Below is an updated version of some notes I made in 2016 on Samuel Decanio’s 2014 AJPS article “Democracy, the Market, and the Logic of Social Choice”.
Decanio’s paper intriguingly brings together important themes from causal inference and social choice in arguing that market mechanisms should be preferred to elections as means of identifying good ideas. In this post I will leave aside the important question of defining the domain of policy questions that is at issue; we may all agree that the market is better than elections at designing a car (Decanio’s example), but the controversy arises when we ask whether the argument applies to e.g. systems of health care provision. Instead I will focus on how and whether the market really is a better means of generating and refining ideas than politics/elections. My main point is that Decanio presents as a difference in kind what seems to me more to be a difference in scale. A consumer evaluating a product faces many of the same problems as a voter evaluating a party or policy. The consumer has some advantages because her evaluation may benefit from a larger data set, but the nature of her evaluation is the same.
First, a bit of exposition: The main theme is the comparison of the market with political choice. Decanio emphasizes that in political contexts parties produce exclusive goods: elections are procedures by which a single party is chosen to produce a bundle of goods (policy outcomes) for a period of time. The problem with this is that the citizen does not get to observe the counterfactual, i.e. what would have happened if another bundle of goods had been produced. He emphasizes a key difference from the market: in the market, firms produce goods and several are available for purchase having been already made: he says that this allows consumers to “evaluate the effects produced” by the scientific knowledge underlying the manufacture of the products “without ever having to understanding [sic] the knowledge responsible for producing such effects”. By contrast, Decanio asserts that in the case of selecting a party to implement a proposed policy, one needs to “accurately predict the effects of rival parties’ policy proposals before the party’s policies can be implemented and their effects observed. Rival parties’ policies cannot be implemented simultaneously, and voters cannot observe the effects of different parties’ policies until after they have made their electoral decisions, the new party is placed in power, and its policies are implemented.”
To me it seems that almost the whole question is one of scale. Decanio says things are easier in the market because firms don’t have exclusive rights over production. But at a large enough scale, a single party doesn’t have exclusive rights of production over policies: there are multiple polities (countries, states, municipalities) trying different things. It’s not that we cannot attempt to approximate an experiment in evaluating policies, it’s that our evaluations have fewer observations. From the perspective of the entity making the choice (the consumer, or voter), it is also a question of scale and aggregation: there may be multiple houses available whose attributes the consumer can compare, but each consumer basically chooses only one house (which is a bundle of different attributes -location, size, decor) and doesn’t know exactly what would have happened if she had chosen another, though she can visit her friends’ houses. In the political realm the choices apply to collectivities, so all of the voters in each electorate get assigned (and collectively get to choose) one bundle of policies; as with families in houses, they only get one outcome, but they can look at what other families chose and try to infer what would have happened if they chose something else. So I think Decanio’s focus on exclusivity, his characterization of the inferential problems facing voters as being so different from those of consumers in the market, is misguided.
Another way of putting this is that the problem of an electorate choosing among parties (i.e. bundles of policies) is not actually that different from the problem of an eater choosing among foods:
The food/diet case: You only get to choose one diet; you don’t know exactly what would have happened if you had chosen something else. You pay attention to how your weight and energy levels covary with different foods and diets, but it’s not easy to infer causal relationships: you’re eating a lot of different things, and there are many things other than diet affecting the same outcomes (e.g. exercise, sleep, stress, disease, relationships with others), and it’s not clear how long each possible effect is likely to take. You get to see what other people are eating, but it’s hard to infer effects because of other (i.e. non-diet) differences between yourself and others. You can try to learn about the effects of diet by looking at scientific studies conducted on observational data; occasionally you get to see a randomized experiment and try to learn from that.
The policy case: A collectivity (or a voter in that collectivity) only gets exposed to one policy bundle at a time, and doesn’t know what would have happened if it had chosen something else. The voter can pay attention to how he feels, or how things are generally in the collectivity, and how this covaries with policies that are attempted, but it’s not easy to infer causal relationships: there are many policies that matter for a given outcome (e.g. economic growth), and there are many things other than policies that affect the same outcomes (e.g. technology, culture, events in other polities), and it’s not clear how long each effect is likely to take. You get to see how policies differ in other collectivities, but it’s hard to infer effects because of other (i.e. non-policy) differences between one polity and another. You can look at scientific studies conducted on observational data; occasionally you get to see a randomized experiment and try to learn from that.
Decanio makes a lot of the price mechanism in markets and how it is so much more informative than electoral margins. To begin with, I think market share, revenue, or sales volume would be better indicators than price if we’re thinking about firms innovating on products, but the specific metric probably doesn’t matter to the argument. But fundamentally (and especially in a common-values setting) I don’t see much of a difference between a vector of vote shares in a PR system and a vector of market shares in a particular product market (e.g. cars). In both cases, the objects of choice (products/parties) differ on multiple dimensions, so it’s not straightforward to figure out why one is “selling” so much more than another, but the competitors will surely try to figure out why, as will the consumers/voters. Again there may be a difference of degree but I see no difference of kind.
Initially I was more convinced by Decanio’s argument that political competition is different from market competition because policy bundles are essentially promises about the future, which require a theory about the state of the world and how policies work, while consumer products are already made, and the effects of the technology can be observed (measured) by the consumer without any specialized knowledge. But on reflection I think the difference is overdrawn. Indeed the consumer can observe some differences among products and choose what she likes without needing a theory about how the firms’ products will end up looking, but this is not true of many important attributes of basically all consumer products (some more than others): reliability, performance, ease of use, suitability for the particular problems the consumer is going to face. The importance of these unverifiable elements is indicated by the willingness of producers to offer guarantees and servicing arrangements and by the existence of publications like Which and Consumer reports, the popularity of websites like Yelp, and the success of Amazon (which includes a huge number of reviews). Warrantees per se don’t exist in politics, but re-election incentives by default play a similar role.
When choosing among parties, voters have information about their track record, their promises, their ideology, their leadership. Some things they can evaluate at the time of “purchase”: what the parties have done in the past; what the leaders look like and say. Other things are harder to evaluate: competence; desire and ability to implement the policies proposed. The claim Decanio makes is that the effects of a firm’s technology are observable at time of purchase, but this is not true for parties; in fact some of the effects of the firm’s technology are not observable at time of purchase (e.g. durability) and some of the effects of the party’s technology are (arguably) observable (e.g. the skill of the leaders, the party’s record).
In conclusion, I think the inferential problem of choosing the right party to govern (i.e. bundle of policies, plus leaders) is very complicated; for a single voter it has all the same features of the problem of choosing the right diet, except that there are fewer observations and the scope for experimentation is smaller. Perhaps the important differences between decisions made by the market and by elections are not about what information the decision-makers can potentially have, i.e. Decanio’s focus. Rather, the important differences are that market decisions are individual and unanimous (i.e. no one is forced to enter into an agreement) while political decisions necessarily are not. This creates distributional issues (e.g. Buchanan & Tullock), and also incentive problems: why go through all the trouble of getting the information necessary to make a good choice when you are unlikely to be decisive? My conclusion is that we should be less worried about the special difficulty of the problem of choice in politics (Decanio’s focus) than about the lack of incentive to solve it properly (as highlighted in classics by Olson and Downs).