VEHICLE RATE GROUP METHODOLOGY An overview of Manufacturers Suggested Retail Price ( MSRP ) and Canadian Loss Experience Automobile Rating ( CLEAR ) Prepared by: INSURANCE BUREAU OF CANADA For: NEWFOUNDLAND AND LABRADOR BOARD OF COMMISSIONERS OF PUBLIC UTILITIES 2005 AUTOMOBILE INSURANCE BENCHMARK HEARING OCTOBER 28, 2004
Table of Contents 1. Introduction... Page 1 History... 1 Definitions of terms... 1 2. Methodology... 3 MSRP... 3 CLEAR... 3 3. Implementation Issues... 7 CLEAR overall premium affect... 7 CLEAR individual policy dislocation... 7 CLEAR balancing to rate level neutrality... 7 MSRP depreciation... 8 4. Appendix A... 9
1. INTRODUCTION History Insurers have historically recognized that different makes and models of automobiles present different risks and have differentiated these risks in various manners, such as price of the vehicle (i.e. MSRP) and type of vehicle (e.g. convertible, sports car). In 1970s, the MSRP system was amended to adjust vehicle rate groups which were based on price alone with the actual experience by different vehicle type, relying on U.S. experience available from Highway Loss Data Institute in Washington, D.C.. This methodology was developed by Insurers Advisory Organization ( IAO ) and was subsequently made the responsibility of Vehicle Code Service ( VCS ), an industry committee formed by IAO, IBC, Association of Canadian Insurers ( ACI ) and Groupement Assureurs des Automobiles ( GAA ). The administration of the methodology through VCS ensured industry-wide support and distribution of the rate group tables. In 1989, the Ontario Automobile Insurance Board ( OAIB ) held hearings in Ontario into Automobile insurance including the classification systems in use by insurers. The industry, through VCS, made a proposal for refining the vehicle rate group methodology and base it on Canadian loss cost experience of individual vehicle make and models. The OAIB accepted the proposal and the industry responded by forming the Vehicle Information Centre of Canada to design and implement a Canadian Experience based vehicle rating system. In 1994 both the Ontario Insurance Commission (successor organization to OAIB) and the Manitoba Public Insurance Corporation accepted the new methodology, named Canadian Loss Experience Automobile Rating system, and implementation of CLEAR started. In 1999 VICC became part of Insurance Information Centre of Canada which subsequently became part of Insurance Bureau of Canada. Definition of terms Claim Frequency Claim Frequency is the average number of claims that occur in a given period per insured vehicle. It is calculated as the number of claims divided by the number of vehicles insured. Insurance Bureau of Canada - 1 - October 28, 2004
Claim Severity Claims severity is the average incurred loss amount per claim; it is a measure of the average cost of automobile claims. It is calculated as the total incurred loss amount divided by the number of claims. Collision Collision coverage provides coverage (subject to a deductible) for damages to the insured s own vehicle arising out a collision. Comprehensive Comprehensive coverage provides coverage (subject to a deductible) for damages to the insured s own vehicle arising out of a peril other than collision (e.g. theft, vandalism, flood, hail, fire, etc.). Loss Cost Loss cost is the average incurred loss amount in a period per insured vehicle. It is the product of Claim frequency and Claim severity. Normalization Normalization is a statistical technique to remove certain effects from data in order to standardize the meaning of the data. Rate group The rate group is a numerical designation of a grouping of vehicles having similar insurance risks. Rate group differential factor A differential factor is a factor associated with a Rate group and is used in premium calculation to reflect the applicable risk of vehicles in the Rate group. Regression Regression is a mathematical technique which can be used to explain or predict an expected outcome. Statistical Model A statistical model is a mathematical statement that describes the relationships between a given set of factors. Insurance Bureau of Canada - 2 - October 28, 2004
2. METHODOLOGY MSRP Methodology This methodology establishes a single vehicle rate group for physical damage coverages (i.e. Collision and Comprehensive). The initial rate group is determined based on the Manufacturers Suggested Retail Price ( MSRP ), this initial rate group is subsequently amended based on the estimated future loss cost determined by CLEAR. This adjustment could vary the rate group by +/- 2 rate groups. In determining MSRP vehicle rate groups, IBC utilizes a differential scale, where each step on the scale (representing a rate group) is approximately 20 percent different from its adjacent step, this changes with higher rate group where the difference between each step is 20 points. Although this scale is used by IBC to establish vehicle rate groups, individual insurers frequently use their own differential scale. MSRP rate groups theoretically comprise up to 50 groups (01-50) with rate group 9 being considered the base (i.e. 1.000 on the differential scale). Practically, rate groups 01 19 represent more than 99% of insured vehicles Prior to 2001, MSRP rate groups would be depreciated initially after three years and then every two years. This depreciation took the form of lowering the applicable rate group by 1. In 2001 a revised balance procedure was introduced to CLEAR and the depreciation of MSRP was halted to provide consistency with the change to CLEAR; this procedure is described in Section 3 (CLEAR balancing to rate level neutral). CLEAR Methodology CLEAR rate groups are calculated separately for Collision, Comprehensive and Accident Benefits reflecting the different risks presented under these coverages. For Collision and Comprehensive separate statistical models are developed for estimating frequency and for estimating severity, these estimates are subsequently combined to estimate expected loss cost. For Accident Benefits a statistical model for Frequency only is used as a surrogate for Loss cost since the variability in severity of the benefits under Accident Benefits and the availability of collateral benefits hinders the opportunity to establish reliable regression models for severity. Insurance Bureau of Canada - 3 - October 28, 2004
