Community Safety/Reports/Wave 2

Community Safety Survey Wave 2 Results

In June and July of 2022, we deployed the 2nd Community Safety survey on 5 language Wikipedias: English, Farsi, French, Portuguese, and Spanish.

The question asked of users was:

“In the last 30 days, have you felt unsafe or uncomfortable contributing to Wikipedia (xx.wikipedia.org)?”

with xx denoting the Wikipedia language (for example, for French it was “fr.wikipedia.org”).

The options to respond were: “Yes,” “No,” and “I'm not sure.”

A few notes about reading this data

Percents reported unweighted reflect the actual responses to the survey.
Percents reported weighted reflect the responses to the survey with weights applied to shape the data to the distribution of editors on the wiki by “edit bucket” (a user having made 5 to 99 edits, 100 to 999 edits, or 1000 or more edits).
- We report weighted data when we see a statistical difference in how different edit bucket users respond to the question. This helps to make the data more reflective of the actual population of the wiki.
If you want to apply this data to the broader population of the wiki, make sure to look for the margin of error! For example, note the difference in meaning between:
- “77% of survey respondents said that they did not feel unsafe or uncomfortable contributing to English Wikipedia in the month before the survey”
- “We estimate that 76% to 79% of English Wikipedians felt safe contributing to English Wikipedia in the month before the survey”
  - The first example tells us about the results of the actual survey, while the second example approximates how the overall population feels (those who were surveyed, and those who were not) based on the survey results, a 95% confidence interval, and margin of error.

With that, here are the results of our Wave 2 analyses for each wiki:

English Wikipedia

For the English Wikipedia Wave 2 data, a chi-squared test found no differences in how users from different “edit buckets” (5 to 99 edits, 100 to 999 edits, and 1000 or more edits) responded to the survey. We thus report unweighted data.

Based on the responses to the survey, we estimate that users on English Wikipedia overall:

14 to 18% felt unsafe or uncomfortable contributing to English Wikipedia in the 30 days before the survey.
74 to 77% did not feel unsafe or uncomfortable contributing to English Wikipedia in the 30 days before the survey.
7 to 11% were unsure about whether they had felt unsafe or uncomfortable contributing to English Wikipedia in the 30 days before the survey.

In the last 30 days, have you felt unsafe or uncomfortable contributing to Wikipedia (en.wikipedia.org)?	Survey Results for En.WP (Unweighted)	Estimates for En.WP population (Margin of Error = ±1.74%, Confidence Interval = 95%)
No	75.3%	73.6 to 77.0%
I'm not sure	8.8%	7.1 to 10.5%
Yes	15.9%	14.2 to 17.6%

Spanish Wikipedia

For the Spanish Wikipedia Wave 2 data, a chi-squared test found differences in how users from different edit buckets (5 to 99 edits, 100 to 999 edits, or 1000 or more edits) responded to the survey^[1]. We thus report weighted data.

Based on the responses to the survey, we estimate that users on Spanish Wikipedia overall:

21 to 28% felt unsafe or uncomfortable contributing to Spanish Wikipedia in the 30 days before the survey.
63 to 69% did not feel unsafe or uncomfortable contributing to Spanish Wikipedia in the 30 days before the survey.
6 to 13% were unsure about whether they had felt unsafe or uncomfortable contributing to Spanish Wikipedia in the 30 days before the survey.

In the last 30 days, have you felt unsafe or uncomfortable contributing to Wikipedia (es.wikipedia.org)?
	Survey Results for Es.WP (Unweighted)	Survey Results for Es.WP (Weighted)	Estimates for Es.WP population (Margin of Error = ±3.32%, Confidence Interval = 95%)
No	65.8%	66.0%	62.7 to 69.3%
I'm not sure	9.6%	9.7%	6.4 to 13.0%
Yes	24.6%	24.4%	21.1 to 27.7%

Persian Wikipedia

For the Farsi Wikipedia Wave 2 data, a chi-squared test^[2] found differences in how users from different “edit buckets” (5 to 99 edits, 100 to 999 edits, and 1000 or more edits) responded to the survey. We thus report data weighted to the edit buckets of the overall user population in 2021.

Based on the responses to the survey, we estimate that users on Farsi Wikipedia overall:

21 to 27% felt unsafe or uncomfortable contributing to Farsi Wikipedia in the 30 days before the survey.
56 to 62% did not feel unsafe or uncomfortable contributing to Farsi Wikipedia in the 30 days before the survey.
14 to 20% were unsure about whether they had felt unsafe or uncomfortable contributing to Farsi Wikipedia in the 30 days before the survey.

