“ First, life [reading and editing], then spaces [discussing and congregating], then buildings [institutions] - the other way around never works. ”

— Jan Gehl, (PPS. 2008)^[1]

Wikimedia affiliates are an essential infrastructure for coordinated movement strategy. They act as both institutions that shape community structures; cultures and activities, and as spaces for the community to congregate and prosper. It is therefore strategic to understand and invest in their institutional resilience if we want to successfully meet shared movement goals. Over the last three years, the Global Data & Insights team has monitored indicators to understand the overall health of affiliate institutions and spaces.

Institutional governance

Affiliates identify themselves in two ways, the first being through structures they set up to govern themselves and the second being the people they elect into those structures. A majority of Wikimedia Chapters and Thematic organizations (ThOrgs) have elected boards, which is a requirement for recognition.

Wikimedia community user groups, on the other hand, continue to vary between having no governance structure (32%), using a democratic process (43%), and having an elected board (26%).

There was a significant year-on-year change in governance structures amongst affiliates.^{[note 1]}

Gender representation in leadership structures continues to be imbalanced, as men are a majority on boards (64% of Chapter/ ThOrg boards and 70% of User Group boards are men) as well as being representatives of affiliate groups (66% of Chapter/ ThOrg primary contacts and 74% of User Group primary contacts are men).

There is an encouraging year-on-year change in Chapter/ ThOrg boards, with women now making up 35% (⬆6%) of board trustees; this makes the overall gender make-up of boards representative to that of the membership base and demonstrates some recent advancement towards achieving gender parity in these leadership structures.

The overall gender composition of user group leadership structures has shown no year-on-year change. However, men account for 52% (⬆8%) and women account for 43% (⬇11%) of user group members; this is a significant reversal in the overall gender composition from last year.

User groups showed a slightly better membership depth when compared to Chapters and Thematic organizations, which means they were better at engaging their membership base to participate, support and lead community related activities.

Institutional Culture

Affiliates have created their own rules of engagement amongst their respective membership communities, as well as dealing with conflict within the affiliate landscape. These rules have evolved in cultural norms that guide how members are admitted into these institutions, how their opinions will be canvassed when making a group decision and how conflict internally amongst members or externally with other institutions are resolved. There were no statistically significant year-on-year change in institutional culture indicators and we continue to see:

Two-thirds of affiliates require that members sign up as well as meet a certain threshold of relevant on-wiki activity to join as members and participate in community activities.
Chapters and Thematic Organizations use more formal processes for decision-making, while User Groups employ both formal processes and less formal processes, to make group binding decisions.
Chapters and Thematic Organizations are more prepared than User Groups to intervene on internal conflicts using both formal and informal conflict resolution mechanisms.

Institutional Capacities

The affiliate’s ability to run its affairs, more often than not, is a function of suitable leadership, adequate resources and availability of members who can carry out the affiliate’s mission and vision (Bartov, A., & Houston, S. 2016). Affiliate primary contacts were asked to assess their organizations against eleven institutional capacities and the following was found.

Affiliates were most confident in their ability of executing programs and events (ranked 1st by Chap/ThOrgs and 2nd by WUGs), developing new contributors (ranked 2nd by Chap/ThOrgs and 1st by WUGs) and building partnerships (ranked 3rd by Chap/ThOrgs and 5th by WUGs). They were significantly less confident in all other capacities related to higher-level organizational development.

There was a statistically significant year-on-year upward trend to the Contributor development capacity to 2nd rank overall up from 4th rank (⬆12%) by Chapter/ThOrgs, while this capacity remained at 1st rank (⬆11%) amongst User Groups.^{[note 2]} A new capacity of On-wiki technical skills was introduced to the survey this year and, we found that in general all affiliates were less confident about this capacity, where Chapter/ThOrgs voted this capacity at 8th rank and User Groups voted it at 10th rank out of 11. There were no statistically significant year-on-year changes observed in the mean and distribution of other capacities as ranked by affiliates.

Community Programs

Organized community members have set up their affiliate organizations for the purposes of advancing the Wikimedia movement and its participation in knowledge creation and dissemination (Affiliations committee, n.d.). On average, we found that more (23%) Chapter/ ThOrgs run programs at a monthly cadence than User Groups (8%).

