Power obsessed

ABSTRACT

Governance is best analysed using rational choice theory to identify the extent to which different groups can systematically punish and reward others. This is roughly in line with Dowding’s important book Rational Choice and Political Power. It is different though in the sense that it does not focus on the measurement of social power, but rather a subset of it. Focusing on the latter avoids difficulties relating to systematic luck, such as distinguishing it from legitimacy and the possession of information, and also clarifies the importance of coordination problems for the analysis.

KEYWORDS:

• Social power
• governance
• systematic luck
• collective action problems

How do we determine who governs in society? In Rational Choice and Political Power Keith Dowding suggests the first step is to determine which groups have the most resources by counting things like the aggregate wealth and the votes of group members. If this first step was all there was to answering the question of governance, we would answer that the people govern, the electorate governs, the 99.9% govern, and be quite content with the distribution of governance in society despite the ever-growing inequalities of the market. While it is lamentably close, the bottom 99.9% (at the time of writing) still have an accumulated wealth greater than the top 0.1% in virtually every country. Our answer, in other words, would be normatively very favourable to the status quo.

Dowding’s argument is that we need to probe further by asking to what extent these groups are hampered by coordination problems. We should first measure a group’s aggregate resources, but then we need to also, as a crucial second step, identify and control for any coordination problems they may face. Are there powerful agenda-setters amongst the top .1% who can use their resources to manipulate the issue-dimensions of the vote (McKelvey 1976, Riker 1988)? Is the reason they face no resistance because the bottom 99.9% are crippled by collective action problems associated with large groups (Olson 1965)? Perhaps it is difficult to identify group members, undermining the possibility of self-regulation (Ostrom 1990) and information pooling (Spiekermann 2020). Whatever the case may be, the upshot of RCPP, as I see it, is that collective action problems are what ultimately determine whether inequalities in resources translate into inequalities in governance.¹

I think this picture is roughly right. We should not conclude that all is well with society if there are diverse groups making policy decisions (e.g. Dahl, 2005) but where large groups are still locked out by collectively sub-optimal yet individually rational non-action. The argument in this paper is an attempt to defend this rational choice picture from some of the controversy associated with the concept of systematic luck, which was first introduced in RCPP in relation to Dowding’s definition of political power. This is in tension with parts of the book. In particular, I question the focus on the concept of social power in the sense that it stretches the rational choice approach too far. Not all inequalities in social power, at least as political power is defined in RCPP, are normatively concerning. Scientific experts during the Covid pandemic or a news service that has built up high viewership based on quality reporting wield considerable political power, yet not in the same normatively relevant sense as the political power wielded by capitalists. It is difficult to see how the power of experts and legitimate authorities might be measured relative to, say, the capitalists’ control over the means of production. Making the question of governance about a precise specification of social power also introduces considerations of responsibility. The concept of systematic luck makes caveats with respect to who exercises social power and underscores that agents should not always be held responsible for predictably getting what they want. But determining which groups are specifically responsible is not the same as determining which groups cannot be responsible due to insufficient resources or coordination problems. The rational choice picture can capture the latter, I suggest, but not necessarily the former.

1.1. Systematic luck

In RCPP Dowding posits definitions for outcome power, social power, luck, and systematic luck, which have proved influential. He defines outcome power, or ‘power to’ as

‘The ability of an actor to bring about or help to bring about outcomes’

(Dowding 2019, p. 48)

A subset of this kind of ability is what he calls ‘social power’,

“The ability of an actor deliberately to change the incentive structure of another actor or actors to bring about or help bring about outcomes.” (Dowding 2019, p. 48)

This can be read as ‘power over’ in the sense that it describes an agent’s power over another. It is taken to roughly capture the conceptual space associated with governance. An agent’s ability to manipulate incentive structures is measured by the resources they have to cover the costs of the means of social power. The means of social power include things like threats, offers, the provision of information, and commands. This suggests resources like wealth, votes, legitimacy, and reputation are all important to measure.

