By Thomas Hargrove
Scripps Howard News Service

The Series Part 1: Analysis says vote system failing Part 2: How the vote study was done Part 3: 12 states don't even tally voters

Scripps Howard News Service conducted an extensive computer study of election returns in the 1996 presidential contest to determine if votes were misreported or miscounted.

The study of election undercounting - using information obtained from Election Data Services of Washington, D.C. - compared how many votes for president were reported in 1,825 counties against the total number of ballots reported cast.

The drop-off rate was calculated by dividing the number of presidential votes counted by the number of ballots reported to have been cast. County officials were contacted and queried about unusually high drop-off rates.

A common source for apparent undercount error came from misunderstandings by county officials over which data to report. Warren County, Ind., for example, reported to state officials in Indianapolis that 5,201 ballots were cast in the Nov. 5, 1996, election. That actually was the number of voters registered in the county at the time.

Thirty-eight states have regulations or statutes requiring county officials to report how many ballots were cast. But county-by-county data was not available in Colorado, Kansas, Virginia or the District of Columbia.

Twelve states do not count the number of voters going to the polls: Alabama, Arkansas, Maine, Mississippi, Missouri, North Carolina, Oklahoma, Pennsylvania, Rhode Island, Tennessee, Texas and Wisconsin.

Scripps Howard also tallied the total number of votes for U.S. senator and governor if such races were held. In states that do not record ballots cast, the study made a comparison of the total presidential votes against votes for other offices.

The calculation of median drop-off rates by various groups (states, types of election equipment used, etc.) was not statistically weighted according to population size. Since the study sought to identify problems in voting procedure, each county was given equal weight in determining the overall health of election practices in the United States.

The study also examined the size of the electorate in each county and compared the number of registered voters against the Census Bureau's estimates of voting-age population in every county to determine the quality of maintenance of the registration rolls.

The data was analyzed using statistical analysis software from the Statistical Package for the Social Sciences - SPSS. Scripps Howard News Service is responsible for all interpretation of the data.