Data Sources and Adjustments The advisory rate groups assigned by the Canadian Loss Experience Automobile Rating (CLEAR) are based on actual vehicle claim experience. The data used are Canadian private passenger insurance claim experience, along with certain vehicle characteristics. Insurance claim experience is obtained from Canada s official statistical agencies, namely Groupement des Assureurs Automobiles (GAA), the Insurance Bureau of Canada (IBC), the Insurance Corporation of British Columbia (ICBC), Manitoba Public Insurance (MPI), Saskatchewan Government Insurance (SGI), and La Société de l Assurance Automobile du Quebec ( SAAQ ). Vehicle characteristics are obtained directly from vehicle manufacturers and importers. Each year, the data are normalized in an attempt to statistically remove influences that are not directly related to the vehicles themselves and are already accounted for in the setting of premiums. Specifically, normalization is accomplished in an attempt to eliminate the influence of such factors as territory, driving records, vehicle use, deductibles and discounts. Methodology Overview All existing vehicle rate groups are re-calculated annually, considering both depreciation and recent loss cost experience of individual vehicle make and model. A six-step process is utilized to develop CLEAR s annual advisory ratings. 1. The first task is to estimate the number of written exposures anticipated during the publication year for every private passenger make/model/model year of vehicle. The exposure projection considers observed aging patterns for model years, these observed changes ( link factors ) are then used to estimate written exposures for each model year for the next year, then anticipated automobile sales numbers are obtained, combined with observed data for the first six month of the current year and, utilized to project exposures for new model years. 2. Second, normalized claim frequencies and severities for each modelyear of every vehicle, for each coverage, are estimated with regression formulae that were derived from the combination of the vehicles characteristics and reported experience. Recently some significant changes were made to CLEAR to improve Insurance Bureau of Canada - 4 - October 28, 2004
the predictive accuracy of the models. Previously, separate models had been formulated for cars and trucks. However, given the number of trucks in today s private passenger fleet, and their increasing commonality with cars, it was deemed that better estimates would be obtained if both were treated within the same models. Similarly, a separate adjustment used to be made for convertible vehicles, the increased number of such vehicles has allowed us to treat them as unique body styles in the development of the new models. Three additional factors, namely the presence or absence of air bags, anti-lock brakes and IBC approved theft deterrent systems, were used in the regression formulas and, to be more consistent with the current literature, a power-to-weight ratio was used instead of a weight-to-power ratio. These new measures allowed the models to be more responsive to current vehicle technology. The five most recently available accident years experience (i.e., accident years 1996 thru 2000) were used to develop the new models and, to reduce the likelihood of adverse effects due to data outliers, accident year/vehicle/model year data were not included unless they represented at least 100 vehicles. This appeared to effectively eliminate most outlier effects. One of the standard measures for evaluating regression model adequacy is the models R 2, a value which represents the percentage of variance the model was able to predict. The table, below, shows the R 2 s that were obtained each for the new statistical models. New Models Coverage Frequency Severity Collision 0.63 0.64 Comprehensive 0.55 0.50 Accident Benefits 0.72 N/A 3. The third step of the process is the annual adjustment of estimated frequencies and severities to correct for discrepancies between the statistical estimates and actually reported experience. Our full credibility used to adjust our statistical estimates with actual experience is determined to achieve a 95.0% confidence level with an error of no more than 5% (i.e. if the actual experience data is sufficient to represent full credibility, the estimated frequency Insurance Bureau of Canada - 5 - October 28, 2004
