Hierarchical Insurance Claims Modeling Travelers PASG (Predictive Analytics Study Group) Seminar
by user
Comments
Transcript
Hierarchical Insurance Claims Modeling Travelers PASG (Predictive Analytics Study Group) Seminar
Hierarchical Insurance Claims Modeling Hierarchical Insurance Claims Modeling Emiliano A. Valdez, PhD, FSA joint work with Jed Frees, U of Wisconsin - Madison Introduction Travelers PASG (Predictive Analytics Study Group) Seminar Tuesday, 12 April 2016 Data The model Model features Model estimation Covariates Models of each component Random effects NB model Multimonial claim type Severity and copula Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion Emiliano A. Valdez, PhD, FSA Department of Mathematics University of Connecticut Storrs, Connecticut page 1 A collection of work Hierarchical Insurance Claims Modeling Emiliano A. Valdez, PhD, FSA Frees and Valdez (2008), Hierarchical Insurance Claims Modeling, Journal of the American Statistical Association, Vol. 103, No. 484, pp. 1457-1469. Introduction Data Frees, Shi and Valdez (2009), Actuarial Applications of a Hierarchical Insurance Claims Model, ASTIN Bulletin, Vol. 39, No. 1, pp. 165-197. The model Model features Model estimation Covariates Models of each component Young, Valdez and Kohn (2009), Multivariate Probit Models for Conditional Claim Types, Insurance: Mathematics and Economics, Vol. 44, No. 2, pp. 214-228. Antonio, Frees and Valdez (2010), A Multilevel Analysis of Intercompany Claim Counts, ASTIN Bulletin, Vol. 40, No. 1, pp. 151-177. Random effects NB model Multimonial claim type Severity and copula Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion page 2 Location of Singapore Hierarchical Insurance Claims Modeling Emiliano A. Valdez, PhD, FSA Introduction Data The model Model features Model estimation Covariates Models of each component Random effects NB model Multimonial claim type Severity and copula Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion page 3 Car ownership in Singapore Hierarchical Insurance Claims Modeling Emiliano A. Valdez, PhD, FSA Owning and driving a car in Singapore is quite expensive. Government has put up measures to manage car ownership: Introduction Data certificate of entitlement (COE) vehicle quota system (VQS) registration fees, road taxes (annual/semi-annual) electronic road pricing (per use, driving in certain areas) The model Model features Model estimation Covariates Models of each component Random effects NB model Multimonial claim type Severity and copula Every motor vehicle must have a valid insurance policy: minimum required is coverage for personal injury to other parties. three major types available in the marketplace: third party, third part + fire & theft, and comprehensive. Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion page 4 Basic data set-up “Policyholder” i is followed over time t = 1, . . . , 9 years Hierarchical Insurance Claims Modeling Emiliano A. Valdez, PhD, FSA Unit of analysis “it” – a registered vehicle insured i over time t (year) Have available: exposure eit and covariates (explanatory variables) xit covariates often include age, gender, vehicle type, driving history and so forth Goal: understand how time t and covariates impact claims yit . Statistical methods viewpoint basic regression set-up - almost every analyst is familiar with: part of the basic actuarial education curriculum Introduction Data The model Model features Model estimation Covariates Models of each component Random effects NB model Multimonial claim type Severity and copula Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion incorporating cross-sectional and time patterns is the subject of longitudinal data analysis - a widely available statistical methodology page 5 Hierarchical Insurance Claims Modeling More complex data set-up Emiliano A. Valdez, PhD, FSA Some variations that might be encountered when examining insurance company records For each “it”, could have multiple claims, j = 0, 