Power here is to be distinguished from luck, which is getting what you want without deliberately exercising power to get it. Dowding (2019; see also 1996; 2017) provoked controversy with the concept of systematic luck to distinguish when groups systematically get more of what they want without trying from when they get it from exercising political power. Some groups will be systematically lucky in the sense they ‘have more luck than others because of the sort of society in which they live’ (Dowding 2019, p. 5). This kind of luck is ‘non-random’ and can be ‘predicted methodically’ (Dowding 2019, p. 134).

So agent A predictably getting what they want at the expense of agent B will not always be due to A intentionally interfering with B’s expected utility calculations. In such cases we should call A lucky and not necessarily socially powerful. Corporate interests getting a tax cut because unions cannot overcome coordination problems to mobilise opposition may be the corporate interests’ luck, not their power in this sense. It could be their systematic luck given we can predict these coordination problems in modern capitalist societies. Capitalists may not be responsible for the tax cut in the sense of deliberately manipulating policymakers to pursue it, and it would be committing the blame fallacy (Dowding 2019, p. 88) to suggest otherwise. They may be lucky, and when this luck is predictable given the kind of society (e.g. capitalist, patriarchal, Christian, secular, etc.) in which they live, then they are systematically lucky.

The substantive thesis that follows from these definitions, as I read it, is that the evidence of a group’s systematic luck should not be conflated with evidence of governance. This is where I disagree. Disproportionate amounts of systematic luck is evidence that certain groups are locked out from governance. At least it is difficult to find a clear case where we can predict a group disproportionately getting what they want given the kind of society under evaluation that is not. This may well be because the kind of society will partly be determined by the inequalities of governance. While a group with a favourable bargaining position may not be directly responsible for systematic outcomes, this does not mitigate the normative issues associated with the inequality. It may be inappropriate to sanction the group, but still appropriate to sanction the status quo. Making governance about the ‘deliberate’ manipulation of incentive structures is in keeping with ordinary language associated with ‘social power’, as is the strict relation between ‘governance’ and ‘social power’. If there were such a strict substantive relation the ascription of systematic luck would be crucial. My hypothesis, however, is that there is no case of systematic luck disproportionately privileging some groups over others that is not evidence of a normatively relevant inequality in governance.

1.2. Collective action problems

Dowding’s own examples do not appear to undermine this hypothesis. He uses Crenson’s study of clean air ordinances in American cities, for instance, to suggest legislation to mitigate the pollution of U.S. Steel was kept off the policy agenda in the small town of Gary, not by the social power of U.S. Steel, but its systematic luck that the citizens of Gary faced a collective action problem. Groups could not coordinate effectively due to problems associated with large numbers (given the interest group is large since clean air is in everybody’s interest).

Olson (1965) argues that in larger groups there is a lower probability an individual will be decisive, which disincentives collective action (following the inequality that it is rational to free-ride when 0 < B*P – C + D, where B is the benefit the individual receives from the collective action, P the probability of being decisive, C the cost of the individual’s action, and D expressive incentives). A smaller group that benefits from this kind of logic will be systematically lucky in a way that is hard to say is caused by an inequality in bargaining ability. But we can say it will itself lead ceteris paribus to an inequality in bargaining ability that will be concerning from a democratic standpoint (see also Ostrom 1990). It will mean smaller groups will be able to mobilise their resources better than larger groups. This will favour smaller more coordinated corporate interests like U.S. Steel in Gary over larger democratically inclusive groups.

The discussion of Gary in RCPP then moves to a comparative analysis. It shifts to explaining why the citizens of Gary overcame their coordination problem so late comparatively to other cities like East Chicago. The answer, again, is luck. But this is not systematic luck. It was not predictable luck, so therefore not systematic. Variance boiled down to a number of ‘contingent political features’ in particular a specific kind of individual in the public attorney’s office (see Dowding 2019: 94, p. 98). In East Chicago ‘no one could understand why this individual worked in a public office when he could have had more lucrative employment in the private sector’ (Dowding 2019, p. 93). Later in Gary, the ‘entrepreneur was more persistent; at one point he is described as a man in search of an issue, but he also had a private interest in clean air – he suffered from a respiratory disease’ (Dowding 2019, p. 94).

