When Saving Is Gambling PDF

JID: FINEC
ARTICLE IN PRESS [m3Gdc;April 19, 2018;13:41]

Journal of Financial Economics 0 0 0 (2018) 1–22
Contents lists available at ScienceDirect
Journal of Financial Economics

journal homepage: www.elsevier.com/locate/jfec
When saving is gamblingR

J. Anthony Cookson
Leeds School of Business, University of Colorado at Boulder, Campus Box 419, Boulder, CO 80309, USA
a r t i c l e i n f o a b s t r a c t
Article history: Prize-linked savings (PLS) accounts, which allocate interest using lottery payments rather
Received 20 December 2016 than fixed interest, encourage savings by appealing to households’ gambling preferences. I
Revised 13 February 2017
introduce new data on casino cash withdrawals to measure gambling, and examine how
Accepted 14 March 2017
individual gambling expenditures respond to the introduction of PLS in Nebraska using a
Available online xxx
difference-in-differences design. After PLS is introduced, individuals who live in counties
JEL classification: that offer PLS reduce gambling by at least 3% more than unaffected individuals. The sub-
G21 stitution effect is stronger in low-frills gambling environments, which most resemble PLS,
D14 indicating that these accounts fulfill the desire to gamble.
L83
© 2018 Elsevier B.V. All rights reserved.
Keywords:
Prize linked savings
Gambling
Risk aversion
Financial literacy
Credit unions
1. Introduction
“Why do people play the lottery or why do people gam-

R
The author gratefully acknowledges helpful comments from confer-
ble, period? You know, it is with the hope of winning
ence and seminar participants at the 2015 Financial Research Associa- something more. There is a sense that this (prize-linked
tion Conference, the 2015 European Finance Association Conference, the savings) actually makes savings fun.” – Derek Kilmer,
2015 UBC Summer Conference, the 2015 Oregon Summer Conference, the Washington State Senator, PBS Newshour, 23 November
2015 Helsinki Finance Summit, the 2015 NBER-SI Workshop on House-
2013
hold Finance, the 2015 Finance Down Under Conference, the 2015 FTC
Microeconomics Conference, the 2015 Northern Finance Association Con-
Some households prefer financial products with high
ference, the 2015 Israel Behavioral Finance Conference, the 2015 Boulder
Summer Conference on Consumer Financial Decision Making, the 2016 variance and skew, despite earning low return (Dorn and
Chicago Quantitative Alliance Spring Meeting, the Federal Reserve Bank Sengmueller, 2009; Boyer and Vorkink, 2014). Gambling
of Cleveland, Colorado State University, University of Nevada – Las Vegas, in financial markets is typically viewed as problematic,
University of Manitoba, Montana State University, and the University of but catering to a gambling preference could also encour-
Colorado. In addition, the paper owes much to helpful suggestions from
age households to engage more with financial markets.
an anonymous referee, Asaf Bernstein, Brad Barber, Jamie Brown, Bruce
Carlin, Shanna Cookson, Rob Dam, Shaun Davies, Daniel Dorn (discus- From this standpoint, financial products that cater to gam-
sant), Cary Frydman (discussant), Rawley Heimer, Ben Iverson (discus- bling preferences could be beneficial, especially given the
sant), Petri Jylha (discussant), Chi Liao, Carsten Murowski (discussant), low level of engagement with household savings products
Sridhar Narayanan (discussant), John Lynch, Genevieve Melford (discus-
(Lusardi et al., 2011). In this spirit, policymakers have pro-
sant), Marina Niessner (discussant), Charlie Sprenger (discussant), Sheri-
dan Titman (discussant), Carly Urban, Ed Van Wesep, Toni Whited (the posed Prize-Linked Savings (PLS) accounts, which reward
editor), and Fernando Zapatero. All remaining errors are my own. savings by offering randomly drawn lottery payoffs in lieu
E-mail address: tony.cookson@colorado.edu of fixed interest payments. PLS accounts are a potentially
https://doi.org/10.1016/j.jfineco.2018.04.001
0304-405X/© 2018 Elsevier B.V. All rights reserved.
Please cite this article as: J.A. Cookson, When saving is gambling, Journal of Financial Economics (2018),
JID: FINEC
2 J.A. Cookson / Journal of Financial Economics 000 (2018) 1–22
appealing way to encourage savings if people with low tifact of the casino cash data, and indicates a broad change
savings rates also like to gamble. Despite the intuitive ap- in gambling behavior in the wake of the introduction of
peal, we still know little about the effects of PLS because prize-linked savings.
these programs have only recently gained academic atten- If the substitution effect works through gambling pref-
tion (e.g., Filiz-Ozbay et al., 2015; Cole et al., 2017). erences, savings lotteries and casino gambling ought to
In this context, I empirically examine how household be weaker substitutes when they are more differentiated
casino gambling responds to the introduction of PLS, and along other dimensions. Consistent with gambling prefer-
I find that PLS substitutes strongly for gambling. Showing ences, I present three heterogeneity tests that show sub-
that PLS substitutes for gambling in other markets is of stitution among similar gambles, and minimal substitution
general interest for at least three reasons. First, this finding when savings lotteries and casino gambling are more dif-
shows that households prefer to take on some kinds of risk ferentiated. First, savings lotteries are a strong substitute
in a financial context, which contrasts with the traditional for local gambling, but not for destination gambling when
view that individuals require compensation for exposures the trip is part of the enjoyment. Second, as the date of
to risk (e.g., Savor and Wilson, 2014). Indeed, exposure to the savings raffle draws nearer, savings lotteries and casino
these risks appears to incentivize saving, rather than dis- gambling become stronger substitutes, consistent with the
courage it. This aspect of my findings is similar to recent gambling payoffs becoming more similar in terms of their
work using the callable options market to study catering to immediacy. Third, savings lotteries are a strong substitute
behavioral preferences (Li et al., 2018). Second, in my anal- for gambling at casinos without nightlife, but not for casi-
ysis, households substitute between gambles in different nos with nightlife, which are better differentiated from
domains (i.e., savings lotteries versus traditional gambling), savings lotteries.
which indicates a general preference for gambling that is The evidence on heterogeneity in the substitution effect
not naturally explained by narrow framing (e.g., Barberis helps rule out a broad class of alternative interpretations.
et al., 2006). In this way, my findings provide novel evi- In particular, the finding that substitution is strongest
dence on individual preferences for skewness (e.g., Boyer when savings lotteries and casino gambling are most sim-
and Vorkink, 2014; Viva et al., 2017). Third, as policies in- ilar contrasts with an interpretation that STW reduced
troduce lottery-like elements to financial products, it is im- gambling through an attention-grabbing effect, or through
portant to understand their effects, not just for personal effective advertising that was targeted toward Nebraska
savings rates, but for other household behaviors as well. consumers in served counties (Becker and Murphy, 1993;
In finding significant effects on gambling in other markets, Barber and Odean, 2008; Hastings et al., 2017). Attention-
my work suggests that these broader effects are important. grabbing effects and advertising cannot explain why gam-
The savings lottery accounts I study—Save-to-Win bling activity at local casinos is more sensitive to the intro-
(STW) accounts—became available to members of select duction of savings lotteries, why the substitution effect is
credit unions in Nebraska in January 2012. My empirical stronger among casinos without nightlife, nor why late-in-
analysis uses proprietary transaction-level data on casino the-month casino transactions are more sensitive to sav-
cash withdrawals to measure how households in affected ings lotteries.1 In a similar spirit, the findings cannot be
counties change their casino gambling activity differently explained by a blanket commitment to spend less because
from households in nearby counties unaffected by the new savings lotteries affect some, but not all types of gambles.
accounts. Using this difference-in-difference approach, I The fact that substitution is strongest when casino gam-
find robust evidence that the introduction of savings lotter- bling is most similar to the experience of a savings lottery
ies reduces the amount of casino gambling. That is, house- strongly suggests that the substitution effect reflects gam-
holds’ newfound opportunity to gamble while saving in bling preferences.
STW accounts is a strong substitute for gambling at com- My findings can be viewed as a partial empirical vali-
mercial casinos. dation of the Barberis (2012) model of casino gambling in
My estimates indicate that the introduction of prize- which patrons exhibit different degrees of self-control at
linked savings led to economically substantial reductions the casino. In the model, high self-control patrons receive
in gambling activity. Relating cash withdrawals to over- a lottery-like payoff from casino gambling (right-skewed
all casino demand, this estimated effect of STW on cash with a few high positive outliers), whereas low self-control
withdrawals translates into a 3.7–10.2% reduction in the gamblers do not receive a right-skewed payoff profile from
amount of gambling for the average affected county. casino gambling because they cannot commit to stop-
In aggregate, I estimate that the introduction of prize- ping while ahead. In this way, the Barberis (2012) model
linked savings reduced gambling by between $175,0 0 0 and motivates why some gamblers would view lotteries and
$396,0 0 0, which is a substantial fraction of the approx- casino gambling as substitutes. This intuition provides the
imately $2 million in additional savings at participating testable prediction that the substitution effect should be
credit unions. This effect arises primarily because of less stronger among high self-control patrons than it is for
visitation as individuals who were exposed to savings lot- low self-control patrons. Indeed, when I examine how the
teries were 15% less likely to visit the casino in the post- substitution effect interacts with proxies for patron self-
period. These results are robust to a battery of alterna-
tive specifications, different subsamples, and accounting 1
Moreover, treated and control counties are likely to be located in the
for pre-trends in the amount of casino gambling. Moreover,
same advertising markets, which is the level at which advertising is de-
I find similar substitution effects away from scratch ticket termined. See Spenkuch and Toniatti (2016) for a detailed discussion of
lottery sales, which shows that my findings are not an ar- advertising markets.
JID: FINEC
J.A. Cookson / Journal of Financial Economics 000 (2018) 1–22 3
control (e.g., low propensity to request unavailable funds lottery-like features. For example, Dorn et al. (2014) find a
on debit transactions, or use a credit card to access cash), negative relation between traditional lottery prize amounts
I find stronger substitution among patrons with greater and retail stock market activity for both the U.S. and Ger-
self-control. Consistent with the model, low self-control many. In a similar vein, Gao and Lin (2015) find a negative
individuals do not substitute at all, and these findings spillover effect in Taiwan. Although these studies show
are robust to controlling for differences in economic that investors appear to exhibit gambling preferences in
conditions and education. the construction of their investment portfolios, they have
My work contributes to the literature on household not quantified the degree to which gambling in financial
saving programs, which has gained significant recent at- contexts replaces gambling in other contexts. By providing
tention (Beshears et al., 2015; Brown et al., 2017). Prior evidence on the substitution away from casino gambling,
work has recognized that households exhibit significant my work suggests that financial alternatives to gambling
inertia in savings decisions, making it difficult for poli- may lead to welfare improvements for individuals who
cies to influence savings rates directly (Madrian and Shea, replace commercial gambling, which yields negative ex-
20 01; Thaler and Benartzi, 20 04; Chetty et al., 2014; Choi, pected returns, with financial gambling, which tends to
2015). Within this literature, my work is most closely re- yield positive returns.2
lated to a small number of studies on prize-linked savings More generally, the study of gambling has been a sub-
accounts (Tufano, 2008; Kearney et al., 2011). For exam- ject of curiosity in the economics literature since the de-
ple, Cole et al. (2017) study South African savings lottery velopment of the expected utility hypothesis (Friedman
accounts (specifically, Million-a-Month-Account; MaMA), and Savage, 1948; Rosett, 1965). Because this paper links
showing their efficacy for increasing savings, and Filiz- formal gambling and gambling through the use of a fi-
Ozbay et al. (2015) provide laboratory evidence on lottery nancial product (savings lotteries), my findings relate to
savings as a mechanism unto itself to encourage greater the general study of gambling behavior, which occurs well
saving. By analyzing actual gambling patterns, my work beyond gambling markets (Becker et al., 2005). In this
complements these analyses by showing that savings lot- broader literature, my findings should be of interest to
teries lead to significant substitution away from casino scholars who study consumer finance outcomes, gambling
gambling, which is an important consideration given the motives in corporate financial decisions, and gambling mo-
broad growth in gambling highlighted by Kearney (2005). tives in occupational choice (Schneider and Spalt, 2017;
My findings are also relevant to the literature on the Peng and Thibodeau, 2017; Zhang, 2017).
relation between financial literacy and household finan-
cial decisions (Behrman et al., 2012; Lusardi and Mitchell, 2. Setting
2014). In this area, prior work links more financial edu-
cation to better personal financial outcomes (Cole et al., To provide context for the empirical analysis, this sec-
2014; Brown et al., 2016), whereas others argue that per- tion describes the historical background of prize-linked
sonal characteristics are important for financial behavior savings accounts, and the introduction of prize-linked sav-
(Fernandes et al., 2014; Kuhnen and Miu, 2017). Viewed ings to Nebraska via participating credit unions.
through this lens, my analysis suggests that the design
of innovative savings programs ought to carefully consider 2.1. Background on prize-linked savings
how personal characteristics and financial education inter-
act with the program’s incentives. For example, my evi- Though uncommon in the United States, lottery sav-
dence on the introduction of Save-to-Win indicates that ings accounts and related financial products exist in other
the STW program was less effective for the sub-population countries, and in some cases, have existed for a long time.
of individuals with low self-control who arguably could For example, since 1956, investors in the United Kingdom
benefit the most from saving. On the other hand, this find- have been able to invest in premium bonds issued by
ing also suggests that improvements to personal character- the National Savings and Investments agency. These bonds
istics like self-control (e.g., through targeted financial edu- pay interest, but instead of the interest being allocated
cation interventions) could complement innovative savings to individual accounts, a lottery determines the set of ac-
programs in a broader program to address low personal countholders who receive the interest payments as well
savings rates. as the payment amounts. In an analysis of U.K. premium
My findings also contribute to the literature on the role bonds, Tufano (2008) notes that investors treat lottery-
of gambling and sensation seeking in financial markets. In style accounts as part savings, part gambling, and he con-
this growing body of research, researchers have examined jectures that the gambling feature of these accounts could
how sensation seeking by individual investors leads to be used to increase savings rates. Similar products are of-
excessive trading to the detriment of retail investors’ port- fered throughout the world in at least 18 countries accord-
folios (Grinblatt and Keloharju, 2009; Barber et al., 2009). ing to Kearney et al. (2011). In a product similar to the
More recently, research has noted how sensation seeking recently introduced savings lottery products in the United
and differences in sophistication have led to incentives
to engage in financial gambling (Barberis and Huang,
2
20 08; Kumar, 20 09; Liao, 2015; Andrikogiannopoulou This is an underappreciated perspective in the finance literature,
which tends to compare financial gambling to textbook investing. On the
and Papakonstantinou, 2016; Doran et al., 2012). Within
other hand, investment losses exceed gambling losses in some contexts.
this literature, some authors have studied the interaction For example, Barber et al. (2009) present a comparison of lottery losses
between traditional lotteries and trading in stocks with to individual trading losses.
JID: FINEC
States, First National Bank in South Africa implemented a grams in other states, members of participating credit
prize-linked savings program as well (studied in Cole et al., unions can open a one-year balance-building certificate of
2017). deposit (CD) with a minimum of $25 on deposit, and each
$25 contributed to the account per month qualifies the in-
2.1.1. Prize-linked savings in the United States dividual for an entry into the raffle (up to ten entries per
In the United States, lottery savings accounts first be- month). As the accounts are designed to encourage sav-
came available in Michigan in 2009 with the launch of ings, there is a $25 fee for the first early withdrawal (be-
the Save-to-Win (STW) program in select Michigan locales. fore the 12-month term), and a second withdrawal from
Subsequently, similar STW programs were introduced in the account closes the account and disqualifies the individ-
Nebraska (2012, described in detail below), North Carolina ual from winning prizes from subsequent raffles. This fee
(2013), and Washington (2013). During this period, savings structure is similar to other certificates of deposit, which
lotteries were unavailable to consumers in other states, can be liquidated in the event of an emergency, but are
largely because federal regulation against gambling pro- intended to be held to term. Beyond retaining the prin-
hibited banks from offering similar products until 2015. cipal and receiving a nominal amount of interest, each
Of these introductions of prize-linked savings, my empir- entry qualifies the accountholder for an entry into vari-
ical work focuses on the Nebraska STW rollout because ous savings raffles—a monthly credit-union-specific draw-
the casino gambling data are available before and after Ne- ing, statewide quarterly drawings, and the annual drawing
braska first offered savings lotteries, which enables a more for the grand prize.
careful empirical design. In 2012, the monthly drawings were for amounts up
Under the STW program, participating credit unions of- to $1500 (with many smaller amounts possible), and the
fer one-year certificates of deposit (CDs) in which each $25 grand prize drawing was $25,0 0 0, paid to one winner. Ac-
deposit qualifies the accountholder for an entry in monthly cording to the Doorways to Dreams Fund, nearly 1600 Ne-
raffles as well as the raffle for the annual grand prize. The braska credit union members opened STW accounts by
amount of the grand prize varies by implementation, year, the end of 2012 (out of around 20 0,0 0 0 eligible mem-
and jurisdiction, but it is generally a significant amount bers). More than a quarter of accountholders (445) won a
of money. For example, Michigan’s grand prize in 2009 prize of some amount through the program, and the to-
was $10 0,0 0 0, while Nebraska’s grand prize in 2012 was tal dollar value of prizes distributed was $51,375, which is
$25,0 0 0. STW accounts also earn a nominal rate of inter- 2.7% of the $1.9 million saved by STW accountholders over
est, but the salient feature of these accounts is the raffle the year. This effective interest rate is generous, but there
component. The fixed interest component of the accounts were notable fixed rate alternatives at the time that were
is typically below comparable products, which helps to off- competitive with this rate. For example, according to an
set the cost of administering the prize pool and marketing archived website of Mutual 1st Federal Credit Union (one
of the program. of the participating credit unions in the sample), a rewards
The participating credit unions pay an upfront fee to checking account that offered 3.4% interest APY was avail-
Doorways to Dreams to administer the drawings, and to able. Although the large prizes receive the most media at-
provide marketing and branding for the STW product (e.g., tention, the average prize amount was $115, conditional on
direct mailings to members, posters at credit union lo- receiving a prize. The average accountholder held $857 by
cations, banner advertising on credit union websites). As July 2012, $1163 by year-end 2012, and $1641 by year-end
Doorways to Dreams offered a turnkey marketing pack- 2013, indicating that there was significant and persistent
age to participating credit unions, the marketing of the ac- growth of the program in its initial couple of years.
counts was relatively standardized, and similar to the kinds
of statements made in casino advertising (e.g., emphasizing
2.1.3. Characteristics of STW accountholders
the large prize amounts, calling the entry in the savings
The stated goal of the STW program is to encourage
raffle “free money”). This advertising almost exclusively fo-
low-to-moderate income individuals to save, and to har-
cuses on the benefits of winning the raffle, and an inter-
ness household gambling demand to achieve this end. In
ested consumer needs to read the fine print to learn that
line with this motivation, recent surveys of low savings
there is a base fixed rate. Moreover, like other kinds of lot-
households indicate a strong gambling preference. For
teries, the amounts of the large prizes are salient, but the
example, according to a 2016 survey by the Consumer
odds of winning these prizes are not. Given this framing
Federation of America, “more than one-fifth of Americans
and the difficulty in organizing the required information
(21%) – 38% of those with incomes below $25,0 0 0 – think
for an apples-to-apples comparison, it is reasonable that
that winning the lottery represents the most practical
individuals would perceive and value the introduction of
way for them to accumulate several hundred thousand
these accounts as a new lottery.
dollars.” Specific to saving at credit unions, the Doorways
2.1.2. Prize-linked savings accounts in Nebraska to Dreams Fund has conducted several surveys of STW
The Nebraska STW program was launched in January accountholders about their motivations to participate
2012 by the Doorways to Dreams Fund in collaboration in savings lotteries, and about their experiences with
with nine participating credit unions.3 As with STW pro- gambling. Information from these surveys suggests that
3
A tenth credit union joined the program during 2012. For the empir- possibility that the other credit union affects additional (control) counties
ical analysis, I focus on the nine initial participating credit unions. The later in the year likely biases against finding any effect of the program.
JID: FINEC
Table 1
Details of the Nebraska Save-to-Win (STW) program.
This table summarizes the rollout of Save-to-Win savings lottery accounts in Nebraska. Panel A provides summary statis-
tics compiled from various surveys reported by the Doorways to Dreams Fund. Source: Doorways to Dreams Fund White
Papers. Panel B provides the names of the nine Save-to-Win participating credit unions, the town of their main branch,
and the counties the credit union serves.
Panel A: Save-to-Win account growth
Total amount saved Average account balance
July 2012 $1.1 million $857

