FIGURE 7.1 Schelling’s participation curve
HOW IT ALL KICKS OFF
In May 2011 protests and demonstrations erupted across austerity-stricken Spain, starting in Madrid’s Puerta del Sol and spreading to consume public squares in over fifty Spanish cities. Protestors demanded radical political changes and a response to the social and economic problems facing Spain, particularly its record-breaking unemployment rate of 27 percent (over half of young people in many regions). Inspired by the revolutions and uprisings of the Arab Spring, the movement was coordinated in the early days almost exclusively via social media, particularly Twitter and Facebook, with only negligible effects of news media until one-fifth of users had already joined the online exchange of protest information.1 Around one-fifth of Spanish people have participated at some point. Called alternatively 15-M (Movimiento 15-M, after the date the first demonstration took place) and Los Indignados (‘the discontented’, to express the widespread feelings of outrage at conditions in Spain), the movement was born out of a myriad of groups, local organizations, and loose associations with highly nebulous and disparate aims, not the most auspicious circumstances for coordination or the growth of a social movement. Yet 15-M survived and became a force for change, even contributing to the Occupy movement and now considered to be ‘part of modern Spain’.2 As we will discuss in the final chapter, 15-M is one of the few movements of 2011—when demonstrations started kicking off everywhere, as Mason put it—that appears to be developing into a sustainable political force with the formation of the political party Podemos (meaning ‘We Can’) in 2014.3
How are mobilizations like 15-M launched? Chapter 4 showed that evidence of a critical mass of participants plays an important role in encouraging people to join in, and Chapter 5 showed how people respond strategically to social information; not participating when it seems as if a good will not be provided or when it is clear that it will be provided without their contribution. But how can a critical mass form if everyone waits for this point? Implicit in the critical mass argument is that some people are willing to join when there are few signals of viability, in order to lay the foundations for critical mass to form by encouraging others to join. If this assumption holds, then the social information effects—which are crucial in indicating both the viability of a mobilization and the most efficient way of influencing it—observed in the last two chapters will act differently on different people. But can we test this assertion? And if we can prove it to be so, what could explain this difference? What sorts of people are willing to start?
To tackle these questions, in this chapter we turn to threshold models of collective action,4 building on the work of the economist Thomas Schelling and the sociologist Mark Granovetter from the 1970s. Both these scholars claim that people vary in their threshold for joining a mobilization (that is the proportion of other people that will propel them to mobilize); some people take the lead by participating when few others have done so, while others require high levels of participation before they will take part. They argue that the distribution of thresholds in a population will be crucial in determining whether a mobilization gains critical mass. If there are enough people with low thresholds to start a mobilization, then evidence of their participation will give a signal of viability to those with slightly higher thresholds, and so on in a series of chain reactions until there is enough of a sense of viability to suggest to those with thresholds somewhere in the middle that it is worth joining. At this point, critical mass will be attained and the majority of potential participants will join. If this series of chain reactions does not occur, then a mobilization will fail. In this way, the threshold model is closely related to the tipping points we identified in Chapter 4, when we looked at the arguments of Marwell and Oliver.5
In this chapter, we investigate the moments when different people join a mobilization (rather than what they contribute), looking for evidence of different thresholds and identifying the types of people associated with them. We use data from our laboratory-based experiment to test Schelling’s assertion that a population will vary in thresholds for joining a mobilization;6 some joining when the number of participants is very low, some when almost everyone is participating, and most joining somewhere in the middle.7 Using measures of personality as we did in the last chapter, we develop the idea that personality may present the clue to understanding an individual’s threshold. If so, then the pattern of starters and followers who progressively join a mobilization may present a way to understand how movements like 15-M can take off and build momentum. Previous research has shown that rather than this kind of collective action relying on a small subset of special individuals with disproportionate influence (such as the mass media), influence derives from a critical mass of ‘common users’ who in aggregate make chain reactions converge.8 González-Bailón et al. usefully analysed chain reactions in the Twitter network relating to 15-M, finding that the protest managed to mobilize so many people in such a short time span because of reinforcing interactions between four categories of participants in the network: influentials (with high visibility), hidden influentials (with low visibility, but a central role), broadcasters, and grassroots users.9 But these categories of protesters are based on roles in the network and in the diffusion of information, rather than any intrinsic characteristics and cannot, therefore, provide insight into differential thresholds.
