. 5
( 7)


A stronger (and more plausible) sense of “free” would require that
no coercion of any sort was involved in generating the exchange. But if
so, markets exist only when the parties are in a condition such that they
need not exchange. Autarchy would be the ideal. But obviously whenever
there is division of labor, people are not autarchic; and if one accepts the
idea (shared by Weber and Marx) that wage labor is de¬ned by the fact
that as Weber put it, workers are compelled to sell their labor power to
some employer or another, then obviously coercion is a systematic aspect
of labor markets.14 The ideology is also clear here: neo-classical theory
would have us believe that we are all buyers (“consumers”) and sellers
(“¬rms”) who are free to reject any and all exchanges. This idea, of course,
is fully reinforced by the intuitive sense with which we began.

Voluntary exchange
We had better abandon any notion that markets require free exchanges.
All sorts of constraints and differential resources “ and thus inequalities
of power and freedom “ are perfectly consistent with markets. But there
is an important point perhaps confusedly seen by those who argue for
“free” exchange and build economic models based on that assumption.
Markets do require that exchanges be voluntary in the standard sense that
the exchanges are not legally compelled. The extraction of surplus from
feudal peasants is not a market exchange. Nor, plainly, is slavery.
It may be, indeed, that the idea of free wage labor contributes substan-
tially to the confusion being addressed here. It might be held, for example,
that although workers sell their labor power, this is not voluntary if the
only choice is to starve. He or she is thus in this sense compelled.
It is critical that we do not think of choices as involuntary simply
because all the choices are undesirable. This would drain the notion of
voluntariness of any usefulness. We need here a conception of freedom
14 This is almost never noticed; sometimes, when it is, as by Charles E. Lindbloom (1977),
the point is obscured. He sees that “freedom depends upon the character of the alter-
natives” and thus, “the generalization . . . is that exchange best supports freedom when
every party can choose among offers that do not greatly differ in value from each other or
from no exchange at all” (1977: 49). He then suggests that this condition is met in either
of two circumstances: exchange is limited to small values or “no single act of exchange
is greatly more advantageous to either party than other available exchange opportuni-
ties. In neither circumstance can anyone be coerced, since he can, without great loss to
himself, easily refuse any offer” (1977: 49). This is a manifest non sequitur since persons
with only their labor power to sell can refuse any particular offer, but must accept some
offer. They voluntarily accept the best offer because they are not free to reject all offers.
See below.
Markets as social mechanisms 135

which precludes this collapse. Roughly, one is free insofar as one can do
(be or have) what one wants to do (be or have). So construed, freedom
is very unequally distributed in society, even though everybody has some
freedoms and nobody is absolutely free. Thus, for example, if Sally lacks
the money to buy a ticket to Maui, she cannot go. Harry has the money
but he cannot go because his boss will not give him the time off. Louis
is free to go to Maui because nothing prevents him from going.15 In this
context, one acts voluntarily in taking a terrible job, but if one could
do as one pleased, one would not take the job even if it were the best
of the awful alternatives. It is easy to see also how confounding “free”
with “voluntary” contributes to the ideological association of markets
and “individual freedom.”16
To see how voluntariness ¬gures in de¬ning a market, a comparison
with practices in the former Soviet Union, where there was no labor
market, is useful here. Every Soviet citizen had to work; one could be
deported to Siberia or receive other punishments if one chose not to.
Workers were con¬ned to localities by the internal passport system and
controlled through the use of labor books and personal ¬les that could be
used punitively. Labor was, in this sense both unfree and involuntary. As

15 In a liberal society, people may have equal rights, but be unequally free. Following Gerald
Feinberg (1973), we should think of freedom as a triadic relationship:
A (some agent) is free from C to be able to do (or be, have, not do, etc.) F where A
is the name of some person or group, C is a speci¬cation of constraints, obstacles, lacks
and F is an action, e.g., going to New York for a holiday, a state of being, e.g., being a
lawyer, or a possession, e.g., a house on Maui, a Mercedes, etc.
Constraints can be classi¬ed roughly as follows:
(1) Internal positive constraints such as compulsive desires and neuroses. Thus A can™t
quit smoking. Each time A lights up, it is correct to say that A acted voluntarily. But
if A simply can™t quit, there is something about A™s mental state which keeps A from
quitting. From the point of view of social theory, these are least interesting sorts of
(2) Internal negative constraints, such as ignorance or de¬ciency of skill. For example,
A can™t read. Some of these, of course, are socially remedial.
(3) External positive constraints. Like those in (2) these depend upon humanly con-
structed social arrangements. They include physical coercion and the threat of phys-
ical coercion, constraints always noticed by liberal social theory. These constraints
make acts both unfree and involuntary, a fact which helps promote the collapse of the
two ideas. But often overlooked are constraints which are rooted in social relations,
e.g., class, racism and sexism. One can say, accordingly, that the propertyless worker
is not free not to work, that glass ceilings restrict opportunities (and thus choices)
for females and African-Americans who would otherwise have such choices.
(4) External negative constraints, such as lack of money, tools, friends, etc.
Some of these, of course, are causally related: if you are born into a poor family, likely
you will live in a poor neighborhood, go to poor schools and thus lack knowledge and
skills which would otherwise be available. Thus, ¬nally, you will lack the money, social
ties, etc., which would give you greater freedom. Again, if liberal society assures equal
rights, it surely does not assure equal freedom.
16 See also Dyck, 1981: chapter 7, who offers a range of conceptions of freedom.
136 A Realist Philosophy of Social Science

Ticktin (1992: 84) argues, we lack a useful way to talk about many of the
relations in the “actually existing socialisms.” Following Oleg Bogomolov,
Ticktin writes that labor in the USSR was “semiforced.” That is, workers
did not sell their labor power even if it was alienated in a particular (and
historically novel) form. Less obvious, but as critical, wages were not
really wages since, as Ticktin writes, “money is not money in the USSR.”
Though we need to develop what this means, here we can say that wages
were nearly meaningless, replaced by party and bureaucratic relations,
the networks which really mattered in getting housing, automobiles and
opportunities to travel and to participate in amenities of life.
But command economies also restrict voluntary exchanges in other
important ways. Still involuntary are exchanges made between produc-
ers under instruction from a central planning authority. Again, failure to
follow the plan brought sanctions. Thus, for example, planners decide
for manufacturers what component parts, electric power, etc. they will
get and whom they get them from. As Nove (1989: 242) writes: “Man-
agers are tied to one supplier, cannot exercise choice, and cannot go else-
where.” Moreover, since there is no stock market, there is no opportunity
for individuals or groups to exchange shares in the material resources of
production, which, for some at least, is the key issue regarding markets.
Of course, command economies do not altogether eliminate voluntary
exchanges. They are minimized in areas of the economy where both indi-
vidual consumers and suppliers are excluded. Of course, as consumers,
people engage in voluntary exchanges at food markets, shops and so on,
even if, given contingent circumstances of development, the imperatives
of production for war, the low priority given to consumption, the failures
of the planning system, etc., choices among consumables are severely
More generally, then, it is a key feature of market economies that
exchanges be voluntary. But all choices, including decisions to exchange,
are enabled and constrained by existing social relations and the positions
of individuals in them. Even where most exchanges are voluntary, as in
all capitalist political economies, people engage in voluntary exchanges
under very different circumstances. As argued in chapter 4, to build a
model we need to know who are the actors, what they want, what they
know and what they can do. What they can do includes identifying alter-
natives, constraints and options as these are structured by class, race
and so on. That is, variation in social structure, political and legal sys-
tems and relations, all thoroughly suffused with culture, make for huge
differences in the conditions of voluntary exchange. It is a critical part
of the empirical problem of understanding markets to be clear on just
what these conditions are. It should go without saying, here, that the
Markets as social mechanisms 137

implications for ideological critique and for normative social theory are

Capitalism and market economies
The foregoing de¬nition of a market as a social institution in which people
“freely” exchange commodities continues with this sentence:

The market presupposes a social division of labor and (at least) de facto private
ownership of the means of production.

If we accept this de¬nition, “market socialism” would be a contradic-
tion in terms. If we mean by socialism some form of community owner-
ship of the means of production, we must, accordingly, reject the view
that markets require private ownership of the means of production. It
may be that the so-called market socialisms cannot solve the problems of
market capitalisms or perhaps, depending on the nature of the market,
they can. In any case, the foregoing de¬nition confuses market economies
with capitalism because it fails to recognize that capitalist property rights
are not the only sorts of rights available which will enable exchange.
And as already noted, market economies can take on many forms; the
differences may make all the difference. Compare the USA, Germany,
Italy, Japan, Indonesia and Brazil. At some level of abstraction these are
all market economies (as well as capitalist), but one needs to explain
why, for example, US school teachers with ¬fteen years™ experience earn
an average of $36,219 compared to Switzerland™s $62,052 and among
OECD countries (thirty nations including most of Europe, North Amer-
ica, Japan, South Korea, Australia and New Zealand) only the Czech
Republic, Hungary, Iceland and Norway pay teachers less relative to
national income (New York Times, June 13, 2001). The point is precisely
that these outcomes can be explained only if we acknowledge that the
prevailing market mechanisms in these economies are very different.

