Power

Power is surely one of the most important concepts in political science and political theory.¹ Political science might be seen as the attempt to understand how political institutions work to produce the outcomes they do. Part of that understanding must involve how actors use those institutions in order to produce the outcomes they judge best. If our most basic understanding of power is Weberian — an actor's capacity to get what s/he wants — then seeing how actors use institutions to get what they want is to analyse their power. Political philosophy can be seen as the attempt to design institutions that produce, overall, the best outcomes: to maximize our power as a community, and, to distribute powers to individuals in ways that seem normatively appropriate. Strangely, perhaps, individual power is not so often viewed as the central concept in discussions of social justice. Debate over what the central distributive concept should be usually centres around 'utility', 'freedom', 'resources', 'capabilities' or 'opportunities for advantage' to name but a few.² One reason power is not often seen as a central normative concept is surely that there seems something wrong with thinking that individuals want, as a matter of justice, power. They may want many things we think appropriate, such as happiness, freedom or material wealth (though we might be wary of them desiring too much of the latter), but not power. That smacks too much of an unhealthy ego. Power should be useful to get things that are appropriate, but it is not something to be desired itself.

One of the candidates listed above is Sen's suggestion for the appropriate candidate for equal distribution — capabilities. Capabilities are the abilities to achieve 'functionings' or achievements: 'doings and beings' (e.g., Sen, 1987, 1993, 12). A capability for Sen is the actual ability of someone to achieve outcomes (Dowding, 2006). In that sense, individual capabilities can be seen as individual powers, though as Peter Morriss (2002, xxiii) points out, Sen rarely if ever uses the term 'power'. Morriss also complains that Sen does not use the ordinary (dictionary) meaning of capability. Capabilities are latent abilities — abilities that a person could develop — not abilities they already have. Morriss further suggests that Sen should have used the term power, or ability, or 'ableness', an archaic word that Morriss rehabilitated in the first edition of his book Power: A Philosophical Analysis. Morriss seems to think that what we should be trying to distribute according to some just distribution rule are ablenesses — a type of power — which is equivalent to Sen's capabilities. This type of power cannot be further analysed. It is not reducible to anything else.

In this paper I concentrate upon Morriss's significant discussion of the vehicle fallacy and power-as-ableness in order to reflect how we might view power or capabilities as an object of justice. In part this question is a measurement one. If we believe that power or capabilities should be distributed in some manner — equally for example — as a matter of justice then how we measure that concept is as important as what it is. Pertinent here are two issues raised by Morriss's analysis. First, is ableness reducible to anything else? Second, is ableness the important sense of power as Morriss claims? The first argument is the point of Morriss's account of the vehicle fallacy. I will argue that we can analyse power as ableness by reducing it to agents' resources. Morriss thinks not and labels this the vehicle fallacy. To the extent that power as ableness can be reduced to agents' resources reflects back upon Sen's capabilities as an alternative to Dworkin's (2002) resourcist account of egalitarian justice. I will then argue that while power-as-ableness is a useful concept it is not the most important type of power. Power-as-ability is the underlying concept, power-as-ableness is what one can do with one's power given one's luck. It is not to deny that power-as-ableness is important for certain types of questions, but it is a mistake to think that power-as-ableness is the basic category. In that sense I also distinguish power and capabilities.

The Vehicle Fallacy

Peter Morriss's important book Power: A Philosophical Analysis was re-issued with a long introduction in 2002 (Morriss, 1987, 2002) and he introduces the concepts of the exercise and vehicle fallacies into modern debates about the nature of power. He bases them on an argument of Anthony Kenny's.³ In the introduction to the second edition he says that many people have followed him in abhorring these fallacies and 'nobody (to my knowledge) has tried to argue that these are not fallacious' (Morriss, 2002, xiii). I have no quibble with the exercise fallacy. To confuse the dispositional power of an object or an actor with the exercise of that power is a mistake. It is not a mistake, of course, to use the evidence provided by the exercise of power to examine the properties or resources of the actor to try to understand the basis or foundations of that actor's power. The resource account of power suggests that we can understand actors' power by looking at their resources. Morriss does not disagree with this sentiment. He believes that examining and comparing resources is an important indirect evidence of actors' powers. He also believes that other (non-resource-based) indirect evidence is also useful. However, these other categories all track actors' resources hence all evidence of power is resource-based. Importantly, though, Morris claims it is a mistake to identify an actor's power with his resources. This is an ontological claim. The resource-based account of power claims that once all actors' resources have been measured we have a complete account of the power structure. Morriss disagrees. This would be to identify power with resources. The advantage of such a reduction is that it allows us to measure a dispositional concept — power — in terms of a non-dispositional one.

This reduction from the dispositional to the non-dispositional will not allow us to produce point predictions. Even if we could accurately measure all actors' resources we could still not predict precisely what the outcomes of their interactions would be. First, we would have to feed in their preferences. Second, even if we fed in their preferences many interactions might have the form of multi-equilibrium n-player games with mixed strategies. We might be able to predict outcomes across probability distributions, but we would not be able to produce point predictions. Some of us might find this reassuring. In other words, seeing social life in terms of non-cooperative game theory will ensure that the resourcist route will be predictive, but only in marginalist as opposed to deterministic terms.

Morriss resolutely avoids considering non-cooperative game theory in his book. I believe that non-cooperative game theory needs to be at the heart of the concept of power as well as its analysis. This is so because what counts as an actor's resources can only be understood in terms of how others view them. This will determine their actions towards him, his reactions to them, and their reactions to his reactions and so on. What others believe, and what the actor believes is also predicated upon each others' actions. Morriss (2002, 139) agrees: 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 effects it produces.

But he does not treat them in this way through his account. Rather he seems to equate resources with the physical attributes of actors. How this happens in his account is subtle, since his argument is complex and nuanced, but we will see how this works when we discuss his ability/ableness distinction. However, to reduce an actor's power to her resources would be to commit the vehicle fallacy according to Morriss.

