## _wp1305

## Source details

**Canonical URL:** [_wp1305](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1305.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1305.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1305.pdf.json)

---

### I. Introduction and purpose
- Objective: gauge degree of fiscal prudence or profligacy for each country over several decades using a newly collected historical dataset of fiscal variables for a large panel of countries.
- Dataset components: fiscal revenues, primary expenditures, the interest bill (and thus both the primary and the overall fiscal deficit), the government debt, and gross domestic product.
- Coverage: 55 countries for up to two hundred years (panel later described as 1800–2011 for many series).
- Conceptual anchor: “fiscal prudence”/“profligacy” measured relative to whether the government’s intertemporal budget constraint is met; practical proxy is the primary fiscal balance (fiscal balance net of interest payments).
- Temporal focus: medium-run judgments over a few years.

### II. Methodological framework
- Core empirical test: Bohn (1998, 2008) fiscal sustainability test: regress primary fiscal surplus on public debt and controls (notation in text: s_t = α + ρ d_t + Z_t + ε_t).
- Extensions/relaxations applied country-by-country:
  - Structural break tests (Bai and Perron (1998)).
  - Recursive searches for influential observations (omitting single years).
  - Rolling-window and expanding-sample Bohn regressions (e.g., 25-year rolling windows; expanding windows starting 1950).
- Complementary criterion: “policymakers’ criterion” — compare actual primary surplus to the primary surplus required to stabilize the debt-to-GDP ratio (debt-stabilizing primary surplus).
- Purpose of methods: capture variation in fiscal policy response functions across countries and over time within countries; relax constant long-term response assumption.

### III. Data, coverage, and construction
- Panel: unbalanced panel of 55 countries (24 advanced economies and 31 non-advanced) over 1800–2011.
- Fiscal series (expressed as share of GDP): government revenue, non-interest government expenditure, interest bill, overall fiscal balance, primary balance, gross public debt.
- Observation counts and series lengths:
  - Revenue, expenditure, overall fiscal balance, and debt: about 5,700 observations each.
  - Interest expenditure (and consequently primary expenditure and primary balance): approximately 4,800 observations.
  - Both debt and primary balance available for about 4,500 country-years.
- Sector coverage: prioritized most comprehensive government sector available (general government when available); sector switches recorded via dummy variables.
- Exclusions in many empirical exercises: country-years with expenditure-to-GDP increases ≥ six percentage points in one year (plus up to two post-war years) and default years (Reinhart and Rogoff (2010)).

### IV. Summary statistics and stylized facts (full sample 1800−2011 and post-WWII)
- Full sample (1800–2011) averages (subject to availability):
  - Revenue: 19 percent of GDP.
  - Expenditure: 21 percent of GDP.
  - Public sector interest bill: 2½ percent of GDP.
  - Primary surplus: ½ percent of GDP.
  - Debt: 50 percent of GDP.
  - Countries generally faced a negative interest-growth differential.
- Selected full-sample distributional statistics (exact figures preserved where listed; No. Obs as given):
  - Revenue: Mean 19.01; Std. Dev. 13.6; Min 0.5; 25% 8.5; Median 14.7; 75% 26.54; 95% 47.45; 99% 55.26; Max 60.55; No. Obs 721
  - Expenditure: Mean 20.9; Std. Dev. 14.4; Min 0.7; 25% 9.6; Median 16.9; 75% 29.9; 95% 50.1; 99% 56.67; Max 71.85; No. Obs 742
  - Interest Expenditure: Mean 2.4; Std. Dev. 2.3; Min 0.0; 25% 0.8; Median 1.7; 75% 3.3; 95% 6.5; 99% 10.32; Max 14.64; No. Obs 862
  - Overall Balance: Mean -1.8; Std. Dev. 4.0; Min -39.1; 25% -3.1; Median -1.0; 75% 0.1; 95% 2.2; 99% 5.4; Max 19.15; No. Obs 739
  - Primary Balance: Mean 0.4; Std. Dev. 3.9; Min -35.5; 25% -1.0; Median 0.5; 75% 2.2; 95% 5.5; 99% 9.5; Max 20.64; No. Obs 834
  - Debt: Mean 49.74; Std. Dev. 42.7; Min 0.0; 25% 19.2; Median 37.9; 75% 66.2; 95% 133.2; 99% 215.52; Max 289.65; No. Obs 670
  - Interest-Growth Differential: Mean -3.9; Std. Dev. 9.1; Min -34.9; 25% -8.4; Median -2.8; 75% 1.5; 95% 9.0; 99% 18.03; Max 34.24; No. Obs 266
- Post-WWII (1950–2011) selected means and counts:
  - Revenue: Mean 25.11; Std. Dev. 13.4; No. Obs 126
  - Expenditure: Mean 27.3; Std. Dev. 13.8; No. Obs 153
  - Interest Expenditure: Mean 2.6; Std. Dev. 2.6; No. Obs 788
  - Overall Balance: Mean -2.2; Std. Dev. 3.9; No. Obs 152
  - Primary Balance: Mean 0.3; Std. Dev. 3.5; No. Obs 803
  - Debt: Mean 45.1; Std. Dev. 34.3; No. Obs 983
  - Interest-Growth Differential: Mean -5.9; Std. Dev. 9.0; No. Obs 471
- Advanced economies (Postwar) sample means:
  - Revenue: Mean 33.31; Std. Dev. 12.8; No. Obs 481
  - Expenditure: Mean 35.41; Std. Dev. 13.6; No. Obs 479
  - Primary Balance: Mean 0.7; Std. Dev. 3.5; No. Obs 467
  - Debt: Mean 49.4; Std. Dev. 34.0; No. Obs 458
  - Interest-Growth Differential: Mean -3.1; Std. Dev. 7.0; No. Obs 360

