Methodology
Preview for press, not yet published. The numbers are final and the citation is stable; the announcement is not out yet, so this page is not indexed.
How Global Employment Facts is sourced, selected, converted, monitored and corrected. This is the published methodology for edition 2026.09.
What the dataset holds
One value per country and field, for 18 statutory fields across 250 countries and territories. Each value carries its unit, the source it came from, the tier that source sits in, the date the figure took effect and the date a person last verified it against the source.
Everything here is a statutory minimum: the floor the law sets, not what employers offer, not what collective agreements add, not what employees actually take. In many labour markets the practical norm sits well above the floor. A reader taking these figures as a description of working life will be wrong; taking them as the legal minimum an employer has to clear will be right.
- Statutory paid annual leave annual_leave_days
- National public holidays public_holidays
- Statutory notice period notice_weeks
- Statutory severance severance_weeks
- Employer social contributions employer_ssc_rate
- Employee social contributions employee_ssc_rate
- Statutory minimum wage min_wage_monthly
- Income tax brackets income_tax_brackets
- Top personal income tax rate personal_income_tax_top_rate
- Corporate tax rate corporate_tax_rate
- Standard VAT rate vat_rate
- Statutory maternity leave maternity_leave_weeks
- Statutory paternity leave paternity_leave_weeks
- Statutory parental leave parental_leave_weeks
- Maximum probation period probation_months
- Statutory weekly hours avg_weekly_hours
- Overtime premium overtime_premium_pct
- State pension age retirement_age
Source tiers
Every source is placed in one of five tiers. The tier is published on every row, because a number is only as good as what stands behind it and a reader should not have to guess.
The legal text itself, or the government body that administers it: a labour code, a tax authority rate table, a social insurance agency circular, a ministry page setting the minimum wage. Nothing outranks this. A value on this tier carries the page it was read from and the sentence it was read out of.
Cross-country compilations assembled under one documented method by bodies with a legal research function: the ILO, the OECD, the World Bank, ISSA, the US Social Security Administration, CLEISS, Eurostat. Comparable across countries by construction, which is exactly what a single national page is not, and the right source where no statute page has been found yet.
Commercial and non-governmental compilers with a real research process, such as WageIndicator and the PwC tax summaries. Useful, and not tied to the statute, so a government row always displaces them.
Values extracted by a model from whatever page was reachable, then reviewed by a person before approval. Good enough to publish where nothing better exists, never good enough to outrank a statute. Each one is a candidate for replacement as the registry closes gaps.
Open community calendars of observances, chiefly Nager.Date. These answer a different question from the one the dataset asks: an observance is not always a paid public holiday, and the counts diverge from statute in both directions. Ranked last deliberately, and used only where a country has nothing else.
The selection rule
2. within a tier, the later effective date wins
3. source identifier, source URL, value as last-resort tie-breaks
One row per country and field. Where a country holds several approved rows for the same field, the highest source tier wins: the statute or the government that administers it, then official multilateral compilations, then secondary compilers, then machine-extracted rows, then open holiday calendars. Within one tier the later effective date wins, and source identifier, source URL and value are last-resort tie-breaks so selection is always total. Nothing is averaged, blended or estimated.
The tier comes first, and it comes first on purpose. A statute does not become less true because a community calendar was re-scraped last Tuesday, and an earlier version of this rule sorted by date, which let a 2026 scrape displace a 1938 Act. Three published figures came out wrong that way before the order was fixed.
Rows that are superseded, rejected or still pending review are dropped before selection, so they can never reach the dataset, not even as a country’s only row. Where no approved row exists for a country and field, the dataset holds nothing and the coverage table counts it as a gap. Nothing is estimated, interpolated, averaged or gap-filled.
Conversion rules
Values are stored in the unit the column names, and a conversion is recorded rather than applied silently. Annual leave is counted in working days: where a statute sets calendar days, the conversion subtracts the weekend days in each completed week and the arithmetic travels with the row. Minimum wage is monthly in the local currency: where a statute sets an hourly or daily rate, the hours assumption used to reach a month is stated on the row. Social contribution rates are employer or employee percentages of gross pay. Leave, notice and severance are in weeks. Where a figure cannot be converted honestly, no value is published.
A conversion is always recorded on the row that carries it, with the arithmetic and the assumption spelled out. Where a statute sets 28 calendar days of leave, the dataset publishes 20 working days and says how it got there. Where a minimum wage is set per hour and the page states no working month, the hours assumption used to reach a monthly figure is stated, because a reader who disagrees with the assumption needs to be able to see it and redo it.
Where a figure cannot be converted honestly, no value is published. That is why the coverage table has gaps in it, and the gaps are the point: a blank is information and a plausible guess is not.
How the dataset stays current
Every government-sourced row stores the registry entry it came from, source identifier and content hash. The monthly country-data monitor re-reads each captured page and opens a review when the hash changes. Nothing updates a published value automatically: a person confirms the new figure against the page before it moves.
Editions are cut quarterly and frozen. A frozen edition is what the public API serves and what a citation resolves to, so a paper citing edition 2026.09 will find the same numbers in it next year. Corrections made after an edition is cut appear in the next edition and in the corrections log immediately.
Corrections policy
Every value we published and later changed is listed in the corrections log on the study page, with the date, the country, the field, the old value, the new value, the reason and the source that settled it. Corrections are never removed from the log and never applied quietly. To report an error, write to hello@employborderless.com with the country, the field and the government page that shows the right figure.
18 corrections are logged so far. They are listed in full, with the old value, the new value and the government page that settled each one, on the study page.
How to cite it
The dataset is published under CC BY 4.0. Reuse it freely, including commercially, with attribution. The canonical name is Global Employment Facts (Employ Borderless) and it will not change.
Schuchmann, R. (2026). Global Employment Facts (Employ Borderless) [Data set]. Employ Borderless. https://employborderless.com/hiring-intelligence/global-employment-facts/
DOI pending. The Zenodo record is being minted; until it resolves, cite the canonical URL above.
If you cite a specific figure rather than the dataset as a whole, name the edition and the field, for example “Global Employment Facts (Employ Borderless), edition 2026.09, notice_weeks”. That is what makes a number checkable.
What the dataset cannot tell you
Coverage is uneven, and the gaps are not random: small territories, and countries whose statutes are not published online in a readable form, are under-represented. The coverage table on the study page names the count for every field.
Entitlements that scale with seniority are taken at the statutory base rate, not at the top of the scale. Federal countries that set holidays or leave regionally are flagged rather than given an invented national figure. Sector-specific and collectively agreed minimums are out of scope entirely.
And every value is as current as its effective date says it is, not as current as the day you read it. The date is published on every row for that reason.
Studies scored from it
Three published studies are views over this dataset: the Global Employer Burden Index, the Termination Cost Index and the Global Paid Time Off Index. Each publishes its own scoring methodology; the numbers they score are the ones selected here.