INTRODUCTION
1. Nicholas Kristof, “President Trump, Meet My Family,” New York Times Sunday Review, January 28, 2017.
2. Vance, Hillbilly Elegy.
3. Sugrue, “Less Separate, Still Unequal.”
4. Danielle Allen, “Toward a Connected Society.”
5. Friedman, The World Is Flat.
6. Carnevale and Smith, “Economic Value of Diversity.”
7. Frey, “ ‘Diversity Explosion.’ ”
8. See, for example, Patricia Leigh Brown, “Silicon Valley, Seeking Diversity, Focuses on Blacks,” New York Times, September 3, 2015.
9. Shetterly, Hidden Figures.
10. Gurin, Lehman, and Lewis, Defending Diversity.
11. Brief of Amici Curiae, Lt. Gen. Julius W. Becton Jr. et al., Grutter v. Bollinger, No. 02-241 (2003).
12. Carnevale, Jayasundera, and Gulish, America’s Divided Recovery.
13. Gender Summit North America, Diversity Fueling Excellence, 2.
14. Grutter v. Bollinger (02-241) 539 U.S. 306 (June 23, 2003), Sandra Day O’Connor, majority opinion, 3–4.
15. Danielle Allen, “Toward a Connected Society,” 90.
PROLOGUE
1. Over the past decade, I have had the opportunity to speak with, among other groups, the US Office of Personnel Management, the Minnesota Association of Independent Schools, Boeing, Google, the Utah Medical Center, Gilead, Northrop Grumman, AB InBev, the US Federal Reserve, the US Treasury, Bloomberg, Microsoft, Yahoo!, Ford, General Motors, Nissan, Caterpillar, Cummins, Molex, Genentech, Legg Mason, the American Medical Association, the American Dental Association, the World Bank, the International Monetary Fund, the World Economic Forum-Davos, the Aspen Institute, Greenhills School, Credit Suisse, First Boston, Motorola, Tyco, the United States Air Force, Louisiana Tech University, Princeton University, MIT, Harvard University, Stanford University, North Dakota State University, Purdue University, Iowa State University, Sandia National Laboratories, Livermore National Laboratories, TotalSAP, Miller-Coors, DARPA, Johnson Controls, US Cellular, PIMCO, the US Department of Justice, and NASA. I have learned from discussions with thought leaders in and outside the academy. Eric Ball, Jenna Bednar, Jon Bendor, Wendell Berry, Lazlo Bock, John Seeley Brown, Daniel Diermeier, Amy Dittmar, John Hagel, Melody Hobson, Lu Hong, Joi Ito, P. J. Lamberson, Sheen Levine, Katherine Phillips, Jeff Polzer, Carl Simon, Daryl Smith, Omar Wasow, and my editor Eric Crahan offered ideas, comments, and challenges. Andrea Jones-Rooy, Juliet Bourke, Nancy Cantor, and Earl Lewis commented on and emended earlier drafts of the book. My family, Orrie, Cooper, and Jenna, provided unwavering support for this project, as our continuous lives moved along. In addition to debriefing me after every talk and helping to frame the entire project, Jenna made line-by-line improvements to the manuscript.
2. Phillips, “How Diversity Makes Us Smarter.”
3. See Economist, “Diversity Fatigue.”
4. Corey, “More Moderate Diversity.”
5. Thomas and Ely, “Making Differences Matter.”
6. Ibid.
7. The data also suffer from two identification problems that I cover at length.
8. Mannes, Soll, and Larrick, “Wisdom of Select Crowds.”
CHAPTER 1
1. This account is borrowed from historian Robert McNamara’s article “Abe Lincoln and His Ax.”
2. Goodwin, Team of Rivals.
3. Bendor and Page, “Optimal Team Composition.”
4. I return to diversity measures when discussing heuristics.
5. Each of his three tools can be paired with each of Barry’s four tools for a total of twelve.
6. National Research Council, Enhancing the Effectiveness of Team Science, 93.
7. See Foresight, “Obesity System Map,” https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/296290/obesity-map-full-hi-res.pdf.
8. Pollack, Only Woman in the Room.
9. Data available from the National Science Foundation and the American Mathematical Society. See Vélez, Maxwell, and Rose, “2013–2014 New Doctoral Recipients.”
10. Accuracy was measured by the squared distance between the predicted rating and the actual rating. If a person rated The Shawshank Redemption as three stars and a participant’s model predicted five stars, the squared error would equal four.
11. In the Netflix Prize contest, more than 98 percent of the one hundred million rankings went into a training set. The remaining data were divided into several testing sets to determine accuracy and the contest winner.
12. Van Buskirk, “How the Netflix Prize Was Won.”
13. As an exercise, take a moment and write down features of movies that might explain customer ratings. Congratulations if you can think up one hundred.
14. Tetlock, Expert Political Judgment.
15. Autor, Levy, and Murnane, “Skill Content of Recent Technological Change”; Autor and Price, “Changing Task Composition.”
16. Katznelson, When Affirmative Action Was White.
17. Argote and Epple, “Learning Curves in Manufacturing.”
18. Bessen, Learning by Doing.
19. Knowledge may not transfer across identity groups equally. Reagans and McEvily, “Network Structure and Knowledge Transfer.”
