Tweet
http://online.wsj.com/article/SB1000...084755588.html
WSJ Life Ins.jpg
WSJ Life Ins.jpg
Deloitte Consulting LLP's effort to persuade life insurers that marketing data can size up people's longevity faces an obstacle: How to know if the method works, when the proof—policyholder deaths—is years away?
So the company developed a test involving 60,000 insurance applicants at the U.S. arm of British insurer Aviva PLC. Deloitte detailed the findings this spring at an industry seminar.
Here's how the test worked. Aviva used traditional underwriting methods—including costly blood and urine tests—to assess the 60,000 applicants. Aviva sorted those people into various risk categories. In addition, some of the individuals had been declined a policy.
Deloitte then took the 60,000 cases and tried to replicate Aviva's traditional underwriting decisions with a new methodology not reliant on blood work.
First, it divided the 60,000 into two equal groups. (Names were withheld.) For the first 30,000, it studied the traditional underwriting decisions and set out to build a data-only "predictive model" that could reach similar conclusions.
Its model didn't use the blood and urine tests. It would be designed to use data such as personal and family medical histories, as detailed on written insurance-application forms. Deloitte's team also had access to industry-shared information from past insurance applications and motor-vehicle reports. On top of that, it used the consumer-marketing data.
The consumer data came from Equifax Inc.'s marketing-services unit, since acquired by Alliance Data Systems Corp. The data noted which of hundreds of attributes applied to each individual—likely hobbies, TV-viewing habits, income estimates. In Deloitte's final predictive model, the nontraditional consumer-marketing data represented about 37% of the predictive ability, it says.
Then came the real test: Running the predictive model on the second batch of 30,000 applicants, to see if it could accurately replicate the underwriters' original assessments. Aviva and Deloitte judged the test largely successful. "The use of third-party data was persuasive across the board in all cases," said John Currier, chief actuary for Aviva USA.
So the company developed a test involving 60,000 insurance applicants at the U.S. arm of British insurer Aviva PLC. Deloitte detailed the findings this spring at an industry seminar.
Here's how the test worked. Aviva used traditional underwriting methods—including costly blood and urine tests—to assess the 60,000 applicants. Aviva sorted those people into various risk categories. In addition, some of the individuals had been declined a policy.
Deloitte then took the 60,000 cases and tried to replicate Aviva's traditional underwriting decisions with a new methodology not reliant on blood work.
First, it divided the 60,000 into two equal groups. (Names were withheld.) For the first 30,000, it studied the traditional underwriting decisions and set out to build a data-only "predictive model" that could reach similar conclusions.
Its model didn't use the blood and urine tests. It would be designed to use data such as personal and family medical histories, as detailed on written insurance-application forms. Deloitte's team also had access to industry-shared information from past insurance applications and motor-vehicle reports. On top of that, it used the consumer-marketing data.
The consumer data came from Equifax Inc.'s marketing-services unit, since acquired by Alliance Data Systems Corp. The data noted which of hundreds of attributes applied to each individual—likely hobbies, TV-viewing habits, income estimates. In Deloitte's final predictive model, the nontraditional consumer-marketing data represented about 37% of the predictive ability, it says.
Then came the real test: Running the predictive model on the second batch of 30,000 applicants, to see if it could accurately replicate the underwriters' original assessments. Aviva and Deloitte judged the test largely successful. "The use of third-party data was persuasive across the board in all cases," said John Currier, chief actuary for Aviva USA.
Comment