Kate Land

As the name implies, when you own a share you own a slice of a business. The market price of your share should depend on the existing value in the business and the future value it is likely to create. In his Nobel prize winning work on the empirical analysis of asset prices, Professor Robert Shiller demonstrated that over the long-term, market prices do indeed track earnings.

However, in the same work Shiller showed that prices fluctuate more than they should if market participants were always responding rationally to new information. This means there are periods when prices and fundamentals dislocate. How you spot such a dislocation is an important, valuable, and difficult question to answer. In Figure 1 we see that the price-to-earnings ratio of the S&P 500 is high with respect to history[1], having only peaked above 30 twice before in the 150 years of data we have. Whether this does or does not represent a dislocation partly comes down to whether these stocks are set to experience a very high level of growth in their future earnings, among other factors.

Much has been written about the relative performance of value and growth stocks during this long-in-the-tooth bull market. In Figure 2 we show the historic performance of a value and a growth portfolio since 1926[2]. In the last ten years (to Dec 2018), the value portfolio has returned 11.0% annualised, and the growth portfolio 15.1%. Such relative performance appears to be remarkable; across history value has rarely underperformed growth in any ten-year period.

But what exactly do the terms ‘value’ and ‘growth’ mean? A very specific methodology is used to define the portfolios shown in Figure 2[3], and many other methodologies exist using different fundamental metrics, metric combinations, datasets, selection criteria, and weightings. All applied to different universes of stocks.

Are the terms ‘value’ and ‘growth’ understood? And are they useful for making investment decisions?

Any metric based on recent realised fundamentals provides only a crude view of what you need to know to make good long-term investment decisions. You need to understand the future fundamentals of a company in order to know what price you should be paying for its shares. At Havelock London we do in-depth data-driven research into companies to understand their potential future earnings, and uncertainties. We are “valuation” investors rather than value investors. A stock doesn’t have to have a low price-to-earnings multiple for us to like it. But it does have to be trading at an attractive price, relative to realistic expectations of its earnings.

By using data to identify companies that we believe are good quality and are trading at attractive prices we are not simply ‘value’ or ‘growth’; we are both.

---

Footnotes

[1] Data courtesy of Prof Robert Shiller (http://www.econ.yale.edu/~shiller)

[2] Data courtesy of Prof Kenneth French (https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/)

[3] The methodology follows that of the Fama-French “HML” factor; US stocks are ranked by price-to-book each year with the top and bottom 30% assigned to ‘growth’ and ‘value’ portfolios respectively. Equal weight is given to small and large stocks within each portfolio, to separate a value effect from a size effect.

Successful active management comes from two components; investment selection and portfolio construction. Very crudely, the former is about maximising long-term returns and the latter is about ensuring there are no surprises on the way. Effective diversification requires identifying feasible scenarios that could impact a portfolio and determining portfolio weights that make it as resilient as possible.

The minimum number of securities required to achieve adequate diversification is a recurring question among investors and academics. It depends on an investor’s risk tolerance, and the scenarios most relevant to the securities in the portfolio. But one scenario that all portfolios are exposed to is that of idiosyncratic shocks in individual securities. Concern for exposure to single-name events drives investors to prefer larger portfolios to smaller ones, however the relevant risk is not usually well quantified, and the distribution of weights is as important as the portfolio size.

To explore this, we develop a metric that quantifies the single-name risk of a portfolio by establishing its effective size. While several metrics exist that quantify portfolio concentration, they generally lack interpretability in terms of risk[1]. Our metric is based on the chance of experiencing losses from single-name shocks, thus it goes straight to the heart of the matter[2].

We review some well-known indices in the table below and find their effective sizes to be much smaller than the headline number of assets[3]. This is a result of weight being unevenly distributed and concentrated in a relatively small number of holdings. In these examples, the effective size is well approximated by 2 x the number of assets covering 50% of the portfolio, which drives home the fact that it is the size of the largest positions that determines the exposure to single-name risk and not the number of assets in the portfolio.

