This post is also available in: Français (French)
Forbes insights publishes a small report on the use of geographic data in insurance that challenges the saying “close enough is good enough” and say that from now on, the GPS approximation is not enough anymore!
Study “Close enough is not good enough”
Geographic data is used by insurers, along with a range of other data to estimate risks and price products. It appears that most of the time, it is a GPS approximation (city, district, postal code, etc.) that is used in the absence of more precise information. Most of the time, the gap between actual and estimated location is insignificant. However, when this is not the case, it can have a significant impact.
A study by Perr & Knight for Pitney Bowes (Note: at first glance, I can not find the original study on the internet ) is focusing on this gap. It appears that about 5% of homeowners contracts and nearly 10% of motor contracts are badly priced because of GPS or geographical approximation.
What is the purpose of improving geographic data?
The temptation would be to think that, because of the law of large numbers, the under-tariffed and over-priced contracts cancel each other out.
However, it’s not a zero sum game! Indeed, when a contract is over-priced (and therefore a priori rather profitable), it is likely that a competitor can come and take it from us. In other words, the contracts that will be actually subscribed will be those for which the price is under-priced, and it is likely to have an over-representation of this type of contracts. Thus, the overall profitability of the portfolio will be affected. This is the very principle of adverse selection. The more time passes, the more the portfolio will contain (in proportion) undervalued risks.
Analysis of results
The study recalculated the premium gap represented by the use of precise geographical location data in relation to the data taken into account by an insurer. This is a contract in Florida, and it is the distance from the coast that was found to be relevant in this case.
- 5.7% of premiums were poorly priced, 3.8% under tariffed and 1.9% over-priced
- Corrections ranged from + 86% ($ 2800) to -46% ($ 2100)
The same exercise is performed by correcting the location of the main parking place of the vehicle.
- 10% of premiums have been corrected, half up, half down.
- Overall, the recalculation slightly reduces the average premium (Note: in a competitive market, this type of information can have a real impact on the acquisition of new customers! )
- Variations ranged from -25% ($ 710) to + 34% ($ 601).
Conclusion of the report “Close enough is not good enough”
The information to remember is of several kinds:
- Geographical approximation is not always enough
- The number of contracts involved is not necessarily high, but the financial impact is important
- Underpricing is not fully offset by over-pricing
- Return on investment can be fast.
The potential impact of this type of correction is very easy to deploy. Indeed, it does not necessarily come into question pricing models, it just correct the data used by these models.
This raises several notable elements:
- We can never say enough, data quality has a cost, but it’s a worthwhile investment!
- In a competitive market with small margins, any niche is to be exploited , especially when it allows both to better manage its portfolio and increase sales.
- GPS trackers are now integrated into smartphones and the accuracy has been significantly improved. In addition, maps are common (starting with Google or Open Street map). It therefore becomes anachronistic not to exploit correctly this information.