price dispersion in the locks data alone, where we have a much larger sample and can further

exploit unique features of the locks data. We ﬁnd that the diﬀerence between the 90th and

10th percentile interest rate that observably identical borrowers lock in for the same loan in

the same market, on the same day, and paying the same points, is 55bp; for a typical loan of

$250,000, this corresponds to an upfront payment of about $6,250. Furthermore, the largest

residual dispersion occurs for borrowers who may be the least ﬁnancially sophisticated. For

example, the 90-10 residual rate gap for FHA borrowers is about 72bp.

We then unpack at which level this dispersion occurs—e.g., across lenders vs. within

lenders, branches, or even loan oﬃcers. Estimated lender ﬁxed eﬀects exhibit quite sub-

stantial variation, but still only reduce the residual rate dispersion by a modest amount,

suggesting limited explanatory power of factors such as (time-invariant) lender reputation

or quality. Allowing lender ﬁxed eﬀects to vary by loan and borrower characteristics and

over time improves explanatory power. However, even after further including branch and

Notably, the lending platform providing our data is mostly used by monoline nonbank mortgage origi-

nators, and wide dispersion remains when we limit the sample exclusively to such lenders. Thus, we can rule

out cross-selling or bundling of other services as explanations behind the observed patterns.

loan oﬃcer ﬁxed eﬀects, almost one-half of the residual dispersion remains, consistent with

an important role for negotiation in this market.

In the ﬁnal part of the paper, we present several ﬁndings that support the idea that

mortgage lenders exert market power, which allows them to charge markups, and that a lack

of borrower ﬁnancial sophistication is a source of lender market power. First, we document

that the largest lenders tend to be the most expensive (i.e., have high residualized rates), and

exhibit the most dispersion in their (residualized) rates, suggestive of large lenders having

market power. Next, we use detailed income and expense data for mortgage lenders from

regulatory ﬁlings merged to our data on transacted rates to show that relatively expensive

lenders have higher proﬁt margins (net income per dollar originated). We also ﬁnd that

expensive lenders spend slightly more on occupancy and technology, but substantially more

on salaries of loan oﬃcers and managers. This could imply a better borrower experience; yet

our evidence from survey data indicates that, on the contrary, borrowers do not get better

service when paying more for their mortgage. Finally, again using survey data, we show

that the most sophisticated borrowers (those who shop the most and have high mortgage

knowledge) get signiﬁcantly cheaper mortgages than the least sophisticated. Moreover, this

gap widens as local mortgage market concentration declines, suggesting that more intense

lender competition is helpful primarily for sophisticated borrowers.

Overall, our empirical results indicate that a large fraction of the borrower population

in the US overpays for mortgages, and a key reason for this seems to be a lack of ﬁnancial

sophistication. The borrowers who fare the worst often get government-guaranteed loans

through the FHA program, which is aimed at lowering the cost of homeownership for lower-

income households. Our results suggest that government entities such as the FHA might

consider ways to reduce price dispersion and excessive markups to help fulﬁll their policy

objectives. Our ﬁndings also suggest that the lack of consumer shopping is important for

the pass-through of monetary policy to the mortgage market: Reduced search eﬀort appears

to prevent borrowers’ rates from falling as much as they could when market rates decrease,

thereby weakening the pass-through of expansive policy.

This paper makes several contributions to the literature on overpayment and dispersion

in the price of mortgages and consumer credit more broadly. Previous work in the US has

assessed dispersion in oﬀer rates (Alexandrov and Koulayev, 2017 and McManus et al., 2018);

transacted mortgage rates (Agarwal et al., 2023 and Ambokar and Samaee, 2019); reset rates

on adjustable-rate mortgages (Gurun et al., 2016); and mortgage broker fees (Woodward and

Hall, 2012). Our data contain key details not available in any other dataset (e.g., exact lock

date, lock period, and points), which provides greater conﬁdence that we are identifying true

dispersion. Moreover, our data span multiple market segments—allowing us to show, for

example, that dispersion is especially wide for FHA loans—and come from recent years after

the implementation of post-crisis mortgage regulations (e.g., loan oﬃcer compensation rules)

that could have reduced price dispersion. While we ﬁnd considerable dispersion in mortgage

rates, our estimates are substantively smaller than other estimates in the literature, as we

discuss in a more detailed review of the literature in Appendix A.1.

Additionally, we

document sizable dispersion within lender, branch, and even loan oﬃcer, suggestive of a role

for negotiation, in contrast to assumptions in previous work that negotiation plays little role

in the US mortgage market (Alexandrov and Koulayev, 2017).

Most notably, this paper is the ﬁrst to compare transacted rates to oﬀer rate distributions

for identical loans, allowing us to measure the gains to additional search across all borrowers,

including those at the bottom of the transaction rate distribution, and how gains vary by

borrower type. We also document that borrower overpayment and shopping intensity change

with market rates over time, consistent with behavioral factors aﬀecting shopping eﬀort.

This paper is also the ﬁrst to link data on ﬁrm-level mortgage pricing with revenues, costs,

and proﬁts. We conﬁrm that our residualized lock rates are strongly correlated with ﬁrm

revenues and also show that expensive lenders earn higher margins per dollar originated.

Alongside these results, we provide novel evidence that borrowers who pay more are less

satisﬁed with their lender, and that a lack of borrower sophistication may be undermining

competition in this market.

Our paper connects to the literature on (in)eﬃciency of consumer choice and price disper-

Some work also exists outside the US, where the institutional details and the mortgage market structure

are diﬀerent. Allen et al. (2014b) study the Canadian market, where there is no dispersion in posted rates,

but there is large dispersion in contracted rates, which they argue arises due to diﬀerences in bargaining

leverage across consumers. Iscenko (2018) and Coen et al. (2023) document substantial heterogeneity in

terms of how well UK borrowers fare when choosing mortgages from the options available at their lender;

Liu (2019) shows that many UK borrowers appear to neglect nonsalient fees and that lenders exploit this

in their price setting. Damen and Buyst (2017) provide evidence that mortgage borrowers in Belgium who

shop more achieve substantial savings.

The result that overpayment increases when market rates are low is distinct from and complements the

ﬁnding of Fuster et al. (2023) that lender oﬀers tend to feature higher markups when market rates are low.

Our ﬁnding that lower market concentration may have limited beneﬁts for less-sophisticated consumers

is consistent with Allen et al. (2014a)’s evidence from mergers in the Canadian mortgage market. Malliaris

et al. (2022) also use the NSMO data to document that indicators of sophistication are correlated with rates

and study how the beneﬁts of knowledge and shopping interact.

sion in various consumer ﬁnance markets, including mutual funds (Horta¸csu and Syverson,

2004; Choi et al., 2010), auto loans (Argyle et al., 2022), and credit cards (Stango and Zin-

man, 2016).

In addition to mortgages being the largest household liability, the composition

of mortgage borrowers spans the income, wealth, and ﬁnancial sophistication spectrums,

providing much cross-sectional variation to help shed light on the factors driving dispersion

and overpayment in consumer ﬁnancial markets more broadly. Along with other work docu-

menting costly “mistakes” in the mortgage market (e.g., Agarwal et al., 2015, 2017; Keys et

al., 2016; Andersen et al., 2020), our results are in line with the growing literature pointing to

ﬁnancial literacy/sophistication as a key driver of diﬀerential outcomes in household ﬁnance

(e.g., Hastings et al., 2013; Gomes et al., 2021).

In recent work, Agarwal et al. (2023) argue that overpayment by certain groups need

not imply that they are unsophisticated (or have high search costs), but could be a rational

response of relatively risky borrowers who fear being rejected. They document that the

relationship between contracted mortgage rate and the number of “inquiries” recorded by

credit bureaus—their proxy for borrower search—is U-shaped. This suggests that borrowers

who search a lot may do so because their application gets rejected, which in turn may lead

these borrowers to accept relatively worse oﬀers. This channel may contribute to some of the

overpayment we document. At the same time, however, we ﬁnd considerable overpayment

even among many well-qualiﬁed borrowers and provide evidence from the NSMO data that

variation in sophistication is important to understanding cross-sectional dispersion.

The rest of the paper is organized as follows. In the next section, we provide some

background on the institutional details of this market. Section 3 describes our data sources.

Section 4 explores how locked rates compare to the oﬀer distribution; here we explain our

EGain measure and study how it varies across borrowers with diﬀerent characteristics and

over time. Section 5 assesses mortgage price dispersion in the locks data alone. Section

6 explores the connections between borrower sophistication, mortgage rates, and market

power. Finally, Section 7 concludes with some potential policy implications.

More broadly, a large literature studies price dispersion in various markets empirically and theoretically

(see e.g., Baye et al., 2006, and Wright et al., 2021, for surveys of relevant work). One important conclusion

from this literature is that a decrease in information cost (making shopping easier) does not necessarily

reduce equilibrium price dispersion—neither empirically nor theoretically.

Over the period we study, underwriting standards in the GSE and FHA segments of the market are

largely dictated directly by these agencies. Thus, for the vast majority of borrowers who get approved for a

loan, it should also be easy to get a loan from a diﬀerent lender. That said, the perception that other lenders

are unlikely to accept one’s application may suﬃce to induce a borrower to accept a relatively “bad” oﬀer.

2 Mortgage Pricing and Originations in the US

In this section, we provide a brief overview of some of the institutional details that will be

important for the rest of the paper.

In the US, there are multiple channels through which a borrower can obtain a loan.

One of them is to go directly to a bank or credit union. An alternative is to obtain a

loan through a specialized mortgage originator, a so-called mortgage bank (or independent

mortgage company). These lenders, contrary to what the name suggests, are not depository

institutions and typically do not keep any of the mortgages on their own balance sheet.

Finally, it is also possible to go through a mortgage broker, who may have relationships with

both bank and nonbank originators, and acts as an intermediary connecting borrowers to

those institutions. When a loan is originated directly by a lender who will either retain the

loan in portfolio or sell it directly in the secondary (mortgage-backed securities, or MBS)

market, this is called a “retail loan”; if a loan is originated via a nonbank entity that originates

the loan for another lender, this is called “wholesale.”

Regardless of the channel, a borrower will generally interact (in person or just by phone/online)

with a loan oﬃcer or broker (henceforth, LO) who will have access to various “rate sheets”

that provide the detailed pricing available at a given point in time (generally updated at

least once a day). Importantly, for any loan type and combination of characteristics, there

is no single interest rate—instead, the rate sheet shows a combination of note rates and

“(discount) points.” To obtain a low note rate, a borrower can pay points (where 1 point

= 1 percent of the loan amount). If the borrower is willing to take a higher rate, they can

receive points (often called rebates or credits), which can be used toward origination costs.

In the case of a retail loan, the available pricing will come directly from the lender’s

pricing desk; in the case of wholesale lending, the rate sheets can come from several diﬀerent

lenders (often referred to as “investors”). Each rate sheet will provide pricing for diﬀerent

loan programs (e.g., GSE loans, FHA, or jumbos) with adjustments depending on a few

loan and borrower characteristics, typically FICO, LTV, loan amount, geographic region,

loan purpose, and property type. Pricing depends on the value that a lender assigns to

the loan—often based on the current value of such a loan in the MBS market, where most

For additional discussion, see e.g., Fuster et al. (2013) or https://files.consumerfinance.gov/f/

201301_cfpb_final-rule_loan-originator-compensation.pdf.

loans are ultimately sold.

Prices also take into account required “guarantee fees” set by the

agencies that securitize the loans and insure the credit risk, namely the GSEs and Ginnie

Mae (for FHA/VA loans).

Furthermore, lenders will add a margin that may depend, among

other things, on the level of demand for loans (Fuster et al., 2023).

On top of the prices from the rate sheet, the costs to the borrower include compensation

of the LO and/or their employer (e.g., the mortgage bank). This compensation may be

explicit (via upfront origination fees) or implicit (via lender proﬁt margins on rate sheets).

Historically, LOs had strong incentives to sell loans with higher interest rates, all else equal,

and thereby generate more compensation not only for the lender but also for themselves

(often called the “yield spread premium,” Woodward and Hall, 2012). However, in the wake

of the ﬁnancial crisis, new regulations were imposed so that LO compensation may no longer

vary with the interest rate and other terms of the loan. But lenders, of course, still proﬁt

when borrowers take higher interest rates.

Importantly, this does not imply that all LOs in

a ﬁrm simply get paid an identical, ﬁxed amount for each loan they originate. In fact, LOs

are frequently given a choice between diﬀerent possible compensation plans, for example,

trading oﬀ ﬁxed salary for higher commission rates per dollar of originated loans.

Finally, it is not the case that the combination of rate sheets and a speciﬁc LO’s com-

pensation plan in all cases determine the ﬁnal rate and points/fees that a given borrower

is oﬀered: There may be “exceptions” granted, for instance, to meet a competitive outside

oﬀer. Lenders generally have speciﬁc procedures for these exceptions, since they want to

avoid violating fair lending laws.

An important step in the origination process is the mortgage rate lock. A lock is a

guarantee that the borrower will be issued a mortgage with a speciﬁc combination of interest

rate and points if the mortgage closes by a speciﬁc date. Borrowers typically lock their

mortgage rates as a protection against rate increases between the time of the lock and the

time when the mortgage closes. A lock can occur at the same time a borrower submits a loan

Generally, prices in the MBS market depend on the yields on alternative investments (especially Trea-

suries) as well as investors’ projections of future prepayments of the underlying mortgages.

In addition to the guarantee fee, which is a ﬂow insurance premium over the life of a mortgage, the

GSEs charge upfront “loan-level price adjustments” that depend on borrower and loan characteristics—see

e.g., https://www.fanniemae.com/content/pricing/llpa- matrix.pdf.

These rules were ﬁrst changed in 2011 as part of the Truth in Lending Act; the Consumer Financial

Protection Bureau published its ﬁnal rule on LO compensation requirements in January 2013.

See e.g., https://www.crai.com/sites/default/files/publications/

Managing-the-Fair-Lending-Risk-of%2DPricing-Discretion-Whitepaper-Oct-2014.pdf or

https://www.mortech.com/mortechblog/pricing-discretion-fair-lending-risk.

application with a lender, but it can also happen at a later time. Not all rate locks ultimately

lead to originated mortgages, since the loan application can still be rejected afterward (e.g.,

because the appraisal of the home comes in lower than expected) or the borrower could

renege. However, the lock is binding for the lender, as long as the characteristics of the loan

and borrower (such as the loan amount or the credit score) remain as speciﬁed at the time

of the lock. Lenders typically do not charge an explicit fee for a rate lock, though there are

generally loan application fees. Also, if a loan does not close by the time the lock period

expires, extending the lock typically requires a fee.

3 Data

In this section, we describe our primary data sources: (i) data on mortgage rate locks and

lender oﬀers from Optimal Blue; (ii) data on nonbank income statements from Mortgage

Call Reports (MCR) collected by state regulators; and (iii) survey data from the National

Survey of Mortgage Originations (NSMO).

3.1 Optimal Blue Data

Our main data come from an industry platform called Optimal Blue that connects over 600

mortgage lenders with more than 200 whole loan investors. Through the platform, mortgage

originators can gather information on mortgage pricing, initiate rate locks, manage pipeline

risk, and sell mortgages to investors. More than 40,000 unique users access the system each

month to search loan programs and lock in consumer mortgages. More than 2.4 million

mortgage locks were processed through this system in 2019, thus accounting for about 30%

of loan originations nationally.

The lenders using the platform tend to be nonbank monoline mortgage lenders. These

lenders have gained substantial market share in the post-crisis period (see e.g., Buchak et al.,

2018); in 2019, they originated 56% of all purchase loans and 58% of reﬁnance loans (CFPB,

2020). Optimal Blue is also used by smaller community banks or credit unions. That said,

many institutions on this platform act as correspondent lenders, meaning that they originate

loans intended to be sold to other ﬁnancial institutions such as a large bank like JPMorgan

or Wells Fargo.

For this study, we use two components of the data generated by the platform: (i)) data

on mortgage products and mortgage prices actually accepted by consumers, and (ii)) data

on mortgage products available and mortgage prices oﬀered by lenders.

3.1.1 Mortgage Rate Lock Data

The ﬁrst source of data is the universe of “rate lock” agreements for the mortgages processed

through the Optimal Blue platform. We have access to all the mortgage locks generated

by the platform since late 2013. Since the market coverage increases over the course of

2013-2014, we start using the data from January 2015; we end in December 2019. The data

have wide geographical coverage of about 280 metropolitan areas as well as rural areas. All

of the standard loan characteristics used for underwriting are included: LTV ratio, FICO

score, debt-to-income (DTI) ratio, loan amount, loan program, loan purpose (purchase or

reﬁnancing), asset documentation, income documentation, employment status, occupancy

status, property type, zip code location, etc.

There are a number of unique features of the data relative to servicing data that are

typically used in mortgage research. First, it includes not only the contracted mortgage

rate, but also the discount points or credits associated with that rate (meaning additional

upfront payments made or received by the borrower). Second, we observe the exact time-

stamp of when the lock occurred, while in most other datasets, only the closing date is

recorded, which generally diﬀers from the pricing-relevant lock date by several weeks or even

months. Finally, we have unique identiﬁers for the lender, branch, and loan oﬃcer processing

each mortgage. For some lenders, we can also observe loan oﬃcer compensation, expressed

as a percentage of the loan amount.

While the lock data features numeric lender identiﬁers, it does not directly provide us

with information on the lenders. However, we are able to classify a subset of lenders into

whether they are an independent nonbank or not by relying on a match between Optimal

Blue locks and Home Mortgage Disclosure Act (HMDA) data.

This will be useful later to

assess whether lender type and cross-selling might be driving the patterns in the data that

we document.

We restrict the sample in various ways to ensure we study a relatively uniform set of

Some lenders process compensation outside of the Optimal Blue system, or do not compensate loan

oﬃcers directly on a per-loan basis.

We use a loan-level merge between Optimal Blue, administrative FHA data, and HMDA used in Bhutta

and Hizmo (2021), which covers years 2014-2015. We also match Optimal Blue directly to the expanded

public HMDA data from 2018-2019 to further classify lender types for lenders entering the Optimal Blue

system after 2015.

loans that represent the type of mortgages originated in recent years. We only keep 30-year

ﬁxed-rate mortgages on owner-occupied single-unit properties, with full documentation of

assets and income, and drop self-employed borrowers. We also drop loans for amounts under

$100,000, and those with implausible values for LTV, DTI, or points/credits. Finally, we

drop VA loans and streamline reﬁnances (which are a small part of the sample). This leaves

us with 3.6 million observations. For the analysis in Section 4, we will further restrict the

sample in order to match the locked mortgages to oﬀers for identical characteristics, as will

be described there.

Table 1 presents some summary statistics from the lock data sample that we use for

the analysis in this paper, separating between the four loan programs in the data, since

they diﬀer substantially in terms of borrower and loan characteristics. The four programs

are: conforming (loans typically securitized through the GSEs, Fannie Mae or Freddie Mac),

super-conforming (with loan amounts above the national conforming limit but below the

local limit; the GSEs can still securitize these loans but potentially at slightly worse prices),

jumbo (loan amount above the local conforming limit, meaning the loan cannot be securitized

through the GSEs), and FHA loans (which require mortgage insurance from the FHA and

are typically pooled into securities guaranteed by the government entity Ginnie Mae).

The table shows that FHA loans are most likely to go to ﬁrst-time homebuyers with

low FICO scores and high LTV and DTI. Jumbo loans, the only loan type where the credit

risk is not guaranteed by the government, tend to go to the most creditworthy borrowers

and feature relatively low LTVs. They only constitute about 2% of our sample. The table

also shows that FHA borrowers on average pay fewer discount points than borrowers in the

other programs; Appendix Figure A-1 displays the cumulative distribution of points paid (or

received) by program.

As noted above, not all lenders use the Optimal Blue platform, and not all rate locks

necessarily result in an originated mortgage. Thus, there is a concern that the distribution

of interest rates recorded in our rate lock data may not accurately represent the rates that

borrowers ultimately end up with. However, in Appendix A.2, we show that the interest

rates observed in the rate lock data mirror the interest rates observed in the well-known

McDash mortgage servicing dataset on originated mortgage loans, both in terms of averages

and dispersion. Furthermore, loan/borrower characteristics in Optimal Blue locks also look

very similar to those in the data on originated loans.

See Appendix Table A-2 and Figure A-4 for details. For jumbo mortgages, the locked interest rates in

3.1.2 Mortgage Oﬀers Data

As our second source of data, we collect data on the menu of mortgage products and mort-

gage rates that lenders oﬀer through the platform’s pricing engine. Optimal Blue’s “Pricing

Insight” allows users (e.g., loan oﬃcers) to retrieve the real-time distribution of oﬀers for a

loan with certain characteristics in a given local market (where an oﬀer consists of a com-

bination of a note rate and upfront fees and points that the borrower pays or receives with

this rate). Importantly, the oﬀers we observe are “customer facing,” i.e., rates inclusive of

margins and fees that borrowers could expect to pay at a particular lender. The Insight

interface is designed for lenders to compare their pricing against that of peers.

For any combination of day, MSA, and loan/borrower characteristics, we measure an

“oﬀer” rate for each lender on the platform. This oﬀer rate reﬂects the interest rate (with

zero points) that the lender could oﬀer a prospective borrower, including fees under the

assumption that the loan is originated by the LO who has locked the most loans for that

lender in that market.

If a lender represents multiple diﬀerent investors, the oﬀer we observe is based on the

most competitive investor oﬀer. Thus, a borrower locking in a loan with this lender would

not necessarily get exactly the observed oﬀer rate for three reasons. First, the locked rate

can vary depending on which LO the borrower goes through, since diﬀerent LOs can charge

diﬀerent markups. Second, the LO may oﬀer a loan that is not based on the rate sheet of the

most competitive investor, but on one from a diﬀerent investor.

Third, as noted earlier,

borrowers may be able to negotiate and get an “exception” or a lower rate from the lender.

We conduct daily searches in one local market (Los Angeles), twice-weekly searches in four

markets, and weekly searches for 15 additional markets.

We collect oﬀer distributions for

Optimal Blue tend to be higher than those in McDash, which could reﬂect that the relatively smaller lenders

that use the Optimal Blue platform may not be as competitive for these types of loans as for FHA and

conforming loans. It is also the case that average jumbo loan amounts are somewhat smaller in Optimal

Blue locks than in McDash originations, which could reﬂect some diﬀerential selection of borrowers. The

dispersion of rates is still very similar, however.

The Pricing Insight data are diﬀerent from earlier Optimal Blue data used by Fuster et al. (2023). Those

data were based on rate sheets that did not include LO compensation and origination charges, unlike the

Pricing Insight data we use here (where oﬀers are “all included”).

