Methodology

Position

This is an independent data project, built from publications by SGK (Statistical Yearbooks), the İSİG Council (press record) and Eurostat (ESAW); it has no ties to these institutions. It does not speak for them: a difference between two series is labelled as the trace of differences in scope and method, never as a claim such as "unrecorded work"; the site analyses, it does not judge. Every figure can be traced from source → raw file (SHA-256) → script → panel (see Verification); open rows are not hidden (see Limitations).

What it shows

• Map: for the selected year (2013–2025), the number of work accidents, deaths from work accidents and occupational diseases in each province.
• Rate: the same figures per 100,000 insured workers.
• Right panel: SGK does not publish the sector (NACE; Rev. 2.1 from 2024) breakdown by province. The panel shows the national sector breakdown for the selected year and category. When a province is selected, the sector breakdown of its insured workers is shown as well. These are not accident data.
• Scope: only insured persons under Article 4/1-a of Act 5510 (workers employed under a service contract). 4/1-b (self-employed) and 4/1-c (public servants) are not included in the map. 4/1-b is shown separately in the province detail (see 4/1-b section).

Sources

All data come from the Statistical Yearbooks of the Social Security Institution (SGK), from the original Excel (xlsx) files for 2013–2025. SGK publishes the yearbooks at sgk.gov.tr/Istatistik/Yillik.

• Section 1 — Insured and Work Place Statistics: insured persons by province (denominator) and insured persons by province × sector.
• Section 3 — Work Accident and Occupational Disease Statistics: cases and deaths by province and by sector (numerator). From 2017, 4/1-a (Section 3-1) and 4/1-b (Section 3-2) are separate files.

According to SGK metadata, the data are produced by classifying administrative records in the SGK database. The reference period is the previous calendar year.

SGK tables used

Table numbers changed over the years. The table used for each year is listed below (4/1-a).

Sector lists use only the 2-digit NACE divisions. Sub-divisions are not added.

NACE Rev. 2.1 (from 2024)

In Section 3 of the 2024 SGK yearbook, the sector classification changed from NACE Rev. 2 to NACE Rev. 2.1. Division 45 (trade and repair of motor vehicles) was removed and spread over divisions 46, 47 and 95. Of the remaining 87 divisions, 21 changed name or scope (e.g. "excluding motor vehicles" was dropped from 46 and 47; 60, 63 and 96 were redefined). The sector list shows each year in its own classification (2013–2023: Rev. 2, 88 divisions; from 2024: Rev. 2.1, 87 divisions). The sector breakdown from 2024 cannot be compared directly with earlier years. The province × sector insured table in Section 1 (Table 1.11) is still Rev. 2 in 2024 and Rev. 2.1 from 2025 (division 45 is removed there too); the sector list in the province detail also shows each year in that table's version. The map, the trend and the international band do not depend on the sector classification.

Definitions

Work accident

The definition in Article 13 of Act 5510: an event that occurs while the insured person is at the workplace, because of work carried out by the employer, while sent on duty outside the workplace, during nursing time, or while travelling to and from work in a vehicle provided by the employer — and that disables the insured person, physically or mentally.

Unit of count

According to SGK metadata, the unit is the number of persons. The tables give the "number of insured having work accident". Accidents with no lost working day are included (the "accident day (at work)" column).

Death from a work accident

SGK uses the European Statistics on Accidents at Work (ESAW) definition: an accident that leads to the death of the insured person within one year.

Occupational disease

The number of 4/1-a insured persons having an occupational disease. The map shows occupational disease cases only. Deaths from occupational diseases do not appear in the province detail. They are kept in the dataset.

Insured workers (denominator)

The number of compulsory insured under 4/1-a in Table 1.8. Apprentices, interns and trainees are not included.

Calculation

Numerator and denominator cover the same scope (4/1-a) and the same year. Example, Kocaeli 2023:

How this differs from SGK's own rate

In Table 3.1.30, SGK calculates the "frequency rate of occupational injuries" from hours worked, not from persons:

For 2023, the SGK values are 14.14 per 1 million working hours and 3.18 per 100 full-time workers (these values were checked by reproducing them from the raw figures). The rate on this site is per registered insured person. The 2023 value for Türkiye is 4.15 per 100 workers. The two methods answer different questions. One should not be used in place of the other.

In Europe (Eurostat ESAW), the standard indicator is also "accidents per 100,000 workers". But ESAW counts only accidents that lead to more than 3 days of incapacity. The figures on this site include all reported accidents. They cannot be compared directly with EU figures.

