A Data Scientist’s introduction to sustainable finance

Nov 21

A marked change in atmospheric carbon has always been incompatible with Earth’s stability, and has been a feature of all 5 mass extinctions. In previous events it has taken volcanic activity up to one million years to increase atmospheric carbon to cause a catastrophe, but an extremely fast and energy-thirsty growth has led to similar events occurring in less than 200 years. The global air temperature had been relatively stable up until the 1990’s due to the ocean’s ability to absorb the excess heat.

The existence and impacts of climate change have been highlighted by scientist’s since the 1980’s, and even though some progress has been made following various UN events (e.g. The 1992 Earth Summit convention to protect biodiversity and the 2010 20 targets to protect biodiversity) there seems to be a lack of global legal account that maintains the unsustainable status quo. Our encroachment further and further into wildlife habitats has been linked to driving emerging diseases, as we have seen with the Covid-19 pandemic. It is clear the sixth mass extinction is well underway as the number of vertebrate species to have gone extinct in the last century would have naturally taken between 800 and 10,000 years, hence a hundred times faster (Ceballos et al 2015). We are still faced with the bleak outlook that if we collectively do not control the heating of the Earth’s surface by less than two degrees by 2100, our water and food supply will be threatened along with, climate control and increased risk of future pandemics.

The last few years have seen an explosion of climate change attention from social movements to countries and organisations setting out ambitious goals to become carbon neutral in the next few decades. Balancing human existence with the natural world in sustainable way, is however, a multifaceted problem requiring a range of approaches. In a reflection of this complexity a recent London climathon, posed four different challenges: 1) Food systems and the climate crisis (Food wastage and transport), 2) Sustainable living and wellbeing (Devising more socio-economic diverse green movements), 3) Carbon Footprint in Supply Chain (Embedding low carbon considerations in the supply chain) and 4)Urbanisation and Development (Zero carbon community through buildings, infrastructure and green spaces).

Before Data Scientists can translate a real world use case into a useful machine learning model, relevant contextual information is required in order to effectively gather the right data, clean it, develop useful features and build a model, hence in this blog post I will introduce and focus on sustainable finance by covering what sustainable finance is, what are the current challenges, what data exists, what models currently exists and the future impact Data Science can have.

What is sustainable Finance and why is it important?

Sustainable finance refers to the process of taking account of environmental, social and governance (ESG) consideration when making investment decisions, leading to increased long term investments into sustainable economic activities.

In 2015 the UN defined 17 Sustainable Development Goals (SDGs), which 193 countries committed to achieving by 2030. The 17 SDGs are: : No poverty, zero hunger, good health and wellbeing, quality education, clean water and sanitisation, decent work and economic growth, industry innovation and infrastructure, reduced inequalities, sustainable cities and communities, responsible consumption and production, life below water, life on land, peace, justice and strong institutions, partnerships for the goals.

Whilst SDG’s did not enter the global lexicon until 2015, investors have been focusing efforts toward socially responsible investing since 2006 using ESG considerations. ESG considerations provide a broader level mapping of SDG’s when considering investment opportunities. Environmental considerations refer to the preservation of biodiversity, pollution prevention and climate change mitigation. Social considerations refer to prevention of inequality, investing in human capital and communities, human rights issues and the relationships with employees, suppliers and customers. Governance considerations to the governance of public and private organisations such as management structures, human resource processes, shareholder rights and the implementation of sustainable corporate strategy.

So why is sustainable finance so important within the wider sustainability conversation? Well, sustainable finance has a key role to play in mobilising the necessary capital to deliver against the sustainability goals. If financial intuitions are incentivised to shift their banking book and contribute to the 600 billion of green investments needed each year for the next 15 years, the goal of limiting global warming below 2.0 degrees by 2100 could be achieved (Vinciguerra et al, 2020).

