Exploitation, Vulnerability, and Social Domination*

Nicholas Vrousalis†

March 2013

And because man is a human being

he doesn’t care for a boot in the face.
He wants no slaves below him
and no masters above him.

Bertolt Brecht
Song of the United Front

This essay attempts to revive a research programme by criticizing it. Its object of
criticism is the powerful research agenda of analytical Marxism. I shall attempt this
revival by criticizing the views of two of the most influential members of that research
programme, G. A. Cohen and John Roemer. I will argue that they are mistaken in their
identification of exploitation with exchange against the background of injustice in the
distribution of assets. Exploitation should be conceived, instead, as a form of domina-
tion, that is, domination for self-enrichment. The latter conception captures intuitions
surplus to the traditional analytical Marxist view, provides a richer, and more plausi-
ble, understanding of socialist goals, is more amenable to integration into a rigorous
Marxist social science, and brings Marxism closer to radical democracy. If I am right,
then the idea of exploitation will have received a new lease of life, and the Marxist
armoury will have been enriched by renewed focus on vulnerability and domination.
The paper is structured as follows. Section 1 offers a general definition of exploita-
tion as the self-enriching instrumentalization of another’s vulnerability. Section 2 re-
stricts the purview of this definition to the economic structure of society. Section 3
defines domination, and argues that exploitation, as conceived antecedently, is a form
of domination. Section 4 discusses the connection between exploitation and class dom-
ination. Section 5 sketches the Cohen/Roemer distributive-injustice conception of ex-
ploitation. Section 6 argues that exploitation as domination for self-enrichment supplies
*
Forthcoming in Philosophy and Public Affairs. Thanks to Dan Halliday, Faik Kurtulmus, Roberto
Veneziani, and two anonymous referees for excellent criticisms of earlier drafts of this paper, John Filling
and Allen Wood for helpful discussion, and audiences at Bristol, Cambridge, Leuven, Manchester, and the
Nordic Network for Political Ethics Conference at Vejle for comments and suggestions.
†
Faculty of Philosophy, University of Cambridge, and Institute of Philosophy, KU Leuven:
nicholas.vrousalis@lmh.oxon.org

Electronic
Electroniccopy
copyavailable
availableat:
at:https://ssrn.com/abstract=2118772
http://ssrn.com/abstract=2118772
a better account than exploitation as distributive injustice. Section 7 responds to an ob-
jection raised by Roemer against the domination-based account. Section 8 concludes.

1
The general definition of exploitation I propose is as follows:

(1) A exploits B if and only if A and B are embedded in a relationship in

which (a) A instrumentalizes (b) B’s vulnerability (c) to extract a net benefit
from B.

In (1) A and B are free variables referring to agents or groups or coalitions of opti-
mizing agents. The rest of this section clarifies elements (a)-(c).

1.1
(a) Exploiting is commonly and rightly associated with using. It is, moreover, a special
kind of using, for it has the ring of inappropriately ‘playing on’ (in Feinberg’s 1988
apt expression) some set of attributes or features S of another. Instrumentalization of
S implies that S is being used as a means. It is, however, only when S comes to cover
a range of sufficiently salient attributes (in virtue of which the agent is who he is or
does what he does) that we can appropriately talk about exploitation of person by per-
son (exploiting someone’s dribbling weakness in football need not amount to exploiting
him). According to Robert Goodin, such ‘whole person exploitation’ obtains just when
the exploitee’s attributes are used unfairly: ‘taking advantage of other people’s honesty
or blindness to steal from them constitutes exploiting those people tout court’ (Goodin
1987, p. 171). As will become clearer later, I do not think unfairness is a necessary
condition for whole-person exploitation (henceforth: exploitation).1 Nor is it necessary
that exploitation be intentional: one can unintentionally, or unknowingly, instrumen-
talize another’s vulnerability, and thereby exploit him.2

1.2
(b) Vulnerability comes in two broad forms: absolute and relational. Unlike absolute
vulnerability, relational vulnerability is semantically close to dependence, but covers a
1
I am here in agreement with Allen Wood’s nuanced discussion of Doyle Lonnegan in Wood 1995, p.
152.
2
Say B is a poor man whose only valuable possession is an expensive coat. He needs to sell it in order to
get food. Mistaking B for an eccentric millionaire, A offers B a very low price for the coat. B accepts because
of the urgency of his need. Here A has unknowingly instrumentalized B’s vulnerability and extracted
a net benefit from him (the value of the coat minus the price A has paid), thereby unknowingly and
unintentionally exploiting B. Assume, further, that B (falsely) believes A to be more needy than himself,
and accepts A’s offer for that reason. Here, it would not be irrational for both A and B to believe that B
exploits A, although both beliefs are false.

Electronic
Electroniccopy
copyavailable
availableat:
at:https://ssrn.com/abstract=2118772
http://ssrn.com/abstract=2118772
broader class of phenomena. I shall unpack this distinction presently. An agent suffers
absolute vulnerability just when he suffers a substantial risk of a significant loss (possibly
relative to some sufficiency threshold) in the relevant metric (welfare, resources, ca-
pabilities, etc.). The absence of absolute vulnerability is guaranteed by security, which
implies such losses will not occur. Now, absolute vulnerability does not make essential
reference to an agent’s power over another, or indeed to other agents whatsoever. Peri-
odic volcanic explosions make villagers situated near the volcano absolutely vulnerable,
but human agency need not intrude on the side of the content of their vulnerability. Re-
lational vulnerability, on the other hand, makes essential reference to other agents. B is
relationally vulnerable to A only when A has some sort of power over B.
Consider the following structural analogy:

Deep Pit – People are stuck in a deep pit, each situated deeper than the other.
Every now and then there are mudslides. The deeper one is, the worse he
is liable to suffering from mudslides.

In Deep Pit the relative position of each person gives us some information as to who
is more vulnerable, in the sense that, if A is situated higher up the pit than B, then
we are justified in inferring either that he is not vulnerable at all, or that he is less
vulnerable than B. But it is not in virtue of B’s position relative to A that B is vulnerable.
Rather, B is vulnerable just because B is deep inside the pit and there are mudslides:
B’s predicament can be fully described without reference to (the positions of) any other
agents. B is vulnerable in the absolute sense.
Now add the following modification to Deep Pit: A has a rope, and it would be very
easy for B to use that rope to climb up, thus avoiding worse mudslides. In this case, B’s
vulnerability proper is no longer just a function of depth and other agent-independent
facts about the world. Indeed, if we assume that it is neither difficult, nor costly, for
B to move upwards once A throws him the rope, and for A to do so, B’s vulnerability
becomes a function of A’s willingness to throw the rope. B’s vulnerability can only be
described with reference to other agents, or, more precisely, the powers such agents
have over B (and their dispositions to use them).3 B is vulnerable in the relational sense.
Thus one set of sufficient conditions for B being relationally vulnerable to A is: (i)
B lacks some desideratum x that is a requirement for, or constitutive feature of, B’s
flourishing4 (in which case x is the object of B’s need),5 (ii) B can only obtain x from
3
It is true quite generally that, if A can prevent sufficiently significant losses to which B is liable, B
does what he needs to do to avoid such losses, and what he needs to do is what A tells him to do, then A
has power over B. This result carries over naturally to the specifically economic case, which I discuss in
section 2 below.
4
Of course not all failures to flourish result from relational vulnerability. That is, I do not deny that
forms of absolute vulnerability (say, due to incurable and unimprovable disease) exist. All I claim is that
a large number, perhaps the majority of affronts to human dignity or frustrations to human flourishing
imply, or result from, relational vulnerability. (A walking disability, it is often said, is no handicap for the
disabled if buildings are equipped with ramps, elevators, and so on. All such disabilities constitute forms
of relational vulnerability.) Exploitation is responsible for a subset of such frustrations and affronts.
5
Note that ‘need’ here does not necessarily denote some sufficiency or decency threshold. If basic need

