Without reliable data on ESG performance, ideally objectively measured and audited, it's difficult to constructively pursue any of the ESG investing strategies we've described or for corporate or boards or investment analysts to assess their choices on ESG issues. The gap between what executives boards analysts perceive is important and the confidence they have in the data at their disposal highlights a key obstacle to ESG integration data reliability. While some norms have evolved there's still no regulatory guidance regarding what to disclose and in what form. Indeed questions pertaining to data integrity and reliability act as a significant constraint on research and practice in the field. The environmental and social activists who developed ESG data reporting standards were concerned about the quantities of output by firms or the number of incidents. And this focus persists in the vast majority of current reporting by companies and third party rating agencies that process their data into rankings or scores. Increasingly however the audience for ESG data, the consumers are financial analysts. They're seeking to take the ESG information and integrated into their investment strategies and into their financial calculations. This presents a significant data challenge analysts aren't well versed in the technical and engineering components of environmental mitigation. Therefore determining the actual environmental impact is going to be subjective even within a rating system. These financial analysts and modelers have adapted to the norms of disclosure but ultimately, they have to integrate esg performance into a dollar denominated financial model. But such performance is measured in different units of output somehow standardized into a score. So they have to take these scores and put it into a financial calculation, they do it without the scientific knowledge necessary. They don't know how to convert tons of carbon into dollars, they rely on correlations, they rely on data analysis rather than on any theory or any framework. Again, these analysts and modelers are left with the ESG equivalent of simple emojis, which are pretty hard to integrate into the analysis of a company's growth rates, its profits or its potential losses. We need to bring new data to the discussion and follow this by asking different more substantive questions. Thanks to much improved standards of corporate disclosure the quantity of ESG data has grown substantially. In recent years for example, bloomberg terminals offer 900 different data fields on ESG factors for 13,000 companies covering over a decade. That amounts to over 140 million data points five years ago, it was only six million and five years before that, one and a half. Five years ago, 81% of S&P 500 companies were issuing sustainability reports that could be evaluated, last year that number had risen to over 90%. Impressive as this rise in the quantity of data sounds, there's still wide variation in the content and format of those disclosures. By one count there are now more than 230 different initiatives trying to standardize sustainability reporting. Another study found that there are 20 different reporting schemes just on the topic of employee health and safety, in order to address the growing proliferation of metrics and standards. The World Economic Forum's International Business Council, in cooperation with the Big Four accounting firms Deloitte, EY, KPMG, and PWC. Has offered a harmonized set of 21 core and 34 expanded metrics for ESG reporting that are aligned with existing frameworks and standards setting bodies. In addition to the challenge of competing standards, a host of private sector data providers offer their own proprietary ratings, each with its own adjustments waded into seas and scores. Examples include companies like Kinder Leitenberg, Dominy and company Sustain Olympics asset for and Vigeo Eiris. All of these providers follow their own distinct methodology, taking in the company's voluntary disclosures, either from sustainability reports or in responses to detailed customized surveys. And they connect this status somehow to construct their own measure of a company's social and environmental performance. Relative to some absolute standard of one of the best practices, processes and outcomes. However, each company differs how many ESG factors do they measure? How do they group them? What are the indicators they assess for each? How did they weight them? Over time each raiders expanded the range of metrics, the indicators they capture they're waiting systems have been adapted. A recent academic paper strongly suggests that this rebalancing and re weighting is not done randomly. It's actually done periodically to improve the measures ability to predict recent stock price performance. Innovest launched in the late 1990s, began with a very different approach and goal rather than trying to capture the goods and bads of a company. They sought to provide scores that would serve as a proxy for the value of a company's intangible assets such as its environmental capital, human capital, stakeholder capital. Or the quality of its strategic governance, which affected the value of all these capitals. Its ratings explicitly compare a company with its industry peers as opposed to all companies in the universe and they're designed to better help investors achieve super normal risk adjusted returns. Through portfolio optimization, tilting or smart beta strategies, RobecoSAM was another company that adopted the same approach seeking to turn ESG data into something that could be monetized. As interest in ESG data has developed among mainstream financial investors. The primary financial service data providers have acquired their own expertise often through the acquisition of these ESG data providers I've already spoken about. Unfortunately, many of these mergers and acquisitions led to aggregations of data structures and methodologies. That were originally designed to measure good and bad practices with those designed to measure intangible value. Such mergers again created the challenge for the acquirers of how do we add the emojis. The goods and the bads onto the value of the intangibles which are measured in dollars and then connect them to balance sheets. More recently, a new breed of data players has been using web crawlers, natural language parsing often combined with artificial intelligence. An attempt to supplement companies voluntarily released information with media event based data and other unstructured text analysis. Since we know that a company's voluntary unaudited disclosure is going to be biased both because it over reports good news