Addressing DC Plan Participant Behaviors With Predictive Analytics

Using “predictive analytics” means using data, technology and tools to look at what has happened to anticipate what an individual might do in the future and address potential behavior that may hurt retirement outcomes.

John Hancock Retirement Plan Services (JHRPS) has expanded its data analytics capabilities to help plan sponsor clients and advisers make plan and platform decisions to help participants save more for retirement.

Lynda Abend, chief data officer for John Hancock Retirement Plan Services in Boston, explains to PLANADVISER that using “predictive analytics” means using data, technology and tools to look at what has happened to anticipate what an individual might do in the future. An example would be considering why participants are taking loans from defined contribution (DC) plans. Predictive analytics looks at which participants are taking loans, what activity were they doing before taking the loan and what activity did they do after taking one. Were participants looking into buying a house then taking a loan? Were they doing a college search then taking a loan? She adds that analytics can determine if there are other participants in a similar situation, and plan sponsors can then segment that population to possibly offer other solutions for financing activities.

So, where does the information come from to do predictive analytics? “Technology has grown significantly over last five or ten years, and consumers produce a significant amount of data with technology that is connected,” Abend says. “In addition, there is a lot of information available to purchase. And, we can go to Bureau of Labor Statistics (BLS) data to get consumer spending habits. While this is not necessarily personalized, plan sponsors can identify participants in similar situations and the challenges they might have. The capabilities today are even different from five or ten years ago.”

JHRPS conducted a predictive analytics pilot with long-term client Farm Credit Foundations (FCF) which had very high participation and retirement readiness but wanted to know why the few non-contributors had opted out after they were auto-enrolled.

Using predictive analytics, JHRPS modeled participant data to identify participant segments—top, normal, and non-contributors—and enriched the data with third-party data to provide broader insight into the personas. It then used machine-learning algorithms to predict future outcomes. The analysis identified who the non-contributors were and provided insight into what might help them save more.

Farm Credit used the data to make targeted plan design changes, decreasing the auto-sweep default to encourage more employees to participate. With the lower default rate, 90% of the non-contributors stayed in the plan after the last auto sweep. In addition, 70% of the new contributors remained at the lower default rate, 16% elected a higher contribution rate and 4% elected to contribute after-tax.

“John Hancock’s advanced analytics team stepped in and helped us change our approach,” said Cynthia M. Burkel, CEBS, SPHR, vice president, Employee Benefits, Farm Credit Foundations, in a statement. “The results of the deep dive into our non-contributing employees surprised us. Without John Hancock’s analysis, we would have continued with our annual auto-sweep unchanged, which would have left some participants behind. John Hancock’s ability to understand the behaviors of these participants enabled us to turn around employee behavior and achieve our goal of getting more people in the plan.”

When asked what is meant by enriching the data with third-party data to provide broader insight into the personas, Abend explains they used data from the recordkeeper about participants’ date of birth, compensation and interactions on the web site with the call center. The company also used data from BLS or other sources to see what challenges this demographic has—such as having debt or being single. “This data helps us create a deep persona about the groups we look at,” she says.

More examples of using predictive analytics

Another example of work JHRPS has done with predictive analytics involves understanding participants’ plans for retirement and the expenses they will have. Looking at what they plan to do in retirement and where they plan to live, predictive analytics can help participants forecast how much their expenses will be and how great an income replacement ratio they will need to achieve their goals. “For example, if a participant is working for a company in New York City, but plans to move to Florida in retirement, expenses will be lower. We can use this to supply advice on a personalized level—how do they need to adjust their savings or investment strategy,” Abend says

JHRPS has also looked at participants who are at risk of lowering their deferral rate. “We looked at the people who have lowered their deferral rates and what is common among them. We identified people who fit into this persona and provided target information and education to try to get in front of that decision, conveying information about why it is important to stay at the higher deferral rate,” Abend says.

She continues, “Predictive analytics allows for plan sponsors, advisers and providers to communicate in a very targeted manner, to get really relevant information in front of participants.”

Abend says JHRPS is using predictive analytics with business partners. “Ideas come from challenges of clients, then we find results are valuable to other clients. The work we do is available across our business and comes from dialogue about what is happening,” she concludes.