Granular Pricing of Workers’ Compensation Risk in Excess Layers Identifying risk at a granular level and pricing it appropriately will put carriers on a path to sound underwriting ability.

Executive SummaryThe net written premiums for the workers’ compensation industry, and its share in the overall commercial lines of business, have steadily declined between 2007 and 2010. Net written pre-miums have plummeted 33% from $48 billion at year-end 2006 to $32 at year-end 2010. In 2011, the industry-wide loss ratio for workers’ compen-sation insurance in the U.S. increased to 118.1%.1 This is the highest level in more than a decade. The combined ratio for workers’ compensation has increased steadily from a value of 100% in 2006 to 119% in 2011 (see Figure 1, on next page). These numbers are influencing insurance carriers to look for creative pricing methodologies.

Pricing for workers’ compensation risk is heav-ily regulated by state regulatory authorities and offers less scope for pricing innovation in the primary layer. However, pricing in the excess layer is not as heavily regulated as pricing in the primary layer and hence provides an oppor-tunity for carriers to gain that competitive advantage by making the pricing more effective. The option to be creative in pricing and the increase in the insured’s buying trend for excess insurance makes excess lines pricing an ideal

place for insurers to concentrate. The onus is now on carriers to develop the competency for clas-sifying risk at a detailed level, which would enable them to price products more efficiently.

Granular pricing would help carriers identify specific components of risk that should be priced in house, components that should be ceded to reinsurers and components that should be avoided. It would also help carriers to identify niche areas of risk and develop appropriate under-writing and pricing techniques to cover them.

One of the integral components in determin-ing the premium for the workers’ compensation line of business (LOB) is excess loss factor (ELF), which is the ratio of expected losses in excess of a limit to the total expected losses. Classifications that use the same ELFs are grouped together to form hazard groups. The National Council on Compensation Insurance (NCCI) periodically publishes ELFs by hazard group for select limits.

NCCI implemented a seven-hazard-group system in 2007, replacing the previous four-hazard-group system. This move was the result of the review of the hazard group mapping by NCCI, which

cognizant 20-20 insights | june 2013

• Cognizant 20-20 Insights

cognizant 20-20 insights 2

Source: “2007 Hazard Group Mapping,” NCCI Research Publication

Figure 2

Distribution of Classes by Prior Hazard Groups (Left) and by New Hazard Groups (Right)

While implementing the seven hazard groups sys-tem, NCCI offered an alter-native collapsed new map-ping for insurers who did not want to switch to the new system immediately. In the collapsed mapping, hazard groups A and B were combined to form hazard group 1; hazard groups C and D were combined to form hazard group 2; hazard groups E and F were com-bined to form hazard group 3 and hazard group G was made hazard group 4.

The collapsed mapping provides a better platform for comparison between the prior mapping and the new mapping structure. If we compare the prior mapping with the collapsed new mapping,

concluded that there was a need for more granular classification of risk. The authors of this white paper believe that there is scope and need for further refinement in excess rating method-ology by determining and using ELFs at a more granular level.

Rate Making Using Hazard GroupsRate making in workers’ compensation is based on hazard groups, which are groupings of class codes. NCCI has defined approximately 900 work-ers’ compensation class codes and four or seven hazard groups, for which it provides rate making service. In the four (I-IV) hazard group mapping system, the bulk of the exposure is concentrated in two hazard groups. Hazard groups II and III con-tained 97% of the total premium (see Figure 2). As the figure indicates, after the implementation of the seven hazard group system there is a more homogenous distribution of premiums by hazard groups, which improved pricing accuracy.

The industry has benefited by moving to the seven hazard groups system, as this system has distributed premium values more evenly. But the opportunity to improve is still immense.

NCCI Hazard Group

Number of Classes

Percent of Total Premium

NCCI Hazard Group

Number of Classes

Percent of To-tal Premium

I 38 1% A 55 9%

II 428 46% B 241 17%

III 318 51% C 160 21%

IV 86 2% D 45 10%

E 224 19%

F 57 19%

G 88 5%

Source: “Best’s Aggregates & Averages - Property/Casualty,” A.M. Best, 2011.

Figure 1

85

95

105

115

125

0

20,000

40,000

60,000

2006 2007 2008 2009 2010 2011

Net Written Premium ($ Mn) Combined Ratio (%)

NWP and CR for Workers’ Compensation Line of Business

cognizant 20-20 insights 3

hazard group 1 (A and B) has substantial portion of total premium value compared with hazard group I of prior mapping (see Figure 3). Hazard groups 2 (C and D) and 3 (E and F) have become slightly smaller compared with their peers in the prior mapping. Hazard group 4 (HG G) has slightly more premium value than prior hazard group IV.

Granular Pricing Through Hazard Groups Decomposition The industry has benefited by moving to the seven hazard groups system, as this system has distributed premium values more evenly. But the opportunity to improve is still immense. With the seven group mapping, 69% of premium value is still distributed between four (HG C, D, E, F) of the seven hazard groups. Also, the ELFs as provided by NCCI are applicable for countrywide class codes. The ELFs should ideally be determined for class codes at the state level for applicable limits as one hazard group can be more hazardous in one state as compared with other states.

