Revolution and Resistance Turning-Point Database

Full report

On this page

Building a turning-point database involves systematically identifying and coding critical events where popular movements altered a regime’s course. We define “turning points” as specific events or episodes that decisively shift the trajectory of a political struggle. For example, the 8 September 1978 “Black Friday” massacre in Iran is widely described as “a turning point” that accelerated the Shah’s fall. Our first task is to compile candidate events from multiple sources:

• Global event datasets: Use large-scale protest/conflict databases to seed event lists. For instance, the NAVCO project catalogs 627 mass-mobilization campaigns worldwide (1900–2021). The ACLED database similarly “collects, analyses, and maps data on conflict and protest” globally. The Mass Mobilization Project records 10,000+ anti-government protests (1990–2020) with descriptions of protester demands and government responses. These sources yield thousands of events (e.g. Clark & Regan’s dataset has ~17,000 entries). We will extract from these datasets any major mobilizations or crises (large protests, strikes, uprisings) that could plausibly have changed political direction.

• Revolutionary episode studies: Use scholarly compilations of revolutionary outbreaks. For example, Beissinger’s Revolutionary Episodes dataset codes 345 episodes of high-intensity contention (1900–2014) with detailed narratives, tactics, and outcomes. Such episodes often contain key turning-point events. We will review those episode reports to identify events like regime concessions or military splits within each case.

• Specialized datasets: Incorporate data from transition-focused projects. The recent ROAD dataset (Resistance Overturning Attempts) documents every attempted coup or rollback in 68 “civil resistance” transitions since 1970. This covers e.g. Egypt 2013, the Philippines 1986, Mali 1991, and others. Using ROAD, we note each “overturning attempt” as a candidate event (success or failure). The dataset provides summaries of each attempt’s actors and outcome, which we will incorporate.

• Historical compilations and secondary sources: Complement quantitative data with expert chronologies and case studies. Histories of revolution (e.g. works by Skocpol, Tilly, Huntington) often highlight key turning points. We will consult timeline monographs, scholarly articles, and encyclopedias to spot events like “massive strike spurs leader’s resignation,” “army refuses orders,” etc. For example, studies note that in the Philippines’ 1986 People Power revolution, mass protests and a military defection forced President Marcos to flee. We will mine such accounts to add or verify events.

By triangulating these sources, we generate a global list of candidate turning-point events. We will also flag events from the question’s example classes – e.g. large protests or general strikes, concessions or pact-signings, coups or failed coups, defection by security forces, constitutional rulings, foreign interventions, etc. – since these often mark changes in political trajectory. (For instance, one study emphasizes that shifting the military’s loyalty is a critical success factor in overthrowing dictators.)

Gathering Sources and Verifying Events

For each candidate event, we collect primary and contemporaneous sources to establish facts. Sources include:

• Media archives: Newspapers, wire services, and journalists’ accounts. Major archives (e.g. New York Times, Washington Post, BBC) often report on protests, crackdowns, and negotiations. Clark & Regan’s Mass Mobilization project, for example, used LexisNexis to retrieve reports from major international papers for each protest episode. We will similarly search news databases for reports on each turning point (using keywords from the event description and date).

• Official documents: Government press releases, decrees, legal rulings, parliamentary records, etc. For instance, a constitutional ruling overturning martial law or a court decision annulling an election can be found in official gazettes or online archives of judiciaries. Military orders (e.g. declarations of martial law) or negotiation transcripts can sometimes be accessed via government archives or freedom-of-information requests.

• Movement publications: Some protests or negotiations are documented by the movements themselves: e.g. manifestos, open letters, union bulletins. These can come from party archives or digital collections of NGOs.

• Eye-witness accounts and memoirs: Leaders’ memoirs, diaries, or oral histories often highlight moments they regarded as turning points. These complement other sources and help with fields like “scholarly interpretation” (see below).

• Academic and secondary analyses: Political science and history papers provide context and interpretation. We will use these to cross-check reported facts and to assign meaning. For example, Erica Chenoweth and colleagues note that large-scale loyalty shifts in a regime’s military tend to precipitate leaders’ downfall. When multiple credible sources (news + academic) confirm an event and its effect, we gain confidence in coding it as a turning point.