and/or severity closely approximates the observed frequency/ severity, respectively). 4. Estimated loss costs for each private passenger vehicle model/model year combination are obtained by computing the product of the adjusted claim frequencies and severities. Then, using the projected fleet, weighted average loss costs for each coverage are calculated in order to estimate the relative loss costs for each separate vehicle make, model and model year. 5. The acceptable range for our statistical models necessitated imposing boundaries on some of the continuous variables and as such, an additional adjustment was developed to reflect the additional risk associated with very expensive vehicles. The starting point for risk loading is dependent upon the data excluded during the development of the statistical models. The starting point, is thus $65,000 in 1988 adjusted dollars and 1/10 of a relative loss cost unit was added for each $5,000 the MSRP (in 1988 dollars) exceeded $40,000 (in 1988 dollars). Various caps and controls exist in CLEAR to ensure that large changes will only occur gradually. 6. Finally, based on the estimated relative loss costs, the vehicles are grouped into common rate groups with corresponding rate differentials, which are then reviewed for reasonableness and unjustified reversals. Insurance Bureau of Canada - 6 - October 28, 2004
3. IMPLEMENTATION ISSUES CLEAR overall premium effect CLEAR vehicle rate groups are used by insurers in calculating policy premiums, each rate group is associated with a differential factor used in premium calculations. Since the CLEAR rate groups (and differential factors) may be considerably different from the previous rate groups used, and each insurer s distribution of insured vehicles may be significantly different from the industry distribution used by IBC, individual insurers may calculate and use an off-balance factor in their premium calculation. Such off-balance factor ensures that a change of vehicle rate group differential factors remain overall premium neutral; i.e. additional (or loss of) premium is not generated by the implementation of CLEAR. CLEAR individual policy dislocation Since the objective of CLEAR is to more accurately reflect the risk of each vehicle make and model, the result of implementing CLEAR is a reallocation of the premium among all policyholders based on the vehicle make and model insured. This reallocation balances the increases with decreases, although there likely are more policyholders seeing decreases rather than increases. These increases and decreases result in dislocation among individual policyholders. This dislocation is expected and is normally mitigated by individual insurers by capping the extent of changes and implementing CLEAR over a longer time. IBC frequently assist insurers to meet their implementation objectives by recalculating and balancing customized rate group tables. CLEAR balancing to rate level neutrality In 2001, a change was made to the balancing of CLEAR rate groups to address concerns from insurers related to the timely annual implementation of new CLEAR tables. The underlying methodology of CLEAR remains unchanged, including depreciation of vehicle makes and models (Age of vehicle is still a significant variable in the regression models). Prior to 2001, CLEAR was balanced to remain rate level neutral at the mid-point of the coming year, by considering the estimated exposures Insurance Bureau of Canada - 7 - October 28, 2004
including new vehicle models in the new year. This would have an immediate adverse impact on insurers at the time of implementation of the new table and such implementation was frequently deferred to coincide with a new rate filing/ adjustment. It was decided to balance each annual CLEAR rate group table to remain rate level neutral at the time of implementation to remove the concern about implementation timing and address the issue of an immediate adverse impact on premium levels. The balancing process consists of reiteratively attempting to balance the new table with a balance factor until no change in the average rate group differential is observed compared to the old table (using a static distribution). As a result of this change, we anticipate that the average differential factor will increase annually and that, to the extent that it would be larger than the increase in loss cost, individual insurers will offset this effect by adjusting their base rates. This balancing procedure is not applicable to Accident Benefits since those rate groups are not age driven. MSRP depreciation In order to maintain similarity with the change in balancing of CLEAR rate group tables, the depreciation procedure was eliminated for MSRP. The depreciation of MSRP rate groups consisted of reducing the assigned rate group by 1, initially after three years and bi-annually thereafter. This approach will yield different results from CLEAR where depreciation is NOT eliminated as it still recognizes different depreciation patterns between different vehicles; for MSRP all depreciation effect is eliminated. Appendix A exhibits the annually estimated average rate group differentials using annual vehicle rate groups weighted by Canada-wide projected exposures. Insurance Bureau of Canada - 8 - October 28, 2004
APPENDIX A Average Rate Group Differentials for 2003 and 2004 CLEAR and MSRP Rate Group Tables 2004 2003 CLEAR MSRP CLEAR MSRP Coll. Comp. Coll. Comp. Coll. Comp. Coll. Comp. C-W 1.6122 1.3727 1.3449 1.3180 1.4719 1.2781 1.2516 1.2222 04/03 9.5% 7.4% 7.5% 7.9% Nfld 1.8202 1.4238 1.4668 1.4276 1.6562 1.3425 1.3657 1.3352 04/03 9.9% 6.1% 7.4% 6.9% The average differentials are estimated using a projected distribution of written exposures for 2003 and 2004, respectively. It is noted that rate group differentials utilized are those used by IBC in determining rate groups, which frequently are different from those used by individual insurers. Insurance Bureau of Canada - 9 - October 28, 2004