WikiSummary, the Social Science Summary Database

Geys: Explaining voter turnout

From WikiSummary, the Free Social Science Summary Database


Geys. 2006. Explaining voter turnout: A review of aggregate-level research. Electoral Studies 25 (4): 637-663.

In Brief

  • Research Question: Why do people turn out on Election Day?
  • Overview: The inability to conclude results about why people turn out to vote in part is the result of heterogeneity in defining variables. This meta-analysis provides a way to assess which variables appear significant (and which appear strongly and consistently significant) over multiple studies. Such triangulation also enables scholars to find holes in the literature (what we do not know yet and what needs to be better operationalized).
  • Method: Meta-analysis of 83 empirical studies in which DV is voter turnout; only aggregate-level studies are examined, not individual-level or experimental studies.
  • DV: Ratio of votes cast to some population measure (this varies; see table 1 pg. 3). Choice of the denominator is the contentious point (see my critique): voters divided by voting age population, voters divided by registered voters, etc?
  • IVs: Three sets of independent variables are considered: 1) socio-economic (population characteristics, 2) political, and 3) institutional


Geys examines three groups of variables in his meta-analysis. Within each group, he identifies possible indicators as strongly significant, significant, or inconclusive/insignificant, based on how frequently previous studies have identified each factor as significant.

Socioeconomic/Demographic Indicators

Strongly significant:

  • Previous turnout (but see my critique, below)


  • Population stability (but see my critique)
  • Size of population (negative effect)

Not significant:

  • Homogeneity
  • Concentration (urban)

Political Variables

Strongly significant:

  • Ex ante measures of closeness (does voter think it will be a close race)
  • Closeness at district level


  • Closeness (both ex ante and ex post - was it a close race - together)
  • Campaign expenditures (but tone may also matter)

Not significant:

  • Ex post measures of closeness
  • Fragmentation (has different impact under alternative electoral rules)

Institutional Variables

Strongly significant:

  • PR (proportional system)
  • Compulsory voting
  • Automatic and Election Day registration; absence of literacy tests and poll taxes.


  • Concurrent elections (pretty weak though)

Comments and Criticism

Case Selection

Most studies in the analysis examine the U.S. case (see the appendix). Does this skew the results? (Geys does not include a dummy for each country, or even for the U.S., which might have addressed this concern.) Still, the strongest conclusion this paper makes is that future studies ought to include these variables as controls.

Measuring Turnout

Variable selection involves normative decisions: ex: should scholars include ex-felons in states where the right to vote has been revoked for life, even though they are no longer serving time? While this is an empirical paper, the normative elements were not addressed, although claims were made: "Clearly one should be registered to be eligible to vote and to be able to register, one must fulfil all other elements of eligibility (e.g. age, civic rights)....Obviously, excluding those legally forbidden to vote should be preferred..." pg. 3. The absence of variables may be problematic (Geys disregarded age, education, and income). If the ecological fallacy can be methodologically overcome (King, 1997), why were these variables still left out? (see footnote 6).

This problem is severe. McDonald and Popkin (2001), which Geys does not cite, demonstrated that replacing the Census's measure of "voting age population" with a refined "voting eligible population" completely changes the trends in turnout. (The improved measure 'excludes' felons and the mentally incapable and 'includes' eligible but uncounted voters, like soldiers stationed overseas). As it turns out (according to McDonald and Popkin), turnout has not steadily declined over the past decades, as many had feared; rather, it spiked in the 1950s, then stabilized by around 1971.

Population Stability

"Population stability" measure actually captures SES (socio economic status). Homeownership and race, as well as homeownership and income are highly correlated. Geys provides a rational choice (seeking of benefits and avoidance of costs) argument for why homeowners show up at the polls more than non-homeowners. It seems a test to control for SES and triangulation of survey data would be necessary to support that this claim, and not other factors (e.g. feeling politically marginalized/ ineffectual) are at work. Further, one could argue that housing policy disproportionately impacts non-homeowners (impacting ability to purchase in the future due to government impacts on housing supply and affordability, property taxes for the provision of public schools and services).

Previous Turnout

Does previous turnout (strongly significant) reveal habitual voting or does it encapsulate numerous other factors (SES, sense of duty to vote)? Does saying "people turn out because they turned out in last election" not just beg the question of "why do people turn out"? Geys assumes this is "habitual voting", but there is insufficient proof of that. Compulsory voting seems to do the same loop: "Why do people turn out? Answer: because they have to". This simply does not help us answer the question of "why do people turn out when they don't have to". Both of these independent variables do not seem to add value in addressing the fundamental question "why do some people vote and others abstain?" While registration requirements suppressing turnout may have valuable policy implications (particularly if certain groups are inordinately affected), the inclusion of this variable does not directly get to the fundamental question either: "why would someone turn out to vote". Some could desire to turn out (show up at the polls but be turned away).

Inclusion of this variable is obviously important (or better yet use of first-differencing) for empirical reasons, but it is not clear that it has the theoretical consequences that we might like to believe.

P term: Closeness of the election

Like many studies, this one finds that turnout goes up when elections are close; and like many studies, this one concludes that this finding validates the rational model of turnout. But recall what other authors have said: Being a foot taller may make you more likely to hit your head on the moon, but you still aren't going to do it. The probability of being decisive is still nill when the election is close. [[]]