Bond managers: can they predict?
Although we have seen quite a bit of research on the usefulness of equity analyst’s buy and sell recommendations for individual shares, far less research has been done that addresses the overall asset class visions of analysts and strategists. Perhaps this can be attributed to the fact that data providers have not maintained quality databases of strategist predictions.
However, this has changed
now, because IPE has been tracking investment manager’s expectations on a monthly basis since February 1997. The IPE database consists of 120 asset managers, with more than 7,000 monthly predictions.
In the December 2005 issue of IPE we presented the results of our research concerning the equity market predictions. In this issue we analyse to what extent general market visions by bond market strategists add value.
In the December 2005 issue, we showed that equity strategists did add value, albeit modest, over a period ranging from February 1997 to August 2005. The strategists did, on average, slightly better than the MSCI indexes for the regions US, Europe, UK, Japan and Asia. We also found substantial differences between individual asset management firms. Some were quite good in predicting equity market movements, whereas others posted poor track records.
We described our research methodology quite extensively in the previous issue, but feel that it is good to give a short summary here. In IPE’s monthly Investment Manager’s Expectation Indicator (see page 10), managers offer their asset class predictions for the upcoming six to 12 months. In our methodology every monthly prediction is considered as separate for both six and 12 months; so we calculate the percentage rise (or fall) of the index for both the six and the 12-month period, and this on a monthly basis according to the following
Dj,k,t * Rk,t
j = manager
k = the asset class/region
t = time period
Rk,t = price return (over the 6 or 12 month period)
Dj,k,t = -1, 0,_1 with Dj,k,t = 0 if the manager expects no price change (neutral), Dj,k,t = -1 if the manager expects the price to drop and Dj,k,t = +1 if the manager expects the price to increase.
The overall manager score is then calculated as:
∑ Dj,k,t * Rk,t
The monthly six and 12 months scores were then averaged. Next, we ranked the n managers from 1 for the best manager to n for the worst manager based on the average results for all four bond regions defined by IPE: the US, Europe, Japan and the UK. The overall top-performers for the asset category ‘Bonds’ were the ones with the best average rankings (equally-weighted) for the four regional analyses. We went through this exercise twice: once in local currency and once in dollar returns.
Currency returns can play a substantial role in bond management. With bonds returns and volatilities being less high than their equity counterparts (in local currency terms), the relative importance of the currency factor (if not completely hedged) will be larger. In other words; when currency fluctuations move against you, your good prediction of local bond market returns can be easily devastated.
Bond managers that are not good in predicting exchange rates should hedge currency risks. On the other hand, managers that do understand exchange rate movements are well advised to use this as an extra source of return. With currency play thus being so important, we wanted to distinguish between return in dollar and local currency terms.
This is especially so because it was not absolutely clear if all managers provided IPE answers based on a purely local, market stand-alone basis or as portfolio investors from an international perspective. There was also no separate survey question asking managers if (and to what extent) they did or did not hedge foreign currency risk in the bond market.
So, to make sure that we capture ‘true quality’, three ranks per region (and for the overall average)
were derived, one in local currency, one in dollar terms, and a combined ranking. The latter was calculated by taking the equally weighted average of the local and dollar return
Via regression models, we then proceeded by trying to explain these return prediction result rankings through a framework that incorporated the following factors:
q Rank in assets under management (SIZE);
q Rank in % of equities in total of assets under management (EQ%);
qRank in % of bonds in total assets under management (%BONDS);
qRank in % of equities in US, Europe, Japan, Asia and Emerging Markets respectively (as percentage of firm assets under management: USEQ%, EUREQ%, JAPEQ%, ASIAEQ% and EMEQ%);
q Rank in % of bonds in specific bond market segments: public bonds Europe, US and Japan, corporate bonds Europe, US and Japan, other (rest of world) public and corporate bonds and money market products (as percentage of firm assets under management: %PUBEUR, %PUBUS, %PUBROW, %PUBJAP, %CORPEUR, %CORPUS, %CORPJAP, %CORPROW, %FI-OTHER and %MMCASH);
q Rank in return prediction quality for the strategic equity market views for the various equity market country/regions and for the overall average equity market prediction quality (RETEQUS, RETEQEU, RETEQJAP, RETEQASIA, RETEQUK, RETEQTOT), so as to test if there are links between prediction results in the various regions and between asset classes;
q Rank in return prediction quality for the strategic bond market views for the various bond market country/regions and for the overall average bond market total views (RETBSUS, RETBEU, RETBJAP, RETBUK, RETBTOT) so as to test if there are links between prediction results in the various regions.