January 2013 $1.9 million $1163
January 2014 $2.4 million $1641
Panel B: Save-to-Win program characteristics
Percent of accountholders who reported

... Never baving before 43%
... Less than $50 0 0 in financial assets 31%
... Inability to pay for 3 months expenses 50%
... Playing the lottery (MI 2011) 63%
... Visiting casinos/racetracks (MI 2009) 38%
Percent of Nebraska population in participating credit unions 20%
Panel C: Save-to-Win participating credit unions
Credit union name Main branch location Counties served
Centris FCU Omaha Douglas, Hall, Lincoln, Sarpy

Gallup FCU Omaha Douglas
Kearney FCU Kearney Buffalo, Dawson
KEE FCU Kearney Buffalo, Dawson
Liberty First FCU Lincoln Lancaster
MembersOwn Lincoln Gage, Lancaster
Mutual 1st FCU Omaha Cass, Douglas, Sarpy, Washington
Omaha Police FCU Omaha Cass, Dodge, Douglas, Lancaster,
Sarpy, Saunders, Washington
SAC FCU Bellevue Cass, Douglas, Sarpy, Saunders, Washington
the program appeals to low savings individuals with high havior, which suffer from well-known underreporting bi-
propensity to gamble. Panel B of Table 1 reports summary ases (Meyer et al., 2009; Li, 2012). Thus, the true amount
information on the characteristics of accountholders. For of gambling among STW accountholders is likely higher
example, 43% of Nebraska accountholders report never than reported to Doorways to Dreams.
having saved before, and a significant fraction report
having very little in savings and an inability to pay three 2.2. Credit union membership and Nebraska
months of expenses from savings.
Doorways to Dreams did not systematically survey Ne- Given that the STW program is only available through
braska STW participants about gambling behavior, and credit unions, it is important to understand the scope of
they do not report a standard set of questions across their the program relative to the population of Nebraska. The
surveys of accountholders conducted in different states. Nebraska STW program launched with the participation of
Thus, I supplement the description of the Nebraska STW nine credit unions, which serve ten of Nebraska’s 93 coun-
program with information from surveys of STW partici- ties. Panel C of Table 1 lists the credit unions that were en-
pants in Michigan to give a broad sense of the propensity listed by Doorways to Dreams to offer STW to their mem-
to gamble among STW accountholders. In general, STW bers, as well as their main branch locations and the coun-
accountholders report an abnormally high propensity to ties served by each credit union. Further, Fig. 1 presents a
gamble. In a survey conducted in 2011, 63% of Michigan map that highlights the geographic distribution of affected
accountholders report having played the lottery in the past and unaffected counties.
six months. Specific to casinos, from a separate survey con- By year-end 2012, there were over 20 0,0 0 0 members of
ducted in 2009, 38% of Michigan STW accountholders re- participating credit unions, a sizable fraction of the 1.4 mil-
ported visiting casinos or racetracks. By comparison, only lion individuals in Nebraska counties covered by my sam-
29% of households in the broadly administered consumer ple (approximately one million gambling-age adults aged
expenditure (CEX) survey report gambling expenditures of 21 and older). Because credit union membership in Ne-
any type during a year (Li, 2012). Piecing this survey ev- braska is so common, my findings should speak more di-
idence together, STW participants are at least twice as rectly to the typical resident’s experience than if the study
likely to gamble as the population at large. Although these were conducted in another state with lower credit union
comparisons give a strong indication that STW participants membership.
are frequent gamblers, the survey information from Door- More than leading to a simple substitution from one ac-
ways to Dreams comes from informal surveys of stigma be- count type to another, the introduction of prize-linked sav-
ings to Nebraska credit unions significantly increased ag-
JID: FINEC
Fig. 1. Map of treatment and control counties. This map portrays the geographic distribution of treatment and control counties used in the empirical
analysis. The treatment counties are counties served by a credit union participating in the Save-to-Win program. Within-Nebraska control counties are
Nebraska counties that were not served by a credit union participating in the Save-to-Win program. Adjacent-state control counties are counties within 20
miles of the Nebraska border. For inclusion in the sample, a county must have at least five transactions in more than ten months of the sample period.
The excluded counties do not satisfy these sample inclusion criteria.
gregate deposits at the credit unions that offered STW. To of qualifying deposits per month, the increase in engage-
quantify the increase in desposits from offering STW, I es- ment with ordinary accounts is reasonable. Although this
timate the following specification using credit union call evidence is suggestive and indirect, the aggregate increase
report data on deposits from the National Credit Union Ad- in engagement with STW credit unions is consistent with
ministration (NCUA): individual-level evidence in Cole et al. (2017), which shows
that depositors who participate in prize-linked savings
log (depositsit ) =γi + γt + β1 postt × ST Wi + it , (1)
tend to open additional accounts beyond those with the
where depositsit equals the amount of deposits at credit prize-linked savings feature.4
union i during date (year-quarter) t, STWi is an indica-
tor variable for whether credit union i offers prize-linked 2.3. The casino gambling industry
savings accounts after 2012, and postt is an indicator for
whether the date is after the January 2012 introduction of The casino gambling industry has experienced signifi-
prize-linked savings accounts. The coefficient of interest in cant growth over the past 25 years. Fig. 2 presents a time-
this specification is the difference-in-differences coefficient line for important events in the growth of the casino in-
β 1 , which quantifies how much participating credit unions dustry. As late as the 1950s, casino gambling in the United
(ST Wi = 1) increase deposits relative to non-participating States was exclusively confined to Nevada. The casino gam-
credit unions (ST Wi = 0) in the period after prize-linked bling industry grew in fits and starts throughout the 1970s
savings is made available. Standard errors are clustered by and 1980s with the start of casino gambling in Atlantic
credit union to account for serial correlation in deposits City (1978) and the construction of high-stakes bingo par-
over time (Bertrand et al., 2004). lors and Indian casinos on a number of prominent Ameri-
Table 2 presents the estimates from this specifi- can Indian reservations (Seminole (FL) 1979; Cabazon (CA)
cation, indicating that deposits increased by 5.7% at 1980; Foxwoods Casino (CT) 1986). After the Indian Gam-
STW-participating credit unions, relative to deposits in ing Regulatory Act of 1989, the casino gambling indus-
non-participating credit unions. This increase is significant, try grew dramatically in the 1990s, both from Indian casi-
amounting to $2.36 million additional deposits in absolute nos that were explicitly authorized under the Act, but also
dollar terms. Columns 3 and 4 restrict the sample to from independent authorizations of riverboat gaming, spe-
relatively large credit unions ( > $5 million in deposits), cial casino districts, and casinos across a large span of
and find quantitatively similar results (4.4% increase in de- states from the Midwest to the Gulf region.
posits, corresponding to an increase of approximately $2.4
million in deposits), which indicates that the estimated
change in deposits is not driven by small credit unions. 4
Although this evidence suggests that it is unlikely that the new de-
This aggregate increase is slightly larger than $1.9 million posits in STW accounts merely displace deposits in other accounts, it does
in deposits in STW accounts reported by Doorways to not rule out the possibility that the additional savings occurs because of
higher debt (e.g., credit card balances). This kind of substitution between
Dreams, suggesting that individuals deposited additional savings and debt is less plausible given that there is virtually no sub-
money into ordinary accounts that were not explicitly part stitution across savings accounts, but it is not possible to speak to this
of the STW program. Given the constraints on the number possibility without detailed data on credit card balances.
JID: FINEC
Fig. 2. Timeline of notable events and growth in the American casino industry. This figure provides a timeline of notable changes to the casino industry
during the recent casino era, as discussed in detail by Rose (1991). The statistics on entry and extent of the casino gambling market post-2003 are taken
from the Gambling Business Directory (Casino City Press, 2012) as computed in Cookson (2017).
Table 2
Credit union deposits after the introduction of prize-linked savings.
Using deposit information from the call reports of the National Credit Union Administration (NCUA), this table
presents results from estimating the difference-in-difference specification in Eq. (1), where the dependent variable
is the natural logarithm of depositsjq , the dollar amount of deposits at the credit union j during quarter q (for years
2010–2012). The variable postq is an indicator that equals one for dates after the introduction of Save-to-Win in
January 2012, PLSj is an indicator for whether credit union j began offering prize-linked savings in January 2012.
The lottery ticket sales data are restricted only to games that had sales before and after January 2012 (for the years
2011 and 2012). Each specification includes credit union fixed effects, and the specifications in Columns 2 and 4
include date fixed effects (year-quarter level). To mitigate concern that the change is driven by very small credit
unions, Columns 3 and 4 focus on the subsample of credit unions with $5 million or more in deposits. Standard
errors are clustered by credit union, and ∗∗∗ , ∗∗ , and ∗ indicate statistical significance at the 1%, 5%, and 10% levels.
All credit unions > $5m in deposits
(1) (2) (3) (4)
Post × # of participating CUs 0.057∗∗∗ 0.057∗∗∗ 0.044∗∗ 0.044∗∗

