The package title is now “Contentious Politics and Civil Conflict Research”.
cn_cite() now returns its result visibly instead of printing it with cat(). The result is a bibentry object with the additional class cn_citation; typing cn_cite("archigos") at the console still shows the BibTeX, and print() shows it explicitly.
The bundled data files are smaller (tarball 4.3 MB): they keep only the columns the package reads, and the MEC and NMC Stata files are stored as compressed .rds. Loader output is unchanged. ‘haven’ is no longer used.
citation("contentiousR") now lists the SSRN preprint describing the package, followed by the software citation.
Bug fixes
load_vdem_data(coding_system = "cow") no longer silently drops East Germany (COW 265, 1949-1990); it was previously excluded only in the "cow" branch, so V-Dem coverage for Germany differed depending on the chosen coding system with no indication why.
add_spell_duration() now computes spells from the actual calendar-year difference instead of row position within a state. A gap in the input years (e.g. left by tidyr::drop_na(event), which the package’s own examples recommend) previously understated the elapsed time.
plot_coverage() now errors with a clear message instead of crashing when drop_empty = TRUE removes every state (e.g. the chosen event never occurs in the supplied panel/year range).
cn_cite() matched keywords by substring, so a short dataset name (e.g. "mm") could also match an unrelated entry whose keyword merely contained it as a substring ("mmad"). Keyword matching is now exact against the comma-separated keyword list.
harmonize_conflict_data(): a factor year_col was coerced with as.integer(), which returns the factor’s internal level codes rather than the printed year. It is now converted to character first. A date_col that fails to parse (e.g. a non-ISO format without a matching date_format) now raises a clear error instead of a generic base-R one.
Bring your own data
Added harmonize_conflict_data() to join externally supplied conflict/ event data (GDELT, ICEWS, ACLED, or anything else not bundled with contentiousR) onto a state panel: it maps the caller’s country column to COW/GW codes via countrycode::countrycode(), extracts a year from either a pre-parsed year_col or a date_col/date_format, optionally aggregates event-level rows to country-year (<prefix>_onset, <prefix>_n_events), and joins the result in. Unlike add_conflict(), it never downloads or bundles data itself, so contentiousR gains no dependency on gdeltr2, icews, or acled.api.
Visualization
plot_coverage() gained a show_labels argument (default TRUE). Set it to FALSE to drop the per-row country/code labels, which otherwise overlap and become illegible once the panel covers more than a few dozen states.
plot_regional_coverage()’s region labels are now horizontal and word-wrapped instead of angled, and its region field’s canonical “Middle East” label is the shorter “Middle East & North Africa” form. plot_temporal_coverage()’s period labels are now vertical instead of angled. Both plots no longer set a ggplot2 title (redundant with a caption in a manuscript).
Added plot_regional_coverage() and plot_temporal_coverage() to compare regional and temporal event coverage across any combination of conflict_data() campaign/episode datasets, plus a “Comparing revolutionary-event datasets” article reproducing the descriptive exercise in Guseinov, Ustyuzhanin & Korotayev (2026).
Added add_regions() to join one or more countrycode region classifications (region, region23, un.region.name, un.regionsub.name) onto a state panel by cow/gw code.
Data
Added the Major Episodes of Contention (MEC) dataset: 2,734 globally covered reformist and maximalist episodes from 1955 through 2018. MEC is available through conflict_data(dataset = "mec") and add_conflict(dataset = "mec"). Episode-level rows are preserved unless the existing aggregate = TRUE policy is requested.
Correctness
fetch_campaign_events() (used by plot_regional_coverage() and plot_temporal_coverage()) now compares NAVCO 2.1 at the same campaign-onset level as the other datasets, using conflict_data()’s own nvc2.1_ONSET flag (keyed on the numeric campaign id) instead of re-deriving onsets from the campaign name, which is not unique and could silently merge distinct campaigns that happen to share a name.
fetch_campaign_events()’s region field now stays at exactly 7 categories. countrycode::countrycode()’s "region" destination labels a few pre-1990 historical entities (Yemen Arab Republic, Yemen People’s Republic, the United Arab Republic) with the World Bank’s older “Middle East & North Africa” string while every other MENA country gets the newer “Middle East, North Africa, Afghanistan & Pakistan” string; since campaign data covering the 1950s-80s includes those historical codes, the region was silently splitting into two bars in plot_regional_coverage()/plot_temporal_coverage().
gdp_growth (load_gdp_data(), dataset = "gapminder") is now NA unless the preceding row is genuinely one year earlier for the same country, the same gap-safety check add_lag() uses.
Renamed add_spells() to add_spell_duration() to stop it silently shadowing (or being shadowed by) peacesciencer::add_spells() when both packages are attached in the same session – the two are similar in spirit but not interchangeable (contentiousR’s version works on any binary _onset/_incidence/_ongoing column, not a fixed set of bundled columns).
conflict_data() now applies start_year and end_year consistently to every bundled source.
UCDP/PRIO output now distinguishes conflict incidence from episode onset. ucdp_prio_onset is based on start_date2; the new ucdp_prio_incidence column identifies all active conflict-years.
SCAD country-year totals now exclude UCDP placeholder records, deduplicate multi-location rows by event ID, and treat documented negative death-count codes as missing.
All-missing groups now produce NA instead of -Inf, NaN, and associated warnings during conflict aggregation.
Mass Mobilization participant strings are retained in mm_participants_reported; mm_participants summarizes only exact numeric reports.
add_conflict() no longer re-aggregates sources that are already unique by country-year, preventing character-valued source fields from being dropped.
API and robustness
Added as_peacesciencer_panel() to supply the state-year key aliases and metadata required by peacesciencer while preserving contentiousR’s API.
Added add_from_peacesciencer() to apply a state-year peacesciencer function while keeping temporary ccode/gwcode keys out of the result.
build_states_cow_panel() is internal again (not exported). It’s the actual COW-panel implementation build_states_panel(coding_system = "cow") dispatches to, but there’s no public reason to call it directly instead of build_states_panel() – and its GW counterpart, build_states_gw_panel(), was never exported in the first place.
Added strict validation for years, logical flags, panel keys, and output-key uniqueness.
build_states_panel() now validates start_year/end_year against the Correlates of War / Gleditsch-Ward state system’s actual coverage: 1816-2025.
Added ucdp_vpp_incidence. The existing ucdp_vpp_onset column remains as a compatibility alias and is now documented as incidence rather than onset.
Removed the magrittr dependency in favor of R’s native pipe.
Release infrastructure
Added broader test coverage, two introductory vignettes, package-level documentation, data provenance and licensing notes, pkgdown configuration, and GitHub Actions for package checks and site deployment.
Added a case study article assembling a panel and modeling revolutionary onset, and a “Comparing revolutionary-event datasets” article (see Visualization above).
Added a package logo, shown in the README and the pkgdown site navbar.
