Swiss Pension Fund Aggregate Indices Performance Monitor

Credit Suisse Pension Fund Index & UBS Pension Fund Index Family

From 31.12.1999 onward, Credit Suisse calculate the first pension fund aggregate index, based on some 100 Swiss pension funds to whom Credit Suisse acts as Global Custodian. While the index has monthly increments, it is only published quarterly.  With an inception date of 31.12.2005, UBS Group began to publish a family of pension fund indices, also based on custody. In contrast to the Credit Suisse counterpart, UBS publish their indices monthly. UBS distinguish pension fund indices by size: small, medium and large pension funds, plus an index reflecting all sizes. The performance of Swiss pension fund aggregate indices is compared to the a medium risk-tier synthetic pension fund benchmark index LPP-C40.

Taking Advantage Of Monthly Performance Updates

The calculation of many quantitative and qualitative metrics requires large data data samples to generate meaningful results. Unlike the data for specific pension funds, any of the aforementioned aggregate indices is available with monthly frequency, thus permitting a depth of analysis and comparisons not possible with quarterly, or annual data. As individual pension funds show a very high degree of congruency with the aggregate indices used here, the analysis of monthly data for the aggregates can be taken as advance insight into pension funds investment performance released much later.

Exhibit-1

As statistics going back to the end of 1999 are only available for the Credit Suisse Pension Fund Index, Exhibit-1 (above) depicts annualised returns for this aggregate alone, alongside that of the benchmark LPP-C40. All returns are calculated to the most recent date. The striking piece of information emerging from this simple, initial glance is that pension funds in general have failed to match, even less exceed, returns from an unselected representation of the investment environment over just about any time horizon that could be considered meaningful.

The table of Exhibit-2 (below) illustrates the evidence from Exhibit-1 in numerical form and adds additional data for the UBS family of Pension Fund Aggregate Indices for ten, five, and three years as well as trailing 12 months and year-to-date performance. Exhibit-2 uses original index values, as published by Credit Suisse and UBS. Please note that the two index groups have different base values: CSPF Index = 100 at 31.12.1999, UBS Indices = 100 at 31.12.2005.

Exhibit-2
Exhibit-3

Exhibit-3 (above) illustrates annual performances back to 2008 together with YTD performance, for the benchmark index, CSPF Index and the UBS PF Index (all sizes). This graph reveals a peculiar pattern: Pension funds tend to outperform the benchmark in bear markets, eroding less, but underperform bull markets, gaining less when financial markets are on the increase. The simplified interpretation of this pattern (which is echoed by the data of most individual pension funds) is that of a continuously ‚underinvested’ stance: Either pension funds hold too many cash reserves regardless of outlook, or constantly hedge their market exposure, suffering the drag of additional costs, even when counter-productive. The pattern certainly merits more detailed analysis. After all, as market tend to rise more often than not, this will most likely be related to the fundamental cause behind lacklustre long-term pension fund returns.

Exhibit-4 (below) looks at patterns of relative performance across the 20 most recent three month periods (5 years). Shown is the net difference in return to the benchmark. To the right of the bar charts, a numerical summary calculates frequencies of performance over or under the benchmark. It also shows the average positive and negative performance differential. If investment decisions were to be taken by means of ‚coin-flipping’, one would expect a 50:50 pattern to emerge, with equal values for average positive and negative performances. Clearly, this is NOT the case here: the data suggests a bias towards underperforming more frequently than mere chance would explain. Worse, the split-analysis of the differences shows that underperformance is not only more frequent, but also more pronounced. This pattern is a formula for certain disastrous long term returns.

Exhibit-4
Exhibit-5

Taking a step back and in order to look at a broader canvas, Exhibit 5 (above) shows the benchmark together with four aggregate pension fund benchmarks, rebased to 120 months ago. The chart is a stark reminder how the negative pattern described above accrues meaningful value differences as time progresses.

Exhibit-6 (below) emphasises the relative performance of pension fund aggregates to the benchmark.  Of all exhibits shown so far, Exhibit-6 best describes the second meaningful pattern arising from this analysis: Large pension funds appear to have noticeably better performance relative to their smaller peers. While most industry observes will argue that this is explained by a cost advantage that large pension funds have over small ones, I suspect that the reason behind this discrepancy are more complex.

