Deaths by Suicide — Crown Dependencies & England/Wales

Isle of Man · Guernsey · Jersey · England & Wales | Annual counts, crude rates, and age-standardised rates | Updated July 2026

⚠ Data sources, methodologies, populations and recording standards differ significantly between jurisdictions. Cross-jurisdictional comparisons require care. See Data Governance Notes below.

View:

Range:

Scotland islands:

Check for confidence bars: Orkney Shetland Na h-Eileanan Siar

Scotland island council areas (NRS) use a 5-year rolling ASMR window (vs 3-year for the Crown Dependencies) and are drawn dotted to flag this. NRS states the 5-year window (rather than 3) directly: small numbers mean "the numbers of suicides can fluctuate markedly from year to year, particularly for the smaller Health Board and Council areas." NRS itself publishes this figure labelled by its full year-range (e.g. "2020-2024 average"), not at a single midpoint year — plotting it at the window's midpoint here is this site's own choice, made for visual comparability with the 3-year Crown Dependencies lines. Population denominators are fixed at recent estimates (Orkney ~22,000 · Shetland ~23,000 · Na h-Eileanan Siar ~26,000). None of the three is currently statistically significantly different from the Scottish average (also shown, dotted, using the same 5-year NRS window) — tick the boxes above to show each island's confidence-interval bars. Annual counts and crude rate are also available (Counts / Crude rate tabs above) — NRS does publish genuine single-year counts for these areas (Table 4A); read them with the same small-numbers caution as the rest of this page, since some years fall to 0.

Reading the 3-year rolling figures. Each Crown Dependency point is a 3-year average plotted at the middle year of its window, so a single year's figure appears in three consecutive points and the plotted position sits one year before the year it describes (the point at 2019 is the 2018–2020 average). A single high or low year therefore reads as a three-point step shifted a year earlier, not as a one-year spike. The Isle of Man's recent rise is dominated by 2020 — 24 deaths registered, of which 19 occurred in 2020; the remainder reflect pandemic-era inquest timing, as deaths are counted by year of registration. As the 2020 window passes, the 3-year figure falls back (the 2021–2023 window is ≈13.7 per 100,000). A fuller analysis of what these figures do and do not show is in preparation.

Demographics & labour market · Crown Dependencies · 2019–2023

IoM male rate (2019–23)

23.8

per 100k/yr · year of death

Jersey male rate (2019–23)

11.1

per 100k/yr · est. sex split

Guernsey male rate (2019–23)

19.6

per 100k/yr · estimated

E&W male rate (2019–23)

13.8

per 100k/yr · ONS registration

IoM — sex & age

76% male

58% aged 30–59 · not cross-tabulated

Jersey — sex & age

72% male

78% aged 25–64 · not cross-tabulated

Guernsey — sex & age

55%

male AND aged 20–59 · confirmed cross-tab

E&W — sex & age

75% male

~65% aged 25–64 · ONS est.

Average annual deaths per 100,000 by sex — 2019–23 · all jurisdictions · same scale

Male Female IoM: confirmed counts (Constabulary FOI 3968233), year of death. Jersey: 2019–23 counts confirmed; sex split estimated from 2010–23 FOI aggregate (72% male applied). Guernsey: both counts and sex split estimated from 2005–24 Mortality Trends 2024 data. E&W: ONS registration counts, ~74% male applied. Rates use sex-specific population denominators. Confidence intervals wide for island jurisdictions given small absolute numbers.

The employment anomaly

The Isle of Man government's own suicide prevention strategy audit (GD2023/0019) examined 65 cases from 2016 to 2021. It found that 33% of those who died were unemployed or unable to work — against 8% of the general working-age population, as measured in the 2021 Census. That is a 4.1× overrepresentation.

That 8% baseline is not unemployment in the headline sense. The 2021 Census recorded five reasons for economic inactivity. Two are relevant here:

Category	Count	% of working-age	Source
JSA claimants	~650	~1.2%	Labour Market Report, 2021 avg
ILO unemployed, not claiming JSA	~660	~1.2%	2021 Census Table 3.2 (ILO total ~1,309 minus JSA)
Unable to work — long-term sick or disabled	2,631	~5.0%	2021 Census Table 3.3
Total	~3,940	~7.5% ≈ 8%

The JSA claimant rate cited in press releases — around 0.6% of total population — represents roughly one person in six of that 8% group. The group 4.1× overrepresented in suicide deaths is roughly thirteen times larger than the figure quoted when politicians say the island is at full employment.

Despite this overrepresentation, the available data does not show a clear correlation between the IoM's unemployment rate and its suicide rate over time. The measurement challenges are substantial — the JSA headline undercounts true unemployment by a factor of several, the ASMR series is short and noisy, and the two series move in opposite directions during 2017–2019. The relationship, if it exists, cannot be demonstrated from what is currently published.

Two cautions on the 4.1× figure itself. First, precision: 33% of 65 cases is 21 people, and the 95% confidence interval around that proportion puts the true overrepresentation anywhere from roughly 2.5× to 5.5×. Clearly elevated, but not precisely "4.1". Second, direction: it runs both ways. "Unable to work" includes people whose illness, including mental illness, produced the labour-market exclusion rather than the other way round. The audit statistic establishes that this group carries heavily concentrated risk and is invisible in headline labour figures. It cannot, on its own, separate "job loss drives risk" from "the same illness drives both". Both readings point at the same population, but they imply different interventions (labour-market re-entry support versus health-led case finding), which is why the Mechanisms tab treats them as competing candidates rather than one finding.

IoM — annual suicide deaths (left) · JSA claimant rate % monthly (right) · census ILO snapshots (right) · 2006–2026

Deaths — registration year (left axis) JSA claimant rate % — monthly (right axis) Census ILO unemployed (right axis — census day snapshot)

JSA = monthly claimant count as % of total population (not seasonally adjusted — winter peaks visible). JSA excludes: people whose contribution-based entitlement has run out; people with savings, redundancy pay, or a working partner above the means-test threshold; the previously self-employed (whose NI contributions don't qualify for contribution-based JSA); and anyone who has stopped actively looking. Census ILO markers (●) = ILO-defined unemployed as % of total population on census day: 2011=1.75%, 2016=1.37%, 2021=1.56% — consistently above JSA because census falls in May/June at the seasonal JSA trough. Deaths = annual registration counts, plotted as a continuous line through mid-year (July) values (left axis). No causal relationship is asserted.

⚠ Pre-2019 IoM demographic breakdown not available from the Constabulary. The suicide audit (GD2023/0019) covers 2016–21 (65 cases): 74% male; peak rate 40–44 in men (~46/100k/yr); 59% of deaths in the 35–59 band. FOI recommended to Public Health IoM / General Registry for pre-2019 annual sex/age data.

📋 Demographics data sources & governance

▼ Click to expand

IoM — sex & age

Constabulary FOI ref 3968233 (25 Jul 2024): year of death, 2019–2023. M=50, F=16, total=66. Pre-2019 not held.
Sex rates calculated as deaths / 5 years / sex-specific population (M~42,000, F~43,000) × 100,000.
Suicide audit GD2023/0019: 74% male; 33% unemployed/unable to work vs 8% general population (4.1×); 59% in 35–59 band.

Jersey

FOI 772310468: annual counts confirmed 2019–23 (12+6+6+11+6=41); sex split estimated by applying 2010–23 aggregate (72% male) to 2019–23 counts.
Rate denominator: M~53,000, F~54,000 (2021 census, ~107,000 total).

Guernsey

Mortality Trends 2024 Table 13: 2005–24 aggregate M=85, F=27. 2019–23 count estimated from 3yr avg series (~40 total). Both count and sex split are estimated for 2019–23.
Rate denominator: M~31,000, F~32,000 (~63,000 total).

