Payment Card Rewards Programs and Consumer Payment Choice

1

• Payment Card Rewards Programs and Consumer
  Payment Choice
• Andrew Ching
• University of Toronto
• Fumiko Hayashi
• Federal Reserve Bank of Kansas City
• April 2, 2008

2
Motivations

• In 2001, 76 of U.S. households hold at least one
  credit card, 70 hold an ATM/debit card.
• Annual earning from credit card industry is
  around 30 billion dollars in 2006 (from
  interchange fee only).
• Card issuers have been offering rewards to
  attract new customers or compete for market
  shares. Rewards account for 44 of the
  interchange fee revenue in 2006.
• Some issuers reported that they see increases in
  spending after launching rewards programs.
• But we know little about the sources of
  increases.
• This marketing tactic has also received a lot of
  attention from policy debates.

3
Proponents view

• Payment Card industry is a two-sided market
  need both consumers and merchants to adopt the
  card (e.g., VISA) in order to generate benefits.
• Network externalities the more consumers use a
  payment card, the more attractive it is for
  merchants to accept it, and vice versa.
• Rewards could induce more consumers to use a
  payment card and ultimately increase the benefits
  of the entire network via network externalities.
• Card issuers certainly benefit from the size of
  the network they earn more revenue from
  interchange fees (and interests for credit card
  issuers).
• Rewards could potentially lead more consumers to
  switch from cash/checks to electronic payment
  methods could increase social welfare.
• Prelec and Simester (2001), Soman and Cheema
  (2002) find evidence that using credit card
  increases willingness to pay rewards could
  benefits merchants.

4
Opponents view

• After the network becomes mature, the effect of
  rewards on the size of the network would
  diminish.
• Rewards distort price signals, might lead
  consumers to choose an (socially) inefficient
  payment method.
• Merchants pay higher interchange fees for some
  rewards credit cards they have to charge higher
  mark-up. Consumer who use other payment methods
  subsidize rewards card users.
• Maintaining rewards programs takes up resources.
• It may get more people into debt.
• Rewards may make consumers single-home in a
  payment method.
• Some people argue that we should not allow
  interchange fees to cover rewards.
• The debate crucially depend on the effect of
  rewards on payment choice.

5

• Objective
• Examine the effects of rewards programs on
  consumer payment choice for in-store transactions
• (1) Estimate a discrete choice model that
  explains consumers most frequently used payment
  method at a given retail type
• (2) Simulate policies that remove rewards from
  payment cards

6
Roadmap

• Previous studies
• Data
• Econometric specifications
• Results
• Conclusion

7
Literature

• Hirschman (1982), Mantel (2000), Jonker (2005),
  Klee (2006) study how payment method attributes
  affect consumer payment choice.
• Kennickell and Kwast (1997), Stavins (2001), Klee
  (2006), Hayashi and Klee (2003) study how
  demographic characteristics affect payment
  choice.
• Rysman (2007) use VISAs Payment System Panel
  Study to analyze the multi-homing behavior among
  credit cards.

8

• Data
• 2005/2006 Study of Consumer Payment Preferences
• by ABA and Dove Consulting
• 3008 responses ? 1979 observations
• Oversample people with higher educational level
  and higher income
• Contains demographic characteristics
• Gender, race, age, education, income, whether
  adopt direct deposit/online banking.
• Contains rich information on consumer payments
• - whether receive rewards on credit/debit card
• - most frequently used method by retail type
• - perceptions toward each payment method
• - which payment method perceived to be accepted
  by retail type
• - five retail types grocery, fast food,
  department store, discount store, drug store
• - five payment methods cash, check, credit
  card, PIN-debit card, signature-debit

9

• Data

Table II Reward card holders
10

• Data

Figure I Share of most frequently used method by
consumer group
ccwob1, if consumers do not carry a credit card
balance dcwr1, if consumers receive debit card
rewards (either PIN, signature, or
both) ccwr1, if consumers receive credit card
rewards.
11

• Data

Figure I Share of most frequently used method by
consumer group
12

• Data

Table III Consumer perceived payment method
attributes
13

• Data

Table IV Consumer perceived acceptance
14

• Model and Econometric Issue
• Multinomial logit model that explains the most
  frequently used method by retail type.
• Utility to consumer i from using method j at
  retail type h
• X-consumer characteristics C-rewards dummies
  e-unobserved preferences e-measurement error.
• But C and e are likely positively correlated.
• The decision to obtain a reward card could be
  endogenous consumers who have good perception
  about payment cards are more likely to have
  rewards cards.
• Rewards may induce consumers to use the card more
  often and as a result the consumers learn about
  the card and improve their perception toward the
  card (i.e., indirect effect).

15

• Model and Econometric Issue
• To handle this endogeneity problem, we use data
  on individual consumers perception toward each
  payment method (Z), which allows us to control
  the unobserved consumer heterogeneity in
  preference (e) Harris and Keane (1999), Horsky,
  Misra, Nelson (2006)
• We also control for the heterogeneity in
  consumers payment choice set.

