It’s Not Fair
My wife and I are raising three young boys. They are already fantastic negotiators and have an arsenal of tactics to use when interested in staying out longer, getting us to spend money, or asking for treats. Their current go-to argument is, “This isn’t fair.” They claim that our parenting tolerance and allowance should be applied equally among the three of them. For us, as parents and decision makers, this sometimes makes sense; other times, not so much.
I often think about risk in a similar context of fairness. I ask, does a risk assessment need to work the same way—perform equally—for everyone? The analytics to examine equality are straightforward: Simply compare the proportion by subpopulations, such as race/ethnicity or location, to see how equal the distribution is. But, because of the principles of risk, populations that frequent the system often have higher risk profiles. This produces a distribution that is not equal—a bad thing, right? Well, perhaps the answer is not that simple.
Actuarial risk is used in many fields, such as medicine, epidemiology, and insurance. The National Cancer Institute offers tools to help individuals understand their risk, or “know your chances,” of a diagnosis for many different types of cancer. These models can help us understand when risk equality may not be appropriate.
Risk for breast cancer can be calculated based on a set of factors such as medical history, family characteristics, lifestyle, and experiences. The calculation uses these factors to estimate the likelihood of a specific outcome or outcomes happening over time. What is interesting about this example is how risk plays out differently for females and males.
On average, one in eight women experience breast cancer over a lifetime—a far greater rate than for men, which is one in 1,000. This is a very important consideration: I am sure that many folks reading this blog know much better than I do that breast cancer treatment, even for preventive measures, can be intense and invasive. Therefore, risk identification plays a critical role in determining when to offer and take up preventive services. So even though far more females will experience breast cancer, risk identification is important for both sexes.
While the process for assessing risk of breast cancer is different for men and women, the goal is the same: Identify individuals who should be concerned about their risk. The mechanism is based on likelihoods of experiencing an outcome. Risk means the same thing regardless of gender. If an individual is higher risk, research shows he or she is more likely to experience breast cancer. But does this mean the groups are equal? Absolutely not. Far more women, in numbers, are at higher risk for breast cancer.
More women are high risk? Isn’t this unfair? I think most folks would agree that it wouldn’t make sense to hold risk of these two populations to a standard of equality. Why is that? I would argue that it has a lot to do with how the information gets used. Many breast cancer prevention and treatment services have been demonstrated to be effective and helpful. They are also, for the most part, voluntary.
The context of child protection is, of course, quite different than the world of medicine. Many agencies struggle daily with disproportionality. A combination of socioeconomic factors and structural inequities lead subpopulations to be overrepresented. This can lead to subpopulations within an agency with unequal risk profiles. I wonder how much of the desire to make risk fair is caught up in the application or utility of the tool. If we are using risk to be helpful, how much fairness would we demand?
Chris Scharenbroch is associate director of research analytics for NCCD.