E&W

ONS 2024: annual counts 2019–23 = 27,554 total. ~74% male applied. Rate denominator M~29.5M, F~30M.
Registration year (not year of death). Standard of proof changed July 2018 (criminal→civil).

Census ILO unemployment snapshots

2011 Census Table 3.2: 1,475 ILO unemployed of 44,609 economically active.
2016 Census Table 3.2: 1,141 ILO unemployed of 42,777 economically active.
2021 Census: 44,875 economically active − 43,566 in employment = 1,309 ILO unemployed.

Sex/age breakdowns for other territories — what exists and what doesn't

Scotland (national) — yes, published annually by NRS. ASMR by sex, 2019–2023 (Table 1): Male 23.7 / 22.1 / 21.6 / 21.2 / 22.4, Female 7.7 / 8.3 / 6.8 / 7.3 / 7.0 per 100k. Simple 5-year average ≈ 22.2/100k male, 7.4/100k female — a ~3:1 male:female ratio, similar to the Crown Dependencies and E&W tiles above.
Orkney, Shetland, Na h-Eileanan Siar — no. NRS does not publish a sex or age breakdown at individual council-area level for this cause of death: annual counts are already in the low single digits per area (see Data Governance Notes), so splitting further by sex would breach small-numbers disclosure control. This is a genuine data ceiling, not an oversight on our part.
Faroe Islands — partial. The Hagstova Føroya causes-of-death export does break suicide deaths down by sex and by age band nationally (see the Faroe_Islands and updated master workbook), but — like the Scotland islands — not by individual settlement/region.

• IoM Constabulary FOI ref 3968233 (25 Jul 2024) · IOM Suicide Prevention Strategy GD2023/0019

• 2011 IoM Census Table 3.2 · 2016 IoM Census Table 3.2 · 2021 IoM Census Part I

• Jersey FOI Request 772310468 · States of Guernsey Mortality Trends 2024

• ONS Suicides in England and Wales 2024 · IoM Labour Market Reports (JSA monthly counts)

Suicide deaths & unemployment — comparative overlays · 2006–

⚠ Unemployment definitions, proxy measures, and temporal resolution differ across jurisdictions. These overlays are exploratory — no causal relationship is asserted. Deaths on left axis, unemployment proxy on right axis throughout.

Isle of Man Deaths reg yr (L) · JSA % pop monthly (R)

Range:

Deaths (left) JSA % pop (right) Census ILO (right) Long-term sick/disabled, 2021 (right)

⚠ JSA covers ~1-in-6 of the unemployment-affected group linked to suicide risk (audit GD2023/0019 found 33% of decedents were unemployed/unable to work vs 8% of working-age population — 4.1× overrepresentation). The JSA series misses most of that group.

England & Wales Deaths reg yr (L) · Claimant Count % 16-64 monthly (R)

Range:

Deaths (left) Claimant Count % 16-64 (right) Census ILO (right) Long-term sick/disabled, UK 2023 (right) UC/2018 breaks

⚠ UC break ~2013: series inflated by Universal Credit rollout. ⚠ 2018 proof standard: civil standard from July 2018 causes structural upward shift in deaths. Try 1992– to see the full recessionary arc and 30-year correlation signal.

Jersey Deaths reg yr (L) · Registered ASW quarterly (R)

Range:

Deaths (left) ASW quarterly count (right) Census ILO count (right)

⚠ ASW covers ~55% of ILO-unemployed (Census 2021 — 1,320 ASW vs 2,058 ILO). Quarterly not monthly. Note the exceptional 2009 count of 26 deaths coinciding precisely with the GFC unemployment peak — striking but statistically uncertain in a small population.

Guernsey 3yr avg deaths/yr (L) · Wholly unemployed % workforce quarterly (R)

Range:

3yr avg deaths/yr (left) Wholly unemployed % workforce (right) Census ILO (right)

⚠ Deaths are 3-year rolling averages — individual-year counts not published. Counter-directional pattern 2015–2021 (unemployment falling, deaths rising) is analytically interesting: may reflect lagged effects, structural economic anxiety, or small-number noise. Quarterly unemployment only.

Scotland & the Scottish islands — a different pairing

These four use 5-year rolling ASMR on the left axis rather than annual death counts — not because NRS withholds individual-year counts for these areas (it publishes them, Table 4A; see the main chart's Counts tab and Data Governance notes), but because a single-digit annual count would be too volatile to read usefully against a percentage-based labour-market axis here. Scotland national uses the same rolling-ASMR basis for direct comparability. The right axis is DWP/ONS claimant count as % of residents aged 16-64 (Nomis NM_162_1) — the same measure used for England & Wales above, not seasonally adjusted. Window runs 2002–2025; claimant count continues to the present, but the pink ASMR line stops at 2022 — NRS's published rolling-ASMR table doesn't yet extend further. Scotland's card also carries a cyan marker for long-term-sick/disabled inactivity (~7.7% of working-age population, most recent published figure).

Scotland (national) ASMR 5yr rolling (L) · Claimant % 16-64 monthly (R)

2002–2025

ASMR /100k, 5yr rolling (left) Claimant count % 16-64 monthly (right) Long-term sick/disabled, 2024-25 (right)

⚠ Different left-axis unit — rate, not count, so don’t compare left-axis height against the four panels above. The cyan marker is long-term-sick/disabled inactivity (~7.7% of working-age population) — it runs above both the UK (~5.8%) and IoM (~5.0%) markers on their own panels, worth reading alongside claimant count since claimant count alone understates true labour-market exclusion just as it does elsewhere on this page.

Orkney ASMR 5yr rolling (L) · Claimant % 16-64 monthly (R)

2002–2025

ASMR /100k, 5yr rolling (left) Claimant count % 16-64 monthly (right)

⚠ Correction: an earlier version of this note wrongly said Orkney avoided an above-average suicide rate — it did not. Claimant count and disability-benefit caseload both run below the Scotland national rate here, but Orkney's suicide rate 2019–2023 was above the Scotland national average — one of five island local authorities (out of six: Orkney, Shetland, Na h-Eileanan Siar, Highland, Argyll & Bute, North Ayrshire) that were, per the Scottish Government's own Scottish Islands Data Overview 2025. Shetland was the sole exception. Worth sitting with: Orkney's labour-market figures look about as good as Shetland's, yet only Shetland avoided the above-average rate.

Shetland ASMR 5yr rolling (L) · Claimant % 16-64 monthly (R)

2002–2025

ASMR /100k, 5yr rolling (left) Claimant count % 16-64 monthly (right)

The one Scottish island council area with a suicide rate below the Scotland national average, 2019–2023 (Scottish Government, Scottish Islands Data Overview 2025) — worth reading alongside its comparatively low claimant count and disability-benefit caseload (~6.0% of population vs Scotland’s ~9.2%).

Na h-Eileanan Siar ASMR 5yr rolling (L) · Claimant % 16-64 monthly (R)

2002–2025

ASMR /100k, 5yr rolling (left) Claimant count % 16-64 monthly (right)

Claimant count has historically run higher here than the other two islands. Disability-benefit caseload (~8.2% of population) sits much closer to the Scotland national rate (~9.2%) than Orkney or Shetland do — the three islands don’t move together on this measure despite being similarly remote.

Isle of Wight & Anglesey — a wider English/Welsh comparison

Two more islands, chosen for the same reason as everything else on this page: real bridge/ferry-defined geography, small enough population that the effect (if real) should show, and — unlike the Scotland council areas above — published annual death counts, so these use the same red deaths / blue unemployment convention as the first four panels, not the pink ASMR workaround. The blue right axis carries two series here: claimant count (solid, DWP/Nomis, same basis as elsewhere) and a dashed modelled unemployment line (ONS local model-based estimate) — the same "claimant count undercounts true unemployment" gap visible on the Crown Dependency panels, but here it's directly measured rather than inferred from a census snapshot. Annual grain throughout (no month/quarter data published at this geography), 2001–2025 where available.