(0.021) (0.022) (0.021) (0.022)
Credit union FE x x x x
Year-quarter FE x x
R2 0.9985 0.9987 0.9985 0.9989

# of credit unions 72 72 53 53
# of quarters 12 12 12 12
N 831 831 614 614
From the 20 0 0s, the industry continued to expand. effect of prize-linked savings as a substitute for casino
As of August 2012, there were casinos in 41 states, and gambling.
the traditional centroid of the casino gambling industry Despite there being considerable growth in the indus-
(Nevada) only comprised approximately 20% of the industry at large, comprehensive casino industry data described
try’s casinos (and a similar percentage of the industry rev- in Cookson (2016) indicate that there was no casino entry
enues). Using an industry database, Cookson (2016) reports in the greater Nebraska region during the sample period
that there were 134 plans to enter the casino industry (2010–2012). Cookson (2016) shows that these entries into
from March 2003 through August 2012. Although not all the casino industry predominantly affect patrons who live
of these plans successfully resulted in a new casino, recent local to the casino. The fact that there is no entry in the
casino industry growth has been sizable. Casino industry Nebraska region alleviates the potential concern that the
growth has reached the point that some observers have ex- effects are driven by differential exposure to entering casi-
pressed concern that the industry is reaching a saturation nos in treatment and control counties.
point, and a number of prominent casino closings (rare for
the industry over the past 20 years) have been taken as
3. Data and measurement
evidence for this (Calvert and Kamp, 2014).
Given the geographic dispersion of casinos in the
This section describes the main data source on cash
United States, willing patrons no longer need to travel
withdrawals at casinos, and outlines the difference-in-
long distances to visit one. Consequently, casino visita-
differences empirical strategy that exploits the introduc-
tion is a part of the expenditure profile of consumers in
tion of prize-linked savings.
every state. Nebraska is no exception. Over the course
of the 26 months covered in the proprietary cash with-
drawal data set (May 2010–June 2012), Nebraska patrons 3.1. Casino cash withdrawal data
withdrew a total of $19.3 million in cash from American
casinos. The pervasiveness of casino gambling in Nebraska The data set on casino gambling activity is a proprietary
sets a quantitatively important backdrop for studying the transaction-level database of cash transactions at casinos
JID: FINEC
throughout the United States.5 For each transaction in the natural urge to under-report gambling activity. Neverthe-
database between May 2010 and June 2012, the data pro- less, it is important to understand the degree to which
vide a casino identifier, a patron identifier, a timestamp cash accessed at the casino is representative of overall
for the transaction, the type of transaction (withdrawal, gambling expenditure. According to information shared by
inquiry, deposit), the form of transaction (credit, debit, the data provider, the cash access data are representative
etc.), casino characteristics, and characteristics of the pa- of the dollars wagered at casinos in the United States, and
tron such as a gender and home ZIP code. The empirical have coverage that is stable across US geographies. The
analysis focuses on cash withdrawals by gambling patrons data come from approximately three-quarters of casinos
whose home ZIP code is in Nebraska or in counties that throughout the United States, and correspond closely to
are within 20 miles of a Nebraska county (Collard-Wexler, revenue fluctuations at the casino-month level. The data
2014). Thus, out of the 22-million-record database, this pa- provider reports that, for casinos that make their revenue
per utilizes 54,375 cash withdrawal records between May data available for validation, a single regression of monthly
2010 and June 2012. casino revenues on aggregated cash withdrawals has an
Beyond measuring overall gambling activity, the data R-squared exceeding 90%. Thus, aggregated withdrawal
also provide insight into the nature of gambling activity. amounts constitute a reasonable proxy for changes in gam-
For example, around 12% of cash withdrawal attempts re- bling activity across regions and across time, though cash
quest more funds than available in the bank account (in- withdrawals do not correspond one-for-one with casino
sufficient funds). In addition, 42% of the transactions in revenues.
the overall sample are transactions where the patron uses At the individual level, the number of patrons who ac-
a credit card to obtain cash, which is a particularly high- cess cash at casinos is 15–20% of overall visitation based
fee method of obtaining cash. Beyond the types of trans- on headcounts. Nonetheless, when the data are paired with
actions, the precise timestamp can be used to compute casino revenue and visitation data, the individuals cov-
the fraction of daytime and weekend transactions, which ered in the cash access data account for more than 50%
helps assess the balance of attributes and compositional of casino revenues. Thus, the sample is representative of
changes to the sample. Finally, the demographic character- frequent casino patrons who account for the majority of
istics of the counties covered by the sample indicate that gambling expenditures at casinos. Indeed, this observa-
the regions in Nebraska and adjacent states are quite sim- tion is consistent with the magnitudes of the summary
ilar in both overall population and average per capita in- statistics in the casino cash data. The average amount per
come, which is useful empirically. transaction was slightly larger than $750, and over the
26-month sample, the average patron withdrew approxi-
3.2. Representativeness of casino cash data mately $3,500, which implies that the average annual cash
withdrawals equal to $1615 for the patrons covered in the
Most research on gambling uses data from surveys casino cash data. As a comparison, Li (2012) reports that
in which it is likely that individuals under-report their frequent gamblers (those who report gambling expenditure
amount of gambling activity due to social stigma (Meyer on all four quarterly surveys of the CEX) have an average
et al., 2009). The most precise surveys on gambling activ- annual gambling expenditure of $633 in 2010 dollars using
ity of households, which attempt to address these well- the Consumer Price Index to account for inflation. Given
known survey biases, are two prominent gambling-specific evidence that the CEX expenditures are under-reported by
surveys that were conducted by the National Opinion Re- half relative to other surveys of gambling, the magnitude
search Center (NORC)—the Gambling Impact and Behav- of cash withdrawals for frequent gamblers is of the cor-
ior Study (1999), and the 2006 California Problem Gam- rect order of magnitude to be representative of the gam-
bling Prevalence Survey. Unfortunately, these surveys cover bling expenditure of frequent gamblers, absent survey bi-
gambling activity from outside of my sample window, and ases. This observation is consistent with the data provider’s
thus, cannot be used for the analysis. The only other large- internal calculations.
scale data set with gambling expenditure information at An analysis of a sample representative of frequent
the individual or household level is the consumer expen- casino gamblers is an interesting exercise unto itself. Af-
diture survey (CEX). As an analysis of the gambling expen- ter all, these individuals account for more than half of
diture data by Li (2012) describes in detail, the CEX likely casino gambling revenues. Thus, frequent casino gamblers
understates “the prevalence of gambling at both the exten- are an important subpopulation to understand. Beyond fre-
sive and intensive margins.” For example, the NORC sur- quent casino gamblers, it would be useful if the response
veys report that the fraction of individuals who gamble of these patrons to the introduction of PLS is similar to
is approximately 60%, whereas the analogous calculation the responses of other types of gamblers. In Section 4.2,
from the CEX reports that 29% of households gamble in a I present an analysis of scratch ticket sales, which speaks
given year, a figure that is half of the NORC surveys. to the degree to which the sample represents the behavior
By comparison to these gambling surveys, the casino of other types of gamblers. Despite the scratch ticket sam-
cash data set does not suffer from biases induced by the ple coming from a broader set of gamblers, I find a simi-
lar magnitude substitution effect to what I find using the
5
casino cash data. Thus, although frequent casino gamblers
DISCLAIMER: All information, data, reports, and other information
differ on some dimensions from other individuals, the con-
used herein was provided to the author in a proprietary, confidential, and
anonymous manner with respect to the identification of any particular clusions drawn from this sample likely extend beyond the
casino or patron. behavior of frequent gamblers.
JID: FINEC
Table 3
Summary of county-level data by treated and non-treated counties.
This table reports averages of key variables in the pre-treatment period, based on cash withdrawal data using
counties in which a participating Save-to-Win (STW) credit union operates to separate treated counties from
non-treated counties in Nebraska, as well as the difference and conditional difference. The comparisons in this
table are based entirely on within-state differences between treatment and control counties. The conditional
difference is computed as the coefficient on a treatment indicator in a regression that also controls for whether
the county has a population greater than 10 0,0 0 0. Standard errors are two-way clustered by county and month–
year. ∗∗∗ , ∗∗ , and ∗ denote statistically significant differences at the 1%, 5%, and 10% levels.
Treated Not treated Difference Conditional difference
# of Counties 10 44
# of Months 26 26
# of Observations 207 1183
... Before 159 907
... After 48 276
Pre-treatment characteristics (all control counties)
Mean transaction amount ($) 537.40 453.97 80.43 76.75
Mean # of transactions 63.85 25.27 38.58 −7.56
% Male 58.44 55.10 3.44 2.99
% Not sufficient funds 14.45 11.53 2.31∗ 1.80
% Use credit card for cash 54.96 40.18 14.78∗∗∗ 14.76∗∗∗
% Daytime transactions 34.22 35.91 −1.69 −1.31
% Weekend transactions 46.41 48.54 −2.13 −2.48∗∗
Per capita personal income ($10 0 0s) 41.54 39.91 1.63 2.47∗
Population (10 0 0s) 122.25 25.47 96.78∗∗ –
% with population > 10 0,0 0 0 30.15 4.61 25.54∗ –
3.3. Empirical approach using the distance of the patron’s home ZIP code from
branches of participating credit unions. In these specifica-
The implementation of Nebraska STW provides quasi- tions, which are presented in the online appendix, the esti-
experimental variation in exposure to savings lotteries be- mated effects using proximity to credit unions are slightly
cause credit union membership is based on whether the stronger, which matches the intuition that the treatment-
individual works or lives in a credit union’s counties. control approach should conservatively estimate the effect.
Moreover, the counties where savings lotteries were intro- My empirical work also employs a sample of observa-
duced are near state borders, which provides a natural set tions at the patron-by-{pre, post} level. As this sample is
of counties in adjacent states that can be used as control useful to condition the effects of savings lotteries on pa-
counties in the empirical analysis. To exploit this variation tron characteristics and to explore heterogeneous effects of
in exposure of savings lotteries around their 2012 intro- savings lotteries, I employ patron-level tests as a comple-
duction, I adopt a difference-in-difference strategy, which ment to the main county-month analysis. Nonetheless, be-
identifies the effect of savings lotteries on gambling by cause there are few observations per patron, patron-level
contrasting whether patrons in treated counties (where tests must be conducted on the patron-{pre, post} level
STW is available) respond differently to the introduction of rather than a more granular patron-month panel. As such,
STW than patrons in nearby control counties. the patron-level tests cannot account for trends. In this
My main sample is an aggregated county-month panel way, it is natural to conduct the main tests on county-
covering 54 Nebraska and adjacent counties across 26 month aggregates, and supplement with patron-level infor-
months from May 2010 to June 2012. To ensure the reli- mation.
ability of the aggregated gambling measure, the final sam-
ple retains month-county observations with at least five 3.4. Balance of attributes and parallel trends
transactions, and counties that have five transactions for at
least ten months of the sample period. Table 3 presents a As the geographic pattern in Fig. 1 indicates, counties
summary of observation counts for both treated and con- where prize-linked savings accounts are available are ge-
trol counties. The monthly frequency of the county-month ographically close to counties where prize-linked savings
panel enables an analysis of differential pre-trends in the accounts are unavailable, even for control counties in adja-
difference-in-difference specification. cent states.
This treatment-control structure is likely to produce a The treatment and control counties are well balanced
conservative estimate of the effects of savings lotteries. across casino gambling attributes. Table 3 presents means
This is because if nearby households in adjacent regions of pre-treatment gambling and demographic characteris-
are affected by the policy (i.e., because they work in the tics, and tests of the difference in means. The treatment
credit union’s county), their inclusion in the control coun- and control counties are well matched on the amount
ties will dampen the difference between treatment and of gambling, the number of transactions, the fraction of
control. In support of this idea, I complement my county- male patrons, the fraction of daytime transactions, and
month tests using patron-level location information, which the fraction of weekend transactions. The only gambling
enables finer measurement of exposure to savings lotteries characteristics that differ significantly between treatment
JID: FINEC
Fig. 3. Testing for differential trends in gambling by treatment and control counties. This figure presents a graphical depiction of the pre-trend in the
dependent variable logged cash withdrawals after residualizing by county and month–year fixed effects, plotted for the years 2011 and 2012 in the sample.
The difference in slopes with respect to the Date variable equals −0.189 (with a p-value of 0.514). This effect is economically insignificant as well in that
it can only explain an effect size of 0.0158 going from December 2011 to December 2012, whereas the jump is an order of magnitude larger.
and control are the propensity of patrons to use a credit that, beyond checking the conditions for the difference-in-
card for cash and to request a transaction amount for difference analysis, I also present a battery of robustness
which there is not sufficient funds. Upon conditioning tests (see Table 6), which further demonstrate the robust-
the difference-in-means test on whether the county has ness of my main results to differential pre-trends, and po-
greater than 10 0,0 0 0 residents, the only gambling char- tential time-varying confounding factors that influence the
acteristic that differs across treatment and control is the desirability of gambling.
propensity to use credit cards to obtain cash.
In comparing the demographic characteristics between 4. Main findings
treatment and control counties, population is significantly
greater in treatment counties, suggesting that the STW This section presents the main findings on substitution
program was targeted to more urban areas. Indeed, 30.15% away from casino gambling and lottery ticket sales after
of treated counties have greater than 10 0,0 0 0 residents, the introduction of Save-to-Win.
while only 4.61% of control counties do. In addition, per
capita income in treated counties is slightly greater, though 4.1. Substitution of cash withdrawals for savings lotteries
only significant when conditioning on the large population
dummy. Thus, it is important to control for these charac- The main specification exploits the fact that savings lot-
teristics when assessing whether the availability of lottery teries are available in some counties rather than others us-
savings accounts reduced the amount of gambling in treat- ing the difference-in-difference specification:
ment counties relative to control counties. Including these
controls mitigates the concern that an omitted factor re- log(1 + cashwdit ) = γi + γt + β1CUi + β2 postt + β3CU i
lated to population or income is responsible for the ob- × postt + ξ Xit + i , (2)
served effects. For this reason, all of the empirical spec-
where cashwdit is the total amount of cash withdrawn by
ifications control for population and per capita income,
patrons living in county i during month–year t,6 γ i and
though the results do not depend upon including these
γ t are county and month-year fixed effects, CUi is a count
controls.
of the number of participating credit unions in county i,7
Fig. 3 presents a plot to evaluate the parallel trends as-
sumption for the response variable—the logged amount of
cash withdrawn. There is a negligible difference between 6
There are a few county-month observations for which patrons had
the slopes of treatment versus control leading up to the cash withdrawals of zero. As log (0) is undefined, adding one enables es-
timating an approximate log specification. Because the added one is small
introduction of savings lotteries at the beginning of 2012.
relative to the typical amount of withdrawals, it has a small impact on the
Indeed, the p-value on the test that the slopes are differ- usual percentage change interpretation. Using the inverse hyperbolic sine
ent equals 0.514, and the magnitude of the slope is not transformation, which is also defined for negative values and also admits
significant relative to the jump at the introduction of sav- an approximate log (x) interpretation, the results are nearly identical.
7
ings lotteries. On this basis and the fact that treatment I also present specifications where CUi is replaced by a treatment
dummy variable that indicates whether a participating credit union serves
and control counties are well balanced, the difference-in-
county i. The main specifications use CUi because it also contains infor-
difference analysis is informative of the causal effect of mation about the intensity of treatment. The online appendix presents
savings lotteries on casino gambling. It is useful to note two other measurement schemes to capture intensity of treatment—
JID: FINEC
Table 4
The effect of the availability of prize-linked savings on gambling demand.
This table presents results from estimating the difference-in-difference specification in Eq. (2),
where the dependent variable is the natural logarithm of one plus the total amount of cash with-
drawn at casinos by individuals in county i during month–year t. In the specification, posti is an in-
dicator that equals one for dates after the introduction of Save-to-Win in January 2012, CU_treatedi
is the number of credit unions that offer STW deposits and accounts, or in some specifications,
an indicator variable for whether there is a credit union in county i that offers STW. The variable
cashwdit is winsorized at the 99th percentile to reduce sensitivities to extreme observations. The
vector of control variables X includes logged population and per capita income measures at the
county-year level from the Bureau of Economic Analysis. Standard errors are clustered by county,
and ∗∗∗ , ∗∗ , and ∗ indicate statistical significance at the 1%, 5%, and 10% levels.
(1) (2) (3) (4)
Post × # of participating CUs −0.188∗∗∗ −0.197∗∗∗