The vehicle fallacy as explained by Morriss does not allow the reduction of power to properties of the objects with that power. Morriss has two takes at this argument. One by quoting Anthony Kenny from whom Morriss gets the idea of the exercise and vehicle fallacies, and one by quoting Willard van Orman Quine, an arch reductionist. Kenny is quoted with approval, Quine is attacked.

Kenny writes: Consider the capacity of whisky to intoxicate. The possession of this capacity is clearly distinct from its exercise: the whisky possesses the capacity when it is standing harmlessly in the bottle, but it only begins to exercise it after being imbibed. The vehicle of this capacity to intoxicate is the alcohol that the whisky contains: it is the ingredient in virtue of which the whisky has the power to intoxicate. The vehicle of a power need not be a substantial ingredient like alcohol which can be physically separated from the possessor of the power. The connection between the power and its vehicle may be a necessary or a contingent one. It is a contingent matter, discovered by experiment, that alcohol is the vehicle of intoxication. (Kenny, 1975, 10, quoted in Morriss, 2002, 14–15)

The necessity or contingency of the relationship is not something that need unduly worry us here.⁴ The relationship between intoxication and the drinking of alcohol is well understood. Some people are much more prone to intoxication (and alcoholism) than others. The bundle of genes that are associated with alcoholism and susceptibility to intoxication are beginning to be identified, though few researchers (and no modern evolutionists) would doubt that one's environment (or 'nurture') may also play a part (NIAAA, 1992, 2000a, 2000b). Other factors may also affect susceptibility to intoxication including age. In one study of rats, the degree of intoxication of rats was clearly age-related. The explanation is probably that young brains are built to learn, whereas old brains are more set in their ways; hence the well-trodden neural pathways may continue to function despite the influence of the alcohol (Holtz, 1997).⁵ People appear intoxicated depending not only on how much they have drunk, but on how much they are used to drinking, how much they have eaten, and their state of mind at the time they drink and so on. Alcohol intoxicates but it does so under all sorts of conditions. The degree of intoxication is affected by aspects of the drinker as well as the strength of the alcohol. So power of whisky to intoxicate depends upon all sorts of factors, and these are analysable such that we can judge the probability of a given bottle of whisky intoxicating different sorts of people under different conditions. This is the way we would analyse the 'disposition' of whisky for intoxication. The mistake of Kenny and Morriss is to think that because the power of the object (in this case alcohol) is affected not only by properties of the object but also by the properties of other objects, we cannot reduce its power to that of its own properties. The thinking seems to be — the properties of an object only have power relative to the properties of other objects, hence the power of the object cannot be reduced to its properties. But this is false, since our judgment about the power of an object is also calculated relative to the environment. We calculate powers with scope restrictions. How we calculate those powers and how we calculate the scope restrictions is fully determined by the properties of all the relevant objects.

In other words, we might calculate for those who drink a given bottle of whisky — that it has the power to kill a child, stupefy a teenager and slow the reactions of the most hardened drinker. And we can compare its strength to another bottle of spirits. The comparative power of another bottle of spirits can be given simply by its alcohol content (and impurities that can affect the speed of uptake of alcohol and short- and long-term damage to the brain affecting, e.g., hangovers).

Of course other things affect how people react to a bottle of whisky but we can analyse its power relative to other bottles simply by its properties. And moreover, its total effects can be measured by reducing all the other elements to their component parts. The explanation of a bottle of whisky's power to intoxicate can be reduced to more basic elements. We can do likewise for the social scientific concept of power.

Morriss argues against this view by taking an example from Quine who suggests that when we say sugar is soluble we are making a claim about its sub-visible structure.⁶ Morriss suggests that there is no reason why dispositionals must refer to sub-visible structures. In fact he does not show there is 'no reason' why dispositionals should not refer to sub-visible structures, rather what he argues is that we can understand sentences containing dispositions without further analysis. This is not the same as suggesting there is no reason why dispositions should refer to sub-visible structures, and certainly not that dispositions may be analysed and examined by looking at other properties of objects. Merely that we do not need to know everything in order to understand a sentence of a natural language. Brian Barry (2002, 161) likewise suggests that the claim that sugar is soluble does not mean that it has a structure suitable for dissolving. It is true of course, that the intension (the meaning) of a statement 'sugar is soluble' is not 'sugar has a structure suitable for dissolving'; rather it means something like 'when you put sugar in liquid it becomes part of the liquid'. However, the statement that sugar is soluble has the extension (the reference) that Quine attributes to it. And Morriss (2002, 18) is correct that we can 'talk perfectly sensibly' about the dispositional properties of objects without understanding the causes of the dispositions the objects have. But I take it that when we do social science we are interested in more than simply talking sensibly about the world. We want to understand and analyse the world. Knowing how the word 'power' can be used sensibly is one thing, but understanding through analysing the power structure of society is another, much more important thing to do.

The extension rather than intension of 'power' is important when doing empirical political science. Outside of the social sciences it is obvious that it is the extension that is important, because causation is extensional. One event does not cause another because of the meanings we happen to attach to the words that describe those events.⁷ In the social sciences it is less obvious that it is the extension that is important since people act based on what they think, so the intensional context is important. However, the intensional context can be captured if we include within the extensional description, belief and desire. In other words, I think the best way of examining the power structure is to examine what sorts of properties (or resources) people have (or can command) given what they want, or may want, to achieve. It is the properties that make sugar soluble, not the dictionary definition of solubility that is of interest to those trying to explain why some objects dissolve in liquid more readily than others. Similarly so it is the properties that lead some people to get what they want more often than others manage and the properties that lead them to want what they want, rather than the dictionary definition of power, that is of interest to the social scientist.

