IESET.
Hypotheses·institutional quality·bukele_mass_incarceration_homicide_impact_2019_2024

El Salvador's ~98% homicide-rate decline from 103/100k (2015) to 2.4/100k (2023) — with the sharpest decline occurring after the Mar 2022 régimen de excepción and the Jan 2023 CECOT opening — is causally attributable in material part to the Bukele-era mass-incarceration regime (régimen de excepción + Plan Control Territorial + CECOT).

The pre-registered claim is that under a synthetic-control design using Central American and Andean peer countries (HND, GTM, NIC, HND, CRI, DOM, COL, ECU, VEN) as donor pool, SLV's post-treatment homicide-rate gap versus its synthetic counterfactual is negative and large (> 50% of observed decline) with permutation inference p-value < 0.05. The hypothesis must also address the pre-existing gang-truce confound (2012-2014 FMLN-era truce + alleged 2019-2022 undisclosed Bukele-administration truce per El Faro reporting) and the demographic factor (mara-age-cohort shrinking). An alternative channel (negotiated truce) must be tested via leaked-message-dated structural-break analysis around the Mar 26-30 2022 massacre that triggered the régimen.

PARTIALengine/runs/bukele_mass_incarceration_homicide_impact_2019_2024

PARTIAL — mean_gap=+27.06, |gap|/pre_sd=2.4, p_perm=0.222 (gap below 0.5×pre_sd or placebo p≥0.10)

confidence cueThe result is useful, but not decisive. Treat it as a clue, not a settled conclusion.

policy briefMixed or noisy

In ordinary language

In plain terms, this asks whether regimen de excepcion active is actually linked to better or worse homicide rate per 100k from 2005 to 2025.

plain answer

The evidence is suggestive but not decisive. mean_gap=+27.06, |gap|/pre_sd=2.4, p_perm=0.222 (gap below 0.5×pre_sd or placebo p≥0.10)

why it matters

This matters because institutional quality claims should change belief only when they survive a pre-declared empirical test.

how the test works

It compares 11 country or place units from 2005 to 2025, using a synthetic control design.

what was measured
What changed
  • Regimen de excepcion active
  • Detention rate per 100k
What we checked
  • Homicide rate per 100k
  • Extortion complaints rate
what this does not prove

A single test is not the whole truth. It narrows the claim under a specific sample, time period, and method. Strong policy conclusions need the pattern to survive nearby tests, alternative data, and serious objections.

verification

5 input datasets, 2 unresolved missing series, provenance status: incomplete.

Results

engine/runs/bukele_mass_incarceration_homicide_impact_2019_2024
1007550250200520152025SLVHNDGTMNICCRIDOMCOL
illustrative sketch · run pending
No coefficients yet. When the model fires, this chart will show homicide_rate_per_100k across 11 sampled countries over 2005–2025.
The shapes above are stylised — none of the lines are real data.
Placeholder for bukele_mass_incarceration_homicide_impact_2019_2024. Published chart will be generated from engine/runs/bukele_mass_incarceration_homicide_impact_2019_2024/chart_data.json.

Who has skin in the game — schools predicting on this

17 schools list this hypothesis as a test of their position. The chips below are school-level scoreboard outcomes, not a second hypothesis verdict.

hypothesis verdict vs scoreboard outcome

The banner verdict judges this hypothesis as written. The scoreboard asks whether each school's polarity-corrected prediction was right. Raw status is not a school win: SUPPORTED supports schools that needed SUPPORTED, but refutes schools that needed REFUTED.

Austrian Business Cycle Theory (ABCT)wins if SUPPORTEDpartial/no-callChicago School Monetarismwins if SUPPORTEDpartial/no-callClassical Liberalwins if SUPPORTEDpartial/no-callDegrowth / Ecological Macroeconomicswins if REFUTEDpartial/no-callDemocratic Socialism (Bernsteinian revisionism + modern)wins if REFUTEDpartial/no-callStructuralist / Developmentalistwins if MIXEDpartial/no-callEco-Socialism / Red-Greenwins if REFUTEDpartial/no-callEmpirical Pragmatist (the framework's house position)wins if MIXEDpartial/no-callNew Institutional Economicswins if MIXEDpartial/no-callMarket Socialism (Lange-Lerner tradition)wins if REFUTEDpartial/no-callMarxian Political Economywins if REFUTEDpartial/no-callMarxism-Leninism (central-planning implementation)wins if REFUTEDpartial/no-callModern Monetary Theorywins if REFUTEDpartial/no-callNew Keynesian (mainstream synthesis)wins if MIXEDpartial/no-callOrdoliberalism (Freiburg School)wins if SUPPORTEDpartial/no-callPost-Keynesianwins if REFUTEDpartial/no-callSocial Democraticwins if REFUTEDpartial/no-call

