Thoughts on the reduced turnout in the 2022 WMF board elections

Summary

This year’s board elections ended with lower global voter participation than in 2021.

By comparing available data of the elections in 2021 and 2022, some possible explanations for this are explored.

It seems as if 2022’s smaller group of candidates, with reduced support from super big communities, had a significant impact on the global voter turnout of the board elections. Their absence is possibly responsible for a significant part of the decline.

A significant second factor was last year’s unique additional outreach in the Japanese, Catalan, and Czech communities, which did not repeat this year. 

Question

This year’s election ended with lower global voter participation than in 2021, though it still was the second-highest number of voters ever. 5957 voters cast their vote in 2022, compared with 6873 voters in 2021, a total loss of 916 votes. This is the first significant reduction since 2013. As the global electorate did not change significantly (2022: 67574, 2021: 67838), it brings up the question why that is.

Diagnosing differences

If we take a look at the election statistics of 2022 and 2021, it is striking that we do not see communities that mobilized heavily above the global standard turnout of ~10% (“champion communities”) in 2022. From former voter statistics, we learn that there were always “champion communities” having voter turnouts way above standard turnouts.

In 2021 we saw the following bigger communities having turnouts of 15% and more:

Wiki	Eligible	Voters	Turnout
ca.wikipedia.org	344	125	36.337%
uk.wikipedia.org	682	153	22.434%
pl.wikipedia.org	1251	280	22.382%
ar.wikipedia.org	464	95	20.474%
cs.wikipedia.org	558	108	19.355%
tr.wikipedia.org	441	73	16.553%
it.wikipedia.org	2113	324	15.334%

2022 is pretty different from that. There were barely any champion communities, and those there are had lower turnouts: 

Wiki	Eligible	Voters	Turnout
ar.wikipedia.org	446	91	20.404%
pl.wikipedia.org	1175	181	15.404%

This is confirmed by checking medians. If we select the Top 30 communities by eligible voters (comprising 84% of the electorate), sort them by their turnouts, and calculate the average of the groups with the highest, medium, and lowest turnouts this is what we can find:

Group	2022	2021	Diff
Highest turnout median	12.60%	19.30%	-6.70%
Medium turnout median	8.20%	7.95%	0.25%
Lowest turnout median	5.40%	5.87%	-0.47%

It is clear that neither the groups with the lowest nor the medium average turnout shows significant changes but the group with the highest turnout does with a loss of 34.7%.

How significant is this?

As shown above, the effect is real, but does it actually affect the change significantly?

To find out, we took the above TOP 30 communities (again: they make up 84% of the electorate) and sorted them by the change in their number of voters in 2021. 

Twelve of the 30 communities experienced losses bigger than twenty voters. For these 12 communities we can see the following:

Wiki	Voters 2022	Turnout 2022	Voters 2021	Turnout 2021	Voter Diff
es.wikipedia.org	237	8.40%	349	12.57%	-112
it.wikipedia.org	215	10.32%	324	15.33%	-109
ru.wikipedia.org	199	6.75%	306	10.39%	-107
pl.wikipedia.org	181	15.40%	280	22.38%	-99
uk.wikipedia.org	64	10.16%	153	22.43%	-89
ca.wikipedia.org	45	13.64%	125	36.34%	-80
ja.wikipedia.org	98	2.71%	175	5.25%	-77
cs.wikipedia.org	56	10.09%	108	19.36%	-52
de.wikipedia.org	604	10.93%	645	11.32%	-41
fr.wikipedia.org	466	11.91%	502	12.81%	-36
pt.wikipedia.org	80	7.03%	115	10.24%	-35
fa.wikipedia.org	52	7.35%	85	11.79%	-33

As mentioned, there was an increase in voter numbers in 6 communities. In three of them only it is significant, these are: 

Wiki	Voters 2022	Turnout 2022	Voters 2021	Turnout 2021	Voter Diff
he.wikipedia.org	94	11.55%	62	7.62%	32
zh.wikipedia.org	153	6.44%	109	4.88%	44
en.wikipedia.org	1963	8.24%	1845	7.95%	118

Remember: the total decrease from 2021 was 917 votes. The above communities’ losses alone make up 870 voters lost and only 194 votes gained in reverse.

Reasons

Most intriguing is the question of the reason for these losses. Just to make it clear: while up to this point all insights are based on data, this is going to be, to some extent, educated guessing.

Last year five of the eight champion communities had a candidate running: ES, IT, PT, RU, FR. All these communities are part of the TOP 10 communities by eligible voters (not counting EN). Their voter decrease makes up for 399 voters lost. 

Usually, in the following elections, the new candidates’ communities compensate for losses in former champion communities. 

With only 6 candidates (2021: 19) that effect was significantly smaller though. With two of them originating from EN, and only one of them from the above-mentioned TOP 10 (Polish), the new candidates communities couldn’t make up for these losses. The gains from the Hebrew and the English communities sum up to 150 voters only.

This insight is strengthened by looking at the results from Indian communities that had four candidates running in 2021 and none this year. Turnouts dropped in 2022 in the Indian language communities, respectively. Due to the number of Indian language versions we took a pass on their quantification, spot checks seem to confirm this effect though.

Other reasons for voter losses were more specific to the respective communities: 

• Outreach: Last year’s hype in the Catalan community, run by a volunteer, did not repeat. (-80 votes)
• Outreach: The Japanese community had some (very basic) outreach by a community member last year, which was missing this year. (-77 votes)
• Outreach: In the Czech community, the chapter was not promoting and supporting the elections as last year (-52 votes)
• Politics: The Ukrainian community lost 89 voters, obviously a consequence of the ongoing war.
• Electorate: The German community had a smaller electorate (-175), and the almost same turnout (2022: 10.9% 2021: 11.3%) led to -41 voters.

The increase of the Mandarin Wikipedia’s turnout is probably a result of localized outreach by the Movement Strategy and Governance team not present in 2021. This led to an addition of 44 voters. 

There are yet no valid explanations available for the losses in the Farsi (-33) and Polish (-99) communities.

Outreach factors, responsible for the losses of 3 of 12 and the gain in 1 of 3 communities, sum up to 209 losses and 44 gains. Again, the gains do not make up for the losses, leading to 165 voters less than last year. Last year’s outreach in the respective communities was driven by local volunteers and affiliates, who did not repeat their engagement this year. 

Possibly last year’s unusual volunteer and affiliate engagement was driven by the euphoria of finally having an election again after four years and slowed down by its repetition just a year later (not to mention other votes like the MCDC and the UCoC Enforcement Guidelines in between). WMF staff outreach was not available to compensate for these losses.

Conclusions

The Wikimedia movement over the years has gotten used to a continuously growing number of voters in board elections. The above data indicates, that various factors cause participation in elections and are influential to the global turnout. Some powerful factors seemingly are the composition of the candidates’ group and localized outreach to communities.

The decreased impact of candidates hailing from big (and due to their size mostly western) communities, might be perceived as less of a loss, but a correction of former imbalances though. A bigger number of candidates though, resulting in a resurrection of the champion community phenomena, will help to raise local and to some extent global turnouts again.

A strong point to learn from the above numbers is the significance of community outreach. Regardless of it being executed by dedicated volunteers, affiliates, or Wikimedia Foundation staff, it seemingly helps to raise general awareness of elections within communities formerly underrepresented and thus equalizing voter turnouts globally.

(All data from the WMF Board Elections statistics)