Search Presentations
STAY TUNED! If you are anticipating additional search filters by attribute and level to align with the CAS Capability Model, it is coming later this Summer. As the CAS begins to code recorded sessions by specific attributes and levels (starting with the 2023 Annual Meeting), these will be tagged in the CAS database of presentations going forward and should be searchable.
But you may use the Capability Model now to help you identify topics. For example, if you want to move up one level under the content area “Functional Expertise,” you may search topics in the particular functional area to expand your knowledge.
Recorded content is searchable by Capability Model attribute and level in the CAS Online Library.
Refine Your Search
The presentation materials are offered in connection with CAS professional education offerings. © 2022 Casualty Actuarial Society. All Rights Reserved. The presentation materials may contain copyrighted content the use of which has not been specifically authorized by the copyright owner. You are permitted to view and print the materials for personal/professional noncommercial research purposes. Except for the foregoing, you agree not to reproduce, distribute, modify, create derivative works, or commercially exploit the presentation materials without prior written permission from CAS. Please direct any copyright permission inquiries regarding use of the presentation materials to acs@casact.org.
STAY TUNED! If you are anticipating additional search filters by attribute and level to align with the CAS Capability Model, it is coming later this Summer. As the CAS begins to code recorded sessions by specific attributes and levels (starting with the 2023 Annual Meeting), these will be tagged in the CAS database of presentations going forward and should be searchable.
But you may use the Capability Model now to help you identify topics. For example, if you want to move up one level under the content area “Functional Expertise,” you may search topics in the particular functional area to expand your knowledge.
Recorded content is searchable by Capability Model attribute and level in the CAS Online Library.
An Overview of the Economy
Nelson Oliver, Research Analyst at the Federal Reserve Bank of Cleveland provides a brief overview of the national economy with a more in-depth look at the regional economy.
Source:
2015 Regional Affiliate - BACE
Type:
affiliate
Panelists:
Nelson Oliver
The Market Cycle and Its Impact on Reserves
Zachary Ballweg, FCAS, MAAA, from Milliman, provides insights on the interrelationships among the underwriting, economic, and reserving cycles.
Source:
2015 Regional Affiliate - BACE
Type:
affiliate
Panelists:
Zachary Ballweg
Improving Rating Accuracy and Segmentation for Commercial Lines Policies
Usage Based Insurance is an ever-more frequent topic of conversation in the property/casualty auto insurance industry. Today a growing number of Commercial Vehicles have telematics devices installed, almost three times as many as in personal vehicles. The goal is to increase profits for employers, yet this data is almost never used to augment commercial lines models. This session will attempt to provide a comprehensive overview of the latest options for modeling and segmentation using this data. The presentation will be vendor neutral and focus on what data is available for traditional and predictive modeling.
Can real time data be affordably implemented in predictive models to create a more accurate, and thus more profitable, rating and pricing?
This session will review why current Usage Based Insurance “driver scores” don’t always provide the expected lift, with an emphasis on what must be done to reduce the noise in these scores. We will look at normalized data from up to 15 different Telematics Service Providers (the fleet management companies who serve business fleets with installed devices), to provide some basic and much needed insight into where, when and how a fleet’s vehicles are being driven. This session will discuss the pros and cons of using a traditional GLM design with driver and fleet scores, versus using non-linear models. Modeling telematics data without using time series data will be discussed with examples of why it does not correlate with actual loss experience, especially for high mileage vehicles. This discussion will cover both business and statistical considerations for personal and commercial lines telematics data, including the challenges of portability. The panelists will offer insight into how data can be normalized across the spectrum of vehicles and the driven environment, and the results fed back for increased lift.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Concurrent Session
Moderators:
Jennifer Balester
Panelists:
Christopher Wilson, Jim Noble
How Predictive Analytics Can Change Your Market Footprint
There are an array of statistical techniques that are available to identify patterns in your existing data. Your existing data reflects the book of business that you have historically written – it represents a cross section of customers that identify with your brand and price. But what if you wanted to expand your brand and grow your book in market niches that don’t currently resemble your existing distribution? The purpose of this session is to discuss analytical approaches to extrapolating beyond your existing distribution thereby altering your market footprint. We will discuss a range of analytical approaches in working on this problem highlighting the benefits and risks of each approach. Finally we will conclude with a General Session discussion of how changing your market footprint is more than just an analytical problem.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Concurrent Session
Moderators:
Sara Hemmingson
Panelists:
Serhat Guven
HealthCare Reform
The Patient Protection and Affordable Care Act (PPACA) was signed into law in 2010. After an overview of the key provisions, the presenters will discuss possible implications of the Act on property casualty liability insurance coverages including Medical Professional Liability and Workers Compensation.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Concurrent Session
Moderators:
Sara Hemmingson
Panelists:
Ann Conway, Anne Petrides
GLM Ratemaking Innovations from the University of Wisconsin
Professors Jed Frees and Peng Shi of the University of Wisconsin will speak about recent research using real data that have applications to predictive modeling of property-casualty insurance.
