Scale Economies at Payday Loan Stores

Mark J. Flannery
Bank of America Professor of Finance
University of Florida
and
Co-director of the Center for Financial Research
Federal Deposit Insurance Corporation
Katherine A. Samolyk
Senior Economist
Federal Deposit Insurance Corporation
June 2007
ABSTRACT
Payday lenders advance small, short-term consumer loans through convenient
storefront shops with limited underwriting. However, their fees convert to very
high annualized rates of interest (APRs > 390%). Using proprietary, store-level
data, we find that a stores age importantly affects its profitability. Younger stores
have fewer customers, make fewer loans, and experience higher losses per loan
than more mature stores. Regardless of age, stores making a larger number of
loans per customer enjoy lower default losses and lower operating costs per loan.
Although these conclusions cannot necessarily be generalized to the entire payday
industry, they do seem to apply to the sort of large, monoline lenders that control
more than a fourth of the industrys storefronts.
Keywords: payday lending, price ceilings, small dollar loans
JEL Classification: G21, D14, D4

* Mark Flannery and Katherine Samolyk, 2007. Reproduction prohibited

without explicit written permission of the authors.
* The views stated here are those of the authors and do not reflect those of the
FDIC or its Board. We want to acknowledge the excellent research assistance of J.
Aislinn Bohren and Brandon Lockhart. We also thank Bob DeYoung, David
Musto, Karyen Chu, Jane Coburn, participants in the Federal Reserve Systems
Fourth Community Affairs Research Conference and in the Federal Reserve Bank
of Chicagos 43rd Annual Conference on Bank Structure and Competition, and
economists at the Federal Reserve Bank of New York for comments on research
using these data. All remaining errors are our own.

Electronic copy available at: http://ssrn.com/abstract=2360233

I. Introduction
Payday, or deferred presentment, loans are small, very short-term loans extended
with minimal underwriting in exchange for a post-dated check. The typical payday loan
matures in two weeks and has a balance of $250-$300. A typical fee of $15 to $20 per
$100 borrowed translates into very high annualized rates of interest (APRs > 390%).
Given normal state usury ceilings, payday lending requires specific enabling
legislation. As of year-end 2006, payday stores were operating in 40 states and in the
District of Columbia. The payday loan industry has grown very rapidly, from fewer than
300 stores in 1994 to more than 24,000 stores in 2006.1 Payday stores are estimated to
have originated $42 billion in payday loans during 2006, to between 10 and 14 million
customers.2 Much of this growth reflects the emergence of large, monoline payday
advance firms, of which the five largest control more than 26 percent of existing stores.
Two monoline companies (Advance America and QC Holdings) had their initial public
stock offerings (IPOs) during 2004. Stephens Inc (2005) estimated that a new payday
store requiring an initial investment of $30,000 would generate more than $258,000 in
operating cash flow in its first five years of operation, implying an annual pretax return of
approximately 170 percent.
Unless operating costs are extremely high, this market seems ripe for entry by new
lenders, who might force down payday loan fees. However, despite rapid growth in the
number of stores, payday loan fees remain close to state statutory ceilings. In the sample
of stores analyzed here, the median store operating at least one full year had fees equal to
the maximum fee permitted by law. Moreover, many payday customers borrow fairly
often. Available data indicate that a majority of customers use the product more than six
times per year and roughly a fourth borrow 14 times or more (Caskey (2005)). The
industrys most strident critics contend that payday lenders would be unprofitable without
frequent borrowers (Skillern (2001), Stegman and Faris (2003)).
The industrys trade group (the Community Financial Services Association, or
CFSA) responds that payday loans are not intended for frequent use, but rather as an
occasional means of avoiding the costs of cash shortfalls. They cite the growth of payday
lending as evidence of strong demand for the product. Payday lenders also argue that
they cannot charge lower fees and still cover their costs. Critics contend that payday
lenders target certain populations, including minorities (King et al. (2005) and Oron
(2005)) and military personnel (Grave and Peterson (2005) and Oron (2005)).3 But,
Morgan (2006) argues that broader state consumer credit patterns are not consistent with
the argument that payday lending is predatory, and Morse (2007) concludes that
communities where payday lenders are present recover more quickly from local disasters.
In some circumstances, credit-impaired individual needing cash might rationally
prefer a payday loan, given the costs of alternatives, such as overdraft protection fees,
bounced check fees, or late payment fees. Fixed operating costs, small loan sizes, and
default losses on loans exchanged for post-dated checks might require a high APR for
lenders to break even. But the extent of repeated borrowing suggests that many
customers are using the product in ways that are not consistent with the industrys best
practices.4 This raises the question of whether the payday loan product is poorly
designed. Would a different loan contractperhaps for a longer term and including

1
Electronic copy available at: http://ssrn.com/abstract=2360233

some amortizationbe better for many customers and be feasible at a lower average
cost? If so, why has competition or competitive entry not fostered new product designs?
Public policy might have an important role to play if the industrys profit motive and
behavioral considerations result in high-frequency use of payday loans; for example, if
borrowers tend to over-estimate the likelihood that they will be able to quickly pay off
their obligations and competition does not lead to lower payday loan fees.5
In this paper, we study store-level data from two large, monoline payday loans
firms. Each firm responded to a detailed questionnaire, providing us with information on
loan activity, revenues, and costs for as many as 300 stores, for the calendar years 2002
2004. The stores operated in 22 states. Although we cannot use these data to assess the
appropriateness of the payday product for specific groups of individuals, we can use this
information to provide evidence about the determinants of store profitability and the
nature of competition in the industry. The public policy debate has been hindered by the
absence of this type of evidence. Our analysis confirms that payday loan stores operate
with sizable fixed costs and scale economies; profitability requires a healthy level of loan
activity. We also emphasize that the industrys rapid expansion makes it important to
evaluate costs and profitability at the store level (rather than for the multi-store firm as a
whole), because there is a strong relationship between store age, loan volume, and store
performance; while new stores incur losses, mature stores (those that have operated for at
least four years) are quite profitable. Using multivariate regressions, we find that a one
percent increase in the number of loans made by a mature store raises its income in the
range of 0.67 to 1.0 percentage point, on average. Default losses are high when measured
as a share of total store operating costs; around 23 percent for mature stores. But loan
losses are not large when measured as a share of the total volume of loans originated
(which determines fee income); in the range of two to three percent for mature stores. As
noted, most stores charge fees close to statutory maximums. Given fees, our tests
indicate that payday store profitability is largely driven by loan volume. But we also find
that stores with more repeat customers have lower loan losses and lower operating costs
on a per-loan basis. And for young stores, repeat business translates into greater
profitability: Young stores making a larger number of payday loans per customer earn
higher profits.
The paper is organized as follows. Section II briefly presents background
information about the payday loan industry, and section III discusses available evidence.
Section IV describes our data and presents univariate statistics that profile store lending
activity, costs, and revenues. Section V discusses our multivariate tests and section VI
presents the results. Section VII presents a model of the industrys operations and the
impact of statutory fee ceilings on store size and profits. The final section concludes.
II. Payday Loan Mechanics
Modern payday lending emerged in the early 1990s, when check cashers and pawn
shops began making these small, short term loans. While the industry remains
fragmented, large monoline lenders operate close to 27 percent of payday storefronts
(Stephens Inc (2007)). Payday loans are extended with minimal underwriting to

