Frequency and Severity Bands Manual
The Department has updated the California Private Passenger Auto Frequency and Severity Bands Manual in order that insurers using the manual may have the most recent available data for use in their rate filings. We will be conducting rulemaking to make the necessary modification to section 2632.9(e) of title 10 of the California Code of Regulations in order to reflect the updating of the manual.
California Private Passenger Auto Frequency and Severity Bands Manual
California Department of Insurance
May 15, 2008
This manual summarizes the experience of California insurers in the private passenger auto insurance market, as specified in Section 2632.9(e) of Title 10 of the California Code of Regulations. The data in the manual are presented for each of the six coverages by three different levels of aggregation: frequency and severity bands, California Automobile Assigned Risk Program (CAARP) areas, and zip codes.
The primary purpose of this manual is to assist insurers in creating rating bands as allowed in Section 2632.5(d). The following language from Section 2632.9(d) describes the data an insurer may use to create the rating bands:
If an insurer elects to use the optional factors average relative frequency band and severity band as listed in Section 2632.5(d)(15) and (16), the bands described in Section 2632.9(e) may be used. . . . An insurer may also elect to use their own data for developing the frequency and severity bands. . . . If the insurer's own zip code data is not fully credible, it shall use the claims frequency and severity for the zip code that is published in the manual described in Section 2632.9(e) either:
(1) directly as it is published, or
(2) to credibility adjust their own data. In the case where the manual indicates that the rate published in the manual has been credibility adjusted, an insurer may:
(A) use the credibility adjusted rate in the manual as the complement of credibility, or
(B) combine the unadjusted data published in the manual with its own unadjusted data. If this combined data is still not credible, then an insurer may elect to use as the compliment of credibility either:
1. the rate published in the manual for the CAARP territory that the zip code is a member of, or
2. the rate based on their own data or data from the manual from another grouping of contiguous whole zip codes, selected by the insurer, that is fully credible and contains said zip code.
This manual is divided into three sections. Section 1 contains frequency and severity rates for the six coverages included in this manual: bodily injury (BI), property damage (PD), medical payments (MP), uninsured motorist (UM), collision (CL), and comprehensive (CM). BI, PD, MP, and UM are classified as liability coverages, while CL and CM are physical damage classifications. Also included is the unadjusted (from a credibility perspective) data. The twenty rating bands were created by grouping zip codes with similar risk levels together. Approximately the same number of total annual exposures are grouped into each band. The specific band assignment for each zip code can be determined by referring to Section 2, where detailed data for each zip code is listed, along with band assignments.
Section 2 is the largest part of the manual. In this section four to five years of market experience is summarized for each zip code using the data collected by the Department per CIC Section 11628(a).1 Five years of data, 1999 to 2003, are used for the four liability coverages . Four years of data, 2000 to 2003, are used for the two physical damage coverages. The zip codes listed in this section were taken from the 2007 zip code directory published by the U. S. Postal Service. Seven new zip codes from 2006 and 2007 had no exposure or claims data. These zip codes, 91008, 92058, 94505, 95811, 93636, 93723, and 93730 were not included in Section 2. As required by Section 2632.9(f)(1) of Title 10, insurers should assign to these zip codes the frequency and severity rates of the bands of the zip code they were created from, 91010, 92054, 94514, 95814, 93638, 93722, and 93720 respectively.
For the frequency rate, a zip code is fully credible if the number of exposures is large enough to pass a statistical test designed to assure a high probability that the frequency rate is at least as accurate as the average difference between all twenty rating bands. For zip code severity rates, a zip code is fully credible if the number of claims exceeds the number of claims consistent with the frequency rate credibility standard adjusted by the coefficient of variation. The coefficient of variation is estimated as the standard deviation of the zip code severities divided by the statewide severity rate. Based on these methods, a zip code's frequency rate is 100 percent credible if that zip code equals or exceeds this level of exposures (exposure years): 27,991 (BI), 8,812 (PD), 34,155 (MP), 122,867 (UM), 4,803 (CL), and 7,935 (CM). The credibility standard for a zip code's severity rate is the following number of claims: 415 (BI), 389 (PD), 415 (MP), 515 (UM), 621 (CL), and 2,480 (CM).
The frequency rate and severity rate shown for each zip code in Section 2 is either fully credible based on its own exposure, claim and loss data or it has been credibility adjusted using the frequency or severity rate from the CAARP territory to which it is assigned. The credibility adjustment is a weighted average of the zip code data and the CAARP data, where the weights are the credibility level and 1 minus the credibility level. The credibility level is calculated as the square root of the zip code's exposures (or claims) divided by the exposures (claims) required for full credibility.
In addition to showing the frequency and severity rate for each zip code, Section 2 also contains:
- the frequency and severity band the zip code is assigned to (the rate for the band is shown in Section 1),
- the raw, or not credibility adjusted data, is the data an insurer would use to combine with their own data, if their own data was not 100% credible and they elected to follow the directions in Section 2632.9(d)(2)(B) (as opposed to Section 2632.9(d)(2)(A)),
- the credibility of the unadjusted data as it relates to estimating the frequency rate and the severity rate, and
- the CAARP territory in which the zip code resides, which was used as the complement of credibility in the case where the raw data did not attain a 100% credibility level.
Section 3 summarizes the four or five years of data by CAARP territory or statistical code. The rates shown in this section were used to credibility adjust those zip codes with less than 100% credibility. Also, as specified in section 2632.9(d)(2)(B)(1) the rates shown here can be used by insurers as the complement of credibility in cases where their own data combined with the unadjusted data from Section 2 is still not 100% credible.
The 2008 California Private Passenger Auto Frequency and Severity Bands Manual - report describing the data and process used to create the manual.
1 The unadjusted claims and capped loss data is not integer data because some insurers report their experience data to the Department on a per claimant basis and other insurers report on a per accident basis. The Department's Statistical Analysis Division standardizes the number of claims and capped losses to a per accident basis using (fractional) adjustment factors for data reported on a per claimant basis. With respect to non-integer exposure data, exposure data is collected in exposure months, thus a policy in force for five months is 5/12ths of an exposure year.