WikiSummary, the Social Science Summary Database

Blais: To vote or note to vote

From WikiSummary, the Free Social Science Summary Database


Blais. 2000. To vote or note to vote: The merits and limits of rational choice theory. Pittsburgh: University of Pittsburgh Press.

In Brief

Research question: Is it possible to account for the decision to vote or abstain from a rational choice perspective?

Method: Multiple empirical studies (his own and others) are utilized to systematically address elements of the voter paradox, identify the strength of various reasons for voting under different conditions, and to analyze the decision calculus of heterogeneous voters.

Chapter overview (by research question):

  • Chapter 1 RQ (research question): When and where do people vote?
  • Chapter 2 RQ: Who votes? What kinds of people vote?
  • Chapter 3 RQ: P term (Do voters believe their vote could be decisive? Do voters make a probabilistic calculus? How and when does this impact turnout?)
  • Chapter 4 RQ: C term (What is the perceived cost of voting?)
  • Chapter 5 RQ: Duty (What are citizens' views on the obligation to vote? How does inclusion of this factor impact rational choice models?)

Conclusion in a nutshell: A weak form of rational choice is generally supported by most studies. Voters tend to behave as if they were making a similar calculus, but many caveats exist and relaxing of the model is needed. A clear finding (consistent throughout the chapters) is that voters are heterogeneous. A promising finding is that voters may be systematically heterogeneous: a consistent portion of the population may use a certain type of voting calculus, while other identifiable groups may utilize other calculus).

Chapter 1: When and Where to Vote

Review of Prior Literature

Blais reviews previous studies and concludes that the existing literature does not provide us with a solid explanation of why turnout is higher in some countries than in others. He does note one consistent result: turnout is substantially higher in countries with compulsory voting.

He notes that there is conflicting evidence for the following hypotheses:

  1. Strong party-group alignments result in higher turnout.
  2. More economically advanced democracies have higher levels of turnout.
  3. Turnout is higher in PR systems.

Blais's Data:

All countries with Freedom House "political rights" scores of 1 and 2. Specifically, 324 democratic elections for the national lower house in 91 different countries between 1972 and 1995.


Blais measures turnout as a percentage of those registered on the electoral list who cast a vote. He notes that others measure turnout as the percentage of the eligible population who case a vote, which is assumed to be the voting-age population. However, he criticizes this because it includes non-citizen residents and aliens.

His analysis excludes the US as a special case, which I assume is because (as he alludes to earlier) the United States is the only country where citizens must take the initiative to become registered.

The general categories of IVs he uses are: (1) Socioeconomic Environment; (2) Institutional Setting; and (3) Party System.

1. Socioeconomic Environment

Indicators: GNP per capita; growth of GNP per capita; average life expectancy; degree of illiteracy; size of population; density of population

Hypothesis: Economic development leads to higher turnout.


  • OLS regression indicates that GNP per capita has a positive correlation with turnout (coefficient of 3.32), and it is significant at the .01 level.
  • The size of the population was negatively correlated with turnout (coefficient of -2.72), and was significant at the .01 level.
  • Illiteracy had a slightly negative correlation.
  • Population density was positively correlated with turnout, but was not significant at the .05 level.
  • Controlling for SES, turnout was higher in Oceania and lower in Africa and North America, which Blais attributes to different political cultures.

2. Institutional Setting

Indicators: Electoral law; electoral system; decisiveness of elections; degree of democracy.

H1: Turnout will be substantially higher in countries where voting is compulsory.

H2: The lower the voting age, the lower the turnout.

H3: Turnout will be higher in PR systems.

H4: Generic: turnout is higher when the election is more decisive. Specific: the more powerful the lower house, the more decisive the election and the higher the expected turnout; turnout will be lower if there is an elected upper house or president or if the country is a federation.

H5: The more democratic a country (as measured by receiving a 1 versus a 2 on the political rights scale), the higher the turnout. Specifically, those countries receiving a 2 will have somewhat lower turnout than those receiving a 1.


  • Compulsory voting appears to boost turnout by 11 points. (.01 level)
  • All else equal, turnout is reduced by almost 2 points when the voting age is lowered by one year. (.01 level)
  • Turnout is reduced by six points when lower house elections are least decisive. (.01 level) [Note: What Blais means here is that the more powerful the national lower house, the more decisive the election].
  • Blais found a difference between PR systems on the one hand and all other systems, on the other hand. All else equal, turnout is three points higher in PR systems. (However, significance was at the .1 level). Also, in a fully proportional system, in which seat shares correspond exactly to vote shares, turnout is six points higher than a non-PR system. (.05 level).
  • Turnout does not appear to be lower in countries where political rights are not as well protected.
  • Economic development and size of population still affect turnout, even after institutional variables are included.

3. The Party System

H1: The greater the number of parties, the more choices voters are offered, and thus, the higher the turnout.

H2: Converse of H1: The greater the number of parties, the smaller the probability of a one-party majority government and the lower the turnout. (Theory behind it: power sharing in coalition governments is produced as much by backroom deals as by the electoral outcome per se).