2
Electronic copy available at: http://ssrn.com/abstract=2360233

borrowers who are often credit-impaired; in some cases, because they are having
difficulty paying their outstanding debts (Elliehausen and Lawrence (2001)). In most
cases, an applicant with verifiable employment and a checking account (in good
standing) can obtain a payday loan.6 In exchange for the loan, the borrower provides the
lender with a post-dated personal check for the loan amount plus the finance charge.
Some lenders, such as internet payday lenders, substitute a debit agreement for the
borrowers check. The borrower can redeem her check by paying off the loan on or
before the loan maturity date. Alternatively, she may renew the loan by paying another
fee. If the loan is not repaid or renewed, the lender can deposit the check after the date
stipulated on its face. Although internet payday lending is growing, the traditional
payday loan storefront remains the dominate delivery channel, where storefront locations
are chosen to maximize convenience and customer traffic (Morgan Keegan & Company
Inc (2006)).
A portfolio of very short-term loans generates some unusual arithmetic for
researchers accustomed to banking data. First, the loans short maturities permit payday
lenders to originate a very large volume of loans with a small amount of balance sheet
funding. The $42 billion in payday loans originated in 2006 translates into 140 million
$300 loans. But the typical two-week loan maturity implies that the industry would have
had outstanding loan balances of only around $1.6 billion on average at any point in
timea portfolio size comparable to that of a relatively large community bank. This
bank, however, would be dealing with millions of customers per year. Now consider
the APR calculation, which attracts so much headline attention. Loan fees are generally
between $15 and $20 per one hundred dollars lent. A two-week, $300 payday loan
would therefore cost $45 to $60. If a payday loan firm makes such a loan every two
weeks for an entire year, it will earn from $1,170 to $1,560 in fees, or an annualized
return of from 390 to 520 percent. Profitability thus depends on the level of store
operating costs and default losses.
Publicly-traded payday companies report annual loan losses in the range of two to
four percent of total loans originated (Stevens Inc (2005)). However, banks generally
report their loan losses as a share of average outstanding loans. Measured in this manner,
payday loan losses will be in the 50-100 percent range because the very short maturity
implies that outstanding loans at any point in time are only a fraction of originations. In
our data, the mean ratio of loan losses to average outstanding loans was 59.1 percent for
mature stores (open more than 4 years) and 86.2 percent for young stores (open 1 4
years).7 Surely, these losses are high in absolute terms. As we report below, other
operating costs are also high; roughly three times store loan losses, on average.
Ultimately however, profitability is what matters; and profitability will depend on how
the costs of providing payday loans compare to the extremely high APRs implied by
loan fees.
III. Existing Evidence and Public Policy Questions
Our data cannot indicate whether payday lenders prey on vulnerable citizens or
increase a local communitys resiliency to natural disasters. However, the cost, balance

sheet, and profitability information we have gathered permits us to evaluate two

questions that are important in the payday lending debate. First, do payday lenders have
an economic incentive to encourage repeat borrowing? Second, is the cost structure of
the payday loan industry consistent with that of a competitive market where relatively
free entry reduces fees and profit margins?
The absence of data has limited researchers ability to study the structure and
performance of the payday loan industry. However, some basic information about the
business is consistent across data sources and with the data that we analyze here.8 Central
tendencies for loan size ($200-$300), loan fees ($15-$20 per $100 borrowed), and loan
maturities (around 14 days) are well documented. As summarized by Caskey (2005),
available data indicate that substantial proportions of borrowers obtain multiple loans
during a given year and consequently pay sizable fees to borrow a relatively small
amount of money. An industry-sponsored survey reported that half of respondents said
their most recent advance was a renewal of an existing loan (Iota Data Corporation
(2002)). Our own data reveal a similar pattern for payday stores operating at least one
full year (see table 1). Moreover, since payday borrowers can take out loans from more
than one store (borrowing from Peter to pay Paul), repeat borrowing at an individual
store will understate the overall incidence of repeated payday borrowing. Elliehausen
and Lawrence (2001) found that 47 percent of their survey respondents had used more
than one payday advance company during the previous 12 month period.
There has been some research about the characteristics of payday customers and
their attitudes about the product. Industry-sponsored surveys indicate that payday
borrowers tend to have moderate incomes.9 Payday borrowers tend not to be extremely
poor because they need to have a job and a bank account to borrow. According to
Cypress Research Group (2004), less than one-third (31%) of surveyed borrowers had
incomes below $25,000 (compared to 30% of U.S. households); while more than half
(52%) had income in the $25,000 to $50,000 range (compared to 29% of U.S.
households).10 But payday borrowers also appear to have a more difficult time paying
their debts. Elliehausen and Lawrence (2001) report that of the 57 percent of respondents
that held bank credit cards, only 25 percent reported that they almost always paid their
balance in full (compared to 49 percent of the broader U.S. population of bank card
holders); and 27 percent of their respondents had monthly debt service to income ratios of
20 percent or more (compared to 10 percent of the broader population).
Consumer advocates object to the high fees paid by repeat borrowers and contend
that the business model is designed to encourage repeat business. As noted, the CFSA
responds that the fees on these loans compare favorably to the costs of alternative ways
of dealing with cash shortfalls and reflect the sizable costs of providing small
denomination, short-term loans through convenient storefront locations.11
Stegman and Faris (2003) have provided the only previous econometric analysis of
payday firms costs and profitability (that we have found).12 Using data collected by the
Office of the Commissioner of Banks in North Carolina for all licensed check cashing
and payday lending companies in 1999 and 2000, they estimated cross-sectional
regressions explaining two dimensions of firm profitability:
(1)

Pi a 0 a1(N i ) a 2(C i ) a 3(% Re peat i ) Z i ~

i

where: Pi = firm is profits per store: alternately revenues (loan fees plus returned check
fees) per outlet or modified gross income (MGI) per outlet;
Ni = the number of outlets operated by firm i;
Ci = the number of customers per outlet at firm i;
%Repeati = the percentage of firm is customers who borrow at least monthly;
Zi = a vector of other control variables.
Stegman and Faris find that a3 > 0 for both profit measures, and conclude that the
financial performance of the payday loan industry, at least in North Carolina, is
significantly enhanced by the successful conversion of more and more occasional users
into chronic borrowers."13 Unfortunately, this analysis of profitability does not
disentangle the influence of repeat business from the influence of greater loan activity.
As indicated in equation (1), for a given number of customers (Ci), an increase in
%Repeat implies more total loans; and it would be quite surprising if firm revenues and
profits did not grow with loan volume.
In contrast to (1), we estimate cost and profit regressions that include separate
measures of a stores total loan originations and its number of customers as explanatory
factors. We also have store-level data albeit for only two firms, but for firms that
operates in many states. Thus, we match stores to zip-code-level census data describing
neighborhood economic and demographic characteristics that could affect payday store
performance. Perhaps most important, we have more complete cost information that
permits us to capture all fixed and variable costs of operations at the store level. As we
discuss below, nonlinear estimated cost and profit functions suggest that our payday
stores operate with sizable scale economies; this has important implications for the
industrys equilibrium.
The nature of industry equilibrium is important to policy discussions because it
describes the impact of new entrants on equilibrium payday loan terms. For example,
standard microeconomic theory predicts that entry to a profitable industry will eventually
drive excess profits to zero. As new firms enter an industry, one normally expects prices
to decline and marginal costs to increase. Does the relatively free entry in states that
permit payday lending imply that payday loan fees reflect marginal production costs?
Competition from new entrants should also address the possibility that the design of the
payday loan product is inherently flawed. Under normal circumstances, firms should
compete by designing more attractive loan products (e.g. longer-term, amortizing), which
presumably would attract customers away from competitors.
IV. New Survey Data: Univariate Age Profiles
In response to a detailed questionnaire (available on request from the authors), two
large monoline payday loan companies each provided us with three years (2002-2004) of
proprietary store-level performance data for a sample of 300 stores. Although the
companies identities have been withheld from us, we do know which stores belong to
the same company. Two large firms cannot fully represent the entire industry,

