Wikipedia:Wikipedia Signpost/2019-11-29/Special report

Special report

How many people edit in your favorite language? Where are they from?

Let's say you are interested in how many active editors from France are editing the English-language Wikipedia; or conversely, you'd like to know how many editors from the UK are editing the French-language Wikipedia. All the necessary information needed to calculate these numbers is recorded, at least temporarily, by the Wikimedia Foundation, but unless you worked for the WMF and had access to the Geoeditors Monthly database you could never find those numbers. The WMF did not wish to disclose this data out of concerns that the numbers were precise enough that governments or others could back out material that might lead to the identification of individual editors.

This month a new dataset was made public by the Wikimedia Foundation Geoeditors/Public, or more informally Active Editors by country. It allows the public to see, more or less, how many active editors (5–99 edits in a month) and very active editors (100+ edits) from about 180 individual countries contribute to active Wikipedia versions, each month from January 2019 onward. For example, if you wanted to know how many people editing from the UK made more than 99 edits to the French version of Wikipedia in September, you can look it up in this dataset. The answer is somewhere between 11 and 20.

Because of privacy concerns exact numbers are not given. Data from 30 countries are excluded, e.g. China, Kazakhstan, Russia, Saudi Arabia and Venezuela. Exact data on the number of editors in each category (editors from country x who edited Wikipedia version y) are not given. Rather these numbers are only given in “buckets” of ten: 1–10, 11–20, 21–30, 31–40, etc. Technical information is available here. The data are available here.

But enough for the preliminaries! What questions can the dataset answer that I’ve been dying to know the answer to? The following analysis is only the briefest overview of data from one month, September, quickly done. It’s not in any sense academic research, but hopefully will allow people to understand what type of data the dataset contains and what type of questions it can be used to address.

My main questions – of personal interest – are:

What countries contribute most to the English-language Wikipedia (enwiki)? Are they the richer, or the more populous English-speaking countries? Or perhaps those countries where English is widely spoken as a second language?
Do these relations differ across different Wikipedia language versions? Answering the above questions for the Spanish-language Wikipedia (eswiki) allows a simple comparison.
And finally, how do contributions across countries to different language versions compare. Edits from the US and UK are examined here.

Who edits enwiki?

Table 1 shows the 11 countries with the most active editors and the 11 with the most very active editors to enwiki (14 countries total), plus two other large English-speaking countries, Ireland and South Africa. Numbers marked * are not in the largest 11.

Editors from	Editors with 100+ edits (lower bound)	% of total reported	Editors with 5–99 edits (lower bound)	% of total reported
United States	1,881	42.9%	25,401	41.0%
United Kingdom	731	16.7%	7,491	12.1%
Canada	271	6.2%	3,321	5.4%
Australia	231	5.3%	2,491	4.0%
India	191	4.4%	5,241	8.5%
Germany	121	2.8%	1,281	2.1%
Philippines	81	1.8%	1,021	1.6%
Netherlands	61	1.4%	621*	1.0%
Italy	51	1.2%	831	1.3%
New Zealand	51	1.2%	441*	0.7%
Sweden	51	1.2%	431*	0.7%
France	41*	0.9%	791	1.3%
Ireland	41*	0.9%	661*	1.1%
Spain	41*	0.9%	681	1.3%
Brazil	31*	0.7%	721	1.2%
South Africa	21*	0.5%	291*	0.5%
Total (in table)		88.8%		83.6%

The countries with the most very active editors in enwiki are the US (43%) and the UK (17%) , or almost 60% of the total reported editors between them. The two large rich countries predominate. Two rich but less populous countries, Canada and Australia, are also well-represented with almost 12% of the total very active editors between them.

The much smaller but still relatively rich New Zealand and Ireland, with about 1% of the total reported very active editors each, trail among those countries where English is the predominant first language.

The proportion of native English speakers by country is shown at English language#Pluricentric English. The four countries with the largest native English-speaking populations are also the largest four contributors to enwiki – in the same order: USA, UK, Canada, and Australia.