The important point is that we can predict the difference between groups that are predictively large and those that are predictively small in capitalist democracies. We cannot, however, predict the difference between Gary and East Chicago. While elites will certainly have coordination problems of their own, their failure relative to more inclusive groups will tend to be unpredictable in light of the specific kind of society in which they live. So while the elite’s coordination problems will be the non-elite’s luck, it is not their systematic luck.

1.3. Preference falsification

Whenever we can predict variation in the strategic dilemmas groups face, my hypothesis is that every case turns out a concern for the state of democratic governance. It will either be the cause or effect of worrying inequalities in bargaining ability. The effects of these inequalities capture the vast majority of what Lukes (2005; see Dowding, 2019: x) calls the third dimension of power. Certain pre-existing inequalities in bargaining ability predictably cause collective action problems like preference falsification amongst the disadvantaged. Those groups that benefit from the preference falsification will often be systematically (i.e. predictably) lucky because they have not deliberately done so.

One of the keys here is understanding the importance of the ‘sucker’ group. Suckers can be created via selective threats and offers, or their anticipation. The most obvious example of this, and an example that is quite clearly structured by inequalities in the ability to issue threats and offers, and not a case of systematic luck, is the suppression of free speech. Once dissenters start being carted off to the gulag as punishment for criticising the government, citizens may clearly desire the state of affairs where there is a large level of dissent, but prefer others do it and so defect (D) rather than risking their own skin in cooperating (C). This is the preference structure of the standard Prisoner’s dilemma, i.e. D,C > C,C > D,D > C,D, which has the suboptimal Nash equilibrium where nobody expresses their true opinion.

While this is a collective action problem, it would be wrong to call it the government’s systematic luck that nobody is speaking out. If anything is power, it is this kind of suppression. But the same analysis applies to more subtle coordination problems where the ascription of systematic luck appears appropriate. Preference falsification, for example, involves groups whose members are unaware they share common preferences, or at least are uncertain of the preferences of others (see Kuran 1997). Preference schedules are not common knowledge like they are in the Prisoner’s dilemma. Yet preference falsification is almost always structured by the sucker possibility to suppress the revelation of this common preference. This possibility will be again structured by inequalities in governance.

This, at least, was Mary Wollstonecraft’s argument in The Vindication of the Rights of Women. She puzzled over the number of middle class women in the late-18^th Century who expressed a preference against an equal education and extending the franchise, policy she thought squarely in their interests. This would appear a prime candidate for a case of systematic luck in the sense that we can predict men will disproportionately get what they want given the kind of (i.e. patriarchal) society. Men are systematically lucky in the sense that they can spend public resources on members of their own sex with no resistance (and at other times, encouragement) from members of the other. Wollstonecraft diagnosed the situation as a collective action problem. She suggested the race to win a hand in marriage incentivised women to falsify their preferences with their ‘false refinement’ favouring modesty ‘gentleness, docility, spaniel-like affection’ and ‘innocence, as ignorance is courteously termed’ (Wollstonecraft 1993, p. 100). If they did not act in this fashion, they would likely be perceived as manly, and (so it was thought) never married off, ostracised to the ‘sucker’ group of ‘spinster’. Spinsters had few rights relative to others in the middle class, and often had to resort to work as maids for married women. The danger of being ostracised from society in this way, Wollstonecraft (1993, p. 276) argued, lead to ‘perpetual rivalships’ between women, where ‘they are all running the same race, and would rise above the virtue of mortals if they did not view each other with a suspicious and even envious eye.’

