Payment Card Rewards: Consumer Choice

The study's primary objective is to quantify the direct effects of rewards card programs on consumer payment choices, specifically focusing on in-store transactions. It aims to determine how consumers would alter their payment preferences if rewards were removed from credit and/or debit cards. The research utilizes a dataset that incorporates consumer perceptions of various payment methods (including speed, convenience, and safety) and perceived merchant acceptance rates to account for the heterogeneity in consumer preferences and choice sets. This detailed approach allows for a more nuanced understanding of how rewards influence consumer behavior than previous studies that lacked this level of granularity in data collection. The findings are directly relevant to current policy debates regarding the fee structures of payment card networks and the impact of interchange fees, ultimately seeking to provide empirical evidence on how rewards programs influence consumer payment choice and contribute to the overall economic welfare. The research is particularly timely given ongoing discussions about the fairness and efficiency of reward structures within the payment card industry.

The study uses a unique dataset that contains detailed information on consumer-perceived attributes of payment methods, including factors like speed, convenience, and safety, as well as perceived acceptance of each method by various merchants. This rich data allows the researchers to control for consumer heterogeneity and variation in choice sets. A crucial methodological aspect addresses the endogeneity problem; the potential correlation between card rewards and unobserved consumer preferences. To mitigate this, the study employs the method proposed by Harris and Keane (1999), using attitudinal data (consumer perceptions) to control for unobserved heterogeneity. Four model specifications are estimated to explore the impact of including perceived attributes and whether to assume a homogeneous or heterogeneous consumer choice set. The researchers found that incorporating consumer perceptions significantly improved model fit and provided more robust estimates, highlighting the importance of understanding consumer attitudes in the analysis of payment choices. This sophisticated approach enables a more accurate assessment of the impact of rewards programs on payment behavior compared to previous studies.

The core of the research involves policy experiments simulating the removal of rewards from credit and debit cards. The results consistently show that the removal of credit card rewards has a more substantial impact on consumer behavior than the removal of debit card rewards. Importantly, only a small percentage of consumers are predicted to switch to paper-based payment methods (checks and cash) if rewards are eliminated. The study finds that removing credit card rewards would primarily lead to a shift towards non-rewards credit cards, while removing debit card rewards would have a more varied impact across retail types, with some increase in debit card usage in certain sectors. Removing rewards from both card types would decrease credit card transactions but increase debit card transactions. The research underscores the importance of considering consumer perceptions, demonstrating that models excluding these factors may yield misleading conclusions about the effects of removing rewards. The findings are compared to the experience in Australia after interchange fee regulations in 2003, which demonstrates that even significantly reduced rewards values didn't greatly alter credit card usage. This comparative analysis strengthens the study's conclusions about the relatively low sensitivity of consumers to changes in reward programs.

The study acknowledges several limitations. The sample excludes consumers without bank accounts or cards (primarily cash users), potentially underestimating the aggregate effects of removing rewards. Additionally, the assumption that all consumers distribute transactions across retail types equally may be unrealistic, potentially affecting the accuracy of some findings. The relatively small proportion of reward card holders in the sample, compared with some other surveys, raises concerns about potential underestimation of the overall effects. Despite these limitations, the study concludes that the direct effect of credit and debit card rewards on consumer payment choice is moderate. The majority of consumers receiving rewards would likely continue using credit/debit cards even without rewards. The findings are consistent with the Australian experience after interchange fee reform. The authors suggest that their estimated effect of rewards might be upwardly biased due to limitations in controlling for consumer heterogeneity in reward sensitivity, and that the policy simulation results should be interpreted as an upper bound on the impact of removing rewards.

A core methodological challenge in assessing the impact of rewards programs is the potential for endogeneity bias. This arises because the decision to obtain a rewards card might be correlated with unobserved consumer characteristics affecting payment choices. For instance, consumers who value convenience might be more likely to both seek rewards cards and frequently use credit or debit cards regardless of the rewards program. This correlation could lead to an overestimation of the true effect of rewards programs on payment decisions. To address this, the study adopts a novel approach using attitudinal data. Instead of relying solely on instrumental variables, a common econometric technique, the researchers leverage the rich dataset including consumer perceptions (speed, convenience, safety, budgetary impact etc.) regarding each payment method. By incorporating these consumer perceptions into the model as controls, the study aims to account for the unobserved heterogeneity in consumer preferences and obtain more accurate and reliable estimates of the effects of rewards programs. This approach offers an alternative to instrumental variables, particularly advantageous for non-linear models such as the multinomial logit model used here.

Another crucial aspect of the methodology is dealing with choice set variation. Consumers may perceive that only a subset of payment methods is available to them in a given retail setting; this is not necessarily the same for every consumer. Ignoring this variation in perceived choice sets can lead to biased parameter estimates. The existing economics literature often makes the simplifying assumption that all payment methods are available to all consumers, which isn't always realistic. While panel data can help address choice set variation, it necessitates strong assumptions about choice set formation. The researchers’ data set, uniquely containing information on each individual's perceived choice set, allows them to avoid these strong assumptions and the potential for misspecifying a model of choice set formation. This careful consideration of choice sets directly enhances the validity and reliability of the estimates concerning the impact of rewards on consumer payment choices. Four different model specifications are explored; two assuming homogeneous choice sets (all consumers have access to all payment methods) and two with heterogeneous choice sets (consumer choice sets vary based on individual perceptions). The superior goodness-of-fit of the heterogeneous choice set model underscores the importance of considering individual consumer perceptions in the analysis.

