## The Coming Great Repression? New Measures and a Century of Evidence. Working Paper No. WP/2026/160

## Source details

**Canonical URL:** [The Coming Great Repression? New Measures and a Century of Evidence. Working Paper No. WP/2026/160](https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026160-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2026/english/wpiea2026160-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2026/english/wpiea2026160-source-pdf.pdf.json)

---

### Introduction — problem statement and contributions
- Fiscal outlooks in many advanced economies feature public debt at high peacetime levels; orthodox responses (fiscal consolidation and growth-enhancing reforms) can be slow and limited.
- Heterodox debt-reduction channels emphasized:
  - Surprise inflation (reduces real value when realized inflation exceeds what was priced).
  - Debt monetization (central bank financing through non-interest-bearing liabilities).
  - Financial repression (regulatory policies inducing intermediaries to absorb government debt at administered or below-market returns).
- Identification challenges:
  - Detailed historical composition and cost data for government debt are limited.
  - Counterfactual sovereign yield absent repression is unobservable.
- Paper’s three contributions:
  - Theoretically-grounded framework identifying repression via sectoral holdings; constructs two indicators (narrow fiscal and consolidated including reserves).
  - New historical dataset consolidating multiple sources on general government debt, central bank liabilities, and sectoral holdings across 17 advanced economies (JST coverage 1870–2020; analytical sample uses 1920–2020).
  - Model-based counterfactuals for sovereign borrowing costs absent repression and embedding implied yield wedge into a debt-decomposition framework; country applications quantify fiscal savings.
- Quantitative highlights:
  - United Kingdom: repression was the single largest deleveraging channel in the postwar decade.
  - United States: fiscal consolidation largest postwar, but repression also major.
  - Panel median annual fiscal savings from repression averaged around 0.57 percent per year, peaking sharply in the mid-1940s.
  - Median fiscal savings rise after the GFC, from about 0.08 percent in 2008 to about 0.51 percent in 2020.

### Concepts, data, and stylized facts — taxonomy and measurement
- Varieties of repression and quantitative indicators:
  - Fiscal repression (narrow): regulated intermediaries absorb government bonds; indicator: Government bonds as share of commercial bank assets.
  - Monetary repression (reserve channel): investors induced to hold central bank liabilities at administered remuneration; indicator: Commercial bank reserves created through central bank purchases of government bonds.
  - Consolidated repression (Fiscal + Monetary): public sector perimeter = General government + central bank; indicator: government bonds held directly and indirectly as share of commercial bank assets.
- Data construction and coverage:
  - Country sample: 17 advanced economies: Australia, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, United Kingdom, United States.
  - JST database span: 1870 to 2020; analytical dataset constructed to cover 1920–2020 to capture wartime episodes.
  - Key data sources: Abbas et al. (2011) and Arslanalp and Tsuda (2014) for sectoral composition (extended to 2020); BIS for central bank assets; Finaeon for currency in circulation; WID for household assets; Graf von Luckner et al. (2024) for restructurings coverage 1823–2024.
- Stylized facts:
  - Stylized Fact 1: Higher sovereign debt burdens are associated with lower, not higher, relative returns on government bonds.
    - UK: relative yields ≈ −4 percent per annum when public debt ≈ 250 percent of GDP in late 1940s.
    - US: relative yields ≈ −3 percent at the postwar debt peak.
    - Japan: as public debt rose toward 250 percent of GDP since the 1990s, relative returns collapsed into deeply negative territory.
  - Stylized Fact 2: Commercial banks hold a disproportionately large share of government debt precisely when relative returns on sovereign bonds are lowest and debt burdens are highest.
  - Stylized Fact 3: Government securities account for a larger share of commercial bank assets when debt burdens are higher and sovereign bond returns are lower relative to private assets; expansion happens via substitution from private lending.

### Theoretical framework — portfolio-demand identification of repression
- Agents: households (h), banks (b), non-financial corporations (c), government (g). Government issues nominal bonds B held by banks (B_b) and households (B_h).
- Portfolio choice:
  - Fréchet-distributed idiosyncratic taste shocks with shape parameters θ_i (i ∈ {b,h}).
  - Time-invariant shifters ̄T_i_j and time-varying regulatory shifters ̃T_i_j. Repression modeled as ̃T_b_g ≥ 1 for banks and ̃T_h_g = 1 for households.
  - Portfolio share formula:
    - s_i_g = ̃T_i_g (̄T_i_g · R^e_g)^{θ_i} / [̃T_i_g (̄T_i_g · R^e_g)^{θ_i} + (̄T_i_c · R^e_c)^{θ_i}]
  - Log-odds linearization yields identification relation for changes:
    - ∆ ln ̃T_b_g = ∆ ln ˜s_b_g − ∆ ln ˜s_h_g − (θ_b − θ_h)(∆r^e_g − ∆r^e_c)
- Normalization and interpretation:
  - Level of repression identified up to scale; normalization sets ln ̃s_b_g = 0 at each country’s minimum observed value.
  - Households assumed not subject to regulatory pressure; if households face non-regulatory pressures, estimates are lower bounds.