Isle of Wight Deaths annual (L) · Claimant % & modelled unemployment % (R)

2001–2025

Deaths (left) Claimant % 16-64 (right) Modelled unemployment % (right) Long-term sick/disabled, 2021 (right)

England's largest island by population (~140,000). 2020–21 claimant spike (COVID) is the sharpest feature; modelled unemployment (only published to 2023) tracks claimant count closely here, unlike some Crown Dependency panels where the two series diverge. Long-term sick/disabled (5.1% of residents 16+, 2021 Census) sits close to the Isle of Man's own marker on the first panel above.

Anglesey Deaths annual (L) · Claimant % & modelled unemployment % (R)

2001–2025

Deaths (left) Claimant % 16-64 (right) Modelled unemployment % (right) Long-term sick/disabled, 2021 (right)

Welsh island, ~70,000 population, connected to the mainland by two road bridges (Menai & Britannia) rather than a tunnel or ferry. Both claimant count and modelled unemployment run consistently higher than Isle of Wight across the whole series, but long-term sick/disabled (4.75% of residents 16+, 2021 Census) runs very slightly lower — the two labour-exclusion measures don't move together here.

Reading these charts. The first four panels (Crown Dependencies & England/Wales) use the same visual language: red for deaths (left axis), blue for unemployment proxy (right axis), and ● for census ILO snapshots. The axes are scaled independently per panel — do not compare axis heights directly across panels. England & Wales shows the clearest long-run correlation: the 30-year arc from the early-1990s recession through GFC through austerity to 2024 tracks closely between unemployment and deaths, with the UC rollout and 2018 proof-standard changes as documented discontinuities. Jersey's 2009 (26 deaths, GFC peak) is the most striking single observation in the Crown Dependencies — but is based on small absolute numbers. Guernsey's smoothed series makes temporal inference difficult. IoM's JSA series captures only a fraction of the true unemployment-affected group; the 4.1× overrepresentation in suicide deaths documented in GD2023/0019 cannot be tracked from headline claimant data. The four Scotland panels below swap the left axis for 5-year rolling ASMR (rate, not count) since the islands' individual-year death counts aren't published — Shetland's below-average rate and comparatively low claimant count/disability caseload stand out against Orkney and Na h-Eileanan Siar, which don't move together despite similar remoteness. Isle of Wight & Anglesey return to the red-deaths/blue-unemployment convention (they have real annual counts) but add a second, dashed blue line for modelled unemployment alongside claimant count — the gap between the two dashed/solid blue lines is itself a measure of how much claimant count understates true joblessness, the same theme raised by IoM's JSA-vs-ILO comparison above.

📋 Data governance — unemployment overlays

▼ Click to expand

IoM — unemployment series

SOURCE: IoM Labour Market Report April 2026 (historic count table). JSA monthly claimant count as % of total population (~85,000). Not seasonally adjusted.
JSA excludes: contribution-lapsed claimants; means-tested out (savings, working partner); previously self-employed; those who have stopped actively looking.
Census ILO snapshots: 2011=1.75%, 2016=1.37%, 2021=1.56% of total population. Consistently above JSA due to census falling at seasonal JSA trough.
⚠ Coverage gap: The 2021 Census found ~3,940 people unemployed or unable to work (~8% of working-age) — JSA captured only ~650 of these (~1.2%).

E&W — unemployment series

SOURCE: ONS/DWP Claimant Count (BCJE series), % of 16-64 resident population. Monthly, not seasonally adjusted. Nomis dataset NM_162_1.
⚠ UC rollout from 2013: Universal Credit progressively replaced JSA. In-work low-hours UC claimants included, inflating the post-2013 series. Designated experimental statistics.
⚠ Deaths break July 2018: Standard of proof lowered criminal→civil. Significant structural upward shift in registrations.
Census ILO snapshots: 2001=3.3%, 2011=3.8%, 2021=4.7% of economically active population.

Jersey — unemployment series

SOURCE: ESSH registered Actively Seeking Work (ASW) quarterly counts. Statistics Jersey open data.
⚠ Significant undercount: 2021 Census found 2,058 ILO-unemployed but only 1,320 registered ASW (64% coverage). Non-claimants, self-employed, and those with working partners excluded.
Census ILO counts: 2011 ~2,100, 2021 = 2,058 ILO-unemployed (plotted as counts on right axis).
No monthly series; labour market report six-monthly since 2025.

Guernsey — unemployment series

SOURCE: Wholly unemployed as % of workforce. Social Security Dept via Quarterly Population, Employment & Earnings Bulletin. Pre-2015: Labour Market Bulletin (same source).
Denominator: workforce jobs (~31,000–32,000). Count taken on a specific week each quarter.
⚠ 2024 methodology revision: States of Guernsey revised recording methodology following cost-of-services survey. Some downward revision in published figures.
Electronic Census 2021 ILO snapshot: ~1.5% of workforce.
Deaths are 3-year rolling averages plotted at midpoint year — individual annual counts not published due to small population.

Scotland, Orkney, Shetland, Na h-Eileanan Siar — unemployment series

SOURCE: DWP/ONS Claimant Count, % of residents aged 16-64. Monthly, not seasonally adjusted. Nomis dataset NM_162_1 — same source and definition as the England & Wales panel above, queried by local authority / region. Series runs Jan 1992–present; chart window runs 2002–2025; the ASMR line itself stops at 2022 where NRS's published table ends.
⚠ Left axis is ASMR (rate), not deaths (count): chosen for readability against a percentage-based labour-market axis, not because the data is unavailable — NRS does publish genuine single-year death counts for these council areas (Table 4A; see the main chart's Counts tab and Data Governance Notes), it's just a very volatile single-digit series to plot against a smooth percentage line. Scotland national plotted on the same rolling-ASMR basis for direct comparability.
Disability-benefit caseload (Adult Disability Payment + residual PIP/DLA), as a share of total population, snapshot 31 Jan 2026: Scotland ~9.2%, Na h-Eileanan Siar ~8.2%, Orkney ~6.4%, Shetland ~6.0%. Source: Social Security Scotland / House of Commons Library local-authority data. Denominator is total population, not working-age — a rough first-pass figure, not a precise like-for-like rate; ADP eligibility ends at state pension age and island populations skew older, so this likely understates the islands' true rate relative to Scotland's.
Long-term sick/disabled economic inactivity, % of working-age (16-64) population, most recent published: Scotland ~7.7% (Oct 2024–Sep 2025), UK ~5.8% (year ending Sept 2023), IoM ~5.0% (2021 Census), Jersey 3% (2021 Census, different denominator — all adults 16+, not working-age). Not directly chartable at Orkney/Shetland/Na h-Eileanan Siar level — Labour Force Survey sample size is too small for a reason-for-inactivity breakdown below Scotland national.
Scottish Government's own Scottish Islands Data Overview 2025 states suicide rates 2019–2023 were above the Scotland national average in five of six island council areas — Shetland was the exception.

Isle of Wight & Anglesey — unemployment series

SOURCE: annual death counts, DWP/Nomis claimant count (% of residents 16-64), and ONS model-based local unemployment estimates (%), all 2001–2025 where published, from the site's own master workbook (previously compiled, cross-checked this session). Annual grain only — no monthly/quarterly series published at this geography.
Modelled unemployment is ONS's small-area model-based estimate, not survey-measured LFS unemployment — it exists specifically because the Annual Population Survey sample is too thin at local-authority level for a direct estimate. Published only to 2023; treat post-2023 as claimant-count-only.
Isle of Wight is England's largest island by population (~140,000); Anglesey is Wales's largest (~70,000). Neither is a Crown Dependency — both sit inside the UK's normal statistical and welfare system, unlike every other territory on this tab, which is a useful "does the pattern need constitutional distinctiveness, or just remoteness" check.
⚠ Not yet cross-checked against ASMR: these two panels plot raw annual death counts, not age-standardised rates — comparing their left-axis height to the Scotland ASMR panels above, or to each other without accounting for population size, isn't valid. IoW's population is roughly double Anglesey's.