The study employs four model specifications to analyze the impact of rewards on consumer payment choice. These specifications vary in two key aspects: whether consumer perceptions of payment methods are included as variables, and whether the choice set is considered homogeneous (all payment methods are available to all consumers) or heterogeneous (consumer choice sets vary based on perceived availability). The inclusion of consumer perceptions improves the model's fit significantly. The results suggest that consumer perceptions capture a large amount of unobserved consumer heterogeneity in preferences for payment methods, highlighting the importance of integrating this data to improve the accuracy of model estimations. Furthermore, modeling heterogeneous choice sets also leads to significant improvements in log-likelihood, indicating that allowing for the variation in consumers' perceived availability of payment methods better captures the complexity of payment decision-making. This emphasizes the importance of incorporating both perceived attributes and choice set heterogeneity when estimating the effects of rewards on consumer payment choices.

Policy simulations focusing on the removal of credit card rewards reveal a significant impact on consumer payment behavior. The study finds that removing these rewards would lead to a reduction in credit card usage across all retail types studied (grocery, department, discount, drug, and fast-food stores). However, the majority of consumers who currently use rewards credit cards would continue to use credit cards, even without the rewards incentive. The reduction in credit card transactions is distributed differently across retail types, with grocery and discount stores showing an almost even split between a shift to debit cards and paper-based methods, while department stores experience a larger shift towards debit card usage. In contrast, drugstores and fast-food restaurants see a more pronounced shift towards paper-based methods. The magnitude of the reduction in credit card usage varies depending on the retail sector; grocery and department stores show the largest reductions (around 10 percentage points), likely due to higher average transaction values resulting in greater reward points earned per transaction. Overall, the reduction in the probability of choosing credit cards remains moderate, highlighting that many consumers value credit card convenience despite the removal of rewards.

The second policy experiment focuses on the removal of debit card rewards. The results indicate a smaller impact on consumer payment choices compared to the removal of credit card rewards. Consumers who received debit card rewards only, or both credit and debit card rewards, were analyzed separately. While there's a reduction in the probability of choosing credit cards across all retail types in both groups, it's notably smaller than the effect observed when credit card rewards are removed. The effect on debit card usage is more complex and varies across retail types. Some retail sectors show a decrease in debit card usage, while others experience an increase, illustrating the nuances of consumer preferences and how rewards impact payment choices across different purchase contexts. The probability of using paper-based methods increases across all retail types in this scenario, but the increase is comparatively smaller than the observed decrease in credit card transactions. These results further underscore the comparatively stronger influence of credit card rewards programs on payment choices, compared to debit card rewards.

The third policy experiment simulates the removal of both credit and debit card rewards. This analysis primarily focuses on consumers who previously received rewards on both card types. The results indicate a decrease in credit card usage and an increase in debit card usage across all retail sectors. This suggests that even without rewards, consumers still exhibit a preference for electronic payment methods over paper-based options, but with a notable shift away from credit cards toward debit cards. The reduction in credit card usage is again moderate, aligning with the results of the previous two experiments. The increase in the share of paper-based transactions remains relatively small, reinforcing the observation that most consumers prioritize electronic payment methods for convenience. This finding further supports the overall conclusion that a significant portion of consumers will maintain their current card usage patterns even without the rewards incentives. The changes in payment choices, even under a complete removal of rewards, remain relatively moderate.

While the individual policy experiments provide insights into consumer behavior under different scenarios, the study also examines the overall effects of removing rewards across all five retail categories (grocery, department, discount, drug, and fast food). The aggregated results reveal that the removal of credit card rewards would reduce the credit card share by a moderate amount (up to slightly above 3 percentage points depending on the retail type). The reduction in credit card transactions is distributed variably across debit and paper-based methods. The removal of debit card rewards demonstrates a much smaller effect on overall market share. The combined removal of both credit and debit card rewards causes a decrease in credit card share, but with an increase in debit card use. The increase in paper-based methods remains moderate in all cases. Three limitations are acknowledged: the sample excludes cash-only consumers, the assumption of equal transaction distribution across retail types may be unrealistic, and the proportion of rewards cardholders in the sample might be an underestimation. These limitations suggest that the observed effects of removing rewards could potentially be smaller than what the results show.

The study's findings are compared to the Australian experience following the 2003 interchange fee regulations, where reward values were significantly reduced. Notably, credit card usage remained largely unchanged after the implementation of these regulations, showing a parallel to the study's findings. While the regulated networks in Australia saw a small decrease in market share, other credit card networks saw a slight increase. Overall, the study concludes that the direct effect of credit and debit card rewards on consumer payment choice is moderate. The majority of consumers who currently receive rewards on their credit or debit cards would likely continue using credit and debit cards even if rewards were no longer offered. However, the authors note that the estimated average direct effect of rewards might be upwardly biased since consumer heterogeneity in reward sensitivity couldn’t be fully controlled for in the analysis; the policy simulation results should therefore be seen as an upper bound estimate.