### Constructed repression indicators and standardization
- Fiscal repression (narrow): FR_F = ln ˜s_{b,F}_g − ln ˜s_h_g, where s_{b,F}_g = B_b / A_b and s_h_g = B_h / A_h; normalized to zero at country minimum.
- Monetary repression (reserve channel): s_{b,M}_g = B_CB / A_b where B_CB are government bonds held by central bank.
- Consolidated fiscal+monetary repression: FR_FM = ln ˜s_{b,FM}_g − ln ˜s_h_g where s_{b,FM}_g = (B_b + B_CB) / A_b.
- Indicators standardized (after normalization) for regressions so coefficients are interpretable in standard deviations.

### Empirical patterns, correlates, and de jure policy links
- Time-series patterns (1920–2020):
  - FR_F and FR_FM show a U-shape: increase during interwar, peak around WWII, elevated postwar through 1980s, decline during 1980s–1990s liberalization, renewed increase after the GFC.
  - Consolidated measure rises more sharply than fiscal in recent years, signaling increased monetary repression.
- Regression associations (all variables expressed in standard deviations; RHS winsorized at 1st and 99th percentiles; standard errors clustered by year):
  - Debt pressures:
    - A one standard deviation greater government debt-to-GDP associated with 0.2 to 0.4 standard deviations higher FR indicators.
    - A one standard deviation greater interest expenditures-to-GDP associated with 0.2 to 0.4 standard deviations higher FR indicators.
    - Higher primary balance associated with significantly higher FR indicators.
  - Interest rates and returns:
    - Low short- and long-term interest rates predict greater repression.
    - Government bond yields lower relative to private asset yields in more repressive episodes.
  - Bank balance-sheet margins:
    - One standard deviation slower nominal or real private credit growth associated with 0.1 to 0.2 standard deviations higher FR indicators.
    - Lower loan-to-deposit ratios associated with higher FR indicators.
    - Bank capital ratios do not predict FR, suggesting leverage is not the primary margin.
  - Investment:
    - One standard deviation lower investment-to-GDP associated with 0.1 to 0.2 standard deviations higher FR indicators.
- Relation to de jure policies (sample: France, Germany, Japan, United States; 1945–2007):
  - Monetary repression indicator higher where reserve requirements larger:
    - A 10 percentage point higher reserve requirement associated with almost one standard deviation higher monetary repression indicator.
  - Total asset requirements strongly correlated with fiscal+monetary repression:
    - A 10 percentage point higher total asset requirement associated with about 0.7 standard deviations higher fiscal+monetary repression indicator.

### Mapping repression into an interest-rate wedge and debt decomposition
- Debt law of motion identity (annual change):
  - ∆d_t =
    1
    1 + g_{Y,t}
    [ (r_{B,free,t} − g_t) d^{pub}_{t−1} ] − (π_t − π^e_t) d^{pub}_{t−1} / (1 + g_{Y,t})
    + (r_{B,t} − r_{B,free,t}) d^{pub}_{t−1} / (1 + g_{Y,t})
    − pb_t − cb_t − h_t + sf_t
- Term interpretations:
  - (r_{B,free,t} − g_t): real interest–growth differential absent repression.
  - (π_t − π^e_t): surprise inflation contribution.
  - (r_{B,t} − r_{B,free,t}): financial repression contribution.
  - pb_t: primary balance contribution.
  - cb_t: central bank purchases (monetary operations) contribution.
  - h_t: haircuts from restructurings contribution.
  - sf_t: stock-flow adjustments.
- Constructing counterfactual r_{B,free,t}:
  - Counterfactual defined by removing policy-induced bank demand shift and restoring market clearing via higher expected relative returns on government bonds.
  - Market-clearing condition: ∆D_b ≡ ∆s_b_g · ̄A_b + ∆s_h_g · ̄A_h = 0, with aggregate assets relative to GDP fixed.
  - Numerical implementation: iterative procedure solving for ∆(r^e_g − r^e_c) required to clear market once ̃T_b_g is reset (Annex A.5).
  - Elasticity calibration: baseline θ = 20; sensitivity checks with θ = 15 and θ = 30.

### Wedge magnitudes and decomposition outcomes
- Distributional magnitudes (1945–2020 examples):
  - Fiscal wedge median: 0.96 percentage points (consolidated wedge larger; country-year-specific and normalized to country minimum, implying lower-bound estimates).
- Case-study five-year cumulative contributions (expressed in percentage points of GDP; debt-reducing contributions):
  - United Kingdom (1945–1955): publicly-held debt fell by roughly 75 percentage points of GDP.
    - Financial repression: 90.7 pp.
    - Surprise inflation: 20.3 pp.
    - Primary balance: 3.9 pp.
    - Consolidated decomposition yields larger total contribution.
  - United States (1945–1955): repression contributed 17.5 pp of the 46 pp decline.
    - Surprise inflation: 15.2 pp.
    - Primary balance: 20.9 pp.
  - Advanced-economy average (2008–2020): financial repression contributed 14.8 pp in one reported column (Table 3 shows AE average 9.8 in another column for specific comparisons).
- Cross-country distribution of annual fiscal savings:
  - Median fiscal saving rises sharply around WWII, peaks mid-1940s, recedes through liberalization era, and turns up after the GFC.
  - 75th percentile truncated at 4 pp in figure note; actual peak 8.1 pp of GDP in 1945.

