Examining Diversity’s Dividends: Can Studies Survive Contact with Peer Review

Share on Facebook
Share on Twitter
Share on

Hubwonk: Examining Diversity’s Dividends: Can Studies Survive Contact with Peer Review 

[00:00:00] Joe Selvaggi: This is Hubwonk. I’m Joe Selvaggi. Welcome to Hubwonk, a podcast of Pioneer Institute, a think tank in Boston. Does diversity pay? Beyond the laudable goals of assembling the best talent pool, irrespective of race or ethnicity, does actively seeking diversity in executive leadership truly redound to a company’s financial success?

[00:00:23] McKinsey, a highly regarded global consulting firm, asserts confidently that data from their four studies conducted in 2016 to 2023 unequivocally asserts it does. Their findings have effectively endorsed diversity, equity, and inclusion initiatives worldwide, lending these programs the weight of scientific validation.

[00:00:43] However, such resolute claims must warrant a critical examination by fellow data scientists who are bound to scrutinize the quality of the methodologies and data employed to arrive at these conclusions. To their credit, McKinsey has been transparent with their research methodology, having only anonymized the firms studied to safeguard client confidentiality.

[00:01:05] Given this transparency and the firmness of their research findings, how do their assertions fare under the scrutiny of peer review? Joining us today is Dr. Jeremiah Green, an Associate Professor of Data Analytics at Texas A&M Mays Business School, and co-author of the recently published research paper, Does Greater Diversity in Executive Race Ethnicity Reliably Predict Better Future Firm Financial Performance?

[00:01:33] Dr. Green and his co-authors delved into McKinsey’s research to examine the extent and causality of the purported financial benefits associated with diverse executive leadership. He will share with us what prompted his team to scrutinize McKinsey’s findings. and unveil the outcomes of their efforts to replicate the results of these four studies that undergird the business case for diversity, equity, and inclusion programs worldwide.

[00:01:58] When I return, I’ll be joined by Texas A& M Associate Professor, Dr. Jeremiah Green. Okay, we’re back. This is Hubwonk. I’m Joe Selvaggi, and I’m now pleased to be joined by Texas A& M Associate Professor, Dr. Jeremiah Green. Welcome to Hubwonk, Professor Green.

[00:02:16] Jeremiah Green: Thanks. Great to be here.

[00:02:17] Joe Selvaggi: All right. It’s a pleasure to have you.

[00:02:19] This is your first time on Hubwonk. We’re going to talk about a paper you recently co-authored titled, Does Greater Diversity in Executive Race Ethnicity Reliably Predict Better Future Firm Financial Performance? It’s a research paper that examines the consulting firm McKinsey’s widely accepted claim that diversity delivers to firms bottom line.

[00:02:41] We’re going to go deep on this, but before we do, I want our listeners to, learn a little bit about you and the science of using accounting data for business analysis. You teach at Texas A& M, you teach data analytics, and you use it in your work. What does this field of data analytics do in the business community?

[00:03:00] Jeremiah Green: Okay, yeah, so in terms of data analytics for accounting, that’s also a little more specific than just data analytics in general. Data analytics for accounting typically means we’re using statistical approaches to analyzing financial reports. If you think about a newspaper article that says, apples earnings per share was, X dollars this year, it’s looking at things like that, what they provide to analysts, what they provide to investors, and trying to understand what that means for the company, for investors, for other groups that are interested in looking at those financial reports.

[00:03:41] Joe Selvaggi: So, you, you examine, differences in how financial data, again, you’re looking at publicly traded companies, I assume. All of that should be, available to, someone who wants to analyze it. And you’re trying to relate how different features of a, of a firm’s balance sheet, either explain current success or perhaps, predict future success.

[00:03:59] let’s see an example. We’re going to talk about your analysis of McKinsey’s analysis in this area. What’s an example of where your work in, maybe a different topic might help our listeners understand, the contours of your work?

[00:04:12] Jeremiah Green: Yeah. So, an example of thinking about how financial information that companies report either explains their performance now or predicts their future performance, we might think about an example could be a paper by Mark Solomon.

[00:04:28] He is a professor at University of Southern California. What he does is look at ratios from the financial statements, so profitability or efficiency, and tries to see whether that predicts future performance. What is he? He tries to predict earnings and future returns, stock returns, and that’s the type of thing we might do.