THRESHOLDS FOR JOINING
The clue to identifying differential willingness to join at the start of a mobilization might come from the threshold model of collective behaviour. Schelling in the book Micromotives and Macrobehaviour develops a model of mobilization, although not for a political context, pointing to examples of where a person’s behaviour will vary depending on information about how many other people are participating,10 such as how many people attend an optional seminar, how many applaud and clap loudly, and how many people leave a failing school.11 He argues that the number representing critical mass varies by context, which can mean a proportion of potential participants (important when considering adopting a certain fashion, for example) or an actual number (which might matter to attendees at a seminar). He assumes that people have different individual thresholds for joining a mobilization and argues that where such thresholds are normally distributed in a population, there will be an S-shaped joining curve for any mobilization, illustrated in its simplest form in Figure 7.1, where the number of people participating in an event is represented on the vertical axis, while the number of people expected to participate is shown on the horizontal axis. There is a relationship between the numbers that people think are going to turn out and how many eventually do so, based on the effect that information about expected numbers has on any one individual’s decision to participate. The eye can follow the line along the axes, so at low levels of expectation, this creates the self-fulfilling prophesy that even lower numbers turn up. As the numbers expected increase, so do the numbers who actually do show up, but at increasing proportion, so the numbers expected and actual numbers equalise. As the numbers get higher, so the numbers coming exceed those expected, which creates the critical mass for mobilization, in Marwell and Oliver’s terms.12 Schelling produces a variety of predictions from this basic form, showing how sometimes the mobilization curve will cross the diagonal line because the number of people expected to participate meets most people’s threshold and the mobilization succeeds, whereas at other times the curve does not rise above the line and the mobilization fails.13
We do not test all aspects of Schelling’s model here. Indeed as we observed in Chapter 2 for the mobilizations relating to US demonstrations against racist policing and noted in Chapter 3 with respect to petition signing, many mobilizations related to political issues show surprisingly little adherence to Schelling’s S-shaped curve. Rather, we focus on his claim regarding thresholds for joining a mobilization. Likewise, Granovetter identified the concept of a threshold as key to the viability of collective behaviour, arguing that the distribution of thresholds within a population was a vital determinant of outcomes.14 The threshold model suggests that individual behavioural disposition does not change before, during, or after collective decisions are made; these are contingent dispositions that act on the situation, even if actual behaviour may change. But there is no attempt to identify the origins of these dispositions; neither Schelling nor Granovetter ‘consider how individuals happen to have the preferences that they do’.15 In contrast, the aim of this chapter is to investigate empirically the distribution of thresholds in a population, focusing on the characteristics of the group of starters who are willing to join a collective action early.
In mobilizations initiated via social media, Schelling’s ‘number expected to attend’ will be informed by the social information that most platforms provide, that is, an indication of the number of participants involved in a petition, a mass email campaign, a Facebook group, or a charitable donation page. Although Schelling makes the questionable assumption that people are able to estimate the ‘number expected’ figure in an offline setting, it may be that for the first time we can now test his claims.
WHO JOINS WHEN?
As discussed in the previous chapter, personality is a fundamental part of the human condition, and different types of people perceive and respond to the same stimuli in fundamentally different ways. As in the last chapter, we follow recent work that has used personality as a predictor of political behaviour and beliefs, in particular the Big Five traits,16 using these variables as part of our research design to investigate whether people with particular personalities show a different propensity to join an action early.
