A GIScience Approach to Analyzing Spatial Patterns of Voter Turnout in Omaha, Nebraska

Transcription

1 A GIScience Approach to Analyzing Spatial Patterns of Voter Turnout in Omaha, Nebraska Kaitlyn Taylor Department of Geography and Geology University of Nebraska-Omaha Omaha, NE *H. Jason Combs Department of Geography University of Nebraska-Kearney Kearney, NE Paul R. Burger Department of Geography University of Nebraska-Kearney Kearney, NE ABSTRACT Scholars from a variety of disciplines have analyzed voter participation, with most studies focusing on socio-demographic issues to explain turnout. Often overlooked, however, is the geography of voter participation. Omaha, Nebraska presents an economically and ethnically diverse study area to examine geographic factors related to turnout. Over 51 percent of the state s registered voters live in three counties Douglas, Sarpy, and Lancaster which contain the Omaha and Lincoln metropolitan areas. This project employs Geographic Information Science (GIScience) along with electoral geography principles and spatial analysis to evaluate voter participation across the Omaha metropolitan area. Getis-Ord Gi* statistic is utilized to demonstrate statistically significant spatial clustering of high and low values of voter participation and turnout at the census block-group level. This study also examines a number of demographic variables through stepwise regression that help explain voting patterns in the Omaha metropolitan area. Results indicate that, independent of party affiliation, educated, middle-class, white populations in Omaha have higher voter participation rates. Key Words: GIScience, Nebraska, voter turnout. INTRODUCTION * Corresponding author The Geographical Bulletin 57: by Gamma Theta Upsilon The right to vote has been described as one of the most fundamental and cherished aspects of democracy (Hajnal and Lewis 2003, 660). Wood (2002, 209) adds that elections are at the core of the American political system and that for most Americans voting is the only form of political participation. As a result, the analysis of voting and elections has undoubtedly become one of the principal areas of investigation for many scholars from a variety of fields (Thomas 1968, 409). 99

2 Kaitlyn Taylor, H. Jason Combs and Paul R. Burger For many citizens voting is a powerful tool and an opportunity to hold elected officials accountable. Voter participation is a key element in the political process as higher voter turnout confers legitimacy for the elected official, provides a mandate for decision making and policy making, and is an indication that voters are connected with their community and government (Wood 2002, 209). On the other hand, low turnout connotes apathy, frustration, anger, disengagement, alienation, and/or lack of a sense of efficacy and confidence in government. Low voter turnout is perceived to be a threat and danger to our democratic way of life (Wood 2002, 209). Hajnal and Lewis (2003, 645) went so far as to state that the fact that almost half of all eligible voters do not vote in presidential elections has been cited repeatedly as evidence of an ongoing crisis in American democracy. Unfortunately, many citizens weigh the costs against the benefits of voting and opt to not participate in the political process (Highton 2004). Despite the importance attached to exercising the right to vote in presidential elections, turnout rates since their peaks in the 1800s have gradually declined over the last two centuries. Somewhat ironically, this downward trend has coincided with the elimination of the poll tax and literacy testing and methods like advance voting, absentee voting, voting by mail, and more flexible and convenient registration laws have made it easier for citizens to register and vote (Karnig and Walter 1983, 491; Wood 2002, 209). Highton (2004, 511) points out, for example, that the National Voter Registration Act of 1993 eliminated the costs of a special trip to register, and thereby eliminated the costs of registration for nearly all eligible voters. Highton (2004, 511) continues, given that all states are now required to have motor voter programs or election day registration, registration costs cannot go much lower. Low turnout has not always been the case though. 1 Following the controversial 1824 presidential election, voter participation rates jumped dramatically. The next three elections witnessed turnout rates greater than 50 percent and in the 1840 election over 80 percent of the potential electorate participated (Shelley et al. 1996, 77). For the remaining elections in the 1800s, voter participation rates remained high and approximately 79 percent of the eligible voters cast ballots in the 1896 election despite wide variations in turnout at the state-level. In the 1920 election, more than 25 million votes were cast for the first time; however, the voter turnout rate dipped below 50 percent. Since the 1940s there has been little variation in participation rates despite the fact that more votes are being cast. In 1992, for example, more than 100 million Americans voted in a presidential election for the first time (Shelley et al. 1996, 119). More recently, over 131 million (62 percent turnout) Americans voted in the 2008 presidential election. In 2012, the number declined to an estimated 124 million (58 percent) despite the fact that there were eight million more eligible voters (Bipartisan Policy Center 2012; Roper Center Cornell University 2016; United States Election Project 2014). Multiple studies have expressed concern over declining civic engagement and voter participation rates in the United States (Twenge et al. 2012). Meirick and Wackman (2004, 1162) contend, for example, that the American electorate is not well informed and that without a certain degree of knowledge, people are unable to make political decisions that reflect their interests and values. Eagles and Davidson (2001) agree and add that in recent decades voter turnout has been troubling, especially among young voters who cite time constraints and lack of interest as factors that contribute to low turnout totals. Several educational endeavors and civic engagement programs have been created to combat issues like low voter turnout. Schmidt (2011, 110, 108) interviewed students in Malawi and demonstrated that there is a strong reciprocal relationship between the meaning of place and activities of citizens and that students who can identify and talk about other places in the world are 100

3 A GIScience Approach to Analyzing Spatial Patterns of Voter Turnout in Omaha, Nebraska better prepared to make political decisions. A separate study examined the Kids Voting USA program, which is designed to educate children about the democratic process and foster political engagement, and found that the program helps students make political decisions that better reflected their attitudes (Meirick and Wackman 2004, 1161). Numerous studies have also explored various factors that impact turnout rates, which vary widely throughout the United States (Cebula and Toma 2006). 2 Many studies have analyzed various socio-economic and demographic characteristics that impact turnout rates, such as age, education, and income (Greene and Nikolaev 1999; Grober and Schram 2006). Karnig and Walter (1983, 503) also focus on socio-economic aspects related to turnout in addition to the percent of the population living in the same residence for the past five years. Endersby et al. (2002) and Settle and Abrams (1976) also studied campaign spending and its relationship to turnout and found somewhat mixed results, while another project examined the decline of traditional (Democrat and Republican) party affiliation and participation (Herrara et al. 2008). Additionally, others have suggested that citizens in the United States, which has the busiest election calendar on earth, experience fatigue due to the frequency of elections at various levels primaries, special elections, local, state, and national elections, for instance which negatively impacts turnout rates (Gerber and Green 2008, 1; Rosenstone and Hansen 2003). Another issue discussed in the literature that impacts turnout is gerrymandering. Decades ago Carl Sauer (1918, ) described gerrymandering as a device for the partial suppression of public opinion and noted that few states are without gerrymandered districts of some sort. More recently, Hayes and McKee (2009, 1006) have argued that gerrymandering and redistricting studies which typically focus on electoral outcomes has caused scholars to overlook the influence of redistricting on citizens exercise of political voice. Hayes and McKee (2009, 1017) continue, the data are clear: redistricting suppresses congressional election participation among redrawn voters. There is, however, an issue discussed in multiple studies that appears to increase turnout the closeness of the contest (Downs 1957; Endersby et al. 2002; Hayes and McKee 2009). Rosenstone and Hansen (2003, 35, 179) point out that people participate more in politics when outcomes hang in the balance in part because the media cover competitive contest more intensely than runaways. Settle and Abrams (1976, 87) agree and offer a simple but accurate assessment: as the winning majority increases, voter turnout decreases as voters consider the impact of their votes on the election s outcome. OBJECTIVES AND DATABASE CONSTRUCTION The objective of this study is to determine how participation and turnout rates from the 2012 General Election vary spatially across the Omaha, Nebraska metropolitan area and what geographic factors contribute to those patterns. This study focuses on Omaha to examine demographic issues on a larger scale in a metropolitan region, in addition to analyzing the spatial factors at work in Omaha elections. Omaha, with a 2010 population of 408,958, was selected for a variety of reasons. First, data were readily available from various election commissions and the Secretary of State s office. Secondly, Omaha s high turnout rate coupled with its demographic diversity and spatially clustered socio-economic groups allows for an in-depth investigation of voter turnout. All geographic layers are 2013 TIGER/Line files obtained from the U.S. Census Bureau as shapefiles. The layers which are available at the county- and state-level include: block groups, counties, edges, places, and states (Table 1). We downloaded and imported data layers into ArcGIS, and merged the countybased layers into a statewide spatial database for each level of geography. 101

4 Kaitlyn Taylor, H. Jason Combs and Paul R. Burger Table 1. GIScience Data Dictionary. Layer Type Source Projection Scale Attributes Places Polygon TIGER/Line 2013 State Plane NE: (NAD 83) Edges Line TIGER/Line 2013 State Plane NE: (NAD 83) Roads Line TIGER/Line 2013 State Plane NE: (NAD 83) Block Groups Polygon TIGER/Line 2013 State Plane NE: (NAD 83) Voters Tabular- Point Nebraska Secretary of State State Plane NE: (NAD 83) Counties Polygon TIGER/Line 2013 State Plane NE: (NAD 83) State Polygon TIGER/Line 2013 State Plane NE: (NAD 83) 1:100,000 Name, Area 1:100,000 Name, Address, Length 1:100,000 Name, Address, Length 1:100,000 Tactician Socio- Demographics 1:100,000 Address, Party, Age, Vote History 1:100,000 Name, Area 1:100,000 Name, Area Tabular data for Nebraska voter registration as of July 2014 was obtained from the Nebraska Secretary of State. Voters whose addresses were not verified using the National Change of Address (NCOA) process utilized by the Secretary of State were removed from the voter file along with new registrants after the 2012 General Election. Finally, we acquired enhanced Census of Population data with projections for 2014 for the block group- and county-levels were acquired from Tactician. METHODOLOGY First, we projected TIGER/Line13 files from the NAD 83 into shapefiles in a File GeoDatabase (GDB). The TIGER/Line Edges layer contains every line feature, including the streets and corresponding address ranges. The streets features and associated attributes were queried and exported as the Roads feature class in the ArcGIS GDB. An address locator was created using the US Streets with Zone style. We then geocoded voters by their home street address for summarizing party participation and turnout for each block group. The overall match rate of Nebraska s 921,581 voters to the streets layers following both automated and manual geocoding was 97.6 percent. We then tabulated voters by block group within the Omaha city limits to examine participation and turnout of members of each party (Republican, Democrat, and Nonpartisan). Summary tables from the voter file were joined to the block group layer where new variables were added as percentages for participation and turnout by party. Participation rates for each party were calculated by dividing those who voted in that party in 2012 by total registered voters in that party. Voter turnout was calculated by dividing the actual votes cast by each party s members by the total number of votes cast in the block group. We analyzed the data using the Getis Ord Gi* statistic, which tests whether a particular location and its surrounding regions have higher-than-average values on a variable of interest (Rogerson 2006; Wilhelm and Steck 1998). G statistics are a family of spatial association measures that when used appropriately deepen the knowledge of the processes that give rise to spatial association, in that they enable us to detect local pockets of dependence that may not 102

5 A GIScience Approach to Analyzing Spatial Patterns of Voter Turnout in Omaha, Nebraska appear when applying global statistics (Getis and Ord 1992, 190). Additionally, an added advantage of the Gi* statistic in ArcGIS is the False Discovery Rate (FDR) option where the p-values determining confidence levels is reduced to account for multiple testing and spatial dependence (ArcGIS Pro). In this particular study, we use the Gi* statistic to demonstrate statistically significant spatial clustering of high and low values of voter participation and turnout. Next, we joined socio-demographic data from 2014 from the Tactician Corporation for Omaha s 454 Census block groups to the voter participation and turnout data, exported from ArcGIS and imported into SPSS statistical software for evaluation. Factor analysis a data reduction technique that clusters variables that are highly correlated with one another and represent a common underlying variable was performed on 105 variables from the categories of age, education, ethnicity, housing value, income, marital status, and occupation (Kachigan 1991). We removed variables with communalities (shared variance) below 0.60 to increase the probability of related variables within the factors. We then rotated the remaining eighty-seven socio-demographic variables with a varimax rotation and an eigenvalue threshold of 1 grouping them into twentyone factors that explain 82.0 percent of the overall variance in the dataset (Burger et al. 2015; StatSoft 2013; Burger and Combs 2011). In addition to reducing the risk of multi-collinearity, a problem when two or more predictor variables are highly correlated with each other, factor analysis provides the added advantage of creating meaningful groups of variables (Kachigan 1991). A varimax rotation aids in the interpretation process by maximizing the variation in the factor (Burger et al. 2015, StatSoft 2013; Burger and Combs 2011; Thrall 2002). Using block-group level participation rates by party and turnout rates for each party as dependent variables, we conducted a series of stepwise linear regressions using the factor scores for the twenty-one factors. Stepwise regression begins with an initial model containing the intercept and adding additional effects (factors) at each subsequent step until a specified number of steps is reached or no additional factors meet the significance threshold in the stepping criteria (Statsoft 2013). Factors with a significance level of 0.05 are added to the model at each successive step and remain in the model as long as their significance level is at or below The stepping process continues until no additional factors meet the significance thresholds or the adjusted R-square increase is below 2 percent. In predicting Omaha participation rates by party or turnout rates for each party, an initial linear model is created through the y-axis (intercept) that includes at least one statistically significant factor (similarly grouped socio-economic variables) that explains voter behavior at the block-group level. Additional factors are added to the model as long as they are statistically significant (alpha = 0.05) and the addition of the variables (factors) makes a significant contribution (increased R-square) to the model s ability to explain voting rates (dependent variable). ANALYSIS In Omaha, for the 2012 presidential contest, 154,043 votes (78.2 percent) were tallied out of a potential 197,001 registered voters. Participation varied slightly between the parties, Republicans had the highest rate at 84.1 percent, followed by Democrats at 77.9 percent and Nonpartisan at 68.6 percent. For a more detailed examination of participation at the block-group level, Figs. 1, 2, and 3 display the results of the Getis-Ord Gi* statistic on voting percentages or participation rate of each party mapped by Z-score. Participation rates are found by dividing each party s votes by each party s registered voters. For example, Figure 1 shows the Gi* statistic on the percentage of actual Democratic votes divided by registered Democratic voters. Areas in dark gray represent a statistically significant high clustering of the proportion of Democrats 103

6 Kaitlyn Taylor, H. Jason Combs and Paul R. Burger 104 Figure 1. Omaha Democrat Voting Percentage in the 2012 General Election.

7 A GIScience Approach to Analyzing Spatial Patterns of Voter Turnout in Omaha, Nebraska Figure 2. Omaha Republican Voting Percentage in the 2012 General Election. 105

8 Kaitlyn Taylor, H. Jason Combs and Paul R. Burger 106 Figure 3. Omaha Nonpartisan Voting Percentage in the 2012 General Election.

9 A GIScience Approach to Analyzing Spatial Patterns of Voter Turnout in Omaha, Nebraska Figure 4. Omaha Democrat Voter Turnout in the 2012 General Election. 107

10 Kaitlyn Taylor, H. Jason Combs and Paul R. Burger 108 Figure 5. Omaha Republican Voter Turnout in the 2012 General Election.

11 A GIScience Approach to Analyzing Spatial Patterns of Voter Turnout in Omaha, Nebraska Figure 6. Omaha Nonpartisan Voter Turnout in the 2012 General Election. 109

12 Kaitlyn Taylor, H. Jason Combs and Paul R. Burger voting in the block group (greater than +2.0 standard deviations). Conversely, light gray block groups depict statistically significant clusters of low proportions of Democrats voting (less than -2.0 standard deviations). Maps by party affiliation (Democratic, Republican, and Nonpartisan), are similar, with statistically significant high-value concentrations of voting percentages for each group in central Omaha from approximately South 55 th Street stretching west to 180 th Street (Fig. 1). Democratic participation strength continues west to North 240 th Street (US Highway 6). L Street marks the southern boundary for each party s concentration with State Highway 64 bounding the northern edge. Republican participation remains significantly high as far north as Fort Street (Fig. 2). Statistically significant clusters of low participation rates by party are found along Omaha s eastern edge from Interstate-480 east to the city limits (Figs. 1, 2, 3). Figs. 4, 5, and 6 portray results of the Gi* statistic on voter turnout for each party. Voter turnout is found by dividing each party s total votes by the total votes cast in each block group. This traditional definition of voter turnout shows the greatest portion of each party voting based upon all those voting. Fig. 4 indicates clusters of high Democratic turnout in the eastern part of Omaha with concentrations of statistically significantly low proportions in west Omaha. Statistically significant high and low clusters of Republicans are not surprisingly the inverse of the Democrats (Fig. 5). Interestingly, the Republican Party is the only one for which the spatial pattern of participation is similar to their turnout (Figs. 2 and 5). Southeastern Omaha contains a clustering of high Nonpartisan turnout with low concentrations found in northeastern Omaha (Fig. 6). The stepwise regressions on the dependent variables by party against the factor scores of the twenty-one rotated components identified three significant factors for participation and six for turnout (Table 2). Adjusted r- square values because multiple R always takes on a positive value, its initially calculated value must be adjusted downward for participation range from a high of 43.7 percent for Republicans to a low of 28.9 percent for Nonpartisan voters while the model for Democrats explains 37.4 percent of their participation (Kachigan 1991). The same three factors are significant for each party affiliation albeit the order for Nonpartisans differs slightly. Factor 1, consisting of married, white, college-educated, middle- to upper-income ($75,000-$200,000) neighborhoods with higher employment, is the most significant component explaining participation regardless of party similar to findings drawn by Rosenstone and Hansen (2003) (Table 3). Middle aged voters primarily between 55 and 64 years of age comprise Table 2. Significant Factors from Regression. Participation R-square Components Republican 43.7% Factor 1 Factor 4 Factor 7 Democrat 37.4% Factor 1 Factor 4 Factor 7 Nonpartisan 28.9% Factor 1 Factor 7 Factor 4 Turnout R-square Components Republican 65.0% Factor 1 Factor 5 Factor 7 Democrat 66.3% (Factor 1) (Factor 5) Nonpartisan 18.5% (Factor 6) (Factor 3) Factor 1 (Factor 4) 110

13 A GIScience Approach to Analyzing Spatial Patterns of Voter Turnout in Omaha, Nebraska Table 3. Factor Loadings of Significant Variables. Factor Rotated component matrix Bachelor s degree.670 High school education Some high school education Employed.831 Unemployed Household income $0-$39, Household income $150,000-$149, Household income $150,000-$199, Household income age 65 to 74 $0-$39, Household income $75,000-$99, Housing value $150,000-$199, Married.624 Race black Race white.856 Age Age Females Females Males Never married Age Females Females Males Graduate degree.499 Household income $200,000 plus.794 Household income age $200,000 plus.879 Household income age 75 plus $200,000 plus.836 Housing value $300,000 plus.817 Age Females Males Less than 9th grade education Race Hispanic Race other

14 Kaitlyn Taylor, H. Jason Combs and Paul R. Burger Factor 4, while Factor 7 has an absence of those with less than a 9 th grade education and Hispanic or other races. Adjusted r-square values for the turnout models were 65.0 and 66.3 percent for Republicans and Democrats respectively with 18.5 percent for Nonpartisans. It is with turnout the proportion from each party voting of the total by block-group that differences among the parties emerge. Factor 1 is again the most important for both Republicans and Democrats though it loads negatively against Democratic turnout. That is, the lower the percentage of married, white, college educated, middle- to upper-income residents in a neighborhood, the higher the Democrat turnout. Factor 5, graduate educated, upper income ($200,000+), retirement age individuals with housing values above $300,000, is the second most important for both Republicans and Democrats. Similar to Factor 1, Factor 5 loads negatively for Democratic turnout. Contrasts between Factors 1 and 5 are evident when comparing the Gi* statistic analysis for Democratic and Republican turnout in Figures 4 and 5. Areas of statistically significant higher Republican turnout (Fig. 5) mirror participation for all three voter categories within central and western Omaha while the same area represents significantly lower turnout among Democrats. Conversely, Omaha neighborhoods with higher proportions of lower income black (northeast) and Hispanic (southeast) populations have statistically significant higher turnout for Democrats and lower for Republicans. Factors 3 and 6 are most important in explaining Nonpartisan turnout both of which loaded negatively. Nonpartisan turnout is highest in neighborhoods with larger proportions of both single and 18-to-34-year-old populations. CONCLUSIONS This project employed GIScience along with electoral geography principles and spatial analysis to evaluate participation and turnout from the 2012 General Election across the Omaha metropolitan area. Ironically, areas with statistically significant high or low participation rates for each party are similar. Voting occurs in higher proportion for each party than their registration in block groups in west-central Omaha and in lower proportions along the eastern edge. Participation rates are spatially consistent even though turnout by party does not necessarily follow a similar spatial pattern. High and low clustering of turnout by party follows traditional urban spatial patterns with Democratic strongholds found in inner-city neighborhoods and Republican concentrations located on Omaha s western suburban edges (Archer 1986). The Getis Ord Gi* statistic revealed block groups within the metro area where turnout is dominated by a single party and a core area where voter participation regardless of party is concentrated. Regression analyses for both participation and turnout by party provide additional insight into these patterns. Participation rates, regardless of party are highest in middle- to upper-income block groups containing higher proportions of white, middleaged married and college educated voters. This explains their similar spatial pattern in the Gi* statistical analysis. Contrasting high and low spatial clustering of Republican and Democratic turnout is also explained through regression factors. While both again share the similar Factors of 1 and 5, for Republicans they load positively while they are negatively related to Democratic turnout. College educated middle-aged and income, white, married neighborhoods high in Republican participation are also high in turnout along with upper-income retired neighborhoods. Conversely, Democratic turnout is strongest where these groups are absent. Nonpartisan turnout is strongest in block-groups with higher concentrations of single and year olds and concentrated in downtown Omaha, specifically within the Interstate-480 loop. Low turnout rates are a concern for the United States. Focusing on the geography of voter turnout and participation can address some of the issues education, gerrymander- 112

