post-autistic economics review
Issue no. 17,  4 December 2002
article 2



    issue 17 contents                           PAE Review index                               home page




Why Neoclassical Economics Explains Nothing At All1

Steve Fleetwood   (Lancaster University, UK)

© Copyright 2002 Steve Fleetwood



Critical realists (c.f. Lawson 1998; Fleetwood 1999a&b, 2001a&b) argue that neoclassical economics is rooted in the deductivist method.2 Deductivism seeks to 'explain' something by deducing or predicting a statement about that something from a set of initial conditions, assumptions, axioms and a covering law and/or some other form of constant conjunction of events which drives the inferential machinery. These conjunctions, where for every event y there exists a set of events x1, x2...xn such that y and x1, x2...xn are regularly conjoined, only occur in, and are constitutive of, closed systems. There are, however, very few spontaneously occurring closed systems in the natural world, and virtually no non-trivial ones in the socio-economic world. Using deductivism, therefore, means engineering artificially closed systems by means of known falsehoods, reducing neoclassical economics to what Hodgson (1999: 11) calls ‘the economics of nowhere’. What is not always appreciated, however, is that the presence of known falsehoods removes all explanatory power from neoclassical economics. This paper illustrates the point via a brief analysis of the theory of labour demand.

Closed system theorising: the example of the theory of labour demand

The law of labour demand, which states that the quantity of labour demanded varies indirectly with the (real) wage, is an example of one kind of constant conjunction of events - a Humean law. Incidentally, the significance of this apparently arcane law should not be underestimated: it underpins the entire neo-liberal project of making labour markets more flexible. This constant conjunction is artificially engineered via (at least) four assumptions known as the Marshall-Hicks conditions. Demand for labour is alleged to be more elastic if:

1.       The elasticity of substitution between labour and capital is high. The demand for labour will be more responsive to a change in wages the more easily labour can be substituted for capital.

2.       The elasticity of demand for the final product is low. The demand for labour will be more responsive to a change in wages if the cost increases caused by wage increases can be passed directly to the consumer without a loss of revenue.

3.       The share of wages in the total cost of production is high. The demand for labour will be more responsive to a change in wages if production is labour intensive because ages  constitutes a relatively high proportion of overall costs.

4.       The elasticity of supply of other factors is high. The demand for labour will be more responsive to a change in wages if, when capital is substituted for labour, the suppliers of this additional capital are able to increase their supply immediately and effortlessly.

The M-H conditions close the system: without them there would be no constant conjunctions between changes in labour demand and changes in the wage. Unfortunately, however, the M-H conditions also introduce known fictions into the theory. Let us consider the four M-H conditions as four closure conditions.

Intrinsic closure conditions (ICC)

To close the system, the internal state of individuals must be artificially engineered so that the individual (person, production system, firm or whatever) always responds in the same predictable way. The ICC is maintained, for example, by assuming ubiquitous substitution between labour and capital.3 Where production involves a relatively fixed crew of workers operating a relatively fixed set of machinery it is often impossible to substitute a worker for a machine. Where production requires human emotion such as a helpful attitude, a machine cannot be substituted for a human. In some cases substitution of labour for capital is not feasible: how does one substitute a machine for a nurse to carefully bath an elderly patient? Even where is it technically possible, substitution is often not socio-politically possible. If, however, the ubiquity of substitutability is not assumed, a change in relative factor prices cannot be said to cause the substitution of labour for capital, and the constant conjunctions of events that constitute the law of labour demand fail to emerge. Assuming ubiquitous substitution is a falsehood. Where non-substitutability is recognised it is treated as a special case. Knowingly false claims are, thereby, treated as the norm, and knowingly true claims (e.g. that firms may offset a legislated wage rise by introducing flexible working practices that raise efficiency and reduce costs) are relegated to an afterthought.

The extrinsic closure condition (ECC)

ensures that the system is completely isolated from any external influences that would violate closure. This occurs by assuming things like: (a) the suppliers of other factors of production can increase their supply immediately and effortlessly should need arise, which is unrealistic, especially when the economy is running near to capacity; and (b) the elasticity of demand for the final product is low so that firms can pass wage increases onto customers, which is a rather tenuous assumption in highly competitive global markets. Maintaining the ECC, then, often requires falsehoods.

The aggregational closure condition (ACC)

ensures that the response remains constant, irrespective of the level of aggregation. Hence, the need to assume that the share of wages is high no matter what the industry. If for example, the industry was, or became, highly capital intensive, an increase in wages might be lost in the overall costs and demand for labour might not fall following a wage increase. In capital-intensive industries, then, the ACC is a falsehood.

