How not to use evolutionary theory in politics: a critique of Peter John*

Transcription

1 British Journal of Politics and International Relations, Vol. 2, No. 1, April 2000, pp How not to use evolutionary theory in politics: a critique of Peter John* KEITH DOWDING In a recent issue of this journal Peter John (1999) suggests we can use an evolutionary account to explain policy change. In particular he suggests we should see the battle of ideas about policy formation as an evolutionary process and gives as an example the introduction and abolition of the poll tax. John is correct in two claims in his article. First, traditional models of policygeneration tend to ignore the role of ideas, concentrating attention upon the bargaining and power struggles between different sets of competing interests. Secondly, he is right that evolutionary explanation has a place in the social sciences. But these two thoughts are best kept apart and the way he packages them suggests a poor understanding of evolutionary explanation and of the role ideas may play within it. There are at least three problems with his account. First, the object at which he directs explanation in his example the poll tax is misspecified. Secondly, he fails to specify a mechanism for the natural selection of ideas, leaving his claim about the promise of evolutionary accounts vague and unsatisfactory. Finally, he fails to distinguish learning as an intentional process from selection as an evolutionary one. Object of explanation The idea of evolution is familiar to everyone. Life is a struggle in which the fittest survive. Fittest here means the lifeforms that flourish are those better equipped than others, which perish in the environment as that environment changes, sometimes incrementally, sometimes dramatically. Political Studies Association Published by Blackwell Publishers, 108 Cowley Road, Oxford OX4 1JF and 350 Main Street, Malden, MA 02148, USA 72

2 How not to use evolutionary theory in politics The process by which some lifeforms are weeded out and others adapt is called natural selection. Underlying natural selection is a random process of gene mutation. Genes are the replicators that encode the traits and attributes by which animals are suited to survival. Genes that produce the traits best fitted for the environment are reproduced at higher rates. Sometimes evolutionary explanation is criticised for being tautological. Fitter is defined in terms of survival and the test of the fittest is their survival rate. (In fact, this led Popper at one time to deny evolutionary explanation scientific status on the ground of its unfalsifiability though he later recanted.) Evolutionary explanation is not very good at suggesting exactly what animals (defined by traits) we should expect to see in which environment. There are simply too many different possible forms and what constitutes the environment includes different mixes of those traits. Or, in gametheoretic terms, in any environment there are multiple equilibria and so we cannot produce predictions about what we should expect to see. All we can ask of evolutionary explanation is a prediction of what sorts of traits we may expect to see, given what we know about the relative advantage of different traits to perform given functions. For example, the trait of producing a single egg may confer advantage to a bird, as the parents can devote all their energies to feeding a single nestling, making it stronger. However, where there are predators, there is danger in putting all one s genes in a single egg; spreading the risk by producing a clutch of eggs may be advantageous. What follows is a falsifiable prediction about the average numbers of eggs in a clutch correlating to the number of predators, controlling for other risks. This prediction does not explain why certain birds should arise and when, but derives from a comparison between the egglaying traits of similar species in different environments. John misspecifies the objects of possible evolutionary explanation in politics with his example of the poll tax. We cannot hope to explain a specific tax policy with an evolutionary model. All we can do is to specify what types of tax policy might be adopted and remain stable over time in different economic and political environments. To that end we can (and tax experts did) predict that a poll tax would not be long-lasting because it was inefficient (very expensive to collect) and unpopular (more people lost than gained with its introduction and its regressive nature led to the perception that it was unfair). However, this is not specifically an evolutionary explanation since it identifies no particular mechanism by which a tax (randomly chosen) should remain stable over time. If a government chooses a tax system because it is likely to confer electoral advantage then we have intentional, not evolutionary, explanation. Political Studies Association

