Twenty-six to one, inside one single market
Eurostat publishes, quarterly, the share of employees in each member state who hold a temporary rather than a permanent contract. For the first quarter of 2026 the table runs from 26.0% in the Netherlands to 1.0% in Lithuania: a factor of twenty-six between two members of the same single market, under the same directives on fixed-term work.
The European Union as a whole reports 12.2%, and the euro area 12.8%. The Dutch figure is more than double either, and it is 10.9 points clear of second place, which is Poland at 15.1%. On a table of twenty-seven countries that is not a lead, it is a different regime.
It is worth saying plainly what this measure counts, because the shape of it decides what the comparison can mean. The denominator is employees, not workers. Somebody self-employed, on contract, or invoicing through a company of their own is outside both halves of the fraction. A country can therefore report a low temporary share while a large part of its workforce holds no employment contract at all.
| Member state | Temporary share of employees | Eurostat flag |
|---|---|---|
| Netherlands | 26.0% | none |
| Poland | 15.1% | none |
| Spain | 14.8% | definition differs |
| France | 14.7% | definition differs |
| European Union (27) | 12.2% | none |
| Lithuania | 1.0% | none |
Nine points between two quarters
The Dutch share did not climb to twenty-six. It stepped there. Between the fourth quarter of 2020 and the first quarter of 2021 it moves from 17.9% to 27.2%, a rise of 9.3 points in a single quarter.
A figure that size needs a comparison, and the fairest one is the country against itself. Across the twenty-eight quarters from the start of 2019, the median size of a Dutch quarter-on-quarter step is 0.3 points, and the largest step other than this one is 1.3 points. The flagged step is thirty-one times the median and seven times the largest. It is not a big move in a series that makes big moves; it is the only move of its kind the series contains.
Eurostat says why, on the observation itself. The value for the first quarter of 2021 carries the flag b, which the dataset's own legend defines as break in time series. A break says the numbers on either side of that date were not produced the same way. It makes no claim about anybody's contract.
And the years since settle the question. If something had happened to Dutch employment in early 2021, the level would have gone somewhere afterwards. It has gone down 1.2 points across the five years to the first quarter of 2026. The lead was not built. It was, in the arithmetic sense, redefined.
| Period | Dutch temporary share | Change |
|---|---|---|
| The flagged quarter, one step | 17.9%27.2% | +9.3 |
| The five years after it | 27.2%26.0% | -1.2 |
the flagged step, fourth quarter of 2020 to first of 2021
+9.3 -1.2
net movement in the five years since that quarter
the median of every other step in the window
0.3 1.3
the largest of every other step in the window
Measured over the twenty-eight quarters from the start of 2019 to the first quarter of 2026. The step across the flagged quarter is compared with the same country's own steps, so the comparison does not depend on any other country behaving normally.
Spain moved the other way, and moved for longer
The country most readers associate with temporary work is Spain, and the association is out of date. The Spanish share peaked at 26.8% in the third quarter of 2019 and reads 14.8% in the first quarter of 2026: a fall of 12.0 points.
The difference between that fall and the Dutch rise is not the size. It is the shape. Spain's step across the flagged quarter is minus 0.6 points, which is an ordinary quarter for Spain. The fall is spread over years, and its steepest passage runs through 2022, from 24.5% to 18.2% across four quarters. Whatever produced it, it did not arrive in one step and it did not arrive at the break.
What produced it is a question this piece does not answer. A labour reform took effect in Spain during that period, and the temptation to draw a line from one to the other is strong. The dataset does not license it: a series shows a shape, not a cause, and nothing else here is a document about Spanish law. The shape is reported; the explanation is left open.
Two countries carry a second flag
The break at the start of 2021 is not a Dutch event. Every one of the thirty-six countries in this dataset carries a break flag at that quarter, the Union and euro-area aggregates included. No comparison that crosses it is clean anywhere, for anyone.
A second flag is rarer and easier to miss. d means definition differs, and unlike a break it sits on individual observations rather than on a date. In the latest quarter exactly two member states carry it: Spain and France. Spain has carried it on every quarterly observation since the second quarter of 2021, twenty-one in a row, and its 2021 value carries both flags at once.