### V. Variance decomposition of debt changes (1950–2011)
- Advanced economies: fluctuations in primary balance, interest-growth differentials, and stock-flow changes each explain roughly one-third of the variance of changes in the debt ratio; high residuals linked to sizable asset accumulation (Norway, Sweden, Finland, Japan) and to Greece (inflation, defaults, restructurings).
- Non-advanced economies: more volatile debt changes largely explained by stock-flow residuals and interest-growth differentials (defaults, high inflation, exchange rate crises), with heterogeneity (e.g., Colombia and India similar to advanced economies).
- Selected country examples (variance components and ∆Debt as in Tables 5 & 6):
  - United States (61 obs): ∆Debt 15.7; Prim. Bal. 6.6; (r-g)d 5.1; Residual 1.2; C(PB,(r-g)d) -0.3; C(PB,Resid) -0.4
  - Japan (61 obs): ∆Debt 41.8; Prim. Bal. 10.1; (r-g)d 7.9; Residual 11.8; C(PB,(r-g)d) 0.5; C(PB,Resid) 0.9; C((r-g)d,Resid) 0.6
  - Argentina (38 obs): ∆Debt 423.5; ∆Prim.Bal. 5.9; (r-g)d 39.9; Residual 346.8; C(PB,(r-g)d) 11.7; C(PB,Resid) -3.5; C((r-g)d,Resid) 22.6

### VI. Methods to measure fiscal prudence/profligacy (operational definitions)
- Bohn test interpretation:
  - If estimated ρ > 0 and significant, fiscal policy consistent with meeting the intertemporal budget constraint under uncertainty.
  - Stationarity condition in Bohn’s formulation: (1/(r−g)) ρ > 0 (paper uses medium-run averages of r and g).
- Complementary procedures identifying time variation and influential episodes:
  1. Structural break tests (Bai and Perron (1998)) to partition history into subperiods with differing ρ.
  2. Influential observation search: recursive omission of single years to find years that change significance of ρ.
  3. Rolling-window and expanding-sample Bohn regressions (e.g., 25-year rolling windows; expanding windows starting 1950).
- Policymakers’ criterion: contemporaneous comparison of actual primary surplus to debt-stabilizing primary surplus.
- Classification rules:
  - “Strongly prudent” country-year: non-war, non-default year meeting three tests (part of at least one 25-year prudent window; influentially prudent; actual primary balance > forward-looking debt-stabilizing primary balance).
  - “Profligate” country-year: non-war, non-default year failing prudence across multiple criteria (not in any 25-year prudent window; actual primary balance < debt-stabilizing primary balance; and either part of ≥1 25-year profligate window or influentially profligate).

### VII. Country- and episode-level findings (selected empirical results)
- Whole-sample Bohn tests:
  - Using full sample, Bohn coefficient positive and significant for three fourths of advanced economies.
  - In many cases estimated coefficient exceeds level necessary for stationarity of public debt ratio.
  - Crisis years (2008–2011) materially affect estimates for many countries (e.g., United States post-2007 loses significance and changes sign when crisis years included).
  - Notable negative and significant Bohn coefficients among large advanced economies: France and Japan (France partly reflects over-accumulation of assets 1950–1977; Japan shows weak primary response despite rising debts from late 1990s).
- Structural breaks:
  - Fiscal policy response to debt is non-constant over time for nearly every tested country.
  - Common break periods: mid 1970s and the 1990s (and earlier around the first decade of the 1900s and the 1930s when using the entire sample).
  - Mid-1970s breaks often mark a shift to a positive and significant response; breaks in the 1990s frequently represent relaxation of fiscal restraint for two-thirds of advanced economies.
- Influential observation search:
  - Many prudent years during 1880–1913 for countries with long data.
  - Few influentially profligate observations in 1950s–1960s.
  - Several advanced economies influentially prudent in mid- or late 1990s; many countries influentially profligate in at least one year since 2008.
  - Early to mid-1970s appear profligate in a fairly large number of countries.
- Rolling and expanding windows:
  - Rolling 25-year windows show substantial time-variation (example: Japan negative and significant in windows ending in late 1970s–1980s; positive and significant in late 1980s–early 1990s; descent to profligacy from 1995 through present).
  - Table 8: countries measured prudent far more often than profligate; most profligacy observations concentrated in last four decades and notably after 2007 for many countries.
  - Expanding-sample windows from 1950: most observed profligacy in mid and late 1970s; by the 2000s most countries measured as prudent, and most remained prudent through the crisis that began in 2008—exceptions include United States, Spain, Portugal, and Iceland.
- Times of strong prudence and profligacy (selected country examples, from Table 13):
  - Sweden — Strong Prudence: 1987-9, 1996-2001, 2006-7; Profligacy: 1915, 1920-2, 1953, 1957; N=202
  - United States — Prudence: 1948-51, 1998-2000; Profligacy: 1934-6, 2009-11; N=177
  - United Kingdom — Prudence: 1830-6, 1841-6, 1850, 1856, 1860, 1906-13, 1954-9, 2002; Profligacy: none; N=160
  - Canada — Prudence: 1881, 1891, 1906, 1909-13, 1933, 1937-8, 1946-8, 1995-2003; Profligacy: none; N=134
  - Japan — Prudence: 1907, 1985-92; Profligacy: 1880, 1997-2011; N=125
  - Noted stimulus/profligacy episodes: United States in 2009–11; Spain in 2010; Japan many years since late 1990s where it did not sufficiently improve its primary balance despite rising debts.

