SEVEN

Coalitions of the Willing

I have never made a friend from whom I could not separate, and I have never made an enemy whom I could not approach.

—Tancredo Neves, president-elect of Brazil, 1985

Fighting and Reconciliation

ALL GOOD PRIMATOLOGISTS KNOW how important it is to be alert to the power of alliances, to the shifting currents of loyalty and betrayal that can make or destroy an individual's standing within the group. And that's just when dealing with their fellow primatologists. When they study the fate of individuals within a group of baboons or chimpanzees, they know that strength, cunning, and luck are not enough. Without the ability to win the support of others and to call on that support in the face of unexpected challenges, a group-living primate is a “poor bare, forked animal” alone in the storm.

Frans de Waal has been studying chimpanzees for nearly four decades and has gained worldwide fame for bringing the political maneuverings of chimpanzees to public as well as professional attention. From 1975 to 1981 he and a group of his students made a detailed study of fighting and reconciliation among the chimpanzees at the colony in Arnhem Zoo in the Netherlands. He believes that both aggression and reconciliation are forms of behavior that reflect adaptive strategies by primates, not merely some inexplicable breakdown of an otherwise rational social order. He found some important differences between the behavior of male and female chimpanzees: in particular, although males in a group are much more likely to fight than females, “reconciliation occurs after 47 percent of conflicts among adult males, but after only 18 percent of those among adult females, with reconciliation between the sexes falling in between.” This difference forms part of a pattern: “Among males, most cooperation seems of a transactional nature; they help one another on a tit-for-tat basis. Females, in contrast, base their cooperation on kinship and personal preference.”¹

The result is that coalitions of males are both unstable and flexible: they form and dissolve according to the needs of the moment, and breakdown of cooperation at one moment is rarely inimical to its reestablishment a short while later. Female coalitions are more stable and loyal, but although female friends rarely turn on one another, grudges are equally rarely settled, and never just in order to take advantage of some passing foraging opportunity. “Whereas females spring into action mostly to defend their offspring or closest friends, male coalitions are much harder to predict, as males frequently team up against individuals whom they normally prefer as grooming and contact partners.”² Compared to females, it seems, males are much more willing to reconcile with their enemies and much more willing to betray their friends.

These findings of gender differences for captive chimpanzees have been broadly corroborated for chimpanzees in the wild in the groups studied by Jane Goodall and Toshisada Nishida and their respective research teams.³ They also appear to hold for rhesus monkeys, among the most aggressive of all primates.⁴ Such findings are not surprising: as the evolutionary ecologist Bobbi S. Low puts it in her book Why Sex Matters, “Coalitions, like so many other phenomena, can be a reproductive strategy; and if this is true, male and female coalitions will tend to be different.”⁵ Sure enough, the size, purpose, and duration of male coalitions have been shown to be different from the size, purpose, and duration of female coalitions across a range of primate species, as well as in some nonprimates, like dolphins.⁶ In the terminology coined in the 1970s by the sociologist Mark Granovetter, male primates appear to invest in networks with a greater proportion of “weak ties,” while female primates invest in networks with a greater proportion of “strong ties.”⁷

It's natural to wonder whether such findings have relevance for human beings. Are similar gender differences found in human coalitions, and if so, do they matter? Do coalitions of male Homo sapiens have the more flexible and opportunistic character that we see in coalitions of male chimpanzees, and might this explain a tendency for men to navigate their way in modern labor markets more profitably than women, on average, can do? There is a lot of evidence testifying to the existence of differences in the way in which men and women construct networks and coalitions. Unfortunately, there is limited agreement in the literature as to what those differences are and to what extent they are stable and systematic, rather than varying randomly across different professional and social contexts. There's also little agreement as to whether differences between men's and women's networks reflect different preferences, as opposed to differences in the opportunities that men and women have, on average, to meet and spend time with others. If we were to learn, for example, that women don't tend to belong to clubs that admit only CEOs of Fortune 500 companies, that would hardly teach us much about why there are so few women CEOs in the first place. It would be the result of whatever factors explained this discrepancy, rather than an explanation of them.