20. Reilly et al., “Randomized Trial of Occlusive Wrap.”
21. Woolley et al., “Evidence for a Collective Intelligence Factor”; Suroweicki, Wisdom of Crowds; Rheingold, Smart Mobs.
22. National Research Council, Enhancing the Effectiveness of Team Science.
23. See Uzzi et al., “Atypical Combinations and Scientific Impact,” and Freeman and Huang, “Collaborating with People like Me.” I take up these studies in more detail later.
24. Jehn, Northcraft, and Neale, “Why Differences Make a Difference.”
25. See Mauboussin, Callahan, and Majd, Organizational Structure and Investment Results. Patel and Sarkissian (“To Group or Not to Group?”) find a 58 basis point advantage for teams of size three over funds run by individuals.
26. Ellison, Invisible Man, 577.
CHAPTER 2
1. Gardner, Frames of Mind; Gardner, Intelligence Reframed.
2. Hewlett, Marshall, and Sherbin, “How Diversity Can Drive Innovation.”
3. Jehn, Northcraft, and Neale, “Why Differences Make a Difference.”
4. Ross and Malveaux, Reinventing Diversity.
5. Kahneman, Thinking Fast and Slow.
6. Gawande, Checklist Manifesto.
7. Johnson, Where Good Ideas Come From.
8. Two varieties might both have white petals and yellow cups, flourish in hardiness zones 3–8, and bloom in mid-spring. The USDA Plant Hardiness Zone Map divides the country into zones based on their average minimum winter temperature. Lower-numbered zones have colder winters.
9. Page, Model Thinking.
10. Dawes, “Robust Beauty of Improper Linear Models.”
11. Tetlock and Gardner, Superforecasting.
12. Tetlock, Expert Political Judgment.
CHAPTER 3
1. See Clarke and Primo, Model Discipline, for a philosophical treatment on the relationship between models and data in the social sciences.
2. Estimates place this between fifty and one hundred thousand feet.
3. Mauboussin, Success Equation.
4. Kleinberg et al., “Prediction Policy Problems.”
5. I thank Sendhil Mullainathan for these examples.
6. See Page, Difference.
7. In the appendix, I present the formal mathematics, along with numerical examples.
8. Romer and Romer, “FOMC versus the Staff.” There exists a more sophisticated variant of the theorem called the bias variance decomposition theorem. Similar logic to that revealed by the diversity prediction theorem can be found in treatments of Bayesian model averaging and ensemble methods.
9. As an experiment, I predicted outcomes in the 2016 NCAA basketball tournament using alphabetical order. I predicted Xavier, Villanova, and Wisconsin would all lose in the first round. I finished in the bottom 1 percent out of thirteen million people on the ESPN Tournament Challenge. Adding my predictions to my family’s pool of predictions reduced the accuracy of the collective prediction.
10. If each of five people predicts that there will be twelve eggs in a dozen and a sixth predicts that there will be only six, then the collective prediction will be eleven. The diversity prediction theorem can then be written as 1 = 6 − 5.
11. Waldron, “Wisdom of the Multitude.”
12. Tetlock, Expert Political Judgment.
13. The two students’ average guess will be 265 miles, an error of only 15 miles.
14. Goldstein, McAfee, and Suri, “Wisdom of Smaller, Smarter Crowds”; Page, “Not Half Bad.”
15. Dietterich, “Ensemble Methods in Machine Learning.”
16. Breiman, “Random Forests.”
17. These are in the form of if-then rules.
18. Brown et al., “Diversity Creation Methods”; Liu and Yao, “Ensemble Learning via Negative Correlation.”
19. Mellers et al., “Identifying and Cultivating Superforecasters”; Satopää et al., “Partial Information Framework.”
20. Manski, “Interpreting the Predictions of Prediction Markets.”
21. Sobel, Longitude.
22. The invention of the mood ring is widely credited to Marvin Wernick, who marveled at thermotropic tape in a hospital emergency room.
23. Weitzman, “Recombinant Growth”; Johansson, Medici Effect.
24. Arthur, Nature of Technology.
25. Leung et al., “Multicultural Experience Enhances Creativity.”
26. McLeod, Lobel, and Cox, “Ethnic Diversity and Creativity.”
27. Triandis, Hall, and Ewen, “Member Heterogeneity and Dyadic Creativity.” The Internet has expanded the number of possible answers to this second question.
28. Guilford, Nature of Human Intelligence.
29. An alternative diversity measure computes the ratio of unique ideas to the total ideas by the group. Two people who come up with the same ideas will have a diversity of zero. Two people who have no overlap in their ideas have a diversity of one because the number of unique ideas equals the number of ideas total. To make this formal, we can let S1 and S2 denote the sets of ideas from two people. Let S1 S2 be the ideas in S1 and not in S2 (define S2 S1 similarly) and let S1 ∪ S2 be the union of the two sets. The diversity, D(S1, S2), can be represented as
A third measure of the diversity of a group computes the ratio of the total number of unique ideas to the sum of the number of ideas from each person.