Pre-registration

pre-registered
first-spec commit bae09ab · 2026-04-29T22:09:42Z
run generated · 2026-05-15T19:53:16Z

El Salvador's ~98% homicide-rate decline from 103/100k (2015) to 2.4/100k (2023) — with the sharpest decline occurring after the Mar 2022 régimen de excepción and the Jan 2023 CECOT opening — is causally attributable in material part to the Bukele-era mass-incarceration regime (régimen de excepción + Plan Control Territorial + CECOT). The pre-registered claim is that under a synthetic-control design using Central American and Andean peer countries (HND, GTM, NIC, HND, CRI, DOM, COL, ECU, VEN) as donor pool, SLV's post-treatment homicide-rate gap versus its synthetic counterfactual is negative and large (> 50% of observed decline) with permutation inference p-value < 0.05. The hypothesis must also address the pre-existing gang-truce confound (2012-2014 FMLN-era truce + alleged 2019-2022 undisclosed Bukele-administration truce per El Faro reporting) and the demographic factor (mara-age-cohort shrinking). An alternative channel (negotiated truce) must be tested via leaked-message-dated structural-break analysis around the Mar 26-30 2022 massacre that triggered the régimen.

Falsification criterion — what would disprove this

set before the run · honoured after

This hypothesis is considered falsified if:

Not supported if any of: (a) synthetic-control estimated treatment effect on homicide rate 2022-2024 is not negative and statistically distinguishable from donor-permutation distribution at p < 0.05, OR (b) the post-2022 gap is smaller than 50% of the observed pre-to-post decline (i.e., the counterfactual donor pool also experienced substantial homicide decline), OR (c) a placebo-in-time test assigning treatment at 2019-06-01 produces a similarly large gap (indicating the decline predates régimen and is attributable to Plan Control Territorial / alleged undisclosed truce rather than mass incarceration), OR (d) the extortion- complaints secondary outcome shows no decline (indicating crime was not substantively reduced, only homicides, implying a body-disposal or reporting-manipulation confound).

formal test & threshold
test:      synthetic_control_plus_placebo_in_time_plus_extortion_secondary
threshold: CATT_2022_2024(homicides) < 0 at p_perm < 0.05 AND |CATT_2022_2024(homicides)| > 0.5 * observed_decline_2015_2023 AND placebo_in_time_CATT_2019_2021 < 0.5 * CATT_2022_2024 AND CATT_2022_2024(extortion_complaints) < 0 at p_perm < 0.10

Method

Template
synthetic_control
Clustering
country
Sample
11 countries · 2005 – 2025
Evidence type
causal

Primary: synthetic control (Abadie 2010) with SLV treated 2022-03-27 onward, pre-period 2005-2021 for donor weighting. Secondary: synth_did for matched pre-trend plus parallel- trends-free estimation. Robustness: Callaway-Sant'Anna DiD with Central American peers as never-treated; placebo-in- space tests applying treatment to each donor unit; placebo-in- time assigning treatment at 2019-06-01 (Bukele inauguration) to test whether the homicide decline precedes the régimen.

Data

VariableSourceTransform
homicide_rate_per_100k
outcome
unodc:intentional_homicidetier 2
world_bank_wdi:VC.IHR.PSRC.P5tier 2
level
extortion_complaints_rate
outcome
sv_pnc:PNCtier 3
level
regimen_de_excepcion_active
treatment
constructed:binary = 1 for SLV from 2022-03-27 onward; 0 otherwisetier 5
binary
detention_rate_per_100k
treatment
constructed from Ministerio de Justicia y Seguridad Pública + Cristosal detention talliescumulative_level
male_15_29_population_share
control
world_bank_wdi:SP.POP.1564.MA.ZStier 2
un_desa:WPPtier 2
level
unemployment_rate
control
world_bank_wdi:SL.UEM.TOTL.ZStier 2
level
remittance_share_gdp
control
world_bank_wdi:BX.TRF.PWKR.DT.GD.ZStier 2
level
gdp_per_capita
control
world_bank_wdi:NY.GDP.PCAP.KDtier 2
log_level