[00:04:52] Some of my research is looking at predicting stock returns also, so we try and use those sorts of pieces from the financial statements to say, what can we learn from them,

[00:05:02] Joe Selvaggi: So, you polished crystal balls to help people who are potentially inventors try to understand what might come in the future.

[00:05:09] We all know, everybody would love such a wonderful device, but it’s challenging, and that’s why you need rigorous analysis and your best effort. So, I want to then talk, rather than the general concept, let’s dive at least a little bit into your paper, the recent paper. You wrote about, the relationship between the diversity, in the leadership positions of a firm and their, probability of being profitable.

[00:05:35] We’re going to talk about that in depth. What was your interest in writing about? Of all the features of a firm, why did you want to examine why diversity might be, among all the other, measures you could use in a company, one that might predict future profitability?

[00:05:50] Jeremiah Green: Yeah, that’s a really good question.

[00:05:52] I, I think it’s probably worth putting in context what got us into that, and that is, we’ve heard lots of claims in, by journalists and by, McKinsey, trying to, that, that say that diversity in various forms improves financial performance. And so, from an academic standpoint, I think we tend to be pretty skeptical of bold claims.

[00:06:16] So that’s one aspect. and the other pa the other issue is if it is the case that diversity improves performance, then that’s something as an accounting person that I want to know so I can tell companies or students that if you increase diversity it’s going to improve your performance and this should be a, a top level concern for something we might want to focus on.

[00:06:41] Joe Selvaggi: So, you either want to challenge a bold claim, see if it’s valid, but you also, if it’s valid, you want to incorporate it into your analysis for the future, right? You want to, as a new, yardstick to add to the other yardsticks used when predicting future performance. But for, I want to provide some background for our listeners.

[00:06:55] We’re going to be looking at a study or several studies, four different studies by the same firm, McKinsey. This is a firm that’s very well regarded in the, a global consulting world. This is the gold standard of consulting. So when you’re talking about challenging or examining, studies or four studies by a very well regarded, accounting, consulting firm.

[00:07:14] These are, you’re taking on a Goliath, if you will. They’ve been around for about 100 years, I think. They’re two years away from 100. so that the relevance when you’re taking on somebody like, a firm like McKinsey is to say that what they’ve put out is almost regarded, I want to use, religious terms, but almost as gospel.

[00:07:31] It essentially is almost presented as it ought to be unquestioning. when we’re talking about you as an academic versus, a firm like McKinsey, which is consulting, Aren’t you more or less, when you’re advising firms or, analyzing what works in businesses, aren’t you in the same business, analyzing the same kinds of data?

[00:07:49] Jeremiah Green: Yeah, okay. Yes. I guess the easiest answer is yes, we are doing, using the same data. We’re trying to make, understand, and make similar conclusions. Can I push back just a little bit in that I would like to not say that we’re trying to take on McKinsey or challenge McKinsey, but, from my viewpoint, if we see research or conclusions based on research that are not, justified by the evidence, then that is something that, as academics, we should be involved in. In terms of what we do with the evidence or decisions we make outside of that isn’t really something I can say anything about, but yes, so in general, we’re doing the same sort of thing and we’re, we have the same sort of, interests. We may not have the same sort of use of it after we make these claims, or whatever they are.

[00:08:45] Joe Selvaggi: I didn’t mean to, I appreciate you pushing back, and I didn’t mean to imply that it was some sort of hostility towards their, the firm. Rather, you’re an academic, and all academics, we hope, present evidence, and then invite their critics. Their colleagues or peers, other, data analysts to scrutinize, meaning you get better.

[00:09:03] You, you don’t want to keep a mistake a secret. You want others to scrutinize your work so that they can point out your mistakes. Every paper, I think, since the dawn of time, has mistakes. And you invite other very intelligent, let’s say colleagues, be they academics or consultants, to scrutinize your work.

[00:09:19] You, what I meant to say, you’re taking on, you’re scrutinizing McKinsey’s work. Can we make that stipulation?

[00:09:26] Jeremiah Green: Yeah, that seems fair enough, yeah.

[00:09:28] Joe Selvaggi: Okay. All right. So, let’s just, before we get into the details, I mentioned there’s four studies by McKinsey. I don’t have the dates right in front of me, but I think the first was 2016.