3 Keith Dowding Mechanisms As Elster (1989a, 1989b; see also Hedström and Swedberg 1998) has made clear, specifying mechanisms through which some institutions and outcomes are selected is the most general form of explanation we are likely to achieve in the social sciences. But not all selection mechanisms are best thought of as providing evolutionary explanation. Mechanisms that do provide evolutionary explanation must, at the least, involve a certain functional feature of an institution which will lead to its survival and the explanation must provide the means by which that institution is replicated. If the institution were specifically designed with that function in mind, we have intentional explanation. If the replication is through deliberate copying because the institution functions well, again we have intentional, not evolutionary, explanation. If the function was undesigned or a byproduct then we have functional explanation and a candidate for evolutionary explanation. In a famous essay designed to rebut Elster s (1978) claim that functional explanations were lacking in the social sciences, Hardin (1980) explained the growth of the US bureaucracy in terms of the beneficial consequences for incumbent Congressmen. A larger bureaucracy means more bureaucratic problems for voters and more complaints to their Congressmen, who are re-elected as they are more able than newcomers to help voters, but thereby have less time for legislative work that devolves to the bureaucracy which therefore grows. In the urban sphere we have a mechanism that suggests that in capitalist societies developers have advantages over ecologists, nimbyists and other opponents of change because local politicians recognise the electoral advantages of local growth strategies (Dowding 1996). Where local growth confers fewer electoral advantages we should expect to see less development or managed growth, and there is corroborating evidence for this hypothesis (Schneider et al. 1995; Clark and Goetz 1994). Despite his early promise to set out the necessary and sufficient conditions for public policy (John 1999, 40), at no time are these more than hinted at. Furthermore, he says The attraction of evolutionary theory is that it maps out causal mechanisms (John 1999, 47), but we also are told the nature of the selection process needs to be further investigated. What is the role of powerful elites in selecting policies? What is the role of groups and public opinion? (John 1998, 188). John argues: this can be the force of argument in deliberation, public opinion or selective benefits or rents that interest groups and policy entrepreneurs differentially excise from the adoption of policy (John 1999, 48). So, if in one case study we have powerful elites that get their way no matter what, that is one selection mechanism. 74 Political Studies Association 2000.

4 How not to use evolutionary theory in politics If we find in another case that groups led to a stable equilibrium (Bentley 1908), that is another selection mechanism. If public opinion turns out to be important in a third example, that is another selection mechanism. Or if one side wins the argument, that is yet another. This is no way to specify natural selection in evolutionary models applied to the social sciences. By not clearly specifying a mechanism, or series of mechanisms, and by choosing the wrong object at which to direct his explanation, John fails to provide an evolutionary explanation of anything. His so-called evolutionary explanation of the poll tax with six moments in local finance history comes down to the claim: there was this bunch of competing ideas, at a given time one was chosen because the time was right and if it was any good it survived but if it was not it got chucked out. That is evolutionary explanation at its tautological worst. Non-intentional and intentional selection I have followed a stricter line regarding what counts as evolutionary explanation than John, but I do not wish to enter a semantic debate. We can, for the sake of argument here, assume that specifying any non-intentional selection mechanism by which policies are generated may be called an evolutionary explanation. What would such an explanation look like? In fact, given that policies are created by people for reasons, it is quite hard to specify what a non-intentional selection mechanism may be and this may be illustrated by considering Downsian natural selection of government policies, in a two-party system dominated by a single ideological dimension. Black s (1958) median-voter theorem demonstrates that, if there is a single policy dimension and voters have single-peaked preferences, the median position is the Condorcet winner. Less formally, Downs (1957) shows there is a centralising tendency for two parties operating in a single ideological dimension, because the party that captures the vote of the median voter will win the election. Assuming that the party that forms the government carries out its manifesto promises and the median voter supported the winning party because of those policies, the policies were selected through the operation of the plurality electoral system, given the democratic single-dimension environment. Is this mechanism that selected these policies intentional or nonintentional? Well, assuming the plurality electoral system was not instituted in order to satisfy the preferences of the median voter, one might think it cannot be an intentional explanation. But that is to move too fast. We can think of several scenarios in which the government policy was selected through the operation of the competitive party system as specified. Political Studies Association

5 Keith Dowding Scenario 1 The policies of the two parties are created by party members. Taking no account of electoral success, they simply decide through their internal procedures according to their pure ideological preferences. We could imagine that policy here was created randomly by each party constrained only by each party s ideology. The winning party is still the one closest to the median voter in ideological space, but need not be that close to the median voter s ideal point. The winning party only needs to be closer than its rival. We have some sort of prediction for government policy based on the selection mechanism, which is that winning parties need to secure the vote of the median voter. But we are far from getting a point prediction, that is predicting that government policy will closely match that of the median voter. In this scenario, even if the electorate s preferences remained fixed over time, the parties policies, the winning party and government policy may vary along almost the entire ideological spectrum. Scenario 2 If we add the assumption that parties choose policy based both on party ideology and the desire to win elections, then they cannot ignore the preferences of the electorate, nor the policies of the rival party. Here, we may expect to see Downsian convergence to the median voter. This is so even if the parties are unaware of Black s theorem even vaguely (securing the vote of middle England ). Trial and error should lead both parties towards the median. Scenario 3 Finally, we have convergence where the parties are aware, at least to some extent, of Black s theorem and deliberately target the median voter and, of course, also try to shape the preferences of the electorate towards their own ideology through argument, persuasion and propaganda. Which, if any, of these scenarios can be considered even in broad terms an evolutionary account of policy formation? All three have a selection mechanism, the electoral system, which was not designed (so we assume) to promote the views of the median voter but which privileges those views. In scenario 1, policy-generation is quite random and, whilst the vote of the median voter is the key to electoral success, the median voter s ideal point 76 Political Studies Association 2000.