### VIII. Panel regressions: time-variation drivers and quantitative results (selected preserved estimates)
- Estimation: extended fiscal reaction function with interactions for long-term growth surprises and real long-term sovereign borrowing costs (10-year domestic-currency yield net of GDP deflator).
- Full sample (examples of preserved coefficients from Table 14):
  - Baseline Bohn coefficient (Lagged Debt): ≈ 0.03 (reported as 0.027*** to 0.030*** across specifications).
  - Debt*Change in HP Growth: coefficients vary (examples: 0.063, 0.179**, 0.33, 0.861** depending on specification).
  - Change in HP Growth: large negative point estimates in some specs (e.g., -13.34; -51.565*).
  - Debt*Real LT Interest Rate: 0.067*** (R3), 0.049 (R4), 0.067*** (R5), 0.045 (R6) across specs.
  - Real Expenditure Gap: consistently negative and significant across specs (e.g., -0.080*** to -0.096***).
  - Number of Observations across full-sample specs: 3774, 3774, 3741, 916, 1916, 1908.
  - Adjusted R-Squared across full-sample specs: 0.457, 0.458, 0.558, 0.558, 0.560, 0.565.
- Post-WWII sample (1950–2011) selected coefficients (preserved):
  - Lagged Debt: 0.022** to 0.020* across specs; some specs not significant.
  - Debt*Change in HP Growth: coefficients variable and sometimes significant (examples 0.35; 0.359; 0.872*; 1.238**).
  - Debt*Real LT Interest Rate: strong positive and significant across specs including it (examples 0.257***, 0.303***, 0.273***, 0.351***).
  - Output Gap: strongly positive and significant (examples 0.173***, 0.172***, 0.241***, 0.240***, 0.308***, 0.291***).
  - Number of Observations across postwar specs: 2394, 2394, 1219, 1219, 1219, 1212.
  - Adjusted R-Squared across postwar specs: 0.204, 0.203, 0.299, 0.299, 0.313, 0.315.
- Threshold and illustrative calculations preserved from source:
  - Full sample threshold where interaction dominates: d′ = 0.60 (60 percent of GDP).
  - Post-WWII threshold: about 20 percent of GDP (since 1950, countries with debt above 20 percent of GDP likely to reduce primary balances when hit by negative long-term growth shocks).
  - Example from model (6) in source (preserved arithmetic):
    - α̂2 = 0.86; α̂3 = −51.6. Interaction of debt with a 1 percentage point unexpected decline in long-term real growth implies 100*0.86*-.01 = −0.86 percent of GDP change in primary surplus from the interaction. Direct effect -.01 * -51.6 = +0.516 percent of GDP. Net effect in that example: −0.34 percent of GDP (−0.86 + 0.516).
  - Sovereign borrowing-cost interaction examples (preserved language):
    - For debt = 100 percent and a 100 basis point increase in real long-term borrowing rates, the full sample interaction implies an increase in the primary balance of 0.045 percent of GDP (source example), and the post-WWII interaction implies an increase by 0.35 percent of GDP (source example).
  - Note from text preserved verbatim: "0.30 percent of GDP for the post-WWII period."

### IX. Interpretation, economic rationale, and policy-relevant implications
- Time-variation interpretation:
  - Policymakers may not perceive declines in potential long-run growth in real time; hence they often fail to respond to such declines with sufficient improvements in the primary balance.
  - Increases in the marginal cost of sovereign borrowing prompt policymakers to tighten fiscal policy (higher real long-term yields associated with stronger fiscal response).
  - Time-variation in ρ helps explain structural breaks and episodes of prudence/profligacy across countries and periods (e.g., mid-1970s and 1990s breaks; Japan’s changing response over postwar windows).
- Policy-relevant takeaways (preserved language and findings):
  - Fiscal policy sustainability is not uniform across time or countries; historical episodes show alternation between prudence and profligacy.
  - Sovereign borrowing costs matter: increases in sovereign borrowing costs lead to a stronger fiscal response to rising debt.
  - Growth surprises matter conditional on debt levels: when debt is above estimated thresholds (65 percent of GDP for full sample; as low as 20 percent of GDP for the post-WWII sample), unexpected declines in long-term growth reduce the strength of fiscal adjustment to rising debt.
  - Assessment of crisis-era policy responses (late 2000s) depends on whether the slowdown in economic growth proves long-lasting or growth returns to pre-crisis levels.
- Conceptual note: “prudence” and “profligacy” are used for presentational simplicity but have positive, technical meanings; a “profligate” fiscal response may be normatively justified in some circumstances (for example, to avoid deep and prolonged recessions).

### X. Concluding summary
- The study assembles a historical record of fiscal prudence or profligacy for 55 countries using a comprehensive database on fiscal flows and stocks spanning up to two centuries.
- Key empirical conclusions:
  - Significant variation across countries, over time, and within country histories in whether fiscal policy behavior is consistent with sustainability.
  - Advanced economies generally fiscally prudent during much of their histories, especially pre-WWI and in the 1990s; mid-1970s marks a departure from prudence for many.
  - Latin America: trend toward greater prudence in the 1990s and 2000s after adjustment following widespread defaults in the 1970s and 1980s.
  - The global financial crisis (late 2000s): mixed policy responses across countries; some signaled increased prudence, others profligacy.
- Outputs: detailed country histories using multiple criteria and an accompanying electronic chartbook (as noted in source).

*Source: IMF working paper excerpt (working paper _wp1305, sections summarized as provided).*

### 1. Country Coverage of Main Variables in this Study ...........................................................26

### 1. Country Coverage of Main Variables in this Study ...........................................................26

### I. Introduction and purpose
- The study gauges the degree of fiscal prudence or profligacy for each country over the past several decades using a newly collected historical dataset of fiscal variables for a large panel of countries.
- Dataset components (as described in the source):
  - fiscal revenues, primary expenditures, the interest bill (and thus both the primary and the overall fiscal deficit), the government debt, and gross domestic product.
  - coverage: 55 countries for up to two hundred years.
- Motivating definition: “fiscal prudence” and “fiscal profligacy” are framed in relation to whether the government’s intertemporal budget constraint is met—that is, whether the expected present discounted value of all future fiscal surpluses matches the existing stock of public debt.
- Practical focus: the fiscal stance within government control is proxied by the primary fiscal balance (i.e., the fiscal balance net of interest payments).
- Temporal nature: prudence and profligacy are medium-term concepts; judgments are made over the course of a few years because neither extreme is built up overnight.

### II. Methodological framework
- Primary empirical anchor: Bohn (1998, 2008) fiscal sustainability test, which regresses the primary fiscal surplus on the public debt and other controls.
- Rationale for Bohn approach:
  - Tests based on univariate time-series behavior of the debt-to-GDP ratio are problematic because of difficulties in detecting mean-reversion when the debt-to-GDP ratio is affected by various shocks.
  - The dataset’s inclusion of flows (primary surplus and interest) adds value relative to tests that use only debt stocks.
- Extensions and relaxations of standard Bohn approach employed in the paper:
  - Three variations of the standard Bohn regression for each country:
    - structural break tests,
    - recursive searches for particularly influential observations,
    - iterations of the standard regression over rolling subsamples.
  - A complementary “policymakers’ criterion” that compares the actual primary surplus to the primary surplus required to stabilize the debt-to-GDP ratio (a special case of stationarity assessed by the Bohn test).
- Purpose of methodological choices: to capture variation in the fiscal policy response function across countries and over time within countries, relaxing the assumption of a constant long-term fiscal policy response.