Sustainable investing can be profitable

While some may argue that focusing on sustainability for profitability is a sinister capitalist outlook, the reality is many countries economic and political systems are engrained and highly dependent on private profitable trade and supply chains are the lifeblood of the global economy, therefore if profitability is not a consideration within the climate change conversation then change is less likely to occur at the speed that is required. Impact investing is the mature segment of sustainable finance and includes sectors such as sustainable agriculture, renewable energy, conservation, affordable housing, healthcare and education. It is now considered a myth that revenue cannot be generated through the implementation of SDGs. A key example of this is the machine learning services the Elastacloud Data Science team created to predict solar irradiance and performance ratios of solar farms for solar farm managers and investors, the outputs of which can be used to select the optimal time to release energy onto the grid and when to store energy according to supply and demand leading to the best price for energy whilst minimising balancing fees. Similarly, if power purchase agreements were not likely to be fulfilled, the insight into future performance allowed for the leveraging of inter and intra- day markets to buy surplus energy at a better price in order to meet obligations. Another example of the lucrative savings that can be achieved through optimising business processes to become more sustainable is the project gigaton that Walmart initiated. After pledging, three years ago, to cut their carbon emissions by 18% by 2025, Walmart discovered that suppliers accounted for 90% of their emissions. Project Gigaton allowed Walmart to scrutinise their supply chains and work with suppliers to cut a gigaton of carbon by 2030. In addition to supply chain logistics, many other companies (e.g. Amazon, Royal Mail, Waste Cycle/Enva) have implemented route optimisation techniques when collecting and making deliveries, which increases productivity and improves fuel efficiency.

In 2018, Bank of America Merrill Lynch found that firms with a better ESG record than their peers, produced high three-year returns, were more likely to become high-quality stocks, were more resilient to price declines, and were less likely to go bankrupt. This resilience was clearly evident during the first few months of this year when the Covid-19 pandemic significantly effected markets. The S&P 500 ESG index, during this period beat the normal S&P index by 0.6%. But, is this the real picture of sustainable investing? It has been suggested that as tech companies make up much of the indices, it is no wonder stocks with high ESG rating have appeared to do well during the Covid-19 pandemic, as more of us turned toward working from home arrangements the reliance on technology was amplified.

Embed from Getty Images

Current Challenges in sustainable Finance

Greenwashing

With the ever growing demand and attention toward making our processes more sustainable, laws, regulation, technological advancements and skill development have not entirely kept up to pace, and so areas of ambiguities have arisen. Many have highlighted, that rather than trying to invest in truly sustainable companies and products, the finance industry simply invest in the least-worst opportunity, such as the least polluting oil and gas company. Similarly, one of Saudi Arabia’s fossil-fuel guzzling electricity monopoly tried to leverage ‘green investment’ to install smart meters. Another questionable trend is within the automobile sector, where car companies are issuing green bonds to fund the development of electric vehicles. The calls for better laws, regulation, data and standardisation to prevent this type of ‘greenwashing’ have been repetitive and numerous. In addition to such formal improvements, the disconnect between corporate sustainability public pledges and business process needs to be addressed by organisations. In some instances, businesses have produced glowing sustainability reports, despite the underlying performance of the organisation being poor (e.g. Enron, see Kulik, 2005). Many have suggested that sustainable investments have been over-inflated in price due to the popularity, and hype surrounding sustainability, rather than tangible positive actions. Caution against such risks have led to events such as Scottish Widows dumping £440m investments that fail to meet ESG standards.

Lack of consistency, transparency and standardisation

Vitaly Nesis, Chief Executive of Polymetal (biggest London-listed gold producer) has criticised methods of ESG scoring for being inconsistent, inaccurate and an exercise to tick boxes based on marketing statements or self-reported statistics. Likewise, chief executive if asset management at BlackRock, Larry Flink has communicated the need for standardised climate reporting. There is a clear lack of consistency and transparency across the industry, and whilst frameworks have been published by the Sustainable Accounting Standard Board (SASB) and the Taskforce for Climate-related financial disclosures do exist, the methods can often be time consuming and lack acknowledgment of fast moving data that can influence investing decisions. Others have gone as far as suggesting sustainability reports are used to erect the facades that enable organised hypocrisy (Cho et al, 2015).