Electronic copy available at: https://ssrn.com/abstract=2118772

A, and (iii) A has it within his discretion to withhold x from B. The account developed
here is similar to that of Goodin (1985), who has argued that states of vulnerability
become morally salient when they generate exploitability, and exploitability obtains if
and only if (i)-(iii) are met.6

1.3
(c) A benefits just when there is an increase in A’s well-being. A obtains a net benefit
when A’s overall well-being increases.

2
2.1
Before going on to study the conceptual connection between exploitation and domina-
tion, I must narrow the scope of definition (1) to economic exploitation. Let me explain
why. Exploitation, in general, has something to do with the beneficial instrumental-
ization of another’s vulnerability. It follows that exploitation can only be completely
removed by eradicating all human vulnerability (leading to benefit), or by getting peo-
ple never to engage in such instrumentalization. But the introduction of such remedies
is generally either impossible, or positively undesirable. Gullibility, or plain ignorance,
constitute (non-economic) forms of vulnerability that render one exploitable. It is,
however, usually impossible to have such traits removed, or to prevent their instru-
mentalization. The disposition to act in good faith towards others, to take another
example, can constitute vulnerability in some contexts. It is positively undesirable that
such a disposition be removed or even discouraged. Indeed, it is no exaggeration to
say that certain (non-economic) forms of vulnerability are constitutive of intrinsically
good relationships (e.g. relationships of affection and love), and tend to be reciprocal
over the course of a life (see Goodin 1985, p. 193). Economic vulnerability, by con-
trast, is never constitutive of such relationships and is never reciprocal (see below). It
pertains centrally, moreover, to the basic structure of society, that is, its fundamental,
publicly observable, institutional features, sometimes said to form the ‘primary subject’
of justice (Rawls 1971).
Now, economic vulnerability can be defined as follows:

(2) B is economically vulnerable to A if and only if B is vulnerable in virtue

of B’s position relative to A in the relations of production.
is defined in terms of such a threshold, and human flourishing requires more than meeting that threshold,
then flourishing requires the satisfaction of non-basic forms of need.
6
Unlike Goodin, I do not think (iii) is a necessary condition for exploitation, for there is nothing con-
tradictory in the thought that A is forced to exploit B, and therefore lacks the said discretion. But I do not
need to press such disagreements here.

Electronic copy available at: https://ssrn.com/abstract=2118772

 The relations of production are systematic relations of effective ownership, and there-
fore of power, over human labour power and means of production in society.7 Say A
owns a water-producing well. If A’s ownership is fully enforced, B needs water, and B
has no independent access to water, then B is economically vulnerable to A. The related
Marxist claim is that, under capitalism, B is vulnerable to A in virtue of A’s ownership
of means of production and B’s lack thereof, which gives A power over B, whether or
not B is exactly forced by his economic circumstances to supply his labour power to A.8
But what does it mean for some agent to have economic power over another?
Philosophers sometimes say that A has power over B only if A can get B to φ, and B
would not otherwise have φ-ed.9 This is too strong. For A can powerover B10 even if B
does what he would have done anyway. What we must say, instead, is that A powerovers
B if and only if A is able to get B to φ for reason S and B would not have φ-ed for S
if A had not done (or failed to do) whatever he did (or failed to do) that bears on B’s
decision to φ.11 All this hardly furnishes a sufficient condition for economic power-over,
however. For on the definition just offered a child can have power over her parents, or
a mouse over an elephant. Economic power-over obtains just when A has the relevant
ability12 (to get B to φ...) in virtue of control over a greater share of resources than B.
That is, if Wa is A’s wealth, Wb is B’s wealth, and P ab stands for ‘A has economic power
over B’, then, ∀a, b (P ab → Wa > Wb ).
Given this set of definitions, (2) implies:
(3) If A instrumentalizes B’s economic vulnerability to A, then in so doing
A takes advantage of his power over B.
Here’s some intuitive support for (3). In the rope variation to Deep Pit (see section
1), relational vulnerability only figures in virtue of the relative position of agents: B is re-
lationally vulnerable just because some agent A somehow stands above B and can throw
him the rope. This is exactly analogous to superior position in terms of wealth.13 Thus
economic vulnerability is a form of relational vulnerability, such that, if B does not own
any means of production (or, more broadly, wealth) and A does, or B owns substan-
tially less than A, then B is economically vulnerable to A and A has economic power
7
See Cohen (1978, chapter 3) for discussion of the distinction between formal and effective power.
Broadly construed, the set of relata in the relations of production will include raw materials and machinery,
but also knowledge and perhaps human talents.
8
I discuss force in section 5 below.
9
See, for example, Lovett 2010, chapter 3.
10
‘Powerovering’ (a wordplay on ‘overpowering’) is a convenient verbalization of ‘power over’, which I
use to refer to instances of the latter type.
11
See Julius (forthcoming) for a similar definition. Note that the inference from ‘A has the ability to get
B to φ’ to ‘B has the ability to get A to get B to φ’ is invalid.
12
Strictly: A has the relevant ableness, that is, the ability and the opportunity to do so. Ability-sentences,
unlike ableness-sentences, imply counterfactual conditionals (see Morriss 2002 for discussion).
13
It follows that there can be no ‘reciprocal’ economic power-over: if Bill Gates and Warren Buffett
own approximately the same amount of wealth, then neither powerovers the other economically. I do
not deny that Gates can be vulnerable to Buffett if, for example, Buffett knows something sensitive about
Gates’ personal life. But this would not count as an instance of economic vulnerability, for it does not
arise systematically from production relations.

Electronic copy available at: https://ssrn.com/abstract=2118772

over B.14 In the case that exercises Marxists: assuming an equal distribution of internal
resources,15 the wealth owned by capitalists systematically gives them a decisive bar-
gaining advantage over workers, which means capitalists always have economic power
over workers and not vice versa. Under capitalism, improvements in the productive
power of workers, or even in their material well-being, systematically rivet them to the
capitalist Juggernaut ‘more firmly than the wedges of Hephaestus held Prometheus to
the rock.’ (Marx 1976, p. 799)

2.2
I turn now to economic exploitation. My flagship definition is as follows:

(4) A economically exploits B if and only if A and B are embedded in a

relationship in which (d) A instrumentalizes B’s economic vulnerability (e)
to appropriate (the fruits of) B’s labour.

