The shape of democratic discourse in Latin America: four decades of media coverage, eleven countries

 

Latin America spans the full democratic spectrum — and that spectrum is visible in the structure of its news coverage.

Latin America is the region where the third wave of democratization made some of its clearest gains, and where the democratic recession of the past fifteen years has bitten back hardest. Eleven countries in the region, with histories ranging from Argentina's military dictatorship to Venezuela's authoritarian consolidation, allow us to look at four decades of news coverage and ask a simple question: what does media-side political deterioration look like, and how does it differ from media-side recovery?

This post extends the global analysis described in my previous post to a Latin American focus. The same measure — β, the conflict framing elasticity — applied to 11 LatAm countries (Argentina, Bolivia, Brazil, Chile, Colombia, Dominican Republic, Ecuador, Honduras, Mexico, Peru, Venezuela) over the GDELT v2 monthly data from February 2015 through May 2026. Four of these countries (Brazil, Mexico, Colombia, Venezuela) have full GDELT v1 coverage back to 1979, allowing a long-run view across the third wave and the democratic recession.

[Insert infographic: Feged-Rivadeneira_Infographic_LatAm_v2.png]

Two structural patterns

The infographic makes two patterns clear that the regional discourse has either obscured or simplified.

Venezuela's authoritarian-consolidation trajectory is the steepest in the dataset. Goldstein scale measurements (which capture the conflict-cooperation balance of events covered, and which are not affected by the GDELT v1→v2 measurement break) decline at a rate of −3.54 per century (p<.001). The decline is most visible in the 2017 constituent assembly era, when the country's monthly z-tone dropped sharply from positive to deeply negative and never recovered. Of all the country trajectories I have looked at across the world, Venezuela's is the cleanest example of media coverage tracking a long-run authoritarian shift.

Colombia's post-conflict recovery is also visible. Goldstein rises from −1.5 in the early 1990s (peak FARC insurgency) to nearly +1.0 by the mid-2010s. The 2016 FARC peace agreement appears as a final consolidation, with the trajectory holding through 2026. This is the clean positive-control case for the method: media coverage tracked a verified political improvement closely.

Mexico is the regional canary. A monthly event study shows a sharp drop in z-tone in late 2018 corresponding to the AMLO administration's inauguration. The drop is sustained through 2026. β for Mexico is rising fastest among the LatAm countries we measured — the only one in the region with a statistically significant rising trend in the original 40-country annual analysis (p = .026).

The other countries

Most of the remaining LatAm countries — Argentina, Chile, Brazil, Dominican Republic — sit in the healthy-discourse attractor of the global field. Their monthly tone distributions are positive on average and have not shifted substantially between 2015–2019 and 2020–2026. They are democracies with active politics, and that activity shows up in the data, but the structural drift is not toward authoritarian floor.

Honduras, Ecuador, and Bolivia sit in the unstable middle. Their monthly tone distributions are negative and widely dispersed. None of these is a Venezuela; none of these is a Colombia. They are exactly the cases where Carothers and Press's argument about the contested middle of democratization applies: countries from which the path to either attractor is short, and the determining factors are political choices made in the present.

What this means and what it does not

The method does not predict regime change. It detects shifts in the way events are framed in the news. Those shifts may follow political deterioration, may anticipate it, or may reflect parallel processes (media consolidation, source-corpus shifts, polarization). Triangulating between β, Δβ, and conventional democracy indices is more informative than any one of them alone.

What β does add is real-time monitoring at near-zero cost. Anyone with BigQuery access can compute the metric for any country in any month. The eight LatAm countries above could be tracked weekly if desired, and the method would flag anomalous shifts earlier than annual survey-based indices.

Source code and queries: github.com/alejandrofeged/conflict-framing.