Each event is verified by cross-referencing at least two independent sources when possible. When accounts conflict (e.g. on casualty counts or who initiated an action), we record all versions in our notes and assign a confidence score reflecting consistency of evidence. For instance, the “Black Friday” shooting in Tehran has widely varying death tolls (state vs. opposition estimates), but all agree it sparked nationwide strikes – this event’s significance is confirmed even if numbers are uncertain.

Event Schema and Coded Attributes

We standardize each event into a structured record with fields as specified. For clarity, fields include:

• Date (day/month/year) and Location (country, and if relevant city or region).
• Movement (name or description of the movement/revolt) and Context (e.g. “Arab Spring – Egypt” or “Velvet Revolution – Czechoslovakia”).
• Description: A brief narrative of what happened (e.g. “Mass strike in Petrograd demanding constituent assembly” or “Presidential resignation amid protests”).
• Actors: Key individuals and groups involved (e.g. “President X; Central Committee of Party; Army High Command; union leaders; UN mediator”). We will standardize actor-types (e.g. state actors like “army, police, king, government”, and non-state actors like “civilians, labor unions, opposition parties”).
• Institutions: Which formal bodies were engaged (e.g. “military, legislature, judiciary, national assembly, secret police”). For example, if a constitutional court ordered a leader’s removal, we would list the judiciary here.
• Mechanism: The means by which the turning point occurred. This reflects the categories given (mass mobilization, strike, coup, negotiation, etc.). For instance, a general strike or mass demonstration would be a mechanism, as would military defection, failed crackdown, or negotiated settlement. (We may adopt a controlled vocabulary or codebook for these; analogous datasets use standardized event codes – e.g., ACLED’s “event type” or CAMEO codes – to ensure consistency.)
• Immediate effect: The short-term outcome right after the event. Examples: “leader resigns”, “minister ousted”, “state of emergency declared”, “military intervenes”, “violence ends protests”. This is derived directly from source accounts.
• Eventual effect: The longer-term impact on the political trajectory (e.g. “regime collapsed”, “new constitution adopted”, “military junta established”, “status quo preserved”). This may be apparent only in retrospect. We differentiate immediate vs. eventual effect because some events (like a crackdown) may momentarily succeed, but later lead to renewed revolt.
• Scholarly interpretation: A summary of how analysts or historians view this event’s significance. For example, the Mass Mobilization Project notes protest demands and regime responses, while Chenoweth et al. emphasize that defection of top military brass often “hasten[s] [the dictator’s] fall”. We will note interpretations like “tipping point” or “turning point confirmed by historians” with citations.
• Confidence: A qualitative rating (high/medium/low or 1–5) of how confident we are in the coding, based on source agreement. For example, well-documented nationwide events get high confidence, ambiguous cases lower.
• Sources: References for the event (primary documents, news articles) and for interpretation (academic studies, histories). We will record bibliographic details or URLs. For example, our Iran Black Friday entry would cite contemporaneous news and later analyses.
• Tags/Concepts: We include tags such as those in the prompt (“mass mobilization”, “negotiated settlement”, etc.) or concepts from theory (e.g. nonviolent resistance, elite defection). These allow thematic queries. (These are “BuiltToResist” concepts in the prompt, presumably referring to a framework of tactics/strategies.)

This schema is similar in spirit to other event collections. For example, Beissinger’s episodes dataset records “forms of contention” (locations/tactics used) and “outcomes” for each episode, which parallels our “mechanism” and “effects” fields. Clark & Regan’s Mass Mobilization data include “protester demands” and “government response”, akin to our mechanism and immediate effect. By aligning our fields with established datasets, we ensure interoperability and ease of comparison.

Coding Process and Cross-Validation

Using the above schema, trained coders will review each source to populate fields. We will draft a codebook defining terms (e.g. what counts as a “general strike” vs. a local labor protest). In ambiguous cases coders note uncertainty and multiple interpretations (later resolved by team review).

To ensure reliability, we will cross-validate interpretations: each event’s coding will be checked against independent scholarly accounts. For instance, if our newspaper research finds that “the army declared neutrality” on a given day, we confirm with academic studies whether that was indeed a turning point. We may find disagreements: some authors might call a negotiation a “successful compromise,” others a “sell-out.” We will document differing views in the “scholarly interpretation” field and may tag that event with lower confidence if consensus is lacking.

