Predictors for bond allocation

Related Categories

Empirical analysis helps to identify variables that historically have had persistent ability to recognise attractive investment opportunities. We have developed a set of global asset allocation indicators that can be used to allocate funds across several bond markets or currencies and to market-time individual markets. These predictor variables attempt to identify value and momentum in markets. They are financial market variables that capture the macroeconomic and monetary policy environment in addition to the time variation in required risk premiums and market sentiment.
Four indicators are used to identify the most and least attractive bond markets on a self-financed (currency-hedged) basis. These include curve steepness, real yield, stock market weakness and recent bond yield trend. Our bond market allocation strategy involves each month buying bonds in an attractive country against selling them, one for one in an unattractive country (for example, buying bonds in the market with the steepest yield curve and selling bonds in the market with the flattest or most inverted yield curve). For the purposes of global asset allocation, the post-Emu global government bond market is proxied by the seven and 10-year sectors of Salomon Smith Barney country indices for the US, Canada, Japan, Germany, UK and Sweden.
Value indicators for the currency markets include short-term interest rate and 10-year forward foreign exchange rate relative to its historic average. Trending is more pronounced in currency markets than bond markets and the previous quarter’s currency return and last month’s change in bond yields can be used as predictors that capture momentum effects. The currency allocation strategies are quite similar to the bond market strategies involving the same six countries.
Out-of-sample historical tests including subperiod analysis show that individual strategies based on the best predictor variables are profitable 60% of the time, on average. This small advantage can be magnified substantially if the strategies based on these predictors are combined together in a diversified portfolio of active strategies.
There are many ways of combining information contained in several predictors into a single trading strategy and several ways of combining individual trading strategies into an active portfolio. Our approach has been to adopt simplicity as the guiding rule for the sake of transparency and intuition.
We use “total score” to summarise the near-term outlook of each bond market based on all four indicators. Each market is given a rank score of between one and six, depending on the attractiveness of the market based on each predictor. The average of each markets’ four rank scores is its total score.
Figure 1 shows the historical performance over the past 12 years of three strategies with different degrees of diversification. Note that, as all trades are self-financed, the profits in Figure 1 are “pure” (they could be earned in excess of income earned on invested cash). Strategy 3 – our flagship strategy – involves four trades (equally weighted): buying the two bond markets and the two currencies at the top of the total score rankings against selling the bottom two bond markets and the bottom two currencies. Other weighting schemes based on volatility or signal strength are of course also possible.
All the strategies produced consistent profits over time. Combining the information in four predictive indicators (that is, using total scores) improved the results. The superior performance of the last strategy partly reflects the higher average returns – but also volatilities – of the currency allocation strategies; yet diversification gains are the main reason for this strategy’s smoother cumulative profit line and higher frequency of profitable months (rising from near 60% to almost 75%).
Both leveraged traders and real-money investors can use the investment recommendations based on our indicators. Hedge funds and other leveraged traders can simply do the recommended trades, buying and selling the recommended bonds and currencies. Real-money managers who are benchmarked against a global government bond benchmark, like the Salomon Smith Barney World Government Bond Index, can underweight or overweight the six markets based on the recommendation of the model. This deviation from the benchmark will of course lead to a tracking error. The degree of permissible tracking error will therefore dictate the exposure the investor may take based on the model's recommendations.
The “total score” Strategy 3 described above has exhibited a rolling annualised volatility of 4–5.5% over the past 10 years. If a tracking error of 1% is acceptable then a 20% exposure to the model’s recommendations may be taken.
Figure 2 shows the results of a simulation of an overlay strategy. The passive index strategy is to be long the Salomon Smith Barney World Government Bond Index unhedged in euro terms. The strategy involves overweighting the two most attractive bond markets and the two most attractive currencies and underweighting the two least attractive markets and currencies based on total score. A 20% exposure means a 5% exposure to each of the four trades. Such an overlay strategy would have outperformed the benchmark on an annualised basis by 154 basis points with a tracking error of 0.93%.
Although it is difficult to predict that future results will be as good as the past, and the strategies were challenged in the high risk aversion environment of 1998–99, we are trying to enhance our asset allocation tools in several ways.
We are looking into better ways of obtaining return forecasts. We are also investigating better ways of constructing individual trading strategies – exploring interaction effects appears promising. Analysing different ways of combining individual strategies into an active portfolio could be another avenue for further research.
Antti Ilmanen is director, European fixed income strategy and Rafey Sayood is director, bond portfolio analysis group, at Salomon Smith Barney in London. They discuss their global asset allocation indicators in greater detail in articles available to SSB customers on:

Have your say

You must sign in to make a comment


Your first step in manager selection...

IPE Quest is a manager search facility that connects institutional investors and asset managers.

  • QN-2467

    Asset class: Search for a broker (mainly ETFs).
    Asset region: Global.
    Size: 250m.
    Closing date: 2018-08-28.

  • DS-2468

    Closing date: 2018-08-24.

  • QN-2469

    Asset class: All/Large Cap Equities.
    Asset region: Global Emerging Markets.
    Size: USD 500m.
    Closing date: 2018-09-04.

Begin Your Search Here