next up previous contents
Next: Sample Uses for Our Up: Overview Previous: Overview

The Units of Analysis: Precincts and ``MCD-Groups''

We provide data at the precinct and MCD-Group level, each of which makes somewhat different information available:

Precinct-Level Electoral Data
This unit of analysis is the ballot box or voting booth. In most counties, eligible citizens in a geographically contiguous area are assigned to a single voting booth to cast their ballot. We refer to the geographical description of the unit of analysis as a precinct, although in some states it is called an election district, and occasionally several voting booths exist within one precinct. The number of precincts nationwide is approximately 170,000, with some variation over time. More detailed official electoral data than precinct returns cannot be created for the entire nation unless the laws governing the secret ballot are revoked.

Our collection includes data from each precinct during 1984-1990 for general elections for all federal offices (President, U.S. House, and U.S. Senate), all partisan statewide elections (such as Governor, Lt. Governor, Secretary of State, Attorney General, Treasurer, Secretary of Insurance, College Trustees, etc.), all state legislative elections (State House and Senate), some statewide ballot initiatives and referenda, and party registration and enrollment where available. The data also include numerical and character descriptions of community, precinct, and district names. The collection includes an average of 33 variables for each state and year, for a total count of 6,628.

Since very few states had established centralized election data collection mechanisms or archives, the original raw precinct-level election data were collected directly from individual local election authorities in every state. Collection was undertaken during 1989-1991. City and town election authorites in New England, and county election officials in other states, were contacted first by letter, and subsequently if necessary by additional letters and phone calls. Missing data in the final collection is due to the lack of responses (after numerous attempts via letters and phone calls) or the absence of physical material to send (i.e., the returns were destroyed or never located). Essentially everything was done to avoid missing data short of physically visiting each locality.gif

``MCD Group''-Level Electoral and Census Data
This geographic designation is one we devised. Each one combines some number of what the Census Bureau calls Minor Civil Divisions or MCDs (or ``Census County Divisions'' when MCDs are not well defined, although we refer to them both as MCDs). Individual MCDs correspond roughly to towns and cities. There are 25,973 of these MCD-groups in the U.S. outside of California and an additional 21,200 within California, and completely tile the U.S. land mass. For each of these units, we provide all our electoral data aggregated from the precinct level, as well as 3,725 variables from the U.S. Census long form. This is the smallest level of aggregation at which publicly-available census data and electoral data can be made to coincide. The ROAD map files make it easy to draw maps of MCD-Groups with any of our variables.gif

Our data include a variety of geographic codes that can be used to create aggregations from these data at many levels (such as various legislative districts).

In addition to cleaning and documenting the precinct-level election data, a fundamental problem this project addresses is that geographic units used by the U.S. Census Bureau to provide socioeconomic demographic data do not match the electoral precincts used by U.S. counties to provide voting data. For this project, two forms of Census data are relevant:

PL94-171
These data were created for redistricting purposes and only include counts of the numbers of people by race and (to some extent) age. This data is form the Census Bureau's ``short form.''
STF3a
These are the Bureau's ``long form'' data, representing 3,725 economic and demographic variables.

Wherever possible, we use the STF3a data.


next up previous contents
Next: Sample Uses for Our Up: Overview Previous: Overview
Copyright © 1997-2004 [ROAD Home] Questions? Contact the ROAD webmaster.