Those predictably benefitting from this collective action problem may be described as systematically lucky in the sense they did not deliberately manufacture it. Notice though that this does nothing to suggest systematic luck was not the consequence of an inequality in governance. That the sucker possibility of ‘spinster’ existed, where women had little in the way of prospects for subsistence short of marriage was, I think, a consequence of women not having the vote. Governments did not face an electoral threat from women and were therefore unlikely to offer much to secure their support. The sucker possibility that structures the relevant collective action problems, in other words, was created and perpetuated by an inequality in the ability to issue credible threats. Women did not have the resource of the vote and could not mobilise any other resources to fight for it. Why did married women not speak up? A simple explanation I think is the oppressive nature of 18^th Century marriage law, where women had no right to property, income, their children (in the event of divorce), etc. They were in a position of domination with respect to their husbands.²

So in relation to the question of governance, the concept of systematic luck does not appear to add anything. While it arguably makes important distinctions with respect to the responsibility of men, it does nothing to mitigate the normative problems associated with the inequalities in society. Crippling coordination problems will certainly begin to skew belief. Insofar as women continued to falsify their preferences, Wollstonecraft conceded people would begin to believe it and start to sincerely infer mistaken generalizations about women. It is easy to mistake the effect of unequal treatment for its cause³ and Wollstonecraft argued this was ubiquitous in the 18^th Century. It is to mistake what Dowding (2019, p. 37) calls an endogenous preference for an exogenous one. This mistake will likely be reinforced by way of information cascades amongst the population. Many in the 18^th Century sincerely believed that equal education and franchise extension was genuinely not in the interests of women. If this belief were truly ubiquitous, those benefitting from it would be systematically lucky. While the sucker possibility is important for explaining the initial mistake, it appears to fade into strategic irrelevance as people start to sincerely update their belief in light of the beliefs and behaviour of others.

The stability of these kinds of skewed beliefs appears to be itself as much conditioned by a collective action problem as the initial preference falsification. It is unlikely every woman bought the 18^th Century dogma; Wollstonecraft would not have been alone in her scepticism. What was more likely was that the sceptical simply faced the very same collective action problem that lead to the cascade effect in the first instance. It was seen as close to scandalous to even mention Wollstonecraft’s name for close to a Century after the publication of the Vindication, for instance, let alone her argument. If they could somehow mobilise and overcome it, an information cascade in the opposite direction was possible. Such cascades are slow processes and we still see considerable unjust discrimination (e.g. along the lines of sex and race) based on past and continuing inequalities in governance. The systematic luck of misogynists here, however, is a symptom of the underlying inequality in ability to issue credible threats and offers between men and women. To say they are systematically lucky really says nothing to mitigate the normative concerns with the inequality in bargaining power between men and women. My hypothesis that disproportionate amounts of systematic luck go hand in hand with disproportionate amounts of governance therefore still holds.

1.4. Provision of information

While preference falsification was one of the best candidates to put pressure on the hypothesis, it is probably not the most controversial. I think this is reserved for the idea that capitalists are systematically lucky when governments pass favourable policies to keep their business. Companies may have no intention of informing the government of their intention to liquidate or move their business offshore in the event of a corporate tax hike, but the government anticipates their exit and therefore leaves the tax rate as is. Is this governance? So stipulated, I think the clear answer in RCPP is ‘no, corporate interests are just systematically lucky’. It therefore looks to be a clear case where systematic luck is not evidence of an inequality in governance.

Once we unpack what is implied here, however, it is less clear. If the government were unaware of the corporation’s intention and the corporation were merely to inform the government, the corporation would go from being systematically lucky to governing in a puff of definitional smoke. The provision of information is counted as a means of social power in RCPP (Dowding 2019, p. 69) given it can be used to deliberately alter the expected utility calculations of other agents. How much success here is due to social power and how much is due to systematic luck really does not make a substantive difference. What might in general be called ‘informational resources’ depend entirely on what others know. If they know what you know, then getting what you want is your systematic luck; if they do not know then it is your power resource. So while the former is technically a sense of systematic luck that is not a case of governance, it does not cut against my hypothesis in the sense that it is still evidence that they could have exercised governance if they were not systematically lucky.

