How many migrant deaths and disappearances will IOM’s Missing Migrants Project record in the Mediterranean region for the period 9 June to 11 August 2026, as listed on 11 August 2026?

• My most likely prediction for migrant deaths and disappearances in the Mediterranean between 9 June and 11 August 2026 is 277.50, with a 50% chance of falling between 172.50 and 442.50.
• While 2026 began with the deadliest first quarter on record (1,022 deaths), data from April and May suggest a significant deceleration in fatalities and total sea arrivals.
• Summer seasonality typically causes a spike in crossings, but recent trends show total arrivals to Italy have dropped by nearly 50% year-over-year.
• The resolution is strictly tied to the database state on August 11, 2026, meaning reporting lags will likely exclude many incidents from late July and early August.
• The Central Mediterranean remains the highest-risk route, accounting for over 65% of all recorded fatalities in the region during 2026.

Key figures

Historical context

The IOM Missing Migrants Project has tracked Mediterranean mortality since 2014, recording over 28,000 deaths. Historically, the Central Mediterranean route (North Africa to Italy/Malta) is the deadliest migration corridor in the world. Annual totals vary wildly: 2016 saw a peak of 5,136 fatalities, while 2025 recorded 1,873 and 2023 recorded 3,105. Seasonal patterns show that crossings and deaths typically ‘skyrocket’ during the summer months (June to September) due to calmer sea conditions, though 2026 has defied this by having its deadliest start on record in the winter. In Q1 2026, 1,022 deaths were recorded, driven by Storm Harry in January. Historically, a two-month summer window in a ‘normal’ year might see between 300 and 600 deaths, but the specific 2026 context of lower arrivals and reporting lags suggests a more conservative listed total.

Tailwinds

• Summer seasonality (June-August) traditionally brings a peak in departure volumes and higher potential for large-scale shipwreck events.
• Tactical shifts by smugglers, such as using longer, more dangerous routes to avoid detection, increase the lethality of each crossing.
• Conflict in Sudan and high youth unemployment in the MENA region continue to drive migration pressure despite tougher enforcement policies.

Headwinds

• Significant reporting and verification lags in the IOM database mean that deaths occurring in the final weeks of the window (late July to August 11) may not be recorded by the deadline.
• Total sea arrivals in Italy for the first five months of 2026 (11,630) are down approximately 50% compared to the same period in 2025 (22,971).
• Increased border controls and the detention of NGO search-and-rescue vessels (e.g., Sea Watch 5, Aurora) may lead to fewer witnessed incidents, potentially causing more ‘invisible shipwrecks’ that never enter the official count.

Detailed reasoning

My analysis is based on three primary pillars: 2026 year-to-date mortality trends, seasonal migration patterns, and the specific ‘database snapshot’ constraint of the resolution criteria.

In the first five months of 2026, the IOM recorded 1,239 deaths and disappearances. However, this figure is heavily skewed by an exceptionally lethal January, where Storm Harry contributed to 430 deaths in a single month. If we exclude this outlier, the monthly average for February through May drops to approximately 202 deaths per month. More importantly, data from April and May 2026 shows a further decrease, with only 217 recorded deaths across both months (roughly 108 per month). This aligns with data showing that sea arrivals to Italy were down nearly 50% year-over-year compared to early 2025. While summer (June–August) typically sees improved weather and a seasonal spike in crossing attempts, the recent downward trend in successful arrivals suggests that the total volume of traffic—and thus the potential for mass-casualty events—is lower than in peak historical years like 2016 or 2023.

A critical factor in this forecast is the ‘listing date’ of August 11, 2026. The IOM Missing Migrants Project relies on a rigorous verification process involving survivor testimonies, NGO reports, and official coast guard data. Historically, there is a significant lag—often several weeks or even months—between an incident occurring and its full casualty count being finalized in the database. Because the resolution depends on the database state at the exact moment the period ends, incidents occurring in late July and early August are highly likely to be undercounted or missing entirely from the record on August 11.

My most likely prediction of 277.50 reflects a 64-day interval (approximately 2.1 months). At the recent 2026 rate of ~108 deaths per month, the baseline would be around 227. I have adjusted this slightly upward to account for the summer seasonal increase, while keeping the 50% HDI between 172.50 and 442.50 to reflect the uncertainty of reporting delays. The distribution is right-skewed to account for the possibility of a single mass-casualty shipwreck (50+ victims), which can occur suddenly and be logged relatively quickly if there are many survivors or recovered bodies.

Key uncertainties

• The frequency of mass-casualty incidents (50+ deaths) is highly unpredictable and can cause the total to jump significantly in a single day.
• Sudden shifts in search-and-rescue (SAR) capacity, such as the release or detention of NGO ships, directly impact how many incidents are witnessed and recorded.
• Weather anomalies (unusually severe summer storms) could replicate the lethal conditions seen during the January 2026 storm surge.

Conclusion

• The most likely outcome for the period between 9 June and 11 August 2026 is 277.50 deaths/disappearances, reflecting a baseline of ongoing maritime risks tempered by a significant database reporting lag.
• Decision-makers should prioritize monitoring Central Mediterranean route developments and humanitarian search-and-rescue capacity, as these factors remain the most volatile predictors of recorded mortality totals.

Question: how many dead/missing will the IOM Missing Migrants Project list for Mediterranean incidents dated 2026-06-09 through 2026-08-11, as recorded in the IOM ArcGIS/CSV snapshot by 2026-08-11 23:59 UTC. Forecasters agree the resolution is an administrative IOM count (not a true mortality total) and that the most relevant drivers are the elevated Jan–May 2026 baseline (1,239 Mediterranean dead/missing to date), strong Libya/Central-Mediterranean departure pressure, rising Libya→Crete eastern-route risk, and the potential for one or more summer mass-casualty shipwrecks. Because historical 64-day summer windows cluster in the mid‑400s but 2026 is already high, teams center the forecast higher—roughly mid‑400s to mid‑600s (many centering around ~550–650) while keeping a wide right tail for a Pylos-scale disaster; all teams assign only a small probability to totals >1,500. A key shared caveat is reporting/verification lag: incidents in late July–early August may be undercounted in the Aug‑11 snapshot unless highly visible, which pulls the listed-by‑Aug‑11 number down relative to true incident‑date mortality. Recommended updates are to monitor IOM MMP rows (ArcGIS/CSV), UNHCR/IOM operational updates, and reputable press/NGO reports for quickly entered mass-casualty events.

