Kategorie: Stellungnahmen
INSIGHTS Stellungnahmen
Hallo! Ich bin Mabelle, KI-Forscherin bei Arabesque AI und arbeite in der Forschung und Entwicklung. Ich bin auf Natural Language Processing (NLP) spezialisiert und verbringe den größten Teil meiner Forschungszeit mit der Analyse von Social-Media-Daten. Während meiner Doktorarbeit extrahierte ich diskrete Emotionen aus historischen Tweets und nutzte sie, um Twitter-Benutzer mit Depressionsrisiko zu identifizieren. Bei Arabesque AI konzentriere ich mich nun darauf, wie wir zukünftige Aktienkursbewegungen mit Hilfe von Twitter-Daten abschätzen können. In diesem Beitrag werde ich Ihnen erläutern, wie dies unter Verwendung von Twitter-Daten für ein anonymisiertes Unternehmen erreicht werden könnte.
von Dr. Mabelle Chen
Mit dem revolutionären Wachstum der sozialen Medien ist die Verwendung großer Datenmengen zum neuesten Trend für Forscher geworden, die Aktienmarktbewegungen analysieren. Unter Verwendung von Techniken aus dem NLP ist die Stimmungsanalyse die vorherrschende Methode zur Extraktion von Merkmalen aus solchen Datenquellen gewesen. Eine kurze Suche von „stock market prediction Twitter“ auf Google Scholar wird zeigen, dass 9 von 10 Arbeiten entweder „sentiment analysis“ oder „mood“ im Titel haben, was eine Allgegenwart der Stimmungsanalyse in NLP-Anwendungen demonstriert. Siehe zum Bollen et al. (2011), Mittal et al. (2011), Yu et al. (2012), Nguyen und Shirai (2015), und Sahana et al. (2019). The concept of using sentiment analysis to predict stock price movements has its origins in behavioural economics. According to Nofsinger (2010), Das Konzept der Verwendung der Stimmungsanalyse zur Vorhersage von Aktienkursbewegungen hat seinen Ursprung in der Verhaltensökonomie. Nach Nofsinger (2010) beeinflusst der Grad des Optimismus oder Pessimismus in einer Gesellschaft die Entscheidungen von Konsumenten, Investoren und Führungskräften von Unternehmen. Dies hat einen Einfluss auf die Gesamtinvestitionen und die Geschäftstätigkeit, was darauf hindeutet, dass die gesellschaftliche Stimmung dazu beitragen kann, zukünftige finanzielle und wirtschaftliche Aktivitäten abzuschätzen. Daher versuchen viele Forscher, die Bewegung des Marktes anhand der Dynamik von Unternehmen, wirtschaftlicher Szenarien und öffentlicher Stimmungen vorherzusagen.
Ist jedoch das Stimmungsbild das einzig sinnvolle Merkmal, das man aus solchen Daten extrahieren kann? Die Antwort ist definitiv nein. In diesem Beitrag untersuche ich, welche anderen Informationen und Merkmale aus Tweets extrahiert werden können und wie effektiv diese zusätzlichen Merkmale in Verbindung mit der Stimmung für die Schätzung künftiger Aktienkursbewegungen sind.
Dieser Blog ist in vier Abschnitte unterteilt:
-
Merkmale: Zu Beginn stelle ich eine Reihe von Features vor, die aus Twitter-Daten extrahiert werden können und sich diesen drei Kategorien zuordnen lassen:
-
Vorhersage:: Ich experimentiere dann mit diesen Merkmalen, um zu beurteilen, wie sie zur Vorhersage der Bewegung von vier Aktienzielen verwendet werden können.
-
Erläuterung: Als Nächstes verwende ich die Modellinterpretation, um zu untersuchen, wie diese Merkmale zu den einzelnen Vorhersageaufgaben beitragen.
-
Schlussfolgerung: Abschließend teile ich meine Schlussfolgerung zur Leistung plattformspezifischer und inhaltsbasierter Merkmale im Vergleich zur Stimmung.
Merkmale: Was können wir also wirklich an Tweets messen?
Während das Extrahieren der Stimmung aus Tweets ziemlich einfach ist, steckt tatsächlich eine Fülle von Informationen in jedem. In diesem Abschnitt gehe ich auf all die verschiedenen Arten von Merkmalen ein, die aus diesen Tweets extrahiert werden können. Einige von ihnen könnten sich für Modellierungszwecke als nützlich erweisen. Zur Demonstration habe ich ein bestimmtes Unternehmen als Beispiel untersucht, das als Firma A bezeichnet wird. Ich habe
tweepy
(einen Python-Wrapper für die Twitter-API) verwendet, um öffentliche Tweets zu sammeln, in denen diese Firma vom 26. November 2019 bis zum 20. Juli 2020 erwähnt wurde. Insgesamt wurden 85.490 Tweets gesammelt, die sich über 242 Tage erstreckten. Die Verteilung dieser Tweets ist in der folgenden Tabelle dargestellt. Die Tweets wurden täglich gruppiert, um eine Reihe von aggregierten Merkmalen für jeden Tag zu erstellen.
Bevor diese Merkmale beschrieben werden, ist es wichtig zu betonen, dass es sich bei allen Merkmalen um Momentaufnahmen handelt, die am Ende eines jeden Tages gemacht werden. Bei jedem Tweet könnte die Anzahl der Likes und Retweets von Tag zu Tag weiter zunehmen. Wenn keine Momentaufnahmen gemacht werden, enthalten die Daten Informationen aus der Zukunft enthalten. Diese häufige Gefahr beim maschinellen Lernen wird als Datenverlust bezeichnet. Da unsere Daten als Zeitreihen vorliegen, ist es absolut entscheidend, dass wir diesen Fehler für zukünftige Prognosen nicht machen. Lassen Sie uns nun die Merkmale untersuchen, die für Modellierungszwecke generiert werden können, wie im folgenden Diagramm dargestellt.
-
Zu Beginn können einige einfache Statistiken relativ leicht extrahiert werden, wie z. B. das Volumen der Tweets und die durchschnittliche Länge der Tweets pro Tag. Man kann jederzeit kreativ werden und weitere Maße in dieser Kategorie definieren. Die Verteilungen der einzelnen Features für den gesamten Datensatz sind in den nachfolgenden Diagrammen unten dargestellt.
-
Funktionen der Twitter-Plattform
Twitter bietet eine Reihe von Funktionen, mit denen Online-Nutzer sich mitteilen und Inhalte austauschen können. Diese plattformspezifischen Funktionen können als Features genutzt werden. Beispielsweise können aus der Verwendung von Hashtags zwei Kennzahlen erstellt werden, indem 1) die Anzahl der verwendeten eindeutigen Hashtags und 2) die Gesamtzahl der verwendeten Hashtags gezählt wird. Da Hashtags zum Markieren des Themas oder der Thematik eines Tweets verwendet werden, beschreiben diese Funktionen, wie viele spezifische Themen an einem Tag diskutiert werden und wie groß der Umfang dieser Themen ist.
Das gleiche Verfahren kann für die Erwähnungen von Benutzern angewendet werden, um die Anzahl der einmaligen Erwähnungen und die Gesamtzahl der Erwähnungen zu erhalten, die den Umfang der Benutzerinteraktionen durch Tweets an einem bestimmten Tag beschreiben.
Eine weitere wichtige Funktion von Twitter ist das Retweet. Durch Extrahieren des Retweet-Volumens und seines Verhältnisses zu allen Tweets an einem bestimmten Tag kann ich erfassen, ob ein Trend-Tweet oder ein Thema viral geworden ist, was die Wahrscheinlichkeit erhöht, dass es soziale Auswirkungen hat. Um diese Informationen zusammenzufassen, berechne ich die durchschnittliche Anzahl der Retweets und das Verhältnis der Retweets zu allen Tweets für jeden Tag.
In ähnlicher Weise enthalten Freigaben eines externen Links (in Form einer URL) oder ähnliche Tweets Informationen darüber, welche Inhalte an einem bestimmten Tag die meiste öffentliche Aufmerksamkeit erregt haben. Um diese Informationen zu erfassen, berechne ich den prozentualen Anteil der Tweets, die URLs enthalten, an allen Tweets an einem bestimmten Tag als URL-Verhältnis und die durchschnittliche Anzahl von Likes pro Tweet für jeden Tag.
-
Tauchen wir nun in den eigentlichen Inhalt von Tweets ein und wenden NLP-Techniken auf den Text an. Wenn wir einen Text lesen, konzentrieren wir uns gewöhnlich auf die Einheiten oder Themen, die diskutiert werden, welche Themen behandelt werden und die Meinung oder das Gefühl, das der Autor zu vermitteln versucht. Dies wird von unseren Gehirnen automatisch verarbeitet. NLP-Techniken ermöglichen es Maschinen, die gleiche Funktion nachzuahmen.
