User:The Vintage Feminist/Antin et al graphics

"Gender Differences in Wikipedia Editing" by Antin et al, describes the experiences of editors in the first three weeks of editing. This essay converts the results tables in their report into a set of at-a-glance graphics to provide a user-friendly way of considering their findings.

It strongly recommended, therefore, that this essay be read in conjuction with their report.

Background

The report authors were interested in whether gender stereotyping rather than "active sexism" was the reason for the discrepency in the number of female editors compared with male editors on Wikipedia.

Gendered Wiki-work

The report considers the different types of "Wiki-work" that exists, from uploading photos to arbitration. It suggests that it is essential to discover which types of work draw which gender, what "are the real or imagined characteristics of tasks that encourage certain individuals to gravitate towards them", and to what extent stereotypes play a part. In order to grow Wikipedia's user base and encourage diversity, the report authors say that, researching wiki-work from a gender perspective "could allow recruitment and educational efforts to be more focused..." giving "potential contributors more specific ideas of the types of work they might like to do.

Method

Base population

Figure 1: 256,190 new user accounts created between 9 September 2010 and 14 February 2011

  242,592 users who did not declare a gender in their Wikipedia profile (95%)

  13,598 users declared a gender in their Wikipedia profile (5%)

Between 9 September 2010 and 14 February 2011, 256,190 users created an account on the English-language Wikipedia. Of these 13,598 optionally declared a gender in their Wikipedia profile (blue segment, figure 1, right).

The 13,598 gender-declaring Wikipedians were made up of, 11,194 (82%) men, and 2,402 (18%) women (figure 2, left). A ratio that, the report authors' note, is higher than the one quoted in the 2010 UNU-MERIT study (13% women).

Figure 2: 13,598 users declared a gender in their Wikipedia profile

  2,402 users who declared themselves to be female (18%)

  11,194 users who declared themselves to be male (82%)

Sample 500 editors

A sample of 500 users was then extracted from the gender-declaring Wikipedians (figure 3).

 

 

 

 

Figure 3: 500 extracted Wikipedians (from the 13,598 gender-declaring Wikipedians)

  250 female Wikipedians - 50 %

  250 male Wikipedians - 50 %

How the four quartiles were formed: The breakdown of the 500 editors
The 500 editors were then divided into four quartiles based on how prolific their revisions (edits) were during their first three weeks on Wikipedia. The four quartiles were made up of editors who had made 0, 1, 2, or 4+ revisions (figure 4a).

Figure 4a: Target sample of 500 new editors:

Number of edits during his/her first three weeks	Number of editors ( % of 500 editors )
0 edits (making up 25% of the bottom 75% prolific editors)	125(25%)
1 edit (making up 25% of the bottom 75% prolific editors)	125(25%)
2 edits (making up 25% of the bottom 75% prolific editors)	125(25%)
4+ edits (making up the top 25% prolific editors)	125(25%)
Source: Antin et al

There were three weeks of data unavailable for 63 editors (figure 4b).

Figure 4b: Editors where the data was unavailable:

Number of edits during his/her first three weeks	Number of editors ( % of 63 editors )
0, 1 or 2 edits (bottom 75% of prolific editors)	62(98.41%)
4+ edits (top 25% of prolific editors)	1(1.59%)
Source: Antin et al

This left an adjusted number of sample editors of 437 (figure 4c).

Figure 4c: Adjusted sample of 437 new editors:

Number of edits during his/her first three weeks	Number of editors ( % of 437 editors )
0 edits	182(41.65%)
1 or 2 edits (both quartiles combined)	131(29.98%)
4+ edits	124(28.37%)
Source: Antin et al

Gender break down of the top and bottom quartiles

Bottom 75% of prolific editors (figures 6 and 7)

Figure 5: Editors making 0 edits in his/her first three weeks^*

M/F	Number ( % of 182 editors )
Editors making 0 edits in his/her* first three weeks	182(100%)
Source: Antin et al (Table 1: middle column, bottom row, by default)

* As this group produces 0 edits, no 'type of revision' data is generated. Therefore, there is no gender breakdown possible for the report's tables (illustrated in figures 9, 10 and 11 below). Put more simply, if someone one makes 0 number of edits in three weeks, there is nothing to analyze.

Figure 6: Editors making 1 or 2 edits in his/her first three weeks: by gender

M/F	Number ( % of 131 editors )
Male editors making 1 or 2 edits in his first three weeks	66(50.38%)
Female editors making 1 or 2 edits in her first three weeks	65(49.62%)
Source: Antin et al (Table 1: middle column, bottom row)

Top 25% of prolific editors (figure 7)

Note: Antin et al do not explain why the gender split in the top group of editors is not even as promised in the method on page 12, where they stated (emphasis added): "We [...] randomly sampled an equal number of men and women from each quartile to build a total sample of 500 editors." The editors for whom data was unavailable only allows for one more editor (see figure 4b).

Figure 7: Editors making 4+ edits in his/her first three weeks: by gender

M/F	Number ( % of 124 editors )
Male editors making 4+ edits in his first three weeks	67(54.03%)
Female editors making 4+ edits in her first three weeks	57(45.97%)
Source: Antin et al (Table 2: middle column, bottom row)

Coding

The report authors decided that there had to be a consensus of two-thirds of coders on a revision type (e.g. add citation) before the classification was accepted.

