White paper:

measuring fraud

Americans live and die by numbers — whether it's political polls, the Dow Jones average, movie ratings or a baseball pitcher's latest earned run average. We have come to expect that all things can and should be measured quickly and easily. Business leaders in the 1980s tried to drive home the point by proclaiming that "if you can't measure it, you can't manage it."

The obsession with measuring nearly every aspect of our lives presents a dilemma for the insurance fraud-fighting community. Simply put, insurance fraud is hard to measure accurately. The hidden nature of the crime — combined with other obstacles that will be discussed in this paper — have stymied efforts to develop easy and efficient methods of gauging the extent of fraud, and the effectiveness of specific solutions in curbing it.

Why Measure?

The most-common rationale often given for the need to measure fraud is a simple one: you need to know the extent of the problem to effectively solve it. Without knowing the size and scope of the problem, how can we know how much and where to deploy resources? The scarcity of money and staff to combat fraud suggests that efficient allocation of resources is vital in order to effectively deter and detect fraud.

A secondary rationale for a consistent measurement system is the need to understand the degree of impact various solutions have on the problem. If a baseline understanding of the problem is lacking, then how can the effectiveness of subsequent solutions be measured? This is especially true in determining whether anti-fraud activities have any impact on deterrence, which may be even harder to measure than the extent of the problem.

Another important reason for consistent measurement is public credibility. Consumers and their legislators are more likely to buy into solutions if they are convinced a problem exists. Anecdotal evidence has helped convince a lot of people that insurance fraud is a problem that requires immediate and forceful remedies. However, cracks are beginning to form in this argument, casting doubt on whether insurance fraud is really as bad as insurers and others say. News articles now are investigating the fraud problem and are questioning how widespread it really is.

Legislators also question whether the problem is in proportion to the solutions advanced by the fraud-fighting community. Advocates for anti-fraud legislation have long said their jobs are much tougher without clear and convincing statistics about insurance fraud.

During the 2000 legislative session in New York, for example, a bill to specifically outlaw "running and capping" middlemen who coordinate fraud schemes was scuttled when the Assembly Majority Leader refused to call up the bill because he deemed this problem wasn't a significant one in the state.

Obstacles To Measuring

Insurance fraud means many different things to different people, and therein lies one of the biggest challenges in measuring fraud: There is no universally understood definition of insurance fraud.

From a strict legal sense, a fraud exists only when a court has deemed it so, either through criminal or civil rulings. While measuring court actions is relatively easy most of the time, the results may not be meaningful in gauging the fraud problem because only a small fraction of fraudulent acts ever end up in court.

Other possible areas to measure include:

• Prosecutions and civil cases brought to light

• Cases referred to prosecutors

• Criminal arrests and civil complaints

• Cases referred to law enforcement

• Claims denied or reduced

• Claims referred to SIUs

• Red flags triggered

Even with these criteria, not all fraudulent acts will be captured because some go undetected during the claims process while some suspected of fraud are established as valid.

Another obstacle is defining the type of fraud to measure. Most fraud estimates focus on claims fraud. Applications, underwriting or premium fraud go largely unmeasured. Fraud by insurers or company insiders also is rarely quantified.

Lines of insurance also define how and what gets measured or even estimated. Property/casualty insurance, specifically automobile and workers compensation, have received the greatest attention. Health insurance fraud has been estimated, although somewhat roughly, thanks to federal research of public health programs. Then again, the definitions remain unclear because government studies usually report the loss due to fraud and abuse, without defining either. Where does abuse end and fraud begin?

Research of life and disability insurance fraud is virtually nonexistent, yet there are signs that criminal activity is growing in both areas.

Not everyone in the anti-fraud community favors clear definitions and measurements, either. During the June 2000 National Insurance Fraud Forum, participants in one of two sessions that focused on this issue said some insurance investigators liked a "fuzzy" definition because it allows flexibility and prevents insurers from being buttonholed into "one-size-fits-all" solutions.

