Can Facebook help to keep politicians in check?

Are digital platforms always bad for democracy? Despite widespread concerns that their algorithms are spreading misinformation and political polarization, new research by IAST’s Horacio Larreguy suggests
they can also enhance electoral accountability. He tells IAST Magazine about the impact of non-partisan Facebook ads informing millions of citizens about government wrongdoing ahead of the 2018 Mexican elections.

While the potential for partisan actors to weaponize misinformation and government propaganda pose critical challenges to democracy, the digital revolution has created low-cost opportunities for targeting information toward citizens en masse. By disseminating credible information about government performance, without needing to rely on under-resourced traditional media outlets that can be vulnerable to political capture, non-partisan actors can help to improve selection and control of elected representatives by voters. This potential is particularly important in the Global South, where the use of internet and social media grew by more than 50% between 2013 and 2018, and where political malfeasance and low-quality public goods are major challenges.
We use a field experiment to estimate the effects of a large-scale social media information campaign during the 2018 general elections. In particular, we evaluate the impact of a non-partisan campaign by Borde Político — a Mexican NGO seeking to promote government transparency using digital tools — that used Facebook ads to inform citizens of the extent of irregularities in municipal expenditures. This information was extracted from publicly available audits conducted by Mexico’s independent government auditing body, and disseminated via 26-second paid-for video ads on Facebook in the week preceding the election. Corruption was a highly salient issue during the 2018 election, in which Andres Manuel López Obrador and his left-wing National Regeneration Movement party defeated traditional incumbents across the country.

In collaboration with Borde Político, we used a randomized ‘saturation’ design to identify the direct effect of being targeted by their Facebook ad campaign, and the indirect or ‘spillover’ effect on untreated areas, as well as how these effect vary with the share of the elctorate targeted by the information campaign, within our sample of 128 municipalities. By generating interactions between voters within treated municipalities, we expected the magnitude of any effects — which are likely to depend on the level of irregularities reported — to be greatest in the municipalities most heavily targeted by the information campaign.
According to Facebook’s ad campaign data, the ads ultimately reached 2.7 million unique Facebook users (appearing three times per person, on average) and resulted in around 15% of targeted voting-age adults watching at least three seconds of the ad. Engagement with the campaign was broadly proportionate with the level of access prescribed by the campaign saturation level.

Precinct-level electoral returns show that the campaign substantially affected voting behavior. Vote shares of the least malfeasant incumbent parties increased by 6-7 percentage points in directly targeted electoral precincts. This effect was greatest in directly and indirectly targeted precincts within municipalities where the campaign targeted 80% — rather than 20% — of the electorate. These results appear to reflect interactions between voters that spill hrough social networks, including to untargeted precincts, rather than responses by politicians or other media outlets. Prior studies set in the Global North have found small, but cost-efficient, effects on party vote shares from partisan political ads on Facebook and Google. Our study shows far larger effects of non-partisan information on electoral accountability in Mexico. This buttresses studies showing that information on social media can increase turnout, political knowledge, and protest against autocratic regimes, while providing a counterbalance to studies showing that social media — often through misinformation — contributes to social harms, including political polarization, hate crime, and poor mental health.
Our findings also add to the literature emphasizing that the media’s impact spills over beyond those directly exposed to its content. Professor David Yanagizawa-Drott, from the University of Zurich, for example found that indirect effects of exposure to “hate radio” in Rwanda on militia violence were at least as large as the direct effects. We further demonstrate that saturation—the share of a market targeted by an information campaign—may help explain why these effects are greater when information dissemination is conducted on a large scale.
We need to understand whether the effects generalize to other contexts and how to maximize the impact of factual non-partisan campaigns in increasingly politically polarized environments. It is not obvious how the lessons from this study implemented in a context of relatively little political polarization apply to the current Mexican context, which has seen a significant increase in pollical polarization in the past three years. Further research is then urgently required given the growth of social media, increasing political polarization, and the need to enhance electoral accountability. Since our results show that online information campaigns can have substantial electoral impacts, they also inform the debate about whether regulation is required to ensure elections are not hijacked by misinformation.

Article published in IAST Magazine #19, Spring 2022

Photo by Deeksha Pahariya on Unsplash