France ÉcoScan strategic study
Why it keeps rising, and what the return of higher financing rates really changes
This analysis remains tied to the data that allowed its conclusion. No figure displayed here is real-time. New data may lead to a refresh, a review or a new version — never a silent change.
The rise in French debt since 1996 is mainly not explained by interest rates but by primary deficits — the years in which revenue did not cover spending excluding interest —, and what is changing today is not its origin but its price, which spreads as the debt is renewed.
Two forces long worked in opposite directions. On one side, repeated primary deficits added debt every year. On the other, growth in current euros above the cost of debt diluted it as it went, while the replacement of old securities by cheaper ones lowered its burden without any decision contributing. The second force masked the first — it did not cancel it. Since 2023, it has changed direction: new financing costs more than the stock average. The extra cost is not seen straight away because it applies only to the part of the debt that is renewed each year. The constraint ahead is therefore not a shock, it is a slow transmission that is already under way.
Full analysis —the quantified diagram of the debt already borrowed against new financing, with the share of the stock renewed each year.
This reading would be invalidated if the State were again to borrow more cheaply than its already contracted debt costs it — refinancing would then become a relief again — or if the economy were to grow, for long enough, faster than the cost of debt, reopening a lastingly favourable gap. It would finally be invalidated if the weight of past deficits ceased to dominate, depending on the year from which one chooses to count. This last test is carried out over 7 different starting years; its quantified result, and what does not withstand it, are in the full analysis.
Sources and scopes
Three producers, and their scopes do not overlap. INSEE (Institut national de la statistique et des études économiques, France's national statistics institute) publishes the national accounts: debt, deficit and gross domestic product come from it, across all of general government — the State, local authorities and social security combined. Eurostat harmonises these same accounts at European level, which makes France comparable with its neighbours. AFT (Agence France Trésor, the State's debt management agency), finally, borrows on behalf of the State and publishes the conditions on which it does so — but for the STATE ALONE, a scope narrower than that of general government. This study never mixes the two.
The most frequent confusion on this subject is to mix what has happened, what is under way, and what might come. Each of these three timeframes is read with its own rules, and the third is not a forecast.
1996 – 2022
1 — What happened
Interest rates have not been the main engine of debt accumulation over the last thirty years: their net effect, once growth in current euros is deducted, amounts to a little over one point of gross domestic product over the whole period. This near-balance does not mean that the cost of debt stayed below growth in most years — it exceeded it more often than it fell short. It comes from a few years of very strong dilution which offset all the others.
since 2023
2 — What is changing
The mechanism is reversing. Since 2023, new financing has cost more than the stock average; the average cost, for its part, reacts only slowly, because only a fraction of the debt is renewed each year. Meanwhile, the stock is at its highest and the primary balance remains in deficit. The gap has widened further: the ten-year yield, 4.00% on average in the latest month published, stood at 4.67% at the reading of 24 September 2026.
2026 – 2035, under assumptions
3 — What may follow, under conditions
With market conditions unchanged, the average cost keeps rising through renewal alone — without any rate needing to rise further. What this produces on the ratio then depends on four quantities: growth, prices, the primary balance, and the price of new borrowing. The trajectories presented later vary the first three; three readings of the rate isolate the fourth, one point above and one point below the level auctioned.
Under assumptions — no forecast
The amounts that follow translate percentages. This is useful for picturing orders of magnitude, and misleading if one forgets that the conversion key changes every year with gross domestic product — which is why this study reasons in points, and converts only when it helps.
Scale reference
In 2025, one percentage point of gross domestic product is worth about €29.9 billion. Half a point is therefore worth around €15 billion, and three points around €90 billion. This conversion changes every year with gross domestic product: it does not apply to the other years of the study.
Order of magnitude 2025 — the conversion follows gross domestic product, so it changes every year.
Source — derived — nominal gross domestic product published by INSEE
The weight of interest payments
€66.6bn a year
Observed · 2025
That is 2.2% of gross domestic product. It is an ANNUAL CHARGE: what general government actually paid in interest that year, on its entire debt.
What this is not — This is not what the debt would cost at today's rate, nor the amount that would have to be repaid: the principal is rolled over without being repaid.