particularly because many payday firms operate on a very small scale.14 Nonetheless,
these data should reasonably describe store-level costs and revenues for the large
monoline lenders, which operate close to 27 percent of the industrys stores (Stephens Inc
(2007)).
In order to avoid survivorship bias, we asked each company to provide the three
years of data for a random sample of the stores it operated as of year-end of 2004. If a
store that existed at year-end 2004 did not exist at the end of an earlier year, we did not
replace it with another store from the earlier year. These data provide unprecedented
detail on loan activity, loan characteristics, fees, borrower frequency, repayment
experience, store-level costs, and revenues. We used each stores zip code to match its
performance data with zip-codelevel census data describing local economic and
demographic conditions in the neighborhood where the store is located.
The industrys rapid expansion is clearly reflected in our data. Of the 586 store
observations we received for 2004, only 461 were operating at the end of 2003 and only
374 at year-end 2002. Even a cursory examination of the data indicated that many
important variables depend on a stores age. We therefore classified stores into the three
age cohorts: "new" stores are those that had been operating for less than a full year (by
the end of a given calendar year);15 young" stores had been open between one and four
years; and mature" stores had been operating at least four years. Table 1 provides
univariate statistics for the three age categories, but we focus our discussion on young
and mature stores.16 Because some payday loan information is already well known, here
we emphasize the more novel information provided by our data.
As indicated in Table 1, loan volume tends to rise sharply with store age: mature
stores extended an average of 8,777 advances per year, compared to 5,742 for young
stores (line 15). The number of customers also rises with store age (line 16), although not
so quickly as loan volume. Mature stores tend to make slightly more loans per customer
(line 5), and a somewhat larger share of their customers tended to borrow more than 12
times per year from a single store. Even mature payday stores, however, are fairly small
and simple operations, having a mean dollar volume of average outstanding loans of just
over $85,700 (line 17). The typical mature store is run by fewer than three full-timeequivalent (FTE) employees (line 18), who earn an average annual salary (with benefits)
of just under $31,000 each. As noted above and reported in line 4 of Table 1, payday
advance stores in our sample tend to charge fees very near state statutory limits.17
Finally, the stores in our sample indicate expansion into rural areas: 19.3 percent of the
new stores and 16.2 percent of young stores operated in rural areas, compared with only
8.5 percent of the mature stores (line 21).
Table 1 also presents statistics summarizing store cost and profitability measures
for our samples of new, young, and mature stores. We concentrate our discussion on the
costs and revenues per loan (rows 29 35); although standardizing costs and revenues by
store (rows 22 28) or by loan balances (not reported) yields similar impressions. Total
revenues per loan average $43.80 ($45.29) for mature (young) stores. operating expenses
per loan (row 30) amount to $19.04 ($26.62) for mature (young) stores and loan losses
per loan (row 31) equal are $6.08 ($9.22).18 Interest expenses on a per-loan basis (line
32) are quite small, reflecting the small size and the very short term of payday loans as
well as the relatively low leverage of parent payday loan companies. Store operating

income (line 33) measures the difference between total store revenue and total store
expenses. Mature stores had average store operating income of $18.68 per loan,
compared to an average of only $9.46 per loan for young stores.
Line 34 reports each stores allocation of its firms general and administrative
(G&A) expenses on a per-loan basis. This is an expense item reported for unspecified
services provided by the corporate headquarters; and is a large item, whose allocation
across stores can substantially affect a stores reported profitability. One of our firms
reported an equal G&A charge per store, while the other appears to have scaled per-store
G&A charges to be proportional to store lending activity. For consistency, we computed
an equal per-store G&A expense for each firm and deducted this expense item to produce
a measure of pretax store profits from the firms perspective. Using this method, young
stores about break even while mature stores earn a mean (median) of $11.24 ($10.83) per
loan originated (line 35).
Overall, the univariate statistics convey a strong impression that store age, loan
volume, and store performance are very inter-related. Therefore, a payday firms (or the
industrys) long-run profitability cannot be evaluated simply by examining the firms
aggregate financials. Because profitability rises quite substantially as a store matures, it
is important to know the age distribution of a firms existing stores and the probability
that new stores will eventually become profitable.
V. Multivariate tests
A large banking literature examines scale economies among mainstream
depository institutions. Cost function specifications have proliferated and grown more
complex (translog, DEA, etc.), largely in an effort to allocate costs within a multi-product
firm. In contrast, the monoline payday loan stores in our sample are very simple
producers of a single product. We therefore use a simple, log-linear cost function
specification similar to that of Greenbaum (1967) and Bell and Murphy (1965). We
include detailed information about portfolio features (average size, average maturity, and
the number of customers served) as well as socio-economic information about the
neighborhoods in which payday lenders are located. We do not estimate cost functions
for stores that have not been in operation for at least one full year (new stores).19 Table 2
reports summary statistics for the variables included in our empirical tests.
A. Cost Functions
We estimated several log-linear regression specifications, of which the most
complete is:
~
(2) ln(OperatingCosts) 0 1 ln(Q ) 2 ln(C) i ln(wi ) j ln( f j ) k ln(dk ) m ln(cm )

where Operating Costs include total store expenses except loan losses and the costs
of collecting delinquent loans,20 and:

Q = the number of loans originated during the year,

C = the number of customers to whom those loans were extended,
fj = features of the stores loan portfolio,
wi = cost of the ith factor input,
dk = measures for local economic and demographic characteristics,
cm = other control variables.
In a log-linear specification, the estimated coefficients measure elasticitiesthe
percentage by which store operating costs change with a one percent increase in the
independent variable. From production theory, we expect 1 > 0 and the i > 0. Constant
returns to scale should yield 1 = 1, while 1 < 1 implies increasing returns to scale.
The impact of repeated borrowing on store costs can be inferred from the
coefficient on C (the number of customers), 2 . From (2), 1 measures the elasticity of
costs with respect to loans, holding C constant. Thus, 1 measures the cost of operating
with a higher rate of repeat borrowers. Conversely, 2 measures the cost of expanding
the customer base while holding the number of loans constant, thereby reducing the
incidence of repeat borrowing. A homogeneous increase in the firms scale (no change in
loans per customer) raises operating costs by the sum of ( 1 + 2 ).
Other loan portfolio characteristics (fi) are the stores average loan size (f1) and its
average loan duration (f2). Operating costs might increase with loan size ( 1 0 ) if
lenders take extra care in underwriting larger loans. As with loan size, origination costs
might also rise with loan duration ( 2 0 ) if lenders devote more resources to longermaturity credits. (Tests for this effect, however, might be weak if loan maturity (f2)
varies little across stores.) Loan features might also be incorporated into measures of
scale economies in (2). The sum ( 1 + 2 ) measures scale economies with respect to
changes in the dollar volume of loan originations (to a given number of customers). The
sum of ( 1 + 1 + 2 ) measures scale economies with respect to proportional changes in
the number of, size of, and the maturity of payday loan originations (which all affect the
dollar volume of outstanding loans at any point in time). We will, however, focus our
discussion on the relative impacts of additional loans (Q) vs. additional customers (C) on
store-level operating costs.
The other explanatory variables summarized by (2) include factor costs where:
w1 = the wage rate, proxied by the stores annual expenditures on wages, salaries
and benefits per full-time-equivalent (FTE) employee;
w2 = the store rental rate, proxied by both the median housing rental rate in the
market (the MSA or rural part of the state) and in the neighborhood (zip code). Including
both measures distinguishes between variations across markets and within a given
market.
w3 = the cost of funds. Because funds are raised at the firm level, the price of funds
is the same for all stores operated by a given firm. We therefore include a firm dummy
variable to capture differences in funding costs. The coefficient on the dummy variable
can also reflect differences between our two sample firms management compensation,
underwriting and loss mitigation policies, or training, IT development, and legal costs; to