Prices and money
What, then of the second objection to the original de¬nition, the idea
that generally, money is the medium of exchange. Here again, there is
a subtle but important point that is probably being obscured. Money is
surely a medium of exchange, but the critical idea as regards most forms
of markets is this: the price system is the mechanism for coordination. Roughly,
everything has a price and buyers and sellers make choices among priced
138 A Realist Philosophy of Social Science

Following Karl Polanyi (1992), we can identify three main forms of
coordination / integration in the exchange of goods and services: (1) reci-
procity, (2) redistribution and (3) a price system. Each of these presup-
poses different sorts of mechanisms of coordination and control. For
example, where reciprocity prevails, there are (if Polanyi is correct), sym-
metrical groupings, for example, a kinship system as a fundamental order-
ing principle. Similarly, redistributive forms require allocative centers of
authority “ from the systems of ancient Egypt to contemporary com-
mand economies. Price systems differ from these in that everything has
an exchange value and prices provide information regarding allocative
and distributive decisions. Concrete societies may involve some mixing
of these three (ideal-typical) forms of coordination / integration, although
as above, generally, one will tend to dominate.
The de¬nition with which we began misses the main point. It is not
the use of money as such which characterizes markets. As Polanyi argues,
money has three critical uses: as payment, as standard, and as measure
of exchange. Money can function as payment and standard where the
market is not the main or essential mode of coordination / integration.
Thus, “bride price” or ¬nes are payments in money. The “standard” or
accounting use of money is essential (Polanyi argues) for redistributive
systems of organization. Here a “price” is ¬xed, usually once and for all,
for speci¬c purposes, including, for example, managing the staples of the
community. Equivalences are established, but they do not function as
in exchange systems and are not, strictly, exchange values. Thus, “they
designate the quantitative relationship between goods of different kinds
that are acceptable in payment of taxes, rents, dues, ¬nes, or that denote
quali¬cations for a civic status dependent on a property census” (Polanyi,
1992: 49). Also, the equivalency can set the ratio at which wages or ratios
in kind can be claimed, at the bene¬ciary™s choosing. Similarly, “under
reciprocative forms of integration . . . equivalences determine the amount
that is ˜adequate™ in relation to the symmetrically placed party” (1992:
49). As Polanyi emphasizes, “clearly the behavioral context is different
from either exchange or redistribution” (1992: 49).
The capacity of money to represent relative values is, of course, essen-
tial to markets, for it allows for extended indirect exchanging. Money
must come to represent a value which allows for commensuration of very
different sorts of objects of exchange “ including concrete labor.17 This
is “commodi¬cation.” In capitalism, we have the universalization of the
17 Because money makes possible generalized exchange, for Marx, it is a precondition for
alienation. Brie¬‚y, from the point of view of the producers, “the relations connecting
the labor of one individual with that of the rest appear [i.e. manifest themselves], not as
direct social relations between individuals at work, but as what they really are, material
Markets as social mechanisms 139

commodity form. With extended exchange, then, prices are critical in
determining what choices will be made by capitalists and by workers as
job-seekers and consumers, and hence, in determining the overall shape
and direction of the economy. But to anticipate: it is critical to see here
that we need a mechanism if we are to understand how prices get ¬xed
and that neo-classical theory is not the only possibility.
Reference again to the former USSR is useful here. When Ticktin
(1992) writes that “money is not money in the USSR,” the force of this
is precisely that rubles make little difference. First, it is not a store of
value or representation of exchange value. In addition to what we have
already said regarding real distributive processes in the (former) USSR,
workers in the same sector tend to have similar real incomes, irrespective
of task, skills or attitudes toward work. Planners do not think of prices
as opportunity costs or signals of relative scarcities and, as a result, pro-
duction targets were based on physical indicators. Whatever real money
there is in the system is, of course, foreign currency precisely because it
does represent exchange value. To summarize: command economies, in
contrast to market economies, restrict voluntary exchanges in many ways;
labor is “semi-forced” and planners determine authoritatively a host of
economic decisions. But in neither are choices generally “free.” Simi-
larly, in command economies, in contrast to market economies, prices
are not used for coordination since in these economies, there is no “real”
Markets provide a coordinating mechanism for economic activity. That
is, in any market system, price will be a dominant consideration for most
participants. It is hardly an earthshaking observation that a potential con-
sumer will not buy if he deems that the price is too high, or that a potential
investor will not borrow if the interest rate is too high.18 Changes in rela-
tive prices also provide information. If the price goes up, participants

relations between persons and social relations between things” (Marx, 1970: vol. I,
part I, Section 4). It is interesting to note also that in chapter 5 of his Two Treatises, Locke
argues that “consent” to the use of money (prior to the contract which constitutes civil
society) overcame God™s limits on appropriation from nature, and had, as its unintended
consequence, class inequality in exactly Marx™s sense.
18 Of course, there are buyers for whom “money is no object”: “If you have to ask the price,
you cannot afford it.” And, worse, there are needs so great that any price will be paid,
for example, medication or life-saving surgery. Since (for most people at least) price is
an obstacle to consumption, it represents an “opportunity” cost: the higher the price
the greater its opportunity cost in terms of other goods that cannot be purchased. Thus,
also, a price reduction will ordinarily spur sales. But as with many other ideas in the
neo-classical toolbox, this was well known before Adam Smith (Oxenfeld, 1963: 72).
See the discussion of elasticity, appendix D.
More generally, we need to be on guard in believing that some theory is true because
it catches a small piece of what was available quite independently of the theory.
140 A Realist Philosophy of Social Science

may consider economizing. Ex post, prices reveal the pro¬tability or
unpro¬tability of economic decisions. Good judgements are rewarded
and poor judgements punished (Boettke, 1997: 26).

Neo-classical price theory
So far, so good. Any market model will acknowledge the foregoing. But
this says nothing about how prices are determined. It is just here where
the troubles begin. On the neo-classical model, demand and supply for
the market can be represented by curves which relate price to quantity.
On the demand side, the higher the price, the less that is demanded; on
the supply side, the higher the price the greater the amount of the good
to be supplied. The curves are merely (sic) the aggregation of the curves
which represent the preferences of the individual consumers or suppliers.
There is nothing empirical about this, although econometric statistical
study does seek to elucidate them. Rather, the curves get constructed
from the premises of the theory, for example, that persons have known
preferences, are rational maximizers, etc. The intersection of the curves,
the equilibrium price, is, ceteris paribus, the market price and, as such, the
social optimum (see appendix D). This is, of course, the utopia of “free
market” fundamentalists “ utopia because not only does it not exist, but
because it almost certainly could not exist.
Indeed, the falsity of the premises does not stand in the way of an even
greater bene¬t to the capitalist market as comprehended by neo-classical
theory. On the theory, market coordination results in ef¬ciency: “A dis-
tribution of goods or a scheme of production is inef¬cient when there are
ways of doing still better for some individuals without doing any worse for
others” (Rawls, 1971: 67). That is, not only is there coordination of all the
many actions of the parties, but the coordination allocates and distributes
so that given the resources, there is no better way to use them. For exam-
ple, at equilibrium, workers are getting paid just what they “deserve” and
employers are ef¬ciently satisfying the wants of consumers.
It is easy to see the attractiveness of this theory of markets. Unfortu-
nately, it is easy to show (appendix D) that even if the conditions set
out by the model could be satis¬ed, a market result need not be ef¬cient
even in the restricted terms of the theory. Markets are important coordi-
nating mechanisms, and indeed, as Hayek has insisted, real markets do
better in rational allocation than planned economies not because prices
do what neo-classical theory says they do, namely, guarantee “ef¬ciency,”
but because by means of decentralization only a fraction of information
about production possibilities and demand needs to be processed at any
one time, in any one place (Hayek, 1978; Elson, 1988; Stiglitz, 2002).
Markets as social mechanisms 141

Whatever stability and ef¬ciency there is in capitalism depends critically
on at least this much information being available.
This leaves us with this de¬nition:
A market is a social institution in which people voluntarily exchange commodities
(goods, resources, services) and coordination of those exchanges is accomplished
via a system of prices.
This de¬nition is admittedly abstract, but that is not a disadvantage since
we must allow for different kinds of markets. All markets require prices
and thus money, and that exchanges be voluntary, but, as noted, we are
not here committed to any particular theory about how prices get formed, to the
assumption, for example, that prices are equal to the marginal costs of
production, nor are we committed to any particular assumptions about
the participants in the market. There are many ways in which prices
get established. Determining how in any concrete historical market they
do so is a critical empirical problem since, indeed, it is precisely these
conditions which enable and constrain voluntary decisions on the part of
One or two general considerations may here be noticed. First, we need
to recognize that in all market societies, the legacy of past practices,
including non-market practices, will have direct effects on exchange val-
ues. History is here critical. This is emphasized by Polanyi (1992: 50)
who notes that price systems
may contain layers of equivalences that historically originated under different
forms of integration . . . Max Weber remarked that for lack of a costing basis
Western capitalism would not have been possible but for the medieval network of
statuated and regulated prices, customary rents, etc., a legacy of gild and manor.
Thus price systems may have an institutional history of their own in terms of the
types of equivalencies that entered into their making.
Indeed, one need not go back to the Middle Ages to see the vital impor-
tance of historical and non-economic factors in the pricing process. This
is con¬rmed by Hicks (1989) who writes, “economic forces do affect
wages, but only when they are strong enough to overcome . . . social
forces.” They are highly in¬‚uenced by “custom . . . or by any other prin-
ciple which affects the parties to the wage-bargain think to be just and
Second, we can follow neo-classical theory and distinguish perfect
competition from imperfect competition. This includes monopoly, the
19 The text is quoted from Hamouda (1993: 119). Sir John R. Hicks, the ¬rst Briton to
receive a Nobel Prize (1972), moved dramatically from the contributions which won
him the prize. He offered that his famous Value and Capital (1939) “was the work of a
˜neo-classical™ economist now deceased.”
142 A Realist Philosophy of Social Science

limiting case of imperfect competition, and oligopoly, where several large
¬rms dominate the market, there are barriers to entry and the suppli-
ers produce relatively similar products. In imperfect competition, prices are
¬xed independently by suppliers: suppliers are price-makers, not price takers
(appendix D). Nor, contrary to the assumptions of mainstream theory,
in the real world, is this the “special case” (Galbraith, 1968; Baran and
Sweezy, 1968).
Conditions of imperfect competition are still conditions of competi-
tion, but as mainstream theory acknowledges, price makers recognize
that price competition would likely lead to a mutually destructive price
war (Baran and Sweezy, 1968: 58). Many solutions are available to pre-
vent this, from outright collusion, sometimes ignored and sometimes
legitimated, to what is called “price leadership,” where a dominating
¬rm sets the price and the rest follow. Since price competition is seen to
be destructive, it gives way to other forms of competition: variation of
the products™ appearance and packaging, planned obsolescence, contin-
uous model changes, brand names, technological innovations and credit
schemes. Indeed, “in an economic system in which competition is ¬erce
and relentless and in which the fewness of the rivals rules out price cutting,
advertising becomes to an ever increasing extent the principal weapon of
the competitive struggle” (Baran and Sweezy, 1968: 115“116). It has
long been recognized that advertising is not merely “informational” but
is essential to the market process (Chamberlin, 1962). Indeed, if market
capitalism is to be reproduced, new needs must constantly be created
(Baran and Sweezy, 1968; Galbraith, 1968). As Schor put the matter,
“consumerism is not an ahistorical trait of human nature, but a speci¬c
product of capitalism” (Schor, 1992: 117).

