Analysis of this experiment suggests that these
interventions increased turnout by approximately
340,000 additional votes. This is approximately 0.5
percent of the total number of votes cast. The Facebook experiment wasn’t designed to change the
outcome. It was simply designed to increase voter
participation. In particular, there was no bias in the
way users were encouraged to vote. Users for the
three difference groups were chosen uniformly at
random. However, it is very possible that the experiment changed the outcome of some of the elections
held that day. We cannot know for sure, as we cannot rerun the election without the intervention. But
the evidence points to an impact.
Consider the 2010 election results for the Windsor–
Orange First District for the Vermont House of Representatives. The 2010 election in this district was
decided by a single vote (Vermont Public Radio
2011). Similarly, the outcome of the 2010 election
in the Rutland 5-4 District for the Vermont House
of Representatives was also decided by a single vote
(Dritschilo 2010). Both elections were won by a female
Democrat running against a male Republican candidate. With such close outcomes, the Facebook experiment could have been critical.
Suppose, for a moment, that Facebook had a younger
and more female demographic in Vermont in 2010
than the voting population of Vermont itself. This
is not an unreasonable assumption. Facebook appeals
best to adult women aged 18–29 (Duggan et al. 2014).
Now suppose younger women in Vermont are more
likely to vote for a female Democratic candidate than
for a male Republican. Again, that is not an unrea-
It follows from these two assumptions that increasing voter participation of Facebook users in Vermont would likely have increased the Democratic
vote. This might easily have got one or two extra
votes for the Democrats. Given the closeness of the
results in these elections, this could have changed
the outcome from what would have been a Republican victory without Facebook’s experiment, to the
Democratic win that was actually witnessed.
It should not have been a surprise that this could
have happened. Thousands of different elections
were held on November 2, 2010, some of which were
very likely to be closely contested. Indeed, Vermont
was one of the more obvious places to see a close result.
The Vermont House of Representatives has relatively
small electorates, making it easier to have a narrow
result, and has historically seen closely fought outcomes. In 1977, 1986, and 2016 there were other districts for the Vermont House of Representatives that
were also decided by a single vote.
Facebook ran further experiments to increase voter
participation in the 2012 US elections. Less is known