the extent that these factors affect store costs (or losses or profits in regressions discussed
below).
Controls for variations in neighborhood economic and demographic conditions
include:
d1 = Median family income in the stores zip code,
d2 = Median family income in the stores MSA or rural part of the state,
d3 = Unemployment rate in the stores zip code,
d4 = Percentage of zip code population with income below the poverty line,
d5 = Owner-occupied housing rate in the zip code,
d6 = Percent of zip-code population with a college degree,
d7 = Population in the zip code,
d8 = Population in the MSA or rural part of the state,
d9 = Number of payday loan stores per 10,000 people in the state (a rough measure
of competition) reported in Stephens Inc (2004). The first eight variables were derived
from zip-code level Census data.
Finally all specifications included the following additional control variables:
c1 = ln(store age), measured in calendar months of operation,
c2 = year dummies for 2002 and 2003 (The 2004 dummy is omitted),
c3 = a rural dummy variable, if the store is not located in an MSA.
B. Loan Loss and Profit Functions
We also estimated a credit losses version of (2), to quantify how portfolio
features and factor costs affect loan losses:
~
(3) ln(Credit Losses) 0 1 ln(Q ) 2 ln(C ) i ln(wi ) j ln( f j ) k ln(dk ) mcm

where Credit Losses is the dollar volume of loan losses and collection costs
reported for the year. Explanatory variables coincide with those in the cost function. (2),
with the addition of the payday loan fee charged by the store per $100 lent. Factor costs
(wi) would affect a stores rational collection efforts, which should affect observed loan
losses. More lending should raise dollar losses (1 > 0), but more loans per customer
should reduce losses (2 < 0) because stores are unlikely to lend to those who have not
repaid or renewed in the past. Evidence of a positive association between size and/or
maturity and loss rates would be consistent with arguments that keeping payday loans
small and very short term reduces losses. (Banking literature often views larger loans or
longer-term loans as more risky.) Regarding the fees charged, in a competitive industry
one might expect that higher fees are associated with higher losses. We expect that older
stores will have lower loan losses because personnel will have more experience building
relationships and identifying reliable borrowers.
Finally, local economic and
demographic variables may influence observed default losses. Hence we include them as
important control variables, in order to obtain unbiased scale economy estimates.
Profitability is perhaps the most important dimension of firm performance, since it
represents a firms bottom line. Unfortunately, estimation of a proper profit function

(Lau (1972), Mullineaux (1977)) requires a measure of fixed assets that we do not have.
We therefore estimate a profit function version of (2):
~
(4) ln(StoreProfits) 0 1 ln(Q ) 2 ln(C) i ln(wi ) j ln(f j ) k ln(dk ) m ln(cm )

where Store Profits is measured by store operating income = revenues operating costs
loan losses costs of collecting late repayments.
Once again, Q (the number of loan originations) is the basic measure of output used
to evaluate scale economies. We continue to be particularly interested in the effect of
additional customers (C) on profits. If 2 < 0, providing a fixed number of loans to more
customers reduces store profits. The explanatory variables in (3) include all of those in
(2), plus the fee that the store charged per $100 lent.
VI. Regression Results
We estimated a number of different specifications of (2) (4). For brevity, we
only report results for full specifications that include controls for economic and
demographic features of a stores market (d1-d9) and other controls (c1-c3), along with the
measures of store output, portfolio features, and factor prices.
A. Mature Stores
Panel A in Table 3 reports coefficient estimates for the subset of mature stores
(those that have been operating for at least four years). The cost function reported in the
first two columns of Panel A indicates that payday loan stores operate with substantial
scale economies. A one percent increase in the number of loans originated raises
operating costs (measured net of loan losses and collection costs) only 0.128 percent,
holding the number of customers constant. The t-statistic of the hypothesis test that
payday loan stores operate with constant returns to scale (that is, that = 1) is close to
19. For a given number of loan originations, operating costs also rise with the number of
customers served (an elasticity of 0.346) indicating that it is more costly to serve a
broader customer base ceteris paribus. If a store increases loans and customers by the
same proportion (keeping the average number of loans per customer constant), store
operating costs increase by 0.128 + 0.346 = 0.474 percent. In other words, it is cheaper
for stores to make additional loans to existing customers than to add loans by attracting
new customers. A stores operating costs also rise with an increase in average loan
maturity or average loan size. Of the three proxies for factor costs, average wages and
the median rent in the market carry appropriately positive and significant coefficients.
We did not find an association between store age and operating costs for mature stores.
We also did not find differences in operating costs for store located in rural (versus
urban) areas or for those operated by firm B (versus firm A). The two year-dummy
coefficients imply that nominal operating costs were higher in 2004 than in 2003 and in
2002.

10

The regression results reported in the middle columns of Panel A indicate the
determinants of loan losses and collection costs for mature stores. Credit loss costs rise
with the number of loans, but less than proportionately (elasticity = 0.189). However,
losses increase substantially when a store serves more customers (elasticity = 0.747),
consistent with the notion that repeat customers tend to be paying customers (most likely
because lenders will not lend to borrowers with a bad track record). Extending larger or
longer-term loans also increases loan losses; but again, less than proportionately.21 Factor
costs have little effect. Controlling for neighborhood characteristics, we did not find that
store age or rural (versus urban) location was associated with variations in credit losses.
But stores owned by one of our firms (that we labeled firm B) had higher losses than
those operated by the other.
The profit-function regression results for mature stores are reported in the two
right-most columns of Panel A in Table 3. Here we find clear evidence of scale
economies measured in terms of the number of loans originated. Holding constant the
number of customers, a one percent increase in the number of loans increases store
operating income by 0.665 percent. The number of customers did not have a systematic
effect on mature stores profitability for our sample; despite the results that increases in
the customer base raise both operating costs and loan losses. We also did not find a
statistically significant relationship between factor costs, store age, or rural (versus urban)
location, and store profitability. Despite firm Bs significantly higher loan losses, its
stores were not less profitable than those operated by firm A. However, higher costs are
reflected in lower operating income on average, for mature stores in 2004, compared to
previous years.
B. Young Stores
Tables 1 indicates that young stores have substantially fewer customers and make
substantially fewer loans than the mature stores in our sample. Panel B of Table 3 reports
the cost-, loss-, and profit- function estimates for our sample of young stores. As was the
case for mature stores, store operating costs rise by more when loans and customers
expand proportionally than when existing customers take out more loans. Similarly, the
number of loans made has no significant effect on the costs associated with credit losses.
But increasing the number of customers increases losses for young stores close to
proportionately. For young stores, age (months in operation) does matter to credit losses
and profitability, reducing the former and increasing the latter. The results for the profit
specification also differ for young firms in that we find that repeat borrowing affects a
stores bottom line. Store operating income increases in proportion with the number of
loans (elasticity = 1.044). However, the significant negative coefficient estimate
associated with the number of customers indicates that, for young stores, serving a larger
customer base lowers the profits associated with extending a given number of loans
(elasticity = -.278). Thus, increasing loan output among a given number of customers
raises operating income by more than increasing the number of loans and the number of
customers proportionately.
C. Summary