[00:09:38] The other just came out in 2023. They had similar, conclusions. They became more confident as they went on. For our listeners, just at the high level, how would you summarize the conclusions McKinsey, makes about diversity? In high level the board and, c-suite leadership and, profitability.

[00:10:01] Jeremiah Green: Yeah. Let me, let me caveat it, or, try and split that into two pieces. I think the first piece is pretty clear about what they do in mechanically and what they say within their papers. when I go to make the step about what conclusions they make. those are slightly different than how they describe what they do.

[00:10:23] In terms of the general conclusions, they say that, diverse, having more diversity, like executive diversity, increases company performance. That’s the general claim, even though they are more nuanced when they get into the discussion, I think.

[00:10:39] Joe Selvaggi: Yeah, I, again, I went into the research, looked at the data, not the way you did, but more generally, but also I looked at the overarching, summary and in the most recent report, I want to quote for our listeners, just to add color to this conversation, how bold McKinsey describes, how boldly they described the observations they made in this most recent paper.

[00:10:58] I’m going to report from, this is a quote from McKinsey’s work. For almost a decade through our Diversity Matters series of reports, McKinsey’s has delivered a comprehensive global perspective on the relationship between leadership diversity and company performance. This year, the business case is the strongest it has ever been since we’ve been tracking, and for the first time in some areas, equitable representation is in sight.

[00:11:22] Further, a striking new finding is that leadership diversity is also convincingly associated with holistic growth, ambitions, greater social impact, and more satisfied workforces. Unquote. So, these are pretty bold claims, based on what they claim is decades of research and four studies. So that’s where I want to start.

[00:11:39] Let’s examine by defining terms. In that passage, McKinsey uses the term diversity. How does McKinsey define diversity?

[00:11:48] Jeremiah Green: Yeah, okay, so broadly, I guess let’s do a, simple explanation. What they really are trying to measure is how, not concentrated their executive team is in any, small or single group or ethnicity.

[00:12:03] They’re looking at ethnicity, different ethnicities, they group it into a few different categories, and then they try and say if it’s, So if it’s all concentrated in one ethnicity, then that’s not diverse. If it’s concentrated, or if it’s spread out across different ethnicities, then it’s diverse.

[00:12:23] Joe Selvaggi: So, for the benefit of our listeners, I thought about this myself. I’m like, okay, what would make something a perfect score? And what would make a firm a, Terrible score. So, a perfect diversity score would be whatever number of diversities Let’s say there’s eight or five. I think those are the two different measures.

[00:12:40] If you have five, one of each, and you have five board members, that’s perfect score. If they’re all white, or for that matter, all black, or Hispanic, or pick it, Asian, that would be a bad score. So, what we’re looking at is how numerous, how many different ethnicities are represented, and if there’s any sort of particular concentration. Is that a fair characterization?

[00:13:00] Jeremiah Green: Yeah, exactly. That’s, I think that’s what they’re doing. Yeah.

[00:13:03] Joe Selvaggi: Okay. And of course, now that the other variable is how does that relate to performance? Again, if diversity is hard to define, so also is performance. But, what yardstick did McKinsey use in this, in these studies?

[00:13:17] Jeremiah Green: Yeah, they used, earnings before interest and taxes divided by revenues. So, it’s a measure of profitability, how, given the revenues that they generate, how profitable are those revenues in generating sort of a bottom-line profitability measure.

[00:13:34] Joe Selvaggi: So, this is your expertise, they’re consultants and they, they have their criteria. As an expert yourself, is that a fair measure? Should you use that profitability score as a measure of performance?

[00:13:48] Jeremiah Green: Yeah, I think it’s okay. You could argue for other measures, right? We might try and measure, return to investors or something like that, but it’s an important part of performance and probably a decent place to start.

[00:14:02] Joe Selvaggi: So, we’ve got two scores. We’ve got diversity and we’ve got performance. And again, this is a layperson, analysis of their work. As I see it, what they did was say they looked at all firms. They didn’t make clear which firms they were looking at. But all the firms, I guess almost 2,000 firms, and they ranked them in quartiles.

[00:14:20] They said, okay, we’ve got bottom performing, bottom middle, upper middle, and top performing. And they compared the diversity score of those in the Yeah, the profitability of those in the top, I reverse this, they had, diversity scores from a bottom quintile, top quintile, and they compared the performance in the top quintile of diversity versus the performance in the bottom quintile of performance, and then looked at the difference between the two. Share with our listeners. How did that look?