H3: The closer the election, the higher the turnout


  • A gap of 10 points or more between the two leading candidates suppresses turnout slightly.
  • Turnout tends to be reduced when the number of parties increases. The important difference seems to be between a system with 2 or 3 parties and one with 6 or 7.
  • All else equal, turnout is approximately 3 points higher in PR systems than in non-PR systems. Blais thinks this is maybe because PR is fairer and more competitive. (But again, this finding appears to be significant at the .1 level).

Tests in the Literature:

Fluctuations over time

Data: The 19 countries that have been deemed democratic every year on the Freedom House political rights scale for which data was available. Turnout for all elections held since 1960.

IVs: impact of the economy and competitiveness

DV: change in turnout in a given election, compared to the previous election.

Findings: Electoral participation declines when there is a slowdown in the economy, but the relationship is modest. There is also a modest relationship between increased competitiveness and increased turnout, but it is not significant.

Turnout in different types of elections

Local vs. National. Median difference in turnout for local and national elections is 13 percentage points. Possible explanation? Local elections are less partisan, less mobilization occurs. But ultimately, Blais has no concrete answers for this.

Legislative vs. Presidential. Hyp: Turnout should be higher the more important the election. Finding: Turnout tends to be higher in presidential elections, the median difference being two percentage points. However, the results were not statistically significant.

Referenda vs. Elections. Finding: Turnout is lower in referenda voting. His proposed reason? More is at stake in elections and/or campaigns for elected officials are more personalized than campaigns for referenda.

Chapter 2: Who Votes?

Blais notes that most people vote in national elections and a good majority are regular voters. Many people vote in just about every election and most of them have already decided that they will vote by the time the election campaign starts. Blais says that these findings do not fit well with the RC model, which predicts that citizens will make their decision whether to vote at the margins, in response to the idiosyncrasies of each election and each election's perceived benefits and costs.

Blais cites Wolfinger and Rosenstone's seminal book, Who Votes? Their most significant finding was that in the United States, education is highly correlated with voting, and age is second. However, studies from other countries have not replicated these findings.

Blais therefore offers a comparative analysis from nine different countries based on a Comparative Study of Electoral Systems. Countries: Australia, Britain, Czech Republic, Isreal, Poland, Romania, Spain, Taiwan and the United States. Average turnout among these countries was 72 percent.


(When the countries were combined)

  • The two most crucial socioeconomic determinants of voting are education and age. The propensity to vote increases substantially with age and education.
  • The third most important characteristic was religiosity. Those who attended religious services regularly and who said they were very religious were more likely to vote.
  • Being married and having a higher income had small but statistically significant correlations.

(In separate regressions for each country)

  • The coefficient for age was positive in each country.
  • The coefficient for education was positive in 8 of the 9 countries.
  • The coefficient for religiosity was positive in 8 of the 9 countries.

Blais's Explanation:

Blais suggests the reason for these correlations is that people with these characteristics tend to be integrated into society. The sociological interpretation suggests that the act of voting expresses one's sense of belonging to the larger community. Blais says that these results do not tell us anything about the validity of the rational choice model, but suggests that they do give support to an alternative sociological interpretation.

Chapter 3: P Term

Does P term "matter" (do data on actual behaviour support rational choice model?)

Specific Research Questions

  1. Do people overestimate the probability of casting a decisive vote?
  2. Do perceptions of P affect the propensity to vote?
  3. Do people turnout because of minimax regret [Ferejohn and Fiorina's thesis 1974].

Data/ Method

Over 32 studies, many by author in Quebec.

Findings and Conclusions

Closeness of a race fosters turnout, and this is not just due to increased GOTV efforts. This is consistent with rational choice model. However, why/ how (p calculation) is still undetermined. P term matters "at the margin". An additive and not multiplicative relationship was found between perceived benefit and probability of being decisive. People have a vague notion that their votes "count" more in a close race. From this Blais concludes that P matters, but not exactly as the rational choice model assumes.

Chapter 4: C Term

Specific Research Questions

  1. How do voters actually perceive the cost of voting?
  2. To what extent does perceived costs affect the propensity to vote?
  3. Are costs so small that people do not bother to do a cost-benefit analysis [Barry's thesis 1978].

Data/ Method

Conducted three surveys (in Canada).

Findings and Conclusions

For most people, most of the time, the subjective cost of voting is extremely small, with some not perceive any cost at all. Even if B*P (or B+P) is extremely small, it is still greater than C, thus rational choice model holds. He concludes that marginal increases in C reduce the propensity to vote only marginally.

Chapter 5: D Term

Specific Research Questions

What are people's actual views on whether voting is a duty or not? How much turnout can be accounted for by a sense of duty? Does the rational choice model perform better for certain types of voters (according to their sense of duty?)

Data/ Method

Face to face interviews, NES surveys, Quebec and BC student surveys.