The labor market: an example
The labor market offers an excellent example of a market where (1)
there are “crowds” of people seeking employment and there are employ-
ers (though not perhaps “crowds”) who are seeking workers. There is,
accordingly, competition, and supply and demand are critical to deter-
mining wages, and (2), as above, wages and salaries are real money: they
are the primary means allowing for voluntary choices by employers as
regards what wages attach to what jobs and what jobs potential employees
will accept. That is, everything, including labor power, is a commodity “
and thus has exchange value. But if supply and demand are pertinent to
determining wages, the mechanism (or mechanisms) runs well beyond
the one offered by neo-classical theory.
Markets as social mechanisms 143

We can distinguish ranking, how jobs, and wages and salaries are cre-
ated, transformed and destroyed, and sorting, the process whereby indi-
viduals get matched to jobs. As Granovetter and Tilly (1988), who I fol-
low here, insist, ranking and sorting go on simultaneously and we must
resist the temptation to reify skills, jobs and occupations as some abstract
market process.20
For neo-classical theory, sorting is based on competition for available
positions by workers who have different skills and competences. As the
mythology goes, a worker™s competences determine the job that he will get
based on his marginal productivity to the ¬rm. Presumably, employees
have information on all jobs available, employers know exactly what they
expect of potential employees and are able to assess the competences of
the pool of potential employees. As rational, they hire the best person
for the job. Ranking depends upon the imperative of pro¬t maximizing
with ¬rms paying wages equivalent to their marginal products.21 Wages
and salaries are unequal because what people earn is commensurate with
what they contribute.
Unfortunately, employees are often ignorant of job possibilities and,
even if known, they are often out of reach; employers may have only vague
ideas of what skills are actually needed; and they are very often unable to
assess the competences of potential employees. Employers are also often
“irrational” regarding whom they hire “ allowing their prejudices, or com-
mitments to friends, etc. to get in the way, but perhaps most importantly,
the idea of a concrete marginal product on which wages are based is a
mathematical ¬ction. Where the “product” is a cooperative product, as
is the case in almost all real-world production, it is dif¬cult, if not impos-
sible, to assess the relative contributions of the cooperators. Indeed, this
simple (and marvelously fair and ef¬cient!) mechanism barely speaks to
The reality of labor markets is extremely complicated and concretely
speci¬c. As Granovetter and Tilly show, talk of “markets” collapses a very

20 As they emphasize, “skill” compounds personal capacities and substitutability: the ease
and expense of replacing the worker. Skill, like productivity, is very dif¬cult to measure,
despite mythology to the contrary. Skill involves tacit knowledge and is not well de¬ned
(contrary to human capital theory). Athletes are the exception, not the rule. Similarly,
jobs (and occupations) are continually being socially constructed (and reconstructed).
As Granovetter and Tilly summarized matters: “What determines outcomes . . . are such
matters as the resources, bargaining power, socialization, cultural and social structural
patterns of negotiating groups, and the state of labor and product markets” (1998: 209).
21 Marginal decision-making considers whether the bene¬ts of “an extra amount” of some-
thing is worth the extra cost. Thus, the marginal product of labor is the extra output of
an additional “unit” of labor. See appendix D.
144 A Realist Philosophy of Social Science

complicated struggle by a host of parties “ capitalists, workers, house-
holds, states and organizations, for example, the American Medical Asso-
ciation, trade and labor unions “ into a misleading abstraction. These
actors have very different capacities, a function of their structural posi-
tions and relations in society. To take some obvious examples, the stun-
ning differences in the range of income inequality in the USA and Japan,
including inequalities in executive compensation, the critical importance
of racism and sexism in the sorting processes, the ¬‚exibility in de¬ning
and rede¬ning occupations, and thus requirements and wages, need to
be explained in terms of the ways that parties have employed resources
which are the product of historically developed structures and relations.
All this is perhaps familiar enough. (Yet, if so, one may be rightly pressed
to explain the grip of neo-classical assumptions on our thinking about
The labor market is indeed a market; prices provide information on
choices, but neither jobs (nor wages) are “a function of” markets as these
have been comprehended by neo-classical theory. If one wants to explain
outcomes in labor markets, one needs to construct a model in which the
beliefs, knowledge, motivations and capacities of typical people looking
for jobs and of typical people hiring workers are identi¬ed. One needs
also to identify the constraints imposed by history, gender and race rela-
tions, credentialing bodies, unions, etc. More generally, the outcomes in
labor markets are the product of a number of interconnected social mech-
anisms, including pertinently, the political system and how it functions,
the mechanisms which explain racism and sexism, and the mechanisms
which give credentialing bodies and unions capacities to in¬‚uence out-
But there is a further point to be emphasized. The neo-classical model
is timeless, but in real markets, time is a critical factor.

Market as process
The importance of time in market processes was recognized by Hayek,
von Mises, Lachman and others. In its most interesting (and radical)
form it rejects the idea that there is or could be general equilibrium. As
Lachman (1984: 304) writes:
The notion of equilibrium which makes very good sense when con¬ned to indi-
vidual agents, like a household or ¬rm, is less easily applied to the description
of human interaction. It still has uses when applied to a very simple type of
market, such as Marshall™s corn market. But “equilibrium of the industry” is a

22 See also, for example, Tilly and Tilly, 1997.
Markets as social mechanisms 145

dif¬cult concept to handle. Equilibrium of the “economic system” is a notion
remote from reality, though Walras and Pareto show its logical consistency. Equi-
librium of an economic system in motion, “equilibrium growth” borders on

The market is not (as per general equilibrium theory) an “end-state”
tending, ceteris paribus, toward equilibrium, nor is it a closed system,
analyzable ahistorically. The main point is that events occur in real time
and what precedes alters conditions in ways which have unpredictable
consequences on what follows. General equilibrium theory has time as a
“variable” in very nearly the same sense that time is a variable in celes-
tial mechanics.23 Nor is it an “open system on which external change
impinges in the form of ˜random shocks™ each of which the system,
possibly with variable time lags, contrives to ˜absorb™.” As Lachman
writes: “The existence of human action consciously designed to pro-
duce certain effects, prompted by expectations which may, and often
do, fail, makes it impossible to look at the market process in this way”
(1984: 305).
The point is that market outcomes are the unintended product of the
conscious action of different actors each acting with very different mat-
erials at hand to bring about their economic goals. This is, of course,
another way of stating what we said about labor markets. For Lachman,
what propels the market as a continuous process is that there is a gap
between actions based on plans embodying a mental picture of the future
and the future itself. If the plans of all the interacting individuals were
consistent (along with some other very strong assumptions, about, for
example, conditions of competition), then a general equilibrium is logi-
cally possible; but because individuals have different mental pictures,
there is continuous unpredictable change and hence no “equilibrium” in
any useful sense. But there are patterns and this implies that there are
social mechanisms that can be identi¬ed.24

23 In response to this, so-called “complexity theory” has now been applied, if modestly,
to economic model-building. See Anderson et al., 1988, especially the essay by John
Holland, “The Global Economy as an Adaptive Process.” Insofar as he (still) ¬nds
valuable “equilibrium dynamics under uncertainty,” Arrow represents the critical, but
still unreformed leading edge of the mainstream.
For neo-classical theory each agent makes a judgement about future prices by means
of a probability distribution. As Little¬eld (1986: 28) summarizes this: “ ˜Tomorrow™
can be characterized as a vector of random variables, where the range the values can take
is known today, and more important, so is the set of variables itself.” Sadly, this is so
far-fetched as to not deserve comment.
24 Lewis and Runde (forthcoming) have opened an interesting line of inquiry in their
examination of the work of Lachman. As they see matters, Lachman addressed “the
challenge inherent in attempting to advance a strong and consistent subjectivist view
of human agency without at the same time undermining the possibility of providing a
146 A Realist Philosophy of Social Science

