## Can Fiscal Consolidation Announcements Help Anchor Inflation Expectations?

## Source details

**Canonical URL:** [Can Fiscal Consolidation Announcements Help Anchor Inflation Expectations?](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024060-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024060-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024060-print-pdf.pdf.json)

---

### Overview and motivation
- Research question: How do fiscal consolidation announcements shape inflation expectations?
- Motivation:
  - Resurgence of inflation after COVID-19 accompanied by signs of de-anchoring of near-term inflation expectations.
  - Fiscal consolidation announcements may reduce inflation through:
    - demand effects captured by the output gap in Phillips-curve dynamics;
    - expectations effects via the “fiscal theory of the price level” (higher expected surpluses → decline in inflation expectations).

### Data, identification, and empirical strategy
- Frequency and identification:
  - Quarterly data used to control for anticipation effects; key identification assumption: fiscal consolidations are unlikely to be adjusted to changes in inflation expectations within the same quarter.
  - Country and time fixed effects included; a set of control variables attenuates omitted variable bias.
- Key data sources and variables:
  - Inflation expectations and dispersion: Consensus Economics consensus forecasts at horizons current year, 1 year, 2–5 years, and 10 years; analysis focuses on horizons 1-year, 3-year, 5-year.
  - Fiscal consolidation announcements: narrative-updated databases (original coverage advanced economies 1978–2013; emerging markets 2000–2018) updated to December 2023, adding 75 announcements (34 advanced economies, 41 emerging markets).
  - Controls: WEO observed inflation and real GDP, INS nominal effective exchange rate, Hamilton-filtered output gaps, fiscal prudence gauges (gap between observed fiscal balances and IMF t-3 projections), global debt database (GDD), AREAER inflation-targeting indicator.

### Baseline regression specification
- Main specification: π_{i,t}^{e,h} = α_i + γ_t + θ (fiscal consolidation)_{i,t} + β X_{i,t} + ε_{i,t}, for horizons h = 1, 3, 5 years.
- Controls X_{i,t} include current and lagged realized inflation; current and lagged output gap; current and lagged y/y changes in nominal effective exchange rate; proxy for short-term expectations dynamics.
- Main coefficient of interest: θ (direct effect of fiscal consolidations on inflation expectations).
- Dispersion regressions: same specification with dependent variable = standard deviation of forecasts (and alternative highest-lowest gap normalized by median).

### Baseline findings on inflation expectations and dispersion
- Direction and significance:
  - Consolidation announcement coefficients are negative across horizons.
  - Statistically significant at conventional levels for the 3-year and 5-year horizons, muted at the 1-year horizon.
- Key coefficients (Table 2, consolidation announcement dummy):
  - 1-year: -0.0294 (std. error 0.0350).
  - 3-year: -0.1160** (std. error 0.0572).
  - 5-year: -0.1078** (std. error 0.0449).
  - Alternate columns:
    - 1-year: -0.0188 (std. error 0.0534).
    - 3-year: -0.1540* (std. error 0.0827).
    - 5-year: -0.1565** (std. error 0.0707).
- Dispersion (Table 3, consolidation announcement dummy):
  - 1-year Dispersion: -0.0109 (std. error 0.0246).
  - 3-year Dispersion: -0.0877** (std. error 0.0337).
  - 5-year Dispersion: -0.0532** (std. error 0.0253).
- High-Low gap (Table 4, consolidation announcement dummy):
  - 1-year High-Low gap: 0.0558** (std. error 0.0270).
  - 3-year High-Low gap: -0.0443** (std. error 0.0219).
  - 5-year High-Low gap: -0.0129 (std. error 0.0215).
- Sample sizes and fit (Table 2):
  - Observations: 3,084 (1-year), 1,411 (3-year), 1,410 (5-year).
  - R-squared: 0.938 (1-year), 0.852 (3-year), 0.748 (5-year).

### Heterogeneity by composition of consolidation
- Tax vs expenditure effects (Table 5):
  - Tax-based consolidations (inflation expectations):
    - 3-year: -0.1627*** (std. error 0.0532).
    - 5-year: -0.1739*** (std. error 0.0367).
  - Tax-based consolidations (dispersion):
    - 3-year Dispersion: -0.1529*** (std. error 0.0426).
    - 5-year Dispersion: -0.0811*** (std. error 0.0268).
  - Expenditure-based consolidations:
    - 3-year inflation expectations: -0.1794* (std. error 0.0937).
    - No statistically significant negative effects on dispersion at conventional levels.
  - Combined tax and expenditure coefficients imprecise (large standard errors).