1, . . . , 5 For each claim yitj , possible to have one or a combination of three (3) types of losses: 1 2 3 losses for injury to a party other than the insured yitj,1 “injury”; losses for damages to the insured, including injury, property damage, fire and theft yitj,2 - “own damage”; and losses for property damage to a party other than the insured yitj,3 - “third party property”. Distribution for each claim is typically medium to long-tail The full multivariate claim may not be observed: Introduction Data The model Model features Model estimation Covariates Models of each component Random effects NB model Multimonial claim type Severity and copula Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion Distribution of Claims, by Claim Type Observed Value of M 1 2 3 4 5 6 Claim by Combination (y1 ) (y2 ) (y3 ) (y1 , y2 ) (y1 , y3 ) (y2 , y3 ) Number 102 17,216 2,899 68 18 3,176 Percentage 0.4 73.2 12.3 0.3 0.1 13.5 7 (y1 , y2 , y3 ) 43 0.2 page 6 The hierarchical insurance claims model Hierarchical Insurance Claims Modeling Emiliano A. Valdez, PhD, FSA Traditional to predict or estimate insurance claims distributions: Introduction Cost of Claims = Frequency × Severity Data The model Model features Joint density of the aggregate loss can be decomposed as: Model estimation Covariates f (N, M, y) = f (N) × f (M|N) × f (y|N, M) Models of each component Random effects NB model Multimonial claim type joint = frequency × conditional claim-type × conditional severity. Severity and copula Parameter estimates The fitted frequency model The fitted conditional claim type model This natural decomposition allows us to investigate and model each component separately. The fitted conditional severity model Conclusion page 7 Model features Hierarchical Insurance Claims Modeling Emiliano A. Valdez, PhD, FSA Allows for risk rating factors to be used as explanatory variables that predict both the frequency and the multivariate severity components. Introduction Data The model Helps capture the long-tail nature of the claims distribution through the GB2 distribution model. Model features Model estimation Covariates Models of each component Provides for a “two-part” distribution of losses - when a claim occurs, not necessary that all possible types of losses are realized. Allows to capture possible dependencies of claims among the various types through a t-copula specification. Random effects NB model Multimonial claim type Severity and copula Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion page 8 Literature on claims frequency/severity There is large literature on modeling claims frequency and severity Hierarchical Insurance Claims Modeling Emiliano A. Valdez, PhD, FSA Klugman, Panjer and Willmot (2004) - basics without covariates Kahane and Levy (JRI, 1975) - first to model joint frequency/severity with covariates. Coutts (1984) postulates that the frequency component is more important to get right. Many recent papers on frequency, e.g., Boucher and Denuit (2006) Introduction Data The model Model features Model estimation Covariates Models of each component Random effects NB model Applications to motor insurance: Multimonial claim type Severity and copula Brockman and Wright (1992) - good early overview. Renshaw (1994) - uses GLM for both frequency and severity with policyholder data. Pinquet (1997, 1998) - uses the longitudinal nature of the data, examining policyholders over time. Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion considered 2 lines of business: claims at fault and not at fault; allowed correlation using a bivariate Poisson for frequency; severity models used were lognormal and gamma. Most other papers use grouped data, unlike our work. page 9 Hierarchical Insurance Claims Modeling Observable data Model is calibrated with detailed, micro-level automobile insurance records over eight years [1993 to 2000] of a randomly selected Singapore insurer. Emiliano A. Valdez, PhD, FSA Year 2001 data use for out-of-sample prediction Information was extracted from the policy and claims files. Unit of analysis - a registered vehicle insured i over time t (year). The observable data consist of number of claims within a year: Nit , for t = 1, . . . , Ti , i = 1, . . . , n Introduction Data The model Model features Model estimation Covariates Models of each component Random effects NB model Multimonial claim type Severity and copula type of claim: Mitj for claim j = 1, . . . , Nit the loss amount: yitjk for type k = 1, 2, 3. exposure: eit Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model vehicle characteristics: described by the vector xit Conclusion The data available therefore consist of {eit , xit , Nit , Mitj , yitjk } . page 10 Risk factor rating system Insurers adopt “risk factor rating system” in establishing premiums for motor insurance. Hierarchical Insurance Claims Modeling Emiliano A. Valdez, PhD, FSA Some risk factors considered: vehicle characteristics: make/brand/model, engine capacity, year of make (or age of vehicle), price/value driver characteristics: age, sex, occupation, driving experience, claim history other characteristics: what to be used for (private, corporate, commercial, hire), type of coverage Introduction Data The model Model features Model estimation Covariates Models of each component Random effects NB model Multimonial claim type Severity and copula The “no claims discount” (NCD) system: Parameter estimates The fitted frequency model rewards for safe driving discount upon renewal of policy ranging from 0 to 50%, depending on the number of years of zero claims. The fitted conditional claim type model The fitted conditional severity model Conclusion These risk factors/characteristics help explain the heterogeneity among the individual policyholders. page 11 Covariates Hierarchical Insurance Claims Modeling Emiliano A. Valdez, PhD, FSA Year: the calendar year - 1993-2000; treated as continuous variable. Vehicle Type: automotive (A) or others (O). Introduction Vehicle Age: in years, grouped into 6 categories - The model Data Model features 0, 1-2, 3-5, 6-10, 11-15, <=16. Model estimation Covariates Vehicle Capacity: in cubic capacity. Models of each component Random effects NB model Multimonial claim type Gender: male (M) or female (F). Age: in years, grouped into 7 categories ages ≥21, 22-25, 26-35, 36-45, 46-55, 56-65, ≤66. The NCD applicable for the calendar year - 0%, 10%, 20%, 30%, 40%, and 50%. Severity and copula Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion page 12 Random effects negative binomial count model Hierarchical Insurance Claims Modeling Emiliano A. Valdez, PhD, FSA Let λit = eit exp x0λ,it βλ be the conditional mean parameter for the {it} observational unit, where xλ,it is a subset of xit representing the variables needed for frequency modeling. Introduction Data The model Negative binomial distribution model with parameters p and r : ! k +r −1 r Pr(N = k |r , p) = p (1 − p)k . r −1 1 Here, σ = is the dispersion parameter and r p = pit is related to the mean through 1 − pit = λit σ = eit exp(x0λ,it βλ )σ. pit Model features Model estimation Covariates Models of each component Random effects NB model Multimonial claim type Severity and copula Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion page 13 Multinomial claim type Hierarchical Insurance Claims Modeling Emiliano A. Valdez, PhD, FSA Certain characteristics help describe the claims type. To explain this feature, we use the multinomial logit of the form exp(Vm ) , Pr(M = m) = P7 s=1 exp(Vs ) where Vm = Vit,m = x0M,it βM,m . For our purposes, the covariates in xM,it do not depend on the accident number j nor on the claim type m, but we do allow the parameters to depend on type m. Introduction Data The model Model features Model estimation Covariates Models of each component Random effects NB model Multimonial claim type Severity and copula Parameter estimates The fitted frequency model Such has been proposed in Terza and Wilson (1990). An