As explained further in Appendix A.3, we observe a distribution of prices (points) for a given note rate,

which we transform into a distribution of rates for zero points.

One reason why an LO might want to do this is to maintain active relationships with multiple investors.

The markets with twice-weekly searches are New York City, Chicago, Denver, and Miami. The markets

with weekly searches are Atlanta, Boston, Charlotte, Cleveland, Dallas, Detroit, Las Vegas, Minneapolis,

100 diﬀerent loan types, diﬀering across the following dimensions: FICO score, LTV ratio,

loan program, loan purpose (purchase or cashout reﬁnance), occupancy (owner-occupied

or investor), rate type (30-year ﬁxed or 5/1 adjustable), and loan amount. The mortgages

require full documentation of income, assets, and employment, and are for single-unit homes.

An important limitation of the oﬀers data is that we are not able to track institutions over

time or match them directly to the lenders in the lock data, since there is no ﬁxed lender

identiﬁer. The time series is also slightly shorter than for the locks data, as we started

systematically tracking oﬀers in April 2016.

Figure 1 shows the dispersion in mortgage rates available from diﬀerent lenders, pooling

data over time and across all of the 20 metropolitan areas for which we obtained data. To

make oﬀers comparable, we subtract the median oﬀer rate across lenders for the same prod-

uct, day, and metropolitan area. Figure 1 indicates wide dispersion in oﬀer rates. There is a

53bp diﬀerence between the 10th and 90th percentile oﬀers, which is similar to what Alexan-

drov and Koulayev (2017) and McManus et al. (2018) also document. In Appendix A.3, we

further show that the degree of oﬀer rate dispersion is quite similar across diﬀerent types of

loans and borrowers, and across all 20 cities in our sample.

3.2 Mortgage Call Report Data

We use ﬁrm-level data on mortgage-origination-related income and expenses of nonbank

mortgage lenders from Mortgage Call Reports (MCR).

Nonbank ﬁnancial institutions (aka

shadow banks) that hold a state license through the Nationwide Multistate Licensing System

(NMLS) must regularly complete an MCR, which consists of data on mortgage loan activity

(e.g., total mortgage origination volume) and data on ﬁrm ﬁnancial conditions, including

detailed information on income and expenses related to mortgage lending.

In order to study how variation in mortgage pricing across ﬁrms relates to ﬁrm income,

expenses, and proﬁtability, we merge the MCR data with the Optimal Blue locks data. The

MCR data cover about 90 percent of nonbank mortgage originations in 2018-2019; our MCR-

Optimal Blue matched data cover 197 ﬁrms operating as independent nonbank mortgage

lenders in 2018 or 2019 that accounted for about 40 percent of nonbank mortgage originations

Phoenix, Portland, San Diego, San Francisco, Seattle, Tampa, and Washington, DC.

These data are available to Federal Reserve researchers through an agreement with the Conference of

State Bank Supervisors, which owns and operates the system that collects MCR data on behalf of state

regulators. In addition to income and expense data, ﬁrms also report data on assets and liabilities. Similar

data are used by Jiang et al. (2020) to study the capital structure of nonbanks.

in 2018 and 2019. This lower coverage in our matched dataset relative to the MCR data

mainly reﬂects the fact that not all nonbank lenders use the Optimal Blue platform.

3.3 NSMO Data

The National Survey of Mortgage Originations (NSMO) is part of the National Mortgage

Database



program, a joint initiative of the Federal Housing Finance Agency (FHFA) and

the Consumer Financial Protection Bureau (CFPB). It surveys a nationally representative

sample of borrowers with newly originated closed-end ﬁrst-lien residential mortgages, fo-

cusing in particular on borrowers’ experiences getting a mortgage, their perceptions and

knowledge of the mortgage market, and their future expectations. We focus on the ﬁrst 26

waves, including borrowers who took out mortgages from 2013 through 2019.

In addition

to its unique focus on new mortgage borrowers, a key feature of the NSMO is that survey

respondents are matched with extensive administrative data from mortgage servicing records

and consumer credit ﬁles. Thus, we observe precise data on the credit risk of respondents

(e.g., their credit score, DTI ratio) and the terms and conditions of the mortgages that

respondents obtained.

For our analyses in this paper, we impose several sample restrictions. We only consider

mortgages on a household’s primary residence and drop mobile/manufactured homes as well

as 2-4 unit dwellings. In addition, we focus on ﬁxed-rate loans with a term of 30 years, and

drop construction loans or those obtained through a builder, mortgages with an associated

additional lien, and those with more than two borrowers on the loan. Finally, we drop

a few observations where the survey respondent was not a borrower on the loan. These

restrictions leave us with 22,567 mortgages for the analysis.

In all our NSMO analyses, we

use the provided analysis weights, which are based on sampling weights and non-response

adjustments.

These coverage ratios use Home Mortgage Disclosure Act (HMDA) data to measure total market size

(see Bhutta et al., 2017 for more on the HMDA data coverage and nonbank market share). We merge the

MCR and Optimal Blue datasets on lender name, after ﬁrst mapping Optimal Blue lender IDs to lender

names by conducting a loan-level match to HMDA data. We match about 450 mortgage lenders operating

in 2018 or 2019 from Optimal Blue to HMDA, of which about 240 are independent nonbanks.

The public version of the data is available on FHFA’s website. We are able to use a conﬁdential version

of the data that includes geographic information.

The full NSMO dataset contains 39,615 loans originated before 2020. The restriction to 30-year mort-

gages drops just over 10,000 observations; most of the non-30-year mortgages have a term of 15 years. These

sample sizes are as of July 2023, but the data may be revised in the future.

4 Comparing Locked Rates to Oﬀer Rates

In this section, we study whether diﬀerent types of borrowers get good or bad deals, relative

to what is available in the market at the time they lock their mortgage. This will allow

us to assess which types of borrowers tend to “overpay” for their loans and test diﬀerent

hypotheses for what is driving diﬀerences across borrowers and over time.

To do so, we use the distribution of lenders’ oﬀer rates (described in Section 3.1.2) by

day, MSA, FICO, LTV, loan amount, and loan program (i.e., conforming, super-conforming,

jumbo, and FHA). We match locked loans to the oﬀers available for loans with nearly identical

characteristics—see Appendix A.4 for additional detail.

We focus the analysis on purchase

mortgages from the 20 MSAs for which oﬀer data are collected and end up with 67,537

matched loans.

To compare the locked rate to the oﬀers, the main metric we consider is the expected gain

from additional search, which we refer to as EGain. This metric is motivated by a simple

search model (e.g., Carlson and McAfee, 1983).

Suppose that there are n mortgage lenders

that are posting mortgage rate oﬀers for a particular borrower type. Rates are ordered from

lowest to highest: r

≤ r

≤ ... ≤ r

. Borrowers see only the mortgage rates available at the

lenders they meet with, and we assume that each borrower has an equal chance of meeting

any one of the lenders. Suppose a borrower has already found a rate r

and is considering

searching to obtain one additional rate quote from a new lender. Given that the probability

of meeting a lender (diﬀerent from lender k) that oﬀers the rate r

is 1/(n − 1), the expected

gain from an additional search is given by:

EGain

k−1

i=1

− r

)

n − 1

−

k−1

i=1

k − 1

n − 1

. (1)

Intuitively, the term in brackets is the locked rate minus the expected rate available at

the k −1 lenders that are oﬀering rates lower than r

. Of course, the borrower does not know

which lenders are oﬀering rates lower than r

, so we have to adjust the expectation by the

share of these lenders in the remaining population, which is (k − 1)/(n − 1). Therefore, this

The oﬀer rates are for a loan with zero points and fees. To compare locked rates to these oﬀers, we adjust

the locked rate for points paid or received by the borrower based on the empirical relationship between points

and interest rates. See Appendix A.5 for more detail on how we estimate this empirical relationship.

We are grateful to Gregor Matvos for suggesting this measure to us.

is a measure of how much money the borrower is leaving on the table, in expectation, by not

conducting one more search. The measure takes into account not just where the borrower’s

rate is relative to the center of the distribution, but it also depends on the width of the oﬀer

distribution—the expected gain from search is higher the more widely dispersed the oﬀers

are. Furthermore, in our implementation, we take into account that some lenders do not

oﬀer a given loan type at all, which would correspond to r

= ∞.

However, it is important

to note that the measure takes the distribution of oﬀers at a point as given—while, if all

borrowers were to search more, the equilibrium distribution of oﬀers might shift. Thus, it is

best to consider it as a measure of overpayment from the point of view of a single borrower.

We also consider an alternative simpler measure where the locked rate is directly com-

pared to the median oﬀer rate for matched loans—we call the diﬀerence between the two the

Locked-Oﬀer Rate Gap. Positive values indicate that a borrower overpays relative to what a

typical lender (as measured by the median) could oﬀer for the same loan characteristics on

the same day.

One might be concerned that the distribution of oﬀers could be a ﬂawed benchmark

for locked rates if the best oﬀers are not “achievable” for some reason. However, we note

that overall, 5.3% of all borrowers obtain a rate in the best (bottom) 5% of their oﬀer

distribution. Even for FHA borrowers, which we will ﬁnd below to do relatively poorly on

average compared to the available rates, this fraction is 3.5%. Thus, even the best oﬀers are

indeed available to borrowers.

4.1 Cross-sectional Diﬀerences in Expected Gain from Additional Search

Panel A of Figure 2 shows the distribution of EGain for all mortgages in our data. The

dashed vertical line denotes the mean of the distribution. EGain is by deﬁnition bounded

below at zero. The ﬁgure shows that about a quarter of locked loans would in expectation

gain less than 5bp from an additional search. However, the distribution is highly skewed,

and there are borrowers who could expect to lower their rate by 50bp or more by randomly

contacting another lender for an oﬀer on the exact same loan type.

Thus, we take as n the number of unique lenders in a given MSA on a given day that oﬀer at least one

loan type across all programs. However, setting n equal to the number of lenders that do post oﬀers for a

given loan type instead does not qualitatively change the results.

In addition, in Appendix Figure A-3, we validate that oﬀer rates derived from Optimal Blue Insights

closely track oﬀer rates for comparable loans published by Mortgage News Daily, an industry website, as

well as by Zillow and Freddie Mac’s Primary Mortgage Market Survey.

Panel B of Figure 2 and the summary statistics in Table 2 show that the distributions

of EGain vary substantially across diﬀerent loan programs. EGain is largest for FHA

borrowers, with an average of 28bp. Moreover, one-quarter of FHA borrowers would in

expectation gain 40bp or more from an additional search. In contrast, the distributions for

super-conforming and jumbo borrowers look very diﬀerent: Even at the 75th percentile, the

expected gain from an additional search is only 12bp (super-conforming) and 3bp (jumbo),

indicating that many borrowers in these segments come close to the best oﬀers posted on

Optimal Blue.

Table 2 further shows that borrowers with lower FICOs and higher LTVs typically have

higher remaining gains from search and overpay the most relative to the median available

oﬀer. Importantly, this overpayment is relative to oﬀered rates that already incorporate

risk-based pricing for low FICOs and high LTVs. First-time homebuyers and those who

pay discount points also tend to fare worse.

Finally, borrowers at independent nonbanks

appear to overpay by more, on average, than borrowers at other lender types; we return to

discussing potential diﬀerences across lender types below.

It is worth noting that within each of the groups in Table 2, there is substantial dispersion

in EGain and the locked-oﬀer rate gap, as shown by the usually large gaps between the 75th

and 25th percentile. Thus, even for high-FICO or low-LTV borrowers, a nontrivial fraction

of borrowers lock rates well above what other lenders could oﬀer them. However, dispersion

tends to be largest for the groups that on average fare the worst.

Appendix Table A-12 provides analogous summary statistics based on the median income,

college education share, minority share, and mortgage market concentration in a borrower’s

location.

Average EGain is largest in areas with low household income, low shares of

college-educated residents, and high minority shares; these are also the areas where the

dispersion in this measure is larger. In contrast, there is little covariation of the EGain

distribution with local mortgage market concentration.

Note that since we adjusted the mortgage note rate for points paid, this relationship is not “mechanical.”

Income, education, and minority shares are measured at the zip code level based on 2017 American

Community Survey data; mortgage market concentration is the Herﬁndahl-Hirschman Index at the county

level averaged over 2016-2019 using the HMDA data.

4.1.1 Regression Analysis

The summary statistics above indicate that borrowers who are ﬁnancially less well-oﬀ, such as

lower-FICO and higher-LTV borrowers, overpay the most. As noted above, this overpayment

is relative to oﬀered rates that already incorporate risk-based pricing for low FICOs and high

LTVs. Instead of reﬂecting risk-based pricing, overpayment could be driven by borrowers

who are less ﬁnancially sophisticated, meaning that they are less eﬀective at searching and

negotiating for low rates. However, an alternative potential explanation for these patterns

unrelated to sophistication is that lower-FICO borrowers and higher-LTV borrowers tend

to have smaller loans and thus less of an incentive (in dollar terms) to shop around. In

columns (1) and (4) of Table 3, we regress EGain on bins for diﬀerent FICO scores and LTV

ratios, respectively, as well as ﬁne loan amount bins and MSA-by-month ﬁxed eﬀects. It is

indeed the case that borrowers with the largest loan amounts have lower EGain (not shown

in table), i.e., they get better rates relative to what is available in the market. However,

conditional on loan amount, lower-FICO borrowers and higher-LTV borrowers continue to

overpay more, to a similar degree as we observed in Table 2. Thus, such borrowers appear to

obtain more expensive loans (relative to available oﬀers) for reasons beyond the diﬀerential

incentive to shop stemming from loan size variation.

Another potential explanation for why low-FICO and high-LTV borrowers are more likely

to pay too much is that they sort into more expensive lenders. Borrowers might choose

expensive lenders because they oﬀer better service or because they spend more on marketing

and are more visible. To investigate this explanation, we add lender-branch ﬁxed eﬀects

in columns (2) and (5) of Table 3. In these columns, the R

jumps sharply to about 40%

from less than 20%, meaning that lender-speciﬁc pricing diﬀerences explain a fair amount

of variation in EGain. Furthermore, the coeﬃcients on FICO and LTV become slightly

smaller, implying that sorting into lenders does explain some of the overpayment by low-

FICO and high-LTV borrowers. Still, even within the same branch, these borrowers tend to

pay diﬀerentially more relative to what is available in the market. This suggests that they

may be less eﬀective at negotiating a good rate.

Along similar lines, low-FICO and high-LTV borrowers may sort into LOs who charge

more and who may be more experienced and oﬀer more assistance in the loan application

process. In order to test for this possibility, in columns (3) and (6), we directly control for

LO compensation, which typically ranges from 1-2% of the loan amount, for the subset of

lenders that report it. The coeﬃcients on this variable are strongly signiﬁcant; however,

the coeﬃcients on FICO and LTV change little, implying that low-FICO and high-LTV

borrowers do not pay higher rates simply because they match with expensive LOs. Rather,

these borrowers appear to pay more than other borrowers even when working with LOs with

the same compensation—suggesting, again, that they may be less eﬀective in shopping and

negotiating for a good rate.

Robustness. Table A-13 in the Appendix reproduces these regressions with the Locked-

Oﬀer Rate Gap as dependent variable; the patterns are similar. In another robustness

check, reported in Table A-14, we restrict the sample to lenders that we can identify as

independent nonbanks. Doing so leaves the coeﬃcients from Table 3 essentially unchanged.

As noted earlier, the nonbank lenders that constitute the majority of our sample are only in

the business of originating mortgages. Thus, diﬀerences in EGain cannot be explained by

potential price advantages that bank lenders might grant to ﬁnancially well-oﬀ (high-FICO,

low-LTV) customers, for instance because they also have signiﬁcant account balances or

other business with the bank.

Finally, it may be that many of the lenders making oﬀers in our dataset are small and

hard to ﬁnd. To ensure our results are unaﬀected by this possibility, we replicate our analysis

using only oﬀers from high-volume lenders, as designated on the Optimal Blue platform. Our

results, shown in Appendix Table A-15, again remain qualitatively unchanged.

4.2 Time-Series Movements in Expected Gain from Additional Search

The last section explored the cross-sectional patterns in the expected gain from additional

search. In this section, we instead study how EGain moves over time, with a particular focus

on how it responds to changes in market interest rates. Are borrowers more likely to end up

with worse rates (relative to what is oﬀered in the market) when market rates are low, and

more likely to get a good deal as rates increase? If so, what might explain this relationship?

Figure 3 plots the average EGain against market interest rates, here measured by the 10-

year Treasury yield.

In the summer of 2016, the level of market interest rates as shown by

Table 2 showed that borrowers who obtain loans from nonbanks tend to have higher EGain and locked-

oﬀer gaps; this could either reﬂect overall advantageous pricing by banks/credit unions, or selection by

borrowers. What we emphasize here is that such diﬀerential pricing, if it exists, does not appear to vary

with borrower creditworthiness.

For the average EGain series, we use the estimated month ﬁxed eﬀects from a regression similar to

Treasury yields was very low. EGain during this time was high, meaning that borrowers were

locking rates from the higher end of the oﬀer rate distribution. As Treasury yields increased,

and as a result lenders increased their oﬀer rates, EGain shrunk, indicating that borrowers

moved toward the cheaper end of the oﬀer distribution. When rates fell again starting in

late 2018, the inverse happened. Overall, the movements in average EGain almost mirror

movements in Treasury yields.

We conﬁrm the statistical signiﬁcance of the relationship between EGain and market

rates in Table 4.

The ﬁrst column regresses EGain on the 10-year Treasury yield, con-

trolling for all borrower characteristics jointly, MSA ﬁxed eﬀects, and MSA-level house price

growth over the past 12 months to account for housing market “hotness,” which could aﬀect

borrowers’ willingness to spend time shopping around. The coeﬃcient implies that as the

10-year Treasury yield increases by 1 percentage point, the average EGain falls by about

5bp. This is sizable, given that over our sample as a whole, EGain averages 18bp.

In column (2), we add month ﬁxed eﬀects (interacted with MSA) and see that the rela-

tionship between Treasury yields and EGain is similarly strong within-month. Column (3)

further adds lender-branch ﬁxed eﬀects to see to what extent the estimated relationship gets

weaker once we control for potentially time-varying selection of borrowers into expensive or

cheap lenders/branches. The coeﬃcient on the Treasury yield is reduced (to 2.6bp), suggest-

ing that some of the overall relationship may be due to borrowers selecting cheaper lenders

when rates are higher (consistent with additional shopping).

In the remaining three columns, we test to what extent this relationship may be driven

by aﬀordability constraints: When market rates rise, constrained borrowers may be forced to

shop more in order to ﬁnd a relatively good rate that will allow them to qualify for a mortgage

(Bhutta and Ringo, 2021). To test this mechanism, we interact the Treasury yield with an

indicator for borrowers with a DTI no higher than 36% (who are likely unconstrained by the

payment burden from higher rates).

The interaction term in columns (4)-(6) is positive

and signiﬁcant, meaning that the relationship between EGain and market rates is indeed

weaker for unconstrained borrowers. However, the magnitude of the estimated coeﬃcients

those in Table 3 but controlling simultaneously for FICO and LTV. We use the 10-year Treasury yield as

our measure of market rates since it is strongly correlated with the 30-year ﬁxed mortgage rate, but avoids

potential endogeneity issues due to the measurement of the latter. However, using the mortgage rate or the

current-coupon MBS yield instead leaves our conclusions unchanged.

Appendix Table A-16 repeats the analysis with locked-oﬀer rate gap as dependent variable.

Using alternative DTI cutoﬀs to separate borrowers, e.g., 43%, leaves the results qualitatively unchanged.

is relatively small, meaning that even for unconstrained borrowers, EGain is higher when

market rates are lower.

This suggests that the relationship may be driven at least partly by “behavioral” factors:

When the level of rates is already low, borrowers may feel less compelled to search for a

good deal or negotiate hard than when rates are higher, even though in dollar terms the

consequences are the same. This might be the case particularly after a recent drop in rates,

as borrowers might compare their oﬀer to a higher reference level.

In Appendix Section A.7.2, we use the NSMO data to provide direct evidence that shop-

ping eﬀort increases with market rates. For example, we ﬁnd that when market rates are

higher, recent mortgage borrowers are more likely to report having considered or applied to

more than one lender. These patterns hold even after controlling for local house price growth

to account for variation in market “hotness,” which might be correlated with interest rates

and aﬀect borrowers’ ability to spend time shopping for a cheaper mortgage.

Importantly, the higher EGain when market interest rates are low is in addition to the

higher “price of intermediation” when rates are low, identiﬁed by Fuster et al. (2023). That

paper shows that oﬀers feature higher lender markups (relative to loan values in the MBS

market) at times of high demand and provides evidence that this is at least in part driven

by lender capacity constraints.

Thus, there are two complementary reasons why after a

drop in market rates, borrowers obtain worse mortgage rates than they could in a frictionless

world: Lenders make worse oﬀers relative to an MBS-market benchmark, and borrowers fare

worse relative to those oﬀers.

5 Understanding Variation in Locked Rates

In the previous section, we observed that many borrowers overpay relative to the rates

available for the same mortgage in the same location on the same day. Furthermore, this

overpayment is ampliﬁed when market interest rates are low. In this section, we take a more

comprehensive look at the dispersion in rates locked in by borrowers that are identical on

observables (including the day of the lock and the location). This analysis is closer to existing

literature on this topic (reviewed in Appendix A.1), but the additional detail available in

Capacity constraints could also aﬀect relative bargaining power and contribute to the patterns docu-

mented in this section. When market rates are low and mortgage demand is very high, a borrower potentially

has less bargaining power because a loan oﬃcer may not care as much about losing a customer, since there

are many others waiting to reﬁnance (and the loan oﬃcer is already at or close to capacity).

our locks data allows us to more precisely measure and unpack price dispersion than what

earlier work was able to do.

To investigate dispersion in locked mortgage rates, we regress locked rates on borrower

and loan characteristics, as well as time eﬀects, and then add an increasingly ﬁne set of ﬁxed

eﬀects. Our outcome of interest is the remaining dispersion in the residual, which we measure

in terms of standard deviations and as the gap between 75th-25th or 90th-10th percentiles.

Comparing the residual dispersion along with the adjusted R

across speciﬁcations allows us

to assess the relative importance of diﬀerent drivers of price dispersion in this market.

Table 5 shows the results from various speciﬁcations, estimated on the same set of nearly

3 million loans locked over the ﬁve-year period 2015-2019.

