What the minimum wage has really gained against prices since 1996, why it is not felt, and how it compares with other wages
France ÉcoScan v9.0.14 · edition 2026.2Published on 2026-09-15Year 2025 · monthly readings up to August 2026
The question
Does the SMIC really protect purchasing power against the rising cost of living — notably during inflationary shocks — and has its progression been comparable to that of other wages?
The measurement covers 1996–2025 at annual steps and January 1996–August 2026 at monthly steps. It does not go back to 1990: INSEE rebased the price index in January 2026, the current base starts in 1996, and the study prefers a living series to a splice of bases that the producer does not publish. For 1990–2025, it merely reproduces the quantified claim that INSEE itself publishes.
The answer
Yes over the long run, no at any given moment: since 1996, the gross hourly SMIC has gained purchasing power against prices, but through successive catch-ups after periods of erosion, and more and more slowly.
What this study does not settle
Three reservations, and the full development follows in the Method area. First, the labour market: none of the 9 relationships tested between the real progression of the SMIC and employment is robust enough to conclude, in either direction. Second, the study measures an hourly FLOOR, never the people who receive it — their number is not in the certified catalogue. Finally, it starts in 1996 and not in 1990, for lack of a price index published before that date in the current base. 4 foundation gaps are declared in all, each with what its absence forbids concluding.
This study would fall if the real SMIC index ended below its 1996 level, if a single one of the 11 starting years tested produced a loss, or if a slump had stayed open for a long time. All three were checked before publication; none occurs. The full criteria, written hypothesis by hypothesis BEFORE the calculation, are in the Method area.
The full answer, quantified, and the mechanism that explains it
The three stages of the answer with their figures — the gain in purchasing power from 1996 to 2025 and its pace, the 27 slumps and their catch-ups, the comparison with the basic wage of blue-collar workers —, then the calendar mechanism that produces them.
Where does the SMIC stand, and why can the long-run gain be hard to feel?
Three questions, three answers. They do not contradict each other: they describe three different objects — the month one lives through, the 30-year average, and the path between the two.
Where do we stand today?
Gap to the best level already reached, in August 2026
−1.33%
The purchasing power of the hourly SMIC is compared with its highest level in the window, reached in January 2026, 7 months ago.
Derived · 2026-08 · France ÉcoScan — derived from the INSEE series cited
The rate has stood at €12.31 an hour since its uprating of June 2026 (+2.41%), which followed 5 months of standstill. During those months prices kept rising: it is this lag, not a decision, that opens the gap.
What this is not — This is not a permanent loss: over this window, every gap of more than 1% opened before this one has closed. Nor is it the purchasing power of an employee: it is that of the gross hourly floor.
Has the SMIC gained purchasing power since 1996?
Purchasing power gained by the hourly SMIC since 1996
+35.1%
Average gross hourly amount for the year, deflated by the price index excluding tobacco, from 1996 to 2025. The 368 months of monthly readings covering the same period serve to describe the SHAPE of the path, never to change its end point.
Derived · 1996-2025 · France ÉcoScan — derived from the INSEE series cited
Over 30 years the SMIC has indeed gained purchasing power, and the result does not depend on any methodological choice: it comes down solely to the gap between the rise in the amount and the rise in prices. At the latest published month, August 2026, the same measure gives +36.2% — the difference from the annual figure is that of the 8 months after 2025 that the closed year does not yet contain.
What this is not — This is not a steady gain, and that is the heart of the study: it was acquired almost entirely before 2010. Nor is it the gain of SMIC workers as a group — the study measures a floor, not a population.
Why can this gain be hard to feel?
Share of time spent below a level already reached
87%
320 months out of 368 in which the purchasing power of the SMIC is below its best previous level.
Derived · 1996-01/2026-08 · France ÉcoScan — derived from the INSEE series cited
This answers the question “why does a measured gain coexist with a feeling of loss?” The SMIC is set by decree; prices rise every month. Between two decrees, purchasing power erodes mechanically — and lived experience is that of those months, not that of the thirty-year average.
What this is not — This is not impoverishment: the starting level of each episode is itself higher than the previous one. It is a saw-tooth trajectory that rises, not a descent. And it is not a statistical anomaly: the proportion is high BY CONSTRUCTION, because an amount that advances in steps cannot keep up with prices that advance every month. That is precisely the mechanism the figure serves to show.
Premium
Three pieces of evidence, measured over the same 368 months
How many times the purchasing power of the SMIC slumps, how far it falls at worst, and how long the longest stretch below a level already reached lasts — each figure with its reading and what it does not mean.
Do prices for low-income households rise faster?
This is the most common objection to any measure of purchasing power, and it can be checked: INSEE publishes the indexA series rebased to 100 in a reference year. It no longer gives the level; it gives how many times the level has been multiplied since that year.Click to read the full reading aid. of prices for the 20% of households with the lowest standards of living — what official statistics call the first quintileA 20% slice. Ranking all households from the least to the most well-off and cutting into five equal parts gives five quintiles; the first quintile is therefore the bottom 20% of the distribution of standards of living.Click to read the full reading aid.. It is even the one that the Labour Code (code du travail) uses to uprate the SMIC.
Premium
Prices for low-income households, over the whole window and during the 2021–2023 shock
The cumulative gap between the price index for the 20% of lowest-income households and that for all households, measured month by month, then isolated over the inflationary episode — and what it changes for the purchasing power of the SMIC.
The SMIC against the rest of the scale
A floor that rises and a scale that compresses produce the same picture and call for opposite readings. Telling them apart requires wage series: six are in the certified foundation, and the cards below place the SMIC relative to the medianThe value that splits a population into two equal halves: half are below, half above.Click to read the full reading aid. of wages and to the decilesThe same idea in ten parts. D1 is the threshold below which the 10% lowest paid are found, D9 the one above which the 10% best paid are found. The D9/D1 ratio sums up the spread of the wage scale in a single number.Click to read the full reading aid. that bound the bottom and the top of the scale.
Premium
The SMIC against the median and the deciles of wages
The race between the SMIC and the basic wage of blue-collar workers and the average net wage, the spread of the wage scale, the test of “smicardisation” broken down into two propositions, and the pace at which the real wage has doubled since 1951.