India, which has the 5th largest group of very active editors (4%) and third largest group of active editors (9%), has a very large population, for whom English is an important medium of instruction but the first language of only a small fraction. The Philippines, with nearly 2% of the reported very active editors, may be affected by similar factors as India. The percentages of reported active editors (5–99 edits) appear to be similar to the percentages for very active editors.

Six rich European Union countries where English is not the mother tongue, Germany, the Netherlands, Italy, Sweden, France and Spain, together account for 8.4% of the reported very active editors. Of the countries in this table, only the rankings of Brazil and perhaps South Africa do not appear to be directly explained by the three factors of mother tongue, population, and wealth.

Who edits eswiki?

Table 2 shows analogous rankings for the Spanish language Wikipedia. While Spain and Argentina combine for slightly over half of the reported very active editors, the very active editors are distributed more evenly over all the reported countries. Only one country without Spanish as its predominant language, the United States, has a fairly large proportion of the very active editors. The same three factors that seem to explain the rankings for enwiki editors, mother tongue, population, and wealth, may very well explain the rankings for eswiki as well.

Nevertheless, wealth – or perhaps dialect – may be playing a stronger role in eswiki than it does in enwiki. The 12 largest countries by native Spanish-speaking population are, in order, Mexico, Colombia, Spain, Argentina, the United States, Venezuela, Peru, Chile, Ecuador, Cuba, Guatemala, and the Dominican Republic. Note that Venezuela and Cuba are excluded by the WMF from the dataset. The population rankings for native English-speaking countries are almost identical to the rankings in Wikipedia contributions of the same countries. But the population rankings for native Spanish-speaking countries are much less similar to their rankings in Wikipedia Spanish-language contributions.

Editors from	Editors with 100+ edits (lower bound)	% of total reported	Editors with 5–99 edits (lower bound)	% of total reported
Spain	211	35.9%	3,881	35.4%
Argentina	101	17.2%	1,421	12.0%
Mexico	71	12.1%	1,471	13.4%
Chile	51	8.7%	831	7.6%
Colombia	41	7.0%	851	7.8%
Peru	31	5.3%	631	5.8%
Ecuador	11	1.9%	231	2.1%
Nicaragua	11	1.9%	81*	0.7%
United States	11	1.9%	251	2.3%
Uruguay	11	1.9%	211	1.3%
unknown	11	1.9%	11*	0.1%
Bolivia	1*	0.2%	101	0.9%
Dominican Republic	1*	0.2%	101	0.9%
Total (in table)		96.0%		91.8%

US and UK editors editing on non-English Wikipedias

Table 3 shows how very active editors from the US and the UK edit the non-English Wikipedias. Altogether very active editors from the US edit in 44 different Wikipedia versions. Those from the UK edit in 29 versions. Among those versions with 11–20 very active editors from the US are an interesting mix of the Chinese, Spanish, Farsi (Persian), Japanese, and Russian Wikipedias. The similar data from UK editors only includes the French Wikipedia.

Version edited	From	Editors with 100+ edits (lower bound)
enwiki	United States	1881
zhwiki	United States	51
eswiki	United States	11
fawiki	United States	11
jawiki	United States	11
ruwiki	United States	11
simplewiki	United States	11
37 others	United States	37
enwiki	United Kingdom	731
frwiki	United Kingdom	11
27 others	United Kingdom	27

So what else can you do with this dataset?

Time is the main variable of interest that was left out of the above examinations. Right now we could see how edit contributions from different countries change over the nine months from January through September 2019. As time goes by, more months of data will be released, and the effect of time will likely be of greater interest. For example, let's say that there was a new program introduced intended to increase the number of editors from country Y. The full effects of the program might not be seen after 9 months, but after 2 or 3 years hopefully any effects could be seen in the data.