When looking for clues as to which personality type is most likely to start, we look for traits that have been associated with leadership and leaders, given the connotations that these concepts have of showing initiative and leading others at the front. Our assumption is that propensity to join early (or start) could be related to some (although not all) of the personality traits associated with leadership. Previous research provides no clear picture of how personality traits relate to leadership,17 and the results of these investigations ‘have been inconsistent and often disappointing’.18 For this reason, we select two alternative typologies of individual-level difference that have been associated with willingness to lead, rather than selecting a single one.
First, as in the previous chapter, we utilize the Big Five personality traits. In that chapter, we followed previous political science research in identifying extraversion and agreeableness as the most significant predictors of contributory behaviour, so we hypothesize here that they will be the most useful in understanding a person’s willingness or reluctance to go early in mobilizations where there are few signals of viability.
Psychological research looks at two broad categories of leadership: leadership emergence and leadership effectiveness,19 the latter being less relevant to the current study. Although much of that work focuses on leaders in the sense of other people’s views of the potential leader, rather than the ‘starter’ criteria that we are interested in here, and some researchers have dismissed trait theory as obsolete in this context,20 two extensive and detailed reviews of leadership research found evidence to suggest that the Big Five typology ‘is a fruitful basis for examining the dispositional predictors of leadership’.21 Of the Big Five characteristics, extraversion in particular has been positively associated to self and peer ratings of leadership.22 In their meta-analysis of leadership and personality research, Judge et al. found that ‘extraversion emerged as the most consistent correlate of leadership’, particularly with respect to leader emergence,23 although there were some findings for conscientiousness and openness to experience. For these reasons, we test all the Big Five personality traits while hypothesizing that extraversion will be the most important and the most likely to lead to a positive association with willingness to start.
We introduce here locus of control as another personality trait. In the field of organizational and applied psychology, internal locus of control (the extent to which people believe that they have control over their own fate) has been identified as an important personality trait with respect to all kinds of social roles, including leadership. Developed by Rotter into a measurable scale (later shortened by Carpenter and employed in Carpenter and Seki),24 locus of control identifies ‘internals’, who are those people who believe that they are masters of their fate and who perceive a strong link between their actions and consequences, and ‘externals’, who do not believe themselves to have direct control of their fate and who perceive themselves in a passive role with regard to the external environment. In work settings, internal locus of control has been positively associated with favourable work outcomes and greater job motivation,25 high job satisfaction and performance,26 and more effective leadership.27 In business studies and management research, people with internal locus of control have been identified as more likely to be leaders and to have superior leadership performance,28 although Judge et al. found locus of control to be a weaker predictor of successful leaders than other personality traits reviewed.29 In economics, Boone et al. found that ‘internals’ in social dilemma situations are more likely than ‘externals’ to try to influence the behaviour of others to achieve their goals, while externals behave less strategically.30 With particular relevance for this study, locus of control has been associated with the tendency for people to exert active control over the environment,31 to participate in social action,32 and to engage in political activism.33
We may hypothesize therefore that extraversion and internal locus of control will be associated with low thresholds for joining a mobilization at an early stage. We may further hypothesize that the likelihood of a mobilization growing in its very early stages will increase as the number of individuals with low joining thresholds in the population increases.
LOOKING FOR STARTERS AND FOLLOWERS
We use data derived from the public goods laboratory experiment undertaken by the authors,34 introduced in Chapter 5. To recap, the subjects were asked to contribute tokens to a collective good under control, visibility, and social information treatments (the instructions to the participants and post-experiment questionnaires are shown in the Appendix). In Chapter 6 we looked at the amount that individual subjects contributed under the different treatments, whereas here we examine the relationship between personality and the order in which subjects participated in each round. We use just those rounds where participants were aware of how many others in their group had contributed (the social information treatment) and whether they were leading or following at the precise moment that they decided to participate, thereby using the number of other participants as a surrogate for ‘number expected’ in Schelling’s model. We look at the rank (first, second, third, and so on) at which individuals contribute, seeking consistent patterns in the rank at which they join rounds, to identify thresholds and then examining the relationship between personality and threshold, controlling for the importance that people ascribe to the issue under consideration. We also examine the relationship between the propensity of a round to get funded and the distribution of personality characteristics across the people in the group playing the round.
