Methodology

The construction of the dataset takes advantage of a unique opportunity: partnership with a multi-university course on Democratic Erosion. The dataset was constructed by aggregating and coding narrative country case studies authored by about 150 students simultaneously enrolled in a Democratic Erosion course across nineteen universities during the 2017-18 academic year. Students in this collaborative course were instructed to select their case study topic from a list of 67 countries, identified by the Capstone Team as potential backsliders using one measure of democratic governance from the Varieties of Democracy dataset (Coppedge et al., 2017) between 2007-2016. The methodology used to select potential cases was determined jointly by the Capstone Team and the DRG. First, any electoral or liberal democracy that experienced a decline in the Liberal Democracy Index over the study period was identified as a potential backslider. From those 108 cases, we eliminated eight island or micro-states and 33 cases in which the mean amount of backsliding was less than 1%. To the 67 remaining cases, we added about ten additional ones we thought were particularly interesting but did not make the original list because they were electoral autocracies when they backslid (rather than electoral democracies).

Each case study details the circumstances through which erosion manifested in a particular country, any precursors that precipitated the erosion, and any resistance and/or recovery in response to the erosion. Students were also asked to provide an assessment of the degree of overall erosion in their case study country on a five-point scale.

Please feel free to navigate this section via following links:

Case Selection
Censoring the dataset
Restricting to electoral democracies
Coding backsliding
Excluding cases
Analytic framework

Case selection

To identify the original list of case studies for the meta-analysis on democratic backsliding, we use the Varieties of Democracy dataset (Coppedge et al, 2017).


Censoring the dataset:

First, the country-year dataset is constrained to only include the past decade, e.g. years 2007-2016.

Restricting to electoral democracies:

Then, because we are looking for cases of democratic erosion, we define democratic backsliding as originating in a country-year in which the country is coded as an electoral democracy. To identify countries-years that qualify as minimal electoral democracies, we use the Regimes in the World index (e_v2x_regime) which has already been coded for all years (rather than just election year). In the original iteration, we require a score of 2 or higher for year t=1. In year t=2, the regime can backslide to a score of 1, which is equivalent to having a score of 2 on the multiparty elections variable. The full coding of this variable is as follows:

  • 0: Closed autocracy: No de-facto multiparty elections for the chief executive).
  • 1: Electoral autocracy: De-facto multiparty elections for the chief executive, but failing to achieve a minimum level of Dahl’s institutional prerequisites of polyarchy as measured by V-Dem’s Electoral Democracy Index (v2x_polyarchy).
  • 2: Electoral democracy: Free and fair multiparty elections and a minimum level of Dahl’s institutional prerequisites for polyarchy as measured by VDem’s Electoral Democracy Index (v2x_polyarchy), but liberal principles of respect for personal liberties, rule of law, and judicial as well as legislative constraints on the executive not satisfied as measured by VDem’s Liberal Component Index (v2x_liberal).
  • 3: Liberal democracy: Free and fair multiparty elections and a minimum level of Dahl’s institutional prerequisites for polyarchy as measured by VDem’s Electoral Democracy Index (v2x_polyarchy), and liberal principles of respect for personal liberties, rule of law, and judicial as well as legislative constraints on the executive satisfied as measured by V-Dem’s Liberal Component Index (v2x_liberal).

Coding backsliding:

To code democratic backsliding, we use the liberal democracy index (v2x_libdem). This measure places special weight on constraints on executive power. From the codebook: “The liberal principle of democracy emphasizes the importance of protecting individual and minority rights against the tyranny of the state and the tyranny of the majority. The liberal model takes a ‘negative’ view of political power insofar as it judges the quality of democracy by the limits placed on government. This is achieved by constitutionally protected civil liberties, strong rule of law, an independent judiciary, and effective checks and balances that, together, limit the exercise of executive power.”