At times Morriss's version of the vehicle fallacy seems to consist in the argument that simply because two things have the same dispositional properties, they do not have to have the same (structural) properties. Of course not, but it does follow that whatever (structural) properties the objects have must include some that give them the disposition in question. One might just deny this. To some extent, that is precisely what Kenny (1975) does and is a standard move in the philosophy of mind where, it is claimed, certain properties simply are not further analysable.⁸ In this sense, the term 'dispositional' tends to mean something different to philosophers than it traditionally does to social scientists. A disposition for a philosopher usually refers to a causal regularity that cannot be mechanistically decomposed. In that sense dispositions are not necessarily unobservable but are contrasted with 'occurrent events'. Thus, individuals beliefs and desires are not objects, and they are not 'occurrent events' so they have to be dispositions. But this just means that under certain conditions q, an individual (or object) i will do (or display) x. Dispositional properties in the social sciences are thought to be unobservables that must be inferred to have causal properties. In which of these senses is a dispositional property of actors? If one takes the first line, then one may resist the reduction of power to resources but only at a cost: the cost of forcing the analysis of power of actors to their beliefs and desires. Morriss will not, I think, want to pay that cost, though below I suggest that the issue deserves very close attention, and the relevance of beliefs and desires to power ascription is more closely bound than many (including myself in earlier work) have recognized. Morriss's important distinction between 'ability' and 'ableness' is particularly pertinent here and will also be discussed below.

It is in the second sense that I have taken it that power is a dispositional property of actors. I take it that the causal properties of the actor to get what she wants are constituted of her resources (that which enables her to get some of the things she wants) and the wants themselves. Her causal properties per se, include all those things that lead her to cause whatever she causes, and could cause. In these contexts, however, the power of an actor is usually conjoined to her desires, intentions or wants (or more formally her preferences) so that power does not simply become the infinitive of 'can' and be too closely associated with causal powers more broadly (see Barry, 2002; Dowding, 2003 for further discussion on this). What makes power dispositional is the fact that the resources that specify the 'can', given the environment, are only switched on under certain circumstances. 'Fragility' is a dispositional property of fine china, which is demonstrated when the china is dropped on to a hard floor. But it is a structural property of china that can be analysed without the china actually being dropped. What 'switches on' the power for actors are their preferences (beliefs and desires). But again, that dispositional property does not need 'switching-on' in order to be analysed. In the same way that 'fragility' can be analysed without dropping the china, the power of actors' can be analysed without their doing anything. At least, by and large it can. (I will say more on this below).

Power in the second sense of disposition is measured by factors in principle observable, but often unobserved — actors' resources; though of course the resources may only be exercised under certain conditions. If we can observe all resources we have reduced the disposition to its (structural) properties and we have completely analysed power. In that sense a resource-based approach to power should be able to analyse the power of all actors to the extent that all resources can be observed. However, the strategic possibilities that open up when actors do not have complete information about the resources of others, which we can safely assume is always the case, means that such a determination of the power structure will remain beyond our means. It does not follow, however, that a resource-based approach, theorized within a strategic framework, is not the best way to analyse the power structure.

Morriss thinks this is false. Because of his belief in the 'vehicle fallacy' he does not think that a resource-based approach to studying power is the right procedure. Morriss does not deny that strategic considerations enter into power relations, but thinks they do not enter in the definition of power. He would be right but for the fact that it is strategic considerations, based on incomplete information, that stops us from fully specifying power in terms of actors' resources.⁹ Thus, strategic considerations cannot be kept out of the meaning of power any more than they can be kept out of its analysis.¹⁰ First let us consider the nature of resources.

Resources

In my own work (Dowding, 1991, 1996; Dowding et al., 1995) I used the term resources to cover five categories in which individuals may use their power. The first four were based on categories used by John Harsanyi (1969a, 1969b) and the fifth the addition of reputation. All are resolutely game-theoretical. The first two are information and legitimacy. The relative advantage that information gives allows some to get what they want more easily. Similarly, others may follow the lead, suggestions or orders of others for content-independent reasons if they recognize the authority or legitimacy of the first to so lead. The third and fourth categories are the abilities of an actor to change the incentive structures of others: through unconditional incentives, and secondly through conditional ones. Unconditionally, changing incentive structures is to change the choice situation of others. This means that others' preferences have not changed, but rather the conditions under which they make choices. For example, a government may try to change the eating habits of citizens. It could try to get people to eat more healthily by persuading them that fried food is bad for them; or it could tax cooking oil. The first uses information, the second unconditional incentives. The first will change the preferences of the citizens (to the extent it works), the other 'preference shapes' — it does not change the preferences, but it changes the choice conditions and so behaviour. Conditional incentives are simply the threats and offers, or combinations thereof that may lead people to behave differently. The fifth resource is reputation. Here people may respond differently to an actor because they believe he has other resources (that he may not in fact have) or because they may predict his actions based on his (or others similarly situated) past behaviour. With the exception of information none of these sets of resources explicitly mention the many attributes of people that might be thought to be resources — money, charisma, a standing army — and so on. Any such list of physical attributes one could produce would almost certainly be incomplete, for many things might be used as a resource in this sense. Rather the five sets of resources provide an organizing principle around which we may study how actors respond to each other in different environments.

Morriss sees actors' powers, correctly, as their property. Their powers are their abilities to get things done. He says: A person's abilities are a property of the person, not of the environment. Like all dispositionals, they carry explicit or implicit references to the conditions in which they apply — and assert nothing about what happens when those conditions do not apply, nor about the likelihood of those conditions occurring. (Morriss, 2002, 80)