-
In Tweets erwähnte benannte Einheiten
Benannte Einheiten (NEs) sind eine Reihe von vordefinierten Kategorien, die in unstrukturiertem Text erwähnt werden, wie z. B. der Name einer Person, Organisationen, Orte, Zeitangaben, Mengen, Geldbeträge und Prozentsätze. Die Aufgabe, diese NEs zu identifizieren, vorherzusagen und zu extrahieren, wird als Named Entity Recognition (NER) bezeichnet. Um zu demonstrieren, wie dies funktioniert, stellen Sie sich vor, ich poste den folgenden Tweet: „Letzte Woche traf ich Tim an der Londoner Börse in London“. Wie auf dem nächsten Bild zu sehen ist, ist NER in der Lage, den Inhalt dieses Tweets in die in der nachstehenden Tabelle beschriebenen NEs einzuordnen.
Davon ausgehend kann ich beginnen, den eigentlichen Inhalt eines Tweets zu verstehen. Mit NER kann ich Merkmale erstellen, die beschreiben, welche Art von Themen diskutiert werden. Diese verschiedenen Einheiten-Typen können für ein bestimmtes Unternehmen unterschiedliche Auswirkungen haben.
TYP BESCHREIBUNG Datum Absolute oder relative Daten oder Zeiträume. Ereignis Genannte Hurrikane, Kämpfe, Kriege, Sportereignisse usw. GPE Geopolitische Einheit, d. h. Länder, Städte, Staaten. Ort Nicht-GPE-Orte, Bergketten, Gewässer. Geld Finanzielle Werte, einschließlich Einheit. NORP Nationalitäten oder religiöse oder politische Gruppen. Organisation Unternehmen, Agenturen, Institutionen usw. Prozentsatz Prozentsatz, einschließlich „%“. Produkt Gegenstände, Fahrzeuge, Lebensmittel usw. (Nicht Dienstleistungen.) Person Personen, einschließlich fiktiver Personen. Mit dem NER-Tagger von
spacy
(einem Python-Paket für NLP), werden insgesamt 20 Objekttypen aus unseren täglichen Tweets erkannt. Ich habe in der nachstehenden Tabelle einige gebräuchliche ausgewählt, die zum Experimentieren sinnvoll erscheinen. Für jeden Objekttyp sammle ich die Zählungen aus Tweets am selben Tag und erstelle 10 NE-Merkmale für jeden Tag. -
Themen / Cluster
Um zu erfassen, worüber die Öffentlichkeit an einem bestimmten Tag spricht, können Themenmodellierungen oder Clustering eingesetzt werden, die den Tweet-Strom in kleinere und spezifischere „Teilströme“ unterteilen. Aus diesen Unterströmen können kompliziertere Funktionen und Modelle erstellt werden. In dieser Kategorie ist allein die Erkennung von Finanzereignissen auf Twitter ein beliebtes und schnell wachsendes Forschungsgebiet. Hier verwende ich jedoch zur Demonstration eine einfachere Methode für die Extraktion von Merkmalen. Ohne allzu sehr ins Detail zu gehen, generiere ich für jeden Tweet sogenannte „Term Frequency Inverse Document Frequency“-Vektoren (tf-idf) und speise sie in
DBSCAN
ein, um die Anzahl der Cluster zu ermitteln. Für alle Tweets (mit Duplikation, d. h. Retweets) weise ich jeden Tweet seinem Cluster zu und berechne das Verhältnis dieser Tweets zu allen Tweets. Diese Merkmale beschreiben, wie viele Mainstream-Themen von der Twitter-Öffentlichkeit an einem bestimmten Tag geäußert werden und wie hoch der Anteil des Engagements zu diesen Themen ist. -
Stimmung
Es gibt eine Reihe von Programmen zur Erstellung der Stimmungsbewertung aus dem Text. Anstatt uns auf ein bestimmtes Programm festzulegen, sollten wir den Durchschnitt aus den Bewertungen zweier populärer Stimmungsanalyse-Programme,
afinn
undvader
, bilden. Jede dieser beiden Stimmungswerte wird auf den Bereich $[-1, 1]$ skaliert, wobei $0$ für eine neutrale Stimmung und $(0, 1]$ und $[-1, 0)$, für eine positive bzw. negative Stimmung stehen. Die Endwerte werden anhand der Länge des Tweets (Wortzahl) normalisiert und über die Tweets desselben Tages gemittelt.
-
Vorhersage: Wie verhalten sich die Funktionen?
Bisher habe ich aus Tweets, die ich für Unternehmen A gesammelt habe, 23 tägliche Features generiert, d.h. eine Matrix von 238 (Tage/Aufzeichnungen) x 23 (Messungen/Features). Lassen Sie uns für diese Studie einige Klassifizierungsziele definieren, die ich mit diesen Merkmalen vorherzusagen versuchen kann. Ich definiere vier binäre Klassifizierungskennzeichen, die angeben, ob der 1) angepasste Schlusskurs, 2) die tägliche Rendite, 3) das Handelsvolumen und 4) die tägliche Volatilität einer Aktie im Laufe des nächsten Tages um einen bestimmten Schwellenwert ansteigen werden. Mathematisch kann dies als Zeichen $\Delta Z$ bezeichnet werden. Mit diesen vier Zielen experimentiere ich mit einem einfachen binären Klassifikationsmodell.
Für alle verfügbaren Daten des Datensatzes (26. November 2019 bis 20. Juli 2020) berechne ich den gleitenden Mittelwert der Merkmale von Datum $t$ bis $t-n$ um störende Fluktuationen zu glätten, wobei die Anzahl der Tage $n$ ein Hyperparameter ist. Dieser Merkmalssatz ($X$) wird mit jedem der vier Zielkennzeichen für den nächsten Tag $t+1$ ($y$) als Datensatz für Klassifikationsexperimente gepaart. Die ersten sieben Monate des Datensatzes (Nov 2019 – Mai 2020) werden für das Training verwendet, die restlichen zwei Monate (Jun und Jul 2020) dienen zum Testen und Bewerten der Wirksamkeit der Merkmale. Ich wähle für diese Aufgabe ein einfaches logistisches Regressionsmodell (aus dem maschinellen Lernpaket scikit-learn), damit sich die Leistung mehr auf die Merkmale als auf das Lernmodell stützt.
Dieser Prozess wurde für jeweils $n$ bis zu $n=14$ Tage und jedes der vier Zielkennzeichnungen durchgeführt. Die obigen Bilder zeigen die Klassifikationsleistung über $n$, gemessen an drei Leistungskennzahlen: AUC (Fläche unter dem erhaltenen Operationsmerkmal), Klassifizierungsgenauigkeit und $f_1$ score. Aus diesen Ergebnissen stelle ich fest, dass die Leistung der Modelle bei größeren Fenstern im Allgemeinen besser ist, mit Ausnahme der Tagesrendite und der Volatilität.
Erläuterung: Welche Funktionen waren wichtig?
Um zu untersuchen, wie die Features zur Vorhersage beitragen, generiere ich die SHAP-Werte für jeden Klassifikator. SHAP ist ein spieltheoretischer Ansatz zum Verständnis der Auswirkungen, die jedes Merkmal auf ein Modell hat. Insgesamt wirken sich die Merkmale unterschiedlich auf die Schätzung unterschiedlicher Ziele aus. Es wird gezeigt, dass eine Reihe von Merkmalen mehr prädiktive Informationen enthalten als der Stimmungswert. Dieses Ergebnis zeigt uns, dass man viel mehr tun kann als, nur eine Stimmungsanalyse für diesen Datensatz durchzuführen.
Fazit
Auf der Grundlage der obigen Ausführungen legt unsere Analyse nahe, dass der Stimmungswert nicht unbedingt das wichtigste Merkmal für die Modellierung von Aktienkursen ist. Aus diesem Feature-Engineering-Experiment ist es mir gelungen, 23 Merkmale zu erstellen, die für die Bewertung der Performance von Aktien nützlich sein können. Die wichtigste Schlussfolgerung ist, dass es immer weitere Möglichkeiten gibt, unsere Prognosen zu verbessern, seien es zusätzliche Daten, Merkmale, Analysen oder Experimente.
Was kann uns Twitter also wirklich über Aktienkurse sagen? Die einfache Antwort ist viel mehr als nur eine Stimmungsanalyse, wie es die Literatur vorschlägt. Alternativ dazu sollten wir vielleicht auch eine andere Frage stellen: Was können wir mit dieser Erkenntnis noch tun und wie machen wir unsere Studie noch aussagekräftiger? Die Möglichkeiten sind endlos.
INSIGHTS Stellungnahmen
The COVID-19 crisis has posed a number of severe challenges for businesses, from reacting to the outbreak, preparing for a potential recession, anticipating an eventual rebound in demand and placing bets against the post-crisis landscape. But it also provides an opportunity for organizations to step back and assess their approach to strategy and their strategic capabilities. As the context in which businesses operate becomes more dynamic and unpredictable, driven by the pace of technological change and a high degree of interconnectedness, we should expect other shocks of a similar nature moving forward, whether the trigger comes from biological pathogens, cyberattacks, market crashes, or another sources. Some will be exogenous to the business world, but some will be endogenous.