There is, however, something normatively problematic with the case of capital flight that the resource approach should be able to capture. That is, whether or not the corporation can provide this information specifically as a credible threat. If the ability of transnational corporations to threaten policymakers outweighs the electoral threat of voter groups, it is deeply problematic for the state of the country’s democracy. If businesses like U.S. Steel can threaten local politicians with exit, they can effectively hold towns like Gary to ransom (see Dowding 2019, p. 95).

A threat, however, is quite different from the CEO’s sincere communication of their contingency plans. For the provision of information to be a threat, moving offshore must not have been what the business would have done anyway. A threat amounts to the commitment – or bluff – to do what one would not otherwise have done in the event of the tax hike in order to punish non-compliance. It is a threat if the CEO would not have moved in response to the tax had they never communicated the information. What is important here is that the CEO’s information could be a credible threat if they have sufficient resources to cover costs associated with moving the company. Local businesses are rarely in this position, but transnational and non-local corporations routinely are given they can move their resources relatively easily. The upshot is that it is often democratically problematic to rely substantially on the business of non-local corporations. They can mobilise their resources to cover the costs of punishment in a way that will often trump the electoral threat of domestic groups. If there is demand for a policy that cuts against the company’s interests, the CEO is in the position to threaten the government and, if they are self-seeking and rational, will. This will likely be known by the government and anticipated, and so following the law of anticipated reactions, if the government is rational the policy will not be entertained in the first place. Non-local businesses like transnational corporations often therefore have the resources and strategic incentive to punish (and reward) policymakers more than any coordinated voter group. This is a worrying inequality for democratic governance.

We can measure the extent of this inequality with the resource approach. This is important given there is a normatively salient distinction between business getting what it wants because of a threat or offer and what it wants because of its interests aligning with government. The latter though is not a clean case of systematic luck because it will often require the provision of information in order to get the desired outcome. Non-local corporations can (and do) use this as a sleight-of-hand in normative evaluation to conflate their fortunes with those of domestic businesses. But they should be kept analytically distinct for they are substantively very different.

The distinction is preserved if we only focus in on the ability of groups to issue credible threats and offers: what I refer to as a group’s bargaining ability (see Harsanyi 1962). There is a difference between companies holding towns to ransom given and business sincerely informing policymakers about their business prospects. One involves a credible threat; the other does not. Bargaining ability, however, is not all there is to social power. The CEO’s sincere provision of information is not in and of itself associated with their bargaining ability, but it is their social power. If we continue to focus on the distinction between systematic luck and social power, then, we are conflating what is normatively distinct. But if we draw the distinction within the extension of social power itself, then I think we can capture the normatively important distinction.

1.5. Incommensurable resources

The resource account is ideally suited to measuring this inequality in ability to issue credible threats and offers. Not so much social power more generally. I take an agent’s resources here to be what roughly determines the extent of their ability to systematically reward and punish others. To maintain credibility, groups will therefore need to collectively mobilise these abilities to stably reward their successful offers and punish their failed threats. If they cannot coordinate to do so, representatives of the group will lose any kind of bargaining ability due to the quickly evaporating credibility of any subsequent threats and offers (see e.g. Selten 1978, Kreps et al. 1982).

Given Dowding defines social power as the ability to change the behaviour of others by deliberately manipulating their utility calculations, anything that covers the cost of the means by which to do this is thrown into the resource basket. So not only wealth and votes count as resources, but reputation and information as well. Dowding follows Harsanyi (1962) in treating legitimate authority and affection as a means of social power and (by way of resources that are supposed to cover the cost of this authority and affection) an object for measurement. It is difficult to grasp what would amount to a resource of legitimate authority. Part of the reason is that it appears to be a category mistake. Legitimate authority and affection are not means of social power. The relevant means to manipulate the expected utility of others in this sense are commands and requests. Commands and requests have effect when they are backed by legitimate authority and affection. In terms of the resource approach to social power, then, this would render legitimate authority and affection an agent’s resource not a means of social power. We could, then, say the capitalists ‘systematic luck’ discussed in the previous section is a case of governance in the sense that it is based on the ‘affection’ of governments and voters towards capitalist industry. Even if they do not explicitly make the request, the anticipation of the request (via the law of anticipated reactions) is enough.