The problem with the confidence with which this claim is asserted is that some properties of people only become properties under certain conditions. It is difficult to denote them without the environmental conditions applying. Take the 'gene for reading'. A gene was once thought to be a 'chunk of DNA'. But as we have learned more about genes it becomes clear that complexities occur. A 'gene for' something is best viewed as an abstract truth about inheritance, and the physical attributes of the gene may be distributed around in the physical attributes of DNA. One might think that a 'gene for reading' then will include any part of the DNA structure that, if missing, will lead to reading abnormalities. But this would be too quick, because of course genetic defects that cause, say, cot death will mean that the baby will not be able to learn to read in its life. But the fragment of DNA which ensures healthy sleeping is not normally thought of as part of the DNA that constitutes the 'gene for reading' — otherwise the DNA for healthy lungs, heart and so on would also be part of that gene. More pertinently, however, the 'gene for reading' was present before people could ever read. It may have been useful for tracking, say, and was selected when hunting was important. Later as humans started using symbols and then writing was invented the same abilities made possible by these genes allowed most people to develop the ability to read. Of course, the ability to read entails more than having this gene. It also entails having been exposed to and been taught how to interpret the symbols. There is no doubt that the ability to read is a property of the person, but the ability to read, only makes sense, in an environment (a culture) where reading exists, even though people who have never lived in such an environment have the requisite genes and so could have developed the ability had they lived in such an environment. The ability to read is not, of course, a property of the environment, but the phrase in parenthesis above is important. The environment in this case is a culture that includes writing and reading. Abilities (at least some of them) are then intimately connected to the environments in which people who have them live. Counterfactuals concerned with such abilities must take into account the mechanisms in which societies (the environment within which our individual with his ability to read) develop such abilities.

Ability and Ableness

Morriss makes a distinction between ability and ableness. While both might be measures of an actor's power he believes that the best measure of an actor's power is ableness. I have criticized Morriss's notion of ableness on several occasions. He suggests that my criticisms are false for two reasons, one elementary and one instructive (Morriss, 2002, xxxv–xxxvi). My elementary error is misunderstanding how to counterfactualize ableness; my instructive one is where I introduce the concept of luck. The two are closely related.

Let us first consider my elementary error of misunderstanding how to counterfactualize ableness. I do not, of course, think I misunderstand the counterfactuality, rather my argument is that the counteractualization of ableness in which Morriss indulges is inappropriate. It comes into the class of intuition pumps that go wrong because some relevant factors are omitted (see, e.g., Dennett, 1984). The example used by both Morriss and myself is whether or not all the judges in a Supreme Court have equal powers, or whether one may have more power than the others. Now in a real Supreme Court it would be unlikely if all had precisely equal powers. Some will be more persuasive, aggressive or charismatic than others, or simply have more authority in some sense, and hence be able to determine the decision of the court more often than other members.¹¹ But this is not what is at issue. What matters is whether any member has more power based entirely upon one resource: their vote. If each member has one vote is there any sense in which one member might have more power than the others. Morriss thinks yes; I think no. Imagine there are two blocs of four voters (one conservative, one liberal) plus one maverick justice (MJ) who sometimes votes with the liberals and sometimes with the conservatives. Morriss says MJ has more 'ableness power'. I say: MJ only appears to have more power because she happens to have been denoted in the example as MJ: that is, she has a different preference structure from the others. Her different preference structure does not, in fact, give her greater power over the results since the two blocs have their reasons for voting together and she has her reasons for voting as she does. She only appears more powerful because the preferences of the other two blocs are taken as given hers is not. (Dowding, 1991, 60)

Morriss says 'this last sentence is just wrong. When investigating the power-as-ableness of an actor, we take the actions of all the others as fixed, and examine the outcomes that can be brought about by the actor whose power is being investigated' (Morriss, 2002, xxxvi). I am not too sure what is 'just wrong' about the final sentence. In my sentence, the preferences of the eight justices are taken as 'given' and in Morriss's the 'actions' of the others are 'fixed' which given that I use the term 'preference' as used in revealed preference theory (the standard axioms of rational choice) they seem to be identical. Of course, if the actions of eight actors are fixed to a tie, then the ninth will determine the result. That is not in question. What is in question is whether this is a measure of the actors' voting power. It is not. Since the counterfactual is inappropriate to this question. Imagine one member of the other blocs, called BJ. When we consider BJ's power keeping all the other (non-MJ) votes fixed, sometimes the vote is 5–3, and sometimes it is 4–4, and 'we can conclude that BJ is only sometimes able to win the vote. MJ therefore does have more power than any BJ' (Morriss, 2002, xxxvi). No we cannot. BJ can no more 'win the vote' than the other four voting his way. As I point out (Dowding, 1991, 59–61) being pivotal in this sense, if it is known that one is pivotal can bring other powers. One is open to side-payments, one is then in a position to get other things than one wants (such as bribe money). But being pivotal in this sense does not bring any more power to bring about the outcome under vote, than the one vote that each of the justices brings to the court.

Morriss misunderstands this counterfactual argument since he follows a poor intuition from an under-specified intuition pump. He says he looks at the 'bribe value' of MJ and BJ to some other actor. He suggests that if the CIA want to bribe a justice, MJ will be better for them than any BJ. The others' votes are predictable, they will split the court, and MJ can be purchased to get the results the CIA want. Not necessarily. Despite being a maverick, MJ may have very strong views on each of the cases brought before her. She may not be prepared to sell her vote to the CIA. She may be honourable, and while everyone may have their price, hers may be more than any given BJ. Indeed the CIA may find they can buy two liberals for each maverick. This is precisely my point. Each justice has their reasons for voting one way or another, and it is those that give us predictions over their behaviour. But their power is to get what they want, whatever they want, given their resources (and the environment they inhabit), and in the voting index of which we are both writing the relevant resource is their one vote each. If we want to look further at their bribe value we need to look at their reasons for voting as they do. It is simply not true that MJ is more easily bribable than the other eight justices. Just because she is less ideological does not mean her preferences are weaker, and hence she is more bribable. The CIA may well get what they want by bribing one of the liberal justices (assuming the CIA is right-wing) than the MJ. Then they can ensure a five-way right-wing vote, whichever way MJ votes. (Hence the CIA will win 5–4 some of the time and 6–3 some of the time.) Assuming strength of preference is identical among all the voters, the bribe value of the actors is identical, hence Morriss' argument demonstrates that in this scenario all actors have the same voting power.