Forecast and Reasoning

Forecast: Probability distribution:

• 10.00% chance of value below 248.065684
• 20.00% chance of value below 358.739391
• 40.00% chance of value below 540.005462
• 60.00% chance of value below 741.759742
• 80.00% chance of value below 1004.007697
• 90.00% chance of value below 1205.716262

Forecast Reasoning

We are forecasting the total (Dead + Minimum Missing) in IOM Missing Migrants Project records with Incident Date in 2026-06-09 to 2026-08-11 (64 days) and Region=“Mediterranean”, as the database appears at 2026-08-11 23:59 UTC. Key drivers:

1. Baseline intensity in 2026: All analysts cite multiple credible reports indicating 2026 began at an unusually deadly pace (hundreds by Feb and ~1,000 by early April). If that elevated level persists even partially into summer, a 64-day window plausibly lands in the mid-hundreds.

2. Seasonality vs early-year spike: Summer typically has more crossings, but early 2026’s high toll may have been partly driven by unusually deadly winter/early-spring conditions. Net effect is ambiguous; I treat summer as modestly higher than an annual average, but not necessarily dramatically higher than the early-2026 realized daily rate.

3. Recording lag at the snapshot date: Because resolution is based on the database state on Aug 11, some late-window incidents (especially “invisible shipwrecks” and cases requiring verification) may not yet be entered, pulling the recorded total down versus the eventual retroactive count. This argues against the very highest medians.

Synthesis: Forecaster 2’s distribution appears somewhat high in the center (median 780) given the snapshot/lag constraint, while Forecaster 1’s median (500) seems low given the documented high 2026 baseline. Forecasters 3 and 4 (median ~650) best balance the elevated 2026 rate with reporting lag and stochasticity. I adopt a median of 650 with a wide right tail to reflect the genuine possibility of one or more mass-casualty incidents being recorded within the window. I also keep a meaningful lower tail in case crossings/deaths drop materially or recording is particularly delayed.

A typical amateur approach would either (a) pro-rate a recent annual total evenly across the year (missing both 2026’s elevated baseline and summer dynamics) or (b) extrapolate the early-2026 daily rate mechanically (missing the Aug 11 database-snapshot lag and regression from an early-year spike). This forecast improves by explicitly combining: (i) evidence of a higher 2026 baseline, (ii) seasonality ambiguity (summer higher crossings but early-2026 unusually deadly), (iii) the mechanical undercount risk from the strict snapshot date, and (iv) heavy-tailed incident risk. Confidence is moderate: the center is guided by multiple consistent signals, but discrete shipwrecks and reporting delays create irreducible variance.

Forecast rationale (numeric):

— Iteration 1 — Across the forecasts, the main drivers are:

• Historical summer patterns: All three anchor on recent Mediterranean summer counts and treat the 64-day window as a period that usually produces several hundred deaths/disappearances.
• 2026 pace appears elevated: Two forecasts explicitly note that early-2026 IOM reporting is running above 2025, which pushes the estimate upward relative to a simple historical average.
• Seasonality matters: There is broad agreement that summer activity increases crossings and fatalities, though the size of that seasonal uplift is debated.
• Snapshot/reporting lag lowers the count: All forecasts reduce their raw seasonal estimate because the figure is taken from the 11 August 2026 snapshot, so incidents near the end of July and in August may not yet be fully recorded.
• Heavy right tail: Each rationale emphasizes the risk of mass-casualty shipwrecks, which could add hundreds of cases in a short time. This leads to very wide upper tails and a strongly right-skewed distribution.

Areas of consensus

• The expected result is well above zero and likely in the mid-hundreds or higher.
• The distribution should be asymmetric, with a modest lower end but a potentially very large upper end.
• Reporting delay is an important correction, preventing the estimate from simply tracking the raw summer base rate.

Main differences

• The forecasts differ mainly on how strongly to weight seasonality versus reporting lag:
  • One model leans higher, putting more emphasis on the elevated Q1 2026 base rate and summer effects.
  • Another is more conservative, grounding the estimate in recent 2-month historical totals and pulling it down for lag.
  • A third sits between these, upwardly adjusting for 2026 activity but still trimming for late-recorded incidents.

Overall synthesis

Taken together, the reasoning points to a mid-hundreds estimate with substantial uncertainty, driven by the interaction of recent elevated migration pressure, summer seasonality, and incomplete same-day database capture, with the possibility of a much higher outcome if a major shipwreck occurs.

— Iteration 2 — Across the forecasts, the core reasoning is highly consistent:

• Elevated 2026 baseline: All models start from the fact that Mediterranean migrant deaths in 2026 were already unusually high by early spring, with roughly 1,000 deaths reported by early April. This pushes the expected summer period above a normal-year average.
• Seasonality matters: The target window, 9 June to 11 August, falls in peak summer crossing season, so the period is treated as disproportionately dangerous relative to its share of the year.
• Reporting lag lowers the snapshot count: Because the question asks for what is listed on 11 August 2026, the forecasts all discount the total somewhat for delayed reporting, especially for incidents occurring in the final days of the window.
• Central estimate in the mid-600s: Despite different calibration methods, the models converge on a median around 650–660 deaths and disappearances for the period.
• Heavy right tail / rare-event risk: The biggest uncertainty is the chance of one or more mass-casualty shipwrecks, which can quickly add hundreds of cases and push totals well above the center.
• Wide uncertainty band: Lower outcomes are possible if summer traffic is quieter or interdiction is effective, while much higher outcomes are possible if catastrophic incidents occur. This produces a broad, right-skewed distribution.