(0.047) (0.049)
# of participating CUs −0.132
(0.097)
Post × STW accounts available −0.567∗∗∗ −0.594∗∗∗
(0.173) (0.182)
STW accounts available −0.757∗∗∗
(0.270)
Month–year FE x x x x
County FE x x
R2 0.488 0.675 0.511 0.676

# of counties 54 54 54 54
# of months 26 26 26 26
N 1390 1390 1390 1390
postt is an indicator that equals one after the Nebraska indicating that choice of aggregation and technique to ac-
STW program was rolled out in January 2012, and Xit is count for serial correlation are not consequential for my
a vector of controls for demographic characteristics (logged ultimate conclusions.
population and per capita income) from the Bureau of Eco- Table 4 reports the results from estimating Eq. (2) us-
nomic Analysis. In this difference-in-difference specifica- ing Ordinary Least Squares with county-clustered standard
tion, the coefficient on the CUi × postt interaction, β 3 , is the errors. The difference-in-difference estimate is highly sig-
coefficient of interest. nificant across specifications, regardless of whether I em-
Standard errors are clustered by county to account for ploy a treatment dummy or CUi to capture the intensity of
serial correlation in the difference-in-differences analysis treatment. According to Column 2, an additional participat-
(Bertrand et al., 2004).8 As an alternative specification to ing credit union decreases casino demand in county i by
account for serial correlation, I follow the alternative rec- 17.88%,9 an effect that is statistically significant at the 1%
ommendation of Bertrand et al. (2004) to ignore time se- level. The typical difference between treatment and con-
ries information by collapsing units into pre-treatment and trol counties is stark: The treatment county experiences a
post-treatment observations. Specifically, I present an anal- reduction of 44.8% in the amount of casino gambling from
ysis of data collapsed to the patron × {pre, post} level in the introduction of Nebraska STW (Column 4). This effect
Section 5. The magnitudes and the statistical significance is significant at the 1% level and robust to the inclusion of
are consistent with one another across types of analyses, county and month–year fixed effects.
4.2. Evidence from scratch tickets

CUi /populationi on the county-month data set, and distance to credit
union (nearest branch, headquarters, or ATM) using the patron-{pre, In this section, I complement the analysis of cash with-
post} sample. The quantitative estimates are similar across measurement drawals using an analysis of scratch lottery ticket sales
schemes.
using data from the Nebraska lottery at the ZIP code-
8
For the number of counties in my analysis (N = 54), the Monte Carlo
simulations in Bertrand et al. (2004) show that clustering by county ac-
game-month level. The analysis of scratch ticket sales helps
counts for the serial correlation problem as well as the alternative of col- validate the casino cash data on three dimensions. First,
lapsing into pre-treatment and post-treatment observations (for N = 50, using scratch ticket sales data alleviates concerns about
Type I error rates of 5.3% for collapsing and 6.3% for clustering by county; representativeness because the data represent all scratch
see their Tables VI and VIII). When the number of clusters is small (N =
ticket sales, which speaks to a broader population than the
10 or smaller), ignoring time series information by collapsing the data
into county x {pre, post} observations produces more accurate Type I error casino cash data do. Second, the lottery results can help
rates, but according to Bertrand et al. (2004), the downside of aggregation confirm that the substitution effect in the previous sec-
as a solution to the serial correlation problem is that the solution suffers tion is genuine substitution across gambles, rather than a
from lower power than clustering. Consistent with the observation of low
power, I find that upon collapsing to the county x {pre, post} level, I ob-
9
tain a similar magnitude estimate (-0.142), but the statistical significance This calculation employs the exact formula provided by Wooldridge
is somewhat lower (at the 10% level). See Table 6 for details on this and (2003) for calculating the percentage effect with a logged dependent vari-
able, e f f ect = eβ − 1.
ˆ
other robustness exercises.
JID: FINEC
change in withdrawal behavior. Third, because the lottery Table 5

The effect of savings lotteries on scratch ticket lottery sales.
data are available at a longer horizon (until January 2013),
Using data on game-month-ZIP code lottery ticket sales from the Ne-
the analysis can help speak to the persistence of the effect braska Lottery, this table presents results from estimating the difference-
of STW.10 in-difference specification in Eq. (3), where the dependent variable is the
To analyze the responsiveness of scratch ticket sales to natural logarithm of salesigt , the dollar amount of scratch ticket lottery
the introduction of savings lotteries,11 I use data on lottery sales in ZIP code i, for scratch ticket game g during month–year t. In
the specification, posti is an indicator that equals one for dates after the
sales by ZIP code, month, and type of scratch lottery game
introduction of Save-to-Win in January 2012, CU_treatedi is the number
from the Nebraska Lottery. Specifically, I estimate the spec- of credit unions that offer STW deposits and accounts. The lottery ticket
ification: sales data are restricted only to games that had sales before and after Jan-
uary 2012 (for the years 2011 and 2012). Each specification includes game
l og(sal esigt ) = γi + γgt + β1CU i × postt + i , (3) by month–year fixed effects to allow for arbitrary game-specific trends
over time. The variable salesigt is winsorized at the 99th percentile to re-
where salesigt is the dollar-value of lottery ticket sales for
duce sensitivities to extreme observations. Standard errors are clustered
ZIP code i, scratch game g, and month t. As in the main by county, and ∗∗∗ , ∗∗ , and ∗ indicate statistical significance at the 1%, 5%,
specifications on cash withdrawals, CUi is the number of and 10% levels.
participating credit unions in the county (ZIP code i is
(1) (2)
mapped to its surrounding county) to capture the intensity
of treatment. The coefficient of interest is β 1 , which gives
∗∗∗
Post × # of participating CUs −0.025 −0.018∗∗∗
(0.009) (0.006)
the difference-in-difference effect of how scratch ticket
gambling in treated counties (those with credit unions par- Game × month–year FE x x
ticipating in STW) differ from the baseline control coun- County FE x
ZIP code FE x
ties. To maintain a consistent sample throughout, the sam-
ple is restricted to scratch ticket games that were offered R2 0.556 0.714
N 2006 2006
both before and after the introduction of STW in Nebraska
in January 2012, and the data run from January 2011 to
January 2013. Each specification includes game × month −
year fixed effects to allow for each scratch lottery game to very strong effect early in the year that evaporates late
have an arbitrarily different time trend, as well as county in the year (i.e., a novelty effect), the effect is slightly
(or ZIP code) fixed effects to capture differences across stronger in the second half of the year. The online ap-
geographic units. To account for serial correlation within pendix presents additional evidence that the substitution
county and over time, standard errors are clustered at the effect for casino gambling does not appear to be driven by
county level. a spike in novelty. Consistent with the evidence from the
Table 5 presents the results from estimating the longer post-period using the lottery data, the estimated
difference-in-difference specification in (3). The estimates effect on casino gambling is of a similar magnitude during
imply a 1.8–2.5% reduction in scratch lottery ticket sales the April–June post-period months as it is for the January–
for each additional participating credit union after the in- March post-period months. In the end, the results cannot
troduction of savings lotteries, and the effect is statistically rule out an intermediate-term novelty effect (lasting more
significant at the 1% level. In addition to showing this than a year), and thus, one should exercise caution when
significant average effect (pre versus post), Fig. 4 presents drawing longer-term policy implications from the findings
a leads and lags plot that also shows the dynamics of (e.g., for long-run savings rates).
the effect over the sample period. The leads and lags plot These findings suggest that the broad conclusions using
serves two purposes. First, similar to the casino cash evi- the cash access data are not primarily reflecting changes
dence in Fig. 3, it shows an absence of pre-trends for the in how patrons access cash, but rather appear to reflect
lottery data. Second, the estimates suggest that the effect a change in gambling behavior. Beyond the evidence in
of STW is relatively long-lived. Rather than exhibiting a this section, my findings in Section 6.1 that substitution
is stronger when savings lotteries are more like gambling
10
further supports the notion that the results reflect substi-
Scratch ticket games are changed over time with turnover from year
to year. As such, the most expansive post-period that includes games
tution.
offered before STW runs until January 2013. As scratch ticket sales are
only available at the ZIP-month-game level, the casino cash access data
4.3. Magnitudes and robustness
enable a more detailed analysis of the substitution effect because I can
construct measures of sophistication and similarity from the transactions
data, and examine whether savings lotteries changed the nature of gam- The findings in the main specification are not sensitive
bling. These tests—presented in Section 6—cannot be run using aggre- to using alternative sample cuts, controlling for differen-
gated lottery data. tial time trends by urban areas and high unemployment
11
The reason to focus on scratch games instead of jackpot games is
areas, or alternative techniques to estimate the effect of
that jackpot games in Nebraska (Mega Millions and Powerball) had no-
table changes in early 2012. First, Powerball doubled the initial size of savings lotteries on cash withdrawals (i.e., quantile regres-
its progressive jackpot from $20 million to $40 million in early 2012, and sion, patron-level analysis, using distance to credit union
that this led to a corresponding increase in jackpot ticket sales. Second, branches to measure treatment). The robustness tests in
the Mega Millions lottery had a record $656 million jackpot in March of this section address two classes of related concerns: (i)
2012, also boosting sales. To account for these changes to the jackpot
games, specifications in the online appendix control for jackpot lottery
uniformity of treatment (e.g., are the findings driven by a
ticket sales in the difference-in-difference specifications. The substitution particular part of the sample?), and (ii) robustness to alter-
effect becomes slightly stronger when controlling for jackpot lottery sales. native specifications, measurement, and omitted factors.
JID: FINEC
Fig. 4. Persistence of the effect of STW on scratch ticket sales. This figure presents a leads and lags graph to illustrate the difference-in-difference effect of
STW on logged scratch ticket sales. Each point in the plot is an estimated β 1τ coefficient from the following specification

