accounted for approximately 2 of all retail store transactions.
1
By 1997, debit cards were used in over 1.4 billion transactions. Visa, U.S.A., has announced it hopes to lead the
industry to the point in 2001 when 10 of all consumer payments will be made with debit cards [Keenan 1999].
Using a survey of gasoline credit cardholders, we examined the consumer’s charac- teristics and reasons for using cash, general purpose credit cards i.e., Visa, MasterCard,
Discover, American Express, proprietary credit cards, and debit cards for retail gasoline purchases.
2
The retail gasoline market is a prime example of transactions traditionally dominated by cash, but which have also long accepted plastic payment devices. The rise
of debit cards is causing many retailers, including oil companies, to re-evaluate the purpose and viability of their proprietary in-house credit card programs [Lunt 1996].
We used the survey evidence to assess how debit will impact the use of alternative payment systems. Specifically, we asked the following questions:
1. What factors determine a consumer’s choice between cash and non-cash methods of payment?
2. Among credit card users, what factors impact the choice of proprietary card versus general purpose cards?
3. What are the characteristics of debit users versus non-debit users and how do these characteristics compare with cash, general purpose card users, and gasoline card
users? 4. Are debit cards more likely to impact the frequency of cash or credit card usage?
II. Survey Techniques and Analytical Methodology
We focused our empirical work on the payment choices of consumers who own at least one gasoline credit card and one general purpose credit card. Restricting the survey to
customers who own both types of cards ensures that consumers had a full range of payment choices. The retail gasoline market also provides a homogenous group of
transactions. Inventory models suggest that payment size influences a consumer’s pay- ment choice.
3
By utilizing the gasoline retail environment, we controlled for payment size by experimental design, as the transactions typically fell into a narrow range of 5 to 25.
The Data
We conducted a mail survey of gasoline credit cardholders during the spring and summer of 1992. The Credit Research Center CRC at Purdue University provided questionnaires,
cover letters and envelopes to twelve participating oil companies with proprietary credit card programs.
4
The companies mailed 24,000 questionnaires to samples of their card-
1
This statistic reflects the use of on-line debit, off-line debit, and prepaid debit cards in 1993. [Caskey and Sellon 1994].
2
Previous literature on credit card usage includes Duca and Whitesell 1995, Boeschoten 1992, Lindley et al. 1989, Hirschman 1982, and Martell and Fitts 1981.
3
See Whitesell 1989, 1992 for a review of inventory models of payment choice.
4
Participating companies include: Mobil, Total, United, Citgo, Phillips, Marathon, Cenex, Fina, Exxon, Unocal, Conoco, and Chevron.
410 K. A. Carow and M. E. Staten
holder base.
5
The particular company sending the questionnaire was not identified to the consumer. Consequently, the consumers’ responses relate to their general use of gasoline
credit cards and not to their behavior with respect to a specific company unless they owned only one gasoline card. Questionnaires were returned directly to the CRC. The
overall response rate was 25.9 6451 total surveys. The sample size was reduced by 965, due to incomplete demographic and credit information. The sample was further
reduced by 322 respondents who did not have both a gasoline credit card and a general purpose credit card, leaving a total of 5,164 useable responses.
6
Logit Estimation Model
To analyze the respondent’s choice between debit and non-debit payment methods, we used a multinomial logit model [Greene 1990, Ch. 21]. To analyze the respondent’s
choice of cash, general purpose card, and gasoline card we used the generalized extreme value GEV model also called the nested multinomial logit model. Although the
multinomial logit model of choice is extremely robust, even when its underlying assump- tions are violated, critics state that the property of independence of irrelevant alternatives,
also called the constant ratio rule, is too restrictive. This property implies that if a new option is added to a choice set, the shares of existing options decrease in direct proportion
to original share size. For a gasoline station this new option may be allowing customers to use a general purpose card, where the choices had previously been limited to only using
cash or the company’s own proprietary credit card. Based on the multinomial logit model, the proportion of customers who will choose to use the general purpose card will draw in
proportion to the market share of those who previously used cash and those who previously used a gasoline card. The GEV model relaxes this assumption. Maddala 1984,
Ch. 3 showed that the GEV model is superior to the multinomial logit model when a sequence of decisions can be analyzed.
The use of the GEV model is a sequence of logit models. First, we analyzed the estimates for a model of choice within a given subset gasoline or general purpose credit
card. Second, we used the sum of the utilities for all the items in the subset also called the inclusive value as an explanatory variable in a higher-level model of choice cash
versus any credit card.
III. Analysis of the Payment Choice