[00:14:49] Jeremiah Green: Yeah, we’ll stick with the general explanation, like you’re talking about, diversity and performance, rather than get into the details of that right now, but they’re talking, by performance, they’re talking about outperformance or underperformance.

[00:15:04] They’re comparing average profitability, relative to their industry. So If you have, if a company were to have Profitability that’s higher than their industry average, those are outperformers. If it’s below their industry average, they’re underperformers. So that’s the Specific profitability part. In terms of the diversity, what they do is rank companies and in the top 25%, They compare those in the top 25 percent of diversity to those in the lowest 25 percent of diversity.

[00:15:39] And then they look and see how much more likely is the top diversity group than the bottom diversity group to outperform their industry. So, they find something, like around, they’re like 30 percent more likely to outperform their industry if they’re in the top, diversity group.

[00:15:59] Joe Selvaggi: So that’s the crux of the claim, which is to say the more diversity, the firms with higher diversity scores were more likely to outperform their average in their industry, and those on the low end of the diversity score were more likely to underperform, and they compared those two, to really, arrive at whether it’s a, a trivial difference or a substantial, when we’re looking at difference between signal and noise, all kinds of things can show up when we do analysis.

[00:16:23] The larger that difference, the more confident they were in their conclusions. I’m not sure I don’t have the numbers right in front of me, but by my analysis, there was a substantial difference in the likelihood of profitability for more diverse firms than less diverse firms, giving them, I would say, again, reflected in the summary of their paper, high levels of confidence that diversity and performance are related, profoundly related. So, say more about. How confident their observations were and why they would be confident.

[00:16:55] Jeremiah Green: Yeah, I think you said two things there. So, one is the magnitude of how much they’re likely to outperform. So, something like a 30 percent likelihood, more likely to outperform is a huge effect. I, I don’t know of hardly anything else that is that magnitude in terms of how important it would be, right? You can’t like, say, if you, let’s say implement a new information system in your company, you’re 30 percent more likely to outperform. That just doesn’t exist, right? Maybe 1 percent or something. that’s in terms of the magnitude of the effect. The other part is the statistical significance, and they report p values that are less than 5% that would mean highly statistically significant.

[00:17:41] Joe Selvaggi: Again, for our listeners who are not statisticians, I’m going to I’m going to take a crack at explaining P values. I’m dusting off my, graduate degree here and say, what we know about P values is we want to be sure that our observation, or we want to be, measure the confidence that our observations weren’t randomly generated.

[00:17:58] Taking any given set of numbers, All kinds of things can happen. the larger the result that you observe, the more likely that it isn’t random. it’s likely to actually be measuring something real rather than a statistical artifact. Is that fair?

[00:18:14] Jeremiah Green: Yeah, that’s, I think that’s a reasonably good way to, state it loosely. I would add one more piece that, it’s about how the results in the sample that you’re using, how likely is that to happen again in another sample? And so, if we have a really small p value, we would expect, that says it’s very unlikely that what we see in our sample is a statistical anomaly. If we go to another sample, we expect to see the, a very similar result. So low P value, highly confident that this is a reliable claim.

[00:18:52] Joe Selvaggi: So, we imagine a large salami, every slice is going to be the same. If we’re low P value, wherever we look, we’re going to find a similar result rather than, some next, the next result to be the opposite. We’re fairly confident that this is a consistent outcome.

[00:19:05] All right. So, then I wanted to talk about one more dimension, which is we’ve established that at least McKinsey asserts that there’s a relationship between, diversity and performance. But there’s another step, right? We have to observe that in their case, there’s a causational relationship, which is to say, as you say, it’s both big, but it also has to, one has to come before the other.

[00:19:28] The diversity has to happen and then the performance has to happen for it to be causal. I hope I’m saying that Share with our listeners. In their analysis, does the way they measure and when they measure diversity and performance, does it really show a causational relationship? You get, you are diverse, then you outperform. Is that what this study really shows?

[00:19:50] Jeremiah Green: Short answer, no. before getting too critical about what McKinsey says, they explain this in their papers, right? They acknowledge that this is a problem. When you read their conclusions, that’s a very different, they don’t match their conclusions to what they’re saying, but they, let’s take the first study for an example. They rank diversity in 2017, and then the performance, the profitability measure that they use is before the diversity ranking. So, they’re, just with their design, it’s impossible for diversity to increase performance because performance happens before the diversity happens.