Findings and Conclusions

About half of the citizens in democracies feel a strong duty to vote. The sense of duty to vote tends to translate into actual voting (turnout impacted by sense of duty). Among those with a weak sense of duty, the rational choice model performs much better. Controlling for perceived benefits and costs of voting as well as level of political interest, duty is strongly correlated with turnout in three studies. The sense of duty increases with women, age, and religiosity. The sense of duty is not static (ex: the exposure to claim that it is irrational to vote produces reduced likelihood of voting; moral relativism translates to reduction in belief that it is a moral duty to vote). There is a trend toward less of a sense of duty to vote.

Comments and Criticism

Chapter 1: Relationship to Rational Choice Theory of Voting?

Blais's most robust finding was that higher GDP leads to higher turnout. Does this tell us anything about the rational choice (RC) model? Maybe the wealthier a country, the lower the C term, because there is a lower cost of acquiring information. But this still doesn't speak to the very low p problem.

The size of the population was negatively correlated with turnout (coefficient of -2.72), and was significant at the .01 level. This fits with the RC model because the higher the population, the lower the p term.

How would the RC model explain the fact that controlling for SES, turnout is higher in Oceania and lower in Africa and North America, which Blais attributes to different political cultures?

Compulsory voting seems to boost turnout by 11 points. (.01 level significance). This means that the costs of NOT voting are a lot higher. This is consistent with the RC model, but again, it does not speak to the problematic p term.

A gap of 10 points or more between the two leading candidates suppresses turnout slightly. This finding is consistent with RC, but the finding is modest.

What does his finding that turnout is higher in a fully proportional system (in which seat shares correspond exactly to vote shares) tell us about the RC model?

Chapters 3-5

Are studies done in Quebec/ Canada generalizable to U.S. populations? Blais takes B term (that there is a benefit to voting) as a given, without sufficient evidence to defend his claim. How to separate out benefit from fulfilling one's duty (e.g. "warm-glow" effect) from non-benefit fulfilment of duty (in the Kantian sense)? How would political scientists go about measuring this accurately? Perhaps by measuring those who dislike voting or get no satisfaction (zero utility) yet vote out of duty regardless?

Blais notes that other theories (e.g. sociological models of voting) could explain the empirical results of turnout as well as rational choice. Given the variation in the findings and unanswered questions, the attempt to save rational choice is questionably successful. For instance, rational choice is only "saved" if closeness of the race is used as a proxy for P. Also seems that C term is only extremely small if cost to prepare for voting (informational costs of becoming informed) is not explicitly asked. I find this the most problematic. Blais does not explicitly ask "what are the perceived costs of preparing to vote in an election" but says he does. What survey actually is asking is "how difficult is it to get information to decide how to vote"? This captures the ease of getting information, but does not capture opportunity costs (the time element involved in gathering, listening, reading, thinking). Yet Blais concludes it does capture these costs. The survey should explicitly ask "do you find it costless, somewhat costly, or very costly, in terms of your time, to prepare yourself to decide on candidates and issues at the polls". The question they ask "if you did not vote, you could do something else with your time" does not explicitly ask "what would you do with the foregone time in preparing to vote (becoming informed)". The question is framed as the 15 minutes in takes to actually vote, not the time which it takes one to become informed, read the propositions, etc. (see pg. 86-87). Further, one cannot assume, as Blais does, that because one comes to the polls with one's mind made up, that one's perceived costs of obtaining information to make such a decision are zero. Again, focusing on the low impact of rain or a queue only measures the actual cost of voting, not the preparation (informational costs).

Further evidence that costs were underestimated in Ch. 4 : pg. 110 in Ch. 5 - those who said it is a duty to vote did not vote in municipal elections due to lack of information and concern (e.g. costs to obtain information on those running are higher than perceived benefits). Blais claims that these folks think these elections are "not really relevant and feel free not to vote". While this is a possibility, the costs of obtaining information are not addressed (Blais brushes this under the carpet). It also does not logically follow that one should not vote if the decision does not directly impact the voter for duty reasons (as he claims is the reason why those claiming a duty vote the same as those who do not in school board elections). Seems he is capturing cost-benefit calculation here, which appears the same in those with and without a sense of duty. The handful of open ended responses he presents, while interesting, are only anecdotal (we do not know that selection bias is not a problem) and do not address whether the stated reason for voting out of duty does not involve: a) pressure to look good for the interviewer, b) propaganda from mobilization efforts, c) social pressures (as opposed to individual duty), d) a cost benefit analysis as well, which may override one's stated belief it is one's duty to vote. Anonymous testing and better research design might be able to overcome these limitations in this particular study.

If voters are truly heterogeneous--inter and intra country, by type of election (municipal versus national), age, gender, religious beliefs, and point in time--then is the best we can hope for a contingent, flexible, contextual model of voter turnout? If one can explain everything with this model, then is it overdetermined? On the other hand, if rational choice only works for one type of voter (those with a weak sense of duty), is the model "saved"? [[]]