Presentation #1: "Rating Endorsements Using General Sessionized Linear Models" -- Jed Frees
Insurance policies often contain optional insurance coverages known as endorsements. Because these additional coverages are typically inexpensive relative to primary coverages and data can be sparse (coverages are optional), rating of endorsements is often done in ad hoc manner after a primary analysis has been conducted. This paper describes a study of the Wisconsin Local Government Property Insurance Fund where it is desirable to have a formal mechanism for rating endorsements. Our goal is to provide prediction algorithms that are transparent and that promote equity among policyholders by determining rates that reflect the appropriate level and amount of uncertainty of each risk. To accommodate potentially conflicting goals of data complexity and algorithmic transparency, we utilize shrinkage techniques to moderate the effects of selected covariates with penalized likelihoods. We find that the rating algorithms using shrinkage techniques have a predictive accuracy that are comparable to unbiased General Sessionized linear model techniques and provide relativities for endorsements that are consistent with sound economic, risk management, and actuarial practice.
Presentation #2: "Ratemaking With a Copula-Based Multivariate Tweedie Model" -- Peng Shi
The Tweedie distribution, featured with a probability mass at zero, is convenient for pure premium determination in property-casualty insurance. In practice, insurance contracts often bundle dependent types of risks, either from different policies or from different coverages within a policy. Moreover, repeated contracting also leads to serial dependence over time. Motivated by the above observations, this paper proposed a multivariate Tweedie model, with the dependence among Tweedie outcomes accommodated by a copula. The model allows for a flexible dependence structure in a General Session multilevel framework, including both cross-sectional and intertemporal dependence among claims. To avoid the computational difficulty in the high dimensional setup, we refer to the composite likelihood approach for model inference. Case studies are provided using an automobile insurance dataset from Canada.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Concurrent Session
Moderators:
Sara Hemmingson
Panelists:
Peng Shi, Edward Frees
GLM III - The Matrix Reloaded
This session will consider new techniques and refinements to the basic GLM which can add material value to the modeling process. It will specifically consider amendments which address some of the purported failings of GLMs in comparison to emerging methods such as machine learning techniques.
The session will include a discussion of: • An innovative approach to detecting subtle and higher dimensional interactions in an efficient way, potentially eliminating the need to consider alternative, harder-to-implement model forms such as nonlinear models.
• The role of such automated methods in comparison with more manual construction of composite explanatory variables.
• Ways to mitigate the risk of over-parameterization through the use of modifications which incorporate elements of credibility within the GLM framework.
• Simple practical modeling steps that can be used to remove distortions created by combining models across claim types.
