Home is Where the Vote Is: A Research Note on the Effects of Changes in North Dakota Voter Identification Laws on College Student Turnout in the 2014 Elections

Since 2005, state legislatures across the nation have passed a series of stringent voter identification laws in the expressed hope of preventing voter fraud. However, some scholars argue that these voter id laws negatively impact people who are less likely to have a photo id, such as racial minorities and young people. This study uses a survey of college students in the North Dakota University System to examine whether changes to North Dakota voter id laws in 2013 had any influence over the ability of these students to successfully cast a vote in the 2014 midterm elections. It finds that the new law did indeed have an impact, with the survey results indicating that several hundred students were unable to vote due to issues related to voter id. Beginning in 2005, state legislatures began passing increasingly stringent voter ID laws (NCSL 2016). Georgia and Indiana were the first states to implement these laws, in 2008, after the U.S. Supreme Court deemed them constitutional. The pace of adoption picked up in the 2010’s. By 2016, 17 states require voters to produce a photo ID in order to be able to vote. North Dakota’s voter ID law, passed in 2013, is one of the strictest in the nation, given the narrow range of IDs that are accepted and the absence of a provisional ballot option. 1 Nicholas Bauroth is an Associate Professor of Political Science, North Dakota State University, 1616 12 th Ave N, Fargo, ND 58108 (Nicholas.bauroth@ndsu.edu). Kjersten Nelson is an Assistant Professor of Political Science, North Dakota State University, 1616 12 Ave N, Fargo, ND 58108 (Kjersten.nelson@ndsu.edu). The authors would like to thank the anonymous reviewers and the Upper Midwest Regional Center on Public Policy for grant support. Bauroth and Nelson Vol. 11, Issue 1 (2016) 2 The national debate over the usefulness of voter ID laws has pitted arguments about preventing voter fraud (e.g., Rousu 2014) against concerns that the laws will discourage voting amongst eligible voters (e.g., Weiser 2014). Those who raise alarm about these kinds of legislation assert that the laws will disproportionately affect populations who are less likely to have a photo ID: in particular, the elderly, racial minorities, the poor, and young people (e.g., Barreto et al 2008; Government Accountability Office 2014). Empirical studies are beginning to unravel the effects of more stringent voter ID laws, though the findings are mixed. For example, some of the first studies, which look at effects in the entire voting population, find little to no effects on voter turnout (e.g., Ansolabehere 2009; Mycoff, Wagner, and Wilson 2009). Others have documented that more stringent non-voter-ID laws reduced turnout amongst African-American and Hispanic voters (Vercelloti and Anderson 2006), while Alvarez, et al (2008) found a disproportionate impact on poorer voters (but not race-based effects). However, studies that were conducted after the most stringent ID requirements were implemented, and in a general election setting, come to different conclusions. Hood and Bullock (2012), for example, find a suppression effect in the 2008 general election, though their findings do not suggest the effect was felt disproportionately amongst ethnic and racial minorities. Ongoing studies that investigate other aspects of more stringent voter ID requirements add to evidence that there are disproportionate effects based on race. Herron and Smith (2014), for example, find race-based effects when early voting is curtailed. Others find race-based differentials in how voter ID laws are applied, with racial minorities being asked to produce photo IDs at higher rates than white voters (Atkeson, et al 2010; Rogowski and Cohen 2014). The most recent investigation of voter turnout and voter ID laws aligns with these related findings. Looking at a general election that took place after many of the more stringent voter ID laws were in effect (e.g., the 2012 general election) and using verified voter turnout records, Hajnal et al (2016) find disproportionate effects of voter ID laws amongst Hispanic, Black, and mixed race voters, as well as Democratic voters. The existing literature, then, has primarily focused on the impacts of these voter ID laws on racial minorities. We have less evidence related to the other groups that were predicted to have greater difficulty voting: the elderly and young individuals. Rogowski and Cohen (2014) do show intersectional effects of these legislative efforts. For example, Black and Latino youth report being asked to show voter ID at higher rates than white youth. Black and Latino youth also indicate that the stringent voter ID laws were more likely to keep them from voting than white youth (17.3 percent versus 4.7 percent, respectively). This study picks up here, assessing the impact of North Dakota’s new voter ID law on a subset of