Another area of interest might involve combining this dataset with other datasets. For example, say a program is undertaken to increase the quality – rather than the quantity – of articles about country Z. Using this data in conjunction with data on readership might give a more complete understanding of the effects of the program.

← Previous "Special report"

Next "Special report" →

In this issue

Special report

Discuss this story

These comments are automatically transcluded from this article's talk page. To follow comments, add the page to your watchlist. If your comment has not appeared here, you can try purging the cache.

Well I am a Wikipedia editor from Sri Lanka and of course English is not a popular language though it is one of the official languages of the country. It is regarded as a link language in the country. Surprised to Portugese speaking Brazil in the top 15 list even ahead of South Africa for gaining popularity in English Wikipedia. Abishe (talk) 03:36, 30 November 2019 (UTC)[reply]

Fascinating. I'm surprised Nigeria did not make the list. -Indy beetle (talk) 08:04, 30 November 2019 (UTC)[reply]

This is all there is from the September file:

language . country quant. lower limit upper limit

arwiki Nigeria 5 to 99 1 10
enwiki Nigeria 100 or more 11 20
enwiki Nigeria 5 to 99 251 260
hawiki Nigeria 5 to 99 1 10
jawiki Nigeria 5 to 99 1 10
yowiki Nigeria 5 to 99 1 10

yowiki is prob. yorba (sp?) . Smallbones_(smalltalk) 14:12, 30 November 2019 (UTC)[reply]

See Yoruba language. MPS1992 (talk) 15:13, 30 November 2019 (UTC)[reply]

These are interesting figures, but they need to be viewed with caution, for a number of reasons, including the following. For a start, the records are of the assumed location, and not necessarily the nationality, of the contributor. So, eg, if a contributor is a foreign student in the USA, the UK or Australia, all of which have big foreign student populations, the records will not be a record of the nationality of the contributor. Secondly, the assumed location may not be correct. So, eg, if a contributor in the PRC is using a VPN that says that the contributor's location is the USA, then the record will show the USA, not the PRC, as the location, even though the true location is the PRC. Thirdly, the records are likely to be skewed not simply towards native speakers (as the article hypothesises), but also towards fluent non-native speakers. That would explain, eg, the high performance of the Netherlands and Germany (the latter of which has more fluent speakers of English than Australia), and possibly also Brazil, a country with a large population (about 2 1/2 times that of Germany) and mandatory learning of at least one foreign language for all 12 grades of compulsory schooling. Fourthly, one needs to be cautious about Canada, where only about 56% of the population speaks English as a mother tongue, and about 21% use French as their mother tongue. Fifthly, the figures indicate that some countries with smaller populations have disproportionately large numbers of contributors. So, eg, New Zealand and Ireland both have populations of about one fifth of that of Australia, but New Zealand has more than one fifth as many prolific contributors, and Ireland has more than one fifth as many small contributors. Similarly, the number of contributors from the UK (both prolific and small) is disproportionately large by comparison with the USA. Bahnfrend (talk) 09:46, 2 December 2019 (UTC)[reply]
- @Bahnfrend: Thanks for this - all datasets have limits or quirks of course, and it's important for everybody to understand the limits. Also you're starting to get into some new hypotheses about the data (e.g. from foreign students, fluent non-native speakers, or VPNs), which is the start of really understanding the data. This dataset has some special quirks, designed right-in to hide identities, and even the most basic measure - what is an "edit"? - is pretty vague covering everything from changing a comma to a semi-colon, to adding in 1,000 words to an article. Beyond simple curiosity, I suppose my motivation has to do mostly with so-called "political bias". A lot of Americans seem to think WP has a liberal bias, but is that due to age, gender, or country of residence of editors? I do think that this dataset will be examined in detail, so getting out all the quirks, biases, and hypotheses now is a worthwhile exercise. Thanks. Smallbones_(smalltalk) 17:19, 2 December 2019 (UTC)[reply]

What do you think of The Signpost? Share your feedback.

Home

About