In the ﬁrst column, as a bench-

mark, we include only lock date-by-MSA ﬁxed eﬀects, in order to document the amount of

overall interest rate dispersion within the same MSA on the same day. These day-by-MSA

ﬁxed eﬀects explain just under 60 percent of the total variation in rates, and the standard

deviation of the residual is 33bp. Given the importance of pure time-variation to explain

variation in rates, in the remaining columns, we also report an adjusted R

within day and

MSA (i.e., after absorbing the ﬁrst set of ﬁxed eﬀects).

In column (2), we add our baseline set of controls: an extensive set of underwriting

variables, which consist of fully interacted bins of values for FICO, LTV, and loan program,

interacted with lock month to allow for time-variation in risk pricing.

We also include

borrower zip code ﬁxed eﬀects, lock period ﬁxed eﬀects, property type ﬁxed eﬀects, cubic

functions of loan amount and DTI, as well as linear controls for FICO and LTV (to allow

for within-bin variation).

This speciﬁcation is similar to regressions one could typically

run with a mortgage servicing dataset.

Within day and MSA, the adjusted R

from these

variables is 0.36, and substantial dispersion remains: The standard deviation in residuals is

The estimation drops “singleton” observations that are completely determined by the set of ﬁxed eﬀects.

There are more such singletons as we add more ﬁxed eﬀects; to ensure that our results are not driven by

changing samples, we use the sample from the most restrictive speciﬁcation (10) in all speciﬁcations. However,

using the largest possible sample for each speciﬁcation instead does not materially aﬀect the results.

We include 13 FICO bins, 9 LTV bins, and 12 dummies for the four loan programs interacted with

three loan purposes (purchase, rate reﬁnance, and cashout reﬁnance). The choice of FICO and LTV bins is

motivated by the loan-level price adjustments set by the GSEs.

The lock period typically varies from 15 to 90 days, with 30 and 45 days being the most common choices.

A longer lock period leads to a slight increase in the fee (or equivalently the interest rate).

It is already somewhat more precise, since here we control for the date in which a loan is locked, along

with the length of the lock period, while in typical dataset loans originated in the same month may have

been locked in diﬀerent months. In Appendix A.1, we report an alternative, more realistic comparison.

0.26, and the borrower at the 90th percentile of the residual distribution pays 58bp more

than the borrower at the 10th percentile.

Column (3) adds bins for the points paid or received by the borrower (interacted with

program by lock month).

This (usually unobserved) variable indeed explains some of the

rate diﬀerences across borrowers, but substantial dispersion remains—e.g., the 90th-10th

percentile diﬀerence is still 55bp. The adj. R

of 0.42 indicates that standard underwriting

variables and upfront payments explain less than half of the variation in interest rates paid

across borrowers within the same day/MSA.

Based on the regression coeﬃcient on discount points (not shown in the table), we can

translate interest rates to upfront points. This coeﬃcient implies that 1 discount point

changes the interest rate by about 21.8bp on average (see Appendix A.5 for details). There-

fore, 55bp in rate is approximately equivalent to 2.5 upfront discount points or 2.5% of the

mortgage balance. In other words, our results imply that a borrower with a $250k mortgage

borrowing at the 90

percentile interest rate should be getting—but in fact is not getting—a

lender credit of about $6,250 relative to someone borrowing at the 10

percentile interest

rate. Alternatively, if one prefers to think in terms of mortgage payments, 55bp correspond

to about $80/month for a $250k loan at the average level of rates over our sample period.

Before discussing the remaining columns of Table 5, we note that Table 6 shows how the

residual dispersion in interest rates varies across diﬀerent loan programs and characteristics.

The middle column of the table uses the residuals from speciﬁcation (3). We see an extreme

amount of dispersion for the two lowest FICO groups. We also see substantial dispersion

for FHA-insured loans, despite the fact that these loans are fully insured by the government

and thus lenders and investors take very little, if any, credit risk. In other words, it is highly

unlikely that unobserved risk factors could explain the wide dispersion in FHA interest rates.

Along the same lines, we also ﬁnd fairly wide dispersion for conforming and super-conforming

loans, which meet the credit standards of the GSEs and will likely be purchased and fully

guaranteed by these institutions.

Finally, we also see wide dispersion even when we focus

We include 8 point bins, as well as a linear function in points to allow for within-bin variation.

One caveat here is that lenders may be worried about so-called put-back risk where loans in default must

be repurchased by the lender due to some defect in the underwriting found by the FHA or GSEs. In the case

of the GSEs, Goodman (2017) documents that put-back risk has been negligible since lenders have stopped

issuing low-documentation and other non-traditional loans. For FHA loans, perhaps the biggest concern

for lenders has been litigation risk under the False Claims Act, which allows the federal government to sue

lenders that knowingly submit false or fraudulent claims to the FHA. Under the Obama Administration,

some of the largest lenders settled with the government, paying ﬁnes close to $5 billion. That said, this risk

just on low-risk borrowers: those with prime FICO scores in excess of 680, and those with

LTVs of less than 75%.

Jumping back to Table 5, in column (4) we add lender ﬁxed eﬀects to allow for the

possibility that some of the rate dispersion may reﬂect diﬀerences in lender characteristics

such as service quality or advertising costs; we study correlates of these estimated ﬁxed eﬀects

in Section 6. We ﬁnd that adding these lender eﬀects decreases the 90th-10th percentile

diﬀerence by 7bp and increases the within-day-and-MSA adjusted R

by 0.1. In columns (5)

and (6), we additionally interact the lender ﬁxed eﬀects with lock day ﬁxed eﬀects and other

controls, to allow for the possibility that lenders’ (relative) pricing may change over time, or

may diﬀer across loan types. This reduces the 90-10 gap by a further 10bp from column (4),

and adds another 0.1 to the adjusted R

. These results suggest that in addition to time-

invariant diﬀerences across lenders, price dispersion may reﬂect lender pricing strategies that

vary over time and across programs. Such variation would make it diﬃcult for borrowers

to ﬁnd low rates simply by following the recommendations of family, friends, or real estate

agents—yet this is a common approach borrowers take to ﬁnding a mortgage.

In columns (7) and (8), we further allow for pricing to vary across diﬀerent branches

of a lender. As discussed earlier, the lenders in our dataset tend to be nonbank monoline

mortgage lenders and community banks. For a typical lender in our data, in a given MSA,

most loans are originated through just 2 or 3 branches located within that MSA. Diﬀerential

branch pricing could reﬂect diﬀerences in convenience of the oﬃce location and/or costs (e.g.,

oﬃce rent). In addition, as noted earlier, diﬀerent branches can have diﬀerent markups and

pricing strategies.

The branch ﬁxed eﬀects in column (7) have noticeable incremental explanatory power,

increasing the adjusted R

and reducing the residual dispersion. Adding branch-by-month

ﬁxed eﬀects in column (8) has more modest eﬀects. Column (8) should come close to looking

at observably identical borrowers getting a loan from the same branch at the same time, yet

the 90-10 gap remains at 31bp, and the interquartile range at 14bp.

Lastly, in columns (9) and (10), we further allow for pricing to vary across diﬀerent

LOs in the same branch, which could reﬂect for instance diﬀerences across LOs in terms of

experience, compensation, or willingness/ability to negotiate. Which LO a borrower matches

is most salient for large banks with signiﬁcant capital at risk, unlike the nonbanks that dominate our data.

Also, this risk has eased in recent years.

We discuss correlates of estimated branch ﬁxed eﬀects (within lender) in Appendix A.6; see also the

discussion in Section 6.1 below.

up with (within a branch) does appear to matter somewhat for the rate they end up with,

since the adjusted R

further increases and the residual dispersion decreases in the last two

columns. Nevertheless, even after including LO ﬁxed eﬀects that are allowed to vary across

time and programs, the 90-10 gap remains at 26bp.

The last column of Table 6 shows that the cross-sectional patterns in residual dispersion,

already discussed above, remain similar in the most restrictive speciﬁcation (10): The dis-

persion is substantially larger for FHA loans, low-FICO borrowers, or ﬁrst-time homebuyers.

The ﬁnal rows of the table show that the residual dispersion is identical if we only consider

loans that were locked with lenders that we are able to classify as independent nonbanks

(as discussed in Section 3.1.1). This suggests that the large dispersion is not driven by un-

observable pricing adjustments that banks or credit unions might make for customers who

already have accounts or other business with them.

To sum up the ﬁndings from this analysis, there is a large amount of dispersion in the

rates that observably identical mortgage borrowers pay, even after controlling for the exact

timing and upfront payments. Adding lender, branch, and LO controls reduces the residual

rate dispersion by about half. However, substantial dispersion remains, implying that two

observably identical borrowers may get quite diﬀerent deals from the same lender branch

or even the same loan oﬃcer at the same time. Furthermore, this appears to be more

pronounced for ﬁnancially less well-oﬀ and potentially less sophisticated borrowers.

6 Borrower Sophistication and Lender Market Power

Thus far, we have presented evidence that many borrowers lock in mortgage rates that are

high relative to available oﬀers in the market. We have also seen that there is considerable

dispersion in locked rates within lender and within location (e.g., zip code). These results

suggest that price diﬀerences across borrowers cannot be fully explained by lender-speciﬁc

attributes (e.g., service quality) or local market area characteristics (e.g., lender concentra-

tion). Moreover, we ﬁnd that rate dispersion and the expected gains from additional search

are highest among borrowers who may be the least ﬁnancially sophisticated, such as low-

FICO and FHA borrowers. These facts together suggest an imperfect market where lenders

are able to charge markups to many borrowers, especially those who may lack the ability to

shop and negotiate eﬀectively.

As noted earlier, in most US cities, there are many active lenders, and market concen-

tration is low (Amel et al., 2018). With many active ﬁrms selling a highly standardized

product, it may be surprising to observe signiﬁcant markups and price dispersion. However,

theoretically, less concentration may not always improve competition and reduce markups.

For example, the model of Armstrong and Vickers (2022), where consumers only consider

subsets of ﬁrms when choosing from which one to buy, has the feature that entry does not

always lower prices (see also Gabaix et al., 2016). As Syverson (2019) discusses, ﬁrms have

“market power” when they can inﬂuence the price they charge as they do not face a perfectly

elastic demand curve, and concentration can be a poor proxy for market power.

In this section, we provide further evidence, using additional data sources, that price

dispersion reﬂects an imperfect market where lenders exert market power, as borrowers

lack the ability to shop eﬀectively. This section proceeds in four subsections. In the ﬁrst,

we measure lender “expensiveness” and document which ﬁrm characteristics (e.g., size) are

correlated with expensiveness. In the second, we use the matched MCR-Optimal Blue data

to test whether expensive lenders earn higher proﬁts or have higher costs. In the last two

subsections, we turn to the NSMO data. We ﬁrst use these data to help assess whether more

expensive lenders diﬀerentiate themselves by providing better service to borrowers. Then,

we test how borrowers’ sophistication relates to the mortgage rates they get and whether

competition may be undermined by a lack of borrower sophistication.

6.1 Which Lenders Are the Most Expensive?

We begin by studying whether variation in lender “expensiveness” is related to other lender

attributes such as size and type of ﬁrm (i.e., bank or nonbank). We measure expensiveness

as the estimated lender ﬁxed eﬀects from speciﬁcation (4) of Table 5. There are 678 unique

lenders in the data, but for this analysis we restrict to the 452 lenders that locked at least

100 loans in the sample used for the regressions in Table 5. Figure 4 displays the distribution

of ﬁxed eﬀects. The standard deviation of these ﬁxed eﬀects is about 0.19, or 19bp. Thus,

there is considerable variation across lenders in their average mortgage rate expensiveness,

in addition to considerable within-lender dispersion as we showed earlier.

Next, we run lender-level regressions of the lender ﬁxed eﬀects on lender size, proxied here

by the number of locks a lender makes in total, and lender type (nonbanks vs. banks). To

facilitate interpretation, we simply divide lenders into size quartiles. Column (1) of Table 7

shows that larger lenders tend to be signiﬁcantly more expensive, and that nonbanks are

more expensive than banks in the sample (for some lenders, type cannot be determined,

but their ﬁxed eﬀects are not signiﬁcantly diﬀerent from the banks’, which are the omitted

category). In column (2), we add the share of locks by a lender that are FHA, jumbo and

superconforming loans (with conforming loans as the omitted category). Lenders with higher

shares of FHA loans are signiﬁcantly more expensive, and adding this information on loan

composition increases the explanatory power of the regression substantially.

In the third column, we ﬁnd that lenders with higher within-lender rate dispersion also

tend to be more expensive.

Furthermore, column (4) shows that once we control for

the lender size, loan type, and the variation in the residual, FHA share and within-lender

dispersion remain highly statistically signiﬁcant, while the eﬀect of lender size is reduced.

This reﬂects in part that larger lenders exhibit more variation in the rate residual (see last

column of Table 7).

In sum, we ﬁnd that larger lenders, and those that originate a larger share of FHA

mortgages, are more expensive on average, and also exhibit more variation in the mortgage

rates their borrowers lock, after ﬁnely accounting for observable characteristics. One possible

interpretation of these ﬁndings is that larger lenders may have market power that enables

them to charge more on average, while still oﬀering competitive rates to some borrowers

(likely those who shop around and/or negotiate).

Alternatively, some lenders may be able

to have high market share and charge higher prices if they can diﬀerentiate themselves by

providing better service to borrowers. Sections 6.2 and 6.3 look at whether loan pricing is

related to lender costs and customer satisfaction.

6.1.1 Branch expensiveness

In Appendix A.6, we describe a similar analysis to understand pricing variation across

branches within lenders. Similar to the ﬁnding for lender expensiveness, we document that

Recall that the lender ﬁxed eﬀects are estimated from regressions where the loan type, as well as many

other variables, are controlled. Thus, this regression is not reﬂecting that FHA loans are more expensive;

rather, it indicates that lenders that make many FHA loans are on average more expensive across all loan

types.

Dispersion is measured as the standard deviation of residuals from speciﬁcation (6) of Table 5—i.e., after

accounting for lender ﬁxed eﬀects that are further allowed to vary across loan types and over time. The 10th

percentile of this variable is 0.07, the 90th percentile is 0.20.

The Armstrong and Vickers (2022) model predicts that larger ﬁrms choose higher maximum prices, in

line with our evidence. In their model (where ﬁrms set a range of prices and play mixed strategies), this

is because some consumers only consider one ﬁrm they are aware of; so large ﬁrms have less to gain from

setting lower prices to attract consumers who consider multiple ﬁrms.

branches that originate more FHA loans are more expensive. We also ﬁnd that the local

characteristics of the areas where branches are originating loans correlate with their pricing.

For example, branches in areas with a lower share of college-educated population are more

expensive. This ﬁnding is reminiscent of ﬁndings in Drechsler et al. (2017) that deposit

spreads vary with local college-education shares, which suggests a lack of depositor sophisti-

cation may be a source of market power in deposit markets. In Section 6.4 below, we use the

NMSO data to measure borrower sophistication more directly and at the individual level,

and we ﬁnd that sophistication strongly predicts the rates borrowers get.

6.2 Do Expensive Lenders Earn Higher Proﬁts?

In this subsection, we ask whether the large diﬀerences in loan expensiveness across lenders

documented above are reﬂected in higher costs or in higher proﬁtability reported by these

lenders. To do this, we merge our measure of lender expensiveness from Optimal Blue with

quarterly ﬁnancial ﬁlings of these lenders (the MCR data; see Section 3.2). For our sample

period of 2015:Q1-2019:Q4, there are 1897 lender-quarter ﬁlings, from 162 unique lenders.

The ﬁrst three columns of Table 8 provide summary statistics on the main line items

from lenders’ ﬁnancial statements.

The median of reported gross income is $4.75 per $100

originated, and most of it comes from secondary market income (or “gain-on-sale”), which

is the income lenders earn from selling the loans they originate in the secondary market.

The median of gross expenses for lenders is $4.16 per $100 originated, with the majority

of expenses going toward personnel expenses. The second-to-last line shows after-tax net

income for only residential origination before corporate allocations. The median of this

measure is $0.49 per $100 originated. The last line of the table shows net income across all

lines of business (including servicing) after taxes and corporate allocations, with a median

of $0.32 per $100 originated. Note that there is also substantial heterogeneity in all of these

variables across lenders within the same quarter. The cross-sectional standard deviation of

gross and net income is $1.97 and $1.87 per $100 originated, respectively.

Next, we investigate whether variation in lender “expensiveness” is related to these mea-

sures of lender income and costs. As in the previous subsection, we measure expensiveness as

the lender ﬁxed eﬀects from speciﬁcation (4) of Table 5. The last column of Table 8 displays

Both income and expenses are only for origination, warehousing, and secondary marketing of mortgages

for 1-4 unit residential and do not include income and expenses from other lines of business such as servicing,

multifamily/commercial, or residential property portfolio management.

results from separate median regressions of each line item on our measure of lender expen-

siveness (expressed in percentage points) and year-quarter ﬁxed eﬀects.

A 1 percentage

point increase in interest rates charged by lenders is associated with an extra gross income

of $4.05 per $100 originated. The size of the eﬀect is in line with expectations, as it implies

that the value of 1 percentage point of interest rates to lenders is about 4 upfront points,

which implies a reasonable point-rate tradeoﬀ of about 25bp. Of the diﬀerent income line

items, lender expensiveness has the largest eﬀect on secondary market income, although it

does also have a moderate eﬀect on origination income (i.e., fees). While these results are

not surprising, they serve to validate that our measure of lender expensiveness is related to

income exactly as we would expect.

Turning to lender costs, a 1 percentage point increase in interest rates charged is associ-

ated with $3.50 of extra gross expenses per $100 originated. Most of these extra expenses

reﬂect increased personnel expenses, particularly payments to loan oﬃcers and managers.

More expensive lenders do spend a bit more on technology, occupancy, and equipment; how-

ever these eﬀects are fairly small.

Lastly, we ﬁnd a positive and signiﬁcant eﬀect on net income: A 1 percentage point

increase in interest rates charged is associated with $0.45 of extra net income for $100 of

residential loans originated, and $0.24 of extra net income across all lines of business after

taxes and corporate allocations.

Overall, we ﬁnd that the substantial variation in our estimated ﬁxed eﬀects (shown in

Figure 4) is also reﬂected in the gross and net income that nonbank lenders report. Thus,

it is not the case that the variation in interest rates across borrowers “nets out” within a

lender such that some borrowers would cross-subsidize others but without an overall impact

on the lenders; rather, some lenders are indeed more expensive and proﬁtable than others.

Although more expensive lenders also have higher costs, these higher costs are not suﬃcient

to oﬀset their higher income. Perhaps even more importantly, the higher costs mostly reﬂect

payments to loan oﬃcers and managers; only to a minor extent do more expensive lenders

spend more on technology or occupancy (e.g., oﬃce rental) that would arguably improve

We focus on median regressions in order to minimize the eﬀect of outliers and possible data errors in the

ﬁlings data. In Appendix Table A-17, we show results from OLS (mean) regressions, which are qualitatively

similar, although less precise.

This point is further illustrated in Appendix Figure A-2, which shows that the shares of borrowers who

get particularly good deals or particularly bad deals are negatively related within lenders—not positively as

in a world with pure cross-subsidization.

borrowers’ experience. Of course, it is still possible that the more highly paid employees at

expensive lenders provide superior service to borrowers; we test this hypothesis in the next

subsection.

6.3 Are Expensive Lenders Better?

Above, we saw that lenders that charge more tend to have higher personnel costs. This

leaves open the possibility that the loan oﬃcers at expensive lenders are of “better quality”

and provide a better service to customers, which in turn might justify their higher costs.

We cannot directly test this hypothesis within our Optimal Blue and MCR data, but we

can do so using the NSMO survey data. This survey asks respondents a battery of questions

about the quality of their experience in obtaining a mortgage. For example, the survey

asks “yes/no” questions about whether borrowers experienced delays in their closing date

and in paperwork processing. It also asks borrowers about their level of satisfaction (“very,”

“somewhat,” or “not at all” satisﬁed) with their lender, with various aspects of the lending

process, and with the interest rate that they got.

To test whether borrowers who pay more receive better service, we run a series of regres-

sions of lender service quality indicators on the interest rate borrowers obtained, conditional

on a detailed set of controls:

ijtw

= βRate

+ ΓZ

+ α

+ δ

+ 

ijtw

. (2)

In equation (2), Y refers to a service-quality outcome for borrower i with mortgage

characteristics j who originated a loan in month t and was surveyed in wave w. Rate

is the

contract rate on the mortgage for borrower i, and Z

is a rich set of borrower and mortgage

characteristics. The full list of controls is provided in the note to Table 9; it contains,

for instance, ﬂexible controls for credit score and LTV, ﬁxed eﬀects for county, program

(e.g., GSE or FHA) and purpose (purchase or reﬁnance), as well as borrower characteristics

capturing income, employment, wealth, race, and more.

We further include origination

month ﬁxed eﬀects α

, which will absorb any economy-wide changes in service quality (e.g.,

when lenders hit capacity constraints), and survey wave ﬁxed eﬀects δ

One limitation of the NSMO data is that they do not contain a direct measure of points paid or received

by the borrower. However, the controls for borrower wealth and expected time in the mortgage should help

absorb diﬀerences in rates due to variation in points.

We emphasize here that most of the right-hand-side variables in equation (2) are drawn

from administrative data rather than being self-reported. Because of the large number of

precisely measured loan and borrower controls, we interpret β as the relationship between

service quality and mortgage expensiveness, as the controls account for variation in mortgage

rates that are due to borrower risk, loan type, and aggregate time series ﬂuctuations.

The results in Table 9 fail to support the notion that more expensive mortgages are asso-

ciated with better service quality. In fact, more expensive mortgages appear to be associated

with more delays (though not statistically signiﬁcant) and less satisfaction. Borrowers with

a mortgage that is 100bp more expensive are 13 percentage points less likely to report be-

ing very satisﬁed with their mortgage rate and about 2 percentage points less likely to be

very satisﬁed with their lender, the application process, and the closing process. The other

coeﬃcients, though smaller in magnitude, also have a negative sign. Overall, rather than

getting better service quality in return for a more expensive mortgage, these results point in

the direction of borrowers getting worse service despite paying more.

6.4 Sophistication, Concentration, and Mortgage Rates

If consumers lack knowledge about mortgages or do not shop around eﬀectively, this could

generate some degree of market power for lenders, even in low-concentration markets where

many lenders oﬀer mortgages. Furthermore, sophisticated borrowers may be better able to

take advantage of more intense competition (lower market concentration). In this section,

we examine these hypotheses based on the NSMO data.