and suppresses the bad. These data providers believe that more signal can be found in the words of the stakeholders of the firms rather than the words of the firms and their managers. They track the words and the sentiment of customers employees, government officials, civil society organizations, what do they say about the firm? How do they act on the firm on separate ESG issues and then they score that sentiment in those actions in real time. Both as a spot market or real time indicator but also as to construct a measure of accumulated capital in the stakeholder relationship with the firm. Finally, they can also explore the momentum, the direction of change between the real time indicator and the accumulated measure or stock measure. These pulse, insight and momentum scores as they're known, can feed into an array of ESG trading strategies that we discussed previously. While this methodology offers important benefits, it also contains important weaknesses. The media decides the relative importance of stakeholders whose identity is not at the current time available to be extracted from these data feeds. Nor is the location of the stakeholder captured, perhaps most importantly, while such stakeholder oriented sentiment and relational capital measures are valuable additions to the analytic tool kit. They still don't allow for a clear assessment of the financial materiality of an issue. There's some effort to do this by using a 01 switch or a dichotomous measure provided by the sustainability accounting standards board. But that's a very crude way to assess the financial materiality of a given ESG issue. Beyond text analysis of the news new alternative data providers are using similar tools to harvest job reviews from glass store hiring and wage data from job boards. Combined with linkedin profiles to measure turn and wage distributions by firm by race by gender by location by year. At Wharton Csg analytics lab and political risk lab we're developing a tool that synthesizes this sort of data. As well as other unstructured text to extract information on the sentiment and structure of relationships, linking affirm not just to its employees but to all the stakeholders in a place. As well as thematically content coding these news stories to identify what drives the positive and negative sentiment between a stakeholder and a firm. Of particular interest are the direct and indirect impacts of multinational firms entry into a foreign country on disadvantaged groups in that country. Including indigenous peoples or other isolated groups but the same analysis can stretch out into different ethnicities, different races even here in the United States and in Europe. Our work to date highlights the tendency for foreign entrance into a country to privilege stakeholders who are elites who have strong ties to the government, the wealthy, the most powerful in a country. That tends to further marginalize stakeholders who already feel excluded and entrenches their resentment directed against firms as well as against globalization and capitalism more broadly. Other firms are using satellites to measure pollution above factories and activists are creating networks of water monitoring devices to map pollution and watersheds back to polluters. While a flood of 3rd party rankings and increased voluntary corporate disclosure is available to analysts, there's no tendency towards convergence. If anything, we're seeing a divergence with the introduction of all these new data sources. Much of the new data that's brought in however is inconsistent and it often struggles to identify the most material information to a company. And remains as a result of limited financial value to investors. New data providers are continually emerging but they're just furthering this divergence in the ratings the company has on ESG issues. Why does the average new raider diverge from the predecessors, well it would be hard to make a business selling the same data. So they have a strong incentive to create different data and try to differentiate themselves from the incumbents in the industry. Each of these data providers is going to report a score in index or account, they're all going to do it at varying levels of aggregation at varying time intervals. As a result, an authoritative recent study of ESG data characterizes the field is one of aggregate confusion. Company's performance scores just differ dramatically depending on which data set we look at even on the same ESG factor. With ratings sometimes structurally biased in favor of large firms or those that fulfill regulatory requirements. While different ESG data providers typically agree on which firms are the worst performing correlations outside the bottom decile of firm performance are worryingly low. To be specific when we look at credit ratings, the correlation between S and P Moody's tends to be above 0.9. But when we look across ESG data providers it tends to be between 0.6 and 0.05. These disagreements allow companies themselves to pick and choose which ratings to highlight. Decisions that are bound to be self serving and which further undermine investor confidence in certain stocks. Consider for example, is visualized here Apple RobecoSAM rates Apple in the lowest quartile overall on ESG. As well as in the lowest quartile on social and governance and not much better on the environmental dimension. However, if we look and turn to Thompson Reuters ratings, they rate Apple and the top quartile overall and on environment governance. And not much below that on social turning to sustain olympics and MSCI, they place apple closer to the median firm. Similar disparities exist for Facebook and Amazon although notably RobecoSAM and sustainable ticks flip in their perspective on Amazon. And sustain olympics and MSCI are more positive on Facebook, while some of this confusion results from different measures being designed for different purposes. Some relates to gaps and available data that are non random and different choices as to how to best impute missing information. As well as combine it with available information by competitive 3rd party data vendors. Overcoming the resulting aggregate confusion is going to require a standardization of corporate reporting, an agreement on standards or best practices in the imputation of missing data. More fundamentally however, we have to agree on a unit of account that allows for the aggregation of ESG scores and ratings into corporate financial data. The obvious answer, which comes from the company's already adopting this type of effort is to measure both ESG performance and financial performance in the same unit of analysis and that is dollars.