30. If we do not know the ideas that people have drawn, then we would expect the group of the most creative people to have the most ideas. For any particular realization of draws, the most creative group could contain someone who is not among the most creative. To identify such a person would require knowing the ideas that everyone drew.
31. Johansson, Medici Effect; Padgett and Powell, Emergence of Organizations and Markets.
32. To arrive at these numbers of possible pairs, take the number of unique ideas (the first idea) times the number of ideas it can be paired with (the number of ideas remaining) and then divide by two to avoid double counting. The eighteen unique ideas combine to make 18 × 17 ÷ 2 = 153 pairs.
33. Weitzman, “Recombinant Growth”; Simonton, Origins of Genius.
34. Von Hippel, Sources of Innovation.
35. Kleinberg and Raghu (“Team Performance with Test Scores”) prove a technical result in a more general setting.
36. Mokyr, Gifts of Athena.
37. Acemoglu and Robinson, Why Nations Fail.
38. Gordon, Rise and Fall of American Growth.
39. Page, Difference; Bendor and Page, “Optimal Team Composition.”
40. Hong and Page, “Problem Solving by Heterogeneous Agents”; Kleinberg and Raghu, “Team Performance with Test Scores.”
41. Sadoway, “PhD Should Be PSD.”
42. Adding zero simplifies expressions. For example, adding zero, written as (6–6), transforms the expression x3 − 3x2 + 3x − 7 into x3 − 3x2 + 3x − 1 − 6 = (x + 1)2 − 6.
43. To calculate the probability of solving the problem, take one minus the probability of not solving the problem. The first person solves the problem with any given tool with probability 0.25. The probability that she doesn’t solve it with a given tool equals 0.75. The probability that no tools work equals (0.75)4 = 0.32. Therefore, she solves the problem 68 percent of the time. The second person doesn’t solve the problem with probability (0.7)3 = 0.34. The third person doesn’t solve the problem with probability (0.7)2 = 0.49.
44. Von Hippel, Sources of Innovation.
45. Hong and Page, “Problem Solving by Heterogeneous Agents”; Hong and Page, “Groups of Diverse Problem Solvers.”
46. To be more precise, the probability that they have identical abilities would be low. Equal abilities would require that each combination of tools has the same probability of solving the problem. If we were to randomly assign potentials to tools, that would be an unlikely event.
47. Hong and Page, “Groups of Diverse Problem Solvers.”
48. Weitzman, “Optimal Search for the Best Alternative.”
49. If we assume uniformly distributed solution values, the expected value equals a little over 6.8.
50. Kleinberg and Raghu, “Team Performance with Test Scores.”
51. One small technical issue arises as well. Their ability measure depends on group size. The most able person for a group of size six may not be the most able for a group of size eight. That is only true given their assumption that solution values are statistically independent. The statistical independence assumption requires diverse repertoires.
52. Tully, Gilmer, and Shugard, “Molecular Dynamics of Surface Diffusion.” I have known Mary Shugard for over twenty years but had not known this paper.
53. The term Brusselator, coined by Nobelist Ilya Prigogine, combines Brussels, the city where the model was developed, with oscillator, the phenomenon produced by the model.
54. Daniel T. Gillespie (“Exact Stochastic Simulation”) describes how the deterministic approach creates a continuous, predictable outcome.
55. The research of Flo Gardipee, a Cherokee Irish wildlife biologist, provides another instance in which a person’s identity contributed to a scientific breakthrough in the case of a constraint. Gardipee studied bison and needed DNA samples. Based on her beliefs, she sought a noninvasive method for collecting DNA samples. She came up with the idea of using fecal samples, now a standard procedure. Gardipee et al., “Fecal DNA Sampling Methods.”
56. Hutchins, Cognition in the Wild.
57. Boulding, Economics as a Science.
58. Sternberg and O’Hara, “Creativity and Intelligence.”
59. Campbell, “Blind Variation and Selective Retention.”
60. Simonton, Origins of Genius.
61. Sternberg and O’Hara, “Creativity and Intelligence.”
62. Ahuja and Lampert, “Entrepreneurship in the Large Corporation”; Fleming, “Recombinant Uncertainty in Technological Search.”
63. Sarah Kaplan and Vakali, “Double-Edged Sword of Recombination.”
64. Youn et al., “Invention as a Combinatorial Process.”
65. The analysis by Youn et al. (ibid.) counts the number of patents P, the number of technology codes T, and the number of unique combinations of codes C. Note that given this encoding, a patent that introduces a new technology code and links to no other codes creates a new combination of size one. They find that up until 1860, the number of patents closely tracks both T and C. Since that time, T has fallen off dramatically. Even if it were the case that the patent office now creates fewer categories, the evidence compellingly shows the preponderance of recombinations.
66. Knox, Lost at Sea.
67. This part of Lost at Sea is a creative task.
68. My former student Ryan Issacs had the idea of using LSAT-type logic questions to demonstrate the concept of knowledge integration.