Existing market-as-process theories de¬ne competition in the same way
that it is de¬ned in neo-classical theory as a condition of many buyers and
sellers, each interested in realizing interests. But the function of compe-
tition is differently understood. For neo-classical theory, a condition is
(fully) competitive only if at equilibrium, price equals marginal cost and
markets clear. Competition serves, on this view, to make it impossible for
producers or consumers to set prices. Imperfect competition gives the
corporation discretion within a price / quantity range. Oligopolistic ¬rms
may be seen as cost economizers (including for Williamson and others,
transaction cost economizers).
On the process conception, it is still assumed that competition makes it
impossible for producers to set prices, but it is held also that competition is
“a method of discovery.” As Hayek (1978: 236) put it: “We have come to
understand that the market and the price mechanism [as staticly analyzed
by neo-classical theory] provides . . . a sort of discovery procedure which
makes the utilization of more facts possible than any other known system,
and which provides the incentives for constant discovery of new facts
which improve adaptation to the ever changing circumstances of the world
in which we live.”
Three points bear emphasis. First, the Hayekian view remains in the
thrall to the competitive model, certainly the marginal case. But this view
does open up some alternative ways to think about markets. That is, not
only does it reject the static equilibrium model, but it shows how price
systems “ even in conditions of imperfect competition “ actually do func-
tion in coordinating economic action. It puts knowledge, more precisely
the highly distributed forms of knowledge, at the center of coordination
problems. As Hayek (1978: 182) writes:
Utilization of knowledge widely dispersed in society with extensive division of
labor cannot rest on individuals knowing all the particular uses to which well-
known things in their individual environment might be put. Prices direct attention
to what is worth ¬nding about market offers for various things.
Or as Boettke (1997: 30), following Kirzner (1985), summarizes matters:
The essence of the coordinating property of the price system lies not in its ability
to convey perfectly correct information about resource scarcity and technological
possibilities, but in “its ability to communicate information concerning its own
faulty information-communication properties”. . . Disequilibrium relative prices,

coherent account of social institutions and socio-economic order.” They argue that while
there are dif¬culties in Lachman™s account, these could be remedied with a version of
Bhaskar-inspired critical realism. If so, the result would be very close to what is being
argued here.
Markets as social mechanisms 147

imperfect as they are, nevertheless provide some guidance in detection and cor-
rection absent from formal models of economic “information” premised on static

Third, and a presupposition of the argument for competition as discov-
ery, is the fact that in market economies, the economic situation is ever-
changing and unpredictable. “Competition is valuable only because, and
in so far as, its results are unpredictable and on the whole different from
those which anyone has, or could have, deliberately aimed at” (Hayek,
1978: 180).26
The process view of markets has much to recommend it even if Hayek™s
particular understanding is ¬‚awed (Lawson, 1997). Process theories are
correct in seeing that there is a historical dynamic to market arrangements
in which what has happened continually alters the conditions for future
decision-making. And even in markets where competition is imperfect,
the price system provides guidance in ¬nding and correcting informa-
tion essential to making economic decisions. Perhaps the most obvious
objection to Hayek-inspired theories is the failure to notice (or at least to
see through) the fact that decisions are structured and that, accordingly,
individuals have widely different (and unequal) resources in making their
choices, a failure carried over from neo-classical theory. These struc-
tures are also a legacy of past activities, currently being reproduced and
transformed. Ignoring this, of course, both ignores the powerful ordering
capacities of institutions (Lewis and Runde, forthcoming) and allows for
the widely shared myth of competition.

25 There is a historical connection here between Hayek™s views and many of the ideas of
Alfred Schutz. Of some importance, Schutz was a member of the Mises™ Privat-Seminar
¨ ¨
in Vienna in the 1930s. See Augier, 1999.
Boettke points out that new information economics, for example, the work of Stiglitz
(1994) and Grossman (1989) remains committed to rational-expectations equilibrium
analysis, concentrating on “market failure” “ the failure of markets to achieve ef¬ciency
as de¬ned by the model. Yet, Boettke notes also “their research on the informational
role of prices has led to a fundamental recasting of many basic questions in orthodox
economic theory” (1997: 29).
26 Hayek notes that “the necessary consequences of the reason why we use competition is
that, in those cases in which it is interesting, the validity of the theory can never be tested
empirically . . . If we do not know the facts we hope to discover by means of competition,
we can never ascertain how effective it has been in discovering those facts that might
be discovered” (1978: 180). The defense of competitive markets is thus global and
historical: “All we can hope to ¬nd out is that, on the whole, societies which rely for this
purpose on competition have achieved their aims more successfully than others” (1978:
180). This is surely contestable, depending hugely, of course, on being clear about the
“aims” of societies. But the point is probably moot since, in a world that has giant
corporations, it may be quite impossible to secure and maintain competitive markets.
148 A Realist Philosophy of Social Science

Better markets?
A number of important conclusions are suggested by the foregoing
account. First, actually existing markets (and market economies) and
differences in these “ differences sometimes of considerable importance “
are the historical products of agents working with materials at hand “ and
their present constitution can be explained (chapter 5).27 It follows also
that there is nothing “natural” about markets which, as human products,
can be changed. As Dugger (1992: 237) writes:
According to market mythology, our fate is due to natural law “ supply and
demand in the market “ and cannot be changed. In this way, the market myth
weakens the credibility of those who would reform the market, dashes the hopes of
those dispossessed by the market, and hides those who bene¬t from the particular
way the market has been instituted.

Granting the constraints imposed by history and nature on the various
market economies of the world, the role of government is here critical.
While this is not the place for extended discussion, it is easy enough to
see that there are a host of policy choices that have dramatic effects on
the constitution of real-world markets. As argued, there is no way for
government not to intervene so as to affect outcomes. In baseball, the
height of the mound has consequences for pitchers and hitters. There is
no way to be “neutral.” So as regards markets, the “rules” inescapably
bene¬t some at the expense of others. The only questions, then, are
what are the goals, and are the means effective? Determining the goals,
of course, is inescapably political. Given that there is no escaping some
policy choice, and thus some concrete institutionalization of the market,
we can only here list the most obvious policy areas and hint at their
Monetary and ¬scal policy. What should be the goals of macro-
economic policy and how should they be carried out?
Industrial policy. Should there be incentives for some industries,
e.g., aircraft. How does the budget for defense shape critical
Labor policy, including the length of the working day, a min-
imum wage, the organization and capacities of unions, anti-
discrimination policy, and laws regarding occupational health
and safety.

27 The task was begun by Marx and Weber, of course. There are a host of other examples
representing differing theories but consistent with the analysis of chapter 5. These include
Tawney, 1998; Polanyi, 1971, 2001; Shon¬eld, 1965; Brenner, 1977; North, 1981, 1990;
Pomeranz, 2000; Davis, 2001; Harvey, 1987.
Markets as social mechanisms 149

Environmental policy. Will the “costs” of destroying the natural
environment be taken as “externalities” from private produc-
tion, or will they be part of the cost analysis of the polluters?
Energy policy. Should some sources be given incentives or
constraints? What is the most effective means to secure the
least expensive, cleanest energy sources, to minimize waste of
Farm policy. As with energy policy, how should these markets be
structured? Can one justify subsidies to “agri-businesses”? To
Financial market policies. Are current Securities and Exchange
Commission rules encouraging investment or promoting waste
and corruption?
Family policy. What are the rights of working women, including
maternity leave policy, facilities for child-care, and safeguards
against discriminatory practices?
Health policy. How should these markets be structured? Is a
system of third-party payment justi¬ed?
Educational policy. What is the commitment of the government
to creating workers with the necessary skills and how should
this be done? Is privatization of mass education, through, for
example, vouchers, feasible or desirable?
Transportation policy. Commerce depends on transportational
infra-structure. Are there ways to utilize properly structured
markets to achieve what is necessary and desirable? Should
these be public monopolies?
Immigration policy. What are the consequences of alternative
policies on labor markets?
To be sure, none of these problem areas has lacked serious inquiry “
despite the obstacles of disciplinary fragmentation.28 But the neo-classical
model of economic behavior and markets has badly misdirected much
intellectual effort in this regard. It is hoped that the foregoing chapters
provide a useful way to conceptualize inquiry in the social sciences and
to avoid some of the more obvious pitfalls. As regards the argument of
the present chapter, it is irresponsible for social scientists to ignore the
fact that markets are differently constituted and that the ideal described

28 The problems, of course, are severely compounded by what is termed “globalization.”
There are also global markets and these, too, are historical products. Again, the topic
is huge, but see, for example, Barnet and Muller, 1974; Robertson, 1992; Barnet and
Cavanaugh, 1994; Appadurai, 1996; Bauman, 1998; Sassen, 1999; Mittleman, 2000;
Ritzer, 2004; Steger, 2005. For an account of the American role in recent globalization,
see Gowan, 1999; Stiglitz, 2002; Brenner, 2003.
150 A Realist Philosophy of Social Science

by neo-classical general equilibrium theory fails to gives us an adequate
understanding of capitalist markets. Indeed, general equilibrium theory
is the myth that “weakens the credibility of those who would reform the
market, dashes the hopes of those dispossessed by the market, and hides
those who bene¬t from the particular way the market has been instituted.”
One ¬nal point: the foregoing insisted that the neo-classical theory of
markets was fatally ¬‚awed, but it also noted that conceiving of markets as
social mechanisms had distinct bene¬ts if properly understood. We might
note here that the criticism of general equilibrium theory leads also to a
critique of centralized planning. The extent to which ideas of central-
ized planning depend upon general equilibrium theory is remarkable. As
Diane Elson (1988) argued in her critique of Mandel (1986, 1988), in
attempting to achieve an ex ante equilibrium, the planners play the role of
a Walrasian auctioneer. But not only must the planner know more than
he could, but things change through time, so theoretically speaking, the
auctioneer could never close shop; he never ¬xes a price “ a point made by
the Austrian critics of general equilibrium theory. Historical experience
suggests that centralized planning cannot be made to work “ although
obviously, our historical experience is limited to undemocratic regimes.
But one must confront the argument that it is theoretically impossible
for a centralized planner to have all the pertinent information.29 On the
other hand, various forms of market socialism remain, at least, consistent
with these arguments.
More generally, the argument of this volume should encourage us to
believe that if social science is sometimes part of the problem, it is, of
necessity, essential to the solution.

