Advertising Expensive Mortgages

• Author: GREGOR MATVOS,AMIT SERU,UMIT G. GURUN
• Date: 01 October 2016
• Published date: 01 October 2016
• DOI: http://doi.org/10.1111/jofi.12423

THE JOURNAL OF FINANCE •VOL. LXXI, NO. 5 •OCTOBER 2016

Advertising Expensive Mortgages

UMIT G. GURUN, GREGOR MATVOS, and AMIT SERU∗

ABSTRACT

Using information on advertising and mortgages originated by subprime lenders,

we study whether advertising helped consumers find cheaper mortgages. Lenders

that advertise more within a region sell more expensive mortgages, measured as

the excess rate of a mortgage after accounting for borrower, contract, and regional

characteristics. These effects are stronger for mortgages sold to less sophisticated

consumers. We exploit regional variation in mortgage advertising induced by the

entry of Craigslist and other tests to demonstrate that these findings are not spurious.

Analyzing advertising content reveals that initial/introductory rates are frequently

advertised in a salient fashion, where reset rates are not.

RECENT LITERATURE SHOWS THE importance of search in the mortgage market

(Mayer and Pence (2009), Scharfstein and Sunderam (2013)). Although mort-

gages are relatively homogeneous products, search frictions create a demand

for information about mortgages that lenders can cater to. There are two broad

views on how lenders use advertising to supply this information to consumers.

On the one hand, the information view claims that advertising allows con-

sumers to find better products (Nelson (1974)). On the other hand, the per-

suasion view suggests that advertising is used to steer consumers into bad

choices (Braithwaite (1928), Thaler and Sunstein (2008)). These views are at

the center of a debate on the role of advertising in the mortgage market in

the aftermath of the housing crisis. Several policy and regulatory changes that

have emerged from these discussions are based on the idea that na¨

ıve con-

sumers were duped by advertising to take an expensive mortgages.1While

∗Umit G. Gurun is at University of Texas at Dallas. Gregor Matvos and Amit Seru are at Booth

School of Business, University of Chicago, and NBER. We thank Geoff Booth, Justine Hastings,

Ali Hortacsu, Devin Pope, Andrei Shleifer, Kelly Shue, Adi Sunderam, Chad Syverson, James

Vickery, Rob Vishny, Richard Thaler, Luigi Zingales, Eric Zitzewitz, and seminar participants at

Chicago Booth, Chicago Fed, New YorkFed, NYU Stern, Rochester, Stockholm, UNC, AEA, NBER

Behavioral Finance meeting, NBER Household Finance Summer Institute, WFA, and European

Conference on Household Finance for helpful suggestions. We are extremely grateful to Robert

Seamans and Feng Zhu for providing some of the data used in Craigslist analysis. Xu Cheng,

Laura Harris, Will Powley, Tricia Sun, and Adithya Surampudi provided outstanding research

assistance. Seru and Matvos thank Fama Miller Center and Initiative on Global Markets for

funding. All errors are our own.

1Regulators have penalized lenders for deceptive practices and implemented explicit regu-

lation of mortgage advertising. The FDIC implemented Regulation Z in 2008 and the FTC

passed the Mortgage Acts and Practices Advertising rule in 2011, both of which directly regulate

DOI: 10.1111/jofi.12423

2371

2372 The Journal of Finance R

anecdotes have been used to justify these claims of deceptive advertising, there

is no empirical study that has systematically investigated this issue.2

In this paper, we provide evidence for deceptive advertising using unique

micro data on lending and advertising from the subprime mortgage market. We

then compare the performance of a rich set of advertising models in explaining

the data. The results reject the canonical models of informative advertising.

Our data set combines the intensity and content of local advertising by sub-

prime lenders with the contract, region, and borrower characteristics of mort-

gages originated by them. We focus on adjustable-rate mortgage (ARM) loans,

particularly ARM reset rates, because they have been at the center of lawsuits

and regulatory scrutiny. The concern is that advertising lures consumers into

bad choices by focusing their attention on the introductory interest rate, fos-

tering the impression that the (low) rate will be permanent rather than reset

after the first few years.

We empirically confirm the view that reset rates are hidden characteristics of

advertised mortgages. In the 37,432 print and direct mail mortgage campaigns

that we analyze, only seven (0.02%) explicitly mention a reset rate. While this

result is suggestive of persuasive advertising that shrouds reset rates, it is not

sufficient to reject informative advertising. In other words, the fact that reset

rates are not being advertised does not imply that consumers are unaware of

them and therefore make worse choices than they should. The obvious difficulty

in separating the informative and persuasive views is that one needs to identify

better and worse mortgages, and then relate the choices of consumers, who may

not be otherwise identical, to lender advertising.