How can companies be better prepared for when those shocks occur? We see 10 strategic lessons emerging from the current crisis:
1. Think in multi-level systems for highly interconnected problems. Companies are embedded in markets, which are embedded in highly interconnected economies and societies, which are in turn embedded in natural ecosystems and the biosphere. In such nested complex adaptive systems, disturbances often cascade up from lower to higher levels and back down again, reshaping the whole system in the process. COVID-19 started as a local outbreak, which many believed would be contained in China. However, by adopting a systems perspective, forward-thinking leaders, forewarned by prior epidemics, might have seen that there was a strong chance that it could spread globally — with cascading effects on economies, markets, and companies. Systemic risk is a common feature of highly interconnected systems, from banking systems to infectious diseases and beyond, begging a systemic approach.
2. Design and manage for resilience. During stable times, companies tend to compete by building scale and static efficiency. From this standpoint, redundant resources and capabilities could be seen as wasteful. However, redundancy is one of six characteristics that we have identified that increase the resilience of natural and social systems. The others are: heterogeneity (a diversity of perspectives and approaches), modularity (creating “firebreaks” to prevent whole system collapse), adaptability (the ability to flex designs in changing circumstances using a process of variation, selection and amplification), prudence (stress testing for plausible tail risks) and embeddedness (coherence with higher-level systems like society and nature). In the case of COVID-19, we see that organizations that failed to build resilience, such as governments that lacked reserves of critical health-care equipment or companies with distressed balance sheets, have much more difficulty responding to the crisis.
3. Create a sense of urgency and avoid complacency. Even after the coronavirus started spreading globally, many countries delayed ramping up testing and adopting mitigation measures, and as a result the outbreak spread and intensified. However, some countries that had experienced prior coronavirus outbreaks, such as Singapore (SARS) and Korea (MERS), were less complacent and intervened more rapidly and forcefully. Many companies found that while they had a “crisis response plan” on paper, they had never used or simulated it, so it was difficult to actually implement. To avoid complacency in future crises, leaders should instill a sense of vigilance and urgency in their organizations by, among other measures, conducting “war games” to create mental and physical preparedness for a range of possibilities.
4. Avoid communications gridlock. Once they eventually grasp the significance of a crisis, many organizations become hyperactive and gridlocked by incessant, ever-changing instructions cascading up and down the hierarchy. There are of course benefits to sharing information, but there are also costs: communication takes time away from the real work that needs to be done, over-communication drowns out critical messages, and the agility of the organization can be diminished. This is an example of a larger issue that pervades many organizations: new instructions, processes or structures get built but few are ever removed, so complexity tends to increase over time. In the case of crisis communications, leaders should employ the military principle of Commander’s Intent (auftragstaktik), focusing on communicating key facts, principles and objectives, and allowing the rest of the organization to decide on appropriate tactics, informed by the local circumstances they face at the time. Effective decision-making requires finding the right balance between leadership and collective action.
5. Match your strategy to your environment. Classical strategies that depend on deliberate, episodic planning cycles, such as those adopted by most governments and businesses, can be effective in stable environments. However, the COVID-19 crisis is extremely unpredictable and fast-moving, due to unknown characteristics of the virus and the exponential trajectory of infections. As a result, many organizations found they needed to use an adaptive strategy in rapidly changing circumstances, and to accelerate decision making accordingly. Furthermore, as the crisis recedes, they will need to adopt shaping and visionary strategies to influence and exploit the post-crisis landscape. Each style constitutes a distinct approach to problem solving and requires very different metrics, processes, capabilities and leadership styles. Companies increasingly need to choose the right approach to strategy at the right time for each part of their business. To effectively implement multiple approaches to strategy in this way, they will need to master the art of strategic ambidexterity.
6. Operate on multiple timescales simultaneously. In early stages of the crisis, companies focused primarily on reacting to immediate threats. However, the outbreak also surfaced other challenges and opportunities: preparing for a potential recession in the near-term, taking advantage of a potential rebound in the medium-term, and eventually reimagining offerings and business models for the post-crisis world. Many companies are learning that they cannot wait to tackle these challenges sequentially, because each requires significant mental and physical preparation — rather, they have to consider multiple timescales simultaneously to both survive and position themselves for the future. The clock speed of most organizations is determined by the annual planning cycle. But at a time when strategically relevant timescales are being stretched from the algorithmic (milliseconds) to the planetary (decades), they need to build the capability to operate at multiple clock speeds.
7. Compete on the rate of learning. The pace of history and strategy is not linear. Sometimes there are long stretches of time when little of strategic significance happens. At other times, every second counts. In a quickly changing, interconnected digital world, with access to the rapid pattern detection capabilities of machine learning, organizations can and need to learn faster. The need is further amplified in rapidly moving crises like COVID-19. Companies increasingly need to compete dynamically and cultivate the capability to learn rapidly. Traditional measures of an economy’s or company’s activity, such as GDP statistics and earnings reports, are far too infrequent and lagging to be of use during the COVID-19 crisis. To understand how activity is changing during the outbreak and see hints of the emerging post-crisis future, organizations need to access and analyze high frequency, granular data, such as daily information on specific products or sub-categories. Digital pioneers that have built massive data ecosystems, and the analytical capabilities to learn from them on a continuous basis, are advantaged in this regard. But it’s not just a technological matter: companies need to build hybrid learning organizations that combine human ingenuity and machine learning in order to compete effectively.
8. Find advantage in adversity. When the economic environment gets challenging, companies tend to act defensively. However, 14% of companies across sectors increase both growth and margins during downturns. As a result, competitive divergence increases during recessions. The companies that emerge stronger from a downturn tend to take a long-term perspective and treat the crisis as an opportunity. Benchmarking suggests that while the majority of companies are focused on reacting to and managing the current crisis, some are seeing and executing against emerging longer term opportunities. For example, Apple made several acquisitions during the heart of the crisis, leveraging its strong cash position to take advantage of depressed valuations.
9. Compete on imagination. There is no tried-and-tested playbook to tackle an unprecedented crisis like COVID-19. And in a strategic sense there could not be, since the impact on each company will be largely determined by its ingenuity and ability to see and shape new possibilities. History shows that new attitudes and patterns of demand emerge during a crisis, prompting many innovations, but the spoils accrue disproportionally to imaginative pioneers. Imagination is often the first victim of a crisis, but it is an essential ingredient for long-term success. Imagination is not, as is often believed, mainly about serendipitous individual inspiration — it can be developed and cultivated as an organizational capability in a structured manner.
10. Shape collaborative
solutions. COVID-19 is a global crisis that touches nearly all aspects of
society, and solving it will require a collaborative response across business
ecosystems, sectors and nations. Atomistic competition within current
governance and regulatory frameworks will not get the job done. When confronting
global issues, such as pandemics and climate change, governments and business
leaders alike will need to build and shape coalitions for collective action at
a global scale. This is necessary both for society as a whole to thrive, and to
preserve the viability and legitimacy of the canvas upon which business
operates. Business ties across nations create the empathy and connection that
can facilitate broader collaboration. In addition, businesses are highly
effective problem-solving entities and have a crucial role to play in tackling
our common problems. This agenda will, however, require innovations in how we
lead, collaborate, govern and collectively mobilize.
About the Authors
Martin Reeves is a managing director and senior partner in the San Francisco office of Boston Consulting Group and the Chairman of the BCG Henderson Institute.
Simon Levin is the James S. McDonnell Distinguished University Professor in Ecology and Evolutionary Biology at Princeton University.
Georg Kell is Chairman of Arabesque, an ESG Quant investment firm, and founder and former head of the UN Global Compact.
Kevin Whitaker is the head of strategic analytics at BCG Henderson Institute.
Saumeet Nanda is a consultant in the San Francisco office of Boston Consulting Group and an ambassador to the BCG Henderson institute.
Read the original article here.
INSIGHTS Stellungnahmen
Georg Kell explores why human wellbeing and the health of the planet are interconnected, and key lessons we can learn from the Coronavirus pandemic. This article was originally published in Forbes.
Neither military power nor wealth can stop the destructive global spread of COVID-19, a tiny member of the Coronavirus family. Its full human impact and economic cost will not be known for months to come. The virus is only now spreading amongst the most vulnerable populations, the millions who are cramped into refugee camps, and the hundreds of millions who live in city slums or in poverty without proper sanitation or medical support. As the pandemic is unfolding, it is revealing human vulnerabilities and showcasing the importance of good leadership and well-functioning, universal social and health care systems.
While the current focus is on responding to the pandemic and on coping with its immediate effects, the lessons we will collectively learn from this crisis are equally if not more important, as we know that the next global crisis – the climate crisis – is already well under way, building up its destructive potential around the globe. This is of particular relevance for the younger generations. They will inherit the political and economic systems that are now being reshaped in response to the COVID-19 pandemic and their future is being mortgaged with enormous debt as governments are mobilizing unprecedented stimuli packages to avoid a deep recession.