29 The formal theoretical similarity of planned and market economies was ¬rst noticed by
Barone (1908) and Pareto (1909). For an account of the misunderstandings of their
work by Samuelson and the formalists, see Boettke, 1997. Boettke notes also that most
current forms of market socialism suffer from sharing assumptions of general equilibrium
theory, but it remains contestable, versus Hayek, whether there are some genuine real-
world alternatives to private property and idealized competitive markets. For a defense
of a distinct and historically novel form of market socialism, see Elson™s remarkable “
and ignored “ essay (1998).
Appendix A The limits of multiple

It is too often thought that multiple regression overcomes the problem
of isolating causes and allows us to weigh their importance to outcomes.
Even careful writers often confuse the following ideas:
1. A (some “variable,” e.g., IQ) correlates with B (some other variable,
e.g., income)
2. A “predicts” B
3. A “explains the variance” in B
4. A “explains” B
5. A “causes” B
We can handle 4 and 5 together. We can say 4, “A explains B” only if
we can say “A causes B.” But, ¬rst, correlations do not establish causes.
Causes are “mechanisms” which produce outcomes. We can have a cor-
relation where there is no conceivable mechanism, e.g., the price of eggs
in a Beijing market and the price of Microsoft on the New York Stock
Exchange. Second, there are always many causes of any outcome. In
order to make a ¬re, we need, in addition to some combustible material,
a source of heat and oxygen. Absent any of these, no ¬re. So which is
more important? We get a ¬re only if the right combination is present. (It
takes a good deal more heat to ignite a vinyl fabric than it does to ignite
cotton.) If we pick out a source of heat as “the cause,” that is because we
assume the presence of oxygen and the combustible material. We forget
about the oxygen and say, the spark “caused” the ¬re. (Weber called this
“adequate causation,” the difference in the existing state which brought
about the effect.) This is both convenient and unsurprising. But the fact
remains: all the factors are important: you will not get a ¬re if any are
absent. Consider then Sarah™s ability to score big on the SAT. What is
“the cause”? Which of the “factors” (causes) will be more important?
Sarah may be “bright,” but she also was well-motivated, got some terri¬c
education “ and she felt good on the day of the test.
1 and 2 also can be treated together. A correlation between A and B can
be between 0.0 and 1.0. 0.0 is no correlation; 1.0 is a perfect correlation:

152 A Realist Philosophy of Social Science

for every change in A there is a commensurate change in B. Perfect cor-
relations are rare indeed. This raises the question: when is a correlation
(or in multiple regression, a “coef¬cient of correlation”) “signi¬cant.”
But we need here to distinguish “policy” or “scienti¬c signi¬cance” from
“statistical signi¬cance” “ roughly, measures designed by statisticians to
ensure “a good ¬t.”1 A statistically signi¬cant ¬nding may or may not have
some policy or scienti¬c signi¬cance. This generally will be a judgement
about causality.
Where we have a statistically signi¬cant correlation, we can predict.
If there is, for example, a positive correlation of 0.7 between the winner
of the Super Bowl and the political party which won the presidential
election, then we can predict the winner once we know the winner of
the Super Bowl. There is no suggestion that there is a causal connection
here. The correlation is a statistical ¬‚uke. But for purposes of prediction,
the relation is all that I need. Nor because there is no causal mechanism
involved here, can I explain the presidential victory by pointing to the
team which won the Super Bowl.
Smoking and cancer is a good example. There is some causal mecha-
nism at work in cancer production, likely several, and smoking is related
to this in ways that we do not yet understand. Some people do smoke all
their lives and never get cancer. And some people who never smoke do.
But we know that the probability of getting cancer signi¬cantly increases
if you smoke: A “predicts” B.
Sophisticated scientists are very often careless when they speak about
3, “explaining the variance.” What they intend by this can be brie¬‚y
summarized. Assume that there are a number of “factors” which taken
together presumably “determine” some outcome. The idea then is to
¬nd out how signi¬cant each factor is in “producing” this outcome. The
language of “producing an outcome” or “determining an outcome” is
causal language. But indeed, such language is entirely inappropriate. We
need to go a little deeper to see what is at issue here.
Assume ¬rst a standard regression equation, a set of dependable, mean-
ingful independent variables (a, b . . . ) with a linear relation to the depen-
dent variable (Y).
Y = a + b 1 + b 2 + b 1 b 2 + e (Equation 1)

1 On “good ¬t,” see below. Ziliak and McCloskey (2004: 333) showed that “of 137 relevant
papers [in the American Economic Review] in the 1990s, 82 percent mistook statistically
signi¬cant coef¬cients for economically signi¬cant coef¬cients.” Indeed, things were get-
ting worse. In their earlier study (1996), 70 percent of papers published in this distin-
guished journal did not distinguish the two, “and fully 96 percent misused a statistical
test in some (shall we say) signi¬cant way or another” (2004: 332). Such is the power of
positivist theory of science.
Appendix A 153

“Y,” the “dependent variable,” presumably is “determined” by the inde-
pendent variables, “a + b1 . . . ” The problem is then one of variable selec-
tion. The goal of the analysis is a “good ¬t.” If we do our work well,
what we end up with is “a useful statistical description defensible against
plausible alternative interpretations” (Achen 1982: 13). It is critical to
emphasize that the very best result is a statistical description, a point nearly
always missed. At best, the result is a highly simpli¬ed picture, a sta-
tistical snapshot, of a fantastically complicated concrete social situation.
For example, as an abstract ratio, the crime rate represents a picture of
crime in the real world. It leaves much out “ obviously. On the other
hand, as Achen remarks: “A picture of a friend is useless if it covers a
football ¬eld and exhibits every pore. What one looks for instead is an
interpretable amount of information, with the detailed workings omitted”
(Achen, 1982: 13). As regards the crime rate, the “detailed workings”
include, of course, the speci¬c structured actions of everyone in society:
both criminals and non-criminals. While it would be agreed that a crime
rate is such a snapshot taken from a very long distance, the same is true
of all other statistical results, including the results of regressions.
A useful description “ a good ¬t “ is not so easy to come by. One test of
this is the “coef¬cient of correlation,” R2 . It is usually said that R2 gives
“the percentage of variance explained” in the dependent variable by the
regression. But as Achen comments: this is an expression that, “for most
social scientists, is of doubtful meaning but great rhetorical value” (1982:
58f.). The rhetorical value lies in the supposition that, ¬rst, a large R2
guarantees “good ¬t” and, second, in the more radical confusion, that the
number represents the causal importance of the factor in the regression.
Neither supposition can be sustained. As Achen says, R2 “is best
regarded as characterizing the geometric shape of the regression points
and nothing more” (1982: 59). It is easy to see why it is nothing more
than this. Achen says: “The central dif¬culty with the R2 for social sci-
entists is that the independent variables are not subject to experimental
manipulation”(1982: 59). In the natural sciences, one tests theories about
causality with an experiment. The experiment seeks to “control” the con-
ditions to see if the hypothesized cause actually produces the outcome
which the theory predicted. This is not possible in the social sciences.
“Regression,” which presumes to “control” variables, mathematically, is
often thought to be an adequate substitute for experiment.
There are several lines of argument that it is not. One regards the
problem that “variances are a function of the sample, not the underlying
relationship” (Achen, 1982: 59). That is, the linear model (Equation 1)
is a local analysis whose result depends upon the actual distributions of
the variables in the population sampled. Thus, “in some samples, they
154 A Realist Philosophy of Social Science

vary widely, producing large variance; in other cases, the observations are
more tightly grouped and there is little dispersion” (Achen, 1982: 59).
(One needs some further understanding of statistical analysis to fully
grasp this criticism.) For this reason, then, “they cannot have any real
connection to the ˜strength™ of the relationship as social scientists ordi-
narily use the term, i.e., as a measure of how much effect a given change
in the independent variable has on the dependent variable” (Achen,
1982: 59).
Second, there is the problem of assuming that the measured variables
“add up” to 1.0, the problem of “additivity” and independence. Achen
offers an example:
If the regression describes, say, domestic violence in countries as a function of vio-
lence in prior years plus economic conditions, can one say which variable is more
important in causing violence? For most purposes the answer is no. The units of
one variable are violence per amount of prior violence; the units of the other are
violence per amount of economic dislocation. One can say only that apples differ
from oranges. As theoretical forces abstracted from any historical circumstances, they
have no common measure. (1982: 70)

Equation 1 makes us believe that the variables are both additive and
independent (with b1 b2 taking into account the interaction effects of the
variables).2 But this is never the case. The best sort of example to illustrate
the general principle is to see the confusion in the mostly meaningless dis-
cussions of the relative effects of heredity and the environment. Consider
a parallel (idealized) biological study, a study that requires a controlled
Take a genotype replicated by inbreeding or cloning. This minimizes
genotypic individuality. Place them in various carefully controlled envi-
ronments. It is then possible to establish rough tables of correspondence
between phenotype on the one hand and genotype-environment combi-
nations on the other. The results, called the “norm of reaction,” are never
predictable in advance.3 They are not predictable since genetic and envi-
ronmental factors are not additive (and hence cannot be represented by
linear equations). They are causes in transaction in exactly the sense that

2 See also “path analysis,” an extension of regression which makes the same assumptions
as does regression, but where “a regression is done for each variable in the model as
dependent on others which the model indicates are causes, direct and indirect.” “Path
coef¬cients,” then, are “used to assess the relative importance of various direct and indi-
rect causal paths to the dependent variable.”
3 This follows Lewontin, 1974. See also Lewontin, 1982 for a fuller account of the impor-
tance of “norms of reaction” and their absence in human quantitative studies: “Except
for such traits as the presence or absence of blood-group antigens . . . we do not have a
norm of reaction for any human trait” (1982: 22).
Appendix A 155

genes cause different phenotypical outcomes in different transactional
If such norms could be experimentally established for persons in their
development, then across the range of controlled environments and
(cloned?) genotypes, one could relate the variances in outcomes with
the changes in the independent variables. This would still not provide
the proportion of causation since causation does not suddenly become
additive. But one could talk sensibly about their relative “importance.”
One could “explain the variance” sensibly. More dramatically, as Achen
says, we might conduct an experiment in which we put some children in
middle-class homes and the others in closets. There surely will be differ-
ences in cognitive ability, personality, etc. Almost certainly, most of the
differences in these realized capacities will be “explained” by environ-
ment. Conversely, put them all (per impossible) in the same environment,
most of the variation surely will be “explained” by heredity. The forego-
ing explains, of course, the importance of (identical) twin studies “ and
their limitations.
It is obvious that except for identical twins, not only are no two geno-
types the same, but that in the concrete real world, there is not any way
in principle to specify all the relevant environmental “variables,” exactly
because these are not independent. The social world is real enough, but the
mere fact that necessarily it is mediated by the consciousness of agents
makes it impossible to say how a condition will be experienced and
understood by the agent, and thus what effect it will have on him and
his behavior. Accordingly, not only will multiple regression not give the
proportion of causality involved in some outcome, but in general, it will
not even allow us “to explain the variance.” Indeed, there are differences
between individuals which are rooted in our genes, but if, for example,
we want to explain inequalities in the real world, we had best ¬nd some
other way.4
But all this is not to say that quantitative methods have a minor place,
or more outrageously, that they have no place in social science. First, they
are enormously useful in providing descriptions of facets of the society.
Second, we need to have numbers of all sorts of things, demographic, eco-
nomic, political and sociological: the number of Americans or Hawaiians,

