Insurance and Actuarial Sciences
Insurance is a data-driven industry, and thus employs a large number of analysts to continuously monitor and analyze data. Analysts in the insurance industry have formal education in a variety of disciplines, including statistics, finance, economics, business, mathematics, and computer science.
Entry-level analysts require a bachelor's degree in one of these quantitative disciplines. Senior analysts may possess an advanced degree. In addition to technical skills, career success requires good project management skills and the ability to communicate effectively with both management and information technology specialists. Qualified individuals have opportunities to move into management roles.
There is much overlap between insurance statisticians and insurance actuaries. For instance, in Progressive Insurance Company, statisticians play a major role in setting the insurance rates, which is traditionally done by actuaries. Being an actuary consistently ranks among the most desirable jobs.
Functions requiring varying levels of statistical skills include the following:
- Pricing and product design
- Multivariate statistical models are used to predict average losses versus driver characteristics (e.g., driver age, gender, marital status, driving record), vehicle characteristics, and geographic location.
- Multivariate models are necessary to separate the contribution of each of these inter-related variables. Such models are used to accurately set the relative price to charge particular segments of customers.
To accurately forecast trends in losses over time, moving averages, linear regression, log-linear regression methods, or time series analysis may be employed.
Does fast response have an impact on the magnitude of claims payments? If a claims adjuster can get right out to the scene of an accident, inspect the vehicle, and meet with the insured and claimants, does that help control the total amount an insurer will pay? A statistical test is used to answer this question.
A new claims adjuster needs to figure out how much to pay for a herniated disk and fractured leg. A number of insurance industry vendors offer models that take into account 20-30 variables (e.g., doctor's diagnoses, type of impact, location of impact, attorney representation) and generate an estimate of how much is typically paid for a particular type of claim feature. These benchmarks are only guidelines, but they provide additional data for adjusters to consider in making an offer.
In marketing, statistical methods are used to model response to advertising campaigns to, for example, target advertising to market segments most likely to respond to the campaigns. Designed experiments also may be used to efficiently test different strategies for increasing sales.
Statistical methods such as logistic regression or survival analysis may be used to identify variables that are predictive of how long a customer stays with the company. For example, such models are used to determine the impact of premium increases on whether a customer renews his policy.
Designed experiments also may be used to efficiently test different strategies for retaining more customers. The results of such customer retention experiments may be used as the basis for actions implemented to increase customer retention.
In operations, computer simulation may be used to model call volume to optimize staffing levels. Quality control statistics may be used to monitor and improve quantitative and qualitative measures of service performance.
Insurance companies invest heavily in information technology to capture and store transactional data in data warehouses in a form suitable for analysis. Thus, analysts have at their disposal massive volumes of data. It is not unusual to work with several hundred thousand, or even millions, of observations and several dozen to hundreds of variables.
As with any data originating from customer transactions, care must be used to recognize possible data errors and understand the limitations of the data. One also must understand the dynamic changes in the data over time.
Knowledge of basic statistics is essential to properly analyze and interpret insurance data. Knowledge of advanced statistics is helpful to implement multivariate methods and develop better methodologies of data modeling. For example, advanced methods such as cluster analysis, regression and classification trees, logistic regression, and survival analysis can be useful to more effectively identify segments of customers for pricing or marketing purposes.
The insurance industry provides many technical challenges to a statistician and opportunities for career growth and advancement.