6 How not to use evolutionary theory in politics may be far from government policy. In scenario 2, the selection mechanism leads to convergence though party strategists do not understand why. In scenario 3, convergence, though still caused by the selection mechanism, is partly intentional, for the parties use the mechanism as part of their strategy. The mechanism remains the same, but its importance grows as the actors respond to it. I do not think any of these cases should be called evolutionary explanation. They all involve a mechanism, and pointing out social mechanisms is important in social science. But we should not confuse the clarity of the social mechanism by making a weak analogy with social evolution. In none of these stories do we have a mimetic equivalent to genes; we have no replicator, unless scenario 3 is the conscious adoption of policies to appeal to the median or swing voter. However, again, this constitutes intentional rather than evolutionary explanation. What would an evolutionary mechanism look like? Social evolution Reproduction of animal traits occurs through genetic transmission. The genes are the replicators through which traits are passed down the generations. John suggests we may be able to offer an evolutionary explanation for the choice of policies seen as ideas and, following Dawkins (1978, ch. 11; see also Dawkins, 1986), he suggests these mimetic equivalents of genes be called memes. But genes replicate traits, and memes in Dawkins model replicate behaviour (or, in modern political science parlance, institutions). An animal is the summation of its traits and a culture the summation of its institutions. John is confusing the object of selection with the mechanism. Perhaps the best critique is to demonstrate how evolutionary theory may be used. John (1998, 1999) explicitly suggests that evolutionary explanation is superior to rational-choice explanation, but one way of applying evolutionary theory to the social sciences is through evolutionary game theory. What is important in evolutionary game theory is that the players are not the animals, but rather the programmes coded in their genes (Binmore 1992, ch. 9). These replicators determine strategies in a game and replicate themselves in proportion to their success (Dawkins 1978; see also Monod 1972, ch. 1). For animals this seems uncontroversial, but if we translate this to human behaviour its interpretation needs great care. We accept that animals behave instinctually but like to believe humans act on reasons. We do, but our reasons are our rationalisations of our actions before, or even as, we carry them out. Thinking strategically is a conscious Political Studies Association

7 Keith Dowding thinking-through of reasons for acting in a given manner. When we act strategically, in true game-theoretic style, it is because our genes include a programme allowing us to do so. This is not to say that what we think in any given situation is encoded into our genes, but rather our ability to think is so encoded. The ability to think is more efficient than trying to encode behaviour directly onto genes and hence the development of brains capable of complex thought processes (Axelrod 1997, 18). This phenotypic plasticity (see Dennett 1991, ch. 7) allows for a greater variety of behaviours, development and learning. The more developed the brain, the more clearly it can plan into the future using what has actually occurred in the past (and not what could have occurred). However, even here, when faced with complex situations, and situations for which history has only partially prepared us, we look for simplifying devices to aid us. We look for patterns in the data and we operate through sets of rules. More generally, we can see rules of thumb, conventions, codes of conduct, ideologies and ideas as replicators. This is what Dawkins calls memes (Dawkins 1978, ch. 11; cf. Dawkins 1986). What sorts of evolutionary explanation can be developed with this process? We can explain human institutions conventions, norms and so on such as the development of the idea of duty and trust in people. Putnam (1993) in his well-known account of trust in Italy, argues that trust enabled northern Italy to develop other institutions, which conferred social and economic advantage over southern Italy. Trust conferred evolutionary advantage, in a way that was not intended (though presumably enjoyed) by both trusters and trustees. The replicator in this case is a form of behaviour being trustworthy which has been promulgated through repetition in which trust matters. This is sometimes described as the internalisation of a norm (Scott 1971). The idea of being trustworthy mattering to me can be seen in several senses. It might matter because of a feeling or a sense. That is, I feel bad if I do not carry out my obligation. In this sense the feeling is instinctual. It might be thought of as a Humean desire. It may be argued that such an idea takes obligation out of obligation, since it would seem that I act thus for non-rational reasons. My motivation is a genetically programmed reaction to the situation I face. In this case, any attempt on my part to explain why I act thus out of duty is just a rationalisation of my actions caused by genetic programming. However, this conclusion is not necessary. We have to examine why I should have such a feeling. I can only have it because I think there is such an obligation. I have this duty because I have so decided, whether I decided through a priori analytic philosophising; whether I learned it through socialising my synthetic experiences, probably 78 Political Studies Association 2000.