What this study does not measure: housing
Many readers spontaneously associate purchasing power with the impossibility of becoming a homeowner. The association is legitimate, and the two measures are different.
The ability to buy everyday goods and the ability to buy a home are not measured in the same way. The latter also depends on housing prices, interest rates, the length of the loan, the deposit and lending conditions — five quantities, none of which enters this study.
The product applies this boundary rather than merely writing it: in the cross-referencing tool, setting a wage against the mortgage rate is flagged as not recommended, with this reason. The subject deserves its own study, not a paragraph borrowed from this one.
The long-run measure, and its robustness
Premium
The long-run measure, and its robustness
The real gain from 1996 to 2025 as a range depending on the starting year, what the date convention changes, the worst year of the window, and the share of the progression carried by a handful of years.
What this result does not prove
The SMIC is legally indexed to a price index, with an automatic catch-up clause during the year: a lasting loss of purchasing power would require a change in the law, and H1 is therefore not a very risky test. What the measurement brings is not the sign of the gain, but its size, its very strong concentration in a few years, and the fact that it far exceeds what price indexation alone would have produced — which would, by construction, give a zero real gain.
Why is the trajectory saw-toothed?
The SMIC does not rise: it jumps. A decree raises it on a given date, then the amount stays still while prices keep rising every month. Between two decrees, purchasing power therefore erodes mechanically — without any decision having been taken, and without any crisis having occurred.
uprating on 1 July every year until 2009, with additional movements in 1996, 2005, 2006 and 2008; moved to 1 January from 2010 (law of 3 December 2008), with two exceptions in December 2011 and July 2012; return of intra-year movements from 2021 to 2024 under the automatic catch-up clause, then no movement for 13 months, from November 2024 to January 2026, and upratings on 1 January 2026 and on 1 June 2026.
Consequence — no residual bias on the end-to-end gain: the measure bears on the annual average of the amount, not on its value at 1 January. The calendar explains the saw-tooth shape of the trajectory, not the level of its end point. The study measures the amount as an annual average, derived from the twelve monthly readings of the producer: the date of decrees therefore no longer distorts the level measured. It does, however, explain the shape of the curve, and that has become the subject.
Calendar recorded month by month on 2026-09-29 from INSEE — Banque de données macro-économiques (macroeconomic database), idbank 000822484, monthly series.
The two stages of an uprating, and what each one catches up
The answer to “does the SMIC react to prices straight away, or with a lag?” is therefore: with a lag of a few months, not a full year. The January decree looks backwards; the mid-year catch-up, for its part, responds to the prices of the moment. It is these months of waiting between two steps, and not a decision, that open the gap.
Diagram of a mechanism: it carries no scale, no year, no value, and no data is plotted in it. The real plot is the chart “The saw-tooth”; the quantified measurement of these two stages is in the Method area.
1 free chart|8 premium chartsThe premium panels work — they are deliberately closed. Click one to see what it contains.
The purchasing power of the SMIC and what happens inside each year
Is the long-run gain experienced as a continuous progression?
FreeObserved 1996-2025index, 100 in January 1996
Reading
Internal range of the year, from low to high
Source — INSEE — monthly SMIC (000822484) and price index excluding tobacco (011814056). Axis: The average and the envelope start from the same month — January 1996, which stands at 100 —: the axis therefore starts near this base, not at zero. Forcing it to zero would squeeze into the upper quarter of the frame the only thing to read — the gap between the highest and the lowest point of each year, which is at most 6.82 points.
What this chart cannot show, and how to use it
Limits
The band is annual: it says by how much the level moved within the year, never in which month or for how long. Those two answers are in the table of episodes.
The first year serves as the base: its internal range is real, but it is measured relative to a starting point that is itself within the year.
Interactions
Hover over the plot: a tooltip gives the point's values, unit and status. Click to hold the selection — it does not disappear when the pointer moves away, so that the reading stays displayed while you read it. On the keyboard, the tab key brings focus to the plot, the arrow keys move the selection, Home and End go to the ends, and Escape closes. On a touch screen, one press selects a point, a second releases it.
Premium
The 27 slumps in purchasing power, one by one
Each slump of at least 1% since January 1996: the month of the last peak, that of the lowest point, the loss reached, the date on which the peak was regained and the number of months spent below it.
The labour market: what the data do not allow us to conclude
Of the 9 relationships tested between the real progression of the SMIC and the labour market, none is robust enough to conclude once the number of tests run is taken into account.
Neither one way nor the other: this study does not show that the SMIC destroys jobs, and it does not show that it does not destroy any either. The result is published here with the three quantities that carry it; the detail of the tests, their probabilities and the scatter plot that makes them visible are further down, in the verification area.
Three quantities, not one: an ILO unemployment rateThe share of people without a job, available for work and actively looking, relative to the labour force. The figures served here are FRENCH: they come from INSEE's Labour Force Survey (enquête Emploi), which applies to French households the statistical definition set by the International Labour Organization. The ILO provides the measurement rule, not the measurement.Click to read the full reading aid. can fall because people found a job, or because they stopped looking. The employment rateThe share of people in work, relative to the whole age group — not only to those looking for a job. Measured, like the unemployment rate, on French households by INSEE's Labour Force Survey, according to the International Labour Organization definition.Click to read the full reading aid. has no such ambiguity. Differences between two dates are written in percentage pointsThe unit of a gap between two percentages. Going from 8% to 7% is a fall of ONE POINT, not of 1%.Click to read the full reading aid., never in percentages.
Premium
Unemployment and the two employment rates, at both ends of the window
The unemployment rate as defined by the ILO, the employment rate of 15–64-year-olds and that of 15–24-year-olds in 1996 and in 2025, with their extremes and what these three figures do not allow us to say.
What is missing, and what its absence forbids concluding
A study that does not say what it lacks suggests that it lacks nothing. The quantities below are not in the France ÉcoScan certified catalogue, and the study refuses to fabricate them — just as it refused to fabricate the splice of bases that would have allowed it to go back before 1996.
The six years 1990-1995. The consumer price index excluding tobacco retained as the deflator is served in the producer's 2025 base, which goes back to January 1996.
