One of the shockers for me in attending last week's ACORD London Forum was learning about the lack of standardization when it comes to working with catastrophe models. I was also surprised to hear that many companies work with multiple models to hedge their bets. It sounds like the industry is still dealing more with wishful thinking than hard science when it comes to their biggest exposure.
ACORD has a number of players participating in a standards-setting initiative on catastrophe data, but panelists at last week's forum lamented that not everyone is on board.
"It is very important that all model vendors are in the boat. Not all of them are at this point," noted Peter Hausmann, head of research and development for catastrophe perils at Swiss Re Group. During an ACORD panel on "Addressing the Challenges of Cat Exposure Reporting," he added that "with more carriers using multiple models to price their risks, we need data standards to assure we are comparing apples to apples."
Paul Nunn, manager of the loss modeling unit at Lloyd's, added that because his is a subscription market, with the same risk being shopped to many syndicates, "you have 20-to-30 teams doing the same scrubbing to clean up the data. It's a terrible waste of time and resources. We're looking to drive out some of that cost for thankless tasks and standardize data to allow time to be better spent analytically."
On the other hand, Mr. Nunn warned, "we have to be mindful to be flexible because the reason our spreadsheets are so messy is that there is a lot of data that might not fit neatly into any category. We don't want to constrain underwriters or deny them relevant information."
I still came away from this session quite alarmed, and was not relieved after speaking with a number of underwriters in attendance following the panel session about their use of cat models.
For those using multiple models so they can provide their boards with a range of worst-case scenarios, again, it sounds more like reading tea leaves than actuarial predictability. The underwriting department is desperate for some quantitative measurement so they can reassure their managers and board members that they have their catastrophe exposures under control.
But I was even more alarmed to encounter a number of underwriters whose firms rely on only one cat model. It seems clear that the various models often conflict--thus providing a "range" of potential losses for those using multiple systems. (If all the models came up with the same numbers, there would be no need to use more than one--you would just pick one according to the best cost and service offered, right?)
If you only go with one, how do you choose which to believe? You are literally betting your company's solvency on the outcome.
In any case, it seems like a no-brainer to me for everyone involved to get on board with the standards effort here. If you can't believe or compare your numbers, it makes it that much harder to come up with any definitive answers on exposure. Garbage in, garbage out, right?
However, much like the ongoing battle to convince all carriers to jump on the SEMCI bandwagon (that's single-entry, multi-company interface), so that independent agents don't have to do multiple data entry to shop their risks, it doesn't surprise me that some will hold out against standardization of their business.
Attendees told me some fear that cat data might become commoditized, making their particular product less "valuable" (translation--too costly, or perhaps no longer even viable), while lowering the entry barriers to new players.
Still, I can't see this particular standards drive being held up for years, the way SEMCI has remained more of a Holy Grail than a reality. It just makes too much sense, and cat exposures are just too big to leave to chance.
Sounds like a no-brainer to me. What do you folks make of all this?
Comments (2)
Sure, some companies have multiple cat models to align their business spread in areas of concern. Then they ask engineering to back up their predictions with hard facts, in the event the models are right--or in some cases, wrong.
For example, we looked at losses that were ZIP code based in certain states, and then based loss predictors on building construction, age, height and, when possible, building codes in effect at time of construction.
We further narrowed them down by geo-coding, distance from navigable bodies of water, elevations and all that goes with potential for hurricane and severe windstorm damage.
Then when all was said and done, decided to cut and run after the first big windstorm, even though our losses were minimal overall.
Maybe the cat models were valuable, and maybe just an exercise in futility? I'd bet they were helpful, but in this case, I'd place my money on engineering expertise and the good old savvy of putting pencil to paper and knowing what you have in terms of each and every property, and the ability to withstand what you think might be coming your way.
Not that it would make a big difference in a Category 5-plus hurricane or an F5 tornado path, but at least a good idea in a "normal" storm season.
Posted by BJ | October 30, 2007 11:49 AM
Posted on October 30, 2007 11:49
In the financial markets, financial data is calculated by one party and then sent out for all to evaluate as they see fit. It is, after all, how individuals interpret those numbers and how they correlate the information to their own or others' experiences. That, in fact, is also the same principle in underwriting.
The degree of success come from the experience and intellectual power behind the analysis of the data.
In the financial world, Peter Lynch and Warren Buffet both have and had the same numbers as the rest of the fund investors, but the way they think through risk showed in the numbers they produced.
The insurance industry needs to deal with some process issues, not just cut cost but to supply that information to the underwriters for their analysis.
Unfortunately, in the quest for process improvements by boxing things up in nice process models, there seems to be a bleed into the thought process that goes in to the very value each company and underwriter adds to the process.
That may be fine in non-catastrophe, homogeneous business, but not in cat or other complicated business.
Models are a great tool and the more views you get from different models the better. However, the models do not come alive until you have a great mind think through what the numbers really mean. Standardize the numbers, let the reviewer add the difference.
Posted by Rosanne Bachman | November 26, 2007 6:49 AM
Posted on November 26, 2007 06:49