• IoM Labour Market Reports (JSA monthly series) · IoM Constabulary FOI 3968233 · IoM Suicide Prevention Audit GD2023/0019

• ONS Employment in the UK (BCJE claimant count series) · Nomis NM_162_1 · 2001/2011/2021 Census England & Wales

• Statistics Jersey, ESSH ASW quarterly series · FOI 772310468 · Statistics Jersey Census 2011 & 2021

• States of Guernsey Quarterly Population, Employment and Earnings Bulletin · Mortality Trends 2021 and 2024

• Nomis NM_162_1 (Scotland/Orkney/Shetland/Na h-Eileanan Siar claimant count, user-exported) · NRS Probable Suicides, Scotland 2024 Table 3B · Social Security Scotland Adult Disability Payment local authority data (House of Commons Library, user-exported) · gov.scot Labour Market Insights Jan 2026 · gov.scot Scottish Islands Data Overview 2025

• Isle of Wight & Anglesey annual deaths, claimant count and ONS model-based unemployment estimates (site master workbook)

Family courts · divorce & family applications over time

Why this is here. Family disruption and social integration are among the oldest suicide risk variables in the literature (Durkheim, 1897) and sit outside the labour-market story the rest of this site tests, so they are worth checking independently rather than assumed. The specific version tested here is the relational cost of leaving a place: not physical distance, but what someone loses by moving (custody arrangements, schooling, housing, care ties). Family court caseload is a rough, procedurally mediated proxy for relationship breakdown, not a measure of it; a proper test would need comparative family-law and relocation data across legal regimes, which this site doesn't have.

Removed in this revision: an earlier version of this tab also charted lone-parent household prevalence from the 2021/22 Censuses. A single cross-sectional snapshot with no Crown Dependency coverage, no time dimension, and no established link to the outcome couldn't test anything, and invited a family-structure league-table reading this site has no basis for. Taken out rather than caveated.

⚠ Four different measures from four different court systems, plotted as an index against a common base year, 2018 = 100 for every territory, not raw counts — Jersey runs in the low hundreds, Scotland in the thousands, England & Wales in the hundreds of thousands, so a shared count axis would flatten three of the four lines. 2018 is used (rather than each territory's own first available year) so every line crosses the same point in the same year, making it easier to read divergence before/after that point — chosen because 2018 is the earliest year all four territories have real data for. This is about shape, not rate: does the curve rise, fall, or spike in the same years across territories, not which line sits highest. Scotland's figures run on the Scottish fiscal year (April–March) and are plotted at their start year, an approximation of up to ~6 months against the calendar-year figures used everywhere else. The underlying measures aren't identical either — Jersey is divorce & judicial separation petitions filed (combined, not split in the published digest), Isle of Man is divorce applications (old-law and no-fault combined across the 2024 reform), England & Wales is divorce case starts (petitions pre-2022, no-fault applications from 6 April 2022), Scotland is the broader "Family" case category (of which divorce & dissolution is roughly 73-75% in every year it's separately published). 2022 is missing for England & Wales — the year the law changed mid-year and no single official combined-regime total was published. Guernsey has no equivalent published data at all.

Read the amplitudes with scale in mind. Against the shared 2018 base, Isle of Man swings hardest (159% in 2016, down to base, a 150% pandemic peak in 2020, then 131–147% through 2022–2024 before 101% in 2025); Jersey and Scotland move in narrower bands, dip through COVID, and decline gently since; England & Wales barely leaves a ~91–100% band across the whole run. But amplitude is partly a function of size before it says anything about strain. England & Wales aggregates over 100,000 cases a year across nearly 60 million people: random year-to-year variation alone is on the order of ±0.3%, so a flat line is close to guaranteed. Isle of Man's series runs at roughly 150–230 applications a year, where pure count noise is already about ±7%, and a single procedural change, listing backlog, or the 2024 no-fault reform can move the total by tens of cases. IoM's observed swings (±30–60%) do exceed pure count noise, so something real moves year to year, but at these numbers the prime suspects are administrative (reform anticipation, court listing timing, a small advocate pool) before they are social. Two honest conclusions rather than one dramatic one: the four caseload series do not visibly co-move, and this chart, at these sample sizes, could not have detected modest co-movement even if it exists. Nor can it be lined up against the suicide series directly: IoM suicide registrations lag deaths by 12–18 months through inquests, and caseload has its own procedural lags, so even a real relationship would be smeared across two to three years here.

📋 Data governance — family structure

▼ Click to expand

Family court caseload — time series

SOURCE: Isle of Man — Courts & Tribunals Service, General Registry Annual Reports: 2019 edition (Figure 8, "2016-2019" — 234, 216, 147, and a mid-year-projected 178 for 2019 not used, see below), 2020/2021 double-issue edition (Figure 8, "2018-2021" — 147, 195, 221, 177) and 2025 edition (Chart 22, "2021-2025", divorce + no-fault divorce applications combined — 177, 216, 202, 193, 148). Cross-checked at two overlap points: 2018 (147 in both the 2019 and 2020/2021 editions, exact match) and 2021 (177 in both the 2020/2021 and 2025 editions, exact match). The 2019 edition's own 2019 figure (178) was explicitly labelled a projection based on actuals through June 2019 only — the ACTUAL full-year 2019 figure (195), as later confirmed in the 2020/2021 edition, is used in the chart instead.
Jersey — Judicial Greffe, Court Service Annual Reports 2014-2024, Appendix B/C Statistical Digest, "Divorce and judicial separation petitions filed": full run, no gaps (212, 225, 231, 219, 200, 211, 175, 186, 169, 161, 151). Every year 2014-2023 corroborated across two independent report editions; 2019 sourced only from the 2020 report's prior-year comparison column since no standalone 2019 edition was found.
England & Wales — HMCTS Family Court Statistics Quarterly, Table 12/Section 10 (Divorce), calendar-year figures as stated directly in each Oct-Dec bulletin: 2017 109,339 · 2018 118,141 · 2019 116,612 · 2020 111,996 · 2021 107,724 · 2023 110,770 · 2024 108,657 · 2025 109,184 (FCSQ Oct-Dec 2025 bulletin, published 26 March 2026 — the bulletin's own narrative states "109,184 applications under the NFD law in 2025... down 1% compared to 2024"). 2022 excluded: the No-Fault Divorce Act took effect 6 April 2022, splitting the year between old-law petitions (Q1, ~30,154) and new-law applications (6 Apr-31 Dec, 89,123) with no single official combined total published for the full year.
Latest data available as of this update (July 2026): Isle of Man runs to 2025 (actual, from the 2025 annual report edition). England & Wales now also runs to 2025 (actual, FCSQ published 26 March 2026). Jersey stops at 2024 — no 2025 Court Service Annual Report exists yet; the 2024 edition itself wasn't published until 23 September 2025, roughly 9 months after year-end, so a 2025 edition is unlikely before around September 2026. Scotland stops at 2024/25 (full year); the SCTS workbook does carry a partial "2025/26 YTD" figure (8,838) and its own extrapolated "2025/26 projected" figure (~10,606), but neither is a completed actual year, so neither is plotted — consistent with this site's practice of using actuals over projections (see the Isle of Man 2019 note above).
Scotland — combined series: gov.scot "Civil Justice Statistics in Scotland" bulletin series, Table 5 "Family procedure cases initiated," 2010/11-2019/20 (14,496 · 13,679 · 13,862 · 13,853 · 13,605 · 12,892 · 13,250 · 12,652 · 12,989 · 12,251 — figures cross-checked as identical across the 2011-12, 2013-14, 2015-16 and 2019-20 bulletin editions) plus Scottish Courts & Tribunals Service "Courts Data Scotland: Civil" workbook, April 2026 edition, user-supplied, "Family" thematic category, 2019/20-2024/25 (12,250 · 10,546 · 12,314 · 11,445 · 11,120 · 10,885). The two series were cross-checked at their 2019/20 overlap: 12,251 (old) vs 12,250 (new) — treated as continuous. Note this is a decade-long gradual decline pre-COVID (14,496→12,989, 2010/11-2018/19), not a flat pre-pandemic baseline, before the sharp 2020/21 COVID dip.
⚠ Not a matched measure across territories — see the warning above the chart. Index values are calculated against a common base year, 2018=100 for all four territories (chosen as the earliest year all four have real data for), so lines can be compared for timing and shape around that shared pivot — but the underlying measures still differ (see above), so absolute index height still shouldn't be read as meaning the same thing across territories, only within one territory's own line over time.
Guernsey — Royal Court of Guernsey checked directly (not blocked); no published family/matrimonial caseload statistics found, consistent with Guernsey being the weakest-data Crown Dependency throughout this whole project.