Main areas of agreement

• The period should be well above a simple pro-rata share of annual deaths because it is a high-risk season.
• The most plausible outcome is around the low-to-mid 600s.
• The distribution is asymmetric, with much more upside risk than downside risk.

Main differences

• The models vary in how aggressively they extrapolate from early-2026 totals to a full-year figure.
• They differ somewhat in how much they discount for August 11 reporting lag and how much weight they assign to catastrophic shipwreck scenarios.

Overall, the collective reasoning points to a summer peak, an elevated 2026 baseline, and substantial uncertainty driven by lag and rare but severe maritime disasters.

— Iteration 3 — The forecasts converge on a mid-hundreds estimate, with the main reasoning built from three factors:

1. Seasonal baseline

   • The June 9 to August 11 window falls in the Mediterranean summer peak, when crossings and fatalities are typically elevated.
   • A simple historical-seasonality approach implies roughly 600–700 deaths/disappearances for this period.

2. 2026 appears above average

   • Early-2026 reported totals are running high relative to recent years, suggesting this may be an elevated-mortality year.
   • That pushes the expected value upward somewhat, with some reasoning centering the outcome in the mid-600s to low-700s rather than the low-600s.

3. Reporting lag and cutoff effects

   • Because the question asks for the figure as listed on August 11, late-period incidents may not yet be fully verified and entered into the Missing Migrants Project database.
   • This creates a downward adjustment versus the “true” event count, pulling the estimate back from a raw seasonal projection.

Consensus

• The most likely outcome is around the middle of the distribution, roughly 550–700.
• There is broad agreement that the number should be higher than a simple historical average-only estimate, but not dramatically so once reporting delays are considered.

Main uncertainty

• The distribution is treated as highly right-skewed because a single mass-casualty shipwreck can add hundreds of deaths/disappearances at once.
• Lower outcomes are possible if crossings fall, rescues improve, or reporting is delayed; very high outcomes remain plausible due to catastrophic incidents.

Overall, the shared logic is: historical summer seasonality + elevated 2026 pace, tempered by database verification lag, with large tail risk from rare major shipwrecks.

Here is a synthesis of the shared reasoning across the ten rationales:

Core Dynamics Forecasters agree that estimating the final recorded number of migrant deaths requires balancing the extreme, escalating lethality of Mediterranean crossings in 2026 against severe administrative reporting lags that will heavily suppress the officially published data by the resolution date.

Heightened Lethality and Shifting Routes The rationales highlight a stark paradox defining migration in early 2026: while overall successful crossings into Europe sharply declined due to aggressive EU-backed border crackdowns in North Africa, the mortality rate surged. Forecasters attribute this baseline lethality to smugglers utilizing highly unseaworthy makeshift “iron boats” and shifting to longer, far more dangerous deep-sea routes—specifically the Tobruk-to-Crete corridor. Furthermore, state restrictions on NGO search-and-rescue vessels (such as Italy’s Piantedosi Decree) have dramatically reduced rescue capacities. Moving into the summer window, forecasters expect calm seas and marine heatwaves to incentivize a surge in departures, further elevating the actual number of fatalities.

The Cutoff Constraint and Reporting Lags (Downward Pressure) The most critical factor identified across all rationales is the strict August 11 resolution cutoff. Because the question resolves based on the database state on the exact final day of the forecast window, there is a massive truncation effect. The IOM’s Missing Migrants Project requires rigorous verification to log fatalities, a process hindered in 2026 by reported IOM budget cuts, field staff reductions, and a “strategy of silence” from coastal states withholding rescue logs. Consequently, forecasters agree that a vast majority of the “invisible shipwrecks” occurring in late July and early August simply will not be processed in time for the snapshot. This reporting lag acts as a massive downward pressure, meaning the recorded deaths will be a fraction of the actual deaths for the period.

Mass Casualty Events (Upward Tail Risk) To account for variance, all rationales emphasize the heavily right-skewed nature of maritime disasters. While the baseline accumulation of recorded deaths is expected to be suppressed by administrative delays, the forecast distribution features a long upper tail to account for the persistent threat of a “mega-shipwreck.” If a massive, overcrowded fishing trawler capsizes early in the summer window, the resulting mass-casualty event would immediately attract global media and NGO attention. Such high-profile tragedies bypass standard bureaucratic delays and would drastically spike the total recorded count well before the August cutoff.

Forecast — Post 413

Modal: 375-563 (25.0%) • partial_shift • 6m20s

Interpretation

Reading: strict

The resolution criteria are highly specific: exact field names (‘Number of Dead’, ‘Minimum Estimated Number of Missing’, ‘Incident Date’, ‘Region’), a precise date range, a named database snapshot time (2026-08-11 23:59 UTC), and a detailed geographic tie-breaker for ambiguous records. This leaves little room for interpretive latitude, supporting a strict reading.

Time bounds

• observation_window: Incidents with ‘Incident Date’ between 2026-06-09 and 2026-08-11 inclusive, as recorded in the IOM database at 2026-08-11 23:59 UTC
• resolution_date: 2026-08-11

Edge cases identified

1. Records added to the database after 2026-08-11 23:59 UTC that describe incidents within the window are excluded even if the incident date qualifies — only the snapshot at the cutoff time counts.
2. Records revised after the cutoff (e.g., updated casualty counts) are also excluded; the resolver uses the value as it appeared at 2026-08-11 23:59 UTC.
3. The Canary Islands / Atlantic route is explicitly classified as ‘Western Africa’ rather than ‘Mediterranean’ — incidents on those routes are excluded even if geographically proximate to the Mediterranean.
4. An incident straddling the Cape Trafalgar–Cape Spartel line with ambiguous region labeling could be classified in or out depending on exact coordinate interpretation.
5. IOM data entry lags mean some incidents occurring during the window may not yet be recorded by the cutoff date, causing an undercount relative to eventual totals.
6. A single large shipwreck with many missing persons can dramatically shift the total; the ‘Minimum Estimated Number of Missing’ field may be conservative relative to later revised estimates.
7. If the database is inaccessible at cutoff, the first available snapshot ‘representing the state as of 2026-08-11 23:59 UTC’ is used — it is unclear how this would be verified if data has since changed.
8. Records where ‘Number of Dead’ or ‘Minimum Estimated Number of Missing’ are blank/null — whether these are treated as zero or excluded is unspecified.