Most of all, the relevance of these primate studies to human behavior is hard to assess because human beings have a much richer range of interactions than do other primates. True, like primates, we feed, fight, and have sex together (not always in that order). But we can also telephone each other; exchange business cards; employ, blackmail, or torture one another; vote for the same political party; attend the same birthday party; smile or sneer at each other; follow each other on Twitter; denounce one another anonymously to the police; work in the same office; tickle each other; pay money to each other; and live on the same street. Primatologists examining chimpanzee coalitions have a shrewd idea which kinds of behavior to look for (mutual grooming, for instance, is often a good predictor of other kinds of interaction between the individuals concerned). Primatologists examining human coalitions barely know where to start.

Strong and Weak Ties: The Evidence

One promising place to begin is by examining coalitions along two dimensions: first, those that involve real links between individuals, as opposed to simply common membership in a group, and second, those in which the links involve significant investments of time and effort by both parties.⁸ The general finding from primate studies is that females tend to invest more than males in links that involve substantial and repeated investments (“strong ties”), while males tend to invest more than females in links that involve lower levels of investment but to have more of these links (“weak ties”). Strong ties can be characterized as close friendships, weak ties as casual friendships or acquaintanceships.

Among women, such a tendency would be an understandable result of a greater selectivity about relationships, as well as a greater degree of investment in those relationships they undertake. Both of these preferences are consistent with (and predicted by) the logic of sexual selection as far as sexual relationships are concerned. What is novel about this evidence is that they appear to influence behavior over a much broader range of ties, both sexual and nonsexual.

The reason that gender differences in the proportions of strong and weak ties might matter for subsequent economic outcomes was first discovered by Mark Granovetter in a study of the importance of people's personal contacts for their ability to find jobs.⁹ You might expect that the sort of links that would really help you to find a job would be your strong ties, because these are likely to be the people who like you best and are most committed to helping you. But these people are often very similar to you and tend to know the same kinds of things and people that you do: consequently their advice and contacts, however sincerely offered, are frequently redundant. Weak ties are much more helpful when you look for a job, because these acquaintances' lower commitment to helping you is more than compensated for by the fact that they are likely to hear of opportunities that you don't already know about. In fact, telling people about job opportunities is a prime example of the benefits of personal networks: it costs very little to do someone a favor, but it may bring them a valuable benefit.

This phenomenon is an instance of what economists call externalities: each time I invest a small effort in a contact with someone else, I bring myself a potential future benefit, certainly, but I also create a benefit for them because of the possible information I may some day pass their way. This benefit for them is something I don't necessarily take into account when I decide what investments of time and effort to make, and as a result all networks tend to be smaller than they would be if we took such benefits fully into account. In other words, we may not build links as often, or invest as much time and energy in them, as it would be collectively beneficial to do.

If it's true, on average, that women tend to invest less than men in weak ties, partly because of constraints on their ability to do so but also partly because of preferences that are a shared component of our primate heritage, that might account for women's networks being systematically less effective at helping them to find jobs. Not all jobs, of course: you can apply for many jobs through standard procedures (such as by answering an advertisement) without needing to receive any information from people in the know. But for some jobs—and particularly very senior positions in companies— recruitment takes place informally, through word of mouth or head-hunting, and if women have systematically less access to this information than men do, this might account for some of the discrepancy in representation that we noted in chapter 5.

So is there any evidence for this conjecture? Let's begin by looking at evidence that women network differently from men before considering evidence about any effect this may have on professional rewards. A study published by the sociologist Gwen Moore in 1990 looked at the answers given by participants in the US General Social Survey when requested to name up to five people with whom they had discussed “important matters” in the previous six months.¹⁰ Moore noted that women named a higher proportion of kin than men, and men named a higher proportion of coworkers. She then showed that a substantial part of the difference could be accounted for by the different opportunities faced by women, notably in their lower rates of employment. Still, some gender differences remained, notably in that the women cited more kin, and more types of kin, than men in similar employment and other situations.