### Heterogeneity by fiscal prudence, government debt, monetary framework, and initial conditions
- Government debt and fiscal prudence (Table 6):
  - Consolid. announcement dummy:
    - 3-year: -0.3208** (std. error 0.1245).
    - 5-year: -0.3034*** (std. error 0.1086).
  - Government debt (percent of GDP) contemporaneous:
    - Column (1): -0.0051** (std. error 0.0021).
    - Column (2): -0.0068* (std. error 0.0036).
  - Government debt × Consolid. ann.:
    - Column (2): 0.0037** (std. error 0.0016).
    - Column (3): 0.0033** (std. error 0.0013).
  - Consolid. ann. × Dev. of fiscal balance relative to WEO t-3:
    - Column (3): -0.0255** (std. error 0.0120).
- Inflation-targeting regime (Tables 7 and 8):
  - Inflation Targeting Countries (full sample):
    - 3-year: -0.0483** (std. error 0.0234).
    - 5-year: -0.0453 (std. error 0.0362).
  - Non-Inflation Targeting Countries: coefficients imprecise and often not statistically significant.
  - Pre-2020 Inflation Targeting Countries show stronger and more significant negative effects (e.g., 3-year: -0.0593** (std. error 0.0238); 5-year: -0.0792*** (std. error 0.0278)).
- Monetary policy stance (Table 9):
  - Consolid. ann. × Tight monetary policy dummy:
    - Column (5): -0.2248** (std. error 0.1065).
  - Consolid. ann. × High Inflation dummy (high inflation defined as y/y inflation above the 75th percentile of the country-specific 1990–2023 distribution):
    - Column (2): -0.8400** (std. error 0.4030).
    - Column (3): -0.9064*** (std. error 0.2699).
    - Column (6): -0.9421*** (std. error 0.2322).
  - Interpretation: tight monetary policy and high-inflation episodes amplify the inflation-anchoring role of consolidation announcements.

### PVAR framework, counterfactuals, and transmission to realized inflation
- PVAR setup:
  - Variables at quarterly frequency: fiscal announcement dummy D_{i,t} (1 in quarter of announcement), 3-year ahead expectations E_{i,t} (y/y percent), realized inflation π_{i,t} (y/y percent).
  - Cholesky identification: fiscal announcements first, expectations second, realized inflation third; timing restriction: fiscal policy does not react contemporaneously to other shocks.
  - Additional assumption a_{i,2,3} = 0 (no contemporaneous effect of realized inflation on expectations).
- Counterfactual methodology:
  - Construct impulse responses holding the response of expected inflation fixed at zero at all horizons (IRF_2(1,h) = 0).
  - Compute sequence of expectation shocks ε_{i,h}^2 required to enforce IRF_2(1,h) = 0; derive modified impulse responses IRF̃_k(1,h).
  - Difference IRF_k(1,h) − IRF̃_k(1,h) measures importance of expectations channel.
- Nonlinearities:
  - Coefficients allowed to vary with inflation level: A_{i,j} = β_{i,j} + ς_{i,j} * πlevel_{i,t}.
  - Impulse responses estimated across the empirical distribution of realized inflation levels.
- Estimation and inference:
  - Each equation estimated by OLS with 4 lags.
  - Runkle (1987) bootstrap for panel inference; 1,000 bootstrap replications used.

### PVAR findings (dynamic responses and counterfactual comparisons)
- High-inflation environments (realized inflation at or above the 75th percentile):
  - Expectations decline materially after announcements: around 10 basis points one year after the announcement.
  - Realized inflation: short-term increase following announcements but inflation falls by four quarters after one year.
  - Counterfactual (expectations held fixed at zero): realized inflation response is higher at most horizons; baseline inflation is almost half the counterfactual response one year after the announcement.
  - Main quantitative summary (from the Introduction):
    - For economies with high realized inflation (≥ 75th percentile), a fiscal consolidation announcement reduces headline inflation by 100 basis points in a one-year window. In the counterfactual where the response of inflation expectations is zeroed out, the reduction in inflation is much more gradual.
- Low-inflation environments (realized inflation below the 25th percentile):
  - Role of fiscal policy and expectations limited; differences between baseline and counterfactual impulse responses are economically small and not statistically significant.
- Bootstrapping and simulation procedure (steps reproduced exactly):
  - 1,000 bootstrap replications 퐴̂ᵢ generated using equation (9) and selected values for inflation spreads in quarter t.
  - Simulation steps include: simulate first period for all variables using bootstrap draws; interact variables with interaction terms and iterate for t = 2, ..., T and i = 1, . . . , N; re-estimate system coefficients on artificial sample and compute IRFs 1,000 times; construct 1,000 adjusted impulse responses holding expectations response at zero; use bootstrap to calculate empirical distribution and one standard deviation confidence bands.
  - Panel bootstrapping: initial conditions generated separately by country (Runkle, 1987) with sampling from the entire panel vector of residuals to account for cross-country correlations.