Why it matters — The question put to the reader begins in 1990; the measurement begins in 1996. Six years of the question are therefore covered by no measurement in this study.
Consequence — The study measures over 1996-2025 and says so. For the period 1990-2025 it publishes only a comparison with INSEE's own quantified statement, built on the all-items index, which does go back to 1990. No splicing of bases is fabricated.
The ARBITRABLE gross disposable income of households — what remains once contractual outlays, rent, insurance and subscriptions have been paid. INSEE published it until 2021 in the 2014 base of the national accounts; the series has since carried the note “Series discontinued” and has not been republished in the 2020 base.
Why it matters — It is the link that would explain why a real rise in wages may produce no effect on what is left to live on: if the constrained share of the budget rises faster than income, measured purchasing power rises while room for manoeuvre shrinks.
Consequence — The study does not fill it, and does not work around it either by borrowing a HOUSEHOLD series to speak of EMPLOYEES: a quintile of households and a population of SMIC employees do not overlap. The question remains open, named, and without a figure.
The proportion of employees actually paid the SMIC is not in the certified catalogue.
Why it matters — It says how many people the floor actually concerns, and its variation says whether the SMIC's grip is extending or tightening. A gain in purchasing power accompanied by a growing share of employees on the SMIC is not read as the same event as a gain accompanied by a shrinking share.
Consequence — The study measures the level of the floor, never the population that receives it. It can therefore say nothing about the change in purchasing power of SMIC employees as a group.
The lags tested are exclusively FORWARD-LOOKING: labour market at t, t+1, t+2 for a SMIC at t. No backward lag was frozen before calculation.
Why it matters — This is the direction that would flatter a reading in which “the SMIC acts on employment”. The reverse direction — the economic cycle acts on the SMIC — is at least as plausible in France, since the legal uprating formula involves a wage quantity that is itself cyclical.
Consequence — The study cannot distinguish a SMIC → labour market association from a labour market → SMIC association. Adding them now, after seeing the results, would destroy the freeze that gives the rest its value: they are recorded for a later batch, before calculation.
Premium
The detail of the 29 annual changes
Each year from 1997 to 2025, with the nominal progression of the SMIC, the year's inflation, the real progression that results and the change in unemployment in the same year — including the 6 years in which purchasing power fell.
Premium
The 7 periods of continuous progression or decline
The cutting rule, then each run of consecutive years in which purchasing power moves in the same direction, with its cumulative change, its annual pace and the change in unemployment over the same interval. This is the direct test of the second half of the question.
Premium
The 9 association coefficients and their decomposition
The real progression of the SMIC compared with three labour-market quantities, at three lags fixed in advance, plus the decomposition that says which of the two terms of the ratio — the SMIC or prices — carries the association observed.
Premium
The test of the 11 starting years
The real gain up to 2025 recalculated from each of the 11 possible bases, with for each its duration and its annual pace — the quantified answer to the objection about the choice of base.
Premium
The test of the 21 starting years on the labour market
For each starting base, the unemployment gap and the two employment-rate gaps up to 2025 — the same robustness as that applied to the real gain, applied this time to the hypothesis that it weakens, with the starts that reverse the sign named.
What INSEE publishes on its side
A measure that cannot be compared with an independent publication is not verifiable. The three comparisons below are open to everyone: charging for the right to check would make the rest unverifiable.
Quantity compared
France ÉcoScan
INSEE, published
Gap
Gross hourly SMIC, multiplier
×2.6
×2.6
aucun
Consumer prices, multiplier
×1.8
×1.8
aucun
SMIC index in constant euros, 2000-2024
+26.4%
+23.5%
+2.89 pts
The first two multipliers are rounded to one decimal place, as INSEE rounds its own: at this precision, they coincide. The third row does not — and its gap is the one the date convention leads one to expect: the study deflates the SMIC in force on 1st January, whereas INSEE builds its index on the annual average. The corresponding card below quantifies this gap and says what it does not prove.
Rise in the hourly SMIC and rise in prices since 1996
SMIC ×2.61 · Prices ×1.80
INSEE writes, on 13 February 2026: “At the fourth quarter of 2025, the gross hourly Smic was multiplied by 2.6 and prices (including tobacco) by 1.8 since March 1990. [INSEE statement, translated from French]” Rounded to one decimal, France ÉcoScan's two multipliers are 2.6 and 1.8.
Derived · 1996-2025 · France ÉcoScan, compared with INSEE — Salaire minimum interprofessionnel de croissance (Smic)
The two multipliers coincide to the decimal, over almost identical windows. This is a check of the two raw inputs, not of the method that combines them.
What this is not — INSEE starts from MARCH 1990 and arrives at the fourth quarter of 2025. The study reproduces exactly these bounds for the SMIC, for which it holds the monthly readings, and compares prices from 1990 to 2025 on the all-items index — the one the sentence cites, tobacco included. INSEE's multipliers are rounded to one decimal, and the comparison is made at that precision.
Gain in purchasing power of the SMIC in constant euros, 2000-2024
France ÉcoScan 26.4% · INSEE 23.5%
Gap of +2.89 percentage points percentage points of cumulative growth over 24 years, between France ÉcoScan's calculation and the index INSEE publishes for the same quantity. Series 001665250, base 100 in 2000, retrieved on 2026-08-17.
Derived · 2000-2024 · France ÉcoScan, compared with INSEE — Indice du SMIC brut horaire - Base 100 en 2000 - Euros constants - Durée légale hebdomadaire de travail de 35 h
The gap goes in the direction announced by the study's stated limit, and it QUANTIFIES it: The DATE convention is now the same on both sides: annual average of the amount. One difference remains, and only one — the DEFLATOR. INSEE deflates by the all-items index, tobacco included; the study deflates excluding tobacco, as does Article L3231-4 of the Labour Code (code du travail). This gap is not tolerated, it is COMPUTED from the two indices of the catalogue, and the tolerance bounds only the residual that remains once tobacco is removed.
What this is not — This is not an endorsement of France ÉcoScan by INSEE, but a comparison made by France ÉcoScan with an INSEE publication. Nor does it cover the whole window: The institutional series begins only in 2000: the benchmark covers 2000-2024, not the whole window of the study. What it establishes therefore holds for these twenty-four years, and the study does not extrapolate its result to the rest of the period.