the population that is expected to be particularly affected: North Dakota college students. Given their age (which is correlated with the likelihood that they are new voters), and the likelihood that they move often (and reside somewhere that is not a permanent address), the requirement that voters produce a photo ID with a current address is, at least in theory, particularly likely to keep this population from voting. And, given that the age of most of this group of voters is the time when voting habits are being established, it is particularly important to understand the impacts of voter ID laws on this subpopulation. The Online Journal of Rural Research and Policy Vol. 11 Issue 1 (2016) 3 Changes to Voter Identification Laws in North Dakota North Dakota is still the only state in the union without voter registration (Grossman 2014). Since 2003, though, the state has required its citizens to show valid identification before they can vote (Lucin 2014). Initially, someone without the proper form could sign an affidavit stating that they were an eligible voter, which the Secretary of State’s office would then verify after the election (Hageman 2014). This was a widely-used option with more than ten thousand North Dakota voters signing affidavits during the 2012 general election (Thompson 2013). However, voter affidavits were eliminated by the state legislature in 2013 over concerns about fraud. Under the new law, voters have to show identification that includes their name, current residential address, and date of birth (North Dakota Century Code 2015). Valid forms include a driver’s license, a non-driver ID card issued by the state Department of Transportation, or a university-issued student identification certificate. In addition, state law requires citizens to live in a precinct for thirty days prior to an election (North Dakota Century Code 2015). Citizens who fail to update their identification to reflect a change in address can only vote in their old precinct. Some observers feared that changes to the voter identification laws would have a negative impact on college students (Michael 2013). Many students have drivers licenses with their parents’ residence listed as the current address rather than their current campus address. In the months preceding the 2014 general elections, the Secretary of State’s office (2014a) and university administrators distributed information to students regarding the need to update their student identification certificate to reflect any change in residency. Despite these efforts, many students remained unaware of voter identification requirements, thereby missing the deadline for changing their addresses. In the aftermath of the 2014 elections, the local media reported anecdotal accounts of students being prevented from voting due to confusion over the thirty-day residential requirements (Hageman 2014). However, the full impact was not quantified. Here, we report on a survey used to assess the impacts of this change in voter ID requirements. Thus, the findings add to the discussion over the extent to which changes in voter identification requirements across the United States have impacted election turnout, particularly amongst populations that were expected to be affected most (Vercellotti and Anderson 2009). Results of a Survey of College Students in North Dakota Shortly after the 2014 midterm elections, researchers invited students at the eleven North Dakota University System institutions to participate in an online survey regarding their voting habits and experiences. Officials at each campus provided current lists of registered students, including their emails. We sent an invitation to participate in a survey related to “civic engagement” to each email address, which included a link to the survey (conducted through Qualtrics), as well as an offer to be included in a drawing for several gift certificates. Responses were collected between November 12, 2014 and January 26, 2015. Of the 48,920 students invited to participate, 1,797 completed the survey for a total response rate of 3.7 percent. Of those, 1,613 answered questions regarding their ability to vote in the most recent election. The University of Bauroth and Nelson Vol. 11, Issue 1 (2016) 4 North Dakota had the greatest number of respondents at 800, followed by North Dakota State University with 479. Survey respondents identified as: 59.1 percent female and 40.9 percent male; 91.1 percent exclusively white; and 20.8 percent freshmen, 16.5 percent sophomores, 16.0 percent juniors, 24.7 percent seniors, and 21.9 percent graduate students. The North Dakota University System (NDUS) reported that 47,660 students were enrolled in these schools in Fall 2014. NDUS identified these students as: 54.0 percent female and 41.0 male; 79.8 percent exclusively white; and 30.7 percent freshmen, 22.0 percent sophomore, 12.4 percent juniors, 21.2 percent seniors, and 13.6 percent graduate students (NDUS 2014; Weber 2016). Students were asked, ‘Did you vote in the 2014 election?’ with three response options: ‘Yes,’ ‘No,’ and ‘I attempted to vote but was unable to.’ Those who selected the third option were then requested to give an open-ended description of what prevented them from voting. We subsequently categorized these explanations as (1) issues related to residential address; (2) problems with absentee ballots; and (3) miscellaneous issues. Miscellaneous issues included such things as: misperceptions regarding polling hours; confusion over the actual date of the election; and involvement in a study abroad program. It should be noted that the survey results are not based upon a random sample. Voter turnout is broken down by type of institution in Table 1. As can be seen, 64.0 percent of respondents voted successfully in the 2014 election, 31.1 percent did not attempt to vote, and 4.9 percent were unsuccessful in their efforts to vote. In contrast, the North Dakota Secretary of State’s office reported that 46.8 percent of all eligible voters across the state participated in the 2014 elections (2014b), though turnout varied considerably by county and by age group. Voter turnout by type of institution is fairly consistent across the survey with the small four-year universities, North Dakota State University, and University of North Dakota students at or just above 64.3 percent. Turnout at the two-year colleges, though, was 57.9 percent. Finally, the percent of respondents unsuccessful in their efforts to vote varied by institution as well, ranging from 4.0 percent for small four-year universities to 5.8 percent for the two-year colleges. Table 1. Full Survey Results by Type of Institution Voted Did Not Vote Tried to vote (unsuccessful) Total Two-year Colleges (N=5) 99 (57.9%) 52 (30.4%) 10 (5.8%) 171 (100%) Small Four-Year Universities (N=4) 112 (64.7%) 54 (31.2%) 7 (4.0%) 173 (100%) North Dakota State University 308 (64.3%) 146 (30.5%) 25 (5.2%) 479 (100%) University of North Dakota 514 (64.3%) 249 (31.1%) 37 (4.6%) 800 (100%) Total 1,033 (64.0%) 501 (31.1%) 79 (4.9%) 1,613 (100%) The Online Journal of Rural Research and Policy Vol. 11 Issue 1 (2016) 5 The ability to successfully cast a ballot by type of institution is presented in Table 2. As the bottom row shows, 92.9 percent of respondents seeking to vote were successful in their efforts, 3.2 percent were prevented from voting due to residency requirements, 1.5 percent could not vote due to issues with absentee ballots, and 2.3 percent were unable to vote for miscellaneous reasons. Respondents from two-year institutions had the greatest problems with residency requirements with 5.5 percent prevented from voting. However, 3.6 percent of North Dakota State University respondents and 2.9 percent of University of North Dakota respondents were also unable to participate due to changes in the voter identification requirements. Table 2. Student Voting Attempts by Type of Institution Successfully Voted Unsuccessful due to address problems Unsuccessful due to absentee voter problems Unsuccessful due to miscellaneous issues Total Two-year Colleges (N=5) 99 (90.8%) 6 (5.5%) 1 (0.9%) 3 (2.8) 109 (100%) Small Four-Year Universities (N=4) 112 (94.1%) 2 (1.7%) 3 (2.5%) 2 (1.7%) 119 (100%) North Dakota State University 308 (92.5%) 12 (3.6%) 4 (1.2%) 9 (2.7%) 333 (100%) University of North Dakota 514 (93.3%) 16 (2.9%) 9 (1.6%) 12 (2.2%) 551 (100%) Total 1,033 (92.9%) 36 (3.2%) 17 (1.5%) 26 (2.3%) 1,112 (100.0%) It should be noted that the thirty-day residency requirement results for two-year institutions was largely driven by respondents from Bismarck State College. Respondents from Dakota College at Bottineau, Lake Region State College, and Williston State College did not report any issues with either the residency requirements or absentee voter procedures. Cross tabulations are presented in Table 3, which shows what percentage of each demographic group reported attempting to vote but failing. We analyze these numbers by comparing the distribution of demographics amongst those who report having tried to vote but failed with the demographic distribution of the entire sample. In most cases, the distribution of those who failed to vote resembles the full sample. However, men were statistically significantly overrepresented in this group of voters (and, likewise, women were underrepresented) (p=.05). Men comprised 41 percent of the overall sample but 55 percent of those who reported not being able to vote. In addition, individuals who reported a medium interest in campaign were also overrepresented in the group who tried but failed to vote, as compared to the overall sample (p<.05) (58 percent versus 41 percent). It is possible that the overrepresentation of men is related to gender differences in students who move away from their permanent addresses for school. For example, in the full sample, 29 percent of men who voted reported voting at a non-permanent school address, compared to 25 percent of women. Unfortunately, we did not ask students who had been denied voting