Colour scale

On the map, colour is set by the square root of the value's ratio to the highest provincial value that year. This keeps very large values, such as İstanbul's, from flattening all other provinces into one colour. The legend is therefore not linear. Provinces with no cases are grey.

Annual national summary

4/1-a, total of 81 provinces. Figures are computed from the verified data file the map uses.

For occupational diseases, the SGK national total also includes cases diagnosed after insurance ended. These are not assigned to any province or sector. The national total is therefore larger than the provincial total. In 2017, 3 work accidents are in the "unknown province" row. They are not in the provincial total.

Verification

On 3 October 2026, the dataset was re-extracted from the SGK xlsx files. The 2024 and 2025 yearbooks were added on 5 October 2026. The data passed the following checks:

• passIn every province row, male + female = total (in all 44 tables).
• passSector rows add up to the table's "Total" row (4/1-a cases and deaths, 13 years; each year with the divisions of its own NACE version).
• passFor occupational diseases, provincial total = sector total. The difference from the national total equals the "after insurance ended" row.
• passDenominator: the Table 1.8 national 4/1-a total matches Table 1.2. The İstanbul value matches Table 1.10 (2025 yearbook).
• passThe site's calculation code gives the same result as an independent calculation in all 17,010 province × year × measure × display cells.

4/1-b (self-employed, Bağ-Kur)

The map shows 4/1-a only. 4/1-b (self-employed) accidents are given in one line in the province detail: 4/1-b: X accidents · Y insured · rate Z/100,000 · W deaths (compulsory scope). The deaths part is left out when there are none. SGK did not publish 4/1-b accident tables by province for 2013–2016, so the line is not shown for those years.

Source tables

Scope and choice of denominator

SGK metadata (2023 Section 3-2) defines the scope as: "Target population: work accidents of insured persons under Article 4/1-b of Act 5510" (Metadata WA, B21) and "Statistical unit: insured persons working on their own account …" (Metadata WA, B28). Voluntary insured persons are not mentioned anywhere in the file. The metadata does not say whether the accident counts include voluntary insured persons. The denominator is therefore the compulsory 4/1-b insured (Table 1.8), and the line is marked "(compulsory scope)".

Legal basis. The choice of denominator rests on the text of the Act (quotations translated from Turkish):

• (a) Work accident and occupational disease insurance applies to 4/1-a and 4/1-b insured persons. The Act lists work accident and occupational disease among the short-term insurance branches — Art. 3, item 4: "Short-term insurance branches: the work accident and occupational disease, sickness and maternity insurance branches" — and rules that these branches do not apply to 4/1-c only — Art. 4, second-to-last paragraph: "The provisions of this Act on short-term insurance branches do not apply to those deemed insured under sub-paragraph (c) of the first paragraph of this article, as long as they remain in that scope." The conclusion "applies to 4/1-a and 4/1-b" is drawn from these two provisions; the Act does not use these words.
• (b) Voluntary insurance covers only the long-term insurance branches. Art. 50, first paragraph: "Voluntary insurance is the insurance that enables persons, by paying premiums voluntarily, to be subject to the long-term insurance branches and to general health insurance."

Voluntary insured persons (IV) in Table 1.21 are therefore outside work accident insurance and are not in the denominator. Source: Act No. 5510 on Social Insurance and General Health Insurance, consolidated text, mevzuat.gov.tr (https://www.mevzuat.gov.tr/mevzuatmetin/1.5.5510.pdf), opened on 4 October 2026; Art. 3 and Art. 4 on PDF p. 1 and p. 6, Art. 50 on PDF p. 57.

Data file

The denominator used in the calculation is the istihdam_il_4b field in panel.json (Table 1.8). The Table 1.21 general total is kept in a separate file for comparison: harita_veri/sgk_il_sigortali_4b.csv (inside the project folder; columns: yil, il_kodu, il, sigortali_4b; 13 years × 81 provinces). This file is not used in the calculation. An independent extract of provincial occupational disease data is in the same folder: harita_veri/sgk_il_mh.csv (yil, il_kodu, il, rejim, mh_erkek, mh_kadin, mh_toplam, mh_olum).