If we look at the movement of classes between the prior four mapping system to the collapsed new mapping system, the great majority (300 classes and 37% of premium) moved down one hazard group (see Figure 4). However it should be noted that the implementation of new mapping was revenue neutral for carriers as there was a general increase in the value of ELFs between the prior mapping and the new mapping.

Even with the seven haz-ard group model, there are numerous classes that have experienced a disproportion-ate number of catastrophic claims that might be inappropriately mapped to a lower severity hazard group. If the carrier is writing risk for these class codes that are placed

Granular pricing in the excess layer would ensure better estimation of overall expected losses, which would improve the combined ratio of the insurer in the long run.

HG I, 1%

HG II, 46%HG III, 51%

HG IV, 2%

HG I, 26%

HG 2, 31%

HG 3, 38%

HG 4, 5%

Source: “2007 Hazard Group Mapping,” NCCI Research Publication

Figure 3

Prior Mapping (Left) Versus Collapsed New Mapping (Right); Percent of Premium by Hazard Group

Source: “2007 Hazard Group Mapping,” NCCI Research Publication

Figure 4

Comparison of Prior Mapping with Collapsed New Mapping with Respect to Movement to Class Codes

63.5%

1.7%

34.5%

0.3%0

100

200

300

400

500

600

No Movement Up 1 HG Down 1 HG Down 3 HGs

552 15 300 3

Nu

mb

er o

f C

lass

Co

des

cognizant 20-20 insights 4

on the lower end of the hazard groups, it may have significantly more excess loss exposure than anticipated and hence the pricing might be

deficient. To further enable homogeneous distribution of premiums within the excess layers, ELFs should be referenced at the indi-vidual class code rather than the hazard group level. The ELFs at the class code level for every state can be deter-mined by the historical infor-mation on severity and fre-quency of losses that each class code had in the state.

An insurer that switched from hazard group to class code level pricing recently compared the underwriting results of deals between the two systems. The insurer found approximately $300 million in extra expected

losses in the excess layer that were not taken into consideration with the earlier hazard group pricing. Granular pricing in the excess layer would ensure better estimation of overall expected losses, which would improve the combined ratio of the insurer in the long run.

Commoditization of personal insurance has been assisted with sophisticated granular pricing tech-niques. Robust underwriting through risk classi-fication at a granular level in commercial lines is key to survival for insurers.

Impact to the Carriers

• Process change: The decomposition of hazard groups will introduce a larger set of factors for actuaries to maintain. Underwrit-ers/actuaries will need to adapt to the new process of reviewing and analyzing pricing worksheets at a deeper level for granular pricing. The amount of time required by actuaries will increase, with a corresponding increase in the accuracy of the price.

• Algorithm logic/calculation mechanism: Excess loss factors are used to calculate loss rating limit and expected losses. The move of assigning ELFs to each individual class code will impact the algorithm logic of calculating expected losses and loss rating limit.

For example, assume that a prospect has 20 class codes in its risk portfolio and that all 20 class codes were previously mapped to hazard group II. With the earlier pric-ing methodology, the respective algorithm logic for calculation of expected losses and loss rating limit used only one ELF assigned to hazard group II; with the granu-lar pricing discussed in this white paper, the carrier would use 20 ELFs, each assigned to an individual class code.

The actuarial algorithms to calculate loss rat-ing limit and expected losses might reside in stored procedures or within the application itself. A thorough impact analysis is needed to determine if the existing processing logic will be able to support the increased size of the reference database. Therefore, the carrier’s rating, pricing and policy issuance systems should be adapted to support the new data structure.

The sizeable increase in the support database will require performance reengineering to prevent the potential slowdown of the calcula-tion mechanism.

• Actuarial models: These models are usually embedded within the carrier’s policy adminis-tration system and receive input from the esti-mated losses/loss rating module and provide output for calculation of the final premium of the policy. Implementation of granular pricing methodology can impact the calculation per-formed by these models handling credit, profit and aggregate charges. The models will need to be modified to cope with the new reference database created to enable granular pricing.

In addition to the modifications to the model, an impact analysis is needed on all modules directly interacting with the models that input data into the model or use the output provided by the model to ensure that they are intact and are able to support the modifications made to the model itself. Also, the actuarial reporting systems will need to be adapted if they report data at the hazard group level.