### Benchmarks, seigniorage, and sensitivity
- Reinhart–Sbrancia benchmark: fixed counterfactual real rates (1, 2, 3 percent) across 1945–1980 used for comparison; Reinhart–Sbrancia (2% counterfactual) reports 60.6 pp for UK 1945–1955.
- Seigniorage measures:
  - Seigniorage on cash: S_1 = i_t · c_t.
  - Seigniorage on unremunerated reserves: S_2 = i_t · η_t · A_{b,t}.
  - Seigniorage shown to be at most a few percentage points of GDP over episodes, substantially smaller than cumulative fiscal savings attributed to financial repression in postwar cases.
- Sensitivity to θ (robustness):
  - Baseline θ = 20; θ = 30 lowers repression estimates; θ = 15 raises them.
  - Example cumulative comparisons (selected entries from robustness table; values are cumulative contributions over the episode, in percentage points of GDP):
    - Financial repression (Baseline): 90.7 (UK 1945–1955), 17.5 (US 1945–1955), 14.8 (AE 2008–2020), 9.8 (other AE column).
    - θ = 30: 60.4, 11.7, 9.9, 6.5.
    - θ = 15: 120.9, 23.3, 19.7, 13.1.
    - Surprise inflation: 20.3, 15.2, 6.7, −9.7.
    - Primary balance: 3.9, 20.9, 1.3, −10.7.
    - Reinhart–Sbrancia (2%): 60.6, 21.9, 22.0, 12.9.
    - Seigniorage on cash: 1.5, 1.8, 2.1, 0.2.
    - Seigniorage on reserves: NA, 2.4, 0.3, 0.0.
  - Notes: baseline estimates use θ = 20; cumulative financial repression computed as sum of d_pub_{t−1} · (r_B,free_t − r_B_t).

### Algorithm and numerical implementation to recover equilibrium relative return (∆(r^e_g − r^e_c))
- Iterative algorithm to “undo” repression shock and find return adjustment that clears market:
  - Step 1: Undo the repression shock: −∆ ln ̃T_b_g = χ, ∆ ln ̃T_h_g = 0.
  - Step 2: Initialize ∆(r^e_g − r^e_c) = 0.
  - Step 3: Update portfolio log-odds: ∆ ln ̃s_i_g = ∆ ln T_i_g + θ_i ∆(r^e_g − r^e_c).
  - Step 4: Map to portfolio shares via logistic transform, compute ∆s_i_g.
  - Step 5: Compute excess demand ∆D_b = ∆s_b_g ̄A_b + ∆s_h_g ̄A_h.
  - Step 6: If |∆D_b| < ε, convergence achieved; repression wedge is converged ∆(r^e_g − r^e_c).
  - Step 7: Otherwise update ∆(r^e_g − r^e_c) by ±δ depending on sign of ∆D_b and iterate.

### Key empirical conclusions and policy-relevant findings
- Historical role:
  - Financial repression has been a recurring and quantitatively important channel for sovereign debt reduction in advanced economies, especially postwar.
  - Repression peaked around World War II, declined during financial liberalization, and rose again after the GFC.
- Contemporary relevance:
  - Post-GFC QE expanded consolidated public-sector claims via reserves; remuneration of those reserves below counterfactual rates and liquidity regulation sustaining captive demand produced a renewed measurable repression contribution.
  - Since 2008, the repression contribution operates significantly through the monetary arm (reserve remuneration), though fiscal indicator also rises post-crisis.
- Tradeoffs and costs:
  - Repression correlates with crowding out of private credit and lower investment.
  - Repression tends to lower deposit rates via greater bank allocation to lower-yielding government debt.
- Enabling conditions and constraints:
  - Repression more effective under restricted capital mobility and captive domestic institutions (Bretton Woods); modern financial integration and nonbank intermediation reduce uniform reach.
  - Emerging financial technologies (CBDCs, stablecoins, tokenization) could change captive creditor bases and repression channels.
- Interpretation caveats:
  - Indicators are lower bounds if households face non-regulatory pressures; normalization and θ calibration influence magnitudes.
  - Results are descriptive and accounting-based; they do not endorse repression as policy nor imply central bank intent when central bank purchases enter the decomposition mechanically.

### Implications and directions for future research
- Measurement extensions: revisit elasticity assumptions, incorporate nonbank financial intermediaries, refine country-specific normalizations, extend sample to emerging markets and post-2020 data.
- Causal identification: exploit quasi-experimental regulatory changes (e.g., Basel III liquidity requirements, reserve remuneration shifts).
- Welfare and distributional analyses: compare incidence and welfare consequences of repression versus fiscal consolidation, surprise inflation, and restructuring.
- Interaction with digital finance: study how CBDCs, stablecoins, and tokenized sovereign debt alter the captive creditor base and effectiveness of repression mechanisms.