whether they were voting at their permanent address, but future surveys could do Bauroth and Nelson Vol. 11, Issue 1 (2016) 6 so to understand the mechanism behind this gender difference. Otherwise, the lack of significant difference between the full sample and this subgroup is some comfort that new voter ID requirements did not sway this election’s outcome. Table 3. Cross-Tabulations Group Percentage of the group that reported trying to vote but not being able to (Frequency) Distribution of the full sample -% (Frequency) Distribution of VoteAttempting Group -% (Frequency) All respondents 4.9 (79) Gender Men 7.0 (41) 41 (611) 55* (41) Women 4.0 (34) 59 (882) 44* (34) Year in School First-Year 3.0 (9) 21 (312) 12 (9) Sophomore 7.0 (16) 17 (248) 21 (16) Junior 8.0 (18) 16 (240) 24 (18) Senior 6.0 (21) 25 (371) 28 (21) Graduate 4.0 (11) 22 (329) 15 (11) Party Democrat 5.0 (26) 37 (519) 35 (26) Independent 5.9 (11) 14 (189) 15 (11) Republican 5.5 (38) 50 (697) 51 (38) Ideology Liberal 4.9 (22) 33 (472) 30 (22) Moderate 3.9 (9) 17 (246) 12 (9) Conservative 6.1 (43) 50 (726) 58 (43) Political Knowledge Low 3.3 (14) 32 (568) 18 (14) Medium 5.6 (35) 36 (649) 44* (35) High 5.3 (30) 32 (580) 38 (30) Interest in the Campaign Low 3.5 (18) 32 (539) 23 (18) Medium 7.0 (46) 41 (692) 58 (46) High 3.4 (15) 27 (451) 19 (15) *Representation significantly differs from the full sample distribution (p≤.05). Any addition within categories that does not sum to 100 is due to rounding. The Online Journal of Rural Research and Policy Vol. 11 Issue 1 (2016) 7 To gauge the full impact of changes in residency requirements on student voting, the findings of this study were extrapolated to the general student population. However, it is possible that survey respondents turned out to vote at a higher rate than the general student population. Consequently, the study assumes that actual student turnout rates reflected those of the county in which their campus is located. For example, if the Secretary of State’s office pegged Cass County voter turnout at 44.4 percent, the study assumes voter turnout for North Dakota State University students was the same percentage. To keep the estimates conservative, the extrapolations were done only for those institutions at which residency requirements proved to be a serious issue. Institutions where residency requirements did not have an apparent impact are excluded. Extrapolating the findings proceeded as follows: Bismarck State College = (5,352 students) X (0.5498 turnout rate for Burleigh County) X (0.077 unable to vote due to residency problems) = 226.6 Minot State University = (3,217 students) X (0.3892 turnout rate for Ward County) X (0.021 unable to vote due to residency problems) = 26.3 North Dakota State College of Science = (3,024 students) X (0.4724 turnout rate for Richland County) X (0.038 unable to vote due to residency problems) = 54.3 North Dakota State University = (13,332 students) X (0.4440 turnout rate for Cass County) X (0.036 unable to vote due to residency problems) = 213.1 University of North Dakota = (15,102 students) X (0.3849 turnout rate for Grand Forks County) X (0.029 unable to vote due to residency problems) = 168.6 The extrapolated total computes as 688.9 students who attempted to vote but were unsuccessful due to residency issues. It is noteworthy that the residency requirements were an issue for respondents from the largest universities. Indeed, the universities in Bismarck, Fargo, and Grand Forks have many students who moved from their home town to attend school. These transplanted students often list their parents’ address as their mailing address, making them vulnerable to exclusion. Students at the other eight institutions are less likely to have moved from their home town to attend school. Consequently, they live either with their parents or only a short distance away, making them considerably less vulnerable to exclusion.

The national debate over the usefulness of voter ID laws has pitted arguments about preventing voter fraud (e.g., Rousu 2014) against concerns that the laws will discourage voting amongst eligible voters (e.g., Weiser 2014). Those who raise alarm about these kinds of legislation assert that the laws will disproportionately affect populations who are less likely to have a photo ID: in particular, the elderly, racial minorities, the poor, and young people (e.g., Barreto et al 2008;Government Accountability Office 2014). Empirical studies are beginning to unravel the effects of more stringent voter ID laws, though the findings are mixed. For example, some of the first studies, which look at effects in the entire voting population, find little to no effects on voter turnout (e.g., Ansolabehere 2009;Mycoff, Wagner, and Wilson 2009). Others have documented that more stringent non-voter-ID laws reduced turnout amongst African-American and Hispanic voters (Vercelloti and Anderson 2006), while Alvarez, et al (2008) found a disproportionate impact on poorer voters (but not race-based effects).
However, studies that were conducted after the most stringent ID requirements were implemented, and in a general election setting, come to different conclusions. Hood and Bullock (2012), for example, find a suppression effect in the 2008 general election, though their findings do not suggest the effect was felt disproportionately amongst ethnic and racial minorities. Ongoing studies that investigate other aspects of more stringent voter ID requirements add to evidence that there are disproportionate effects based on race. Herron and Smith (2014), for example, find race-based effects when early voting is curtailed. Others find race-based differentials in how voter ID laws are applied, with racial minorities being asked to produce photo IDs at higher rates than white voters (Atkeson, et al 2010;Rogowski and Cohen 2014). The most recent investigation of voter turnout and voter ID laws aligns with these related findings. Looking at a general election that took place after many of the more stringent voter ID laws were in effect (e.g., the 2012 general election) and using verified voter turnout records, Hajnal et al (2016) find disproportionate effects of voter ID laws amongst Hispanic, Black, and mixed race voters, as well as Democratic voters.
The existing literature, then, has primarily focused on the impacts of these voter ID laws on racial minorities. We have less evidence related to the other groups that were predicted to have greater difficulty voting: the elderly and young individuals. Rogowski and Cohen (2014) do show intersectional effects of these legislative efforts. For example, Black and Latino youth report being asked to show voter ID at higher rates than white youth. Black and Latino youth also indicate that the stringent voter ID laws were more likely to keep them from voting than white youth (17.3 percent versus 4.7 percent, respectively). This study picks up here, assessing the impact of North Dakota's new voter ID law on a subset of the population that is expected to be particularly affected: North Dakota college students. Given their age (which is correlated with the likelihood that they are new voters), and the likelihood that they move often (and reside somewhere that is not a permanent address), the requirement that voters produce a photo ID with a current address is, at least in theory, particularly likely to keep this population from voting. And, given that the age of most of this group of voters is the time when voting habits are being established, it is particularly important to understand the impacts of voter ID laws on this subpopulation.

Changes to Voter Identification Laws in North Dakota
North Dakota is still the only state in the union without voter registration (Grossman 2014). Since 2003, though, the state has required its citizens to show valid identification before they can vote (Lucin 2014). Initially, someone without the proper form could sign an affidavit stating that they were an eligible voter, which the Secretary of State's office would then verify after the election (Hageman 2014). This was a widely-used option with more than ten thousand North Dakota voters signing affidavits during the 2012 general election (Thompson 2013). However, voter affidavits were eliminated by the state legislature in 2013 over concerns about fraud.
Under the new law, voters have to show identification that includes their name, current residential address, and date of birth (North Dakota Century Code 2015). Valid forms include a driver's license, a non-driver ID card issued by the state Department of Transportation, or a university-issued student identification certificate. In addition, state law requires citizens to live in a precinct for thirty days prior to an election (North Dakota Century Code 2015). Citizens who fail to update their identification to reflect a change in address can only vote in their old precinct.
Some observers feared that changes to the voter identification laws would have a negative impact on college students (Michael 2013). Many students have drivers licenses with their parents' residence listed as the current address rather than their current campus address. In the months preceding the 2014 general elections, the Secretary of State's office (2014a) and university administrators distributed information to students regarding the need to update their student identification certificate to reflect any change in residency. Despite these efforts, many students remained unaware of voter identification requirements, thereby missing the deadline for changing their addresses.
In the aftermath of the 2014 elections, the local media reported anecdotal accounts of students being prevented from voting due to confusion over the thirty-day residential requirements (Hageman 2014). However, the full impact was not quantified. Here, we report on a survey used to assess the impacts of this change in voter ID requirements. Thus, the findings add to the discussion over the extent to which changes in voter identification requirements across the United States have impacted election turnout, particularly amongst populations that were expected to be affected most (Vercellotti and Anderson 2009).