Verification

• passThe CSV provincial total equals the "Total" row of Table 1.21 in all 13 years (e.g. 2023: 3,100,834).
• passIn every province and the national row, general total = sum of sub-blocks (2019–2025: compulsory (I+II+III) + voluntary (IV); verify.py §5b).
• pass2023 İstanbul and İzmir: the compulsory part of Table 1.21 = Table 1.8 4/1-b (627,778 / 167,118); the difference from the general total is voluntary insured (42,070 / 10,634).
• passThat voluntary insured persons are outside work accident insurance: SGK metadata does not say; closed by Art. 50 of Act 5510 (above).

İSİG Council layer

The fourth view on the map ("İSİG · press record") shows the provincial breakdown in the annual work-death reports of the Health and Safety Labour Watch (İSİG Meclisi). Source: isigmeclisi.org; the 2014–2023 reports were read on 4 October 2026, the 2024–2025 reports on 5 October 2026. These data are kept apart from SGK data (public/data/isig.json); they are not in panel.json.

Method (from İSİG's own reports)

• Figures are the deaths that could be identified and are given as "at least".
• According to the 2023 report, 66% of cases came from the national press and 34% from workers' colleagues, families, safety experts, workplace doctors, unions and the local press (national press share: 2020 66%, 2021 65%, 2022 68%). The reports for 2013–2019 do not publish this share.
• No SGK registration is required: uninsured, child and migrant workers and deaths abroad are included. Deaths abroad are not in the provincial breakdown, so they are not on the map; they are part of the national total in the right panel.
• The 2013 and 2017 annual reports give only the top 11 / top 10 provinces, so the layer is off for those years. The 2023 value for Tunceli is empty (the report says "all cities"; Tunceli is not in the list).
• In 2024 the report's total within Türkiye (1,897 − 35 abroad = 1,862) is 2 higher than its provincial breakdown (1,860); the report's province map gives the same figures in all 81 provinces. These 2 deaths are not assigned to any province: the provincial total is shown as 1,860 and the national total as 1,897.
• In 2025 the value for Bayburt and Tunceli is 0: the report says it recorded work deaths "in 79 cities of Türkiye (except Bayburt and Tunceli)"; this zero is given by the report itself.

Difference in scope from SGK

In the province detail, the İSİG record is shown next to SGK 4/1-a deaths from work accidents. The difference runs both ways. It is a trace of differences in scope and method, not a measure of error. In 2023, İstanbul: İSİG 274, SGK 342 (−68); Adıyaman: İSİG 58, SGK 11 (+47). In 2023 the İSİG figure is higher in 46 provinces and the SGK figure in 29; they are equal in 5. Türkiye: İSİG provincial total 1,899; SGK 4/1-a provincial total 1,966.

Extraction chain

• The provincial breakdown is a paragraph, not a table. The parser starts from each year's opening sentence and reads groups of the form "N deaths in Province" / "N workers in Province" / "in Province N workers".
• Three-way check: CSV provincial total = provincial total in the report (group × number of names) = headline total − abroad (in 2024, − 2 not assigned to a province). Equal in all eleven years; if it breaks, no file is written.
• Typographical faults in the reports (missing separator, "ise", "+56", "5’ ölüm", "Yurdışında") are recorded in a correction table: harita_veri/isig_kaynaklar.md.
• Raw files: ham_veri/isig/ (11 files); SHA-256 sums in ham_veri/SHA256SUMS_isig. build_isig_csv.py --check and build_isig_json.py --check reproduce the chain byte for byte.

Attribution

According to the footer of İSİG's reports, the content may be used freely with attribution. Source: İşçi Sağlığı ve İş Güvenliği Meclisi (Health and Safety Labour Watch), annual work-death reports 2014–2016 and 2018–2025, isigmeclisi.org.

International comparison

The band on the main page sets the official number of deaths from work accidents in Türkiye against the official numbers for the EU-27 and five EU countries (France, Germany, Italy, Poland, Spain), 2014–2024: fatal accidents only, all sectors, absolute numbers.