But why should we not do the counterfactual analysis that Morriss wants in this example? Why not hold the environment, including the votes of the other justices constant, and then look at what someone can do. What is wrong with ableness here? The answer is that holding other things constant in this way ignores the fact that parts of the environment are reactive. In many contexts what a BJ will do may depend on what an MJ will do. Many, if not most, power situations are strategic. That is not to say that one way of working out what an MJ should do to get what she wants is to imagine what the others would do, but she should work out what they are likely to do under several scenarios, and scenarios that include her actions varying. Obviously this takes us into game theory, and without a game-theoretic analysis of such interactions our concept as well as our analysis of power will be poorer. It is Morriss (2002, 91), who seems to make the elementary error: Your ableness to win the competition would not be lessoned because many other people had this ableness too. When we assess powers person-by-person like this, we take the actions of others as given and counterfactualize the preferences and actions only of the actor whose we are considering (and, of course, the preferences of any others that we predict will be altered if the actor under consideration changes her preferences).

How can we hold all others' actions as given, if we are allowed to vary their preferences (as we surely must) with the preferences of the actor under consideration? (That is not to say, however, that keeping others' preferences and actions constant while we vary those of the individual under consideration is always illegitimate, but in many strategic contexts it would be. The only way one can find out how to play a strategic game is by working out what all the other players will do depending upon what one does. Here we look at all the logically possible scenarios but it is the feasible set that matters for analysis.)¹²

It might be thought that Morriss and I are simply looking at the issue from two different perspectives. He is looking at the issue as one of the power of coalitions, and I am looking at the issue as one of the power of individuals. Imagine the committee of nine is not a court but a parliament. There are two parties composed of four MPs, and one party composed of one MP. In that case it is quite natural to claim that each of these three parties has the same power. In either the Shapley–Shubik or Banzhaf power indices each party has the power of 1/3. Here surely the member of the party composed of one member (MJ) has more power than any member of the parties composed of four. We might think that each member of the party blocs has a quarter of the power of 1/3 that is 1/12. But that is not correct. MJ may have the power of 1/3 seen as a party bloc in this manner. But MJ's power disappears if one member of either bloc votes against his bloc. Each member of each bloc has the same power as MJ. Each bloc member only appears to have less power then MJ because we are considering the bloc as necessarily voting together. In my terms, MJ is lucky, not powerful.

It is this that Morriss calls my instructive mistake. Following Barry I suggest that luck is getting what you want without trying, or as Morriss correctly amends it, one is lucky to the extent that given one's preferences, one can get more than one might expect. MJ is lucky in that she is always decisive. As I point out that luck might have been based on someone else's power (a President who appointed her, who shares precisely her 'maverick' preferences) but for her it is luck. Morriss seems to want to equate 'ableness power' and luck when he says '[w]here Dowding goes wrong is in thinking that luck (in his sense) and power are incompatible' (Morriss, 2002, xxxvii). I make explicit the point that an actor may be both powerful and lucky (e.g., Dowding, 1991, 83; see also Dowding, 2003), that is someone may get more than the effort would suggest, but they could have put more effort in. But this is not what Morriss means here. He agrees that MJ is lucky, but why beyond the fact that she gets what she wants all the time because she is pivotal? In that sense, luck and ableness seem to be the same thing. Similarly my extension of 'luck' to 'systematic luck' where the amount of luck a person has is determined by the structures of society (Dowding, 1991, 137–138, 152–157) means that some people are lucky simply to be who they are: male, white, upper-class or whatever. Morriss equates systematic luck with his concept of 'passive power'. Are Morriss and Dowding having an uninteresting spat over the term to use for a concept? I think not: at least not in other ways in which Morriss uses the notion of ableness and his critique of realist approaches to power.

For Morriss the distinction between ability and ableness is vital for he holds power as ableness is the important sense of the term. (For me, luck is much less important than power. Indeed it is a residual used to explain why some people get more of what they want than we might expect given their resources — or given how they have spent those resources.)¹³ Morriss (2002, 83) suggests that 'an ability-sentence contains (explicitly or implictly) counterfactual descriptive and manifestation conditionals' whereas an 'ableness-sentence cannot contain counterfactual descriptive or manifestation conditionals'. The former is descriptive under which the conditional holds (e.g., the conditions under which sugar will dissolve in water) and the latter denotes the conditionals that hold when those conditions obtain (when the water and sugar actually come together). He goes on: For this reason, social and political power is usually a sort of ableness and not an ability, in social philosophy we are not usually interested in what people could do if they had resources that in fact they do not have. (Morriss, 2002, 83)

There are, I think, two things wrong with this claim. First, it is not clear that ableness sentences do not contain counterfactual descriptive or manifestation conditionals. The sentence 'I cannot read it today (my glasses are broken)' implicitly contains the descriptive conditional 'if my glasses were not broken, I could read it today'. The importance of the counterfactual conditions is that they specify the resources that I require in order to be able to read. In this sense ableness is what we are interested in for it is the resources that I require in order to be able to read. The only way I can see a distinction between ableness and ability is some physical feature (the gene) of an individual that implies that an ability could be there. But as we saw above genes cannot simply be treated in that way, for the phenotypes (us and our behaviour) are a relationship between our genes and the environment, and an environment, moreover, we have partly fashioned. Reading is something we have fashioned, given our genetic inheritance to be sure, but one cannot read off ableness from anything simply physical about a person such as a fragment of their DNA. Imagine, for example, that Homo erectus had the fragments of DNA that are identified as 'the gene for reading'. Would we say Homo erectus had the ableness to read, 300,000 years before writing was invented? And this leads to the second problem.

Typically when analysing the power structure we are interested in the resources that people do not have, since we are explicitly or implicitly comparing the powers of classes or types of people. One of the reasons that Morriss does not get to the heart of the difficulties of counterfactual analysis in the main part of his book is that he only considers somewhat contrived examples of what people might do if the CIA bribed them or threatened them. The real problems in social science do not concern such token individuals but types of individuals. Typically, the sorts of questions we ask in the social science are: what is the US President able to do when another party controls the legislature? How does his power differ from that of the French President under those same circumstances? Or, can institutions be designed to counteract the power of developers in local politics? Is the march of globalization inevitable? A lot more factors enter into these considerations, than whether someone will do their best or not, to try to help the CIA (see Morriss, 2002, 77–79). And many of these factors involve the mechanisms that may (or may not) exist within the structures of the legislatures and the political systems of the respective countries (in the presidents example), as well as considering the actions and reactions of the multitude of actors involved (in all the examples). Here, more control is required than that allowed by Morriss's approach to counterfactuals.