### Robustness and additional specifications
- Country-period fixed effects across subperiods (1990–2000; 2001–2008; 2009–2019; post-COVID) do not qualitatively change patterns: consolidation announcements associated with lower medium-term inflation expectations (5-year) and reduced dispersion, though magnitudes smaller and some significance lost.
- Leads and lags: including fiscal consolidation t-1 and t+1 leaves the consolidation coefficient significantly negative; effects materialize mostly on impact with no evidence of protracted impacts.
- Excluding pandemic/post-pandemic quarters: connection stronger in pre-2020 sample; pre-2020 coefficients larger for expectations and dispersion.
- Additional robustness tables (selected highlights):
  - Table 10 (time-varying fixed effects): 3-year: -0.0988* (std. error 0.0542); 5-year: -0.0896** (std. error 0.0358).
  - Table 11 (leads and lags): fiscal consolidation t: 3-year: -0.1074* (std. error 0.0573); 5-year: -0.1130** (std. error 0.0448).
  - Table 12 (pre-2020 sample): 3-year: -0.1389** (std. error 0.0640); 5-year: -0.1446*** (std. error 0.0454).

### Mechanisms and interpretation
- Three principal channels through which fiscal prudence can affect inflation:
  - (i) Demand channel: curb aggregate demand contemporaneously, easing price pressures.
  - (ii) Monetary channel: reduce pressures to finance deficits via inflationary monetary expansions.
  - (iii) Expectations channel: affect inflation expectations and thereby current inflation.
- Empirical emphasis: the expectations channel plays an important role—anchoring medium-term expectations via consolidation announcements helps contain inflationary pressures and influences medium-term inflation dynamics.
- Complementarities: inflation anchoring by fiscal announcements is enhanced where fiscal and monetary policy frameworks are credible and when monetary policy is tight, especially during high-inflation episodes.

### Policy conclusions and suggested future research
- Main policy-relevant conclusions:
  - Fiscal consolidation announcements are associated with lower medium-term inflation expectations, with stronger effects for emerging economies.
  - Consolidation announcements reduce dispersion of inflation expectations, consistent with enhanced anchoring.
  - Through expectations, fiscal consolidation can be a tool to contain inflationary pressures and shape medium-term inflation dynamics.
  - Strong complementarities exist between fiscal and monetary policies, particularly during high-inflation episodes.
  - A turn to a more prudent fiscal stance could help anchor expectations and improve fiscal sustainability.
- Suggested avenue for future research:
  - Investigate symmetry: whether fiscal expansions push up inflation expectations and whether the link between fiscal policy changes and inflation expectations is symmetric for fiscal tightening and fiscal stimulus.

*Source: IMF Working Paper "Can Fiscal Consolidation Announcements Help Anchor Inflation Expectations?" — Introduction, Section III, PVAR analysis, Annex 1: Figures and Tables*

### Introduction ...........................................................................................................

### Introduction

### Overview and motivation
- The resurgence of inflation following the COVID-19 pandemic and other subsequent global shocks was accompanied by signs of a de-anchoring of near-term inflation expectations (IMF, 2023a).
- This paper explores the role of fiscal consolidation announcements in shaping inflation expectations using quarterly data and a novel database of fiscal consolidation announcements.
- Fiscal adjustment announcements can reduce inflation through:
  - effects of austerity on aggregate demand (captured by the output gap term in the standard Phillips curve);
  - effects on inflation expectations via the “fiscal theory of the price level” (Leeper and Leith, 2017): higher nominal debt not backed by future surpluses could lead agents to anticipate higher inflation; conversely, higher expected fiscal surpluses would lead to a decline in inflation expectations.

