Evaluating nudge policies in sales processes of high risk financial products

This study investigates whether presenting additional explanatory materials on top of the standard product brochure affects customers’ investment decisions when purchasing Equity-Linked Securities (ELS), a high risk financial product. Sixty branches are randomly assigned to four groups (A to D), with 15 branches per group. Arm A serves as the control and maintains business as usual, while Arms B, C, and D receive the interventions described below. All activities take place at the bank’s trust product sales counters. The study focuses on KRW denominated ELS sold during the intervention period whose underlying structure combines KOSPI200, S&P500, and Eurostoxx50. To comply with regulatory requirements for investment solicitation, the added materials are delivered uniformly by supplementing the standard brochure with an additional explanatory insert, and data are collected from transactions during the three month intervention window.

The first intervention addresses potential behavioral biases induced by the profit and loss (P&L) charts commonly used in bank presentations and emphasizes the salience of potential losses when customers consider subscribing to ELS. In existing brochures, the chart at times uses inconsistent or split vertical scales between gain and loss regions, which can lead investors to overestimate gains and underestimate risks. Such design can trigger probability misjudgments and create a dark nudge toward overstating upside and understating downside. The intervention therefore presents loss information first, corrects scale distortions by using consistent axes for gains and losses, and clarifies labels and annotations (for example, explicitly noting that “8.2 percent per annum” is an annualized rate and adding plain language explanations for states such as early redemption success or failure).

The second intervention provides a comparative table across ELS variants, highlighting that higher first barrier levels imply a lower likelihood of early redemption and therefore higher risk. For multiple underlying, high risk ELS, a higher barrier typically offers a higher fixed coupon yet comes with greater risk, and this trade off should be made explicit. Existing sales routines often rely on historical simulations for individual products and do not clearly convey the systematic relation between barrier levels and redemption probabilities. The insert therefore summarizes, under identical underlying and fee conditions, key items by first barrier level, namely 75, 80, 85, and 90. It presents the fixed coupon, the loss range, the barrier level, and the historical probability of first redemption in a single table for side by side comparison.

The third intervention introduces a comparative table between ELS and ELB/DLB. Whereas the second intervention compares barrier levels within ELS, the third contrasts ELS that are not principal protected with ELB/DLB that are principal protected, which are relatively safer except under extraordinary circumstances such as issuer default during a severe financial crisis. Previously, there was no explicit guideline requiring advisors to present principal protected alternatives alongside ELS. This intervention therefore presents, in one place, a comparison of fixed return, loss range, redemption conditions, underlying assets, and risk grades, enabling customers to clearly understand the risk and return differences between ELS and ELB/DLB.

2025-08-01

2025-10-31

[Fixed return (coupon rate) of contracted ELS]
Unlike deposits with fixed returns at the time of sale, ELS determines payoffs automatically based on the path and level of the underlying indices after contracting, while the contracted coupon rate is specified at contracting. We define this outcome as the annualized coupon rate (%) of the ELS actually contracted, measured at the transaction level. The sample is restricted to contracts during the intervention period at the trust product counters that are KRW denominated ELS with combined underlyings of KOSPI200, S&P500, and Eurostoxx50.

[Contract amount of ELS]
This outcome is the contract amount in KRW for the ELS actually executed. It is observed at the transaction level and uses the initially recorded amount for eligible ELS contracts during the intervention period.

[No purchase rate after an ELS consultation visit]
This visit level indicator equals 1 if a customer who comes for an ELS consultation does not contract to an ELS during that visit, and 0 otherwise. Due to the additional recording burden on bank staff, this outcome is measured only during the final two weeks of the intervention, rather than over the entire experimental period. In addition, applicable regulations prohibit retaining personal information of non purchasing customers. Accordingly, the measure is recorded without personally identifiable information as aggregates by branch and day, with the denominator being all ELS consultation visits during the two week window and the numerator being the number of those visits that result in no contract.

Sixty branches are randomly assigned to four groups A to D with fifteen branches per group, and data are collected during the three month intervention. Our primary outcomes are designed to quantify how supplemental materials affect risk taking through the coupon rate, transaction scale through the contract amount, and avoidance of purchase through the no purchase rate. The coupon rate reflects the relative risk level of the chosen ELS under a common menu, while the contract amount captures the economic magnitude of the decision. The no purchase rate is collected only during the final two weeks of the intervention and is recorded without personally identifiable information as aggregates by branch and day due to staff workload and privacy regulations. This approach allows inclusion of the full universe of visitors, including non purchasers. Accordingly, we use the measures that span the full intervention period, the coupon rate and the contract amount, to identify effects on risk and scale, and we use the two week no purchase rate as a complementary indicator of deferral or avoidance. Taken together, the three outcomes provide a coherent view of how the interventions shift risk preferences, transaction size, and final contract choices.

[Heterogeneous Treatment Effects (Seniors, Repeat ELS)]
To capture differential impacts across customer types, we predefine two subgroups: seniors aged 65 or above and repeat ELS customers with prior ELS purchase history. Using interaction terms or separate subgroup analyses, we assess where the intervention has larger impacts on key outcomes. The senior dummy equals 1 if age is 65 or above and 0 otherwise. The age 65 threshold is defined in current financial regulations as the policy cutoff for vulnerable consumers, who receive additional protections such as extra paperwork when purchasing financial products, which makes 65 a salient and operational boundary. The repeat ELS dummy equals 1 if the customer has any prior ELS contract and 0 for first time customers. Repeat customers may tilt toward riskier choices or pay less attention to loss information, so this subgroup is central for assessing debiasing effects.

