Benchmarking vs portfolio diversification

October 18, 2016

Benchmarking can encourage sub-optimal behavior in the medium to long term. This is particularly topical given just 4 stocks represent approximately 40% of year-to-date (mid-October 2016) Asia ex Japan equity performance. Since these 4 are all large-cap technology “growth” stocks, this concentration of performance raises risks for benchmark performance going forward.


Benchmarking vs. portfolio diversification

It is common for investment product performance to be measured relative to some form of appropriate benchmark. Whether it is a global bond product measured against an index of bonds, a value manager measured against a value-tilted index, or some other related comparison, investors typically want to measure whether that product is doing better or worse than a basket of similar assets. This of course makes a great deal of sense – to limit the ability of the product to “free ride” on performance of the market. However, it is an approach that is not without its limitations, and some of those limitations have been clearly evident in Asia this year.

Generally speaking, portfolio diversification is considered to be a good thing. Subject to constraints on efficient diversification (for example, the limitations imposed by transaction costs), investors will avoid placing all eggs in one basket. However, if a benchmark index for the product is heavily weighted towards a small number of assets, and a manager is incentivised to control risk relative to the benchmark, this can encourage the manager to reduce portfolio diversity in the interests of anchoring performance around the benchmark.

I began my career in markets in New Zealand in the mid-1990’s. At that time Telecom NZ was more than a third of the main NZ equity benchmark. So a manager needed to be either very “aggressive” in some senses to hold a highly diversified portfolio, or hold around one third of their assets in a single stock. In those two cases which is the more risky? The highly diversified portfolio or the portfolio with 1/3 in Telecom NZ? The answer depends on who is being asked and factors such as the investment horizon of the investor. Arguably, the more diversified portfolio would better suit a longer term investor (who may be less benchmark-focused) and the more concentrated portfolio may better suit an investor looking for a shorter-term index- “plus-a-bit” return.

Importantly, these considerations are particularly critical in Asia today. This is because roughly 40% of the performance of the MSCI Asia ex Japan index for the calendar year to date (as at mid-October 2016) is attributable to just four stocks – four large cap technology stocks: Alibaba, Tencent, TSMC and Samsung Electronics. By weighting we estimate these represent approximately 17% of the MSCI Asia ex Japan Index. Contrasting this with the rest of the region, the median stock performance over the same time period was closer to 9% (compared with the MSCI index up 14%).

Several points can be made about this. Firstly, it indicates that average stock performance has not been as strong as the benchmark index performance would at first glance imply. Secondly, aggregate market performance represents a concentrated momentum-biased return. This is somewhat concerning given price momentum, in all its shapes and forms, has historically been a poor factor to back in Asia for anything other than select isolated periods. Thirdly, and returning to my original point, it suggests that investors were potentially rewarded (penalised) for more concentrated (diversified) portfolios.

Fourthly, these four stocks, on average, are high PE stocks and would generally be classed as “growth” stocks. For example, Alibaba and Tencent are trading on FY1 forecast PEs (according to Bloomberg estimates) around 34x and 40x, respectively. This matters because the higher a stock’s PE multiple, the more sensitive that stock is to any change in expectations underlying that multiple (including growth expectations, risk perception and, topically, interest rate expectations).

By way of example, take two theoretical stocks; stock A on a PE of 10x, and stock B on a PE of 35x. Let’s assume that both have a payout ratio of 50% and investors apply a cost of equity of 12% to both. The simple Gordon Growth dividend discount model states that the forecast PE multiple should equal the forecast payout ratio divided by the cost of equity less the expected perpetual growth rate of dividends:

Plugging in the above numbers for our two imaginary stocks, and solving for growth, stock A has an implied expectation of 7% dividend growth and stock B has implied growth of nearly 11%.

Now, let’s imagine a 50bp change in the cost of equity (e.g. a rise in interest rates and/or an increase in perceived risk). If the growth expectation is unchanged this would imply a 9% fall in the PE (i.e. a 9% fall in the stock price) for stock A and a 25% fall for stock B. That is, stock B is much more sensitive to a change in underlying expectations. Compounding this effect, high PE growth stocks typically have much lower payout ratios (given they are retaining funds to reinvest in ROE-accretive projects). If stock B had a payout ratio of 10% the implied fall in stock price from a 50bp increase in the cost of equity would be 64%. Note the current payout ratios of Tencent and Alibaba are roughly 11% and 0%, respectively. This makes Alibaba, in one sense, akin to a long-dated zero coupon bond. Bond traders will appreciate the implied impact on expected volatility of such an asset. The payout ratio would need to increase substantially to mitigate such an effect, but estimates posted to Bloomberg suggest no expectation of that happening in at least the next few years.

In addition, rising interest rates would typically result in reduced growth expectations resulting in an even larger price response amongst high PE names.

Of course stocks do not trade perfectly according to a Gordon Growth DDM. It is also effectively impossible to identify whether it is changes in the cost of equity or changes in expected growth which are driving returns at any one point in time – they are two sides of the same coin. Nonetheless, it is reasonable to expect a substantially greater reaction amongst high PE names to any adverse change in underlying expectations.

Further highlighting this is not a purely academic exercise, a stock’s PE multiple can be split into a growth component and a valuation component, represented by the stock’s ROE and Price to Book ratio (PB), respectively (given PE = PB * (1/ROE)). By plotting ROEs and PBs for stocks in the market we can obtain a sense of the market pricing of ROE. We should expect a non-linear relationship whereby higher ROEs receive exponentially higher PBs. This is indeed observable in Figure 2 for Asia ex Japan stocks. However, we can also employ such an approach to identify potential outliers, where stocks are trading on unusually low or high PBs relative to their ROEs and relative to other stocks trading on similar ROEs.

Stocks included in this chart are all those that are active and available on Bloomberg with an Asia ex Japan domicile, ROE and Price to Book ratio estimate data, market cap > USD1b and excluding mainland China exchanges. Stocks with negative ROEs are removed from the sample and the chart is constrained to stocks with forecast ROEs < 50%. An exponential trend line is fitted to the data.

Source: Bloomberg, Firth Investment Management

Using such an approach we can see in Figure 2 that Alibaba and Tencent are relative outliers. They are trading on PBs of 6.7 and 11.3 versus ROEs of 18.3 and 30.8, respectively. Compare these with market average PBs (from the fitted trend line) for those ROE levels of 2.6 and 6.9. The implication is, if the market is highly rational and efficient, that investors are expecting large ROE increases for these stocks (to over 30% for Alibaba and over 37% for Tencent) to be on par with other stocks on similar PBs. Average analyst forecasts do not incorporate such an expected increase over at least the next 3 years.

For compliance reasons this research series is unable to provide financial advice. So keeping discussion to a purely factual level, it is worth noting that Tencent has seen its ROE trend lower since peaking in 2009, its ROA has fallen in every one of the last 6 quarters and the one factor that has prevented its ROE from falling even faster has been an upward trend in leverage.

Relating this all back to the Firth Asia Systematic Equity Strategy (FASE), we remain underweight large cap Asian tech, although roughly neutral Asian tech overall. Our process emphasises diversification as a key driver of portfolio construction, and our historic testing provides evidence of the success of that strategy over the long run. Our strategy has been specifically engineered to generate expected outperformance through the cycle. Unless an investor is able to identify and trade regime shifts in the cycle before they occur, we believe a high level of portfolio diversification is critical to inter-cycle performance.

Written by: Dr. Hamish Macalister