(e) needs clarification: A appropriates B’s labour when B toils for H hours, and
A appropriates use values produced by B embodying H − X hours of toil, for any
H > X ≥ 0. A appropriates the fruits of B’s labour when B produces an amount of
H widgets, and A appropriates H − X of these widgets, for any H > X ≥ 0.16 For
simplicity, the discussion will proceed in terms of the fruits of one’s labour, which I
assume to be an increasing and continuous function of labour time.
Now assume that B produces H widgets in a rudimentary economy without capital,
where H represents total widget production. Assume, further, that, for whatever rea-
son, A appropriates H − X, X ≥ 0. This is an instance of unequal exchange. Unequal
exchange obtains just when there is an unreciprocated net transfer of goods or labour
time from one party to another. Some philosophers attribute to Marx a ‘technical’ no-
tion of exploitation, according to which unequal exchange is not only a necessary, but
also a sufficient condition for economic exploitation. It is doubtful that Marx held such
a notion.17 But whatever Marx thought, the content of the attribution is implausible.
For gift-giving implies unequal exchange, but noone thinks (even systematic) gift-giving
14
The form of relational vulnerability I am working with here is therefore individuated independently of
sufficiency or decency thresholds (see section 5 for related discussion). It follows that superrich capitalists
can exploit, indeed dominate, rich capitalists (in Marxist idiom: ‘big’ capital can dominate ‘medium’
capital). It does not follow that the exploitation suffered by workers, or other non-owners of means of
production, is not morally more urgent.
15
Such as bodily and mental capacities, including talents. For the significance of this caveat, see Cohen
1995, pp. 149-151.
16
None of what I say implies, or requires, the labour theory of value, which postulates a correspon-
dence between value and labour time. Moreover, although my definition makes no reference to economic
surplus, it is consistent with (indeed, is entailed by) the view that capitalists exploit workers by appro-
priating surplus labour time. That is, if N is the net product, V is the sum of wage goods, and S the
surplus in a competitive capitalist economy, such that S = N − V , then exploitation normally consists in
instrumentalization of the workers’ vulnerability to appropriate S.
17
For vindication of these doubts see Arneson (1981) and Geras (1992).

Electronic copy available at: https://ssrn.com/abstract=2118772

exploitative. If one part of society freely18 decides to pass on a large part of whatever
use-values it creates (with its own labour power) to another part of society, the re-
sulting inequality in the consumption of (surplus) labour need not be objectionable.
Only a set of conditions over and above unequal exchange can establish the presence
of exploitation (I claim that the conjunction of (d) and (e) furnishes such a condition).

3
In this section I argue that there is a conceptual connection between exploitation and
domination. To do this, I first offer a set of sufficient conditions for domination. I then
show that the set of necessary and sufficient conditions for exploitation ((d) and (e))
put forward in (4), implies that set of sufficient conditions.
Domination, in its most general form, is subordination offensive to equality of status.
More precisely, I take a part of domination to be constituted by disrespectful (that is,
degrading, or demeaning, or humiliating) powerovering:

(5) A dominates B if A and B are embedded in a relationship in which (f) A

takes advantage of his power over B, or the power of a coalition of agents
A belongs to, in a way that is (g) disrespectful to B.

A comment on (f): that A takes advantage of his power over B to get B to φ (or to φ for
different reasons) does not imply that A exercises that very power. For B might φ strate-
gically, e.g. in order to avert A’s exercise of that power. The proverbial playground
bully takes advantage of his power to beat up other children when they independently
do what he wants them to do, without exercising that power.
Two comments on (g): First, any sufficient condition for domination requires some-
thing surplus to (f). For, as any parent knows, mere power-over another does not suffice
for domination. (g) completes the relevant set of sufficient conditions. Second, some
action-type, or form of relationship, is disrespectful to B just when it fails to express
respect to B. A’s φ-ing, for example, respects B if and only if it expresses a rational
(nonpropositional) attitude R vis-à-vis B,19 and R is dialogically endorsable. The ques-
tion then becomes which putative attitudes count as dialogically endorsable. Moral
philosophers have proposed a number of plausible answers, so I shall simply state the
one I believe to be most promising in this context: R is dialogically endorsable if and
only if the considerations giving rise to R can be advanced as putative justifications
for action in the context of embarassment-free dialogue among interested parties.20
‘Embarassment-freeness’ obtains when people have no reason to feel shame, or guilt,
for putting, or allowing, a particular form of justification for some putative act on the
table of discourse. An implication of this ‘expressionist’ condition is that the extent to
18
‘Freely’: not by dint of domination, coercion, or force.
19
On rational attitudes, see Anderson (1995).
20
See Cohen (2008), Habermas (1989), Pettit (2001).

Electronic copy available at: https://ssrn.com/abstract=2118772

which a particular (set of) interaction(s) between persons embodies respect or recogni-
tion depends crucially on their implicit relationship seen in light of a conversation mode
(see Darwall 2010 and Pettit 2001), independently of their actual interactions, prior
intentions, etc.
I shall not dwell further on (5). Suffice it to say that standard instances of domi-
nation (the relation between slave and slaveowner, serf and lord, wife and husband in
the patriarchal family, etc.) all involve disrespectful powerovering in the broad sense
I am interested in here.21

3.1
I now want to show that (4) implies (5), or, more precisely, that (d) and (e) jointly
imply (f) and (g). If I am right, then economic exploitation implies domination.
A first step towards this demonstration has already been taken in section 2.1: if (3)
is true, then (d) implies (f). That is, when A instrumentalizes B’s economic vulnerability
he is eo ipso using (or ‘taking advantage of’) his power over B.

3.2
I must now defend the move from (d) and (e) to (g), that is:

(6) If A instrumentalizes B’s economic vulnerability to enrich himself, then

A affects B in a way that is disrespectful to B.

Note, first, that the appropriation of the fruits of another’s labour is a form of self-
enrichment, for it confers on the appropriator a net material gain: A enriches himself
in that minimal sense.22 Exploitation thus consists in someone’s enrichment from the
instrumentalization of another’s vulnerability.23
21
The most advanced students of domination in analytic philosophy are republicans. Republicans define
domination as subjection to the arbitrary power of another (see Pettit 1997, Skinner 1998, Lovett 2010).
I disagree with those republicans that domination is a form of unfreedom, for domination is consistent
with an increase in one’s (valuable or valued) options, and therefore consistent with (increased) freedom
(if this was not the case, moreover, then there would seem to be no interesting distinction between being
dominated and being coerced or forced). But the main reason I do not adopt the republican definition
is that I do not know of any sufficiently robust republican account of arbitrariness. Pettit (1997), for
example, claims that A has arbitrary power over B if and only if A has power over B and A’s (exercise
of that) power fails to track B’s interests. What these interests are, and why failure to track them is a
necessary and sufficient condition for domination, remains largely unsettled.
22
Of course such appropriation need not involve enrichment only for A: B might also gain from the
transaction.
23
One can instrumentalize another’s vulnerability not, strictly, to enrich himself but, more broadly, to
serve his own interests. A’s interests are served both when A instrumentalizes B’s vulnerability to get B to
φ, where φ-ing improves A’s position and where B’s φ-ing leaves A as well-off as he would otherwise have
been (assuming A is sufficiently well off). These two notions of ‘serving one’s interests’ correspond to what
may be called self-enriching and conservative vulnerability-instrumentalization, respectively. Exploitation
by definition rules out conservative vulnerability-instrumentalization.