• Isle of Man Courts & Tribunals Service, General Registry Annual Reports (2020/2021, 2025 editions) · Jersey Judicial Greffe, Court Service Annual Reports 2014-2024 · HMCTS Family Court Statistics Quarterly, Table 12 · gov.scot Civil Justice Statistics in Scotland bulletins (2011-12 through 2019-20 editions), Table 5 · Scottish Courts & Tribunals Service, Courts Data Scotland: Civil (April 2026 edition)

Mental health · recorded depression prevalence, 2025

⚠ Not suicide data, and not a direct test of "does mental illness predict suicide" — that relationship is close to definitional and not informative on its own. This tests something narrower: does the same island pattern visible in the suicide/labour-market data also show up in recorded mental-health prevalence, or does it diverge? Currently Scotland-context only (Orkney, Shetland, Na h-Eileanan Siar, Scotland national) — Isle of Wight, Anglesey and the Crown Dependencies are not yet covered; England & Wales prescribing data is being sourced separately and will be added if it comes together.

The pattern doesn't run the way a simple story predicts. Shetland has both the highest recorded depression prevalence of the three islands (15.1%, matching the Scotland national rate) and the lowest suicide rate — the one island council area below the Scotland national average 2019–2023 (see Labour Market tab). Orkney is the reverse: the lowest depression prevalence of the three (13.2%) but an above-average suicide rate. If population mental-illness burden were the main driver of the suicide gap, this ordering should run the other way. It doesn't prove burden is irrelevant — recorded prevalence depends on GP diagnosis and registration practice, which can itself differ by area (see mechanisms 5 and 7 on the Mechanisms tab) — but it's a real divergence, not a confirmation, and worth reading that way rather than as evidence either mechanism is settled. One reading consistent with the ordering: recorded prevalence is partly a measure of case-finding, and case-finding is protective — the direction the Gotland GP study points (see mechanism 7 on the Mechanisms tab). ⚠ Worth reading Shetland's side of this cautiously: its 5yr ASMR carries a wide CI (10.4, 4.5–16.3), and its lowest published points (2015-18) are amplified by two consecutive zero-count years (2016, 2017 — see Counts tab) sitting inside four overlapping windows, not four independently quiet years. The below-average ranking itself is externally sourced (Scottish Islands Data Overview 2025), but the depth and flatness of the trough shouldn't be read as more settled than that.

📋 Data governance — mental health

▼ Click to expand

SOURCE: Public Health Scotland, National Therapeutic Indicators dashboard raw data export (2025-12-31 snapshot, user-exported). "Prevalence" sheet, indicator group Mental Health, condition Depression — recorded depression as % of registered patients, by NHS Board, 2025.
⚠ The two prescribing indicators below are not a neutral cross-board comparison. "National Therapeutic Indicators" (NTIs) are an active Public Health Scotland benchmarking programme, not passive monitoring — PHS's own Quality Prescribing Strategy guides explicitly instruct boards to review their data against other boards' and address "unwarranted variation." A mechanism actively pushing boards toward convergence is currently operating on this indicator family (confirmed via PHS/gov.scot documentation), so any snapshot gap between boards may understate the "true" underlying difference by an unknown amount — there's national-average trend data for these indicators, but nothing found quantifying how much the inter-board gap itself has narrowed. Read as context under caution, not as evidence: long-term antidepressant use (adults on antidepressants ≥2 years, as % of all adults on antidepressants) — Orkney 57.2%, Shetland 65.6%, Na h-Eileanan Siar 64.0%, Scotland 60.1%; combination antidepressants (people on 2+ antidepressants per 1,000 registered patients) — Orkney 5.9, Shetland 5.0, Na h-Eileanan Siar 8.1, Scotland 5.7. Not mixed into the main chart.
Geography here is NHS Board, not council area — but Orkney, Shetland and Na h-Eileanan Siar are each their own single-council-area health board, so this lines up exactly with the geography used everywhere else on this site for these three.
Depression prevalence (the measure actually charted above) sits on firmer ground than the prescribing indicators: it's a GP-recorded diagnosis register, and Scotland dropped QOF (the pay-for-performance scheme that would have created a comparable convergence incentive) in 2016 with no equivalent "unwarranted variation" review programme put in its place for diagnosis coding. It still depends on GP diagnostic practice and patients presenting for care, not the true underlying population prevalence of depression — that ordinary caveat still applies — but it isn't subject to the same active management loop as the antidepressant indicators above.
⚠ Not yet covered: Isle of Wight, Anglesey (England/Wales use a different prescribing data system — OpenPrescribing.net — not yet pulled), Isle of Man, Jersey, Guernsey (no comparable published measure found).

• Public Health Scotland, National Therapeutic Indicators dashboard (user-exported raw data, snapshot to 31 Dec 2025) · PHS NTI data visualisation methodology notes · Scottish Government Quality Prescribing Strategy: Antidepressants, guide for improvement 2024–2027 (board convergence/"unwarranted variation" review process) · PHS national data catalogue, QOF (closed 2016)

Why does remoteness correlate with suicide risk? · candidate mechanisms, tested against the evidence on this site

⚠ Nothing on this tab is a finding. These are candidate explanations, generated to be tested against real data — several contradict each other, and the evidence above supports some better than others. No causal relationship is asserted anywhere on this page.

The starting question, and why the simple version doesn't survive contact with the data

The intuitive story is psychological entrapment — it's harder to physically get away from a source of distress if you live somewhere remote or insular. But the territories on this site don't sort the way that simple version predicts — and the comparator set doesn't actually separate cleanly on "physically cut off" in the first place: the Faroe Islands, Gotland, Isle of Man, Jersey and Guernsey are all reachable by ferry/air only with no bridge or tunnel to a mainland, so that criterion alone doesn't distinguish any of them from each other. Gotland is, if anything, the best-connected territory in the whole set (up to 16 ferry sailings and ~4 flights a day, sub-3-hour crossings) — yet it still shows a below-average suicide rate and near-lowest unemployment, alongside the more sparsely-connected but similarly low-unemployment Faroe Islands. Meanwhile the Isle of Man, which has the highest recent 3-year ASMR in the Crown Dependencies/E&W set (a figure inflated by the 2020 window — see the chart note above), has decent transport links and isn't remote in any physical sense either. Isolation-as-distance doesn't explain the pattern on its own — worth reading alongside the Family courts tab's caseload chart, which tests a different version of "distance": not how far away somewhere is, but what it costs someone, relationally, to leave.

1 · Thin labour markets — economic lock-in

The claim: the risk factor isn't being physically stuck on an island — it's being economically stuck. A thin local labour market means losing a job (or being unable to work) means losing access to the entire local market, because relocation is costly. Remoteness matters only insofar as it shrinks the number of employers. (Earlier drafts called this monopsony, but the load-bearing feature is thinness: monopsony proper predicts suppressed wages for people in work, whereas thinness predicts a catastrophic cost of losing work. The second is the suicide-relevant claim, and the two are separately testable.)