Consistency of the inflation measure used to deflate
Maximum gap 0.053 percentage point
Maximum absolute gap, year by year, between the inflation implied by the price index retained as the deflator and the inflation rate INSEE publishes under another identifier, over the 29 years that are genuinely independent (1997-2025).
Derived · 1997-2025 · France ÉcoScan, compared with INSEE — Les comptes de la Nation en 2025 (National accounts, 2025)
The price level retained as the deflator reproduces the inflation rate that INSEE publishes separately, to within a few hundredths of a point.
What this is not — This is not a validation of the deflator but a non-regression check: the control series is rounded to the tenth of a point, and the gaps observed fall essentially within this rounding envelope. Years before 1997 are excluded, the published series there being itself derived from the tested index.
The SMIC against hourly wages and labour cost, second quarter 2026
SMIC +2.0% · wages +2.4% · labour cost +2.5%
Year-on-year changes for the second quarter of 2026. INSEE writes, on 15 September 2026: “In the second quarter of 2026, hourly wages rise by 2.4% over one year, the hourly labour cost by 2.5% [INSEE statement, translated from French]”, and attributes the acceleration to the June uprating of the SMIC.
Derived · second quarter 2026, year on year · France ÉcoScan, compared with INSEE — Indice du coût du travail (ICT) — résultats détaillés, deuxième trimestre 2026 (Informations rapides n° 219)
Over one year to the second quarter of 2026, the quarter's average SMIC rose slightly less than hourly wages across the whole scope, and about as much as prices excluding tobacco; but the June step puts the amount in force +3.6% above its level a year earlier. INSEE reads in this step the reason for the acceleration of hourly wages: “This acceleration is driven notably by the uprating of the Smic in June 2026 (+2.4%) [INSEE statement, translated from French]”.
What this is not — It is neither the wage of a SMIC employee, nor a measure of the share of employees affected by the uprating: it is the cost of an hour of work for the employer, across all employees in scope.
Gross hourly SMIC, quarterly average€11.88 → €12.12 — calculated from INSEE monthly readings (idbank 000822484)
+2.0%
Gross hourly SMIC, June alone€11.88 → €12.31 — the step of 1 June 2026, not the average
Hourly labour cost (labour cost index)observed — INSEE, same scope, wages, contributions and taxes net of subsidies
+2.5%
Prices excluding tobacco, quarterly averagecalculated from INSEE monthly readings (idbank 011814056)
+2.1%
These comparisons are made by France ÉcoScan against INSEE publications. They constitute no endorsement of France ÉcoScan by INSEE, or by anyone.
Method, limits and robustness
The formula, to copy
Dividing an amount by a price index is deflatingThe price index by which an amount is divided to remove the rise in prices. This is the operation that turns current euros into purchasing power.Click to read the full reading aid. : the operation that turns current euros into constant eurosAn amount from which the rise in prices has been removed, so that two years can be compared on an identical basket.Click to read the full reading aid., that is, into purchasing power.
real index(a) = 100 × [ SMIC(a) / SMIC(1996) ] / [ prices(a) / prices(1996) ], where SMIC(a) is the AVERAGE of the twelve gross hourly amounts in force during year a (INSEE, idbank 000822484, monthly readings) and prices(a) the average of the twelve monthly readings of the consumer price index EXCLUDING TOBACCO, all households, France, in the producer's 2025 base (INSEE, idbank 011814056). Neither series is rebased at ingestion: the ratio to the year 1996 is formed here, where it is visible and reversible.
The operation this study forbids itself
subtracting or directly comparing an inflation RATE (%) and a SMIC AMOUNT (€): these are two different dimensions, and their difference has no unit.
How France ÉcoScan qualifies its claims
Every quantified claim in this study carries one of these four natures. They are not equal: they say what one is entitled to conclude from them.
Fact
Value published directly by a source, at its date.
Calculated result
Result calculated by France ÉcoScan from published data, with its formula displayed.
Association
Two quantities move together statistically, without this indicating which acts on the other, or whether a third explains both.
Causality
Assertion that one quantity produces an effect on another. In this study: no causal assertion — none of its methods identifies a causal effect.
The 4 sentences this study forbids itself to write
These are not hypotheses: they cannot be refuted by the data. They constrain the pen, and a guard checks that no served text breaks them.
No causal sentence, in either direction. No method used here — averages, ratios, correlations on changes — identifies an effect, and no instrumental variable, discontinuity, natural experiment or control of confounders is implemented.
No operation mixing two dimensions: subtracting or directly comparing an inflation RATE (%) and a SMIC AMOUNT (€): these are two different dimensions, and their difference has no unit.
No conclusion resting on a gap between end points of less than one point for the Labour Force Survey series: four documented overhauls separate the two ends of the window.
No sentence presenting this study as a measure of the purchasing power of employees in general, nor of the average wage, nor of the net wage.
A · The SMIC and prices — measuring the lag
The diagram of the two stages of an uprating, further up, says that the SMIC reacts to prices with a lag of a few months rather than a full year. Here is the measurement that proves it. It covers only two series — the nominal progression of the SMIC and inflation — and it has nothing to do with the 9 labour-market tests.
Premium
The measurement of the SMIC's lag behind prices
The two coefficients that say whether the nominal progression of the SMIC follows the prices of the same year or those of the previous year, with their period, their lag and their scale.
What would invalidate each hypothesis, written before the calculation
The gross hourly SMIC has gained purchasing power against prices since 1996.
Refuted if — the real index of the latest year is less than or equal to 100, OR a single one of the starting years tested between 1996 and 2006 produces a real loss at the end, OR correcting the date-convention bias suffices to cancel the gain.
Measured from end point to end point, the labour market is not in a worse state at the end of the window than at the start.
Refuted if — the unemployment rate is higher at the end than at the start for the MAJORITY of the starting years tested, OR the employment rate of 15-64-year-olds is lower at the end for any one of them.
At annual frequency, years of strong real SMIC growth are not associated with a deterioration of the labour market.
Refuted if — one of the nine association coefficients survives the correction for the number of tests conducted AND keeps its sign and order of magnitude when the two most influential years are removed AND is not explained by the SMIC uprating calendar.
