Transcript
Speaker: Welcome to HSBC Global Viewpoint, the podcast series that brings together business leaders and industry experts to explore the latest global insights, trends, and opportunities.
Speaker: Make sure you're subscribed to stay up to date with new episodes.
Speaker: Thanks for listening.
Speaker: And now onto today's show.
Speaker: Hello everybody and welcome to the ESG Brief, our regular review of the environmental, social, and governance landscape by HSBC Global Research.
Speaker: I'm your host, Jack Reed.
Speaker: In this edition, we're looking at the ESG team's recent report on corporate diversity and data science.
Speaker: Using the powers of natural language processing, what can we learn from the transcripts of corporate conference calls with investors about employee gender diversity?
Speaker: Are the words from executives about diversity translating into actions that actually diversify the workplace?
Speaker: To answer these questions, I'm joined here in London by Mark McDonald, the head of data science and analytics, Marina Cobal, our corporate governance analyst, and in New York by our ESG analyst, Camilla Sarmiento.
Speaker: A governance note of our own first, this report was recorded for publication on the 20th of September.
Speaker: All the disclosures and disclaimers associated with it must be viewed on the link attached to your media player.
Speaker: Irina, let's start with you.
Speaker: What did you set out to investigate in this report and why?
Speaker: Yes, we set out to investigate corporate diversity.
Speaker: According to our most recent EHD sentiment survey, diversity remains a top social issue investors are planning to spend most of their time in the next 12 months.
Speaker: Diversity is complex and we have investigated diversity in many different ways in the past, but this time, together with our data science team, we looked at the frequency of diversity mentions on earning calls in different sectors and regions.
Speaker: As there is a limited amount of time for discussion on earning calls, we believe that frequency of diversity mentions is a good proxy for the degree to which management and investors are prioritizing diversity.
Speaker: And then we compared this frequency with workforce and board gender diversity progress to better understand where the focus should be.
Speaker: So effectively, you're looking at whether the words of company executives are being matched by deeds.
Speaker: Yes, yes, indeed.
Speaker: We looked at the progress between 2018 and 2021.
Speaker: That's when the frequency of diversity mentions actually changed a lot.
Speaker: And we specifically selected gender as it has been probably the most extensively studied and easily measurable characteristics.
Speaker: So we think that it can provide a useful basis for our diversity analysis.
Speaker: Now before we look at the broad findings of the report, I want to turn to Mark McDonald.
Speaker: Mark is the head of our data science team.
Speaker: Mark, you have used natural language processing for a number of different teams in global research.
Speaker: Take us through what the process has been in this case.
Speaker: So again, in this situation, we're returning to what companies are saying on their earnings calls.
Speaker: And we keep returning to this because it's fantastically rich data sets.
Speaker: Now, of course, analysts and investors will often dial into a conference call and listen to what's being said.
Speaker: And the reason that they do this is that often on an earnings call, information is disclosed, which may not be officially disclosed anywhere else.
Speaker: Now, we're not listening to the calls.
Speaker: We're using earnings call transcripts.
Speaker: So after the call, what's said on the call gets transcribed.
Speaker: And then we apply our natural language processing techniques to us.
Speaker: And the reason that we do this is you get a vast amount of information.
Speaker: In each reporting season, you get thousands of new transcripts.
Speaker: And this is millions of sentences.
Speaker: That's obviously far too much material for any human to even really scratch the surface.
Speaker: But of course, that's perfect for a machine.
Speaker: Our natural language processing algorithms can just as easily be applied to thousands of earnings calls as they can to one.
Speaker: So it's a very powerful analytical technique that can bring out sort of macro themes from very micro data sets.
Speaker: And presumably you can strip out a lot of the noise, for example, words like diverse or diversification, which could be applied to a lot of things that aren't necessarily about gender diversity.
Speaker: Yes, that's exactly right.
Speaker: The key aspect of any of these natural language processing analyses.
Speaker: The first step is to hone in accurately and precisely on the subset of the sentences that are about the theme at hand.
Speaker: And that's often the most time-consuming part of the analysis is improving that until that works.
Speaker: Turning to you, Camilla, what does the data show us?
Speaker: So when looking at our data, we found that diversity discussions on earning calls have increased exponentially over the past decade.
Speaker: So looking at 2021, they are about 15 times more likely to occur than a decade ago.
Speaker: And we found similar trends relating to just specifically speaking about board diversity as well.
Speaker: Now, looking at 2022, we don't have obviously a complete year of data yet, but it is coming in slightly below 2021 data.
Speaker: Now, we think this could be attributed to various global events that may have shifted company focuses towards managing these emerging risks.
Speaker: But we still think that diversity and inclusion is very much remains a top priority for corporates and investor engagement, as Yurina had said.
Speaker: From this vast amount of data gleaned from transcripts and the specific mentions of diversity in conference calls, what's the takeaway, Irina?
Speaker: So our main takeaway for different sectors and regions is that investors should consider the risk of diversity washing.
Speaker: And I'm saying this because when we match the frequency, the high frequencies of diversity mentions, we realize that it's not matched by gender diversity progress.
Speaker: So most of the diversity discussions, they are actually driven by management on the calls.
Speaker: And we see a lot of general comments.
Speaker: For example, they're talking about maintaining a strong focus on diversity, remaining actively engaged to foster gender diversity or being satisfied with a diverse qualified boards or peers.
Speaker: progress made.
Speaker: So as you can see, those are really all general comments and they often give no detail as what it actually means.
