This Election Atlas provides background for understanding the 2013 mayoral election in New York City. It helps you visualize where the votes will come from and who is more likely to receive them. It particular, it visualizes election results for all the Democratic mayoral candidates who have previously run for citywide office. For comparison and contrast, it also includes maps of recent gubernatorial and presidential elections.
Recent primary, runoff, and general elections for Mayor, Comptroller, and Public Advocate show us how 4 of the 8 candidates fared in the past. These voting patterns provide a quantifiable, visual record against which to gauge the 2013 campaigns.
Voting data is presented here in a new way — by Census block, instead of election districts. The City currently has about 5,300 election districts, but 30,000 inhabited city blocks. Displaying the information at this level reveals voting patterns literally city block by block in some areas, whereas election districts (EDs) often cover much larger areas. Census blocks also enable us to compare voting outcomes directly with Census demographics. And blocks provide a common geographic unit for comparing elections consistently across time, even when ED boundaries change.
To allocate ED-level voting results to blocks, the Center for Urban Research geocoded more than 11 million voter registration records over the past eight years to determine how those actually casting ballots each year overlap between EDs and blocks. See the Methodology section in the More... tab above for details.
The different colors on each map depict the winners and losers. The shade of each block's color represents how strongly voters supported the candidate, while the intensity of the colors portrays the level of turnout. This enables you to understand not only how strongly voters in a given area favored the candidate, but also how well each candidate mobilized their voter base to come to the polls.
Each map can also be viewed in relation to demographic patterns, neighborhoods, and Council districts. Use the dynamic overlays above each map:
We provide a summary analysis next to each map about each election, and we encourage you to examine the maps closely to reach your own conclusions.
This page analyzes the results of the 2013 mayoral race, and also compares these results with other elections.
2013 election results (NYC BOE)
Source: Associated Press (10/1/13)
City Councilmember Letitia James won the runoff with just under 60% of the vote, against State Senator Daniel Squadron. Turnout was relatively low — almost 188,000 votes cast, or roughly 6% of registered Democrats. By comparison, turnout in the 2009 runoff for Public Advocate was just over 230,000 votes, and in 2001 the Public Advocate's runoff election brought out almost 670,000 voters (in part due to the mayoral runoff election that year).
The slider above the map at right reveals predominant race/ethnicity patterns compared with the vote results (and the drop-down list will show other demographic patterns). The interactive map available by clicking the "Public Advocate runoff" tab above enables you to click on the map to reveal actual vote counts, vote share, and turnout percents in each Election District.
Bill de Blasio's victory in the Democratic primary was suprising to many, partly because he was an early underdog candidate in a crowded field (eight other candidates including the former NYC Comptroller who also ran for mayor in 2009 and almost won, the current Comptroller, City Council speaker, and a high profile former Congressman).
But most observers also assumed the city's Democratic voters would largely follow earlier electoral patterns of "identity politics" in which black voters would support William Thompson, the former comptroller and an African American, recent immigrants would support John Liu, the current comptroller and an immigrant himself from Taiwan, and native-born whites of various ethnicities and ancestries would split their vote among de Blasio, Christine Quinn, Anthony Weiner, and perhaps several of the lower profile candidates.
De Blasio's support (more than 260,000 votes) was widespread and seemed to run counter to the "identity politics" narrative. In particular, it exceeded Thompson's support in predominantly black communities where Thompson himself did well in 2009. At the same time, Liu's support was limited to communities with predominantly Chinese and other Asian populations (an example of identity politics, but backed up with relatively few votes — just under 45,000). And Quinn (99,700 votes) and Weiner (just over 31,000 votes) received fewer votes combined than Thompson (168,000).
The maps on this page highlight these patterns and explore the variations within.
2013 primary results (NYC BOE)
These maps display the most recent citywide Democratic primary & runoff results for 4 of the 7 current Democratic mayoral contenders: William Thompson, John Liu, Bill de Blasio, and Anthony Weiner.
2009 election results (NYC BOE)
2005 election results (NYC BOE)
Three current Democratic mayoral contenders ran for mayor (William Thompson), Comptroller (John Liu), and Public Advocate (Bill de Blasio) in the 2009 general election. These maps show the results.