log salesigt = γi + γgt + 12τ =−5 β1τ Ii{lag=τ } × CUi + i jt ,
which allows for the effect of STW to vary depending on the lead or lag (−5 months to +12 months) relative to the January 2012 introduction (t = 0 ) of
STW in Nebraska. In the underlying regression, the period used to benchmark pre-treatment effects (i.e., where all of the lead and lag indicators equal
zero) is January 2011–June 2011. The dashed horizontal lines on the plot represent the mean effect pre-treatment and post-treatment.
4.3.1. Uniformity of treatment dress this concern, I undertake a systematic subsampling

Because gambling is a potentially abnormal behavior, analysis, which is an example of the specification curve
one might worry that a few extreme and anomalous in- analysis recently proposed in Simonsohn et al. (2015).12
dividuals in the sample could drive the pattern of sub- In this spirit, I construct all subsamples across three sub-
stitution. In this case, there would be a substitution ef- sampling choices: (i) dropping one, two, or three of the
fect, but it would not be relevant to the typical gambler. I ten treatment counties, (ii) using within-state, out-of-state,
present two complementary analyses to address this point. or all control observations, and (iii) different stances on
First, I present evidence using quantile regression to eval- seasonality (all month–year observations, dropping all July
uate heterogeneity in the substitution effect across indi- through December observations). For each subsample, I es-
viduals. Second, I conduct a systematic subsampling anal- timate the effect of introduction of STW (60 0 0 specifica-
ysis, dropping treatment counties one, two, or three at a tions in total). The empirical cumulative distribution func-
time. tion of these estimates is the specification curve, which
Specifically, I estimate the difference-in-difference ef- gives a comprehensive view of how sensitive the result is
fect using quantile regression, both for the main county- to these subsampling choices.
month data set as well as the patron-{pre, post} data an- Fig. 5 plots the specification curve, plotted against the
alyzed in Section 5. Together with other robustness mea- 5th, 50th, and 95th percentiles of 500 bootstrapped spec-
sures, the quantile regression results for the 20th and 80th ification curves. Regardless of the subsample, I obtain a
percentiles are presented in Panel A of Table 6. Regardless negative point estimate, and 79.7% of these estimates are
of the percentile of the distribution considered, there is a
significant and economically meaningful effect of the intro- 12
To construct a specification curve, the researcher identifies a set of
duction of STW. For example, the smallest magnitude esti- defensible specifications over which the researcher would usually choose
mate is found using the 20th percentile of county-month a subset to present in tabular form. Rather than running a selective sam-
gambling amounts. The estimated effect is 9.7%, which is ple of these specifications, the researcher runs all of them and reports
statistically significant at the 5% level. The online appendix the distribution of coefficient estimates. The properties of the distribu-
tion of tests across specifications give a comprehensive, and ex ante, bal-
presents the full results using quantile regression.
anced view of robustness to a set of choices. The goal in my applica-
An alternative concern about sensitivity to a few obser- tion of the specification curve is to present a comprehensive view of the
vations is that the results are potentially driven by individ- robustness of the findings to subsampling choices, but the method can
uals from a few treatment counties, rather than an anoma- be applied broadly to evaluate robustness to a wider set of specification
lous part of the distribution of cash withdrawals. To ad- choices. Simonsohn et al. (2015) provide several examples of robust and
non-robust published results.
JID: FINEC
Table 6
Alternative specifications and robustness checks.
Panel A presents results from estimating alternative specifications and sample cuts. Each entry in this table reports a separate robustness check in which
month-year and county fixed effects are employed. In the case of patron-level specifications, ZIP code fixed effects are employed. The coefficient reported
for each test is the coefficient estimate on the post × # o f part icipat ing CUs term. Standard errors are clustered by county. Full regression detail is provided
in the online appendix. Panel B presents summary information on the systematic subsampling analysis (specification curve) in which I estimate the model
on all subsamples that (i) exclude one, two, or three of the ten treatment counties, (ii) drop within-state control counties, drop adjacent-state control
counties, or include all control counties, (iii) drop data from months July–December, or retain these dates (accounting for seasonality, or not). There are
60 0 0 such subsamples. To evaluate statistical significance, I follow the bootstrapping procedure in Simonsohn et al. (2015) with 500 replications to generate
500 resampled specification curves. ∗∗∗ , ∗∗ , and ∗ indicate statistical significance at the1%, 5%, and 10% levels.
Panel A: Summary of alternative specifications and robustness exercises
Robustness test Estimated effect Robustness test Estimated effect
Alternative sample cuts Controlling for differential time trends

Main specification −0.197∗∗ Difference relative to 2011 trend −0.104∗∗
(0.049) (0.049)
Within-Nebraska controls only −0.135∗∗ Differential trend by > 50K residents −0.133∗∗
(0.062) (0.058)
Adjacent-to-Nebraska controls only −0.185∗∗∗ Differential trend by > 100K residents −0.164∗∗
(0.055) (0.064)
January–June Obs (seasonality) −0.137∗∗∗ Differential trend by > median unemp. −0.195∗∗
(0.049) (0.058)
April–June post period (persistence) −0.194∗∗∗ Differential trend by > 90th pctile unemp. −0.176∗∗
(0.056) (0.050)
Quantile regression specifications Other controls and measurement
20th pctile on county-month data −0.097∗∗ Controlling for Jackpot Lottery Sales −0.204∗∗
(0.045) (0.048)
80th pctile on county-month data −0.273∗∗ (−1)∗ Log distance to branch (patron-level) −0.148∗∗
(0.048) (0.051)
20th pctile on patron-level data −0.148∗∗∗ Credit unions per capita (Z) −0.143∗∗∗
(0.032) (0.053)
80th pctile on patron-level data −0.187∗∗∗ Patron × {pre, post} −0.156∗∗∗
(0.034) (0.055)
County × {pre, post} −0.143∗
(0.082)
Inverse hyperbolic sine of cashwd as dependent variable −0.199∗∗∗
(0.047)
Panel B: Robustness to systematic subsampling (specification curve)
Percentiles of the specification curve
% of Significant results (α = 0.10) 5th 25th 50th 75th 95th
Characteristics of specification curve

Estimate 0.797∗∗ −0.225∗ −0.189∗∗∗ −0.148∗∗∗ −0.120∗∗∗ −0.076∗∗∗
p-value (0.012) (0.064) (0.0 0 0) (0.004) (0.004) (0.008)
Bootstrap summary statistics
Mean 0.110 −0.070 −0.030 0.003 0.035 0.074
Median 0.028 −0.064 −0.028 0.004 0.034 0.072
Std Dev 0.166 0.067 0.057 0.053 0.054 0.059
statistically significant at the 10% level. Using the boot- nomenon throughout the sample. It does not appear that
strap procedure suggested by Simonsohn et al. (2015), only the effects I document reflect abnormal or anomalous be-
1.2% of the bootstrap specification curves yield higher rates havior of a few individuals or regions.
of statistical significance, indicating that it is unlikely that
the degree of uniformity of the effect would appear by 4.3.2. Additional robustness and quantifying magnitudes
chance. Panel B of Table 6 summarizes the distribution of I also present a series of tests to evaluate whether the
the effects across the 5th, 25th, 50th, 75th, and 95th per- results are sensitive to other specification choices about
centiles of the specification curve. The magnitude of the measurement, aggregation, or how to account for pre-
estimated effects across these percentiles ranges from ap- trends. Panel A of Table 6 summarizes this set of ro-
proximately −7.6% to −22.5%, and all of these percentiles bustness exercises, presenting estimates for the difference-
are significantly significant according to the Simonsohn in-difference effect across each of the alternative specifi-
et al. (2015) bootstrap procedure. cations I have considered. The online appendix presents
Taken together, these results indicate that substitution more detail on each of these tests. Across specifications, I
between gambling and savings lotteries is a pervasive phe- find savings lotteries significantly reduce casino cash with-
JID: FINEC
Fig. 5. Specification curve analysis for the effect of prize-linked savings. This figure presents the empirical cumulative distribution function (ECDF) for
coefficient estimates across 6,0 0 0 subsamples that (i) drop one, two, or three out of ten treatment counties, (ii) use either all control observations, within-
state control counties only, or adjacent state control counties only, and (iii) use all year–month observations, or drop July–December observations to account
for seasonality. The plotted ECDF is compared with 5th, 50th, and 95th percentiles of bootstrapped specification curves from a bootstrapping procedure
suggested by Simonsohn et al. (2015) that bootstraps the specification curve itself. The dashed lines represent 90% confidence bands from the bootstrapping
procedure.
drawals, with a percentage effect that ranges from 9.2% to dramatically larger than the substitution effect shown in
23.9%. For the average control county with 2.4 participat- the lottery results (1.8–2.5%), which reflect substitution
ing credit unions, the substitution effect ranges from 20.8% among all lottery gamblers. Given that the casino cash
to 48.1%.,13 data reflect the behavior of frequent gamblers (see discus-
These estimated magnitudes are large as a percentage sion in Section 3.2), these somewhat larger substitution
of cash withdrawals, but are moderate relative to overall magnitudes are plausible.
gambling. To quantify the effect on overall gambling, note In addition, Table 7 presents several back-of-the-
that there is approximately $45 billion annually wagered envelope calculations for the substitution effect in terms
at casinos in the data set, and approximately $10 bil- of dollar values, giving reasonable magnitudes from that
lion withdrawn annually in the data.14 Thus, each dollar standpoint as well. For example, multiplying the percent-
withdrawn corresponds to approximately $4.50 of overall age reduction by the average withdrawal amount over
gambling. If gamblers reduce gambling by first reducing six months ($794.86), we obtain dollar-for-dollar substi-
cash withdrawals, and subsequently wagering less outside tution effects. By this calculation, the conservative esti-
money, the percentage decline in cash withdrawals is 4.5 mate gives an estimated $132.64 reduction in cash with-
times larger than the percentage decline in gambling. drawals. To compute the aggregate substitution effect, mul-
Using this link between withdrawals and gambling, the tiply the average dollar-value effects by the number of
95th percentile of the specification curve corresponds treated patrons (1323 patrons from treated counties made
to a 3.71% decline in casino gambling. This substitution withdrawals in the post-period). In aggregate, this calcula-
effect is similar to the percentage increase in deposits at tion implies a substitution away from casino gambling of
Save-to-Win credit unions relative to non-participating $175,689 on the conservative side, and $396,443 from the
credit unions (4.4–5.7%, Table 2). In addition, the sub- main specification. In comparison with the approximately
stitution effect captured in the casino cash data is not $2 million saved in participating credit unions over 2012
(both from the STW white papers, and from the credit
union deposits regression evidence), these are reasonable
13
For the estimated β 3 from Panel A of Table 6, this calcula- magnitudes.
tion employs the exact formula provided by Wooldridge (2003) for Beyond these comparisons, there are at least two rea-
calculating the percentage effect with a logged dependent variable,
sons to expect a large effect relative to the relatively small
e f f ect = eβ − 1.
ˆ
14
The $45 billion figure is reported as an aggregate summary statistic
number of prize-linked savings accounts opened. First,
from the casino cash data provider. treatment is not necessarily the opening of an account, but
JID: FINEC
Table 7
Relating cash-withdrawal effects to gambling demand.
This table presents back-of-the-envelope calculations for several scenarios to relate the documented substitution effect on cash withdrawals to dollar value
and percentage substitution effects on overall casino gambling. The “Effect for average patron” case is constructed using the estimated treatment effect
.73
from Table 3, Column 4, and the $794.86 average amount withdrawn every six months by patrons in the sample (794.86 = 3448 26
× 6). The “Conservative
effect for average patron” case is constructed using the 95th percentile of the specification curve, summarized in Panel B of Table 6, which equals −7.6 log
points, scaled by the average number of credit unions in treatment counties (2.402), and then applying the percentage effect formula (exp{β} ˆ − 1). Finally,
the effect from the median patron is taken by computing the percentage effect from the median regression (reported in the online appendix), and applying
this effect to the amount of cash withdrawn by the median patron over six months ($184.56, computed from the summary statistics). The “% of Casino
gambling” is computed assuming that initially $3.5 of cash are brought per one dollar of cash accessed, a figure that matches aggregate casino revenue
and withdrawal statistics. All the calculations in this table maintain the assumption that the entire dollar-value substitution effect comes from a decrease
in withdrawals (and not a decrease in the amount of cash brought to the casino).
Per credit union effect (%) % of cash withdrawn % of gambling $ Withdrawn (per 6 mo.) Effect in $
Average patron −17.9 −37.7 −9.28 794.86 −299.66