There are at least four lessons we should learn from the pandemic:
1) Human history and natural history can no longer be separated – human health and the health of the planet go together
“Mother nature is striking back, and humans are caught on their back feet,” is how a senior finance executive recently summed up the pandemic. Indeed, the pandemic should above all be a wakeup call that our wellbeing is closely tied to the health of the planet. Despite scientists’ warnings about the high risk of animal-borne infectious diseases, we continue to destroy natural habitats. The evidence of the destructive human impact on the natural environment from water to soil to the air, and its negative impact on human health and wellbeing, is overwhelming. Yet, we find it difficult to change course. Despite the many warning signs, humans have become a geophysical force as we continue to destroy, pollute and poison on a massive scale the very foundation we depend on for survival and wellbeing. Every year we dump over 30 billion tons of carbon into the atmosphere. We destroy entire animal and plant species at an alarming rate. We have cut down forests everywhere. We poison the soil and the water, and our garbage covers the floors of the oceans. And yes, every year, we kill over 100 billion animals to feed our carnivorous appetites. Our industrial-era mindset of ‘growth at any cost’ has become a recipe for self-destruction.
We have long known that markets cannot succeed in failing societies. Now we must learn that healthy societies and markets depend on the health of the natural environment. We know in principle what needs to be done and we have the means to do it: shifting the goalposts and the incentives that put a premium on clean and healthy growth instead of subsidizing the destruction of the environment, and putting decarbonization of economic activity and material reuse center stage, while restoring natural habitats and forests. With the enormous stimuli packages now being rolled out, we have a unique opportunity to change course. We can respond to the current crisis while at the same time building a healthier, greener and safer future. Going back to business as usual would be short-sighted and self-destructive – we would respond to one crisis by fueling the next one. Green and inclusive growth is no longer a nice thing to have. It is the only way to prevent the next crisis which, according to scientists, could well turn out to have an even greater destructive impact than COVID-19.
2) Prevention is better than cure
– we must learn to listen to science
The pandemic is a strong reminder that ignoring science carries steep costs. Scientists have long warned about infectious diseases, especially since the recent outbreaks of Ebola, SARS and the bird flu. Only last September did the WHO’s Global Preparedness Monitoring Board issue an authoritative report urging governments to better prepare. Alas, their call was widely ignored. Scientists’ consistent and overwhelming warnings about the human impact on the global climate, on soil and water, and on plant and animal diversity have equally been ignored.
In our era of deliberate misinformation, fake social news and divisive political propaganda, we now have an opportunity to rediscover science as a reliable arbitrator and a guide to informed decision making. COVID-19 has already helped to elevate trust in credible mainstream scientists and thereby reversed the decade-old trend of eroding trust in established institutions. According to a recent New York Times article, the voice of science has become indispensable to inform policymakers and the public on how to deal with the pandemic. The fact that scientists collaborate across borders, even in the light of political divides, adds further credentials to their efforts, and their pursuit of the public good often trumps narrow interests. Moving forward, scientists should be encouraged to play a greater role in public debates and policymaking and the public should be encouraged to take a greater interest in science and support their work.
3) Global threats need global collaboration
Former UN Secretary General Kofi Annan referred to climate change, diseases and terrorism as ‘problems without passports’ that cannot be stopped at the border and that can only be tackled if we cooperate. The premium for cooperation is growing as global threats are becoming national security threats. A virus can’t be stopped at borders and climate change does not respect national sovereignty. Yet, sadly, international cooperation has given way to strategic rivalry and fragmenting power blocks. The Paris Agreement of 2015, which had China’s and the USA’s cooperation, seems like from another era. The United Nations’ recent call for cooperation to deal with the pandemic was hardly noticed and its most important organ, the Security Council, is “missing in action”. Some political observers fear that the pandemic may accelerate existing divergences, supercharging nationalism and undermining free trade even further, leading to a world in even greater disarray. Indeed, ancient power concepts and myths of the past still set the tone and have not evolved since Thucydides, despite the growing interdependence of humanity and the many lessons history has taught us. Policymakers seem to have no grasp of “global public goods” and their importance for national security.
Over time nature will force our hands, whether we are prepared or not. The idea of a “common enemy” may sound outlandish to many at this point, but COVID-19 demonstrates that conventional power concepts are no longer useful when dealing with global threats. The notion of “global stewardship” and the imperative to build stronger collaborative bonds across and between nations will ultimately become a necessity. Human security needs more than military deterrence: it needs a new focus on good stewardship for the life-supporting services that nature provides and a collective willingness to improve the state of affairs everywhere. No country is prepared for the next pandemic if the rest of the world is not. And no country can stop the impact of climate change alone. COVID-19 gives us an opportunity to change course in this direction. Heeding UN Secretary General Antonio Guterres’ call to better coordinate and mobilize efforts to deal with the impact of the pandemic in less developed countries would be a good first step for the better.
4) The pivotal role of the private sector
Commerce has long acted as a bridge-builder between nations by connecting cultures and people, not through military power, but by spreading knowledge, mutual understanding and economic benefits. The world will remain deeply interconnected also in the post-COVID-19 era, and economic interdependence will remain the most viable pathway to secure peaceful co-existence and prosperity. In a fragmenting world where policymakers seem to have forgotten the lessons of history and largely ignore common interests that are shared by all of humanity, the private sector is playing an ever more critical role.
While coping with the pandemic in a struggle to survive, many well-managed companies have put the health and safety of their workers first while cooperating and working with clients and customers across borders and along supply chains. Many companies have retooled their manufacturing capacities to supply medical supplies and have mobilized community-based efforts to cope with the pandemic. One such example is Volkswagen, which in late January already donated medical supplies to Hubei Province of China. China in turn is now providing massive supplies to Germany. Leading technology companies are collaborating to develop “contact apps” and pharmaceutical companies have initiated unprecedented cooperation in the race to develop vaccines.
Good corporate citizenship practices are much needed to complement government efforts. The idea and practice of corporate sustainability and its financial equivalent – sustainable investing – is now playing an important role in coping with the crisis, affirming once again that values and purpose are the enduring features of resilient organizations. Equally, if not more important, is the fact that corporate sustainability and responsible investing now also serve as counterweights to the dark forces of economic nationalism and protectionism.
The sustainability movement will gain further relevance in the post-COVID-19 era. For corporations and investors, the need to align strategies with a broader purpose that speaks to the needs of society will be the key to growing and building trust. The pandemic has put a spotlight on human vulnerability and the fact that human safety and the health of the natural environment go hand in hand. This may well reinforce existing consumer trends towards healthier and more sustainable lifestyles. Environmental priorities are bound to gain greater strategic relevance over time and far-sighted executives will use the current crisis to accelerate decarbonization and to resource-efficiency measures. And arguably most important of all, the pandemic has acted as an accelerator for everything digital. Innovation and new business models will enjoy a premium and will give a boost to automation, resource efficiency and decarbonization, touching all segments of the economy. Moreover, digitalization and better smart data analysis are the fuel that drives ESG investing. Early evidence is already suggesting that ESG investing is gaining greater relevance in the light of the pandemic. Moving forward, the convergence between corporate sustainability and sustainable investing offers unprecedented opportunities to renew markets from within.
Are we capable of learning?
While coping with the crisis, we have an opportunity to rediscover basic values of humanity and the bonds that connect us. We now have it in our hands to lay the foundation for a safer, healthier and cleaner life on planet earth. We have the technology and the means to come out stronger if we understand that human wellbeing and the health of the planet are two sides of the same coin. Now is the time to retire old dogmas and to give way to a fresh start.
Georg Kell is Chairman of Arabesque, and the founding Executive Director of the UN Global Compact.
INSIGHTS Stellungnahmen
Von Georg Kell
Verantwortungsbewusstes Investieren wird weithin als Integration von Umwelt-, Sozial- und Governance-Faktoren (ESG-Faktoren) in Investitionsprozesse und Entscheidungsfindung verstanden. ESG-Faktoren decken ein breites Spektrum von Themen ab, die traditionell nicht Teil der Finanzanalyse sind. Sie können aber dennoch finanzielle Relevanz haben können. Dazu kann gehören, wie Unternehmen auf den Klimawandel reagieren, wie gut sie Wassermanagement betreiben, wie wirksam ihre Gesundheits- und Sicherheitspolitik beim Schutz vor Unfällen ist, wie sie ihre Lieferketten managen, wie sie ihre Mitarbeiter behandeln und ob sie eine Unternehmenskultur haben, die Vertrauen schafft und Innovationen fördert.
Der Begriff ESG wurde erstmals 2005 in einer wegweisenden Studie mit dem Titel „Who Cares Wins“ geprägt. Heute werden die ESG-Investitionen auf über 20 Billionen Dollar in AUM oder rund ein Viertel aller professionell verwalteten Vermögen weltweit geschätzt. Ihr rasches Wachstum baut auf der Bewegung für sozial verantwortliche Investitionen (Socially Responsible Investment, SRI) auf, die es schon viel länger gibt. Aber im Gegensatz zu SRI, das auf ethischen und moralischen Kriterien basiert und meist Negativ-Raster verwendet, wie z. B. nicht in Alkohol, Tabak oder Schusswaffen zu investieren, basieren ESG-Investitionen auf der Annahme, dass ESG-Faktoren finanziell relevant sind. Im Jahr 2018 haben Tausende von Fachleuten aus der ganzen Welt die Berufsbezeichnung „ESG-Analyst“ inne, und ESG-Investitionen sind Gegenstand von Nachrichtenartikeln der Finanzseiten weltweit führender Zeitungen.
