In Focus: Sector Tilts

As you may remember, when we rebalanced our recommended portfolios at the end of last year, we used a fairly broad definition of an asset class. Specifically, because the benefit from diversification comes from risk reduction, we required that the "asset classes" we used could have no more than a .60 correlation of returns with each other. That definition eliminated from use a number of groupings of stocks and bonds that other commentators call "asset classes." Examples of these include small cap stocks or large cap growth stocks, and short-term bonds. In our view, all of these represent various "tilts" that one can make in order to enhance the risk/return trade-off within an asset class. At the time of our rebalancing, we promised that we would be taking a closer look at these "tilts" to see which, if any of them, made sense. Last month we looked at tilts based on market capitalization and growth vs. value. This month we look at sector investing. Next month we’ll look at country investing, in August we’ll look at investing in different bond maturities, and in September we’ll look at momentum investing. As was the case last month, the fundamental question we’re trying to answer is whether or not you can improve on the risk/return trade-off for the asset class as a whole by making a sector tilt in your portfolio.

To answer this question, we ran analyses using both the Dow Jones U.S. Sector Indexes and the Dow Jones Global Market Sector Indexes. We chose the latter over the new FTSE indexes because they included companies from developing as well as developed countries, and therefore seemed to be more representative of the potential benefits of sector tilts. Our data set for both analyses covered the period from January, 1992 to December, 2000.

Let’s look first at the U.S. results, and start with a description of the indexes we used. The Dow Jones U.S. Sector Indexes are based on the Dow Jones U.S. Total Market Index, which includes companies that comprise 95% of U.S. equity market capitalization. The indexes divide the total market into ten sectors: Basic Materials, Consumer Cyclicals, Consumer Non-Cyclicals (called staples by others), Energy, Financials, Health Care, Industrials, Technology, Telecommunications, and Utilities. At the end of 2000, The three biggest sectors (in terms of their market capitalization as a percentage of the total market index) were Technology (23.4%), Financial Services (17.3%), and Health Care (14.0). In the period covered by our data, the results for various sector indexes varied widely. At the low end of the distribution, U.S. Basic Materials delivered average annual returns of only 9.40%, with a standard deviation of 21.12%, or only .445% of return per unit of risk. At the other end of the spectrum, Financials generated average annual returns of 22.71%, with a standard deviation of 22.34%, or 1.017% of return per unit of risk. In terms of highest annual returns, the Technology Sector was the leader at 30.54%. However, this came at a price – the standard deviation of those returns was 35.99%, for an all in result of only .849% of return per unit of risk taken on. By way of comparison, the overall Dow Jones U.S. Total Market Index delivered average annual returns of 16.52% during the 1/92 to 12/00 period, with a standard deviation of 15.45%, or a very respectable 1.069% of return per unit of risk. Obviously, the fact that the overall index was diversified across all the sectors accounted for the fact that its risk adjusted performance was better than that of any single sector.

However, the extent to which the overall index was diversified varied over time throughout the year, and was directly related to the relative underlying performance of the different sectors, as evidenced by the still heavy weighting of the Technology Sector at the end of last year. The question we must ask ourselves then, is whether or not a fixed weighting of different sectors would have delivered a superior risk adjusted performance.

The first step for testing this was a look at the extent to which the returns on different sector indexes were correlated with each other during the period covered by our data. The results were very encouraging, with a large number of very low correlations. Here are a few examples: Basic Materials/Health Care = .17; Basic Materials/Utilities = .10; Consumer Cyclicals/Utilities = .07; Energy/Telecomms = .19; Energy/Health Care = .23; Utilities/Technology = (.16).

Using these inputs, we used our optimization software to construct two portfolios. The goal of the first was to exceed the total market portfolio’s average annual return, while matching its 15.45% standard deviation. The goal of the second was to match its 16.52% return with a lower standard deviation. In both cases, we set the further limit that no single sector index could account for more than 20% of our portfolio.

The results were impressive. For the 15.45% target standard deviation portfolio, we were able to achieve expected annual returns of 20.58% versus the total market portfolio’s 16.52%, or 1.332% of return per unit of risk. For the 16.52% target return portfolio, we were able to reduce standard deviation to 12.37%, or 1.335% of return per unit of risk taken on. In the case of the former, our result was achieved with a mix of 20% in each of Energy, Financials, Health Care, and Technology, 8.5% in Industrials, and 11.5% in Utilities. In the second case, our portfolio included a mix of 20% each in Consumer Cyclicals, Energy, Health Care and Utilities, along with 5% in Consumer Non-Cyclicals, 7% in Technology, and 8% in Telecomms.

We achieved similar results in our global experiment. In this case, the index against which we benchmarked ourselves was the Dow Jones World Index, which covers 34 developed and emerging markets. Over the 1/92 to 12/00 period, this index had delivered average annual returns (in U.S. dollars) of 11.62%, with a standard deviation of 14.36%, or .809% of return per unit of risk. By comparison, our 14.36% target standard deviation portfolio had expected annual returns of 16.84% (1.173% of return per unit of risk), while our 11.62% target return portfolio had a standard deviation of 11.09% (1.048% of return per unit of risk). The allocation of the former was 20% each to Energy, Health Care, Technology, and Telecommunications, 8% to Financials, and 12% to Utilities. For the latter portfolio, the allocations were 20% each to Consumer Cyclicals and Non-Cyclicals, Health Care and Utilities, 8% to Energy, and 6% each to Basic Materials and Telecomms.

We repeated this analysis five more times, using the perspectives of investors whose functional currencies were Australian Dollars, Canadian Dollars, Euros, Yen, and U.K. Pounds. In each case, we were able to achieve similar improvements versus the performance of the total market index, though with somewhat different sector allocations. Obviously, something was going on in the past. This raised the next two logical questions: what accounted for the performance improvements delivered by the sector portfolios, and would these causal factors continue to operate in the future?

We believe that the superior performance of the sector fund portfolios is due to two factors. First, they make heavy allocations to sectors whose returns have very low correlations. More importantly, it would seem that a substantial portion of these low correlations is due to a low correlation between the fundamental demographic and economic factors that affect growth and profitability in the respective sectors, and is not just a historical market phenomenon that is unlikely to repeat itself in the future.

Second, the portfolio allocations to different sectors are stable over time. In contrast, the portfolio allocations in the Total Market Index change dynamically as the sectors’ relative market capitalizations change. At the margin, this means that the total market index portfolio will tend to overreact in ways that hurt performance – overinvesting in sectors that have become overvalued, and underinvesting in sectors that have become undervalued. However, lest this be construed as an argument against indexing, we should also point out that, by definition, the total market index fund will consistently (and that is important) correct these mistakes faster than 50% of active managers.

The crux of the issue then, comes down to this. If you believe that (a) the correlations between some sectors will continue to remain low in the future, and (b) the risk/return characteristics of those sectors will remain similar to what they have been in the past, then you should consider building a portfolio around a carefully chosen mix of sector indexes, rather than buying a total market index fund. On the other hand, if you don’t believe that these two conditions will hold true in the future, then you are better off going the total market index route which, despite its shortcomings, will still over time deliver returns that are superior to all but a very few actively managed funds that are impossible to identify in advance.

| Performance Update | In Focus: Sector Tilts | Fund Performance and Fees |