Explains well: the IoM audit's 4.1× overrepresentation of unemployed/unable-to-work among suicide decedents (see Demographics tab); the Faroese paradox — near-zero unemployment, near-zero of the "stuck" state; Gotland's isolation-plus-full-employment combination.

Struggles with: Orkney, Shetland and Na h-Eileanan Siar don't sort by employment health the way this predicts — Shetland (best labour-market figures) is the one island below the Scotland average, but Orkney (also decent figures) is not. Its headline evidence is also causally ambiguous: the 4.1× figure includes "unable to work", where illness can produce the labour-market state rather than the reverse (see the caution now attached to that figure on the Demographics tab).

Untested prediction, not yet evidence: this mechanism predicts Faroese regions cut off from Tórshavn by ferry/helicopter (pre-tunnel Suðuroy; pre-2020 Eysturoy/Norðoyar) should show measurably higher unemployment than tunnel-connected regions. Hagstova Føroya publishes a regional monthly unemployment table (StatBank AM02020) that could test this directly, but nobody has pulled the actual pre/post-tunnel figures yet — so this stays a testable prediction on this site, not a confirmed finding, until someone does.

Policy implication: stop targeting headline unemployment — JSA/claimant count captures roughly 1-in-6 of the true unemployed-or-unable-to-work group on the IoM. Build active labour-market and re-entry support around the Census-measured excluded population instead, ideally delivered through health settings (individual placement-and-support style), not the jobcentre.

2 · Prosperity / wealth buffering (largely subsumed into 1 — retained as a tested-and-narrowed candidate)

The claim: islands per se do nothing; suicide tracks economic security in general, and islands just happen to span the full range of it. Faroe and Gotland are rich and fully employed. IoM and the Scottish islands have pockets of real exclusion hiding under prosperous headline statistics.

Explains well: the Faroe/Gotland pattern cleanly, and is broadly consistent with mechanism 1.

Struggles with: the IoM is wealthy on GDP-per-capita terms, and its long-term-sick inactivity rate (~5.0%) is lower than Scotland's (~7.7%) — yet IoM has the highest recent 3-year ASMR in its comparator set. Average prosperity doesn't track the outcome; this mechanism mostly collapses into mechanism 1 once you look at distribution rather than averages.

Policy implication: a distributional audit, not an aggregate one — find who is invisible in the official statistics (the JSA-vs-Census gap on the Labour market tab is the clearest example) and make that population, not the claimant count, the headline the territory's labour statistics report against.

3 · The "wrong-sized settlement" gap

The claim: risk peaks not in the most remote places but in remote small towns — too big for the dense informal mutual support of a village, too small and cut off to sustain proper services or the anonymity-plus-opportunity of a city. A U-shaped settlement-size curve, and islands are risky only when their main settlement sits in the trough.

Explains well: this is the mechanism with the most direct supporting evidence already on this site — Scotland's own urban-rural classification shows "remote small town" areas with a higher ASMR than "remote rural" areas in multiple years (see Data Governance Notes, main chart). Shetland's below-average rate fits if Lerwick, plus strong internal connectivity, functions as a genuinely complete hub rather than an isolated small town.

Struggles with: Tórshavn is also a remote small town by population — unless its role as national capital (full government, hospital, university, media) makes it functionally "complete" despite its size, which is really a refinement of this mechanism (hub completeness, not settlement population) rather than a contradiction of it.

Policy implication: counterintuitive — don't spread services thin across remote areas. Concentrate into one genuinely complete hub per territory (24/7 crisis capacity, psychiatric liaison near A&E, non-clinical "third places" open evenings) and solve transport to that hub, rather than trying to replicate services everywhere.

4 · Out-migration selection

The claim: islands export their most mobile, resourced people. Whoever stays — or returns after failing to launch elsewhere — is selected for fewer options and worse outcomes. Territories with strong economies retain or re-attract people and escape this selection effect; territories with weak economies concentrate it.

Explains well: the Faroese and Gotland patterns read as retention success stories under this framing. Could also explain divergence between Orkney/Shetland/Na h-Eileanan Siar if their demographic pipelines differ — Shetland's oil and renewables economy plausibly retains working-age people better than the other two.

Struggles with: the IoM has net in-migration in many years, which cuts against a pure selection story — though in-migrants are themselves selected for economic attachment (work permits), so a "stuck native" subgroup could coexist with net inward migration. Testing this needs age-structure and cohort data this site doesn't yet have.

Policy implication: track and target the "returner/stayer" cohort specifically — working-age men who never left or returned early, concentrated in non-mobile occupations — with transition support timed to known choke points (end of a fishing or construction season, a redundancy event), rather than population-wide campaigns.

5 · Means, alcohol, and health-system practice

The claim: some of the cross-territory gap is boring and proximate rather than structural — alcohol culture, prescribing practices, means availability, and how different coroners or registrars classify deaths. Faroe has historically restrictive alcohol policy; small jurisdictions may code differently than ONS/NRS.

Explains well: part of the Faroese paradox; potentially part of the Scotland-vs-England gap (Scotland's alcohol-specific death rate runs roughly double E&W's).

Struggles with: the within-jurisdiction findings (mechanisms 1 and 3) can't be classification artefacts, since they compare areas using the same coroner and registration system.

Policy implication: a jurisdiction-specific means/alcohol audit before anything else — for the IoM, extend the GD2023/0019 audit's questions on alcohol involvement, prescribed-drug access and method, then act on whichever is overrepresented. Separately: commission a classification-comparability check between Manx, Channel Island and ONS/NRS coding before trusting any cross-territory league table, including the ones on this site.

6 · Occupational composition

The claim: islands differ in what people do for a living, and occupation itself carries risk. National occupational analyses (ONS's suicide-by-occupation work for England; comparable findings elsewhere) consistently find elevated rates in agriculture, fishing and parts of construction: solitary work, seasonal and precarious income, occupational access to lethal means, and cultures of self-reliance that suppress help-seeking. Island economies are built disproportionately from exactly these occupations, so part of any "island effect" may be compositional rather than geographic.

Explains well: potentially the Orkney/Shetland divergence that defeats mechanism 1. Orkney's economy leans agricultural; Shetland's leans oil, gas and renewables — more salaried, more unionised, more shift-structured. Two islands with similar remoteness and similar headline labour figures can still differ substantially in occupational risk mix.

Struggles with: the Isle of Man, whose economy is dominated by finance, e-gaming and professional services — low-risk occupations on paper — yet carries the highest recent 3-year ASMR in its comparator set. Composition can modulate the island pattern; it can't be the whole story.

Untested prediction, not yet evidence: occupation of decedents. The GD2023/0019 audit could code this from records it already holds; NRS publishes Scotland-level occupational analyses that could be set against the three islands' industry mix. If the island gap shrinks materially after adjusting for occupation, this mechanism is doing real work.

Policy implication: target prevention by occupation, not just geography — the farming and healthcare sectors both have established occupation-specific prevention literatures (means-access protocols, peer outreach timed to seasonal pressure points) that a small jurisdiction can adopt off the shelf.

7 · Detection & treatment contact

The claim: what varies between territories is not the amount of distress but whether distress gets found and treated. Primary-care recognition of depression, waiting times, and willingness to present all differ by place — and in small communities the confidentiality problem is real: the GP's receptionist may be a neighbour, which suppresses presentation exactly where visibility is highest.

This mechanism has the most famous natural experiment in the field, and it happened on one of this site's own comparators. The Gotland study: in 1983–84 Sweden ran a systematic depression-education programme for every GP on Gotland (Rutz and colleagues, published in Acta Psychiatrica Scandinavica). Suicides fell in the years immediately after, with the clearest effect among women — then drifted back as the trained GPs left the island. Gotland's present-day below-average rate should be read with that history in mind: it isn't only a well-connected, fully-employed island; it's arguably the most intensively studied primary-care suicide-prevention site anywhere.

Explains well: the Mental health tab's inversion. Shetland records the highest depression prevalence of the three Scottish islands and has the lowest suicide rate; Orkney records the lowest prevalence and an above-average rate. If recorded prevalence partly measures case-finding rather than burden, that ordering is what a detection mechanism predicts: high recorded prevalence can be a sign the system is catching people. Caveat: Shetland's low point estimate itself is a wide-CI, small-numbers figure whose lowest years are partly a rolling-window artefact (two consecutive zero-count years, see main chart Counts tab) — the inversion is real and worth noting, but shouldn't be leaned on as if either side of it were a precise measurement.

Struggles with: recorded prevalence is ambiguous by construction (see the Mental health tab's governance notes), and the Gotland effect itself faded — a caution against reading any single-period rate as a stable property of a place rather than of its current service configuration.

Untested prediction, not yet evidence: contact-before-death rates. An audit of the GD2023/0019 type can establish what fraction of decedents had recent primary-care or mental-health contact; that figure differs between systems in the wider literature and is the direct test of this mechanism. FOI-able in principle for the IoM.

Policy implication: the one mechanism on this list with a proven, exportable, cheap intervention: systematic GP education on depression recognition, plus confidential access routes that bypass small-community visibility (off-island telehealth, out-of-area appointments).

8 · Entrapment, salvaged — the working synthesis

The original intuition isn't wrong, just imprecise. Entrapment is a psychological state produced when someone hits a life shock — job loss, relationship breakdown, disgrace — in a context with no exit and no anonymity: everyone already knows, and there's nowhere to reconstitute yourself. Islands raise risk only when both conditions hold at once: the shock rate is high (a weak or thin local economy) and the escape valves — alternative jobs, a complete service hub, a feasible relocation — are absent. Faroe and Gotland are islands where those valves work; the Isle of Man and several Scottish islands are places where, for some residents, they don't.

This isn't armchair psychology dressed up, either. Defeat-and-entrapment is the central motivational pathway in the current academic models of suicide — Williams' "cry of pain" and O'Connor's integrated motivational-volitional (IMV) model — which treat entrapment as what generates suicidal ideation, and separate factors (means access, acquired capability, impulsivity) as what governs whether ideation becomes action. Mechanisms 5, 6 and 7 sit on that second stage; this synthesis sits on the first. The two stages need different interventions, which is why "combine them" is a structure, not a fudge.

Working headline: "islands don't trap people; failed local labour markets and incomplete hubs do — islands just make those failures harder to escape."

This is really mechanisms 1 and 3 combined (with 7 as the service-side lever acting on the same pathway), and it's the only entry on this list that survives contact with every finding on this site rather than most of them.

On aggregating small territories. Every territory here has a small-numbers problem for suicide counts specifically — Orkney, Shetland, Na h-Eileanan Siar, Faroe, Gotland and PEI all sit at single digits to low double digits per year. Comparing territories one at a time, as most of the charts on this site necessarily do, is statistically underpowered by design; it's presented as description, not inference. A more defensible approach — not yet built — would pool territory-years by mechanism category rather than by raw geography: group territory-years into strata like {tight labour market / slack}, {complete service hub / incomplete}, {well-connected / ferry-only}, and compare pooled rates across strata using 5-year rolling windows with Poisson-based confidence intervals on the raw counts (Scotland already publishes its island figures this way). That design uses Faroe and Gotland as the natural control cases they are, and would let each island be matched against 2-3 comparable mainland regions on population, age structure and industry mix — rather than compared directly against a large national aggregate it was never a fair match for. One hard rule for any future version of this: never pool raw counts across coding jurisdictions (NRS vs ONS vs Hagstova vs Statistics Canada) without a classification-comparability check first (see mechanism 5) — a pooled trend that's partly a coding artefact would be worse than no pooled trend at all.

And on aggregating the time series into a single overall trend: no — but for a fixable reason. A summed or averaged line across territories is the weakest version of the idea. It would be dominated by whichever series is largest, it would inherit every coding difference the paragraph above warns about, and it answers a question nobody is asking ("what does the average island line look like") instead of the real one ("do suicide, labour-market exclusion, and family disruption move together"). The design that answers the real question is a within-territory panel model: Poisson regression on annual death counts with a log-population offset, a fixed effect per territory, common year effects, dummies for documented breaks (E&W's 2018 proof-standard change), and the economic covariate of interest entered with 0–2 year lags. The territory fixed effects absorb every stable between-territory difference — coroner practice, coding, demography — so identification comes only from whether deaths move with the covariate within each territory over time. Cross-jurisdiction incomparability then drops out of the estimate instead of contaminating a pooled line, though within-territory coding breaks still need their dummies.

Why single territories look uninformative even though the underlying relationship is well established: the aggregate literature (e.g. Nordt and colleagues, Lancet Psychiatry 2015) associates major unemployment shocks with relative rises in suicide on the order of 20–30%, and roughly one in five suicides worldwide with unemployment. On an IoM baseline of about ten deaths a year, a 20–30% effect is two or three extra deaths against random year-to-year noise of roughly ±3 — invisible by construction. Pooled across ten small territories and two decades of territory-years, the same effect becomes detectable. That is the statistical version of the intuition that suicide, divorce and economic wellbeing "should move together with a large enough sample": they probably do, and the sample exists — it's just spread across territories rather than years, which is exactly what a fixed-effects panel is for. Divorce/family caseload can join the same model as a covariate, but only with reform-year dummies of its own: filing counts respond to law changes, fees and court capacity as much as to underlying relationship stress (IoM's 2024 no-fault reform and E&W's 2022 change both move those series for purely administrative reasons — see the Family courts tab).

📋 Data governance — mechanisms

▼ Click to expand

📋 Data Governance Notes

▼ Click to expand

England & Wales — ONS

METRIC (ASMR tab): Age-standardised mortality rate per 100,000, 2013 European Standard Population, ages 10+. Annual single-year figures.
METRIC (counts & crude rate tabs): Annual registration counts from ONS bulletin. Crude rate = count ÷ mid-year population × 100,000.
COVERAGE: 1981–2024. ICD-9 before 1993; ICD-10 from 1993 (X60–X84 + Y10–Y34).
⚠ SERIES BREAK 26 July 2018: Standard of proof lowered from criminal ("beyond all reasonable doubt") to civil ("balance of probabilities"). Significant upward shift in registered deaths. Pre- and post-2018 series are NOT directly comparable. Chart marks this break.
Figures are for year of registration, not year of occurrence. Inquest delays mean registration may lag death by months or years.

Isle of Man — IOM Public Health Directorate

COUNTS: Annual registration counts, ICD-10 X60–X84 + Y10–Y34, from IOM Mortality Report 2022 Figure 28 and FOI responses. Year of registration.
CRUDE RATE: Calculated as count ÷ 85,000 × 100,000. Population is stable at approximately 85,000.
ASMR: 3-year rolling, 2013 European Standard Population, from Figure 29 of IOM Mortality Report 2022 and confirmed FOI responses. Plotted at midpoint year of window.
⚠ SMALL NUMBERS: With ~85,000 population, one death shifts the crude rate by ~1.2/100k. Annual counts range 2–24. Treat single-year figures as highly volatile.
⚠ 2020 SPIKE: Registration-year count was 24; police year-of-death count for 2020 was 19. Inquest delays cause significant registration lag on IOM (12–18 months typical).
ASMR values confirmed from FOI: 2014–16=7.8, 2015–17=10.2, 2016–18=10.7 (FOI 1228481); 2020–22=17.8 (Mortality Report 2022 text); 2021–23=13.9 (Mortality Report 2023, GD2025/0091). Other values read from Figure 29 chart image (±1–2 units).
⚠ COUNT CORRECTIONS (June 2026): 2013 corrected 12→11; 2015 corrected 11→8, both per GD2025/0091 Figure 28.