[00:20:28] Joe Selvaggi: So, the proverbial cart before the horse, or this is something that academics in your field wrestle with all the time, right? Correlation is not causation. I might use a, I’ll use something silly and say, umbrella use, and rainy days seem to correlate, but it doesn’t mean umbrellas cause it to rain. You know, for our listeners, just briefly explain why, what a data scientist has to do to establish causation rather than mere correlation.

[00:20:54] Jeremiah Green: Yeah, what do we have to do to establish causation? That’s difficult. the best-case scenario would be we’re in a lab where we can randomly assign Something to happen, and then we see that the effect happens. a couple of things we can do. One is the direction that you’re taught, that we had just talked about, right?

[00:21:14] It, the thing that causes something else should happen beforehand. the other part is we have to deal with other, what we call correlated omitted variables, something else that could explain the results that, we were trying to claim are causal. So, you use the example of the umbrella. Another typical one that I find easy to conceptualize is, ice cream sales and shark attacks are correlated. And we don’t think ice cream sales drive shark attacks. Because we think something else is driving both of those, and that’s warmer temperatures. And we have to deal with those. I guess two things. We have to go through the logic of, is this plausible? And then we also have to try and deal with other things that could cause that. In the case of ice cream sales, we would need to find a way to control for warm temperatures. If we found, for example, that, using diversity, that, if we have some 3rd variable, but we can find that it happens outside of when that third variable could affect it, then okay, we’re, at least more confident that there’s a cause and effect happening.

[00:22:29] Joe Selvaggi: Yeah. and so they acknowledge that I’d say in the footnotes in this study, they do acknowledge the challenge with establishing causation. So, I thought, the, people who are doing the analysis Unlike the people who write the headline for McKinsey, acknowledge that the data does not support a causational, but rather just a correlational.

[00:22:45] I want to caveat that if we’re going to criticize McKinsey for something. Is that fair? Meaning that the, the people doing the research acknowledge this limitation?

[00:22:53] Jeremiah Green: Yeah, perfectly fair. And I have no problem with their research at all. I have a problem with the connections of the, like you said, the headlines that they’re putting out there don’t match their research, right? And so that’s right. I think they’re pretty fair when you actually read inside the research there. I think they’re pretty straightforward about it.

[00:23:16] Joe Selvaggi: All right, so we’ve set up this whole sort of conversation. Let’s get to it. okay, you say you, connect more with the, the research of McKinsey rather than the headlines of McKinsey, but let’s get into the research. Was McKinsey’s report, were you able to essentially get their data sets and do analysis of the simple math, or was that something that’s, difficult to, for them that they’re reluctant to share or they were unclear about?

[00:23:39] Jeremiah Green: Yeah, okay, so a few things, in terms of the method they use, they’re very clear. So that, it’s not hard to go in and, if you have the data, to do what they did. The piece that they don’t provide is the companies that they’re using. we don’t have access to, their sample is something like, 300 or 500 companies, nobody but them, they’re the only ones that know what those companies are. And being able to perfectly replicate what they do is not possible unless they give us their companies. But other than that, the rest is very straightforward. You can just copy and do exactly what they do.

[00:24:19] Joe Selvaggi: And of course they didn’t share which companies, not because they’re trying to keep it a secret, but it’s discretion. They don’t want to, in a sense, out their clients, right? They, that’s reasonable. They have a responsibility to keep that confidential. You did find your own companies to analyze. As we mentioned earlier, the P value is very low, suggesting that you’re going to find this result more or less everywhere you look. How did you develop, again, if you’re trying to replicate without the actual list of companies, you’re just replicating the methodology rather than the actual dataset, where did you go to find companies to analyze?

[00:24:50] Jeremiah Green: Yeah, sure. We tried a few things. It’s, this is really the biggest challenge, I think, in terms of trying to replicate their research. We there are so many companies out there. If we look globally, it’s hard to know where to start, but we know that they consult with a bunch of the largest companies, S&P 500 is a pretty common company.

