Why Are (Some) Consumers (Finally) Writing Fewer Checks?: The Role of Payment Characteristics Scott Schuh and Joanna Stavins Federal Reserve Bank of Boston October 26, 2007 Economics of Payment Systems Telecom Paris
2 Total Volume of Paper Checks in the US SOURCE: Benton, Blair, Crowe, and Schuh. (2007) “The Boston Fed Study of Consumer Behavior and Payment Choice: A Survey of Federal Reserve System Employees.” Federal Reserve Bank of Boston Public Policy Discussion Paper #07-1.
3 Shift to Electronics SOURCE: Survey of Consumer Finance (1995, 2004). Consumers are shifting from paper checks to electronic payments
4 Motivation & Overview Limited research on consumer payment choice Most econometric studies: Schuh and Stavins econometric study: Few other studies also use payment characteristics to explain payment choice (Carow and Staten 1999, Jonker 2005, Klee 2006, Borzekowski, Kiser and Ahmed 2007), but focus on subset of payments, small set of characteristics, lack of individual data.
5 Payment Choice Variables (Y ij ) ADOPTION (0 or 1, logit): USE/SHARE (OLS):
6 Consumer Payment Data Surveys Partial characteristics data: –Boston Fed (FRS employees) –Boston Fed/AARP (U.S. consumers) Complete characteristics data: –Boston Fed/RAND survey (U.S. consumers) To be collected in –Other sources Dove Consulting/ABA (U.S. consumers) FirstData (U.S. consumers) Jonker (2005) (Dutch consumers) Existing data on consumer payment behavior are inadequate for testing models of payment demand –Public data: few, infrequent, limited payments variables –Private data: proprietary or expensive (or both), not representative
7 Federal Reserve System Survey Data: Demographics 6 age categories: <25, 25-34, 35-44, 45-54, 55-64, over 65 4 education categories: HS or less, some college, college, post-graduate 2 homeownership (“wealth”) categories: own, rent 4 income categories: <$50K, $50-75K, $75-100K, over $100K
8 Federal Reserve System Survey Data: Payment Characteristics We measure consumers’ assessments of: –Cost (out-of-pocket only) –Convenience (or ease) –Safety –Privacy –Errors –Timing/control –Record keeping NOTE: RAND Survey will have expanded, refined list
9 Payment Characteristics Relative CHAR for each payment method: –Credit cards vs. checks –Debit cards vs. checks –ACH vs. checks –Online banking vs. checks –Stored value cards vs. checks Asked if better (+1), worse (-1) or same (0) as check for each characteristic type
10 Payment Characteristics We DERIVE characteristics (k) relative to other payment methods (j, j’) from OBSERVED characteristics relative to checks: DERIVED relative characteristics may not reveal valid differences when payment methods have the same OBSERVED characteristic rating relative to check (see diagonal below) start with subtract from it …
11 Example: Online Bill Payment SOURCE: AARP (2006). Unconditional age profiles of adoption and use differ
12 Example: Online Bill Payment SOURCE: AARP (2006). Average use is similar across ages but varies widely within age; characteristics help explain this large within-group variation
13 Econometric model summary CHAR add a lot –Higher R 2 when CHAR included True for observed, derived or both types –Tests show that all CHAR should be included Especially in share regressions –CHAR reduce significance of demographics
14 Model Evaluation: Model Fit Adoption of Payment Methods (Pseudo R^2) Payment TypeObservationsFull Model Original CHAR Only Derived CHAR OnlyDEMOG Only Credit Card Debit Card ACH Internet Banking Share of Payment Methods (R^2) Payment TypeObservationsFull Model Original CHAR Only Derived CHAR OnlyDEMOG Only Check Credit Card Debit Card ACH Internet Banking
15 Payment characteristics reduce significance of demographics Significance in Econometric Model of Adoption Without CharacteristicsWith Characteristics Credit Card Debit CardACH Online Banking Credit Card Debit CardACH Online Banking Age Education Income Percent of Data Explained
16 Model Evaluation: Restriction Tests Adoption of Payment Methods (P values) Payment TypeObservations Exclude DEMOG Exclude Derived CHAR Exclude Derived and Original CHAR Exclude Original CHAR from model of DEMOG and Original CHAR Credit Card Debit Card ACH Internet Banking Share of Payment Methods (P values) Payment TypeObservations Exclude DEMOG Exclude Derived CHAR Exclude Derived and Original CHAR Exclude Original CHAR from model of DEMOG and Original CHAR Check Credit Card Debit Card ACH Internet Banking
17 Adoption Regression Results Demographics: –very few significant variables OBSERVED characteristics (relative to checks): –ease (+) and timing (+) highly significant –cost (+) significant for CC, OBP but not DC, ACH –safety (+) significant for DC only DERIVED characteristics (relative to other methods) –ease –timing –cost –record keeping
18 Use Regression Results Demographics: –Age: young use fewer checks, more CC, DC –Education: graduate degrees use fewer checks, more CC OBSERVED characteristics (relative to checks): –ease (+) important, especially for DC –record keeping (+) important for all but DC –errors (-) important for CC DERIVED characteristics: –ease; cost; record keeping; timing –privacy and safety only important in OBP
19 Assessments of Characteristics Much Worse WorseSameBetter Much Better SOURCE: Benton et al. (2007). Characteristic AdoptersNon-adopters Credit Card Debit CardACH Stored- Value Card Online Bill Pay Credit Card Debit CardACH Stored- Value Card Online Bill Pay Cost Convenience Safety Privacy Errors Timing Record keeping Assessments vary widely between adopters & non-adopters (Relative to paper checks)
20 Econometric Concerns Potential problems: Model with time (subscript t): for several potential reasons
21 Endogeneity of Characteristics Likely endogenous: –Cost Likely exogenous: –Errors, timing/control Mixed?: –Convenience, security, privacy, record keeping
22 Econometric Solutions 1.Instrumental variable (IV) estimation –Shortage of IV candidates A few survey questions may be valid –DEMOG not promising IV’s 2.Data collection from multiple surveys? –Import instruments from other payments data –LHS, RHS variables from different surveys (a solution used in marketing literature)
23 DEMOG as Instruments? CHAR not explained well by DEMOG in 1 st stage: R 2 are all below 0.05 DEMOG explain very little cross-section variation in CHAR (same as Jonker (2005)) cost and ease slightly better explained by demographics; AGE almost uniformly significant, INCOME not usually significant DEMOG mostly unimportant for ACH, OB
24 IV Estimation Results Point estimates generally remain about the same as in non-IV estimation but… Not much is statistically significant (as is usual for IV estimation)
25 Theoretical Musings Key questions to be answered by theory: What are the primary payment methods? What are the main payment characteristics? How do we model payment demand?
26 Conclusions Payment characteristics are much more important than demographics in explaining consumer payment demand Consumer payment decisions consistent with their assessments of characteristics Existing data on consumer payment behavior are inadequate for testing models of payment demand Need to develop better theory and data for research on consumer payment demand –Boston Fed/RAND new-and-improved survey ( )