4 Whit¬eld and McClearn recognize that “the causal nexus for any particular phenotype
may be viewed as a complex network with inputs from multiple genetic loci and from
multiple environmental factors” (2005: 106). But they nevertheless put con¬dence in
the idea that “increasingly sophisticated statistical designs of structural equation models
in quantitative genetics can be enormously ampli¬ed by incorporating measurements
of theoretically relevant environmental variables, speci¬c genetic loci, and physiological
mediators of the causal nexus” (2005: 112).
156 A Realist Philosophy of Social Science

the income distribution, the number of voters, by income, by ethnicity,
etc., crime rates, etc. Of course, there are considerable methodological
and indeed epistemological issues here, best handled by sophisticated
methodologists. We need only repeat that there is considerable danger in
confusing causal explanation with description. Description is an essential
¬rst step and lacking an adequate picture of social reality, explanation is
pointless even where it is not dangerous.
Second, these methods give us capacities to generalize, including gen-
eralizations discoverable only through the use of regression and similar
methods. As Kemp and Holmwood (2003: 12) argue, “quantitative and
statistical techniques may be used to reveal patterns . . . that are obscured
by the range of in¬‚uence operating on them . . . Likewise, statistical tech-
niques can sometimes be used to extract revealing patterns in data even
when the precise parameters of the various in¬‚uences are not known prior
to analysis.”5 As argued, generalizations are materials for explanatory
inquiry: for example, how do we explain differences in variance regard-
ing ethnicity or income in voting behavior, etc. As above, identifying such
patterns does not give us causality, but as Kemp and Holmwood write,
˜“the existence of such a pattern suggests that there may be structural
in¬‚uence at work, a claim that can be investigated further to examine its
plausibility” (2003: 12).

5 See also Olsen and Morgan (forthcoming).
Appendix B Comparison, Mill™s methods
and narrative

Historical sociology typically employs comparison as a research strategy.
In what follows we concentrate on the idea of comparison and try to
get clear on its use “ and misuse. Inspired by Theda Skocpol™s in¬‚uen-
tial States and Social Revolutions (1979), recent sophisticated discussions
argue that the methods codi¬ed by John Stuart Mill are appropriate tech-
niques toward advancing a more “scienti¬c” approach to comparative
analysis. Finally, as part of this, we look at the idea of “narrative” as
a mode of explanatory sociology. The issues are at the heart of a lively
recent debate between Skocpol and William Sewell.1 While the two writ-
ers seem often to be talking past one another, once the key assumptions
and confusions are clari¬ed, a resolution is readily available.
We can develop these issues by considering the ideas put forward
by James Mahoney (1999) regarding “Nominal, Ordinal, and Narrative
Appraisal in Macrocausal Analysis,” the title of his important essay.2
Mahoney is interested in identifying explanatory generalizations and
offers that the three strategies identi¬ed in the title of his essay can be, and
are, employed jointly by researchers, despite the view often taken that the
work uses only one strategy. On his view, “each of these three different
strategies represents a different technique that can be used for assessing the
same causal relationship.” Thus States and Social Revolutions is wrongly
taken to employ one basic strategy, “nominal appraisal.” Mahoney argues
that, indeed, Skocpol employs successfully all three. Mahoney provides
by far the most sophisticated attempt to sustain this view. We conclude
by arguing that his effort fails, and that, indeed, while Skocpol did use
these three different “strategies,” this introduced incoherence into her

1 See Sewell, 1996. The essay won a prize in 1991 and, still unpublished, was vigorously
attacked by Skocpol in the concluding essay of her Social Revolutions in the Modern World
2 See also his “Strategies of Causal Assessment in Comparative Historical Analysis,” in
Mahoney and Rueschaemeyer (eds.), 2003. Portions of this essay in this volume were
adapted from an earlier essay, “Strategies of Causal Inference in Small-N Analysis,”
Sociological Methods and Research, 28 (May 2000).

158 A Realist Philosophy of Social Science

account, and remarkably, gave her book a persuasive power which was
quite alien to some critical speci¬c claims regarding the use to which she
put nominal and ordinal methods.3

Nominal comparison
Mahoney writes: “Nominal (or categorical) comparison entails the use of
categories that are mutually exclusive cases . . . and collectively exhaus-
tive” (1999: 339). Thus Skocpol™s concern is revolution versus non-
revolution. Following John Stuart Mill, she hopes to establish “valid
causal associations.” To do this, one can seek to establish what several
cases have in common with the phenomenon to be explained “ Mill™s
Method of Agreement. Or one can contrast cases in which the phe-
nomenon to be explained and the hypothesized causes are present to cases
where both the hypothesized cause and the effect are absent, but which
are otherwise similar. This is Mill™s Method of Difference. As regards
macro-historical phenomena, Skocpol notes that “in practice . . . it is
often possible, and certainly desirable, to combine these two compara-
tive logics” (1979: 37).
She has three positive cases to be explained, the social revolutions in
France in 1789, Russia in 1917 and China in 1911. In very interesting
chapters she undertakes a comparative-historical analysis in which she
considers these, and though brie¬‚y, three “negative” cases, or situations
where there was no social revolution. While her emphasis is on the posi-
tive cases, strictly, she employs Mill™s joint method.4 She concludes that
the three positive cases have in common “(1) state organizations sus-
ceptible to administrative and military collapse when subjected to inten-
si¬ed pressures from more developed countries abroad and (2) agrarian
sociopolitical structures that facilitated widespread peasant revolts against
landlords” (1979: 154). Taken together, she concludes that these are “the
suf¬cient distinctive causes” of these revolutions.

3 To anticipate, in an early essay review of her book, I argued that her “narrative” made
her argument convincing, but that it was, indeed, incoherently joined to what Mahoney
calls “nominal appraisal.” See Manicas, 1981. This is substantially the criticism that has
been reasserted by Sewell, 1996. I consider this below.
4 Schematically,

ABC (x1 , x2 . . . ) E (y1 , y2 . . . )
’ E (y1 , y2 . . . )
ADF (x1 , x2 . . . )

GH (x1 , x2 . . . ) not-E (y1 , y2 . . . )

MN (x1 , x2 . . . ) not-E (y1 , y2 . . . )
Probably A is cause (or “determining condition”) of E
Indeed, this better ¬ts Skocpol™s use of Mill™s methods.
Appendix B 159

The logic is clear enough. One hypothesizes various potential “causes,”
then eliminates some as neither necessary nor suf¬cient. Mahoney argues
that nominal methods provide “a sound logical basis for eliminating
potential necessary and suf¬cient causes” (1999: 241f.). Unfortunately,
this is an unduly optimistic conclusion, for many reasons.
We can begin with the Method of Agreement. In Mill™s own words:
If two or more instances of a phenomenon have one circumstance in common,
the circumstance in which alone all the instances agree is the cause (or effect) of
the given phenomenon.

ABC (x1 , x2 . . .) ’ E (y1 , y2 . . . .)
ADF (x1 , x2 . . .) ’ E (y1 , y2 . . . .)
Probably A is cause (“or determining condition”) of E
Capital letters ABC etc. are hypothetically identi¬ed possible “causes”
of E. The small letters in parenthesis represent unknowns present in the
situation but not part of the analysis. The inference to A as the probable
cause depends on the assumption that B, C, D and F are not necessary
conditions (since E occurs when they are absent) and that E is not the
product of the joint operation of ABC, BC, DF and ADF. These have
not been eliminated as suf¬cient conditions for E.
Mill™s method of difference aims at eliminating conditions as not suf-
¬cient. In his words:
If an instance in which the phenomenon under investigation occurs and an
instance in which it does not occur have circumstance in common save one, that
one occurring in the former, the circumstance in which alone the two instances
differ is the effect, or the cause, or an indispensable part of the cause, of the

ABC (x1 , x2 . . .) ’ E (y1 , y2 . . . .)
BC (x1 , x2 . . .) ’ non-E (y1 , y2 . . . .)
Probably A is cause (“or determining condition”) of E
In the real world, it is next to impossible to ¬nd cases satisfying this method.
Indeed, Sewell (1996) is quite correct to insist that Skocpol assumes what
he calls “experimental temporality,” and that, “in order for Skocpol™s
revolutions to be subjected to her comparative method, they must be
conceptualized as analogous to separate ˜trials™ of an experiment. This
means that the trials must be both equivalent and independent” (1996:
160 A Realist Philosophy of Social Science