Research (3/3 variants, shared evidence pool)

Total evidence registered (shared pool): ?

Research Brief

Evidence confidence: medium

Scenario 1: Moderate-to-high total: 350–700 deaths/disappearances [high evidence]

Conditions favoring

The arrival-forecast mean (~504) and median (~423) both fall in this range, anchored by the 2026 daily rate of ~8.3/day over 63 days. Summer is peak season for Central Mediterranean crossings, historically producing elevated casualties. The 2026 rate has been elevated relative to 2025. No exceptional single catastrophic event occurs, but the baseline rate persists.

Conditions against

If the elevated early-2026 rate was front-loaded and the summer pace moderates (as suggested by the seasonal pattern of crossings being higher in summer but deaths-per-crossing varying), the actual count could fall below this range. IOM reporting lags could also suppress the recorded count below the true incident count.

Scenario 2: Lower total: below 350 deaths/disappearances [medium evidence]

Conditions favoring

The p25 of the arrival forecast is ~242. If 2026 summer conditions moderate (reduced crossings, improved rescue operations, or policy changes), or if IOM data-entry lag is significant and the snapshot at August 11 captures fewer incidents than actually occurred, the recorded count could fall well below 350. The 2025 full-year total of 2,185 (averaging ~182/month) implies a 63-day window would yield ~383 at average pace — but some months are below average.

Conditions against

The 2026 trajectory has been running at roughly double the early-2025 pace. Multiple large incidents have already been recorded in 2026. Summer is historically the highest-casualty season for the Mediterranean route.

Scenario 3: Very high total: above 700 deaths/disappearances [medium evidence]

Conditions favoring

The arrival forecast p75 is ~679 and p90 is ~981. One or more large shipwreck events (e.g., like the April 7, 2026 event with 180+ feared dead) could push the total well above 700. The 2023 annual total of 3,155 (~8.6/day average) with summer spikes could yield 700+ in a 63-day window. The elevated 2026 pace, if sustained or accelerating, supports higher totals.

Conditions against

A total above 700 requires either multiple large catastrophic events or a sustained very high daily rate. While 2026 has been elevated, the p75 suggests roughly a 25% chance of exceeding ~679 under the modeled distribution.

Scenario 4: Near-zero or very low total: below 100 deaths/disappearances [low evidence]

Conditions favoring

The arrival forecast p5 is ~89. A major policy shift or dramatic reduction in Mediterranean crossings (comparable to or exceeding the 2016–2017 Turkey-EU deal effect on the Eastern Mediterranean), combined with IOM reporting lags, could yield a very low recorded total. 2025 saw a full-year total of only 2,185 despite summer peaks, suggesting some windows can be low.

Conditions against

The 2026 pace has been among the highest since 2014, with 1,246 deaths/disappearances recorded by end-May. A collapse to near-zero in summer would require extraordinary and unprecedented change. June–August is historically the highest-casualty period for the Central Mediterranean route.

Key uncertainties

• Whether the elevated 2026 death rate (~8.3/day through May) will persist, moderate, or accelerate during the June 9–August 11 window
• Whether any single large shipwreck event (100+ deaths) occurs during the resolution window, which would be the dominant driver of variability
• IOM data-entry lag: how many incidents occurring before August 11, 2026 will be recorded in the database by the August 11 23:59 UTC snapshot
• Whether the Mediterranean crossing volumes and death rates in summer 2026 follow seasonal 2025/historical patterns or continue the anomalously elevated 2026 trend
• Policy changes or SAR (search-and-rescue) capacity changes between June and August 2026 that could affect both crossing volumes and fatality rates

Forecast variants (3/3)

Variant 0: inside_view (inside_view_v1)

Inside-View Causal Chain Analysis

Current state → mechanism → outcome:

The IOM Missing Migrants Project has recorded ~1,246 Mediterranean deaths/disappearances through end-May 2026 (source 18), implying a 2026 daily rate of ~8.3/day. The resolution window covers 63 days (June 9–August 11), which falls entirely in the peak Mediterranean summer migration season — historically the highest-casualty period for the Central Mediterranean route.

Causal mechanism:

• Elevated 2026 crossing mortality rate (~8.3/day vs ~6.0/day in 2025) driven by overcrowded boats + restricted rescue operations (source 19)
• Summer seasonality historically amplifies Central Mediterranean crossings and casualties
• Single large shipwreck events (like April 7, 2026: 180+ feared dead, source 17) can dominate any 63-day window
• IOM data-entry lag means some incidents may not appear in the snapshot by August 11, creating undercount risk

Arrival forecast (source 20, HIGH evidence strength): At 8/day with overdispersion k=0.5 (negative binomial, appropriate for bursty shipwreck events):

• p5 = 89, p25 = 242, median = 423, mean = 504, p75 = 679, p90 = 981, p95 = 1,197

Scenario weighting per evidence strength tags:

1. “Moderate-to-high total: 350–700” — tagged HIGH evidence strength (arrival forecast mean/median centered here). This range (bins 3-4: 375-750) should carry the most mass.
2. “Lower total: below 350” — tagged MEDIUM evidence strength. Possible via summer rate moderation or IOM lag.
3. “Very high total: above 700” — tagged MEDIUM evidence strength. Requires one or more major shipwrecks; not implausible given 2026 trajectory.
4. “Near-zero/very low: below 100” — tagged LOW evidence strength. Would require unprecedented disruption.

IOM data-entry lag consideration: The snapshot is taken at August 11 23:59 UTC. For incidents near the end of the window (late July–early August), IOM typically records them within days to weeks. This creates a modest downward pull on the recorded count vs. true incident count.

UNHCR comparison (source 14): July–August 2025 alone saw ~401 Mediterranean deaths/disappearances per UNHCR. Since 2026 is running at roughly 38% higher pace than 2025, a June 9–August 11, 2026 window (64 days including June portion) at comparable summer rates would suggest 500–700 range.

Distribution shape:

• The arrival forecast (HIGH evidence) centers on bins 3-4 (375-750), so I concentrate mass there
• MEDIUM evidence scenarios for both lower and higher tails merit meaningful but smaller allocations
• LOW evidence for near-zero scenarios gets minimal mass
• The overdispersion (k=0.5) creates a heavy right tail, so bins 5-8 merit non-trivial weight
• Bin 0 (below 0) = 0; bin 1 (0-188) = very low given 2026 elevated pace

Bin mapping:

• < 0 (bin_0): 0.00
• 0-188 (bin_1): 0.03 — very unlikely given elevated 2026 pace
• 188-375 (bin_2): 0.12 — possible if lag suppresses recorded total or summer moderates
• 375-563 (bin_3): 0.22 — arrival forecast median/mean both fall here, high evidence
• 563-750 (bin_4): 0.22 — strong continuation of central mass
• 750-938 (bin_5): 0.14 — one major shipwreck could push into this range
• 938-1125 (bin_6): 0.10 — ~p85-p92 of arrival forecast
• 1125-1313 (bin_7): 0.08 — fat tail from overdispersion
• 1313-1500 (bin_8): 0.05 — p95+ territory

• 1500 (bin_9): 0.04 — extreme but non-negligible with bursty data

Variant 1: outside_view (outside_view_v1)

Reference class: 63-day Mediterranean summer windows (counts of IOM-recorded dead + minimum missing for a specific 63‑day period). Base-rate anchor: use the arrival-forecast distribution derived from the recent 2026 run‑rate — median ≈ 423, mean ≈ 504 for the June 9–Aug 11, 2026 window (per source 20). Adjustment: modest upward tilt and fattening of tails because 2026 has run well above 2025 pace (IOM ~1,246 dead/missing through end‑May 2026 → ≈8.3/day; source 18) and because single large shipwrecks (e.g., Apr 7, 2026, ~180+ feared dead; sources 4, 17) can produce heavy right‑tail outcomes. I also widen the distribution slightly to reflect medium confidence in the evidence and known IOM data‑entry lags that can both undercount (if incidents are not entered by the cutoff) and create post‑cutoff revisions (evidence confidence = medium). Reading: strict — I interpret the resolution rules literally (only incidents with Region = ‘Mediterranean’ or tie‑broken by the Cape Trafalgar–Cape Spartel line are included; snapshot at 2026‑08‑11 23:59 UTC). Weighting: the arrival-forecast (source 20) is treated as the strong quantitative signal (high relative evidence strength for the moderate-to-high scenario), so most mass lies in the middle bins (375–750). Probabilities: concentrated in bins that contain the median/mean (bin_3 375–563 and bin_4 563–750) with appreciable probability down to historical-low windows (bin_1 and bin_2) and a nontrivial right tail for one-or-more catastrophic events (bins 5–7+). Specific fact citations used in this rationale: arrival-forecast quantiles (mean=504, median=423, p5=89, p25=242, p75=679, p90=981, p95=1,197) — source 20; IOM 2026 YTD ≈1,246 through end‑May (source 18); April 7, 2026 single event ~180+ feared dead (sources 4, 17). This distribution is the outside‑view trajectory: anchored to historical/quantitative base rates and adjusted only modestly for 2026-specific evidence and uncertainty.

Variant 2: contrarian_v2 (contrarian_v2)

Adversarial reading

The brief’s bulk evidence and high-evidence scenario converge on a moderate-to-high total of 350–700 (centered on the arrival-forecast mean/median of ~423–504) driven by the observed 2026 pace of ~8.3/day persisting through the 63-day summer window.

Counter-evidence found (or absence)

Counter-evidence exists in the brief: the elevated 2026 rate may be front-loaded (early-year conditions), summer crossing patterns and death rates historically vary, IOM data-entry lags will suppress the recorded snapshot below true incidents, and single large shipwrecks introduce burstiness that the negative-binomial model already flags but that could easily undershoot if rescue/SAR conditions improve. The “lower total” scenario receives medium evidence strength.

Distribution implications

This shifts materially more mass into 188–375 (bin_2) than a naive centering on the forecast median would imply, while trimming mass from bin_5+ (above 750) relative to the forecast tail; the high-evidence moderate scenario still anchors the core probability in bin_3, with modest spread to bin_4 to respect burstiness.

Consolidate

Strategy: logit_mean • Drafts: 3/3

Reviewers (3/3)

anchoring → shift_lower (low magnitude, medium confidence) anchoring_v1

The draft rationale anchors the p50 firmly at the arrival-forecast median (~423) and mean (~504), treating the “moderate-to-high total: 350–700” scenario as the high-evidence central case. However, the arrival forecast was computed at a rate of 8/day, which is itself anchored to the 2026 year-to-date pace through end-May. Two systematic biases push this toward optimism about the upper range being the modal outcome: (1) The 2026 elevated pace (~8.3/day) reflects an exceptionally deadly January–May period described as “one of the deadliest starts to a year since 2014” and “more than double” the early-2025 pace — there is regression-to-mean risk if the conditions driving that spike (overcrowded boats, restricted rescue in early 2026) normalize somewhat by summer. (2) The resolution snapshot is at August 11 23:59 UTC, and the brief itself flags IOM data-entry lag as a known downward suppressor — incidents near the end of the window may not yet appear — yet the draft “only qualitatively discounts” this (“creating a modest downward pull”) without shifting the median lower. A flat-prior on the rate would triangulate between the 2026 pace (~8.3/day) and the 2025 baseline (~6.0/day), suggesting ~7/day × 63 days ≈ 441 true incidents, but with the lag discount the recorded snapshot could plausibly center closer to 350–420. The draft places its modal mass in 375–563 (bin_3) which is reasonable, but the secondary mass in 188–375 (bin_2) at ~22% feels slightly under-weighted given the lag and potential rate regression, while bins 5–6 at ~16% combined may be slightly over-weighted relative to modal expectation.

Flagged concerns

• Rate anchoring at 2026 YTD elevated pace: The arrival forecast uses 8/day as the base rate, anchored to the 2026 YTD pace of ~8.3/day. This rate reflects an exceptionally deadly early-2026 period (described as ‘one of the deadliest starts to a year since 2014’), and a flat prior over plausible summer rates (regressing toward the 2025 level of ~6.0/day) would imply a somewhat lower modal outcome. The draft acknowledges regression risk but does not shift the forecast median downward to reflect it.
• IOM data-entry lag not fully discounted in central estimate: The brief explicitly flags that ‘IOM data-entry lag: how many incidents occurring before August 11, 2026 will be recorded in the database by the August 11 23:59 UTC snapshot’ is a key uncertainty. The draft calls this ‘a modest downward pull’ and qualitatively discounts it, but the modal scenario (median ~423) is not adjusted. For a fixed-window snapshot question, this lag creates a systematic undercount that should shift the modal estimate lower, not just widen the lower tail.
• Best-case 2026-pace persistence treated as modal: Placing peak mass at the arrival-forecast median derived from the current 2026 elevated pace essentially treats the best-case scenario (the elevated 2026 pace persisting through summer) as the modal scenario. The arrival-forecast median would be ~63 × 6.0 = 378 if the rate reverts to 2025 levels, suggesting the modal expectation may be closer to bin_2/bin_3 boundary (~350–450) rather than solidly in bin_3.

ceiling → shift_lower (low magnitude, medium confidence) ceiling_v1

The key structural constraint the forecaster partially addresses but doesn’t fully integrate is the data-entry lag ceiling: the resolution snapshot is taken at 2026-08-11 23:59 UTC, meaning incidents that occurred during the window but not yet recorded by IOM will be excluded. The brief explicitly notes “IOM data entry lags mean some incidents occurring during the window may not yet be recorded by the cutoff date, causing an undercount relative to eventual totals.” Given the window ends on August 11 (the same day as the snapshot), incidents from the final ~1–2 weeks (late July through August 11) are particularly vulnerable to lag-induced exclusion. This is a structural downward pressure that the forecaster acknowledges qualitatively but does not fully translate into a lower effective mean. At 8.3/day with, say, 1–2 weeks of lag suppression (~10-14 days × 8.3 = ~83-116 events excluded), the effective recorded count would be closer to 63-day total minus ~100, pulling the distribution left. The draft’s modal mass in bins 3-4 (375-750) is appropriate, but the right tail above 750 (bins 5-9 sum to ~13.3%) may be slightly overweighted given the lag constraint acts as a soft ceiling on the snapshot count. There is no hard numeric ceiling being violated — the question has no consensus target or watermark — so this is a moderate rather than severe concern.

Flagged concerns

• Data-entry lag soft ceiling: The snapshot is taken at 2026-08-11 23:59 UTC — the same day the window closes. The brief explicitly states IOM entry lags mean incidents occurring near the window end will not appear in the snapshot. At ~8.3/day, a 10–14 day lag suppresses ~83–116 events from the recorded total vs. the true incident count. This acts as a soft structural ceiling on the snapshot count, pulling effective mass leftward. The draft’s right tail (bins 5-9, summing ~13.3%) is modestly overweighted relative to this constraint.
• No hard upper constraint violated: There is no stated consensus target, watermark, or discrete cap that would impose a hard ceiling. The bin structure extends to ‘Above 1500’, which is theoretically reachable but structurally unlikely given annual totals of 2,185–3,155 spread across 365 days — a 63-day window at peak rates could in theory yield 500+ at base rate, so no arithmetic ceiling is breached.
• Lag-adjusted effective daily rate: The 8.3/day rate is derived from cumulative data through May 30, where older incidents have had time to be entered. For the resolution window (June 9 – Aug 11), incidents from the final 1-2 weeks will likely be underrepresented in the August 11 snapshot. The effective recorded rate for the full 63-day window is likely closer to 6.5–7.5/day rather than 8.3/day, shifting the distribution mean down by roughly 50–100 units and warranting slightly more mass in bins 2-3 vs. bins 4-5.

math → redistribute (medium magnitude, medium confidence) math_v1

The draft distribution exhibits a significant internal inconsistency. The rationale cites the arrival forecast (source 20) with mean=504 and median=423, placing most mass in bins 3–4 (375–750); however, the published bins use different labels than the forecast’s stated percentile values. Critically, the probabilities sum to 1.0088 rather than ~1.0, violating the closure constraint. When normalized, the distribution allocates only ~26.5% to bin_3 and ~19.3% to bin_4, which together comprise ~46% of the mass — this is below what the “high evidence strength” arrival forecast (centered on median=423, mean=504) should anchor. The draft’s heavy allocation to 188–375, 22% (bin_2) reflects the adversarial reviewer’s concern about front-loading and IOM lag, but this pull is not calibrated against the stated “HIGH” evidence for the central scenario.