We start by studying the relationship between a borrower’s mortgage contract rate and

their shopping and knowledge. We estimate OLS regressions of the form

Rate

ijtw

= βX

+ ΓZ

+ α

+ δ

+ 

ijtw

, (3)

where Rate

ijtw

is the contract rate on the mortgage for borrower i with loan characteristics

j, loan origination month t, and responding to survey wave w. X

are diﬀerent measures

of borrower i’s shopping eﬀort or knowledge about the mortgage market based on several

questions in the NSMO. Z

is a rich set of borrower and mortgage characteristics that could

inﬂuence the pricing of the loan and is similar to the set of controls described above in

equation (2) (see the note to Table 10 for the full list of controls; recall that these are mostly

drawn from loan-level administrative data, i.e., not self-reported).

The ﬁrst column of Table 10 shows how mortgage rates obtained by borrowers correlate

with 8 shopping and knowledge variables simultaneously. We think of the ﬁrst four items

as capturing shopping eﬀort, while the other four capture knowledge about the mortgage

market and their own mortgage. All are individually signiﬁcant, suggesting that there are

diﬀerent dimensions to shopping and knowledge that can contribute to a borrower obtaining

a low rate.

For instance, borrowers who are very familiar with market conditions may not

need to consider more than one lender, if they can negotiate a good rate purely based on

their knowledge.

Conversely, shopping alone does not guarantee a good rate if a borrower’s

knowledge is low (see also Malliaris et al., 2022).

In the second column, we regress Rate

ijctw

on a composite “sophistication index”measure,

which we construct as the sum of six of the shopping and knowledge dummy variables, and

then divide by six so that it ranges from 0 to 1.

The coeﬃcient on the sophistication

index of -0.226 implies that the diﬀerence in rates between the most and least sophisticated

borrowers is nearly 23bp. This result suggests that variation across consumers in shopping

and knowledge is an important contributor to mortgage rate dispersion, especially considering

that our sophistication index is based on coarse responses to qualitative survey questions,

likely leading to individual-speciﬁc noise and attenuation of the resulting coeﬃcients.

In Appendix A.7, we provide a more complete description and summary statistics for each shopping and

knowledge variable, and regression results for each of these variables individually. The coeﬃcients are only

slightly larger relative to the simultaneous regression in Table 10. Borrowers may apply to 2+ lenders for

reasons other than “better terms” (e.g., because they got turned down at one lender), which we condition

on. In the appendix, we show that applying to 2+ lenders for other reasons is positively related to rates, in

line with the ﬁndings of Agarwal et al. (2023).

One may be concerned that borrowers who use a mortgage broker report only considering one lender

even though the broker may be shopping across many lenders on the borrower’s behalf. When we control

for whether a borrower used a mortgage broker, our results remain virtually unchanged, and using a broker

is associated with getting a slightly higher mortgage rate.

In an earlier version of this paper (available at https://doi.org/10.17016/FEDS.2020.062), we provide

a complementary analysis using data from the 2016 Survey of Consumer Finances. Consistent with the NSMO

results, we ﬁnd that borrowers who report shopping more, and borrowers with high ﬁnancial literacy—based

on their answers to the Lusardi-Mitchell ﬁnancial literacy questions—get signiﬁcantly lower interest rates,

even after controlling for loan characteristics, borrower credit risk, and borrower demographics.

We include the dummy “considered 3+ lenders” but not the dummy for considering exactly 2 lenders; we

also exclude the dummy for whether “most lenders oﬀer the same rate” since this question was not asked in

early waves. The median borrower has a sum of 3 (i.e., an index value of 0.5); about 14 percent of borrowers

have a sum of 1 or 0, and about 8 percent of borrowers have a sum of 5 or 6.

For instance, respondents likely diﬀer in what they view as using an information source “a lot” vs. “a

little,” or being “very” vs. “somewhat” familiar with a topic. In Appendix A.7.1, we provide evidence that

mortgage knowledge and shopping is higher for jumbo borrowers relative to FHA borrowers, and high-FICO

borrowers relative to low-FICO borrowers, consistent with our earlier results that overpayment is correlated

In column (3) of Table 10, we interact the sophistication index with market concentration,

measured as the county-level HHI of lender concentration in the year before origination and

standardized to have mean zero and standard deviation of one. Bear in mind that most

of the US population live in counties with a low HHI, limiting the variation and range

of HHI over which we can estimate the role of concentration. Nonetheless, we estimate a

statistically signiﬁcant and positive coeﬃcient on the interaction term HHI×Sophistication.

This means that sophisticated borrowers tend to pay lower rates in less concentrated markets

(HHI low) than in more concentrated ones (HHI high), while for less sophisticated borrowers,

the diﬀerences across market types are smaller. Thus, this ﬁnding is consistent with the idea

that a reduction in market concentration may be less eﬀective in promoting competition and

limiting markups when borrowers are less sophisticated.

Further support for this story is provided in Appendix A.8. We use the Optimal Blue

locks data to assess whether the propensity of borrowers to get “good deals” (a rate residual

less than -20bp) or “bad deals” (a rate residual larger than +20bp) varies with county HHI.

We ﬁnd that with higher market concentration, the likelihood of getting a good deal is

signiﬁcantly lower (while the average rate is not signiﬁcantly related with HHI).

7 Conclusion and Policy Implications

Our empirical results provide evidence that many borrowers from the most vulnerable part

of the borrower population in the US seem to overpay for mortgages: those that are most

likely to be relatively low income, low net worth, and more likely to be ﬁrst-time homebuyers.

These are the exact borrowers that various government programs attempt to subsidize. If

they were to obtain mortgages from the lower end of the oﬀer distribution, this would make

their mortgage payments more aﬀordable and leave them with more disposable income.

Given our ﬁndings, future research might focus on the design of policies that would

help borrowers search and negotiate more eﬀectively. Alternatively, future research could

study whether the problem can be alleviated if the guaranteeing agencies were to impose

requirements on the maximum locked-oﬀer rate spreads they allow for loans to be securitized.

Of course, to understand the eﬀectiveness of such policies one would need to consider general

equilibrium eﬀects on the oﬀers that lenders make (as in Agarwal et al. 2023, Alexandrov

and Koulayev 2017, or Guiso et al. 2022).

with loan program and credit score.

The negative relationship between overpayment and the level of market rates that we

document in Section 4.2 also matters for monetary policy transmission. Our ﬁndings imply

that as rates fall (e.g., in response to central bank actions), borrowers tend to do worse

relative to the rates available in the market, likely at least in part due to less shopping

or negotiation. It follows that the contract rates they end up with do not fall as much as

they could, based on lenders’ oﬀers, adding another friction to the pass-through of expansive

monetary policy to the mortgage market.

On the other hand, the pass-through of increases

in policy rates to rates on new mortgages may be dampened by more intense borrower

shopping. This could be good or bad news for monetary policymakers, depending on whether

slowing the housing market through higher mortgage rates is seen as desirable in a given

situation or not.

See Amromin et al. (2020) for a review of related work.

References

Agarwal, Sumit, Itzhak Ben-David, and Vincent Yao, “Systematic mistakes in the mortgage

market and lack of ﬁnancial sophistication,” Journal of Financial Economics, 2017, 123 (1), 42

– 58.

, John Grigsby, Ali Horta¸csu, Gregor Matvos, Amit Seru, and Vincent Yao, “Searching

for Approval,” Econometrica, 2023, forthcoming.

, Richard J. Rosen, and Vincent Yao, “Why Do Borrowers Make Mortgage Reﬁnancing

Mistakes?,” Management Science, 2015, 62 (12), 3494–3509.

Alexandrov, Alexei and Sergei Koulayev, “No Shopping in the U.S. Mortgage Market: Direct

and Strategic Eﬀects of Providing Information,” Working Paper, CFPB 2017.

Allen, Jason, Robert Clark, and Jean-Fran¸cois Houde, “The Eﬀect of Mergers in Search

Markets: Evidence from the Canadian Mortgage Industry,” American Economic Review, October

2014a, 104 (10), 3365–96.

, , and Jean-Francois Houde, “Price Dispersion in Mortgage Markets,” The Journal of

Industrial Economics, 2014b, 62 (3), 377–416.

Ambokar, Sumedh and Kian Samaee, “Mortgage Search Heterogeneity, Statistical Discrimi-

nation and Monetary Policy Transmission to Consumption,” Working Paper, University of Penn-

sylvania 2019.

Amel, Dean, Elliot Anenberg, and Rebecca Jorgensen, “On the geographic scope of retail

mortgage markets,” FEDS Notes, 2018.

Amromin, Gene, Neil Bhutta, and Benjamin J. Keys, “Reﬁnancing, Monetary Policy, and

the Credit Cycle,” Annual Review of Financial Economics, 2020, 12 (1), 67–93.

Andersen, Steﬀen, John Y. Campbell, Kasper Meisner Nielsen, and Tarun Ramado-

rai, “Sources of Inaction in Household Finance: Evidence from the Danish Mortgage Market,”

American Economic Review, 2020, 110 (10), 3184–3230.

Argyle, Bronson, Taylor Nadauld, and Christopher Palmer, “Real Eﬀects of Search Fric-

tions in Consumer Credit Markets,” Review of Financial Studies, 2022, 36 (7), 2685–2720.

Armstrong, Mark and John Vickers, “Patterns of Competitive Interaction,” Econometrica,

2022, 90 (1), 153–191.

Baye, Michael R., John Morgan, and Patrick Scholten, “Information, Search, and Price

Dispersion,” in T. Hendershott, ed., Economics and Information Systems, Elsevier, 2006, pp. 323–

375.

Bhutta, Neil and Aurel Hizmo, “Do Minorities Pay More for Mortgages?,” Review of Financial

Studies, 2021, 34 (2), 763–789.

and Daniel Ringo, “The eﬀect of interest rates on home buying: Evidence from a shock to

mortgage insurance premiums,” Journal of Monetary Economics, 2021, 118, 195–211.

, Steven Laufer, and Daniel Ringo, “Residential Mortgage Lending in 2016: Evidence from

the Home Mortgage Disclosure Act Data,” Federal Reserve Bulletin, 2017, 103 (6).

Buchak, Greg and Adam Jørring, “Do Mortgage Lenders Compete Locally? Implications for

Credit Access,” Working Paper, Stanford University and Boston College 2021.

, Gregor Matvos, Tomasz Piskorski, and Amit Seru, “Fintech, Regulatory Arbitrage, and

the Rise of Shadow Banks,” Journal of Financial Economics, 2018, 130, 453–483.

Carlson, John A. and R. Preston McAfee, “Discrete Equilibrium Price Dispersion,” Journal

of Political Economy, 1983, 91 (3), 480–493.

CFPB, “Data Point: 2019 Mortgage Market Activity and Trends,” Technical Report, Consumer

Financial Protection Bureau 2020.

Choi, James J., David Laibson, and Brigitte C. Madrian, “Why Does the Law of One

Price Fail? An Experiment on Index Mutual Funds,” Review of Financial Studies, 2010, 23 (4),

1405–1432.

Coen, Jamie, Anil Kashyap, and May Rostom, “Price Discrimination and Mortgage Choice,”

Working Paper 31652, National Bureau of Economic Research 2023.

Damen, Sven and Erik Buyst, “Mortgage Shoppers: How Much Do They Save?,” Real Estate

Economics, 2017, 45 (4), 898–929.

d’Avernas, Adrien, Andrea L. Eisfeldt, Can Huang, Richard Stanton, and Nancy Wal-

lace, “The Deposit Business at Large vs. Small Banks,” Working Paper 31865, National Bureau

of Economic Research 2023.

Drechsler, Itamar, Alexi Savov, and Philipp Schnabl, “The Deposits Channel of Monetary

Policy,” Quarterly Journal of Economics, 2017, 132 (4), 1819–1876.

Fuster, Andreas, Laurie Goodman, David Lucca, Laurel Madar, Linsey Molloy, and

Paul Willen, “The Rising Gap between Primary and Secondary Mortgage Rates,” Federal Re-

serve Bank of New York Economic Policy Review, 2013, 19 (2), 17–39.

, Stephanie H. Lo, and Paul S. Willen,“The Time-Varying Price of Financial Intermediation

in the Mortgage Market,” Journal of Finance, 2023, forthcoming.

Gabaix, Xavier, David Laibson, Deyuan Li, Hongyi Li, Sidney Resnick, and Casper G.

de Vries, “The impact of competition on prices with numerous ﬁrms,” Journal of Economic

Theory, 2016, 165, 1–24.

Gomes, Francisco, Michael Haliassos, and Tarun Ramadorai,“Household Finance,” Journal

of Economic Literature, 2021, 59 (3), 919–1000.

Goodman, Laurie S., “Quantifying the tightness of mortgage credit and assessing policy actions,”

BCJL & Soc. Just., 2017, 37, 235.

Guiso, Luigi, Andrea Pozzi, Anton Tsoy, Leonardo Gambacorta, and Paolo Emilio

Mistrulli, “The cost of steering in ﬁnancial markets: Evidence from the mortgage market,”

Journal of Financial Economics, 2022, 143 (3), 1209–1226.

Gurun, Umit G., Gregor Matvos, and Amit Seru, “Advertising Expensive Mortgages,”

Journal of Finance, 2016, 71 (5), 2371–2416.

Hastings, Justine S., Brigitte C. Madrian, and William L. Skimmyhorn, “Financial

Literacy, Financial Education, and Economic Outcomes,” Annual Review of Economics, 2013, 5

(1), 347–373.

Horta¸csu, Ali and Chad Syverson, “Product Diﬀerentiation, Search Costs, and Competition

in the Mutual Fund Industry: A Case Study of S&P 500 Index Funds,” Quarterly Journal of

Economics, 2004, 119 (2), 403–456.

Hurst, Erik, Benjamin Keys, Amit Seru, and Joseph Vavra, “Regional Redistribution

through the US Mortgage Market,” American Economic Review, 2016, 106 (10), 2982–3028.

Iscenko, Zanna, “Choices of dominated mortgage products by UK consumers,” Occasional Pa-

per 33, Financial Conduct Authority 2018.

Jiang, Erica, Gregor Matvos, Tomasz Piskorski, and Amit Seru, “Banking without De-

posits: Evidence from Shadow Bank Call Reports,” Working Paper 26903, National Bureau of

Economic Research 2020.

Keys, Benjamin, Devin Pope, and Jaren Pope, “Failure to Reﬁnance,” Journal of Financial

Economics, 2016, 122 (3), 482 – 499.

Liu, Lu, “Non-Salient Fees in the Mortgage Market,” Working Paper, Imperial College London

2019.

Malliaris, Steven, Daniel A. Rettl, and Ruchi Singh, “Is competition a cure for confusion?

Evidence from the residential mortgage market,” Real Estate Economics, 2022, 50 (1), 206–246.

McManus, Doug, Liyi Liu, and Mingzhe Yi, “Why Are Consumers Leaving Money On The

Table?,” Freddie Mac Economic & Housing Research Insight, http: // www. freddiemac. com/

research/ insight/ 20180417_ consumers_ leaving_ money. page , 2018.

Scharfstein, David and Adi Sunderam, “Market Power in Mortgage Lending and the Trans-

mission of Monetary Policy,” Working Paper, Harvard University 2016.

Stango, Victor and Jonathan Zinman, “Borrowing High versus Borrowing Higher: Price Dis-

persion and Shopping Behavior in the U.S. Credit Card Market,” Review of Financial Studies,

2016, 29 (4), 979–1006.

Syverson, Chad, “Macroeconomics and Market Power: Context, Implications, and Open Ques-

tions,” Journal of Economic Perspectives, 2019, 33 (3), 23–43.

Woodward, Susan E. and Robert E. Hall, “Diagnosing Consumer Confusion and Sub-optimal

Shopping Eﬀort: Theory and Mortgage-Market Evidence,” American Economic Review, 2012,

102 (7), 3249–76.

Wright, Randall, Philipp Kircher, Benoˆıt Julien, and Veronica Guerrieri, “Directed

Search and Competitive Search Equilibrium: A Guided Tour,” Journal of Economic Literature,

March 2021, 59 (1), 90–148.

Table 1: Summary Statistics of the Rate Lock Data

Conforming Super-Conforming Jumbo FHA

Mean St. Dev. Mean St. Dev. Mean St. Dev. Mean St. Dev.

Loan Amount (000) 255 94 544 71 720 262 222 92

Interest Rate 4.33 0.51 4.31 0.47 4.21 0.50 4.30 0.61

Discount Points Paid 0.15 0.95 0.28 0.97 0.19 0.74 0.06 1.14

FICO 742 47 750 41 763 33 669 47

LTV 81 14 80 12 77 10 93 8

DTI 35 9 36 9 31 9 42 10

First-time Homebuyer % 24 23 11 49

Reﬁnance Share % 31 33 33 17

N. observations 2,316,400 119,894 76,941 1,092,535

Data Source: Optimal Blue

Notes: Sample includes 30-year ﬁxed-rate mortgages on owner-occupied single-unit properties, with full documentation of assets and

income. Self-employed borrowers, loans for amounts under $100,000, VA loans, and streamline reﬁnances are excluded.

Table 2: Summary Statistics of the Expected Gain from Search and Locked-Oﬀer Rate Gap

Expected Gain from Search Locked-Oﬀer Rate Gap

Observations Mean 25

pct. 75

pct. Mean 25

pct. 75

pct.

All Mortgages 67,637 0.18 0.04 0.25 0.10 -0.08 0.25

Program

FHA 14,715 0.28 0.08 0.40 0.25 0.01 0.46

Conforming 46,535 0.17 0.04 0.22 0.09 -0.06 0.22

Super-Conforming 4,448 0.10 0.01 0.12 -0.05 -0.21 0.08

Jumbo 1,939 0.04 0.00 0.03 -0.23 -0.35 -0.08

FICO



640, 660



7,629 0.27 0.06 0.40 0.23 -0.02 0.46



680, 700



9,617 0.22 0.05 0.31 0.16 -0.06 0.35



720, 740



10,666 0.19 0.04 0.26 0.11 -0.06 0.26

740+ 39,725 0.15 0.04 0.21 0.07 -0.09 0.20

LTV



75, 80



22,270 0.13 0.03 0.17 0.02 -0.12 0.16



85, 90



7,295 0.15 0.04 0.21 0.05 -0.09 0.20



90, 95



16,573 0.16 0.04 0.22 0.08 -0.08 0.21



95, 97



21,499 0.27 0.08 0.38 0.23 0.01 0.42

First-Time Homebuyer

No 33,378 0.15 0.03 0.20 0.06 -0.09 0.20

Yes 34,253 0.21 0.05 0.30 0.15 -0.05 0.31

Discount Points



-5, -0.2



14,133 0.14 0.02 0.19 0.00 -0.17 0.19



-0.2, 0.2



23,430 0.16 0.03 0.21 0.08 -0.10 0.21



0.2, 5



30,074 0.22 0.06 0.30 0.18 -0.01 0.31

Lender Type

Independent Non-bank 56,483 0.19 0.05 0.26 0.12 -0.05 0.27

Other/Unclassiﬁed 11,154 0.13 0.02 0.17 0.00 -0.15 0.16

Data Source: Optimal Blue

Notes: The expected gain from an additional search, given by equation (1). The locked-oﬀer rate gap is the diﬀerence

between each locked rate and the median oﬀer rate in the same market on the same day for an identical mortgage. In the

discount points category, negative values mean that the borrower receives points (also known as a rebate or credit), while

positive values mean that the borrower pays points.

Table 3: Regressions of the Expected Gain from Search on Observables

(1) (2) (3) (4) (5) (6)

FICO (omitted cat.: [640,660))

680≤F ICO<700

-0.039

∗∗∗

-0.029

∗∗∗

-0.024

∗∗∗

(0.006) (0.005) (0.008)

720≤F ICO<740

-0.069

∗∗∗

-0.047

∗∗∗

-0.045

∗∗∗

(0.008) (0.006) (0.009)

F ICO≥740

-0.097

∗∗∗

-0.067

∗∗∗

-0.056

∗∗∗

(0.008) (0.006) (0.008)

LTV (omitted cat.: (60,80])

85<LT V ≤90

0.015

∗∗∗

0.011

∗∗∗

0.011

∗∗∗

(0.003) (0.003) (0.004)

90<LT V ≤95

0.033

∗∗∗

0.025

∗∗∗

0.021

∗∗∗

(0.004) (0.004) (0.004)

LT V >95

0.128

∗∗∗

0.102

∗∗∗

0.074

∗∗∗

(0.009) (0.008) (0.011)

Loan Oﬃcer Comp (%) 0.107

∗∗∗

0.096

∗∗∗

(0.025) (0.027)

Loan amount F.E. ($10k bins) Yes Yes Yes Yes Yes Yes

MSA x Month F.E. Yes Yes Yes Yes Yes Yes

Lender-Branch F.E. Yes Yes Yes Yes

Adj. R-Squared 0.122 0.403 0.403 0.160 0.429 0.417

Observations 67637 65757 15444 67637 65757 15444

Data Source: Optimal Blue

Notes: The dependent variable is the expected gain from an additional search, given by equation (1). The data

cover mortgage rates for 20 metropolitan areas during the period 2016-2019. We focus on 30-year, ﬁxed-rate,

fully-documented purchase mortgages. Standard errors shown in parentheses are two-way clustered at the month

and lender level. Signiﬁcance: * p<0.1, ** p<0.05, *** p<0.01.

Table 4: The Relationship Between the Expected Gain from Search and Treasury Yields

(1) (2) (3) (4) (5) (6)

Treasury Yield -0.051

∗∗∗

-0.044

∗∗

-0.026

∗∗

-0.057

∗∗∗

-0.051

∗∗∗

-0.031

∗∗

(0.007) (0.018) (0.012) (0.007) (0.018) (0.012)

Treasury Yield x DTI ≤ 36 0.013

∗∗

0.015

∗∗∗

0.012

∗∗∗

(0.005) (0.005) (0.004)

Borrower and Loan Controls Yes Yes Yes Yes Yes Yes

MSA F.E Yes Yes Yes Yes Yes Yes

MSA x Month F.E. Yes Yes Yes Yes

Lender-Branch F.E. Yes Yes

Adj. R-Squared 0.145 0.163 0.430 0.146 0.165 0.430

Observations 67241 67241 65349 67241 67241 65349

Data Source: Optimal Blue

Notes: The dependent variable is the expected gain from an additional search, given by equation (1). Treasury

Yield is the daily 10-year yield on the day of the mortgage rate lock. All speciﬁcations include controls for

FICO, LTV, and loan amount, as well as for MSA house price growth (from Zillow) over the past 12 months.