*Source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026160-source-pdf.pdf*

### 1.   Introduction3

### 1.   Introduction3

### 2.   Concepts, Data, and Stylized Facts9
- Section title: "2.   Concepts, Data, and Stylized Facts"
- Page: 9

### 3.   Theoretical Framework21
- Section title: "3.   Theoretical Framework"
- Page: 21

### 4.   Estimates and Correlates of Financial Repression26
- Section title: "4.   Estimates and Correlates of Financial Repression"
- Page: 26

### 5.   Financial Repression in Sovereign Debt Dynamics32
- Section title: "5.   Financial Repression in Sovereign Debt Dynamics"
- Page: 32

### 6.   Conclusions46
- Section title: "6.   Conclusions"
- Page: 46

### Annex51
- Annex
- Page: 51

*Source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026160-source-pdf.pdf*

### References63

### References63

### Glossary of Acronyms
- AE Advanced Economy
- AR(n) Auto-regressive by order n
- BIS Bank for International Settlements
- CBDC Central Bank Digital Currency
- FR Financial Repression
- GFC Global Financial Crisis (2007–2009)
- GDP Gross Domestic Product
- GMD Global Macro Database (Müller-Plantenberg et al., 2024)
- JST Jordà–Schularick–Taylor Macrohistory Database (Jordà et al., 2017)
- LCR Liquidity Coverage Ratio
- PFIMH Public Finances in Modern History (Mauro et al., 2013)
- QE Quantitative Easing
- UIP Uncovered Interest Rate Parity
- WID World Inequality Database
- WWI World War I (1914–1918)
- WWII World War II (1939–1945)

### Introduction — problem statement and contribution
- Fiscal outlooks in many advanced economies have become increasingly challenging, with public debt at high peacetime levels.
- Standard orthodox approaches (fiscal consolidation and growth-enhancing reforms) can be slow and politically difficult; they may have limited effects on debt ratios if not well-timed.
- Heterodox channels historically used to reduce debt burdens:
  - Surprise inflation: reduces real value of nominal claims when realized inflation exceeds what was priced at issuance.
  - Debt monetization: central bank financing of government deficits through non-interest-bearing liabilities (currency or reserves).
  - Financial repression: regulatory policies compress sovereign borrowing costs by inducing financial institutions to absorb government debt as captive buyers.
- Key empirical and identification challenges motivate the paper:
  - Detailed historical data on composition and cost of government debt are limited.
  - The counterfactual sovereign yield absent repression is unobservable.
- Paper’s three contributions:
  - Theoretically-grounded framework identifying financial repression via sectoral holdings of public liabilities; constructs two indicators (narrow fiscal measure based on government bond holdings and consolidated measure including reserves).
  - New historical dataset consolidating multiple sources on general government debt, central bank liabilities, and sectoral composition of holdings across 17 advanced economies; descriptive evidence that repression peaked around WWII, remained elevated through Bretton Woods, declined post-1971, and rose again since the GFC.
  - Model-based counterfactuals for sovereign borrowing costs absent repression and embedding implied yield wedge in a unified debt-decomposition framework; country applications highlight sizable fiscal savings from repression.
- Quantitative highlights:
  - In the United Kingdom, repression was the single largest deleveraging channel in the postwar decade, accounting for more observed debt reduction than surprise inflation or fiscal consolidation.
  - In the United States, fiscal consolidation played the largest role postwar, but repression was also a major contributor.
  - Across the panel, median annual fiscal savings from financial repression averaged around 0.57 percent per year, peaking sharply in the mid-1940s.
  - Median fiscal savings rise after the GFC, from about 0.08 percent in 2008 to about 0.51 percent in 2020.
- Empirical correlates of repression:
  - More pervasive when government debt and interest expenditures are high.
  - Associated with higher primary balances.
  - Coincides with low short- and long-term interest rates and depressed government bond yields relative to private asset returns.
  - Correlates with lower private credit growth, lower loan-to-deposit ratios, and reduced investment.
  - Post-WWII de jure reserve and portfolio requirements line up systematically with the repression indicators.

### Related literature and framing
- Situates work within three literatures:
  - Measurement and historical correlates of financial repression (McKinnon 1970; Shaw 1973; Reinhart and Sbrancia 2015).
  - Debt monetization and monetary financing (Sargent and Wallace 1981; Reis 2016; Feldberg and Lawson 2020).
  - Decomposition of public debt dynamics (Eichengreen and Esteves 2022; International Monetary Fund 2023a; Reis 2022).
- Methodological advances relative to prior work:
  - Quantity-based indicators exploiting differential sectoral absorption of public liabilities, mapping quantity shifts into an explicit repression wedge to enable systematic cross-country and historical comparisons.
- Clarifying distinctions:
  - Financial repression studied here is scoped to advanced-economy debt-reduction settings and focuses on repression as a sovereign debt reduction channel.
  - Distinguishes outcome-based measurement from stated policy intent—indicators focus on balance-sheet outcomes (who holds public liabilities at administered or below-market returns), not on declared motives.
  - QE is not classified as repression by construction, but the framework evaluates empirically whether QE episodes produce repressive balance-sheet outcomes (i.e., whether resulting reserves exceed decline in government bonds held by banks).

### Concepts, Data, and Stylized Facts — Varieties of Financial Repression
- Core mechanism: policies induce financial intermediaries to hold government liabilities at administered or below-market returns, transferring resources from the private sector to the state.
- Taxonomy based on perimeter of public sector liabilities placed onto regulated intermediaries:
  - Fiscal repression: regulated intermediaries absorb government bonds, compressing sovereign borrowing costs.
    - Instruments: liquidity regulation, capital regulation (preferential risk weights), collateral eligibility rules, portfolio/portfolio requirements, interest-rate ceilings.
    - Quantitative repression indicator: Government bonds as share of commercial bank assets.
  - Monetary repression: investors induced to hold central bank liabilities (commercial bank reserves) at administered remuneration below market alternatives.
    - Instruments: seigniorage on currency (legal tender laws), reserve requirements, remuneration policies, QE (evaluated by outcome).
    - Quantitative repression indicator: Commercial bank reserves created through central bank purchases of government bonds.
  - Consolidated (Fiscal + Monetary) repression:
    - Public sector perimeter: General government + central bank.
    - Quantitative repression indicator: Government bonds held directly and indirectly as share of commercial bank assets.
- Methodological note: outcome-based approach focuses on realized balance-sheet allocations rather than coding instruments or inferring intent.

### Concepts, Data, and Stylized Facts — Data construction and coverage
- Measurement strategy: indicators based on portfolio shares of government liabilities held by domestic financial institutions relative to non-captive investors; requires balance-sheet aggregates for commercial banks, households, and the consolidated public sector.
- Data challenges: harmonizing multiple sources with different accounting conventions, country coverage, and time spans to produce consistent sectoral compositions and portfolio shares over a long historical horizon.
- Country sample and temporal coverage:
  - Analysis restricted to 17 advanced economies covered by the JST database: Australia, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, United Kingdom, and United States.
  - The JST database spans 1870 to 2020.
  - The paper elsewhere applies the repression measures to a dataset covering 17 advanced economies since 1920 (dataset construction is designed to reach wartime episodes when repression was most pronounced).

*Source: wpiea2026160-source-pdf - References63*

### 1880.  Data on the sectoral composition of government debt holdings—the share held by

### 1880.  Data on the sectoral composition of government debt holdings—the share held by

### Data sources and construction
- Sectoral composition of government debt (shares held by commercial banks, the central bank, households, and nonresidents) drawn from Abbas et al. (2011) and Arslanalp and Tsuda (2014), extended to 2020.
- Central bank total assets: Bank for International Settlements. Currency in circulation: Finaeon. Where JST or PFIMH macro-fiscal aggregates are missing, gaps filled from the Global Macro Database (Müller-Plantenberg et al., 2024).
- Household total asset holdings: World Inequality Database (WID) (Piketty and Zucman, 2014); historical wealth-to-output series through 1979 where available, later observations harmonized as net market-value national wealth divided by average GDP.
- Sovereign debt restructurings and haircuts: Graf von Luckner et al. (2024). Coverage: 329 distressed restructuring events with external private creditors worldwide over 1823–2024. Annual haircut term constructed as NPV haircut × debt restructured, converted to local currency and scaled by GDP.
- Expected returns:
  - Government bonds: nominal bond yields from JST combined with expected inflation estimated using country-specific autoregressive model (rolling AR(1) estimated on data available up to year t-1, forecasts from 1900 onward; Germany: 1921–1924 excluded, 1925 forecast re-initialized using median 1900–1920 inflation).
  - Private assets: dividend yields for equities and rental yields for housing from JST.
- Balance-sheet and return estimation details: provided in Subsection A.2. Summary of data sources in Table 2 (paper’s analytical sample: 17 advanced economies over 1920–2020).

### Stylized facts motivating the model
- Stylized Fact 1: Higher sovereign debt burdens are associated with lower, not higher, relative returns on government bonds.
  - Example magnitudes:
    - United Kingdom: relative yields reached approximately −4 percent per annum when public debt peaked near 250 percent of GDP in the late 1940s.
    - United States: relative yields dropped to around −3 percent at the postwar debt peak.
    - Japan: as public debt rose toward 250 percent of GDP since the 1990s, relative returns collapsed into deeply negative territory.
  - Relative return definition: government bond yield minus an equally-weighted average of equity dividend yields and housing rental yields.
- Stylized Fact 2: Commercial banks hold a disproportionately large share of government debt precisely when relative returns on sovereign bonds are lowest and debt burdens are highest.
  - Bank share of government debt rises when government bond relative returns fall and when government debt-to-GDP rises.
- Stylized Fact 3: Government securities account for a larger share of commercial bank assets when debt burdens are higher and sovereign bond returns are lower relative to private assets.
  - Banks expand sovereign exposure at the expense of private lending (substitution from private to public assets), not by balance-sheet expansion.

### Theoretical framework (overview)
- Agents: households (h), banks (b), non-financial corporations (c), government (g). Government issues nominal bonds B held by banks (B_b) and households (B_h).
- Portfolio demand model:
  - Fréchet-distributed idiosyncratic taste shocks with shape parameters θ_i (i ∈ {b,h}).
  - Time-invariant shifters ̄T_i_j and time-varying regulatory shifters ̃T_i_j. Repression modeled as ̃T_b_g ≥ 1 for banks and ̃T_h_g = 1 for households (regulatory pressure on intermediaries).
  - Portfolio share for investor i in government bonds:
    - s_i_g = ̃T_i_g (̄T_i_g · R^e_g)^{θ_i} / [̃T_i_g (̄T_i_g · R^e_g)^{θ_i} + (̄T_i_c · R^e_c)^{θ_i}]
  - Log-odds transformation ln ˜s_i_g ≡ ln[s_i_g /(1− s_i_g)] yields a linear representation.
- Identification of repression (changes between two periods):
  - ∆ ln ̃T_b_g = ∆ ln ˜s_b_g − ∆ ln ˜s_h_g − (θ_b − θ_h)(∆r^e_g − ∆r^e_c)
  - Interpretation:
    - First two terms: raw quantity signal—change in banks’ government bond log-odds relative to households’.
    - Third term: adjustment for return-driven portfolio rebalancing and differences in substitution elasticities.
  - When θ_b = θ_h, return adjustment drops out and repression changes equal relative holdings changes.
- Normalization and interpretation:
  - Level of repression is identified only up to scale; normalization sets ln ̃s_b_g = 0 for each country’s lowest observed value.
  - Households assumed not subject to regulatory pressure; if households face non-regulatory pressures (e.g., moral suasion), estimates are lower bounds.
  - Quantity-based indicators measure repression as excess bank demand for public liabilities relative to returns and household behavior; these map into an interest-rate wedge via market-clearing.

