An Economic Theory of Democracy
An Economic Theory of Democracy (ISBN: 0060417501) was written by Anthony Downs in 1957.
Part 1
Part 1 establishes a model for rational voting behavior. The first several chapters detail the mode of rationality and the motives for parties and voters that the author requires.
Parties are assumed to be vote maximizers. They are treated as teams with the uniform preferences of the centralized decision maker.
Voters are assumed to have a utility function in terms of government actions. To 'soften' the assumption that voters are capable of evaluating the expected utility of each candidate, the author argues that the problem simplifies in a two party system with substantial policy continuity. (More on continuity of party ideologies in part 2.) Voters only have to perform a retrospective evaluation of the incumbent party and compare that to an estimate of the opposition platform. Furthermore, the author allows that voters only pay attention to the "few areas of government activity where the difference between parties is great enough to impress"; they are otherwise indifferent.
The author has a notable response to the critique that voters are unaware of how some government actions create utility income for them, so there cannot be a direct causal relationship between government actions and voter utility. "[A]n important political strategy of governments is making voters aware of benefits they are already receiving." (p. 38). It must be that there is a causal relationship tempered by voters' awareness. (More on how voters become informed in part 3.)
Independence of issue spaces is assumed.
In chapter 4, the author builds out a convoluted model of party politics where
2 parties, X representing the incumbent and Y the opposition
- incumbent party moves first, selecting all positions simultaneously
all issue spaces are binary, M representing the majority opinion and N the minority
- all voters have at least one issue space where they hold the minority opinion
- voters care more about an issue space when they are among the minority
- issue spaces are independent at least insofar as there is no majority with 'perfect consensus'
Under these assumptions, a coalition of minorities is a dominant strategy. The author then adds a third level and explores Arrow's theorem; there is no dominant strategy for the first mover because the second mover always wins.
At the same time, a passionate majority is effectively guaranteed to get its way. Empirically what we see is a clear consensus on both what the important issue spaces are and what the consistent majority wants in those spaces. The two parties then converge on these policy positions.
Then part 1 concludes with a mathematical statement of the author's hypothesis for political behavior:
the government actions strategy at time t (At) is selected dependent on expected voter actions (V'e) (not known) and other parties' strategies (Pt)
other parties' strategies (Pt) are not a reflection: they are selected dependent on voters' utility income from government actions (Ue) and the government actions strategy (At)
voters' utility income from government actions at time t (Ut) are of course dependent on the government actions strategy (At)
voter actions at the time of an election e (Ve) are selected dependent on utility income from government actions (Ut) and other parties' strategies (Pt)
parties only have access to estimates of voter actions, denoted by the apostrophe in V'e.
Part 2
Part 2 begins with an exploration of uncertainty in the model. Most of this is tedious, but there is a notable assertion that political leadership comes from uncertainty. Such leaders gain a following proportional to the followers' lack of knowledge. (p. 83)
The author discusses a variety of matching mechanisms used by government, parties, and voters under uncertainty.
- Agitators (voters who, for a variety of self-oriented reasons, work to persuade other voters) effectively hold the votes for a group larger than themselves.
- Intermediaries (lobbyists), whose claims are discounted.
- Ideology as a platform.
The author particularly focuses on the latter. Voters use ideology as a shortcut, but interpreted through their perceptions of the party's reliability and responsibility. Governments must align their actions with their ideology to be seen as reliable. This also forces a coherence of ideologies, as governments cannot be reliable to contradictory interests. Parties must have continuity in their ideology to be seen as responsible. This despite the temptation to view a lost election as a sign that the ideology should be changed.
The author predicts that the nature of representative democracy depends on where uncertainty exists. If there is uncertainty as to how opinions vary across geographic regions, then representatives will be selected from districts. If there is uncertainty as to how opinions vary over race/ethnicity, then representatives will be selected from each such group.
The author also predicts that party systems develop toward an equilibrium where each election is competitive and they have little incentive to ever revise their ideologies. Changing the ideology at that point would poison their perceived responsibility.
The author introduces a model of voting behavior. It builds on Hotelling's law, taking the maximum of each voters' utility income function as the ideal position. Through this, two party systems are expected to converge to a centrist ideology while multiparty systems are expected to remain differentiated.
The author attempts to predict the number of parties a party system will have based on the distribution of voters' ideal positions. A single-peaked distribution that is monotonic on either side of the peak will always develop two parties. A uniform distribution will always develop multiple parties. Multiple modes (peaks) tend to cause multiparty systems.
The author then adapts the model in absurd ways for coalition governments, crucially not allowing for coalitions to dissolve, restructure, or hold votes of no confidence. The author's point is to say that multiparty elections are more challenging for voters to evaluate for utility, and are more likely to feature preference (protest) voting.
As an example, consider a voter that is pleased with the actions of coalition government B-C-D in period 1. They vote for party C in election 1 because they are part of the coalition and their ideology is nearest. However, in part due to strong performance of party C, party B is dropped from the coalition for period 2. The coalition's actions in period 2 lower that voter's utility income. In election 2, despite the voter's nominal preference for party C, they may vote for party B expecting a return of coalition government B-C-D in period 3.
Ultimately, the author predicts that multiparty systems behave just like two party systems: governments converge on centrist policies based on a coalition of minorities strategy. The parties project an ideology that is not centrist, both because opposition parties want to capture voters away from the government and because coalition members want to capture a larger proportion of the coalition's seats. But these ideologies are not very reliable (coalition governments select a compromise actions strategy that misaligns with all member ideologies) or responsible (centrifugal incentive is constant).
Lastly in part 2, the author demonstrates that there is no reason to expect a democratic government to put forward Pareto optimal actions. This is a result of the uncertainty, and the fact that it is illegal to buy votes to manage that uncertainty.
Part 3
Part 3 discusses the costs involved in this model of behavior.
First they explore information costs, that is the costs that voters face when becoming informed. He proposes that they seek information from a small set of vetted sources (e.g., publishers, interest groups, parties, etc.), and only on a small set of differential areas by which parties can be discriminated between.
The author also explores how various actors can influence outcomes by taking on the information costs for voters. This includes governments providing 'free information'.
The author explores non-negligible costs of voting, by which abstention can be rationally explained.
Part 4
Just a set of derivative hypotheses for future work.
Reading Notes
A pivotal work that joined the fields of political science and economics.
Like all work revolving on the median voter theorem, it has not aged well as a predictive theory. But it did lay the groundwork for more refined and attuned models.
Parts of this work sounded quite like constructivism, but ultimately no constructivist would agree with the way utility income is used to glue party and voter behaviors together, nor with the predictions made about how distributions of voter preferences translate to number of parties in a party system.