• In-flight change: By changing the pricing methodology, underwriters will need to set the correct expectations with brokers/clients. For any policy that has already been quoted, pricing it again using the new mechanism

A thorough impact analysis is needed to determine if the existing processing

logic will be able to support the increased size

of the reference database. Therefore,

the carrier’s rating, pricing and policy issuance systems

should be adapted to support the new

data structure.

cognizant 20-20 insights 5

might change the premium values — which might be difficult to explain to the broker. Policies that have already been quoted or have reached a particular stage of underwriting should not be automatically impacted while the granular pricing functionality is imple-mented in the system. Underwriters should be given the choice, in the system, to be able to use the old mechanism or new mechanism to price the in-flight policies. This dual pric-ing can be achieved by implementing a simple solution of factor-date logic at the instance of calculation of premiums where the system would provide the underwriter with the option of using previous factors or new factors. The database would need to store both sets of factors for a certain period of time and, if the database does not have time-stamps, it will need to be implemented to support application of factor-date logic.

Any policy that has already been quoted sho-uld automatically follow the old mechanism if repriced in the system to avoid any adverse effect on the quote. Any new policy priced in the system should automatically follow the new mechanism.

Competitive Differentiation With intense competition and the intent to place risks more accurately, there is a need for insurance carriers to optimize workers’ compen-sation pricing. Carriers are trying to improve their underwriting practices to achieve a distinguished

position. Identifying risk at a granular level and pricing it appropriately will put carriers on a path to sound underwriting ability.

Several advantages of this approach include:

• More accurate pricing: With better insight into risk classification, carriers would be able to determine appropriate expected losses in excess layers for every state and hence would be able to price in risk more accurately than before.

• Ability to assume higher levels of risk: With the higher accuracy of risk classification, there will be some hazard groups, for specific states, that will move down to a lower hazard group. This will make the hazard less hazardous, so to speak, and make it assumable for the insurance carrier.

• Ability to avoid insufficiently priced risk: Insurance carriers will be able to identify poor risks easily and thereby avoid them, or engage in appropriate risk transfer mechanisms. Classes that have exhibited the propensity to produce higher or more frequent losses can thus be avoided.

As with the advancements in any field, we believe that this functionality will become a hygiene factor in the years to come. Insurance carriers proactively adopting this approach will obtain a competitive edge.

References

• “Revised Hazard Group Assignments,” National Council on Compensation Insurance Actuarial Committee Agenda, Item ACT-92-57, April 20, 1993.

• John P. Robertson, “2007 Hazard Group Mapping,” National Council on Compensation Insurance Research Paper.

Footnote1 “Best’s Aggregates & Averages - Property/Casualty,” A.M. Best, 2011.

About Cognizant

Cognizant (NASDAQ: CTSH) is a leading provider of information technology, consulting, and business process outsourcing services, dedicated to helping the world’s leading companies build stronger businesses. Headquartered in Teaneck, New Jersey (U.S.), Cognizant combines a passion for client satisfaction, technology innovation, deep industry and business process expertise, and a global, collaborative workforce that embodies the future of work. With over 50 delivery centers worldwide and approximately 162,700 employees as of March 31, 2013, Cognizant is a member of the NASDAQ-100, the S&P 500, the Forbes Global 2000, and the Fortune 500 and is ranked among the top performing and fastest growing companies in the world.

Visit us online at www.cognizant.com for more information.

World Headquarters

500 Frank W. Burr Blvd.Teaneck, NJ 07666 USAPhone: +1 201 801 0233Fax: +1 201 801 0243Toll Free: +1 888 937 3277Email: inquiry@cognizant.com

European Headquarters

1 Kingdom StreetPaddington CentralLondon W2 6BDPhone: +44 (0) 207 297 7600Fax: +44 (0) 207 121 0102Email: infouk@cognizant.com

India Operations Headquarters

#5/535, Old Mahabalipuram RoadOkkiyam Pettai, ThoraipakkamChennai, 600 096 IndiaPhone: +91 (0) 44 4209 6000Fax: +91 (0) 44 4209 6060Email: inquiryindia@cognizant.com

© Copyright 2013, Cognizant. All rights reserved. No part of this document may be reproduced, stored in a retrieval system, transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the express written permission from Cognizant. The information contained herein is subject to change without notice. All other trademarks mentioned herein are the property of their respective owners.

About the AuthorsRamanujam Venkatesan is a Senior Consulting Manager within Cognizant Business Consulting’s Insur-ance Practice. He specializes in the property and casualty space and has worked with multiple carriers across North America, Europe and Asia-Pacific. Ramanujam has a post graduate degree in management from the Indian Institute of Management (IIM) Indore and has a bachelor’s degree in engineering from the University of Madras. He also holds certifications from the American Institute for Chartered Property and Casualty Underwriters (AICPCU). He can be reached at Ramanujam.Venkatesan@cognizant.com.

Shishir Narula is a Senior Consultant within Cognizant Business Consulting’s Insurance Practice. He specializes in commercial insurance underwriting and policy administration for workers’ compensation, auto, general liability and excess lines of business. He holds an M.B.A. from Institute of Management Technology (IMT) Ghaziabad and a bachelor’s degree in engineering from Indraprashtha University, Delhi. He also holds certifications from the American Institute for Chartered Property and Casualty Underwriters (AICPCU). He can be reached at Shishir.Narula@cognizant.com.