Some investigators also believe their performances could be misinterpreted by measurements that might not adequately apply to the structure of their operations or the type of fraud they investigate.

Finding adequate funding to install a consistent measurement system over time is another clear obstacle to measuring fraud.

Past Measurements

Attempts to measure insurance fraud have either been broad estimates, one-time snapshots, or narrowly focused on a single area, such as bodily injury fraud. The most-notable efforts include:

• Coalition Against Insurance Fraud. Publishes annual estimates for claims fraud in four areas of insurance (auto, homeowners, health and business). Estimates are based on extrapolations of previous industry and government estimates of fraud. Latest national estimate for claims fraud in the U.S. is $79 billion.

• National Insurance Crime Bureau. Estimates that insurance fraud totals $18 billion to $20 billion. Estimate focuses on property-casualty insurance, but news media and others often misinterpret this figure to include fraud across all lines.

• Conning & Company. Published a landmark study on insurance fraud in 1996 that estimated claims fraud at $120 billion, mostly based on other estimates.

• U.S. Government Accounting Office (GAO). Published a 1992 study that estimated fraud and abuse in Medicare and Medicaid reached as much as $100 billion annually.

• Insurance Research Council. Conducted a closed-claim study in 1995 of automobile bodily injury claims that concluded about 36 percent of those claims had signs of fraud or buildup. While 18 percent of the claims were suspected of being fraudulent, only one in five were thought to be planned as a staged or caused accident

• Automobile Insurers Bureau of Massachusetts. Conducted the original closed claim study in 1991 that found 11 percent of auto bodily injury liability claims were suspected of fraud, with an additional 21 percent judged to be buildup claims.

• Rand Corporation. Conducted a study of the Insurance Research Council data of bodily injury liability and no-fault claims that concluded that about 28 percent of all claims submitted by auto accident victims are exaggerated for the purpose of collecting insurance payments.

• U.S. Chamber of Commerce. Reported that 25 percent of all workers compensation claims are fraudulent.

The varying estimates are confusing and often contradictory, and the statistical methods do not always hold up to rigorous analysis. When examined by the public, media and legislators, these figures fail to definitively measure how large the fraud problem really is.

Measuring SIUs & Fraud Bureaus

The performance of special investigation units (SIUs) and state fraud bureaus also are often considered ripe for benchmarking, or at least consistent and continual measurement of various aspects of their operations and results.


There's a common belief that most SIUs systematically measure their return on investment to their companies. A variety of measurements involving cases opened, files forwarded, referrals from claims department, savings based on reserves, cases referral to prosecution, etc. are used.

One large personal lines insurer, however, does not measure its SIU operation, contending that whatever the savings, an aggressive anti-fraud program is in the best interest of policyholders and is "the right thing to do." Other insurers downplay measurement for fear that such a system could be used against them in civil litigation that the company unfairly denies claims, perhaps even legitimate ones.

Some insurers have sophisticated systems of measurement that let them measure savings and understand the benefits of adding staff investigators versus the cost of outside contracted services.

Still, even with an outstanding system of measurement, a single insurer cannot measure itself against industry-wide standards, because none exist.

During the Fraud Forum, the two discussion groups were split on the idea of a uniform system of measurement to gauge performance of SIUs. The first group thought benchmarking would help management better compare performances from insurer to insurer. This would give management more confidence in the SIUs and thus attract more resources to anti-fraud activities. Benchmarking also could spotlight fraud units that excel, and highlight the reasons for the success, possibly encouraging more innovation and experimentation.

But the second discussion group wasn't convinced a uniform benchmarking system is now possible because carriers now use so many different measurement systems.

Designing a system that is valid across different lines of insurance was another obstacle. Can an SIU who mostly deals with workers comp cases be compared with one who investigates automobile insurance fraud? The differences between property/casualty insurance and health/disability/life are even starker, potentially making a benchmark even less-useful.