The data cover mortgage rates for 20 metropolitan areas during the period 2016-2019. We focus on 30-year,

ﬁxed-rate, fully-documented purchase mortgages. Standard errors shown in parentheses are two-way clustered

at the month and lender level. Signiﬁcance: * p<0.1, ** p<0.05, *** p<0.01.

Table 5: Unpacking the Dispersion in Locked Interest Rates

Underwriting Grid Add Lender Controls Add Branch Controls Add LO Controls

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Standard Deviation 0.33 0.26 0.24 0.22 0.21 0.18 0.16 0.15 0.15 0.13

75-25 Percentile 0.36 0.28 0.26 0.22 0.21 0.17 0.15 0.14 0.13 0.12

90-10 Percentile 0.79 0.58 0.55 0.48 0.44 0.38 0.33 0.31 0.30 0.26

Underwriting Variables Grid

Lock Date x MSA F.E. Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

FICO x LTV x Program x Lock Month F.E. Yes Yes Yes Yes Yes Yes Yes Yes Yes

ZIP Code F.E. Yes Yes Yes Yes Yes Yes Yes Yes Yes

Discount Points x Program x Lock Month F.E. Yes Yes Yes Yes Yes Yes Yes Yes

Add Lender Controls

Lender F.E. Yes Yes Yes Yes Yes Yes Yes

Lender x Lock Date F.E. Yes Yes Yes Yes Yes Yes

Lender x FICO x LTV x Program x Lock Month F.E. Yes Yes Yes Yes Yes

Lender x Points x Lock Month F.E. Yes Yes Yes Yes Yes

Add Branch Controls

Branch F.E. Yes Yes Yes Yes

Branch x Lock Month F.E. Yes Yes Yes

Add Loan Oﬃcer Controls

Loan Oﬃcer F.E. Yes Yes

Loan Oﬃcer x Program F.E. Yes

Loan Oﬃcer x Lock Year F.E. Yes

Adj. R-Squared 0.58 0.75 0.77 0.81 0.82 0.85 0.88 0.88 0.89 0.90

Adj. R-Squared (Within Date x MSA) 0.36 0.42 0.52 0.55 0.62 0.69 0.70 0.72 0.74

Observations 2996149 2996149 2996149 2996149 2996149 2996149 2996149 2996149 2996149 2996149

Data Source: Optimal Blue

Notes: The dependent variable is the mortgage interest rate locked. The data cover mortgage rates locked for 277 metropolitan areas during the period 2015-2019. We focus on 30-year,

ﬁxed-rate, fully-documented mortgages. “Program” refers to 12 dummy variables representing four loan programs interacted with three loan purposes. Speciﬁcations (2)-(10) also include lock

period f.e., property type f.e., cubic functions of loan amount and DTI, as well as linear functions of FICO, LTV, and (from speciﬁcation (3) onward) discount points. For MSAs that span

across multiple states, we include MSA x State ﬁxed eﬀects.

Table 6: Summary Statistics of the Residualized Locked Rate

Observations

− 10

Percentile Gap

Spec. (3) of Table 5 Spec. (10) of Table 5

All Mortgages 2,996,149 0.55 0.26

Program

FHA 884,681 0.72 0.31

Conforming 2,001,083 0.48 0.25

Super-Conforming 72,419 0.43 0.21

Jumbo 37,966 0.48 0.24

FICO

< 600 42,369 0.93 0.47



600, 640



221,431 0.82 0.37



640, 680



490,343 0.68 0.30



680, 740



933,110 0.55 0.27

≥ 740 1,308,896 0.44 0.23

LTV

≤ 75 506,400 0.47 0.23



75, 80



624,315 0.46 0.23



80, 95



968,492 0.50 0.26

>95 846,179 0.71 0.33

First-Time Homebuyer

No 1,991,677 0.50 0.24

Yes 1,004,130 0.64 0.31

Loan Purpose

Purchase 2,294,776 0.55 0.27

Cashout 351,269 0.50 0.23

Rate Reﬁ 350,104 0.57 0.25

Lender Type

Independent Non-bank 1,897,661 0.55 0.27

Other/Unclassiﬁed 1,098,488 0.54 0.26

Data Source: Optimal Blue

Notes: This table summarizes the residualized locked mortgage rate from speciﬁcations (3) and (10) of Table 5.

Table 7: Which Lenders Are the Most Expensive?

(1) (2) (3) (4) (5)

Dep. Var.: Lender FE Lender FE Lender FE Lender FE Within-lender

dispersion

Size Quartile 2 0.0533** 0.0488* 0.0253 0.0216***

(0.0258) (0.0256) (0.0238) (0.00672)

Size Quartile 3 0.0740*** 0.0812*** 0.0282 0.0487***

(0.0256) (0.0243) (0.0234) (0.00577)

Size Quartile 4 0.127*** 0.131*** 0.0465* 0.0777***

(0.0244) (0.0235) (0.0257) (0.00573)

Nonbank 0.108*** 0.0622*** 0.0579*** 0.00395

(0.0220) (0.0215) (0.0202) (0.00499)

Lender type unknown 0.00916 0.0151 -0.0147 0.0273***

(0.0215) (0.0210) (0.0208) (0.00550)

FHA share 0.385*** 0.267*** 0.108***

(0.0763) (0.0767) (0.0169)

Jumbo share -0.154 -0.253** 0.0910

(0.202) (0.105) (0.110)

Superconf. share 0.0405 0.0460 -0.00502

(0.214) (0.206) (0.0449)

Within-lender dispersion 1.637*** 1.091***

(0.177) (0.203)

Constant -0.205*** -0.296*** -0.317*** -0.355*** 0.0540***

(0.0234) (0.0266) (0.0251) (0.0267) (0.00687)

Adj. R2 0.16 0.28 0.22 0.34 0.37

Obs. 452 452 452 452 452

Data Source: Optimal Blue

Notes: Table displays lender-level regression results using data for 452 lenders that use the Optimal Blue

platform and have at least 100 mortgage locks. The outcome variable in columns (1)-(4) are the lender ﬁxed

eﬀects from speciﬁcation (4) in Table 5. Within-lender dispersion is measured as the standard deviation of

residuals from speciﬁcation (6) of Table 5. Robust standard errors in parentheses. * p<.1, ** p<0.05, ***

p<0.01.

Table 8: Nonbank Income and Expenses (from MCR Data) and Their Relationship with Estimated Lender Expensiveness

Regression on

Lender Expensiveness

Mean Median St. Deviation β St. Err.

Income

Origination Income 1.33 0.80 1.53 0.78*** (0.19)

Interest Income 0.28 0.23 0.32 0.03 (0.04)

Secondary Market Income (Gain-on-Sale) 3.35 3.52 1.76 3.37*** (0.43)

Other Income 0.35 0.02 2.11 0.05** (0.02)

Gross Income 5.03 4.75 1.97 4.05*** (0.41)

Expenses

Loan Production Oﬃcers (Sales Employees) 1.37 1.36 0.58 1.06** (0.47)

Loan Origination (Fulﬁllment/Non-Sales) 0.65 0.58 0.42 0.45*** (0.13)

Origination-Related Management and Directors 0.28 0.15 0.36 0.25*** (0.07)

Employee Beneﬁts 0.27 0.25 0.19 0.33*** (0.05)

Other Personnel Expenses 0.44 0.31 0.54 0.24* (0.13)

Interest Expenses 0.25 0.21 0.37 0.06 (0.04)

Occupancy and Equipment 0.25 0.23 0.15 0.30*** (0.03)

Technology 0.10 0.08 0.09 0.12*** (0.03)

Outsourcing, Professional, and Subservicing Fees 0.15 0.09 0.22 0.10*** (0.02)

Other non-interest expenses 0.65 0.51 0.64 0.26 (0.23)

Gross Expenses 4.33 4.16 1.76 3.50*** (0.41)

Corporate overheads

Total Corporate Expenses 0.35 0.11 1.22 0.28*** (0.10)

Net Income

Net Income (residential originations, pre-corporate allocations) 0.67 0.49 1.21 0.45** (0.18)

Net Income (all lines, after corporate allocations and taxes) 0.31 0.32 1.87 0.24** (0.12)

Data Source: Conference of State Bank Supervisors Mortgage Call Reports; Optimal Blue

Notes: This table is based on 1897 quarterly ﬁlings from 162 unique lenders between 2015:Q1 and 2019:Q4. All variables are shown as

dollars per $100 originated. Some categories for income and expenses are combined or not shown since they only have zeros for all ﬁrms. The

standard deviation reported in the third column is the cross-sectional standard deviation, computed after taking out year-quarter ﬁxed eﬀects.

The reported coeﬃcients in the last two columns are from separate median regressions of each line item on lender expensiveness (in percentage

points) as measured from lender ﬁxed eﬀects (from speciﬁcation (4) of Table 5), and year-quarter ﬁxed eﬀects. The standard errors determining

statistical signiﬁcance are clustered at the lender level. * p<.1, ** p<0.05, *** p<0.01.

Table 9: Are Higher Mortgage Rates Associated with Better Service Quality?

β SE N Outcome mean

Did you experience delays in...

...Your closing date .008 (.008) 22,567 0.17

...Paperwork processing .009 (.009) 22,567 0.23

Were you very satisﬁed with...

...Your interest rate -.13*** (.01) 22,567 0.69

...Your lender -.023*** (.009) 22,567 0.77

...The application process -.019** (.01) 22,567 0.67

...The documentation process -.008 (.011) 16,188 0.62

...The loan closing process -.019* (.01) 22,567 0.68

...The timeliness of disclosures -.007 (.01) 22,567 0.68

Overall satisfaction -.035*** (.009) 16,188 0.69

Source: Authors’ calculations based on National Survey of Mortgage Originations and

the National Mortgage Database

Notes: Each row represents a separate regression with the dependent variables listed in

column (1). The displayed coeﬃcients are those on the borrower’s mortgage rate, in per-

centage points. All outcomes are dummy variables, except “Overall satisfaction,” which is

the sum of the previous six satisfaction questions divided by six (thus, its range is 0 to

1). Sample restricted to ﬁrst-lien 30-year FRM loans for single-family principal residence

properties, with no more than two borrowers, originated from 2013 through 2019 (the doc-

umentation question was not asked in every wave and thus has fewer observations). All

regressions control for origination month ﬁxed eﬀects, survey wave ﬁxed eﬀects, county ﬁxed

eﬀects, credit score (linear term plus dummies for 11 credit score bins), LTV (linear term

plus dummies for 6 LTV bins), indicators for loan purpose (purchase, reﬁnance, or cashout

reﬁnance), 9 loan amount categories, loan program (Freddie, Fannie, FHA, VA, FSA/RHS,

other), ﬁrst-time homebuyer status, single borrowers, log borrower income, self-employment

status of borrower(s), respondent gender, race, ethnicity, whether the household owns 4

diﬀerent types of ﬁnancial assets, whether the household could pay their bills for 3 months

without borrowing, metropolitan CRA low-to-moderate income tract status, self-assessed

creditworthiness, and the likelihood of moving, selling, or reﬁnancing within a couple years.

Observations weighted by NSMO sample weights. Robust standard errors in parentheses. *

p<.1, ** p<0.05, *** p<0.01.

Table 10: Sophistication, Lender Concentration, and Mortgage Rates

(1) (2) (3)

Considered 2 lenders -0.022**

(0.009)

Considered 3+ lenders -0.044***

(0.012)

Applied to 2+ lenders for better loan terms -0.075***

(0.019)

Used web/broker/friends to get info? A lot -0.021***

(0.008)

Familiar with mortgage rates? Very -0.042***

(0.015)

Most lenders oﬀer same rate? No -0.022**

(0.010)

Knows their interest rate -0.060***

(0.007)

Answered whether rate is ﬁxed or variable -0.056***

(0.019)

Sophistication Index -0.226*** -0.219***

(0.019) (0.019)

Sophistication Index × County HHI (last year) 0.050**

(0.020)

County HHI (last year) -0.025*

(0.014)

Adj. R2 0.53 0.53 0.53

Obs. 22567 22567 22563

Source: Authors’ calculations based on National Survey of Mortgage Originations and the National

Mortgage Database

Notes: Dependent variable is the borrower’s mortgage interest rate in percentage points. Sample

restricted to ﬁrst-lien 30-year FRM loans for single-family principal residence properties, with no

more than two borrowers, originated from 2013 through 2019 (waves 1-24; “most lenders oﬀer same

rate” was not asked in the ﬁrst 6 waves of the survey so we include a separate dummy for these

missing observations). All regressions control for origination month ﬁxed eﬀects, survey wave ﬁxed

eﬀects, county ﬁxed eﬀects, credit score (linear term plus dummies for 11 credit score bins), LTV

(linear term plus dummies for 6 LTV bins), indicators for loan purpose (purchase, reﬁnance, or

cashout reﬁnance), 9 loan amount categories, loan program (Freddie, Fannie, FHA, VA, FSA/RHS,

other), ﬁrst-time homebuyer status, single borrowers, log borrower income, self-employment status

of borrower(s), respondent gender, race and ethnicity, whether the household owns 4 diﬀerent types

of ﬁnancial assets, whether the household could pay their bills for 3 months without borrowing,

CRA low-to-moderate income tract status, self-assessed creditworthiness, applying to 2+ lenders for

reasons other than “better loan terms,” and the likelihood of moving, selling, or reﬁnancing within a

couple years. Observations weighted by NSMO sample weights. Sophistication index is a composite

of shopping/knowledge questions and ranges from zero to one (see text for details). HHI is market

concentration at the county-year level, as of the year prior to origination, standardized to have mean

0 and standard deviation 1. Robust standard errors in parentheses. * p<.1, ** p<0.05, *** p<0.01.

1.5

Density

-.7 -.6 -.5 -.4 -.3 -.2 -.1 0 .1 .2 .3 .4 .5 .6 .7

Offered Rate Minus Median Rate

Figure 1: Oﬀer Dispersion for Identical Mortgages

Data Source: Optimal Blue

Note: Figure shows the distribution of real-time oﬀered interest rates, in percentage points, where for each oﬀer rate we

subtract the median oﬀered rate across lenders for an identical mortgage in the same metropolitan area. The histogram

includes data between April 2016 and December 2019 from 20 metropolitan areas for 52 combinations of loan characteristics

(FICO, LTV, program, loan amount).

A. All Programs

.05

.15

Fraction

0 .2 .4 .6 .8 1 1.2

Expected Gain from Additional Search (Interest rate, in %)

B. Diﬀerences Across Programs

Cumulative Share

0 .2 .4 .6 .8 1 1.2

Expected Gain from Additional Search (Interest rate, in %)

Conforming

Super-Conforming

Jumbo

FHA

Figure 2: Distribution of the Expected Gain from Additional Search for Identical Mortgages

Data Source: Optimal Blue

Note: Panel A shows the distribution of the expected gain from additional search, deﬁned in equation (1),

for all borrowers in the sample. Being at (or close to) zero means that the rate locked by a borrower is near

the lowest available rate, so that the borrower would have little to gain from additional search. Larger

values mean that a borrower would have more to gain from additional search. The dashed line denotes the

mean of the distribution. Panel B shows the cumulative share of loans below a certain expected gain across

the four loan types in the sample.

10-Year Treasury Yield (Right)

Expected Gain from Search (Left)

1.5

2.5

3.5

Percent

.12

.14

.16

.18

.22

.24

.26

Percentage Points

2016m1 2017m1 2018m1 2019m1 2020m1

Figure 3: The Evolution of Average Expected Gain from Additional Search and Treasury

Yields

Data Source: Optimal Blue

Note: The dashed red line is the 10-year Treasury yield. The solid black line is the monthly average

expected gain from additional search (EGain) after controlling for borrower and loan characteristics

(FICO, LTV, loan amount, and MSA ﬁxed eﬀects).

Figure 4: Distribution of Estimated Lender Fixed Eﬀects from Interest Rate Regressions

Data Source: Optimal Blue

Note: Lender ﬁxed eﬀects, in percentage points, come from speciﬁcation (4) in Table 5. Only lenders with

at least 100 locked loans in Optimal Blue are included in the ﬁgure; size of circles is proportional to total

number of locks.

Internet Appendix for

“Paying Too Much? Borrower Sophistication and

Overpayment in the US Mortgage Market”

A.1 Relationship to the Literature

Table A-1 contains a summary of other papers that have studied price dispersion in the US mortgage

market.

These other papers look either only at oﬀers or only at originations. Among those that

look at originations, arguably the results in Agarwal et al. (2023) and Ambokar and Samaee (2019)

are most closely related to ours.

There are at least ﬁve major distinctions between our work and these papers. First, we have

data from all segments of the mortgage market (conforming, FHA, jumbo)—this is important as it

allows us to document diﬀerences in dispersion across segments. Second, our data come (only) from

the post-ﬁnancial-crisis period, when new regulations as well as the increased popularity of online

lending (and online mortgage shopping) could have been expected to reduce or even eliminate price

dispersion. We show that substantial dispersion remains.

Third, we are able to observe identiﬁers for branch and loan oﬃcer within lender. This allows

us to unpack at which “level” price dispersion occurs (across vs. within lenders, and then further

across branches and across loan oﬃcers within lender). We are thereby able to quantify remaining

dispersion for observably identical borrowers who get loans not only from the same lender at the

same time, but even in the same branch and from the same loan oﬃcer.

Fourth, we are able to observe discount points and the exact lock date, variables that are

both highly pricing relevant. Without these variables, it is not possible to know how much of the

estimated price dispersion is simply an artifact of missing controls. We can compare results for

the conforming segment, which these other papers also study. We ﬁnd a lower dispersion, which

may reﬂect either the diﬀerent time period or the additional controls we are able to include.

The

standard deviation of rate residuals in our conforming mortgage subsample is 19bp, after controlling

for lender ﬁxed eﬀects and 12bp after controlling for a full set of observables including (time-varying)

branch and loan oﬃcer ﬁxed eﬀects.

To illustrate to what extent our ﬁner controls matter for the amount of residual price dispersion,

There also exist comparable studies from non-US markets; cf. footnote 3 in the main text. Since

institutional arrangements diﬀer quite substantially across countries, we do not include the quantitative

ﬁndings of these papers in our discussion in this section.

Gurun et al. (2016) look at “reset rates” on privately securitized adjustable-rate mortgages prior to the

ﬁnancial crisis; Woodward and Hall (2012) study broker fees on a sample of FHA mortgages in 2001.

Note that Agarwal et al. (2023) could control more ﬁnely for loan characteristics and timing in their

data, but measuring price dispersion is not the main focus of their paper.

Papers, including ours, usually also report R-squared values from the regressions, but these are not easily

comparable across studies because they are to a large extent driven by the amount of time-series variation

in overall rate levels in the sample period considered.

we consider a version of our analysis in Table 5 where we do not directly control for points and

only control for an approximate closing month.

In this case, the standard deviation of the rate

residual from a speciﬁcation similar to (4) in Table 5 is 24bp, which is about 26% higher than the

19bp when we control for points and exact lock date. Similarly the 90th-10th percentile gap in the

rate residual is 53bp when we do not include the additional controls versus the 42bp when we do.

Fifth, and ﬁnally, we are able to match the prices (rates) that borrowers obtain with the

distribution of available rates at the same time for loans with identical characteristics. This allows

us to cleanly measure expected (unrealized) gains from additional search, and how those vary with

borrower and loan characteristics. Note that one could consider doing something similar based on

transacted rates only, by taking the best accepted rate by a borrower with certain characteristics as

a benchmark, and determine overpayment relative to that. However, this strategy would require a

large number of borrowers with certain characteristics getting a loan in a given location on a given

date. In our data, if we group loans by MSA, day, FICO bin, LTV bin, and loan program, there are

on average 2.4 loans per bin, and 90% of the bins have 5 or fewer loans in them. Clearly, using the

lowest rate obtained by one of these borrowers would not provide a robust benchmark. Moreover,

without matching to the oﬀer data, it would not be possible to detect if every borrower of a certain

type is overpaying relative to what is available in the market.

To be exact, we generate a closing date to be a random number between (lock date + lock period)/2

and (lock date + lock period), and then convert that date to a month.

Table A-1: Other Estimates of Price Dispersion in US Mortgage Market

Paper Data Main Findings Comments

Agarwal et

al. (2023)

Mortgages originated between 2001 and 2011

and insured by one of the GSEs. 85% of orig-

inations are from 2001-2009.

Residual standard deviation of 39bp (after

controlling for origination quarter, state, and

various loan and borrower characteristics).

90th-10th percentile gap in residual is 90bp.

Interest rates are adjusted for points and

fees. Adding lender-by-quarter ﬁxed eﬀects

appears to have little eﬀect on residual dis-

persion (see their Figure A1.C).

Alexandrov

and

Koulayev

(2017)

Rate sheet data (oﬀers data) for 31 lenders

from Informa. Calculate dispersion in oﬀered

rates for characteristics of 1.3M loans origi-

nated in 2014.

Focus is on the spread between highest and

lowest oﬀer across lenders. The average of

this spread is 50bp, though with variation

across loan types (10th percentile is around

33bp, 90th percentile around 69bp).

No data on transactions. Argue that wide

dispersion in oﬀers implies borrowers don’t

shop much, and that search costs, incorrect

beliefs about dispersion, and non-price lender

preferences matter (with some evidence from

NSMO survey data).

Ambokar

and Samaee

(2019)

Fannie Mae and Freddie Mac public loan-

level data on originations from 1999 to 2016.

Residual standard deviation of 27bp (after

controlling for origination month, FICO bins,

LTV bins, 3-digit zip code and other loan

characteristics available in these data, as well

as originator ﬁxed eﬀects).

Data do not contain information lock date or

fees/points. Only relatively large originators

are individually identiﬁed.

Gurun et al.

(2016)

Privately securitized adjustable-rate mort-

gages through 2006 (data source: LoanPer-

formance)

Find 95th-5th percentile gap in residual re-

set rate of 310bp; also construct “lender ex-

pensiveness”measure from residual; 95th-5th

gap there averages 280bp.

Reset rate typically applies 2+ years after

origination, if borrower does not reﬁnance

sooner. Main focus is on relationship be-

tween lender expensiveness and advertising.

McManus et

al. (2018)

PMMS survey data (oﬀer rates at lender level

for a ﬁrst-lien, prime, conventional, conform-

ing home purchase mortgages with a loan-to-

value of 80%) from 1995 through 2017.

For most periods, the standard deviation of

oﬀer rates ranges from 15 to 25 bp (mostly

below 20). During the global ﬁnancial crisis,

the rate dispersion peaked at about 45 bp.

On a speciﬁc date in 2018, show that range

of oﬀers is about 90bp.

No data on transactions. Calculate potential

gains from shopping simply based on calcu-

lating the minimum rate obtained from X

random searches (for X = 1 to 5).

