Sage Journals: Discover world-class research

Abstract

People vary in their preferred times of day for activity. Notably, as individuals age, their daily energy and attention typically peaks earlier in the day. When voting is permitted may then affect voters’ age distribution, even when holding constant the number of hours polls are open. Data from along the time-zone border in Kentucky, where poll-availability hours vary, supports this hypothesis: places where voting hours are later see higher turnout rates among younger voters and lower turnout rates among older voters. The one-hour delay in voting hours reduces older registrants’ turnout, and boosts younger registrants’, by roughly three percentage points.

Keywords

Turnout age convenience voting time zone chronobiology

Controversy rages over the consequences of extending poll hours to make voting more convenient. Academics debate how much such accessibility measures affect turnout (e.g. Gronke et al., 2008; Springer, 2012), while policymakers and judges grapple with the effects of extended hours for partisan or discriminatory aims. Such discussions have focused more on the number of hours voting is available than on the specific hours that are chosen. This is somewhat surprising, as poll-opening hours vary substantially across and within countries; changing voting hours may change accessibility for different groups. Age, in particular, is implicated. Older populations typically prefer relatively earlier schedules; the proportion of younger voters may accordingly decrease when poll-opening hours are earlier in the day.

This hypothesis is tested below using data from along the time-zone border that divides Kentucky. Eastern Time Zone areas, where the polls open and close an hour earlier relative to the sun, see significantly higher turnout among older populations and lower turnout among younger populations. This has implications for election results and for equality in ballot-box access as populations age across the rich world (Burden et al., 2017). More pointedly, though, given the divergent partisan and policy preferences of voters of different age groups, these results suggest that choice of polling hours may influence substantive political outcomes in ways readily manipulated by electoral rule-makers (Dassonneville et al., 2017).

Age and active hours

Biology provides reason to expect earlier polling hours to affect the electorate’s age distribution. One of chronobiology’s most consistent findings is that age correlates positively with “morningness”—i.e. being a morning person—and negatively with “eveningness” (Adan et al., 2012; Carrier et al., 1997; Tankova et al., 1994). Indeed, older people typically perform better on cognitive tasks earlier in the day, while younger people, if anything, reverse this pattern (Schmidt et al., 2007). This helps older people to muster the energy for voting earlier in the solar day, so that having polls open more in the mornings and less in the evenings is more congenial for older than for younger would-be voters.

Older people’s greater tendency to be (semi-)retired from the workforce further increases their flexibility to vote during the day, putting less pressure on the post-work evening hours. In jurisdictions where election days are not holidays or during weekends, working (or preparing for the workday) typically crowds younger voting-age populations’ schedules. Given the widespread preference for voting during daylight hours (Rallings et al., 2003),¹ full-time workers on a typical daytime shift will, compared with others, see relatively greater expansion of voting opportunities by having more polling hours later in the day.

These theories primarily relate to solar time: circadian rhythms and increased voting before sunset respond to the timing of daylight (“sun time”), not the numbers on a clock (“clock time”).² This is not to say that behavior ignores clock time. Potential turnout influences such as television schedules follow the clock, not the sun, and could also connect to age-related variations in turnout. For example, age correlates with more television-watching and less use of time-shifting technology like digital video recorders (Mares and Woodard, 2006). Clock time’s effects are, however, analytically separable from those of solar time, even though in most contexts sun time and clock time are essentially interchangeable. Continuing with the television example, time-zone boundaries do not necessarily correspond to television market areas, so that broadcasts at a fixed clock time may reach audiences at varying solar times.³

Previous studies have hinted that, within given polling hours, younger voters disproportionately show up relatively later in the day (Brown et al., 2006; Busch and Lieske, 1985). Of course, patterns observed within a given period of hours need not extend outside that period, and comparing turnout across different elections is complicated by the vast range of election-administration and candidate- or party-related factors that change voters’ age distribution (Blais and Rubenson, 2013; Herron and Smith, 2013; Pomante and Schraufnagel, 2015). Any single election, conversely, typically shows little variation in poll-opening hours—but the exceptional cases where election hours vary for different voters can provide empirical traction on the hypothesis linking votes earlier in the (solar) day to disproportionately older voting populations.