### III. Data advantages and scope
- Novelty claim: for the first time, a large cross-country historical data set covers both fiscal stocks and flows, enabling country-level application of the Bohn test over long historical stretches.
- Contrast with prior work:
  - Bohn’s own work analyzed long-run U.S. time series.
  - Mendoza and Ostry (2008) used panel data for 34 emerging markets and 22 advanced economies over 1990–2005 but constrained the fiscal policy response coefficient to be the same across groups due to short sample length.
- Implication: the new dataset makes it possible to run country-specific tests for a large number of countries and to analyze within-country temporal variation.

### IV. Main empirical findings (summary)
- General result for advanced economies:
  - “For most advanced countries, particularly prior to the global economic and financial crisis that began in 2008, we find evidence that the response of the primary fiscal surplus to variation in government debt is consistent with meeting the intertemporal budget constraint, as well as stationarity of the debt.”
- Temporal and cross-country variation:
  - A given country’s fiscal policy response to changes in debt is not constant through history; significant variation exists across countries and over time within a given country.
  - Periods of a few or more years are distinguishable as clearly “prudent” or “profligate,” with multiple techniques often giving consistent messages.
- Examples of documented episodes of prudence:
  - widespread fiscal prudence in most advanced economies during the mid-1990s until at least the mid-2000s;
  - prudence in emerging economies becomes more widespread after the year 2000.
  - country-specific episodes:
    - United States: strong prudence in the late 1990s (discussion of possible disappearance of public debt).
    - Canada: prudence since the mid-1990s (ambitious and successful fiscal adjustment plan).
    - several Euro area countries: prudence during the mid-1990s (coinciding with the Maastricht Euro entry process).
    - Ireland: prudence in the late 1980s and early 1990s (known fiscal adjustment episode).
    - Japan: prudence in the mid-1980s to early 1990s (sought to stabilize the debt).
    - Turkey: prudence in the mid-1990s and at several points in the 2000s (improved its primary balance significantly).
- Examples of notable episodes of fiscal stimulus (profligacy) noted in the text:
  - United States in 2009–11,
  - Spain (episode mentioned but date cut off in provided excerpt).

### V. Analytical contributions and interpretation
- Contribution: documents how individual countries fare with respect to fiscal prudence and profligacy using multiple methods (Bohn variations and policymakers’ criterion) and provides country-level results in the working paper’s tables and charts and in an accompanying Chartbook.
- Conceptual note: the terms “prudence” and “profligacy” are used for presentational simplicity but are given positive, technical meanings in the analysis; a “profligate” fiscal response may be normatively justified in some circumstances (for example, to avoid deep and prolonged recessions).
- Empirical strategy highlights:
  - Use of long historical panel to detect structural breaks and time variation in fiscal response functions.
  - Algorithmic combination of criteria to generate a reasonable gauge of prudence/profligacy over time for each country.

*Source: IMF working paper excerpt, section I. INTRODUCTION and table of contents for the working paper.*

### 2010. And Japan is found not to sufficiently improve its primary balance despite rising

### _wp1305 - 2010. And Japan is found not to sufficiently improve its primary balance despite rising

### Data, coverage, and construction
- Panel: unbalanced panel of 55 countries (24 advanced economies and 31 non-advanced) over 1800–2011.
- Fiscal variables: government revenue, non-interest government expenditure, interest bill, overall fiscal balance, primary balance, gross public debt (all expressed as a share of GDP).
- Observation counts and series lengths:
  - Revenue, expenditure, overall fiscal balance, and debt: about 5,700 observations each.
  - Interest expenditure (and consequently primary expenditure and primary balance): approximately 4,800 observations.
  - Both debt and primary balance available for about 4,500 country-years.
- GDP and historical series sources: WEO, IFS, OECD, Statistical Yearbooks of the League of Nations and the United Nations, Flandreau and Zumer (2004), Mitchell’s International Historical Statistics, MOXLAD, country-specific sources.
- Sector coverage: prioritized most comprehensive government sector available (general government when available); sector switches recorded via dummy variables.
- War and default exclusions: country-years with expenditure-to-GDP increases ≥ six percentage points in one year (plus up to two post-war years) and default years (Reinhart and Rogoff (2010)) excluded in many empirical exercises.

### Summary statistics and stylized fiscal facts
- Averages (full sample, 1800−2011, subject to availability):
  - Revenue: 19 percent of GDP.
  - Expenditure: 21 percent of GDP.
  - Public sector interest bill: 2½ percent of GDP.
  - Primary surplus (sample average): ½ percent of GDP.
  - Debt (average): 50 percent of GDP.
  - Countries generally faced a negative interest-growth differential.
- Distributional notes:
  - Primary surpluses in the top percentile exceed 9½ percent of GDP (largely commodity producers or countries with large government assets).
  - Top five percent of primary surplus distribution: above 5½ percent of GDP.
- Historical patterns:
  - Sharp primary balance decreases and debt increases during World Wars; primary deficits reverse quickly after wars while postwar debts decline more gradually.
  - Post-WWII: debts decline until the 1970s, then generally increase from the 1970s onward; strong primary balance improvements in the second half of the 1990s (Maastricht-related consolidations).
  - The 2008–09 global crisis produced the most pronounced peacetime worsening of the primary fiscal balance in their long-run series.

### Variance decomposition of debt changes (1950–2011, advanced vs non-advanced)
- For advanced economies: fluctuations in primary balance, interest-growth differentials, and stock-flow changes each explain roughly one-third of the variance of changes in the debt ratio. High stock-flow residual variances correspond to countries with sizable asset accumulation (Norway, Sweden, Finland, Japan) and to Greece (inflation, defaults, restructurings).
- For non-advanced economies: more volatile debt changes are largely explained by stock-flow residuals and interest-growth differentials (defaults, high inflation, exchange rate crises), though heterogeneity exists (e.g., Colombia and India similar to advanced economies).

