{"id":184531,"date":"2013-04-29T18:55:14","date_gmt":"2013-04-29T18:55:14","guid":{"rendered":"http:\/\/www.esri.com\/arcgis-blog\/?post_type=blog&p=184531"},"modified":"2018-12-18T10:25:13","modified_gmt":"2018-12-18T18:25:13","slug":"red-blue-and-purple-mapping-the-2012-us-presidential-election","status":"publish","type":"blog","link":"https:\/\/www.esri.com\/arcgis-blog\/products\/arcgis-online\/national-government\/red-blue-and-purple-mapping-the-2012-us-presidential-election","title":{"rendered":"Red, Blue and Purple: mapping the 2012 US Presidential Election"},"author":4951,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","format":"standard","meta":{"_acf_changed":false,"_searchwp_excluded":""},"categories":[23341,22851,22941],"tags":[26471,29471,29481,25581,32661],"industry":[],"product":[36551,36611],"class_list":["post-184531","blog","type-blog","status-publish","format-standard","hentry","category-analytics","category-national-government","category-mapping","tag-kennethfield","tag-election-results","tag-elections","tag-spatial-statistics","tag-thematic-mapping","product-arcgis-online","product-story-maps"],"acf":{"short_description":"Every time an election occurs, maps become a key component in telling the story","flexible_content":[{"acf_fc_layout":"content","content":"

E\"\"very time an election occurs, maps become a key component in telling the story, but what type of map best tells the story of the winners and losers? Red\/blue choropleths? Areas shaded in an array of purples? Value by alpha maps? Dot density by County? Ultimately, the areas used (e.g. Counties) are arbitrary, exhaust space and dictate the visual pattern we see.\u00a0 We can warp them into cartograms but these sometimes distort geography too much for them to make much sense. The patterns we see are as much a product of the boundaries as the voting patterns of real people in real places. This blog entry explores different ways to map election results and describes a different type of map we made to show the 2012 Presidential election results\u2026it\u2019s a multiscale dasymetric dot density web map (viewable on\u00a0ArcGIS Online<\/a>).<\/p>\n

Debate about how best to represent election results is nothing new, fuelled principally by the default two-colour map commonly used to show red vs blue States (Figure 1).<\/p>\n

\"\"<\/p>\n

Figure 1: 2012 US Election result at state level as a unique values map (source: Mark Newman 2012)<\/em><\/p>\n

You only have to look at the map and ask the question \u2018who won the election?\u2019 to see the problem.\u00a0 There is more red than blue which suggests a win for the Republicans when the opposite was the case. The electoral college voting system means that a state will be mapped as either red or blue.\u00a0 Large states become visually dominant and while they might not contribute many electoral college votes to the total they appear visually prominent. The other problem here is that the different sized areas alters the visual prominence of some States relative to others, regardless of how many people or voters live there. The data should be normalized to take account of the population distribution because it doesn\u2019t allow for the fact that the large red states generally have smaller populations. \u00a0At least Figure 1 used an Albers equal area conic projection\u2026the key part being \u2018equal area\u2019 so it ensures that each of the states is in its correct proportion compared to all the others.<\/p>\n

\"\"<\/p>\n

Figure 2: 2012 US Election results mapped using different techniques (sources: Newman, 2012; Axis Maps, 2012; Nelson, 2012)<\/em><\/p>\n

The use of a Mercator projection (or Web Mercator on a web map) further distorts the areas of the states and compounds the visual problems seen in Figure 1 (Figure 2a). Figure 2b shows how at County level, the red\/blue approach seems to magnify the wrong visual message even further. A population density equalizing cartogram can be used to modify the sizes of the states by rescaling according to the number of electoral votes (Figure 2c). This accommodates the visual problem of red dominating the map but modifies the map shape and can cause difficulty for readers unfamiliar with this map type.<\/p>\n

Figure 2d shows the percentage of vote with vivid red showing areas of strong Republican support, vivid blues showing strong Democrat support and shades of purple for areas with a strong balance between voting. Because the map shows percentages of votes by total vote, the data is normalised using a consistent denominator and the map becomes a true choropleth allowing for visual comparisons to be accurately made across the map. The value by alpha approach is shown in Figure 2e which uses saturation of colour so areas of low population fade to black. A dot density approach is illustrated in Figure 2f which goes some way to showing the mix of vote per county.\u00a0 Each blue or red dot represents 1,000 votes. Density of voter turnout is reflected in the different dot densities seen in each county.\u00a0 If we consider voter turnout a proxy for population then we can see where the more densely populated counties are and vice versa.\u00a0 We can also see counties with predominantly blue or red dots and those where the dots are mixed (giving rise to the illusion of purple hues because of the close proximity of dots), showing a more balanced vote between the candidates.<\/p>\n

However, a drawback of all the above examples is that they use States or Counties as a basic unit in which to represent the data visually. With much of the US unpopulated, large Counties with sparse populations are not easily compared to small counties with densely populated areas and vice versa.<\/p>\n

An alternative approach is to create a dasymetric map; a map popularised by the American geographer John Kirtland Wright in the 1940s (who incidentally also coined the term \u2018choropleth\u2019 in 1938). A dasymetric map is a type of thematic map that attempts to correct the errors we\u2019ve seen that result from mapping population data using choropleths.\u00a0 It is a form of spatial classification of data where ancillary data is used to reapportion data captured in one geographic area to another geographic area.\u00a0 In terms of our map of election votes, if we think of a county as the area used as the original container for the data, a better way to map the data might be to display it using boundaries that reflect the distribution of the population within that area.\u00a0 This overcomes the visual assumption of choropleths that the population is evenly distributed within each area.<\/p>\n

Take, for example, San Bernardino County in California.\u00a0 The county covers approximately 20,000 sq miles in Southern California yet the proportion of urbanised area is extremely small. These areas are where virtually all of the 450,000 voters live (Figure 3a).<\/p>\n

\"\"<\/p>\n

Figure 3. San Bernardino county in Southern California<\/em><\/p>\n

The entire area would be coloured blue in the traditional red\/blue map since the Democrats won the County with 51% of the vote\u2026a narrow margin (Figure 3b). If we used a dot density approach (with 1 dot = 200 votes) then the voting pattern would be distributed throughout the sparsely populated County which contrasts with the adjacent Los Angeles County and Orange County (Figure 3c). However, a dasymetric map reapportions the data into the populated areas. In Figure 3d, the vote count data has been reapportioned into different urban categories to reflect high density, low density and non-urban (inhabited) areas. Certain other land uses such as airports and parks are masked out to avoid areas where we know people do not live and the data rendered as a dot density. This approach has several advantages over the alternatives namely:<\/p>\n