Exhibit-6
Exhibit-7

In Exhibit-7 (above) trailing annualised returns over 36 months is shown, comparing pension fund aggregates with LPP-C40. During the 10 years of history covered by this chart, no pension fund aggregate’s trailing return has ever exceeded that of the benchmark, except when the benchmark’s trail was travelling in negative territory.

Matching the approach behind the previous chart, Exhibit-8 (below) depicts ex-post risk, also calculated over trailing 36 months. As much as returns of pension fund aggregates are below the benchmark, as much their risk is also noticeably lower, confirming an observation made in the comments on Exhibit-3. Superficially, this is an argument in favour of pension funds mission to avoid risks. However, another aspect of the analysis will give an altogether different angle of interpretation. A general caution regarding this, or any other ex-post risk metric. Ex-post risk is a contrarian indicator. Very low readings should be viewed as a warning, as risk will likely rise, and vice versa. For this reason, Exhibit-8 shows low risk at the upper end of the graph.

Exhibit-8
Exhibit-9

In Exhibit-9 (above), the values for trailing risk have been deducted from the values for trailing returns, showing the net result, or the risk-adjusted rate of return. In this metric historic comparisons and cross-comparisons at identical points in time matter most. Between any two identical rates of return, the one achieved with less risk is of higher worth.

A numerical approach was taken in Exhibit-10 (below) to emphasise the data from the previous chart. In addition to delineating the adjusted rate of return, Exhibit-10 shows supplemental statistics to give a more comprehensive profile of risk and return. Attention is drawn to ‚Horizontal Distortion’ which indicates by how much monthly values differ from ‚normal distribution’. Large distortions render the concept of ‚volatility’ obsolete, a metric which is wrongly used as a measure of risk.

Exhibit-10
Exhibit-11

Considering that pension funds are under a legal obligation to seek minimal risks in their investment strategies, it is more than odd, that they have collectively opted to invest extremely closely to their/a performance benchmark, or even to outright clone it. This philosophy is entirely incompatible with risk management. The degree to which this hazardous pseudo-strategy is being implemented is revealed by measures of congruency.  R-squared is the statistical measure of dependence from an independent variable, here LPP-C40.  Exhibit-11 (above) shows scatters of the most recent 120 months of paired returns of the Credit Suisse and the UBS Pension Fund Indices. Both show an extremely high degree of dependence to LPP-C40, with the UBS Index at 97.2% being even higher than the Credit Suisse Index at 96.6%.

Put differently: the negative performance differentials illustrated in previous exhibits are generated with merely 3% of ‘discretion’ exercised in deviating from the benchmark. That suggests that whatever genuine investment decisions are being taken, must indeed have catastrophic outcomes. At this point readers are reminded of the comments made on pension funds seeming lower risk. It is conceivable that the true cause of seemingly lower risk readings  is that pensions funds raise cash after having suffered losses After all, over the long run, the lower risk benchmark index has meaningfully outperformed the higher risk benchmark index. For details of benchmark performance, click here.

Exhibit-12 (below) shows the intercept (Alpha) and slope (Beta) of the regression line, and the degree of dependence (R-Squared) of all pension fund aggregates from LPP-C40. While all but small pension funds are able to generate a mildly positive ‘alpha’, the corresponding sensitivity to moves in the benchmark are so low, that the cumulative differences in performance remains quite negative.

Exhibit-12
Exhibit-13

Exhibit-13 (above) shows the failure of Swiss pension funds to create value from their investments in an almost brutal clarity. It list, across a variety of time horizons the ‚success ratios of performance better than LPP-C40. Readings below 50% (equal to coin-flipping) are shown in red, those above 50% in green. Currently, there is not a single ‚green’ figure in that table.

Conclusion

Benchmark indices (such as LPP-C40) are a very crude and rather simplistic reference to sum up developments in a given investment environment. Nothing more, nothing less. Such benchmarks are meant as a representation of choices but the are NOT reflective of investment selection of any kind. Benchmarks require absolute no genuine investment expertise and are, by definition, free from any attempt to create investment ‘value’ in the proper sense of the term. Unsurprisingly, benchmarks are fairly easily replicated, requiring no more than basic skills in arithmetic and statistical methods.

Yet, Swiss pension funds, as represented by the various aggregate indices used in this analysis collectively fail to match, even less exceed such a value-free measure. If nothing else, this must be taken as evidence for structural and systemic deficiencies.

Data Sources

Raw data underlying all calculations and illustrations was sourced from:

Download Exhibits

Contact Author