Source — Eurostat — excessive deficit procedure table, interest paid (D.41)
The State's financing requirement
€310.0bn for 2026
State only — budgetary and treasury accounting · voted in the finance act
Observed · fiscal year 2026, as set by the finance act
Securities fall due: the State repays their holders and they leave the outstanding amount. To finance this repayment, it issues new securities, at the rate of the day. On this part, the stock is unchanged — its PRICE, however, changes.
A new requirement of the fiscal year: it replaces no maturing security. It is the GENERAL BALANCE OF THE STATE BUDGET, distinct from the public deficit in the Maastricht sense, which covers all of general government and is read in the national accounts.
Assumption of SNCF Réseau's debt and miscellaneous treasury needs — the last two lines of the table, grouped here because neither carries a reading issue on its own.
The financing requirement adds up operations of different natures: REPLACING securities that have fallen due — the State repays them, then issues new securities in their place — and FINANCING the deficit of the year. This whole sum is therefore not new debt: of the €310.0bn, €169.9bn replace securities that expire at the same moment, €134.6bn finance the deficit of the year, and €5.5bn cover the assumption of SNCF Réseau's debt and treasury needs. The whole is financed by €310.0bn of medium- and long-term issuance net of buybacks, +€2.0bn of net change in short-term securities and €−2.0bn of other treasury resources: these last two lines have opposite signs and cancel out, so that total financing comes back to the €310.0bn of the requirement.
What this is not — This is not the financing requirement of all of general government: neither local authorities nor social security are included. Nor is it an amount of new debt — repayment replaces securities that disappear at the same moment. Finally, SHORT-term securities enter only through their net change over the year: their outstanding amount is refinanced continuously within the fiscal year, and this rollover is not a financing requirement.
Source — Finance Act for 2026 (Law No. 2026-103 of 19 February 2026), Article 147, II, 1°
Reading aid
Teaching example — round figuresAn example with round figures, unrelated to any French year. Suppose €50 billion of securities fall due during the year. The State repays these securities to their holders and they disappear from the outstanding amount: the stock falls by the same amount. To finance this repayment, the State issues new securities for €50 billion, at the rate of the day: the stock rises by the same amount. On this part, the stock is therefore unchanged — but the rate, coupon and maturity of the new securities may differ from the old ones. The deficit of the year, assumed here to be €10 billion, replaces nothing: it adds on. In total, the stock rises by €10 billion, not by €60 billion.
Teaching example with round figures — these are not French amounts, and the real maturity profile is not spread evenly across the year.
The ten-year rate the newspapers talk about
4.00%
Monthly average, August 2026 — not a rate of the day · annual average 2025: 3.37%
Observed · August 2026 — latest month published
The price at which French securities with a maturity close to ten years trade between investors. It applies neither to the stock already borrowed, nor even to all of the year's issuance: it indicates market conditions, not the State's bill.
What this is not — A monthly average does not say whether a threshold was crossed on a given day: the daily ten-year reading is served, dated, in the marginal-cost section.
Source — European Central Bank, convergence series — compiled by Banque de France, transmitted to the ECB
Full analysis —the quantified gap between the price of new borrowing and the average cost of the stock, and what a loan costs at today's rate.
What the interest burden reduces
The interest burden is a constrained expenditure, inherited from past financing: it is due whatever the budgetary choices of the year. With an unchanged budget, each euro paid in interest is a euro that does not finance a new public policy, does not reduce the deficit and does not lighten any levy.
That euro does not disappear for all that: it pays the holders of the debt, residents and non-residents alike, and stays in the economy. What it reduces is not the country's wealth — it is the room for decision the public budget has today.
Why we do not compare this €66.6bn with a ministry's budget
This €66.6bn covers the interest paid by all of general government, in national accounting. A ministerial budget, for its part, corresponds to the appropriations voted for the State, in budgetary accounting: a narrower scope and different conventions. Placing the two amounts side by side without restatement would give a spectacular and accounting-wise misleading comparison.
France ÉcoScan prefers to forgo the comparison rather than mix two scopes — even when it would be striking.
A rise in rates does not make everything the State owes more expensive at once. It makes new securities more expensive, then those that fall due, then refinancing — gradually. The cards that follow separate what is owed, what is borrowed, what new borrowing costs that day, and the speed at which one reaches the other. Each carries its status.
The ten-year rate — latest value published
4.67% · 24/09/2026
TEC 10 index recorded on 24 September 2026 — one day, not an average
Observed · daily reading of 24 September 2026
last month published in the previous edition of the study (15 August 2026)
last month published by the European Central Bank
+0.48 percentage points up to the reading of 24 September 2026
The French ten-year yield tightened by +0.82 percentage points between the average of July 2026 and the reading of 24 September 2026. It is the price at which the secondary market trades ten-year securities that day: it sets the cost of the NEXT issues, not that of the outstanding stock.
What this is not — It is neither the rate on French debt nor what the State pays: debt already issued keeps its coupons. A daily reading can also ease as fast as it tightened; the study serves it dated, and makes it say neither a trend nor a threshold.
Source — Agence France Trésor — TEC 10 index of the day; European Central Bank — monthly averages
The curve as it was paid
3.86% · 4.37% · 4.55%
Last auction at ≈ 10, ≈ 20, ≈ 30 years — excluding syndications
Observed · auctions of 6 August 2026 and of 4 June 2026
€1.996bn allotted on 6 August 2026 · 9.8 years of residual maturity · price 79.06
€2.040bn allotted on 6 August 2026 · 17.8 years of residual maturity · price 52.68
€1.090bn allotted on 4 June 2026 · 31.0 years of residual maturity · price 97.56
In its latest sessions, the State paid 3.86% to borrow at ten years and 4.55% at thirty years: the curve is STEEP, and it is on its long end that the repricing is most marked. These are prices paid, line by line — not quotations.
What this is not — Three lines do not make a full curve, and two of them do not date from the same day. An auction rate is a yield, never a coupon: the ≈ 20-year line goes far below par because its coupon is 0.5%.
Source — Agence France Trésor — OAT auction history, line-by-line detail
Full analysis —the maturity schedule of medium- and long-term securities over the next five years, and the grid stock · flow · price of new borrowing · pass-through.
The average cost of the stock is not assumed: it is derived from the price of new borrowing and the speed of renewal. Three rate paths, all else equal — today's held, one point higher, one point lower — and what they do to the ratio over ten years.
Premium —the 3 readings of the rate — price of new borrowing held, one point higher, one point lower — with the arrival ratio of each, are part of the paid analysis.
To understand the effect of rates, two things must be distinguished: the debt already accumulated — the stock — and the financing that renews it each year — the flow. The rate of the day applies to the second, never to the first.
THE STOCK
Everything that has already been borrowed and is still outstanding: €3,595.5 billion of Maastricht debt for all of general government at the end of Q2 2026 — INSEE's latest quarterly observation —, after €3,460.1 billion at the end of 2025, the last closed fiscal year. The largest share of it is the State's negotiable debt — €2,903.8 billion at 31 August 2026 — made of securities that the State issues in two kinds: medium and long term, the OAT (obligation assimilable du Trésor, a medium- or long-term French government bond); short term, the BTF (bon du Trésor à taux fixe et à intérêt précompté, a short-term French Treasury bill). Each carries the rate at which it was issued, sometimes ten or fifteen years ago. The rest — local authorities, social security, other bodies — finances itself differently.
THE FLOW
What is borrowed this year — the STATE'S FINANCING REQUIREMENT, €310.0 billion for 2026, as the finance act sets it. It adds up two things of a different nature: the repayment of securities that have fallen due (€169.9 billion in medium and long term), which leaves the stock unchanged, and the financing of the State budget deficit (€134.6 billion), which adds to it. It is to this flow, and to it alone, that the rate of the day applies. Mind the scope: this State budget deficit is not the public deficit of €151.6 billion measured across all of general government in 2025 — two correct quantities, two different scopes, never to be added together.
Hence the least intuitive fact on the subject: when the rate at which the State borrows rises, the burden of debt does not rise by the same amount, nor immediately. It rises by the share of the stock that is renewed each year, and takes several years to carry most of it.
The image has a limit that must be stated before using it: a reservoir empties uniformly, whereas a debt is renewed in tranches of unequal ages. The speed of pass-through is therefore estimated, not observed: its two calibrations and the resulting half-life are in the full analysis. It is calculated, in addition, on the State's negotiable debt, the only one whose outstanding amount and average life are published, then applied to all of general government: the real pass-through is probably a little slower than the one described here.
Each tab carries a plot, what to see in it, the reading of the selected year and the conclusions it does not allow.
How to explore these charts
Choose a tab, then hover over the plot or select a year. The plot, the values and the associated interpretation update together. On the keyboard, tabbing brings focus to the plot and the arrow keys move the selection.
How has French debt evolved since 1981?
Reading
Source — INSEE — general government accounts. Axis: The axis starts at zero. A debt ratio has an absolute scale that means something: truncating the axis would exaggerate inflections that the real slope does not carry.
Limits
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.
1996-2025
The widely shared intuition is that French debt comes from a snowball effect: interest would beget interest.
Over 1996-2025, primary deficits weigh much more heavily than the snowball effect in the cumulative rise of the ratio. The count by channel, in percentage points of gross domestic product, and what it does not prove, are in the full analysis.
Full analysis —the 4 quantified results, each with its period, its calculation and what it does not prove, and the breakdown of the cumulative rise between the three channels.
None of these trajectories is a forecast, and their range is not a confidence interval: each answers “if these assumptions, then this path”. They are produced by the same accounting identity as the historical decomposition, rolled forward.
On the charts, what has been measured is drawn as a solid line and what is assumed in dotted lines, separated by a “start of projection” marker. This is not a decorative convention: a trajectory drawn in a single stroke would be read as an observation, and the reader would have no way of knowing where what was measured stops.
Premium —the detailed table of the 5 scenarios, with their full assumptions, the arrival ratio, the peak point and its year, and the stabilisation verdict, is part of the paid analysis.
Limits to keep in mind — what these trajectories do not model
None is absurd and several are partly true. The point is neither to validate them nor to reject them wholesale, but to establish what these data allow one to say about them — and to name, each time, the part of the question that remains out of reach.
Analysed in detail in the full study
The paid analysis puts figures on both sides of the gap — primary expenditure and revenue, at the start and the end of the window —, measures the stabilisation effort that would have to be found, says to which year spending would have to return to achieve it, and sets out the three competing economic readings that these series cannot tell apart.
9 quantified figures and 5 stated limits. The paid analysis measures and explains why the question remains open — it does not settle it.
Analysed in detail in the full study
The paid analysis gives the level of revenue that would have to be reached with spending unchanged, compares it with the maximum ever observed in France and its year, shows how much this level moves depending on the assumed growth and cost of debt, and names the calculation assumptions that make this figure more fragile than it looks.
6 quantified figures and 5 stated limits. The paid analysis measures and explains why the question remains open — it does not settle it.
Analysed in detail in the full study
The paid analysis gives the nominal growth that stabilisation would require, compares it with the median and the ninth decile observed since 1996, counts the number of years in which France has reached it and the longest consecutive run — the quantity that really decides, since stabilisation requires holding a level, not reaching it once.
7 quantified figures and 2 stated limits. The paid analysis measures and explains why the question remains open — it does not settle it.
Quantified verdict in the full study
The paid analysis quantifies, over recent years, what prices and volumes each removed from the ratio, isolates the mechanical rebound of 2021 that distorts the reading, replays those years with a deflator brought back to its pre-shock median, and counts the points of debt that inflation actually avoided — before saying why this advantage is overstated.
12 quantified figures and 3 stated limits. The verdict exists and is quantified; it is reserved for the paid analysis.
Quantified verdict in the full study
The paid analysis relates the voted financing requirement — already given in the preview — to gross domestic product, separates medium- and long-term maturities from short-term debt that is rolled over, gives the share of the stock whose rate resets each year and the average life of negotiable debt — and then names the three things these series cannot see, including the structure of holdings.
8 quantified figures and 4 stated limits. The verdict exists and is quantified; it is reserved for the paid analysis.
The data contradict this claim. The average cost of the stock is 2.01% in 2025, whereas the composite rate of 2026 auctions reaches 2.75%: the gap is still 0.74 percentage points. With the issuance rate unchanged, the pass-through model estimates that the average cost would keep rising as the stock is renewed — which would make no sense if the rise had already passed through.
What the data establish
The gap between the cost of the stock and the price of new borrowing changed sign in 2023, after twenty-four years in which it was negative — a period during which France was replacing expensive debt with cheap debt and saw its interest payments fall without any decision being taken. That movement is over and is reversing. From 11.8 to 13.3% of the stock has its rate reset each year: with the issuance rate unchanged, the average cost would reach 2.39% in five years, that is 0.44 percentage points of GDP of additional interest payments — without any market rate having needed to rise further.
The figures that carry it
Each value carries its status: observed — published as is by a producer; derived — calculated here by a declared formula; counterfactual — engine output under an explicit assumption, which was never observed; bounded estimate — bounded for lack of exact published data; regulatory — written into a law that has been voted.
derived · Eurostat — interest paid by general government relative to the previous year's debt outstanding
observed · Agence France Trésor — auctions published as of 15 September 2026, excluding syndications
derived · difference of the two lines above — two scopes and two years: a direction, not a like-for-like subtraction
derived · same gap, calculated on the only year for which both terms are closed
derived · first year in which the issuance rate exceeds the cost of the stock
bounded estimate · two calibrations, the second of which borrows its weighting from the first — the gap bounds what remains unknown
bounded estimate · ln(0.5)/ln(1−ρ), at the two ends of the range of ρ above
counterfactual · pass-through model r(t) = (1−ρ)·r(t−1) + ρ·m, issuance rate held at the 2026 level
counterfactual · rise in average cost projected above, applied to the debt ratio of the latest year
What we cannot settle
The terms actually used in this study, and only those. Each is also explained where it is used: these cards exist to check a word in passing, without leaving the page or interrupting the reading.
INSEE— Institut national de la statistique et des études économiques
The French public institute that produces the national accounts: gross domestic product, the public deficit and the outstanding amount of public debt all come from it.
Eurostat
The statistical office of the European Union. It collects and harmonises the public accounts of the member states, which makes French debt comparable with that of its neighbours.
AFT— Agence France Trésor
The State service in charge of managing its cash position and its negotiable debt: it is the agency that borrows, every week, on behalf of the French State.
Not to be confused with — Its figures cover the STATE alone. They therefore include neither local authorities nor social security, and do not describe all French public debt.
general government
The set formed by the State, local authorities, social security and various public bodies. It is on this scope — wider than the State alone — that public debt is measured in the European sense.
OAT— obligation assimilable du Trésor
The French State's medium- and long-term debt security — from two to fifty years. It is the main form of negotiable debt: when people speak of “France's ten-year rate”, an OAT is what is meant.
BTF— bon du Trésor à taux fixe et à intérêt précompté
The State's short-term debt security, of one year at most. It finances current cash needs and is therefore renewed several times a year.
Not to be confused with — Because it is renewed quickly, it passes a change in rates through to the cost of debt much faster than a long bond — even though it weighs little in the total outstanding amount.
auction
The auction by which the State places its securities with investors, on a calendar fixed in advance. The rate that comes out of it is the price at which the State borrows that day.
Not to be confused with — Part of issuance goes not through auction but through SYNDICATION — a placement organised with banks. An average auction rate therefore does not cover all that the State borrowed during the year.
Maastricht debt
The common European definition of public debt: the GROSS, CONSOLIDATED debt of all general government, valued AT NOMINAL VALUE and fixed at 31 December. It is the figure divided by gross domestic product to obtain the famous ratio.
Not to be confused with — Three words do all the work. GROSS: the assets held by government are not deducted — net debt would give another level, and is not the reference for European surveillance. CONSOLIDATED: what government bodies owe each other is eliminated, otherwise the same euro would be counted twice. AT NOMINAL VALUE: securities count for what will be repaid, not for their price of the day — debt therefore does not fall when its market price falls.
stock and flow
The stock is everything that has already been borrowed and is still outstanding. The flow is what is borrowed this year — new debt and the replacement of securities that have matured. The rate of the day applies to the flow, never to the stock already built up.
Not to be confused with — This is the costliest confusion on the subject: when a borrowing rate rises, the burden of the debt does not rise by the same amount, nor immediately.
primary balance
What general government collects minus what it spends in the year, BEFORE paying the interest on past debt. It measures the effort of the year, stripped of the weight of history.
Not to be confused with — It is negative when government borrows for anything other than interest. Total deficit and primary balance differ in one thing only: interest payments.
average cost of the stock
What the whole of the debt already borrowed costs on average: the interest paid in the year, relative to the debt at the end of the previous year.
Not to be confused with — It is NOT the rate at which the State borrows today. It is the inherited average of those rates, and it takes years to approach them.
Technical name: implied interest rate, or apparent rate on the debt.
issuance rate
The average rate of the year's AUCTIONS, weighted by the amounts actually raised. It is the price of what the State is borrowing at the moment, on its own scope.
Not to be confused with — It applies only to what is borrowed this year, and not even to all of it: issues placed by syndication are absent, and the current fiscal year is not closed. The rest of the stock continues to carry the rate at which it was issued, sometimes ten or fifteen years earlier.
r − g
The gap between the average cost of the debt (r) and the growth of the economy in current euros (g). When negative, debt is diluted in an economy that grows faster than the debt costs; when positive, it grows heavier on its own, without any new spending having been decided.
Not to be confused with — The g is NOMINAL growth — volume and prices together. Substituting real growth for it changes the result by several points of debt over a decade.
snowball effect
What r − g produces on the debt ratio: the share of its change that comes from the interplay between the cost of debt and growth alone, independently of any budgetary decision of the year.
GDP deflator
The price measure that corresponds to gross domestic product: it covers everything the country produces. It is the only one consistent with the debt ratio, whose denominator is that same GDP.
Not to be confused with — It is not the consumer price index, which covers only what households buy. In 2022, the two measures diverge by more than three points.
stock-flow adjustment
The part of the change in debt that does not go through the deficit: purchases of financial assets, cash movements, issuance premiums and discounts.
Not to be confused with — We know exactly how much it is worth, and not what it is made of: its composition is not in our sources. This study says so rather than inventing a cause for it.
Technical name: debt–deficit gap.
percentage point of GDP
The unit in which this whole study is counted. Relating a sum to the size of the economy makes it possible to compare 1996 with today, and France with its neighbours, which an amount in euros never allows.
stabilising primary balance
The primary balance that would leave the debt ratio exactly where it is, neither higher nor lower.
Not to be confused with — It is neither a forecast, nor a target, nor a recommendation: it is an arithmetic condition, calculated under displayed assumptions, and it ignores the feedback effect of a budgetary adjustment on growth.
Each figure in this study carries its status, its source, its scope — and, for public accounts, its revision status.
Year provisional —The 2025 accounts are provisional: INSEE will revise them. All the figures that carry this year — the debt ratio, the outstanding amount, interest payments, the average cost of the stock and the starting point of the trajectories — may move.
Observed
Value published as is by its producer.
Derived
Value calculated by France ÉcoScan from published series, using a displayed formula.
France ÉcoScan scenario
Value produced by our engine under explicit assumptions. It is not a forecast.
Institutional projection
Forecast published by an institution, cited with its date. It bounds, it does not steer.
Illustrative order of magnitude
Illustrative arithmetic intended to give a sense of scale. It is not an exact accounting calculation.
| Producer | What comes from it | Scope and period |
|---|---|---|
| INSEE (Institut national de la statistique et des études économiques, France's national statistics institute) | Debt outstanding in the Maastricht sense, public deficit, nominal and volume gross domestic product | All of general government — annual accounts 1981 to 2025; quarterly debt up to Q2 2026 (published on 29 September 2026) |
| Eurostat | Interest paid, primary balance, public revenue and expenditure | General government — 1995 to 2025 |
| AFT (Agence France Trésor, the State's debt management agency) | Average auction rates, line-by-line detail, outstanding amounts and average life of negotiable debt, TEC 10 index of the ten-year yield | State only — 1999 to 2026; outstanding amount as of 31 August 2026; TEC 10 as of 24 September 2026 |
| European Central Bank | Market yield of ten-year government securities | Secondary market — annual averages through 2025, monthly averages through August 2026 |
The 2026 issuance rate and the 2026 financing table concern the current year: they are execution or legislative data, distinct from the closed annual accounts.
Reference to copy
France ÉcoScan v9.0.14, “France's public debt”, edition 2026.3, 2026-09-29, annual accounts closed at 2025, quarterly debt as of Q2 2026. https://france-ecoscan.fr/en/dossier/dette-publique
Method. Accounting identity of debt applied with no residual term, decomposition into three channels, a rate pass-through model with one parameter calibrated on published data and checked against a statistical fit, conditional trajectories whose cost of the stock is derived and not assumed. All sources are primary and cited with their identifier.