We also incorporated two dummy variables, one for the country of origin of the asset manager (COUNTRY) and one for the firm’s investment style (STYLE)1. We extracted information on these variables from IPE’s July/August supplement issue on the Top 400 Asset Managers. The indices used for the bond and equity market are from MSCI. Both single-variable and multivariate regressions were used to detect patterns in the score and distinguish between good, average, and poor managers.
In our analysis of the value of equity signals by investment strategists, we, as said before, found that equity strategists did - on average - add value. This conclusion is no longer valid when looking at bonds as an asset class.
On average, ie, taking into account all managers that have contributed at least 80 months of expectations in all regions in both local and dollar returns, the managers lost slightly when looking at the average monthly return contributions. There was however a difference between local and dollar returns. In local currencies there was on average a (slight) positive score, suggesting that currencies are indeed an important determinant.
These results did not change when we added the managers that have contributed less than 80 months of predictions. The results even deteriorated a bit, suggesting that recent years have been difficult to predict. Just like it was the case for Equities, we noticed quite a bit of dispersion in results among the analysed managers. The top10 (based on a combination of local and dollar returns) are listed in table 1.
The top four from this list had a good rank in both local and dollar returns, the ranking of the others was a result of a good rank in either local or dollar returns and a (considerably) lower rank in the other. Remarkable: the table incorporates two French and two Swiss asset managers. France and Switzerland aren’t exactly the largest nations within the industry when looking at country of origin, but it confirms a more general feeling within the asset management industry. French and Swiss asset managers are good in bonds, but much less so in equities.
In the original table, in which we included asset management firms that cease to exist, Credit Lyonnais was a top10 player as well. Remarkable: AG Bisset & Co, Morgan Stanley and Robeco were also in the top10 for equities. AG Bisset is a currency overlay manager, which further corroborates our conclusion that currency views are of tremendous importance in bond investing.
In table 2 we show the results for the single-variable, or univariate regressions based on the overall total ranking in which both local currency and US dollar rankings are taken into account.
When looking at the underlying univariate regressions in local currency and dollar terms separately, we see that they differ slightly. In the overall regression (combined), we see that SIZE is a negative factor for US bonds, implying that the smaller houses tend to better predict US bonds. In dollar terms we see that SIZE is also a significant negative factor for Europe and the UK, but a positive one for Japan. This implies that bigger houses do only outperform smaller boutiques in Japanese government bond strategies (in local currency terms).
Apparently, the peculiarities of the Japanese bond market are such that only large specialists with sufficient resources are capable of outperforming their smaller competitors. This conclusion is justified by the fact that the percentage of fixed income (%FI) of total assets is a significant positive factor for Japan only, implying that an asset manager needs to concentrate on Japanese bonds in order to be able to add value with his predictions. Non-specialists do not seem to get a firm hold of this complicated market. Surprisingly, table 2 also shows that US asset managers tend to be less effective in both the US and the European bond markets.
This phenomenon could be attributed to the fact that American managers often seemed to neglect the currency factor, unlike for instance the successful French and Swiss managers who were actively playing with it. The final conclusion that can be drawn from the univariate regressions is that there seems to be a high correlation between prediction results for the US, European and UK bond markets. The Japanese bond market on the other hand seemed to represent a more segregated segment of the global market.
