The Democratic Virtues of Teaching Statistics

Authored By 
Jerome Schafer
Blog contributor 
Policy Fellow

It is well-established that most voters know little about politics. Although a significant proportion of Americans is remarkably well informed, many more are almost entirely ignorant of political facts. This pattern is strikingly stable over time. The potential implications for democracy include failure to select the politicians that serve voters’ interests best, and to hold them accountable for their actions in office. While many political scientists advocate compulsory civics classes in high school and in college, evidence suggests that teaching statistics may be a more promising path to improving voter competence. Basic statistical knowledge can be intuitively transmitted and proves very useful to political decision-making.

An optimistic view of democratic accountability holds that citizens do not necessarily need much knowledge in order to make informed voting decisions. For example, they need not know where candidates stand on every issue. They may use heuristics or “information shortcuts,” such as partisanship or cues from interest groups, to determine which candidate they feel closest to. Furthermore, the American two-party system alleviates voters from much of the burden of policy knowledge. Although Americans are slightly less informed on average than their peers in other advanced democracies, the level of “correct voting” – that is, the fit with a voter’s preferences measured by a long battery of questions – is higher in the United States than in most countries that do not have a two-party system.  

While voters may rely on partisan cues when they select politicians, they still need to keep track of policymakers’ record in office in order to hold them accountable. Yet, recent studies suggest that voters make systematic mistakes in their retrospective assessments of incumbent performance. For example, they put more weight on the final year of economic growth relative to the first three in a presidential term. This may allow strategic politicians to fool them. Richard Nixon famously launched counter-cyclical high-spending policies in the year preceding his re-election in order to boost citizens’ subjective assessments of his economic record.

Scholars have suggested that compulsory civics classes in secondary schools and in college might increase political knowledge. Length of formal education appears to augment electoral turnout and other forms of democratic participation. However, the effect of teaching civics on democratic accountability is less certain. Voters who went to college tend to be more interested in politics than less educated individuals, but biology and English majors are on average just as knowledgeable as social science majors. This indicates that more knowledge about political institutions and political history does not necessarily imply that voters have more relevant information available when they make their decisions.

Educating citizens to process political information may be a more promising path to improving democratic accountability. Assessing the incumbent’s performance in office involves computations that are in essence similar to solving statistical problems. In an ideal world, voters would precisely define their prior beliefs about the incumbent’s ability and update them in light of all new evidence that they have observed. In practice, they need to rely on mental shortcuts to quickly come to decisions. The mistakes that they make follow predictable patterns and may be reduced by teaching basic statistical concepts.

The cognitive challenge of political decision-making consists in identifying relevant information and distinguishing it from irrelevant information. Yet, voters tend to overemphasize recent events even when they remember the data from previous periods. It constitutes a mild form of a well-known inclination to neglect base rates when making statistical inferences. Voters also struggle to restrict their attention to relevant facts, and let unrelated events influence their decisions. If the local football team wins Sunday before the election, the incumbent governor gains two percentage points in votes on average.

Training in statistical reasoning can reduce mistakes in political decision-making. It provides formal tools to make complex inferences. But it also allows individuals to present information in a format that facilitates intuitive reasoning. For example, natural frequencies are much easier to understand than conditional probabilities and rates of change, especially when supported by pictorial representations. Instead of analyzing yearly GDP growth rates to assess the president’s economic performance, citizens may be better off by looking at how the size of economy has changed from a base of 100. Basic statistics can be intuitively taught even to primary school children and provides literacy in probabilities and risk that can improve political decisions.

Jerome Schafer is an ISPS Policy Fellow and a third-year graduate student in political science at Yale.