### Data and identification strategy
- Use of quarterly (rather than annual) frequency allows control for anticipation effects of fiscal policy implementation, important for variables that respond quickly to news, such as inflation expectations.
- Key identification assumption: fiscal consolidation measures are unlikely to be adjusted to changes in inflation expectations within the same quarter, given that design, announcement, and implementation typically takes several months and often involves extended executive-legislative negotiations (Lienert, 2010).
- Omitted variable bias is attenuated by inclusion of a number of control variables as well as country and time fixed-effects.

### Baseline findings on expectations
- Fiscal consolidation announcements reduce inflation expectations over the medium-term, but not in the short-term.
- The medium-term effect appears driven by the dynamics of realized inflation after announcements and their effect on near-term inflation expectations in emerging markets.
- Consolidation announcements reduce “disagreement” about expected future inflation at longer horizons.

### Heterogeneity by composition and channels
- Composition matters:
  - Expenditure-based consolidation announcements reduce inflation expectations at shorter horizons.
  - Revenue-based announcements reduce medium- to long-term inflation expectations (possibly because of the initial impact of taxes on prices).
- Fiscal and monetary frameworks and initial conditions matter:
  - Response of inflation expectations to consolidation announcements is more muted in countries with higher government debt.
  - Consolidation announcements have larger impacts on expectations in countries that display more fiscal prudence than initially forecasted.
  - Among emerging markets, countries with inflation targeting regimes see a larger reduction in medium-term inflation expectations compared to others.
  - Consolidation announcements have a larger impact on inflation expectations in countries where monetary policy is tight.
  - Consolidation announcements have a larger and more immediate impact on expectations in times of high inflation.

### Inflation expectations in transmission to realized inflation (PVAR analysis and counterfactual)
- The paper estimates relationships between fiscal consolidation announcements, inflation expectations, and realized inflation in a panel vector autoregression (PVAR) framework.
- A counterfactual methodology is built following Bernanke and others (1998), Sims and Zha (2006), Kilian and Lewis (2011), and Bachmann and Sims (2012), conditioning impulse responses on the level of inflation at the time of announcements.
- Main quantitative result for high-inflation economies:
  - For economies with high realized inflation (at or above the 75th percentile of the distribution), a fiscal consolidation announcement reduces headline inflation by 100 basis points in a one-year window.
  - In the counterfactual scenario where the response of inflation expectations is zeroed out, the reduction in inflation is much more gradual.

### Novelty and contributions
- First paper to use well-identified fiscal policy announcements at a quarterly frequency to quantify the link between fiscal consolidations and inflation expectations for a large set of countries comprising advanced economies and emerging markets.
- Contributes by:
  - investigating how a country’s macroeconomic policy frameworks shape the impact of fiscal actions on expectations;
  - exploring monetary–fiscal policy interactions;
  - highlighting the key role of medium-term expectations in price formation processes.
- Places findings in context of related literature:
  - Prior work often focuses on inflation outcomes rather than expectations (e.g., Catão and Terrones, 2005; Banerjee and others, 2022; Jorgensen and Ravn, 2022; Cevik and Miryugin, 2023).
  - Recent complementary work includes Dabla-Norris, Goncalves, and Nguyen (2023) using narrative tax changes for 10 advanced economies, and household-level or narrative studies of expectations (Coibion, Gorodnichenko and Weber, 2021; Brandão-Marques and others, 2023; Cloyne and others, 2023).

*Source: Introduction section, IMF Working Paper "Can Fiscal Consolidation Announcements Help Anchor Inflation Expectations?"*

### section III presents the empirical strategy and findings on the effects of fiscal announcements on inflation

### wpiea2024060-print-pdf - section III presents the empirical strategy and findings on the effects of fiscal announcements on inflation

### Data sources and key variables
- Inflation expectations and dispersion:
  - Source: Consensus Economics consensus forecasts (survey-based forecasts by professional forecasters).
  - Horizons available: current year, one year ahead, 2-5 years ahead, and 10 years ahead.
  - This paper focuses on three horizons: one year ahead (short-term), three years ahead (medium-term), and five years ahead (medium-term).
  - Dispersion measured as the standard deviation of forecasts; alternative measure used is the gap between highest and lowest forecast normalized by the median forecast.
- Fiscal consolidation announcements:
  - Narrative approach updating databases from Beetsma and others (2021) for advanced economies and David, Guajardo, and Yépez (2022) for emerging markets.
  - Original coverage: advanced economies 1978-2013; emerging markets 2000-2018. Updated to include announcements up to December 2023, adding 75 announcements (34 for advanced economies and 41 for emerging markets).
  - Identification rules: include only announcements representing net consolidation efforts that imply an expected improvement in the fiscal balance; exclude announcements that are fully offset or revenue-neutral shifts.
  - News search: Dow Jones’ Factiva database for keywords such as “fiscal consolidation”; sample limited to availability of English, Spanish, or Portuguese articles; sample covers 32 countries.
- Additional control variables and data sources:
  - Observed year-over-year inflation and real GDP: World Economic Outlook (WEO) database.
  - Nominal effective exchange rate: IMF’s Information System Notice (INS) database.
  - Output gaps: Hamilton filter on real quarterly GDP.
  - Fiscal prudence gauges: distance between observed fiscal balances and IMF projections from WEO vintages (gap between observed year t balance and IMF projection made t-3); global debt database (GDD) from Mbaye, Moreno Badia and Chae (2018).
  - Inflation-targeting regime indicator: IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions (AREAER).