Woodward

and Hall

(2012)

Sample of 1,500 FHA loans from six weeks in

2001; brokered mortgages only.

90th-10th percentile gap in fees paid to bro-

kers (upfront fee + yield spread premium) is

about 300bp.

300bp in upfront costs would correspond

to about 75bp in rates just from broker

costs; there might be additional rate varia-

tion across borrowers.

A.2 Comparing Oﬀer and Locked Interest Rates in Optimal Blue

to Other Data Sources

In this section, we assess whether the interest rates we observe in the Optimal Blue data align

with other data sources. To begin, we compare median oﬀer rates from Optimal Blue to oﬀer rates

from Mortgage News Daily (MND) for various 30-year ﬁxed-rate loan programs. MND uses several

sources of information to estimate typical oﬀer rates, including directly obtaining rate sheets from

the largest lenders. The three panels in Figure A-3 plot median oﬀer rates from Optimal Blue

against MND’s oﬀer rates for conforming, FHA, and jumbo mortgages, respectively. In the top two

panels, the Optimal Blue median oﬀer rates for conforming and FHA loans—which are the bulk of

our data—move almost in lockstep with the MND oﬀer rates. For jumbo loans, the Optimal Blue

median oﬀer rate exhibits a little more variation from trough to peak, but on average, the level is

quite similar. Overall, these results help establish that our median oﬀer rates from Optimal Blue

are representative of the overall market.

Next, we compare lock rates to interest rates on closed mortgages. A concern with the locks

data is that high and low lock rates may systemically be less likely to actually proceed all the way

to origination. For example, borrowers who lock in a high rate at one lender may continue to shop

around and ultimately ﬁnd a better rate.

The top panel of Table A-2 compares unconditional distributions of interest rates from the

Optimal Blue locks data with interest rate distributions from other administrative data sources on

closed mortgages, by loan type. If the Optimal Blue locks are representative of closed loans, then

the rate distributions across these datasets should be very similar.

The ﬁrst four columns compare distributions for FHA loans locked or closed in 2014-15. For

these years, we have access to administrative data from the Department of Housing and Urban

Development (HUD) on the universe of originated FHA loans, which serves as an ideal benchmark.

In addition, we compare the locks data to well-known and widely used Black Knight McDash

servicing data, which contains loans serviced by the largest mortgage servicers in the US. We

can see in the top left portion of Table A-2 that average and 90th percentile locked rates line up

identically to both the HUD and McDash data. This remains true whether we look at all FHA

locks in Optimal Blue (column 1) or only those that we were able to match to an originated loan in

the HUD data (column 2), based on the procedure of Bhutta and Hizmo (2021). The fact that the

distributions in columns (1) and (2) are almost identical (also for other characteristics) implies that

there is little evidence of “selection” in terms of which locks end up in originated loans.

Moreover,

the full HUD and McDash data are slightly lower at the 10th percentile, suggesting an even wider

distribution than in Optimal Blue. Table A-2 also indicates that the distribution of FICO scores

and LTVs in Optimal Blue almost mirrors the HUD data, whereas the McDash data are skewed

This remains true if we plot the distribution of rates in the Optimal Blue locks over time: The distribution

of all locks and the matched (i.e., originated) locks are almost always nearly identical.

slightly toward less risky borrowers.

The remaining columns compare Optimal Blue locks to McDash loans in 2016-18, separately

for FHA, conforming, and jumbo loans. The most notable diﬀerence is for jumbo loans, where

we observe higher interest rates in Optimal Blue by 30-40bp, although the amount of dispersion

is similar to McDash. In Figure A-4, we plot the average, 10th, and 90th percentile rates over

time from Optimal Blue locks and McDash. Rates move closely together across the distribution,

with McDash rates lagging locked rates a bit—as expected since mortgages typically do not get

originated until a few weeks after the rate mortgage rate is locked in. Again, while the levels of

rates are very similar across the two datasets for FHA and conforming mortgages, Optimal Blue

rates tend to be higher than McDash for jumbo loans, although the amount of dispersion is similar.

A.3 Price Dispersion in Mortgage Oﬀers

In this appendix, we provide additional detail on our analysis of price dispersion in oﬀer interest

rates across lenders, already brieﬂy discussed in Section 3.1.2 of the main text.

There are two things to consider when thinking about the “price” of a mortgage with certain

characteristics. First, lenders do not oﬀer a single mortgage rate to borrowers but rather a menu

with diﬀerent combinations of mortgage rates and discount points to choose from. Borrowers can

pay discount points, each equal to 1% of their mortgage balance, in order to lower their mortgage

interest rate. Alternatively, they can choose negative points, known as lender credits or rebates,

in return for a higher mortgage rate. In this case, borrowers receive cash from the lender, which

can be used toward closing costs. Either way, one point in upfront payments corresponds to about

20bp in mortgage rate (so a borrower could get, e.g., a 4% mortgage rate with no points, a 4.2%

rate but receive one point, or a 3.8% rate by paying one point).

Second, lenders also charge origination fees. While fees are not typically considered part of the

price of the mortgage, they are part of the total cost of securing the mortgage. We can think of

lender fees and discount points as interchangeable: From the borrower’s perspective, a lender that

charges an origination fee of 1% to originate a mortgage at 4% interest is equivalent to a lender

that charges no fees but requires the borrower to pay 1 discount point for a mortgage rate of 4%.

In the Optimal Blue Pricing Insights interface, we observe how lenders compare in terms of the

sum of points and fees that they charge for a given mortgage rate, on a given day in a given location,

and for certain borrower and loan characteristics. The interface allows users to specify the key

underwriting and loan characteristics, including location (MSA), FICO score, LTV, loan amount,

DTI, loan type and term (e.g., 30-year ﬁxed), loan purpose (e.g., cashout reﬁnance), program (e.g.,

FHA or conforming), as well as details about the property (e.g., whether it is a single-family home

or a condo) and whether it will be owner occupied or not. Furthermore, the user speciﬁes the

desired lock period (e.g., 30 days). One could furthermore specify a given mortgage rate for which

oﬀers should be compared (e.g., 4%), but by default the system instead shows the comparison of

points/fees for the mortgage rate at which the median lender that makes an oﬀer does so at (as

close as possible to) zero points and fees.

An example of the resulting output is shown in Figure A-5. Lenders are sorted based on the

“price” they oﬀer for a loan with the desired characteristics, where the price equals 100 minus the

points/fees the borrower would be charged. Thus, a price of 101 means the borrower would receive

1 point, while a price of 99 means the borrower would have to pay 1 point to get this loan. As can

be seen in the screenshot, the range of oﬀers in this example spans almost 4 points, which for a

typical loan of $250,000 would correspond to a diﬀerence between the cheapest and most expensive

lenders of $10,000.

As noted in the main text, we conduct searches for 100 diﬀerent combinations of FICO, LTV,

program, loan amount, loan purpose, occupancy, and rate type, across 20 MSAs (at diﬀerent

frequencies). For each of these searches, we then receive the underlying individual price oﬀers for

the mortgage rate the system chooses (as explained above).

For our main analysis, we then transform these prices into the rate each lender would oﬀer at

zero points and fees, by converting points into rates using a conversion factor that we estimate

based on the lock data. As explained in detail in Section A.5, we allow for this conversion factor

vary by loan program × lock month. The estimated conversion factor averages about 22bp in rate

per 1 point upfront, which is also in line with what is typically observed in lender rate sheets. So

for instance, a lender that is shown as oﬀering a price of 100.5 for a 4.25% mortgage rate is assigned

a rate of 4.14%.

A.3.1 Dispersion in Oﬀer Rates

We start by documenting the dispersion in mortgage rates available from diﬀerent lenders for

identical mortgages in Los Angeles, since we have daily searches for this MSA. The ﬁrst panel of

Figure A-6 shows the distribution of rates oﬀered by diﬀerent lenders for conforming mortgages

with an amount of $300k, FICO=750, LTV=80 and DTI=36. There are about 120 diﬀerent lenders

oﬀering this mortgage in Los Angeles on any given day. The histogram shows the daily oﬀer rates

after subtracting the median (for the same day) over the period of April 2016 to December 2019.

The rate diﬀerence between the cheapest and the most expensive lender is about 100bp. More-

over, even though much of the mass is in the middle of the distribution, the tails of the distribution

are rather fat. These patterns can also be seen in the other two panels of Figure A-6, which plot the

dispersion for a typical FHA mortgage and a jumbo mortgage. The exact shape of the distribution

diﬀers across these mortgage types, but the amount of dispersion is similar.

Figure 1 in the main text shows the dispersion in mortgage rates available from diﬀerent lenders

in all of the 20 metropolitan areas. Table A-3 shows more detailed summary statistics of the rate

dispersion in this pooled oﬀer data, broken down by mortgage types. There are typically over

100 unique lenders on any given day making oﬀers for each mortgage type in each location. The

median mortgage rate is higher for jumbo loans than for conforming loans reﬂecting in part the

fact that conforming loans are guaranteed by Fannie Mae or Freddie Mac in exchange for a low

guarantee fee, which is rolled into the mortgage rate. FHA mortgages have lower interest rates than

other products since borrowers also have to pay upfront (175bp) and ongoing mortgage insurance

premia (85bp) which are not part of the quoted mortgage rate. Generally, the price dispersion is

a bit higher for mortgages with low FICO scores, high LTVs, and FHA mortgages. Overall, there

is about a 50-55bp diﬀerence in mortgage rates between the 10

percentile lender and the 90

percentile lender, and a 90bp diﬀerence between the 1

and the 99

percentile lender.

Table A-4 compares the rate dispersion for a “plain vanilla” conforming mortgage with LTV of

80% and FICO of 750 across MSAs. We see that, while there are some diﬀerences in the exact

amount of dispersion across MSAs, the qualitative points from above generalize across all of the

cities, and Los Angeles is not an outlier.

A.3.2 Dispersion in Oﬀered Points and Fees

In this subsection we focus on the points and fees charged by lenders to originate a mortgage with

a median interest rate. The median interest rate for each mortgage type is deﬁned exactly as in

the previous subsection: It is the interest rate at which the median lender oﬀers a mortgage (with

given characteristics) at zero points or fees. Figure A-7 shows the distribution of points and fees

charged by diﬀerent lenders to originate this median interest rate mortgage, with discount points

and fees measured as a percent of the mortgage balance. This ﬁgure shows that the range of oﬀers

in the screenshot in Figure A-5 appears representative of the universe of oﬀer distributions.

Table A-5 summarizes this dispersion for diﬀerent mortgage types. The diﬀerences in the upfront

costs of a mortgage with an identical rate across lenders are very large. The diﬀerence between

the 90

percentile and 10

percentile lender is around 2.2% to 2.5% of the mortgage balance. For

a typical conforming loan of $250k, that amounts to roughly a $6,000 diﬀerence in upfront costs

between these lenders. Even going from the 75

percentile lender to the 25

percentile lender

would save about $3,000 for a typical borrower with a $250k loan.

A.4 Matching Oﬀers and Locks

As described in Section 3.1.2, we collect data on mortgage oﬀers for 20 MSAs (some daily, others

twice or once per week) and for diﬀerent loan programs (conforming, super-conforming, jumbo,

and FHA) and borrower/loan characteristics. In particular, we collect rates for FICO scores of

640, 680, 720, and 750, and LTV ratios of 70, 80, 90, 95, and 96%. When matching locks to these

oﬀers, we allow for some variation in the characteristics around the values that we collect rates for,

but do so in a conservative way. What this means is that (with two small exceptions noted below)

we match locks with FICO scores slightly above the FICO value from the rate oﬀer and with LTV

ratios slightly below the LTV value from the oﬀer, as follows:

• Oﬀer FICO 640: Lock FICO range 640-659

• Oﬀer FICO 680: Lock FICO range 680-699

• Oﬀer FICO 720: Lock FICO range 720-739

• Oﬀer FICO 750: Lock FICO range 740-850 (maximum FICO)

• Oﬀer LTV 70: Lock LTV range 60.01-70

• Oﬀer LTV 80: Lock LTV range 75.01-80

• Oﬀer LTV 90: Lock LTV range 85.01-90

• Oﬀer LTV 95: Lock LTV range 90.01-95

• Oﬀer LTV 96: Lock LTV range 95.01-97

In choosing these ranges, we follow Fannie Mae’s loan-level pricing adjustment (LLPA) grid (https:

//www.fanniemae.com/content/pricing/llpa-matrix.pdf). This grid is also why we decided to

assign FICO scores of 740-749 the FICO 750 oﬀer as well, and similarly for LTVs of 96.01-97 for the

LTV 96 oﬀer. (LTV values above 95 are uncommon for GSE loans but are very common for FHA

loans, where the modal LTV is 96.5.) We do not include some intermediate values (e.g., FICO

660-679, 700-719; LTV 80-85) since LLPAs can be diﬀerent and do not always change linearly;

however, matching less conservatively in that regard does not materially aﬀect the results. Simi-

larly, restricting more conservatively to matches where FICO and LTV are within 1 point between

datasets does not aﬀect the results shown in Table 2 but considerably reduces the sample size from

67,637 to 4,156.

In addition to matching on date, FICO, LTV, MSA and loan program, we also only retain

purchase mortgages with a 30-day lock period (since that is what the rate search is for); 30 days is

also the most common lock period in the data.

A.5 Estimating the Discount Point to Interest Rate Tradeoﬀ

In some parts of the paper, we need to know how much 1 discount point (or credit) buys down (or

up) the interest rate borrowers pay. For example, we use this tradeoﬀ to match each rate locked

to the zero point oﬀer rate, or for back-of-the envelope calculations of how much money borrowers

leave on the table. We estimate the relationship between discount points and interest rates in a

regression speciﬁcation identical to column (10) of Table 5, with the only exception that discount

points are allowed to only enter linearly and are only allowed to vary by loan program × lock

month. Appendix Figure A-8 shows in a binned scatter plot that the relation between points and

rate is indeed close to linear. The slope of the line in the chart is the average point-rate tradeoﬀ,

which implies that 1 discount point changes the interest rate by about 21.8bp on average.

In Table A-6, we show how the point-rate tradeoﬀ varies with FICO, DTI, and LTV by adding

several interactions of discount points to the same regression we run in the paper. As a baseline, the

average coeﬃcient on discount points over all of the data in the paper is 0.218, so 1 discount point

paid buys down the interest rate by 21.8bp. Relative to this magnitude, the coeﬃcients shown in

Table A-6 are small and economically insigniﬁcant, all of them being 0.1 to 0.8bp. It is safe to say

that the point-rate tradeoﬀ does not vary in a meaningful way with these loan characteristics.

Another possibility is that the point-rate tradeoﬀ may vary across lenders. In Table 5, we

do interact binned discount points with lender ﬁxed eﬀects. However, in the parts of the paper

where we consider whether borrowers are overpaying, we think it would be overcontrolling for us

to soak up the variation across lenders in the point-rate tradeoﬀ. The fact that some lenders might

oﬀer worse tradeoﬀs than others reﬂects the fact that those lenders are just more expensive to do

business with. In other words, we should not control for lender speciﬁc point-rate tradeoﬀs since

that is part of how expensive a lender is relative to others, which is what we are trying to capture

when comparing locked and oﬀered rates.

A.6 Variation in Pricing Across Branches Within Lender

This section complements Section 6.1 in the main text, where we study covariates of the estimated

lender ﬁxed eﬀects from the regressions in Table 5. Here, we instead focus on variation in average

ﬁxed eﬀects across branches within a given lender. The goal is to see whether the “expensiveness”

of branches within a lender covaries with characteristics of the branch and its lending, or with local

characteristics based on where the branch originates loans. This analysis is thus related to work

that studies the pricing of deposits across branches of the same bank (e.g., Drechsler et al., 2017;

d’Avernas et al., 2023).

We again restrict to lenders with at least 100 loans locked in the sample, as in Section 6.1.

Those 452 lenders have a total of 15,881 branches.

In our baseline speciﬁcation below, we restrict

the sample to branches that made at least 50 loans. We are interested in correlates of variation in

branch ﬁxed eﬀects (i.e., eﬀectively the average rate residual per branch) within lenders.

As in Table 7 in the main text, we start by considering the role of branch size and the share

of diﬀerent loan types originated by a branch. Column (1) of Table A-7 indicates that while size

does not appear to correlate with the branch’s expensiveness (as measured by its ﬁxed eﬀect),

branches that originate a larger share of FHA loans are more expensive. The standard deviation

(SD) of FHA shares across branches in the sample of column (1) is 0.2 (or 0.15 within lender) such

We do not know whether all of these are truly “branches” with a separate physical location; in some

cases, what is recorded as a “branch” in the data could also be small teams of loan oﬃcers. Conversely,

a few lenders, including larger ones, have only one or two branches recorded in the data, suggesting that

perhaps they do not use diﬀerent branch IDs in Optimal Blue (instead simply relying on LO IDs). Below,

we try diﬀerent sample restrictions to ensure that our results are not sensitive to removing what appear to

be “unusual” branches.

The estimated ﬁxed eﬀects we take for this exercise come from a speciﬁcation like in column (7) of

Table 5 of the main text, except that we replace the lock date × MSA ﬁxed eﬀects and the zip code ﬁxed

eﬀects by lock date × state ﬁxed eﬀects. This is because we want to use the local variation associated with

diﬀerences across counties in the regressions in this appendix section.

that the strength of the economic association is meaningful, with a one-SD move in FHA share

corresponding to a 5.5bp larger branch ﬁxed eﬀect (relative to a SD of ﬁxed eﬀects across branches

of 14bp).

In the second speciﬁcation, we relate the expensiveness of a branch with the characteristics of

the counties where it lends (weighted by how many loans a given branch made in each county).

We observe that branches that lend in counties with lower mortgage market concentration (HHI),

lower average income, lower shares of college-educated households, and higher minority shares tend

to be more expensive. In terms of the economic magnitude of the coeﬃcients in column (2), the

implied eﬀect per one-SD move in a variable is largest for the share of college-educated households

(where a one-SD increase is associated with a 3.3bp decrease in the branch ﬁxed eﬀect), while

the smallest eﬀect is for the HHI (where a one-SD increase is associated with a 0.5bp decrease in

the branch ﬁxed eﬀect). Thus, even though the direction of the estimated eﬀect for local market

concentration is surprising (as one might expect a positive relationship), the economic magnitude

of the association is very small. This remains the case across the other columns of the table, where

we control for FHA share and the demographic variables jointly. Not surprisingly, given that FHA

share and socioeconomic characteristics are correlated, the associated coeﬃcients are attenuated

somewhat but remain signiﬁcant. To study the robustness of these conclusions, in column (4),

we remove the minimum-loan-count restriction, while in column (5) we retain only lenders where

the largest branch makes no more than 20% of that lender’s total loans (so we remove dominant

branches). In either case, results remain qualitatively similar.

In sum, this evidence suggests that the pricing variation across branches within lender is mean-

ingfully related to the branch’s typical borrower clientele. Branches that make predominantly FHA

mortgages or are active in locations with a less educated population (as well as lower incomes and

higher minority share) tend to be more expensive. The latter ﬁnding is related to the ﬁnding of

Drechsler et al. (2017) that bank branches in counties with a lower share of college-educated adults

tend to increase deposit spreads by more (i.e., pay lower deposit rates to their customers) in re-

sponse to a Fed funds rate increase. As a caveat to our analysis, we are not able to disentangle

to what extent these within-lender diﬀerences are driven by variation across branches in their “ex-

ante” (strategic) price setting, versus variation arising during the negotiation process (e.g., with

borrowers in less economically well-oﬀ locations being less willing or able to negotiate for a better

deal).

A.7 Additional Details on NSMO Analysis

In Section 6.4 of the main text, we consider the following mortgage shopping and mortgage knowl-

edge variables in our analysis of how shopping and knowledge relates to the mortgage rates that

We have also considered speciﬁcations where instead we deﬁned for each branch the county where it

made the most loans as its “home county,” and used that county’s characteristics. Results are similar.

borrowers get:

1. The answer to the question, “How many diﬀerent lenders/mortgage brokers did you seriously

consider before choosing where to apply for this mortgage?” 48% of respondents (weighted)

answer 1; 36% 2; 13% 3; 2% 4; and 1%, 5 or more. We combine the last three groups into

“3+”.

2. The answer to, “How many diﬀerent lenders/mortgage brokers did you end up applying to?”

Here, nearly 75% answer 1; 20% 2; 4% 3; 0.9% 4; and 0.4%, 5 or more. We combine the last

four groups into “2+”.

3. Those who indicated that they applied to two or more lenders are asked which of four non-

exclusive reasons were driving the multiple applications. We create an indicator for those

who indicate that “searching for better loan terms” was a reason (about 82% of those that

apply to more than one lender, or 21% of the sample overall).

4. A series of questions are asked about possible sources of information on mortgages and

mortgage lenders that the borrower could have used. For each source, a respondent can say

they used it “a lot,”“a little,” or “not at all.” As a proxy for search eﬀort, we create a dummy

variable that equals one if the respondent says they used one of “Other lenders or brokers,”

“Websites that provide information on getting a mortgage,” or“Friends/relatives/co-workers”;

about 38% of respondents said they used at least one of these information sources “a lot”.

5. The answer to the question, “When you began the process of getting this mortgage, how

familiar were you (and any co-signers) with [t]he mortgage interest rates available at that

time?” About 55% respond “Very,” whereas 38% say “Somewhat,” and 8% say “Not at all”.

6. An indicator for whether a borrower agreed with the statement, “Most mortgage lenders

would oﬀer me roughly the same rates and fees.” This question was only added in Wave 7;

of those who were asked, almost 70% agree with the statement.

7. We create an indicator variable for whether a borrower knows their interest rate by compar-

ing the self-reported contract rate to the contract rate available from the linked loan-level

administrative data. We set this variable equal to one if the borrower’s self-reported rate is

within 5bp of the actual rate. Just over 49% know their rate by this standard.

8. An indicator for whether the borrower said, “Don’t know” in response to the question “Is this

an adjustable-rate mortgage?” (About 5% answered “Don’t know.”)

In Table A-8, we regress mortgage rate against each of these measures one at a time, and then

in a ﬁnal speciﬁcation jointly (the joint speciﬁcation is also shown in the main text in Table 10).

In all speciﬁcations, we control ﬁnely for other factors (e.g., credit risk, loan type) that likely

inﬂuence loan pricing. We see that most proxies for intense shopping and better mortgage market

knowledge are associated with lower mortgage rates. In the ﬁnal column, where we include all X

jointly, some of the coeﬃcients are slightly attenuated relative to the earlier columns, but all remain

individually signiﬁcant, suggesting that there are diﬀerent dimensions to shopping and knowledge

that can contribute to a borrower obtaining a low rate.

