By Timothy Murphy
On March 23, 2010 President Obama signed into law the Affordable Care Act (ACA). One of the primary objectives of the ACA was to reduce the number of uninsured Americans. And depending on the source, it appears that the number of uninsured has decreased by approximately 25 percent within the first year. However, as noted in a recent Wall Street Journal article, 30 million Americans are still uninsured, despite the fact that nearly 20 percent would be eligible for coverage at almost no cost beyond the time it takes to enroll. And in an effort to reach many of those who are still uninsured, federal officials intend to spend $32 million to increase enrollment and close this gap.
This poses an interesting problem. For many, choosing health care coverage would likely be financially more beneficial than forgoing it, especially for those eligible for near-free coverage. Yet so many do not take action despite the millions of dollars spent to increase enrollment.
An opportunity exists to increase participation through the combination of two disciplines, data science and behavioral science. Jim Guszcza, a chief data scientist at Deloitte, refers to scenarios like this as last-mile problems. These are situations that can benefit from analytical models for targeting and segmenting, but additionally, behavioral nudges are implemented to entice the desired action.
Last-mile problems: From presidential campaigns to fraud detection
Recently, combining data analytics and behavioral economics has yielded some encouraging outcomes.
Guszcza highlights President Barack Obama’s 2012 presidential campaign as an effective application of this last-mile approach.  At the time, the Obama campaign employed a team of data scientists to target undecided voters who were the most likely to be persuaded to vote in his favor. But absent the right intervention, the model could not inform the proper actions to take.
To overcome this hurdle, the campaign implemented behavioral-based outreach strategies to encourage the intended action: vote. One strategy involved commitment devices. The campaign understood from behavioral research that they could increase the likelihood of achieving their goal by requesting that people complete and sign a “commitment card” bearing a photograph of President Obama. Reinforcing this strategy, they capitalized on the power of social proof by explaining to these “persuadable” individuals how their neighbors intended to cast their votes.
Fraud detection has also benefited from these complementary sciences. In 2015, the New Mexico Department of Workforce Solutions (DWS) applied this thinking to reduce unemployment insurance fraud. Specifically, they found that solely applying algorithms to detect dishonest behavior was not enough to adequately reduce unintended behavior. The reason: Algorithms are not 100 percent accurate and a false accusation of fraud can result in dire consequences for both the accuser and accused.
The DWS identified an opportunity to still use the algorithms to target those likely to commit undesirable behavior but they turned to more subtle nudge tactics to reduce the unintended behavior. First, they looked to lessons from priming research to encourage honest behavior; for example, they required people to certify accuracy since research suggested priming individuals with assertions of honesty increases the likelihood that honest behavior will follow. The work that the Behavioural Insights Team accomplished with the “190 million pound sentence” also proved helpful. By explaining to claimants that “9 out of 10 people” accurately report their earnings, they invoked social proof to nudge greater compliance. And through randomized control trials, the DWS was able to point to evidence that the low-cost nudges effectively kindled the honesty they desired.
Applying the last mile in health care
These last-mile lessons may be applicable to those tasked with reducing the number of uninsured. To start, data analytics and segmentation may identify individuals with the highest likelihood to opt-in to insurance coverage. It could be those eligible for free insurance or even more specifically, the algorithm could suggest a subset of zero-cost eligible individuals who would be even more likely to enroll due to special circumstances. In contrast, the data might suggest that young, relatively healthy people may be less inclined to seek or accept coverage. A tiered system could help policy makers determine how aggressive the enrollment tactics would need to be to alter behavior for each segment. Armed with this information, we can start hypothesizing the necessary tactics.
Low-effort segments: A well-targeted group may just require a gentle, low-cost nudge. These could include an easy to access form that employs smart default options when filling out paperwork. Slightly more aggressively, they could be presented with commitment cards for enrollment that include the “first three steps” to gaining coverage.
Medium-effort segments: These individuals may require more convincing or help. They may be prime candidates to offer a health insurance “coach” who would guide them through the decision making process, akin to a fitness coach. Another option could involve equipping the decision maker with a health cost insurance calculator that takes into account a variety of attributes such as age and family size. Doing so could help decision makers overcome the present bias that overly weighs the immediate cash outlay.
High-effort segments: This would be a more resource-intensive group. Nudges may not be effective; instead, this group would require more of a “push” approach. Consequently, policy makers may need to consider the more traditional “carrots and sticks” of traditional economics theory to incentivize behavior. However, implementing a targeting mechanism such as an algorithm would hopefully reduce the overall universe of those who require more costly interventions.
Without testing, it’s difficult to predict the efficacy of these interventions. The good news is that we possess both a roadmap and a wide variety of success stories to help guide a well-informed strategy for reducing the number of uninsured citizens—often in a cost-effective, non-intrusive manner.
 Margot Sanger-Katz et al., “Is the Affordable Care Act working?” The New York Times, October 26, 2014, http://www.nytimes.com/interactive/2014/10/27/us/is-the-affordable-care-act-working.html#/.
 Louise Radnofsky, “Millions eligible for Medicaid go without it,” The Wall Street Journal, January 31, 2016, http://www.wsj.com/articles/millions-eligible-for-medicaid-go-without-it-1454277166.
 James Guszcza, “The last-mile problem: How data science and behavioral science can work together,” Deloitte Review Issue 18, January 25, 2015, http://dupress.com/articles/behavioral-economics-predictive-analytics/.
 Joy Forehand and Michael Greene, “Nudging New Mexico: Kindling honesty among unemployment claimants,” Deloitte Review Issue 18, January 25, 2016, http://dupress.com/articles/behavior-change-among-unemployment-claimants-behavioral-economics/?coll=11936.
Deloitte refers to one or more of Deloitte Touche Tohmatsu Limited, a UK private company limited by guarantee (“DTTL”), its network of member firms, and their related entities. DTTL and each of its member firms are legally separate and independent entities. DTTL (also referred to as “Deloitte Global”) does not provide services to clients. Please see www.deloitte.com/about for a detailed description of DTTL and its member firms. Please see www.deloitte.com/us/about for a detailed description of the legal structure of Deloitte LLP and its subsidiaries. Certain services may not be available to attest clients under the rules and regulations of public accounting.