Morriss's arguments are directed almost exclusively to the focus of looking at a token individual's power. However, my focus is almost exclusively on examining the power structure of societies and thus on the power of types of actors. Thus, part of what is an actor's ableness — his skill 'is, of course, itself a dispositional term...' (Morriss, 2002, 138). From my perspective however, we can assume that the unobservable elements such as 'skill' within a type are normally distributed across the set of actors within the type, and therefore need not worry too much about this aspect of their resources. We may assume their luck is normally distributed also. It is only when we come to compare across types, that we become interested in their relative resources and luck.¹⁴ For someone writing a book on a theoretical concept like power Morriss has a very inductivist approach to understanding and truth. Time and again he suggests that we can only discover actors' true powers (and the dispositions of objects such as sugar) by experiment. Theoretically, however, once we understand why salt dissolves in water we should be able to predict that substances with similar 'sub-visible' structures such as sugar will also dissolve in water. Demonstration in science and the social sciences is not always through experiment or empirical observation.¹⁵ Morriss's athoeretical inductivist approach leads him to discount realism and underlies his approach to the ability-ableness distinction.

Anti-Realism

Part of Morriss's attack on the vehicle fallacy is his dismissal of realist approaches to power. In this context a realist approach is not merely a claim about the truth-values of particular power claims, but of mechanisms that underlie power ascriptions. Ted Benton (1988, 492) calls Morriss's approach one of power-attribution that can function as an explanation-sketch or explanation-attribution. In Morriss's inductivist approach one cannot go further in causal claims than Hume's constant conjunction. The idea underlying realism is that, to take the fragile china example, there is something structural in the nature of fine china that makes it fragile. This entails that when dropped it will (usually) break. Morriss argues against this view of underlying structures entailing that china is fragile by considering 'finkish dispositionals'.

We can follow an intuition-pump of C.B. Martin (1994) discussed by Morriss. Let us say that the fragility of china is related to temperature. The warmer the china the less breakable it is. Martin suggests that if the china is warmed and then dropped and does not break it is not fragile.¹⁶ But with Martin imagine that between being dropped and breaking 'a divine agent' intervenes to make it break. In Martin's example this divine agent 'makes it fragile' again — perhaps by cooling it. However, they might do it by any sort of means, speeding it up just before it hits the ground, or zapping it with sub-atomic waves that causes it to break up at the same moment it touches the ground. Is this 'finkish disposition' a real disposition to fragility? According to Morriss, we sometimes should ascribe fragility under these finkish circumstances and sometimes not. We should not if we were doing a scientific analysis of the china, but should if we are labelling this special china (with its 'divine agent' mysteriously attached) for sale to the public. The problem with these absurd intuition-pumps is that authors do not take them seriously enough. If one is to entertain that a 'divine agent' can intervene between dropping and breaking, we must consider how we, as scientists, would analyse that condition and how we would then consider it in our scientific analysis of china. I suggest that we would categorize the fragility of objects in terms of their properties — all their properties — and relate these to the conditions under which they causally lead (under some probability distribution) to certain outcomes. In the case of the china, we would measure the structural properties of the china, and relate this to the probability of its breaking when dropped under different gravitational forces, temperature, humidity and other conditions. Given the fact that some types of china break under conditions that our theory does not allow, we reanalyse in appropriate ways, discover the divine agent who intervenes, note the type of china that the divine agent is attached to and modify our laws accordingly. In Martin's example as I have further specified it — by suggesting ways in which our divine agent intervenes — we do not have to modify our account of fragility, rather the conditions under which the special china is dropped. These always alter in some manner caused by the divine agent to ensure the china breaks. On the other hand, we might find that the divine agent has a new way of breaking china and then we might have two rather different mechanisms for the same outcome. Generally speaking in natural laws, such mechanisms can often be related with more general accounts underlying both. However, prior to the discovery of such underlying accounts we can easily work with different mechanisms. As pointed out earlier, the fact that different structural properties may underlie the same dispositional properties does not mean that those structural properties do not fully explain the dispositions. (To repeat, they do so under certain conditions, but these conditions can be specified as part of the dispositional property.) So even under the finkish circumstances we can still reduce the disposition to natural properties (taking the intuition-pump seriously) and can discover real properties and mechanisms.¹⁷

Morriss makes strange use of the finkish disposition example. He suggests that the powers of actors might indicate different things 'It may be that each of two leaders has the power to do the same specific thing but the explanations of their powers could be very different: the power of one derives from the social structure, whilst the other's would come from the leader's personality' (Morriss, 2002, xxx). Yes. But why is this statement a criticism of realism? Actors may have very similar powers based on very different resources, which is precisely why it is of interest to carry out the reduction of explaining their powers in terms of their resources rather than leaving them as explanation-indicators. Where Morriss and some of the realists he criticizes may both go wrong is insisting on a strong ontological distinction between an actor and an individual such that we either have individualist or structural explanations of power in society. As I have tried to explain elsewhere the two are not contraries. Only individuals act, but their preferences are shaped and their actions enabled and constrained by their resources and those of others around them (Dowding, 1991, 1994, 1996, 2001; Dowding and Hindmoor, 1997). One person's action is (part of) another's structure deep down into their preferences (Dowding, 2008a). Again I repeat that in the social sciences most explanation is given in terms of types — leaders, dictators, capitalists and so on. We are interested in the types of resources that these actors have, and the distributions of these resources relative to other distributions.¹⁸ This does not imply, of course, that we cannot use these types when we investigate the powers of token actors. But our knowledge of their resources will be informed by what we have discovered, both theoretically and empirically from the resources of others in similar and often dissimilar situations.

What is the Fallacy of the Vehicle Fallacy?