A third obstacle is the inability to measure the intangible benefits of SIUs. The mere creation of a fraud-fighting entity should help deter crime, yet actually measuring that deterrence is nearly impossible. SIUs also provide functions other than investigating claims for the purposes of denials: Reviewing claims files that are quickly determined to be legitimate claims allows insurers to pay claimants promptly and close files quickly.

Fraud Bureaus

Fraud bureaus should use consistent and continual measurements, the participants agreed. Perhaps because of the culture and requirements of state government, developing a uniform system would be easier and less-demanding, especially since many of the current 45 units already capture data necessary to file annual reports to legislatures, governors and state insurance commissioners.

Fraud forum participants generally agreed that state agencies should be held accountable for their performance whether they are funded by tax dollars or by industry assessments. While measuring fraud bureaus is easier because only 45 exist, they still face some of the same problems as measuring SIUs.

The structure of fraud bureaus varies from state to state, as does their focus, funding and capabilities. Some deal with all types of insurance and insurance fraud, while others narrowly focus on workers comp or auto insurance.

Capabilities vary even more. A few state units have more than 100 investigators, full police powers and the ability to prosecute, while others employ just a handful (as few as four or five) investigators with few or no special powers.

The challenge is to develop a consistent and fair system that effectively compares the performances of all the state agencies, and gauges changes in performance, capabilities and structure from one year to the next.

Previous attempts to measure fraud bureaus have included the following data points:

• Number of employees

• Number of investigators

• Years of existence

• Where in state government is the agency housed

• Total budget, budget as a fraction of insurance lines covered premium

• Number of pending cases, arrests, cases referred to prosecution, cases prosecuted and convictions — criminal and civil.

To help provide a fair comparison, various ratios have been published, including number of cases per investigator, cases per capita and prosecutions per million of dollars of premium written in the state.

While such measurements are far from perfect, they do offer a fairer method of comparison from state to state. However, they still do not measure the intangibles that a fraud bureau brings to the table, especially deterrence.


Measuring Fraud

• Develop concise definitions of fraud for the purposes of uniform measurement, and promote the understanding and use of the definitions with industry, government, academia and the media. Terms might include "suspected fraud," "referred fraud" or "convicted fraud," depending on the action taken.

• Make clear the role of exaggerated or buildup claims as they relate to the definitions of fraud.

• Investigate the feasibility of developing methods of conducting closed claims studies in non-auto lines, including homeowners, workers compensation and health insurance.

• Investigate securing the aggregate reporting data insurers file annually with state insurance departments, to determine whether data could be extrapolated for measurement purposes.

• Develop methodologies for measuring the extent of fraud in the application process, especially in automobile and life insurance.

• In the interim until more accurate measurements can be ascertained, major anti-fraud organizations should seek agreement on more-consistent estimates, or perhaps resist publishing estimates that are not based on some level of scientifically valid empirical studies.

Measuring SIUs

• Publish a guidebook to assist insurers in measuring the performance of SIUs that would include case studies and the various measurement systems used by insurance companies. Provide an in-depth analysis of the benefits and drawbacks of various systems. Underscore the importance that intangibles such as deterrence play in fraud prevention, and that the total worth of a special investigations unit cannot be calculated by dollar savings alone.

Measuring State Fraud Bureaus

• Continue publishing statistical studies that measure the state fraud bureau performances, and compare them from state to state, using a variety of ratios that help keep results in perspective. To better understand the capabilities of each bureau, such studies also should describe the bureaus' structures and discuss how they operate.


Measuring insurance fraud will never be easy, and likely will remain controversial. Reaching consensus on definitions and methods will be hard, especially when people focus narrowly on their own operations instead of keeping the big picture in mind.

But such consensus is essential if effective measurement programs are to help spotlight the damage caused by this crime, and ultimately convincing the public and decision-makers how much insurance fraud affects the U.S. economy and lives of Americans everywhere.

A unified, all-industry approach to measurement is needed if we expect to prove that the fraud-fighting community is serious about managing this responsibility diligently, and that, indeed, insurance fraud is a severe social and economic problem in the United States.