Their position in the table is what makes this worth printing. Spain reads 14.8% and France 14.7% - first and second in a run of adjacent countries, one tenth of a point apart. A reader ranking Europe by this measure is placing two differently-defined series against twenty-five standard ones, and the gap between them is smaller than the thing the flag is warning about.
break in time series
36
countries carrying a break flag at the first quarter of 2021, out of thirty-six reporting
definition differs
2
countries whose latest observation carries a definition flag: Spain and France, adjacent in the ranking at 14.8% and 14.7%
Two different flags, and the piece turns on the difference. A break says the figures on either side of a date were not produced the same way. A definition flag says this country's figures are not built to the standard definition, and it is attached to individual observations rather than to a date.
What the flags license, and what they do not
None of this makes the table useless, and reporting it that way would be its own error.
Within the latest quarter the comparison stands: every country is on the far side of the same break, so 26.0% against 1.0% is a real difference between two labour markets today, whatever produced the definitions on either side. The two definition flags qualify two rows of that comparison, and the article prints which two.
What does not stand is the sentence a reader is most likely to form from a chart of this series: that the Netherlands became Europe's most temporary labour market over the past five years. On the numbers, it arrived there in one quarter, has drifted down since, and the quarter in question is marked by the publisher as one where the counting changed.
The useful version is narrower and holds. Before comparing two labour markets across time, check whether one of them changed its ruler. Eurostat wrote that check into the data as a single letter, on every observation where it applies, and it costs nothing to read.
What the data establishes
- In the first quarter of 2026 the Dutch share is 26.0% and the Lithuanian share is 1.0%, both on the far side of the same break.
- The Dutch series rises 9.3 points across the single quarter Eurostat flags as a break, against a median step of 0.3 points and a largest other step of 1.3 over the same window.
- The Dutch level is 1.2 points lower now than immediately after that step.
- All thirty-six reporting countries carry a break flag at the first quarter of 2021.
- The Spanish share fell 12.0 points from its 2019 peak, and its step across the flagged quarter was minus 0.6 points.
- Spain and France are the only member states whose latest observation carries a definition flag.
What it does not
- That the Dutch labour market became more temporary in early 2021. The flag says the counting changed; it says nothing about contracts.
- What Eurostat or the Dutch statistical office altered in the survey, or why. The dataset carries the flag, not the reason.
- That Spain's 2022 labour reform caused the fall in its share. The shape is measured here; no document cited establishes the cause.
- That a low temporary share means secure work. The denominator is employees, so a country with many self-employed people can report a low share for reasons this measure cannot see.
- How the two flagged series would rank against the other twenty-five if they were built to the standard definition.
Method and what is not claimed
- One document, linked below: the Eurostat dataset lfsq_etpga, "Temporary employees - % of total employees - quarterly data", read on 28 August 2026 and stamped as updated on 12 June 2026. The latest reference period it carries is the first quarter of 2026, so this is a quarterly series with a lag, not a fresh monthly release. Nothing here comes from our own platform.
- Figures are for age 15 to 64, both sexes, unit percent, as published. The measure is temporary employees as a share of employees, not of all people in work: the self-employed are outside both the numerator and the denominator.
- The flags are the publisher's, not ours. Eurostat attaches an observation status to individual values, and the dataset ships its own legend: `b` is "break in time series", `d` is "definition differs (see metadata)", `bd` is both. Every flag quoted in this piece was read from the dataset response, not inferred from the shape of a line.
- One comparison is derived: the Dutch step across the flagged quarter is set against the median and the maximum of that same country's other quarter-on-quarter steps over the twenty-eight quarters from the first quarter of 2019. The median is 0.3 points and the largest other step is 1.3 points, against a flagged step of 9.3. The arithmetic is ours; the values are published.
- No cause is claimed for either move. A labour reform took effect in Spain during the period in which its share fell most steeply, and this piece does not connect them, because no document cited here establishes the connection. Likewise the Dutch break is reported as a break, with no claim about what was changed in the survey or why.
- We do not say the Dutch labour market became more temporary in early 2021. The flag forbids reading the step that way, and the level has fallen 1.2 points in the five years since.
- Country rows in the table are six of twenty-seven: the highest, the second, the Union aggregate, the two countries carrying a definition flag in the latest quarter, and the lowest. The full table was read; the exhibit is a selection and says so.
- Percentage points throughout for differences between shares, never percent.
Primary sources
We publish how our own matching is tested for bias, and mark the dimensions we cannot measure rather than filling them with a plausible number - the method and the numbers are here. A market whose own statistics carry warning flags deserves tools that carry theirs.