RDP 2015-03: The Value of Payment Instruments: Estimating Willingness to Pay and Consumer Surplus 5. Willingness to Pay for the Use of Credit Card Features

5.1 Model

The value that consumers place on debit card and credit card payments appears to be related to the type of card held. To explore this further, we expand the model underlying the econometric analysis in Section 4.1 to include both the characteristics of the respondent and the key features of the credit card that they hold as determinants of a consumer's willingness to pay.

Specifically, the model used to explain a consumer's willingness to pay (WTPi) is:

where xi are an individual's list of personal characteristics and preferences and zi refers to the features of the card: the rewards rebate per dollar spent and the interest rate. Given the structure of the data, the willingness to pay is necessarily measured in relation to the use of a different payment method, in this case cash. Therefore, the constant term, α, captures the net benefit associated with paying with a card instead of using cash for the baseline individual. This reflects the net effect of the various costs and benefits of cash and cards discussed previously.

Given that an individual's characteristics and attitudes to payment methods may be correlated with their willingness to pay, we favour including a wide range of personal characteristics and preferences as controls in the regression (xi). These include household income, a person's employment status, education and characteristics of their household. Information about what factors – security, speed, privacy, the ability to use their own funds or the desire to avoid charges – the person considers when choosing a payment method at the point of sale was also included to help control for unobserved heterogeneity, following Ching and Hayashi (2010). Two additional card characteristics (whether the card is a premium card or charge card such as American Express or Diners Club) are also included as controls.

Again, the question as to a plausible minimum willingness to pay is relevant to understanding the results. In this analysis, the marginal effects are calculated assuming that the minimum willingness to pay is zero.

5.2 Sample Selection and Endogeneity

The model in this section is estimated for the sample of people who hold a credit card and thus the results in this section are limited to this group. We do not speak to how the value of a credit card payment may differ between credit card holders and non-credit card holders.

Another issue is whether there is a relationship between a person's choice of card and potentially unobserved characteristics that also influence a person's willingness to pay for the features of that card. Ching and Hayashi (2010) suggest such endogeneity may arise because: consumers may choose rewards programs because they use credit cards more often and so expect a higher net benefit from rewards after search costs or an annual fee are taken into account; or consumers with rewards programs may have better knowledge of credit card features, causing them to view credit cards more favourably.[13] However, these issues are not likely to cause meaningful endogeneity in our regression for similar reasons as those discussed in Simon et al (2010). First, the payment function of credit cards is very similar to that of debit cards (which are ubiquitous in Australia), and so it is highly unlikely that consumers will learn anything about the non-price features of credit cards from greater use of them. Second, once an individual chooses to hold a credit card, the primary reason to select a card with a rewards program is the monetary benefit of the reward points; which is exactly the feature we wish to value.

Although we conclude that any sample selection or endogeneity issues are minimal, for the interested reader we provide some information about the demographics of credit card and rewards program holders. We model the choice of holding a credit card or not using a probit model, where the latent variable is the utility of holding the card and is determined by demographic and preference characteristics. For the group of credit card holders, we model the choice of reward program membership in a similar fashion. The regression results are presented in Table C1 and some notable results are discussed here.

5.2.1 Credit card holding

Our results on the likelihood of holding a credit card and the likelihood of being a member of a rewards program (see Table C1) are consistent with other papers that link demographic characteristics with credit card holding and use, for example, Klee (2006) and Simon et al (2010). Respondents with high incomes and high levels of education are more likely to hold a credit card, whereas respondents aged under 30 years, as well as respondents who are unemployed or not in the labour force, are less likely to hold a credit card. One reason for this result is that credit card application criteria typically include income and employment status, and may include age. The positive income effect may also reflect the possibility that those who have greater expenditure have greater incentive to own a credit card, as they are more likely to regain the annual fee paid on the card through reward points. Those who are retired are just as likely as those who are employed to hold a credit card.

The results for preferences in Table C1 are intuitive and significant, indicating that these often unobserved personal characteristics are important for determining payment use.[14] Respondents who stated that they value rewards are significantly more likely to own a credit card. In contrast, those that prefer to draw from their own liquidity when making payments and those that value a higher level of privacy are significantly less likely to own credit cards.

5.2.2 Membership of rewards programs

Conditional on holding a credit card, respondents from low-income households are less likely to participate in a rewards program (Table C1). As discussed by Simon et al (2010), this is likely to reflect two factors. First, households with lower incomes may be uncertain of paying off their credit card each month, and so may choose to hold low-rate cards, which are much less likely to have rewards programs. Second, households with lower income may have lower expected expenditure and, therefore, might not expect to gain enough reward points through use of credit cards to offset the cost of holding such cards that are likely to have higher annual fees. It may also be more difficult for lower-income households to be approved for cards with more generous reward programs. Those who value spending from their own funds are also less likely to participate in a rewards program, even controlling for income effects and conditional on holding a credit card. This supports the need to control for these factors in our regressions of willingness to pay.