69. The probability that the second team makes the correct move can be computed as follows: All five will be correct with probability (0.8)5 = 0.33. There are five combinations in which four choose correctly and one is incorrect. Each of these has probability (0.8)4(0.2) = 0.08. And there are ten combinations in which three are correct and two incorrect. Each of these has probability (0.8)3(0.2)2 = 0.02. Combining gives 0.33 + 5 (0.08) + 10 (0.02) = 0.93. The same intuition drives Condorcet’s jury theorem, which assumes a set of voters, each of whom independently knows the correct answer with the same probability. If that probability exceeds one-half, four results follow: the majority identifies correctly with a higher probability than each individual, collective accuracy increases in individual accuracy and in group size, and large groups approach but never achieve perfect accuracy.
70. Marcolino, Jiang, and Tambe, “Multi-agent Team Formation.”
71. Simon, Administrative Behavior.
72. De Bono, Six Thinking Hats.
73. Bourke, Which Two Heads?
74. These examples can all be found in a report on analytic tradecraft. US Government, Tradecraft Primer.
75. CIA, Diversity and Inclusion.
76. Steven Kaplan and Lerner, “It Ain’t Broke.”
77. Jurvetson, “Brainiac Steve Jurvetson.”
78. Ban, “Role of Serendipity.”
79. Chivian and Bernstein, “Role of Traditional Medicine.”
80. Hollister, “21 Different Interpretations.”
81. Knight and Johnson, Priority of Democracy, 1.
82. Gurin, Nagda, and Zuniga, Dialogue across Difference.
83. Kahneman, Thinking Fast and Slow.
CHAPTER 4
1. Phillips, “How Diversity Makes Us Smarter”; Jackson and Joshi, “Work Team Diversity.”
2. These can be found at www.metamia.com, an analogy website.
3. Medin and Ortony, “Psychological Essentialism.”
4. Zimmer, “White? Black?”
5. Some identity attributes, including sexual orientation, remain the subject of heated debate within academic, scientific, and religious communities.
6. Wood et al., “Role of Common Variation.”
7. How the Dutch got so tall (besides being a good name for a children’s book) intrigues scientists. Early evidence supports a genetic contribution. Tall Dutch have been reproducing at a faster rate than short Dutch, meaning that tall genes have been reproducing at a faster rate than short genes. Stulp et al., “Does Natural Selection Favour Taller Stature?”
8. Hall, Beyond Culture.
9. More than 80 percent of the students in the more than fifty evangelical student groups at the University of California’s two main campuses, Berkeley and Los Angeles, identify as Asian American. Korean Americans represent a majority of these evangelicals. Kim, God’s New Whiz Kids.
10. Harris and Sim, “Who Is Multiracial?”
11. Sen and Wasow, “Race as a ‘Bundle of Sticks.’ ” Sen and Wasow rely on a bundle-of-sticks analogy. I use the timber-framed house analogy to highlight that some pairs of attributes will be more closely connected than others.
12. Cole, “Intersectionality and Research in Psychology.”
13. Ibid.
14. Crenshaw, “Demarginalizing.”
15. Setting aside the social maladroitness of such a request.
16. Appiah, “Uncompleted Argument”; Michael James, “Race.”
17. Pattillo, Black Picket Fences.
18. Nisbett, Geography of Thought.
19. Cattell, Abilities.
20. Flynn, “Massive IQ Gains in 14 Nations”; Flynn, What Is Intelligence?
21. Cox, Navarro-Rivera, and Jones, “Race, Religion, and Political Affiliation?”
22. Anderson, Imperative of Integration.
23. Heath and Heath, Made to Stick.
24. Bourke, Which Two Heads?
25. Google Ads’ settings page is located at https://www.google.com/settings/u/0/ads/authenticated.
26. Algorithms that can predict race and gender can discriminate based on those attributes by biasing the ads people see or by offering different interest rates. See Sweeney, “Discrimination in Online Ad Delivery,” and Consumer Finance Protection Bureau, Using Publicly Available Information.
27. The oil applied to clean guns contaminates the process.
28. Buchanan, “Stars Who Were Almost Cast.”
29. I present a similar version of the story in The Difference. I repeat it here because it is so provocative.
30. Martin, “Egg and the Sperm.”
31. Program head Robert O. Bernard would later include her formulation in an academic paper.
32. Kaufman, “Watch David Bowie.” MTV eventually integrated due in part to Michael Jackson’s “Billie Jean,” “Beat It,” and “Thriller” trilogy.
33. Antonio et al., “Effects of Racial Diversity”; Gurin, Nagda, and Zuniga, Dialogue across Difference.
34. In 2010, President Obama challenged the federal government to increase its employment of differently abled workers by one hundred thousand. That action will have substantial direct effects given the different cognitive repertoires of differently abled people. It will also have large indirect effects through the increased awareness of others.
35. Levine et al., “Ethnic Diversity Deflates Price Bubbles.”
CHAPTER 5
1. Mannix and Neale, “What Differences Make a Difference?”
2. Data from Kopf, “How Many People Take Credit?”
3. Freeman and Huang, “Collaborating with People like Me”; Patel and Sarkissian, “To Group or Not to Group?”; Tetlock and Gardner, Superforecasting.