[00:25:15] Benchmark for all companies. So that’s where we started. So, S&P 500, it’s about the same size as the, their sample. And we know that they consult with many of them. And so, to us, that was as close as we could get. as a side note, we also did a random sample, we thought, of U.S. company, sorry, public companies, and tried the same thing.

[00:25:38] And no matter what we did, we get similar results, but it, I don’t know, is that good enough? We tried the S& P500, and we tried a random sample, to try and approximate what we think they did.

[00:25:50] Joe Selvaggi: Yes, and of course, if their, observations would be valid, and your observations would be valid, you, they couldn’t, in a sense, Use selection bias to craft their results if they did. They would say, among those companies, among the companies we looked at, this is true, but who knows about any other company? If that were the case, then no assertion would be generalizably true. It would be a worthless observation. In order to have their observations have any weight, they have to be generalizable your analysis and their analysis, though they’re different companies, in theory, should show the same approximate result. Is that fair?

[00:26:22] Jeremiah Green: Yeah, yeah, I would say even more than in theory. Their P values aren’t valid if Their sample is not a random sample, and so what their P value should be is a reflection of what we should see elsewhere. And yeah, if we can get a sample that is meant to approximate the same types of companies that they’re looking at, we should see something similar.

[00:26:45] Joe Selvaggi: So, the companies used, provided it’s a large enough sample, are virtually irrelevant to an observation like this. and you replicated their methodology, which you say they were very candid about how they went about evaluating. So we’ve buried the lead enough. Given all your research, you had two co-authors and a substantial amount of time to analyze the data, did your research see a correlation, not, we’re not going to get to causation yet, between diversity and performance?

[00:27:15] Jeremiah Green: I just want the short answer.

[00:27:16] Joe Selvaggi: No.

[00:27:18] Jeremiah Green: No.

[00:27:18] Joe Selvaggi: Okay. I don’t know if our listeners were anticipating that answer. It’s the shortest of all the answers you’ve offered. So I want to go, a little bit further and say, okay, we established early on that what we’re talking about diversity, we’re talking about diversity in the higher level ranks. We’re talking about a large signal that they saw that a large result of confidence in their results. You found no such, relationship between How diverse a firm was and how profitable a firm was. Did you use, the same exact, let’s say, I think it was earnings before interest and taxes. did you use, did you try to tweak the numbers in any way to look at different metrics that maybe you were missing some, some variable that they accounted for?

[00:27:55] Jeremiah Green: Yeah, let me give a little, more than you would see in the paper. So, in the paper we tried, I can’t remember, six or something different measures. Outside of that, I, we were convinced there’s got to be something to this. So, I tried every Measure that I could think of, so I don’t know what it was 25, 30 measures. There’s got to be something there was my thinking and no, no matter what we did, we couldn’t find it. and this is replicating their approach. We mentioned that they measured diversity and then looked at performance prior to that. And so, I tried that with every measure, we tried it with every measure we could think of.

[00:28:36] No association at all. And then in another paper, so this paper that we’ve been talking about is with John Hand. He’s at North Carolina. We also have another paper with Sekou Burmes, who is a strategy professor at North Carolina, and we flipped the direction. So, then we tried to say, what if we measure diversity, let’s say, in 2017 and looked at performance after 2017.

[00:29:02] And, maybe we could get this sort of the right direction, and then we would find results. Nothing. So, no matter what we did, and this is You know, you don’t see it all in there, but we spent an enormous amount of time trying to find any association that we could and just can’t do it.

[00:29:22] Joe Selvaggi: So, we’re not going to even talk about causation because where there’s no correlation, of course, there could be no causation in either direction that this is an extraordinary revelation in my view. It’s an extraordinary revelation. one would think even randomness might generate some sort of result, you could draw a line in the middle of a room and say, did the people on the left hand in the room have anything different than the people on the right-hand room? You’re going to find something somewhere. Guys named Joe may be different than guys, everybody else, you didn’t find anything.

[00:29:49] Jeremiah Green: I tried, I guess the only thing I didn’t try, we did the S& P 500 and the random sample. I guess another thing we could have tried is, just keep shifting around the companies and see if in any sample we could happen to find it, right?

[00:30:06] If it was just, you’ve looked through the data too much and randomly you should find a result. That, I guess that’s the other thing we could try, but yeah, everything we tried, there’s just nothing there. Oh, we also expanded it when we started, we did it with one year, just like a single one of their studies, and then we expanded it to many years, and no matter what, we just can’t find it.