5.3 Results: Willingness to Pay for Credit Card Features

Table 4 provides the estimation results of the model specification set out in Section 5.1 and methodology detailed in Appendix B. The marginal effects can be interpreted as the average additional willingness to pay across respondents for a given change in the variable of interest, holding the respondents' other characteristics at their observed values. The change will either be a one unit increase for continuous variables or a change from the base case to the alternative for indicator variables. As a crosscheck of the results, we also present the equivalent regression on the willingness to pay for debit cards (where the card feature variables are the respondent's credit card features). We include the card features of the primary credit card of that individual with the expectation that the effect of a respondent's credit card features on the willingness to pay for the use of debit cards will be insignificantly different from zero. This expectation is supported and, further, we find that similar demographic factors affect the willingness to pay for credit cards and debit cards. This correspondence with expectations supports the validity of our interpretation of the results for credit card features.

In line with expectations, the level of rewards rebate for a respondent's credit card has a positive effect on an individual's willingness to pay for credit cards. The marginal effect is estimated to be around 0.6; that is, an increase in the rewards rebate rate of 1 basis point increases the price that individuals in our sample will pay to use their card by 0.6 basis points.

As expected, the rewards rebate does not influence an individual's willingness to pay for debit cards; the estimated marginal effect is zero. This crosscheck provides us with a degree of confidence that the rewards rebate variable is capturing effects specific to the gaining of reward points from the use of credit cards. Contrary to expectations, we find that the posted interest rate on the credit card is not a statistically significant predictor of the willingness to pay for the use of credit cards, although the estimated marginal effect is negative as expected. The effect is not significant even when interacting with the revolver status.

Table 4: Marginal Effect on Willingness to Pay
Sample of credit card holders
Credit card Debit card
Basis points Basis points
Card features
Rewards rebate (basis points) 1.1*** 0.6   0.2 0.0
Interest rate (per cent) −5.2 −2.7   0.9 0.1
Financial status
Revolver 40.0* 24.0   66.8** 12.4
Premium status 41.4* 24.8   55.7* 8.7
Charge card −94.3 −31.6   −1.3 −0.2
Age (base = 30–39 years)
18–29 years 55.0 36.7   81.3* 21.3
40–49 years 11.1 6.2   16.0 1.9
50–64 years −10.7 −5.2   −10.4 −0.9
65+ years −10.0 −4.9   −9.8 −0.9
Household income (base = 3rd quartile)
1st quartile −12.4 −4.7   1.4 0.1
2nd quartile 37.4 19.8   44.3 6.0
4th quartile 45.0* 24.9   40.7 5.2
Education (base = year 12)
Year 11 or below −13.4 −4.8   −13.8 −0.8
Trade certificate 59.7 32.5   65.2 9.2
Diploma 63.7* 35.4   49.6 5.9
Bachelor degree or higher 19.8 8.5   31.7 3.0
Labour force status (base = employed)
Unemployed −23.2 −11.2   3.6 0.6
Not in labour force 10.5 5.9   −2.0 −0.3
Retired 6.1 3.3   −47.9 −4.6
Gender (base = female)
Male −12.9 −7.0   −21.1 −2.8
Life stage (base = couple with children)
Couple, no children 13.4 8.5   −4.0 −0.7
Couple, children left home −28.3 −14.5   6.7 1.2
Single, no children −40.4 −19.4   −69.6* −6.2
Single, children −41.3 −19.7   −18.6 −2.7
Single, children left home 78.1 61.1   76.7 25.4
Other 1.6 0.9   −9.6 −1.5
Location (base = capital city)
Regional −23.3 −11.9   −1.2 −0.2
Speed −24.8 −13.8   −8.5 −1.3
To avoid charges −5.1 −2.8   −19.8 −2.8
Greater security 13.0 7.0   −45.7 −5.9
Draw from own funds −27.9 −15.0   −11.5 −1.7
Greater privacy 5.5 3.0   36.8 6.5
Constant 19.5     −138.9  
Sample size 605   605
Notes: ***, ** and * denote significance at the 1, 5 and 10 per cent level, respectively
(a) Calculated as the average marginal effect on the sample from a one unit change in the continuous independent variables and a change from the base category to specified case for categorical independent variables; the average is calculated after truncation of willingness to pay at zero for any predictions less than zero (Appendix B); significance not shown

Two additional results are of interest. Premium card holders were found to be willing to pay more for the use of both debit cards and credit cards. While premium cards provide additional benefits like concierge services and travel insurance, the majority of such benefits would not be realised in the hypothetical $50 purchase in question. This result is also independent of the price features of the credit card, which were controlled for by including the interest rate and rewards rebate in the regression. One possible explanation of this result is that premium card holders value the ‘prestige’ that is a marketing feature of such cards, although this should only influence the result for credit cards.