Do multivariate regressions in which the combined effects of the variables are tested change the picture? All the aforementioned variables were taken into account, including the ones related to the regional percentages of equities and bonds under management. The most important conclusions we can draw from the multivariate regressions were related to:
q Correlations: the high correlation between predictive qualities in various regions, with the exception of Japan, remained. In local currency returns the correlation coefficient between the European and Japanese rankings were even negative. This implies that bond managers that tend to be less good in Europe are among the best in Japan.
q Size: size is again of no importance (see table 3). In the US case it actually even hurts: the best bond managers in this market tend to be relatively smaller players.
q Style: being a structured manager was only significant in the US; with style being irrelevant in most other markets. There is more than one route that leads to good performance. The more fundamental managers did a better job in Japan. This is yet another indication of the Japanese bonds market clearly being a different story.
q Percentage of bonds under management: this factor is once again only important in Japan; you need to be a Japanese bond specialist in order to be good. Apparently, you need to be positioned within the right information networks within the land of the rising sun. Another interesting phenomenon is the negative correlation between the bond and the money market/cash segment in Europe. If a manager’s percentage of money market funds/cash under management (%CASHMM) is high, the average performance in the European bond market tends to slip. At the same time the correlation between Euro-zone, UK and US bond markets remains high, which implies that analysis of international long-term interest rate and yield curve differentials is more important for European bond managers than thorough yield curve play which focuses mainly on analysis of the European region itself. Once again, we have yet another indication of the Japanese markets being truly different. When a manager specialises in European bonds (high %PUBEUR), this will, on average, hurt his predictive performance in the Japanese government bond market.
q Percentage of equities under management: The equity factors do not play a big role, suggesting that bond managers can do a good job without looking at or managing other the other main asset category. In our previous contribution in the December 2005 issue of IPE, we saw that this was quite different when analysing equities where too much specialisation on equities did not add to a manager’s top-down macroeconomic qualities. For the UK excess specialisation on equities did even seem to hurt (in dollar returns). It appears that equity managers need the broader macro experience of other asset classes, whereas bond managers do not necessarily need the equity market experience. In exchange, their skills in the currency market have to be more developed than those that equity managers would need. Emerging market equity and Asian equity specialists are really another ball game: managers who are excelling in these areas are almost without exception (except maybe for Robeco) not as good in the bond area.
The explanatory power of our regression models was quite good (see table 4). Only for Japan (the dollar-denominated regression) did our model post a relatively poor result. This confirms our earlier finding that the Japanese market should be treated differently. We probably need to incorporate other variables, not tested here, to really capture what is going on in the Japanese bond market.
In summary, using the extensive IPE Managers’ Expectations database, our research indicates that bond strategists did, on average, add less value over the period 1997-present than the equity strategists did. Bond markets were either more efficient or more difficult to predict than equity markets.
There is, however, a clear distinction between good and bad bond managers. The average manager might not add value, but there is clearly a super league of managers that do, and our univariate and multivariate regressions can help in finding them. The French and Swiss do a good job as bond managers. Another interesting conclusion is that firm size is not really an important factor when it comes to manager selection. And if it is, the indications are more in favour of boutique managers than against them. In our equities study, it did hurt to be big in Japan or the UK. In the bonds case this seemed to hold for the US market only.
In order to score in the Japanese bond market you really need to be a true Japanese bond specialist. Taking into account currency effects it seems justified to conclude that the Japan specialist needs to be a local, Japan-based player as well.
Generally, currency effects can make a difference in most bond regions. There are some good local players that ‘think’ in local currency terms, but when faced with an international clientele they can do quite badly. It is therefore not surprising that this holds true for the two largest markets (with the highest likelihood of local players catering to a local clientele of institutional investors), namely the US and Japan. The excellent performance of US-based currency overlay specialist AG Bisset is therefore, not surprising.
In the third article in a future issue, we will have a closer look at the usefulness of currency predictions by strategists working in the asset management industry.
1The transformation into ranking variables helped us solve problems related to the non-normal distribution of some variables. This was especially a problem for the skewed assets under management variables. It did not apply to the dummy variables for country (US = 1 and Others = 0) and style (Quantitative/Structured = 1, fundamental/macro = 0, mixed = 0,5).
Erik L van Dijk is CIO and partner and Harry J Geels is senior investment manager at Compendeon based in Bunnik in the Netherlands.