Viele Anleger erkennen, dass ESG-Informationen über Unternehmen unerlässlich sind, um den Unternehmenszweck, die Strategie und die Managementqualität von Unternehmen zu verstehen. Aber was erklärt den bemerkenswerten Anstieg von ESG-Investitionen und was bedeutet dies für die Zukunft?
Die Geschichte der ESG-Investitionen begann im Januar 2004, als der ehemalige UN-Generalsekretär Kofi Annan mehr als 50 CEOs großer Finanzinstitute anschrieb und sie einlud, an einer gemeinsamen Initiative unter der Schirmherrschaft des UN Global Compact mit Unterstützung der Internationalen Finanz-Corporation (IFC) und der Schweizer Regierung teilzunehmen. Ziel der Initiative war es, Wege zur Integration von ESG in die Kapitalmärkte zu finden. Ein Jahr später legte diese Initiative einen Bericht mit dem Titel „Who Cares Wins“ vor, dessen Autor Ivo Knoepfel war. Der Bericht machte deutlich, dass die Einbettung von Umwelt-, Sozial- und Governance-Faktoren in die Kapitalmärkte wirtschaftlich sinnvoll ist und zu nachhaltigeren Märkten und besseren Ergebnissen für die Gesellschaft führt. Gleichzeitig erstellte UNEP/Fi den sogenannten „Freshfield-Bericht“, der zeigte, dass ESG-Fragen für die finanzielle Bewertung relevant sind. Diese beiden Berichte bildeten das Rückgrat für die Einführung der Grundsätze für verantwortungsbewusstes Investment (Principles for Responsible Investment – PRI) an der New Yorker Börse im Jahr 2006 und die Einführung der Sustainable Stock Exchange Initiative (SSEI) im darauf folgenden Jahr.
Heute ist die von den Vereinten Nationen unterstützte PRI eine blühende globale Initiative mit über 1.600 Mitgliedern und über 70 Billionen Dollar an verwalteten Vermögenswerten repräsentiert. Die Rolle der PRI besteht darin, die Integration von ESG in die Analyse und Entscheidungsfindung durch Thought Leadership und die Schaffung von Instrumenten, Anleitung und Engagement voranzutreiben. Der SSEI, der von der in Genf ansässigen UNCTAD unterstützt wird, ist im Laufe der Jahre gewachsen, und viele Börsen schreiben nun die Offenlegung von ESG für börsennotierte Unternehmen vor oder bieten Leitlinien für die Berichterstattung über ESG-Themen an. Trotz seines raschen Wachstums in den Mainstream ist der Anstieg der ESG-Investitionen jedoch weder glatt noch linear verlaufen.
Institutionelle Investoren zögerten anfangs, sich dieses Konzept zu eigen zu machen, da sie argumentierten, ihre Treuepflicht beschränke sich auf die Maximierung der Aktionärswerte, unabhängig von den Auswirkungen auf die Umwelt oder die Gesellschaft oder allgemeineren Governance-Fragen wie Korruption. Unglaublicherweise werden solche Argumente immer noch angeführt. Doch da sich die Beweise dafür mehren, dass ESG-Themen finanzielle Auswirkungen haben, hat sich das Blatt gewendet. In vielen wichtigen Märkten, darunter die USA und die EU, wird die ESG-Integration zunehmend als Teil der treuhänderischen Pflicht angesehen, siehe z. B. Al Gores Update zu relevanten Entwicklungen.
Ein weiteres großes Hindernis war der Mangel an Daten und den notwendigen Instrumenten, um die zersplitterten und unvollständigen Informationen in den Griff zu bekommen. Die Offenlegung von ESG-Themen durch Unternehmen hat sich jedoch seit der Einführung der Global Reporting Initiative (GRI) im Jahr 2000 stetig verbessert. Heute verwenden 80 % der weltweit größten Unternehmen GRI-Standards. In jüngerer Zeit haben die International Integrated Reporting Initiative (IIRC) und das in den USA ansässige Sustainability Accounting Standard Board (SASB) dazu beigetragen, die branchenspezifische Berichterstattung und ihre Relevanz für Investoren voranzubringen. Insgesamt ist der Markt für ESG-Informationen reifer geworden, und die Qualität, auch wenn sie noch unvollkommen ist, wird immer besser. Neue Technologien, die auf maschinellem Lernen und großen Datenmengen basieren, können bereits wertvolle Erkenntnisse liefern und einfache Möglichkeiten bieten, ESG-Daten zusätzlich zu herkömmlichen Finanzinformationen anzuwenden.
Das stetige Wachstum der ESG-Investitionen wurde um 2013 und 2014 stark beschleunigt, als die ersten Studien veröffentlicht wurden, die zeigten, dass eine gute Nachhaltigkeitsperformance von Unternehmen mit guten finanziellen Ergebnissen verbunden ist. Arbeiten von Akademikern wie George Serafeim, Bob Eccles und Ioannis Ioannou zeigen die Bedeutung von ESG-Informationen für die Bewertung von Unternehmensrisiken, Strategien und operativer Leistung.
Die Idee, dass Investoren, die Umwelt-, Sozial- und Governance-Risiken von Unternehmen integrieren, die Renditen verbessern können, verbreitete sich nun rasch auf den Kapitalmärkten aller Kontinente. In Europa zum Beispiel hat eine kritische Masse von Pensionsfonds und Versicherern damit begonnen, ihr Neugeschäft ausschließlich an Vermögensverwalter mit ESG-Fähigkeiten zu vergeben. Die globale Anlegergemeinschaft hat eine Vielzahl von Methoden entwickelt, um ESG-Informationen optimal zu integrieren, wie z. B. in A Practical Guide To ESG Integration for Equity Investing. Zu den vielen ESG-Faktoren, die als finanziell relevant angesehen werden, gehören insbesondere solche, die mit dem Klimawandel zusammenhängen. Der Grund dafür ist, dass der Klimawandel nicht mehr eine ferne Bedrohung am Horizont ist, sondern eine Bedrohung, die hier und jetzt besteht und wirtschaftliche Folgen in Milliardenhöhe hat. Viele Investoreninitiativen drängen nun auf eine Dekarbonisierung und die Task Force on Climate-related Financial Disclosures (TCFD) hat viele Anstöße zur Verbesserung der Risikobereitschaft und damit auch zu Maßnahmen der Dekarbonisierung gegeben.
Zyniker mögen argumentieren, dass verantwortungsbewusstes Investieren nur eine Modeerscheinung sei. Aber ein genauerer Blick auf die Kräfte, die die Bewegung in den letzten 15 Jahren vorangetrieben haben, legt etwas anderes nahe. Erstens: Die Technologie und die Zunahme der Transparenz sind von Dauer. Das Sammeln und Verarbeiten von Daten wird immer einfacher und billiger werden. Intelligente Algorithmen werden zunehmend eine bessere Interpretation nicht-traditioneller Finanzinformationen ermöglichen, deren Volumen sich alle paar Jahre zu verdoppeln scheint. Zweitens werden Umweltveränderungen, insbesondere der Klimawandel, mit wissenschaftlicher Gewissheit einen wachsenden Wert auf gute Verantwortung und kohlenstoffarme Praktiken legen, da natürliche Ressourcen im Laufe der Zeit an Wert gewinnen werden. Und drittens werden die Menschen überall auf der Welt immer mehr durch die Technologie gestärkt. ESG-Investitionen ermöglichen es ihnen, ihre eigenen Werte zum Ausdruck zu bringen und sicherzustellen, dass ihre Ersparnisse und Investitionen ihre Präferenzen widerspiegeln, ohne Kompromisse bei der Rendite einzugehen.
Der Anstieg von ESG-Investitionen kann auch als ein Indikator dafür verstanden werden, wie sich Märkte und Gesellschaften verändern und wie sich die Bewertungskonzepte an diese Veränderungen anpassen. Die große Herausforderung für die meisten Unternehmen besteht darin, sich an ein neues Umfeld anzupassen, das intelligentere, sauberere und gesündere Produkte und Dienstleistungen begünstigt, und die Dogmen des Industriezeitalters hinter sich zu lassen, als Umweltverschmutzung kostenlos war, Arbeit nur ein Kostenfaktor und Größe und Umfang die vorherrschende Strategie. Für Investoren werden ESG-Daten immer wichtiger, um diejenigen Unternehmen zu identifizieren, die für die Zukunft gut positioniert sind, und um diejenigen zu meiden, die wahrscheinlich unterdurchschnittlich abschneiden oder scheitern werden. Für Einzelpersonen bieten ESG-Investitionen die Möglichkeit, mit ihrem Geld eine Wahl zu treffen. Und für politische Entscheidungsträger sollte es eine willkommene marktgeführte Entwicklung sein, die sicherstellt, dass das Gemeinwohl nicht um jeden Preis in kurzfristigen Profiten verloren geht. Heute sind ESG-Investitionen so weit gereift, dass sie die Markttransformation zum Besseren erheblich beschleunigen können. In dem Maße, wie Unternehmen und Investoren wachsenden Einfluss und Macht erfahren, prägen ihre Handlungen und Entscheidungen zunehmend die Zukunft. Unter der Voraussetzung, dass sich die auf Offenheit und globalen Regeln basierenden politischen Rahmenbedingungen nicht weiter verschlechtern, werden marktgetriebene Veränderungen in wirklich großem Umfang als eine Kraft des Guten wirken.