### Methods to measure fiscal prudence/profligacy
- Bohn (1998) fiscal reaction function (time series per country): s_t = α + ρ d_t + Z_t + ε_t (notation from source).
  - If estimated ρ > 0 and significant, fiscal policy consistent with intertemporal budget constraint under uncertainty; stationarity condition requires (1/(r−g)) ρ > 0 in Bohn’s formulation (paper uses medium-run averages of r and g).
- Complementary approaches used to capture time variation:
  1. Structural break tests (Bai and Perron (1998)) to partition history into subperiods of differing fiscal response.
  2. Search for influential observations: recursive omission of single years to identify “most profligate” or “most prudent” years that change significance of ρ.
  3. Rolling-window and expanding-sample Bohn regressions (e.g., 25-year rolling windows; expanding windows starting 1950) to capture evolving prudence detectable in shorter samples.
- Policymakers’ criterion: comparison between actual primary surplus and debt-stabilizing primary surplus (from the debt motion equation); used as a contemporaneous benchmark.

### Findings on fiscal prudence/profligacy (country and historical results)
- Whole-sample Bohn tests:
  - Using full sample, the Bohn coefficient is positive and significant for three fourths of advanced economies.
  - In many cases the estimated coefficient exceeds the level necessary for stationarity of the public debt ratio.
  - The addition of crisis years (2008–2011) materially affects estimates for many countries (e.g., United States post-2007 loses significance and changes sign when crisis years included).
  - Notable exceptions with negative and significant Bohn coefficients among large advanced economies: France and Japan (France partly reflects over-accumulation of assets 1950–1977; Japan shows weak primary response despite rising debts from late 1990s).
- Structural breaks:
  - Fiscal policy response to debt is non-constant over time for nearly every tested country.
  - Common break periods: mid 1970s and the 1990s (and earlier around the first decade of the 1900s and the 1930s when using the entire sample).
  - Mid-1970s breaks often mark a shift to a positive and significant response (coinciding with oil shocks, slower growth, higher real rates). Breaks in the 1990s frequently represent relaxation of fiscal restraint for two-thirds of advanced economies.
- Influential observation search:
  - Many prudent years during 1880–1913 for countries with long data.
  - Few influentially profligate observations in 1950s–1960s (debts broadly declining).
  - Several advanced economies influentially prudent in the mid- or late 1990s (consolidation episodes); many countries influentially profligate in at least one year since 2008 (global financial crisis impact).
  - Early to mid-1970s appear profligate in a fairly large number of countries.
- Rolling windows and expanding-sample windows:
  - Example, Japan: rolling 25-year windows show negative and significant response coefficients for windows ending in late 1970s–1980s; positive and significant in late 1980s–early 1990s; descent to profligacy from 1995 through present (as primary balance trends worsened while debt rose).
  - Table 8: countries are measured prudent far more often than profligate; most profligacy observations concentrated in last four decades and notably after 2007 for many countries.
  - Expanding-sample windows beginning 1950: most observed profligacy in mid and late 1970s; by the 2000s most countries measured as prudent, and most remained prudent through the crisis that began in 2008—exceptions include United States, Spain, Portugal, and Iceland.
- Summary classification:
  - “Strongly prudent” country-year: non-war, non-default year meeting three tests (part of at least one 25-year prudent window; influentially prudent; actual primary balance > forward-looking debt-stabilizing primary balance).
  - “Profligate” country-year: non-war, non-default year failing prudence across multiple criteria (not in any 25-year prudent window; actual primary balance < debt-stabilizing primary balance; and either part of ≥1 25-year profligate window or influentially profligate).
  - Empirical instances:
    - Strong prudence: United States in late 1990s; Canada since mid-1990s; several Euro area countries in mid-1990s (Italy, Belgium, Netherlands, Greece in early 1990s); Ireland in late 1980s/early 1990s; Japan mid-1980s to early 1990s; Turkey mid-1990s and points in 2000s.
    - Notable profligate/stimulus episodes: United States in 2009–11; Spain in 2010; several years since late 1990s where Japan did not sufficiently improve its primary balance despite rising debts.

### Drivers of time-variation in fiscal responsiveness (panel estimates and interpretation)
- Extended reaction function estimated in panels with country fixed effects (equation forms and notation preserved as in source). Key variables: debt, long-term growth surprises (g'—difference between full-information long-term growth and contemporaneous limited-information forecast), and real long-term sovereign borrowing costs (10-year domestic-currency yield net of GDP deflator).
- Main panel regression findings (Table 14, full sample and post-WWII):
  - Estimated baseline Bohn coefficient (α̂1) ≈ 0.03 for full sample (interpreted: debt-to-GDP = 100 => predicted primary surplus = 3 percent of GDP absent other controls).
  - Post-WWII Bohn coefficient is lower and sometimes statistically insignificant, suggesting weaker prudence after WWII.
- Impact of long-term growth surprises:
  - Interaction coefficient estimates (α̂2) range from 0.1−0.9 across models and are statistically significant when controlling for growth surprises in some specifications (e.g., models (2) and (6)).
  - Example (model (6)):
    - α̂2 = 0.86 (interaction of debt with a 1 percentage point unexpected decline in long-term real growth implies 100*0.86*-.01 = −0.86 percent of GDP change in primary surplus from the interaction).
    - α̂3 = −51.6 (direct effect of the negative growth surprise implies an upward correction in primary balance of 0.52 percent of GDP: -.01 * -51.6 = +0.516).
    - Net effect in that example: decrease of 0.34 percent of GDP (−0.86 + 0.516).
  - Threshold logic:
    - For countries with debt above d′ = (−α̂3/α̂2) the interaction term dominates and negative long-term growth surprises lead to overall worsening of the primary balance.
    - Full sample threshold: d′ = 0.60 (60 percent of GDP).
    - Post-WWII threshold: about 20 percent of GDP. Thus since 1950, countries with debt above 20 percent of GDP are likely to reduce primary balances when hit by negative long-term growth shocks.
- Impact of sovereign borrowing costs:
  - Interaction estimates (α̂4) ~ 0.05 (full sample) and ~ 0.35 (post-WWII).
  - Example: for debt = 100 percent and a 100 basis point (1 percentage point) increase in real long-term borrowing rates:
    - Full sample interaction implies increase in primary balance by 0.045 percent of GDP (100 * 0.05 * 0.01 = 0.05? — preserving source text example: "0.045 percent of GDP based on the full sample estimate").
    - Post-WWII interaction implies increase by 0.35 percent of GDP (source example).
  - Direct impact point estimates: +0.02 percent of GDP for full sample; −0.05 percent of GDP (insignificant) for post-WWII.
  - Overall net impacts reported in source (full sample and post-WWII) combine direct and interaction effects (source text provides partial numeric examples; full-sample and post-WWII net impacts discussed in context).