11

The results of our empirical tests support two important conclusions about the costs
and profitability of payday loan stores. First, operating scale economies are substantial,
implying that fixed costs at the store level require a reasonable volume of loan activity
for profitability. (We discuss this more below.) Second, frequent borrowers are less
costly to serve, impose lower credit losses, and for young stores, are more profitable than
more occasional borrowers. The largest effect is for credit losses, which are much lower
for stores with more repeat borrowing.
These results are consistent with a fast food economic model of payday loan
stores. New stores take a while to become profitable, after which they generate
considerable positive cash flow. Operating costs include a sizable fixed cost
component, which makes profits quite sensitive to volume sold. Unlike fast food
restaurants, however, payday stores care to whom they sell their product. Compared to a
new customer, repeat customers are cheaper to service, generate lower loan losses, and
(for new stores) yield higher profits.
VII. Industry Equilibrium
Our multivariate cost and profit functions for stores operated by large monoline
payday firms can be used to speculate about the nature of competition in the payday
industry, and how entry might affect equilibrium profits and prices. The estimated cost
functions reported in Table 3 imply a substantial fixed cost of operating a payday loan
store, given the very small size of these loans. These functions are plotted in Figure 1.A.
Extrapolating the regression parameters to a low output level and evaluating them at the
mean value of the other independent variables indicates that both young and mature
stores have a fixed operating cost in the range of $80,000 to $100,000. Figure 1.B shows
the corresponding average operating cost curves, which slope down monotonically by
virtue of the regressions log-linear specification. The steep initial section reflects the
high fixed operating costs.
Available anecdotal evidence also implies that a payday store confronts a rather
inelastic demand for loans. The industry touts quick and convenient access as a major
advantage of its product, which conforms to conventional notions about convenienceoriented retail businesses. Location is critical. The presence of these two features
large fixed operating costs and relatively inelastic demand suggests that the industry
can be characterized as monopolistically competitive.
Figure 2 depicts a representative monopolistically competitive firm in an out-ofequilibrium industry. The firm produces a quantity where MR = MC and sets a price
given by the demand schedule. Price exceeds average cost and the resulting profits will
attract entry into the stores market area. The entrance of new competitors will shift the
representative incumbent firms demand schedule leftward. However, the effect of
competitive entry on each firms demand elasticity is not clear; and it is this response that
determines the effect of entry on equilibrium price. Recall that we find that fees tend to
be at the statutory maximum set by the state where the store is located. DeYoung and
Phillips (2006) concludes that a statutory payday fee ceiling serves as a focal point that
impedes price competition among payday lenders.

12

If the demand elasticity remains low even with new entrants, entry may have little
effect on equilibrium price.22 Figure 3 depicts a possible new equilibrium where the
representative store earns zero profits with a higher price than in Figure 2. As the
demand schedule shifts, each firms output falls and its price must actually rise to cover
higher average costs per loan. Price continues to exceed marginal cost. This is the classic
deadweight cost of market power. Too many firms operate in the industry and too few
customers are served. Free entry assures that stores earn zero economic rents, but
monopolistic competition provides no necessary pressure to produce output at the lowest
attainable cost (Chamberlain (1947)).23
In equilibrium, Sandberg (1963) shows how government intervention can improve
the real allocations associated with monopolistic competition. Most states impose
ceilings on payday loan fees, and we find that stores set their prices at (or very near to)
the ceiling. Figure 4 illustrates the imposition of a very special ceiling on the typical
payday store, at the representative stores minimum average cost of production. The
stores relevant demand schedule is now XYZ and its marginal revenue schedule
(XYVW) is discontinuous at Y. The binding ceiling induces the representative store to
increase its output by removing the slope of the marginal revenue schedule over the range
XY. The representative store in Figure 4 suffers losses because average cost exceeds
price. These losses will drive some stores from the industry, until the representative
demand schedule intersects the average cost curve at minimum cost, as in Figure 5. By
setting a binding ceiling equal to the minimum average cost, regulators could induce
more payday loans from the same set of industry firms. If demand is very inelastic,
reducing the maximum fee may have little effect on the total number of loans taken.
However, social costs are lower because the ceiling reduces the number of store locations
and hence the fixed costs of providing payday loans.
The most important policy implication from this exercise in microeconomic theory
is that competition from new or potential entrants can be expected to drive profits to zero,
but it may not lower prices. This implies that states with higher (binding) fee ceilings
should have more payday stores, each operating with lower loan volume. Put another
way, with a higher rate ceiling, a store doesnt need to make as many loans to cover its
fixed operating costs.
VIII. Summary and Conclusions
We have analyzed a proprietary data set obtained from two monoline payday
companies, each of which operates stores in many states. We cannot be certain of the
extent to which our results apply beyond these two firms. However, neither do we have
reason to believe that our sample does not represent the large monoline firms that control
a sizable segment of the industry. At a minimum, our analysis expands the available
information about the costs of delivering small and relatively homogeneous consumer
loans through storefront offices.
As others have found previously, our data indicate that customers who borrow
repeatedly are common in our sample. Rollover/renewal transactions account for more
than 45 percent of loan originations; and 45 to 50 percent of customers borrow more than

13

six times per year. Our data also indicate that nearly all loans tend to be priced at
statutory state ceilings. These detailed, proprietary data permit us to estimate cost and
profit functions. Regression analysis indicates that monoline payday loan stores face
sizable scale economies in making these very small, short-term loans and that repeat
borrowers are cheaper to serve and more likely to repay than more infrequent customers.
This is particularly relevant for young stores (in operation for between one and four
years), because their profits actually decline with the number of customers they service
(for a given number of loan extensions). At least part of the industry, therefore, has a
profit motive to encourage repeat borrowing
Our estimated cost functions and anecdotal evidence that loan demand is relatively
price-inelastic carry important implications for the impact of competitive entry on payday
loan fees. In this sort of monopolistically competitive industry, free entry should
generate zero profits in equilibrium, but it will not necessarily lower loan fees or drive
costs to their social minimum. Indeed, entry is already quite easy in states that permit
payday lending! We conjecture that binding fee ceilings could improve the social value
of payday lenders, at least to a limited extent.
Large G&A expenses per store suggest that there also may be scale economies at
the firm level, which we cannot assess with data for only two firms. Stegman and Faris
(2003) found that per-outlet profitability was positively related to the number of outlets
operated by the firm. They conjecture that part of the firm-level fixed costs involves the
growing number of legal challenges brought by increasingly aggressive consumer
interests and the growing complexity of regulatory environments.24 More broadly,
payday firms confront political and legislative risks, which may be reduced by
centralized lobbying efforts. Of course, the cost structure of providing payday loans is
likely to differ for multi-product firms. There also is little information about the elasticity
of the demand for payday loans. These factors limit the conclusions we can draw about
the dynamics of the payday loan industry.
The size of payday loan fees and the apparent demand for the product make it
noteworthy that depository institutions have not entered the business. Bair (2005) points
out that depository institutions have already incurred fixed operating costs, and might
therefore be able to make small, short-term loans profitably at substantially lower fees.
Yet few banks (or even credit unions) appear to have entered this market. Bair posits that
the profitability of fee-based overdraft protection programs may deter banks from
offering alternatives to payday loans. Another puzzle concerns the lack of product
diversity. The monoline lenders almost exclusively offer a two-week, non-amortizing
loan,25 although even the industrys trade group admits that payday loans make sense
only for individuals with reliably transitory cash shocks. The frequency or repeated
borrowing suggests that some customers might prefer a different loan product, such as an
amortizing, multi-period having a lower APR. More research is needed to help policy
makers understand why the industry has produced so little product innovation. Clearly,
much more research is needed about this growing, high-profile portion of the financial
services industry.