### Constructed financial repression indicators
- Fiscal repression (narrow): FR_F = ln ˜s_{b,F}_g − ln ˜s_h_g, where s_{b,F}_g = B_b / A_b and s_h_g = B_h / A_h. Normalized to zero at country minimum.
- Monetary repression (reserve channel): s_{b,M}_g = B_CB / A_b where B_CB are government bonds held by central bank.
- Consolidated fiscal + monetary repression: FR_FM = ln ˜s_{b,FM}_g − ln ˜s_h_g where s_{b,FM}_g = (B_b + B_CB) / A_b.
- Standardization: after normalization, indicators standardized for regressions so coefficients are interpreted in standard deviations.

### Empirical patterns and correlates
- Time series patterns (1920–2020):
  - Both fiscal (FR_F) and consolidated (FR_FM) indicators show a U-shape: increase during interwar period, peak around World War II, widespread postwar through 1980s, decline during financial liberalization in 1980s–1990s, and a renewed increase following the Global Financial Crisis (GFC).
  - Consolidated measure rises more sharply than fiscal in recent years, suggesting increased monetary repression.
  - Normalization note: measures normalized to country minimum; series are lower bounds if residual repression exists in baseline period.
- Associations (regressions: FR_{i,t} = β X_{i,t} + α_i + α_t + ε_{i,t}; FR and RHS variables expressed in standard deviations; RHS winsorized at 1st and 99th percentiles; standard errors clustered by year):
  - Debt pressures:
    - A one standard deviation greater government debt-to-GDP ratio associated with 0.2 to 0.4 standard deviations higher FR indicators.
    - A one standard deviation greater interest expenditures-to-GDP associated with 0.2 to 0.4 standard deviations higher FR indicators.
    - Higher primary balance associated with significantly higher FR indicators.
  - Interest rates and asset returns:
    - Low short- and long-term interest rates predict greater use of repression.
    - Government bond yields are lower relative to yields on private assets in countries with more repression.
  - Crowding out and bank balance sheets:
    - One standard deviation slower nominal or real private credit growth associated with 0.1 to 0.2 standard deviations higher FR indicators.
    - Lower loan-to-deposit ratios associated with higher FR indicators (deposit funding channel for government liabilities).
    - Bank capital ratios do not predict FR, suggesting leverage is not the margin used.
  - Investment and saving balances:
    - One standard deviation lower investment-to-GDP associated with 0.1 to 0.2 standard deviations higher FR indicators.
- Relation to de jure policies (sample: France, Germany, Japan, United States; post-WWII before Basel requirements, 1945–2007):
  - Monetary repression indicator is higher where reserve requirements are larger.
    - A 10 percentage point higher reserve requirement associated with almost one standard deviation higher monetary repression indicator.
  - Total asset requirements (reserve + portfolio requirements) strongly correlated with fiscal+monetary repression.
    - A 10 percentage point higher total asset requirement associated with about 0.7 standard deviations higher fiscal+monetary repression indicator.

### Mapping repression into an interest-rate wedge and debt decomposition
- Law of motion for publicly-held debt (d_pub = B_pub / Y). Key decomposition identity (annual change):
  - ∆d_t =
    1
    1 + g_{Y,t}
    [ (r_{B,free,t} − g_t) d^{pub}_{t−1} ] − (π_t − π^e_t) d^{pub}_{t−1} / (1 + g_{Y,t})
    + (r_{B,t} − r_{B,free,t}) d^{pub}_{t−1} / (1 + g_{Y,t})
    − pb_t − cb_t − h_t + sf_t
  - Interpretation of terms:
    - (r_{B,free,t} − g_t) term: real interest–growth differential absent repression.
    - (π_t − π^e_t) term: surprise inflation contribution.
    - (r_{B,t} − r_{B,free,t}) term: financial repression contribution (difference between observed and counterfactual free-market real rates).
    - pb_t: primary balance contribution.
    - cb_t: central bank purchases (monetary operations) contribution.
    - h_t: haircuts from restructurings contribution.
    - sf_t: stock-flow adjustments.
- Constructing the counterfactual free-market rate r_{B,free,t}:
  - Counterfactual defined as the real interest rate that would prevail if the policy-induced bank demand shift were removed and market clearing were restored via higher expected relative returns on government bonds.
  - Market-clearing condition for bond demand changes:
    - ∆D_b ≡ ∆s_b_g · ̄A_b + ∆s_h_g · ̄A_h = 0, holding aggregate assets relative to GDP fixed.
  - Numerical implementation:
    - Iterative procedure (Annex A.5) solving for the increase in expected relative returns required to clear the market once ̃T_b_g is reset (repression removed).
  - Elasticity calibration:
    - Baseline θ = 20 (motivation: Eren et al. (2023) report θ ≈ 33 within debt securities; model uses lower value to reflect denominator of total assets).
    - Sensitivity: alternative θ values (15 and 30) considered; main conclusions robust though magnitudes vary (lower θ → larger repression contribution).
- Distribution and magnitudes of estimated wedges (1945–2020):
  - Fiscal wedge median: 0.96 percentage points (distribution shown in Figure 10; consolidated wedge larger).
  - Wedge estimates are country-year-specific and time-varying; normalized so country minimum defines zero-repression benchmark (implies estimates are lower bounds if residual repression exists in baseline).

