Lower prices produce higher demand… or do they? A bank’s direct marketing to holders of “free” checking accounts show that a large discount on 60% APR overdrafts reduces overdraft usage, especially when bundled with a discount on debit card or auto-debit transactions. In contrast, messages mentioning overdraft availability without mentioning price increase usage. Neither change persists long after messages stop. These results do not square easily with classical models of consumer choice and firm competition. Instead they support behavioral models where consumers both underestimate and are inattentive to overdraft costs, and firms respond by shrouding overdraft prices in equilibrium.
Why would a bank hide information on overdraft costs? After all, a classically rational consumer would simply infer that shrouded prices are high prices. But recent behavioral theories show that shrouded and high prices can persist if consumers tend to underestimate their add-on costs and firms cannot profit from de-biasing consumers with more transparent pricing or information about competitors’ high add-on prices (Gabaix and Laibson 2006; Grubb 2015; Heidhues, Köszegi, and Murooka 2017).
What drives overdraft pricing, advertising, and usage?
Retail banking in much of the world: “free if nonnegative balance, very expensive if in overdraft”.
the experiment varies promotions Yapi Kredi sent via SMS from September-December 2012 to 108,000 existing checking account clients who had not overdrafted in the previous few months.
|with debit card/auto bill payment||B||D|
H1: The key comparison is between overdraft-promoting messages that mention price (A & B) and those that do not (C & D). Our key hypothesis is that drawing attention to the cost of overdrafting will depress demand for it. We test this hypothesis by comparing overdraft usage, during the experimental period, between customers sent an Overdraft Interest Discount message and customers sent an Overdraft Availability message (A vs C) (debit card and auto bill payment were separate conditions I grouped in the matrix).
H2: A second behavioral hypothesis is that bundling the overdraft discount with a discount on debit card or automated bill-payment usage will further depress demand (A vs B)
H3: Third, we hypothesize that promoting overdraft availability, without mentioning price, will change demand. We test whether and how promoting overdraft availability changes demand by comparing overdraft usage among customers sent the Overdraft Availability message to customers sent messages that promote only debit card or automated bill payment usage and do not mention overdraft at all. (C vs D)
H4: To test the dynamics of attention and overdraft behavior, we examine data from the post-campaign period (January-May 2013). Treatment effects will persist if consumer learning or attention re: add-ons is durable. Treatment effects will not persist if consumers quickly forget about add-ons or only attend to them when induced to by external stimuli like advertising.
We still find that mentioning overdraft price lowers overdraft demand. E.g., the likelihood of overdrafting during the experiment is 1.2 percentage points lower (se=0.4 pp) for those receiving the discount offer relative to those receiving a message that mentions overdraft without mentioning its price, on a baseline likelihood of 31%. (H1 not true, C overdrafts more than A). These results support the hypothesis that drawing attention to overdraft costs reduced demand (even while offering a 50% discount!).
Offering the overdraft discount alone reduces overdraft likelihood by only 0.7pp (se=0.5) relative to messages that mention overdraft without mentioning price, while the bundled reductions are 1.4pp for automated bill payment (se=0.5) and 1.7pp for debit card (se=0.5). (H2; A overdafts more than B). Do All Promotional Discounts Backfire? No. Offering the debit discount alone (cell D) weakly increase debit card usage, by 0.5pp (se=0.4), the auto-pay discount alone increases auto-pay signup (by 0.4pp, se=0.1).
The overdraft availability message increases overdraft likelihood, by about 0.9 pp (se=0.4). (H3, overdraft C > overdaft D).
Ad H4: Treatment effects do not persist: the lack of persistent effects suggests that consumer learning and/or attention concerning overdrafts depreciates quickly, and hence that advertising and de-biasing campaigns must persist to be effective.
One message is not enough to generate an effect, and repeating messaging does influence demand, with diminishing marginal effects from messaging every 10 vs. 20 days.
Altogether our results are consistent with the models of shrouded equilibrium and limited/reactive consumer attention. In particular, they support:
- the key modeling assumption that consumers tend to underestimate add-on costs (if consumers’ estimates were unbiased then offering a discount would weakly increase demand);
- the key assumption that firms lack incentives to unshroud prices;
- a key prediction of reactive attention models that consumers respond differently when advertising highlights different add-on attributes (price or availability).
our results should give pause to third parties seeking to improve overdraft markets with messages (like social marketing campaigns) that draw attention to overdraft costs. To fix ideas, imagine messaging around the theme of “Beware of big overdraft fees!”, delivered by an entity that might actually benefit from unshrouding; e.g., a regulator, a firm with social objectives or a product-differentiation strategy, or a personal financial management service.
Our results also suggest that unshrouding could be quite costly to sustain, since its effects do not persist. Moreover, our results suggest that incumbent suppliers could effectively counter unshrouding campaigns by advertising non-price attributes (like availability/credit lines in our case). Hence we are sympathetic to Heidhues, Köszegi, and Murooka’s conjecture that third-parties, or deviating firms, will be outgunned in a messaging arms race with incumbent add-on suppliers.
I should probably read more on Heidhues & Koszegi to understand that:
Ethics of Experimenting with High-Cost Credit
We are frequently asked in seminars whether researchers should partner with a lender that is seeking to sell more high-interest rate loans. We think yes, for four key reasons:
- First, an ethical concern here presumes that high-cost consumer credit harms consumers. We emphasize the presumption; extensive research on this question suggests that a different assumption is warranted– (weakly) beneficial impacts for consumers (Karlan and Zinman 2010; Zinman 2014; Banerjee, Karlan, and Zinman 2015).
- Second, YK’s advertising was truthful and its terms were competitive. Thus, combining the first and second points, the experiment was not trying to convince consumers to accept a bad deal in either absolute terms or compared to market alternatives.
- Third, YK was going to promote overdraft usage among its existing customers with or without the participation of the research team; we helped convince bank management to feature prices despite its skepticism about the effectiveness of past overdraft price promotions.
- Fourth, YK and the research team contracted ex-ante that the academic co-authors would have unrestricted intellectual freedom to report the results and disseminate them publicly to benefit regulators and further scientific knowledge.