14

Notes
1

See Stephens Inc (2007) and Snarr (2002).

See Stephens Inc (2007) and Stephens Inc (2006). In addition, Stephens Inc (2007) estimates that
Internet payday lending advanced $5.65 billion in loans in 2006.
3

In October, 2006, President Bush signed the Talent-Nelson amendment to the John Warner
National Defense Authorization Act for Fiscal Year 2007, which outlaws loans to military
personnel at APRs exceeding 36 percent.

4
5

See the CFSA Web site, www.cfsa.net/industry_best_practices.html

See discussions in Mann and Hawkins (2006) and Samolyk (forthcoming 2008).

Many lenders use subprime credit reporting agencies (such as Teletrak) to identify applicants
with a history of bouncing checks or failing to repay payday loans. Payday loan registries in
several states (including Florida and Illinois) are intended to limit the number of concurrent loans
a borrower can have outstanding.
7

Here losses are the sum of loan losses reported as an annual expense plus recovery costs
incurred.
8

State governments are the source of much of the available data on the payday loan industry.
Industry-sponsored surveys have tended to be the source of demographic and economic
information about payday loan customers.
9

See IOTA (2002), Elliehausen and Laurence (2001), and Cypress Research Group (2004)).

10

The remaining 17% of respondents had incomes of $50,000 (compared to 41 % of U.S.

households).

11

See the CFSA Web site, www.dfsa/net/myth-vs-reality.html#4.

12

Ernst and Young (2004) consider similar questions, using survey data from 19 Canadian payday
firms. Although 19 data points cannot support econometric analysis, they report scale economies
at the firm level. They also report substantially higher costs associated with serving new
customers compared to repeat customers, although some of the data on the number of new
customers is estimated by the researchers (page 15).
13

Stegman and Faris (2003), p.8.

14

According to Stephens Inc (2007), around 55 percent of payday loan storefronts are operated by
firms having fewer than 100 stores.

15

We calculated store age in terms of the number of years that the store had been operated by the
firm in the indicated year.

15

16

Table 1 reflects data that excludes observations having significant missing data or clearly
problematic data. Incomplete or problematic data and the exclusion of new stores reduced the
number of store-year observations used in our multivariate analysis to 846 observations.

17

This tendency is also reported by Stegman and Faris (2003) and DeYoung and Phillips (2006).

18

Because predictable losses can be factored into loan pricing, however, each stores loan loss
variability is also relevant.
19

In a similar vein, banking studies often exclude young banks (less than five year old) from
performance studies.

20

We exclude G&A expenses from our cost measure because they are a fixed per-store charge to
cover costs incurred at the corporate center. Including such fixed expenses in operating costs
would tend to increase measured scale economies, but we do not have enough observations on
different firms to study differences in these corporate-level costs across firms.

21

Loans that default may be kept on the books for a few days while employees determine whether
the customer means to default.

22

Ausubel (1991) offers another reason why entry to a financial market may not reduce prices. He
observes that credit card loan rates in the 1980s were not greatly affected by the presence of many
potential entrants, and suggests that price competition might have perverse information and sorting
effects in this market, as in Stiglitz and Weiss (1981).

23

Mankiw and Whinston (1986) and Prescott and Visscher (1977) consider whether a
monopolistically competitive equilibrium produces too much or too little product diversity. We
do not consider this aspect of the equilibrium here.
24

Stegman and Faris (2003), p 11.

25

However, in states where regulations limit repeat borrowing, lenders have begun to introduce
installment products albeit with very high fees.

16

References
Bair, Sheila. 2005. Low-Cost Payday Loans: Opportunities and Obstacles, Anne Casey
Foundation. June 2005.
John P. Caskey. 2002. The Economics of Payday Lending. Filene Research Institute.
. 2005. Fringe Banking and the Rise of Payday Lending. In Credit Markets for
the Poor; editors Patrick Bolton and Howard Rosenthal, Russell Sage Foundation. pp1745.
Chamberlin, E.H. 1947, The Theory of Monopolistic Competition (5th ed., Harvard
University .Press, 1947).
Chessin, Paul. 2006., Borrowing from Peter to Pay Paul: A Statistical Analysis of
Colorados Deferred Deposit Loan Act, 83 DENV. U.L. REV. 387, 393 (2005).
DeYoung, Robert and Ronnie Phillips, 2006, Strategic Pricing of Payday Loans:
Evidence from Colorado, 2001-2005, Working paper.
Elliehausen, Gregory, and Edward C. Lawrence. 2001. Payday Advance Credit in
America: An Analysis of Customer Demand. Monograph no. 35, Credit Research Center,
Georgetown University.
Ernst and Young, 2004. The Cost of Providing Payday Loans in Canada. Tax Policy
Services Group Report. October 2004.
Fox, Jean Ann. 2004. Unsafe and Unsound: Payday Lenders Hide behind FDIC Bank
Charters to Peddle Usury. Consumer Federation of America.
Fox, Jean Ann, and Edmund Mierzwinski. 2000. Show Me the Money! A Survey of
Payday Lenders and Review of Payday Lender Lobbying in State Legislatures. U.S.
PIRG and Consumer Federation of America. February.
Fox, Jean Ann, and Edmund Mierzwinski. 2001. Rent-A-Bank Payday Lending: How
Banks Help Payday Lenders Evade State Consumer Protections. Consumer Federation of
America and the U.S. Public Interest Research Group. November.
Graves, Steven M. 2003. Landscapes of Predation, Landscapes of Neglect: A Location
Analysis of Payday Lenders and Banks, The Professional Geographer 55(3) at p312
(2003).
Graves, Steven M., and Christopher Peterson. 2005. Predatory Lending and the Military:
The Law and Geography of "Payday" Loans in Military Towns, University of Florida
Law School, March 2005.

