What this page shows
- The model
- The paper models the public debt of the seven G7 economies, with China modeled beside them, from 2026 through 2036. A government's borrowing cost includes a sovereign spread over the risk-free rate; the spread depends on how far the debt stands from a fiscal limit, the present value of the largest primary surplus the government can sustain; and the debt depends in turn on the borrowing cost. The price of the debt in a year is a root of that loop.
- The crisis equilibrium
- Over a range of debt, the crisis zone, the loop has three roots, two of them stable: a benign one and a crisis one at a higher yield, with an unstable one between them that the model never selects. Which of the two the market prices is a matter of coordination among investors, drawn in the model as a random event, the sunspot. Beyond the zone only the crisis root exists. An economy is on the crisis equilibrium in a year in which its debt is priced at the crisis root.
- The baseline and the two regimes
- The baseline is the path on which investors never coordinate on the crisis root. The paper follows two configurations of U.S. policy, each in force in every year of the horizon. Under regime M the U.S. primary balance follows its projected path, the automatic stabilizers, and a response to the sovereign spread, and the policy rate responds more than one for one to inflation. Under regime F the U.S. primary balance does not respond to the spread or the debt, the policy rate responds less than one for one, and the price level bears the adjustment. The other economies keep the rules of regime M under either, so their paths under the two regimes differ only through the U.S. crisis paths, which reach them through long-term interest rates and trade.
- Fiscal capacity
- The fiscal capacity is the largest primary surplus (the budget balance before net interest) a government can sustain. The paper reads it from each economy's own record, as the largest average primary surplus over 10 consecutive years the government has run, since 1950, or since the record's first year where it begins later, with receipts from abroad excluded. China's record shows no 10-year average of the primary balance above zero, and its capacity is scaled from its revenue.
- A share of paths
- The model is run 10,000 times, each run with its parameters and its annual shocks drawn at random within stated ranges. A share of paths is the percentage of those runs in which an event occurs in at least one year of 2027 through 2036. It is a share of the model's simulated paths under the stated assumptions, not a forecast and not a probability estimated from a record.
The central scenarios and the simulated paths, by economy
Each row of the strip is one run of the model for one economy: the baseline, regime M, and regime F. A dark cell is a year on the crisis equilibrium; a mid-gray cell, a year in which a crisis root exists and the economy is not on it; and a pale cell, a year with no crisis root.
Years on the crisis equilibrium
First year on the crisis equilibrium (years on it), and the share of simulated paths with at least one such year
Fiscal capacity: where it is read, and where a result turns on it
The capacity is read from each economy's long record of its primary balance, as the largest average over 10 consecutive years, receipts from abroad excluded. The chart shows the record for the economy selected, with the 10 years shaded.
The historical fiscal capacities
The largest capacity at which each economy spends a year on the crisis equilibrium, percent of GDP
One economy's capacity is set to each value of a grid, every other economy's at its historical value, and the model is run whole. A point is the largest capacity tried at which the economy spends at least one year on the crisis equilibrium in that run; above it, it spends none. Where the tick, the historical capacity, lies to the left of a point, the economy has a crisis year in that run.
The baseline and the limit
This section sets the model's baselines beside published projections and shows what the crisis equilibrium of the baseline and the central scenarios rests on (the candidate spread, the spread the pricing equation is solving for, in the rate at which the fiscal limit is discounted, the United Kingdom's crisis year of 2036 apart; and the discount rate's converging to the year's marginal rate as the stock of debt reprices) and what the timing and the reach of the crisis rest on (the pace of that convergence).
The baseline of the United States beside the Congressional Budget Office's projection, percent of GDP
The model's baseline is not fitted to the published projections; the distance is reported. The model's ratios are over calendar-year GDP and the CBO's over fiscal-year GDP.
Debt held by the public
Net interest
The baselines of the six beside the IMF's projected net debt
The model holds each of the six economies' primary balance at its 2026 value, where the published projection has a path of its own; the primary-balance column is what that difference contributes to the gap.
The crisis root exists beside the benign root because the fiscal limit is discounted at a rate that includes the candidate spread on the stock the market holds. The table reads the pricing equation (the loop of the overview) of one economy and year four ways; the readings with the limit held describe no economy. With the limit held at the value it takes at the crisis root, the crisis root is the only one; with the limit held at the value it takes at the benign root, or discounted without the candidate spread, the only root is at zero spread, the benign root.
First year on the crisis equilibrium (years on it): the model's limit beside the limit discounted without the candidate spread in every economy and year
First year on the crisis equilibrium (years on it) under regimes M and F: the model's limit, the limit at the rate on the stock, and the limit along the repricing path
The model's limit is the present value of the fiscal capacity, discounted at the financing rate of the limit less terminal trend growth. The financing rate of the limit, the marginal rate in what follows, is the year's real marginal financing rate at a neutral policy rate, which includes the candidate spread; discounting at it in perpetuity values the whole stock of debt as though it paid that rate from the year the limit is formed, while only the year's share of the stock, one over its average maturity, reprices in a year. The table sets the model's limit beside two limits that differ from it in the financing rate alone, the capacity and terminal trend growth being the same, under regimes M and F; the baseline is not in the table, and its crisis years are given below.