Average patron (conservative) −7.3 −16.7 −3.71 794.86 −132.64
Median patron −22.7 −46.1 −10.24 184.56 −85.01
the option to open an account. Even for individuals who 5. Patron-level evidence
plan to use this option later, the introduction of PLS could
have led the individual to substitute away from gambling. The main empirical tests employ county-month aggre-
Second, gambling is a social activity in which peer effects gated measures for which there is enough data to use
are likely. Encouraging one individual to forsake gambling the transaction data to reliably represent casino demand
is likely to lead to a reduction of gambling by that individ- for that county. A potential concern is that this sample
ual’s friends, amplifying the original effect. construction could bias the estimated substitution effect.
To address this possibility, I analyze data at the patron-
by-{pre, post} level using all patrons in the region. This
data set has two observations per patron, one observa-
4.4. Evidence on the nature of gambling tion aggregated over the patron’s pre-treatment transac-
tions, and the other aggregated over post-treatment trans-
Using the detail on the frequency and timing of transactions. When aggregating the data in this way, approxi-
actions from the cash access data, I deepen the analysis mately 30% patrons make transactions both before and af-
by examining how the introduction of savings lotteries after the introduction of savings lotteries in January 2012.
fected the nature of gambling. Specifically, I can evaluate Using this patron-level data set, I estimate a difference-in-
whether the decline in gambling was due to fewer transac- difference specification that is analogous to the difference-
tions to the casino, or withdrawing less money per trans- in-difference specification for the county-month panel:
action. In addition, I can examine whether the nature of
Ykt = γz∗ + β1∗ postt + β2∗CU _treatedi × postt + γ ∗ Xi + kt∗ ,
casino gambling changes along other dimensions—timing
during the day (morning, afternoon, evening), timing dur- (4)
ing the week, types of transactions (credit card, debit card, where the outcome variable Ykt reflects one of three out-
mistakes), and for a subsample of observations, gender. To comes: (i) the logged amount withdrawn by patron k in
evaluate these potential changes to the nature of gambling, period t ∈ {pre, post}, (ii) an indicator for whether patron
I estimate specifications analogous to Eq. (2), separately k had positive withdrawals in post-treatment period, and
using different outcomes related to the nature of gambling (iii) the logged amount of cash access fees paid by pa-
as response variables. tron k in period t ∈ {pre, post}. As in the county-month data
Panel A of Table 8 illustrates how casino gambling be- specifications, the coefficient of interest is the difference-
havior changed after the introduction of savings lotteries. in-difference coefficient, β2∗ , which captures the effect of
The overall reduction in demand comes most robustly from being treated by greater exposure to participating credit
a reduction in the number of transactions. Average trans- unions that offer savings lotteries. To evaluate the inten-
action size declines by approximately the same magnitude sive margin of gambling, I focus on the persistent sample
as the number of transactions, but the effect is estimated of patrons who are observed in both the pre-period and
imprecisely. These findings hint at extensive versus inten- the post-period. To the extent that treatment reduces the
sive margin effects, which are more readily analyzed in the propensity of the patron to gamble at all, these estimates
patron-{pre, post} sample. On the nature of gambling, the understate the true effect.
frequency of insufficient funds transactions and frequency Columns 1 and 2 of Table 9 present the results from
of credit-card-for-cash transactions do not exhibit statisti- estimating Eq. (4) with logged cash withdrawn as the de-
cally significant changes. Although females appear to sub- pendent variable. According to these specifications, an ad-
stitute away from casino gambling more than males (ow- ditional participating credit union reduces the amount of
ing to the slight rise in %Male in the post-period), Panel gambling in the post-period by 14.4%. This is a similar es-
B of Table 8 indicates similar time-of-day and day-of-week timate to the estimate from the county-month aggregated
patterns of gambling. In sum, the introduction of savings sample, and the effects are significant at the 1% level using
lotteries influences the amount of gambling (somewhat standard errors clustered by county.
more robustly through visitation), but it does not appear To evaluate the extensive margin of gambling, I focus
to influence the qualitative nature of gambling. on the sample of patrons who were observed in the
JID: FINEC
Table 8
The effect of savings lotteries on the extent and nature of gambling.
This table presents results from estimating the difference-in-difference specification in Eq. (2), where the dependent vari-
able is either a different measure of gambling activity or a patron characteristic. In Panel A, the dependent variable is a
type of observed gambling activity—either the logged number of transactions, the logged average transaction size (in dol-
lars), the fraction of transactions where the patron had insufficient funds for the cash withdrawal (%NSF), or the fraction of
transactions where the patron used a credit card for cash in county i and month–year t. In Panel B, the dependent variable
is a characteristic of the patron or timing of gambling—either the fraction of weekend transactions, the fraction of daytime
transactions, or the fraction of transactions by males in county i and month–year t. As for the right-hand side variables,
posti is an indicator that equals one for dates after the introduction of Save-to-Win in January 2012, CU_treatedi is the
number of credit unions that offer STW deposits and accounts, or in some specifications, an indicator variable for whether
there is a credit union in county i that offers STW. The vector of control variables X includes logged population and per
capita income measures at the county-year level from the Bureau of Economic Analysis. Standard errors are clustered by
county, and ∗∗∗ , ∗∗ , and ∗ indicate statistical significance at the 1%, 5%, and 10% levels.
Panel A: Changes in gambling behavior
Log(#t ransact ions ) Log(transsize) % NSF % credit card
Post × # of participating CUs −0.089∗∗∗ −0.117 0.004 0.011

(0.024) (0.072) (0.006) (0.010)
Month–year FE x x x x
County FE x x x x
R2 0.836 0.352 0.193 0.447

# of counties 54 54 54 54
# of months 26 26 26 26
N 1390 1390 1390 1390
Panel B: Changes in sample composition of patrons
% daytime % weekend % male
Post × # of participating CUs 0.0 0 0 −0.008 0.038∗

(0.005) (0.005) (0.023)
Month–year FE x x x
County FE x x x
R2 0.088 0.112 0.339

# of counties 54 54 54
# of months 26 26 26
N 1390 1390 1091
Table 9
The effect of savings lotteries on gambling, propensity to gamble, and cash access fees.
Each observation in this table is a patron × (before, after) treatment. This table presents results from estimating the
difference-in-difference specification in Eq. (4) where the dependent variable is either one plus the total amount of
cash withdrawn at casinos by the patron k during treatment period t (transactions are aggregated before versus after
treatment) or an indicator for whether the patron had any withdrawals during the treatment period t, γ z are ZIP code
fixed effects, posti equals one post-treatment observations that were aggregated for transactions after the introduction
of Save-to-Win in January 2012, CU_treatedi is the number of credit unions that offer STW deposits and accounts, or
in some specifications, an indicator variable for whether there is a credit union in county i that offers STW. For each
specification, the cashwdk variable is winsorized at the 99th percentile to reduce sensitivity to extreme observations.
The estimates for the logged withdrawal amount are constructed using the persistent subsample, which contains only
patron × treatment for which I observe pre- and post-observations. The estimates for the indicator for no withdrawals
are constructed using a sample of patron transactions for which the patron was observed in the pre-period. The vector
of control variables X includes logged population and per capita income measures at the county-year level from the
Bureau of Economic Analysis interacted with the post-period. Standard errors are clustered by county, and ∗∗∗ , ∗∗ , and ∗
indicate statistical significance at the 1%, 5%, and 10% levels.
Log(cash withdrawn) No withdrawals dummy Log(fees)
(1) (2) (3) (4) (5) (6)
Post × # of participating CUs −0.156∗∗∗ 0.036∗∗∗ 0.007

(0.055) (0.007) (0.016)
Post × STW accounts available −0.674∗∗∗ 0.154∗∗∗ 0.004
(0.223) (0.026) (0.074)
ZIP code FE x x x x x x
2
R 0.149 0.150 0.479 0.479 0.404 0.404
# of ZIP Codes 482 482 654 654 482 482
N 7262 7262 18730 18730 7262 7262
JID: FINEC
pre-period and use the no_gamblingkt indicator as the Table 10

The effect of savings lotteries is greater when savings lotteries and casino
dependent variable. In this context, the difference-in-
gambling are similar.
difference estimate reflects the differential propensity to This table presents results from estimating the main difference-in-
avoid gambling in the post-period, given that the pa- difference specification in Eq. (2), but aggregated among similar trans-
tron was observed to withdraw cash in the pre-period. actions versus dissimilar transactions for three similar/not cuts of the
Columns 3 and 4 of Table 9 present the results from this sample of transactions. For each specification, the cashdit variable is win-
sorized at the 99th percentile within each subsample. For the construc-
estimation. These estimates highlight that the introduction
tion of subsamples, (i) close is the sample of transactions by patrons
of savings lotteries was important to the extensive margin within 120 miles of their home ZIP code, while far is the sample of trans-
of gambling as well. An additional participating credit actions where the patron is more than 120 from his home ZIP code, (ii)
union increases the probability of not gambling in the short time until lottery is for transactions that occur after the 21st of the
month, while long time until lottery are transactions that occur before
post-period by 3.6 percentage points. In aggregate, being
the 7th of each month, and (iii) casinos with/without nightlife are trans-
in a county that has at least one participating credit union actions that occur at casinos that do / do not offer nightlife. The vector
reduces the propensity to gamble in the post-period sig- of control variables X includes logged population and per capita income
nificantly. Patrons in affected counties are 15.4 percentage measures at the county-year level from the Bureau of Economic Analysis.
points more likely to not gamble at all in the post-period. Full results are reported in the online appendix. Standard errors are clus-
tered by county, and ∗∗∗ , ∗∗ , and ∗ indicate statistical significance at the
It is also informative to analyze the effect of the in-
1%, 5%, and 10% levels.
troduction of savings lotteries on the amount of cash ac-
cess fees paid because it indirectly indicates the amount of Sample split Post × # of
participating CUs
sophistication among those who substitute. The results in
Columns 5 and 6 show that there is no change in the fees Similar
Close transactions (within 120 miles) −0.222∗∗
paid by affected patrons relative to unaffected patrons. This
(0.101)
finding suggests that the customers who substitute away Short time until lottery (week 4 transactions) −0.239∗∗∗
from casino gambling are those who pay minimal or no (0.083)
fees to access cash. That is, the substitution effect appears Casinos without nightlife −0.291∗∗∗
to most strongly affect individuals who exhibit more so- (0.098)
Differentiated
phisticated cash-access behavior. This finding foreshadows
Far transactions (outside of 120 miles) −0.063
the analysis of the heterogeneity in how patrons respond (0.086)
to the introduction of savings lotteries. Long time until lottery (week 1 transactions) −0.157
(0.113)
Casinos with nightlife 0.044
6. Heterogeneity
(0.098)
This section provides greater insight into why con-

