A number of papers have been developed in response to calls by the Casualty Actuarial Society's Committee
on Ratemaking and the Committee on Management Data and Information. Selected papers will be presented
and discussed by their authors during sessions at the 2003 Ratemaking Seminar.
The completed papers are in the CAS 2003 Winter
Forum, which is available online.
Ratemaking Research Call Papers
RCP-1 Ratemaking Call Papers Presentation 1
Moderator:
Neal M. Leibowitz, Commercial Lines Actuarial-Director, The Hartford
Capital Consumption: An Alternative Methodology for Pricing
Reinsurance
By Donald F. Mango, American Re-Insurance
Company
This paper introduces a capital consumption methodology for the price evaluation of reinsurance in a stochastic environment. It differs from the common practice of risk-based capital allocation and release by: (i) evaluating the actual contract cash flows at the scenario level; (ii) eliminating the need for contract-level supporting capital allocation and release; (iii) evaluating each scenario's operating deficits as contingent capital calls on the company capital pool; and (iv) reflecting the expected cost of contingent capital calls as an expense load. This method eliminates the need for capital allocation and release; creates scenarios that more closely model actual contract capital usage; allows more flexibility in stochastic modeling; and makes risk-return preferences an explicit part of the pricing decision.
Statistical Learning Algorithms Applied to Automobile Insurance
Ratemaking
By Charles Dugas, Chief Actuary, Insurance Corporation of British Columbia
Yoshua Bengio, Apstat Technologies Inc.
Nicolas Chapados, Apstat Technologies Inc.
Pascal Vincent, Apstat Technologies Inc.
Germain Denoncourt, Actuary, L'Alpha Compagnie d'Assurances Inc.
Christian Fournier, Insurance Corporation of British Columbia
We recently conducted a research project for a large North American automobile insurer. This study was the most exhaustive ever undertaken by this particular insurer and lasted over an entire year. We analyzed the discriminating power of each variable used for ratemaking. We analyzed the performance of several models within five broad categories: linear regressions, generalized linear models, decision trees, neural networks and support vector machines. In this paper, we present the main results of this study. We qualitatively compare models and show how neural networks can represent high-order nonlinear dependencies with a small number of parameters, each of which is estimated on a large proportion of the data, thus yielding low variance. We thoroughly explain the purpose of the nonlinear sigmoidal transforms which are at the very heart of neural networks' performances. The main numerical result is a statistically significant reduction in the out-of-sample mean-squared error using the neural network model and our ability to substantially reduce the median premium by charging more to the highest risks. This in turn can translate into substantial savings and financial benefits for an insurer. We hope this paper goes a long way towards convincing actuaries to include neural networks within their set of modeling tools for ratemaking.
RCP-2 Ratemaking Call Papers Presentation 2
Moderator:
Charles H. Boucek, Ernst & Young, LLP
Dynamic Pricing Analysis
By Charles H. Boucek, Ernst & Young, LLP
Thomas P. Conway, Ernst & Young, LLP
This paper presents a methodology that represents a significant enhancement to current pricing practices. The goal of this methodology is to estimate the impact that a rate change will have on a company's policyholder retention and the resulting profitability of this transformed book of business. The paper will present the basics of this methodology as well as where future work will need to be done to bring this methodology into mainstream pricing. The work that the authors have done in this area has focused on Private Passenger Auto Insurance but these techniques could be applied to other lines of business.
Estimating the Cost of Commercial Airlines Catastrophes-
A
Stochastic Simulation Approach
By Romel G. Salam, Transatlantic Reinsurance Company
Actuaries are increasingly finding more applications for stochastic simulation in pricing, reserving, DFA, and other insurance and financial engineering problems. For instance, stochastic simulation has gained acceptance as a pricing tool for property catastrophe coverage in the insurance, reinsurance, broker and investment communities. This has required primary companies to compile and provide information at a more detailed level than they did only a few years ago. Various commercial simulation products have emerged to help companies assess and price their property catastrophe exposures. Although there are many parallels between the catastrophe exposures of property and commercial aviation risks, the use of simulation is not widespread in the assessment of commercial aviation catastrophic exposures. In this paper, we present the framework for a simulation model for commercial aviation catastrophes and we discuss various aspects of designing such a model including the level and type of information needed.
RCP-3 Ratemaking Call Papers Presentation 3
Moderator:
Kim Ward, Chief Actuary, American Association of Insurance
Services
A Unifying Approach to Pricing Insurance and Financial Risk
By Andreas Kull, Converium Ltd.
The actuarial and the financial approach to the pricing of risk remain different despite the increasingly direct interconnection of financial and insurance markets. The difference can be summarized as pricing based on classical risk theory (insurance) vs. non-arbitrage pricing (finance). However, comparable pricing principles are of importance when it comes to transferring insurance risk to financial markets and vice versa as it is done e.g., by alternative risk-transfer instruments or derivative products. Incompatibilities blur business opportunities or may open up the possibility to arbitrage. For these situations, the paper aims to bridge the gap between insurance and finance by extending the non-arbitrage pricing principle to insurance. The main obstacle that has to be tackled is related to the incompleteness of the insurance market. It implies that equivalent martingale probabilities are not uniquely defined. By the information theoretical maximum entropy principle, a sensible way to choose a particular equivalent martingale measure is found. This approach parallels the successful application of the maximum entropy principle in finance.