The limit at the rate on the stock discounts in perpetuity at the rate the stock carries once the year's share has repriced at the marginal rate, so that the candidate spread enters that rate with the weight of that share alone; under it no economy has a crisis year under either regime or in the baseline. The limit along the repricing path discounts at the rates the stock pays as it reprices, year by year, toward the marginal rate: the candidate spread, held in every later year, stands on the part of the stock that has repriced, and that part grows at the pace of the stock's maturity. Under it the crisis equilibrium remains, no onset is earlier, and no economy has more crisis years than under the model's limit: the U.S. crisis begins in 2029 under regime M and in 2030 under regime F, against 2027 under both, Germany has no crisis year, and Italy has 1 under regime M against 5. In the baseline the crisis years along the repricing path are those of the model's limit, France's and the United Kingdom's.
In the central scenarios the crisis equilibrium of the United States, France, the United Kingdom, and Japan therefore does not rest on valuing the whole stock at the marginal rate from the year the limit is formed, and in the baseline and the central scenarios the crisis equilibrium of every economy vanishes where the discounting holds in perpetuity the rate the stock carries after the year's repricing: in those scenarios it rests on the discount rate's converging to the marginal rate as the stock reprices. Its timing and its reach rest on the pace of that convergence.
The central scenarios, by economy
Each chart follows one quantity for the economy selected, under the baseline and the two regimes. Where the three lines coincide, the regimes do not differ for that economy in that year.
Across the simulated paths
A bar is a share of the simulated paths of the simulation selected, not a forecast and not a probability estimated from a record: the percentage of paths on which the economy spends at least one year on the crisis equilibrium, and the percentage on which it arrives there by coordination while a benign root exists.
Share of paths, 2027 through 2036, percent
How much of the variance of an outcome lies between the parameter draws
The line is the median path; the bands hold the middle half and the middle nine tenths of the paths in each year. They describe the spread of the model's paths under its draws.
Settings under which the crisis equilibrium vanishes: the resolution map
For each economy the map covers, the table gives the setting of one instrument, announced in the year shown and held, at which the crisis equilibrium vanishes from that economy’s path from that year on. Three instruments are solved one at a time: a fiscal capacity, a ceiling on the sovereign spread, and an increment to trend growth.
The same, as bars, by the year of the announcement
The combination: the smallest capacity that eliminates the crisis equilibrium at each spread ceiling announced in 2027 (ceiling in percentage points across; capacity in percent of GDP up)
The swing against the revenue and expenditure base: percent of GDP, percentage points of GDP, and percent of the WEO April 2026 general government revenue and outlays of 2027
The assumptions of the model
Each parameter with its value, its basis, and its treatment in the simulated paths: fixed, or drawn on the range shown.
Sources and limits
The paper
Fiscal Capacity and Sovereign Debt Crises in the G7, 15 November 2026. The paper states the model, its equations, its data, and the results this page draws from it; the footer names the values shown here that the paper does not print.
Public data sources
Limits
- A share is a share of the simulated paths on which the event occurs in at least one year of 2027 through 2036. It is a share of the simulated paths under the stated assumptions, not a forecast and not a probability estimated from a record.
- The results hold under the assumptions the Assumptions panel lists. Several of them rest on no record and are drawn on ranges of convenience; each is marked as such there.
- The settings in the section “Settings under which the crisis equilibrium vanishes” are announced and priced by the market and not delivered on the path on which the crisis equilibrium vanishes, and at the central calibration each must be held through the horizon and beyond it; they state what each instrument would have to be, not whether it is attainable. Each is reported at two decimals, on the side of the solved threshold on which the crisis equilibrium vanishes.
- The largest capacity at which an economy has a crisis year is found for one economy at a time, the others at their historical capacities, with the capacity set from 2027 and read as a counterfactual record rather than as a policy the market prices.
- The fiscal limits discounted without the candidate spread, at the rate on the stock, and along the repricing path are comparisons with the model's limit and not models of their own, and the limits held at the value they take at one root describe no economy; none of them, the model's limit included, is derived from a government's budget constraint with default.
- An estimated response of the primary balance to lagged debt with a cubic term, from which Ghosh, Kim, Mendoza, Ostry, and Qureshi build a debt limit, supplies none on any of three panels of the OECD's Economic Outlook database, 1963 through 2022: its cubic coefficient, on whose sign such a limit depends, is separated from zero at a level of 0.05 on none of them, and at its point estimate the function fitted on the panel of 21 economies on net financial liabilities yields no debt limit. The estimate shares a limitation of the fiscal capacity, which is read from each economy's record: both measure what governments have delivered and not what they could deliver.