sumers substitute from casino gambling by examining het-
erogeneous effects. The findings are consistent with gen- 6.1.1. Local versus destination gambling
uine substitution that is stronger when the gambling prod- According to the substitution-among-similar products
ucts are more similar, giving a strong indication that the hypothesis, savings lotteries should be more substitutable
substitution effect reflects gambling preferences. In addi- with local casino activity if destination gambling involves
tion, the pattern of substitution across different levels of other elements of entertainment, such as visiting a new
patron self-control is consistent with a leading behavioral area, whereas gambling at local casinos is more easily
model of casino gambling (Barberis, 2012), further validat- substituted using the gambling feature of savings lotter-
ing the empirical findings by grounding them in theory. ies. Consistent with substitution among similar attributes,
Table 10 demonstrates that close transactions (within 120
6.1. Substitution across similar gambles miles) are much more responsive to the introduction of
savings lotteries than far transactions (farther than 120
If consumers substitute between savings lotteries and miles). An additional STW-participating credit union re-
casino gambling on account of a gambling preference, the duces the amount of local casino gambling by approxi-
substitution effect should be stronger when savings lotter- mately 20% for patrons in affected counties, but for trans-
ies are more similar to casino gambling, ex ante. Using in- actions far away from the patron’s home ZIP code, the in-
formation on casino attributes and the location and tim- troduction of savings lotteries does not seem to matter.
ing of transactions, this section presents three subsampling The finding that local gambling is more substitutable
tests that show stronger substitution among similar gam- with savings lotteries suggests a sensible pattern of substi-
bles. Notably, the substitution effect is (i) stronger for lo- tution across gambling products. In theory, a similar effect
cal gambling than for destination gambling, (ii) stronger could arise mechanically if gambling trips are planned in
for more immediate lottery payoffs than those more dis- advance and locked in, rather than being about attribute
tant in time, and (iii) stronger for low-frills casinos than substitution as I argue. If the effect is mechanical due
casinos with nightlife. These findings on heterogeneity are to locked-in plans, the substitution effect will eventu-
consistent with substitution among similar products, and it ally show up in far regions later in the sample when
cannot be jointly explained by a broad attention-grabbing patrons are not locked into their destination trip. The
effect (Barber and Odean, 2008), the effects of advertising online appendix addresses this possibility by splitting
(Becker and Murphy, 1993), nor a blanket commitment to the post-period into two subsets: the first three months
spend less. and the last three months. In these specifications, there
JID: FINEC
is no significant substitution effect in any part of the exhibit no statistically significant change in withdrawals
post-period for far transactions, a finding that supports post-introduction.
the attribute-substitution interpretation. In fact, the statis- Taken together, these results on the similarity of casino
tically insignificant effects for destination gambling appear gambling and savings lotteries are difficult to reconcile
to diminish even further in the last three months of the with rationales unrelated to the gambling feature of the
post-period for non-local gambling. STW product. In particular, if the accounts merely grabbed
investor attention, there would be no reason to expect sim-
6.1.2. Immediacy of payoff and substitution with casino ilar gambles to substitute more strongly. Similarly, it would
gambling be difficult to explain the stronger substitutability among
The timing of monthly drawings also provides useful similar gambles if the accounts were simply attractive for
variation in how substitutable the savings lotteries are their non-gambling features (e.g., high expected interest
with casino gambling. As much of the thrill of casino gam- relative to other accounts). The findings are also inconsis-
bling is the immediate payoff, lotteries with more im- tent with savings lotteries presenting merely a broad com-
mediate payoffs should intuitively exhibit greater substi- mitment to spend less on unplanned expenditures. Speak-
tutability with casino gambling. Because Nebraska STW has ing to this alternative, my finding of no substitution away
monthly drawings for which accountholders qualify with a from casinos with nightlife is most relevant. Expenditures
deposit during that month, payoffs from the savings lottery are just as likely to be unplanned at casinos with nightlife
become more immediate closer to the end of the month. versus without nightlife, but casinos without nightlife are
According to this logic, the substitution effect should be less differentiated from gambling in savings lotteries. Thus,
magnified later in the month relative to earlier in the the findings are most naturally explained by consumers
month. substituting across products that are less differentiated in
Table 10 presents results from estimating Eq. (2), where attribute space.
the dependent variable only aggregates cash withdrawals
over part of the month. Consistent with the close-versus- 6.2. Substitution and self-control
far results, the time-until-lottery results indicate a greater
effect on casino cash withdrawals that occur when savings The Barberis (2012) model of casino gambling implies
lotteries are more substitutable with casino gambling. For that individuals with high self-control tend to substitute
example, the reduction in week-one gambling due to an more strongly away from casino gambling under competi-
additional credit union is 15.7 log points and statistically tion from lotteries than low self-control individuals. In the
insignificant, while the reduction in week-four gambling Barberis (2012) model, individuals with high self-control
(22nd and after) is 23.9 log points and significant at the who gamble at casinos can commit to a stopping rule at
1% level. Moreover, the effect becomes more statistically which to leave the casino, but low self-control individu-
significant and generally larger in magnitude as the time als cannot. As a result, the casino gambling payoff pro-
period shifts from week one to week four. file for high self-control individuals looks similar to a lot-
This pattern of results is consistent with the idea that tery payoff (right-skewed with a few large positive out-
the immediacy of payoff is an important factor behind liers). By contrast, low self-control gamblers in the Barberis
casino demand, which suggests that there is more to the (2012) model cannot commit to stopping while ahead,
substitution away from casino gambling than offering an and thus do not receive a lottery-like payoff profile from
attention-grabbing alternative or merely advertising the gambling at a casino (eventually losing their maximum
STW programs to members of the credit union. stake). Based on the similarity of their payoff profiles, lot-
teries and casino gambling will appear similar to high self-
6.1.3. Product differentiation: casinos with and without control individuals, but not to low self-control individuals.
nightlife In light of the Barberis (2012) model intuition, find-
Beyond information about transactions, I also utilize in- ing that savings lotteries substitute for casino gambling al-
formation about casino attributes. If the effects are due ready provides indirect evidence that gamblers, on average,
to attribute substitution, transactions at casinos that are exhibit some degree of self-control. Yet, beyond an aver-
differentiated from pure gambling should be less affected age effect, it would be reassuring from the standpoint of
than transactions at relatively unadorned casinos. One di- the Barberis (2012) model to find that relatively low self-
mension that distinguishes casinos from each other is control gamblers exhibit weaker substitution between sav-
whether the casino has nightlife (i.e., a bar, dance club, ings lotteries and casino gambling. To speak to this point,
etc.). the transaction-level detail in the cash access data provide
We should expect stronger substitution away from direct insight into gambling patrons’ self-control based on
transactions at casinos without nightlife because those the nature of their cash access behavior. For example, a pa-
casinos are less differentiated from the experience of a tron likely lacks self-control if he or she uses a credit card
savings lottery. Table 10 presents the findings separately to obtain cash at the casino, incurring the high cash ad-
for transactions at casinos with nightlife and for transac- vance fees. Similarly, a patron who—at the casino ATM—
tions at non-nightlife casinos. Consistent with the other requests funds in excess of the account balance (insuffi-
attribute-based substitution findings, I find that the substi- cient funds) likely has low self-control.
tution effect is particularly strong for transactions that oc- As both the frequency of credit card use (to obtain cash
cur at casinos without nightlife (25.2% for each additional at casinos) and insufficient funds transactions are behav-
participating credit union), whereas casinos with nightlife iors related to low self-control, I employ the frequency of
JID: FINEC
Table 11 the effect is robust to controlling for a triple-difference

