Tag Archives: Modeling

Survey: Predictive Modeling Boosts Insurers’ Bottom Line

Towers Watson just released its annual survey on predictive modeling with some notable results.

The percentage of U.S. property/casualty executives reporting a positive impact on profitability has dramatically increased over the past six years, while the breadth and depth of predictive modeling applications has grown.

Some 87 percent of property/casualty executives report that predictive modeling had a favorable impact on profitability in 2014, an increase of eight percentage points over 2013. The increase continues a pattern of growth over several years, and is up significantly from 57 percent six years ago.

A positive impact on rate accuracy helps explain the boost in profits, Towers Watson said.

In fact, the percentage of carriers citing a positive impact on rate accuracy has increased every year since 2010, when 70 percent cited a positive impact. By 2014, 98 percent of insurers reported that predictive modeling has improved their rate accuracy.

More accurate rates also positively impact loss ratios, which have improved in parallel, according to p/c insurance executives. In 2014, 91 percent cited the favorable impact of predictive modeling on loss ratios, an increase of 14 percentage points over 2013.

The survey shows the use of predictive modeling in risk selection and rating/pricing has increased significantly for all lines of business over the last year, continuing a long-term trend.

For personal lines, auto saw the largest increase with 97 percent of participants saying they used predictive modeling in underwriting/risk selection or rating/pricing in 2014, up from 80 percent in 2013 — a 17 percentage-point increase.

Even more  noteworthy is the increased  use of predictive modeling in commercial lines.

For commercial property/commercial multiperil (CMP)/business-owner peril (BOP) as well as  commercial auto the use of modeling increased 19 percentage points  — to 51 percent and 41 percent respectively, year-to-year.

But it was specialty commercial lines that saw the largest increase, where 44 percent of p/c executives said they use predictive modeling in risk selection and rating/pricing in 2014, up from 13 percent in 2013 — a 31 percentage point increase.

us-insurers-bottom-line-predictive-modeling-boosts-profits-fig1

While the survey suggests that insurers are increasingly comfortable with predictive modeling and using it in a growing number of capacities, more progress is still possible, according to Towers Watson.

Treating data as an asset and more effectively using predictive modeling applications to improve claim and other functional results could make a significant difference in the profitability of insurance companies, it suggests.

More on the survey results  in this  Insurance Journal article.

Towers Watson gauged the views of 52 U.S. insurance executives in personal lines and commercial lines carriers for the survey.

 

 

What Insurers Can Learn From Errant Forecasts

Most actuaries know about projections that go awry, so we have quite a bit of sympathy for the weather forecasters who missed the mark early this week, says I.I.I.’s Jim Lynch:

Weather forecasts have improved dramatically in the past generation, but this storm was odd. Usually a blizzard is huge. On a weather map, it looks like a big bear lurching toward a city.

This storm was relatively small but intense where it struck. On a map, it looked like a balloon, and the forecasters’ job was to figure out where the balloon would pop. They were 75 miles off. It turned out they over-relied on a model — the European model, which had served them well forecasting superstorm Sandy, according to this NorthJersey.com post mortem.

There are lessons for the insurance industry from the errant forecast and the (as it turns out) needless shutdown of New York City in the face of the blizzard that wasn’t:

  • – Models aren’t perfect. Actuaries, like weather forecasters, have multiple forecasting models. Like forecasters, actuaries have to know the pros and cons of each model and how much to rely on each one given the circumstances. Actuaries and forecasters both bake their own experience into their final predictions.
    Property catastrophe models are considerably cruder than the typical weather forecasting model. By crude I mean less accurate. Cat models project extreme events, where data are sparse and everything that happens has an oversize influence on everything else that is happening. Woe to the insurer that over-relies on cat models, something cat modelers themselves say regularly.
  • – It’s hard to pick up the flag once you have planted it. Forecasters suspected late Monday that New York City would be spared the brunt of the storm, but acknowledge now they were reluctant to make too big a change because it could hurt their credibility, particularly if the new forecast had proved too mild. This is a human failing both by the forecaster and its recipient, both of whom worry about crying wolf.
    The tendency also helps explain why it is hard to project market turns, whether they are from growth to recession or from rising insurance rates to falling.
  • – Policymakers have egg on their faces today, but they appear to have been following sound risk management principles. It’s not unusual to prepare for disasters that don’t happen, something to think about next time you unbuckle a seatbelt or unlock a door. The scale this week was much larger, but the principle was the same. Needlessly closing a subway is better than stranding hundreds on it, and the occasional forecaster’s error is certainly better than the crude prognostication that gave us the Galveston hurricane or the Schoolchildren’s Blizzard.

I.I.I. has Facts and Statistics about U.S. catastrophes in general and winter storms in particular.

Check out this timelapse video of the blizzard hitting Boston: