The years 1966, 1970, 1974, 1978, 1982, 1986, 1990 and 1994 were chosen for this study because it would have been unmanageable to study every year from 1966 to 1994. Also, this stratified sample affords a look at trends for four periods before the 1978 settlement of the Women's Caucus lawsuit against The New York Times -- and for four periods following the settlement. It was thought 1978 could be counted as a pre- settlement period because the settlement did not come until Oct. 6, 1978 and The New York Times was then in the midst of a four- month strike that lasted until mid-November 1978. Since the settlement did not call for any immediate hiring or promotion of women, it seemed the affect of the settlement would not show before the end of 1978, but would be blooming by 1982, the next year studied.
Along with the front page of each New York Times in the sample, also studied were the business section front, the first page of arts news or reviews, the op-ed page, letters to the editor, obituaries and the Metropolitan section front. The New York Times, however, did not have a labeled metro section front before late 1978, or in many Sunday editions after that. For those dates in the sample, no metro news was counted. Also, on the obituary pages, only the staff-generated obituaries in regular-size type were counted, not the paid obits in agate. All of The New York Times issues in the sample were examined on microfilm or microfiche. It took about 20 minutes to analyze each edition for gender.
This study replicated the counting protocol used in the Women, Men and Media studies. This involves counting the number of male and female bylines, pictures and references on each section front (jumps of stories are ignored).
The trickiest, most arduous part of the content analysis involved counting the references to men and women within stories. Following the guidelines of the Women, Men and Media studies, it was counted each time a man or woman was referred to in a bylined story by name. Subsequent gender-tinged references such as "he said" "she ran" or even "the President said he would decline..." were not counted. To be counted, the name had to be readily identifiable as that of a man or a woman, either by the first name or a subsequent reference to gender. The small percentage of names that could not be identified in that manner were ignored. The gender of names was not identified by context alone; e.g. it was not assumed all officials from the Middle East were men or all school teachers named Leslie were women. Each gender-identifiable name was counted once on first reference and again every time the name was used. So a story referring to "Mayor Koch," "Koch," "the Mayor" and "the Koch Administration" would be counted as having three references to men (not four, because "the Mayor" does not explicitly include a male name).
In counting photos (and line drawings or other graphics that depicted humans), any picture with at least one man and one woman visible was counted twice -- once for males and once for females. This was true even if there was only one woman whose face or figure was barely identifiable far in the background of a portrait of a man, or among a sea of men. (The opposite also held true, as was the case in a photo of one man tagging along in a women's march). By contrast, if the sex of the people in the picture could not be identified by sight or cutline -- as in a photo showing construction "workers" in silhouette -- it was not counted for men or women.
As for bylines, multiple bylines on a story (or letter to the editor) were counted separately, tallying one for each man or woman listed among the authors. Bylines that did not readily indicate the sex of the writer were recorded in a separate column. After all dates in the sample were analyzed, all of the uncategorized bylines were submitted to two longtime staffers of The New York Times, Nan Robertson (personal communication, February 3, 1995) and Carolyn Lee (personal communication, March 20, 1995). They identified the sex of 33 writers, leaving just 10 bylines uncounted for gender among the 6,310 encountered in the sample. That means the sample ultimately included 6,300 articles or letters to the editor, not 6,300 different writers. Many reporters wrote multiple stories included in the sample.
Also, an uncounted but small percentage of the names of letter-to-the-editor writers had to be thrown out, because there was no practical way to contact them or anyone who knew their sex. Many of these letter writers signed their names with initials for their first and middle names. These may have been women hoping to improve their chances of publication by concealing their sex, or they may have been men impressed by the initials-only convention of many prominent Times writers, such as A.M. Rosenthal, C.L. Sulzberger and R.W. Apple Jr. In any case, the sample still included 1,476 letter-to-the-editor writers whose sex was readily identifiable by their names.
All of the coders were given one hour of training on these methods and they all coded the same front page for practice. The intercoder reliability on that page was 92 percent. Spanning the sample, the intercoder reliability would be even higher, because 135 of the 224 days were coded by the researcher and his wife. The 37 undergraduates who participated coded a total of 89 days, with no one of them coding more than four editions. To further standardize the analysis, every coder used copies of the same tally sheet (Appendix 1) for every day.
The data were keyed into a Works spreadsheet, initially sorted in a Works database and then translated into SPSS/PC+ Studentware Plus. Independent t-tests were performed on all variables for both sexes, comparing the means for data from 1966 to 1978 with those from 1982 to 1994. In other words, The New York Times' coverage of women before the Women's Caucus settlement was compared to its coverage since then, to see if there have been statistically significant increases. Also, linear regression was applied to all the variables to see if hypothesized trends were significantly evident. Lastly, multiple regression was used to measure which variables correlated most with increases in the Times' references to women.
Back to Times women index page.