This study is not endorsed by any institution and is not certified by any outside expert. What is audited, versioned and reproducible are its sources, its calculations and its guards — not its authority.
Why pay per study
How the State borrows, in practice
The rates this study talks about come out of precise operations, held on a fixed date and published line by line. Here is one, with what it contains and what it does not prove.
The price of a long OAT (obligation assimilable du Trésor, a medium- or long-term French government bond), on a given day
3.95% and 4.37%
Auction of 6 August 2026 — two lines of the session
Observed · auction of 6 August 2026, settlement on 10 August 2026
€2.119bn allotted · weighted average price 98.72 — below par, so a coupon below the yield
€2.040bn allotted · weighted average price 52.68 — below par, so a coupon below the yield
4 long-term OAT lines, excluding syndications.
On 6 August 2026, the State placed €12.5 billion of long-term bonds. In this single session, the two longest tranches went at 3.95% and 4.37%. This is no average: it is the price at which those tranches found buyers that day.
What this is not — This does not mean that all French debt costs 4%. The average cost of what is already borrowed remains well below that, and most of the year's financing is done at shorter, cheaper maturities. An auction rate is moreover a YIELD, not a coupon: the weighted average prices above, on either side of par, show this directly.
Source — Agence France Trésor — OAT auction history, line-by-line detail
Coupon, price, yield: three different numbers
A single security carries three rates, and they do not say the same thing. This is why a bond showing 0.5% can have gone at 4.37%.
OAT 25 June 2037 — coupon 3.8%
OAT 25 June 2044 — coupon 0.5%
OAT 25 June 2037 — 98.72 per 100 of face value, below par
OAT 25 June 2044 — 52.68 per 100 of face value, below par
OAT 25 June 2037 — 3.95%
OAT 25 June 2044 — 4.37%
Hence the rule, and its reason: paying more than face value gives a yield LOWER than the coupon, because the repayment at €100 takes back the difference; paying less gives a yield HIGHER than the coupon, because that same repayment adds it. The security most generous in coupon is therefore not the most profitable for its buyer — it is simply the most expensive to buy.
Source — Agence France Trésor — line-by-line detail of the auction session and characteristics of the securities
Reading aid
How a government bond is issued, and what becomes of it afterwards
The Agence France Trésor announces, four working days ahead, the lines put up for auction and a range of amounts. The Spécialistes en Valeurs du Trésor (SVT, the primary dealers), the only ones authorised to bid, pass their price offers to the Banque de France, which forwards them to the Agency without revealing who made them. The auction is held “pay-as-bid” (“au prix demandé”): the Agency first serves the offers most advantageous to the State, up to the amount retained, and each bidder pays exactly the price it proposed. The yield follows from this — nobody chooses it. At the session of 6 August 2026, the two longest lines went at 3.95% and 4.37%. After issuance, the securities trade between investors on the secondary market: their price and yield move there every day, without creating any new debt whatsoever for the State.
1 · The Agence France Trésor announces
Four working days before the session
It designates the lines put up for auction and a range of amounts. It sets neither price nor rate.
2 · The Spécialistes en Valeurs du Trésor (primary dealers) bid
15 institutions
2025-2027 mandate — the only ones authorised to bid
Each proposes a price and a volume. Offers go through the Banque de France, which passes them on to the Agency without revealing who made them.
3 · The auction decides
At the price bid
Technique used since 1985
The offers most advantageous to the State are served first, up to the amount retained. Each bidder pays the price it proposed.
4 · The yield follows
3.95% and 4.37%
Session of 6 August 2026 — the two longest lines
Nobody decides it: it falls out of the price paid, the coupon written into the security and the time remaining to maturity.
5 · Then the secondary market
After issuance, between investors
Securities change hands; their price and yield move every day. A resale creates NO new debt for the State: it changes the holder, not the amount owed or the coupon.
Mechanism and calendar published by the Agence France Trésor; yields taken from the line-by-line detail of the session. Securities placed by syndication follow another route, reserved for launches and the least deep segments.