alternative model to claim type, multivariate probit, was considered in: The fitted conditional claim type model The fitted conditional severity model Conclusion Young, Valdez and Kohn (2009) page 14 Hierarchical Insurance Claims Modeling Severity We are particularly interested in accommodating the long-tail nature of claims. Emiliano A. Valdez, PhD, FSA We use the generalized beta of the second kind (GB2) for each claim type with density f (y ) = exp (α1 z) α +α , y |σ|B(α1 , α2 ) [1 + exp(z)] 1 2 where z = (ln y − µ)/σ. Introduction Data The model Model features Model estimation Covariates Models of each component µ is a location, σ is a scale and α1 and α2 are shape parameters. Random effects NB model Multimonial claim type Severity and copula Parameter estimates With four parameters, distribution has great flexibility for fitting heavy tailed data. Introduced by McDonald (1984), used in insurance loss modeling by Cummins et al. (1990). The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion Many distributions useful for fitting long-tailed distributions can be written as special or limiting cases of the GB2 distribution; see, for example, McDonald and Xu (1995). page 15 GB2 Distribution Hierarchical Insurance Claims Modeling Emiliano A. Valdez, PhD, FSA Introduction Data The model Model features Model estimation Covariates Models of each component Random effects NB model Multimonial claim type Severity and copula Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion Source: Klugman, Panjer and Willmot (2004), p. 72 page 16 Heavy-tailed regression models Hierarchical Insurance Claims Modeling Emiliano A. Valdez, PhD, FSA Loss Modeling - Actuaries have a wealth of knowledge on fitting claims distributions. (Klugman, Panjer, Willmot, 2004) (Wiley) Introduction Data are often “heavy-tailed” (long-tailed, fat-tailed) Data Extreme values are likely to occur The model Extreme values are the most interesting - do not wish to downplay their importance via transformation Model estimation Model features Covariates Models of each component Random effects NB model Studies of financial asset returns is another good example Rachev et al. (2005) “Fat-Tailed and Skewed Asset Return Distributions” (Wiley) Healthcare expenditures - Typically skewed and fat-tailed due to a few yet high-cost patients (Manning et al., 2005, J. of Health Economics) Multimonial claim type Severity and copula Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion page 17 Hierarchical Insurance Claims Modeling GB2 regression Emiliano A. Valdez, PhD, FSA We allow scale and shape parameters to vary by type and thus consider α1k , α2k and σk for k = 1, 2, 3. Despite its prominence, there are relatively few applications that use the GB2 in a regression context: Introduction Data McDonald and Butler (1990) used the GB2 with regression covariates to examine the duration of welfare spells. Beirlant et al. (1998) demonstrated the usefulness of the Burr XII distribution, a special case of the GB2 with α1 = 1, in regression applications. The model Model features Model estimation Covariates Models of each component Random effects NB model Multimonial claim type Severity and copula Sun et al. (2008) used the GB2 in a longitudinal data context to forecast nursing home utilization. We parameterize the location parameter as µik = x0ik βk : Thus, βk ,j = ∂ ln E (Y | x) /∂xj Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion Interpret the regression coefficients as proportional changes. page 18 Claim losses by type of claim Hierarchical Insurance Claims Modeling Emiliano A. Valdez, PhD, FSA Table 2.3. Summary Statistics of Claim Losses, by Type of Claim Statistic Third Party Own Damage (y 2 ) Third Party Injury (y 1 ) non-censored all Property (y 3 ) Number 231 17,974 20,503 6,136 Mean 12,781.89 2,865.39 2,511.95 2,917.79 Standard Deviation 39,649.14 4,536.18 4,350.46 3,262.06 Median 1,700 1,637.40 1,303.20 1,972.08 Minimum 10 2 0 3 Maximum 336,596 367,183 367,183 