Results

Figure 8: Revisions made during the study period (total = 4,549)

  1,700 revisions made to Talk pages, User pages, and other non-article pages (37.37%)

  2,565 coded revisions (56.38%)

  284 revisions where coders were unable to reach a consensus on one or more substantive categories (6.24%)

Bottom 75% of prolific editors by gender

Figure 9: Revision count by gender in the bottom 75% of sample editors

Type of revisions	Number of revisions: female	Number of revisions: male
Add citations	5(56%)	4(44%)
Add new content	26(45%)	32(55%)
Change wiki markup	28(48%)	30(52%)
Create new article	3(43%)	4(57%)
Delete content	4(40%)	6(60%)
Fix typo(s) / grammar	9(56%)	7(44%)
Reorganize existing text	3(50%)	3(50%)
Rephrase existing text	20(51%)	19(49%)
Vandalism	6(60%)	4(40%)
Unsure	3(60%)	2(40%)
Source: Antin et al (Table 1: first column)

Figure 10: Number of editors making particular types of revision in the bottom 75% of sample editors

Type of revisions	Number of editors: female	Number of editors: male
Add citations	5(62%)	3(40%)
Add new content	24(46%)	28(50%)
Change wiki markup	22(51%)	21(50%)
Create new article	3(50%)	3(50%)
Delete content	4(40%)	6(60%)
Fix typo(s) / grammar	8(53%)	7(50%)
Reorganize existing text	3(50%)	3(50%)
Rephrase existing text	16(48%)	17(50%)
Vandalism	5(62%)	3(40%)
Unsure	3(60%)	2(40%)
Source: Antin et al (Table 1: middle column)

Figure 11: Revision size for bottom 75% of sample editors
(using the standard metric Levenshtein distance)

Type of revisions	Size of revision: female	Size of revision: male
Add citations	190.8	1,949.5
Add new content	796.4	710.7
Change wiki markup	341.4	174.8
Create new article	2,509.7	899.2
Delete content	59.5	621.2
Fix typo(s) / grammar	16.7	254.6
Reorganize existing text	788.3	386.7
Rephrase existing text	387.2	105.9
Vandalism	987.8	288.2
Unsure	10	1.5
Source: Antin et al (Table 1: third column)

Top 25% of prolific editors by gender

• The gender split for the number of revisions in this quartile was 27% female to 73% male.

Figure 12: Revision count by gender in the top 25% of sample editors

Type of revisions	Number of revisions: female	Number of revisions: male
Add citations	30(27%)	81(73%)
Add new content	162(34%)	315(66%)
Change wiki markup	231(29%)	562(71%)
Create new article	10(26%)	29(74%)
Delete content	34(37%)	58(63%)
Fix typo(s) / grammar	40(32%)	86(68%)
Reorganize existing text	9(21%)	33(79%)
Rephrase existing text	146(31%)	330(69%)
Vandalism	8(28%)	21(72%)
Unsure	20(33%)	41(67%)
Source: Antin et al (Table 2: first column)

Figure 13: Number of editors making particular types of revision in the top 25% of sample editors

Type of revisions	Number of editors: female	Number of editors: male
Add citations	14(30%)	33(70%)
Add new content	43(46%)	50(50%)
Change wiki markup	48(45%)	59(60%)
Create new article	9(38%)	15(60%)
Delete content	21(47%)	24(50%)
Fix typo(s) / grammar	23(43%)	31(60%)
Reorganize existing text	9(31%)	20(70%)
Rephrase existing text	42(44%)	53(60%)
Vandalism	8(35%)	15(70%)
Unsure	15(36%)	27(60%)
Source: Antin et al (Table 2: middle column)

Figure 14: Revision size for top 25% of sample editors
(using the standard metric Levenshtein distance)

Type of revisions	Size of revision: female	Size of revision: male
Add citations	542.9	230.0
Add new content	583.0	276.8
Change wiki markup	190.1	138.1
Create new article	2,019.8	1,102.5
Delete content	569.1	590.2
Fix typo(s) / grammar	489.4	175.5
Reorganize existing text	1,046.2	761.5
Rephrase existing text	265.7	133.7
Vandalism	281.4	242.2
Unsure	23.8	102.9
Source: Antin et al (Table 2: third column)

Discussion and conclusion

• Male Wikipedia editors drastically outnumber female editors overall, however the report found that 18% of editors female, compared with the 2010 UNU-MERIT study which stated only 13% of editors were female.
• In the bottom three quartiles of Wikipedians, men and women made similar numbers of revisions in nearly every category of Wiki-work (fig. 9).
• In the top 25% of Wikipedians, there is evidence that only 27% of revisions were completed by women (fig. 12).
• In the bottom 75% of Wikipedians, women appeared to make larger revisions in many categories (fig. 11), however, only 9% of revisions were made by editors in the bottom 75%, so this may not be statistically significant.
• In the top 25% of Wikipedians, women tended to make more sizeable revisions, particularly in the "Add New Content" and "Rephrase Existing Text" categories (fig. 14).
• We did not see evidence that men and women are attracted to different types of editing work.
• Two areas of work, where women made significantly larger revisions, involved creative production, synthesis, and reorganization of text, suggesting that, compared to men, women often develop more successful solutions to R&D problems posed through innovation brokers such as Innocentive.
• Despite limitations (for example, sample size, and the inability to make claims about patterns over time) this analysis suggests that the story of Wikipedia's "Gender Gap" is perhaps not as straightforward as initial reports have suggested.
• The results are encouraging in respect that, there was significant gender parity in number of revisions, and women are better represented than the UNU-MERIT study would have led us to believe.
• Either Wikipedia has been successful in attracting more women, or there may be inaccuracies in prior studies.
• The results are discouraging in respect that, there are far fewer women editing Wikipedia than men, and there was a particular gender skew in revision quantity among the Wikipedians who do most of the work.
• Problematically, the most active Wikipedians are also those who largely set policies, arbitrate disputes, and do other high-level tasks into which biases of worldview and temperament can subtly creep. This is the very group among which women may need more representation, but also one which may be more difficult to break into.