Correcting for this category mistake puts one of the best attacks on the resource approach into sharper relief. It begs the question how we should measure legitimate authority and affection relative to other resources like votes and wealth. In Chapter 19 of Power: A Philosophical Analysis Peter Morriss argues that resources are too intangible and incommensurable with one another to be implicated in the empirical measurement of governance. We might be able to measure legitimate authority with something like surveys and interviews. But it is difficult to know how we would square the chosen measure with the measure of a group’s wealth. Morriss’ argument here is worth quoting at some length given it is instructive as much as for what it gets wrong as it gets right. He suggests,

“The main difficulty with resources is that a resource is not an empirical datum, like a chain of office or a palace: we cannot observe resources directly. We have to infer that things are resources by examining other people’s reactions to them; one cannot simply measure resources since the worth of a resource is determined by the effect it produces. (Compare money, which, as the economics textbooks tell us, is only money – as opposed to lumps of metal or pieces of paper – when it is widely accepted as such within the economy.) Studying resources is every bit as complicated, and indirect, as studying power itself. In order for the investigator to identify a resource, he needs to have a theory of others’ motivations.” (Morriss, 2002, p. 139)

The first point, however, does not appear to hold unless one is sceptical of the measurement of social objects and kinds more generally. The paradigmatic cases of ‘empirical datum’ Morriss uses are not distinguishable from resources in the sense he suggests. They are likewise what they are only by virtue of ‘people’s reaction to them’. A building is a palace only if individuals have, at some point, treated it as the official residence of a ruler. A chain of office is likewise only a chain of office if individuals have taken it to be one.

Morriss’s second criticism though is telling and, if we are to count information, reputation, legitimate authority, and affection as an individual’s resource, decisive. Measuring social power by way of its vehicle (i.e. resources like reputation and legitimate authority) will be ‘every bit as complicated, and indirect, as studying power itself’. To know somebody has an important resource, we would need to observe the reactions of others to that resource to establish its relative value. This leads to epicycles in debate, as it has, Morriss suggests, in the classic elitist (i.e. those who believe the wealthy govern) and pluralist (i.e. those who believe the vote mitigates the effect of wealth) debate in political science. He argues,

“For one of the differences between these schools of thought hangs on the question how important a political resource the vote is. Pluralists and elitists agree that, by and large, votes are evenly distributed throughout Western societies, whilst almost all other conceivable resources are very unevenly distributed … As yet, we are not very far advanced towards developing a theory of the political process that allows us to evaluate these divergent sorts of resources. If we are to choose between these two competing accounts of the distribution of power within Western societies, we need to employ one of the other approaches, rather than a resource-based one.” (Morriss, 2002, p. 143-144)

The vote will certainly be an irrelevant resource if policymakers are corruptible, uninterested in winning another term in office, and only interested in the kickbacks they can secure when they leave. The vote will be very relevant, on the other hand, if politicians are incorruptible, refuse to listen to lobbyists, and have little concern for how they might fund their re-election campaigns. Morriss is right in arguing we will only know which is which by directly observing which resources are effective with respect to policymakers. It begs the question why we should rely on such an indirect method as counting resources rather than, say, directly counting policy success.