Premium
What this study does not settle, and what would change its conclusion — in full
The open questions one by one, with the labour-market quantities and the foundation gaps that leave them open, then the invalidation criteria for each part of the answer, with the measured values that keep them at bay.
Where these figures come from
All the figures in this study come from INSEE, and from no other source.
Five series enter it: the amount of the gross hourly SMIC and the annual consumer price index, read by web service in the macroeconomic database (Banque de données macro-économiques); the unemployment rate and the two employment rates, read in the long series of the Labour Force Survey. A sixth, the published inflation rate, serves only to check the fifth and enters no calculation.
Gross hourly SMIC
the legal minimum wage, in current euros
Producer
INSEE
Official series
000822484
Unit
€
Scope
Amount of the gross hourly SMIC (cross-sector growth-linked minimum wage) — effective step: every uprating, including mid-year, is a monthly reading from the producer, and the applicable value is the latest reading — value in force on 1 January
Latest year in the catalogue
2026
Read by the study up to
August 2026 monthly; 2025 annual average
Revision status
FINAL
Retrieved on
2026-09-29
Transformation
average of the twelve amounts in force during the year, derived from the monthly series, then a base-100 index at the first year of the window formed in the formula
Does not measure — the NET wage received, neither the employer cost nor the median wage — it is a gross legal floor, and a minority of employees receive it
· Average of the twelve amounts in force during the year, derived from the monthly series: uprating on 1 July every year until 2009, with additional movements in 1996, 2005, 2006 and 2008; moved to 1 January from 2010 (law of 3 December 2008), with two exceptions in December 2011 and July 2012; return of intra-year movements from 2021 to 2024 under the automatic catch-up clause, then no movement for 13 months, from November 2024 to January 2026, and upratings on 1 January 2026 and on 1 June 2026. Consequence — no residual bias on the end-to-end gain: the measure bears on the annual average of the amount, not on its value at 1 January. The calendar explains the saw-tooth shape of the trajectory, not the level of its end point.
· The base year of the window itself contains an uprating during the year: it is the base where the date shift matters most, and the study publishes what a start one year later costs.
· GROSS amount. The change in the net depends on levies, which have themselves changed over the period.
· Years before 2002 are expressed in euros by conversion from the franc, at the legal fixed rate: the conversion affects no ratio calculated here.
Consumer prices excluding tobacco (base 100 in 2025)
the price level excluding tobacco, the only deflator admitted
Producer
INSEE
Official series
011814056
Unit
Index
Scope
Consumer price index excluding tobacco, all households, France — base 100 in 2025, average of the twelve monthly readings for the year — average of the observations of the year
Latest year in the catalogue
2025
Read by the study up to
August 2026 monthly; 2025 annual average
Revision status
FINAL
Retrieved on
2026-09-29
Transformation
average of the twelve monthly readings for the year, then the ratio to the year 1996 formed in the formula
Does not measure — the inflation RATE, which is its annual change and not its level; nor purchasing power, which would relate these prices to incomes; nor the cost of living of any particular household, which an index weighted on average consumption cannot describe
· Series published in the producer's 2025 base, going back to January 1996. The study does NOT splice it to the previous base, which would go back to 1990: INSEE does not publish that splice, and over the 30 years of overlap the two bases give the same rise to within 0.01 point.
· Calendar-year average. It is compared with an average income for a year, not with a fixed-date value.
· All-households index. It describes the basket of no particular household, and notably not that of a SMIC employee. The study publishes separately the index of the first quintile, which approaches it without being the same thing — a quintile of HOUSEHOLDS is not a population of EMPLOYEES.
· Excluding tobacco. For a smoker, the index understates the rise experienced; that is the price of not attributing to the cost of living a rise that stems from a public-health policy.
· The geographic scope of the consumer price index was extended during the period (metropolitan France, then all of France, then the integration of Mayotte). The producer ensures the splice; the study does not redo it and cannot check it point by point.
Prices for low-income households — first quintile (base 100 in 2025)
the inflation experienced by modest households — a distinct question, never a second deflator for the main result
Producer
INSEE
Official series
011818264
Unit
Index
Scope
Consumer price index excluding tobacco for households in the first quintile of standard of living, France — base 100 in 2025, average of the twelve monthly readings — average of the observations of the year
Latest year in the catalogue
2025
Read by the study up to
August 2026
Revision status
FINAL
Retrieved on
2026-09-29
Transformation
average of the twelve monthly readings for the year
Does not measure — the LEVEL of prices paid by these households, which is not measured, nor the basket of a SMIC employee: a quintile of households and a population of employees do not overlap
· Series published in the 2025 base from January 2011 only: the comparison covers 15 years, not the whole window of the study.
· Households, not employees. The first living-standard quintile includes pensioners, students and households without employment; it does not describe the population of SMIC employees.
· Weights updated each year by the producer: the index follows the observed consumption structure, which itself shifts with prices.
Basic hourly wage of blue-collar workers (base 100 in 2000)
the only wage directly comparable with the SMIC — gross, hourly, excluding bonuses
Producer
INSEE
Official series
001665241
Unit
Index
Scope
Index of the gross basic hourly wage of blue-collar workers, base 100 in 2000, current euros — France, private sector — annual value published as is
Latest year in the catalogue
2024
Read by the study up to
2024
Revision status
FINAL
Retrieved on
2026-09-29
Transformation
served as is
Does not measure — the wage actually paid, from which bonuses, overtime and seniority are excluded; nor a wage level, which a base-rate index does not give
· Scope of MANUAL WORKERS (ouvriers), not manual and clerical workers. Widening the label would widen the population without changing the data.
· Index of the BASE rate: it ignores the variable part of pay, which grew over the period.
· Published on a base 100 in 2000: the comparison with the SMIC covers 2000-2024, not the whole window.