First, we looked at the propensity of individual subjects to contribute tokens in the early stages, when there was little or no evidence that others were participating, to see if we could identify subjects who consistently started or followed, providing evidence of stable personal thresholds. We examined the minimum, median, and mean rank at which subjects participated in a round as their threshold for participation (Figure 7.2). First, with respect to minimum rank, if an individual subject has a minimum rank of 1, it indicates that he or she is willing to start a mobilization with no other participants. Alternatively, if the participant’s minimum rank across the twenty-eight rounds is 5, then a higher threshold is suggested. Figure 7.2a shows the distribution of minimum rank across our 185 subjects, showing that just over half of them were willing to start at some point. This result was quite surprising, suggesting a far from normal distribution of people’s willingness to start a mobilization. To identify threshold, however, we need to determine consistent starters, rather than just people who are willing to start at some point. Therefore, second, we looked at median rank (Figure 7.2b). Here the figures are more evenly distributed, with a few subjects starting often (and therefore having a median rank of 1 or 2), a few almost always joining only at the end of the mobilization, and most being somewhere in between. The most common median rank is 3, suggesting people who are most likely to contribute when over 25 percent (two out of eight, the most common group size) have already done so. Third, Figure 7.2c shows the results for mean rank, again illustrating a distribution that appears normal and that fails all tests for non-normality.
Finally, we construct a variable to count the number of times a subject starts a round (propensity to start) normalized by the number of rounds the subject played. This variable exhibits an extremely skewed distribution (Figure 7.2d), with just under half of the participants never starting a round, and for whom the ‘propensity to start’ variable is therefore zero. Of those subjects who do start a round at some point (i.e., who have at least one round with rank one and therefore a minimum rank of 1), the distribution is still left-skewed, with most individuals starting only one or two rounds.35
FIGURE 7.2 Distribution of subjects’ (a) minimum, (b) median, and (c) mean ranks and distribution of (d) the number of times subjects started a round
Each subject played between seven and fourteen rounds in the conditions investigated here. We calculated the standard deviation of each subject’s ranks across these rounds as a measure of their consistency. Plotting this standard deviation score against the median rank of each participant (Figure 7.3) shows that nearly all subjects with a low or high average rank consistently contributed early or late. In contrast, while some subjects with a median rank towards the middle also had a low standard deviation score, other subjects were much less consistent and had higher standard deviation scores. Standard deviations for subjects with a mean rank between 4 and 6 ranged from 0.488 to 3.827, with a mean deviation of 2.276 across subjects. These distributions show that the subjects have heterogeneous propensities to start or join a collective action and suggest that a few people have consistently low thresholds for joining, and some have consistently high thresholds. There is greater variability in subjects going near the middle, but early movers and late movers act more consistently.36
FIGURE 7.3 Median rank and standard deviation of rank per subject
Note: The best fit line is a locally weighted scatterplot smoothing (LOWESS) line with a 95 percent confidence interval shaded around the line.
WHO STARTS, WHO FOLLOWS?
The analysis above has identified four possible ways to identify the threshold for individual subjects: minimum rank, median rank, mean rank, and propensity to start. All four measures exhibit variance across individual subjects, supporting our aim to find some proxy for personal thresholds, by which we mean consistent behaviour in terms of participants joining a petition early, late, or somewhere in the middle. Before we could examine the relationship between these threshold measures and personality, we first examined the personality variables in our dataset for collinearity (Table 7.1). These results show that in general the variables are not highly correlated; even the correlations of .34 and .35 between Rotter score (the variable used to measure internal locus of control) and three of the Big Five variables (extravert, open, and emotionally stable) are not high enough to be problematic. Nonetheless, we ran separate regressions for the Big Five and Rotter score when testing our hypotheses, shown in separate panels of Table 7.2.
TABLE 7.1 Correlation matrix for personality measures (Big 5 & locus of control)
TABLE 7.2 Tobit regressions showing relationships between threshold measures and personality traits
*p < .05. **p < .01. ***p < .001.
