A model to work through fiscal scenarios for Japan using the standard debt equation.
There are a few ways to measure the fiscal deficit, depending on which parts of the government are included (central government, local government, social security funds), and what exclusions are made. For debt dynamics, the measure is the difference between revenue on the one hand, and expenditure excluding net interest payments on the other. The IMF puts the gap at of GDP in 2025. That is still a much lower deficit than used to be typical in Japan, or is common in most OECD economies of today.
The current government's “Growth Strategy” promises to raise real GDP growth to near 2%. But the “structural reform” third arrow of the Abenomics project also promised stronger growth, and there was no change in the pre-Abe trend of a bit under 1%. According to the BOJ, Japan's potential real growth is just 0.7%. In the last 12M, the economy in real terms has grown by 0.5%.
What has shifted since Abenomics, and particularly since the covid pandemic, is inflation, and so GDP growth in nominal terms — the g in the debt equation. Inflation before 2020 averaged essentially nothing: the GDP deflator −0.5% a year over 2000–19, core CPI 0.1%. It then increased, and since 2022 has only rarely been below 2%. That is important, because 2% is the BOJ's inflation target.
The BOJ's main tool to control inflation is the policy rate, the uncollateralised overnight call rate. That is currently 1%. Over an economic cycle, that should settle at neutral, a level that can be thought of as r* plus inflation at the 2% target. That sounds simple, but in reality isn't. One complication is that r* isn't observable. The estimated rate can be used as a benchmark for where neutral is — but the BOJ has no fewer than six estimates of r*.
A second complication is that inflation expectations have historically been anchored below 2%. The BOJ has therefore kept monetary policy accommodative via a policy rate below neutral (and QE before 2024) to push inflation expectations up towards 2%. So policy does not simply jump to neutral. It converges on it, and how long that takes is a judgement; the path in between can sit either side.
Interest rates on government debt — in Japan being the yield on JGBs — are not set directly by the policy rate. Instead, yields are the combination of two things. First is the average expected short rate over the next ten years, which from the previous card you have set at . Second is the term premium, the amount investors demand for the risk of holding longer debt. In this calculator, that is defined at the 10-year level, and today stands at .
The MOF doesn't just issue 10-year debt. It sells debt as short as 3 months, and also super-long bonds of 30 and 40 years, giving an average maturity M in 2026 of 9.4 years. This number is important, because it drives a wedge between the marginal rate of interest that faces the government — the yield on JGBs, forecast from the policy rate and the term premium — and the average rate, the rate accruing on the whole stock of JGBs outstanding. M is generated as a function of the maturity of the existing stock, and of new debt issued in the future.
The combination of the policy rate, the JGB marginal rate and the maturity structure produces an estimated average gross rate on all government debt of in 2026, and on the assumptions you've made above, going forward.
r is measured net of all the interest the state receives, which is forced by the primary balance rather than chosen — see Which interest rate on the Checks page. Move the credit to zero and the projection reverts cleanly to a gross basis.
The debt stock can also move because of valuation changes, accrual and cash accounting, and the government's acquisition and the disposal of assets. The latter should be important in Japan, because the government has financial assets of over 140% of GDP. That is the reason some commentators think net debt — which includes these financial assets — is a better measure of Japan's true fiscal vulnerability.
In setting sfa, the question is how much of this 140% of GDP can really be sold. Almost half sit in the GPIF and other social security funds, but they need to be set against the contingent liabilities not yet included in d that will be incurred as the population ages. Around 30% of financial assets consist of equity claims on listed companies and unlisted corporations and reserves of local governments.
Neutral is where policy rests once inflation is already at target. It is not what you set while bringing inflation down. With inflation above target, policy has to sit above neutral for a period, and nominal growth is squeezed below trend while it does — then both return. Set the phase length to zero and the paths are the plain glides again.
Growth here is nominal and has to be set by hand: this page has no monetary transmission, so tightening does not slow the economy on its own. That is a limitation, not a result — the scenario is you asserting both legs, not the model deriving one from the other.