258). But it is quite impossible to satisfy these conditions. It was just this
problem which led Mill to offer his Joint Method, a method which is not a
combination of the methods of agreement and of difference. Mill™s point
precisely was that it was the fallback method when one did not have con-
ditions of closure optimally demanded for the method of agreement and
all but required for the method of difference. The point is that since most
outcomes are not the product of a single non-trivial suf¬cient condition,
the methods can offer little help with trying to sort this out. On the other
hand, while they can offer some con¬dence in eliminating factors as not
necessary, if there are alternative paths to outcomes, this will not be of much
help either. This is the point about equivalence. Consider, for example,
alternative paths to modernization (Moore, 1966). That is, at some point
in time, given existing conditions, some condition may be “necessary” for
some outcome. But it may not be necessary at some other time and place
given other existing conditions. For example, at some time and place,
breaking the capacity of a landed nobility to resist private property may
be necessary for there to be commercial development, but at some other
time and place, given (say) abundant merchant capital in cities, this is
not necessary.
But there is a far more serious problem. Mill intended his methods
to be used to establish causes in Hume™s understanding of causality: a
is the cause of b means that there is law-like, but contingent, association
(“constant conjunction,” “invariant relation”) between a and b. The real-
ist conception developed in chapter 2 has it that a causes b means that
a produces or brings about b, and the relation is not contingent. On the
Humean reading, causality can be analyzed in terms of necessary and
suf¬cient conditions; on the realist reading, it cannot. Mahoney is not
insensitive to the problem which this objection presents, even if he fails
to see that it is fatal.
As he says, “even if logical methods exist for identifying necessary and
suf¬cient causation, some analysts contend that it is still not a productive
way to think about causation” (1999: 348). Why not? To say that C is a
suf¬cient for E is to say: “If C, then E” with the “if . . . then” analyzed
as a material conditional: the conditional is false only when C is true and
E false. C is a necessary condition for E means, “If not-C, then not-A.”
(It follows, logically, that if C is a suf¬cient condition for E, then E is a
necessary condition for C “ they are contrapositives.) Thus, paying your
parking tickets is a necessary condition for graduation; graduating is a
suf¬cient condition for having paid your parking tickets.
The analysis of causality in terms of necessary and suf¬cient conditions
suffers from being both too wide and too narrow. It is too wide since it
Appendix B 161

includes cases where there is no suggestion of causality. One doesn™t grad-
uate because one pays one™s parking tickets and graduating did not cause
the paying of the parking ticket. On the other hand, and more seriously,
genuine causes may be neither necessary nor suf¬cient conditions.5 Con-
sider an example where causation (and explanation) is at issue.
Suppose that you are Louis Pasteur interested in determining the cause
of fermentation. You select a variety of fermented liquids, including beer,
wine, vinegar and cider. Microscopic examination shows that each has a
characteristic micro-organism (which, as it turned out, was Mycoderma
aceti, a kind of yeast). You conclude that this micro-organism is the cause
of the fermentation. Compare, then, Skocpol™s use of Mill™s methods.
First, in both examples, it is clear that considerable theory is involved
since theory is essential if we are to have an idea of what to look for, the
character of the likely causes and how they are to be identi¬ed. Pasteur
looked for micro-organisms; Skocpol looked for “structural conditions”
with particular reference to the political economy, the institutions of the
state and the international political and economic environment. She could
have surely looked elsewhere. She might, for example, have not looked
at “structural conditions” or merely put them in the background. She
might, instead, have looked at the psychology of actors. Or she might have
theorized “structures” differently, and instead of focusing on political
economy, she might have considered long-term cultural facts, e.g., the
role of the Catholic Church, the Reformation, etc.6
Second, Mycoderma aceti is the cause of fermentation in a very ordinary
sense. It is an identi¬able “thing” which, in appropriate circumstances,
produces fermentation. It is the difference in the prevailing state of affairs
which brings about change in the wine. Is there an analogy in Skocpol™s
account? Skocpol recognizes, of course, that her “suf¬cient distinctive
causes” are not at all like Pasteur™s Mycoderma. Indeed, if anything com-
pares to the Mycoderma, it might be King Louis XVI! She herself writes:
“as everyone knows, the summoning of the Estates-General [by the king]

5 More generally, causality does not submit to an extensional analysis. Thus, on the exten-
sional analysis, inconsistent counterfactuals are both true: “Had Hitler not invaded Rus-
sia, he would have won the war” and “Had Hitler not invaded Russia, he would have lost
the war” are both true, since the antecedents of both are false.
6 Her structuralism, of course, was speci¬cally in response to what she termed “aggre-
gate psychological theories” (1979: 9). But as I argue subsequently, she could not escape
assumptions about motivations. Similarly, like many who advocate structuralist (and
causal) accounts, her anti-interpretativist bias leads her to dismiss “culture” as a crit-
ical part of the account. See McDonald, 1996. But as argued in chapter 3, despite the
bias shared by Parsonians and structuralist Marxists, “culture” can hardly be separated
from “structure” properly understood.
162 A Realist Philosophy of Social Science

served not to solve the royal ¬nancial crisis but to launch the Revolution”
(1979: 65). “Launch” is causal. Could not one say, as many historians
would, that this was a cause of the Revolution (Hexter, 1971)? Of course,
he did not intend “to launch a Revolution” and we can only guess at how
things would have developed had the king chosen to do otherwise (as he
surely could have). On the other hand, there is something right about
Skocpol™s interest in structural conditions, even if, in no stretch of the
imagination can we think of them as causes in Hume™s sense, and even if
they are, for her, causes in the realist sense. That is, the king™s decision
to summon the Estates-General had the consequences it did because of
the existing “structural conditions.” Compare a ¬re: the “structural con-
ditions” are the presence of combustible materials and plenty of oxygen.
To have a ¬re, then, one needs only a lit match. In both cases, there were
pertinent causes at work, both events and causal mechanisms, and once
we see how they came together, we can explain the outcome.
Third, in the Mycoderma example, the analysis of causes in terms of
necessary and suf¬cient conditions utterly breaks down. We noted earlier
that the method of agreement eliminated conditions as not necessary. In
this case, can we say that the Mycoderma aceti is a necessary condition for
fermentation? In fact, we cannot. That is, there is a whole set of organisms
which, in the right environments, produce fermentation. “Fermentation”
is an abstract term and we need to be aware that there are also very
different kinds of fermentation. We see, accordingly, the pertinence of
the unknowns in our schema, y1 , y2 . . .
Nor (even assuming we now employ the method of difference) can we
say that Mycoderma aceti is a suf¬cient condition for fermentation for in
addition to a fermenting agent, fermentation requires the presence (or
non-occurrence) of many other conditions as well. Some of these may
be too obvious to make a fuss over, e.g., the presence of a fermentable
liquid; others may be less obvious, e.g., a temperature range which will
permit the process to begin and proceed, or a host of non-events, e.g., the
absence of an accident in the laboratory which would affect the outcome.
Mahoney acknowledges this problem by introducing Lieberson™s
example of the drunk driver. He writes: “These methods correctly show
that drunk driving by itself is neither a necessary nor a suf¬cient condition
for an automobile accident” (1999: 349). But indeed, in some particu-
lar case, drunkenness was the cause of the accident, and one surely does
explain it by noting that the driver was drunk.
This also suggests severe limits on providing true and non-trivial causal
generalizations. Thus it is false that “Whenever a driver is drunk, he has
an accident,” “People with cancer are all smokers,” “If you put salt in
Appendix B 163

water, it dissolves.” Consider, then, the structure of Skocpol™s covering
law “explanation” of the three social revolutions:
If a state organization susceptible to administrative and military collapse is
subjected to intensi¬ed pressures from developed countries abroad and there is
widespread peasant revolt facilitated by agrarian sociopolitical structures, then
there will be a social revolution.
In 1789, France was subjected to such pressures and had an agrarian social
structure which facilitated widespread peasant revolt.
Hence, France in 1789 had a social revolution.

The ¬rst premise is the “explanatory generalization.” We can replace
China or Russia for France in the second premise and thus also “explain”
their social revolutions.
The argument is a perfectly valid deduction and hence if the premises
were true, then assuming the covering law model, Skocpol would have
achieved her hoped for “valid, complete explanation of revolution.” But if
the conditions are not suf¬cient, the ¬rst premise is false. Thus a collapse
of the state coupled with widespread peasant revolts, along with inevitable
other conditions, could lead to a restoration of the old order and not “the
emergence of new sociopolitical arrangements.” Consider, for example,
Iran in 1953 where Mohammed Mossadegh™s revolutionary attempt was
thwarted with assistance from the CIA. On the other hand, one could
have a social revolution even where the state has not “collapsed,” for
example, the Cuban or Sandinista revolution which succeeded because
armed insurgents were able to defeat the forces of the existing state.7 On
the other hand, one could “save” the explanation by providing pertinent
ceteris paribus clauses, for example: “If a state organization susceptible to
administrative and military collapse is subjected to intensi¬ed pressures
from developed countries abroad, then unless it receives support from other
international actors . . .” But there is now the real danger of trivializing the
Which is not to say that we can never offer true non-trivial generaliza-
tions expressing necessary and/or suf¬cient conditions: “Whenever there
is combustion, oxygen is present”; “When a bullet passes through the
brain, the person dies.” The presence of oxygen is a non-trivial “factor”
in combustion; bullets through the brain are suf¬cient “causes” of death.
In both cases, it is worth emphasizing, we do not need inductive argu-
ments to have con¬dence in these generalizations exactly because we have
an excellent understanding of the pertinent causal mechanisms.

