Computer technology and getting out the vote: new targeting tools - Political Professional
Florida didn't have to be nearly' as close as it was in the 2000 presidential race. That's the obvious verdict from examining turnout statistics in several of that state's large counties. Legions of registered voters simply failed to show up at the polls in the state's major metropolitan areas.
Had Democratic or Republican campaign mobilization efforts been more effective at reaching irregular voters and non-voters, this election, like many others, could have been won by thousands of votes.
Many elections turn on voter mobilization, and the parties have recently taken greater interest in getting people out to vote. Whether we are talking about the Republicans 72-Hour plan, or the Democrats' newly unveiled Project 5104, get-out-the-vote (GOTV) has recaptured the attention of politicians and consultants as never before.
One reason for the renewed interest GOTV is that revolutionary new computer software has made it possible to do more precise targeting than ever before. In the last five years, campaigns have discovered that Geographic Information Systems (GIS) is not just for redistricting, for which it has been used to map tailor-made districts for lawmakers.
Ten years ago, GIS was the province of a few specialists, housed mostly in engineering firms and operable only by those with highly specialized training. But GIS has gone desk top, and political consultants and campaign personnel now can obtain a working knowledge of these powerful software tools with just a few days training and some practice.
For years, the most trustworthy tool in the GOTV tool kit has been the voter list. These lists are valuable because they contain residential and mailing addresses, vote histories and miscellaneous demographic attributes of registered voters. These lists vary in quality, but the better ones also contain voters' party registration. From these massive databases, walking lists for individual precincts can be printed out, addresses can be merged onto mass mailings, and telephone banks can spring into operation. No seriously competitive campaign can do without these lists.
But voter outreach efforts are costly and, in the past, highly imprecise. By mapping voters' residences, GIS technology has made it possible to do far more precise targeting than could be done before. After all, only the very wealthiest and/or most naive campaigns would indiscriminately target every voter for a contact. As nice as "inclusiveness" sounds as a slogan, carpet bombing entire neighbor hoods, cities and even states, with direct mail and phone calls risks mobilizing voters of the opposite party without converting them, countering the very purpose of campaign efforts.
Moreover, what many campaigns fail to recognize is that lint every favorably inclined voter needs a contact. Many regular partisan voters from California to Maine report being barraged with junk mail that is completely wasted on them. These reliable citizens intended to vote regardless of how often they were touched. These voters' longstanding voting habits are usually evident on the lists, and the vote history information suggests that they need few, if any, reminders.
The more sophisticated state parties have already purchased Caliper Corp.'s Maptitude software, or ESRI's Arcview and ArcGIS software, and begun to map their voter lists, pinpointing the longitude and latitude coordinates of residences on the neighborhood streets where they live.
Missouri's Republican party is among the most advanced, having mapped virtually the entire state using Maptitude. Political Maptitude, a more elaborate version of the basic Maptitude package, contains useful tools and base maps specifically for doing political work. ESRI's Arcview, a well established product, has been used by forward-thinking party organizations and consultants at least since the mid- '90s. But until recent years, the use of these products has been limited to redistricting, and for the mapping of county and precinct data. ESRI's newer software, including ArcGIS, has made it easier to map the voter lists themselves. Once the lists are mapped, the possibilities are only limited by the information provided on the lists.
For example, a campaign might want to display each person's voting history on the map for the explicit purpose of targeting irregular voters for" contacts. Campaigns could map only the 18- to 29-year-old voters, or only those voters who have most recently registered. In states and locations where party registration is available on the lists, highly precise targeting is possible, whereby a campaign can focus its attention only on voters belonging to specific parties.
To obtain a general impression of where the neighborhoods of high and low turnout appear, it usually is not necessar5 to map an entire list. Mapping a 10 or 20 percent random sample of the very largest lists would be sufficient to save computation time, and yet still get a realistic picture of locations where voters are in greatest need of contacts. For precinct canvassing purposes, maps of all voters in each neighborhood can be generated.
A map of a particular location might show notable clusters of non-voters and irregular voters that did not participate in the 2000 presidential election. But it is still important to try to minimize the costs of reaching these households by routing canvassers in the most efficient way possible. Sometimes clusters of like-minded voters are present, but not readily apparent to the eye.
To confront this challenge, I am working on applying cluster detection software used in epidemiology research to detect groups of voters with similar traits that are not readily visible by visual inspection. This approach scans the surface of the map to detect point patterns that differ from randomness, telling us where, for instance, significant blocs of Democrats, Republicans, 18- to 29-year-olds, or newly registered voters might be concentrated. The software program will then circle, or mark, the locations that should be targeted first. GIS can also design optimal routing plans to canvass neighborhoods in the most time saving and effective way.
For fringe suburbs and rural areas, these techniques will make it possible to do the kind of voter contact work that was once only done in the most densely packed urban areas.
To be sure, these software tools can be used in the traditional way to map county and precinct level election results also. But in politically divided areas, the utility of simple county and precinct level targeting quickly breaks down. While county and precinct results are a good place to begin, this level of information does not help to reach party loyalists in precincts where the other party dominates. Using precinct information alone, these voters would be missed and commonly are in campaign efforts.
Moreover, some recent political science research has suggested that minority partisans in areas where the opposite party dominates may" not turn out in large numbers. The social pressure associated with being a minority keeps them away from the polls unless they are contacted by a GOTV effort. Voters with diverse social networks may also be less inclined to participate, according to recent research by Diana Mutz, a political scientist at Ohio State University. These are precisely the kinds of voters who most need the stimulus of a campaign contact.
One of the newer, more exciting innovations we have worked on here at the University of Maryland is the merging of precinct and census information with the voter lists. Using GIS in connection with common database management software, it is possible to locate each voter within their respective precinct and census tract. We can then understand what types of neighborhoods each voter lives in. Why would this be worth the effort? Mainly because neighborhoods tell us a lot about the voters who live there. Neighborhoods reveal housing preferences, spending habits, racial and ethnic corn position, lifestyles, levels of geographic mobility, voting habits and other traits relevant to predicting political participation and attitudes.
Neighborhood characteristics don't always matter in the same way everywhere, but given what little information is available on the voter lists, adding in the neighborhood and precinct data can be a real boon. It might allow a campaign to do better persuasion-oriented targeting. Suppose that Democrats wanted to target female voters in affluent suburban neighborhoods. The lists alone do not contain household income information, but the combination of gender and high income is easily accessible once the census and voter list information are merged together.