This is exactly the reason. Big-name insurers trade in volume business. The more they standardize the product, the cheaper they can sell it - not a whole lot different to any other industry. That's not to say they don't attempt to have sophisticated pricing models that ingest all kinds of underwriting data points, but there are costs to that complexity too.

There is a broad market in specialist insurance which caters to the non-volume cases. It just costs more because you have to expend more effort-per-sale (and of course because non-volume also tends to mean that the risks genuinely are higher - e.g. homeowners with prior subsidence claims).

Given how easy it would be to data mine this out of the claim history I find this explanation hard to buy. There must be a whole set of specific claim types that show no correlation with future claims. If you can’t find these in your data you are not looking very hard.

Of course if no one else is looking why give up the easy money - after all it is not like the customer can buy a new policy without the massive markup. Collusion does not need to be explicit, it can be accomplished just through everyone using the same systems and rules.