56,156.51 Introduction Data The model Model features Model estimation Covariates Models of each component Random effects NB model Multimonial claim type Severity and copula Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion page 19 Dependencies among claim types Hierarchical Insurance Claims Modeling Emiliano A. Valdez, PhD, FSA We use a parametric copula (in particular, the t copula). Suppressing the {i} subscript, we can express the joint distribution of claims (y1 , y2 , y3 ) as F(y1 , y2 , y3 ) = H (F1 (y1 ), F2 (y2 ), F3 (y3 )) . Introduction Data The model Model features Model estimation Here, the marginal distribution of yk is given by Fk (·) and H(·) is the copula. Modeling the joint distribution of the simultaneous occurrence of the claim types, when an accident occurs, provides the unique feature of our work. Some references are: Frees and Valdez (1998), Nelsen (1999). Covariates Models of each component Random effects NB model Multimonial claim type Severity and copula Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion page 20 Hierarchical Insurance Claims Modeling The calendar year effect Emiliano A. Valdez, PhD, FSA Introduction Data Count 0 1 2 3 4 5 Number by Year 1993 91.5 7.9 0.5 0.1 4,976 Table 3.2. Number and Percentages of Claims, by Count and Year Percentage by Year 1994 1995 1996 1997 1998 1999 2000 2001 Number 89.5 89.8 92.6 92.8 90.8 88.0 89.2 87.8 178,080 9.6 9.2 7.0 6.7 8.4 10.6 9.8 11.0 19,224 0.9 0.9 0.4 0.5 0.7 1.3 0.9 1.1 1,859 0.1 0.1 0.0 0.0 0.1 0.1 0.1 0.1 177 0.0 0.0 0.0 0.0 11 0.0 0.0 5,969 5,320 8,562 19,344 19,749 28,473 44,821 62,138 199,352 Percent of Total 89.3 9.6 0.9 0.1 0.0 0.0 100.0 The model Model features Model estimation Covariates Models of each component Random effects NB model Multimonial claim type Severity and copula Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion page 21 The effect of vehicle type and age Hierarchical Insurance Claims Modeling Emiliano A. Valdez, PhD, FSA Table 3.3. Number and Percentages of Claims, by Vehicle Type and Age Percentage by Count Count Count Count Count Count Count Percent =0 =1 =2 =3 =4 =5 Number of Total Vehicle Type Other 88.6 10.1 1.1 0.1 0.0 0.0 43,891 22.0 89.5 9.5 0.9 0.1 0.0 155,461 78.0 Automobile Vehicle Age (in years) 0 91.4 7.9 0.6 0.0 0.0 58,301 29.2 86.3 12.2 1.3 0.2 0.0 44,373 22.3 1 2 88.8 10.1 1.1 0.1 20,498 10.3 3 to 5 89.2 9.7 1.0 0.1 0.0 41,117 20.6 6 to 10 90.1 8.9 0.9 0.1 0.0 33,121 16.6 11 to 15 91.4 7.6 0.7 0.2 1,743 0.9 16 and older 89.9 8.5 1.5 199 0.1 Number 178,080 19,224 1,859 177 11 1 199,352 100.0 by Count Introduction Data The model Model features Model estimation Covariates Models of each component Random effects NB model Multimonial claim type Severity and copula Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion page 22 Hierarchical Insurance Claims Modeling The effect of gender, age, NCD Emiliano A. Valdez, PhD, FSA Table 3.4. Number and Percentages of Claims, by Gender, Age and NCD for Automobile Policies Percentage by Count Count Count Count Count Count Count Percent =0 =1 =2 =3 =4 =5 Number of Total Gender Female 89.7 Male 89.5 Person Age (in years) 21 and younger 86.9 22-25 85.5 26-35 88.0 36-45 90.1 46-55 90.4 56-65 90.7 66 and over 92.8 No Claims Discount (NCD) 0 87.7 10 87.8 20 89.1 30 89.1 40 89.8 50 91.0 Number 139,183 by Count 9.3 9.5 0.9 0.9 0.1 0.1 12.4 12.9 10.8 9.1 8.8 8.4 7.0 0.7 1.4 1.1 0.8 0.8 0.9 0.2 0.2 0.1 0.1 0.1 0.1 0.1 11.1 10.8 9.8 10.0 9.3 8.3 14,774 1.1 1.2 1.0 0.9 0.9 0.7 1,377 0.1 0.1 0.1 0.1 0.1 0.1 123 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 3 0.0 1 34,190 121,271 22.0 78.0 153 3,202 44,134 63,135 34,373 9,207 1,257 0.1 2.1 28.4 40.6 22.1 5.9 0.8 37,139 13,185 14,204 12,558 10,540 67,835 155,461 23.9 8.5 9.1 8.1 6.8 43.6 100.0 Introduction Data The model Model features Model estimation Covariates Models of each component Random effects NB model Multimonial claim type Severity and copula Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion page 