Data Quality: lots of noise, little meaning

Many organisations struggle to accurately measure sustainability impact, because relevant data is not being collected or is of poor quality. There is a disparity between affluent and less-affluent companies, with the former being able to invest in various tracking devices and processes more so than the former. This coupled with the complexities and variations in sustainability reporting provide prominent roadblocks for investors hoping to make informed choices (Elastacloud whitepaper). There are over 600 ESG data providers and a continuously growing number of sustainability indexes (the Dow Jones Sustainability world/Europe index and the FTSE4Good index being two very common references), with variable correlation among them. With varying differences between indexes, it is difficult to know which is better without being able to evaluate the methodology upon which they were formulated, and the weight applied to different metrics. As methodologies are largely kept confidential evaluation is difficult (Delmas and Blass, 2010).

One size does not fit all

To some degree a variable number of sustainability indexes will be required, as depending on the investment sector being considered, one generic sustainability index will not be appropriate. For example, the impact of agricultural pesticide processes polluting rivers and clearing parts of forests is much easier to measure through satellite data than the impacts of the electronic industry or sustainable housing industries. Evidently, the same data cannot be used to measure impacts of all industries.

Lack of overlay between environmental and economic data and professionals

There is a clear lack of overlay between finance professionals and environmental scientists, which would allow for collective data and considerations to provide a better holistic, accurate picture of sustainable finance opportunities. The two professions often speak in different languages with prominent focus in either financial or the environment impact. Part of the reason for this pattern can be attributed to the speed at which the sustainable finance industry has grown over the last few years. It is hoped that Data Science can facilitate bridging this communication gap through robust analytical approaches applied to large volumes of data. Data Scientist’s often work with industry professionals to understand different domains and relevant data. Through this understanding, data can be enriched and processed to be used to develop useful insights and predictive models that can improve decision making.

Embed from Getty Images

Data Sources, Current Models and the impact of Data Science

So what kind of models and data is being used for reporting, where are the gaps and what available data is not yet being utilised? These questions are quite broad and outputs are constantly updating, therefore the information presented here is by no means exhaustive, but provides some insight into what is available.

Data and Models used in Sustainable Investing

Many investment research companies have screening methodologies to help socially responsible investors select companies. As previously mentioned, whilst financial performance indicators are well defined and structured there remains heterogeneity when considering environmental performance indicators. Financial KPI’s (e.g. alphas factors (excess return on an investment relative to the return on a benchmark index), beta Factors (the measure of relative volatility)) are just one part of the puzzle when considering portfolio risk analysis, and the transitional (credit loss from adjusting towards lower-carbon investments) and physical risks (climate disruptions) of sustainable investments need to also be considered (Vinciguerra et al, 2020). Some the environmental performance indicators considered, broadly include: environmental impact (toxicity, emissions, energy use, revenues from coal oil derivative products, etc), regulatory compliance (violation fees, number of audits, etc), organisational processes (environmental accounting, audits, reports, environmental management systems, pollution prevention, recycling, removal of hazardous waste) (Delmas and Blass, 2010). NLP methods, such as text classification and sentiment analysis are actively being used to extract key information from annual produced sustainability reports in an efficient manner (KPMG at PyData 2019).

Generic Models overlaying economic and environmental data

Whilst specific sustainable finance investment opportunities will go through vigorously extensive life-cycle analysis processes of individual organisations, others have proposed more generic models. In a recent webinar I attended, presented by Daniel Moran from NTNU (From satellite to supply chain: New approaches to connect earth observation to economic decisions) the importance and complexities of overlaying environmental and economic data were considered. Using country level statistics and remote sensing, Moran and colleagues have proposed a method to gain insight into economic impacts on sustainability. Billions of pounds are invested in Earth observation infrastructure allowing for near real-time observation of the planet, but most of this information never leaves the scientific silo. Using satellite sensor technologies and hyperspectral image-processing, real-time data on a wide range of ecosystem changes can be monitored, such as crop type production, urban road expansion, surface water seasonality, deforestation and biodiversity loss. Input-output tables describe the sale and purchase relationships between producers and consumers across different industries. Environmentally extended multi-region input-output tables offer a framework for linking upstream and downstream economic activity to environmental impacts. Extensive academic work has allowed for the creation of such data tables, through approaches such as maximum entropy models and Bayesian approaches. For example, the Eora MRIO documents 5 billion supply chains, connecting 15,000 sectors across 190 countries between the year 1970-2020 in monetary terns and with over 2000 itemized emission and ESG indicators. This generic output has still proven to be tremendously useful to organisations such as the world bank, Accenture, Amazon, London Stock Exchange and more.