### Empirical specification and baseline regressions
- Baseline regression:
  - Dependent variable: inflation expectations π_{i,t}^{e,h} at horizons h = 1, 3, 5 years ahead.
  - Main regression (equation 1): π_{i,t}^{e,h} = α_i + γ_t + θ (fiscal consolidation)_{i,t} + β X_{i,t} + ε_{i,t}.
  - Controls X_{i,t}: current and lagged realized inflation; current and lagged output gap; current and lagged (y/y) changes in nominal effective exchange rate; a proxy for short-term inflation expectations dynamics (weighted sum of end-of-year and last-quarter-next-year expectations).
  - Fixed effects: country fixed effects α_i and time fixed effects γ_t.
- Main coefficient of interest: θ (direct effect of fiscal consolidations on inflation expectations).
- Dispersion regressions: same specification replacing dependent variable with dispersion (standard deviation) and alternatively highest-lowest gap normalized by median.

### Baseline findings on inflation expectations and dispersion
- Direction and significance:
  - Coefficients linked to fiscal consolidations are negative across horizons, indicating consolidation announcements are associated with declines in inflation expectations.
  - Statistical significance: consolidation coefficient is only statistically significant at conventional levels for the three-year and five-year horizons (medium-term), but muted at the one-year horizon (short-term).
- Heterogeneity by country group:
  - Interaction with an advanced-economy dummy indicates a more muted link between fiscal consolidation announcements and inflation expectations in advanced economies.
  - Possible reasons noted: advanced economies have relatively low inflation expectations and more credible policy frameworks.
- Dispersion results:
  - Fiscal consolidation announcements significantly reduce dispersion of expectations at three- and five-year horizons, but not at the one-year horizon.
  - Normalizing the highest-lowest gap by the median yields consistent results: consolidation announcements significantly reduce dispersion of medium-term forecasts.
- Composition of consolidations:
  - Announcements labeled as revenue-based (tax hikes), expenditure-based (spending cuts), or combined.
  - Both revenue- and expenditure-based consolidations linked with lower inflation expectations, with timing/magnitude differences:
    - Expenditure-based consolidations: negative impact across horizons; statistically significant for 3-year ahead expectations.
    - Tax-based consolidations: lower estimated short-term coefficient (even positive but not significant for dispersion), strong medium-term impact.
    - Dispersion effects: expenditure-based consolidations have no statistically significant effect on dispersion; tax-based and combined consolidations do reduce dispersion.

### Role of fiscal prudence, monetary framework, and initial conditions
- Interaction specification (equation 2): π_{i,t}^{e,h} = α_i + γ_t + θ (fiscal consolidation)_{i,t} + ϖ P_{i,t} + δ (fiscal consolidation * P)_{i,t} + β X_{i,t} + ε_{i,t}.
  - Coefficient δ captures how P_{i,t} amplifies/dampens the impact of consolidations.
- Fiscal prudence and debt:
  - Higher government debt (share of GDP) dampens the decline in inflation expectations following consolidation announcements for all horizons.
  - A larger positive gap between contemporaneous fiscal balances and IMF t-3 forecasts (indicating higher fiscal prudence) amplifies the reduction in inflation expectations after consolidation announcements; statistically significant for 5-year ahead expectations.
- Monetary credibility / inflation targeting:
  - Sample split for emerging markets by inflation-targeting (IT) regime:
    - For IT countries, consolidation announcements lower medium-term inflation expectations (significant at 3-year horizon).
    - For non-IT countries, the negative effect is not statistically significant.
  - Pre-2020 sample: larger absolute point estimates for IT countries and most horizons for non-IT countries; estimated effects stronger and more often statistically significant in the pre-2020 sample.
- Monetary policy stance:
  - Dummy for tight monetary policy defined as ex-post real interest rate above an 8-quarter moving average.
  - Tight monetary policy is associated with lower medium-term inflation expectations and boosts the inflation-anchoring role of fiscal consolidation announcements in the medium-term.
- Initial inflation conditions:
  - Dummy for high inflation: year-on-year inflation at announcement above the 75th percentile of country-specific inflation distribution 1990-2023.
  - Consolidation announcements during high inflation periods have an additional statistically significant impact on lowering inflation expectations at all horizons; magnitude increases with horizon.
  - Pre-2020 sample: interaction term remains positive and significant at all horizons.