1.6. Rewards and punishments

I suggest this is difficult to overcome if we want to measure social power as it is defined in RCPP. If we restrict the rational choice picture to bargaining ability, however, we can measure bargaining ability in a far more robust way indirectly via resources than directly via the behaviour of policymakers. Doing so allows us to determine which groups are locked out of governance and make normative evaluations of states of affairs on that basis. I take bargaining ability here to be the threats and offers agents can make to one another. I suggest this ability is best measured by which groups can systematically reward and punish others agents. We capture the systematic relation (contrary to Dowding’s discussion of systematic luck) by measuring resources like money and legal rights and identifying the coordination problems groups face.

Insofar as resources can be used to reward or punish individuals, they will be actively desired or feared. Individuals will therefore have an in-built incentive to take or destroy them. I think this rather obvious point is all we need in way of a ‘theory of others’ motivations’ and ‘theory of the political process’ to reply to Morriss’ objection. That is, something only counts as my resource if there is some active threat or offer in play to dissuade them from taking or destroying it. A plot of land only counts as my resource if there is a law of trespass excluding others by way of threat of legal punishment. A vote is an individual’s resource by virtue of the threat of civil disobedience in the event the government chooses to ignore the result. Digits on my internet banking page represent a resource by virtue of the threat to inform others in the event the bank does not honour their offer to return my deposit, potentially triggering a bank run and, at the very least, legal sanction.

So resources for systematically punishing and rewarding others will be constituted out of the active exercise of threats and offers. In advanced democracies these threats and offers will be relatively stable with respect to what is traditionally counted as a resource. Were Donald Trump to ignore the result of the 2020 Presidential Election, we can predict groups will coordinate to put pressure on him, his administration, and his family in a way that is an effective deterrent. If a government were to violate principles associated with private property it is likely they would be either voted out of office or, failing that, violently removed. We can predict that groups will overcome coordination problems to punish this level of non-compliance amongst elites and so bank on the on-going decisiveness of resources like votes and money for rewarding and punishing others. These resources are systematically structured by coordination problems in a way legitimate authority, information, and reputation are not. As Dowding persuasively argues in RCPP, solutions to coordination problems will be predictable in a relatively robust way. So wealth, votes, and other legal protections should count as resources; reputation, legitimate authority, affection, and information should not.

Once we narrow our attention to the extent to which individuals are able to systematically reward and punish others, Morriss’ objection to the resource approach in political science dissolves away. While the preferences of policymakers will ultimately determine the importance of the vote relative to wealth, the extent to which groups can coordinate their vote and wealth can be equally said to determine the kinds of policymaker who will be in office in the first place. Given the stability of threats and offers that secure the vote and property, we can rely on a couple of important assumptions. First, while corrupt politicians may be insensitive to the electoral sanction, voters will still have the aggregate resources to vote them out. The wealthy still have the wealth to groom and finance rival candidates to run against the incorruptible. Where voter groups are gripped by collective action problems and there are large inequalities in wealth, you will predictably have politicians open to financial incentives and vice versa when there is a well-mobilised electorate. If voters cannot mobilise, it is predictable politicians will be sensitive to the whims of wealthy elites. When voter groups can mobilise and overcome collective action problems, it is predictable they will vote in politicians who will be less sensitive to wealthy elites and more sensitive to electoral sanction. So inequalities in the extent of groups’ ability to systematically reward and punish will determine the kinds of policymakers in office. The resource approach is therefore a robust way to indirectly measure inequalities in bargaining ability.

1.7. Conclusion

I think this is a consistent response to Morriss. It is important, however, to emphasise it does not seamlessly follow from the argument in RCPP. Morriss’ criticism still holds if we wish to measure social power as it is defined in RCPP rather than restricting our attention to the inequalities in only those resources that enable groups to systematically reward and punish others. The latter is what is important for the democratic quality of a society (pace Pansardi, 2021) given rewards and punishments are what determine the respective credibility of threats and offers issued on behalf of groups.