Electronic copy available at: https://ssrn.com/abstract=2118772

 Now, (6) is not unlike most substantive moral claims, in that it is not directly
amenable to proof. All one can do is propose intuition-eliciting examples and show
why these intuitions withstand the test of counterexample in reflective equilibrium
with other judgements and theories. I claim that paradigmatic instances of exploita-
tion, such as those represented in Zola’s Germinal, Dickens’ Oliver Twist, Lars Von
Trier’s Dogville and David Lynch’s Elephant Man all share this property of vulnerability-
instrumentalization. And I also claim that the reason we are appalled by the content of
these relationships is precisely that they are humiliating, demeaning or degrading (in a
word: disrespectful) to the underdog, whose vulnerabilities are treated as opportunities
for self-enrichment.24 As I shall argue in sections 5 and 6, alternative explanations for
our disapproval of these relationships do not withstand scrutiny at reflective equilib-
rium.
If (3) and (6) are true, then (d) and (e) suffice for (f) and for (g). It follows from
this25 that (4) implies (5), that is, A’s instrumentalization of B’s economic vulnerability
for A’s enrichment is a form of domination.26 This concludes my positive argument,
which I must now summarize.

A economically exploits B.
∴ A instrumentalizes B’s economic vulnerability to appropriate B’s labour.
(from (4))
∴ A instrumentalizes B’s economic vulnerability to enrich himself. (from
the fact that labour-appropriation enriches)
∴ A takes advantage of his power over B in a way that is disrespectful to B.
(from (3), (6))
∴ A dominates B. (from (5))27
24
It does not follow that exploitation is wrongful domination-independently. Our rational disapproval
vis-à-vis exploitative practices is conceptually entangled with our disapproval vis-à-vis a particular form
of power-over, namely domination.
25
Which further implies that they suffice for (f) and (g), since if d & e → f and d & e → g, then d &
e → f & g.
26
Several authors have noticed that exploitation implies power-over. Wolff (1998, p. 111) identifies
‘exploiting someone’s circumstances’ with ‘exploiting the power one has over another’ and notes that ‘the
notion of having exploitable circumstances seems equivalent to that of being vulnerable’. Both are, he
says, necessary for exploitation. Goodin (1987, p. 184) similarly claims that exploitation is a kind of
‘abuse of power’. Neither of these authors claim there is a necessary connection between exploitation
and domination. Indeed, Goodin explicitly locates the wrongmaking feature of exploitation not in the
nature of the form of power-over the actualization of exploitation implies, but rather in the violation
of a duty to protect the vulnerable. But one can protect and exploit, indeed, protect the vulnerable by
exploiting them. Suppose A is after B, and will kill B if he gets him. I can protect B (only) by hiring
him in my factory for a sweatshop contract. In these circumstances I would be protecting B by exploiting
him. This possibility is not mere high-flown fancy: the quintessentially capitalist institution of the labour
market systematically enforces a social distinction between exclusion from, and inclusion into, work. Even
where exclusion does not imply starvation (the ‘kill’ option), the principal means capitalism mobilizes for
protecting the vulnerable is the commodification of their labour power through inclusion in the labour
market (the ‘exploit’ option).
27
Roberto Veneziani points out (in private communication) that if exploitation is the unfair taking of
advantage (à la Wertheimer 2001) and if disrespect is a special case of unfairness, then domination is but

Electronic copy available at: https://ssrn.com/abstract=2118772

4
I have so far argued that there is a conceptual connection between economic exploita-
tion, on the one hand, and domination, on the other:

(7) If A economically exploits B, then A dominates B.

I now face a preliminary objection. The putative objector sets out from an implica-
tion of (7), to the effect that individual capitalists dominate individual workers. But this,
the objector says, is implausible: even if capitalist A exploits his worker B, he does not
eo ipso dominate B.28 So (7) is false. Now consider:

Three Thugs - A man, B, is threatened by a gang of three thugs, A1 , A2 , and

A3 . The thugs are weak, such that none of them could take B down indi-
vidually. But all three could: their being (at least) three is a necessary and
sufficient condition for taking B down. Assume this is common knowledge.
By threatening to take B down if he does not hand over his money, A1 gets
B to give him the money. A1 runs away, never to share the loot with A2 and
A3 .

In Three Thugs it is true of each Ai , i ∈ {1, 2, 3}, that he dominates B. But a necessary
condition for him dominating is that the others be present and, if instructed, take B
down. Thus Ai can dominate B even if Ai cannot individually get B to do what he
would not otherwise have done. It is therefore not incoherent to maintain, by analogy,
that individual capitalists dominate individual workers, even when individual capitalists
cannot dominate individual workers individually. Indeed, the example seems to vindicate
the common-stock socialist idea that the capitalist class dominates the working class,
and it is in virtue of such collective domination that individual domination, and indeed
exploitation, are possible.
Let PaN b be the power possessed by AN , the thugs taken as a coalition, that is, their
collective ability to get B to do things. When thug Ai gets B to hand over his money, he
is taking advantage of that very power, PaN b , and it is in this sense that he dominates B.
In the capitalist-worker case the structure is exactly thug-like: Ai takes advantage, not
(necessarily) of Pai b , but of PaN b , that is, the collective ability of coalition AN (to which
Ai belongs) to get B to produce a surplus destined for appropriation by some capitalist.
(A sufficient condition for PaN b to apply is the presence of some form of collective, or
‘conjunctive’ vulnerability, of the sort discussed in Goodin 1985, p. 136).29 This is how
capitalist Ai dominates B.
a special case of exploitation, and not vice versa. I argue against the first premiss of this argument in
section 6 below. I discuss the second premiss in a postscript to this paper, entitled ‘Why Marxists should
be interested in exploitation’, which is available from me upon request.
28
Indeed, denying this seems to contradict revered Marxian canon, according to which the individual
worker is ‘property, as it were, of the entire bourgeois class’ and not of any individual capitalists (Engels
1969, Marx 2006, p. 20).
29
There seems to be a disanalogy between the typical form of interaction between capitalist and worker
and that between thug and thug-victim. For thug Ai can’t get B to hand over the loot to him individually,

10

Electronic copy available at: https://ssrn.com/abstract=2118772

 Under realistic assumptions, therefore, when Ai economically exploits B, Ai domi-
nates B, and Ai dominates B in virtue of domination of Ai ’s class over B’s class.30 This
is, moreover, why many socialists believe that the abolition of class domination suffices
for the abolition of both class-class exploitation and man-man exploitation.
It is therefore safe to conclude that the preliminary objection to (7) fails. My rebuttal
of this objection shows why the notion of class has an important and subtle role to play
on any domination-based account of exploitation. I now turn to alternative accounts
of exploitation. My purpose is to show that the domination-based account is superior.