4. Hunt, Layton, and Prince, “Why Diversity Matters.”
5. Dawson, Kersley, and Natella, CS Gender 3000.
6. Note that foreign nationals should not be equated with underrepresented minorities.
7. Dezsö and Ross, “Female Representation.”
8. Sparber, “Racial Diversity and Aggregate Productivity”; Florida and Gates, “Technology and Tolerance.”
9. Sparber (ibid.) also analyzes the correlation between the effect of racial diversity and responses to the Department of Labor O*NET survey. He selected four questions that asked whether people make decisions and solve problems, think creatively, serve customers, or work as part of a team. The four questions he used from the O*NET survey were the following: How important is making decisions and solving problems to the performance of your current job? How important is thinking creatively to the performance of your current job? How important is customer and personal service knowledge to the performance of your current job? How important are interactions that require you to work with or contribute to a work group or team to the performance of your current job? He finds that racial diversity improves performance on problem solving, thinking creatively, and serving customers. He finds a negative effect of working as part of a team. Here as well, he finds meaningful effect sizes. The effect of a one standard deviation increase in racial diversity for an industry one standard deviation above average in decision making would produce a 4 percent increase in productivity. He estimates the corresponding effects for creative problem solving and customer service at 2–3 percent and 6–7 percent. The loss for an industry one standard deviation above the mean on the team problem-solving question is also in the 2–3 percent range.
10. Bettencourt, Samaniego, and Youn, “Professional Diversity.”
11. Sen and Wasow, “Race as a ‘Bundle of Sticks.’ ”
12. Bertrand and Mullainathan, “Are Emily and Greg More Employable?”
13. Ahern and Dittmar, “Changing of the Boards.”
14. The primary explanatory variables here must include experience and age. If we were to compare cognitive repertoires of the women appointed to boards to those of the men in their cohorts, the germane differences could not be so large as to explain a 20 percent drop-off in performance. The cohort of successful midcareer businessmen and businesswomen in Norway attended similar schools, have similar work experiences, and for the most part share a common ethnicity. If one attempts to account for systematic differences in board members’ attributes, gender effects remain, though they are less pronounced. The direct measurement of gender effects is subject to challenge because it assumes separability of effects. That assumption runs counter to the timber-framed house model, which argues that the contributions of a woman CEO cannot be decomposed into a woman component and a CEO component.
15. Ahern and Dittmar, “Changing of the Boards.”
16. Matsa and Miller, “Female Style in Corporate Leadership?” A recent study that chooses a different start date from all of the other papers finds little to no effect of gender. That study explains the negative finding as driven by a small number of firms that were reliant on government contracts, and those firms had greater gender equity prior to the law’s passage. Those firms performed well during the period of the study, a downturn, because of their government support. Eckbo, Nygaard, and Thorburn, “Gender-Balancing.” Those firms were coded as performing better yet not increasing in diversity. I mention this study not because I find it more compelling but because it reveals a subtlety of interpreting regression coefficients. The regression shows the effect of adding women to a board. The data include firms that already had substantial female representation. Those firms performed better than average. When you fit a regression line, those firms will be data points with no added women and strong performance. Those data will cause the regression line to slope downward. In other words, successful firms with women already on boards increase the negative coefficient for adding women to boards that do not have them.
17. Ahern and Dittmar, “Changing of the Boards”; Matsa and Miller, “Female Style in Corporate Leadership?”
18. If disruption was a contributing factor, we might expect smaller performance dips for companies with more experienced boards. Ahern and Dittmar, “Changing of the Boards.” We might also expect the opposite. The data do not support that hypothesis. If disruption hindered monitoring, it did so uniformly.
19. Open Science Collaboration, “Estimating the Reproducibility.” Findings from experiments involving people fail to replicate because what was true for one set of subjects at one point in time may not be true for another subject pool. People behave idiosyncratically and also conform. The behaviors of a few could steer an entire population to take an action. Physical scientists have fewer challenges with replicability. A measurement of the tensile strength of copper does not depend on the attitude of the copper that day.
20. Woolley et al., “Evidence for a Collective Intelligence Factor.”
21. Jackson and Joshi, “Work Team Diversity”; Mannix and Neale, “What Differences Make a Difference?”
22. Jackson and Joshi, “Work Team Diversity”; Williams and O’Reilly, “Demography and Diversity in Organizations”; Mannix and Neale, “What Differences Make a Difference?”
23. Lakhani and Jeppesen, “Getting Unusual Suspects”; Jeppesen and Lakhani, “Marginality and Problem-Solving Effectiveness.”
24. Woolley et al., “Evidence for a Collective Intelligence Factor.”
25. Woolley, Aggarwal, and Malone, “Collective Intelligence and Group Performance.”
26. For surveys of the literature, in addition to Williams and O’Reilly, “Demography and Diversity in Organizations,” see Jehn, Northcraft, and Neale, “Why Differences Make a Difference”; Jackson and Joshi, “Work Team Diversity”; and Van Knippenberg and Schippers, “Work Group Diversity.”
