Inaccurate data can add to scheme administration costs and increase fund liabilities warns Andrew Scrimshaw, KPMG
Accurate record keeping is vital if pension schemes are to fulfil their basic purpose: to pay the right benefits to the right people at the right time.
There has long been an awareness within the pensions industry that poor quality member data is a big problem, but it is only relatively recently, through a concerted campaign by the Pensions Regulator, that we have had any real impetus to tackle it.
Of course, it is not only members who can lose out as a result of poor record keeping. Incomplete or inaccurate records add to scheme administration costs and can even increase fund liabilities – by as much as 5% on some estimates. With the growing trend towards buying out scheme liabilities, any doubt over the quality of the scheme data can add greatly to the buyout cost. For these purposes, data needs to be adequate at all times; data cleansing merely at the point of benefit vesting is not sufficient.
Where poor record keeping becomes publicly known, there is also a clear reputational risk not only to the trustees but also to the sponsoring employer. (Nobody would want the kind of headlines that HM Revenue & Customs (HMRC) generated earlier this year as a result of a data matching error: “£1.3bn national insurance error could hit thousands of pensions”, being one of the milder examples. The fact that the problem lay primarily with the quality of data supplied by employers did not prevent HMRC from having to take the rap.)
In strengthened guidance issued in 2010, the Pensions Regulator (TPR) recommended specific targets and timescales for record keeping, as well as an industry wide mechanism for measuring and assessing member records.
TPR identifies 11 common data items that should be present for all members of all schemes, eg name, date of birth, address. It also sets targets for the presence of this data, to be achieved by the end of 2012:
(There is certainly a case for arguing that schemes should aim for 100% in both categories.)
TPR guidance divides member data into two distinct categories:
Conditional data items are those required for the administration of the scheme. Their presence will depend on many factors – for example, the type of scheme, scheme design, the member’s current status and events that have occurred during the individual’s membership. TPR considers that administrators and trustees are best placed to decide the constituents of this data, based on an understanding of their own scheme and administration system.
Trustees need to understand the extent to which the data needed to administer the scheme is present and the quality of that data. So, tests should be designed to indicate not only the presence of data but also areas where it may be inadequate. Robust targets for the standards of conditional data should be set. Where tests indicate there may be inconsistencies or errors, trustees should work with their administrators to develop action plans for further investigation and correction as appropriate. TPR requires trustees to make all reasonable endeavours to meet the targets above, with a timetable for corrective action by December 2012.
After completing their plans, or where the tests indicate the records are acceptable, trustees should conduct measurement tests annually as evidence that their internal controls are operating effectively.
A commercial data testing service can help trustees to meet these requirements, reviewing scheme data in bulk to identify not just information gaps but also anomalies. These anomalies might include such things as incompatible retirement and birth dates, as well as more complex items such as “spurious” national insurance numbers and “unusual” dates. (Do we really have a member born in 1864?) It should also be possible to customise tests to address any particular scheme- or member-specific concerns.
Where remedial action is required, data testing tools can facilitate updates using both public and private information sources to resolve data deficiencies and improve data quality.
A recent analysis1 of member records produced by KPMG, using its KLean data quality tool, revealed some worrying findings. Among the various failings highlighted, the most common was that a third of records were missing some or all of the entries in the members’ national insurance records.
Other commonly occurring errors included missing guaranteed minimum pension figures (17%), retirement dates that were inconsistent with the members’ expected retirement ages (13%), missing member contributions (6%) and missing salary data (5%). It was even found that some 6% of members were recorded as leaving the scheme on or before the date they were recorded as having joined it.
As might be expected, many of the data errors highlighted above relate to problems with legacy data, for example where the scheme is relying on decades old, paper based systems or where there have been repeated changes of administrator. Another common reason for error is the adoption of new administration systems with new data fields that require completion; these often end up with dummy entries (that also serve to mask the extent of the problem).
While TPR’s lower target for older data reflects the legacy issue, the problem is not simply that older data is more likely to go missing. The marked concentration of errors in particular data fields shown in the KLean analysis indicates that there are often systemic failings as well.
For new member records created now, there is no real excuse for common data to be missing – but data errors still occur, often through human error.
Earlier this year, TPR carried out a survey to test the take up and effectiveness of its record keeping guidance.
The proportion of occupational schemes with administrators who have read the guidance and taken action increased from 31% in 2010 to 42% in 2011, still leaving a significant majority who have apparently yet to take any action towards meeting the 2012 target date.
Perhaps not surprisingly, schemes where the administrator had read and acted on TPR’s guidance were much more likely to have alerted trustees to a common data problem. So, it is hard to avoid the rather obvious conclusion that data errors are most likely to be found in schemes that check for them… (There are grounds to think that the schemes that have checked data so far have often done it because they are confident their data is relatively sound. There must be a concern that those that have not checked are avoiding doing so because they fear their data is poor. Analyses to date may therefore be underestimating the problem.)
Still, those schemes whose processes are revealing record-keeping errors can at least take comfort from the fact that TPR is much less likely to escalate to enforcement action where there is evidence of data problems being treated seriously and plans put in place to deal with them. Schemes which have so far ignored this issue need to start taking action soon, if they are to meet the 2012 deadline.
|(1) For more details of the member records analysis, please email firstname.lastname@example.org|
Author: Andrew ScrimshawAndrew Scrimshaw is technical services manager, pensions, at KPMG.