RDP 2009-04: Price Incentives and Consumer Payment Behaviour 3. Consumer Payments Literature

The academic literature on consumers' use of payment methods remains relatively scant, although this field of payments research has grown considerably in recent times. Earlier literature in this area focused on the influence of demographic variables and payment instrument attributes on consumers' choice of payment instrument; see Stavins (2001) and Klee (2006), for example. More recent papers, such as by Borzekowski, Kiser and Ahmed (2008), Ching and Hayashi (2008) and Zinman (2008), build upon this earlier literature to incorporate price-related variables into econometric modelling. Humphrey, Kim and Vale (2001) use aggregate transaction and price data for Norway to estimate own-price, cross-price and substitution elasticities for cash, debit cards and cheques. They conclude that consumers are quite sensitive to explicit pricing for payment services and also find that debit cards are quite a strong substitute for cheques.

The approach taken in this paper is similar to the approach in Borzekowski et al (2008), which estimates a series of probit models of debit card holding and use. That paper used 2004 data from a special component of the Reuters/University of Michigan Surveys of Consumers. A nationally representative sample of around 1,500 households were asked about their use of PIN and signature debit cards, and open-ended questions to elicit information on the underlying reasons households choose to use debit cards (for example, security and convenience attributes). In addition to demographic variables, data on fees some banks charged cardholders for PIN debit transactions were also incorporated into probit models of debit card use. The results of these models indicate that the imposition of fees for PIN debit transactions encourages households to use signature debit cards instead of PIN debit cards but reduces the overall likelihood of households using debit cards. In fact, the paper finds that a transaction fee of less than 2 per cent of the average transaction amount is associated with a 12 per cent reduction in the likelihood of using a debit card. The authors note that the price response to merchant surcharging is likely to be larger than this, given that transaction fees are generally not disclosed to the consumer at the point of sale.

Ching and Hayashi (2008) estimate a series of multinominal logit models to analyse the effects of demographic variables (including individuals' adoption of technology), payment method attributes (including convenience and security) and loyalty program rewards on consumers' choice of payment methods at the point of sale. The paper uses data from the 2005/2006 Study of Consumer Payment Preferences, conducted by the American Bankers Association and Dove Consulting. Around 3,000 consumers completed the questionnaire, mainly via the internet. While not a nationally representative survey, it provides a richer source of payments information (for example, data on consumers' credit card loyalty program participation) than can be obtained from sources such as the Federal Reserve's Survey of Consumer Finances. The questionnaire asked consumers about their most frequently used payment method across five merchant categories.

Ching and Hayashi run a hypothetical experiment of abolishing loyalty program rewards and conclude that only a small percentage of consumers would switch from electronic to paper-based payment methods, while consumers would increase their use of debit cards and decrease the use of credit cards. Their paper also finds that those consumers who normally incur an interest charge on their credit card balance are far less likely to make payments with a credit card. Zinman (2008) also estimates probit models of the use of payment instruments and finds that revolvers are at least 21 per cent more likely to use debit cards than transactors, based on 1995 to 2004 data from the Survey of Consumer Finances. Zinman's results suggest that debit cards are a close substitute for credit cards, cash and cheques.

This paper builds on the existing literature that examines the influence of price-related variables on the use of payment instruments. It takes advantage of a unique transaction-level database that provides rich data not typically available in the questionnaire-style approach underpinning much of the existing literature. In this regard, our dataset is most similar to that used in Bounie and François (2006); a paper that estimates a multinominal logit model of cash, cheque and payment card use incorporating demographic and transaction explanatory variables (for example, transaction amount and merchant type). Our paper extends this type of work by incorporating price-related variables such as loyalty program participation and consumers' transactor/revolver status. The dataset used in Bounie and François (2006) consisted of transaction-level data on 17,000 payments made by around 1,400 French adults over an eight-day period in 2005, captured through a payments diary. Consumers were asked to record the details of the payments they made – the amount, merchant type, type of good, any payment restrictions faced, and the channel used (for example, face-to-face, internet or telephone). The paper discusses the importance of factors such as the transaction amount and merchant environment in determining the payment instrument used; for example, they find that controlling for other factors, the probability of cash use declines as the transaction value increases.