Kentucky’s time-zone borderland

One such case can be found in the US state of Kentucky. Different parts of the state diverge in their poll-opening times. In particular, Kentucky polling hours extend from 6:00 AM to 6:00 PM according to local clock time. As the eastern part of the state lies in the Eastern Time Zone and the western part of the state is in the Central Time Zone, voting in eastern parts of Kentucky begins and ends an hour earlier.

The region just on either side of Kentucky’s time-zone boundary therefore allows observation of the effects of solar time distinct from that of clock time, and serves as a relatively clean quasi-experimental set-up for empirical analysis (Holmes, 1998; Keele and Titiunik, 2015, 2016; Mattingly, 2017). The time-zone boundary does not follow natural barriers such as rivers or mountain ranges that might lead to stark differentiation between territories on either side of the border. Instead, it separates areas that are demographically, culturally, and economically similar, with no statistically significant differences in observable characteristics such as population density or poverty rates (though such factors are further considered as potential confounding factors below). Both sides of the border, importantly, also vote for most of the same races in any given elections: not only do all statewide votes span the time-zone border, so too do most of the region’s state-legislative seats and, to a somewhat lesser extent, those to the federal Congress.⁴ Focusing on areas near the border minimizes contextual differences across larger geographical areas, then, and provides generally similar background conditions and races at issue—while having voting hours that are, relative to the sun, 1 hour later in the western part of the state. Figure 1 maps the counties along the time-zone border in Kentucky.

Figure 1.

Kentucky counties along the time-zone border. Light gray counties are in the Eastern Time Zone; darker gray counties are in the Central Time Zone.

For most⁵ election years since 1998, Kentucky provides turnout rates among registered voters⁶ for each county in each of five age groups: those under 25 years of age; those 25 to 34; those 35 to 49; those 50 to 61; and those 62 and over. Even though casting a ballot is an individual-level choice and county-level (or, more precisely, county-age group-year-level) data is aggregated, using it raises few risks of ecological inference, as each individual registrant in a county is matched with certainty to polls with known opening hours. This demographic information about voters is released only for counties, not for units with finer geographical grain such as precincts. Though there are advantages of considering narrower or wider bands on either side of a discontinuity (Arai and Ichimura, 2016; Imbens and Kalyanaraman, 2012), this data limitation means that the analysis here considers voters within roughly 40 km of the time-zone border.⁷

The quasi-experimental set-up reduces the need for control variables, though aggregating across elections does necessitate controlling for races being contested. This is especially true in Kentucky, where state-level offices—including the governorship—are elected in odd-numbered years, separately from all the even-year, federal elections. Furthermore, different parts of Kentucky’s time-zone border region are in different congressional districts and so face differing elections for the House of Representatives. Accordingly, all models use congressional district-year fixed effects.

Results

Table 1 reports models estimating turnout as predicted by the variables discussed above, including models treating the categorical age-group variable as ordered, evenly spaced categories (in Columns I and III) and as a battery of dummy variables (in other columns). The dependent variable is percent of registrants in a given county-year-age group who vote. This variable is most often predicted using ordinary least-squares linear regression, which for ease of interpretation is retained in Table 1’s Columns I and II. However, turnout percentage is constrained to being between 0% and 100%, and theory suggests nonlinear effects: if 95% of a population was going to vote anyway, an additional factor that encourages voting will probably have a smaller effect than if only 50% would otherwise vote. This is accounted for by using generalized linear model with a binomial link function, analogous to a logistic regression model but predicting outcomes over the range of potential percentages. Columns III and IV report these generalized models.

Table 1.

Ordinary least squares (Columns I and II) and generalized linear (Columns III and IV) models of turnout among registered voters for age groups in Kentucky counties, 1998–2016. N = 980. * indicates (two-tailed) p < 0.05. All models include congressional district-year fixed effects.