Results of a Survey of College Students in North Dakota
Shortly after the 2014 midterm elections, researchers invited students at the eleven North Dakota University System institutions to participate in an online survey regarding their voting habits and experiences. i Officials at each campus provided current lists of registered students, including their emails. We sent an invitation to participate in a survey related to "civic engagement" to each email address, which included a link to the survey (conducted through Qualtrics), as well as an offer to be included in a drawing for several gift certificates. Responses were collected between November 12, 2014 and January 26, 2015. ii Of the 48,920 students invited to participate, 1,797 completed the survey for a total response rate of 3.7 percent. Of those, 1,613 answered questions regarding their ability to vote in the most recent election. The University of North Dakota had the greatest number of respondents at 800, followed by North Dakota State University with 479.
Survey respondents identified as: 59.1 percent female and 40.9 percent male; 91.1 percent exclusively white; and 20.8 percent freshmen, 16.5 percent sophomores, 16.0 percent juniors, 24.7 percent seniors, and 21.9 percent graduate students. The North Dakota University System (NDUS) reported that 47,660 students were enrolled in these schools in Fall 2014. NDUS identified these students as: 54.0 percent female and 41.0 male; 79.8 percent exclusively white; and 30.7 percent freshmen, 22.0 percent sophomore, 12.4 percent juniors, 21.2 percent seniors, and 13.6 percent graduate students (NDUS 2014; Weber 2016).
Students were asked, 'Did you vote in the 2014 election?' with three response options: 'Yes,' 'No,' and 'I attempted to vote but was unable to.' Those who selected the third option were then requested to give an open-ended description of what prevented them from voting. We subsequently categorized these explanations as (1) issues related to residential address; (2) problems with absentee ballots; and (3) miscellaneous issues. Miscellaneous issues included such things as: misperceptions regarding polling hours; confusion over the actual date of the election; and involvement in a study abroad program. It should be noted that the survey results are not based upon a random sample.
Voter turnout is broken down by type of institution in Table 1. As can be seen, 64.0 percent of respondents voted successfully in the 2014 election, 31.1 percent did not attempt to vote, and 4.9 percent were unsuccessful in their efforts to vote. In contrast, the North Dakota Secretary of State's office reported that 46.8 percent of all eligible voters across the state participated in the 2014 elections (2014b), though turnout varied considerably by county and by age group. Voter turnout by type of institution is fairly consistent across the survey with the small four-year universities, North Dakota State University, and University of North Dakota students at or just above 64.3 percent. Turnout at the two-year colleges, though, was 57.9 percent. Finally, the percent of respondents unsuccessful in their efforts to vote varied by institution as well, ranging from 4.0 percent for small four-year universities to 5.8 percent for the two-year colleges. The ability to successfully cast a ballot by type of institution is presented in Table 2. As the bottom row shows, 92.9 percent of respondents seeking to vote were successful in their efforts, 3.2 percent were prevented from voting due to residency requirements, 1.5 percent could not vote due to issues with absentee ballots, and 2.3 percent were unable to vote for miscellaneous reasons. Respondents from two-year institutions had the greatest problems with residency requirements with 5.5 percent prevented from voting. However, 3.6 percent of North Dakota State University respondents and 2.9 percent of University of North Dakota respondents were also unable to participate due to changes in the voter identification requirements. It should be noted that the thirty-day residency requirement results for two-year institutions was largely driven by respondents from Bismarck State College. Respondents from Dakota College at Bottineau, Lake Region State College, and Williston State College did not report any issues with either the residency requirements or absentee voter procedures. Table 3, which shows what percentage of each demographic group reported attempting to vote but failing. We analyze these numbers by comparing the distribution of demographics amongst those who report having tried to vote but failed with the demographic distribution of the entire sample. In most cases, the distribution of those who failed to vote resembles the full sample. However, men were statistically significantly overrepresented in this group of voters (and, likewise, women were underrepresented) (p=.05). Men comprised 41 percent of the overall sample but 55 percent of those who reported not being able to vote. In addition, individuals who reported a medium interest in campaign were also overrepresented in the group who tried but failed to vote, as compared to the overall sample (p<.05) (58 percent versus 41 percent). iii It is possible that the overrepresentation of men is related to gender differences in students who move away from their permanent addresses for school. For example, in the full sample, 29 percent of men who voted reported voting at a non-permanent school address, compared to 25 percent of women. Unfortunately, we did not ask students who had been denied voting whether they were voting at their permanent address, but future surveys could do so to understand the mechanism behind this gender difference. Otherwise, the lack of significant difference between the full sample and this subgroup is some comfort that new voter ID requirements did not sway this election's outcome. *Representation significantly differs from the full sample distribution (p≤.05). Any addition within categories that does not sum to 100 is due to rounding.