Average monthly net wage, full-time (base 100 in 2000)
the middle of the wage scale — context, never a direct term of comparison with the gross SMIC
Producer
INSEE
Official series
001665246
Unit
Index
Scope
Index of the average monthly net wage of full-time jobs, base 100 in 2000, current euros — France — annual value published as is
Latest year in the catalogue
2024
Read by the study up to
2024
Revision status
FINAL
Retrieved on
2026-09-29
Transformation
served as is
Does not measure — the change in the wage of the same person: an average also moves when the composition of employment changes, without any employee having had a raise
· NET when the SMIC served is GROSS. Comparing the two directly would measure in part the change in social contributions, and attribute it to the wage scale.
· FULL-TIME posts only: part-time work, whether involuntary or chosen, is absent.
· Composition effect not corrected — it is the main limit of any wage average.
Annual net wage — first decile (D1)
the bottom of the wage scale — a threshold, not the average wage of the lowest paid
Producer
INSEE
Official series
010752336
Unit
€
Scope
First decile (D1) of the annual net wage in full-time equivalent, private sector, France — 10% of employees earn less — annual value published as is
Latest year in the catalogue
2024
Read by the study up to
2024
Revision status
FINAL
Retrieved on
2026-09-29
Transformation
served as is
Does not measure — the wage of the lowest-paid 10%, which is LOWER than this threshold; nor the public sector, excluded from the scope
· Full-time equivalent: part-time work is neutralised by construction, and therefore invisible.
· Private sector only.
· NET annual wage, bonuses included — its comparison with the SMIC requires a declared conversion.
· Data collection changed during the period (switch to the nominative social declaration, DSN). The producer ensures the splice and publishes a continuous series; the study does not redo it and cannot check it year by year.
Median annual net wage (D5)
the exact middle of the wage scale
Producer
INSEE
Official series
010752342
Unit
€
Scope
Median (D5) of the annual net wage in full-time equivalent, private sector, France — half of employees earn less — annual value published as is
Latest year in the catalogue
2024
Read by the study up to
2024
Revision status
FINAL
Retrieved on
2026-09-29
Transformation
served as is
Does not measure — the average wage, systematically higher because high wages pull it up
· Full-time equivalent, private sector.
· NET annual wage, bonuses included.
Annual net wage — ninth decile (D9)
the top of the wage scale — an entry threshold, not the level of very high wages
Producer
INSEE
Official series
010752348
Unit
€
Scope
Ninth decile (D9) of the annual net wage in full-time equivalent, private sector, France — 10% of employees earn more — annual value published as is
Latest year in the catalogue
2024
Read by the study up to
2024
Revision status
FINAL
Retrieved on
2026-09-29
Transformation
served as is
Does not measure — the dispersion ABOVE the ninth decile, far greater than below it and invisible here
· Full-time equivalent, private sector.
· An interdecile ratio says nothing about what happens beyond the ninth decile.
Monthly net SMIC (annual average)
the amount actually paid to the employee on the SMIC
Producer
INSEE
Official series
000883658
Unit
€
Scope
Monthly SMIC net of social contributions, CSG and CRDS, for 151.67 hours, as an annual average — France — annual value published as is
Latest year in the catalogue
2025
Read by the study up to
2024
Revision status
FINAL
Retrieved on
2026-09-29
Transformation
multiplied by twelve to be compared with the annual thresholds of the distribution
Does not measure — the disposable income of a household on the SMIC, which includes benefits and the activity bonus (prime d'activité) and may depend on them to a significant degree; nor employer cost; nor income after tax
· Published from 2005 only.
· Without bonuses or overtime, whereas the observed wages include them: the ratio of the SMIC to the distribution thresholds therefore OVERSTATES the real position of the floor.
· The definitions of “net” in the two sources are close without being guaranteed identical.
Purchasing power of the average net wage (base 100 in 1951)
the only series deep enough to speak of generations
Producer
INSEE
Official series
010752375
Unit
Index
Scope
Index of the purchasing power of the average annual net wage of full-time jobs in the private sector, base 100 in 1951 — deflated by INSEE — annual value published as is
Latest year in the catalogue
2024
Read by the study up to
2024
Revision status
FINAL
Retrieved on
2026-09-29
Transformation
served as is
Does not measure — the wage of the same person over a career, nor that of a generation: the average moves with the composition of employment, which has changed beyond recognition since 1951
· ALREADY deflated by the producer. The study does not deflate it again, and the cross-reference register forbids superimposing it on a price index.
· The “private sector” scope in the label is accurate only from 1983: until 1982, the series carries values strictly identical to those of the “all employees” version (idbank 001665161), and the two diverge afterwards by at most 3.5 index points. The postwar doubling therefore belongs to the wider scope.
· COMPOSITION effect not corrected, and it is major over 74 years: the shift to services, feminisation, longer education. Part of the doubling of the Trente Glorieuses (the 1945-1975 postwar boom) measures a shift in the structure of employment, not individual raises.
· Average, not median. Full-time posts in the private sector.
Unemployment rate (ILO)
the labour market, unemployment side
Producer
INSEE
Official series
Taux de chômage au sens du BIT par sexe et âge (T302)
Unit
%
Scope
All of France, including Mayotte
Latest year in the catalogue
2026-Q2
Read by the study up to
2025 (closed year)
Revision status
PROVISIONAL
Retrieved on
2026-08-08
Transformation
served as is
Does not measure — the number of registered jobseekers, which is an administrative definition and not the International Labour Office definition
· Population aged 15 or over — a different scope from that of the employment rate used here.
· Labour Force Survey: whole-of-France scope, Mayotte included, with back-casting by the producer.
· The Labour Force Survey (Enquête Emploi) has undergone at least five overhauls or changes of scope over the window: the switch to the continuous survey in 2003, Eurostat alignment on age and trainees in 2011, a change of questionnaire in 2013 — which INSEE writes produced breaks in the unemployment-rate series —, extension to the overseas departments in 2014, a major overhaul of the questionnaire and protocol in 2021 under the European regulation on social statistics, extension to Mayotte in 2024. The producer back-casts the series each time; the comparability uncertainty that remains is not quantified by it, and the study does not quantify it either.
· No gap between end points smaller than 1 point can support a conclusion over this window.
· In 2020, the definition of the International Labour Office (ILO) classes as inactive people prevented from looking for a job, and counts as employed employees on short-time working. The 2020 year of the three labour-market series is therefore not comparable with the others, and it pushes the measures in both directions at once.