Sources

• EU and countries: Eurostat, European Statistics on Accidents at Work (ESAW), dataset hsw_n2_05 ("Accidents at work by NACE Rev. 2 activity and size of enterprise"). Selection: NACE Rev. 2 = all activities, size of enterprise = total, severity = fatal, unit = number. Full name of the EU series: "European Union - 27 countries (from 2020)". The file was extracted from Eurostat on 5 October 2026 (dataset last updated 25 September 2026). The default view shows 2015–2024, so the time axis was extended to start in 2014; the other dimensions are as in the default view. The previous extract (13 June 2026, 2014–2023) stays in the archive; between the two extracts only the EU-27 value for 2023 changed (3,298 → 3,322). Files: ham_veri/eurostat/, SHA-256 in ham_veri/SHA256SUMS_eurostat.
• Türkiye: SGK Statistical Yearbooks, insured persons under 4/1-a who died from a work accident, total of the 81 provinces (the same figure as "Deaths" on the map). In all 13 years this total equals the national total in the SGK NACE table. Deaths from occupational diseases are not included: the ESAW definition of a fatal accident (death within one year of the accident) covers work accidents only, and the SGK metadata gives this ESAW definition in every yearbook from 2013 to 2025. The row for cases "after the end of the insurance" exists only in the occupational-disease tables. It does not affect the total of accident deaths. Values for 2013–2016 come from the tables headed "4-1/a" in SGK's single-volume Section 3 yearbook (key comb in panel.json); values for 2017–2024 come from the separate 4/1-a volume (Section 3-1). The 2014–2016 part of the band falls in the first period.

Difference in scope

SGK counts insured persons only; ESAW covers persons in employment. Deaths of uninsured workers are not in the SGK series. The site has a separate layer for them (see İSİG section). The two official series are therefore not on the same scale: the band compares the official SGK figure with the official ESAW figure.

Why absolute numbers only

hsw_n2_05 gives case counts only. It has no number of workers (no denominator). Rates cannot be computed for the EU side, so the band shows absolute numbers. Absolute numbers also carry the differences in the size of each country's workforce. The height of a line is not a level of risk.

Why non-fatal accidents are left out

For non-fatal accidents ESAW uses a threshold of "more than 3 days" of incapacity (at least 4 days); the SGK metadata gives the same definition. The threshold is not applied the same way in every country. In Türkiye, the practice of hospital sick notes can bring cases that would fall below the threshold into the 4-days-or-more category. This is a practitioner's note from the author's field experience. It is not a measured rate or a published finding. The size of the effect is not known, so non-fatal accidents are left out of the comparison.

Breaks in series (Eurostat flag b)

Eurostat marks years with a change in method or coverage with the flag "b" ("break in time series"). The value for such a year cannot be compared directly with the year before. The 77 selected cells carry only "b"; there is no "u" ("low reliability"). These years are shown as open circles in the band.

Extraction chain

The script stops if the selected block, a series or a cell is missing. It never substitutes another region or scope. build_karsilastirma.py --check rebuilds the CSV and the JSON byte for byte; verify.py checks the 77 Eurostat cells against the xlsx with a separate reader, and the Türkiye values against panel.json. Source notes: harita_veri/uluslararasi_kaynaklar.md.

Limitations and open points

• limitNo province × sector case data. SGK publishes the sector breakdown for Türkiye only.
• limitSmall numbers. Where cases are few, a handful of cases moves the rate a great deal. Read these provinces' rates with caution.
• limitMinor accidents included. Accidents with no lost working day are counted. This raises the rates and prevents comparison with the EU.
• openWhich province a case is assigned to. SGK metadata does not say whether a case is recorded in the province where the workplace is registered or where the accident happened.
• openReference date of the denominator. The metadata does not say whether the insured count is year-end or a yearly average. The numerator accumulates over the year; the denominator is a single count. Because people join and leave jobs during the year, the risk per person may look higher than it is.
• openInterns and apprentices. Since 2016 they are insured against work accidents, but they are not in the denominator. Whether their accidents are in the numerator could not be verified.
• open2016 revision. The 2016 yearbook gives the 4/1-a total as 13,775,188. The 2016 column in the 2023–2025 yearbooks gives 13,415,843. Each year uses its own yearbook.
• openReporting level. These data cannot separate how much of the difference between provinces comes from real risk and how much from completeness of reporting.

Known limitations

• 2016 revision: the 4/1-a total in the 2016 yearbook (13,775,188) differs from the 2016 value in the 2023–2025 yearbooks (13,415,843); each year uses its own yearbook.
• Accident assignment rule: SGK metadata is silent on whether a case is recorded in the province where the workplace is registered or where the accident happened.
• Reference date of the denominator: SGK metadata is silent on whether the insured count is year-end or a yearly average.
• Interns and apprentices: SGK metadata is silent on whether accidents of interns and apprentices, who are not in the denominator, are in the numerator.

Source: SGK metadata and the checks run in this session.