The vehicle fallacy is itself fallacious. It is 'ignoratio elenchi' — supposedly refuting one thing, while in fact arguing against something else. Kenny and Morriss argue that one cannot reduce the dispositional concept of 'power' to the properties of the object to which the power is attributed. But one can. What one cannot do is analyse the casual effects of those properties without also taking into account the properties of other objects. That is what they should have been demonstrating. But that is true of all features of all objects. Indeed any unary property of an object can only be defined in terms of binary properties — the relations of that object (and its properties) to other objects (and their properties) (Dowding, 1991, 15). Hence we cannot identify an object without taking into account other objects, nor can we measure its physical properties, causal properties, powers or whatever. Not understanding this logical feature of our descriptions of the universe causes almost as much confusion in social scientific theoretical discourse as misunderstanding the importance of the type/token distinction in explanatory contexts (see Dowding, 1994, 2001).

Power undoubtedly is a resource of actors, but we can only analyse that resource in relationship to the resources of other actors. How much we can buy with the money in our possession in part depends upon how much money is in the possession of others. The power of money is relative to its distribution. This means that strategic considerations must enter into the very essence of the concept of power. How much power people have, what resources they have, may in part be composed of their luck. We may have no control over what resources we have, or what resources others have. Nor indeed over our preferences or those of others (though we may have some power in both cases). For that reason I think the analysis of both power and luck is important, but we must start that analysis by looking at actors' resources, and the strategic situations in which they find themselves.

I began this essay by suggesting (following the lead of Morriss) that what Sen means by the term 'capabilities' is actually a form of individual power, ability or ableness. And for Sen, crudely speaking, justice requires we maximize functionings (what people do with their capabilities) and equalize capabilities.¹⁹ To the extent that capabilities are power, and to the extent that I am right that we can best measure individual power in terms of resources, then Sen's capabilities account of what egalitarian justice requires is not so distinct from the Dworkin's (2002) resource account that Sen criticizes in building up his capabilities approach. To be sure, my account of what individual resources are is broader than Dworkin's, and in that broader account might lie the difference between Sen and Dworkin. Nevertheless, I believe that working through the measurement of power, or capabilities, will lead us to examining individual resources and in that sense, Sen's capabilities account at base, is resourcist.

Notes

¹ A version of this paper was first presented at the PSA conference in Leicester 2003. I thank the participants, particularly Patrick Dunleavy and Peter Morriss for their comments. I also thank Matthew Braham, and an anonymous referee for comments and especially Martin van Hees for many discussions of related issues over the years.

² Brighouse and Fleurbaey, MS is the closest counterexample I can think of, but is not really about this topic.

³ All references are made to the second (2002) edition of Morriss's book first published in 1987. The second edition is identical apart from minor typographical changes and a new long Introduction.

⁴ I would say it is a 'naturally necessary' feature of alcohol. Alcohol was developed and drunk because it helped destroy germs. It also has the pleasant effects of intoxication. Thus, whisky and its alcohol content are not simply contingent features of our social world. They were developed for a purpose and they fulfilled a function, and thus their presence and why we drink whisky are not entirely contingent. How people react to its effects is also not contingent. People whose heritage is European are less prone to the effects of alcohol than those from the east or Americas since those in the east and the Americas used other techniques to free their drinking water from germs — largely through heating and using herbs (see e.g., Diamond, 1998).

⁵ Alcoholics on the other had may be strongly affected by the first drink, as the alcohol immediately sets their brain on its 'alcohol-directed path'.

⁶ This and the next two paragraphs are taken from Dowding (2003).

⁷ Although what we recognize as a 'cause' depends upon what causal question we are asking. And so what we define as 'cause' and what as 'background conditions' does attach to the meaning of the words by which we describe those events.

⁸ Such as qualia.

⁹ See Dowding (2006) for a related argument about capabilities, and Dowding (2008c) for an extension of the strategic problems over what to equalize for luck egalitarianism.

¹⁰ Braham and Holler (2005) argue that the concept of power is contained with the game form though it is exercised through the game. This is a subtly different account from Morriss's. Here the game form will specify the power of the player given how the game will be played once preferences are assigned. (A game form is the rules of a game without the assignation of preferences). But how the preferences are assigned is, in my terms, luck, since that will specify which coalitions are likely to form. An agent's power is then partly a matter of luck, and partly a matter of the game form. We can identify resources with the rules of the game (the game form).

¹¹ Studies of voting patterns on supreme courts in US States have shown that where the votes are taken in order of seniority, decisions go with the first voters more often than those where votes are taken in reverse seniority or random order.

¹² Dowding and van Hees (2007) consider the strategic inter-relationship between preferences and the concept of freedom.

¹³ It is not an ad hoc device used to 'save' any purported explanation. Rather luck — especially in its systematic form — is analysable in terms of the convergence or divergence of actors' preferences. And, I think, other structural features of society can help explain such convergence or divergences. I show in Dowding (2008c) that if outcomes are equalized (by whatever that equalization is measured) luck disappears.

¹⁴ See Dowding and van Hees (2003) for more on the importance of type and token in the measurement of rights and freedoms, in relationship to different measures of welfare in Dowding (2008d) and in the social sciences more generally in Dowding (2008e).

¹⁵ Of course, in the sugar example, empirical demonstration may well be easier than theoretical demonstration! But this is not so for all cases in the sciences or social sciences.

¹⁶ To begin with this just seems like a mistake. As Morris recognizes dispositions need to be defined in terms of the conditions under which they hold. An object is still fragile under the conditions in which it easily breaks. We consider fragility of ordinary objects under the temperatures in which we handle them. But scientifically fragility would be defined relative to temperature.

¹⁷ If the divine agent only sometimes intervenes then we either analyse the conditions under which He does — perhaps it always happens on wet Tuesdays — or recognize the intervention is random under some probability distribution.