The main model says that the policy rate doesn't directly determine the rate of interest on government debt, which is true. But the true impact isn't quite as indirect as that statement suggests. In the BOJ's Quantitative Easing (QE) programme, it purchased huge quantities of JGBs from financial institutions, in exchange for reserves credited to their accounts with the central bank. The bank pays an interest rate on those reserves — which currently stand at of GDP — which rises with the policy rate.
The BOJ is now reversing QE via Quantitative Tightening, with the default settings in the model being: that the BOJ continues to run down its reserves for 10 years; that the end-state for reserves is 300tn, about 44% of today's GDP; that each 1% of GDP of run-off, by raising the supply of bonds that private investors need to absorb, raises the term premium by 1.5 basis points (0.015ppt); and that the premium converges on that level with a half-life of 8 years. All four can be tweaked here.
QT feedback assumptions
r is the net interest rate facing the government. That differs from the gross rate, because of the government's financial assets. Not all of those are held in fixed income instruments, but some are. The average interest rate earned by the government in 2024 was , enough to reduce the interest rate faced by the government from a gross to a net . This is calculated using Japan's SNA national accounts, which include the government's interest payments and receipts. The gap between them matches the difference between the overall fiscal balance and the primary balance in the IMF data, to 0.00ppt on average since 2000.
Interest rates on liabilities and assets, %
The fact that financial assets generate income further complicates the decision as to whether they can — or should — be sold. Selling an asset that generates a high return can lead to r rising, leading to a deterioration in debt dynamics. It also matters whether g > r at the time of a sale that reduces the size of d. If it is, then the excess of g over r is being multiplied by a smaller d, and so the change in the debt ratio is smaller than it otherwise would have been.
Since 2012, on the eve of Abenomics, the value of the government's financial assets has risen . A lot of that represents a one-off change as GPIF shifted its portfolio from low-yielding local assets to foreign equities and bonds. The value of these assets could continue to rise. Obviously, however, they could also fall. The model doesn't try to forecast valuations from here.
Revaluation of government financial assets, % of GDP
In the last 29 years, sfa has averaged +0.4ppt of GDP a year. That positive contribution isn't just valuations. Flow of funds data from the BOJ show the government has been a persistent net buyer of financial assets, adding to them in 19 of the last 27 years. So this is one important quirk in Japan's fiscal position: the government has been running a fiscal deficit and so borrowing money in part to finance the purchase of assets.
SFA and net acquisition of financial assets, % of GDP
The 10-year term premium is set as an input in the model. The 30-year rate is set using the 10s30s gap today, and used to define the convexity of the whole curve from 0. That is convenient, but the term premium at the long end of the curve isn't in reality fixed relative to the 10-year segment. This method of determining overall convexity also ignores the reality that the short end is currently steeper, as the market prices yet-to-happen policy rate hikes.
JGB rates by maturity, %
The model assumes that 1/M of the stock matures and is reissued each year, a geometric run-off. With the MOF extending maturity from 5.3 to 9.4 years since 2005, 1/M has fallen from 18.8% to 10.6%. That can be checked against the bonds MOF actually sells to refinance maturing debt, which have fallen from 21% to 12% of the stock over the same period. Directionally, then, the two are the same. The levels differ for three reasons. First, the published 9.4 years covers ordinary JGBs only, while refunding also refinances other debt too, principally FILP bonds. Second, during the pandemic a large part of the emergency funding was raised in Treasury bills, which sit outside the 9.4-year figure but enter the refunding programme once termed out into bonds. Third, the model does not track which bonds redeem when. Against MOF's published redemption schedule the real profile is front-loaded — 13.8% of the stock falls due in the first year against the model's 10.6%. The model therefore reprices around 3pp too slowly in the first two years, but is back in line by year six, so doesn't have a material impact — the peak effect on r is 0.06pp, and around 1ppt of GDP on the 2039 ratio.
Average maturity of the JGB stock, years
It is likely that a rising debt ratio pushes up the term premium further, but the model doesn't include such a mechanism. Nor does the model include the impact on the government's fiscal position caused by changes in the BOJ's financial position resulting from QT.
IMF Fiscal Monitor and World Economic Outlook. Cabinet Office national accounts and fiscal projections. MOF budget releases. BOJ flow of funds and accounts.