Comparative literature will guide interpretation. For example, Kara Neu’s research on nonviolent revolutions shows that even unified military defections (like Mali 1991, Bangladesh 1990) have different implications: a unified defection can let generals control the transition, whereas a fractured defection often enables democracy. We can apply such insights: tagging Bangladesh 1990 as a “fractured defection – democratic transition” versus Mali 1991 as “unified defection – junta interim.” Similarly, Elmore et al. note that 72% of “civil resistance” transitions faced at least one coup attempt. We use this fact to check completeness: if our list of a certain transition (say, Serbia 2000) missed a known overturn attempt, we re-examine sources.

Whenever possible, we cite peer-reviewed or well-regarded analyses. For example, ROAD finds that Egypt’s 2013 coup “not only led to a new authoritarian regime”. In our Egypt database entry, that interpretation would be noted (Egypt’s transition → authoritarian outcome) with the citation. In contrast, the Philippines’ 1986 People Power is often classified as a successful nonviolent ouster; Neu writes that a faction of the military defected “leading Marcos to leave the country”. We record that as a positive outcome and cite Neu for mechanism.

Database Assembly and Query Examples

With coded records assembled, the database can be implemented in a structured format (e.g. SQL tables or a well-indexed CSV/JSON). Each row is an event; columns correspond to the fields above. A user interface or query tool can then answer comparative questions. For example:

• Military defections: A query for WHERE mechanism LIKE '%military defection%' would return events such as the 1986 Philippine uprising (military unit breaks with Marcos) and the 1990 Bangladesh crisis (junior officers withdrew support), among others.

• Mass protests with failure: Filtering mechanism = 'mass protest' AND outcome = 'regime survives' could surface cases like certain large demonstrations that were repressed or petered out despite turnout (e.g. failed uprisings in Argentina 2001 or Belarus 2020 – to be coded if included).

• General strikes: A search for mechanism = 'general strike' might list events like the 1980 Solidarity strikes in Poland, the 2006 France protests, etc. We could then examine which of those strikes forced concessions vs. which saw the incumbent survive.

• Negotiated settlements: Query mechanism = 'negotiated settlement' and outcome = success would identify transitions resolved through pacts. Historical examples include the South African talks (Mandela release → negotiations) and Portugal 1974 (negotiated transition after Carnation Revolution). Each such entry would note the pact’s terms and effect.

• Courts or elections: Searching mechanism IN ('constitutional ruling','election','referendum') would find turning points like the 1988 Chile plebiscite (Pinochet accepted exit) or a constitutional court annulling a crackdown order. We can then see which crises were resolved legally.

• Authoritarian successors: By tagging final outcomes (democratic vs. authoritarian), one can answer “Which revolutions produced authoritarian successors?” The database would show cases like Egypt 2011–13, Iran 1979, or Albania 1990 (in ROAD’s terms, Albania’s post-1990 path became “de facto single-party dictatorship”), contrasted with events that led to lasting democracy (e.g. 1974 Portugal led to a progressive era).

Importantly, every query result will be backed by the coded fields and sources. For example, the Philippine entry might be tagged with Outcome=Democracy, Mechanism=MilitaryDefection, with Neu’s quote in the description. The “authoritarian successor” query can use the Outcome field to filter.

Illustrative example (not exhaustive): Consider the question “Which revolutions involved military defection?” In our database, we would retrieve all events where Mechanism includes military defection or neutrality. Cited evidence indicates that in the Philippines (1986), “a reformist faction of the military defected, leading Marcos to leave”. In Bangladesh (1990), junior officers’ defection helped oust Ershad without the military seizing power. Conversely, some attempts had no defection (e.g. Bahrain 2011, where the military stayed loyal). Our schema allows those differences to be queried directly, enabling systematic comparison.

Throughout, we avoid any prescriptive language; this is an analytical tool. It is structured to facilitate queries and comparisons, not to advise action. By combining quantitative event data with qualitative scholarly interpretation, the database will serve as a comprehensive reference on how and when popular resistance has truly shifted regimes worldwide.

Sources: We rely on established datasets and scholarship for grounding (see citations above). All event entries will include footnotes to original reporting and secondary analysis, so users can trace how we coded each turning point.