This is certainly consistent with our hypothesis, though because the focus was on discussion of “important matters,” we can't be sure whether men were investing in additional weak ties: indeed, the average numbers of contacts named by men and women were about the same (almost exactly three in each case). The evidence is also consistent with the hypothesis that men have as many strong ties as women but that their strong ties happen to include more coworkers. One or two other studies provide some corroborating support, but overall there is a frustrating lack of direct evidence for the crucial part of our hypothesis, namely the claim that, whatever their investment in strong ties, men tend on average to have more weak ties of the kind that may be useful to them in their careers.¹¹

We lack relevant data partly because the most detailed evidence on networks and their role in career development often comes from studies of single organizations, which tend therefore to focus on ties to coworkers and pay less attention to how these contacts fit into individuals' overall networks. The question of how representative such organizations are of the overall world of work is also hard to escape. Such studies have, however, suggested some important qualifications to the simple view that men focus more on weak ties while women focus more on strong ones. The first qualification is that both men and women display a preference (other things being equal) for networking with members of their own sex. Although unsurprising in itself, this preference has an interesting consequence in organizations where women are underrepresented, because networking primarily with their own sex tends to shut women out of networks of power and influence. The result is that while those professional ties that are most instrumentally useful to men are also the ones that coincide with their social ties, women tend to interact with one group of colleagues (largely female) for personal support and a different group (largely male) for professional help, advice, and advancement.¹² This tendency, documented for the United States, has also been corroborated in a study of Chinese managers, suggesting that it is unlikely to be due to purely local and cultural factors.¹³

The result is that when women do seek professional advice, they do so from colleagues with whom their other interactions are relatively few, and this may sometimes be a handicap. If so, it suggests a second qualification to the simple view: it is not necessarily someone's weak ties as such that are helpful professionally but the right balance of weak and strong ties, with strong ties in appropriately strategic points. The work of the sociologist Ronald Burt has shown in particular how women in organizations in which they are a minority do not necessarily lack weak ties; rather, they lack the legitimacy to make effective use of them. Such women therefore depend on what he calls “borrowing the network of a strategic partner”;¹⁴ they need one or two strong ties to be able to benefit from the weak ties of others, whereas men's legitimacy in the hierarchy allows them to use weak ties wherever they are found.

The overall picture, therefore, seems to be one in which different preferences of women and men play some role in the formation of personal and professional networks, but so do the different constraints on and opportunities available to them, in ways that are not easy to disentangle. It's useful, then, to look at a different source of evidence for the role of preferences, which comes from measurement of the ways in which men and women communicate. Many surveys have noted that men and women report using the telephone, for example, in different ways, with women tending to hold longer phone conversations on more personal matters and men tending to use the phone for practical purposes, such as making logistical arrangements or undertaking negotiations.¹⁵ But these data rely on self-reporting, which means that they may reflect stereotypical views of expected gender patterns of behavior rather than actual behavior; many fewer studies have examined actual records of telephone use.¹⁶

Together with Guido Friebel of the University of Frankfurt, I have looked at direct billing evidence (from anonymized billing records) of the use of phones in both private and professional contexts. We wanted to see if this evidence could tell us something about how men and women communicate, since it is through communicating that they build up their networks. The evidence is not conclusive, but it is certainly intriguing.¹⁷ First, we compared the lengths of calls made by men and women in a random sample of subscribers of a mobile phone company in Italy and Greece during a two-year period from 2006 to 2008. Controlling for other factors such as age and income, calls made by women lasted 16 percent longer than calls made by men; women also made fewer calls. The same was true within each age category: women in their twenties made longer calls than men in their twenties, and so on for women and men in other age groups.