We code a country-year, t, as backsliding if the country received a lower score on the Liberal Democracy Index in year t than in year t-1. In addition, the country had to receive a score of at least 2 on the Regimes in the World index (indicating an electoral democracy) in year t-1 and a score of at least 1 in year t.

In addition to coding whether or not backsliding occurred in that country-year, we also code how much backsliding occurred in percentage terms (change in Lib Democracy Index divided by last year’s score).

Excluding cases:

This exercise elicited a list of 108 countries that had at least one year of backsliding in the last decade. To prioritize cases, we constrained the list using several criteria.

  • We eliminated island or micro-states (8 total).
  • We eliminated cases in which the mean amount of backsliding was less than 1.5% (33 total).

Analytic framework

The first six weeks of the capstone course focused on the theoretical literature on democracy and democratic erosion to enable the team to develop an appropriate framework for the coding of the event data. Through these readings, particular characteristics of democracy and its erosion were identified that could be used for the coding methodology. The entire team then read the same five country case studies and created a joint inventory of events from each of the cases. Using the completed inventory, the team identified similar logged events to construct conceptually-distinct groupings and create more reliable variable categories. Together, the team debated the potential categorization of events into the characteristics from the theoretical literature. It was decided that there was a fundamental difference between events that seemed to be leading to severe erosion, or precursors, and events where erosion had been institutionalized, or symptoms of erosion. The precursors were split into civic, economic, political, institutional, and violent/security events with a final “other” category to capture events that did not fit into the other subcategories. The symptoms were split into a reduction in vertical accountability, horizontal accountability, and a change in societal norms. Lastly, there were a number of events in the case studies in which citizens resisted these forms of erosion. To capture them, the team coded resistance as the antithesis of the symptoms, resistance to horizontal accountability, and resistance to vertical accountability as well as an “other” category. The team also decided to only code dynamic rather than static events. For example, if inequality had been a consistent challenge, it was not coded, but if there was a sudden increase in inequality the event would be coded. Our final event framework is depicted in Table 1 below. Following this, we briefly describe the theoretical background of each of these subcategories.


Precursor Symptom Resistance
Civic
Lack of legitimacy
Media bias
Polarization
Increasing control of civil society
Reduction in horizontal accountability
Suspension of rules/constitution
Relaxing of term limits
Circumventing the rule of law
Reducing judicial independence
Reducing legislative oversight
Weakening integrity institutions
Increase in horizontal accountability
Check on central power by subnational government
Check on executive by judiciary
Check on executive by legislature
Economic
Corruption
Economic inequality
Economic shocks
Reduction in vertical accountability
Media repression
Repression of opposition parties
Systemic reduction in electoral freedom/fairness
Curtailed civil liberties
Increase in vertical accountability
Nonviolent protest
Violent protest
Increase in civic capacity
Coalitions or elite pacts
Political
Cooptation of the opposition
Extremist/populist parties
Malapportionment
Party weakness
Electoral fraud
Changing societal norms
Lack of confidence/public disillusionment
Threats and intimidation
Other
Pressure from outside actors
Exit of people or money
State attempts to prevent backsliding
Institutional
Delegitimizing or weakening judiciary
Coup or regime collapse
State restructuring
Manipulation of civil service
Constitutional reforms
Violence/security
Non-state violence
State-sponsored violence or abuse
Electoral violence
Other
Refugee crisis
External realignment
Prior failed attempts at erosion

 

Coding

With the variable categories defined, the Team created a coding instrument in Google Forms to streamline the process of coding all student case studies. The form allowed each teammate to quickly input data from new case studies in a uniform fashion that enables comparability and will facilitate future analysis of cross-country data. In addition to coding every event documented in each narrative case study (according to the above categorization scheme), Team members added a short description of each event, the year(s) of the event, and a final ranking on a five-point scale that assesses the severity of backsliding in the country case. After testing the form with 65 case studies to gauge both reliability and comprehensiveness, minor adjustments were made prior to the complete review of case studies produced during the Spring 2018 semester.