Speaker: So we think that corporates, they actually feel stakeholder pressure to make a statement on diversity as if to check a box without strategically acting on improving diversity without their businesses.
Speaker: And just to add there on some examples from our analysis, we found that when looking at sectors specifically, the utility sector had the highest frequency of discussions of diversity on their earnings calls.
Speaker: However, they had very little progress on their gender diversity across their board and senior and organizational management levels.
Speaker: So, you know, here we're seeing, you know, higher risks of diversity washing.
Speaker: Now, looking at more of the regional areas, we also found risks of diversity washing specifically within the LATAM area.
Speaker: This is because there has been very little progress and also engagement by governments on diversity within Latin America.
Speaker: However, they also had the highest frequency of diversity discussions on earnings calls as well.
Speaker: One of the things that you have included in the report is a recommended set of questions that investors can ask corporates to prize out more information about their diversity agenda.
Speaker: What is the goal here, Irina?
Speaker: Yes, we propose eight questions for investors to help them assess how corporates incorporate diversity considerations into their organizational practices.
Speaker: So these questions focus on a number of issues.
Speaker: So that's, for example, succession planning and automation challenge or some sector-specific considerations.
Speaker: We think these are really useful questions because we believe that there's opportunity for investors and earning calls to
Speaker: hold corporates to greater transparency on the progress made against the diversity strategies.
Speaker: And these questions, we believe, will help them to focus on specific diversity actions and progress, helping to address diversity washing and drive better progress on diversity globally.
Speaker: And Camilla, aside from the issue of diversity washing, what else stands out in this research at a sector level?
Speaker: I think one of the main differences in the sectors that we found was kind of these sector-specific differences in historical gender makeup.
Speaker: So we have a lot of sectors that have been the majority female or majority male.
Speaker: And what this means for diversity is a question that we asked ourselves.
Speaker: A prime example here is the basic materials sector, which has been very male-dominated.
Speaker: This has been due to many of the characteristics, like physical strength, that are needed in the industry.
Speaker: So there are very different views on what the gender balance should be in different sectors.
Speaker: You know, we believe there's just a continuous focus on diversity will promote gender diversity.
Speaker: Again, to the basic material sector, while they still are majority male, they have had some of the largest progress percentages across the sectors.
Speaker: And there also is the opportunity here for different technologies and automation, which could allow these
Speaker: these sectors become more gender balanced as different characteristics like physical strength with new technologies might not be as important.
Speaker: And how does the data that you're looking at differ on the regional level?
Speaker: So I think the largest consideration regionally is the regulatory initiatives that are being put in place to promote diversity by different regulatory bodies around the world.
Speaker: So we have seen this through disclosure requirements and through board and senior management level quotas and targets, and more specifically relating to the UK and the US areas that have been promoting diversity.
Speaker: And when we look at our diversity discussions on earning calls, we found that those regions which had a high frequency of diversity discussions also had regulatory initiatives in place.
Speaker: And so we found that these kind of regulatory initiatives that are going on have been key drivers, we think, of these discussions that are happening on earnings calls.
Speaker: So lots to digest in this report and some helpful guidance for investors who may want to push corporates on the precise actions underway on diversity.
Speaker: A final question to you, Mark.
Speaker: The use of data science and natural language processing in HSBC Research's products and services has clearly grown over the past couple of years.
Speaker: What are you learning in the process here?
Speaker: Well, in many ways, the way that we've been doing this is actually the ESG team have been helping out the data science team.
Speaker: So often with these natural language processing analyses, one of the hard things is to identify accurately and precisely
Speaker: the sentences that are about that topic.
Speaker: And with many ESG themes, the language that would be used to discuss that theme is often used metaphorically elsewhere.
Speaker: So as an example, if you simply search for the word environment on an earnings call, the vast majority of those are false positives, where it's people talking about the funding environment or the competitive environment.
Speaker: And similarly, mentions of sustainable are typically talking about dividend sustainability rather than ESG sustainability.
Speaker: And so the only way to accurately hone in on many of these ESG themes is to train a machine learning model how to identify them.
Speaker: And what this means is that you need trained experts to go through thousands of examples of sentences and to mark them down that this one is relevant and this one is not.
Speaker: And
Speaker: with enough examples, these models can hone in and accurately identify, yes, this sentence is about that theme and no, this sentence is not.
Speaker: And so we've been working closely with the ESG team.
Speaker: They are the trained experts that I referred to that have done this analysis.
Speaker: So we now have these models that can hone in at a very granular level and say, yes, this sentence is about this theme and then we can start applying that analysis
Speaker: in a variety of ways.
Speaker: So really, I'm fantastically grateful to my colleagues in the ESG team for that work.
Speaker: Mark, Camilla and Yurina, thank you very much.
Speaker: Super interesting report.
Speaker: Thanks, Jack.
Speaker: Thanks for having us on the podcast.
Speaker: For Global Research clients, the full report is available on the research website at research.hsbc.com.
Speaker: The report's also showcased on the Global Research app.
Speaker: Thank you for joining us.
Speaker: If you have any questions, please do email us at askresearch at hsbc.com.
Speaker: And don't forget, if you're listening to this podcast at your desktop, you can also find the ESG Brief on Spotify and Apple, along with our weekly macro viewpoint and our newest podcast, Under the Banyan Tree.
Speaker: Just search for the HSBC Global Viewpoint channel and you'll find us there.
Speaker: Thanks very much for listening.
Speaker: We'll be back soon.