	I	II	III	IV
Central Time Zone	4.40* (0.97)	2.64* (0.90)	0.21* (0.06)	0.13* (0.05)
Age group (1 = youngest, 5 = oldest)	8.26* (0.49)		0.37*(0.01)
Central Time Zone × Age group	−1.44*(0.09)		−0.07*(0.02)
Age 25–34		5.09*(0.85)		0.26*(0.02)
Age 35–49		17.50*(1.26)		0.82*(0.03)
Age 50–61		28.24*(1.65)		1.28*(0.04)
Age 62 and over		29.74*(2.08)		1.35*(0.05)
Central Time Zone × Age 25–34		−0.99*(0.18)		−0.06(0.03)
Central Time Zone × Age 35–49		−2.13*(0.33)		−0.11*(0.04)
Central Time Zone × Age 50–61		−4.00*(0.35)		−0.19*(0.05)
Central Time Zone × Age 62 and over		−5.71*(0.33)		−0.26*(0.07)

Being in the Central Time Zone, with its later voting hours relative to solar noon, associates with higher turnout for the baseline group (registered voters under the age of 25). That effect of later voting, however, interacts with age; progressively older groups see smaller turnout increases from being in the Central rather than Eastern Time Zone. Indeed, among the oldest age group, with potential voters aged at least 62, turnout is higher in the Eastern Time Zone. This interaction between age and time zone is consistent across all six models, and meets standard thresholds for statistical significance. It is also substantively appreciable, implying that the effect of the later hour of voting differs by approximately six percentage points between the youngest and oldest groups.

Figure 2 illustrates this graphically, using predictions from Table 1’s Columns II (in the left panel) and IV (in the right panel), leaving district-year as-is when calculating the predicted value. For each age group, the predicted turnout for a county on the Central Time Zone side of the border is denoted with a solid circle, while the hollow circles show predicted turnout on the Eastern Time Zone side of the border. As the figure shows, turnout is essentially indistinguishable across the time-zone boundary for those between the ages of 35 and 61. For the two youngest groups, though, Central Time Zone counties saw turnout rates three to five percentage points higher, while among the most senior voters, Eastern Time Zone counties’ turnout was approximately three percentage points higher. While these differences are dwarfed by the underlying differences in turnout across age groups (Binstock, 2000; Persson et al., 2013), they do appreciably shape electorate demographics. Indeed, the implied effect size of this 1-hour shift in voting hours is roughly comparable to that seen when changing other convenience-voting measures such as voter-identification requirements or election-day registration (Hood and Bullock, 2012; Neiheisel and Burden, 2012). By having these effects not on aggregate turnout but on specific demographic groups, though, Table 1’s results suggest larger effects on election outcomes than would broader-based shifts in turnout.

Figure 2.

Turnout (as percent of registered voters) by age in Kentucky counties along the time-zone border, as predicted by Table 1. Solid circles indicate locations in Central Time Zone; hollow circles indicate Eastern Time Zone. Bars indicate 95% confidence intervals.

Moreover, the relationship between time zone and turnout behaves as one might expect in other respects. Consider, for instance, how elections’ importance might modulate Table 1’s effects. Polling hours that are uncongenial for members of an age group might especially be expected to dissuade members of that group from casting ballots in elections perceived to be low-stakes, even if those group members think it worth the effort to turn out for higher-stakes votes. This suggests that one might expect time-zone effects on different age groups to be smaller in higher-stakes elections: in low-stakes votes, only people whose scheduling preferences coincide with local voting hours cast ballots, while in higher-stakes situations, citizens of all ages cast ballots. Results in the Appendix show this to hold true; while older voters see relatively higher turnout with earlier voting hours in all elections, the effects are larger in non-presidential years, when most Americans find voting and elections less pressing. The Appendix also shows that, within age groups, partisan orientation generally does not affect propensity to vote depending on time zone. Since there is little theoretical reason to expect Democrats to differ from Republicans in morningness, this serves as a placebo test of the hypothesized mechanism.

Are county characteristics responsible?

While proximity likely makes counties on either side of the border similar, the quasi-experimental design might be questioned. Any consistent southwest-to-northeast gradient in economic or cultural geography would correlate with the time-zone treatment. This section explores whether controlling for contextual characteristics affects the previous section’s results. Because few such variables are available at the level of the county-age group, these models’ unit of analysis is the county-year. The dependent variable accordingly becomes the difference in turnout rates between the oldest and youngest age groups; an alternative reported specification uses differences between the youngest two and oldest two groups.