5
Perhaps the most influential–and certainly the most analytically sophisticated–contemporary
theory of economic exploitation is due to John Roemer. In a series of innovative arti-
cles, Roemer has argued against traditional Marxist accounts of exploitation and proven
a series of correspondence theorems pertaining to exploitation and asset inequality, ex-
ploitation and class location, and so on. I shall focus on Roemer’s account of capitalist
exploitation.
According to an influential (pre-Roemerian) Marxist account:

(8) A exploits B if and only if A extracts forced, unpaid, surplus labour from
B31

This definition has the exegetical advantage that it accords with much of what Marx
says about worker ‘compulsion’ under capitalism, and suggests a rough-and-ready set of
necessary and sufficient conditions for economic exploitation. It is the forced nature32
of the labour flow between capitalists and workers that makes, or breaks, the case for
(capitalist) exploitation.
But (8) is (exegetically and)33 definitionally inadequate. Here is an argument as
to why forced transfer (of unpaid surplus labour) furnishes no sufficient condition for
exploitation. Societies with welfare states generally provide for the sick and disabled,
but can get him to do so collectively, that is, in the company of Aj , j 6= i. Yet capitalist Ai can do neither:
he cannot get B to work for him, whether collectively or individually. Someone might therefore infer that
Ai can’t possibly be dominating B. The inference would be invalid. For it is true of each thug Ai that he
dominates B even if it is not true of each Ai that he is capable of getting B to hand over the loot to him
(e.g. because the others would prevent such a transfer, or because he can’t handle loot, or because he is
an idiot). There is, therefore, no relevant disanalogy between individual thug and individual capitalist. I
thank an anonymous referee for raising this issue.
30
The set of assumptions for successful class domination must include the following: no candidate dom-
inator can dominate individually, candidate dominators have common interests, and they dispose of mech-
anisms to enforce such interests through the prevention and/or deterrence of free-riding (one such coor-
dination mechanism under capitalism is naturally the state). See Elster (1983, chapter 7) for discussion.
31
For well-known defences of (8) see Holmstrom (1977), Peffer (1990), Reiman (1987).
32
A is forced to φ if and only if A has no reasonable or acceptable alternative to φ-ing.
33
Exploitation as domination for self-enrichment does a better job exegetically, or so I argue in ‘Why
Marxists...’.

11

Electronic copy available at: https://ssrn.com/abstract=2118772

among others. Those welfare beneficiaries receive a net transfer of labour time from
able-bodied taxpayers. The able-bodied are, moreover, forced, because coerced by the
state, to engage in these net transfers. But no one wants to say that the disabled or the
sick exploit the able-bodied.34
What about forced transfer as (part of) a necessary condition for exploitation? Con-
sider a variation on an example due to Roemer (1996):

Two Plots - A and B own different plots of land, A’s more productive than
B’s, and have identical cardinal utility functions of the form u = ax, where
x is the amount of widgets consumed. If they do nothing, then their land
will magically generate x̄ for each, such that they both enjoy a perfectly
decent level of utility ax̄. A offers B work in A’s land, which is much more
productive than A’s when worked on by human hands. If B accepts A’s
offer, he will produce N widgets and consume x̂, where x̂ > x̄ (x̂ is also
sufficiently large to compensate for the disutility of labour, if any). A will
then consume N − x̂, where N − x̂ >> x̂, without working at all. B accepts
the offer.

Roemer argues, plausibly, that this sort of interaction is exploitative. But B is forced
neither by his economic circumstances nor by third parties, to enter into it. Hence force
does not furnish a necessary condition for exploitation.
The scenario just envisaged raises the further, interesting question whether there
can be exploitation in a world with a substantial unconditional basic income (UBI). I
think Two Plots can be amended to show that this is perfectly possible. Consider:

Basic Income - Same ownership and preference structure as above. The main
difference lies in A’s offer to B: A will work for a total of 2 hours on his own
land and produce x̄, to be passed on to B for consumption. B, on the other
hand, will work for 12 hours on A’s land and produce N , of which he will
keep an amount . B’s final consumption is therefore x̂ = x̄ +  (x̂ is also
sufficiently large to compensate for the disutility of labour, if any). A’s final
consumption will be N −  >> x̂. B accepts the offer.

As before, A’s superior cut in the means of production allows him to consume much
more labour than he expends by appropriating B’s surplus product (here N − ). If Two
Plots is exploitative, then this is exploitative. Moreover, B’s position in this example
seems exactly analogous to that of the UBI beneficiary receiving x̄ under conditions of
advanced capitalism (for an argument that UBI and advanced capitalism are compossi-
ble, see Van Parijs 1993, part IV). Hence basic income capitalism can be exploitative.35
34
And Marx certainly does not (see, for example, ch. 1 of Marx (1968)). Note that the truth of ‘A exploits
B’ does not imply that A himself disposes of the means to exploit B (e.g. a monopoly over the legitimate
use of violence).
35
If I am right, then UBI is, at best, only a necessary condition for human emancipation. Opponents
of UBI object that it is not even that. Their argument tends to go roughly as follows: Under UBI, non-

12

Electronic copy available at: https://ssrn.com/abstract=2118772

 In light of all this, Roemer argues that what is wrong with examples like Two Plots
is injustice in the distribution of alienable assets. Thus, for any coalition of agents A
and its complement B, Roemer claims:

(9) A exploits B if and only if

(i) were B to withdraw from the society, endowed with its per capita share of society’s
alienable property (that is, produced and nonproduced goods), and with its own
labor and skills, then B would be better off (in terms of income and leisure) than
it is at the present allocation;

(ii) were A to withdraw under the same conditions, then A would be worse off (in
terms of income and leisure) than it is at present;

(iii) were B to withdraw from society with its own endowments (not its per capita
share), then A would be worse off than at present. (Roemer 1996, p. 40)

In this connection, Roemer (1985, 1996) has argued that exploitation itself is, at
best, a morally secondary phenomenon (if morally significant at all). According to
Roemer, the locus of our normative interest should be the (in)justice of property rela-
tions (see also Cohen 1979).

6
I now want to argue that the Cohen/Roemer account fails to supply a necessary condi-
tion for exploitation, whether or not it furnishes a sufficient condition. Roemer attacks
the notion of exploitation as unequal exchange in ‘Should Marxists be interested in
Exploitation?’. He returns a negative answer to that question:

...for the sake of clarity and consistency, I think exploitation conceived of as

the unequal exchange of labour should be replaced with exploitation con-
ceived of as the distributional consequences of an unjust inequality in the
distribution of productive assets and resources. (Roemer 1996, p. 96)

What constitutes unjust inequality? Roemer affirms a theory of equality of oppor-

tunity that has come to be called ‘luck egalitarianism’: the distribution of x (where x
is some desideratum of justice) between A and B is just if and only if any difference
workers appropriate the fruit of workers’ labour. All such appropriation is exploitative. Therefore UBI
is exploitative (see, for example, Van Donselaar 2008). As it stands, this argument is untenable. For its
major premiss implies, implausibly, that sick and disabled nonworkers who benefit from a UBI (or any
redistribution, for that matter) exploit workers. The premiss can, of course, be amended to apply only to
the able-bodied. (This restriction would raise the question how, if at all, the revised major premiss can
be defended in a principled way, given that its main source of support is some reciprocity requirement.)
Even if such a restricted version of the argument can form the proper basis for a theory of exploitation
(which there is reason to doubt), the positive account defended here can afford to remain agnostic about
its merits. I thank an anonymous referee from bringing this argument to my attention.