Second, revolvers are willing to pay more to use both debit and credit cards.[15] Focusing first on the credit card result, it could be expected that revolvers would be willing to pay less than others to use a credit card, as they are likely to immediately incur interest costs for the purchase. However, if the revolver is liquidity constrained and must use credit, they could be more willing to pay, though we do not believe this would be applicable for the $50 purchase posed in the DCE. A possible explanation could be that the revolver variable is capturing a an unrelated effect not accounted for in our other demographic and preference variables. This is supported by the results on the willingness to pay for the use of debit cards, which should not otherwise be influenced by the credit features of the respondents' credit cards. Additionally, while revolvers have a higher willingness to pay for both credit and debit cards than do transactors, the additional benefit that revolvers place on using a credit card above using a debit card is larger than the same for transactors.

Broadly speaking, demographic and preference characteristics appear to play a limited role – after the inclusion of other variables – in explaining the variation in willingness to pay, with most coefficients being statistically insignificant in determining willingness to pay for either credit or debit cards. Similar factors appear to affect the willingness to pay for debit cards and credit cards. An income effect is suggested as respondents in the highest household income quartile are willing to pay more to use their credit cards relative to the other quartiles. Individuals aged 18–29 years are also willing to pay more than older individuals to use their debit or credit cards. While preferences were important factors for whether respondents held a credit card and joined rewards programs, most did not appear to influence how much respondents were willing to pay to use a credit card. Again, the similarity of the estimated coefficients across the debit card and credit card regressions supports the results of the credit card regression.

Table 5 draws out the estimated willingness to pay for card features. Starting from a baseline where each credit card respondent has a rewards rebate of 40 basis points (but otherwise has their observed characteristics unchanged), a 40 basis point increase in the effective rewards rebate to 80 basis points results in an increased benefit of 31 basis points. While our measure of reward rebates is not sufficiently precise (since it is based on the assumed redemption of $100 gift cards) to conclude that this result is evidence of a one-for-one valuation, it is possible that cardholders do not value the rewards rebate at its full redemption value. Rewards rebate points are not as liquid as cash; cardholders typically need to accumulate enough reward points to be able to benefit from the rewards and the timing of payoff is uncertain. However, this does not preclude some individuals from obtaining one-to-one or greater benefits from rewards.

Table 5: Marginal Willingness to Pay for Card Features
Marginal increase in willingness
to pay Basis points
Credit card
Rewards rebate(b)(basis points) 40 80 31
Interest rates (per cent) 17 13 12
Notes: (a) The baseline respondent's willingness to pay is 98 basis points
(b) The coefficient for the rewards rebate is significant at the 1 per cent level; the rewards rebate is calculated as $100 divided by the spending required to obtain a $100 major store gift card

A 4 per cent decrease in the interest rate from the baseline of 17 per cent to 13 per cent, a level similar to a low-rate card, is associated with a 12 basis point benefit, although this coefficient was not significantly different from zero at even the 10 per cent level. This insignificant result is somewhat surprising given the potential interest savings. However, it may be due to the fact that 73 per cent of respondents reported that they typically pay off their balance before any interest is due.

Although we argue that rewards program membership is not endogenous as argued by Ching and Hayashi (2010), we may speculate on how the results may be affected if this behaviour were true. If true, rewards program holders are likely to have chosen a credit card with features they find desirable and, therefore, value more highly. Consequently, the estimated value of the rewards rebate in our model could be overestimated. Given the expected direction of the bias, the hypothesised bias does not weaken our finding that the incremental value placed on use of credit cards is relatively small.

We also repeat this exercise with the assumption of a log-normal distribution for the willingness to pay for the use of credit cards (see Appendix D for full details). The results are qualitatively similar in terms of the statistical significance and direction of the coefficients regardless of which distribution is assumed. Under a log-normal assumption, the marginal effect of a basis point increase in the rewards rebate is a 0.2 basis point increase in willingness to pay.[16]


Ching and Hayashi (2010) also consider a third source of endogeneity; that consumers who use credit cards more frequently may receive pre-approved offers to join rewards programs. This is not relevant in the context of the scenario, which is based on the known qualities of the existing primary card. [13]

Preferences to avoid charges did not appear to influence credit card ownership. This gives greater validity to the use of a surcharging scenario to gather the willingness to pay for the use of debit cards and credit cards, as respondents that hold credit cards do not appear to have a fundamentally different view of surcharges to those that only hold debit cards. [14]

Revolvers are individuals who typically allow their credit card balance to roll over from month to month and, therefore, incur interest charges; transactors typically pay the balance before the end of the interest-free period. [15]

In comparison to the normal distribution specification with truncation, the effect of the rewards rebate under the log-normal specification may be underestimated because of the skewness in the log-normal distribution. [16]