Affiliates found creative ways to continue with their normal community programs during the pandemic and generally managed to hang on to the top four programs they ran between 2018 and 2020.

The most significant slides down the program ranks were on Technical Events (⬇21%) amongst Chapter/ Thematic Organizations and Conference Organizing (⬇13%) amongst User Groups, sliding down 6 places to be ranked last.^{[note 3]}

In general, Affiliates in economically affluent areas tend to run programs at a more frequent cadence.

So what?

To strengthen affiliates as effective institutions for building a diverse and inclusive movement

AffCom should continue to encourage affiliates to adopt gender equity in community governance structures.

The Wikimedia Foundation should prioritize capacity building in communications, contributor development, and community governance amongst affiliates.

Notes

↑ When comparing prevalence of governance structures across affiliate tenure, we found a significant correlation with affiliate tenure (χ²(5) =38.344, p = .000) with significant differences in tenure between groups with no process (mean rank = 85.11, p = .005) or informal discussion process (mean rank = 60.81, p = .012), and/or popular vote (mean rank = 83.07, p = .001) compared to Full democratic processes to the board (mean rank = 120.56), or between informal discussions and board voting (mean rank = 129.49, p = .000). Arranged in descending order of average tenure in years, in 2020 we found:

a. Board voting processes:                 mean = 6.7years;     N = 76 affiliates

b. Full democratic process to board:  mean = 5.6 years;     N = 18 affiliates

c. Partial democratic:                          mean = 4.9 years;     N = 13 affiliates

d. Popular vote by the membership:   mean = 4.15 years;   N = 37 affiliates

e. No decision process reported:        mean = 4.1 years;     N = 32 affiliates

f. Informal discussions as a group:     mean = 2.7 years;     N = 27 affiliates
In 2021 when looking more narrowly at single select response options (forced choice) between no governance structure, a democratic process, and an elected board we again found significant differences in medians (χ²(2) =18.391, p = .000) and distributions (χ²(2) =22.510, p = .000), this difference related primarily to groups having an elected board (mean rank = 73.43) and those without and having either no process (mean rank = 42.00, p = .000) or a democratic process (mean rank = 45.21, p = .002):

g. Elected board process:  mean = 5.9 years;   N = 41 affiliates

h. Democratic process:      mean = 2.6 years;   N = 40 affiliates

i. No process:                     mean = 2.7 years; N = 28 affiliates
↑ When comparing the ranking of Contributor Development as a capacity amongst affiliates between 2020 and 2021, an Independent-Samples Mann-Whitney U test found a significant change in the distribution (χ²(X) = 4974.50, p = 0.018), while the difference in medians did not reach significance (p = 0.057).
↑ When comparing an overall prevalence of program types amongst affiliates between 2020 and 2021, an Independent-Samples median and Mann-Whitney U tests were used to identify significant differences for the following program areas:

a. Presence of Conference Attendance and Presenting in distribution (χ²(1) = 2563.0, p = 0.000) while a change in medians could not be calculated.

b. Presence of Wikipedia Library

medians (χ²(1) = 5718.5, p = 0.000) and in distribution (χ²(1) = 12.382, p = 0.000). Demonstrating a higher mean rank in 2021 (107.46), compared to 2020 (mean rank = 82.10)

c. Presence of Other Partnerships

medians (χ²(1) = 2893.00, p = 0.000) and distribution (χ²(1) = 23.645, p = 0.000). Demonstrating a lower mean rank in 2021 (81.54), compared to 2020 (mean rank = 116.14)

d. Presence of Policy and Advocacy

medians (χ²(1) = 3721.00, p = 0.008) as well as in distribution (χ²(1) = 6.28, p = 0.012). Demonstrating a lower mean rank in 2021 (89.14), compared to 2020 (mean rank = 106.17)

e. Presence of Meet-ups in distribution(χ²(1) = 3875.50, p = 0.024) while a change in medians could not be calculated. Demonstrating a lower mean rank in 2021 (90.56), compared to 2020 (mean rank = 104.31)

References

↑ PPS. (2008, December 31). Jan Gehl. Project for Public Spaces. https://www.pps.org/article/jgehl

Statistical References:

Dunn, O. J. (1964). Multiple comparisons using rank sums. Technometrics, 6, 241-252.
Kruskal, W. H., & Wallis, W. A. (1952). Use of ranks in one-criterion variance analysis. Journal of the American Statistical Association, 47(260), 583-621.
Laerd Statistics (2015). Statistical tutorials and software guides. Retrieved from https://statistics.laerd.com
Lehmann, E. L. (2006). Nonparametrics: Statistical methods based on ranks. New York: Springer.
Mann, H. B., & Whitney, D. R. (1947). On a test of whether one of two-random variables is stochastically larger than the other. The Annals of Mathematical Statistics, 18(1), 50-60.
Sheskin, D. J. (2011). Handbook of parametric and nonparametric statistical procedures (5th ed.). Boca Raton, FL: Chapman & Hall/CRC Press.

[2] When comparing prevalence of governance structures across affiliate tenure, we found a significant correlation with affiliate tenure (χ²(5) =38.344, p = .000) with significant differences in tenure between groups with no process (mean rank = 85.11, p = .005) or informal discussion process (mean rank = 60.81, p = .012), and/or popular vote (mean rank = 83.07, p = .001) compared to Full democratic processes to the board (mean rank = 120.56), or between informal discussions and board voting (mean rank = 129.49, p = .000). Arranged in descending order of average tenure in years, in 2020 we found:

a. Board voting processes: mean = 6.7years; N = 76 affiliates

b. Full democratic process to board: mean = 5.6 years; N = 18 affiliates

c. Partial democratic: mean = 4.9 years; N = 13 affiliates

d. Popular vote by the membership: mean = 4.15 years; N = 37 affiliates

e. No decision process reported: mean = 4.1 years; N = 32 affiliates

f. Informal discussions as a group: mean = 2.7 years; N = 27 affiliates
In 2021 when looking more narrowly at single select response options (forced choice) between no governance structure, a democratic process, and an elected board we again found significant differences in medians (χ²(2) =18.391, p = .000) and distributions (χ²(2) =22.510, p = .000), this difference related primarily to groups having an elected board (mean rank = 73.43) and those without and having either no process (mean rank = 42.00, p = .000) or a democratic process (mean rank = 45.21, p = .002):

g. Elected board process: mean = 5.9 years; N = 41 affiliates

h. Democratic process: mean = 2.6 years; N = 40 affiliates

i. No process: mean = 2.7 years; N = 28 affiliates

[3] When comparing the ranking of Contributor Development as a capacity amongst affiliates between 2020 and 2021, an Independent-Samples Mann-Whitney U test found a significant change in the distribution (χ²(X) = 4974.50, p = 0.018), while the difference in medians did not reach significance (p = 0.057).

[4] When comparing an overall prevalence of program types amongst affiliates between 2020 and 2021, an Independent-Samples median and Mann-Whitney U tests were used to identify significant differences for the following program areas:

a. Presence of Conference Attendance and Presenting in distribution (χ²(1) = 2563.0, p = 0.000) while a change in medians could not be calculated.

b. Presence of Wikipedia Library

medians (χ²(1) = 5718.5, p = 0.000) and in distribution (χ²(1) = 12.382, p = 0.000). Demonstrating a higher mean rank in 2021 (107.46), compared to 2020 (mean rank = 82.10)

c. Presence of Other Partnerships

medians (χ²(1) = 2893.00, p = 0.000) and distribution (χ²(1) = 23.645, p = 0.000). Demonstrating a lower mean rank in 2021 (81.54), compared to 2020 (mean rank = 116.14)

d. Presence of Policy and Advocacy

medians (χ²(1) = 3721.00, p = 0.008) as well as in distribution (χ²(1) = 6.28, p = 0.012). Demonstrating a lower mean rank in 2021 (89.14), compared to 2020 (mean rank = 106.17)

e. Presence of Meet-ups in distribution(χ²(1) = 3875.50, p = 0.024) while a change in medians could not be calculated. Demonstrating a lower mean rank in 2021 (90.56), compared to 2020 (mean rank = 104.31)

[1] PPS. (2008, December 31). Jan Gehl. Project for Public Spaces. https://www.pps.org/article/jgehl

[1]

[note 1]

[note 2]

[note 3]

Global Data and Insights/Reports/2021 Affiliates data survey report/Institutions