Menu Adam R Brown

Notes navigation: Browse by titleBrowse by authorSubject index

Lupia and McCubbins: The Democratic Dilemma

Disclaimer. Don't rely on these old notes in lieu of reading the literature, but they can jog your memory. As a grad student long ago, my peers and I collaborated to write and exchange summaries of political science research. I posted them to a wiki-style website. "Wikisum" is now dead but archived here. I cannot vouch for these notes' accuracy, nor can I say who wrote them.

Lupia and McCubbins. 1998. The Democratic Dilemma. Cambridge: Cambridge University Press.

In Brief


Driving through an intersection would require perfect information about where all the other cars are going, but we use a traffic signal as a substitute for all this information. We can still make a rational, reasoned choice without perfect information. In politics, we frequently hear that people (voters especially) lack the information required to make good decisions; this criticism has led to serious critiques of democracy. Yet politics has much in common with this traffic signal: "Using similar logic, it follows that limited information precludes reasoned choice only if people appear to be stuck at complex political intersections and lack access to effective political traffic signals" (page 12, in chapter 1). This book's main claim is that people do have access to "effective political traffic signals."

Main Point

In effect, citizens are "principals" who elect "agents" to government. But these agents can abuse their principals' trust because it would be excessively costly for principals (i.e. voters) to monitor every move that their agents (politicians) make. This principal-agent problem can be solved if there is a third party, a "speaker." This speaker assumes responsibility for telling the principals what they need to know about their agents. Applied to politics, this speaker might be a news commentator, a politician, a candidate, an interest group, a trusted friend, or some other opinion leader. Principals choose to heed advise from "persuasive" speakers.

Institutions can determine whether particular speakers are persuasive or not--that is, whether principals will view particular speakers as reliable political traffic lights. What matters is whether the speaker has knowledge and appropriate incentives, and crucially, whether the principal perceives the speaker's knowledge and incentives. Thus, we can overcome difficult principal-agent problems if we have a speaker who we trust telling us whether the agent is doing its job in our best interest.

This is largely a formalization and experimental analysis of Popkin's (1994) arguments about cue-taking and information shortcuts. For example, Popkin uses the vignette of Gerald Ford's ignorance of how to eat a tamale. As Lupia and McCubbins point out, it is significant that this gaffe didn't hurt Ford everywhere, only among those voters who (1) knew how to eat a tamale and (2) connected ignorance of Latino culture with ignorance of Latino issue concerns (see p 67). Similarly, partisanship, ideology, and other information doesn't affect voters uniformly. "Persuasion is a function of perception, context, and choice" (p 67).

Place in Literature

The big debate: Earlier scholars have noted that voters don't know much about current issues, have difficulty recalling or recognizing the names of their representatives, and know little about the governmental process. Similarly, scholars have argued that legislators have little knowledge about the policy impacts of the legislation they pass. In general, most of these scholars conclude that low information means (1) voters cannot make reasoned choices at the ballot box and (2) legislators get captured by the bureaucracy and special interests. But:

"We reject this conclusion because it is based on an erroneous, though prevalent, assumption. The assumption is that people can make reliable predictions about the consequences of their actions only if they know a detailed set of facts about these actions. If this assumption is true, then it must also be true that reasoned choices can be made only by ambulatory encyclopedias--people who can store and quickly retrieve a detailed set of facts about every decision they make. If, however, the assumption is false, then even individuals who cannot answer simple survey questions or explain the details of proposed legislation may nevertheless be capable of reasoned choice." (p 18)


Two New Assumptions

  1. Learning is active and goal-driven. We have a scarce ability to pay attention to stimuli around us, so we direct our attention only where it is most worthwhile--meaning that we ignore much of the information we could, conceivably, gather.
  2. Connectivity. People compare new bits of information to preexisting mindsets or theories they might embrace. For example, I observe dark clouds; in general, I have connected dark clouds with rain. Thus, I conclude that it might rain soon. If my goal is to remain dry, I will therefore seek cover. I do not need a detailed understanding of meteorology to make this connection. Through trial and error, we learn over time what connects with what.

The General Logic

For an overview of the model, see the diagram on p 77. In general, the theory makes four broad points, quoted from page 20:

  1. Learning is active: People choose when and what to learn.
  2. Knowledge is the ability to predict accurately thc consequences of choices, and information is the data from which knowledge may be derived. Therefore, knowledge requires information, but large amounts of information do not ensure knowledge.
  3. Information is valuable only when it improves the accuracy of predictions about the consequences of choices.
  4. Reasoned choice requires only that people make accurate predictions about the consequences of their choices.

Learning from Others

In gathering political information, we can learn from our own experience or from others. In reality, we often rely on others. This insight is not new:

"Some scholars argue that people rely on factors such as party identification (Downs 1957), known issue biases (Calvert 1985), likability (Brady and Sniderman 1985), certain histories of observed behaviors (Sobel 1985), being in a competitive situation (Milgrom and Roberts 1986), shared policy interests (Krehbiel 1991), or elite status (Zaller 1992). Other scholars argue that people learn from the aggregate actions of others such as are contained in history (Downs 1957, Fiorina 1981, Key 1966), polls (McKelvey and Ordeshook 1986), the media (Iyengar and Kinder 1987; Page, Shapiro, and Dempsey 1987), levels of campaign expenditure (Lupia 1992), the size of public protests (Lohmann 1993), certain campaign events (Lodge, Steenbergen, and Brau 1995; Popkin 1991), and public mood (Rahn, Kroeger, and Kite 1996)" (from page 40).

What is new is this: People choose which of these information sources to use. We need a theory that explains why and how people make this choice among potential information sources. More broadly, we must explain who can persuade whom.


"Learning from others requires persuasion. We define persuasion as one person's successful attempt to change the beliefs of another. In settings where reasoned choice requires learning from others, persuasion is a necessary condition for reasoned choice" (p 40).

Speakers are persuasive if principals perceive them as knowledgable and having congruent interests. The term "perceive" is crucial. Details:

Both these conditions are reduced by external factors:

You can also learn from a speaker with interests opposite to your own if you witness that speaker addressing people with common interests. For example, a Democratic can watch Rush Limbaugh address Republicans and pick up important information.

Main Conclusions

"We conclude that:

Empirical Test

An experimental design, in which respondents can rely on a "speaker" for help in predicting the outcome of an unobserved coin toss. The experimenters manipulate the speaker's incentives and knowledge about the coin toss; they also manipulate whether the respondents know the speaker's incentives and knowledge. All hypotheses are verified, at least in this experimental setting. For evidence that this theory works in the real world, consider Lupia (1994).

Research by the same authors

Research on similar subjects


Lupia, Arthur (author)McCubbins, Mathew D. (author)American PoliticsExperimentResearch DesignDemocracyVotingInformationPrincipal-AgentRational ChoiceShortcutsSignaling

Wikisum home: Index of all summaries by title, by author, or by subject.