Potential Impacts of Extended Advance Voting on Voter Turnout
The question is whether turnout is likely to increase if there is greater opportunity for advance voting and if Canadians have two similar polling days instead of one, especially if the first day is a Sunday.
The first response that we can give with some confidence is this: this is very unlikely to reduce turnout. The logic is simple: the easier it is to vote, the more likely people are to vote. The more difficult questions pertain to the probable magnitude of the impact (is it going to boost turnout a lot or just a little?) and to the comparative effects of various methods through which electoral participation can be facilitated.
Before reviewing the cross-national literature, it is useful to examine how responsive people are to the cost (or difficulty) of voting. Using individual survey data, Blais (2000) measures costs through three questions: how much time respondents think it will take them to vote, how easy or difficult they feel is it to go and vote, and how easy or difficult they find it to get information and decide how to vote. Blais finds that the perceived cost of voting is very small. He also shows that costs have a statistically significant effect on turnout: the easier it is to vote, the lower the propensity to abstain. But the impact is weak. He concludes that "marginal increases in (costs) reduce the propensity to vote only marginally" (Blais 2000, 91). His results suggest that measures designed to facilitate the vote would help, but only a little.
Let us now consider the more typical cross-national evidence. A series of studies focuses directly on administrative measures for facilitating the vote; their goal is to determine whether, everything else being equal, turnout tends to be higher in countries that have adopted such measures.
Franklin has most extensively examined the impact of voting facilities. In a first piece (Franklin 1996), he reports a positive effect for the presence of such facilitiesFootnote 5 as well as for Sunday voting; the estimated impact, four and six percentage points respectively, is substantial. Franklin also finds a counterintuitive negative effect of the number of voting days.Footnote 6 In a second study published six years later, Franklin (2002) again finds that these two factors give a substantial boost (six percentage points) to turnout. Number of voting days is not included this time.
The third and last study is Franklin's book (2004), where the emphasis is on explaining turnout variation from the previous election. Franklin finds no significant effect this time for either voting facilitiesFootnote 7 or weekend voting.
In short, Franklin's initial results suggested relatively strong effects for voting facilities as well as for Sunday voting, but his later work reports no significant impact.
Norris (2002) also looks at the impact of voting facilities on turnout. Her analysis covers a wider range of countries, including non-democratic ones. She finds a positive effect for rest-day voting, a negative impact for proxy voting and the number of days that polling stations are open, and no effect for other special voting facilities.
Blais et al. (2003) conducted a study covering 151 elections in 61 democratic elections for Elections Canada. The authors examine the impact of holiday voting and "ease of voting." The latter indicates whether it is possible to vote by mail, in advance or by proxy. It corresponds to the proportion of options available in a given country (the variable equals 0.66 in Canada, where mail and advance voting exist but there is no proxy voting). The study reports no significant effect for holiday voting, but a relatively strong impact for ease of voting. Everything else being equal, turnout would be 11 points higher in a country with all three options (Sweden is the only such case), compared with countries with none available (these countries being mostly in South America).
Finally, Rose (2004) examines turnout in the (then) 15 European Union member countries in both national (1945–2002) and European Parliament (1979–1999) elections. He looks at the effect of holding the election on a day of rest and finds a strong positive impact (4 percentage points for national elections and 11 points for the European Parliament elections).
The findings of studies based on cross-national comparisons are not very consistent. In the case of voting facilities, Franklin's initial studies and Blais et al. report that they boost turnout, while Franklin (in his most recent research) and Norris find no effect. In the case of Sunday or holiday voting, Franklin (early work), Norris and Rose indicate that it contributes to a higher turnout, while Franklin's most recent book and Blais et al. report no impact. As for two regular days of voting, some studies suggest a negative effect, which does not seem very credible. Such ambiguous findings make it difficult to reach definitive conclusions. They suggest, at the very least, that these administrative measures are unlikely to have a big impact. At the same time, there is rather strong evidence that convenience helps a little, and so it seems plausible that measures that make it easier for voters to vote at a more convenient time should slightly increase turnout.