• Innovative ways of modeling bodily injury claims and other miscellaneous refinements.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Concurrent Session
Panelists:
Claudine Modlin, Brent Petzoldt
GLM - II
GLM I provided the case for using GLMs and some basic GLM theory. GLM II will be a practical session outlining basic modeling strategy. The discussion will cover topics such as overall modeling strategy, selecting an appropriate error structure and link function, simplifying the GLM (i.e., excluding variables, grouping levels, and fitting curves), complicating the GLM (i.e., adding interactions), and validating the final model. The session will also include a discussion on diagnostics that help test the selections made.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Concurrent Session
Panelists:
Paul Bailey, Brent Petzoldt
GLM - I
Do terms such as “link function,” “exponential family,” and “deviance” leave you puzzled? If so, this session will clarify those terms and demystify General Sessionized linear models (GLMs). The session will provide a basic introduction to linear models and GLMs. Targeted at those who have modest experience with statistics or modeling, the session will start with a brief review of traditional linear models, particularly regression, which has been taught and widely applied for decades. Session leaders will explain how GLMs naturally arise as some of the restrictive assumptions of linear regression are relaxed. GLMs can model a wide range of phenomena, including frequencies and severities as well as the probability that a claim is fraudulent or abusive, to name just a few. The session will emphasize intuition and insight in addition to mathematical calculations. Illustrations will be presented using actual insurance data.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Concurrent Session
Panelists:
Ernesto Schirmacher
Getting Started With Price Optimization - Concepts, Models and Hurdles
Demand modeling and insurance price optimization are quickly gaining attention in North America as a growing number of companies acknowledge the need for more sophisticated pricing. This session will provide guidance on what companies should consider when they start adopting these techniques, for example the techniques available in the market, the results of academic research, and diagnostics available to evaluate the models. Presenters will also explore the key challenges companies should expect to face theoretically, technically and operationally, and the solutions that can be considered to deal with them.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Concurrent Session
Moderators:
Sara Hemmingson
Panelists:
Yves Colomb
Effective Methods for Measuring and Evaluating the Success of Newly Implemented Pricing Strategies
There’s truth in the saying “If you cannot measure it, you cannot improve it.” But what if some of those things you think are immeasurable canactually be measured and subsequently improved?
The designed experiment with a randomly selected control group is the gold standard for evaluating competing alternatives. However, formal experimentation is usually infeasible because of ethical, legal, or practical considerations.
It is common for insurance companies to invest a significant amount of time and effort in coming up with new pricing strategies; however, often the effectiveness of these strategies is in question due to the difficulty in disentangling the “treatment” effect from other factors which can’t be controlled for directly.
In this session, using North American insurance pricing constraints, we will explore different methods, such as “Differences in Differences” and Propensity Scoring, to measure the impact of newly implemented pricing strategies under such situations. We will also discuss the application of some of these methods including building robust new business demand models and, in particular, estimating the price effect on new business conversion rates without the ability to perform price tests.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Concurrent Session
Moderators:
Sara Hemmingson
Panelists:
Reuven Shnaps
Dr. Frankenstein Created the GLM – What Could Go Wrong?
Although seldom discussed, GLMs, along with many other regression approaches, suffer from two unpleasant side-effects. One is the that GLMs are no more successful at distinguishing signal from noise. The other is the stopping problem; when have we put enough variables in. During this session, empirical studies on real experience data are presented for both problems. Finally, the experimental approach is also set out, allowing practitioners to satisfy themselves by replicating the results on their own books. This session will also cover ensembles, and what's next in Analytics.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Concurrent Session
Panelists:
Christopher Cooksey, Paul Beinat
Determining the Cost Savings from a Workers Compensation Claims Predictive Modeling Project
Many workers compensation claims predictive modeling projects are expected to produce results that can be used to reduce a payer’s claim costs. However, measuring the cost impacts of initiatives implemented from these projects can be difficult. These initiatives are typically introduced in situations where other changes are occurring at the same time, consideration needs to be given to the interactions between the amount of medical care and the duration of lost work time, it may difficult to control for the many other factors that can influence claim costs, and economic conditions may influence return-to-work outcomes.
This session will describe four types of studies commonly used to perform cost-impact evaluations and that can (and have been) used for workers compensation claims: (a) randomized, controlled trials, (b) contemporaneous, parallel evaluations, (c) pre-post evaluations, and (d) time-series evaluations. The session will describe the strengths and weaknesses of each type of study. The session will describe the outcome measures that can be used for each type of study, including considerations for breaking down medical costs into utilization and price-per-service components. The discussion will cover the design of participant and control groups, controlling for confounding factors (such as the introduction of other competing cost-control initiatives and differences in medical treatment protocols), and adjusting for differences in claim characteristics (such as worker demographics and employer and employment characteristics). The discussion will also cover the significance for controlling for state regulations (such as medical fee schedules, treatment guidelines, and prior authorization), which could be important if there were changes in state regulations during the period covered by the evaluation or if claims from several states are included in the analysis.