### Robustness exercises
- Country-period fixed effects:
  - Four subperiods: i) 1990-2000, ii) 2001-2008, iii) 2009-2019, iv) post-COVID period.
  - Inclusion of country-period fixed effects does not qualitatively change patterns: consolidation announcements associated with lower medium-term inflation expectations (5-year ahead) and reduced dispersion, though magnitudes are smaller and some significance is lost.
- Leads and lags of announcements:
  - Including t-1 and t+1 announcements to control for anticipation and persistence.
  - Consolidation coefficient remains significantly negative; effects materialize mostly on impact with no evidence of protracted impacts.
- Excluding pandemic/post-pandemic quarters:
  - Results show the connection between consolidation announcements and inflation expectations was stronger in the pre-2020 period; larger coefficients for consolidation dummy in pre-2020 sample than full sample for expectations and dispersion.

### PVAR framework and counterfactual analysis (transmission to realized inflation)
- PVAR variables and frequency:
  - Three variables at quarterly frequency: fiscal announcement dummy D_{i,t} (equals 1 in quarter of announcement), 3-year ahead inflation expectations E_{i,t} (y/y percent), and realized inflation π_{i,t} (y/y percent).
- Identification and ordering:
  - Cholesky decomposition with fiscal announcements ordered first, inflation expectations second, realized inflation third.
  - Timing restriction: inflation variables react contemporaneously to fiscal shocks; fiscal policy does not react contemporaneously to other shocks.
  - Additional identifying assumption: a_{i,2,3} = 0 (no contemporaneous effect of realized inflation on expectations), so expectation shock ε_{i,t}^2 and realized inflation shock ε_{i,t}^3 are identified.
- Counterfactual methodology:
  - Construct hypothetical impulse responses of realized inflation to a fiscal announcement by holding the response of expected inflation fixed at zero at all horizons: IRF_2(1,h) = 0.
  - Compute required sequence of expectation shocks ε_{i,h}^2 to enforce IRF_2(1,h) = 0 (equations 7 and recursive formula).
  - Modified impulse responses IRF̃_k(1,h) computed to quantify the contribution of inflation expectations to realized inflation response.
  - Difference between actual IRF_k(1,h) and IRF̃_k(1,h) measures importance of expectations as a transmission channel.
- Nonlinearities and interaction with inflation level:
  - Allow coefficients A_{i,j} to vary with inflation level: A_{i,j} = β_{i,j} + ς_{i,j} * πlevel_{i,t}.
  - Impulse responses estimated across full empirical distribution of realized inflation levels.
- Estimation details:
  - Each equation estimated by OLS with 4 lags (quarterly data).
  - Runkle (1987) bootstrap employed for statistical inference in the panel with interaction terms.

*Source: IMF WORKING PAPERS — Can Fiscal Consolidation Announcements Help Anchor Inflation Expectations? — Section III*

### 1. Estimate the PVAR(p) in equation (4) and generate 1,000 bootstrap replications 퐴

### 1. Estimate the PVAR(p) in equation (4) and generate 1,000 bootstrap replications 퐴̂ᵢ using equation (9) and the values of the selected values for the levels of inflation spreads in quarter t