Several factors potentially encourage the turnout of younger rather than older registrants. Many of these are demographic. Local preponderance of a particular age group may create greater collective-action difficulties in mobilizing that group to vote (Morrison, 2014). Conversely, locally predominant populations may have the clout to make it easier for their members to vote, as by getting polling places near a university or retirement community. Census Bureau estimates of the percentage of a county-year’s population that is younger than 25, and the percentage that is 65 or older, account for these possibilities.⁸

The Census Bureau further reports the ethnic distribution of different ethnic and racial groups in each county, in particular the shares of Blacks and Hispanics. Ethnic or racial features may affect propensity to mobilize (Holbrook et al., 2016; Uhlaner and Scola, 2016), so that any correlation between age distributions and ethnic mix—likely, given ongoing demographic shifts—might produce seemingly age-related effects that instead reflect ethnic-group effects. Measures used to control for this are the percentages of the population that are Black (alone or in conjunction with other races, regardless of Hispanic status) and Hispanic (of all races), respectively.

Economic factors may also affect different age groups’ turnout. For example, consider population density. A denser, more urban population distribution likely reduces most voters’ distance to their polling place. Such proximity to the polls varies in its effects on turnout to the extent voters of different age groups diverge in their degree of mobility (Haspel and Knotts, 2005). This is measured with population density, with a logarithmic transformation to reduce variable skew. Unemployment and local incomes also correlate with both age and electoral participation rates (Burden and Wichowsky, 2014; Choi et al., 2015); income is measured here using Census estimates of median household income deflated to 2010 dollars.

Table 2 brings these variables together in ordinary least-squares models with panel-corrected standard errors and, as with Table 1, congressional district-year fixed effects. The table’s first column predicts differences in turnout between the oldest and youngest registrant categories, while the second column aggregates the youngest two and oldest two categories.

Table 2.

Ordinary least squares models of turnout gap between older and younger age groups in Kentucky counties, 1998–2016. * indicates (two-tailed) p < 0.05. N = 196; all models include district-year fixed effects.

	Over 61 vs. Under 25	Over 49 vs. Under 35
	I	II
Central Time Zone	−1.45*(0.57)	−0.98*(0.40)
Population 25 or younger (%)	1.16*(0.57)	0.96*(0.47)
Population 65 or older (%)	0.61(0.49)	0.43(0.39)
Population Black (%)	0.06(0.13)	0.07(0.10)
Population Hispanic (%)	−0.60(0.44)	−0.24(0.30)
Ln (population per km²)	1.05(1.12)	1.47*(0.67)
Unemployment rate	0.03(0.12)	−0.04(0.12)
Median household income (1000s)	0.76*(0.09)	0.49*(0.06)

In both models, being in the Central Time Zone again associates with a smaller gap in turnout between younger and older registrants. Being on the western side of the time-zone border appears to reduce the turnout gap across age groups by one-and-a-half to two percentage points. The effect continues to attain standard benchmarks of statistical significance, and is roughly comparable in magnitude to the predicted effect of increasing population density by one standard deviation.

Conclusion

Making poll-opening hours earlier or later may change the age distribution of voters: when polls are open earlier in the solar day, turnout increases among older but decreases among younger voters. This is true despite the difference in voting hours here, shifted by a mere hour on either side of a time-zone boundary, being relatively small in the universe of plausible voting hours; the logic of how different groups react to time extends to later plausible voting hours as well. That is, although the case examined here cannot directly test the claim, presumably the difference between opening a 12-hour polling day at 6:00 AM and opening it at 8:30 AM would be somewhat larger than that examined here.

Those concerned about ensuring equal ballot access, or about predicting (or altering) electoral turnout, may accordingly find it fruitful to consider not just how many hours voting is allowed in, or on what day those hours are, but also to what time of day the hours involve. Such concerns may extend beyond young or old voters. Income may relate to casting a ballot earlier in the day, perhaps because higher-income jobs offer more flexible work schedules (Fuchs and Becker, 1968). Poorer or richer voters may then comprise a larger share of voters depending on the hours that polls are open, with potential implications for representation, electoral outcomes, and exit-poll interpretation. Future research can explore such other demographic dimensions.