Credit & Surety Pricing and the Effects of Financial Market
Convergence
By Althula Alwis, American Re-Insurance Company
Christopher M. Steinbach, Swiss Reinsurance America Corporation
This paper describes how the convergence of the insurance and financial markets is affecting Credit & Surety insurance. The paper explains why prior experience has become an unreliable measure of exposure and how this paradigm shift affects the pricing of Credit & Surety products. The authors propose a new exposure-based method for analyzing Credit & Surety that combines the best practices of insurance and financial market pricing theory. Discussions about its implementation as well as sample calculations for both primary and reinsurance pricing are included. This paper also discusses the new breed of Commercial Surety bonds that have been recently developed to compete with traditional financial products. Finally, the paper addresses the need for better and more sophisticated risk management techniques for the industry.
Data Management, Quality, and Technology Call Papers
DCP-1 Emerging Technology Call Papers Presentation 1
Moderator:
Suzanne E. Black , Mercer Risk, Finance, and Insurance
Does Credit Score Really Explain Insurance Losses? Multivariate
Analysis From a Data Mining Point of View
By James C. Guszcza, Deloitte & Touche LLP
Cheng-Sheng "Peter" Wu, Deloitte & Touche LLP
One of the most significant development in insurance ratemaking and underwriting in the past decades has been the use of credit history in personal lines of business. Since its introduction in late 80's and early 90's, the predictive power of credit score and its relevance to insurance pricing and underwriting have been the subject of debate. The fact that personal credit is widely used by insurers strongly suggests its power to explain insurance losses and profitability. However, critics have questioned whether the apparently strong relationship between personal credit and insurance losses and profitability really exists. Surprisingly, even though this is a hot topic in the insurance industry and in regulatory circles, actuaries have not been actively participating in the debate. To date, there have been few actuarial studies published on the relationship of personal credit to insurance losses and profitability. A possible reason for the lack of published data is that many insurers view credit scores as a confidential and cutting-edge approach to help them win in the market place. Therefore, they might be reluctant to share their results with the public. In this paper, we will first review the two published studies and comment on their results. We will then share our own experience on this topic.
DCP-2 Emerging Technology Call Papers Presentation 2
Moderator:
Sara Schlenker, Actuary, Allstate Insurance Company
Where is My Market? How to Use Data to Find and Validate New
Commercial Lines Market Niches
By Lisa Sayegh, ISO
Entering a new insurance market is not a decision to be taken lightly. Market segment analysis is a lengthy process, and finding the right data is just the beginning. Being able to make meaningful comparisons of data from various sources and across insurance lines is the key to identifying profitable markets. Fortunately, there are data sources and tools available that can help with the analysis, as well as provide quantifiable assessments of your niche-market recommendations. This paper discusses critical elements to keep in mind as you go through the process.
Rainy Day: Actuarial Software and Disaster Recovery
By Aleksey S. Popelyukhin, Commercial Risk Reinsurance Co., Ltd.
Tragic events with disastrous consequences that are happening all around the globe made disaster recovery and continuity planning a much higher priority for every company. Scenarios, in which data centers, paper documents, and even recovery specialists themselves may perish, became more probable.
Both actuarial workflow and actuarial software design should be affected by disaster recovery strategy. Actuaries may simplify recovery tasks and insure higher rates of success if they properly modify their applications' architecture and their approaches to documenting algorithms and storing structured data.
The paper attempts to direct actuaries to strategies that may increase chances of complete recovery: from separation of data and algorithms to effective storage of actuarial objects to automated version management and self-documenting techniques.
The matter of continuity of actuarial operations is in the hands of actuaries.
DCP-3 Emerging Technology Call Papers Presentation 3
Moderator:
Holmes M. Gwynn, Senior Actuary, Texas Department of Insurance
Modeling Hidden Exposures in Claim Severity via the EM Algorithm
By Richard A. Derrig, Automobile Insurers Bureau of Massachusetts
Grzegorz A. Rempala, University of Louisville
The authors consider the issue of modeling the so-called hidden severity exposure occurring through either incomplete data or an unobserved underlying risk factor. They use the celebrated EM algorithm as a convenient tool in detecting latent (unobserved) risks in finite mixture models of claim severity and in problems where data imputation is needed. The paper provides examples of applicability of the methodology based on real-life auto injury claim data and compares, when possible, the accuracy of the authors' methods with that of standard techniques.
Martian Chronicles: Is MARS Better Than Neural Networks?
By Louise A. Francis, Francis Analytics & Actuarial Data Mining Inc.
A recently developed data mining technique, Multivariate Adaptive Regression Splines (MARS) has been hailed by some as a viable competitor to neural networks that does not suffer from some of the limitations of neural networks. Like neural networks, it is effective when analyzing complex structures which are commonly found in data, such as nonlinearities and interactions. However, unlike neural networks, MARS is not a "black box", but produces models that are explainable to management.
This paper will introduce MARS by showing its similarity to an already well-understood statistical technique: linear regression. It will illustrate MARS by applying it to insurance fraud data and will compare its performance to that of neural networks.