Quant managers are mining ‘unstructured data’ as part of the search for the ESG factor
- A lot of work is going into discovering a systematic way of identifying ESG stocks but significant roadblocks remain
- Conventional ESG scoring methodologies do not fit into traditional factor models
- There are great possibilities in the potential use of ‘unstructured data’ that would transform our ability to glean valuable ESG insights from a company
European institutional investors are increasingly allocating large swathes of their portfolios to ESG investments. Whereas once this was a niche sector, with investors prepared to accept potentially lower returns as the price of their convictions, they are now expecting more from their managers.
The incorporation of ESG factors has historically only been associated with fundamental strategies but this has been changing. As the appetite for ESG products continues to surge, quant managers are rising to meet the challenge and upping the ante in their search for a reliable way of identifying stocks with strong or improving ESG characteristics that are likely to generate alpha. As quantitative investment processes are likely the most efficient at incorporating ESG data, it is perfectly achievable for quant investors to do well through doing good.
To discover an ‘ESG factor’, namely a systematic way of identifying such stocks, would be a huge win for the industry. A lot of work is going into discovering one but there remain some significant roadblocks.
The first, and most significant, is that ESG is not a factor in the traditional sense. A factor is defined as a set of quantifiable characteristics that can explain stock risk and returns. Traditional factors are robust, having been tested across different geographies and market conditions for a long period of time. Moreover, they are well documented and, although they may be defined differently by different investors, a consensus view exists on basic definitions. The same cannot be said about ESG.
As a quant, I am used to looking at multiple decades’ data, to use as a basis for an investment strategy. There is not ESG data on companies that goes back far enough to decisively say that it is a factor that is driving risk and return in a systematic fashion. Clearly more (and higher quality) ESG data is needed. While ESG data coverage has improved, it is still sparse when one compares it to the data available on traditional financial factors.
At most, ESG data spans a little over a decade and is available only for a fraction of companies in the investable universe. Data on factors such as value, momentum and others has been collected and analysed for several decades and covers a universe of thousands of stocks.
It is also necessary to consider that conventional ESG scoring methodologies do not fit into traditional factor models. For example, larger companies tend to have better corporate governance and disclosure policies when compared to smaller companies, and (due to regulatory requirements) European companies tend to be more transparent than their North American counterparts. All of this affects the ultimate ESG scores. If more companies disclosed more ESG data, particularly using similar metrics to peers and allowed their ESG data to be audited and released information on a more regular basis, investors would be better placed to identify alpha opportunities using ESG metrics.
It is possible to develop a quantitative approach which combines the material ESG items and ESG expansion. Companies with better ESG metrics have higher valuations than lower-scoring companies but comparable future returns. Financial analysts may misprice the returns of good ESG companies by expecting their higher valuations to continue, in much the same way that good-quality companies often enjoy both current higher valuations and potentially higher future returns.
Moreover, our research shows that companies with better ESG scores have similar returns to those with poor ESG scores. This suggests it should be possible to systematically tilt a portfolio toward better-scoring companies without detracting from performance.
The hunt for that elusive ESG factor continues, but in the meantime, it would help to focus less on the effort a company exerts (such as policies and committees) and more on the measurable impact of their behaviour. Here, alternative (or big) data could help generate the necessary ESG insights. In fact, given time, quants could fundamentally overhaul the way the investment industry views ESG and the ability of an investor to identify stocks that meet more specific ESG requirements.
There are great possibilities in the potential use of ‘unstructured data’ that would transform our ability to glean valuable ESG insights from a company. This unstructured data could include information taken from websites such as Glassdoor, a forum for employees to anonymously review their company and share salary information, and social media outlets. The ability to systematically harvest candid insights from employees about their inner workings of their companies could prove to be far more illuminating when it comes to corporate social responsibility programmes than what the company is likely to formally disclose.
Until a consistent means to capture the ESG factor is identified, investors may need to rethink how ESG is evaluated. However, the world is changing, and in the future, maybe thanks to pressure from governments and regulators, there will be longer and more robust data sets about ESG. These, in turn, will enable us as quants to feel more confident about the statistical significance of the data sets that we are able to access. In other words, all hope is not lost, and quants should take heart from the fact that, while in many ways ESG scores are in the eye of the beholder, quant strategies provide a consistent lens for data.
Gavin Smith is a portfolio manager at QMA