Guest blog from Rahul Agarwal

The problem we are trying to solve

We start with the aim of creating universally applicable metrics which help understand the value people have got for their money and the likelihood of future value for money.

The dashboard above refers.

We also produce reports for customers who want to know how their workplace pensions have done. The dashboard below refers

Understandably, those who commission AgeWage reports often design their own default. They feel their default is most  appropriate for their membership and should have its own benchmark– such conviction is admirable but a common means of expressing value for money requires a single benchmark.

Allowing everyone to benchmark against their strategic asset allocation means everyone can say they won, which is what happens in most value for money assessments. This is the problem of many forward looking measures of value for money, they are not grounded in the evidence from the past.

We look at internal rates of return (IRR) achieved by savers and compare them with the rate of return the same contribution set achieved invested in the Morningstar UK Pension index. This says a lot about the “value” groups of savers got for their “money” invested but can it tell us about the suitability of the default -its future capacity to deliver value for money?

The same can be said for dynamically de-risking defaults, either target-dated or life-styled. Can we  use IRRs for all savers to understand how well they have de-risked?  Can we provide an assessment of value for the risk savers have experienced within such funds? Can this measure be helpful in assessing the likelihood of getting VFM in the future?

Rather than create multiple benchmarks, we have started thinking about risk in terms of the outcomes of investing in a particular default relative to the outcomes of investing in an industry standard way.

If we think of this conceptually, the IRR embeds the risks actually experienced by the fund into which contributions were made. In other words, it is the saver’s risk-experienced performance. In order to produce the score, we compare this IRR with the distribution of other IRRs which might have been experienced by the saver. These comparator IRRs consider all of the risks which crystallized during the period, including those experienced by the member. The score does not consider risks which might have occurred but did not.

By examining the outcomes one by one, we should be able to build a picture of the individual risk experience of a group of savers. Comparing this collective experience against the risks experienced in the benchmarks by all savers, we might be able to produce an analysis of suitability of the default to the particular risks experienced by the group of savers under analysis.

We have started to conduct attribution analysis’ which will first figure out the level of risk providers have taken.  Once this has been achieved ,we can then measure the realized deviation between the Strategic Asset Allocation (SAA) implemented in the default, and that created in the AgeWage Index.

We are challenging ourselves to do this work without knowledge of the funds we are analysing. We are letting the data doing the talking. This approach eliminates prejudices we may have about named funds, ABI categories or of the fund’s asset allocation. We want to test whether a methodology can be created that works universally and does not have to be refined for each fund or dynamic fund matrix.

The Methodology we employ to solve the problem

We start with a set of savers’ data with only the contribution history and Net Asset Value (NAV)

Step 1: Grouping savers into different cohorts

1A: We group individual savers into cohorts based on contributions and returns; individuals with extreme returns will be grouped separately into an “outlier” cohort

1B: Savers with life styling may get grouped together or dropped into the outlier bucket or (if age is given) eliminated from the cohort  (if they get grouped wrongly, this will be identified in Step 2)

Step 2: Creating a price index for each group

2A: We create a price tracker using the contribution history of savers within a group. The price tracker will  have no values on dates where there we see no contributions

2B: We compare the price tracker with the AgeWage Index. The depth of data set will determine if additional calculations are needed to generate more data points to perform a regression analysis

2C: In the calculations, we may find that people who have been de-risked (lifestyled) will generate abnormal index prices in early stages which will help us eliminate and potentially group those people separately (if they are falsely identified in Step 1)

Step 3: Performance Attribution

3A: Using the price tracker for the cohort, we calculate beta comparing it to the AgeWage return index and perform a regression analysis

3C: The co-efficient of regression will give us an indication of the level of risk savers have taken, and the drag will indicate the realised deviation from Strategic Asset Allocation decision (SAA), compared with that made in AgeWage Index

3C: Each cohort will get a separate attribution analysis

Limitations

 This methodology is meant for fiduciaries and providers to analyse the performance of their scheme in a detailed manner. It includes approximations and assumptions for lack of detailed data such as age, funds etc. but it gives a reasonable estimation of overall performance based on risk levels. A more accurate analysis can be performed if greater depth of data is available.

Rahul Agarwal   AgeWage COO  – July 2020