Earlier this year, we analysed over 3-million-member records across nearly 70 UK pension schemes. The findings, as featured in our inaugural Pension Pulse report, reveal the scale of the risks and costs of inaccurate member data.
Our analysis of member data has served as a valuable reality check for these schemes, removing, in some cases, a false sense of security linked either to data scores submitted to the Regulator previously, or to a scheme having completed data cleansing exercises in the past.
As one of our data experts points out, this can never be a “once and done” project.
“Even if schemes are quite diligent, data is never going to be completely accurate,” they explain. “People move and forget to tell their pension provider; people pass away and no-one tells the pension scheme. Every time you get your data accurate it’s going to start to degrade the next day.”
It’s worth noting that whilst some of the average error percentages listed in our findings are small – for a scheme with tens or hundreds of thousands of members, they may represent significant increases in costs and risks. In addition, records containing different types of mistakes may not always overlap, suggesting that higher percentages of member records may contain one or more errors.
We analysed data linked to more than 3m individual members in nearly 70 pension schemes. Most are Defined Benefit (DB) schemes with an average membership of about 44,000. Mortality statistics were gathered through separate, client-specific work.
This will increase administrative costs and workloads for schemes and providers and could cause unnecessary stress for people trying to find lost pensions. This figure is based on the need for multiple items of data related to an individual and a pension to match in order to achieve a “match” rating and avoid a “possible match”. Combining average inaccuracy rates for surnames and addresses (see below), along with the average inaccuracy rate for National Insurance numbers shows that 18.54% of queries are likely to include at least one error.
Our data also reveals that, for these schemes, on average:
Women’s surnames are more likely to be wrong, often because someone has got married or divorced and failed to inform a pension provider or scheme. Errors may also be due to input errors. Sometimes names are mis-spelled, with one common problem being confusion around a Mac or Mc. Multiple schemes had incorrect data for more than 10% of surnames.
This average hides a huge range of accuracy/ inaccuracy. In one scheme, a remarkable 43.01% of forenames were wrong, but the next two highest levels of inaccuracy were 21.91% and 8.31%. At the other end of the scale, some clients’ data contained no incorrect forenames at all.
Again, this average hides huge variations: at one scheme more than one in three (35.94%) of DOBs were incorrect. In most schemes the figure is somewhere between 1% and 2%, while the scheme with the highest accuracy had only 0.26% of errors.
But even a few errors in percentage terms represent significant risks for a scheme with tens or hundreds of thousands of members. They may affect decisions about investment strategies, or member retirement dates; and could distort visibility of a scheme’s true liabilities, or the accuracy of benefit calculations.
The average share of pension scheme members who change their address each year is thought to be about 8% [2], but clearly members often fail to inform schemes or providers. Incorrect address data exposes schemes to risks linked to data privacy, and increases waste and costs.
Sometimes National Insurance numbers are duplicated and assigned to the wrong person. That could have serious consequences for the individuals in question and may increase reputational and regulatory risks for schemes and employers. It’s worth noting that our analysis only accounts for formatting errors and obvious temporary NI numbers, leaving 0.31% a conservative minimum.
Continuing to pay benefits to this share of pensionable members who have died would cost a scheme over £250,000 each month per 10,000 pensioners, based on an average monthly payment of £1,100. If deaths are unreported and unknown to a scheme this increases the incidence of fraud, reputational and regulatory risks, as well as financial loss. Clients who had not completed data cleanses for some time discovered worryingly high percentages of member deaths of which the scheme was unaware – in some cases more than 6% of members have died.
Scheme trustees or managers may be reluctant to pursue bereaved spouses or other family members to recover money paid into the deceased’s bank account following their death – although a failure to do so may mean the scheme is failing in its fiduciary duty to the other members.
As we move into a DC-dominated pensions landscape and more people are likely to build up pension savings in multiple pension pots during their careers, this problem will become more common and more acute, because informing every scheme or provider of a death can be difficult and time-consuming for relatives of the deceased.
Our findings also show that one in 50 scheme members below pensionable age had already died. If schemes were unaware of these deaths before the date when those members would have reached pensionable age this might expose a scheme to additional risks, such as fraud. It will also mean some beneficiaries could be owed money, arguably when they most need it.