16

• Regression Results
• Table V Log-likelihood

17

• Regression Results
• From log-likelihood
• Consumers perceptions toward each payment method
  capture a large amount of consumer heterogeneity
  in preferences for payment method at all five
  types of stores
• Controlling the variation of individual
  consumers choice set is also important

18

• Regression Results
• Table VI Coefficients for reward dummies
• Grocery

19

• Regression Results
• Table VI Coefficients for reward dummies
• Department Store

19
20

• Regression Results
• Table VI Coefficients for reward dummies
• Drugstore

20
21

• Regression Results
• From coefficients for reward dummies,
• The rewards dummies (in particular, PIN-debit
  rewards) have become less significant and their
  point estimates have consistently reduced after
  incorporating the perception variables.
• This suggests the endogeneity problem is not
  merely a theoretical concern.
• Most of the rewards dummies remain statistically
  significant.
• This suggests rewards does affect consumer
  payment choice, but the economics significance
  might be small.

22

• Regression Results
• Table VII Coefficients for perceptions (Grocery)

23

• Regression Results
• From coefficients for perceptions,
• Most of the perception variables are significant
  except Money taken right away
• Comfortable and Convenient seem to be the most
  crucial perception variables.
• Safe is not as significant as other perception
  variables.
• For small amounts is not significant at
  department and discount stores .

24

• Figure II Effect of removing credit card
  rewards based on specification 4
• G-Grocery De-Department Di-Discount Dr-Drug
• F-Fast Food.

25
Effects of Removing Rewards on Credit Card

• Both groups of consumers reduce the probability
  of choosing credit cards.
• The percentage point reductions vary across
  retail types.
• Overall, their reductions are moderate. That
  means the majority of reward credit card
  transactions are replaced by non-reward credit
  card transactions.
• Among switchers who only have rewards on credit
  cards, they tend to substitute more towards
  paper-based methods.
• Among switchers who have rewards on both credit
  cards and debit cards, they tend to substitute
  more towards debit cards.

26

• Figure III Effects of removing credit card
  rewards based on specification 3 (without
  perception variables)

27
Effects of Removing Rewards on Credit Card when
Ignoring Consumers Perception

• The prob. of choosing credit cards tends to be
  slightly overestimated even before removing
  rewards.
• The reductions in prob. of choosing credit cards
  are much larger.
• More reward credit card transactions are
  estimated to be replaced by paper-based
  transactions.
• Without perception data, the estimation results
  could cause misleading policy implications
• w/o perception data ? policy may reduce welfare
• w perception data ? policy may increase welfare

28

• Figure IV How does the policy affect consumers
  with credit card balance?
• ccwb with credit card balance
• ccwob without credit card balance

29
Results (contd)

• Consumers with credit card balance would also
  reduce the prob. of choosing credit cards after
  the policy is implemented.
• Rewards may potentially increase their credit
  card debts (Removing rewards on credit cards may
  potentially improve their financial situations).

30

• Figure V Removing debit card rewards

31
Effects of Removing Rewards on Debit Card

• Both groups of consumers reduce the probability
  of choosing debit cards (except at fast food,
  which is statistically insignificant).
• The reductions are much smaller than those of
  choosing credit cards under the first policy.
  That implies the majority of reward debit card
  transactions will be replaced by non-reward debit
  card transactions
• For switchers who only have rewards on debit
  cards, they tend to substitute more towards
  paper-based methods.
• For switchers who have rewards on both credit
  cards and debit cards, they tend to substitute
  more towards credit cards.

32

• Figure VI Removing credit and debit card
  rewards

33
Effects of Removing Rewards on Credit Cards and
Debit Cards

• Affects consumers who have rewards on both credit
  cards and debit cards.
• Reduce the prob. of choosing credit cards at all
  types of stores, but the reductions are much
  smaller than under policy 1.
• The prob. of choosing debit cards remains fairly
  unchanged.
• Increase the prob. of choosing paper-based at all
  types.
• Most of them would keep using credit/debit cards
  if there were no rewards!

34

• Table VIII Overall effects of removing rewards
  based on the entire sample

35
Overall effects on the population

• The percentage of transactions that would be
  switched from electronic- to paper-based method
  is likely quite small (at most slightly over 2
  percentage points).
• The actual impact might be even smaller because
• Our sample excludes consumers w/o a bank account,
  credit card, or debit card.
• We do not allow the reward dummies to be
  heterogeneous, and therefore we are estimating
  the upper bound of the direct reward effect.

36
Removing Credit Card Rewards at Grocery Stores
The Effects According to the initial Probability
of Choosing Credit Cards
37

• Conclusion
• (1) Controlling for consumer heterogeneity in
  preferences and choice sets significantly
  improves the fit of our model, and allows us to
  alleviate the endogeneity problem of rewards.
• Removing rewards today would only cause a small
  percentage of consumers switching from electronic
  payment methods to paper-based methods and the
  increase in share of paper-based transactions at
  a given type of store is likely quite small
  (consistent with Australias experience).
• But, we do not claim that the indirect effect of
  rewards is small.
• Removing rewards could potentially reduce
  consumers credit card debts.

38

• Australias experience

39

• Australias experience

40

• Australias experience