Tag Archives: big data

Addressing problematic defined benefit data

Defined benefit (DB) plan sponsors often express how complicated their plans are to administer.  Many times, the reasons are due to unique plan provisions, but more frequently are related to missing and/or incomplete historical participant data.  As plan sponsors move to outsourcing their DB plan administration, these data issues need to be addressed during the conversion process. These issues directly affect the efficiency of the ongoing administration.  Tackling data issues can become overwhelming, but with thoughtful planning and consulting, data concerns can be mitigated, if not eliminated for the most part. 

What exactly is bad data?   A common request during the vendor search process is to have the plan sponsor provide some indication of the state of the data.  The initial reaction to this request is typically along the lines of “we have data and our actuary is using it to produce a valuation report each year, so what’s the concern?”  Actuarial quality data and administration quality data are not the same.  During the annual valuation process, the calculations used to determine the plan liabilities are based upon approved actuarial assumptions and are not designed to be exact; they only need be reasonable, justifiable, and within defined parameters. By contrast, individual participant calculations are based on the plan’s formula and historical data specific to each participant.  The participant is relying on these calculations to make an important financial decision; as such, it is critical and that the data used for benefit calculations be accurate.

“Bad” data can cause problems. For example, bad data could result in the participant receiving overstated or understated amounts when modeling benefits or when electing to begin receiving benefit(s).  The data needed to calculate a benefit that can be “bad” is specific to each DB plan.  The data elements frequently found to be problematic include historical salary and hours needed to calculate accruals as well as employment status details needed for service calculations.  This is true for both traditional and cash balance plans.  In the event the plan has had other plans merged into or divested from it, the historical participant indicators, grandfathered benefit amounts, and identification of deferred participants and vested amounts can be lost.  For plans that have existed for a long time, certain data elements may only exist on paper files in a basement storage space or another questionable location.  It is no wonder plan sponsors are anxious when this question is posed or do not fully understand the importance of a candid response. 

How can we clean it up?  Identifying what you don’t know is impossible; however, the annual valuation data is a good place to start.  This file will include any participants used to determine the plan’s liability and should list all participants actively accruing a benefit, those with a deferred benefit, and those currently receiving a payment from the plan.  This can be the starting point to determine what might be missing or incomplete.   Your actuary can also document the assumptions being used and can often indicate missing underlying data needed for calculations and provide direction on how to determine a value for these participants.   

Analysis of the data elements used in your calculation can be extracted from the valuation file.  If this review is taking place during a conversion to a new administrator, your conversion team can identify which participants are missing data elements and which elements could be suspect.  Once identified, data can be collected from other sources to be aggregated and used within the calculation.   Participants can also be flagged to indicate data is missing that must be collected before any projected benefits are presented to the participant.  For deferred participants missing an accrued benefit amount, once the data is identified, the benefits can be calculated and subsequently certified.  This clean-up process can be performed in conjunction with or following a conversion project, depending on resources.

Data challenges are common. Many DB plans have experienced the transition from paper-based data to digital data and, as a result, the plan’s underlying data has likely existed in many forms over time.  Once the issues are clearly identified, the results of the clean-up effort will benefit both the participants and the plan.  The data review could identify administrative errors that lead to the need for Voluntary Compliance Program (VCP) updates or the identification of missing participants, which will be beneficial to identify early for valuation purposes and compliance concerns in preparation for plan termination. Spending the effort to address data issues provides confidence in ongoing administration of the plan and will allow for more accurate valuations in the future.

Knowing participants’ profiles is becoming increasingly important

The debate about a new pension system in the Netherlands is becoming more and more complicated because of issues including solidarity, labor market flexibility, indexation security and uncertainty about the level of pension income. These subjects are complicated. The question regarding whether pension income from retirement date is high enough in relation to income received in active employment or more relevant to the spending pattern is not often mentioned in this context. The questions about how long pension is to be paid out (lifelong) and how much premium participants are willing to pay for their retirement are rarely discussed.

We suspect that one of the reasons that we find these questions so difficult to answer is because we do not really know about the (ex) participants (workers, retirees and former participants with vested pensions). As a consequence, the pension debate becomes an abstract compensation and benefits discussion focused on a complicated financing component.

Having relevant knowledge about our stakeholders could provide significant benefits. If we know and understand our participants well, then:

• Pensions, even without specific customization, could be fitted to stakeholders more appropriately.
• Choosing the most appropriate financing (in terms of risk, duration and reservation) could be ensured.

Getting knowledge and information about our pension stakeholders can be accomplished in various ways. This may include:

• The pension stakeholders ask the right questions at the right level of knowledge-estimated by using available data (such as salary level and job title)-and in understandable language
• Combining knowledge of our pension stakeholders with external data to gain more insight and to better understand their needs.

A good example is the correlation between education level and life expectancy of participants. The Dutch Central Bureau of Statistics (CBS) regularly publishes that the life expectancy of a Dutch man with a highly qualified education at the academic level is much higher than that of a man who has enjoyed a maximum of elementary school education. Milliman calculated that the remaining life expectancy at the age of 68 for the more highly educated group was more than two years greater than for the other group.

In practice, it appears that data about the training of individual participants is often not available to pension funds. If this information were adequately collected and stored in the near future, then additional analyses could be performed using this data. This contributes to the necessary knowledge and insight into the needs of our pension stakeholders. As a result, not only the expected duration of benefits can be determined, but also, by combining this data with other available data, we could estimate the individual’s income needs. The combination of data and analysis of connections between data can create even greater insight. For example, it makes a big difference whether a participant in a retirement scheme has a physically demanding occupation or a light one, whether he travels regularly or stays at home reading, and whether he maintains a healthy lifestyle or just the opposite.

Collecting knowledge about our participants and analyzing already available knowledge or information (big data) could ensure that we design better pension schemes and that their funding takes place in the most appropriate way.

Let’s start with that today. More knowledge and insight into participant profiles helps both the employer and the performer get better “demonstrable in control” information regarding their pension commitments, provisions, and HRM policies.