[00:30:28] Joe Selvaggi: This is pretty powerful. When one finds no results, one can’t say there isn’t some relationship, but it’s just not measurable, particularly given the data and the methodology. This is big news, because I think McKinsey has a, their influence is powerful. corporate boards and, C suite people use that as I won’t say gospel, but essentially as a north star for many of their decisions. To make the assertion that they do, that it essentially is almost. not almost. It is proved that diversity leads to greater profitability. If I’m a leader of a firm or a member of a board, I have a fiduciary responsibility to my shareholders, right? I don’t own the company. I may be the president, but I don’t own the company. I work for the shareholders. It’s my job. I have to make choices for their benefit. If it’s true, what McKinsey asserts, then I would be, violating my oath as a fiduciary to not make my firm more diverse. Without this evidence, in a sense, now we can’t privilege diversity over others, let’s say, of course, we want to make the firm profitable.

[00:31:29] We’re not saying that this is morally or ethically a bad idea, diversity, we’re not going there. We’re just saying as a measure of profitability, as a head of a firm, it’s either imperative, it’s a financial imperative that I use diversity, but without that evidence, no such imperative exists. We can essentially use Pro more traditional, profitability criteria when making our decisions. I don’t want to go too bold here, but I want to say if this is, if this, data is not valid or observation is not valid, I’ll say maybe the data is, would this, upset the entire, diversity push in financial markets?

[00:32:10] Jeremiah Green: Yeah, let me, be careful in how I state this here. I think I agree that, if that were the case, and I think this is how what you were saying, has been used that way. If it is the case that increasing diversity increases performance, managers must increase diversity because that is going to improve returns to their shareholders. Fantastic, right? It’s a good story. If it works, then the path forward is clear. let me try and claw back some space for McKinsey here.

[00:32:44] In the last study, I, the one that just came out in 2023, so that was actually after we had, looked at the first three studies, they add other things that, that diversity might influence. for example, employee satisfaction. It’s completely believable that diversity could influence employee satisfaction.

[00:33:09] You could think of a lot of reasons that might be the case in terms of that, great, right? that may be something that companies should care about. It may be something that employees care about, and if a company decides to make decisions based on that, then no problem. The challenge is, though, they can’t rely on this evidence to say, it’s going to increase performance because that’s not the case. Let me give the other side of that is we don’t find that it decreases performance either. So, it, it doesn’t seem to hurt or help in terms of bottom-line performance. It could be as benefits in other ways, but just not, the sort of bottom-line profitability approach, or type of outcome.

[00:34:00] Joe Selvaggi: I’m glad you point out the double-edged sword of finding no results, which is diversity, when pushed doesn’t help, of course, but diversity when pushed doesn’t hurt it maybe is a Rorschach test to say which, which, support you want to, observe. that, that’s fair, and I’m glad you did point that out. Again, no signal gives us all latitude to, to look for other things. And as you say, employee satisfaction. But of course, if it improved employee, being or satisfaction or something like that, that theoretically, we want that. Of course, it’s nice, it’s inherently a good idea, but we would also think that would improve Thank you.

[00:34:31] Profitability. Again, if you want to signal, employee satisfaction isn’t correlated with profitability. It’s good, but it’s not a, a financial imperative. so what I really want to, again, I’m going to ask you to go on a limb here and say, okay, McKinsey has produced these four reports with glowing, summaries, and which are touted, when there’s legendary slide decks to, to C suite executives and boards, why would McKinsey, with, has built a reputation, I think they hire the best, smartest people coming out of the best schools, why would, McKinsey, make claims that their own data really wouldn’t support?

[00:35:07] It seems, or at least it is impossible to reproduce. What do you see as a sort of incentive for a firm to, say, jeopardize their reputation for data, quality data and integrity?

[00:35:20] Jeremiah Green: Yeah, I’m glad you said you’re going to ask me to go out on a limb because this is just speculation, right? I don’t know. Although we have examples in academic peer reviewed research where, researchers do things that would have bad consequences in terms of their reputation. So This does happen everywhere, so I can’t, claim that, we’re peer reviewed, so we’re so much better, but in terms of, I think, the incentives, we do have, some amount of, we are different, and I think the big difference, if I were to take a I don’t want to take a view like they were purposefully trying to, misstate what they are saying, but I think they, they want to be able to tell companies you should do this, and we don’t want to do that as academics, we just want to evaluate the evidence, and in the jump from them. Here’s what we found should. I think that’s where the misinterpretation of their evidence came from. That’s my guess. if you’re trying to tell a company you should do this, but then you tell them, if you have diversity, you previously would have performed better, that isn’t very compelling to get them to do something, I guess.