27. One study of sixty-four teams that perform intelligence analysis spanning six agencies found that six attributes explained three-fourths of the variation in performance. Successful teams have stable membership, a well-defined objective, supportive context, productive norms of conduct, resources, strong support, and access to coaching. Hackman and O’Connor, “What Makes for a Great Analytic Team?” A second study of 120 international business teams finds that by having the right people on the team, they can explain more than half of performance variation. Wageman et al., Senior Leadership Teams. A more recent series of laboratory studies also shows that the most effective teams include individuals able to recognize the reactions and emotional responses of other group members. Woolley et al., “Evidence for a Collective Intelligence Factor.”
28. Homan et al., “Bridging Faultlines by Valuing Diversity”; Thomas and Ely, “Making Differences Matter”; Swann et al., “Finding Value in Diversity.”
29. I served as an adviser to one of the teams.
30. Cattell, Abilities.
31. Mellers et al., “Identifying and Cultivating Superforecasters.” The elite teams of forecasters became even more accurate, whereas the individuals in the top 3–5 percent regressed to the mean.
32. Mannes, Soll, and Larrick, “Wisdom of Select Crowds.”
33. University of Michigan alumni will recognize the section heading as the words of Bo Schembechler. The quotation from which it is excerpted reads as follows: “No man is more important than The Team. No coach is more important than The Team. The Team, The Team, The Team, and if we think that way, all of us, everything that you do, you take into consideration what effect does it have on my Team? Because you can go into professional football, you can go anywhere you want to play after you leave here. You will never play for a Team again. You’ll play for a contract. You’ll play for this. You’ll play for that. You’ll play for everything except the team, and think what a great thing it is to be a part of something that is, The Team.” “Team Speech,” transcript.
34. Steinbeck, East of Eden, 131.
35. Janssen and Renn, “History.”
36. Shenk, Powers of Two.
37. Singh and Fleming, “Lone Inventors.”
38. Uzzi et al., “Atypical Combinations and Scientific Impact.”
39. Phillips, “How Diversity Makes Us Smarter.”
40. Shi et al., “Impact of Boundary Spanning”; Freeman and Huang, “Collaborating with People like Me”; Tetlock and Gardner, Superforecasting.
41. National Research Council, Enhancing the Effectiveness of Team Science.
42. Wuchty, Jones, and Uzzi, “Increasing Dominance of Teams.”
43. Patel and Sarkissian, “To Group or Not to Group?”
44. Wuchty, Jones, and Uzzi, “Increasing Dominance of Teams.”
45. National Research Council, Enhancing the Effectiveness of Team Science. Science and engineering rely more on teams than social science. This may be because science and engineering research involve more problem solving and prediction. Many social science papers describe history, derive theories, or offer interpretations and reviews of existing literatures. These activities require a coherent narrative, a trait that advantages single authors.
46. Jones, Uzzi, and Wuchty, “Multi-university Research Teams.”
47. Singh and Fleming, “Lone Inventors.”
48. Removing self-citations (which stacks the deck against larger teams, as coauthors often write much of the related research) does not alter the results. Teams still receive more citations in 159 of the scientific subfields and 51 of the social science categories.
49. Far fewer than one in a thousand papers earns one thousand citations. Fewer than 1 percent of social science papers earn one hundred citations.
50. Singh and Fleming, “Lone Inventors.”
51. Patel and Sarkissian, “To Group or Not to Group?” A basis point equals one-hundredth of a percent. So, this translates into a 0.6 percent increase in risk-adjusted return. To put this in perspective, investing $1,000 in a fund that returns 5 percent annually yields $11,467 after fifty years. A fund returning 60 basis points higher, or 5.6 percent, would yield $15,247.
52. Kozlowski and Bell, “Work Groups and Teams,” 4.
53. Adamic et al., “Individual Focus and Knowledge Contribution.”
54. Jones, Uzzi, and Wuchty, “Multi-university Research Teams.”
55. Freeman and Huang, “Collaborating with People like Me.” They use the Herfindahl Index as a measure of diversity. See Page, Diversity and Complexity, for a summary of diversity measures.
56. Shi et al., “Impact of Boundary Spanning.”
57. Uzzi et al., “Atypical Combinations and Scientific Impact.”
58. If novelty and conventionality were equally likely, each bin would contain one-fourth of all papers. That is not the case. While the papers not classified as conventional are roughly evenly split between novel and not novel, almost 44 percent of the conventional papers are not novel. Fewer than 7 percent of all papers are conventional and novel.
59. Schilling and Green, “Recombinant Search.”
60. Cummings et al., “Group Heterogeneity.”
CHAPTER 6
1. McDonald’s Corporation, “Our Ambition.”
2. REI Co-op, “About REI.”
3. I interpret the word business broadly to include research institutes, nonprofits, government agencies, and educational institutions. Efforts to alleviate poverty, cure cancer, and improve educational outcomes can gain as much from diversity as attempts to design driverless cars, safer airplanes, and user-friendly web-based platforms.