Index	Number of assets	Effective size	Largest weight	Sum (number) of weights > 5%	Number covering 50%	Gini coefficient
FTSE100	100	25	11.3%	24.7% (3)	11	0.59
NASDAQ	100	22	9.7%	36.9% (4)	9	0.61
S&P500	500	99	3.6%	0% (0)	50	0.60

Rather than working with simulations, heuristics often form the basis of practical approaches to managing diversification. One well established rule is the 5/10/40 restriction placed on most UCITS funds which says that no single weight can be greater than 10%, and weights greater than 5% must total less than 40%. This equates to a minimum effective size of ~15-18[4]. It is interesting to note that these rules are relaxed specifically for UCITS schemes that replicate indices, and the FTSE100 and the NASDAQ are frequently at odds with them[5]. Therefore, products that track these indices can, and do, take positions that are prohibited for an active manager.

We conclude that it is important to look past the number of assets in a portfolio in order to assess risk. Equity funds that contain ~25 stocks will generally be referred to as ‘concentrated’ by the industry, but as we have seen here these portfolios may have lower risk levels than those with a much larger number of assets. Working with metrics that meaningfully quantify the risk you are concerned with is key to effective risk management.

---

[1] For example, the Gini coefficient compares portfolio weights to those of an equally-weighted portfolio of the same size. This does not help to understand the risk associated with portfolios of different sizes, as seen in the table.

[2] We simulate 10⁶ 10y futures in which each asset in a portfolio has a 1% chance of experiencing complete loss in a single year. We quantify exposure to large losses with a weighted-count of 10y losses >15%. Effective size is defined as the size of an equally weighted portfolio with the same exposure to large losses, using the same methodology. Using like-for-like analysis makes the effective size metric robust to the underlying assumptions of the simulations.

[3] Index weights are as of 31/12/18, 15/02/19, and 14/02/19 for FTSE100, NASDAQ and S&P500 respectively. Multiple share classes for the same issuing body are combined.

[4] The higher of these considers that it is not practical to maintain weights exactly at the boundaries of what is permissible.

[5] For example, the FTSE100 largest position is currently 11.3%. Further, as of 16/11/2018, the NASDAQ had 41.0% of its weight in just 4 positions with the largest at 11.9%.

What is data science?

A lot of people are talking about data science at the moment, but they aren’t always clear about what they mean. Here at Havelock London, we consider data science to be the activity of collecting and processing data, to gain insight and guide decisions, often within a business. A data scientist combines computing, statistics, and domain knowledge – in our case, knowledge of investment management – to identify where and how data can be useful.

You mean AI, big data, machine learning and all that jazz?

Not necessarily. Whilst the ‘data science’ term is new, what it is describing is not. Scientists have always made discoveries about the natural world from observations, and there is a long history of ‘quants’ working in finance and businesses employing analysts. What has changed is the amount of data that most businesses can now access, and an increased demand for people who can identify ways to leverage this data.
The technical activities data scientists are involved in vary with the domain, but the goal is usually to improve the effectiveness of decisions, or to automate them. In some areas useful contributions come in the form of technology that can capture and organise data. In other areas advances are made by better understanding complex data that already exists.

Ok, so what has data science got to do with long-term investing?

Let’s start by identifying some of the challenges faced by an investor:

Too much data, too much noise!

Someone making an investment decision today does not lack data; there are newspapers, TV channels, websites, businesses and individuals dedicated to providing data and commentary about the financial markets. What is difficult is knowing which pieces are relevant. Investors need to know how to filter the massive amounts of information at their disposal.

Vast amounts of uncertainty

No one knows exactly what is going to happen in the markets tomorrow, let alone over the next few years. Some events are more likely than others but there always remains the possibility that you will lose money. An investor needs to be able to understand the range of potential outcomes that they face.

As for the solutions… watch this space!