[00:36:37] Joe Selvaggi: There’s so many ways to interpret it. Again, we’re not inside the minds of McKinsey consultants, but of course, we all respond to incentives. Academics, hopefully academics’ reputation for integrity and accuracy is the sort of coin of the realm. In consultancy, perhaps, There is an incentive to tell your client what they want to hear, or I’ll say, again, I’m going to relate it to many topics we cover here on Hubwonk is, unfortunately, this sounds like a political, an aspirationally political position, which, and my, it’s going to go on our, my tombstone, perhaps, is the unfortunate fact that when you mix science and politics, you get politics.

[00:37:13] This seems to be just one more instance where a revered institution has Perhaps, wandered off the path and compromised integrity for, results that people prefer. you know, do you see this sort of perhaps as a potential victim of either group thinking or motivated reasoning?

[00:37:31] Jeremiah Green: Yeah, I hesitate to make any judgment about the moral or ethical good or bad, is this good or bad? It does seem, I like your statement of if you mix science with politics, you get politics. That seems pretty accurate to me. So, what I do think is that we should be aware of when we’re interpreting what research says, and maybe particularly when it comes from, people like consulting firms or groups like consulting firms that we need to be careful of how it’s being interpreted. And I think, we, I hope we’ve shown that, at least in these studies, you can’t make the conclusions that McKinsey has been making, those are just not, they’re just not correct, they’re not what is shown in the data.

[00:38:23] Joe Selvaggi: And that’s fair, and one can understand why McKinsey, having found nothing in the data, that, the headline cannot read, we found nothing they took the ball and ran with it. I appreciate you taking your valuable time with us today, Professor Green, it’s a dry subject, it’s a difficult subject, but I think peer reviews, scrutiny of, academics with integrity is absolutely essential. We’re not asserting whether diversity has a moral benefit.

[00:38:44] We’re just examining whether it has the. the financial benefit that the study claims. So at least in this regard, you’re shedding some light in what seems to be a fairly, dark or cloudy area of analysis. Where can our listeners, find either your study or your paper? I was able to access it, there were no, paywalls or anything. Where might, are, and you’ve published quite a bit, where could our listeners read more about your work?

[00:39:10] Jeremiah Green: Yeah, a few places. If you just search Jeremiah Green at Texas A&MI pop up so you can find my webpage and my CV there. if you go to econ journal, watch, or Journal of Economics Race and Policy, that’s where one of our other ones is. Or you can look up some of the co-authors, John Hand at UNC or Kuber at UNC. I think it’s pretty easy on academics to find that stuff.

[00:39:36] Joe Selvaggi: Wonderful. I appreciate your time and thank you again for joining me today on Hubwonk. Professor Green, you really, I think shed some light on a difficult topic.

[00:39:43] Thank you. for having me.

This has been another episode of Hubwonk. If you enjoyed today’s show, there are several ways to support Hubwonk and Pioneer Institute. It would be easier for you and better for us if you subscribe to Hubwonk on your iTunes podcatcher. It would make it far easier for others to find Hubwonk if you offer a five star rating or a favorable review.

[00:40:03] We’re of course grateful if you share Hubwonk with friends. If you have ideas or comments or suggestions for me about future episode topics, you’re welcome to email me at hubwonk@pioneerinstitute. org. Please join me next week for a new episode of Hubwonk.


Joe Selvaggi talks with business data scientist Dr. Jeremiah Green about his peer review work examining consulting firm McKinsey’s studies on the measurable financial benefits of diversity in corporate executive leadership.



Jeremiah Green is an Associate Professor of Accounting and holds the Ernst & Young Professorship of Accounting at the Mays School of Business at Texas A&M University. He does research on executive race/ethnicity in US public companies, and capital markets research that focuses on the use of accounting information. Dr. Green also studies hedge funds, equity and debt analysts, auditors, managers, the business press, and equity trading strategies. His teaching centers on data analytics and analytics for financial reporting.