The data and calculations presented in Table 1 support this conclusion and extend the measure to 1930. The table presents the percentage of Republican candidate votes for Governor and Senate, for both Tulsa County and the balance of the state, plus Tulsa County's percentage lead for those races and the percentage differences between leads as an indication of partisan deviation. In 1974, 7.7 percent more people voted Republican in Tulsa County compared to the rest of the state for Governor than Senate. The data show that Tulsa County is more different from the rest of the state in 1974 than in any time since 1938; that it is considerably different from the post-war pattern; and that its difference in 1974 is over twice as high as the mean of the differences (4.1) from 1930 to 1966.

With more time available for data collection and analysis, a measure of ticket-splitting has also been developed which is more informative than a discussion of partisan deviations and which is comparable to those employed in reputable and previously-cited studies comparing the American states. The measure also takes a greater proportion of the ballot (s) into account (beyond Senate and Governor) by analyzing all contested statewide races, except judicial contests. It estimates the magnitude of ticket-splitting as the difference between the highest Republican percentage of the vote for a statewide office and the lowest Republican percentage of the vote for an office in the same election. For example, if the highest Republican percentage of the vote was cast for Governor (e.g., 60 percent) and the lowest Republican percentage for any statewide office was cast for Corporation Commissioner (e.g., 40 percent), the difference between the percentages

would measure the magnitude of ticket-splitting (20 points). In the above example, it is estimated that 20 percent of those people who voted for the Republican Gubernatorial candidate also voted for a Democratic Corporation

1974 Senate

Governor

Table 1

Difference Between Tulsa County's Republican Percentage
Lead Over the Balance of the State for Coterminous

Gubernatorial and Senatorial Races: 1930-1974

Commissioner. Such measures were calculated for all statewide races which

included both Senate and Governor since 1930; during which period party circles or machine levers were used (with the exception of Tulsa County in 1974). Separate calculations were made for Tulsa County and the remainder of the state. As ticket-splitting increases substantially in Tulsa County in 1974, it is indicative of the impact of discrepant ballot/machine arrangements over any other peculiarities (uniqueness) associated with events, issues, and candidates which vary widely from year to year over the entire period.

The data presented in Table 2 confirm the impact of the absence of party "levers" in Tulsa County in 1974. The amount of ticket-splitting in Tulsa County that year is substantially larger than it had ever been in Tulsa County or the rest of the state in any previous years in over four and one-half decades. Its value (29.1) is 10.7 percentage points higher in 1974 than ever before in Tulsa County, and ticket-splitting in the remainder of the state was only 4.2 percent higher than ever before. The amount of ticket-splitting in Tulsa in 1974 was 14.3 percent higher than the mean ticket-splitting in the county from 1930 to 1966. In addition, the average ticket-splitting for all observations in Table 2 where party levers/circles were used (state balance and Tulsa values without 1974) is 13.6--a full 15.5 points lower than the one current case without party

levers.

Roll-Off and the Effects of Erroneous Instructions

In addition to the effects of the absence of party levers on splitticket voting there is evidence of voter confusion in this case resulting from (1) the general absence of levers which are normally present, or (2)

Table 2

Percentage Difference Between the Highest Republican
Percent of the Vote for a Statewide Office and the
Lowest Republican Percent of the Vote for a
Statewide Office in Tulsa County Compared
To the Balance of the State: 1930-1974

the presence of additionally confusing instructions about how to use nonexistent levers in all but 46 of the 315 Tulsa county precincts. to the latter, it has been contended that many voters who attempted

With regard

to follow such instructions would register a vote for the "bottom of ballot" candidates in the false hope of effectuating a straight ticket vote for

the entire column.

In order to address such issues, we must first validate the concept of "voter fatigue" or ballot "roll-off," i.e., the tendency for the electorate to vote for higher/prestige offices but not for lower offices on the same ballot at the same election. This phenomenon has been well-established in the professional political science literature encompassing a large number of states over time (Walter Dean Burnham, "The Changing Shape of the American Political Universe," American Political Science Review, 59 (1965), 7-28) and in studies of the impact of ballot form (e.g., Walker, op. cit.). Such rolloff has been attributed to public indifference to a large number of elective administrative offices, to an increasing number of peripheral voters in the active electorate, and especially to the form of the ballot. Roll-off as

Norman:

an empirical fact is well-established for Oklahoma--since the 1920's the average percentage spread of vote totals for the top and bottom offices on the ballot is approximately 15% (Burnham, op. cit., p. 20; and Samuel A. Kirkpatrick, David R. Morgan, and Thomas Kielhorn, The Oklahoma Voter. University of Oklahoma, forthcoming). Although roll-off or fatigue may not be perfectly linear, an increase in total votes cast for bottom of ticket offices appears to be a rare phenomenon. If fatigue is operating normally there should not be a disproportionate increase in bottom lever voting.

63-007 O-76-7