13

Electronic copy available at: https://ssrn.com/abstract=2118772

between A’s and B’s enjoyment of x reflects a difference in their choices, deserts, or
faults. Any inequality that violates this principle implies involuntary disadvantage (see
Roemer 1996, pp. 179-198, Cohen 1989):36
(10) A exploits B if and only if the exchange in which they are engaged
occurs against the background of an unjust distribution, i.e. a distribution
involving involuntary disadvantage.37
Assuming individual choices reflect (unmanipulated) preferences, a corollary of (10)
is that:
(11) Any distribution that reflects nothing but differences in (unmanipu-
lated) preferences is just, and therefore not exploitative.
The rest of this section proceeds by arguing that (10) and its corollaries should be
rejected.38

6.1
Consider:
Rescuer - A finds B in a pit. A can get B out at little cost or difficulty. A
offers to get B out, but only if B agrees to pay a million euros, or to sign a
sweatshop contract with A. B signs the contract.
This exchange is paradigmatic of exploitation: if this is not an instance of exploita-
tion, then nothing is. Yet thinking that B’s acceptance of the offer results in A exploiting
B is inconsistent with (10). For A’s offer can be forthcoming even in the absence of in-
voluntary disadvantage, or indeed any sort of distributive injustice: B may have found
himself there through just steps, from a just starting position. B may, for example, have
chosen not to purchase cheap insurance against falling in pits, or against being rescued
afterwards. Alternatively, the pit may be the bottom of a huge vanilla ice-cream cone,
and B may have ended up there by intentionally licking all of the vanilla. It does not
follow, and it is false, that the offer is not exploitative: A can rescue B without asking
for anything, but instead uses B’s vulnerability in order to improve her own lot.39 It fol-
lows that asset injustice furnishes no necessary condition for exploitation, and therefore
no proper basis for a charge of exploitation.40
36
Roemer does not formulate his own view in terms of involuntary disadvantage (as opposed to ‘equal
shares’) but this terminological difference is insubstantial.
37
(9) follows from (10) if any lack in one’s per capita share constitutes involuntary disadvantage.
38
The next subsection draws upon Vrousalis (2013).
39
Someone might object that A’s offer is exploitative because coercive. But, as I argued in section 5,
force and/or coercion are neither necessary nor sufficient for exploitation. In my view the wrongmaking
features of Rescuer must be sought neither in coercion, nor in force, nor in asset injustice.
40
More generally, this triad is inconsistent: (i) exploitation is unfair advantage-taking, (ii) unfairness is
responsibility-constrained equality, (iii) exploitation can arise from any material inequality. (ii) and (iii)
are very compelling. I claim that (i) must be rejected. On my view, exploitation is not tantamount to
unfair advantage-taking, for there can be fair exploitings.

14

Electronic copy available at: https://ssrn.com/abstract=2118772

6.2
By way of illustration of this rather general complaint against (10), I must now rehearse
the possibility of a ‘cleanly generated capitalism’, a form of capitalism that does not
arise from ‘primitive accumulation’, through massacre, plunder, forced extraction, or,
quite generally, through transgression of some norm of distributive justice. Rather, it
arises from ‘clean’ social interactions: a labourer, or a class of labourers, manages to
accumulate significant quantities of capital through toil and savings, thereby turning
himself into a capitalist.41
Consider a limiting case of cleanly generated capitalism:

Ant and Grasshopper - Grasshopper spends the summer months singing, whereas
Ant spends all his time working. When the winter comes, Grasshopper needs
shelter, which he presently lacks. Ant has three options: he can do noth-
ing to help Grasshopper, he can offer him shelter which costs Ant nothing,
or he can offer him costless shelter on condition that he signs a sweatshop
contract. The possibilities, in the relevant metric, are:
(Ant, Grasshopper)
(i): Do nothing: (10, 1)
(ii): Sweatshop: (12, 2)
(iii): Shelter: (10, 3)

Now, it is plausible to think that Ant has an obligation to help Grasshopper. But
one need not have a view on that to believe that (ii) is morally worse than (iii). That
is, if Ant decides to help (decides to do something other than (i)) then he must not
opt for (ii), in part because doing so constitutes exploitation. I take the widely-shared
intuition expressed in this consequent as evidence that the vulnerability-based defini-
tion of exploitation is plausible. But now note that, if (ii) involves exploitation, then
asset injustice furnishes no necessary condition for it. For, according to (10), all of (i),
(ii), (iii) are equally acceptable at the bar of asset justice.42 Roemer, and other Marxist
advocates of (10), have nothing to oppose to such a cleanly-generated capitalism on ex-
ploitation grounds, and this is an embarrassment to their claims to holding a socialist,
or a socialism-friendly, theory of equality. This result generalizes, I believe, to any the-
ory of exploitation that posits some defensible account of asset injustice as a necessary
condition for exploitation.43
41
Marx himself envisages a similar possibility, arising from the labourer’s ‘own labour and that of his
forefathers’ in Marx (1976, p. 728). On Marx’s account he still exploits the worker.
42
(ii) is, after all, an upshot of Grasshopper’s own choices or faults. One can construct examples where
the demands of distributive justice are not only unnecessary for, but also contradict (what are, intuitively)
claims of exploitation. See Vrousalis 2013.
43
Take another prominent example: leximin. Leximin does indeed dub (iii) as the most just distribution
of widgets, and therefore seems to give the intuitive answer. But the result is misleading. Assume the
shelter outcome turns out to provide no substantial improvement in Grasshopper’s condition, as follows:
(Ant, Grasshopper)

15

Electronic copy available at: https://ssrn.com/abstract=2118772

 The reason why exploitation survives in the absence of distributive injustice, I sub-
mit, is that exploitation constitutes procedural injury to status, and status-injury is not
reducible to distributive injury (see White 2004 for some discussion of the former claim,
and Anderson 1999 for a defence of the latter claim).44
In this section I have argued that judging (ii) as exploitative supports (4) and un-
dermines (10). If I am right, and if (4) has the conceptual connection with (5) that
I have claimed, then exploitation is best viewed as a form of (injustice-independent)
domination. On the view defended here, domination is the distinct vice that sits in
the (hitherto largely unexplored) conceptual realm between coercion and unfairness.
But now I face another objection, articulated by Roemer himself, to the effect that
exploitation is broader in scope than domination and cannot, therefore, be a form of
domination. I turn to this objection presently.

7
In his early work, Roemer (1982, p. 195) affirms a variation on (9), in which (iii), the
‘withdrawal with one’s own share’ condition, is replaced by ‘A being in a relation of
dominance to B.’ The reason some such condition is conceptually indispensable is that
(i) and (ii) do not jointly suffice for exploitation. If they did, then two individuals who
never interact with each other, but control unequal amounts of assets, could exploit
one another.45 Roemer’s dominance condition fills this gap, by making the exploiter’s
well-being dependent on (the nature of the) interaction with the exploitee. In his later
work Roemer offers some hints as to why he came to reject this earlier emphasis on
domination:

It is necessary to distinguish two types of domination by capitalists over

workers, domination in the maintenance and enforcement of private prop-
erty in the means of production, and domination at the point of production
(the hierarchical and autocratic structure of work). The line between the
two cannot be sharply drawn, but let us superscript the two types domination1
and domination2 , respectively... each of domination1 and domination2 im-
plies exploitation, but not conversely. Hence if our interest is in dom-
ination, there is no reason to invoke exploitation theory, for the direc-
tion of entailment runs the wrong way... In certain situations, exploita-
tion requires domination1 , but since we cannot know these cases by an-
(i): Do nothing: (10, 1)
(ii): Sweatshop: (12, 2)
(iii’): Shelter: (10, 2)
In this case, leximin dubs (ii) as the most just distribution, and therefore produces a counterintuitive result
(and so does maximin, which recommends choosing between (ii) and (iii’) on a coin toss). Exploitation
and distributive injustice again part ways.
44
Status equality (causally) requires some measure of equality in distribution, and may sometimes man-
date compensation for involuntary, or brute luck, disadvantage. But then again it may not.
45
For an early critique of the Roemerian definition, see Bertram (1988).