[First/Final barrier of contracted ELS]
For each executed contract, we record the first barrier level, one of 75, 80, 85, or 90, at the transaction level. Under a common underlying, a higher first barrier such as 90 generally implies a higher fixed coupon and higher risk. We also record the final barrier, either 60 or 65. If any index falls below the final barrier at maturity, losses arise. The loss range is 40 to 100 percent when the final barrier is 60 and 35 to 100 percent when it is 65. In practice, first barrier levels of 75 or 80 map to a final barrier of 65, while first barrier levels of 85 or 90 map to a final barrier of 60. A lower final barrier of 60 is interpreted as a riskier choice. These variables serve as auxiliary indicators of risk taking.

[Withdrawal within the cooling-off window]
Under current rules, ELS contracts can be withdrawn within three business days after execution. This outcome is a binary indicator recorded in the contract row, capturing whether the ELS is withdrawn during the cooling off period. Unlike no purchase after consultation, this measures cases where a contract is executed but subsequently withdrawn.

[Portfolio diversification among ELS contractors]
Beyond reducing high risk selections, we assess whether the intervention encourages substitution toward safer products such as ELB or DLB or co purchases that diversify the portfolio during the same visit. In particular, we consider increases in safe asset selection, concurrent selection, and greater variety in product composition as evidence of improved diversification. The operational definition will be finalized separately.

Sixty branches are randomly assigned to four groups A to D with fifteen branches per group, and data are collected during the three month intervention. These secondary outcomes show for whom, when, and through which channels the interventions operate. The heterogeneity analysis focuses on seniors aged 65 or above and on repeat ELS customers, two groups that are salient for policy and behavior, to test whether information provision delivers greater protection for vulnerable customers and debiasing for experienced but risk prone customers. The first and final barriers capture risk choices at the contract level, and the final barrier of 60 versus 65 conveys differences in the loss range and thus in risk intensity. The cooling off withdrawal outcome distinguishes post execution reversals from no purchase after consultation and reveals whether information prompts reconsideration after commitment. Portfolio diversification looks beyond risk reduction to assess substitution toward safer assets and diversification, providing evidence of qualitative improvement in consumer protection and sales practices.

[Group Assignment and Treatment Overview]
We conduct a field study with 60 sales branches of Customer Bank, organized into four groups (A to D) with 15 branches per group.

Control (A) maintains business as usual, and staff are not informed about the experiment.
Treatment 1 (B) appends to the standard brochure an amended profit and loss chart that presents loss information first and corrects scale distortions.
Treatment 2 (C) adds an ELS comparison table clarifying that a higher first barrier raises the fixed coupon but lowers the probability of early redemption and widens the loss range.
Treatment 3 (D) appends a comparison table between ELS and principal protected products (ELB, DLB) to make risk and return differences clear.

[Sales procedures and supplemental information timing]
All branches follow the same baseline counseling flow. Upon visit, staff assess the customer’s risk profile and determine the investor suitability grade before recommending products. The ELS products studied are the bank’s highest risk grade and, under suitability rules, are eligible only for customers with the most aggressive investor profile. In treatment branches, during recommendation and brochure explanation, staff present the group specific supplemental insert alongside the standard brochure and use it in counseling. Apart from the group specific materials, all other sales procedures remain unchanged.

[Weekly cycle of experimental progress]
The weekly cycle for the supplemental inserts is as follows. Each Monday, the bank finalizes the product lineup for the following week; on Tuesday, the bank sends the product list to the research team. On Wednesday, the research team returns the supplemental inserts to the bank. After the bank obtains the regulatory document review, Friday features distribution of group appropriate brochures to staff (control materials for A and supplemental versions for B, C, and D). The following Monday, products launch, and staff use the distributed group specific brochures in on site sales and counseling.

Not available

Using April to June 2025 branch-level operational data and customer demographic characteristics at each of the 60 branches, we implement computer based random assignment into four groups (A to D) with 15 branches per group. After assignment, we conduct balance F-tests and adopt an assignment that satisfies F-test p-values ≥ 0.10 across the following variables.

Number of staff selling high-risk products, Seoul indicator, Risk-adjusted operating income, Branch headcount (size), Number of customers, ELS sales count, highest risk product sales count, Corporate customer ratio, Average customer age, Average customer male ratio, Customer annual income distribution, and customer investment-to-total-assets ratio distribution.

Branch (cluster) level randomization. Treatment is assigned at the branch (cluster) level.

Yes

A total of 60 branches (clusters), with 15 branches per arm across the four groups A through D.

Based on the preceding three months and the first month of the trial, we expect 700 to 800 observations per month, yielding 2,100 to 2,400 observations over the three month intervention window. Observations primarily refer to eligible KRW ELS contracts; visit level auxiliary outcomes are measured over a separate window.

15 branches per arm (A, B, C, and D). In observation counts, we anticipate approximately 525 to 600 per arm.

The study features branch level (cluster) randomization with transaction level outcomes. Inference will use cluster robust standard errors. We will include month or day fixed effects to improve precision by reducing residual variance.