8 How not to use evolutionary theory in politics heavily influenced by my parents and early peers; or a mixture of both. At the very least the genotype encoded in our genetic structure leads to such feelings, given our rationalisations about the situations in which we find ourselves. We simply drop any thought of requiring a motivation beyond thinking that x is the right thing to do given our moral beliefs. But this does not take us beyond evolutionary selection. It takes us into mimetic selection. This is the level of cultural transmission of the types of beliefs and ideas that people create and learn from each other. Societies with some types of moral behaviour will do better than others in given environments. The (selective) environment is shaped, at least in part, by the unintended consequences of such moral behaviour. The transmission of these institutional replicators can still be represented as the equilibria of games. Phenotypic plasticity allows for this learning facilitated by our ability to communicate through our highly developed languages. This is the way in which to bring evolution into the explanation of the development of policies. Policies may be viewed as institutions that enable some communities, or societies, or states, or nations to do better than others militarily, technologically, politically or economically. The time-scale for such emergent advantages is measured in decades or centuries, as in Putnam s account, and not the few years suggested by John s poll tax example. If we want to use a mimetic evolutionary explanation we need to specify precisely what advantages an institution confers in relation to rival institutions and also the mechanism by which it is replicated. A long timescale allows us to recognise how a selection process provides advantages and contrast it with deliberate decisions of agents and the power play of interests in standard bargaining models. Rather than claiming that everything that social scientists do can be synthesised into evolutionary accounts (John 1998, ch. 8) which are therefore superior, evolutionary accounts of society need to be kept well apart from standard accounts of the policy process. This is so for three reasons. Firstly, the time-scale is different. Secondly, the type of questions best answered are very different. Finally, where selection mechanisms can be found, we need not subsume these under the general rubric of evolutionary explanation ; rather, we can see them for what they are, selection mechanisms that may operate outside intentional behaviour, or be used intentionally by actors for their own advantage. Note *I would like to thank Anne Gelling and an anonymous referee for comments helpful in revising this piece. Political Studies Association

9 Keith Dowding Bibliography Axelrod, R. (1997) The Complexity of Cooperation (Princeton, NJ: Princeton University Press). Bentley, A. (1908) The Process of Government (New York: Alfred Knopf). Binmore, K. (1992) Fun and Games: A Text on Game Theory (Lexington, MA: D. C. Heath and Co.). Black, D. (1958) The Theory of Committees and Elections (Cambridge: Cambridge University Press.) Clark, T. N. and Goetz, E. G. (1994) The anti-growth machine: can city governments control, limit or manage growth?, in Terry Nichols Clark (ed.), Urban Innovation: Creative Strategies for Turbulent Times (Thousand Oaks, CA: Sage), Dawkins, R. (1978) The Selfish Gene (Oxford: Oxford University Press). Dawkins, R. (1986) The Blind Watchmaker (Harmondsworth: Penguin). Dennett, D. C. (1991) Consciousness Explained (Harmondsworth: Penguin). Downs, A. (1957) An Economic Theory of Democracy (New York: Harper and Row). Dowding, K. (1996) Power (Buckingham: Open University Press). Elster, J. (1978) Ulysses and the Sirens (Cambridge: Cambridge University Press). Elster, J. (1989a) The Cement of Society: A Study of Social Order (Cambridge: Cambridge University Press). Elster, J. (1989b) Nuts and Bolts for the Social Sciences (Cambridge: Cambridge University Press). Hardin, R. (1980) Rationality, irrationality and functional explanation, Social Science Information, 19, Hedström, P. and Swedberg R. (eds) (1998) Social Mechanisms: An Analytical Approach to Social Theory (Cambridge: Cambridge University Press). John, P. (1998) Analysing Public Policy (London: Pinter). John, P. (1999) Ideas and interests; agendas and implementation: an evolutionary explanation of policy change in British local government finance, British Journal of Politics and International Relations, 1:1, Monod, J. (1972) Chance and Necessity (New York: Knopf). Putnam. R. D. with Leonardi, R. and Nanetti, R. Y. (1993) Making Democracy Work: Civic Traditions in Modern Italy (Princeton, NJ: Princeton University Press). Schneider, M. and Teske, P. with Mintrom, M. (1995) Public Entrepreneurs: Agents for Change in American Government (Princeton, NJ: Princeton University Press). Scott, J. F. (1971) Internalization of Norms (Englewood Cliffs, NJ: Prentice-Hall). Dr K. M. Dowding Department of Government The London School of Economics and Political Science Houghton Street London WC2A 2AE (Submitted April 1999, accepted for publication June 1999) 80 Political Studies Association 2000.