16

Electronic copy available at: https://ssrn.com/abstract=2118772

 alyzing the exploitation accounts alone, there is no reason to invoke ex-
ploitation if, indeed, our interest in exploitation is only as a barometer of
domination1 . Furthermore, our interest in domination1 is essentially an in-
terest in the inequality of ownership of the means of production, for the
purpose of domination1 is to enforce that ownership pattern. I maintain if
it is domination1 one claims an interest in, it is really inequality (however
defined) in the ownership of the means of production which is the issue.
(Roemer 1996, p. 73)

Given the significance of this objection to the overall integrity of my thesis, I pro-
pose to study it at some length. I proceed as follows: first, I reconstruct Roemer’s
account of domination1 and domination2 . I then rebut his anti-domination conclusion
on the grounds that the forms of domination he discusses do not exhaust the sphere of
domination tout court. Exploitation is simply a ‘third form’ of domination, equivalent
to neither domination1 , nor domination2 . In the course of this rebuttal, several points
of agreement with Roemer will surface.

7.1
Here’s a reconstruction of Roemer’s argument against the view that exploitation implies
domination2 :

(12) domination2 occurs only within the point of production, P.

(13) Exploitation does not occur only within P.
∴ (14) Some instances of exploitation are not instances of domination2 .
∴ (15) Our interest in domination2 cannot justify an interest in exploitation.
(Roemer 1996, p. 77)

Roemer defines domination2 in terms of the ‘hierarchical and autocratic structure of

work’: (12) follows from this definition. (13) is trickier to defend. Unlike most Marxists
before him, Roemer claims that A can exploit B even in the absence of labour or credit
markets (see Roemer 1982). Here’s a brilliant illustration of that claim:

Friday and Robinson - There are two producers, Friday and Robinson. Fri-
day is poor in capital, and Robinson is rich. If they do not trade, Robinson
will work 8 hours and Friday will work 16 so that each can satisfy his or
her needs (assuming they’re both rational, the fact that they decide to trade
shows that they both benefit from trading). They only trade in final goods,
and there are no labour or credit markets. In equilibrium under free trade,
Friday works 12 hours, Robinson works 4, and both attain subsistence. (Roe-
mer 1996, pp. 52-53)

Roemer claims there is exploitation here. Having worked for 4 hours, Robinson
can relax for the rest of the day, while Friday toils to produce what Robinson would
otherwise have produced only with an extra 4 hours of work:

17

Electronic copy available at: https://ssrn.com/abstract=2118772

 Robinson benefits from Friday’s presence, and is able to use his wealth as
leverage, through the market, to get Friday to work for him, which Friday
would not have to do if he had access to his per capita share of the produced
capital. (Roemer 1996 [1985], p. 52)

I agree with Roemer’s conclusion, and I also agree that it vindicates (13). Thus ex-
ploitation does not presuppose domination2 ((14) is true) and an interest in domination2
cannot justify an interest in exploitation ((15) is true).

7.2
I now proceed to Roemer’s more general argument against the view that exploitation
and domination are related in the way advocated in (7):

(16) domination1 arises from unjust property relations (i.e. asset injustice).
(17) Exploitation is only a ‘barometer’ for domination1 .
∴ (18) We should be interested in domination1 , not exploitation.

(16) is true by definition. (17) says that exploitation provides an indication or

prediction of domination1 , just as barometers indicate or predict the weather (Roemer
1996, p. 73). And since we are only interested in predictions because we are interested
in the weather, it follows that exploitation is of no intrinsic interest, which is what (18)
says. Adding the premiss:

(19) domination1 and domination2 are the only forms of domination tout
court.

It follows that:

(20) An interest in domination tout court cannot justify an interest in ex-

ploitation. (from (15), (18) and (19))

Whatever one thinks about the inference to (18), the inference to (20) seals off
any appeal to domination as explanans of our interest in exploitation and concludes
Roemer’s attack on that front.

7.3
My response to Roemer consists in showing that his account of domination is too nar-
row, that is, consists in refuting (19). For it is plausible, I think, that there exist in-
stances of domination (i) that do not arise from asset injustice (i.e. are not instances
of domination1 ), and (ii) that do not take place at the point of production (i.e. are not
instances of domination2 ).
I begin by defending claim (ii). Roemer’s own Friday and Robinson example is pur-
ported to show that exploitation can take place outside the point of production. But

18

Electronic copy available at: https://ssrn.com/abstract=2118772

why is the same not true of domination? Roemer says that Robinson ‘is able to use his
wealth as leverage to get Friday to work for him,’ adding that this is something ‘Friday
would not have to do if he had access to his per capita share of the produced capital.’
(emphases mine) It follows that Robinson can instrumentalize Friday’s vulnerability to
get Friday to do something he would not otherwise have done (or not have done for
the same reasons), in order to obtain a material gain. As I argued in section 3, there is
nothing counterintuitive about saying that such instrumentalization is tantamount to
domination.
One reason why Roemer wants to resist the domination-based description seems to
be that he takes domination to be coextensive with the absence of perfect competition:

...where markets for particular assets or commodities are thin... one agent
has power over another which he would not have in a fully developed, per-
fectly competitive market economy (Roemer 1996, p. 74)

Now, given domination implies power over, it follows by modus tollens from the pas-
sage just quoted that ‘fully developed’ perfect competition excludes the possibility of
domination. Yet Roemer (and indeed Marx) thinks that exploitation supervenes even in
a perfectly competitive capitalist economy (which I take to be a system of generalized
commodity production in which there are no externalities and in which all producers
face infinitely elastic demand schedules, i.e. have no market power). Therefore not
all instances of exploitation are instances of domination. In response to this line of
argument, it bears repeating that even under ‘fully developed’ perfect competition a
class of people (owners of the means of production) can ‘get’ others (nonowners) to do
things those others wouldn’t otherwise have done, by instrumentalizing their vulnera-
bility (i.e. their nonownership of such means), in order to make money. All these are
instances of power over and, indeed, instances of domination (as betrayed by Roemer’s
own language). The first premiss of Roemer’s argument, that domination is coextensive
only with the absence of perfect competition, is therefore false.46
I now turn to (i), the view that domination always arises from asset injustice. Con-
sider:

Slavery contract – A and B are identical twins starting from exactly equal
bargaining positions. Each freely agrees to the following contract: on a
50/50 coin toss, the one who gets heads enslaves himself for life to the one
who gets tails. B gets heads. (see Cohen 1995, p. 47)

In this example, domination arises from conditions of full asset justice through the
voluntary, unmanipulated choices of free agents. The example shows why complaints
of (Roemerian) asset injustice are unnecessary for a charge of domination (and therefore
why (i) is false).
I tentatively conclude that it is not implausible to think that there are forms of
domination that correspond neither to domination1 (asset injustice), nor to domination2
46
I am here also opposed to Lovett (2010, p. 79).