INSIGHTS Stellungnahmen
In diesem Webcast sprechen Dr. Rebecca Thomas und Emily Matthews von Arabesque S-Ray über die aktuellen Herausforderungen und Chancen im Zusammenhang mit Kohlenstoffdaten und deren Relevanz für die Investorengemeinschaft.
In den letzten Jahren haben Unternehmen aufgrund der gestiegenen Berichtspflicht vermehrt über den Klimawandel berichtet. Aber viele Unternehmen melden immer noch nicht ihre Emissionen. Trotzdem integrieren Investoren Klimapunktzahlen in ihre Investitionsentscheidungen. Dies geschieht durch Schätzungen von Emissionen anhand von Modellen; das aktuelle Klima-Rating-System bietet Metriken für mehr als 10.000 Unternehmen.
Diese Strategie hindert allerdings Investoren daran, ein genaues Risikomanagement zu betreiben, da es keine Möglichkeit gibt, zwischen den Unternehmen zu unterscheiden, die berichten und eine Führungsrolle beim Klimaschutz übernehmen, und denen, die dies nicht tun. Folglich gibt es für nicht berichtende Unternehmen mit hohen Emissionen keinen Anreiz zur Berichterstattung. Darüber hinaus deuten unzureichende Offenlegungen darauf hin, dass die Unternehmen möglicherweise nicht vollständig über ihre Auswirkungen Bescheid wissen. Es ist auch möglich, dass Unternehmen Maßnahmen ergreifen, um ihre Auswirkungen auf den Klimawandel zu reduzieren. Dies kann aber ohne eine öffentliche Prüfung der Daten nicht verifiziert werden.
Laden Sie den Webcast über FactSet herunter, um mehr zu erfahren (link)
INSIGHTS Stellungnahmen
Speech given by Mark Carney, Governor of the Bank of England
European Commission Conference: A global approach to sustainable finance, 21st March 2019
A New Horizon?
A few years ago, I spoke of the Tragedy of the Horizon – how the catastrophic impacts of climate change will be felt beyond the traditional horizons of most banks, investors and financial policymakers, imposing costs on future generations that the current one has no direct incentives to fix.1 Once climate change becomes a clear and present danger to financial stability it could already be too late to stabilise the atmosphere at two degrees.
The paradox is that risks will ultimately be minimised if the transition to a low-carbon economy begins early and follows a predictable path. But for markets to anticipate and smooth the transition to a 2-degree world, they need the right information, proper risk management, and coherent, credible public policy frameworks.
Today, catalysed by the COP21 Paris Agreement, and national policies such as the UK Government’s Clean Growth Strategy, some of these elements are coming into place, creating a potential path to break the Tragedy of the Horizon. But the task is large, the window of opportunity is short, and the stakes are existential.
In pursuit of that New Horizon, let me briefly discuss progress and prospects in three critical areas – reporting, risk and return.
First, reporting
Three years ago in response to a call from G20 leaders, the FSB began addressing the financial stability risks associated with climate change by ensuring the market had the right information to price climate risk and reward climate innovation. The FSB established the Task Force on Climate-Related Financial Disclosures (TCFD) led by businesses from a wide range of industries across the G20. Eighteen months later, the TCFD delivered to the Hamburg G20 Leaders Summit its recommendations for voluntary disclosures of material climate-related financial risks. Since then there has been a step change in both demand and supply of climate reporting.
On the demand side, current supporters of the TCFD include three-quarters of the world’s globally systemic banks, 8 of the top 10 global asset managers, the world’s leading pension funds and insurers, major credit rating agencies and the Big Four accounting firms.2 In total, these financial firms manage almost US$110 trillion in assets.
As a consequence, the incentives for companies to disclose and manage climate-related risks have increased dramatically. Moreover, climate change claimed its first S&P 500 bankruptcy last year,3 climate related shareholder resolutions spiked to 90 last year,4 investment managers controlling over 45% of global assets under management now back shareholder actions on carbon disclosure, and companies representing over 90% of all shareholder advisory services now support the TCFD.
Not surprisingly, the supply of disclosure is responding. Over 600 organisations, with a total market capitalisation of US$9 trillion, have endorsed the TCFD recommendations since 2017.
The TCFD’s September 2018 Implementation Report assessed, using artificial intelligence, some 1800 companies, and analysed in detail an additional 200 of the largest companies, drawn from eight representative sectors from across the G20.5
In both cohorts, the majority of companies were already disclosing information in their 2017 filings that aligned with one or more of the TCFD’s recommendations. This is commendable given companies only had six months to respond to the final TCFD recommendations, but more progress is needed.
In particular:
Financial implications are often not yet disclosed;
Disclosures are often in multiple reports making comparisons harder; and
Disclosure varies by industry and region, with higher percentages of European firms and higher shares of those on the climate frontline – such as the energy sector – disclosing more information aligned with the recommendations.
The next milestone will be the TCFD implementation report for the G20 Leaders Summit in Osaka, which should set out:
The growing momentum behind disclosure;
The types of disclosures that are most decision-useful for investors; and
Best practice examples, including examples of scenario analysis so that firms can test their strategic resilience to different climate outcomes.
The momentum behind TCFD’s voluntary disclosure is creating a virtuous circle by encouraging learning by doing. As companies apply the recommendations and investors increasingly differentiate between firms based on this information, adoption will continue to spread, disclosure will become more decision-useful and efficient, and its impact will grow.
As firms work to enhance their disclosures, they are being supported by various TCFD Preparers’ Forums from energy to finance.6 The TCFD will also continue to work with market participants to refine metrics so that they are consistent, comparable and decision-useful; and it will share best practices on the disclosure of risk management and governance.
In the future, disclosure will move into the mainstream, and it is reasonable to expect that more authorities will mandate it. IOSCO could play a constructive role in coordinating such mandates and in any event, the current iterative process of disclosure, reaction and adjustment will be critical to ensure that these eventual market standards are as comparable, efficient and effective as possible.
Second, risk analysis
The second step on the path to a new horizon is better climate change risk management.
Climate change creates both physical and transition risks.7
Physical risks arise from the increased frequency and severity of climate- and weather-related events that damage property and disrupt trade.
Transition risks result from the adjustment towards a lower-carbon economy. Changes in policies, technologies and physical risks will prompt a reassessment of the value of a large range of assets as costs and opportunities become apparent. The longer meaningful adjustment is delayed, the more transition risks will rise.
Climate risks also have a number of distinctive elements, which, in combination, require a strategic approach. These include their:
Breadth, as climate risks affect multiple lines of business, sectors and geographies;
Magnitude, as the full impacts of climate risks are large, potentially non-linear and irreversible;
Foreseeable nature;
Dependency on short-term actions given that the size of future impacts will, at least in part, be determined by the actions taken today; and
Uncertain time horizon which may stretch beyond traditional business planning cycles.
The nature of these risks means that the biggest challenge in climate risk management is in assessing the resilience of firms’ strategies to transition risks.
Part of the genius of the private sector-led TCFD is its recognition that disclosure needs to go beyond the static to the strategic. Markets need information to assess which companies can seize the opportunities in a low carbon economy and which are strategically resilient to the physical and transition risks associated with climate change.
The Bank of England has also become increasingly active in such assessments, consistent with our financial stability and prudential mandates.
As the supervisor of the world’s fourth largest insurance industry, we know that general insurers and reinsurers are on the front line of managing the physical risks from climate change. Insurers have responded by developing their modelling and forecasting capabilities, improving exposure management, and adapting coverage and pricing.8 In the process, insurers have learned that yesterday’s tail risk is closer to today’s central scenario.
Sadly with climate, history repeats not as a farce but as tragedy and with growing frequency.
For banks, the financial risks from climate change have tended to be beyond their planning horizons. The PRA’s survey of 90% of the UK banking sector, representing over $11trn of assets, found that these horizons averaged four years – in other words, before risks would be expected to be fully realised and prior to ambitious climate policies taking effect. 9
That notwithstanding, the PRA’s latest survey finds that almost three quarters of banks are starting to treat the risks from climate change like other financial risks – rather than viewing them simply as a corporate social responsibility issue.
Banks have begun considering the most immediate physical risks to their business models – from the exposure of mortgage books to flood risk, to the impact of extreme weather events on sovereign risk. And they have started to assess exposures to transition risks in anticipation of climate action. This includes exposures to carbon-intensive sectors, consumer loans secured on diesel vehicles, and buy-to-let lending given new energy efficiency requirements.