### Debt decomposition results and fiscal savings
- Case studies (five-year cumulative contributions, 1945–2020 windows):
  - United Kingdom (postwar): publicly-held debt fell by roughly 75 percentage points of GDP in the decade following World War II.
    - Financial repression accounts for approximately 90.7 percentage points of this decline over 1945–1955.
    - Surprise inflation contribution: 20.3 percentage points.
    - Primary balance contribution: 3.9 percentage points.
    - Consolidated decomposition (including monetary channels) yields an even larger total contribution.
  - United States (postwar): repression contributed roughly 17.5 percentage points of the 46 percentage point decline over 1945–1955.
    - Surprise inflation: 15.2 percentage points.
    - Primary balance: 20.9 percentage points.
  - Advanced-economy average (2008–2020): financial repression contributed 14.8 percentage points (Table 3 reports AE average for 2008–2020 as 9.8 in one column and 14.8 in another for specific comparisons; Table 3 layout shows United Kingdom, United States, AE average columns).
- Cross-country distribution of annual fiscal savings (Figure 13):
  - Median fiscal saving from repression rises sharply around World War II, peaks mid-1940s, recedes through liberalization era, and turns up after the GFC.
  - Cross-country interquartile range widens in postwar decades; 75th percentile truncated at 4 pp in figure note (actual peak 8.1 pp of GDP in 1945).
- Benchmarks and comparisons (selected episodes, cumulative contributions in pp GDP; all reported as debt-reducing contributions):
  - Table 3 highlights:
    - UK 1945–1955: Financial repression 90.7; Surprise inflation 20.3; Primary balance 3.9.
    - US 1945–1955: Financial repression 17.5; Surprise inflation 15.2; Primary balance 20.9.
    - AE average 1945–1955 and 2008–2020 reported in table alongside Reinhart–Sbrancia benchmark and seigniorage measures.
  - Reinhart–Sbrancia benchmark uses fixed counterfactual real rates (1, 2, 3 percent) across 1945–1980; the model-implied wedge is time-varying and disciplined by observed portfolio allocations.
- Seigniorage comparisons:
  - Seigniorage on cash: S_1 = i_t · c_t.
  - Seigniorage on unremunerated reserves: S_2 = i_t · η_t · A_{b,t}.
  - Table 3 shows seigniorage measures are at most a few percentage points of GDP over episodes, substantially smaller than cumulative fiscal savings attributed to financial repression in postwar cases (e.g., UK repression >> S_1 + S_2).
- Sensitivity to θ:
  - Baseline θ = 20; alternatives θ = 30 and θ = 15 considered (Annex A.6 Table 9). Lower θ increases estimated repression contribution; main qualitative conclusions robust.

### Key empirical conclusions and policy-relevant findings
- Historical role:
  - Financial repression has been a recurring and quantitatively important channel for sovereign debt reduction in advanced economies, especially in the postwar decades.
  - Repression peaked around World War II and remained elevated in early postwar decades; it declined during financial liberalization and rose again following the GFC.
- Contemporary relevance:
  - Post-GFC and post-2008 quantitative easing expanded the consolidated public-sector claim via reserves; remuneration of those reserves below counterfactual rates and liquidity regulation that sustains captive demand have produced a renewed, measurable repression contribution.
  - The repression contribution since 2008 operates significantly through the monetary arm (reserve remuneration) though the fiscal indicator also rises post-crisis.
- Tradeoffs and costs:
  - Financial repression is associated with crowding out of private credit and lower investment (estimates show negative associations with private credit growth and investment-to-GDP).
  - Repression tends to lower deposit rates via greater bank allocation to lower-yielding government debt and further yield compression on government bonds.
- Enabling conditions and constraints:
  - Repression historically more effective under restricted capital mobility and captive domestic institutions (e.g., Bretton Woods era). Modern financial integration, wider substitute assets, and migration of intermediation outside traditional banking may limit the reach and uniformity of repression today.
  - Novel financial technologies and instruments (CBDCs, stablecoins, tokenization) could alter the scope and channels through which public liabilities are held, creating new margins relevant to repression.
- Interpretation caveats:
  - Indicators measure regulatory-induced demand shifts relative to household benchmarks and are lower bounds if households face non-regulatory pressures.
  - Normalization choices and elasticity calibration influence magnitudes; results are robust to reasonable alternatives but should be read as conservative estimates.
  - Results are descriptive and accounting-based; they do not endorse repression as a policy and do not imply policy intent by central banks when central bank purchases enter the decomposition mechanically.

