Flood Cover Take-Up Below Plan: Behavioural Diagnosis and Repricing
Northern Shield, a UK general insurer, launched a household flood product in a coastal region eighteen months after a one-in-fifty-year flood. The government's subsidised scheme had been withdrawn. The actuarial pricing model gives an expected annual loss cost of £240 per policy and expenses of £80; with a target combined ratio of 95% the indicated premium is about £337, and the product was launched at £340. A stated-preference survey of 1,200 households suggested that 35% would buy at that price. Six months after launch, take-up is 14% and the renewal rate is 62% against a plan of 80%. Intermediaries report that customers say 'it won't happen again in my lifetime' and that many compare the price with a competitor offering £290 cover with a £500 excess. The board wants to know whether the price is wrong, the product is wrong, or the model is wrong.
1. Using behavioural economics, explain the gap between the survey's 35% and the observed 14% take-up, and state the implications for the demand assumptions in the pricing model.
Read the worked answer
Stated-preference surveys systematically overstate purchase of low-probability, high-severity risks. Several mechanisms explain the gap:
- Hypothetical bias: saying yes costs nothing. Revealed preference at £340 is the only reliable evidence.
- Availability and probability neglect: the flood is memorable, but people infer 'it happened, so it won't happen again soon', or focus on the vivid outcome rather than its probability.
- Optimism bias: 'my house is on higher ground' or 'the defences will hold'.
- Present bias and loss aversion: the premium is a certain, immediate loss; the benefit is distant and uncertain. Framing the £500 excess as a potential loss makes the £290 offer more attractive.
- Anchoring and social proof: £290 is a strong anchor, and if neighbours have not bought, buying feels unusual.
Implication: the demand curve in the model is wrong. Take-up elasticity is higher than assumed, and price is not the only lever. The model should use observed take-up and renewal data segmented by risk and channel, and treat the survey as qualitative evidence only.
2. Recommend product and pricing changes that could raise take-up without underpricing the risk, and explain the behavioural mechanism for each.
Read the worked answer
Options and mechanisms:
- Reduce friction and bundle: offer cover at mortgage or utility purchase, exploiting the status quo/default effect. This is likely the largest single lever.
- Monthly instalments: mental accounting and present bias mean £28 a month feels smaller than £340 upfront.
- Reframe the cover: sell 'flood recovery' - alternative accommodation, rebuild management, temporary business support - rather than a probability. This addresses probability neglect and loss aversion.
- Adjust the excess: a £500 excess is a salient potential loss. A £250 excess at a premium between £290 and £340 may work if the technical cost supports it.
- Use social norms: 'X% of households in your postcode have cover' can raise take-up.
- Offer a multi-year premium guarantee to address uncertainty aversion.
Do not match £290. With loss cost £240 and expenses £80, technical cost is £320; £290 gives a combined ratio above 110% before any risk margin. Cutting price would grow a loss-making book. Test changes with controlled experiments rather than repricing on judgement.
3. Explain how the observed take-up and behavioural factors affect adverse selection and the capital the insurer should hold.
Read the worked answer
Low take-up may produce an unrepresentative pool:
- Self-selection: households that know they are exposed, or have had a loss, are more likely to buy, so the insured pool's average risk exceeds the market average. A 14% take-up could produce a worse loss ratio than the 35% plan assumed.
- Anti-selection spiral: if the insurer raises prices to cover the worse pool, remaining low-risk customers leave, worsening the pool again.
The capital model should include demand risk and parameter uncertainty, not only claims volatility. Model take-up by risk decile, allowing correlation between risk and propensity to buy. Run a scenario of a second flood within three years, testing both claims capital and lapse behaviour. Add a margin to the pricing and reserving bases for possible adverse selection. Behavioural factors also create moral hazard where maintenance of flood defences is unobservable, so conditions and excesses should be designed with that in mind.