It may also examine the effects of poll-availability times in other places and contexts. Kentucky is somewhat unusual in the earliness of its voting hours: 6 AM to 6 PM leaves relatively little time for those with a typical working day to vote on their way home. The age-and-turnout effects of shifting polling times may vary in size for places that have later, or shorter, voting hours. Other features of electoral administration may matter as well. Varying access to postal balloting, for example, may shape how dependent would-be voters are on election-day polling hours. So, too, might the season of the election and the polling location’s latitude and longitude; these features affect solar-time benchmarks.

The results here fit in with recent scholarly recognition that time is not fungible in its voting impact. Context can make the minutes needed to vote seem more or less costly (Gibson et al., 2013; Newman et al., 2014). As time is often central to the cost side of political participation’s cost-benefit decision calculus, taking such temporal factors seriously is likely to improve understanding of many important political behaviors. The findings also emphasize the degree to which the decision to vote may be contingent on circumstances on election day itself. The effects observed here suggest that the decision to vote for marginal voters (as opposed to habitual ones) depends on mood, attention, and energy at the time when voting is possible rather than long-term planning. This may contribute to the difficulties of forecasting who will vote (Rogers and Aida, 2014). Better understanding how circumstances like time and context affect propensity to vote can help allay these difficulties.

Footnotes

Declaration of Conflicting Interest

The author declares that there is no conflict of interest.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Supplementary Material

The supplementary files are available at http://journals.sagepub.com/doi/suppl/10.1177/2053168017720590 and the replication files are available at: .

Notes

Carnegie Corporation of New York Grant

This publication was made possible (in part) by a grant from Carnegie Corporation of New York. The statements made and views expressed are solely the responsibility of the author.

References

Adan

Archer

Hidalgo

. (2012) Circadian typology: A comprehensive review. Chronobiology International 29(9): 1153–1175.

Arai

Ichimura

(2016) Optimal bandwidth selection for the fuzzy regression discontinuity estimator. Economics Letters 141: 103–106.

Binstock

(2000) Older people and voting participation: Past and future. Gerontologist 40(1): 18–31.

Blais

Rubenson

(2013) The source of turnout decline: New values or new contexts? Comparative Political Studies 46(1): 95–117.

Brown

Docherty

Henderson

. (2006) Exit polling in Canada: An experiment. Canadian Journal of Political Science 39(4): 919–933.

Burden

Fletcher

Herd

. (2017) How different forms of health matter to political participation. Journal of Politics 79(1): 166–178.

Burden

Wichowsky

(2014) Economic discontent as a mobilizer: Unemployment and voter turnout. Journal of Politics 76(4): 887–898.

Busch

Lieske

(1985) Does time of voting affect exit poll results? Public Opinion Quarterly 49(1): 94–104.

Carrier

Monk

Buysse

. (1997) Sleep and morningness-eveningness in the “middle” years of life (20–59 y). Journal of Sleep Research 6(4): 230–237.

10.

Choi

Janiak

Villena-Roldán

(2015) Unemployment, participation and worker flows over the life cycle. Economic Journal 125(589): 1705–1733.

11.

Dassonneville

Hooghe

Lewis-Beck

(2017) Do electoral rules have an effect on electoral behaviour? An impact assessment. West European Politics 40(3): 503–515.

12.

Eisinga

te Grotenhuis

Pelzer

(2012) Weather conditions and voter turnout in Dutch national parliament elections, 1971–2010. International Journal of Biometeorology 56(4): 783–786.

13.

Fuchs

Becker

(1968) A brief report on the time of day when people vote. Public Opinion Quarterly 32(3): 437–440.

14.

Gibson

Kim

Stillman

. (2013) Time to vote? Public Choice 156(3–4): 517–536.

15.

Gronke

Galanes-Rosenbaum

Miller

. (2008) Convenience voting. Annual Review of Political Science 11: 437–455.