### Bootstrapping and simulation procedure
- Estimate the PVAR(p) in equation (4).
- Generate 1,000 bootstrap replications 퐴̂ᵢ using equation (9) and the values of the selected values for the levels of inflation spreads in quarter t.
- Simulation steps:
  - 1. Simulate the first period for all variables in the system using the bootstrap draws.
  - 2. After the first period is simulated for all variables, interact the variables with the interaction terms in (9) and then repeat step 1 for t = 2, ..., T and i = 1, . . . , N, where T is the sample length and N is the number of countries.
  - 3. Use the artificial sample, together with the interaction variables, to re-estimate the coefficients of the system. Compute IRFs 1,000 times for each generated variable k to the first structural shock (that is the fiscal announcement) at horizon h = 1, . . . , H.
  - 4. Construct 1,000 adjusted impulse responses holding the response of inflation expectations fixed at zero at each horizon h.
  - 5. Use the bootstrap simulations to calculate the empirical distribution for the difference between the baseline and counterfactual scenarios and construct one standard deviation confidence bands around the simulated estimates.
- Bootstrapping detail:
  - Bootstrapping for the panel was done by generating initial conditions separately for each country as in Runkle (1987), but sampling from the entire panel vector of residuals to account for possible cross-country correlations.

### PVAR results and dynamic responses
- Setup for impulse response estimation:
  - Conditioned impulse responses to an austerity announcement are plotted in an 8 quarter window, with the coefficients of the PVAR varying depending on the inflation level.
  - The fiscal announcement dummy is set to value 1 on impact and zero thereafter (hence a cumulative value of 1 over the 8-quarter window).
  - Shaded gray regions are one standard error confidence bands from Runkle’s (1987) bootstrapping method.
- Key empirical findings:
  - For periods of high inflation (defined to be at or above the 75th percentile of the empirical distribution):
    - Inflation expectations significantly decline in the aftermath of the announcements by around 10 basis points one year after the announcement.
    - Austerity announcements tend to be followed by an increase of realized inflation in the short term, but inflation quickly falls four quarters after a year of the announcement.
  - Dashed hypothetical impulse responses (holding the response of expected inflation fixed at zero):
    - The direct response of realized inflation without the endogenous response of expectations is higher at most horizons.
    - The differences in the response of realized inflation following the fiscal announcements are economically important and statistically significant, with inflation in the baseline scenario being almost half than the counterfactual response a year after the announcement.
  - For low inflation environments (realized inflation below the 25th percentile):
    - The role of fiscal policy and inflation expectations appears more limited.
    - Differences between impulse responses are economically small and not statistically significant.

### Quantification and inference
- Bootstrap usage:
  - 1,000 bootstrap replications are used to generate IRFs and adjusted impulse responses.
  - One standard deviation confidence bands are constructed from the bootstrap simulations.
- Comparative experiment:
  - Baseline scenario vs. counterfactual scenario that holds expected inflation fixed at zero to isolate the role of expectations in transmitting fiscal announcements.

### Interpretation and mechanisms
- Channels through which fiscal prudence can affect inflation:
  - (i) Demand channel: curb aggregate demand contemporaneously, easing price pressures.
  - (ii) Monetary channel: reduce pressures to finance fiscal deficits through inflationary monetary expansions.
  - (iii) Expectations channel: affect inflation expectations and, as a result, current inflation.
- Role of expectations:
  - Results emphasize the importance of the expectations channel: anchoring expectations via fiscal consolidation announcements can contain inflationary pressures and influence medium-term inflation dynamics.
- Importance of initial conditions and frameworks:
  - The inflation anchoring role of fiscal consolidation announcements is enhanced by the credibility of a country’s fiscal and monetary frameworks, and when fiscal and monetary policy work in tandem.

### Conclusions and implications for policy and future research
- Main conclusions:
  - Fiscal consolidation announcements are associated with lower medium-term inflation expectations, with stronger effects found for emerging (economies).
  - Fiscal consolidation announcements help anchor expectations, as reflected by a lower dispersion of inflation expectations.
  - Fiscal consolidation, by anchoring expectations, can be a powerful tool to contain inflationary pressures and drive inflation dynamics in the medium-term.
  - Strong complementarities exist between fiscal and monetary policies, particularly during high-inflation episodes.
- Policy implication emphasized:
  - A turn to a more prudent fiscal stance could help anchor expectations going forward in addition to putting fiscal trajectories on a more sustainable path.
- Suggested avenue for future research:
  - Explore whether fiscal expansions push up inflation expectations i.e., whether the link between fiscal policy changes and inflation expectations is symmetric (the same in the cases of fiscal tightening and fiscal stimulus).