17

Jefferies & Company Inc. 2006. Recap of CFSA Payday Lending Conference, Consumer
Finance Report. March 9, 2006.
King, Uriah, Wei Li, Delvin Davis, and Keith Ernst. 2005. Race Matters: The
Concentration of Payday Lenders in African American Neighborhoods in North Carolina.
Center for Responsible Lending.
Mann, Ronald and J. Hawkins. 2006. Just Until Payday, Berkeley Electronic Press
Express, Preprint Series. Draft Paper no. 1863, 2006. Released November, 2006.
Morgan Keegan & Co Inc. 2006. Payday Lending Update, Industry Note, March 6, 2006.
Morgan, Donald P. 2007. Defining and Detecting Predatory Lending, Federal Reserve
Bank of New York Staff Report no. 273.
Morse, Adair. 2006. Payday Lenders: Heroes Or Villains? Working paper, Ross School
of Business, University of Michigan, December.
Mullineaux, Donald J. 1978. Economies of Scale and Organizational Efficiency in
Banking: A Profit-Function Approach, The Journal of Finance, Vol. 33, No. 1. (Mar.,
1978), pp. 259-280.
Oron, Assaf. 2005. Easy Prey: Evidence for Race and Military Related Targeting in the
Distribution of Pay-Day Loan Branches in Washington State, Department of Statistics,
University of Washington, June 2005.
Roth Capital Partners 2006. Rentcash Inc, Equity Research, July 19, 2006.
Samolyk, Katherine, 2008 (forthcoming). Payday Lending: Evolution, Issues, and
Evidence, in Household Credit Usage: Personal Debt and Mortgages, edited by Sumit
Agarwal and Brent W. Ambrose, New York: Palgrave/MacMillian.
Skillern, Peter. 2001. How Payday Lenders Make Their Money. Community
Reinvestment Association of North Carolina. April 18.
. 2002. Small Loans, Big Bucks: An Analysis of the Payday Lending Industry in
North Carolina. Community Reinvestment Association of North Carolina. .
Snarr, Robert W., Jr. 2002. No Cash til Payday: The Payday Lending Industry.
Supervision, Regulation and Credit (Q1), Federal Reserve Bank of Philadelphia.
Stegman, Michael A., and Robert Faris. 2003. Payday Lending: A Business Model That
Encourages Chronic Borrowing. Economic Development Quarterly, Vol. 17, No. 1,
February. pp. 8-32.

18

Stephens Inc. 2000. Payday AdvanceThe Final Innings: Standardizing the Approach.
Industry Notes (September 22), 5.
. 2004. Industry Report: Payday Loan Industry.
. 2006. Payday Loan Industry Report , April 6, 2006.
. 2007. Payday Loan Industry Report , March 28, 2007.

F igure1.B S toreO peratingC os ts

P erL oan

200,000

A verag ec os tperloan
($s )

T otaloperatingcos ts ($s )

F igure1.AS toreO peratingC os ts

150,000

100,000

50,000
Mature
Y oung

4,000

8,000

12,000

Numberofloans madeperyear

300
250
200
150
100
50
0

Mature
Y oung

4,000

8,000

Numberofloans madeperyear

19

MC

AC

MC

P*

AC

P**
P*

Demand Curve

Demand Curve

MR

MR

Figure 2: Monopolistic competitor, out of long-run equilibrium.

MC

Figure 3: Monopolistic competitor, long-run equilibrium.

AC

MC

P**
X

AC

Demand Curve: XYZ

Demand Curve: XYZ

X

MR
W

Figure 4: Price ceilings effect on monopolistically competitive equilibrium (Sandberg

(1963)). Negative profits earned.

MR: XYVW
W

Figure 5: Price ceilings effect on monopolistically competitive equilibrium (Sandberg

(1963)). Zero equilibrium profits, minimized average cost.

20

Table 1: Payday Loan Store Characteristics and Performance Measures, by Age Cohort
Source: Compilation of responses to authors' questionnaire
New Stores
Operating less than 1 year
Mean
Std Dev Median
Loan Characteristics
1.
Avg loan size ($)
255.5
47.6
246.8
2.
Loan duration (days)
15.1
1.8
15.0
3.
Fee per $100 borrowed
14.98
2.46
15.00
4.
Fee (% of state max) (1)
90%
22%
98%
5.
# Loans per customer
5.2
14.7
3.7
6.
% of customers with:
7.
1 advance per year
28.0
17.5
25.6
8.
2-6 per year
48.9
12.3
50.5
9.
7-12 per year
15.8
10.2
18.6
10.
>12 per year
7.2
11.0
0.6
11. Rollovers/renewals
12.
as % of total # of loans
44.6
19.2
46.9
13.
as % of total $ volume of loans
39.0
24.1
43.5
14.
15.
16.
17.
18.
19.
20.
21.
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40

Store Characteristics
303
Total loans ($1000s)
1,152
Total loans (#)
270
# customers
12.8
Avg. outstanding loans ($1000s)
2.0
# FTE employees
Number of loans made
1,152
Years store operated by company
0.5
=1 if store not in MSA (percent)
19.3
Store Costs and Revenues; $1,000s per Store
40.9
Total store revenues
69.4
Operating costs excl losses & collection costs
11.0
Loan losses plus collections expense
1.4
Interest expenses
-39.7
Store operating income
53.1
G&A expenses (1)
-94.3
"Store profits" (3)
Store Costs and Revenues, $'s per loan
34.67
Total store revenues
96.79
Operating costs excl losses & collection costs
9.27
Loan losses plus collections expense
2.06
Interest expenses
(71.39)
Store operating income
107.86
G&A expenses
(181.31)
"Store profits"
Store Operating Cost Components, % of Total Store Operating Costs
33.6
Wages & salaries
24.8
Rent, maint., util, taxes
12.5
Advertising
4.9
Depreciation
24.2
Other expense (nec)
Total
100.0

Young Stores
Operating 1 - 4 years
Mean
Std Dev Median

Mature Stores
Operating more than four years
Mean
Std Dev Median

257.7
14.8
16.76
96%
7.3

51.2
1.9
2.80
9%
1.4

244.7
14.2
17.23
100%
7.2

227.6
15.3
18.27
93%
7.9

47.6
3.4
6.87
14%
1.8

225.1
14.3
17.59
100%
7.9

16.7
36.1
21.3
24.2

6.6
6.3
3.7
12.5

17.1
37.1
21.2
20.5

13.6
32.8
21.1
30.6

5.9
7.1
4.3
14.1

13.8
32.8
21.2
26.6

46.4
34.2

25.1
29.9

38.7
21.5

46.7
39.8

24.0
26.4

52.5
45.2

277
919
167
11.8
0.7
919
0.2
39.6

232
868
233
9.7
2.0
868
0.5
0.0

1,549
5,742
799
62.6
2.5
5,742
2.3
16.2

864
2,731
379
35.6
1.1
2,731
0.9
36.9

1,382
5,260
708
55.0
2.0
5,260
2.2
0.0

2,068
8,777
1,117
85.7
2.7
8,777
6.1
8.5

894
3,670
436
40.1
1.2
3,670
1.6
27.9

2,014
8,465
1,041
79.6
3.0
8,465
5.8
0.0

36.9
26.8
16.5
1.1
19.4
13.3
23.3

31.4
66.6
5.3
2.0
-38.1
43.6
-90.8

254.2
127.0
50.1
2.3
75.4
55.9
17.2

126.6
28.5
38.1
1.1
86.5
13.6
85.5

225.5
121.8
40.4
2.4
62.2
48.9
7.1

350.4
147.2
45.1
2.5
155.7
51.0
102.2

126.2
38.3
34.6
0.9
84.2
12.0
86.6

334.2
140.2
35.9
2.4
143.6
43.6
90.0

35.22
75.96
6.12
1.23
(50.33)
55.28
(109.72)

45.29
26.62
9.22
0.47
9.46
12.29
(3.30)

34.4
24.3
13.4
4.3
22.7

47.5
25.3
6.1
4.3
16.8
100.0

10.42
72.48
9.29
2.83
73.25
147.11
212.82
8.5
7.7
8.1
3.0
6.3

12.91
13.00
5.29
0.30
14.12
7.76
19.88
6.7
5.8
4.6
1.9
4.7

42.56
23.47
8.27
0.42
11.29
10.33
1.25
47.4
25.6
5.7
4.1
16.4

43.80
19.04
6.08
0.35
18.68
7.09
11.24
51.3
22.3
7.0
1.9
17.4
100.0

18.19
8.08
4.94
0.21
10.76
4.21
10.91
7.5
5.8
4.4
1.5
6.5

39.90
16.93
4.56
0.29
17.10
5.74
10.83
51.6
22.2
6.6
1.5
15.7

(1) The ratio of the store's fee to the state maximum was calculated only for stores in states that had a fee ceiling.