Informed by these findings, the PRA will soon publish its final Supervisory Statement for banks, insurers and investment firms.10 This statement will set out the PRA’s expectations regarding firms’ approaches to managing the financial risks from climate change, including with respect to:
Governance, where firms will be expected to embed fully the consideration of climate risks into governance frameworks, including at board level, and assign responsibility for oversight of these risks to specific senior role holders;
Risk management, where firms will need to consider climate change in line with their board approved risk appetites;
The regular use of scenario analysis to test strategic resilience; and
Developing and maintaining an appropriate disclosure of climate risks.
Recognising the need for industry to build capacity and to develop best practices, the PRA has established a Climate Financial Risk Forum, jointly with the FCA, to work with firms from across the financial system.11
The responses to our supervisory consultation reflect the urgency and significance of the issues. Perhaps for the first time in financial regulation, firms are both thanking their supervisors for raising an issue and pushing us to go further; with some asking for more prescriptive recommendations and others for mandatory disclosures.12
Certainly, while climate risk management is improving, there is more to do particularly when assessing strategic resilience.
For companies, that means conducting scenario analysis.
The TCFD 2018 Status Report found that non-financial industries (energy, transport, building and agriculture) were the most advanced at measuring strategic resilience, including some examples of scenario analysis.13,14
The TCFD review found that the financial sector is also moving toward enhanced strategic analysis. For example half of all insurance companies reviewed used the 2°C scenario, and the majority of banks described the potential impact of climate-related issues on their businesses.
However, the September TCFD report showed that while firms were starting to consider strategic resilience, few systematically conducted scenario analysis.
Indeed, the PRA has found that despite the sophistication of insurers in modelling climate risks, there are still gaps in their own risk management. The PRA is increasingly focused on cognitive dissonance in some insurers whose careful management of climate risks on the liability side of their balance sheets is not always matched by similar considerations on the asset side.
And the PRA’s banking survey last September found that, although almost three quarters of banks recognised the risks of climate change, only one in ten were taking a long term, strategic approach to them.
With that in mind, we expect firms to consider scenario analysis as part of their assessments of the impact of climate risks on their balance sheet and broader business strategy.
An important question is the form these scenarios should take. Climate scenarios aren’t forecasts, but data driven narratives that help companies think through different possible futures. The scenarios should be comprehensive, rigorous and challenging. The assumptions and methodologies in the models – such as the assumed global temperature rise, the energy mix, or whether the transition happens smoothly or abruptly – should be sufficiently transparent to allow for comparisons and external challenge. And finally, scenarios should be implemented consistently across the business, linking identification of risks and opportunities to both strategy and disclosure.
To do this, firms will need either to develop their own transition scenarios or build on commonly available models. The TCFD report signposts existing models that firms can use, and the PRA’s Climate Financial Risk Forum will work with industry to review tools and metrics, with the view to publishing reference scenarios and standard assumptions.15
For supervisors, assessing strategic resilience will require climate-related stress testing. This involves linking high-level data-driven narratives on the evolution of physical and transition risks to quantitative metrics to measure the impact on the financial system.
Next month, the PRA will ask UK insurers, as part of a market-wide insurance stress test, to consider how their businesses would be affected in different physical and transition risks scenarios.
Testing the banks, and possibly other participants in the financial system, with climate-change scenario stress tests would have two objectives:
To consider whether, across the financial system, financing flows are consistent with an orderly transition to the climate outcome set out in the Paris agreement. These long-term scenarios can facilitate discussions between firms and their clients about possible risks across different sectors and geographies; and
To consider whether the financial system would be resilient to shorter-term shocks – including a climate “Minsky moment” when climate risks materialise suddenly.These long and short-term risks are, of course, linked – any overall misalignment with climate goals increases the short-term risks from a disorderly transition, possibly caused by extreme weather events or abrupt shifts in climate policy. A system-wide stress test can help supervisors and climate policymakers judge the adequacy of the current transition and whether further actions could be expected.
As the Bank of England considers the timing and design of such a stress test, we are working with colleagues in the Network for Greening the Financial System (NGFS) to develop a small number of highlevel scenarios.16 And in our Climate Financial Risk Forum we will work with banks, insurers and asset managers to ensure these scenarios are rolled out effectively within their organisations. Together with our work on this year’s insurance survey, these initiatives will provide a basis for our future assessments of the system-wide exposure to climate risks.
The third and final area is return
A new horizon brings new opportunities.
The IEA estimates that the low-carbon transition could require $3.5trn in energy sector investments every year for decades – twice the rate at present. Under their scenario, in order for carbon to stabilise by 2050, nearly 95% of electricity supply will need to be low carbon, 70% of new cars electric, and the CO2 intensity of the building sector will need to fall by 80%.
With an estimated US$90 trillion of infrastructure investment expected between 2015 and 2030, smart decisions now can make sure that investment is both financially rewarding and environmentally sustainable. 16 The voluntary network was set up by 8 central banks and supervisors in December 2017 at the One Planet Summit, and has since grown to 29 members, representing countries accounting for nearly half of global emissions, and five observers. It is a voluntary, consensus-based forum whose purpose is to share best practices, contribute to the development of climate- and environment-related risk management in the financial sector and mobilize mainstream finance to support the transition toward a sustainable economy. The analytical work is split into three work streams and the research will be published in April 2019: WS1 microprudential/supervisory; WS2 macrofinancial; and WS3 Scaling up green finance.
Regulators and market participants are collaborating to facilitate cross-border investments in green infrastructure. The European Commission’s Sustainable Finance Action Plan is developing a classification system for sustainable economic activities, a harmonised green bond standard and methodologies for lowcarbon indices.17 The three major credit rating agencies have all integrated environmental risk and green certification into credit ratings. And international organisations such as the Climate Bonds Initiative (CBI) and International Capital Markets Association (ICMA) have developed definitional frameworks, certification and validation methods for green financing.18
This work is helping the green bond market to gather pace, with issuance quadrupling from $45bn in 2015 to $168bn in 2018.19 Last year also saw inaugural sovereign green issues from five countries.20
For investors, green bond markets offer stable, rated and liquid investments with long duration. For issuers, green bonds are a way to tap the huge US$100 trillion pool of patient private capital managed by global institutional fixed-income investors. The shift to the capital markets from banks will also free up limited bank balance sheet capacity for early-stage project financing and infrastructure lending.
Over the last two years, the City of London has been glowing green with sixteen renewable infrastructure funds with a value of $7bn listed on the LSE. The City has been the centre of a series of landmark global green bond issuances, from China’s first Green Covered Bond – the country’s first ever international issuance of a green bond – to the first green Masala Bond worth INR 20bn. In our view, such local currency green bonds will be particularly important to the climate transition in emerging market economies (EMEs).
However, while they are important catalysts, green bonds will not be sufficient to finance the transition to a low carbon future. They accounted for only 3% of global bond issuance in 2018.
Achieving the transition will require mobilising mainstream finance.
Advances in reporting and risk analysis are paving the way for investors to realise the opportunities in climate-friendly investment by re-orienting their focus to broader, more sustainable long-term value creation.
Such investment approaches are becoming increasingly common. There are now almost 2000 signatories, with over $80 trillion in assets under management, to the UN Principles for Responsible Investment (UN PRI), an international network of investors committed to considering ESG factors in their work.21
This swell of support is driven by the expectation that sustainable investment can generate excess returns in three ways.
First, companies that score well on ESG metrics could better anticipate future climate-related risks and opportunities. This makes them more strategically resilient and therefore able to anticipate, and adapt to, the risks and opportunities on the horizon, generating true alpha from ESG.
Second, strong ESG scores could signal that a firm is more naturally disposed to longer-term strategic thinking and planning. Climate disclosure is increasingly seen not only as necessary in and of itself, but also as informative about the extent to which companies are focused on long-term value creation.
And third, strong ESG firms may enjoy valuation premiums consistent with shifting investor preferences. Millennials, keenly focused on company values and sustainability, are set to inherit $24trn of wealth in the US alone over the next 15 years and will seek the investment opportunities to match.22 Already, assets are moving to ESG strategies at 20 per cent annual growth.23
A review of over 200 sources on ESG performance by Oxford University and Arabesque showed that in the overwhelming majority (88%) of companies that focused on sustainability, operational performance was improved, translating to higher cash flows.24
And meta-analysis of over 2000 studies confirms that the responsible, as well as the economic case for ESG investment is tangible. 90% of studies find that there is no penalty on return on ESG investment, and the majority suggest that focusing on ESG criteria generates a positive return.25
The outperformance of strong ESG companies is uncorrelated with underlying factors such as return on equity or capital employed, and reflects greater earnings stability and lower share price volatility. While “screening” – excluding poor ESG performers – is still the most common tool among investors, some research finds that a more proactive consideration of ESG factors may pay off. 26,27, 28
“Tilt” strategies, which overweight ESG stocks, and “momentum” strategies, which focus on companies that have improved their ESG rating, have outperformed global benchmarks for close to a decade.29
This suggests that there is more to well-regarded ESG companies than simply better management of downside risk.