Employment rate (aged 15–64, %)
the labour market, employment side
Producer
INSEE
Official series
Taux d'emploi par sexe et âge quinquennal (T207)
Unit
%
Scope
All of France, including Mayotte
Latest year in the catalogue
2026-Q2
Read by the study up to
2025 (closed year)
Revision status
PROVISIONAL
Retrieved on
2026-08-08
Transformation
served as is
Does not measure — the volume of hours worked, nor the quality of jobs, nor their duration
· Population aged 15 to 64 — a different scope from that of the unemployment rate used here.
· The Labour Force Survey (Enquête Emploi) has undergone at least five overhauls or changes of scope over the window: the switch to the continuous survey in 2003, Eurostat alignment on age and trainees in 2011, a change of questionnaire in 2013 — which INSEE writes produced breaks in the unemployment-rate series —, extension to the overseas departments in 2014, a major overhaul of the questionnaire and protocol in 2021 under the European regulation on social statistics, extension to Mayotte in 2024. The producer back-casts the series each time; the comparability uncertainty that remains is not quantified by it, and the study does not quantify it either.
· No gap between end points smaller than 1 point can support a conclusion over this window.
· In 2020, the definition of the International Labour Office (ILO) classes as inactive people prevented from looking for a job, and counts as employed employees on short-time working. The 2020 year of the three labour-market series is therefore not comparable with the others, and it pushes the measures in both directions at once.
· The rise of this aggregate over the window is largely a fact of demography and retirement — women's participation, the raising of the retirement age — and not a fact of the bottom of the wage distribution. It tells little about the labour market in the neighbourhood of the SMIC.
Youth employment rate (aged 15–24, %)
the labour market, the population most exposed to the minimum wage
Producer
INSEE
Official series
Taux d'emploi par sexe et âge quinquennal (T207)
Unit
%
Scope
All of France, including Mayotte
Latest year in the catalogue
2026-Q2
Read by the study up to
2025 (closed year)
Revision status
PROVISIONAL
Retrieved on
2026-08-08
Transformation
served as is
Does not measure — youth unemployment: an employment rate relates employed young people to ALL young people, students included, whereas an unemployment rate relates only to young people present on the market
· Population aged 15 to 24, overwhelmingly in education: the employment rate moves there as much with the length of studies as with the economic cycle.
· The Labour Force Survey (Enquête Emploi) has undergone at least five overhauls or changes of scope over the window: the switch to the continuous survey in 2003, Eurostat alignment on age and trainees in 2011, a change of questionnaire in 2013 — which INSEE writes produced breaks in the unemployment-rate series —, extension to the overseas departments in 2014, a major overhaul of the questionnaire and protocol in 2021 under the European regulation on social statistics, extension to Mayotte in 2024. The producer back-casts the series each time; the comparability uncertainty that remains is not quantified by it, and the study does not quantify it either.
· No gap between end points smaller than 1 point can support a conclusion over this window.
· In 2020, the definition of the International Labour Office (ILO) classes as inactive people prevented from looking for a job, and counts as employed employees on short-time working. The 2020 year of the three labour-market series is therefore not comparable with the others, and it pushes the measures in both directions at once.
· The contract retained this series as the one most exposed to the minimum wage. That is true under an ordinary regime and false since 2020: an apprentice is paid as a percentage of the SMIC, and is therefore exempt from the floor, and the growth of apprenticeship shifts this employment rate through a channel that is not the SMIC. The movement of +5.8 points between 2020 and 2022 is moreover superimposed on the 2021 overhaul and the rebound following the health crisis; the study cannot separate these three effects with the series at its disposal.
A sixth series, inflation, serves as control witness — never an input to the calculation. Years 1991-1996 DERIVED from idbank 001764363, that is, from the index retained as the deflator. The comparison is independent only from 1997, the first year that INSEE publishes under the 2025 base (idbank 011814640). Series not served before 1991. Series published rounded to the tenth of a point (envelope ±0.05), whereas the implied inflation of the deflator is calculated to the hundredth. The gaps observed fall essentially within this envelope: the witness is a non-regression check, not a validation.
Premium
Everything this study asserts, and what bounds each claim
The 10 frozen claims of the study, each with its nature, its value, its period, its scope, its source, its formula, its revision status, its confidence and its limits. None is of a causal nature.
Reading aids
The terms actually used in this study, and only those. Each is also explained where it is used: a word underlined with dots in the text opens its definition on hover or by keyboard, and leads here with one click.
The legal minimum gross hourly wage in France. An employer cannot pay below it, and the amount is set by decree.
Not to be confused with — It is neither the average wage, nor the median wage, nor a disposable income. A minority of employees receive it.
Gross
The amount before social contributions. This is what the law sets, and therefore what the official series publishes.
Not to be confused with — Take-home pay moves differently: contribution rates changed several times over the period, and this study does not track them.
Index
A series rebased to 100 in a reference year. It no longer gives the level; it gives how many times the level has been multiplied since that year.
Not to be confused with — An index of 145 does not mean “145%” or “€145”: it means “1.45 times the level of the base year”.
Constant euros — or purchasing power
An amount from which the rise in prices has been removed, so that two years can be compared on an identical basket.
Not to be confused with — It is not an amount you can read on a payslip: it is a current amount divided by a price level.
Inflation
The CHANGE in the price level over one year, expressed as a percentage. 2% inflation means prices are 2% higher than a year earlier.
Not to be confused with — Inflation is not the price level. When inflation falls from 5% to 2%, prices keep rising, just more slowly. An inflation rate is never subtracted from an amount in euros.
ILO unemployment rate — ILO: International Labour Organization
The share of people without a job, available for work and actively looking, relative to the labour force. The figures served here are FRENCH: they come from INSEE's Labour Force Survey (enquête Emploi), which applies to French households the statistical definition set by the International Labour Organization. The ILO provides the measurement rule, not the measurement.
Not to be confused with — It is not the number of registered jobseekers, which follows an administrative definition. The two do not move together.
Employment rate
The share of people in work, relative to the whole age group — not only to those looking for a job. Measured, like the unemployment rate, on French households by INSEE's Labour Force Survey, according to the International Labour Organization definition.
