All Events by Year
The Democratic Erosion Event Dataset v.3 contains 1772 unique events on democratic erosion in 98 countries, occurring between 2000 and 2018. Precursors and symptoms of erosion make up 42 percent and 38 percent of the dataset respectively, while cases of resistance to erosion comprise the last 20 percent. The number of events coded jumps after 2012, due both to the rise of the phenomena and bias towards more recent news in searches for case studies. While the data cannot claim to encompass all erosion-related events in these countries, it is the most systematic attempt so far to produce disaggregated event data with such level of detail. This is particularly helpful in understanding trends across time and space, as well as type of erosion patterns. A description accompanies each event, making the dataset a powerful research tool for analyzing democratic erosion both quantitatively and qualitatively.
All Events by Type
Events are further classified into 44 “types,” with vertical corruption (corrupt deals between government officials and citizens), nonviolent protest, and media repression dominating the three main categories. Media repression is the most common event in the dataset, logged nearly 200 times. The fact that the data is compiled from news reports suggests a possible bias towards media-related democratic erosion, as exposing these events is directly in the interest of news sources.
Media Repression Worldwide
Selecting a specific type of event, like media repression, one can compare its prevalence around the world. Media repression appears in all regions, but concentrates in Russia and India, the only countries with over 10 events. This visualization also flags Guyana, Ecuador, and Turkey as having high levels of media repression. The qualitative descriptions of the events reveal that government officials in Guyana have brought several high-profile lawsuits against the press for investigations into corruption and pressured private news sources to conform to the state media. While the dataset is limited to the 98 countries included in the student case studies, the data allows for specific, robust analysis of which types of democratic erosion prevail and where.
Horizontal and Vertical Events by Region
Each event in the dataset can be characterized as horizontal or vertical erosion. Horizontal events relate to checks on power between government branches and officials, while vertical events relate to checks on the government by citizens and civil-society organizations. This characterization aids analysis of what causes a breakdown of democracy and where resources should be focuses to strengthen a democratic system. Worldwide, vertical erosion is far more common, likely due to the fact that horizontal erosion requires structural changes in government while vertical erosion, such as media repression, can occur more spontaneously. Looking at horizontal erosion, Latin America and South/East Europe display similarly high levels, both driven by many cases of reduction in judicial independence. They differ in that horizontal erosion events in Latin America tend to be led by the president with the goal of extending term limits, while in Eastern Europe the legislature seek to erode checks and balances and stifle opposition parties. One explanation is that Nicaragua and Bolivia account for most of the horizontal erosion in Latin America, both of which have longstanding, strongmen presidents. In South/East Europe, however, events are concentrated in countries with recently-elected executives, such as in Hungary and Macedonia.
Precursors v. Symptoms v. Resistance
The relationship between the three core variables merits analysis. In these plots, bubble size represents the number of resistance events, ranging from zero to 14 (South Korea). As evident by the R2 of 0.035, the number of precursors in a country has a very weak linear relationship with the number of symptoms there. When the data is staggered as pre-2012 precursors and post-2012 symptoms, the correlation is even weaker, suggesting that precursors don’t predict symptoms in the conglomerated dataset. One explanation is that students non-randomly select which events to record, and that symptoms, which are more evidently democratic erosion, are more likely to be chosen for a case study. This may explain why some countries, like Bangladesh, exhibit five times as many symptoms as precursors, while very few, like Macedonia, show a substantial amount of precursors.
Precursors (pre-2012) v. Symptoms (post-2012) v. Resistance