### Interpretation and economic rationale (as presented)
- Policymakers may not perceive declines in potential (long-run) economic growth in real time; hence they often fail to respond to such declines with sufficient improvements in the primary balance (growth surprises associated with weaker fiscal response).
- Conversely, increases in the marginal cost of sovereign borrowing prompt policymakers to tighten fiscal policy (higher real long-term yields associated with stronger fiscal response).
- The time-variation in fiscal reaction (ρ) helps explain observed structural breaks and episodes of prudence/profligacy across countries and periods (e.g., mid-1970s and 1990s breaks; Japan’s changing response over postwar windows).

*Source: IMF working paper chapter content as provided.*

### 0.30 percent of GDP for the post-WWII period. Because both the interaction term and the

### _wp1305 - 0.30 percent of GDP for the post-WWII period. Because both the interaction term and the

### VI. Conclusions — overall summary
- The study assembles a historical record of fiscal prudence or profligacy for 55 countries over the past two centuries using the most comprehensive database available to date on fiscal flows and stocks.
- Evidence of significant variation across countries, over time, and within individual country histories in the extent to which fiscal policy behavior is consistent with sustainability.
- Advanced economies: generally fiscally prudent during most of their histories; especially prudent in the first era of global finance (pre-WWI) and the 1990s; mid-1970s represents a departure from prudence.
- Global financial crisis (late 2000s): policy responses varied — some signaled increased prudence, others profligacy — possibly reflecting differences in market conditions and sovereign borrowing costs.
- Latin America: broad trend toward greater prudence in the 1990s and 2000s after adjustment following widespread defaults in the 1970s and 1980s.
- The paper provides detailed country histories using various criteria and is accompanied by an electronic chartbook.

### Main empirical findings (panel regression and interactions)
- When a country has a public debt stock above an estimated threshold:
  - Thresholds: 65 percent of GDP for the whole sample period; as low as 20 percent of GDP for the pre-WWII period.
  - Unexpected declines in long-term growth lead to weaker increases of the primary fiscal balance in response to rising debt.
  - Increases in sovereign borrowing costs lead to a stronger policy response to rising debt.
- Note from text: "0.30 percent of GDP for the post-WWII period." (preserved verbatim as in source).
- Footnote style result: If one were to consider ߙොହ൏0 despite lack of statistical significance, the net overall impact would be positive as long as the debt is above ߙොହߙୋସ⁄, which equals 16 percent in the post-WWII sample.

### Data coverage and country list (Table 1 — country coverage)
- Sample: up to 55 countries; observations and coverage vary by country. Examples (preserve exact figures as listed):
  - Argentina: 148 observations; coverage 1864-2011; 89.0 percent other hand-collected sources.
  - Australia: 99 observations; coverage 1913-2011; 0.46 percent country-specific sources.
  - Sweden: 212 observations; coverage 1800-2011; 91.0 percent country-specific sources.
  - United States: 212 observations; coverage 1800-2011; 75.3 percent country-specific sources.
- Notes: Table provides number of observations for which Debt, Primary Balance and macro variables are available and share of observations drawn from country-specific and other hand-collected sources.

### Summary statistics (Full sample 1800–2011; Tables 2–4)
- All countries (Full sample):
  - Revenue: Mean 19.01; Std. Dev. 13.6; Min 0.5; 25% 8.5; Median 14.7; 75% 26.54; 95% 47.45; 99% 55.26; Max 60.55; No. Obs 721
  - Expenditure: Mean 20.9; Std. Dev. 14.4; Min 0.7; 25% 9.6; Median 16.9; 75% 29.9; 95% 50.1; 99% 56.67; Max 71.85; No. Obs 742
  - Interest Expenditure: Mean 2.4; Std. Dev. 2.3; Min 0.0; 25% 0.8; Median 1.7; 75% 3.3; 95% 6.5; 99% 10.32; Max 14.64; No. Obs 862
  - Overall Balance: Mean -1.8; Std. Dev. 4.0; Min -39.1; 25% -3.1; Median -1.0; 75% 0.1; 95% 2.2; 99% 5.4; Max 19.15; No. Obs 739
  - Primary Balance: Mean 0.4; Std. Dev. 3.9; Min -35.5; 25% -1.0; Median 0.5; 75% 2.2; 95% 5.5; 99% 9.5; Max 20.64; No. Obs 834
  - Debt: Mean 49.74; Std. Dev. 42.7; Min 0.0; 25% 19.2; Median 37.9; 75% 66.2; 95% 133.2; 99% 215.52; Max 289.65; No. Obs 670
  - Interest-Growth Differential: Mean -3.9; Std. Dev. 9.1; Min -34.9; 25% -8.4; Median -2.8; 75% 1.5; 95% 9.0; 99% 18.03; Max 34.24; No. Obs 266
- Post-WWII (1950–2011) — All countries:
  - Revenue: Mean 25.11; Std. Dev. 13.4; Min 2.1; 25% 13.9; Median 21.9; 75% 34.7; 95% 49.8; 99% 56.4; Max 60.53; No. Obs 126
  - Expenditure: Mean 27.3; Std. Dev. 13.8; Min 1.7; 25% 16.1; Median 24.0; 75% 37.5; 95% 52.6; 99% 68.67; Max 81.83; No. Obs 153
  - Interest Expenditure: Mean 2.6; Std. Dev. 2.6; Min 0.0; 25% 0.8; Median 1.9; 75% 3.6; 95% 7.4; 99% 12.02; Max 14.62; No. Obs 788
  - Overall Balance: Mean -2.2; Std. Dev. 3.9; Min -31.3; 25% -4.0; Median -1.8; 75% 0.0; 95% 2.8; 99% 7.0; Max 19.13; No. Obs 152
  - Primary Balance: Mean 0.3; Std. Dev. 3.5; Min -28.2; 25% -1.5; Median 0.3; 75% 2.2; 95% 5.9; 99% 6.2?; Max 20.62; No. Obs 803
  - Debt: Mean 45.1; Std. Dev. 34.3; Min 0.1; 25% 19.7; Median 37.8; 75% 60.6; 95% 106.7; 99% 169.92; Max 289.62; No. Obs 983
  - Interest-Growth Differential: Mean -5.9; Std. Dev. 9.0; Min -34.9; 25% -10.4; Median -4.5; 75% 0.0; 95% 5.7; 99% 13.53; Max 34.22; No. Obs 471
- Advanced economies (Postwar):
  - Revenue: Mean 33.31; Std. Dev. 12.8; No. Obs 481
  - Expenditure: Mean 35.41; Std. Dev. 13.6; No. Obs 479
  - Primary Balance: Mean 0.7; Std. Dev. 3.5; No. Obs 467
  - Debt: Mean 49.4; Std. Dev. 34.0; No. Obs 458
  - Interest-Growth Differential: Mean -3.1; Std. Dev. 7.0; No. Obs 360