Note: Standard errors are in parentheses
We ran Tobit regressions to test the association between the threshold measures and the various personality traits discussed above, in order to examine our hypotheses that people scoring highly on extraversion and internal locus of control would be more likely to start or go early. These results are shown in Table 7.2.37 We tested for possible associations between demographic variables (income, age and ethnicity) and propensity to start, but none were found to be significant. We also found that social value orientation (as discussed in Chapter 6) was unrelated to people’s propensity to start, in spite of the strong relationship between this variable and contribution behaviour identified by economists and ourselves in the previous chapter.38 In all regressions we controlled for the relative importance of the scenario at hand (as asked in the post-experiment questionnaire) for our subjects, as clearly the importance ascribed to an issue will be an important factor in the motivation of an individual to participate, something we have established in earlier work and in Chapter 4.39 The post-experiment questionnaire included questions on whether the subjects agreed with the issue and how important they considered it to be. Both were significant, but as they were correlated we controlled for importance as it had the most predictive strength.
First, we found locus of control, the personality variable most often associated with leadership in the psychology and management literature, to be significantly associated with going early in mobilizations. The locus of control variable was a predictor of median rank, mean rank, minimum rank, and propensity to start across all issue subjects. So our hypothesis that internal locus of control would be associated with willingness to start is confirmed, although, as found in some earlier work,40 locus of control was found to have a weaker effect than some of the other personality traits we investigated.
We found significant results for two of the Big Five personality traits. Across our measures of threshold, extraversion was consistently associated with joining early, and agreeableness was consistently associated with joining late. Table 7.2b shows agreeable people as being strongly associated with higher median ranks and higher mean ranks, and with a lower propensity to start. Extraversion is also an important personality variable. Extraverts have a significantly lower median rank (p < .01) and mean rank (p < .01) and are significantly more likely to have a high number of starts (p < .01). None of the other Big Five personality traits have a significant effect on these threshold variables. These results confirm our prediction that extraversion would be a predictor of low threshold to start or join early, with the additional finding for agreeableness, which has a consistently higher coefficient than those for extraversion.
If successful mobilizations rely on a number of people with low thresholds in order to start them and if these people are those with extravert personalities (or a high internal locus of control), then it should be that rounds without people with one or other of these personality types will consistently fail. To test this hypothesis we collapsed the data at the round level, looking at the maximum, minimum, and mean scores for external/internal locus of control and extraversion/intraversion of the people participating in each round and the relationship between these variables and whether a round was funded or not.
We ran regressions for all personality variables separately while controlling for the mean importance attributed by all participants on the team to the scenario at hand and the amount of the first contribution. Basically, the higher the importance ascribed to the issue across the group of participants, the more likely it is to be supported. The amount of the first contribution in a round is also significant, reaffirming results of previous experimental research in economics investigating conditional cooperation in charitable giving, which showed that the larger the first contribution, the larger the total contributions are likely to be.41 We control for these two variables in the regressions that follow.
Among all the personality variables tested, only extraversion was significant in explaining the likelihood of a round being funded. Two measures of the extraversion of the group, the mean and minimum, were significant, as shown in Figure 7.4. If we picture Schelling’s mobilization curve (Figure 7.1), it seems that extraversion correlates with willingness to join the movement (Schelling’s x-axis). Those high in extraversion are so motivated that they are willing to start, while those with more moderate levels of extraversion have a medium threshold to then join. Those with low levels of extraversion have a very low willingness to join, and too high a proportion of these individuals within a population hampers the possible success of a movement. A high minimum value for extraversion in any given round suggests that there are lower numbers of people with low extraversion (high thresholds), which is associated with a higher likelihood of a round being funded. Mean extraversion was also significant, showing again that generally high numbers of extraverts in the group were associated with a higher chance of a round being funded. We did not, however, find similar results for internal locus of control. Similarly, the minimum, maximum, and mean levels of agreeableness (and all the other personality variables) were not associated with the likelihood of a round’s being funded.