7 For critical discussion along these lines, see Burawoy, 1989 and Sewell, 1996, drawing on
Burawoy (1989). Skocpol (1994) takes issue with these criticisms, but she seems unwilling
to acknowledge that, given her explicitly stated goals, they are entirely fair.
164 A Realist Philosophy of Social Science

Ordinal comparison
Mahoney could reply that the foregoing argument shows only what he
acknowledges, that nominal methods are not suf¬cient. On his view, these
need to be supplemented with ordinal comparison.
Ordinal analysis involves rank ordering cases using variables with three or more
values based on the degree to which the phenomenon is present. This kind of
analysis facilitates the use of J. S. Mill™s method of concomitant variation,8 in
which the analyst tries to establish causation by looking at the relationship between
scores on an ordinally measured explanatory variable and scores on an ordinally
measured outcome variable. (1999: 353)

This is what Tilly had in mind in talking about “textbooks and learned
essays” which hold that all valid comparison is variation-¬nding. It is also
the basic logic of those quantitative methods which talk of “dependent”
and “independent” variables.9 (See appendix A)
Important here is the fact that causality is inferred not by identify-
ing necessary or suf¬cient conditions, but in terms of correlations.10
Mahoney points to a host of problems in the attempt to combine this
method with nominal methods and the interested reader should turn to
his careful account. But the critical point to be made here is acknowledge-
ment that while a correlation may be evidence for a causal relationship,
causality must be inferred. Thus, “the discovery that the explanatory vari-
able is related to the outcome variable in an ordinal assessment does not
indicate how one should interpret the nature of the association” (1999:
354). This is familiar ground, but what is most remarkable is his idea
that “process tracing” is a critical part of the analysis.11 This involves
“identifying the causal mechanisms that link explanatory variables with
the outcome variable.” “Causal mechanisms,” the text continues, “can
be de¬ned as the processes and intervening variables through which
an explanatory variable exerts a causal effect on the outcome variable”
(1999: 363).

8 In Mill™s formulation: “Whatever phenomenon varies in any manner whenever another
phenomenon varies in some particular manner is either a cause or an effect of that
phenomenon, or is connected with it through one fact of causation.”
9 It is of more than historical note that Durkheim (1982) was very sensitive to the problems
of using Mill™s versions of nominal strategies, and insisted that what Mahoney calls
“ordinal appraisal” was the most useful technique. Durkheim wrote before the invention
of linear regression models and followed what Mill called the Method of Concomitant
Variation. As Mahoney writes: “Ordinal analysis is in fact the strategy of inference that
comparative historical researchers turn to when they seek to identify linear correlations
across a small number of cases” (1999: 353).
10 This is true also of more sophisticated linear regression models, discussed in appendix
11 See also “path analysis,” appendix A.
Appendix B 165

This is a puzzling remark since it seems to reproduce the system-
atic ambiguity regarding causality between Humean and realist views.
“Processes” suggest a realist conception of causality, but “intervening
variables” suggest otherwise. Again, consider the example of the drunk
driver. Assume a correlation between explanatory variable A, drunken
driving, and outcome variable E, automobile accidents. This is non-
spurious because we know that there is a mechanism which explains the
correlation. It will be a complicated causal story, involving what alcohol
does to the brain, how this affects motor control and perception, etc.
But we know that drunkenness is causally related to accidents because
we know that drunkenness is a cause in the realist sense: it is not merely
“conjoined” to accidents nor is it an “intervening variable”; drunkenness
can produce an accident (again, via well-known mechanisms).
Notice also that the correlation does no explaining. After the accident,
we explain it not by appealing to the correlation, but by noting that in
this particular case, the driver was drunk, carrying along in the back-
ground what we know about alcohol and our neurophysiology. Similarly,
the probability of getting lung cancer is higher if you smoke, so it is very
good policy not to smoke. But Sam who doesn™t smoke wants to know
why he got cancer, and Charlie, who does smoke, did not. The correla-
tion is of no help. As above, smoking is neither a necessary nor suf¬cient
condition for getting cancer, but here we remain in the dark regarding
the mechanism which makes this probability what it is. Still, we believe
strongly there is one. Indeed, identifying the mechanism or mechanisms
is precisely the goal of inquiry. There is a strict analogy in historical soci-
ology. But to be clear on this, we need ¬rst to look at Mahoney™s ideas
regarding “causal narrative.”

Causal narrative
The idea has recently become fashionable, but as Skocpol (1994: 332)
rightly remarks, since “narratives can be structured in many, many
ways,” “to advise people to write ˜narratives™ is really to advise noth-
ing.” Mahoney offers that the idea “has been extensively examined,” and
that “a consensus has emerged that narrative can be a useful tool for
assessing causality in situations where temporal sequencing, particular
events, and path dependence must be taken into account” (1999: 1164).
But how narrative does this remains unclear.12 There are, perhaps, two

12 Mahoney provides a number of citations to the literature (1999: 1164). He hopes to
remedy two fairly clear shortcomings: the absence of concrete illustrations, and the
gap between narrative and “other” strategies of causal assessment. He is remarkably
ingenuous with regard to his illustrations, but it seems to the present writer, at least, that
the gap remains un¬lled. See below.
166 A Realist Philosophy of Social Science

very different conceptions of what is involved.13 The following sums up
Mahoney™s view of the matter. With the technique of causal narrative, he
the analyst attempts to validate aggregated cross-case associations by “breaking
apart” variables into constituent sequences of disaggregated events and com-
paring these disaggregated sequences across time. The purpose of unpacking
aggregated variables through narrative is not only to provide a contextualized
description of cases; rather, the goal is to support a cross-case argument at a
more disaggregated level.
This technique relies on historical narrative . . . However, the procedures
through which analysts decide whether a narrative account lends support to a
causal pattern have not been well speci¬ed . . . Event-structure analysis is the
most developed statement on how narrative can be wedded to causal inference.
(2003: 365f.)

Two problems may be noticed. First, disaggregation, a reductionist
strategy, is intended to provide temporality, but as Sewell writes (follow-
ing Burawoy), “by fracturing the congealed block of historical time into
arti¬cially interchangeable units,” history is “frozen” (1996: 258); the
temporality assumed, as Sewell says, is akin to the sort available in exper-
imental contexts. This is not merely a matter of “grouping events” so as to
compare slices of history, but a matter of eliminating time as process. As
Burawoy insists, it fails to make sense of unique and sequentially unfold-
ing processes within cases. Second, “causal patterns” established by nom-
inal and ordinal methods are to be given “support” by means of narrative,
but how “narrative” is to do this remains unclear, except to notice that
there is a “complex trade-off” between “the clarity of informal narrative
presentations and the rigor of explicitly diagrammed narrative accounts”
(Mahoney, 2003: 367). Mahoney™s efforts at “explicit diagrams” simply
do not translate into causal narratives “ except as chronologies, exactly
because we have conjunctions and not causes, realistically understood.
That is, instead of a causal story, we have “a series of connected events.”
One might note, as well, that common sense overcomes the most careful
of writers. In moving to narrative, it is very easy to lapse, inconsistently,
into hints of references to causal mechanisms, to promissory notes unful-
¬lled. Indeed, despite the “of¬cial” doctrine regarding causality, such

13 Sewell helpfully identi¬es three ways that temporality is conceptualized: “teleological
temporality” “experimental temporality” and “eventful temporality.” We do not consider
here “teleological temporality” which aims “to abstract transhistorical processes leading
to some future historical state” (1996: 247). By now, it is hard to ¬nd anybody who
would defend this idea although it may be in the background of some accounts. But
“experimental temporality” conforms nicely to Mahoney™s view of the matter, while the
option defended here conforms to Sewell™s “eventful temporality.”
Appendix B 167

references are commonplace in both the social science literature and in
ordinary thought and discourse.
A very different concept follows from rejection of the Humean notion of
causality assumed in the foregoing. Brie¬‚y, explaining outcomes requires
seeing how, through time, events and decisions were constrained and
enabled by prevailing social mechanisms (processes). The only way to
do this is with a historical narrative which integrates actions, events and
mechanisms into an evidentially convincing “story.”14 On this view, com-
parative method is still employed, but it is not employed to ¬nd necessary
and suf¬cient conditions, but rather, to identify the causal mechanisms
at work in particular cases.
The difference in these conceptions of “causal narrative” is at the root
of the debate between Sewell and Skocpol. As Mahoney says, Skocpol did
employ nominal and ordinal methods, and she did employ narrative, but
as argued here, Mill™s methods did not and could not yield the outcomes
that she had hoped for. On the other hand, her narrative ¬ts very well
into the description in the previous paragraph. Sewell rightly remarks
that “the bulk of her book is composed not of a rigorous weighing of
comparative evidence but of carefully constructed causal narratives spec-
ifying how social revolutions are brought about in her three cases” (1996:
260). And as he points out, the best statement of her narrative strategy
is in a footnote, where she says: “social scienti¬c analyses of revolutions
are never . . . given suf¬cient analytic weight to the conjunctural, unfold-
ing interactions of originally separately determined processes” (Skocpol,
1979: 320, quoted by Sewell, 1996: 260). Indeed, if we drop all talk
of identifying the “suf¬cient distinctive causes” of social revolutions by
means of Mill™s methods, what remains of Skocpol™s book is, as Sewell
agrees, of considerable value.
One of the strengths of Skocpol™s account is response to what she takes
to be the existing social scienti¬c theories of revolution. For her, “in con-
trast to modes of explanation used by currently prevalent theories, social
revolutions should be analyzed from a structural perspective” (1979: 5).
But this means that explanatory generalizations have been abandoned for
a realist conception of explanation. She does not give a systematic analysis
of her use of the concept of structure, but it is clearly in¬‚uenced by Marx,
and clearly, the “currently prevalent theories” to which her structuralism
is opposed are the ones she refers to as “aggregate-psychological theo-
ries.” Nevertheless, from what she does say about structure, and more
14 Describing an account as a “story” identi¬es its rhetorical moves. Despite post-modernist
appropriations, “stories” may be assessed for their truth-value, here a function of evi-
dence, plausibility and coherence. But indeed, there may be several quite good, convinc-
ing, plausible accounts. See Hexter, 1971.
168 A Realist Philosophy of Social Science

crucially, how it functions in her account, it is possible to make some
important inferences about what she has in mind. For her, structures
are determinate relations, objective and impersonal. They “condition,”
“shape” and “limit” situated actors, though they are not to be identi-
¬ed with the actual actions or transactions of acting agents. Structures
also have a “dynamic” and a “logic” and these can be discovered (see
especially pages 14ff.).


. 5
( 7)