A leader of a political party may have won legitimate authority over their members through honesty and good management. Does their ability (by means of a request) to get their supporters to turn out to vote in a by-election suggest a problematic inequality in governance? I suggest this is unlikely. There is only really a problem in terms of democratic governance if the party leader won their credibility and legitimacy by way of inducements bestowed to them by inequalities in wealth and political office. This is what we ought to be measuring and it is exactly what the resource approach captures. Dowding (2015) argues that we should always ensure we are measuring only what is at issue in the normative argument: no more, no less. I think we are therefore justified restricting our attention to inequalities in the ability to credibly threaten and offer (i.e. inequalities in bargaining ability). If Dowding were to drop systematic luck from his argument going forward, and focus less on capturing the semantic space of terms like ‘power’ and ‘luck’, he would have a less controversial approach to the governance literature, but one that is still decisive over a wide range of political argument.⁴

One of the misimpressions of the resource account I think is in the name. While it is important to be analytically clear with what is meant by a resource, the measurement of resources is simple once we have a consistent theory. My claim (contrary to RCPP) is that the relevant resources in most formally democratic countries roughly amounts to wealth, votes, and other legal rights. There is not a lot of difficulty counting these, at least compared with Dowding’s more intangible inclusions of reputation, information, and legitimate authority. The empirical meat and nuance of the resource theory is with the identification of collective action problems. Valid empirical disagreement (short of routine observational errors) amounts to alternative yet ultimately falsifiable hypotheses regarding coordination problems. This amounts to the progressive empirical research program championed in RCPP – one that I believe is as pressing now as when the book was first published.

Disclosure statement

No potential conflict of interest was reported by the author.

Additional information

Notes on contributors

William Bosworth

Will Bosworth is LSE Fellow in Political Theory at the London School of Economics. His research interests range over the field of PPE with a focus on conceptual analysis and methodological questions in political science and political philosophy.

Notes

1. A broadly representative distribution of governance is assumed to be necessary for a just society, but not always sufficient. As will be discussed, there are injustices caused by past historical inequalities in governance, even where those inequalities have now been corrected.

2. Dowding, Wollstonecraft, and Lukes all seem to be on the same page here. The latter two, however, do not need the concept of systematic luck to make the point. One of the standout features of RCPP in my view is to show how we can use revealed preference theory to empirically identify even the most obscured collective action problems. But it is important to emphasise that Wollstonecraft and even Lukes also endorse the barebones of the theory. The idea is that our inferences about the interests and preferences of individuals should be ultimately grounded in their observable choice. Lukes argued the Dalit Indians – the so-called ‘Untouchables’ – who actively seemed to endorse the Indian caste-system, claiming it was the natural way of things despite it strongly discriminating against them, were oppressed by the third dimension of power. He suggests that with careful observation of their behaviour, the third dimension reveals itself: when the opportunity to convert to alternative religions like Islam and Christianity presented itself, i.e. ‘when the apparatus of power was removed or relaxed’ (Lukes 2005, p. 50), the Untouchables converted en masse. The Dalit’s public support for the Caste system was likely (Kuran 1997: 131–3; see also Dowding 2017, p. 42) due to a sub-optimal coordination problem associated with the sucker group created by the Dalit’s tendency to ostracise the small number who sought higher stations in normal times. Wollstonecraft likewise looks for natural experiments to identify preference revelation. When young boys and girls were given the same toys to play with (e.g. dolls and doll clothing), they revealed identical preferences (Wollstonecraft 1993, p. 109–10); when men are nurtured by their career (e.g. soldiers) into obsessing over their body as much as women were (Wollstonecraft 1993, p. 82), they revealed similar ‘rivalships’ as women; and so on. She argued this indicated a large number of differences between men and women were socially and strategically conditioned. Both Lukes and Wollstonecraft appear to be in keeping with the spirit of revealed preference theory here.

3. Aristotle’s account of slavery being a case in point.

4. In the new edition, Dowding (2019: xi, p. 165) endorses the method of elimination (Chalmers 2011, Bosworth 2020), which should incline him to rephrases of his position that do not rely on the terminology of ‘social power’ and cognates like ‘systematic luck’ (pace Pansardi, 2021).