Guernsey — Public Health Intelligence Unit

COUNTS: Annual figures are NOT published due to small numbers. The count series shown here uses 3-year rolling average annual counts from bar charts in Guernsey Mortality Trends 2021 and 2024. These are averages, not individual-year counts.
CRUDE RATE: Not shown for Guernsey — cannot derive reliable annual crude rates from 3-year averages.
ASMR: 3-year rolling, 2013 European Standard Population. Confirmed: 2013–15=7.0/100k; 2019–21=15.4/100k (Mortality Trends 2021). All other values from published charts.
⚠ 2022–24 ASMR: Two values shown — 9.6/100k (confirmed closed inquests only) and 11.0/100k (maximum if all outstanding inquests recorded as suicide). True value lies between these.
Total deaths 2002–2021: 104. Total 2005–2024: 112. 2021 peak: 11 deaths (highest single year in 20 years). 2024: 6 confirmed deaths. Population ~63,000–64,000; includes Alderney.
Source: Guernsey Mortality Trends 2021 (Aug 2023) and Mortality Trends 2024 (Oct 2025), Public Health Intelligence Unit, States of Guernsey.

Jersey — Health and Care Jersey / Statistics Jersey

COUNTS: Annual counts from Jersey Mortality Statistics reports (Figure 14 in 2020 report; Figure 16 in 2024 report). Definition: intentional self-harm + undetermined intent, consistent with ONS. Year of registration.
CRUDE RATE: Calculated as count ÷ ~105,000 × 100,000. Population grew from ~103,000 (2007) to ~107,000 (2023); 105,000 used as approximation throughout.
ASMR: 3-year rolling, 2013 ESP, from Jersey Mortality Statistics 2024 Figure 17. Plotted at midpoint year of window.
⚠ OUTSTANDING INQUESTS: Jersey reports note that comprehensive data is only available once inquest process is complete. 2023 figures confirmed; some years may be subject to revision.
Age and sex breakdown suppressed under Freedom of Information (Jersey) Law 2011 Art.25 due to small numbers and privacy risk.
Source: Jersey Mortality Statistics 2020 and 2024, Health and Care Jersey / Statistics Jersey.

Scotland island council areas — Orkney, Shetland, Na h-Eileanan Siar (overlay)

SOURCE: National Records of Scotland, Probable Suicides, Scotland 2024 (Tables 4A & 4B), published 16 Sept 2025, updated 28 Oct 2025. Same definition as the Crown Dependencies and E&W: ICD-10 X60–X84 (intentional self-harm) + Y10–Y34 (undetermined intent).
COUNTS: Table 4A — genuine single-year registration counts, Persons, 1974–2024. Not rolling averages.
CRUDE RATE: Calculated from Table 4A counts against fixed recent population estimates (Orkney ~22,000, Shetland ~23,000, Na h-Eileanan Siar ~26,000) — populations held constant across the whole series, as with IoM's crude rate above.
⚠ ASMR WINDOW DIFFERS: Table 4B gives a 5-year rolling ASMR (2013 ESP), not the Crown Dependencies' 3-year window — NRS states this reasoning directly: small numbers mean, in NRS's own words, that suicide counts "can fluctuate markedly from year to year, particularly for the smaller Health Board and Council areas" (NRS Notes, note 5). NRS itself publishes this figure labelled by its full year-range (e.g. "2020-2024 average"), not at a single midpoint year. Plotting it at the window's midpoint here is this site's own choice, for visual comparability with the 3-year Crown Dependencies lines — drawn dotted throughout to flag both the different window length and that the midpoint placement is our construction, not NRS's.
⚠ NOT STATISTICALLY SIGNIFICANT: Checked directly against NRS's own non-overlapping-confidence-interval test (2019–2023 window): none of Orkney (19.5, CI 10.7–28.2), Shetland (10.4, CI 4.5–16.3) or Na h-Eileanan Siar (18.1, CI 10.3–25.9) has a confidence interval sitting entirely above Scotland's (14.1–15.0) — despite Orkney and Na h-Eileanan Siar showing higher point estimates. Only Dundee City, Highland and East Ayrshire are significantly elevated council areas in that window, and none of those three is an island. Error bars on the ASMR view show this directly — this is a genuine complication for a simple "islands are worse" reading, not a reason to discard it: wide CIs from small populations mean a real island effect can't currently be ruled out, but it isn't proven at council-area level either.
NRS's own strongest defensible geographic finding is about "remote small towns" (Urban Rural Classification UR4), not islands specifically — UR4 areas had the highest ASMR of all six UR categories in 2022 and 2023, ahead of "remote rural" (UR6) areas in every year checked. That is arguably a better match for a compounding geographic + social-density/relational entrapment hypothesis than pure isolation would predict.
SCOTLAND (NATIONAL) REFERENCE LINE: plotted from the same Table 4B-equivalent (Table 3B, Health Board summary, "Scotland" row), same 5-year rolling window and 2013 ESP standardisation — so it's directly comparable to the three island lines above, not to the Crown Dependencies' 3-year series. This is the line the islands' confidence intervals are being tested against; its own CI isn't plotted (tick boxes above are scoped to the three islands) but is available in its tooltip.
CONFIDENCE-INTERVAL CHECKBOXES: off by default to keep the ASMR view legible — each island's 95% CI error bars can be switched on individually next to the island toggles. The Scotland (national) line has no CI checkbox of its own (it's the reference line the three islands are tested against) but its interval is available in its tooltip.
COUNTS / CRUDE RATE NOW SHOWN: the three islands and Scotland (national) appear on the Counts and Crude rate tabs too, drawn from the same Table 4A genuine annual counts noted above — not derived from the 5-year ASMR. Read them with real caution: at these populations (~22-26k) a single-digit annual count is expected to swing sharply, and some years are genuinely 0 (e.g. Shetland 2016 and 2017) — that is normal Poisson variation at this population size, not a data error or a sign anything unusual happened.

⚖ Comparability across jurisdictions

The 3-year rolling ASMR series (all using 2013 ESP) are the most directly comparable metric across IoM, Guernsey, Jersey, and E&W. Even so, differences in coroner practice and inquest timelines mean like-for-like comparison requires caution.
Annual counts and crude rates are available for IoM and Jersey — these are honest about volatility. Guernsey publishes only 3-year averages for this cause due to small numbers.
E&W ASMR series has a methodological break in 2018 (standard of proof change) that does not apply to the Crown Dependencies, which follow their own coroner systems.
All jurisdictions include undetermined intent (Y10–Y34) alongside confirmed suicide (X60–X84), which is standard practice and broadly consistent across series.

Sources

• ONS, Suicides in England and Wales: 2024 registrations — ons.gov.uk

• IOM Public Health Directorate, Mortality Report 2022 (GD 2024/0073), Figures 28 & 29

• IOM Cabinet Office FOI responses: 1228481 (2020), 4081089 (2024)

• Guernsey Public Health Intelligence Unit, Mortality Trends 2021 (Aug 2023) and Mortality Trends 2024 (Oct 2025) — gov.gg/publichealthinfo

• Jersey Mortality Statistics 2020 and 2024, Health and Care Jersey / Statistics Jersey — gov.je

• National Records of Scotland, Probable Suicides, Scotland 2024 (Tables 4A & 4B, 6), published 16 Sept 2025, updated 28 Oct 2025 — nrscotland.gov.uk

• A full data file with sources per row is available: suicide-data.csv

Compiled June 2026, revised July 2026. Values read from published charts carry visual-estimation uncertainty of ±1–2 per 100,000.