Given this growing track record, companies are developing ways to better score ESG performance and invest accordingly. For example, Arabesque uses machine learning models to assess the performance and sustainability of companies, and stock selection strategies to tailor portfolios to a wide range of investor ESG preferences. This week BNY Mellon adopted such an approach motivated in part by the EU’s Directive on Pensions (IORP II).30 Earlier this year UBS launched a pilot project that will allow investors to rate how much weight they want to place on different ESG factors.31 And last month, BlackRock launched six new Exchange Traded Funds (ETFs) that combine an ESG uplift and a 30% reduction in carbon emissions.32 These sustainable building blocks can be substituted into many traditional portfolios, improving ESG scores and reducing greenhouse gas (GHG) intensity without sacrificing performance.
In the future, climate and ESG considerations will likely be at the heart of mainstream investing. Investors will tailor their investments and fulfil their fiduciary duties through: better quality and more widely available data on sustainability and performance; superior data analytics through the advent of AI and Machine Learning; and more informed judgements of strategic resilience.
Conclusion: a New Horizon
Financial policymakers will not drive the transition to a low-carbon economy. Governments will establish the climate policy frameworks, and the private sector will make the necessary investments.
Nonetheless, financial policymakers do have a clear interest in ensuring the financial system is resilient to any transition hastened by those decisions. Our role is to develop the frameworks for markets to adjust efficiently.
A market in the transition to a two-degree world is being built. It will reveal how the valuations of companies could change over time as climate policies adapt and carbon intensity declines.
It will expose the likely future cost of doing business, of paying for emissions, and of tighter regulation.
It will help smooth price adjustments as opinions change, rather than concentrating in a climate “Minsky moment”.
And it will allow feedback between the market and policymaking, making climate policy a bit more like monetary policy, with policymakers learning from markets’ reactions, and markets internalising policymakers’ objectives, strategies and instruments.
In this way, recent progress in disclosure, risk management and return optimisation is creating a path to a New Horizon. A virtuous circle is becoming possible where companies disclose more information, investors make better informed decisions, and sustainable investment goes mainstream.
But the speed with which this market develops will be heavily influenced by the coherence and credibility of climate policies. Finance will complement – and potentially amplify – but never substitute for climate policy action.
The policy frameworks with the greatest impact will be: time consistent (not arbitrarily changed); transparent (with clear targets, pricing and costing); and committed (through treaties, nationally determined contributions (NDCs), domestic legislation and consensus).
As countries build their track records and their credibility grows, the market will allocate capital to deliver the necessary innovation and growth and pull forward the adjustment to a low carbon future.
The more prolific the reporting, the more robust the risk assessment and the more widespread the return optimisation, the more rapidly this transition will happen, breaking the Tragedy of the Horizon.
Originally published by the Bank of England.
All speeches are available online at www.bankofengland.co.uk/publications/Pages/speeches/default.aspx
- References
1 Carney, M. (2015). Breaking the Tragedy of the Horizon. Available: https://www.bankofengland.co.uk/speech/2015/breaking-thetragedy-of-the-horizon-climate-change-and-financial-stability
2 Full list of current TCFD supporters available on: https://www.fsb-tcfd.org/tcfd-supporters/
3 WSJ. (Jan 2019). PG&E: The First Climate-Change Bankruptcy, Probably Not the Last. Available: https://www.wsj.com/articles/pg-e-wildfires-and-the-first-climate-change-bankruptcy-11547820006
4 Horster, M and Papadopoulos, K. (2019). Climate Change and Proxy Voting in the U.S. and Europe. Available: https://corpgov.law.harvard.edu/2019/01/07/climate-change-and-proxy-voting-in-the-u-s-and-europe/
5 Task Force on Climate-Related Disclosures (TCFD). (2018). TCFD:2018 Status Report. Available: https://www.fsbtcfd.org/publications/tcfd-2018-status-report/
6 For example the Oil and Gas industry group convened by the World Business Council on Sustainable Development and the Institute of International Finance for banks.
7 The other channel concerns liability risks. These stem from parties who have suffered loss from the effects of climate change seeking compensation from those they hold responsible. Such claims could arise well into the future, as the science and evidence of climate change hardens, though some are already taking action against companies on the grounds of failure to disclose the risks posed to their business models by climate change.
8 Prudential Regulation Authority. (Sept 2015). The impact of climate change on the UK insurance sector. Available: https://www.bankofengland.co.uk/-/media/boe/files/prudential-regulation/publication/impact-of-climate-change-on-the-uk-insurancesector.pdf
9 Prudential Regulation Authority. (Sept 2018). Transition in thinking: The impact of climate change on the UK banking sector. Available: https://www.bankofengland.co.uk/-/media/boe/files/prudential-regulation/report/transition-in-thinking-the-impact-of-climate-change-onthe-uk-banking-sector.pdf?la=en&hash=A0C99529978C94AC8E1C6B4CE1EEC
10 Prudential Regulation Authority. (Oct 2018). Consultation Paper on Enhancing banks’ and insurers’ approaches to managing the financial risks from climate change. Available: https://www.bankofengland.co.uk/prudential-regulation/publication/2018/enhancingbanks-and-insurers-approaches-to-managing-the-financial-risks-from-climate-change.
11 Prudential Regulation Authority. (March 2019). PRA and FCA’s joint Climate Financial Risk Forum. Available: https://www.bankofengland.co.uk/news/2019/march/first-meeting-of-the-pra-and-fca-joint-climate-financial-risk-forum.
12 Forthcoming April 2019
13 As described in TCFD September 2018 report: These companies disclose the inputs to and outputs of their scenario analyses including strategic responses to the low-carbon transition, such as changes in portfolio mix or investment
14 Encouragingly, all members of the oil and gas preparer forum used the 2-degree energy transition scenario to inform strategic decisions. The materials and building sector also had the highest percentage of companies disclosing information about strategic resilience and most provided some information on the climate-related scenarios they used to make these assessments.
15 The most widely used and well-known are the IEA transition scenarios, which model six different assumed pathways and associated temperature increases. For modelling physical risks, the IPCC’s four Representative Concentration Pathways (RCPs) fix greenhouse emissions and analyse the resulting change to the climate.
16 The voluntary network was set up by 8 central banks and supervisors in December 2017 at the One Planet Summit, and has since grown to 29 members, representing countries accounting for nearly half of global emissions, and five observers. It is a voluntary, consensus-based forum whose purpose is to share best practices, contribute to the development of climate- and environment-related risk management in the financial sector and mobilize mainstream finance to support the transition toward a sustainable economy. The analytical work is split into three work streams and the research will be published in April 2019: WS1 microprudential/supervisory; WS2 macrofinancial; and WS3 Scaling up green finance.
17 For more information on the Commission’s Sustainable Plan, see: https://ec.europa.eu/info/business-economy-euro/banking-andfinance/sustainable-finance_en
18 See: CBI https://www.climatebonds.net/about and ICMA https://www.icmagroup.org/green-social-and-sustainability-bonds/greenbond-principles-gbp/
19 Climate Bonds Initiative. (2018). Green bonds: The state of the market 2018. Available: https://www.climatebonds.net/resources/reports/green-bonds-state-market-2018
20 By Indonesia, Belgium, Lithuania, Ireland and Seychelles
21 See: https://www.unpri.org/pri
22 Deloitte. (Nov 2015). The future of wealth in the United States. Available: https://www2.deloitte.com/content/dam/insights/us/articles/us-generational-wealth-trends/DUP_1371_Future-wealth-inAmerica_MASTER.pdf
23The Cerulli Edge, Global Edition, Issue 206 (Apr 2018). Available: https://www2.deloitte.com/content/dam/insights/us/articles/usgenerational-wealth-trends/DUP_1371_Future-wealth-in-America_MASTER.pdf
24 Clark, G, Feiner, A, and Viehs, M. (March 2015). From the Stockholder to the stakeholder: how sustainability can drive financial outperformance. (Oxford University and Arabesque)
25 Friede, G, Busch, T, and Bassen, A, (2015) ESG and financial performance: aggregated evidence from more than 2000 empirical studies, Journal of Sustainable Finance & Investment, 5:4, 210-233
26 Nordea Markets. (Sept 2017). Cracking the ESG code. Available: https://nordeamarkets.com/wp-content/uploads/2017/09/Strategyand-quant_executive-summary_050917.pdf
27 BlackRock. (Feb 2019). Sustainability: The Future of Investing. Available: https://www.blackrock.com/us/individual/literature/whitepaper/bii-sustainability-future-investing-jan-2019.pdf
28 A recent review by Hermes Investment Management shows that companies with good or improving social factors have outperformed other companies by 15bps per month over a decade and good governance generates a 24bps per month elevated return. A focus on the E in ESG – environmental – meanwhile has no penalising effect on returns, and companies with strong environmental policies do better in downturns by 19bps than their peers. See: https://www.institutionalassetmanager.co.uk/2018/11/13/270456/hermes%E2%80%99-esg-study-reveals-social-characteristicsoutperforming
29 Nagy, Z, Kassam, A, and Lee, L-E,. (June 2015). Can ESG add alpha?.Available: https://www.msci.com/documents/10199/4a05d4d3b424-40e5-ab01-adf68e99a169.
31 See: https://www.ubs.com/global/en/ubs-news/r-news-display-ndp/en-20190121-wef.html
32 See: https://www.ishares.com/us/strategies/sustainable-investing#esg