The reducibility closure (sub) condition (RCsC)

The RCsC requires the existence of assumptions whose sole purpose is to ensure mathematical tractability. These are merely technical assumptions used to ensure the relevant functions are well behaved, thereby preventing perverse outcomes. Even where substitution of capital for labour is possible, it is often not continuous or ‘smooth’ but ‘lumpy’. Production functions would have ‘lumps’ in them and could not be differentiated. 

If any of these four closure conditions are not met (and there are, of course, more ways of meeting them than mentioned here) constant conjunctions will not emerge.  Incidentally, t
hat there are four M-H and four closure conditions is merely coincidence. Moreover closure requires far more than the M-H conditions: the latter are merely those explicitly mentioned in the theory. Other assumptions lie buried within the ceteris paribus clause; are attached to sub-components of the theory, such as diminishing marginal returns; or are made by omission.

Neoclassical economists know perfectly well that they are using falsehoods (hence the reference to known falsehoods) but often ignore the causes and consequences of constraints on their freedom to choose the M-H conditions, or assumptions in general. They cannot choose assumptions on the grounds of their truth-likeness because the need to maintain systemic closure often overrides these (and other) considerations – such as descriptive adequacy. Faced with a decision between adopting an assumption that is known to be false yet closes the system, and one that is known to be true yet violates closure, the known falsehood must be chosen or the constant conjunctions will not emerge.

Removal of explanatory power

A damaging consequence of adopting known falsehoods is that their presence removes explanatory power, for (at least) three reasons - on explanation see Runde (1998).

Explanation is not merely efficient causality

Many critical realists share with Lipton (1993; 33) the thesis that to ‘explain a phenomenon is to give information on the phenomenon’s causal history’. The causal history of a phenomena is not merely (if at all) one couched in terms of the event(s) that precede the phenomena, but in terms of the underlying causal mechanisms. One does not, for example, adequately explain (the event of) a lamp becoming illuminated simply by pointing to the (event of) flicking of the switch that preceded it. One does not adequately explain an increase in the demand for labour by pointing to the fall in wages that (allegedly) preceded it. Yet this form of ‘explanation’ is all deductivism offers. The overriding necessity of closure requires the removal (theoretically of course) of all causal mechanisms that might violate the closure conditions. So, for example, the theory of labour demand omits reference to trade unions, the introduction or abolition of labour laws and responses to them, government policy, political ideology, management systems, different working practices and so on, mechanisms that have considerable causal impact on labour demand. But here is the rub: once removed from the theory these causal mechanisms cannot subsequently be recalled and offered as part of the causal explanation. Relevant causal mechanisms are either included in the theory, in which case they can contribute to the causal explanation, or they are excluded, in which case they cannot.

Explanation is not prediction

Prediction does not constitute explanation. The conflation of prediction and explanation is referred to as the ‘symmetry thesis’ whereby the only difference between explanation and prediction relates to the direction of time (Caldwell 1991; 54). Explanation entails the deduction of an event after it has (or is known to have) occurred. Prediction entails the deduction of an event prior to (knowledge of) its occurrence. One can, however, predict without explaining anything at all. One can predict the onset of measles following the emergence of Koplic spots, but the latter does not explain measles. Even supposing an econometric model successfully predicted an event (and the predictive power of neoclassical economics is arguably weak), the regression might be grounded in no economic theory whatsoever, or be grounded in a theory that contains known falsehoods. In this case, even a successful prediction would not constitute an explanation.

Explanation does not allow known falsehoods

If, as part of a causal account, one includes a known falsehood, or, leaves out some important causal mechanism (falsehood by omission) then the ‘explanation’ can immediately be rejected as a bone fide explanation by pointing to this falsehood. Consider an analogy. In explaining how my rubbish bags get ripped during the night, I might hypothesise that it is the work of a fox or I might hypothesise that it is the work of a ghost. The explanation involving the fox is advanced because I believe it to be true. The ‘explanation’ involving the ghost is known to be false but is advanced for a pragmatic reason: I want to frighten my young nephew and stop him playing with the bin bags. Whilst the explanation involving the fox is valid (even if it turns out to be mistaken) the ‘explanation’ involving the ghost, pragmatically useful as it is, is invalid because it is known to be a falsehood. One only has to reflect upon this for a moment to see this conclusion is self-evidently correct: if known falsehoods are allowed to constitute ‘explanations’ imagine the bizarre explanations that could be advanced!

