The Responsibility100 Index (Beta)
Category: Best data-driven reporting (small and large newsrooms)
Country/area: United Kingdom
Organisation: Tortoise Intelligence
Organisation size: Small
Publication date: 23/09/2019
Credit: James Harding, Alexandra Mousavizadeh, Andrew Haynes, Alex Clark, Luke Gbedemah, Alexi Mostrous, Chris Newell
There is an accountability gap for Britain’s most powerful companies. Annual reports often declare a commitment to sustainability goals, but rarely are these pledges followed up. The Responsibility100 Index is designed to close this gap and keep companies honest.
It’s also a first. Previous attempts to measure corporate responsibility have used proprietary data, or direct company surveys, but the Responsibility100 uses only publicly available data. And unlike previous attempts that measure only what a company says, this one directly compares a company’s commitments – in the form of their ‘talk’ score – with their actions – their ‘walk’ score.
The Responsibility100 has inspired a race to the top on sustainability reporting. The first step on the road to solutions – for the climate crisis, for equality and sustainability – is knowing the extent of the problem. The Responsibility100 is a landmark project in terms of accountability, providing comprehensive data across multiple issues for public use.
We’ve built the index to incentivise transparency. When a company fails to report data for a particular indicator, they are treated as if they had the worst possible datapoint. For instance, though around half of the FTSE100 report their renewable energy use, those that don’t are assumed to use 0%. But this assumption means a company can never lose points by disclosing data – instead, they are likely to move further up the ranking by becoming more transparent.
It’s this incentive structure that means the Responsibility100 has already encouraged changes in the way the FTSE 100 report their data. A major supermarket is now considering publishing their waste-to-landfill figure in response to the Index. Another company has told us that though it already pays a living wage it is now seeking accreditation to increase their Index performance.
To build Responsibility100, we first developed a methodology for a team of data journalists and scientists to manually examine the FTSE 100’s annual reports in a consistent way. The team relied on a pre-agreed collection template when reading sustainability material, awarding points for particular indicators when finding sufficient evidence.
However, we also developed automated tools to collect data. For example, to create a first estimate of the gender proportions of the FTSE 100’s boards of directors, we wrote a Python script that used the Companies House API to download each company’s register of directors and then guessed each director’s gender based on their name. Likewise, to get a sense of how often the FTSE 100 were being taken to a tribunal by employees, we used the web scraping framework Selenium to gather cases available from the gov.uk website.
One particular technological challenge during data collection was to account for the large number of subsidiaries that each of the FTSE 100 has significant control in. We mined the Companies House ‘People with significant control snapshot’ as well as undertaking tables in annual reports to create a parent-subsidiary network for each FTSE 100. In total, we found the FTSE 100 had majority control in around 20,000 distinct companies in 2018. Having this data meant we could not only search for, for instance, the employment tribunal cases involving the FTSE 100 but also any case involving one of their subsidiaries.
To visualise our investigation, designer Chris Newell created custom templates using the Flourish SDK, including a full ranking table that showcased each of the FTSE 100’s strengths and weaknesses. Another highlight of Chris’s work was a dumbbell chart to demonstrate which of the FTSE 100 had the biggest gap between their ‘talk’ on sustainability and their ‘walk’.
What was the hardest part of this project?
The hardest part of the project was the technical side of data collection. At best, we faced well-populated data tables provided by companies that were nonetheless contained in impenetrable formats like PDFs. Examples included gathering scope 1, 2 and 3 emissions values as well as the addresses of subsidiaries for checking whether they were in tax havens.
At worst, we had to discard indicators that seemed obvious choices for assessing corporate responsibility but proved unfeasible in practice. For instance, though we could gather the number of meetings the FTSE 100 were having with UK ministers and EU commissioners, we were unable to tell what had been discussed in those meetings. This meant we couldn’t assess whether the company was making a positive contribution on sustainability issues and so we stopped short of including this data in the final ranking.
However, while time-consuming, the benefit of the manual approach we were often forced to adopt has been to lay the foundations of an AI algorithm capable of extracting this data instead. As we’ve been going through annual reports we’ve been noting down page numbers, extracting text and taking screenshots – building a training set with an eye to automation. We hope to deploy such a model by the final quarterly update of the Responsibility100 in 2020 as we continue amassing training data in the meantime.
What can others learn from this project?
The first lesson to take away from Responsibility100 is that more transparency is needed in corporate sustainability reporting. While there are examples of good progress – for instance, the UK government’s required publishing of gender pay gap data – in other areas we faced huge gaps. Responsibility100 shines a particularly bright light on issues of data availability because, in being an index, by definition we require like-for-like, quantifiable data to fairly compare companies.
But another lesson from Responsibility100 is that having complete data on a company only gets you part of the way there on judging their sustainability efforts. In building the index and deciding on weights, we grappled with issues such as: if you are in an inherently environmentally damaging industry like mining or energy, but committed to sustainability in word and deed, perhaps you should be able to outperform a tech or media company with a much smaller environmental footprint that’s doing little to improve the prospects for people or the planet.
However, the lesson is not that these comparisons between companies shouldn’t be made – rather, Responsibility100 shows us that though no two companies face exactly the same sustainability demands, all can be assessed on whether they’re meeting a bare minimum as well as on how well they’re trying to meet their own distinct duties.