A.7.1 Who Pays More Because of a Lack of Shopping or Knowledge?

The previous subsection provides evidence that more intense mortgage shopping and more mortgage

knowledge is associated with lower contracted rates. We next ask which observable borrower and

loan characteristics are associated with stronger shopping intensity and mortgage knowledge.

In Table A-9, we test whether a composite “sophistication index” is correlated with loan type

(FHA, GSE, jumbo, etc.), credit score, and ﬁrst-time homebuyer status. The sophistication index

is the same measure we use in Section 6.4, in which we construct as the sum of six of the shopping

and knowledge dummy variables shown in Table A-8, divided by six so that the index ranges from

0 to 1.

Column (1) of Table A-9 tests for a correlation between sophistication and loan type, and

indicates that FHA borrowers tend to be the least sophisticated (the omitted category is jumbo

and other non-conforming conventional loans). In column (2), we ﬁnd a monotonic relationship

between credit score and sophistication. In column (3), we do not ﬁnd a correlation between

ﬁrst-time homebuyer status and sophistication. In column (4), we include loan type, credit score,

and ﬁrst-time status simultaneously, and continue to ﬁnd that FHA and low-score borrowers are

the least sophisticated. Finally, in column (5), we control for loan amount bins to account for

the possibility that borrowers getting smaller loans, which may correlated with FHA status and

lower credit scores, may be have less incentive to shop around. Of course, loan amount also likely

reﬂects income, education, etc., and therefore might be a proxy for sophistication. We ﬁnd that

the coeﬃcients on loan type and credit score are only somewhat smaller and remain statistically

signiﬁcant.

Overall, we believe the ﬁndings here lend support to the mechanism we postulated in our

earlier analysis using the rate locks and oﬀers data. Namely, at least some of the overpayment by

many borrowers and dispersion in observed locked rates is likely due to ineﬀective shopping and

negotiation, reﬂecting a lack of ﬁnancial sophistication and knowledge of the market.

It is interesting to note that the coeﬃcient on “applied to 2+ lenders” ﬂips sign if we simultaneously

control for having applied to 2+ lenders in search of better loan terms. This likely reﬂects that those who

applied to multiple lenders but not in search of better terms got turned down on their previous application

(or learned negative news in the process), in line with the ﬁndings of Agarwal et al. (2023).

We include the dummy “considered 3+ lenders” but not the dummy for considering exactly 2 lenders;

we also exclude the dummy for whether “most lenders oﬀer the same rate” since this question was not asked

in early waves. We include “applied to 2+ lenders for better loan terms” but not “applied to 2+ lenders”

since the latter is associated with higher rates conditional on the former. The median borrower has a sum

of 3 (i.e., an index value of 0.5); about 14% of borrowers have a sum of 1 or 0, about 8% of borrowers have

a sum of 5 or 6, and the standard deviation of the index is 0.2.

A.7.2 Time-series Variation in Shopping Intensity

In Section 4.2 in the main text, we documented that our measures of borrower overpayment in the

Optimal Blue data, namely the expected gain from additional search (EGain) and the locked-oﬀer

rate gap, decrease when market interest rates are higher, even for borrowers who do not appear

constrained. We speculated that this may be driven in part by an increase in shopping intensity

when interest rates are higher. The NSMO enables us to test this hypothesis directly. We estimate

linear probability models of the form:

Shopping

ijctw

= β · P MMS

+ ΓZ

+ ΘW

+ δ

+ 

ijctw

, (A1)

where Shopping

ijctw

is a binary measure of shopping intensity (discussed below) by borrower i

with loan characteristics j, located in county c, loan origination month t and responding to survey

wave w. P MMS

is our main variable of interest, the market mortgage rate on average during

the month of loan origination. Z

are borrower and mortgage controls, similar to the controls in

equation (3). W

includes county ﬁxed eﬀects, and controls for county house price growth in the

12 and 24 months prior to the survey to account for “hotness” of the local housing market, which

might inﬂuence mortgage shopping behavior. Finally, δ

are survey wave ﬁxed eﬀects.

As dependent variables, we use binary versions of the shopping variables that were associated

with lower contract interest rates in Table A-8: (i) whether a borrower seriously considered at least

two lenders; (ii) whether a borrower applied to at least two lenders in search of better terms; (iii)

whether a borrower used other lenders/brokers to get information “a little” or “a lot”; and (iv)

whether a borrower used websites that provide information on getting a mortgage “a little” or “a

lot.” For each of these variables, we report regressions with and without other covariates (except

for survey wave ﬁxed eﬀects).

Table A-10 reports the results of these regressions. Across the diﬀerent measures in the full

sample (panel A), a higher level of market mortgage rates is associated with more shopping eﬀort,

with statistical signiﬁcance for three of four outcomes. The coeﬃcients are largely unaﬀected by

the addition of controls, which alleviates concerns that the relationship is driven by variation in the

type of borrower who applies at diﬀerent points in time (and at diﬀerent levels of market rates).

Columns (1)-(2) imply that a 1 percentage point increase in market mortgage rates increases

the probability that a borrower considered more than one lender by 4-5 percentage points, relative

to a sample average of 52%.

Columns (3)-(4) indicate that the likelihood of applying to 2+

lenders for better terms rises by about 5-6 percentage points when rates rise by 100bp, relative to

the sample average of 21%. Finally, the coeﬃcients in the last four columns on the likelihood of

using other lenders and the web to gain information are positive but not quite as large. In panel

B, we restrict the sample to borrowers whose DTI ratio ends up below 36 percent, suggesting that

Over our sample period (2013-2019), the market mortgage rate as measured by PMMS varied from about

3.3% to nearly 5%.

they had additional room to make larger payments. The coeﬃcients are somewhat larger than in

panel A, and thus it does not appear that the positive relationship between market interest rates

and shopping is mainly driven by aﬀordability constraints.

A.8 Is Getting a “Good Deal” Related to Market Concentration?

In Section 6.4, we found evidence using NSMO data that higher lender concentration in a county

is associated with a relative increase in the rate that highly “sophisticated” borrowers get relative

to less sophisticated borrowers. In other words, market concentration seems to matter more for

borrowers who are more likely to shop around and have strong knowledge about the mortgage mar-

ket, whereas for less sophisticated borrowers, lower concentration may be less eﬀective in increasing

borrower surplus.

In this section, we study the relationship between concentration and Optimal Blue mortgage

lock rate residuals, and in particular whether the propensity of borrowers to get particularly “good”

or “bad” deals varies with local (county-level) market concentration, as measured by the Herﬁndahl-

Hirschman Index (HHI). We regress either a loan’s rate residual, its absolute value, or indicators for

good and bad deals (deﬁned as rate residuals of less than -20bp or more than +20bp, respectively),

on the county-level HHI from the previous calendar year, various other local controls (the log

number of loans originated in the county in the previous year, as well as median household income,

the share of college-educated individuals, and the minority share, all at the zip-code level), as well

as loan type-by-month ﬁxed eﬀects. In some speciﬁcations, we further add lender-by-month ﬁxed

eﬀects; this allows us to study eﬀects of market concentration on rates within a lender at a point

in time (relying on lenders that are active in multiple counties).

For the absolute rate residuals, as well as the good-deal and bad-deal dummies, we rely on

the residuals from column (3) of Table 5 (i.e., the regression speciﬁcation that ﬁnely controls for

loan characteristics but not lender ﬁxed eﬀects). For the rate residual itself, we use residuals

from a slightly diﬀerent regression speciﬁcation, since the one in column (3) of Table 5 uses zip-

code ﬁxed eﬀects that would absorb most county-level variation in average residuals. Therefore,

we generate residuals based on a speciﬁcation with only state-by-month ﬁxed eﬀects, so we are

eﬀectively studying within-state variation in residuals as a function of county-level concentration.

When interpreting the results below, keep in mind that the HHI in the mortgage market is generally

low: Across the loans in our sample, the median lagged county-level HHI is 0.020, with a 90th

percentile at 0.037 and a 10th percentile at 0.006.

Table A-11 shows that local concentration is not signiﬁcantly related to average residuals

These HHI numbers may appear low, but that is partly because counties are weighted by the number

of loans in our sample, and large counties tend to be less concentrated. If we keep only one observation per

county-year in our sample, the median HHI increases to 0.035, with 90th and 10th percentiles at 0.076 and

0.016.

(columns 1-2).

However, columns (3) and (4) indicate a strong relationship between local concen-

tration and the absolute value of the residual: In more concentrated markets (higher HHI), there

is less dispersion in rate residuals (absolute values are lower).

The remaining columns show that

what is particularly strongly aﬀected is the likelihood of getting good deals: In more concentrated

markets, borrowers are signiﬁcantly less likely to obtain a rate residual of -20 bp or lower. For

instance, the coeﬃcient in column (6) implies that moving from the 90th to the 10th percentile of

HHI changes the probability of a good deal by about 1.7 percentage points, or 10% of the mean.

Columns (7) and (8) show that higher concentration also reduces the probability of bad deals,

although this eﬀect is only signiﬁcant without lender ﬁxed eﬀects.

A potential rationalization of these ﬁndings is that when a market is less concentrated (which

could be seen as more competitive), lenders may try harder to earn high rents from unsophisticated

borrowers, because they know that they need to compete harder for sophisticated borrowers (e.g., it

becomes more likely that a borrower obtains an outside oﬀer). Therefore, it appears that sophisti-

cated borrowers may indeed beneﬁt from increased competition, while less sophisticated borrowers

do not beneﬁt or may even be better oﬀ in more concentrated markets.

This ﬁnding is in line with Hurst et al. (2016) and Scharfstein and Sunderam (2016), who also do not

ﬁnd a relationship between local concentration and mortgage rates. Buchak and Jørring (2021) similarly ﬁnd

no relationship between concentration and rates, but do ﬁnd that lenders charge higher fees (as reported in

HMDA since 2018) in more concentrated markets.

This ﬁnding is in line with Allen et al. (2014a), who ﬁnd that after concentration-increasing mergers in

the Canadian mortgage market, there is a decrease in price dispersion.

Table A-2: Comparing Mortgage Locks in Optimal Blue to Closed Mortgages

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

FHA Loans, 2014-15 FHA Loans, 2016-18 Conventional Conforming Conventional Jumbo

Loans, 2016-18 Loans, 2016-18

Optimal Blue HUD McDash Optimal Blue McDash Optimal Blue McDash Optimal Blue McDash

All Matched

Interest Rate

10th 3.75 3.75 3.625 3.625 3.625 3.5 3.75 3.625 3.75 3.375

mean 4.14 4.14 4.11 4.09 4.40 4.26 4.44 4.29 4.33 3.95

90th 4.625 4.625 4.625 4.625 5.25 5.125 5.125 5 4.875 4.625

FICO Score

10th 628 629 630 641 620 629 681 686 719 726

mean 679.4 679.3 680.5 688.9 672.3 684.0 745.2 750.0 766.1 771.3

90th 744 742 745 754 738 751 800 802 801 803

LTV

10th 93.7 95 94.3 87.6 93.4 87.9 66.6 64.4 66.7 65.0

mean 95.3 95.5 95.8 93.7 95.4 93.7 83.6 82.0 77.6 82.7

90th 96.5 96.6 96.5 96.5 96.5 96.5 95.0 95.0 85.0 85.0

Loan Amount

10th 89,745 92,640 84,000 81,987 100,360 97,697 116,000 113,715 482,000 485,100

mean 187,624.3 186,804.2 180,450.1 173,106.5 204,065.3 203,275.5 255,892.9 255,738.2 729,963.4 850,403.6

90th 300,000 294,325 293,250 276,892 321,985 325,004 417,000 418,125 1,060,000 1,260,000

N 282,933 162,244 1,318,700 777,763 860,579 1,468,968 1,547,776 2,695,218 61,430 190,993

Data Source: Optimal Blue, HUD, Black Knight McDash

Note: All statistics are for 30-year, ﬁxed-rate, home-purchase mortgages for owner-occupied properties. Conventional conforming include super-

conforming loans that have loan amounts under the higher loan limits in high-cost geographies. “McDash” refers to Black Knight McDash data.

“Matched” in column (2) means Optimal Blue locks that matched to originated FHA loans in the HUD data.

Table A-3: Interest Rate Dispersion for Oﬀered Mortgage Products with No Points and Fees

Median Median Standard Percentile Diﬀerences

No. Oﬀers Rate Deviation 75

h − 25

− 10

− 1

All Oﬀers 118 4.67 0.20 0.27 0.53 0.90

Program

FHA 117 4.08 0.22 0.32 0.59 0.93

Conforming 122 4.54 0.19 0.27 0.51 0.88

Super-Conforming 144 4.68 0.20 0.27 0.52 0.88

Jumbo 106 5.06 0.20 0.26 0.53 0.92

FICO

640 107 5.23 0.21 0.29 0.54 0.92

680 118 4.64 0.20 0.28 0.53 0.90

720 122 4.48 0.20 0.27 0.52 0.90

750 122 4.44 0.20 0.27 0.52 0.90

LTV

70 122 4.67 0.20 0.27 0.52 0.90

80 117 4.78 0.20 0.28 0.53 0.91

90 105 4.78 0.20 0.27 0.52 0.91

95 128 4.63 0.20 0.27 0.51 0.88

96 119 4.27 0.21 0.30 0.55 0.91

Data Source: Optimal Blue

Notes: This table compares real-time interest rates for identical oﬀered mortgages (same FICO, LTV, DTI,

loan amount, location, time etc.) with no points and fees. Column 1 shows the median number of lenders

oﬀering each mortgage product in a location on a speciﬁc day. Columns 4-6 show the diﬀerence between various

percentiles of the oﬀer distribution.

Table A-4: Interest Rate Dispersion for Oﬀered Conforming Mortgages with No Points and Fees, Across MSAs

Median Median Standard Percentile Diﬀerences

No. Oﬀers Rate Deviation 75

h − 25

− 10

− 1

Atlanta, GA 112 4.68 0.20 0.28 0.54 0.92

Boston-Worcester-Lawrence, MA-NH-ME-CT 77 4.49 0.21 0.30 0.56 0.93

Charlotte-Gastonia-Rock Hill, NC-SC 93 4.67 0.21 0.28 0.55 0.93

Chicago-Gary-Kenosha, IL-IN-WI 103 4.57 0.20 0.28 0.53 0.90

Cleveland-Akron, OH 61 4.71 0.21 0.30 0.57 0.92

Dallas-Fort Worth, TX 136 4.67 0.21 0.29 0.55 0.93

Denver-Boulder-Greeley, CO 119 4.69 0.19 0.25 0.49 0.88

Detroit-Ann Arbor-Flint, MI 76 4.68 0.21 0.29 0.56 0.94

Las Vegas, NV 87 4.88 0.21 0.28 0.55 0.92

Los Angeles-Riverside-Orange County, CA 147 4.69 0.20 0.27 0.52 0.89

Miami-Fort Lauderdale, FL 95 4.66 0.21 0.30 0.56 0.93

Minneapolis-St. Paul, MN 73 4.65 0.19 0.26 0.51 0.89

New York-Northern New Jersey-Long Island, NY-NJ 93 4.60 0.21 0.30 0.56 0.92

Phoenix-Mesa, AZ 117 4.80 0.21 0.29 0.54 0.91

Portland-Salem, OR 88 4.77 0.20 0.27 0.52 0.88

San Diego, CA 103 4.71 0.19 0.26 0.51 0.89

San Francisco-Oakland-San Jose, CA 112 4.75 0.19 0.26 0.51 0.88

Seattle-Tacoma-Bremerton, WA 101 4.79 0.19 0.26 0.51 0.88

Tampa-St. Petersburg-Clearwater, FL 124 4.80 0.20 0.27 0.53 0.92

Washington-Baltimore, DC-MD-VA 116 4.61 0.21 0.28 0.55 0.93

Data Source: Optimal Blue

Notes: This table compares real-time interest rates for 30-year, ﬁxed-rate conforming mortgages with a LTV=80, FICO=750, DTI=36, and with no

points and fees. Column 1 shows the median number of lenders oﬀering mortgages in a location on a speciﬁc day. Columns 3-5 show the diﬀerence

between various percentiles of the oﬀer distribution.

Table A-5: Dispersion in Points and Fees Charged to Orig-

inate at the Median Interest Rate, from Lender Oﬀers

Percentile Diﬀerences

h − 25

− 10

− 1

Program

FHA 1.42 2.59 3.83

Conforming 1.19 2.22 3.69

Super-Conforming 1.23 2.35 3.79

Jumbo 1.13 2.31 3.84

FICO

640 1.30 2.41 3.83

680 1.22 2.35 3.77

720 1.19 2.30 3.78

750 1.20 2.30 3.78

LTV

70 1.19 2.28 3.77

80 1.24 2.37 3.81

90 1.19 2.29 3.81

95 1.20 2.26 3.72

96 1.32 2.44 3.80

Data Source: Optimal Blue

Notes: This table compares real-time points and fees charged by

diﬀerent lenders to originate identical mortgages at the median inter-

est rate. Points and fees are given as percent of the mortgage balance.

The median interest rate is chosen such that the median lender charges

no points and fees at this interest rate.

Table A-6: Variation of the Point-Rate Tradeoﬀ with FICO, LTV, and DTI

(1) (2) (3) (4)

DiscountP oints × I

F ICO≥720

0.001 0.001

(0.000) (0.000)

DiscountP oints × I

D T I>36

-0.002

∗∗∗

-0.002

∗∗∗

(0.000) (0.000)

DiscountP oints × I

LT V >80

0.008

∗∗∗

0.008

∗∗∗

(0.000) (0.000)

Average Coeﬃcient on Discount Points 0.218 0.218 0.218 0.218

Observations 3001321 3001321 3001321 3001321

Data Source: Optimal Blue

Notes: We estimate the relationship between discount points and interest rates in a regression

speciﬁcation identical to column (10) of Table 5, with the only exception that discount points are

allowed to only enter linearly. Discount points are allowed to vary by loan program x lock month,

and the average across all data is 0.218. Signiﬁcance: * p<0.1, ** p<0.05, *** p<0.01.

Table A-7: Pricing Variation Across Branches Within Lender

(1) (2) (3) (4) (5)

Size Quartile 2 0.00319 0.00264 0.00791 0.00739

(0.00608) (0.00589) (0.00517) (0.00627)

Size Quartile 3 0.00137 0.000757 0.00657 0.0116**

(0.00504) (0.00487) (0.00445) (0.00518)

Size Quartile 4 -0.000998 -0.000242 0.00896 0.0120*

(0.00584) (0.00578) (0.00577) (0.00703)

FHA share 0.278*** 0.201*** 0.115*** 0.121***

(0.0304) (0.0346) (0.0167) (0.0205)

Superconf. share -0.0273 -0.0176 -0.0639*** -0.0377*

(0.0442) (0.0393) (0.0201) (0.0202)

Jumbo share -0.0194 0.143 0.00720 0.0108

(0.112) (0.106) (0.0396) (0.0441)

Avg. county HHI -0.799* -1.371*** -1.191*** -1.425***

(0.466) (0.470) (0.436) (0.512)

Avg. income (1000s) -0.000890** -0.000711* -0.000381 -0.000626*

(0.000384) (0.000368) (0.000314) (0.000366)

Avg. BA+ share -0.426*** -0.244*** -0.198*** -0.163**

(0.0546) (0.0592) (0.0545) (0.0667)

Avg. minority share 0.111*** 0.0633*** 0.0554*** 0.0593***

(0.0221) (0.0228) (0.0183) (0.0209)

Lender FE? Yes Yes Yes Yes Yes

Mean Dep. Var. 0.00 0.00 0.00 0.00 0.00

SD Dep. Var. 0.14 0.14 0.14 0.19 0.19

Adj. R2 0.11 0.10 0.13 0.10 0.10

Obs. 6815 6815 6815 15881 12059

Sample restriction? Branch loans > 50 Branch loans > 50 Branch loans > 50 - Share largest branch < 20%

Data Sources: Optimal Blue, HMDA, ACS

Notes: Standard errors in parentheses are clustered by lender. * p<.1, ** p<0.05, *** p<0.01.

Table A-8: Relationship Between Mortgage Rates and Measures of Shopping and Knowledge

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Considered 2 lenders -0.028*** -0.022**

(0.007) (0.009)

Considered 3+ lenders -0.062*** -0.044***

(0.011) (0.012)

Applied to 2+ lenders -0.022** 0.065***

(0.009) (0.017)

Applied to 2+ lenders for better loan terms -0.040*** -0.075***

(0.010) (0.019)

Used web/broker/friends to get info? A lot -0.034*** -0.021***

(0.007) (0.008)

Familiar with mortgage rates? Very -0.058*** -0.042***

(0.015) (0.015)

Most lenders oﬀer same rate? No -0.028*** -0.022**

(0.010) (0.010)

Knows their interest rate -0.066*** -0.060***

(0.007) (0.007)

Answered whether rate is ﬁxed or variable -0.079*** -0.056***

(0.019) (0.019)

Adj. R2 0.52 0.52 0.52 0.52 0.52 0.52 0.53 0.52 0.53

Obs. 22567 22567 22567 22567 22567 22567 22567 22567 22567

Source: Authors’ calculations based on National Survey of Mortgage Originations and the National Mortgage Database

Notes: Dependent variable is the borrower’s mortgage interest rate in percentage points. Sample restricted to ﬁrst-lien 30-year, ﬁxed-rate loans for single-family principal

residence properties, with no more than two borrowers, originated from 2013 through 2019 (waves 1-24; “most lenders oﬀer same rate” was not asked in the ﬁrst 6 waves of

the survey so we include a separate dummy for these missing observations). All regressions control for origination month ﬁxed eﬀects, survey wave ﬁxed eﬀects, county ﬁxed

eﬀects, credit score (linear term plus dummies for 11 credit score bins), LTV (linear term plus dummies for 6 LTV bins), indicators for loan purpose (purchase, reﬁnance, or

cashout reﬁnance), 9 loan amount categories, loan program (Freddie, Fannie, FHA, VA, FSA/RHS, other), ﬁrst-time homebuyer status, single borrowers, log borrower income,

self-employment status of borrower(s), respondent gender, race and ethnicity, whether the household owns 4 diﬀerent types of ﬁnancial assets, whether the household could pay

their bills for 3 months without borrowing, CRA low-to-moderate income tract status, self-assessed creditworthiness, and the likelihood of moving, selling, or reﬁnancing within

a couple years. Observations weighted by NSMO sample weights. Robust standard errors in parentheses. * p<.1, ** p<0.05, *** p<0.01.