Counter arguments considered

Two counter-arguments are commonly deployed to legitimise the use of known falsehoods. The first runs as follows: ‘all theory has to leave out the inessential, has to abstract from reality, has to make unrealistic assumptions, so all theory is inevitably false in the strict sense of the word’. Now whilst abstraction is legitimate, the process is complex and cannot be elaborated upon here (c.f. Sayer 1998). I defend my claim with the following observation. Theories like that of labour demand are replete with such obvious falsehoods that to suggest they are really (legitimate) abstractions is merely a rhetorical ploy to avoid methodological discussion. In any case, as noted above most neoclassical economists admit to knowingly using falsehoods. The second counter-argument invokes the ‘method of successive approximation’ (Sweezy 1968; 11) or the ‘method of isolation’ (Maki 1992, but see Pratten 1999), and runs as follows. ‘The initial stages of theorisation use known falsehoods. Explanatory power is added in stages as realistic assumptions are successively substituted for false ones’. There are two objections to this.

  1. The method of successive approximation or isolation might be appropriate when the successive analytical steps merely involve the mechanical addition of factors that were previously assumed away. This mechanical addition is, however, not appropriate for systems where the elements possess emergent properties. When, for example new technology is introduced to a workplace or a new management regime is installed, its behaviour often evolves, giving rise to properties that were not present before. Many theoretical propositions derived on the basis of pre-emergent properties provide no grounds for the analysis of post-emergent forms of behaviour.

  2. Theory is still reliant on closed systems. All that has happened is that one closed system has been added to another (slightly larger) closed system. A succession of closed systems does not, however, add up to an open system. Consider the following example:

·         Closed system1 assumes demand for labour is determined solely by wages.

·         Closed system1 generates the deduction/prediction1 that the introduction of a minimum wage will cause a fall in labour demand.

·         losed system2 now allows labour demand to be determined by wages and (say) aggregate demand.

·         System2 is, however, still a closed system: it just contains more variables. Many of the previous (false) assumptions remain in place and, new (false) ones are added to ensure closure in this more complex system. Falsehood is then piled upon falsehood – and the dream of one day removing all false assumptions evaporates. 

The method of successive approximations, or successive closures might, therefore, be more accurately termed the ‘method of successive falsehoods’ or the ‘method of successive closed systems’. In short, the counter-arguments do not evade the critical realist critique.


To the extent that neoclassical economic theory is rooted in the deductivist method, constant conjunctions of events, artificially closed systems and known falsehoods, it explains nothing at all.


1. I wish to thank Paul Lewis for his careful comments.

2. Deductivism is also found in some heterodox (Austrian, Institutionalist, Marxist and Post-Keynesian) economics whereupon these perspectives also become vulnerable to the following critique.

3. Neoclassical theorists do, of course, recognise that substitution between labour and capital is not ubiquitous and attempt to deal with it via non-convex isoquants. ‘L’ shaped isoquants imply only one production technique based upon one capital-labour combination and allow no substitution. Isoquants with n ‘flat’ sections imply n-1 production techniques and allow limited substitution. But, where tangency between the isocost curve and the isoquants is at a corner, factor prices could change without ‘causing’ substitution. Where tangency occurs along the face of one of the ‘flat’ sections of the isoquant, then the choice of technique becomes indeterminate. 


Caldwell B. (1991) Beyond Positivism: Economic Methodology in the Twentieth Century, Unwin Hyman.

Lipton P. (1993) Inference to the Best Explanation, Routledge.

Fleetwood S. (1999a) ‘The Inadequacy of Neoclassical Theories of Trade Unions,  Labour Vol.13, No.2, pp 445-80.

----- (1999b ed.)  Critical Realism in Economics: Development and Debate, (1999) Routledge.

----- (2001a) ‘Causal Laws, Functional Relations and Tendencies’, Review of Political Economy, Vol. 13, No. 2, pp 201-220, reprinted in P. Downward (forthcoming 2002) Applied Economics and the Critical Realist Critique, Routledge.

-----(2001b) ‘What Kind of Theory is Marx’s Labour Theory of Value? A Critical Realist Inquiry’  Capital & Class 73, pp.41-       77.

Hodgson G. (1999) Economics and Utopia: Why The Learning Society is not the End of History, Routledge.

Lawson T. (1998) Economics and Reality, Routledge

Maki U. (1992) ‘On The Method Of Isolation In Economics’, Dilworth C. (ed)  Intelligibility in Science IV, Rodophi.

Pratten S. (1999) ‘The “Closure” Assumption as a First Step’, in S. Fleetwood (ed) Critical Realism in Economics;
       Development and Debate
, Routledge.

Runde J. (1998) ‘Assessing Causal Economic Explanations’, Oxford Economic Papers No. 50, pp 151-172

Sayer A. (1998) ‘Abstraction: A Realist Interpretation’, M. Archer, R. Bhaskar, A. Collier, T. Lawson, A. Norrie, (eds) 
       Critical Realism: Essential Readings, Routledge

Sweezy P. (1968) Theory Of Capitalist Development, Modern Reader.

Steve Fleetwood,  "Why Neoclassical Economics Explains Nothing At ”, post-autistic economics review,
issue no. 17, December 4, 2002, article 2.