Heterogeneity by self-control and education.
term allowing for heterogeneity by the employment-to-
The results in this table come from estimating the triple-difference spec-
ification modification of Eq. (2), where the difference-in-difference effect population ratio in the county (emp_to_ pop × postt × CUi ).
is interacted with a characteristic of interest, characteristici . In Panel A, In these richer econometric specifications, indicators of
characteristici is one of two indicators of short-run patron self-control: (1) greater sophistication continue to amplify the substitution
fraction of credit card transactions in the county-month, and (2) fraction effect of savings lotteries beyond the terms that control for
of insufficient funds transactions. In Panel B, characteristici is one of two
relatively stronger local economies. Indeed, the findings on
indicators of educational attainment from the US Census: (1) fraction of
county residents who attended at least some college in 2010, and (2) frac- sophistication are quite robust in magnitude and signifi-
tion of county residents with a high school diploma in 2010. The vector cance (see the online appendix for details).15 Moreover, the
of control variables X includes logged population and per capita income stronger effect among high self-control individuals is con-
measures at the county-year level from the Bureau of Economic Analy-
sistent with my finding of a negligible substitution effect
sis, as well as interactions between the effect and the employment-to-
population ratio. For ease of interpretation, this table reports separately
on fees in Section 5. The individuals who substitute away
the estimated difference-in-difference effect for a high level of these char- from the casino pay little or no cash access fees, indicative
acteristics (+1 sd) versus a low level of these characteristics (−1 sd). Full of greater self-control. Taken together, these results indi-
regression results—including estimates for interaction terms—are reported cate that patrons who have relatively more self-control–as
in the online appendix. Standard errors are clustered by county, and ∗∗∗ ,
∗∗ exhibited by their cash access behavior at the casino—tend
, and ∗ indicate statistical significance at the 1%, 5%, and 10% levels.
to substitute more strongly away from casino gambling.
Panel A: Differential effects by proxies for self-control In addition, the heterogeneity across proxies for self-
Characteristic Post × # of
control does not reflect differences in educational attain-
participating CUs ment, broadly measured. Panel B of Table 11 reports how
higher education interacts with the substitution effect be-
High self-control
Infrequent use of credit card for cash −0.329∗∗ tween PLS and gambling. These specifications indicate that
(0.076) educational levels are largely unrelated to the substitu-
Infrequently requesting unavailable funds −0.353∗∗∗ tion effect. Beyond the specifications included in the on-
(0.066) line appendix, the findings on heterogeneity by self-control
Low self-control are robust even after including triple-difference terms for
Frequent use of credit card for cash −0.017
educational attainment. Viewed through the lens of re-
(0.076)
Frequently requesting unavailable funds −0.039 cent work showing that education promotes financial lit-
(0.066) eracy (Cole et al., 2014; Brown et al., 2016), these findings
suggest that self-control might not be influenced by gen-
Panel B: Differential effects by education
eral education or even financial education, which can work
Characteristic Post × # of through other channels to affect financial behavior (e.g.,
participating CUs better numeracy or general financial knowledge). Rather, as
High education a personal characteristic, self-control and its effects appear
Large % attending college −0.186∗∗ to be relatively unrelated to educational attainment.
(0.082) Finally, one interpretation of the self-control proxies is
Large % high school graduates −0.170∗
that they may broadly proxy for the generic quality of fi-
(0.102)
Low education nancial behaviors. Intuitively, individuals who exhibit bet-
Small % attending college −0.210∗∗ ter financial behavior ought to respond to the introduc-
(0.082) tion of a reasonable gamble.16 From the standpoint of par-
Small % high school graduates −0.170∗ simony, the self-control view not only provides a ratio-
(0.102)
nale for the heterogeneity in the response, but also, it pro-
vides clear intuition for why savings lotteries would sub-
stitute casino gambling in the first place. The view that
these behaviors as proxies for low self-control in my em- the findings are driven by generic quality of financial be-
pirical tests. Specifically, I estimate triple-difference spec-
ifications that interact the effect of savings lotteries with
15
these self-control proxies to allow the substitution effect Interpreting the coefficient estimate on the employment-to-
population ratio from the online appendix, I find that regions with
to differ by high self-control versus low self-control. Based
better economic conditions exhibit stronger substitution due to the
on these triple-difference specifications, Panel A of Table 11 introduction of savings lotteries. This finding complements the main
reports the estimated substitution effect for high self- finding in that it highlights that the effects of savings lotteries are
control versus low self-control individuals. Consistent with less pronounced among financially sensitive individuals (low financial
the Barberis (2012) model, high self-control individuals re- sophistication, worse economic conditions).
16
For example, individuals who do not frequently request unavailable
spond more to the introduction of savings lotteries than
funds may also have greater awareness of financial alternatives, which
low self-control individuals, and the difference is striking. may also contribute to the heterogeneous effects. Unlike self-control in
The substitution effect ranges from -21.9% to -29.7% for the Barberis (2012) model, financial awareness has potentially an am-
high self-control individuals, whereas the effect is an in- biguous relationship with the substitution effect. On one hand, high-
significant −1.7% to −3.8% for low self-control individuals. awareness patrons may see that savings lotteries and casino gambling are
kindred ways to gamble, whereas low-awareness gamblers may not see
This heterogeneity does not arise because of differ- the connection. On the other hand, low-awareness gamblers could over-
ing economic conditions because the base specification react to the savings lottery by mistaking how much of a gamble has been
accounts for population and per capita income, and staked (just the interest, not the whole deposit).
JID: FINEC
havior does not provide intuition for the substitution ef- Behrman, J.R., Mitchell, O.S., Soo, C.K., Bravo, D., 2012. How financial lit-
fect, and would need a separate motivation for why the eracy affects household wealth accumulation. Am. Econ. Rev. Papers
Proc. 102 (3), 300–304.
two types of gambling are substitutes. Thus, although some Bertrand, M., Duflo, E., Mullainathan, S., 2004. How much should we trust
of the heterogeneous effects can be explained by better differences-in-differences estimates? Q. J. Econ. 119 (1), 249–275.
quality financial behavior, self-control is a more parsimo- Beshears, J., Choi, J., Laibson, D., Madrian, B.C., Milkman, K.L., 2015. The
effect of providing peer information on retirement savings decisions.
nious and unifying explanation for the substitution effect J. Finance 70 (3), 1161–1201.
between savings lotteries and casino gambling. Boyer, B.H., Vorkink, K., 2014. Stock options as lotteries. J. Finance 69 (4),
1485–1527.
7. Conclusion Brown, D.C., Cederburg, S., O’Doherty, M.S., 2017. Tax uncertainty and re-
tirement savings diversification. J. Financ. Econ. 126, 689–712.
Brown, M., Grigsby, J., van der Klaauw, W., Wen, J., Zafar, B., 2016. Finan-
Saving whilst gambling is a promising idea to influ- cial education and the debt behavior of the young. Rev. Financ. Stud.
ence household savings rates (Tufano, 2008). This paper 29 (9), 2490–2522.
Calvert, S., Kamp, J., 2014. Casino boom pinches Northeast-
introduces new transaction-level data on casino cash with-
ern states. Wall Street J https://www.wsj.com/articles/
drawals to study how those affected by the introduction casino- boom- pinches- northeastern- states- 1403220428.
of PLS change their gambling behavior. Because I am able Casino City Press, T., 2012. Gaming business directory. Online. http://
to evaluate gambling decisions by individuals, this paper online.casinocity.com
Chetty, R., Friedman, J.N., Leth-Petersen, S., Nielsen, R.H., Olsen, T., 2014.
offers a new perspective not just about the likely effects Active vs. passive decisions and crowd-out in retirement savings ac-
of PLS, but also about how individuals evaluate gambling counts: evidence from Denmark. Q. J. Econ. 129 (3), 1141–1219.
across different contexts. The lottery component of lottery- Choi, J., 2015. Contributions to defined contribution pension plans. Ann.
Rev. Financ. Econ. 7, 161–178.
linked savings accounts genuinely substitutes for consumer Cole, S., Iverson, B., Tufano, P., 2017. Can gambling increase savings? Em-
gambling, and the effects are large, amounting to at least pirical evidence on prize-linked savings accounts. Unpublished Work-
3% of overall gambling. ing Paper.. Harvard Business School, Brigham Young University, and
University of Oxford.
These findings suggest that offering innovative savings Cole, S., Paulson, A., Shastry, G.K., 2014. Smart money: the effect of edu-
products like Save-to-Win have under-appreciated effects cation on financial outcomes. Rev. Financ. Stud. 27 (7), 2022–2051.
beyond influencing savings rates. In showing that signif- Collard-Wexler, A., 2014. Adjacent county data. Technical Report. New
York University, Stern School of Business. available online http://
icant gambling resources were redeployed when savings pages.stern.nyu.edu/∼acollard/Data.html
lotteries become available, my findings imply that there Cookson, J.A., 2016. Anticipated entry and entry deterrence: evidence
are important effects of prize-linked savings on actual from the American casino industry. Manag. Sci. Forthcoming https:
//pubsonline.informs.org/doi/abs/10.1287/mnsc.2017.2730.
gambling expenditures. These effects ought to be consid-
Cookson, J.A., 2017. Leverage and strategic preemption: lessons from entry
ered as policymakers consider prize-linked savings as a plans and incumbent investments. J. Financ. Econ. 123 (2), 292–312.
tool to improve consumer financial outcomes (Filiz-Ozbay Doran, J., Jiang, D., Peterson, D., 2012. Gambling preference and the
et al., 2015; Cole et al., 2017). On the other hand, prize- New Year effect of assets with lottery features. Rev. Finance 16 (3),
684–731.
linked savings accounts are not a panacea for improving Dorn, A.J., Dorn, D., Sengmueller, P., 2014. Trading as gambling. Manag.
household savings and gambling behavior. An important Sci. 61 (10), 2376–2393.
subpopulation of casino patrons—those with low self- Dorn, D., Sengmueller, P., 2009. Trading as entertainment? Manag. Sci. 55
(4), 591–603.
control—do not respond to the introduction of savings Fernandes, D., Lynch, J.G.J., Netemeyer, R.G., 2014. Financial literacy, finan-
lotteries. Although this finding implies that prize-linked cial education, and downstream financial behaviors. Manag. Sci. 60
savings accounts are not a full replacement for other sav- (8), 1861–1883.
Filiz-Ozbay, E., Guryan, J., Hyndman, K., Kearney, M.S., Ozbay, E.Y., 2015.
ings enhancement programs, the significant effects that I Do lottery payments induce savings behavior? Evidence from the lab.
find for those who exhibit greater self-control suggest that J. Pub. Econ. 126 (1), 1–24.
PLS accounts and other programs can be complements. Un- Friedman, M., Savage, L.J., 1948. The utility analysis of choices involving
risk. J. Pol. Econ. 56, 279–304.
der the right conditions, savings lotteries can dramatically Gao, X., Lin, T.-C., 2015. Do individual investors trade stocks as gambling?
reshape consumption patterns, thereby increasing saving. Evidence from repeated natural experiments. Rev. Financ. Stud. 28 (7),
2128–2166.
References Grinblatt, M., Keloharju, M., 2009. Sensation seeking, overconfidence, and
trading activity. J. Finance 64 (2), 549–578.
Andrikogiannopoulou, A., Papakonstantinou, F., 2016. History-dependent Hastings, J., Hortacsu, A., Syverson, C., 2017. Advertising and competition
risk preferences: evidence from individual choices and implications in privatized social security: the case of Mexico. Econometrica 85 (6).
for the disposition effect. Unpublished working paper.. London School Kearney, M.S., 2005. State lotteries and consumer behavior. J. Publ. Econ.
of Economics and Imperial College. 89, 2269–2299.
Barber, B.M., Lee, Y.-T., Liu, Y.-J., Odean, T., 2009. Just how much do indi- Kearney, M.S., Tufano, P., Guryan, J., Hurst, E., 2011. Making savers win-
vidual investors lose by trading? Rev. Financ. Stud. 22 (2), 609–632. ners: an overview of prize-linked savings products. In: Mitchell, C.,
Barber, B.M., Odean, T., 2008. All that glitters: the effect of attention and Lusardi, A. (Eds.), Financial Literacy: Implications for Retirement Secu-
news on buying behavior of individual and institutional investors. Rev. rity and the Financial Marketplace. Oxford University Press, New York,
Financ. Stud. 21, 787–818. pp. 218–240.
Barberis, N., 2012. A model of casino gambling. Manag. Sci. 58, 35–51. Kuhnen, C.M., Miu, A.C., 2017. Socioeconomic status and learning from fi-
Barberis, N., Huang, M., 2008. Stocks as lotteries: the implications of prob- nancial information. J. Financ. Econ. 124, 349–372.
ability weighting for security prices. Am. Econ. Rev. 98, 2066–2100. Kumar, A., 2009. Who gambles in the stock market? J. Finance 64,
Barberis, N., Huang, M., Thaler, R.H., 2006. Individual preferences, mone- 1889–1933.
tary gambles, and stock market participation: a case for narrow fram- Li, G., 2012. Gamblers as personal finance activists. Working Paper
ing. Am. Econ. Rev. 96 (4), 1069–1090. 2012-18. FEDS.
Becker, G.S., Murphy, K.M., 1993. A simple theory of advertising as a good Li, X., Subrahmanyam, A., Yang, X., 2018. Can financial innovation succeed
or bad. Q. J. Econ. 108, 941–964. by catering to behavioral preferences? Evidence from a callable op-
Becker, G.S., Murphy, K.M., Werning, I., 2005. The equilibrium distribution tions market. J. Financ. Econ. 128 (1), 38–65.
of income and the market for status. Journal of Political Economy 113, Liao, C.M., 2015. Risk taking begets risk taking: evidence from casino
282–310. openings and investor profiles. Working Paper. SSRN 2639700
JID: FINEC
Lusardi, A., Mitchell, O.S., 2014. The economic importance of financial lit- Schneider, C., Spalt, O.G., 2017. Acquisitions as lotteries: Do managerial
eracy: theory and evidence. J. Econ. Lit. 52 (1), 5–44. gambling attitudes influence takeover decisions? Crit. Finance Rev. 6
Lusardi, A., Schneider, D., Tufano, P., 2011. Financially fragile households: (1), 77–132.
evidence and implications. Brookings Papers on Economic Activity. Simonsohn, U., Simmons, J.P., Nelson, L.D., 2015. Specification curve: de-
Spring, pp. 83–134. scriptive and inferential statistics on all reasonable specifications.
Madrian, B.C., Shea, D.F., 2001. The power of suggestion: Inertia in 401(k) Working Paper SSRN 2294998.
participation and savings behavior. Q. J. Econ. 116 (4), 1149–1187. Spenkuch, J.L., Toniatti, D., 2016. Political advertising and election out-
Meyer, B.D., Mok, W.K., Sullivan, J.X., 2009. The under-reporting of trans- comes. Working Paper. Kilts Center for Marketing at Chicago Booth.
fers in household surveys: its nature and consequences. Working Pa- Nielsen Dataset Paper Series 1-046
per No. 15181. NBER. Thaler, R.H., Benartzi, S., 2004. Save more tomorrow: using behavioral
Peng, L., Thibodeau, T.G., 2017. Idiosyncratic risk of house prices: evidence economics to increase employee saving. J. Pol. Econ. 112, S164–S187.
from 26 million home sales. Real Estate Econ. 45, 340–375. Tufano, P., 2008. Saving whilst gambling: An empirical analysis of U.K.
Rose, I.N., 1991. The rise and fall of the third wave: gambling will be out- premium bonds. Am. Econ. Rev. Papers Proc. 98, 321–326.
lawed in forty years. In: Eadington, W., Cornelius, J. (Eds.), Gambling Viva, L.D., Kasanen, E., Trigeorgis, L., 2017. Real options, idiosyncratic
and Public Policy: International Perspectives. University of Nevada skewness, and diversification. J. Financ. Quant. Anal. 52 (1), 215–241.
Press, Reno, NV, p. 70. Wooldridge, J.M., 2003. Introductory Econometrics: A Modern Approach,
Rosett, R.N., 1965. Gambling and rationality. J. Pol. Econ. 73, 595–607. second ed. Southwestern Publishing, Thompson Learning, Mason, OH.
Savor, P., Wilson, M., 2014. Asset pricing: A tale of two days. J. Financ. Zhang, H., 2017. Pre-matching gambles. Unpublished Working Paper..
Econ. 113 (2), 171–201. Michigan State University.

When Saving Is Gambling PDF

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

When Saving Is Gambling PDF

Загружено:

Авторское право:

Доступные форматы

JID: FINEC

ARTICLE IN PRESS [m3Gdc;April 19, 2018;13:41]

Contents lists available at ScienceDirect

Journal of Financial Economics

When saving is gamblingR

“Why do people play the lottery or why do people gam-

2 J.A. Cookson / Journal of Financial Economics 000 (2018) 1–22

J.A. Cookson / Journal of Financial Economics 000 (2018) 1–22 3

4 J.A. Cookson / Journal of Financial Economics 000 (2018) 1–22

J.A. Cookson / Journal of Financial Economics 000 (2018) 1–22 5

Panel A: Save-to-Win account growth

Total amount saved Average account balance

July 2012 $1.1 million $857

Panel B: Save-to-Win program characteristics

Percent of accountholders who reported

Panel C: Save-to-Win participating credit unions

Credit union name Main branch location Counties served

Centris FCU Omaha Douglas, Hall, Lincoln, Sarpy

6 J.A. Cookson / Journal of Financial Economics 000 (2018) 1–22

J.A. Cookson / Journal of Financial Economics 000 (2018) 1–22 7

All credit unions > $5m in deposits

(1) (2) (3) (4)

Post × # of participating CUs 0.057∗∗∗ 0.057∗∗∗ 0.044∗∗ 0.044∗∗

R2 0.9985 0.9987 0.9985 0.9989

8 J.A. Cookson / Journal of Financial Economics 000 (2018) 1–22

J.A. Cookson / Journal of Financial Economics 000 (2018) 1–22 9

Treated Not treated Difference Conditional difference

10 J.A. Cookson / Journal of Financial Economics 000 (2018) 1–22

J.A. Cookson / Journal of Financial Economics 000 (2018) 1–22 11

(1) (2) (3) (4)

Post × # of participating CUs −0.188∗∗∗ −0.197∗∗∗

R2 0.488 0.675 0.511 0.676

4.2. Evidence from scratch tickets

12 J.A. Cookson / Journal of Financial Economics 000 (2018) 1–22

change in withdrawal behavior. Third, because the lottery Table 5

J.A. Cookson / Journal of Financial Economics 000 (2018) 1–22 13

4.3.1. Uniformity of treatment dress this concern, I undertake a systematic subsampling

14 J.A. Cookson / Journal of Financial Economics 000 (2018) 1–22

Panel A: Summary of alternative speciﬁcations and robustness exercises

Robustness test Estimated effect Robustness test Estimated effect

Alternative sample cuts Controlling for differential time trends

Panel B: Robustness to systematic subsampling (speciﬁcation curve)

Percentiles of the speciﬁcation curve

% of Signiﬁcant results (α = 0.10) 5th 25th 50th 75th 95th

Characteristics of speciﬁcation curve

J.A. Cookson / Journal of Financial Economics 000 (2018) 1–22 15

16 J.A. Cookson / Journal of Financial Economics 000 (2018) 1–22

Average patron −17.9 −37.7 −9.28 794.86 −299.66

J.A. Cookson / Journal of Financial Economics 000 (2018) 1–22 17

Panel A: Changes in gambling behavior

Log(#t ransact ions ) Log(transsize) % NSF % credit card

Post × # of participating CUs −0.089∗∗∗ −0.117 0.004 0.011

R2 0.836 0.352 0.193 0.447

Panel B: Changes in sample composition of patrons

% daytime % weekend % male

Post × # of participating CUs 0.0 0 0 −0.008 0.038∗

R2 0.088 0.112 0.339

Log(cash withdrawn) No withdrawals dummy Log(fees)

(1) (2) (3) (4) (5) (6)

Post × # of participating CUs −0.156∗∗∗ 0.036∗∗∗ 0.007

18 J.A. Cookson / Journal of Financial Economics 000 (2018) 1–22

pre-period and use the no_gamblingkt indicator as the Table 10

This section provides greater insight into why con-