Table A-9: Is Mortgage Knowledge and Shopping Correlated with Loan Type and

Credit Score?

(1) (2) (3) (4) (5)

Loan Program

GSE -0.038*** -0.036*** -0.017***

(0.006) (0.006) (0.006)

FHA -0.068*** -0.053*** -0.034***

(0.007) (0.007) (0.007)

VA/RHS/FHLB -0.044*** -0.034*** -0.024***

(0.007) (0.007) (0.007)

Credit Score

720-779 -0.013*** -0.011*** -0.011***

(0.004) (0.004) (0.004)

620-719 -0.039*** -0.033*** -0.029***

(0.004) (0.004) (0.004)

< 620 -0.055*** -0.045*** -0.038***

(0.007) (0.008) (0.008)

First-time homebuyer

Yes 0.000 0.006 0.012***

(0.004) (0.004) (0.004)

Loan amount bins Yes

Adj R-sq 0.05 0.05 0.04 0.05 0.06

N 22567 22567 22567 22567 22567

Data Source: Authors’ calculations from the National Survey of Mortgage Originations and the

National Mortgage Database

Notes: Dependent variable is the “sophistication index,” which is a composite of six mortgage

knowledge and shopping questions and ranges from zero to one (see text for more details), and

has a mean of 0.46 and standard deviation of 0.2. On the right-hand side, the omitted loan type

category is nonconforming (e.g., jumbo) loans; the omitted credit score category is 780-850. Sample

is the same as described in Appendix Table A-8. All regressions control for origination month ﬁxed

eﬀects, survey wave ﬁxed eﬀects, and county ﬁxed eﬀects. Robust standard errors in parentheses.

* p<.1, ** p<0.05, *** p<0.01.

Table A-10: Do Borrowers Shop More When Mortgage Market Interest Rates Are Higher?

Considered 2+ lenders Applied to 2+ lenders Used other lenders Used web

for better terms to get info to get info

(1) (2) (3) (4) (5) (6) (7) (8)

A. All borrowers

PMMS rate 0.050** 0.040* 0.058*** 0.045*** 0.033 0.028 0.036* 0.045**

(0.020) (0.021) (0.016) (0.017) (0.020) (0.021) (0.020) (0.021)

Controls? Yes Yes Yes Yes

Mean of Dep. Var. 0.52 0.52 0.21 0.21 0.42 0.42 0.54 0.54

Obs. 22567 22567 22567 22567 22567 22567 22567 22567

B. DTI < 36

PMMS rate 0.074** 0.073** 0.077*** 0.088*** 0.061** 0.070** 0.032 0.053*

(0.029) (0.031) (0.023) (0.025) (0.029) (0.032) (0.029) (0.031)

Controls? Yes Yes Yes Yes

Mean of Dep Var 0.52 0.52 0.20 0.20 0.42 0.42 0.56 0.56

Obs. 10561 10561 10561 10561 10561 10561 10561 10561

Source: Authors’ calculations based on National Survey of Mortgage Originations and the National Mortgage Database

Notes: Dependent variables are listed above each column; all are dummy variables. Sample restricted to ﬁrst-lien, 30-year, ﬁxed-

rate mortgage for single-family principal residence properties, with no more than two borrowers, originated from 2013 through 2019

(the documentation question was not asked in every wave and thus has fewer observations). Controls include survey wave ﬁxed

eﬀects, county ﬁxed eﬀects, house price growth in the 12 and 24 months before the survey, credit score (linear term plus dummies for

11 credit score bins), LTV (linear term plus dummies for 6 LTV bins), indicators for loan purpose (purchase, reﬁnance, or cashout

reﬁnance), 9 loan amount categories, loan program (Freddie, Fannie, FHA, VA, FSA/RHS, other), ﬁrst-time homebuyer status,

single borrowers, log borrower income, self-employment status of borrower(s), respondent gender, race and ethnicity, whether the

household owns 4 diﬀerent types of ﬁnancial assets, whether the household could pay their bills for 3 months without borrowing,

metropolitan CRA low-to-moderate income tract status, self-assessed creditworthiness, and the likelihood of moving, selling, or

reﬁnancing within a couple years. Observations weighted by NSMO sample weights. Robust standard errors in parentheses. *

p<.1, ** p<0.05, *** p<0.01

Table A-11: Market Concentration and Price Dispersion

(1) (2) (3) (4) (5) (6) (7) (8)

Rate residual |Rate residual| I(residual< −0.2%) I(residual> 0.2%)

County HHI (year-1) -0.280 0.167 -0.307*** -0.380*** -0.419** -0.794*** -0.363* -0.147

(0.246) (0.218) (0.104) (0.0827) (0.209) (0.207) (0.188) (0.166)

Loan type × month FE? Yes Yes Yes Yes Yes Yes Yes Yes

Local controls? Yes Yes Yes Yes Yes Yes Yes Yes

Lender × month FE? No Yes No Yes No Yes No Yes

Mean Dep. Var. -0.00 -0.00 0.17 0.17 0.17 0.17 0.15 0.15

Adj. R2 0.00 0.16 0.08 0.13 0.03 0.13 0.02 0.08

Obs. 2995231 2995229 2995231 2995229 2995231 2995229 2995231 2995229

Standard errors in parentheses two-way clustered by lender and county.

* p<.1, ** p<0.05, *** p<0.01

Notes: Loan type × month FE are indicators for a loan being FHA, jumbo, or superconforming, interacted with the lock month.

Local controls are the county’s one-year lagged log number of loans originated (from HMDA), as well as the loan zip code’s median

household income, population share with at least a BA degree, and minority share. Rate residuals and HHI are winsorized at the 0.5th

and 99.5th percentile.

Table A-12: Summary Statistics of Expected Gains from Additional Search by Zip

Code Demographics

Observations Mean St. Deviation

Percentiles

All Mortgages 67,637 0.18 0.21 0.04 0.25

Median Household Income

First Tercile 22,585 0.21 0.23 0.06 0.29

Second Tercile 22,491 0.18 0.20 0.04 0.24

Third Tercile 22,536 0.16 0.20 0.03 0.21

Percent College Educated

First Tercile 22,569 0.21 0.23 0.06 0.29

Second Tercile 22,539 0.19 0.21 0.04 0.25

Third Tercile 22,518 0.15 0.18 0.03 0.19

Minority Share

First Tercile 22,560 0.16 0.19 0.03 0.21

Second Tercile 22,587 0.17 0.20 0.04 0.23

Third Tercile 22,479 0.22 0.23 0.05 0.30

County HHI

First Tercile 23,011 0.18 0.21 0.04 0.25

Second Tercile 22,117 0.18 0.21 0.04 0.24

Third Tercile 22,509 0.19 0.21 0.04 0.26

Data Source: Optimal Blue, American Community Survey (ACS), HMDA

Notes: This table summarizes the expected gain from an additional search, given by equation (1)

in the main text. The median household income, percent college educated, and minority share (share

of Hispanic/Latino plus non-Hispanic Black) are based on ACS data and observed at the zip code

level. The mortgage market HHI (Herﬁndahl-Hirschman Index) is computed at the county level,

averaging over 2016-2019 using the HMDA data.

Table A-13: Regressions of the Locked-Oﬀer Rate Gap on Observables

(1) (2) (3) (4) (5) (6)

FICO (omitted cat.: [640,660))

680≤F ICO<700

-0.060

∗∗∗

-0.046

∗∗∗

-0.039

∗∗∗

(0.008) (0.006) (0.009)

720≤F ICO<740

-0.092

∗∗∗

-0.059

∗∗∗

-0.056

∗∗∗

(0.010) (0.008) (0.013)

F ICO≥740

-0.128

∗∗∗

-0.081

∗∗∗

-0.064

∗∗∗

(0.011) (0.009) (0.013)

LTV (omitted cat.: (60,80])

85<LT V ≤90

0.018

∗∗∗

0.009

∗

0.018

∗∗∗

(0.005) (0.005) (0.006)

90<LT V ≤95

0.050

∗∗∗

0.034

∗∗∗

0.029

∗∗∗

(0.005) (0.005) (0.007)

LT V >95

0.186

∗∗∗

0.143

∗∗∗

0.102

∗∗∗

(0.012) (0.011) (0.019)

Loan Oﬃcer Comp (%) 0.172

∗∗∗

0.154

∗∗∗

(0.037) (0.039)

Loan amount f.e. ($10k bins) Yes Yes Yes Yes Yes Yes

MSA x Month f.e. Yes Yes Yes Yes Yes Yes

Branch f.e. Yes Yes Yes Yes

Adj. R-Squared 0.111 0.481 0.446 0.151 0.508 0.461

Observations 67637 65757 15444 67637 65757 15444

Data Source: Optimal Blue

Notes: The dependent variable is the mortgage interest rate locked minus the median oﬀer rate in the same

market and day for an identical mortgage. The data cover mortgage rates for 20 metropolitan areas during

the period 2016-2019. We focus on 30-year, ﬁxed-rate, fully documented purchase mortgages. Standard errors

shown in parentheses are two-way clustered at the month and lender level. Signiﬁcance: * p<0.1, ** p<0.05,

*** p<0.01.

Table A-14: Regressions of the Expected Gain from Search on Observables, for Independent

Nonbank Originators Only

(1) (2) (3) (4) (5) (6)

FICO (omitted cat.: [640,660))

680≤F ICO<700

-0.037

∗∗∗

-0.028

∗∗∗

-0.025

∗∗

(0.007) (0.005) (0.010)

720≤F ICO<740

-0.067

∗∗∗

-0.047

∗∗∗

-0.041

∗∗∗

(0.009) (0.007) (0.010)

F ICO≥740

-0.096

∗∗∗

-0.068

∗∗∗

-0.051

∗∗∗

(0.009) (0.007) (0.008)

LTV (omitted cat.: (60,80])

85<LT V ≤90

0.015

∗∗∗

0.013

∗∗∗

0.013

∗∗∗

(0.003) (0.003) (0.005)

90<LT V ≤95

0.033

∗∗∗

0.026

∗∗∗

0.018

∗∗∗

(0.004) (0.004) (0.005)

LT V >95

0.131

∗∗∗

0.106

∗∗∗

0.065

∗∗∗

(0.010) (0.009) (0.009)

Loan Oﬃcer Comp (%) 0.107

∗∗∗

0.097

∗∗∗

(0.026) (0.027)

Loan amount f.e. ($10k bins) Yes Yes Yes Yes Yes Yes

MSA x Month f.e. Yes Yes Yes Yes Yes Yes

Branch f.e. Yes Yes Yes Yes

Adj. R-Squared 0.130 0.400 0.410 0.169 0.428 0.420

Observations 56483 55053 13123 56483 55053 13123

Data Source: Optimal Blue

Notes: The dependent variable is the expected gain from an additional search, given by equation (1). The data

cover mortgage rates for 20 metropolitan areas during the period 2016-2019. We focus on 30-year, ﬁxed-rate,

fully-documented purchase mortgages. Standard errors shown in parentheses are two-way clustered at the month

and lender level. Signiﬁcance: * p<0.1, ** p<0.05, *** p<0.01.

Table A-15: Regressions of the Expected Gain from Search, Constructed Using Oﬀers from

High-Volume Lenders Only, on Observables

(1) (2) (3) (4) (5) (6)

FICO (omitted cat.: [640,660))

680≤F ICO<700

-0.037

∗∗∗

-0.026

∗∗∗

-0.021

∗∗

(0.006) (0.004) (0.008)

720≤F ICO<740

-0.069

∗∗∗

-0.047

∗∗∗

-0.044

∗∗∗

(0.008) (0.006) (0.009)

F ICO≥740

-0.096

∗∗∗

-0.067

∗∗∗

-0.052

∗∗∗

(0.008) (0.006) (0.008)

LTV (omitted cat.: (60,80])

85<LT V ≤90

0.014

∗∗∗

0.010

∗∗∗

0.010

∗∗

(0.003) (0.003) (0.004)

90<LT V ≤95

0.033

∗∗∗

0.025

∗∗∗

0.020

∗∗∗

(0.004) (0.004) (0.005)

LT V >95

0.126

∗∗∗

0.100

∗∗∗

0.069

∗∗∗

(0.009) (0.009) (0.012)

Loan Oﬃcer Comp (%) 0.106

∗∗∗

0.096

∗∗∗

(0.024) (0.026)

Loan amount f.e. ($10k bins) Yes Yes Yes Yes Yes Yes

MSA x Month f.e. Yes Yes Yes Yes Yes Yes

Branch f.e. Yes Yes Yes Yes

Adj. R-Squared 0.107 0.377 0.374 0.142 0.402 0.385

Observations 67637 65757 15444 67637 65757 15444

Data Source: Optimal Blue

Notes: The dependent variable is the expected gain from an additional search, given by equation (1), and it

is computed using oﬀers only from high-volume lenders as identiﬁed in the Optimal Blue Pricing Insights data.

The data cover mortgage rates for 20 metropolitan areas during the period 2016-2019. We focus on 30-year,

ﬁxed-rate, fully-documented purchase mortgages. Standard errors shown in parentheses are two-way clustered

at the month and lender level. Signiﬁcance: * p<0.1, ** p<0.05, *** p<0.01.

Table A-16: The Relationship Between the Lock-Oﬀer Rate Gap and Treasury Yields

(1) (2) (3) (4) (5) (6)

Treasury Yield -0.059

∗∗∗

-0.081

∗∗∗

-0.051

∗∗∗

-0.067

∗∗∗

-0.089

∗∗∗

-0.058

∗∗∗

(0.008) (0.015) (0.015) (0.008) (0.016) (0.016)

Treasury Yield x DTI ≤ 36 0.020

∗∗∗

0.020

∗∗∗

0.017

∗∗∗

(0.006) (0.006) (0.006)

Borrower and Loan Controls Yes Yes Yes Yes Yes Yes

MSA F.E Yes Yes Yes Yes Yes Yes

MSA x Month F.E. Yes Yes Yes Yes

Branch F.E. Yes Yes

Adj. R-Squared 0.141 0.150 0.503 0.143 0.152 0.504

Observations 65938 65936 64025 65938 65936 64025

Data Source: Optimal Blue

Notes: The dependent variable is the mortgage interest rate locked minus the median oﬀer rate in the same

market and day for an identical mortgage. The data cover mortgage rates for 20 metropolitan areas during the

period 2016-2019. We focus on 30-year, ﬁxed-rate, fully-documented purchase mortgages. Standard errors shown

in parentheses are two-way clustered at the month and lender level. Signiﬁcance: * p<0.1, ** p<0.05, *** p<0.01.

Table A-17: The Relationship Between Lenders’ Expensiveness and Their Income and Expenses — Mean (OLS) Regression vs.

Median Regression

Mean Regression Median Regression

Observations β St. Err. β St. Err.

Income

Origination Income 1,897 1.02 (0.69) 0.78*** (0.19)

Interest Income 1,897 −0.07 (0.10) 0.03 (0.04)

Secondary Market Income (Gain-on-Sale) 1,897 2.44*** (0.76) 3.37*** (0.43)

Other Income 1,897 0.71 (0.46) 0.05** (0.02)

Gross Income 1,897 3.47*** (0.70) 4.05*** (0.41)

Expenses

Loan Production Oﬃcers (Sales Employees) 1,897 0.86*** (0.29) 1.06** (0.47)

Loan Origination (Fulﬁllment/Non-Sales) 1,897 0.36** (0.15) 0.45*** (0.13)

Origination-Related Management and Directors 1,897 0.36** (0.14) 0.25*** (0.07)

Employee Beneﬁts 1,897 0.31*** (0.09) 0.33*** (0.05)

Other Personnel Expenses 1,897 0.37* (0.22) 0.24 (0.13)

Interest Expenses 1,897 −0.08 (0.12) 0.06 (0.04)

Occupancy and Equipment 1,897 0.29*** (0.05) 0.30*** (0.03)

Technology 1,897 0.08* (0.04) 0.12*** (0.03)

Outsourcing, Professional, and Subservicing Fees 1,897 0.07 (0.10) 0.10*** (0.02)

Other non-interest expenses 1,897 0.44 (0.32) 0.26 (0.23)

Gross Expenses 1,897 3.05*** (0.65) 3.50*** (0.41)

Corporate overheads

Total corporate expenses 1,897 0.50 (0.31) 0.28*** (0.10)

Net Income

Net Income (residential originations, pre-corporate allocations) 1,897 0.45* (0.25) 0.45** (0.18)

Net Income (all lines, after corporate allocations and taxes) 1,897 0.46 (0.57) 0.24** (0.12)

Data Source: Conference of State Bank Supervisors Mortgage Call Reports; Optimal Blue

Notes: The table displays results from OLS and median regressions of lenders’ ﬁnancial ﬁling line items on lender expensiveness (in percentage points),

as measured from lender ﬁxed eﬀects column 4 of Table 5, and year-quarter ﬁxed eﬀects. All line items are shown as dollars per $100 originated. Some

categories for income and expenses are combined or not shown since they only have zeros for all ﬁrms. The standard errors are clustered at the lender level.

* p<.1, ** p<0.05, *** p<0.01.

Empirical CDF

-2 -1 0 1 2

Discount Points Paid

Conforming

FHA

Jumbo

Figure A-1: The Empirical Cumulative Distribution of Discount Points Paid, by Program

Data Source: Optimal Blue

Note: Figure shows cumulative share of borrowers who paid up to a certain amount of discount points; negative values

represent credits/rebates. Data include purchase and reﬁnance rate locks in 2015-2019.

Share of loans with a bad deal (rate residual > +0.2)

0 .2 .4 .6 .8 1

Share of loans with a good deal (rate residual < -0.2)

Figure A-2: Are Lenders That Give Good Deals More Likely to Also Give Bad Deals?

Data Source: Optimal Blue

Note: The chart shows lender level shares of locks considered to be good and bad deals. We deﬁne

whether a lock is a good or bad deal by using the rate residual from speciﬁcation (3) of Table 5. A good

deal is deﬁned as a lock with a rate residual more than 20bp below the average rate lock for that mortgage

product in a particular date after controlling for borrower and loan characteristics. Conversely, a bad deal

is a lock with a rate residual of more than 20bp. The size of the bubble represents the size of the lender.

The Spearman correlation of the share of loans with good and bad deals is -0.64.

2.75

3.25

3.5

3.75

4.25

4.5

4.75

5.25

5.5

Jan 2016 Jan 2017 Jan 2018 Jan 2019 Jan 2020

Mortgage News Daily

Freddie Mac PMMS

Optimal Blue Insight

Conforming Mortgages

2.75

3.25

3.5

3.75

4.25

4.5

4.75

5.25

5.5

Jan 2016 Jan 2017 Jan 2018 Jan 2019 Jan 2020

Mortgage News Daily

Optimal Blue Insight

FHA Mortgages

2.75

3.25

3.5

3.75

4.25

4.5

4.75

5.25

5.5

Jan 2016 Jan 2017 Jan 2018 Jan 2019 Jan 2020

Mortgage News Daily

Zillow

Optimal Blue Insight

Jumbo Mortgages

Figure A-3: Comparison of Average Oﬀer Rates from Optimal Blue with Mortgage News

Daily Data

Data Source: Optimal Blue, Mortgage News Daily, Freddie Mac, Zillow

Note: The Optimal Blue Data are for borrowers with FICO=750, DTI=36, with no points/fees, and LTV=80 for conforming

and jumbo, and LTV=96.5 for FHA. The Mortgage News Daily (MND) data reﬂect rates for “top-tier” borrowers, and we

adjust the MND rates assuming they include 0.5% points and fees.

Figure A-4: Comparison of Locked Interest Rates from Optimal Blue with Interest Rates on

Closed Originations in McDash

Data Source: Optimal Blue, Black Knight McDash

Note: The Optimal Blue series lead the McDash series because for Optimal Blue we observe the date when the loan terms are

locked, while in McDash we observe when a loan is originated.

Figure A-5: Screenshot of Sample Oﬀer Distribution from Optimal Blue Pricing Insights

Data Source: Optimal Blue

Note: Figure shows an example of the real-time distribution of oﬀers across lenders in the same metropolitan area for a loan

with given characteristics and at a note rate of 5.125%. Lenders are sorted by “price,” which equals 100 + the points

(rebate/credit) the lender pays to the borrower (so “102” means the borrower receives two points at closing, while “98” means

they would have to pay two points). The mortgage note rate for which oﬀers are shown is chosen such that the median lender

oﬀers a price as close as possible to 100. For actual lenders using the interface, an orange dot would show their position in the

distribution.

.01

.02

.03

.04

.05

Fraction

-.6 -.5 -.4 -.3 -.2 -.1 0 .1 .2 .3 .4 .5 .6

Offered Rate Minus Median Rate

Conforming, $300k, FICO=750, LTV=80, DTI=36, No Points/Fees

.01

.02

.03

.04

Fraction

-.6 -.5 -.4 -.3 -.2 -.1 0 .1 .2 .3 .4 .5 .6

Offered Rate Minus Median Rate

FHA, $300k, FICO=680, LTV=96, DTI=36, No Points/Fees

.02

.04

.06

.08

Fraction

-.6 -.5 -.4 -.3 -.2 -.1 0 .1 .2 .3 .4 .5 .6

Offered Rate Minus Median Rate

Jumbo, $700k, FICO=750, LTV=80, DTI=36, No Points/Fees

Figure A-6: Interest Rate Oﬀer Dispersion for Identical Mortgages in Los Angeles

Data Source: Optimal Blue

Note: The spread is deﬁned as the diﬀerence between real-time mortgage rate oﬀers and the median oﬀer rate for identical

mortgage products. The histogram includes daily data between April 2016 and December 2019.

.01

.02

.03

.04

Fraction

-3 -2.5 -2 -1.5 -1 -.5 0 .5 1 1.5 2 2.5 3

Points and Fees (% of mortgage balance)

Distribution in Points/Fees Charged for Mortgages at the Median Interest Rate

Figure A-7: Dispersion in Points and Fees Lenders Charge for Identical Mortgages at the

Median Interest Rate

Data Source: Optimal Blue

Note: Points and fees are given as percent of the mortgage balance. The median interest rate is calculated as the rate at

which the median lender charges no points and fees.

-.3

-.2

-.1

Residualized Rate

-1.25 -1 -.75 -.5 -.25 0 .25 .5 .75 1 1.25

Residualized Points

Figure A-8: The Relationship Between Discount Points Paid and Mortgage Rates

Data Source: Optimal Blue

Note: Binned scatter plot. Discount points and mortgage rates are ﬁrst residualized using a regression speciﬁcation identical

to column (10) of Table 5, with the only exception that we are not controlling for discount points.