### Decomposition of variance of debt changes (Tables 5 & 6)
- Advanced economies (1950-2011) — selected country examples:
  - United States (61 obs): ∆Debt 15.7; Prim. Bal. 6.6; (r-g)d 5.1; Residual 1.2; C(PB,(r-g)d) -0.3; C(PB,Resid) -0.4
  - Japan (61 obs): ∆Debt 41.8; Prim. Bal. 10.1; (r-g)d 7.9; Residual 11.8; C(PB,(r-g)d) 0.5; C(PB,Resid) 0.9; C((r-g)d,Resid) 0.6
  - Average (advanced): ∆Debt 21.08; Prim. Bal. 8.9; (r-g)d 14.3; Residual 14.6; C(PB,(r-g)d) 0.1; C(PB,Resid) -5.3; C((r-g)d,Resid) -11.5
  - Median (advanced): ∆Debt 17.3; Prim. Bal. 8.7; (r-g)d 6.9; Residual 8.7; C(PB,(r-g)d) -0.1; C(PB,Resid) -2.1; C((r-g)d,Resid) -1.5
- Emerging economies (1950-2011) — selected country examples:
  - Argentina (38 obs): ∆Debt 423.5; ∆Prim.Bal. 5.9; (r-g)d 39.9; Residual 346.8; C(PB,(r-g)d) 11.7; C(PB,Resid) -3.5; C((r-g)d,Resid) 22.6
  - Chile (30 obs): ∆Debt 56.6; ∆Prim.Bal. 15.7; (r-g)d 48.8; Residual 74.1; C(PB,(r-g)d) 13.5; C(PB,Resid) -4.8; C((r-g)d,Resid) -90.8
  - Average (emerging): ∆Debt 72.8; ∆Prim.Bal. 8.2; (r-g)d 28.1; Residual 76.1; C(PB,(r-g)d) 0.4; C(PB,Resid) -5.6; C((r-g)d,Resid) -37.0
  - Median (emerging): ∆Debt 45.5; ∆Prim.Bal. 6.0; (r-g)d 19.8; Residual 41.2; C(PB,(r-g)d) 2.0; C(PB,Resid) -4.7; C((r-g)d,Resid) -31.5

### Bohn (1998) sustainability tests and results (Tables 7–13)
- Bohn regressions (country-by-country OLS with robust SEs; debt coefficients reported). Selected country coefficients (full-sample/postwar windows preserved as in table):
  - Sweden: 0.038***; 0.010; 0.085***; 0.084***; N=202
  - United States: -0.006; 0.042***; -0.012; 0.064***; N=177
  - United Kingdom: 0.043***; 0.005; 0.016***; 0.017***; N=160
  - Canada: 0.056***; 0.077***; 0.077***; 0.087***; N=134
  - Japan: -0.026***; -0.055***; -0.03***; -0.027***; N=125
  - Korea: 0.149***; 0.148***; N=53
  - Philippines: 0.092***; 0.093***; N=57
  - Brazil: 0.001; 0.002; N=28
- Rolling 25-year window Bohn regressions (Table 8) — iteration results:
  - Reports windows where debt response coefficient (ρ) is significantly positive or negative, changes to positive, changes from positive, percent positive, percent negative, number of windows, first window (final year reported).
  - Example: United Kingdom — Significantly Positive windows include 1854-64, 1866-70, 1872-7, 1880-1913, 1978-86, 1990-94, 2003-10; none significantly negative; Percent Positive 75; Percent Negative 0; Number of Windows 104; First Window 1854.
  - Example: Japan — Significantly Positive windows 1906-33, 1988-96; Significantly Negative windows 1902-3, 1939-41, 1976-81, 2002-11; Percent Positive 45; Percent Negative 26; Number of Windows 82; First Window 1899.
- Iterative expanding-sample Bohn regressions (Table 9):
  - Example: Australia — Significantly Positive 1976-2002; Significantly Negative none; Postwar Obs. 62.
  - United States: Significantly Positive 1974-2008; Significantly Negative none; Postwar Obs. 62.
  - Japan: Significantly Positive 1990-4; Significantly Negative 1976-83, 2002-11; Postwar Obs. 59.
- Structural breaks in Bohn test (post-war and full sample; Tables 10 & 11 — Bai and Perron tests):
  - Postwar structural breaks — selected examples:
    - Australia: breaks at 1965 and 1985; estimated ρ subsamples: -0.078*, 0.097***, 0.082**; N=58; Sample 1950-2007.
    - Canada: breaks at 1974, 1993, 1975? (as listed); ρ estimates include 0.06**, 0.035*, 0.247***; N=58; Sample 1950-2007.
    - United States (full-sample Table 11): three breaks with ρ estimates 0.157***, 0.076***, 0.09***; Obs. 184; Sample 1801-2011.
  - Full sample structural breaks — selected examples:
    - Sweden: 2 breaks; 1852 and 1972; ρ estimates -0.072, 0.016**, 0.141***; Obs. 198; Sample 1801-2011.
    - United Kingdom: 3 breaks; years including 1867, 1931, 1986?; ρ estimates 0.046***, 0.065***, 0.031***; Obs. 158; Sample 1830-2011.
- Influential years (recursive Bohn test, Table 12):
  - The study identifies specific years that make previously significant ρ estimates become insignificant (or vice versa). Examples:
    - Sweden influential years include 1803, 1981-2, 1987-9, 1993-4, 1996-2000, 2006-7; original estimate 0.038***.
    - United States influential years include many episodes (1936-37, 1939, 1941, 1948-51, 1959, 1975, 1983-88, 1990-93, 1998-2000, 2003-05, 2008-11); original estimate -0.006.
    - Ireland original estimate 0.062***; influential years include 1937, 1960, 1979-83, 1985-91, 1992-3, 1997, 2006, 2009-11.
- Times of strong prudence and profligacy (Table 13) — selected country examples:
  - Sweden — Times of Strong Prudence: 1987-9, 1996-2001, 2006-7; Times of Profligacy: 1915, 1920-2, 1953, 1957; N=202
  - United States — Prudence: 1948-51, 1998-2000; Profligacy: 1934-6, 2009-11; N=177
  - United Kingdom — Prudence: 1830-6, 1841-6, 1850, 1856, 1860, 1906-13, 1954-9, 2002; Profligacy: none; N=160
  - Canada — Prudence: 1881, 1891, 1906, 1909-13, 1933, 1937-8, 1946-8, 1995-2003; Profligacy: none; N=134
  - Japan — Prudence: 1907, 1985-92; Profligacy: 1880, 1997-2011; N=125