The session will also discuss how attention may need to be given to issues other than cost impacts. For example, for a workers compensation initiative to reduce medical costs, there may be interest in the impact on the quality of medical care and return-to-work outcomes. This session will discuss considerations for data limitations, and differences in approaches when detailed cross-sectional data are not available and the evaluation must rely on aggregate data.
For each type of study, examples from published literature will be presented.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Concurrent Session
Panelists:
Philip Borba, Terry Wade
Demystifying External Data: Sources, Maps  (Including Premium Competitiveness) and Reports
In today’s world there is a lot of data available, but being able to visualize the data and present it in an organized fashion is a powerful skill. In this session Garrett Bradford, a cartographer who helps insurance companies visualize data, will show you how to obtain census data, attach it to a data set, and use mapping software to visualize the data. This will be a how to class taught using the ArcGIS mapping software and real data. The how to class will be followed by a demonstration of the power of data visualization presented by Howard Botts, PhD VP Chief Scientist CoreLogic Spatial Solutions.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Concurrent Session
Panelists:
Sheri Scott, Howard Botts, Garrett Bradford
Customer Lifetime Value - Opportunities and Challenges
Customer lifetime value (CLV) is a useful tool in marketing and customer relationship management (CRM). CLV has been gaining ground in the insurance industry over the last several years. Despite the theoretical simplicity of CLV, it is fraught with difficulty when applied in practice. In this talk, we will discuss critical issues to consider when modeling and implementing CLV applications within the insurance industry.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Concurrent Session
Panelists:
Mohamad Hindawi
Current Regulatory and Market Advancements in the China P&C Insurance Market
China’s insurance market has experienced explosive growth over the last decade. As the market continues to grow and mature, China insurance carriers have made progress in various areas in competing in the market place. In this session, the panelists will discuss recent regulatory and market advancements in the China P&C insurance market, including market trend, advanced analytics, predictive modeling, pricing deregulation, and regulatory environment.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Concurrent Session
Moderators:
Sara Hemmingson
Panelists:
Jun Yan, Zhigang Xie, Joe Liang
Crowdsourcing Models – The Kaggle Challenge
One of the most interesting developments in predictive modeling is “crowdsourcing”. One example of this phenomenon is the relative success of betting markets (compared to traditional polling methods) in predicting events such as the probability of a candidate’s winning an election. Kaggle is a company that facilitates crowdsourced competitions to solve problems that have been sponsored by private companies. Although this phenomenon has impacted many fields earlier than it has insurance (as is typical), it has arrived in the P&C world. There have now been two major challenges sponsored by top P&C companies; the most recent was this year, when one such insurer sponsored a competition to predict fire peril losses. The ensuing competition ran from July 8 through September 2, and predictive modelers from around the world including statisticians, machine learning experts, and actuaries, competed for the top prize.
We’ll tell the story of this competition from the angle of the modelers. How do they work with a dataset that has been "anonymized"? What types of strategies work for the modelers – and how do ensemble models and coalition building come into it? How is the standings board calculated and updated? What are the possible implications for the way that companies will develop predictive models in the future? And what kind of domain knowledge will be most useful in tomorrow’s analytic world?
Source:
2015 Ratemaking and Product Management Seminar
Type:
Concurrent Session
Moderators:
Sara Hemmingson
Panelists:
Mark Goldburd, Dmitriy Guller, Owen Zhang
Crossing the Big Data Chasm: Accessing, Assessing and Implementing Competitive Advantage Insights
For several years now Big Data has been making Big Headlines in technical and customer analytics media. However for many, if not most, of us it remains an elusive concept. This session will cover how this paradigm could be approached to provide the actuarial community with a pragmatic reference point. It will also address why this data paradigm should be relevant to ratemaking actuaries with the aim of helping actuaries access, assess and implement Big Data applications.