FIGURE 7.4 Logistic regressions to show the effect of personality on the round being funded
Note: N = 217.
STARTERS AND FOLLOWERS
We have identified heterogeneity in our experimental subjects in their willingness to join a collective action at an early stage, when there is little or no evidence that other people will join or that the common goal is likely to be attained. Some people were willing to start a mobilization more than once, some joined mobilizations only during the later stages, and most varied their behaviour across a middle range. We have also identified that for a proportion of subjects in our experiment, the rank (or willingness) at which they join is consistent, and we found most evidence of this consistency at the early and late stages of a mobilization. That is, we identified some people with consistently low thresholds, and some with consistently high thresholds for joining.
We have identified the personality characteristics associated with low or high thresholds, although not for all the personality variables that we investigated in our analysis. Internal locus of control is positively associated with a willingness to start, supporting the findings of research on leadership in psychology and management studies.42
Likewise, we find that extraversion is consistently associated with willingness to start. This joins the growing body of evidence from political science that points to a relationship between extraversion and political participation.43 We might have expected that in online contexts where participants do not interact socially, such as the one we simulated here, extraversion would be less predictive (as Mondak et al. and Gerber et al. found for various ‘nonsocial’ types of participation, and we found in the previous chapter),44 so the fact that it remained significant throughout our analysis in this chapter renders this finding more interesting and suggests that this personality trait acts endogenously to predispose individuals to start or initiate action even when there is no indication that others will do so. The positive association we found between the concentration of extraversion in a subject group, and the likelihood of a round getting funded, suggesting that a mobilization with higher numbers of people with high extraversion has a higher chance of success, reinforces this finding.
We have made the additional finding that agreeableness is positively associated with a tendency to go late, suggesting that agreeable people wait until it is clear that the mobilization will be funded before joining themselves. This joins the mixed results of previous research for this personality trait; agreeable people seem drawn to some types of participation and not to other, more conflictual activities. Starting an action where there are few other participants could fall into this category, confirming the importance of context and ‘personality-relevant characteristics of each participatory act’, as we found in the last chapter.45 None of the other Big Five personality variables investigated is significantly associated with an individual’s joining threshold, a result that adds to the mixed results for these traits in most previous political science research.46
These findings also support previous research on leadership showing that the Big Five typology is a fruitful basis for examining predictors of leadership and that extraversion is the most consistent correlate of leadership.47 This previous research focused on leadership in organizations, whereas our findings for extraversion in the different context of collective action in a non-organizational setting are stronger and less equivocal. It may be that the strength of our results derives from extraversion being best suited to the starting element of leadership, and less so to the other activities required for leading in offline contexts, which require other dimensions of personality.
MOBILIZATION WITHOUT LEADERS, REVOLUTIONS WITHOUT ORGANIZATIONS
Analysis of large-scale transactional data from electronic petitions presented in Chapter 3 showed that the early days of a mobilization are crucial; success is reliant on a significant number of people being willing to act at this early point, when there is no social information to indicate viability. On social media, where people can be rapidly notified of new mobilizations through large-scale social networks, and social information is readily available, mobilizations can quickly reach critical mass; likewise, a movement may fail almost immediately as news of the mobilization slips down the news feed on social media platforms, and the starting moment is lost. By identifying heterogeneous thresholds for our subjects, consistent at least at the early and late stages of mobilizations, we have provided some evidence for this mechanism in action.