● ready  · ● pending  · ● reconstruct-needed

Detailed result card

Result card — bukele_mass_incarceration_homicide_impact_2019_2024

Verdict: PARTIAL — mean_gap=+27.06, |gap|/pre_sd=2.4, p_perm=0.222 (gap below 0.5×pre_sd or placebo p≥0.10)

Pre-registration

  • Claim: El Salvador's ~98% homicide-rate decline from 103/100k (2015) to 2.4/100k (2023) — with the sharpest decline occurring after the Mar 2022 régimen de excepción and the Jan 2023 CECOT opening — is causally attributable in material part to the Bukele-era mass-incarceration regime (régimen de excepción + Plan Control Territorial + CECOT). The pre-registered claim is that under a synthetic-control design using Central American and Andean peer countries (HND, GTM, NIC, HND, CRI, DOM, COL, ECU, VEN) as donor pool, SLV's post-treatment homicide-rate gap versus its synthetic counterfactual is negative and large (> 50% of observed decline) with permutation inference p-value < 0.05. The hypothesis must also address the pre-existing gang-truce confound (2012-2014 FMLN-era truce + alleged 2019-2022 undisclosed Bukele-administration truce per El Faro reporting) and the demographic factor (mara-age-cohort shrinking). An alternative channel (negotiated truce) must be tested via leaked-message-dated structural-break analysis around the Mar 26-30 2022 massacre that triggered the régimen.
  • Falsification rule: Not supported if any of: (a) synthetic-control estimated treatment effect on homicide rate 2022-2024 is not negative and statistically distinguishable from donor-permutation distribution at p < 0.05, OR (b) the post-2022 gap is smaller than 50% of the observed pre-to-post decline (i.e., the counterfactual donor pool also experienced substantial homicide decline), OR (c) a placebo-in-time test assigning treatment at 2019-06-01 produces a similarly large gap (indicating the decline predates régimen and is attributable to Plan Control Territorial / alleged undisclosed truce rather than mass incarceration), OR (d) the extortion- complaints secondary outcome shows no decline (indicating crime was not substantively reduced, only homicides, implying a body-disposal or reporting-manipulation confound).

Synthetic-control estimate

  • shape: synth_did
  • treated_country: SLV
  • event_year: 2015
  • n_donors: 8
  • donor_weights (top): {'CRI': 0.37, 'GTM': 0.2979, 'NIC': 0.2863, 'MEX': 0.0458, 'HND': 0.0}
  • pre_rmse: 41.18240091295425
  • pre_period_sd: 11.146324269091995
  • mean_post_gap: 27.060528633957627
  • end_period_gap: 2.4831568214287163
  • post_period_years: [2017, 2021]
  • placebo_p_value: 0.2222222222222222
  • n_placebos: 8
  • method: synthetic-control via NNLS, permutation inference

Variables resolved

  • unodc:intentional_homicide; world_bank_wdi:VC.IHR.PSRC.P5 (backfill via WDI mirror) → homicide_rate_per_100k (outcome, n=4372)
  • constructed: binary = 1 for SLV from 2022-03-27 onward; 0 otherwise → regimen_de_excepcion_active (treatment, n=231)
  • world_bank_wdi:SP.POP.1564.MA.ZS (proxy); un_desa:WPP → male_15_29_population_share (controls, n=16965)
  • world_bank_wdi:SL.UEM.TOTL.ZS → unemployment_rate (controls, n=8071)
  • world_bank_wdi:BX.TRF.PWKR.DT.GD.ZS → remittance_share_gdp (controls, n=8867)
  • world_bank_wdi:NY.GDP.PCAP.KD → gdp_per_capita (controls, n=14066)

Generated by scripts/run_synth_did.py at 2026-05-15T19:53:16+00:00

Strongest opposing argument

Every hypothesis ships with its charitable opposing argument. The framework earns credibility by handling objections at their strongest, not weakest.

Notes

Data-gated: UNODC and national-police homicide series for the donor pool must be current through 2024. Cristosal detention dataset (monthly series of régimen-de-excepción detentions) is the primary administrative-detention metric. Extortion- complaints data thinner but essential for the body-disposal confound test. Placebo-in-time design at 2019-06-01 is critical for separating Plan Control Territorial effect from régimen effect.

Authored framework. Read the transparency note.