To determine whether such differences might reflect different professional and other opportunities open to women, we examined the records of calls directed (at random) to male and female employees of a call center of a large German consumer services company. The male and female employees work under identical conditions. Calls allocated to women lasted 15 percent longer on average than those allocated to men, controlling for other factors. (Again, the differences between men and women are large and statistically highly significant within each age category as well as for the sample as a whole.) We tested whether this difference might reflect less enthusiasm among women for getting on with the job by looking at operations involving sales. It turns out that women make slightly more sales per shift than men, so they appear to be using systematically different communications strategies and are no less effective as employees.

The one thing that prevents us from concluding that the longer calls of women reflect their different preferences is that they could reflect the preferences of the incoming callers (both male and female) for speaking to women. There's no real way to disentangle the two motivations because, after all, every conversation is negotiated between two people. No one, perhaps not even the parties themselves, can be sure to what extent the overall conversation resembles the conversation they would most have preferred. But though this impossibility may be what gives guaranteed employment to psychoanalysts, it may not really matter for our purposes, because the same is true of all conversations in life. So it seems reasonable to take these data as an indication that communication strategies involving women lead to slightly fewer, slightly longer conversations, and that difference may affect the kinds of network links women make as a result. But it would be good to have more evidence, and no doubt research on social networks in the coming years will make this a priority.¹⁸

Networking and Professional Success

If the evidence for women's networks being different from those of men is suggestive but not conclusive, can we look at the problem from the other end? Can we look for evidence that women's networks, whatever their structure, are less effective at delivering professional benefits? Research on this question has been inconclusive up to now, largely because of the difficulty of measuring networks and of getting data on both networks and career performance for a sufficiently large and representative group of people.¹⁹ In work with Marie Lalanne of the Toulouse School of Economics, I've been analyzing this very question, using information on around 16,000 individuals who are board members or senior executives of American and European companies (only 9 percent of whom are women). This is a large data set by the standards of most research on this topic, but its size comes at a cost: we don't have information about the actual networking activities of all those people, or even about the people who belong to each other's active social networks. What we do have is information about who has had the opportunity to network with whom, because they have worked for the same employer in the past.²⁰ We don't know whether they chose to make anything of this opportunity. But if it is indeed true that women's networks are somehow less effective at bringing them professional benefits than those of men, this fact should be reflected in our data, because women should be making systematically less of their opportunities than men. In particular, we expect to see two things: first, we can calculate, for each person in our data set, the number of currently powerful people with whom that individual has crossed paths in the past (we can call this the IMS, the Index of Movers and Shakers). We expect that after allowing for the influence of such factors as age and education, those individuals with a larger IMS should have higher salaries if the contacts they have been able to make have yielded professional benefits. Second, if our hypothesis about the greater effectiveness of men's networks is correct, we expect that the effect of IMS on salaries should be larger for men than for women.²¹

Our data do indeed show a very clear effect of IMS on people's salaries. It's not simple to measure, because there's a possibility that both the numbers of opportunities a person has had in their past career and their current salary are joint effects of a third cause, namely their talent. If this were true, simply correlating salaries with the IMS might overstate the effect of one on the other. Adjusting our estimates to take account of this possibility is not straightforward (there are various different ways of doing it, none of which is ideal),²² but the end result is that we find that people with an IMS value of 250 compared to the average of around 150 would have somewhere between a 2 percent and 4 percent higher salary than the average in 2008, depending on which particular estimate you favor. When we look at the other forms of remuneration, such as stock options, the returns to network size are around twice as large. This is a useful return to being well connected, though not a massive one. Being well connected won't make you rich: it just might make you somewhat better off than you are already. And connections seem to help in two ways. First, being well connected helps you learn about job opportunities out there that would suit you better than your current job. Second, it acts to make you a bit more attractive to your current employer and able to negotiate a little more ambitiously on your own behalf.²³