Not to be confused with — An unemployment rate can fall because people found a job, or because they stopped looking. The employment rate has no such ambiguity; this is why the two are served together.
Correlation coefficient — denoted r
A number between −1 and +1 that measures how far two series rise and fall together. The SIGN gives the direction: positive, they move the same way; negative, in opposite ways. The ABSOLUTE VALUE gives the strength: 0 means no association, 1 a perfect association.
Not to be confused with — It is neither a percentage, nor a multiplier, nor an effect. An r of 0.5 does not mean “50%”, and it does not say by how much the second series moves when the first one moves. A correlation also never says which of the two acts on the other, nor whether a third factor causes both. No method in this study identifies a causal link.
Quintile
A 20% slice. Ranking all households from the least to the most well-off and cutting into five equal parts gives five quintiles; the first quintile is therefore the bottom 20% of the distribution of standards of living.
Not to be confused with — It is not “the poor” in the sense of a poverty line: it is a rank in a ranking, and the rank exists whatever the actual standard of living of the people who occupy it.
Decile — D1, D9
The same idea in ten parts. D1 is the threshold below which the 10% lowest paid are found, D9 the one above which the 10% best paid are found. The D9/D1 ratio sums up the spread of the wage scale in a single number.
Not to be confused with — D1 and D9 are THRESHOLDS, not groups or averages. The D9/D1 ratio says nothing about what happens above D9, where dispersion is far greater.
Median
The value that splits a population into two equal halves: half are below, half above.
Not to be confused with — It is not the mean. A handful of very high wages pulls the mean upwards without moving the median by a euro — this is why the median describes “the middle” better.
Deflator
The price index by which an amount is divided to remove the rise in prices. This is the operation that turns current euros into purchasing power.
Not to be confused with — The choice of deflator changes the result. This study uses the price index excluding tobacco, and says so everywhere the figure depends on it, rather than filing it in a footnote.
Slump
A period during which the purchasing power of the SMIC stays below the best level it has already reached. It opens at the last peak, falls to a low point, and closes in the month when that peak is regained.
Not to be confused with — A slump is not a fall in the amount: the nominal SMIC never fell over the window. It is a lag behind prices, and it is made up.
Percentage point
The unit of a gap between two percentages. Going from 8% to 7% is a fall of ONE POINT, not of 1%.
Not to be confused with — Confusing the two changes the order of magnitude: going from 8% to 7% is a fall of 12.5% in relative terms, and of one point in absolute terms. This study always writes “point”.
Correction for the number of tests — or multiplicity correction
When several tests are run instead of one, the chances of stumbling on a link that does not exist rise mechanically. The probability obtained is therefore corrected by multiplying it by the number of tests run, and only what survives this correction is retained.
Not to be confused with — A link that does not survive the correction is not “weak”: it cannot be distinguished from what chance would produce, and therefore cannot carry anything at all.
What this study contains
Sold as a single purchase, edition by edition, when sales are open.
Accessible to everyone
The answer to the question asked, in one sentence, and what bounds it: what the study does not settle, and what would invalidate it
Three dated benchmarks: the gap to the best level already reached at the latest published month, the long-run gain over the closed years 1996–2025, and the share of time spent below a level already reached
1 interactive chart as an example — the saw-tooth of purchasing power, with the high, the low and the average of each year
The exact formula for purchasing power, written to be copied, and the 4 sentences the study forbids itself to write
The 4 quantities that the certified foundation does not carry, and what their absence forbids concluding — including the share of employees paid at the SMIC
The actual calendar of SMIC upratings since 1996, recorded month by month from the producer up to the latest step, and what it explains about the saw-tooth shape
The 13 series used, each with its producer, its official identifier, its scope, the period the study reads from it, its revision status and what it does not measure
The three comparisons with INSEE publications: the official SMIC index in constant euros, the published claim on multipliers, and the check of the deflator over 29 independent years
Reading aids: index, constant euros, ILO unemployment rate, employment rate, defined on the page
The reference to copy, the version, the data cut-off date and the method — to cite this study in a note
Reserved for the paid study
The full answer in three steps, with its figures — the long-run gain and its pace, the slumps and their catch-ups, the comparison with the basic wage of blue-collar workers —, the mechanism that explains it, then the open questions and the invalidation criteria in full
The quantified benchmarks of the demonstration, each with its reading and what it does not mean: slumps, prices for low-income households, the wage scale, the long-run measure and its robustness, the labour market, indexation to prices
The 10 frozen claims of the study, each with its nature — fact, calculated result or association —, its value, its period, its scope, its source, its formula, its revision status, its confidence and its limits
8 premium charts, each with what to see in it, what it does not mean, its limits and the reading of the selected year
The 29 annual changes broken down between the nominal progression of the SMIC and inflation, including the 6 years in which purchasing power actually fell
The 7 periods of continuous progression or decline, the rule that cuts them, and for each the change in unemployment over the same interval — this is where the second half of the question is decided
Each slump in purchasing power since 1996: the month of the last peak, that of the lowest point, the loss reached, the date on which the peak was regained and the number of months spent below it
The 9 association coefficients between the real progression of the SMIC and 3 labour-market indicators, at three lags, each with its probability corrected for the 9 tests run and the interval obtained by removing one year at a time — this tells whether a coefficient holds for the whole or for two years
The test of the 11 starting years on the real gain, with for each the cumulative figure and the annual pace — the answer to “you picked a base that suits you”
The same robustness applied to the labour market: 21 starting years, and for each the unemployment gap and the two employment-rate gaps up to 2025 — including the starts that reverse the sign
A study is read once and serves for a long time: charging for it individually costs less than a subscription for someone who reads only one.
No commitment, no renewal, no collection of a payment method as long as payment is not open.
No subscription plan is offered: studies will be paid for one at a time.
Automatic alerts are not available yet.
Scenario customisation is not offered.
Cite this study
France ÉcoScan v9.0.14, “The purchasing power of the SMIC since 1996”, edition 2026.2, 2026-09-15.
Year 2025 · monthly readings up to August 2026.
France ÉcoScan is neither a statistical institute nor an accreditation body. This study calculates from series published by INSEE and compares its results with INSEE publications; it is not validated by any third party.