Impacts of Nonpartisan Political Information on Electoral Accountability in Mexico

Researchers:
Jose Ramon Enriquez
Sample:
2.7 million Facebook users
Timeline:
2018 - 2018
Target group:
  • Voters
  • Adults
Outcome of interest:
  • Electoral participation
  • Transparency and accountability
  • Voter Behavior
Intervention type:
  • Social networks
AEA RCT registration number:
AEARCTR-0003135
Partners:
  • Borde Politico

Ensuring elected officials are held accountable for financial wrongdoing is a persistent challenge in low- and middle-income countries. Leveraging the increasingly widespread use of social media, researchers conducted a large-scale randomized evaluation to test the direct and indirect impact on electoral accountability of a nonpartisan information campaign delivered via Facebook ads during the 2018 Mexican municipal elections. Incumbent parties with negligible corruption levels saw their vote share increase by 6-7 percentage points in localities targeted by the Facebook ads. Researchers show this increase was driven by direct and indirect effects in municipalities where a high proportion of residents were targeted by Facebook ads. This appears to reflect both online and offline interactions within those communities.

الموضوع الأساسي

Internet and social media use has skyrocketed in low-income countries, growing by more than 50 percent between 2013 and 2018. The reach of misinformation—the deliberate or unintentional creation or spread of false or inaccurate information—has been growing in tandem with this trend. Additionally, government propaganda is often circulated online. As such, citizens may be influenced by false or politically motivated information they encounter online. Even if they are aware of this phenomenon, they may be unsure how to tell which sources of information are reliable and which are not. Furthermore, the spread of misinformation or propaganda can be particularly pronounced during election seasons, when there are heightened political consequences. 

Meanwhile, the misuse of public expenditures continues to be an obstacle in low- and middle-income countries. Though citizens may be generally aware of this kind of corruption, it can be challenging to know which politicians are misusing funds—and to what extent. Furthermore, even when this information is available, voters may not be aware of it or have easy access. It can also be difficult for traditional media outlets to meaningfully highlight corruption, as they are often under-resourced and sometimes vulnerable to government influence. As such, politicians frequently do not face consequences for financial misuse at the ballot box. 

In light of these challenges, nonpartisan groups could promote electoral accountability by providing reliable information to citizens. Leveraging low-cost digital communication technologies, they can target information toward many citizens in specific localities. However, there is limited evidence on the impact of targeted nonpartisan social media campaigns on individuals and communities, especially where many residents receive similar information. Can providing nonpartisan information to citizens on the use of public expenditures by the current administration increase the likelihood that voters support the best-performing politicians and sanction the worst-performing, particularly when it is delivered en masse to the community simultaneously? 
 

سياق التقييم

Mexico has 2,463 municipal governments led by mayors, typically elected to three-year terms. Until recently, consecutive re-election of mayors was not allowed. As such, municipal election campaigns in Mexico are generally oriented around political parties rather than specific candidates, and voters likely draw inferences about parties’ current candidates based on previously elected officials from that party. Additionally, according to the Comparative National Election Project’s 2018 survey in Mexico, 31 percent of Mexicans reported talking to neighbors, 44 percent to friends, and 56 percent to family about electoral campaigns sometimes or often in 2018, demonstrating the frequency of political discussions occurring within municipalities.

The Municipal Fund for Social Infrastructure’s (FISM) federal transfers represent around a quarter of the average municipality’s budget and must be spent on infrastructure projects that benefit marginalized localities, citizens in extreme poverty, or priority zones1. To increase transparency, audits are conducted by the Federal Auditor’s Office on the use of FISM funds in 150-200 municipalities every year to identify irregularities in expenditures. Between 2009 and 2018, 17 percent of FISM expenditures were determined to have irregularities, which may entail spending on unauthorized projects or funds unaccounted for in the municipal budget. In 2018, of the 128 municipalities in this study, 48 percent had no irregular spending. The majority reported irregularities well below 10 percent. Despite lower levels of irregularities than in previous years, during the 2018 municipal election, corruption was a salient issue nationwide. Indeed, the recently-founded left-wing, anti-corruption party, the National Regeneration Movement (MORENA), won federal and local elections across the country. 

In 2018, 64 percent of Mexican adults used social media, spending an average of more than three hours a day using it, and Facebook and WhatsApp were the prime channels for spreading political content. Additionally, misinformation was a particular concern during the 2018 election campaign, where political parties were accused of spreading misinformation online to influence voter behavior. One non-profit organization that aims to counter the growing trend of misinformation is Borde Político (BP), a Mexican NGO which promotes government transparency using digital tools, like Facebook and Twitter. 
 

An example of the Facebook ad used in the evaluation showing high and low spending irregularities.
An example of the Facebook ad used in the evaluation, showing cases of high and low spending irregularities by municipal governments.
Photo: Borde Político

معلومات تفصيلية عن التدخل

Researchers partnered with BP to test the impact of disseminating nonpartisan information about municipality expenditures on voter behavior. Using the publicly-available audits, BP developed a set of 26-second-long nonpartisan Facebook video ads to inform citizens about the extent of irregularities in their municipality’s expenditures. 

In order to understand how the level of ad targeting within a community might impact voting behavior, researchers targeted different percentages (saturation levels) of the electorate with videos. They selected 128 municipalities with elections in 2018 where the mayor in office before the election had presided over the expenditures audited by ASF. These municipalities were split into 42 groups of three municipalities, each governed by the same incumbent party (the party currently holding office), and then randomly selected into one of the following three groups:

  1. Comparison group: In these municipalities, no one was targeted with Facebook ads. 
  2. Low saturation: 20 percent of all voting-age Facebook users in these municipalities were targeted by BP’s nonpartisan Facebook ads during the week before the elections. 
  3. High saturation: 80 percent of all voting-age Facebook users in these municipalities were targeted with the Facebook ads during the week before the election. 

To achieve the 20 and 80 percent saturation levels, the researchers divided each municipality into multiples of five contiguous and largely compact segments of roughly equal population size. Four of every five such segments were targeted with ads in high-saturation municipalities, while one of every five such segments were targeted with ads in low-saturation municipalities.

To determine who viewed ads and how many times an ad was viewed, researchers used data from Facebook’s ad analytics. To measure impact, researchers used precinct-level electoral results, focusing primarily on the municipal incumbent party’s vote share as a share of total votes cast. 

In addition to receiving ethical review and approvals from institutional review boards, researchers made further efforts to address and account for ethical questions. For example, they complied with Mexican electoral law2, ensured the campaign remained nonpartisan3, leveraged credible and publicly accessible information, and did not ask citizens to respond to the information in any way. 

النتائج والدروس المستفادة بشأن السياسات

In municipalities where voting-aged adults were targeted with the nonpartisan Facebook ads, incumbent parties with zero or negligible financial irregularities received more votes relative to incumbent parties in comparison group municipalities. Much of the increase in votes can be attributed to voters in municipalities that received the high saturation ad campaign, where there were also significant spillover effects.

Reach of Facebook ads: In practice, the ads reached close to 14 percent of voting-age adults in low-saturation municipalities and 56 percent in high saturation municipalities. Additionally, the ads appeared three times per person on average, with 15 percent of targeted voting-age adults watching at least three seconds of at least one ad.

Direct electoral impacts: Across precincts within targeted population segments in both high- and low-saturation municipalities, incumbent parties with little to no irregularities experienced a 6 to 7 percentage point increase in their vote share relative to the 28 percent mean vote share for the incumbent party in comparison group municipalities. There was a 7 to 8 percentage point increase for incumbent parties with little to no financial irregularities in precincts in the high-saturation municipalities versus only a 2 to 3 percentage point increase in precincts in the low saturation municipalities. However, while negative, there were no significant impacts on parties that had an above-average level of financial irregularities. Given the public’s low expectations of politicians in Mexico and the lower levels of FISM irregularities detected in 2018 than in previous years, BP’s ads likely reported better performance than many citizens expected. 

Indirect electoral impacts: Voting behavior was also affected in precincts within targeted municipalities where no adults in a segment were targeted by the Facebook ads. This effect was significantly larger in high-saturation municipalities, where untargeted precincts were 7 percentage points more likely to vote for the best-performing incumbent parties as compared to precincts in untargeted municipalities. This means that those who were not directly targeted with the ads on Facebook still changed their vote after ads were disseminated in their municipality, and were more likely to do so if a higher proportion of people in their municipality received the ads. Researchers posit this is a result of interactions between voters within more socially-connected municipalities, based on more coordinated vote choices and larger effects observed in municipalities where more individuals are friends on Facebook.

Taken together, these findings show that nonpartisan information campaigns on social media can support electoral accountability, especially when a large proportion of the electorate is targeted. Such dynamics may also arise for mass broadcast media, although these technologies may have higher costs and barriers to entry. In contrast, targeted social media ad campaigns are a relatively low-cost intervention that may be particularly impactful in middle- and low-income countries, where social media use is rapidly growing while governance challenges remain. Researchers note that the nonpartisan nature of this information campaign was likely a crucial detail: partisan ad campaigns in high-income countries have not produced as meaningful results.

The implementing partner, Borde Político, did not end up using the information from this evaluation because it became an NGO accredited with Mexico's National Electoral Institute, which prevented it from engaging in related activities. However, the results from this evaluation led another partner, Data Cívica, to start to run a similar evaluation, which is pre-registered here.

1.

In 2010, the CONAPO defined 79 percent of localities as marginalized. Eligible projects include investments in the water supply, drainage, electrification, health infrastructure, education infrastructure, housing, and roads.

2.
Mexican electoral law allows NGOs to exercise the freedom of expression they enjoy as collectives of citizens in order to disseminate nonpartisan information.
3.
Researchers used the full set of municipalities for which audit reports pertaining to the incumbent mayor’s government were available, randomized group assignment within blocks of municipalities governed by the same political party, and avoided the use of color schemes associated with any particular political party in the ads.