Another series of studies have looked at the consequences of advance or postal voting facilities adopted in certain American states. Three studies (Karp and Banducci 2000, Berinski, Burns and Traugott 2001, Gronke and Miller 2007) ascertain the impact of postal voting in the United States, specifically in the state of Oregon, where all elections are now conducted by postal voting alone. Gronke and Miller also study the state of Washington, where postal voting is allowed in some counties but not others. The common conclusion is that the introduction of postal voting increased turnout by making it easier for people who generally, but not always, vote to cast ballots. However, it did little to cause habitual non-voters to cast ballots. The effect, then, is positive but modest. Gronke and Miller report similar results from Washington state, where counties that conduct elections by mail experience an average turnout 4.5 percentage points higher than counties that do not.
Oliver (1996) examines the effects of the provision of advance voting opportunities on turnout. He finds that advance voting is higher in states with liberal voting laws, but only if parties actively mobilize citizens to take advantage of these increased opportunities to vote. A recent study by Gronke et al. (n.d.) looks at the consequences of advance voting in 50 states over a 24-year period. Their study reports a positive effect of about five percentage points for postal voting, but no significant effect for in-person early voting. Again, the general theme is that these measures increase turnout only among specific subgroups or under some conditions, and so the overall impact is limited.
Scholars have also analyzed the impact of changes in American electoral administration since the introduction of the Help American Vote Act in 2002. Mycoff et al. (2007) find that there is no relationship between the strictness of identification requirements and turnout at either the aggregate or the individual level.
There are also a number of interesting studies on the relationship between convenience and voting. For example, scholars have studied the impact of weather on turnout. The most systematic study was published by Gomez et al. (2007). They examine turnout in over 3,000 counties for each American presidential election from 1948 to 2000, and they relate it to estimates of rain and snow on the day of the election. The authors show a statistically significant effect for both rain and snow. The impact, however, is quite small. They find that one additional inch of rain decreases turnout by almost one percentage point; the effect of one additional inch of snow is almost 0.5 percentage point. A similar result is reported by Ben Lakhdar and Dubois (2006), who look at the effect of the climate in 43 departments for five French legislative elections.
Finally, there is some research on the impact of distance from polling stations on turnout. Gimpel and Schuknecht (2003) examine turnout in 363 precincts from three suburban Maryland counties in the 2000 presidential election. They relate turnout to distance between polling sites and voter district centroids. They find a statistically significant effect, controlling for a host of other factors: "a five mile increase in accessibility increases turnout by an average of 1.74 percent (sic)."
A second study by Dyck and Gimpel (2005) is even more interesting, because it is based on individual-level data. The authors look at distance between voter residences, their precinct and nearest early voting sites in Clark County, Nevada, for the 2002 mid-term election. They estimate the impact of distance on the propensity to abstain or to vote at the precinct, early, or by mail. They find that distance matters. But the precise effects are somewhat surprising: the greater the distance from precinct site, the greater the propensity to vote by mail, and the greater the distance from early voting sites, the greater the propensity to vote at the precinct. Distance from voting sites primarily affects which form of voting is chosen (precinct, early or mail). There is an impact on overall turnout, but it is modest; living one standard deviation beyond mean distance from a precinct site (about 1.75 miles) increases abstention by 1.3 percentage points.
Finally, a recent study by Baker (2007) examined the effect of the consolidation of polling stations in Ohio and Kansas. While she expected the consolidation of polling stations to reduce turnout, she found no effect.
Research on the impact of weather and distance thus comes up with consistent findings. Nicer weather and shorter distance from the polling stations contributes to a higher turnout. But the effect appears to be small. Taken together, all these findings suggest that measures that make it easier to vote can be expected to increase voter turnout, but only slightly.
Return to source of Footnote 5 Franklin refers to "postal voting," but the data he uses (Katz 1997, Table 13.2) concern postal voting, advanced voting and other "special" provisions. The term "voting facilities" would better reflect what is being measured.
Return to source of Footnote 6 We suspect that this counterintuitive finding is due to the fact that India, which has a very low turnout, is construed as having five voting days. The fact is that there are different voting days for different regions of the country, but an elector may vote only on a given day. Therefore, India should be considered as having one voting day.
Return to source of Footnote 7 Franklin is now referring to "absentee" voting, but the data still cover postal and advanced voting plus other special provisions.