16.

Haspel

Knotts

(2005) Location, location, location: Precinct placement and the costs of voting. Journal of Politics 67(2): 560–573.

17.

Hassid

Watson

(2014) State of mind: Power, time zones and symbolic state centralization. Time & Society 23(2): 167–194.

18.

Herron

Smith

(2013) The effects of House Bill 1355 on voter registration in Florida. State Politics & Policy Quarterly 13(3): 279–305.

19.

Holbrook

Sterrett

Johnson

. (2016) Racial disparities in political participation across issues: The role of issue-specific motivators. Political Behavior 38(1): 1–32.

20.

Holmes

(1998) The effect of state policies on the location of manufacturing: Evidence from state borders. Journal of Political Economy 106(4): 667–705.

21.

Hood

III Bullock

III (2012) Much ado about nothing? An empirical assessment of the Georgia voter identification statute. State Politics & Policy Quarterly 12(4): 394–414.

22.

Imbens

Kalyanaraman

(2012) Optimal bandwidth choice for the regression discontinuity estimator. Review of Economic Studies 79(3): 933–959.

23.

Keele

Titiunik

(2015) Geographic boundaries as regression discontinuities. Political Analysis 23(1): 127–155.

24.

Keele

Titiunik

(2016) Natural experiments based on geography. Political Science Research and Methods 4(1): 65–95.

25.

Lamare

(2013) Mobilization and voter turnout: Should canvassers worry about the weather? PS: Political Science & Politics 46(3): 580–586.

26.

Mares

M-L

Woodard

(2006) In search of the older audience: Adult age differences in television viewing. Journal of Broadcasting & Electronic Media 50(4): 595–614.

27.

Mattingly

(2017) Colonial legacies and state institutions in China: Evidence from a natural experiment. Comparative Political Studies 50(4): 434–463.

28.

Mishra

Ramgopal

(2013) Field studies on human thermal comfort: An overview. Building and Environment 64: 94–106.

29.

Morrison

(2014) Quantifying the effect of age structure on voter registration. Social Science Quarterly 95(1): 286–294.

30.

Neiheisel

Burden

(2012) The impact of Election Day registration on voter turnout and election outcomes. American Politics Research 40(4): 636–664.

31.

Newman

Johnson

Lown

(2014) The “daily grind”: Work, commuting, and their impact on political participation. American Politics Research 42(1): 141–170.

32.

Persson

Wass

Oscarsson

(2013) The generational effect in turnout in the Swedish general elections, 1960–2010. Scandinavian Political Studies 36(3): 249–269.

33.

Pomante

Schraufnagel

(2015) Candidate age and youth voter turnout. American Politics Research 43(3): 479–503.

34.

Rallings

Thrasher

Borisyuk

(2003) Seasonal factors, voter fatigue and the costs of voting. Electoral Studies 22(1): 65–79.

35.

Rogers

Aida

(2014) Vote self-prediction hardly predicts who will vote, and is (misleadingly) unbiased. American Politics Research 42(3): 503–528.

36.

Schmidt

Collette

Cajochen

. (2007) A time to think: Circadian rhythms in human cognition. Cognitive Neuropsychology 24(7): 755–789.

37.

Springer

(2012) State electoral institutions and voter turnout in presidential elections, 1920–2000. State Politics & Policy Quarterly 12(3): 252–283.

38.

Tankova

Adan

Buela-Casal

(1994) Circadian typology and individual differences: A review. Personality and Individual Differences 16(5): 671–684.

39.

Uhlaner

Scola

(2016) Collective representation as a mobilizer: Race/ethnicity, gender, and their intersections at the state level. State Politics & Policy Quarterly 16(2): 227–263.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.14 MB

Youthful hours: Shifting poll-opening times manipulates voter demographics

Abstract

Keywords

Age and active hours

Kentucky’s time-zone borderland

Results

Are county characteristics responsible?

Conclusion

Footnotes

Declaration of Conflicting Interest

Funding

Supplementary Material

Notes

Carnegie Corporation of New York Grant

References

Supplementary Material