Do Price Comparison Sites Increase Premiums?

Hypothesis: the rising cost of car insurance is actually driven by marketing costs.

Gross premiums have increased 14% in the past year, and the car insurance industry would have us believe this is driven by the 'rising cost of personal injury claims and fraud'. But that would suggest net premiums (the proportion of the gross premium that actually covers the insured event) are being priced wrongly, which I find difficult to believe. Actuaries must be fairly good at predicting accident rates, deaths and injuries and so on by now. And these actually appear to be in decline, according to the Office of National Statistics research published in April 2009:

"The total number of deaths in road accidents fell by 7 per cent to 2,946 in 2007 from 3,172 in 2006. However, the number of fatalities has remained fairly constant over the last ten years...

The total number of road casualties of all severities fell by 4 per cent between 2006 and 2007 to approximately 248,000 in Great Britain. This compares with an annual average of approximately 320,000 for the years 1994-98.

The decline in the casualty rate, which takes into account the volume of traffic on the roads, has been much steeper. In 1967 there were 199 casualties per 100 million vehicle kilometres. By 2007 this had declined to 48 per 100 million vehicle kilometres."

So I wonder if there's a variable cost in there that's proving difficult to control. A chief culprit might be marketing. We're certainly being inundated with TV advertisements for insurance price comparison sites, so I wonder if that is in turn being paid for by higher 'costs per click' associated with advertising on those sites? While some insurers are refusing to list their products on the price comparison sites, they may still face competition for key search terms, for example.

Needless to say, I haven't yet been able to find much public analysis on this, but there is a helpful post on analysing the efficiency of cost per click by The Catalsyst. It's easy to over-spend, and the competition may force you to.

If my hypothesis is right, the case is building for semantic web applications to enable people's computers to find deals simply by interrogating insurers' computers - without all the expensive advertising noise.