4. Cameron and Quinn, Diagnosing and Changing Organizational Culture.
5. See University of Wisconsin, “Mission Statement,” and Goodyear Corporate, “Our Responsibilities.”
6. Based on data taken from a widely reported analysis of LinkedIn. Business Insider, “20 Schools.”
7. Woolley et al., “Evidence for a Collective Intelligence Factor.”
8. Steele, Whistling Vivaldi; Carney et al., “Implicit Association Test (IAT).”
9. The intelligence community develops and maintains diversity in three ways. First, it casts a wide net to attract diverse employees. Second, it encourages those employees to develop deep, diverse repertories. The analysts working behind the walls at Fort Meade and Langley receive constant opportunities for training in new analytic tools and frameworks. Third, the community structures activities to build cognitive diversity. They constitute competing red and blue teams to assess vulnerabilities. The red teams attack. The blue teams defend. Each team develops distinct sets of skills and knowledge bases.
10. Quoted in Maheshwari, “Big Brands Ask Ad Agencies.”
11. In 2007, Beaner’s rebranded as Biggby’s at a cost of over $1,000,000. By 2011, it was the fastest-growing coffee chain in the country.
12. Cohen, Gabriel, and Terrell, “Case for Diversity.”
13. Sapienza, Zingales, and Maestripieri, “Gender Differences in Financial Risk Aversion.”
14. The same logic applies to diversity of nationalities. The Bank of New York Mellon operates in over one hundred markets spread across thirty-five countries. Google has seventy offices in more than forty countries. The Ford Motor Company exports to more than one hundred countries. Caterpillar sells farm equipment in more than one hundred countries.
15. Thomas and Ely, “Making Differences Matter.”
16. These data are all readily available from multiple sources. See Lewis and Cantor, Our Compelling Interests, for one overview.
17. Valantine and Collins, “National Institutes of Health.”
18. These men were very fast. The world record they set at the Olympics lasted more than twenty years.
19. Shetterly, Hidden Figures.
20. In this section, I lean on the prescient article by Thomas and Ely (“Making Differences Matter”), who describe the discrimination and fairness paradigm and the access and legitimacy paradigm and introduce the idea that diversity can produce benefits in an integrative fashion. My construction of diversity bonuses as a third narrative expands and formalizes their ideas.
21. This is his word choice, which I repeat out of respect for his witty opinions and not out of sarcasm.
22. Evidence for that claim can be found in support for affirmative action in admissions. In the 1960s and 1970s, elite universities admitted minority students in larger numbers but did little to promote inclusion or produce bonuses. Because the students from underrepresented groups were not integrated into the university community, they graduated at low rates. They also contributed less to their universities than they might have had the universities changed their cultures to be more inclusive. Sander and Taylor (Mismatch) argue that law schools continue to lack nuance in their admissions decisions. They also make a more disputed claim that this has proved disadvantageous to minority law students. The effort at inclusion produced insufficient bonuses. As a result, affirmative action policies soon lost public support. A 2013 Gallup poll found that Americans oppose affirmative action in college admission by more than two to one.
23. Personal correspondence from Sheen Levine, February 15, 2017.
24. Thomas and Ely, “Making Differences Matter”; Liff, “Two Routes to Managing Diversity.”
25. Swann et al., “Finding Value in Diversity.”
26. I thank Mark Wiseman for a fun interchange on horses and camels that resulted in this section of the book.
27. See again Thomas and Ely, “Making Differences Matter.” Mark Nivet refers to Diversity 1.0 as about numbers: discrimination and composition. Diversity 2.0 emphasized changing processes so that everyone could succeed individually. What he calls Diversity 3.0 links to mission. It seeks to identify and achieve diversity bonuses. See American Academy of Family Physicians, “Diversity 3.0.”
28. Thomas and Ely, “Making Differences Matter.”
29. Microsoft Corporation, “Mission Culture.”
30. Groysberg and Connolly, “BlackRock.”
31. Patel and Sarkissian, “To Group or Not to Group?”; Lutton and Davis, Morningstar Research Report.
32. Fink, “Annual Letter to CEOs.”
CHAPTER 7
1. Dover, Major, and Kaiser, “Diversity Policies Rarely Make Companies Fairer”; Dobbin, Schrage, and Kalev, “Rage against the Iron Cage.”
2. Gurin, Nagda, and Zuniga, Dialogue across Difference.
3. To see why requires thinking like a statistician. Imagine a process that produces potential employees. The cognitive repertoires of the potential employees can be thought of as coming from a distribution. Increasing the number of people considered increases the expected best employee. It also increases the maximum expected diversity between any two, three, or four people. Thinking in extremes helps see the logic. If there exist only two potential hires, neither may be very good and the two might not be diverse. From a pool of a million potential hires, there should exist some fantastic individuals and the potential for amazing, diverse teams.
4. Taylor and Binder, “Washington Talk.”
5. African Americans, in particular, have suffered from multiple forms of institutional discrimination. For example, African Americans did not share equally in New Deal programs or in the postwar GI Bill. Katznelson, When Affirmative Action Was White. These surely contributed to today’s economic and educational disparities. See also Roithmayr, “Reproducing Racism.”