15 A GIScience Approach to Analyzing Spatial Patterns of Voter Turnout in Omaha, Nebraska ing, and income, for example that negatively impact participation rates. Advocacy groups and civic organizations can examine spatial relationships and target low turnout areas to encourage citizens to engage in the political process. Moreover, understanding these spatial patterns enables candidates and political parties to efficiently utilize resources and coordinate efforts to engage voters in the political process which is key. Ultimately, mobilizing significant numbers of voters can be decisive at all levels of elections (Gerber and Green 2008). NOTES As low as presidential election turnout rates are, mid-term and local election rates are typically even worse (Karnig and Walter 1983; Rosenstone and Hansen 2003; Wood 2002). In 2012, participation rates varied dramatically in the presidential election. Five states had rates higher than 70 percent (Minnesota was the highest at just over 76 percent) while four states had turnout rates below 50 percent (Hawaii was the lowest at just below 45 percent) (United States Elections Project). REFERENCES ArcGIS Pro. Tool Reference: Hot Spot Analysis (Getis-Ord Gi*). Date Last Accessed, September 6, Archer, J. Clark, and Shelley, Fred American Electoral Mosaics. Washington, D.C.: Association of American Geographers. Bipartisan Policy Center Voter Turnout Report. library/2012-voter-turnout/. Date Last Accessed, February 24, Burger, Paul R., Combs, H. Jason, and John T. Bauer Market Area Analysis of Robinwood Retirement Community: How GIScience Integrates Marketing Geography and Spatial Analysis. Papers in Applied Geography. 1(4): Burger, Paul and Combs, H. Jason Cleaning Up the Competition with GI- Science: A Market Area Analysis of Yellow Van Cleaning Services. The Pennsylvania Geographer 49(2): Cebula, Richard and Toma, Michael Determinants of Geographic Differentials in the Voter Participation Rate. Atlantic Economic Journal 34: Douglas County Election Commission. election_results.aspx. Date Last Accessed 2/12/16. Eagles, Munroe, and Davidson, Russell Civic Education, Political Socialization, and Political Mobilization: Can Kids Voting, USA Diminish Inequalities in Voter Participation? Journal of Geography 100: Endersby, James, Galatas, Steven, and Rackaway, Chapman Closeness Counts in Canada: Voter Participation in the 1993 and 1997 Federal Elections. The Journal of Politics 64(2): Getis, Arthur, and J.K. Ord The Analysis of Spatial Association by Use of Distance Statistics. Geographical Analysis 24: Green, Donald and Gerber, Alan Get Out the Vote. Washington, D.C.: Brookings Institute. Greene, Kenneth and Nikolaev, Oleg Voter Participation and the Redistributive State. Public Choice 98: Grober, Jens and Schram, Arthur Neighborhood Information Exchange and Voter Participation: An Experimental Study. The American Political Science Review 100(2): Hajnal, Zoltan and Lewis, Paul Municipal Institutions and Voter Turnout in Local Elections. Urban Affairs Review 38(5): Hayes, Danny and McKee, Seth The Participatory Effects of Redistricting. American Journal of Political Science 53(4): Herrara, Helios, Levine, David, and Martinelli, Cesar Policy Platforms, 113

16 Kaitlyn Taylor, H. Jason Combs and Paul R. Burger Campaign Spending and Voter Participation. Journal of Public Economics 92(3-4): Highton, Benjamin Voter Registration and Turnout in the United States. Perspectives on Politics 2(3): Kachigan, Sam Multivariate Statistical Analysis. New York: Radius Press. Karnig, Albert and Walter, Oliver Decline in Municipal Voter Turnout. American Politics Quarterly 11(4): Meirick, Patrick, and Wackman, Daniel Kids Voting and Political Knowledge: Narrowing Gaps, Informing Votes. Social Science Quarterly 85: Rogerson, Peter Statistical Methods for Geography. Thousand Oaks, California: Sage Publications. Roper Center, Cornell University Popular Votes cornell.edu/polls/us-elections/popularvote/. Date Last Accessed, February 24, Rosenstone, Steven, and Hansen, John Mobilization, Participation, and Democracy in America. New York: Macmillan. Sauer, Carl Geography and the Gerrymander. The American Political Science Review 12(3): Schmidt, Sandra Making Space for the Citizen in Geographic Education. Journal of Geography 110: Settle, Russell and Abrams, Buron The Determinants of Voter Participation: A More General Model. Public Choice 27: Shelley, Fred, Archer, J. Clark, Davidson, Fiona, and Brunn, Stanley Political Geography of the United States. New York: The Guilford Press. StatSoft, Inc Electronic Statistics Textbook. Tulsa, OK: StatSoft. WEB: Date Last Accessed, February 15, Thomas, Norman Voting Machines and Voter Participation in Four Michigan Constitutional Revision Referenda. The Western Political Quarterly 21(3): Thrall, Grant Ian Business Geography and New Real Estate Market Analysis. New York: Oxford University Press. Twenge, Jean, Freeman, Elise, and Campbell, W. Keith Generational Differences in Young Adults Life Goals, Concern for Others, and Civic Orientation, Journal of Personality and Social Psychology 102(5): United States Election Project National General election VEP Turnout Rates, 1789-Present. Date Last Accessed, February 24, Wilhem, A. and Steck, R. Exploring Spatial Data Using Interactive Graphics and Local Statistics. Journal of the Royal Statistical Society: Series D The Statistician. 47(3): Wood, Curtis Voter Turnout in City Elections. Urban Affairs Review 38(2):