Flagged concerns

• Probability mass does not sum to 1.0: Published probabilities sum to 1.0088 (excess of 0.0088), violating the ±0.01 tolerance. Bins: 0.00106 + 0.08828 + 0.21956 + 0.26475 + 0.19294 + 0.10284 + 0.05687 + 0.04280 + 0.02301 + 0.00791 = 1.00792 (after accounting for label mapping).
• Distribution shape inconsistent with stated evidence strength: The rationale identifies the arrival forecast (mean=504, median=423) as ‘HIGH evidence strength’ for a ‘moderate-to-high total: 350–700’ scenario. However, bins 3–4 (375–750, which contain the forecast median/mean) receive combined weight ~0.458 after normalization, while 188–375 (bin_2) receives ~0.218. The heavy tilt toward lower bins suggests the draft is over-weighting the ‘MEDIUM evidence’ lower scenario (below 350) and under-weighting the HIGH evidence scenario. If the arrival forecast’s central tendency (mean~504, median~423) is truly high-evidence, the modal bins should be 3–4 with combined mass >0.50, not ~0.46.
• Bin label mapping ambiguity: The draft uses labels ‘0-188’, ‘188-375’, etc., while the requirements specify exact keys ‘bin_1’, ‘bin_2’, etc. The mapping is recoverable (bin_1=0-188, bin_2=188-375, … bin_9=Above 1500), but this inconsistency introduces transcription risk.
• Tail weight not calibrated to forecasted percentiles: The arrival forecast specifies p75=679, p90=981, p95=1197. The draft allocates only ~0.081 combined to bins 5–7 (750–1313), implying P(>750) ≈ 0.081. However, the arrival forecast with k=0.5 overdispersion should imply P(>679, i.e., >p75) ≈ 0.25 at minimum. The draft’s right tail (~0.141 to bins 5–9) is tighter than the forecast’s dispersion warrants.

Reconcile: partial_shift

Reviews consumed: 3

Two reviewers (anchoring and ceiling) independently suggested shift_lower of low magnitude, both citing regression-to-mean from the elevated 2026 YTD pace (~8.3/day) toward 2025 baseline (~6.0/day) and the IOM data-entry lag that systematically suppresses the August 11 snapshot (incidents in late July/early August unlikely to be entered yet). The math reviewer flagged that the draft sums to ~1.008 (closure violation) and that the right tail is under-weighted relative to the arrival forecast’s stated p75/p90/p95 — these are mechanical issues that must be fixed. I made a partial shift: I lowered the central anchor slightly (mean target ~430-450 instead of 504 to reflect lag suppression and partial regression), shifted modest mass from bin_4 down into bin_2/bin_3, and rebalanced the right tail to better match the overdispersed arrival forecast (P(>750) ~ 0.15 rather than 0.08), since the math reviewer correctly noted the tail was too thin given k=0.5 overdispersion. I rejected the math reviewer’s suggestion to push combined bin_3+bin_4 above 0.50 because the anchoring and ceiling reviewers’ lag/regression concerns push the median lower, not higher — keeping bin_3 as mode but with meaningful bin_2 mass. Final distribution sums to 1.00 exactly.

Rejected critiques

• [math] Combined bin_3+bin_4 should exceed 0.50 to honor ‘HIGH evidence strength’ of arrival forecast central tendency

  • Why rejected: Two other reviewers (anchoring, ceiling) independently argued the arrival forecast’s 8/day rate is itself anchored to the anomalously deadly early-2026 period and should be discounted for regression-to-mean plus IOM data-entry lag. Concentrating >50% in bins 3-4 would ignore those structural downward pressures. I kept bins 3-4 combined at ~0.47 as a compromise.

• [ceiling] Right tail (bins 5-9) overweighted given lag-induced soft ceiling

  • Why rejected: The math reviewer correctly noted the arrival forecast’s overdispersion (k=0.5) with p75=679, p90=981 implies the right tail should be heavier, not lighter. A single large shipwreck during peak summer (precedent: April 7, 2026 with 180+ feared dead) can dominate the window. I held the right tail roughly at the draft level (~0.16 combined for bins 5-9) rather than trimming it as the ceiling reviewer suggested.

Final Distribution

Evidence Sources (23)

