According to IBM, we create 2.5 quintillion bytes of data every day, with 90% of the world’s data having been produced in the last two years alone. Businesses across the spectrum, from social media and entertainment to banking and investments, have recognised the benefits of harnessing big data.
Advances in intelligent technologies have resulted in data screening tools (quantitative analysis) gaining a strong foothold in the investment management industry. When it comes to finding good companies in which to invest, information and timing are key. This requires the ability to speedily process huge volumes of information from multiple sources.
A quantitative approach uses mathematical and statistical modelling that collects immense volumes of data at lightning speed to help identify good investment ideas and assess portfolio risks. To put it in perspective, more than 16 million discrete pieces of information feed into constructing a portfolio from a 4000-stock universe. Quant analysts are often described as data analysts working in finance, as they need to be highly skilled in maths, statistics, computer science and finance.
Why human insight remains key
When it comes to big data and machine learning, public debate tends to pit humans against machines, reinforcing the stereotype of an ‘us versus them’ scenario, rather than entertaining a ‘marriage of two minds’. Even within the asset management industry, it is common to find that active equity managers only employ quantitative analysis as an initial screening tool to identify good investment ideas based on a specific set of criteria. So, quants is often used as a filter to narrow a large investment universe, after which fundamental analysts do a deep dive on these stock ideas.
Figure 1: Harnessing human insight and technology
However, in our view, each approach (quantitative and fundamental analysis) has key strengths and weaknesses. To put it into perspective: quants is ‘a mile wide, but an inch deep’, while fundamental research is a ‘mile deep, but an inch wide’.
Table 1: Both approaches have strengths and weaknesses
Idea generation – finding the right balance
Over the last eight years we have refined our investment process, blending fundamental insight with quantitative analysis to produce more consistent investment returns for our investors. We employ a two-fold process to generate ideas:
- Qualitative: bottom-up fundamental research from our experienced team of analysts, where the best investment ideas are identified based on fundamental company research. For instance, our analysts will examine the market’s earnings forecasts for a company. The aim is to determine whether a company is likely to have higher (or lower) earnings than other market participants expect. When profit forecasts are revised upwards or downwards, we believe it can have a material impact on a company’s share price.
- Quantitative: a stock screen process which identifies the best investment ideas, based on very specific data-driven, fundamental criteria. This screening includes finding companies based on fundamental investment data, with favourable dynamics (earnings/profit expectations) and reasonable valuations.
Figure 2: Idea generation in the Investec General Equity strategy
It is important to note that our quantitative stock screen research and fundamental analysis run parallel to each other, so the one process isn’t relegated to a supporting act: both have a star billing. The investment ideas that are identified by both our quantitative and fundamental analysts typically represent our high conviction stock picks in the Investec General Equity strategy. For example, Anglo American and BHP Billiton came up as top picks based on our quantitative and fundamental research. These stocks, together with platinum stocks, currently represent a material holding in the Investec General Equity strategy.
A difference of opinion can be healthy
As these two research processes run independently, the one may identify a stock as a good or bad investment idea, which might not be supported by the other. For example, Naspers scores poorly on our quant-based metrics (valuation). But a more detailed fundamental analysis reveals that Naspers represents reasonable value, when valuing each business within the internet and media giant separately. Both processes may fuel debate, highlighting the need for further analysis of an investment idea.
This integrated approach reduces the risk of ‘over-confidence’, which can be a pitfall where idea-generation favours only one of these two research processes. Essentially, we capitalise on the best attributes of both. Quantitative screening offers discipline, repeatability, objectivity and efficiency, while bottom-up fundamental research provides depth, human insight and judgement.
Managing risk together
Investments may result in financial losses, which is why it’s so important to manage the tension between investment conviction and risk. Portfolio construction and risk management are complex processes for humans. Constructing a simple 30-stock portfolio requires managing hundreds of pieces of expected return, risk and related information, easy for a machine but difficult for a human. Some asset managers will try and manage risk by limiting the weighting of an individual stock or by imposing sector-specific exposures. We believe these measures are a blunt way of managing risk as they constrain potential outperformance and do not consider the multiple correlations between stocks and sectors.
Having quant expertise within our Investec General Equity investment team has enabled us to develop an extensive risk and portfolio management process over years to address these challenges. Proprietary quant tools allow us to combine key risk and return metrics to optimise our portfolio and maximise diversification. Our internally developed system provides crucial information to us on a pre-trade basis. For example, before we implement a trade to reduce the allocation to one stock in favour of another, the system tells us how it will impact the overall portfolio risk.
This process is based on the principals of mean-variance optimisation introduced by American economist Harry Markowitz (for which he won the 1990 Nobel Prize). By introducing the concept of a co-variance matrix as a risk model we are able to ensure that our position sizes are risk-aware and diversified. The co-variance matrix captures:
- how volatile or risky an individual stock is; and
- how similar or dissimilar stocks behave relative to each other – in much the same way as sector classifications attempt to do, but far more efficiently.
Our optimisation algorithms take care of blending the return and risk features of all the ideas that we identify during our research.
While our proprietary models have greatly enhanced the risk management and portfolio construction process, human insight and common sense remain crucial. Fundamental analysts are best placed to interpret breaking company news, market dynamics, regulatory or tax changes, and environmental, social and governance (ESG) issues. For instance, fundamental analysis allowed us to assess the risks to Sasol’s earnings when carbon taxes were introduced this year. Given deteriorating fundamentals, we reduced our position in Sasol. When governance issues hit Glencore last year, we took the decision to exit our position.
Figure 3: Investec General Equity strategy