### Implications and directions for future research
- Measurement extensions:
  - Revisit elasticity assumptions, the composition of the non-captive investor base (including nonbank financial intermediaries), and country-specific normalizations.
  - Extend sample to emerging markets and post-2020 observations (through end-2025) to assess post-COVID inflation and monetary policy tightening effects.
- Causal identification:
  - Exploit quasi-experimental regulatory changes (e.g., introduction of Basel III liquidity requirements, shifts in reserve remuneration policy) to sharpen causal estimates of repression’s mechanisms and effects.
- Welfare and distributional analyses:
  - Systematically compare incidence and welfare consequences of repression relative to fiscal consolidation, surprise inflation, and restructuring.
- Interaction with digital finance:
  - Study how CBDCs, stablecoins, and tokenized sovereign debt change the captive creditor base and the effectiveness of repression mechanisms.

*Source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026160-source-pdf.pdf*

### 1.  Undo the repression shock. Reverse the policy-induced preference shifter for banks:

### 1.  Undo the repression shock. Reverse the policy-induced preference shifter for banks:

### Algorithm to find equilibrium relative return (∆(r_e_g − r_e_c))
- Step 1: Undo the repression shock. Reverse the policy-induced preference shifter for banks:
  - −∆ ln  ̃T_b_g = χ, ∆ ln  ̃T_h_g = 0.
- Step 2: Initialize the return adjustment.
  - Start from no change in relative returns: ∆(r_e_g − r_e_c) = 0.
- Step 3: Update portfolio log-odds for each investor i ∈ {b, h}:
  - ∆ ln  ̃s_i_g = ∆ ln T_i_g + θ_i ∆(r_e_g − r_e_c),
  - combining the policy shock (banks only) with the price response.
- Step 4: Update portfolio shares.
  - New log-odds levels: ln  ̃s_i′_g = ln  ̃s_i_g + ∆ ln  ̃s_i_g,
  - Map to portfolio shares using the logistic function:
    - s_i′_g = exp( ln  ̃s_i′_g ) / (1 + exp( ln  ̃s_i′_g )),
    - ∆s_i_g = s_i′_g − s_i_g.
- Step 5: Compute excess demand (market-clearing residual):
  - ∆D_b = ∆s_b_g ̄A_b + ∆s_h_g ̄A_h.
- Step 6: Check convergence.
  - If |∆D_b| < ε, the equilibrium relative return has been found.
- Step 7: Update returns and iterate.
  - If ∆D_b < 0, increase ∆(r_e_g − r_e_c) by δ;
  - if ∆D_b > 0, decrease it by δ.
  - Return to step 3 and repeat until convergence.

- Definition: The repression wedge for period t is the value of ∆(r_e_g − r_e_c) at convergence, which gives the increase in the sovereign borrowing cost required to clear the market absent regulatory pressure on banks’ portfolio allocation.

### Robustness checks: Financial repression and related mechanisms (Table 9 summary)
- Table scope: Robustness Checks: Financial Repression and Related Mechanisms
  - Periods/columns referenced in table header: 1945–1955, AE Average, United Kingdom, United States, 1945–1955, 2008–2020.
- Selected entries (cumulative contributions over the episode, expressed in percentage points of GDP and reported as debt-reducing contributions):
  - Financial repression (Baseline): 90.7, 17.5, 14.8, 9.8
  - θ = 30: 60.4, 11.7, 9.9, 6.5
  - θ = 15: 120.9, 23.3, 19.7, 13.1
  - Surprise inflation: 20.3, 15.2, 6.7, −9.7
  - Primary balance: 3.9, 20.9, 1.3, −10.7
  - Reinhart–Sbrancia benchmark (2% real-rate counterfactual): 60.6, 21.9, 22.0, 12.9
  - Seigniorage on cash: 1.5, 1.8, 2.1, 0.2
  - Seigniorage on reserves: NA, 2.4, 0.3, 0.0

- Notes on computation and definitions (preserve technical formulas and wording):
  - All entries represent cumulative contributions over the episode, expressed in percentage points of GDP and reported as debt-reducing contributions. A positive value indicates a reduction in public debt.
  - Financial repression is computed as the cumulative sum of d_pub_{t−1} · (r_B,free_t − r_B_t), where r_B,free_t is the model-implied counterfactual free-market rate.
  - Surprise inflation is computed as the cumulative sum of (π_t − π^e_t)· d_pub_{t−1} /(1 + g_Y_t).
  - The primary balance contribution is computed as the cumulative sum of pb_t.
  - The Reinhart–Sbrancia benchmark assumes a fixed counterfactual real interest rate of 2 percent throughout each episode.
  - Seigniorage on cash is defined as S_1 = i_t · c_t.
  - Seigniorage on unremunerated reserves is defined as S_2 = i_t · η_t · A_b_t, where η_t denotes the reserve requirement applicable to unremunerated reserves.
  - The baseline financial repression estimates are based on the θ value of 20. The Reinhart–Sbrancia benchmark uses a 2 percent real-rate counterfactual.
  - “AE” refers to the advanced economies in the sample. NA indicates that the series is not available for the relevant country-period.

*The Coming Great Repression? New Measures and a Century of Evidence. Working Paper No. WP/2026/160*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026160-source-pdf.pdf_
