/ Home


Third Party Motor Insurance in Sweden

Keywords: Poisson regression, gamma regression, compound Poisson, generalized linear models, offset, dispersion modelling, exact zeros.


The data give details of third party motor insurance claims in Sweden for the year 1977.

"In Sweden all motor insurance companies apply identical risk arguments to classify customers, and thus their portfolios and their claims statistics can be combined. The data were compiled by a Swedish Committee on the Analysis of Risk Premium in Motor Insurance. The Committee was asked to look into the problem of analyzing the real influence on claims of the risk arguments and to compare this structure with the actual tariff."

Variable Description

Kilometres Kilometres travelled per year
1: < 1000
2: 1000-15000
3: 15000-20000
4: 20000-25000
5: > 25000
Zone Geographical zone
1: Stockholm, Göteborg, Malmö with surroundings
2: Other large cities with surroundings
3: Smaller cities with surroundings in southern Sweden
4: Rural areas in southern Sweden
5: Smaller cities with surroundings in northern Sweden
6: Rural areas in northern Sweden
7: Gotland
Bonus No claims bonus. Equal to the number of years, plus one, since last claim
Make 1-8 represent eight different common car models. All other models are combined in class 9
Insured Number of insured in policy-years
Claims Number of claims
Payment Total value of payments in Skr

Make 4 is the Volkswagen 1200, which was discontinued shortly after 1977. The other makes could not be identified because of the potential for the data to impact on sales of those cars.


Data File (tab-delimited text)


Hallin, M., and Ingenbleek, J.-F. (1983). The Swedish automobile portfolio in 1977. A statistical study. Scandinavian Actuarial Journal, 49-64. The actual data are not listed in this paper.
Andrews, D. F., and Herzberg, A. M. (1985). Data. A collection of problems from many fields for the student and reseach worker. Springer, New York, pages 413-421. Only the data from Zone 1 are listed.
The data was obtained electronically from the Statlib database.


The number of claims in each category can be treated as Poisson to a good approximation. The amount of each claim can be treated as gamma. The total payout is therefore compound Poisson.



Home - About Us - Contact Us
Copyright © Gordon Smyth