Publications

The survey asked respondents to report on the number of articles they had published in refereed journals and in refereed conference proceedings during the 3 years prior to the survey. Data for sole authorship and co-authorship were combined into a single variable.

We looked first at journal articles published in refereed journals. Overall, male faculty published marginally more journal articles in the past 3 years than female faculty (men, 8.9 articles; women, 7.4 articles). It is important to note that these statistics and those that follow related to publications could be misleading, given the significant interactions discovered in our multivariate analysis of gender, dis­cipline, publications, and other variables. Disaggregated by field (see Figure 4-7), men appear to publish more papers than women in chemistry (men, 15.8; women, 9.4). The differences between men and women in mathematics and physics were smaller, with women publishing more than men in electrical engineering.

We then looked at the total number of publications in refereed journals and conference proceedings combined. Overall, there appears to be no difference between male and female faculty in the total number of publications (men, 13.9; women, 12.8). Disaggregated by field, men published significantly more than women in chemistry, but not in any other field; women published marginally more in electrical engineering, but not in any other field.

There are two differences between our survey data and the data from the NSOPF:04. The first is that we only asked about articles in refereed journals and in refereed conference proceedings, while the NSOPF:04 asked about articles both in refereed and nonrefereed journals, as well as books, textbooks, reports, and presentations. The second is that we asked respondents to sum information over the previous 3 years, while the NSOPF:04 asked about the past 2 years. A sum­mary of the NSOPF:04 data is shown in Table 4-1. In these data, male faculty had significantly more publications than female faculty in the previous 2 years (men, 10.9 publications; women, 8.2 publications). Looking at the gender differences in the various subcategories, however, we find that the only significant difference between men and women was in articles in nonrefereed journals, a category we did not include in our survey.

Both the faculty survey and the NSOPF use simple numerical counts as measures of publications. Counts are one sensible approach, but they have their problems. “Simple counts of articles and books published account for neither

quality nor the importance of scholarship” (NSF, 2004b:8). Alternative approaches include weighting publications by the prestige of the source (e. g., top journals or university versus commercial presses for books) and counting citations of the publications to measure the impact of the faculty member’s research. Both of these approaches are very difficult, but as the debate over quantity is increasingly clari­fied, taking an approach such as one of these may be fruitful in the future.

Next, we asked which variables might contribute to the number of articles a faculty member published in refereed journals and conference proceedings in the past 3 years. (Again, given the interactions, this more conditional look is more likely to accurately reflect the nature of the impact of gender and discipline on number of publications. Specifically, because disciplinary area interacts with gender and number of publications, one cannot directly interpret the effect of discipline in isolation from gender, and gender in isolation from discipline.) First we looked at the number of refereed journal publications. This model was fit to 1,404 observations corresponding to full-time faculty, tenured or tenure-track. Only 934 (of the 1,404 faculty) had complete information on all covariates in the model and had reported a number of journal articles. The number of journal publications is a count variable, making a Poisson model plausible for this out­come variable. We found, however, that a normal distribution was also a plausible model because the number of journal publications varied between zero and 40 with a mean of about 9. Therefore, to facilitate interpretation of results, we fitted an ordinary linear model to the number of journal publications and included the

following covariates: gender, discipline, faculty rank, type of institution (public or private), prestige of institution, percent of time spent in research activities, having or not having a mentor, and all of the two-way interactions between gender and the other factors. The R2 for the model was 19.0 percent (0.19).

Significant effects in the model were discipline (^-value < 0.0001), gender (^-value = .0001), rank (^-value < 0.0001), prestige (^-value = 0.0012), indica­tor for mentor (^-value = 0.005), percentage of time spent in research (^-value = 0.0001); and the three interactions gender with discipline (^-value = 0 .037), gender with rank (^-value = 0.042), and gender with mentor (^-value = 0.049).

Appendix 4-19 contains the least-squares (or marginal) mean number of journal publications for each level of each combination of fixed effects, and the lower and upper bounds of the 95 percent confidence intervals around each mean. Discipline has a very significant impact on the number of publications, as does gender, rank, prestige, and presence of a mentor. Also, discipline, rank, and pres­ence of a mentor had modestly significant interactions with gender. Regarding the interaction of discipline and gender, we can assert that men publish more journal articles than women in biology; men publish more than women in chemistry; there is no significant difference between men and woman in mathematics, in electrical engineering, or in civil engineering; and men publish a borderline significant more than women in physics. Regarding the interaction of rank by gender, men increase the number of journal publications between the ranks of assistant and associated more than women do. The difference in the degree of increase from associate to full professor is less pronounced between the two genders. Regarding the interac­tion of mentor with gender, the difference between number of journal publications between men and women is more pronounced when faculty have mentors. Finally, the number of journal articles increases by 0.06 when a faculty member spends an additional 1 percent of his or her time in research activities.

The same analysis was then carried out in modeling total number of refereed publications (journal articles and refereed proceedings). Due to data quality issues, this analysis was conducted on 1,019 faculty members who reported both their refereed journal publications and their refereed proceedings articles, and whose values were not unrealistically high. Of the 1,019, only 774 had complete covariate information. The least-squares regression model with the same covari­ates obtained an R2 of 23 percent. Since in this case no interactions were signifi­cant, it is easier to interpret the main effects. The significant effects were those of discipline (p-value < 0.00001), gender (p-value = 0.04), rank (p-value < 0.0001), prestige (p-value = 0.0002), mentor (p-value = 0.01), and percent time spent in research (p-value = 0.04). The marginal means and their 95 percent confidence intervals are provided in Appendix 4-19. Again, since none of the interactions among the main effects was significant, all the effects in this appendix can be interpreted in a straightforward manner. For instance, electrical engineers publish the most, followed by chemists, physicists, and civil engineers. Also, men publish more than women, and full professors publish more than associate professors, who in turn publish more than assistant professors. Furthermore, those at prestigious institutions publish more than those at less prestigious ones, and having a mentor increases the number of publications. Finally, since the regression coefficient of percent research time on fitting total number of refereed publications was 0.045 (with a p-value of 0.04), when research time increases by 1 percent, one can estimate that that will be accompanied by an increase in total publications of.045 per year.