19

Electronic copy available at: https://ssrn.com/abstract=2118772

(point-of-production autocracy). If such a ‘third form’ of domination exists, then (20)
fails to follow. The thesis that an interest in exploitation can be justified by an interest
in domination has not been refuted by Roemer.

8
Let us take stock. I began by defining economic exploitation as the instrumentaliza-
tion of someone’s economic vulnerability for the appropriation of (the fruits of) his
or her labour. Under capitalism, economic vulnerability is determined by relations of
effective control over labour power and the means of production. Such relations per-
mit capitalists (owners of the means of production) to instrumentalize the vulnerability
of workers (nonowners of such means) to appropriate the surplus the latter produce.
I then argued that exploitation implies domination of workers by capitalists, and de-
fended the view that domination is distinct from, and not reducible to, distributive
injustice. I also claimed that class domination cannot be abolished without abolish-
ing capitalist relations of production. Such abolition is not, however, normally suffi-
cient for the removal of class domination. Differences in talents and knowledge, for
instance, can generate domination, thus paving the way for ‘post-capitalist’ forms of
economic exploitation.47 Under technocracies, economic vulnerability is determined
by relations of effective control over knowledge and technical expertise. Such control
permits technocrats (owners of knowledge assets)48 to instrumentalize the vulnerabil-
ity of non-technocrats (nonowners of such assets) to appropriate the surplus the latter
produce.49
This essay has largely focused on distribution and class domination. For all I have
shown, however, it is still possible that class division under capitalism is but the ex-
pression of a deeper form of social domination immanent to this social system: that
of ‘dead’ over ‘living’ labour, or of capital over human beings. If there is a nontriv-
ial, ontologically nonsuspicious sense in which this is right, then capitalist domination
can conceivably survive the removal of capitalists (a possibility crisply expressed in the
much-vaunted oxymoron of ‘market socialism’). A satisfactory critique of capitalism
would, I think, need to study the antecedent and the consequent of this conditional by
drawing upon the rich traditions of both critical theory and analytic political philoso-
phy.50
47
In Roemer’s (1982) schema, means-of-production equality abolishes ‘capitalist’ exploitation, but not
‘socialist’ exploitation, that is, unjust flow generated by differences in status, knowledge or talent. Non-
exploitative societies transcend both capitalist and socialist forms of exploitation (see also Wright 1985).
48
Wright (1985) uses the (broader) term organization assets.
49
I discuss technocracy, and some of the implications of the account defended here for socialism, in
‘Why Marxists...’.
50
For rudiments towards such a critique in the former tradition, see Postone 1993.

20

Electronic copy available at: https://ssrn.com/abstract=2118772

References
Anderson, E. 1995. Value in Ethics and Economics. Harvard University Press.
Anderson, E. 1999. What’s the point of equality? Ethics 109: 287-337.
Arneson, R. 1981. What’s Wrong with Exploitation? Ethics 91: 202-227.
Bertram, C. 1988. A Critique of John Roemer’s Central Theory of Exploitation. Political
Studies 36: 123-130.
Cohen, G. A. 1978. Karl Marx’s Theory of History: A Defence. Princeton: Princeton
University Press.
Cohen, G. A. 1979. The Labour Theory of Value and the Concept of Exploitation. Phi-
losophy and Public Affairs 8: 338-360.
Cohen, G. A. 1983. Review of Wood. Mind XCII: 440-445.
Cohen, G. A. 1989. On the Currency of Egalitarian Justice. Ethics 99: 906-944.
Cohen, G. A. 1995. Self-Ownership, Freedom, and Equality. Cambridge: Cambridge
University Press.
Darwall, S. 2010. The Second-Person Standpoint. Cambridge, Mass: Harvard Univer-
sity Press.
Elster, J. 1983. Making Sense of Marx. Cambridge: Cambridge University Press.
Engels, F. 1969. The Principles of Communism, Marx-Engels Selected Works, Vol I,
Progress Publishers, Moscow.
Feinberg, J. 1988. Harmless Wrongdoing. Oxford: Oxford University Press.
Geras, N. 1985. The Controversy on Marx and Justice, New Left Review 150: 47-85.
Geras, N. 1992. Bringing Marx to Justice, New Left Review. 195: 37-69.
Goodin, R. 1986. Protecting the Vulnerable. University of Chicago Press.
Goodin, R. 1987. Exploiting a Person and Exploiting a Situation. In Reeve, A. Modern
Theories of Exploitation. Sage Publishers.
Holmstrom, N. 1977. Exploitation. Canadian Journal of Philosophy 7: 353-369.
Julius, A.J. Forthcoming. The Possibility of Exchange. Mimeo.
Lovett, F. 2010 A General Theory of Domination and Justice. Oxford University Press.
Marx, K. 1968. Critique of the Gotha Programme. In Marx and Engels, Selected Works
in One Volume, London.
Marx, K. 1976. Capital. Vol. I Penguin.
Marx, K. 2006. Wage-Labour and Capital. New York: International Publishers.
Morriss, P. 2002. Power. Manchester University Press.
Peffer, R. 1990. Marxism, Morality and Social Justice. Princeton University Press.
Pettit, P. 2001. A Theory of Freedom. Oxford University Press.
Postone. M. 1993. Time, Labor, and Social Domination. Cambridge University Press.
Rawls, J. 1971. A Theory of Justice. Harvard University Press.
Reiman, J. 1987. Exploitation, Force, and the Moral Assessment of Capitalism. Philos-
ophy and Public Affairs 16. pp. 3-41.
Roemer, J. 1982. A General Theory of Exploitation and Class, Cambridge: Cambridge
University Press.

21

Electronic copy available at: https://ssrn.com/abstract=2118772

Roemer, J. 1985. Should Marxists be Interested in Exploitation?, Philosophy and Public
Affairs 14, 30-65.
Roemer, J. 1989. What is Exploitation? A Reply to Jeffrey Reiman, Philosophy and
Public Affairs 18, 90-97.
Roemer, J. 1994. A Future for Socialism. Verso.
Roemer, J. 1996. Egalitarian Perspectives. Cambridge University Press.
Sample, R. 2003. Exploitation: What it is and why it’s wrong. Rowman and Littlefield.
Van Parijs, P. 1993. Marxism Recycled. Cambridge University Press.
Vrousalis, N. 2013. G. A. Cohen on Exploitation. Politics, Philosophy and Economics.
Forthcoming.
Wertheimer, A. 2001. Exploitation. Stanford Encyclopedia of Philosophy.
White, S. 2004. The Citizen’s Stake and Paternalism. Politics and Society 32: 61-78.
Wright, E. O. 1985. Classes. Verso.
Wolff, J. 1998. Marx and Exploitation. Journal of Ethics 3: 105-120.
Wood, A. 1995. Exploitation. Social Philosophy and Policy 12: 136-158.
Wood, A. 2004. Karl Marx. Routledge.

22

Electronic copy available at: https://ssrn.com/abstract=2118772