Organisation specific tools: carbon calculators, blockchain and GPS tracking

Whilst, it would be a game changer to over lay more specific data with environmental data, that covers the entire supply chain of business processes (production, transportation, manufacturing and consumption), it is not a legal requirement to disclose such data and remains a complex issue of trade-offs between corporate privacy/transparency vs environmental issues. The WWF have a wealth of data to but explicitly do not attempt to shame sectors or industries. As will be further demonstrated, some organisations have utilised blockchain to overcome the sensitivities of this debate. Current tools are either limited to broad country level analyses or require a narrow-focused, laborious life cycle assessment. As Moran et al (2020) highlight, tools which allow for transparency of supply chains, while keeping costs low are needed and should be data driven, third party services, thus removing the need for self-reporting.

With the growing trend of sustainability becoming more mainstream, technology and environmental organisations have also started collect various types of data to analyse and predict various trends occurring as a result of climate change. For example, Microsoft and AWS have developed carbon calculators that easily integrate with their cloud platforms. As well as developing tools to facilitate organisations to track their environmental impacts, tech giants are also dedicating parts of their businesses to focus solely on how AI can be used for good. These AI for good initiatives look to apply machine learning and develop tools covering everything from: health, Earth conservation, accessibility, humanitarian action to cultural heritage.

Nestle have also partnered with Open SC (founded by WWF Australia and the Boston Consulting Group Digital Ventures) to pilot open blockchain technology to trace milk from farms in New Zealand to Nestle factories and warehouses in the Middle East. Nestle plan to later test the platform using palm oil sources in the Americas, allowing consumers to make better informed choices.

Open SC have also leveraged ship GPS tracking data from sensors to collect information such as the speed of fishing vessels and how deep the sea floor. This information if fed into machine learning models to establish in an automated way, whether the ship is fishing in an area it is supposed to be fishing in, thus validating sustainability claims. Each fish is also RFID tagged allowing a unique serial number to be attributed to each fish, which can be tracked through the entire supply chain. The sustainably verified and traced fish data can also be shared with consumers using blockchain. Open SC have developed different digital tools depending on whether the consumer is in a supermarket or a restaurant to account for how different cognitive and emotional states will impact how people want to consume information in different environments, thus further promoting better choices to be made. Other companies such as Fishcoin are providing similar services.

Concluding Remarks

The tip of the iceberg has demonstrated the use of generic models which overlay economic and environmental data, to be more specific tools to guide investment decision making or facilitate organisations to develop better interventions in the supply chains using NLP, image processing, blockchain, Bayesian methods and many more. It is clear through robust analytical approaches Data Science could be of paramount importance in the fight against climate change and guiding our communities toward a more sustainable future.

I hope you enjoyed reading my first blog post- any feedback or twitter shares will be appreciated. If you are interested and want to learn more about sustainable finance, please feel free to sign up to the Data Science London Hackathon we will be running in January 2021.

References and useful resources

Papers and Blogs

ESG on a Sunday

Elastacloud white paper

How banks can save the world! (Green RWA) ((Vinciguerra et al, 2020)

Overview of sustainable finance

ESG to SDGs: Connected Paths to a Sustainable Future

Sustainability and profitability can co-exist. Here’s how

Why ESG investing is a megatrend no asset manager can ignore

Amazon’s Last Mile Delivery: steps to finding your key to improved delivery times

Ethical investment if about morals not markets

Gold chief call for common ESG reporting standard