In the last 30 days, have you felt unsafe or uncomfortable contributing to Wikipedia (fa.wikipedia.org)?	Survey Results for Fa.WP (Unweighted)	Survey Results for Fa.WP (Weighted)	Estimates for Fa.WP population (Margin of Error = ±2.99%, Confidence Interval = 95%)
No	58.6%	58.8%	55.8 to 61.8%
I'm not sure	17%	17.1%	14.1 to 20.1%
Yes	24.4%	24.1%	21.1 to 27.1%

French Wikipedia

For the French Wikipedia Wave 2 data, a chi-squared test found no differences in how users from different edit buckets (5 to 99 edits, 100 to 999 edits, or 1000 or more edits) responded to the survey. We thus report unweighted data.

Based on the responses to the survey, we estimate that users on French Wikipedia overall:

8 to 15% felt unsafe or uncomfortable contributing to French Wikipedia in the 30 days before the survey.
81 to 88% did not feel unsafe or uncomfortable contributing to French Wikipedia in the 30 days before the survey.
1 to 8% were unsure about whether they had felt unsafe or uncomfortable contributing to French Wikipedia in the 30 days before the survey.

In the last 30 days, have you felt unsafe or uncomfortable contributing to Wikipedia (fr.wikipedia.org)?
	Survey Results for Fr.WP (Unweighted)	Estimates for Fr.WP population (Margin of Error = ±3.16%, Confidence Interval = 95%)
No	84.4%	81.1 to 87.5%
I'm not sure	4.4%	1.2 to 7.6%
Yes	11.4%	8.2 to 14.0%

Portuguese Wikipedia

For the Portuguese Wikipedia Wave 2 data, a chi-squared test found no differences in how users from different “edit buckets” (5 to 99 edits, 100 to 999 edits, and 1000 or more edits) responded to the survey. We thus report unweighted data.

Based on the responses to the survey, we estimate that users on Portuguese Wikipedia overall:

23 to 29% felt unsafe or uncomfortable contributing to Portuguese Wikipedia in the 30 days before the survey.
62 to 69% did not feel unsafe or uncomfortable contributing to Portuguese Wikipedia in the 30 days before the survey.
5 to 12% were unsure about whether they had felt unsafe or uncomfortable contributing to Portuguese Wikipedia in the 30 days before the survey.

In the last 30 days, have you felt unsafe or uncomfortable contributing to Wikipedia (pt.wikipedia.org)?
	Survey Results for Pt.WP (Unweighted)	Estimates for Pt.WP population (Margin of Error = ±3.43%, Confidence Interval = 95%)
No	65.4%	62.0 to 68.8%
I'm not sure	8.6%	5.2 to 12.0%
Yes	26.0%	22.6 to 29.4%

Additional Notes

Wave 1 Data
Wiki	Response Rate^[3]	Number of Responses	% of Users Displayed the Question
English Wikipedia	~24%	3,180	3%
Spanish Wikipedia	~18%	869	10%
Persian Wikipedia	~26%	1,074	20%
French Wikipedia	~15%	960	10%
Portuguese Wikipedia	~17%	816	20%

Past Results

Please note, if you are comparing current results to past results, to pay attention to the margins of error. Overlapping margins of error between waves mean that a change seen may not be statistically significant.

Wave 1 Results

↑ X² (4, N = 869) = 11.887, p = .018.
↑ X² (4, N = 1074) = 26.031, p = <.001.
↑ We estimate response rates by calculating the number of unique responses after data cleaning divided by the number of unique survey session tokens from impressions. This gives a slightly smaller response rate than the actual, as users who clear their browser cookies or switch devices may have the survey shown to them again and WMF staff also force the survey to be shown in order to test for quality assurance.

[1] X² (4, N = 869) = 11.887, p = .018.

[2] X² (4, N = 1074) = 26.031, p = <.001.

[3] We estimate response rates by calculating the number of unique responses after data cleaning divided by the number of unique survey session tokens from impressions. This gives a slightly smaller response rate than the actual, as users who clear their browser cookies or switch devices may have the survey shown to them again and WMF staff also force the survey to be shown in order to test for quality assurance.

[1]

[2]

[3]