### Panel regressions: fiscal response as function of growth surprises and borrowing costs (Table 14)
- Estimation framework: primary balance reaction function sd_t = α + β1 d_t + β2 (interactions and shocks) + controls Z_t; two sample periods reported: full sample 1800-2011 (top) and postwar 1950-2011 (bottom).
- Full sample (selected coefficients, exact preserved):
  - Lagged Debt: 0.027*** (R2), 0.027*** (R3), 0.029*** (R4), 0.030*** (R5), 0.030*** (R6), 0.029*** (R1)
  - Debt*Change in HP Growth: 0.063 (R2), 0.179** (R3), 0.33 (R4), 0.861** (R5) — coefficients vary by specification
  - Change in HP Growth: -13.34 (R2), -51.565* (R3) — large negative point estimates in some specs
  - Debt*Real LT Interest Rate: 0.067*** (R3), 0.049 (R4), 0.067*** (R5), 0.045 (R6)
  - Real LT Interest Rate: 1.859 (R3), 2.403 (R4)
  - Output Gap: 0.046** (R2), other specs 0.031, 0.053, 0.053, 0.084*, 0.043
  - Real Expenditure Gap: -0.080***, -0.081***, -0.096***, -0.096***, -0.093***, -0.093*** (across specs)
  - Non-fuel Commodity Prices: -0.001, -0.001, -0.085***, -0.088***, -0.080***, -0.068***
  - Number of Observations: 3774, 3774, 3741, 916, 1916, 1908 (as listed for each spec)
  - Adjusted R-Squared: 0.457, 0.458, 0.558, 0.558, 0.560, 0.565 (by spec)
- Postwar (1950-2011) (selected coefficients, exact preserved):
  - Lagged Debt: 0.022** (R2), 0.022** (R3), 0.015 (R4), 0.014 (R5), 0.020* (R6), 0.018 (R1)
  - Debt*Change in HP Growth: 0.35, 0.359, 0.872*, 1.238** (coefficients vary by spec)
  - Change in HP Growth: -0.769, -26.515 (large negative estimates in some specs)
  - Debt*Real LT Interest Rate: 0.257***, 0.303***, 0.273***, 0.351*** (strong positive and significant across specs that include it)
  - Real LT Interest Rate: -3.235, -5.432 (negative point estimates where included)
  - Output Gap: 0.173***, 0.172***, 0.241***, 0.240***, 0.308***, 0.291*** (strong positive)
  - Real Expenditure Gap: -0.129***, -0.129***, -0.273***, -0.274***, -0.270***, -0.270*** (strong negative)
  - Non-fuel Commodity Prices: -0.006*, -0.006*, -0.343***, -0.340***, -0.345***, -0.329*** (negative and sometimes strongly significant)
  - Number of Observations: 2394, 2394, 1219, 1219, 1219, 1212 (by spec)
  - Adjusted R-Squared: 0.204, 0.203, 0.299, 0.299, 0.313, 0.315 (by spec)
- Interpretation preserved from source: higher sovereign borrowing costs (real LT interest rate) are associated with stronger fiscal responses to rising debt; unexpected declines in long-term growth weaken fiscal responses to rising debt once debt is above estimated thresholds.

### Visual and illustrative findings (figures)
- Figure 1 (Government Debt and Primary Fiscal Balance, 1850−2011):
  - Presents unweighted and GDP-weighted mean and median of government debt and primary fiscal balance as shares of GDP; 15th-85th percentile ranges; country coverage bars over time.
- Figure 2 (Japan rolling regressions, 1875-2011):
  - Rolling 25-year Bohn coefficient series and p-values: blue line = coefficient value; red line = p-value; criterion for fiscal sustainability: blue > 0 and red < .05.
- Figures 3–11:
  - Periods of prudence and profligacy across countries and time periods, marking influentially prudent/profligate years, structural breaks, war years, and default years; figure panels for multiple country groups and time windows.

### Policy-relevant implications and interpretation (preserved language)
- Fiscal policy sustainability is not uniform across time or countries; historical episodes show alternation between prudence and profligacy.
- Sovereign borrowing costs matter: increases in sovereign borrowing costs lead to a stronger fiscal response to rising debt.
- Growth surprises matter conditional on debt levels: when debt is above estimated thresholds (65 percent of GDP full sample; as low as 20 percent pre-WWII), unexpected declines in long-term growth reduce the strength of fiscal adjustment to rising debt.
- Assessment of crisis-era policy responses (late 2000s) will depend on whether the slowdown in economic growth proves long-lasting or growth returns to pre-crisis levels.

*Source: Authors’ estimates and tables/figures as presented in the supplied content.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1305.pdf_