As soon as we discovered this effect, we wanted to know whether it worked for women as effectively as it worked for men. And the surprise was that there was no apparent difference at all between men and women on this measure. But when we looked more carefully at our data, we noticed something striking. Board members of companies fall into two distinct categories: executives and nonexecutives. Executives run the company and typically work full-time. Nonexecutives show up for board meetings and sometimes for meetings of various subcommittees of the board; they almost always work part-time, although they may and often do sit on the boards of several companies at the same time. Nonexecutives usually earn much less than executives (in our sample, nonexecutive pay was less than 30 percent of executive pay). And it may not surprise you to learn that only 32 percent of women in our sample are executives, whereas 49 percent of men hold executive jobs. Only 6 percent of our executive sample are women. We have no idea whether this is because prejudice is keeping women out of executive positions or because women are more interested in applying for positions with more flexible working conditions. (Most of the women in our sample are over forty-five, so they are less likely to be constrained by young children than more junior women in the organization.) But it's important to take this difference into account when comparing salaries.

Comparing the salaries of nonexecutive women with those of nonexecutive men does not show a large difference, either in average salaries or in the effect of IMS on salaries. Network contacts are very important here: by our best estimate, having an IMS of 250 instead of 150 raises your salary by 8 percent. And the nonexecutive women clearly use their contacts as effectively as the men. This doesn't mean there's no possibility of discrimination in access to such positions: women still make up only 12 percent of our sample. But the salaries of those who hold nonexecutive positions seem (almost) to match those of similarly qualified men. We therefore can't rule out the possibility, discussed in chapter 6, that women's lower representation is due in part to different preferences for such occupations.

With executive board members, it's a different story. In our sample, women executives earn on average 30 percent less than similarly qualified men. When we include data on other forms of remuneration (such as stock options), the difference looks even bigger.²⁴ But when we look at the effect of contacts on remuneration, we see something striking. Male executives get a benefit from their contacts: those with 250 contacts instead of the average 150 have 3-4 percent higher salaries and around 10 percent higher levels of stock options and other such indirect forms of remuneration. But female executives don't. It's as though all those contacts earlier in their careers were bringing them no benefit at all. The exceptions are a minority of women who happen to have a lot of currently powerful women among their contacts: those contacts turn out to be fairly valuable to them after all. And after you adjust for the generally different productivity of men's and women's contacts in influencing their salaries, the effect of gender as such becomes small and no longer statistically significant. It looks as though our hypothesis is confirmed for executives and rejected for nonexecutives.

To put it another way, male networks seem to be giving men an edge compared to women when it comes to running companies, but when it comes to just sitting on boards, the two sets of networks are more or less on a par. Once again, women are most disadvantaged when it comes to positions of real power. And this suggests, by the way, that even if policies designed to increase women's representation in the boardroom result in substantial increases in the proportions of women among nonexecutive directors, they may make little difference to the proportions of women among executives, particularly among those who wield real power, such as CEOs.²⁵

The fact that these results hold only for executives means there can't be some universal difference in behavior between men and women, which operates regardless of context. Whether women's networks bring them professional benefits depends not just on how women behave but also on how men behave in their turn. It's worth remembering, too, that the people we're looking at in this study are a very privileged minority even of business executives, let alone the population as a whole. So we don't know to what extent we can draw general conclusions about the behavior of men and women from the results of studies such as these. But in some ways the results we've found are stronger than we might have expected, because these are people who are very well connected indeed and probably wouldn't have become board members in the first place unless they were pretty good at networking. To return to one of the themes of chapter 5, it's a bit like looking at the effect of height on basketball scores. If you want to test whether being tall makes you a better scorer at basketball, you shouldn't look just at professional basketball players, because they're all very tall already, and indeed have been selected precisely because of how tall they are. It turns out that among such a tall group, variations in basketball prowess are related mainly to other things and hardly at all to the remaining variation in height.²⁶

