European pension funds and institutional investors are used to substantial domestic biases in their portfolios. This is rational in risk/return terms, since all or almost all their liabilities are denominated in the domestic currency.
The creation of euroland will effectively extend the definition of 'domestic' from an asset/liability management perspective. This is a positive development. The larger 'home spectrum' will create more opportunities for diversification. And the chances of outperforming relevant benchmarks will increase automatically within a larger spectrum. This latter is true not simply for statistical reasons. It is also related to the skewness of market cap distributions over stocks in European countries.
With the UK probably the only exception, European countries are characterised by the fact that a relatively small group of companies represent 70-90% of free-float corrected market capitalisation. This causes tremendous problems when those stocks are on the 'sell' list. It can be very difficult to do more than marginally underweight those stocks compared to the relevant benchmarks. Shifting from a large cap tilt to small caps is almost impossible. A substantial part of performance problems in European country mandates and current excess volatility can be explained by this 'skewness factor'. The transition to a pan-European benchmark and equivalent mandates will definitely make life easier.
An analysis of broad-based European indices indicates that none of the European large caps dominate those indices. Individual company weights are never higher than 2-3%, making life much easier for large-scale investors. Stocks on a 'sell' list can be skipped and an optimisation with only those stocks on 'buy' or 'hold' will result in a well-behaved portfolio with an attractive risk/return profile and acceptable liquidity.
But things will not get easier automatically. Pension funds will first need to make the transition towards the new European benchmark. That transition will be tedious, complicated and full of potential pitfalls. We have constructed a model to assist pension funds with t his process. This also enabled us to do some very realistic simulations.
Important elements of the model that should be taken into account are:
q Risk control How to control the portfolio tracking error when moving from the home-biased portfolio towards the new 'euroland' portfolio.
q Transaction costs and market impact How to control transaction costs and market impact in a situation where others have to undergo comparable transitions within similar time spans, which might increase risks.
q Differential return expectations The 'normal' asset management of portfolios will continue during the transition process. Return expectations between countries might differ and a naive predetermined shift that ignores these differentials might introduce opportunity costs in the form of lost returns.
q Other costs For example, costs related to the fact that pension funds manage domestic portfolios to a larger extent internally than foreign/international portfolios. This might make it necessary to restructure the fund. These costs should be taken into account.
q Current position of the fund The extent to which transition should take place differs for each fund and each country (see also table 1). This also holds for the targeted risk profile of the fund. We also have to take into account factors such as the maximum acceptable transition time and the moment at which a new European benchmark will be introduced.
So there is a lot of work to be done, with some remarkable country differences as shown in Figures 1 and 2.
Figure 1 shows that the advent of 'euroland' will result in a substantial outflow of pension assets from the Netherlands. The well-funded domestic pension market - number four in the world - in this relatively small country faces a very difficult transition. The domestic bias is relatively large (see also Figure 2) and stock market liquidity relatively low, except for the 30 or so largest stocks. Crossings with foreign pension funds without enough Netherlands exposure in existing portfolios might mitigate the problem, but not solve it. Ireland will witness something similar. Germany and to a lesser extent France and Italy will see a substantial net inflow of assets.
Figure 2 shows that the average equity allocation to the home country is approximately 50%, but in countries such as Italy and Norway all equity exposure is in domestic stocks. Admittedly, equity allocations in many countries are still relatively low from an ALM perspective, but simply allocating more assets to Europe instead of the home country will not solve the transition problem.
It is sometimes argued that futures and options strategies and crossings through major brokerage houses can solve the transition problem. Derivatives can be important tools, but they are just that: tools. They will help pension funds gain time, but no European pension fund can abstain from transforming its portfolio to bring it more in line with new European country and sector weightings. And derivative strategies may be much more expensive than expected. We'd like to point to experiences with portfolio insurance back in 1987. Portfolio insurance, in itself a good idea, was based on heterogeneous expectations. The transition will initiate synchronous buy and sell transactions in more or less the same direction all over Europe. This will increase the probability that derivatives prices will rise dramatically or, even worse, that the market for some derivatives will cease to exist. That is exactly what happened with portfolio insurance during the October 1987 crash. Crossings through major brokerage houses are of interest, but again pension funds need to be aware of the kind of problems described above when relying on derivatives.
We used our transition model to run some realistic simulations using data for the period January 1995- October 1998. We assumed that the transition moment for our fictitious pension funds would be somewhere during this period and analysed the process from a starting domestic bias of 50% to a position without domestic bias. We preferred to use real data instead of the widely used Monte-Carlo simulations with fictitious data. Real-life financial markets will generate and process information about such a large event as the European transition on a continuous basis. Expectations about the transition were indeed incorporated in share prices during the test period. The increase in stock market volatility during this period was probably also to a substantial extent related to the event.
In our simulation we used three different pension funds ('small', 'medium' and 'large'), simulated the 'normal' day-to-day asset management process and analysed various scenarios for tracking error, introduction moment of the new benchmark and the acceptable maximum transition period. Space limitations prevent us going into detail, but the major conclusions are remarkable:
q Relationship between transition costs and tracking error The more risk-averse a pension fund and the lower the acceptability of excess tracking error, the higher the costs of the transition.
q Relationship between transition costs and the introduction moment of the new benchmark/ acceptable maximum transition period Speed is costly. Funds that plan carefully and take their time will face much lower costs than those that want to shift quickly and on an ad hoc basis. Those that take longest (two or more years) may actually gain from it. The additional costs of transition will be offset by benefits related to the shift from a home-biased portfolio towards a more broadly based, diversified international portfolio.
q Relationship between transition costs and pension fund size In basis point terms, smaller funds pay a more for transition than larger ones. Smaller funds normally hold portfolios with more small- and mid-caps. As a result they need to do more transactions than larger funds. This disadvantage is not completely offset by the advantage of having a lower market impact if the large fund transitions rationally. This advantage of the larger funds might easily be destroyed if they transition sub-optimally and run into market impact problems.
q Non-linearity of the transition process It is wrong to transition in one month, unless the pension fund is very tiny. It is also wrong to perform a 'naive transition' where a fund that wants to transition in x months simply transitions a portion of 1/x every month. This approach ignores non-linearities in the cost function (transaction costs and market impact, tracking error penalty, etc). Our analysis indicates that the bulk of the transition will have to be done during a three- to 12-month period around the moment at which the new benchmark is introduced, with gradual portfolio shifts starting earlier.
To conclude, the transition operation is a complex one whose costs can only be kept in check through timely planning and a structured optimisation approach taking into account all relevant factors. It will not be cheap for the average pension fund. Our simulations show that costs (additional transaction costs/lost portfolio returns) could be as high as DM4m (Ecu2m) for 'smaller' pension funds (European plus domestic equity stakes of up to DM1bn) on an annual basis during the transition period. For the largest pension funds those costs could be as high as DM 100m.
Erik van Dijk is chief executive officer of Palladyne Asset Management in Amsterdam