An analysis of differences in success rates of male and female debaters.

• Author: Bruschke, Jon

The past few years have witnessed a heightened awareness of gender issues in NDT debate. There are a wide variety of issues which have come to the surface recently and it is important that we extend the discussion and debate on the multiplicity of concerns in order to comprehend the subject more fully. The particular issue this paper will address is the possibility that females and males have their debate performances evaluated differently. In NDT debate it may be difficult to identify an overall pattern of results that can be called sex bias. It is difficult to distinguish between a lack of success due to discrimination and a lack of success to a lack of talent, experience, or evidence. However, we can gain sight of consistent bias by examining the success of a large number of participants and trying to identify differences that can be attributed to gender and not any of the many other factors that come into play in determining who wins or loses a debate round. We fee6l that much can be learned by identifying the ways that debaters may have to compete on an uneven playing field and attempt to understand how these conditions are created.

Gender differences have been pervasively examined in persuaded research. For example, it is not a typical for researchers to focus on differences in how men and women are persuaded Cooper, 1979; Cantor, 1976; Eagly, 1978; Leventhal & Cupchik, 1976; Losco & Epstein, 1975; Zillman & Stocking, 1976), differences in styles of communication according to gender (Andrews, 1987; infante, 1989, Mulac, Torborg & Bradac, I986), and how audiences react differently to male or female speakers (Burgoon, Dillard & Duran, 1983; Richmond & McCroskey, 1975; Ward, Seccombe, Bendel & Carter, 1985).

In relation to competitive speaking events, there have been examinations of differences in speaking styles (Larson & Vreeland, 1985), different rates of participation for competitors (Dean & Dean, 1985; Friedley & Manchester, 1985; Nadler 85) judges (Kay & Aden, 1984), and differences in judging styles according to gender of the judge and speaker (Friedley & Manchester, 1987).

Although there is a growing awareness among some in NDT about gender issues, the activity has not been very self reflexive in the recent past. Although Friedley and Manchester (1985) explored the female participation rate in elimination rounds during the 1984 National Debate Tournament, we wanted to examine the overall success rate of participants in relation to their gender. We were able to find three studies which examined this question, the most recent of which was about fifteen years old and studied high school debaters (Rosen, Dean, & Willis, 1978). Collegiate success rates have not been examined since 1972 (Hayes & McAdoo, 1972; Hensley & Strother, 1968). The most recent study has concluded that the state of research on the topic reveals that "mixed-sex teams appear to be superior, as indicated by the won-lost records in the present study ... An examination of neither ... rankings nor quality points reveals a sex difference in the present study" (Rosen, Dean, & Willis, 1978, p. 20).

The need for a more recent examination is manifestly clear if for no other reason than much has changed in the intervening two decades. The average speaker points found by Rosen, Dean, and Willis, for example, ranged between 16 and 18. Additionally, speakers in the first position could, on average, out-point speakers in the second position, something implausible in the modern NDT context. in short, one is pressed to find contemporary relevance of the Rosen et al. data to collegiate NDT debate because, in addition to the differences noted above, they drew from a high school sample using a 25 point ballot. Further, in the Hayes and McAdoo study, ranks rather than speaker points were considered, a decision which would seem odd in the modern debate context. Ranks, when considered at all, are usually only the fourth or fifth tie-breaker for speaker awards.

In particular, we wanted to discover if there were differences in speaker points and the number of wins that could be attributed primarily to the gender of the speaker. The obvious difficulty is how to control for the influence of talent. Our approach was twofold. First, we examined debates that occurred in high-high power matched rounds. The effect of this process was that rounds where competitors were matched as evenly as possible in relation to success rates were heavily weighted. Second, we statistically controlled for differences due to prior record by treating prior success at the tournament and victory in the round as covariates, that is, the differences due to those factors were statistically "pulled out" of the data and the remaining differences should be due primarily to gender.

Methods

Sample and Collection Procedures

Data were collected from results sheets from several large intercollegiate NDT debate tournaments which were held during the 1989 through 1992 debate seasons. The tournaments were conducted in geographically diverse regions, but to generate a sufficient sample size larger national tournaments were predominantly used. Every team listed on the results sheet which contained a female debater was included. In instances where we were unable to confirm the gender of the debater or the judge the round was excluded from analysis. A total of 136 teams were analyzed. Because teams may have competed at more than one tournament some debaters and judges may be over-represented in the sample. The speaker points for rounds two through seven were analyzed for each speaker on the team, producing 1904 speaker ratings for analysis. One-hundred and seventy-two cases were excluded because of data anomalies, leaving 1720 cases included the final analysis.

Because there were more male debaters than female debaters we randomly excluded male/male teams where necessary to even out cell sizes. Although there was a temptation to take the stance that "more data is better", the random exclusion of male/male teams was guided by a balancing teams was guided by a balancing of the benefits and drawbacks of their inclusion. Since there were more males than females in the sample, including all of the male/male teams would grossly exaggerate differences in cell sizes. We were guided by Keppel's (1991) advice:

Unfortunately, there are no useful rules of thumb that specify the point at which unequal sample sizes and variance heterogeneity become a serious problem. Milligan, Wong, and Thompson (1987) discuss several possible alternative courses of action but none is as effective as avoiding unequal sample sizes in the first place. (p. 283)

The main advantage to including more data is that you can reduce error due to sampling quirks, thus reducing the chances of type II error (which is accepting the null when true differences exist). In other words, you increase the "power" of the statistical test or its ability to detect true differences. As Pearson and Hartley (reproduced in Keppel, 1991) demonstrate, once sample sizes rise above 60 there are only slight increases in power to be gained by adding elements to the sample. Since the male/male cells had over 200 data points already, adding further data would only minimally increase power. Therefore, adding all male/male teams to the data would further unbalance cell sizes but would not substantially increase power. Random exclusion to attempt to mitigate cell size imbalances seemed the...