The talk aims to be the end of a new beginning by also providing feedback from an ASTIN working party which will be producing, subject to ASTIN endorsement, a short note about examples of crossing the chasm of this emerging paradigm, starting small with the aim of actuaries building on this initial work. This, together with input from attendees, aspires to encourage our peer group to embrace this distinct data discipline with applications which positively impact the lives of our customers. In so doing well strive to keep the actuarial profession at the forefront of emerging business practices.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Concurrent Session
Moderators:
Sara Hemmingson
Panelists:
Prashant De, Jason Paschalides
Cost of Capital and Capital Attribution in Ratemaking — A Primer for the Property-Casualty Actuary
This session will provide a historical primer on the subject and address past and current research of the practical ways in reflecting the cost of capital in ratemaking. This primer will discuss aspects as developed by Merton-Perold, Myers-Read, Mango's asset share model, and the RMK procedures. Also considered will be current research as it relates to how actual capital should perhaps be considered in percentile layers and/or traunches, based on the risk metrics considered. We have come a long way in ratemaking since using premium to surplus ratios of 30 years ago. Let's continue the evolution.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Concurrent Session
Moderators:
Sara Hemmingson
Panelists:
Glenn Meyers, David Ruhm, Neil Bodoff
Beyond Symbol: How Ensemble Models and Vehicle Build Are Transforming Vehicle  Ratemaking
In this session, we weigh the pros and cons of traditional methods which look at vehicles' experience vs. advanced methods which consider vehicles' attributes in a multivariate setting. We then use a case study to illustrate the extent to which an ensemble approach incorporates the strongest features of each. Finally, we look at how new sources of vehicle information are evolving beyond traditional VIN decodes, and are making it possible to include very detailed information about safety features, such as lane departure warning and adaptive cruise control into insurers' symbol processes.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Concurrent Session
Moderators:
Sara Hemmingson
Panelists:
Jim Weiss, Eric Bower
Basic Ratemaking: Part IV — Large Account Pricing
This session will examine the differences between large accounts and small accounts, and how they can and should be treated differently from an insurance pricing perspective. Participants will be introduced to pricing methods that both utilize the increased credibility of large account experience and recognize the uniqueness of each risk. Specifically, this session will serve as an introduction to experience rating, schedule rating, and retrospective rating.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Workshop
Moderators:
Sara Hemmingson
Panelists:
Angelo Bastianpillai
Basic Ratemaking: Part III — Estimating Claim Liabilities
This session will focus on estimating claim liabilities including development factor techniques, advantages and disadvantages of different methods, and diagnostics of the various methods.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Workshop
Moderators:
Sara Hemmingson
Panelists:
Alejandro Ortega
Basic Ratemaking: Part II — Ratemaking Relativities
This session will examine the various methods that actuaries use to allocate overall average rates to various subdivisions of a line of business, including territories, classifications, and tiers. Some of the methods discussed will consist of univariate, multivariate, and General Sessionized linear modeling techniques.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Workshop
Moderators:
Sara Hemmingson
Panelists:
Christopher Cooksey
Basic Ratemaking: Part I — Exploring the Fundamental Insurance Equation
The fundamental insurance equation is the foundation of the ratemaking process. This session will include techniques for adjusting historical data to estimate its various components (premiums, losses, expenses, and profit) in the relevant pricing time period. The session will conclude by briefly exploring other considerations company management should make, along with the cost-based rate indication derived from the fundamental insurance equation, to determine what rates to charge in practice.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Workshop
Moderators:
Sara Hemmingson
Panelists:
Samantha Steiner
Autonomous Vehicles and the Impact on the Insurance Industry
Many new vehicles offer advanced crash avoidance features. The systems started out as options on a few luxury models and have steadily spread to more of the fleet. Advanced technologies assist the driver with warnings or automatic braking to help avoid or mitigate a crash. Advances are also being made in intelligent transportation systems that allow vehicles to communicate with one another or with road infrastructure to help avoid crashes. As we move from an operation based upon a human responding to an environment they observe, to an operation where a machine responds to data it records, how that data will be handled and used is a large and currently unresolved question. This panel will supply an update on CAS research, discuss these technical opportunities, the political and legal risks, and the potential impacts to insurance that should be considered by insurance companies and actuaries.
Source:
2015 Ratemaking and Product Management Seminar
Type:
Concurrent Session
Moderators:
Sara Hemmingson
Panelists:
Morgan Bugbee, Robert Peterson, Jonathan Charak