23 Comparing the alternative conditional claim type models Hierarchical Insurance Claims Modeling Emiliano A. Valdez, PhD, FSA Table 3.7. Comparison of Fit of Alternative Claim Type Models Model Variables Number of -2 Log Parameters Likelihood AIC BIC Intercept Only 6 25,465.3 25,477.3 25,538.5 Automobile (A) 12 24,895.8 24,919.8 25,042.2 A and Gender 24 24,866.3 24,914.3 25,159.2 Year 12 25,315.6 25,339.6 25,462.0 Year1996 12 25,259.9 25,283.9 25,406.3 A and Year1996 18 24,730.6 24,766.6 24,950.3 VehAge2 (Old vs New) 12 25,396.5 25,420.5 25,542.9 VehAge2 and A 18 24,764.5 24,800.5 24,984.2 A, VehAge2 and Year1996 24 24,646.6 24,694.6 24,939.5 Introduction Data The model Model features Model estimation Covariates Models of each component Random effects NB model Multimonial claim type Severity and copula Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion page 24 Hierarchical Insurance Claims Modeling The fitted copula model Table 3.8. Fitted Copula Model Type of Copula Independence Normal copula Parameter Third Party Injury σ1 1.316 (0.124) 1.320 (0.138) α11 2.188 (1.482) 2.227 (1.671) 500.069 (455.832) 500.068 (408.440) α12 βC,1,1 (intercept) 18.430 (2.139) 18.509 (4.684) Own Damage σ2 1.305 (0.031) 1.301 (0.022) 5.658 (1.123) 5.507 (0.783) α21 α22 163.605 (42.021) 163.699 (22.404) 10.037 (1.009) 9.976 (0.576) βC,2,1 (intercept) βC,2,2 (VehAge2) 0.090 (0.025) 0.091 (0.025) βC,2,3 (Year1996) 0.269 (0.035) 0.274 (0.035) βC,2,4 (Age2) 0.107 (0.032) 0.125 (0.032) βC,2,5 (Age3) 0.225 (0.064) 0.247 (0.064) Third Party Property σ3 0.846 (0.032) 0.853 (0.031) 0.597 (0.111) 0.544 (0.101) α31 α32 1.381 (0.372) 1.534 (0.402) βC,3,1 (intercept) 1.332 (0.136) 1.333 (0.140) βC,3,2 (VehAge2) -0.098 (0.043) -0.091 (0.042) βC,3,3 (Year1) 0.045 (0.011) 0.038 (0.011) Copula ρ12 0.018 (0.115) ρ13 -0.066 (0.112) ρ23 0.259 (0.024) r Model Fit Statistics log-likelihood -31,006.505 -30,955.351 number of parms 18 21 AIC 62,049.010 61,952.702 Note: Standard errors are in parenthesis. t-copula Emiliano A. Valdez, PhD, FSA 1.320 (0.120) 2.239 (1.447) 500.054 (396.655) 18.543 (4.713) 1.302 (0.029) 5.532 (0.992) 170.382 (59.648) 10.106 (1.315) 0.091 (0.025) 0.274 (0.035) 0.125 (0.032) 0.247 (0.064) 0.853 (0.031) 0.544 (0.101) 1.534 (0.401) 1.333 (0.139) -0.091 (0.042) 0.038 (0.011) 0.018 (0.115) -0.066 (0.111) 0.259 (0.024) 193.055 (140.648) Introduction Data The model Model features Model estimation Covariates Models of each component Random effects NB model Multimonial claim type Severity and copula Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion -30,955.281 22 61,954.562 page 25 Concluding remarks Model features Hierarchical Insurance Claims Modeling Emiliano A. Valdez, PhD, FSA Allows for covariates for the frequency, type and severity components Captures the long-tail nature of severity through the GB2. Provides for a “two-part” distribution of losses - when a claim occurs, not necessary that all possible types of losses are realized. Allows for possible dependencies among claims through a copula Introduction Data The model Model features Model estimation Covariates Models of each component Random effects NB model Multimonial claim type Allows for heterogeneity from the longitudinal nature of policyholders (not claims) Severity and copula Parameter estimates The fitted frequency model Other applications Could look at financial information from companies The fitted conditional claim type model The fitted conditional severity model Conclusion Could examine health care expenditure Compare companies’ performance using multilevel, intercompany experience page 26 Hierarchical Insurance Claims Modeling Emiliano A. Valdez, PhD, FSA Introduction Data The model Model features - Thank you - Model estimation Covariates Models of each component Random effects NB model Multimonial claim type Severity and copula Parameter estimates The fitted frequency model The fitted conditional claim type model The fitted conditional severity model Conclusion page 27