Why would one pursue an active allocation investment policy within a fixed income portfolio? The overall risk of the fixed income portfolio is low compared to the other traditional asset classes. Even the most risky fixed income asset classes like Emerging Markets and High Yield have a volatility which is only half of that of equities. From an equity perspective, volatilities of all fixed income asset classes are more or less equal. And given that all fixed income classes, regardless of their specific characteristics, are exposed to general interest rate movements, it is not surprising that perceived correlations are high. So why bother?

Looking at correlations of bonds within a specific fixed income asset class this conclusion could be valid. Fixed income markets have developed rapidly especially in Europe since the introduction of the Euro. Derivatives and information technology have greatly increased the efficiency of the EMU government bond and Euro credit market. At the same time new markets like Euro High Yield, Inflation Linked and Converging Europe have emerged as new investment opportunities.
Analyzing the return patterns of all the sub-asset classes suggests however that Euro fixed income markets are not that similar at all. Various fixed income markets do react differently to changes in the economic environment. Return distributions are not stable over time and thus correlations are not. These time-varying correlations pose investment opportunities by tactically allocating assets to those asset classes that promise the highest risk adjusted returns. But then again, are the return differentials large enough to create alpha in the portfolio?
We have chosen a simulation approach to answer that question. Not that performance attribution cannot provide a convincing analysis, but results can depend heavily on the model chosen and classifications used. Furthermore, it can be hard to pin-point the nature of a trade. The addition of a specific corporate bond to a portfolio implicitly has a duration impact, is also an allocation to credits and selection of this specific issuer. To overcome this we have chosen the route of a model and classification independent analysis.
We have collected monthly excess returns versus EMU government bonds of indices of the following fixed income asset classes: Euro Inflation Linked, Euro Credits, Converging Europe (in local currency), Emerging Markets and Global High Yield, the latter two hedged to Euro. Excess returns versus EMU government bonds are used since this is the natural opposite for an asset allocation decision for a Euro-benchmarked investor. We assume that the portfolio has a strategic holding in each asset class. The period covers the start of EMU at 1st January 1999 till 31st December 2005. The following allocation strategy is applied: if over the next month an asset class outperforms EMU government bonds, an overweight position is implemented. This implies, however, that the investment manager always chooses the “correct” (e.g. overweight when an asset class outperforms and underweight when underperforming) allocation. To reflect the imperfections of the investment manager, the “correct” allocation is modified in such a way that its frequency equals the Information Coefficient or skill of the manager. We have chosen 0.55 as the Information Coefficient of the investment manager. The strategy is repeated month after month throughout the whole observation period and rerun for a 1000 cycles to achieve statistically stable results. The allowed allocations are typical for the size of the over or underweight positions in those markets in a normal Euro Aggregate portfolio context. It does not necessarily mean that each allocation decision consumes ex-post an equal part of the risk budget. As we use index returns all results with respect to returns and risk are due to the allocation strategy and do not include any selection effects, duration effects nor costs.
Table 1 shows the allocation, the achieved annual excess return, ex-post tracking error and information ratio when the strategy is applied to an asset class in isolation and for all asset classes together.
We conclude that an active allocation policy only applied to the less risky asset classes of Inflation Linked and Credits, adds about 10 basis points annually. Allowing for investments in more risky asset classes increases the excess return by an additional 26 basis points to about 36 basis points annually. Moreover it greatly enhances the efficiency of the portfolio. To illustrate this, we have plotted (see graph 1) the distribution of excess returns, when allocating in isolation to an asset class and when all asset classes are part of the investment universe.
As can be seen the distributions of excess returns, when allocating only to the individual asset classes, are concentrated, except for Converging Europe. The excess return distribution in the broad portfolio is centered much more to the right. Although its variance is somewhat larger, the probability of generating an excess return below zero is only 8%1. The probability of having negative excess returns when allocating to just one asset class is +/-25%, regardless of the asset class. These diversification benefits cause the information ratio to increase from roughly 0.25 for each asset class separately to 0.54 when all asset classes are allowed.
In the above simulation, we have assumed that investment decisions are made every month and positioning is adjusted accordingly. Although this may seem very active for an asset allocation policy, graph 2 indicates that it pays off to actively manage the portfolio. Again, the Information Coefficient is set at 0.55 and allocations to all asset classes are allowed.
If the investment decisions are made less frequent, the excess return declines rapidly. When the forecast period is one month, the excess return is 36 basis points. Increasing this period to about once a quarter, the excess return declines to only 20 basis points. As the Tracking Error is stable (since the allocations do not change) the Information Ratio declines with the excess return. If decisions are being made only once a year, the Information Ratio is just one third of the Information Ratio of the monthly investment cycle. This phenomenon has a statistical basis. As the number of investment decisions decreases, the variance around the expected Information Coefficient increases, e.g. the number of poor investment decisions increases (with an Information Coefficient less than 0.5) and thus the final excess return decreases.
So far we have assumed the investment manager is right in 55% out of one hundred decisions. Improving your skill dramatically improves the investment results. Table 2 shows the Information Ratio in relation to the Information Coefficient and the Forecast Period.
If the investment cycle is one month, an increase of only 5% of the Information Coefficient from 0.55 to 0.60, results in an increase the Information Ratio from 0.54 to 1.04! Furthermore a manager with roughly a quarterly investment cycle has to have 5% more skill to be as efficient as a manager with skill of 0.55 who makes monthly decisions. An even less active manager, with a 6-9 month cycle, needs another 5% of additional skill to be equally efficient. Reversing the argument: if an investment manager does not have any skill (Information Coefficient < 0.5) he’d better be very passive.
We realize that a simulation study cannot replace actual investment results. By recognizing that an investment manager has limited, but positive skills, combined with realistic portfolio positioning we have shown that fixed income tactical asset allocation does add value, both in terms of excess returns and in enhancing the portfolio efficiency.
In doing so we have illustrated the value of the fundamental laws of active management:
q Expanding the investment universe results in higher (although not unlimited), but most importantly in more stable excess returns;
q If an investment manager has skill, it pays off to be active.
Lombard Odier Darier Hentsch has a long track record in fixed income asset management. Fixed income tactical asset allocation is an integrated part of the investment tools used in our discretionary, tailor-made mandates as well as in funds like LODH EMS Plus Rentefonds and LODH Invest The European Bond Fund (listed in Luxembourg).
1 Given the assumptions made and data used in this study
Niels de Visser is senior portfolio manager at Lombard Odier Darier Hentsch & Cie (Nederland) N.V. and responsible for the asset allocation policy in fixed income briefs