Slowing down to add it up: Using behavioural insights to support decision-making about add-on insurance

Participants completed an online survey with elements mimicking a real-world scenario, in which they purchased a primary product (e.g., plane ticket, mobile phone) and were shown an advertisement for an add-on insurance product (e.g., travel insurance, extended warranty, consumer credit insurance) towards the end of the purchase. In the treatment conditions, participants were shown the ASIC information sheet immediately before the sale of the add-on product. The information sheet was the key intervention.

2020-07-29

2020-08-19

Whether respondents indicate that they will 'buy' the add-on insurance or not.

This is a binary measure. All respondents will be required to decide whether they will 'buy' or 'not buy' add-on insurance in a hypothetical purchasing scenario. We will calculate sample proportions from this response.

Whether respondents indicate on the information sheet that they would like to 'opt-out' of follow-up about the add-on insurance

This is a binary measure. When participants are shown the information sheet the first time (in the hypothetical scenario), they may or may not click on the 'opt-out' box. We will calculate sample proportions from this response.

This is a multi-celled framed-field experiment, randomised at the individual level. There will be 21 different cells that respondents can be allocated to.

This study uses a 2x3 factorial design: We will have two independent variables (IVs) creating six versions of the information sheet. In addition, we have a no-information-sheet control condition.

IV 1: Colour of information sheet. Participants will receive either a blue information sheet (education condition), or a red information sheet (warning condition).

IV 2: Claims ratio. The information sheet will either show no claims ratio for the add on product, or it will show that the add-on product has a low claims ratio, or it will show a moderate claims ratio.

Finally, there are three separate purchasing scenarios (phone, travel, loan). 'Scenario' is not an independent variable; we are averaging across scenarios for our primary analyses.

Randomisation done by the survey software (Qualtrics).

individual

No

1

6300 individuals

300 per cell of the design (21 cells = 3 x (3 x 2 + 1))

Cohen's h (standardized effect size for comparing proportions) = 0.09. This corresponds to the intervention having an effect of about 3 to 5 percentage points.

Preanalysis plan BETA 041