¹⁸ This is what we are interested in when considering the power structure. In mainstream political science, 'power' as a concept has almost disappeared as much is concerned with explaining specific outcomes rather than the relative outcomes (Dowding, 2008b), though as the literature on the power of different institutions of the EU demonstrates, power does re-enter when those relative comparisons are made.

¹⁹ Sen denies that he has a 'theory of justice' and would, I think, say this aphoristic representation of his argument is simplistic; nevertheless the upshot of his considerations seems to me to boil down to this claim.

References

1. Barry, B. (2002) 'Capitalists rule OK? Some puzzles about power', Philosophy, Politics and Economics 1(2): 155–184. | Article |
2. Benton, T. (1988) 'Review of Peter Morris Power', Sociology 22: 491–493. | Article |
3. Braham, M. and Holler, M. (2005) 'The impossibility of a preference-based power index', Journal of Theoretical Politics 17(1): 137–157. | Article |
4. Brighouse, H. and Fleurbaey, M. (MS) 'On the fair allocation of power', http://philosophy.wisc.edu/brighouse/PapersonWeb.htm.
5. Dennett, D. (1984) Elbow Room: The Varieties of Free Will Worth Wanting, Cambridge, MA: Bradford Books.
6. Diamond, J. (1998) Guns, Germs and Steel: A Short History of Everybody for the Last 13,000 Years, London: Vintage.
7. Dowding, K. (1991) Rational Choice and Political Power, Aldershot: Edward Elgar.
8. Dowding, K. (1994) 'The compatibility of behaviouralism, rational choice and the "new institutionalism"', Journal of Theoretical Politics 6(1): 105–117. | Article |
9. Dowding, K. (1996) Power, Buckingham: Open University Press.
10. Dowding, K. (2001) 'There must be end to confusion: policy networks, intellectual fatigue, and the need for political science methods courses in British universities', Political Studies 49(1): 89–105. | Article | ISI |
11. Dowding, K. (2003) 'Resources, power and systematic luck: a response to Barry', Philosophy, Politics and Economics 2(3): 305–322. | Article |
12. Dowding, K. (2006) 'Can "capabilities" reconcile freedom and equality?', Journal of Political Philosophy 14(3): 323–336. | Article |
13. Dowding, K. (2008a) 'Agency and structure: interpreting power relationships', Journal of Power Studies 1(1): 21–36. | Article |
14. Dowding, K. (2008b) 'Rational Choice Perspectives', in S. Clegg, M. Haugaard and C. Lester (eds.) Handbook of Power, London: Sage.
15. Dowding, K. (2008c) 'Luck and responsibility', Critical Review of Social and Political Philosophy, forthcoming.
16. Dowding, K. (2008d) 'What is Welfare and How might it be Measured?', in D. Ross and H. Kincaid (eds.) Oxford Handbook of the Philosophy of Economics, Oxford: Oxford University Press.
17. Dowding, K. (2008e) '"Defensa de la Preferencia Revelada" ("In defence of revealed preference theory" trans. Fernando Aguiar)', Revista Internacional de Sociología LXVI 49(Enero-Abril): 9–31.
18. Dowding, K., Dunleavy, P., King, D. and Margetts, H. (1995) 'Rational choice and community power structures', Political Studies 42(3): 265–277. | Article |
19. Dowding, K. and Hindmoor, A. (1997) 'The usual suspects: rational choice, socialism and political theory', New Political Economy 2(3): 451–463. | Article |
20. Dowding, K. and van Hees, M. (2003) 'The construction of rights', American Political Science Review 97(2): 281–293. | Article |
21. Dowding, K. and van Hees, M. (2007) 'Counterfactual success and negative freedom', Economics and Philosophy 23: 141–162. | Article |
22. Dworkin, R. (2002) Sovereign Virtue: The Theory and Practice of Equality, Cambridge, MA: Harvard University Press.
23. Harsanyi, J.C. (1969a) 'Measurement of Social Power, Opportunity Costs, and the Theory of Two-person Bargaining Games', in R. Bell, D.V. Edwards and R.J. Wagner (eds.) Political Power: A Reader, London: Collier-Macmillan.
24. Harsanyi, J.C. (1969b) 'The Measurement of Social Power in n-person Reciprocal Power Situations', in R. Bell, D.V. Edwards and R.J. Wagner (eds.) Political Power: A Reader, London: Collier-Macmillan.
25. Holtz, A. (1997) 'Study: young brains especially susceptible to alcohol', CNN Interactive Health, http//www.cnn.com/HEALTH/9701/09/nfm/youth.alcohol/index.html, accessed 4 March 2003.
26. Kenny, A. (1975) Will, Freedom and Power, Oxford: Blackwell.
27. Martin, C.B. (1994) 'Dispositions and conditionals', Philosophical Quarterly 44: 1–8. | Article |
28. Morriss, P. (1987) Power: A Philosophical Analysis, 1st edn., Manchester: Manchester University Press.
29. Morriss, P. (2002) Power: A Philosophical Analysis, 2nd edn., Manchester: Manchester University Press.
30. NIAAA (1992) 'The genetics of alcoholism, national institute on alcohol abuse and alcoholism', No. 18. PH 357, July: http://www.niaaa.nih.gov/publications/aa09-text.htm, accessed 4 March 2003.
31. NIAAA (2000a) 'Why do some people drink too much? Role of genetic and psychological influences, 10th special report to the United States Congress on alcohol and health', 24(1): http://www.niaaa.nih.gov, accessed 4 March 2003.
32. NIAAA (2000b) 'From genes to geography: the cutting edge of alcohol research', National Institute on Alcohol Abuse and Alcoholism, no 48, July: http://www.niaaa.nih.gov/publications/aa09-text.htm, accessed 4 March 2003.
33. Sen, A. (1987) 'The Standard of Living: Lecture II: Lives and Capabilities', in A. Sen (ed.) The Standard of Living, Cambridge: Cambridge University Press, pp. 20–38.
34. Sen, A. (1993) 'Capability and Well-being', in M.C. Nussbaum and A. Sen (eds.) The Quality of Life, Oxford: Oxford University Press, pp. 30–53.