Similarly, we are dealing here with a group of formidably successful women and men, and they are all probably formidable networkers. So we might have expected that looking for variations in networking ability to explain the remaining differences in professional success would be as misguided as looking for variations in height among professional basketball players to explain the remaining differences in basketball scores. But we have found such an effect all the same. All of this suggests that there's a lot of work left to do before we can confidently conclude that men's networks are responsible for the high price that women pay for the different career choices they make across many different occupations in modern society. But given the evidence that men's networks play a strikingly different role from women's networks when it comes to allocating positions of power in leading companies in both America and Europe, networks remain a leading suspect in the wider world as well.

Network Choices: Prudence or Preference?

It might seem natural to ask whether men and women choose their social and professional networks with an eye to what will be useful for them, or whether they just indulge their preferences for the kinds of people with whom they want to interact, as they might indulge their preferences for the kind of furniture that decorates their homes or offices.²⁷ But this may be a misleading question to ask, and not just because the answer, if there is one, is almost certainly that people have a mixture of both motivations. The question may be misleading because we have good reasons to think that our preferences for the kind of people with whom we interact have been formed by millions of years of evolution to reflect the kinds of people with whom our ancestors found it most prudent to interact. As we saw in chapter 3, our preferences were a kind of shorthand for what prudence would have dictated, a shorthand suitable for a world of imperfect cognitive abilities and imperfect capacity for commitment. We are attracted to people who smile, so we prefer to be around them; but we also trust them more, and the evidence suggests we're right (on average) to do so.

It seems likely, as Low suggests, that male and female preferences for coalitions developed because these coalitions served important reproductive purposes in prehistory.²⁸ In particular, individuals chose their coalition partners because those were the individuals they could most trust to bring them the reproductive benefits they needed in the conditions of prehistoric hunting and gathering. If men preferred to interact with men, other things being equal, and women with women, such preferences presumably made sense in the light of the strong division of labor we discussed in chapter 4. But this division of labor is no more. Are the preferences that went with it now entirely out of date? Or might it be true that men still prefer to interact with other men at least partly because they understand them better and can judge better how much they can trust each other? And that women prefer to interact with other women for the same reasons? There is scant and conflicting evidence on whether men are better judges than women of male character and whether women are better judges than men of female character, at least in the workplace.²⁹ Likewise, there is conflicting evidence from laboratory trust-game experiments about whether men and women behave in a more trustworthy fashion when paired with another player of their own gender.³⁰

One possibility is that both men and women interpret signals about the behavior of others in the light of rather coarse-grained generalizations about the groups to which they belong, which in the cases of men judging women and women judging men are more likely to involve judgments about gender (possibly stereotypical ones) than when either group is judging individuals of their own sex. Male employers assessing how seriously other men are committed to the company may well use a wider and more sophisticated set of criteria than those same male employers assessing the commitment of women. We return to the finding of Marianne Bertrand, Claudia Goldin, and Lawrence Katz in chapter 6: women pay a high price for those career choices that they are more likely than men to make. Could it be because they, and the men who employ them, are stuck in a signaling trap in which commitment and talent are simply harder for women to signal than for men?

I return to this question in chapter 9, but before doing that we need to broaden the picture. As Charles Darwin was only too aware, the signals and strategies that men and women use to play the game of sexual selection are just a sample of the signals and strategies that all individuals use to navigate the groups and coalitions that determine the fitness of members of a group-living primate species. As if the anxieties that beset our determination to be accepted by our sexual partners were not burden enough, we all have to worry about how our talents will endear us to our potential collaborators in the workplace independently of concerns about gender. You might have thought that in the modern information economy, these kinds of anxiety might be likely to diminish, as matching technologies in everything from online dating to job search make the process of finding professional and personal partners more efficient. Chapter 8 investigates whether this optimistic vision is too good to be true.