In this environment, leadership is the aggregate of many low-cost actions undertaken by those willing to start, rather than the high-cost actions and rare attributes usually associated with leadership. Of course, the group of starters will usually include at least one leader more in the traditional mould who has taken a higher cost action, for example, the person who sets up a petition and circulates it to close associates in her immediate social networks. But the number of starters needed to get the mobilization off the ground will be beyond that which could be obtained with strong ties to the initiator alone, but will be attained with weak ties, such as the friend of a friend on a social networking site, or the retweet of the retweet of a tweet. To obtain the five hundred signatures used as a measure of success for the first UK petition platform analysed in Chapter 3, an individual initiating a petition would probably have to go beyond his immediate social network; in 2011, the median number of friends on Facebook was 100,48 and the median number of followers on Twitter was 85.49 In the United States, where petitions must have 150 signatures before they can even be public on the petitions platform, a petition initiator could struggle to gain visibility for her petition using her own immediate social network alone. As requests to sign arrive through weaker and weaker ties, the raw information of how many signatures have been obtained will be a relatively important piece of social information, rather than the network attachment to the initiator. Many petitions on similar issues fail while one might succeed; it is not the existence of the leader who brings the success, but the existence of a sufficient number of starters with low thresholds, and the readily available and disseminated social information about other participants that draws in the followers with higher thresholds. By providing this social information, social media platforms circumvent the need for other activities traditionally performed by leaders.
If extraversion is a good predictor of low threshold, then the likelihood of a mobilization obtaining some measure of success will depend in part on the distribution of extraversion in the pool of people who care about the issue at hand and who are aware of the mobilization. A possible corollary of this finding is that mobilizations initiated by people with high levels of extraversion are more likely to succeed. Given that we might hypothesize that extravert people tend to select extravert friends;50 given evidence to suggest that extravert people have more friends on social media;51 and given that an individual is most likely to kick off a mobilization via her own online social networks, then an extravert’s initiative might be immediately disseminated to a pool of people with greater than normal levels of extraversion and, if our findings are correct, low thresholds for joining.
The earlier work by Schelling and Granovetter discussed in the first section assumed that different individuals have different thresholds for joining a mobilization and based their models on this assumption. They did not investigate why individuals have the thresholds that they do, or what types of people have which types of threshold. In contrast, we have looked at the characteristics of people with low thresholds and identified a relationship between extraversion (and, to a less clear extent, internal locus of control) and low thresholds. From the evidence reported here, we cannot attempt to make any claims as to what the distribution of thresholds might be in any given context, as our findings are based on experimental data that did not purport to be a representative sample of any population. However, the findings suggest that future research aimed at understanding the distribution of thresholds might include secondary analysis of research into the distribution of extraversion or locus of control in different populations.
These findings also have a wider significance. The argument presented here combined with the evidence of previous chapters suggests that contemporary political mobilizations can become viable without leading individuals or organizations to undertake coordination costs, and proceed to critical mass and even achieving the policy or political change at which they are aimed. Our conclusions provide one plausible explanation for how the Spanish 15-M movement could mobilize without any national organization or clear leadership. It could be argued that this is the case with some of the uprisings of the Arab Spring, where dictators and their administrative apparatus were toppled by huge political mobilizations of people, but without institutions, political parties, or embryonic leaders in the wings to undertake the process of rebuilding the state.
The existence of movements like 15-M makes it more likely that discontent or other strong political emotions crystallise into concrete expression. The large resources of followers are potential voters, or can be mobilized to protest again, bringing further turbulence to political life. But Internet-based mobilization of this kind is no guarantee of sustained political change, as evidenced by some of the disappointments of the nascent political movements that emerged from 2011. In Egypt, for example, in spite of the huge waves of popular protest that toppled the Mubarak regime, the Muslim Brotherhood, as the only organized force on the political scene but with limited commitment to democracy, gained power in the first elections, leading to Mohamed Morsi becoming the first democratically elected head of Egypt in 2012. Even after a further year, when renewal of popular protests and demonstrations led to the removal of Morsi, it was the Egyptian military—again a preexisting organization rather than a new organizational force—that returned to power with the election of President Abdel Fattah el-Sisi, previously the minister of defence.
In the final chapter, we pull together the arguments and evidence in this book, to see how new forms of collective action impact on the practice of politics more generally. We combine the experimental and big data evidence we have gathered about collective attention, critical mass, tipping points, thresholds, and personality effects to characterize the relationship between social media and collective action. We highlight how greater turbulence is the consequence of the processes we have traced. We identify a model of democracy that encapsulates this turbulence, and provides the basis to understand politics in the age of social media.