2009 election results (NYC BOE)
The purpose of these two maps is to show how strongly one African American candidate, Barack Obama, won the 2008 general election, while a second, William Thompson, had much more partial support only one year later in 2009. In short, many areas that were willing to support a black Democratic nominee for high office in one year shifted to the incumbent Mayor, running on the Republican and Independence lines, in 2009.
2009 election results (NYC BOE)
2008 election results (NYC BOE)
Election results for the Democratic mayoral primary and runoff election, and general election, in 2001.
2001 primary results (NYC BOE) [PDF]
2001 runoff results (NYC BOE) [PDF]
2001 general election results (NYC BOE) [PDF]
How did Michael Bloomberg's opponents (and therefore Bloomberg) do in each general election: 2001, 2005, & 2009?
2009 election results (NYC BOE) [PDF]
2005 election results (NYC BOE) [PDF]
2001 election results (NYC BOE) [PDF]
These maps show that Thompson was not the only candidate for top office to do less well than Barack Obama in 2008. Indeed, Obama himself did less well in 2012 than he had in 2008, while Democratic gubernatorial nominee Andrew Cuomo also did less well. At the same time, these maps show that voting in elections for chief executive of the state and country has a different make-up than voting for the city's chief executive.
2012 election results (NYC BOE)
2010 election results (NYC BOE)
2008 election results (NYC BOE)
These maps of demographic and socio-economic characteristics are provided for reference purposes. The Center for Urban Research prepared them originally for the NYC Districting Commission as part of a paper on the concept of "communities of interest" in New York City.
UPDATE: includes turnout in 2013 mayoral primary.
Turnout in the 2009 Democratic mayoral primary, 2009 mayoral general election, and 2012 presidential election.
In order to visualize the patterns of voting results from 2001 to 2012 at the local level across the city, we used the smallest geographic area possible. The NYC Board of Elections records local voting results by Election District. There are currently almost 5,300 election districts covering the city with an average population of 1,500 people. (In comparison, Census tracts are larger – there are 2,100 tracts with an average of 3,700 people – and Census blocks are much smaller – there are approximately 30,000 populated blocks with an average population of less than 300.)
But election district (ED) boundaries change each year, especially after redistricting. In 2010, for example, there were 6,300 EDs. By 2012 that number had dropped to less than 5,300. Therefore we could not directly compare the vote counts by ED from one year to the next. Also, election district boundaries are not coterminous with Census geography, making it difficult to directly compare vote results to demographic data reported by Census tract or block.
Therefore, we decided to allocate ED-level vote counts to Census blocks. This provides an apples-to-apples spatial comparison of the local voting patterns from one year to the next, and also allows for easy analysis of voting patterns with Census data.
Allocating population data (in this case, voting population) from one type of geography to another can be accomplished using several methods. This is a common practice with redistricting, when voting data by precinct needs to be allocated to Census blocks, which are literally the building blocks of legislative districts.
The methodology we used was to determine the proportion of each ED's population that was located in each overlapping area between EDs and Census blocks. If an ED was wholly contained within a block, 100% of the votes from the ED were allocated to the block. If 50% of the ED overlapped a block, only half the votes were allocated to the block.
This effort was more involved than other attempts to allocate population from one geography to another. First, we wanted to allocate the voting results from at least eight years worth of election districts to the current (2010) Census blocks. We allocated the results from elections from 2001 to 2012 (with the exception of the few years when there were no citywide elections of significance). We included primary (and runoff) elections and general elections for Mayor, Comptroller, and Public Advocate at the city level, and Governor and President at the state and federal levels – we wanted to focus on executive offices (consistent with analyzing the race for mayor).
Two other factors adding to the complexity of the allocation process were the number of candidates and the population base that we used to determine the overlap between EDs and Census blocks. All told, we allocated voting results for more than 40 candidates. In 2009 alone there were 13 major-party candidates for the elections noted above.
Regarding the overlap between EDs and blocks, there are several approaches that are typically used. Our approach was to determine the number of voters registered in each year we examined, and use each ED's count of registered voters as the denominator and the number of active registered voters in each overlapping area between EDs and blocks as the numerator. If five voters were located in an overlapping area, and the ED's voting population was 25, twenty percent of the ED's votes in that year for each candidate were allocated to the overlapping block. The results of those calculations were then summed by block, for each candidate in each election in each year, to determine the block-level totals.
(As an aside, each year's ED boundaries from the Bytes of the Big Apple program had to be cleaned using ArcMap's "repair geometry" tool before comparing the ED geometry with Census blocks. Otherwise, the allocation results were bing confounded by stray, seemingly random and invisible ED boundaries. The Repair Geometry tool removed these phantom EDs and the subsequent results were internally consistent and sensible.)
The schematic outline below summarizes the allocation process using EDs and tracts:
To determine the exact count of voters in each overlapping area, we geocoded each year's active registered voters from registration files provided by the NYC Board of Elections. Although there were duplicate records from year to year (people who remain registered at the same address from one year to the next), we needed a separate population base for each year's voting results. Therefore, we geocoded more than 18 million voter registration records over the period from 2001 to 2012. We successfully matched more than 99% of these records, due to the address quality in the Board of Election files and our geocoding system refined over several years of use with city data.
Maps of election district results tell only part of the story of electoral outcomes. Without also accounting for voter turnout, choropleth maps that just show voting results treat all outcomes as equal (in terms of color intensity), and ignore the importance of a candidate's ability to turn out his/her base of support.
After testing different color styles and cartographic techniques, CUR used the value-by-alpha approach that uses color to indicate value (in this case, a candidate's share of votes) and levels of transparency of those colors (the "alpha" component of a color scheme) to indicate intensity of that vote share as measured by turnout.
The Center for Urban Research obtains election data from the NYC Board of Elections and uses SPSS to restructure the files into a database format suitable for analysis. The Board of Elections provides election results by Election District (ED), and all registered voters citywide.
CUR's John Mollenkopf, Steven Romalewski, and Kristen Grady worked together to develop a methodology to allocate ED results to Census blocks (see Methodology section). Steven geocoded the voter registration records for active voters to develop a voter population base for the allocation process. Kristen, a PhD student in Earth and Environmental Science at the CUNY Graduate Center, used ESRI's ArcGIS for Desktop to allocate the ED-level results to Census blocks. The allocation is a tedious and painstaking — but essential! — multi-step process involving several spatial calculations comparing each year's ED boundary file (typically 5,000 to 6,000 ED polygons) with the 2010 Census block file (almost 40,000 polygons) and attribute calculations for each candidate in each year's elections. So Kristen developed a set of Python scripts and ArcMap modules to automate the process of allocating ED data to blocks for each of the candidates in each year's primary, runoff, and general elections (more than 40 candidates between 2005 and 2012).
The ED boundaries from each year were downloaded from the NYC Department of City Planning's Bytes of the Big Apple program. Census block boundaries are from the Census Bureau's TIGER files. CUR used the MapInfo Professional GIS application to geocode voter registration records against the Department of City Planning's LION file representing street centerlines.
CUR's Romalewski produced the maps with ArcGIS for Desktop based on a technique developed by Andrew Wheeler, a PhD student at the University at Albany, SUNY. Other cartographers and election analysts offered very helpful feedback along the way.
Interactive versions of the maps are in the works! Stay tuned.
SPSS and ESRI's mapping software applications were used for data analysis and cartography. MapInfo Professional was used for geocoding. David Burgoon, CUR's application architect, customized Twitter's Bootstrap framework to create a website providing easy comparisons across maps. He also customized a slider plugin for Bootstrap to integrate the overlay maps of demographic patterns, neighborhoods, and Council district boundaries.
The CUNY Graduate School of Journalism and its Center for Community and Ethnic Media have partnered with the Center for Urban Research at the Graduate Center / CUNY to develop the Atlas. The Journalism School and CCEM have provided generous funding support to provide public access to the Atlas's maps and data and to ensure that Journalism School students and CCEM media outlets benefit directly from this work.
The Graduate Center's information technology team and communications and marketing staff have provided essential support for the project.