1. IOM: 21 Migrants Died Every Day in 2025 — IOM’s Missing Migrants Project recorded at least 2,185 people dead or missing in the Mediterranean in 2025, equating to approximately 182 per month on average.
2. Data | Missing Migrants Project - IOM — IOM Mediterranean deaths and disappearances by year: 2025: 2,185; 2024: 2,573; 2023: 3,155; 2022: 2,411.
3. Mediterranean - Missing Migrants Project - IOM — IOM Missing Migrants Project ‘Mediterranean’ region page exists at https://missingmigrants.iom.int/region/mediterranean and the search snippet shows ‘Total Dead and Missing. 34906.’
4. IOM: Over 180 Feared Dead as Mediterranean Death Toll Nears 1000 in 2026 — As of early April 2026, at least 990 deaths and disappearances had been recorded in the Mediterranean in 2026, described as “one of the deadliest starts to a year since 2014.”
5. untitled — In the first two months of 2026 (January–February), at least 655 migrants died or went missing in the Mediterranean — more than double the 287 recorded in early 2025 — despite a 52% drop in irregular crossings per Frontex.
6. Data | Missing Migrants Project - IOM — IOM data page (as of late March/early April 2026 snapshot) shows Mediterranean dead/missing in 2026 at 831 (with latest incident recorded on 28 March 2026).
7. Download Missing Migrants Project data - IOM — IOM provides a ‘Download Missing Migrants Project data’ page at https://missingmigrants.iom.int/downloads offering direct access to their dataset (CSV/GeoJSON) and indicates structured data is available for analysis.
8. Mediterranean | Missing Migrants Project - IOM — The IOM Mediterranean region page showed “The latest incident recorded occurred on 30 May 2026” with a cumulative total of 34,906 Missing Migrants recorded in the Mediterranean since 2014, as of the page access date near June 2026.
9. Missing Migrants Project Data | Humanitarian Dataset | HDX — IOM’s Missing Migrants Project offers data downloads and the dataset is also mirrored on HDX (Humanitarian Data Exchange) at https://data.humdata.org/dataset/iom-missing-migrants-project-data.
10. IOM Missing Migrants Project 2025 Annual Global Report - ReliefWeb — ReliefWeb published IOM’s 2025 annual global report on 22 April 2026 stating nearly 8,000 migrants were reported dead or missing worldwide in 2025, and a total since 2014 exceeding 82,000.
11. Situation Europe Sea Arrivals - UNHCR Operational Data Portal — UNHCR’s Europe Sea Arrivals data page showed 585 dead and missing on Mediterranean and West Africa Atlantic routes combined, last updated 24 May 2026 (separate figure from IOM’s MMP total).
12. Deaths of migrants in the Mediterranean Sea 2026 - Statista — Statista reported that as of March 2026, 658 casualties had been recorded in the Mediterranean, with the full-year 2025 total at 2,185.
13. UNHCR Europe Situations Data and Trends August 2025 — UNHCR data from August 2025 report showed cumulative (Jan 1–Aug 31, 2025) Mediterranean dead/missing: Central Mediterranean 845, Eastern Mediterranean 53, Western Mediterranean 395; total ~1,293 for all Mediterranean routes through end-August 2025.
14. UNHCR Europe Situations Data and Trends June 2025 — UNHCR data from June 2025 report showed cumulative (Jan 1–Jun 30, 2025) Mediterranean dead/missing: Central Mediterranean 611, Eastern Mediterranean 40, Western Mediterranean 241; total ~892 for all Mediterranean routes through end-June 2025, implying July–August 2025 saw roughly 401 additional deaths/disappearances.
15. AFP and media coverage March 2026 — Multiple news reports in early 2026 cited IOM data noting a record number of migrant deaths in the Mediterranean in Jan–Feb 2026, with figures like ‘at least 655 migrants died or went missing’ in the first two months.
16. Mediterranean migrant deaths mounting towards 1,000 in 2026: UN - Al Jazeera — AlJazeera reported (April 7, 2026) that about 765 people had died in the Central Mediterranean so far in 2026 (as of early April 2026), and the overall Mediterranean total was at least 990.
17. Over 180 Feared Dead as Mediterranean Death Toll Nears 1000 - IOM (7 Apr 2026) — IOM press release on 7 April 2026 reported ‘Over 180 Feared Dead as Mediterranean Death Toll Nears 1,000 in 2026’ and stated in the Central Mediterranean around 765 people have died this year (2026).
18. IOM Missing Migrants Project Homepage - current data — IOM Missing Migrants Project data (latest incident 30 May 2026) showed Mediterranean dead/missing in 2026 at 1,246 — through approximately end-May 2026 (about 5 months into the year, or ~150 days), implying a 2026 rate of ~8.3 per day.
19. Why the Mediterranean route is becoming more deadly for migrants - InfoMigrants — InfoMigrants (April 23, 2026): At least 1,003 people died in the Mediterranean in 2026 as of ~April 23, 2026 — described as the highest number IOM had recorded since it began keeping records in 2014, driven by overcrowded boats, restrictive border policies, and limits on private rescue operations.
20. untitled — Arrival forecast for Mediterranean migrant deaths/disappearances in the 63-day window June 9–August 11, 2026: using rate of 8/day (based on 2026 pace of ~8.3/day) with overdispersion 0.5 (bursty shipwreck events): mean=504, median=423, p5=89, p25=242, p75=679, p90=981, p95=1,197. Wide distribution reflects large single-incident variability.
21. UNHCR Situation Europe Sea Arrivals — UNHCR data shows that approximately 2,500 migrants died or went missing in the Mediterranean between January 1 and September 24, 2025 (a 9-month cumulative total). The full 2025 UNHCR total was 2,950 for Mediterranean and Atlantic combined.
22. IOM news item Feb 2026 — IOM published a news item on 23 Feb 2026 titled ‘30 Migrants Dead or Missing off Greece in Latest Tragedy in the Mediterranean’ (IOM website referenced in media and UN feeds).
23. Missing Migrants Project – IOM homepage — The IOM Missing Migrants Project database, as of the most recent snapshot (around early June 2026), shows the latest incident recorded occurred on 30 May 2026. The resolution window of June 9–August 11, 2026 therefore falls entirely in the future relative to currently available data.

wooley_rhino • run a3abea36 • $2.86 • 6m20s

The public incident register is revised irregularly and lacks a reliable ‘date added’ marker, so snapshot counts near the cut‑off tend to understate true events.

Seasonal maritime exposure is a slow‑moving structural constraint concentrated on central, eastern, and western corridors, producing persistent variance in losses.

Recorded losses through late spring are elevated relative to recent years, which raises the baseline expectation for the June–August interval.

Contemporaneous indicators of sea arrivals and detections trended lower year‑on‑year in early 2026, pulling the typical summer outcome down absent a mass event.

A single large wreck or mass‑fatality incident can add several hundred cases at once, creating a heavy‑tailed distribution for the window.

The August‑11 snapshot is likely to miss late revisions and unverifiable cases, biasing the recorded total downward relative to eventual revised totals.

Primary unresolved uncertainties are reporting lag, verification capacity, and reduced information sharing by coastal authorities, each able to shift the recorded figure materially.

Overall sensitivity is dominated by tail risk from one large incident, while flow reductions and snapshot immaturity control the central tendency.

Forecast (median): 371.63 migrant deaths and disappearances

• Summer season drives the baseline up: The June 9–Aug 11 period falls squarely in peak crossing months, so the underlying daily death/missing rate is expected to be relatively high.
• 2026 is structurally riskier than 2025: Factors like more dangerous boat conditions, longer routes, and reduced rescue capacity are expected to keep fatality levels elevated even if crossings are somewhat lower.
• Snapshot timing likely misses some cases: Because the total is taken from the database as of 11 Aug 2026, late-July and early-August incidents may be undercounted versus the eventual “final” tally, pulling the recorded number down.
• The main uncertainty is “big accidents”: The outcome is right-skewed—most days add modest numbers, but one major (or delayed-to-enter) incident could push the total much higher.
• Net result: Combining a higher-than-2025 baseline (but excluding an earlier 2026 anomaly outside this window) with the reporting-lag downside yields a median around the high-300s, with a meaningful chance of going well above 400 if a large incident is recorded in time.