We measure whether mortgages are relatively better or worse for the con-

sumer by computing the extent to which identical consumers pay different

prices for otherwise similar mortgages in a given market. In particular, as-

suming that, all else equal, cheaper mortgages are better products from the

perspective of the consumer, we measure the relative expensiveness of a given

mortgage as the excess reset rate of the mortgage after accounting for a broad

set of borrower, contract, and regional characteristics associated with the mort-

gage, including the initial interest rate. The idea is that if identical consumers

obtain the same mortgage with different reset rates in the same market, then

the difference in the reset rates measures how much worse the choice of the

consumer with the higher reset rate was.

We find large differences in average reset rates charged by lenders within

geographic regions (designated market areas, or DMAs) after conditioning on

borrower and loan characteristics and the initial interest rate: the average

advertising of mortgages. An October 1, 2000 New York Times article summarized this prevailing

view: “One of the most important lessons of the mortgage collapse is that potential borrowers

need clear explanations of exactly what kind of commitment they are making.” The Fed fined

Wells Fargo $85 million for steering consumers into expensive mortgages, and the Department

of Justice reached a $175 million settlement with Wells Fargo to resolve fair lending claims (see

http://www.justice.gov/opa/pr/2012/July/12-dag-869.html; accessed on February 29, 2013).

2In general, lack of data has precluded research on advertising mortgage products (see Agarwal

and Ambrose (2011)).

Advertising Expensive Mortgages 2373

difference between the 95th and 5th percentile lenders in a given region is 2.8

percentage points. This result suggests that loans originated by some lenders

are, on average, more expensive than others.

We next find that lender expensiveness is positively correlated with adver-

tising within a given market. Thus, lenders that advertise more intensively

also charge more for the same mortgage. To show the above results are not

spurious, we first exploit variation in the relative advertising of lenders within

a given location using region fixed effects. Our results continue to hold. Thus,

our findings are not driven simply by lenders advertising more in regions with

higher mortgage prices. Exploiting within-lender variation also allows us to

allay concerns that the results are driven by lender characteristics (such as the

lender’s brand, propensity to renegotiate or securitize, or marginal costs, other

lenders’ activities aimed at attracting customers) that may be correlated with

advertising.

We next examine whether advertising attracts borrowers who are charged

higher reset rates due to a lower ability to repay. This alternative could ex-

plain our findings if true borrower quality is not captured by our rich set of

conditioning variables. If advertising lenders extend loans to borrowers who

are less likely to repay a loan, then such borrowers should be less likely to

repay a loan in the future. We find that advertisers lend to consumers who, all

else equal, default less, which implies that our results are not likely driven by

unobservable borrower quality.

Second, we exploit variation in mortgage advertising induced by the stag-

gered entry of Craigslist across different regions and times regions and years.

Mortgage classifieds represent over 8% of all financial services posts on

Craigslist. Thus, Craigslist entry into a market serves as a potentially viable

source of variation in mortgage advertising in that market. Indeed, Craigslist

entry has a significant impact on paid mortgage advertising, with the high-

est impact on classified advertisements in newspaper advertising, for which

Craigslist online classifieds substitute most directly. We continue to find a

positive relationship between the intensity of local advertising and the ex-

pensiveness of mortgages extended by lenders. Importantly, Craigslist entry

is unrelated to borrower characteristics in that region. These results indicate

that catering costs are not likely driving our findings, and thus further support

the view that advertising attracts borrowers of different riskiness.

The magnitudes of our results are large and suggest that a consumer who

obtains a mortgage from a lender who advertises pays on average roughly

$3,000 more in present-value terms if the consumer pays the reset on an ARM

for one year before refinancing. These estimates are on the same order of

magnitude as the estimated losses faced by mortgage borrowers who do not

properly account for broker service fees given by Hall and Woodward (2012).

Having established a positive relationship between the expensiveness of a

lender and the advertising intensity of that lender within a market, we can

compare the performance of models of the information view and the persuasion

view of advertising. Canonical models of informed advertising (e.g., Butters

(1977), Robert and Stahl (1993), Bagwell and Ramey (1994)) cannot generate