6. Moss-Racusin et al., “Science Faculty’s Subtle Gender Biases”; Bertrand and Mullainathan, “Are Emily and Greg More Employable?” In considering the extent of discrimination by race in the workplace, we can distinguish between discriminatory actions and behaviors in hiring and promoting and differences in educational attainment. Differences in attainment in turn are partly the result of residential and social segregation. Discrimination results from a constellation of mutually reinforcing situations, policies, and behaviors. Therefore, we cannot end discrimination by reducing bias. O’Flaherty, Economics of Race. A more unified, multifront approach will be required. We see broader efforts by governments, foundations, universities, and some businesses. We see governments pushing for equality of resources in education and foundations coordinating on rebuilding communities. In Detroit, the Ford Foundation ($125 million), the Kresge Foundation ($100 million), the W. K. Kellogg Foundation ($40 million), the John S. and James L. Knight Foundation ($30 million), and the William Davidson Foundation ($25 million) participated in a grand bargain to salvage the pensions of city workers. We see universities devoting substantial resources to developing racially diverse student bodies and faculties and, as part of the anchor institutions movement, to building strong integrated communities.
7. Ross and Malveaux, Reinventing Diversity; Carney et al., “Implicit Association Test (IAT)”; Greenwald, Banaji, and Nosek, “Statistically Small Effects.”
8. Danielle Allen, “Toward a Connected Society.”
9. Fryer and Jackson, “Categorical Model of Cognition.”
10. Moss-Racusin et al., “Science Faculty’s Subtle Gender Biases”; Bertrand and Mullainathan, “Are Emily and Greg More Employable?”
11. Ayers, Banaji, and Jolls, “Race Effects on eBay.”
12. The National Football League’s Rooney Rule requires interviewing at least one minority for each head coaching vacancy.
13. My cousin Terry Page earned those second three degrees and went on to a faculty position at Johns Hopkins Medical School.
14. Scott, Radical Candor.
15. Bock, Work Rules!
16. Cameron and Quinn, Diagnosing and Changing Organizational Culture.
17. The company profile states “Recognized in 2015 as Canada’s Outstanding CEO of the Year, Mr. Cope has earned a reputation as an innovative communications strategist and builder of high-performance teams.” BCE, “Executive Team: George Cope.”
18. See Friedman, “How to Get a Job at Google.”
19. Woolley et al., “Evidence for a Collective Intelligence Factor”; Woolley, Aggarwal, and Malone, “Collective Intelligence and Group Performance.”
20. Uzzi et al., “Atypical Combinations and Scientific Impact.”
COMMENTARY
1. Phillips, Liljenquist, and Neale, “Is the Pain Worth the Gain?”; Sommers, “On Racial Diversity.”
2. Apfelbaum, Phillips, and Richeson, “Rethinking the Baseline.”
3. Phillips, Kim-Jun, and Shim, “Value of Diversity in Organizations,” 255.
4. See, for example, Phillips et al., “Diverse Groups and Information Sharing”; Gruenfeld et al., “Group Composition and Decision Making”; and Stasser, Stewart, and Wittenbaum, “Expert Roles and Information Exchange.”
5. Phillips, Kim-Jun, and Shim, “Value of Diversity in Organizations,” 255.
6. See, for example, Vernon L. Allen and Wilder, “Group Categorization,” and Tajfel and Turner, “Integrative Theory of Intergroup Conflict.”
7. Williams and O’Reilly, “Demography and Diversity in Organizations.”
8. Phillips, “Effects of Categorically Based Expectations.”
9. For research on groupthink, see Janis, Groupthink, and Turner and Pratkanis, “Twenty-Five Years of Groupthink Theory.”
10. See Sommers, “On Racial Diversity”; Phillips, Liljenquist, and Neale, “Is the Pain Worth the Gain?”; and Antonio et al., “Effects of Racial Diversity.”
11. Phillips, “Effects of Categorically Based Expectations”; Phillips and Loyd, “When Surface and Deep-Level Diversity Collide.”
12. Phillips and Loyd, “When Surface and Deep-Level Diversity Collide.”
13. Ibid.
14. Sommers, “On Racial Diversity.”
15. Antonio et al., “Effects of Racial Diversity.”
16. Loyd et al., “Social Category Diversity Promotes Pre-meeting Elaboration.”
17. See, for example, Vernon L. Allen and Wilder, “Categorization, Beliefs Similarity”; Vernon L. Allen and Wilder, “Group Categorization”; Heider, Psychology of Interpersonal Relations; and Holtz and Miller, “Assumed Similarity and Opinion Certainty.”
18. Phillips, “Effects of Categorically Based Expectations.”
19. Phillips, Liljenquist, and Neale, “Is the Pain Worth the Gain?”
20. Ibid.; Phillips, “Effects of Categorically Based Expectations”; Phillips et al., “Diverse Groups and Information Sharing”; Phillips and Loyd, “When Surface and Deep-Level Diversity Collide.”
21. Phillips, Liljenquist, and Neale, “Is the Pain Worth the Gain?”
22. Williams and O’Reilly, “Demography and Diversity in Organizations.”
23. Phillips, “Effects of Categorically Based Expectations.”
24. Ibid.
25. Ibid.; Loyd et al., “Social Category Diversity Promotes Pre-meeting Elaboration”; Antonio et al., “Effects of Racial Diversity.”
26. Phillips, “How Diversity Makes Us Smarter.”
27. Rosette, Leonardelli, and Phillips, “White Standard.”
APPENDIX
1. Page, Difference.
2. Ibid.