21

Table 2: Summary Statistics: Payday Store Questionnaire Data

Mature stores (n = 494)
Mean

Std dev

Young stores (n = 321)

Mean

Std dev

Store performance measures

Operating Costs: Operating costs net of credit losses ($'s)
Credit Losses: Loan losses and collection costs ($'s)

147,196

38,306

127,030

28,480

45,055

34,627

50,145

38,053

155,659

84,222

75,440

86,541

Number of loans

8,777

3,670

5,742

2,731

Number of customers

1,117

436

799

379

Average loan size ($'s)

Store Profits: Store operating income ($'s)

Measures of store output and loan features

244.28

69.90

269.24

66.53

Loan duration (days)

15.32

3.41

14.83

1.86

Fee per $100 ($'s)

18.27

6.87

16.76

2.80

Factor prices
30,940

12,573

27,682

10,824

Median rent in ZIP code ($'s)

Average wage per employee ($'s)

594

161

566

140

Median rent in market ($'s)

609

169

587

144

0.23

0.42

0.43

0.50

Control variables
Owned by Firm B
Store age (months)

73.71

19.63

27.86

11.10

Year = 2002

0.22

0.42

0.26

0.44

Year = 2003

0.36

0.48

0.29

0.45

Economic and demographic variables

Median family income in zip code ($'s)

47,514.1

11,600.8

44,830.5

10,167.2

Median family income in market ($'s)

50,905.5

8,953.3

48,451.2

8,283.9

Unemployment rate in zip code

6.38

2.99

6.45

3.16

Unemployment rate in market

6.04

1.90

6.10

1.89

Percent in poverty in zip code

13.85

7.57

14.37

7.55

12.67

3.72

13.56

4.05

Median housing value in zip code ($'s)

121,893

73,939

108,614

48,033

Median housing value in market ($'s)

135,965

78,481

122,603

57,975

60.66

13.01

62.49

12.29

Percent in poverty in market

Owner-occupancy rate in zip code (%)

Owner-occupancy rate in market (%)

66.08

6.49

66.26

6.36

Pct with college degree in zip code

21.86

10.78

19.76

9.77

Pct with college degree in market

23.20

6.77

22.48

7.03

Total zip code population (#)

34,820

15,587

35,883

16,950

Total market population (1,000's)

2,282

3,342

2,168

3,116

Stores per 10,000 pop in state (1)

1.21

0.73

1.25

0.58

(1) From Stevens Inc (2005).

22

Table 3: Cost and Profit Functions Including Economic and Demographic Variables
(OLS estimates of log-linear specifications)
Panel A: Mature Stores
Dependant variable:
Dependent variable mean:

Costs: Store operating

costs less loan losses and
collection costs

Credit losses: loan loss

expenses and collection
costs

11.729

10.579

Panel B: Young Stores

Store profits: store
operating income
11.949

Costs: Store operating

costs less loan losses and
collection costs
11.729

Credit losses: loan loss

expenses and collection
costs
10.579

Store profits: store

operating income
11.949

R-square:

0.674

0.707

0.745

0.563

0.799

0.753

Adjusted R-square:

0.659

0.693

0.733

0.533

0.784

0.743

Variable

Coefficient Std Error

Coefficient Std Error

Coefficient Std Error

Coefficient Std Error

Coefficient Std Error

Coefficient Std Error

Number of loans

0.128

0.045 ***

0.189

0.138

0.665

0.089 ***

0.109

0.052 **

0.036

0.134

Number of customers

0.346

0.046 ***

0.747

0.146 ***

0.089

0.094

0.221

0.058 ***

0.975

0.147 ***

Average loan size ($loan/#loans)

0.142

0.045 ***

1.259

0.112 ***

0.455

0.058 ***

0.080

0.056

1.152

0.136 ***

0.714

0.105 ***

Average loan duration (days)

0.187

0.059 ***

0.911

0.211 ***

-0.477

0.114 ***

0.206

0.103 **

1.799

0.269 ***

-0.472

0.167 ***

0.086 ***

-0.107

0.148

0.879

Median family income, zip code

-0.285

0.147 *

-0.063

0.364

-0.158

0.166

Median family income, market

-0.356

0.147 **

0.156

0.370

0.095

Unemployment rate, zip code

-0.004

0.041

0.186

0.121

-0.021

Poverty rate, zip code

-0.078

0.051

-0.082

0.145

Pct owner-occ housing, zip code

-0.024

0.060

-0.154

0.163

Fee per $100 borrowed

-0.215

0.207

0.836

-0.170

0.124

0.285

0.294

-0.122

0.211

0.145

0.194

0.167

0.535

0.336

0.001

0.250

0.034

-0.066

0.045

0.021

0.114

-0.037

0.080

-0.051

0.052

0.023

0.057

0.075

0.086

0.143

0.070 **

0.238
-0.095

0.174 ***

0.141 *

0.056

0.089

0.143

0.031

0.108

0.014

0.089

0.002

0.038

0.022

0.030

0.070

0.090

-0.010

0.051

-0.008

0.020

-0.064

0.057

0.001

0.022

0.005

0.018

0.144

0.068 **

-0.086

0.035 **

Population, market

0.007

0.009

-0.019

0.022

-0.017

0.029

-0.017

0.020

Stores (#) in state per 10,000 pop.

0.036

0.015 **

0.130

0.052 **

Average wage per employee

0.102

0.023 ***

0.034

0.059

Median rent, zip code

0.029

0.111

0.145

0.275

Median rent, market

0.406

0.139 ***

0.046

0.377

Store age (months)

-0.071

-0.156

0.125

Store owned by firm B

-0.019

0.045

0.916

0.013

0.051

-0.194

Store not in MSA

0.035 *

0.102 ***
0.107 ***

Population, zip code

Pct w/ college degree, zip code

0.065

1.044
-0.278

0.052

0.031

0.010 *

-0.026

0.014 *

0.020

0.023

0.019

0.025 *

0.113

0.032

0.119

-0.081

0.160

0.079

0.060

0.096 ***

-0.025

0.061

0.069

0.162

-0.021

0.056

-0.005

-0.041

-0.008
0.222

0.080 ***

0.096

0.028 ***

-0.154

0.062 **

-0.047

0.123

0.122

-0.208

0.327

0.330

0.074

0.160

-0.079

0.345

-0.186

-0.027

0.022

-0.195

0.061 ***

0.178

0.041 ***

0.821

0.097 ***

0.013

0.065

0.125 *

0.178

0.077 **

0.038 *
0.050

-0.210

0.043 **
0.052
0.183 *
0.233

Year = 2002

-0.078

0.040 **

-0.073

0.096

0.163

0.039 ***

0.093

0.034 ***

0.207

0.090 **

0.073

0.069

Year = 2003

-0.053

0.023 **

-0.061

0.050

0.085

0.025 ***

0.006

0.025

0.112

0.061 *

0.027

0.039

Intercept

10.548

1.539 ***

-5.982

4.725

3.200

2.297

5.449

1.762 ***

3.378 ***

0.983

2.717

-13.758

Standard errors robust to heteroskedasticity and clustering within store

*,**,*** indicate statistical significance at the 10%, 5%, and 1% significance levels, respectively

23