Cross tabulations are presented in
To gauge the full impact of changes in residency requirements on student voting, the findings of this study were extrapolated to the general student population. However, it is possible that survey respondents turned out to vote at a higher rate than the general student population. Consequently, the study assumes that actual student turnout rates reflected those of the county in which their campus is located. For example, if the Secretary of State's office pegged Cass County voter turnout at 44.4 percent, the study assumes voter turnout for North Dakota State University students was the same percentage. v To keep the estimates conservative, the extrapolations were done only for those institutions at which residency requirements proved to be a serious issue. Institutions where residency requirements did not have an apparent impact are excluded. Extrapolating the findings proceeded as follows: Bismarck State College = (5,352 students) X (0.5498 turnout rate for Burleigh County) X (0.077 unable to vote due to residency problems) = 226.6 Minot State University = (3,217 students) X (0.3892 turnout rate for Ward County) X (0.021 unable to vote due to residency problems) = 26.3 North Dakota State College of Science = (3,024 students) X (0.4724 turnout rate for Richland County) X (0.038 unable to vote due to residency problems) = 54.3 North Dakota State University = (13,332 students) X (0.4440 turnout rate for Cass County) X (0.036 unable to vote due to residency problems) = 213.1 University of North Dakota = (15,102 students) X (0.3849 turnout rate for Grand Forks County) X (0.029 unable to vote due to residency problems) = 168.6 The extrapolated total computes as 688.9 students who attempted to vote but were unsuccessful due to residency issues.
It is noteworthy that the residency requirements were an issue for respondents from the largest universities. Indeed, the universities in Bismarck, Fargo, and Grand Forks have many students who moved from their home town to attend school. These transplanted students often list their parents' address as their mailing address, making them vulnerable to exclusion. Students at the other eight institutions are less likely to have moved from their home town to attend school. Consequently, they live either with their parents or only a short distance away, making them considerably less vulnerable to exclusion.

Conclusion
The data reported here provide initial evidence that North Dakota's voter ID law had a small but real effect when it comes to preventing college students from voting. In the case of a lopsided election, as was the case in North Dakota in 2014, such a small percentage may not be consequential. However, in the event of a very close electionfor example, a Florida 2000-type scenarioanecdotal and empirical evidence of a similarly-sized problem could bring the legitimacy of the election outcome into question. Regardless of whether those who are kept from voting systematically would have preferred one candidate over the other, if the number of people turned away for lack of ID exceeds a candidate's win margin, citizens may question its legitimacy.
These findings amongst college-aged respondents are particularly troubling, however, despite the small affected percentage. Voting habits are set early in an individual's voting career (see, e.g., Campbell et al. 1960;Green and Shachar 2000;and Aldrich et al. 2011 for discussions of voting as a habit). Additional hurdles to vote may alter the calculus for an individual who may otherwise have become a habitual voter. Future research could examine whether being turned away from the polls has lingering effects in subsequent elections. In addition, the election examined here was a midterm election; it is reasonable to extrapolate that voting problems would be more widespread in a presidential election year when a larger, typically less-attentive, portion of the population turns out to vote. In any event, the study here offers another case study in the accumulating evidence of the effects of stringent voter ID laws.

Endnotes
i There are eleven NDUS public institutions: five two-year colleges and six four-year universities. In addition, there are five tribal colleges in the state; students enrolled only at these tribal colleges were not included in the sampling frame.
ii Full wording of the survey questions, as well as additional descriptions of the data, are available on the Upper Midwest Regional Center on Public Policy's web-site at www.ndsu.edu/centers/publicpolicy/.
iii "Medium" political interest indicates that the respondent noted he or she had been "somewhat interested" in the political campaigns this year. iv "Medium" political knowledge represents the 31 st to 65 th percentile on a nine-question standard battery of political knowledge questions. v Young adults tend to vote at lower rates than older demographics, though voting rates increase with levels of formal education. These two countervailing trends may bring the NDUS college student turnout relative close to the county-wide voter turnout rate, though it remains an estimation.