*IMF WORKING PAPERS Can Fiscal Consolidation Announcements Help Anchor Inflation Expectations? — INTERNATIONAL MONETARY FUND*

### Annex 1: Figures and Tables

### Annex 1: Figures and Tables

### Figures: timing, impulse responses, and counterfactuals
- Figure 1. The Timing of Fiscal Consolidation Announcements.
- Figure 2. PVAR Impulse Responses Fiscal Consolidation Announcements.
- Figure 3. Difference in Realized Inflation in Baseline and Counterfactual Scenarios.
  - Quarters shown: 1, 2, 3, 4, 5, 6, 7, 8.
  - Series shown: Difference in High Inflation Environments (75th percentile); Difference in Low Inflation Environments (25th percentile).

### Sample composition
- Countries included in the sample (listed): United Kingdom, Austria, Belgium, Denmark, France, Germany, Italy, Netherlands, Sweden, Finland, Ireland, Portugal, Spain, Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, Guatemala, Mexico, Paraguay, Peru, Uruguay, India, Indonesia, South Africa, Slovak Republic, Hungary, Poland.

### Key regression findings: Inflation expectations at different horizons (Table 2)
- Dependent variables: 1-year, 3-year, and 5-year inflation expectations (columns (1)–(6)).
- Consolid. announcement dummy coefficients:
  - Column (1) 1-year: -0.0294 (std. error 0.0350).
  - Column (2) 3-year: -0.1160** (std. error 0.0572).
  - Column (3) 5-year: -0.1078** (std. error 0.0449).
  - Column (4) 1-year: -0.0188 (std. error 0.0534).
  - Column (5) 3-year: -0.1540* (std. error 0.0827).
  - Column (6) 5-year: -0.1565** (std. error 0.0707).
- Consolidation announcement*AE dummy:
  - Column (4) 1-year: -0.0312 (std. error 0.0600).
  - Column (5) 3-year: 0.1269 (std. error 0.1038).
  - Column (6) 5-year: 0.1627* (std. error 0.0885).
- Inflation (contemporaneous) coefficients:
  - Column (2) 3-year: -0.2663*** (std. error 0.0693).
  - Column (3) 5-year: -0.1482*** (std. error 0.0472).
  - Columns (5),(6) similar: -0.2656*** (std. error 0.0696); -0.1472*** (std. error 0.0474).
- Exchange rate (contemporaneous) significant at *:
  - Column (1): 0.0121* (std. error 0.0065).
  - Column (2): 0.0185* (std. error 0.0097).
- Short-term inflation expectations strongly significant:
  - Column (1): 0.6680*** (std. error 0.1448).
  - Column (2): 0.8423*** (std. error 0.0907).
  - Column (3): 0.4790*** (std. error 0.0609).
- Observations and fit:
  - Observations: 3,084 (1-year), 1,411 (3-year), 1,410 (5-year).
  - R-squared: 0.938 (1-year), 0.852 (3-year), 0.748 (5-year).
- Note: Driscoll-Kraay standard errors in parentheses. All specifications include country and time fixed effects. Significance: *** p<0.01, ** p<0.05, * p<0.1.

### Dispersion of inflation expectations (Table 3)
- Consolid. announcement dummy coefficients (dispersion):
  - Column (1) 1-year Dispersion: -0.0109 (std. error 0.0246).
  - Column (2) 3-year Dispersion: -0.0877** (std. error 0.0337).
  - Column (3) 5-year Dispersion: -0.0532** (std. error 0.0253).
  - Column (4) 1-year Dispersion: -0.0176 (std. error 0.0318).
  - Column (5) 3-year Dispersion: -0.1265*** (std. error 0.0450).
  - Column (6) 5-year Dispersion: -0.0634** (std. error 0.0316).
- Consolidation announcement*AE dummy:
  - Column (2): 0.1364** (std. error 0.0560).
- Inflation contemporaneous:
  - Column (1): -0.0992*** (std. error 0.0299).
  - Column (2): -0.1146*** (std. error 0.0343).
- Exchange rate (contemporaneous) significance:
  - Column (2): 0.0122*** (std. error 0.0039).
  - Column (3): 0.0137* (std. error 0.0070).
- Short-term inflation expectations positive and significant for dispersion:
  - Column (1): 0.2023*** (std. error 0.0238).
  - Column (3): 0.1468** (std. error 0.0704).
- Observations and fit:
  - Observations: 1,786 (1-year Dispersion), 1,116 (3-year), 1,115 (5-year).
  - R-squared: 0.844 (1-year Dispersion), 0.622 (3-year), 0.396 (5-year).

### High-Low gap of inflation expectations (Table 4)
- Consolid. announcement dummy coefficients (High-Low gap):
  - Column (1) 1-year High-Low gap: 0.0558** (std. error 0.0270).
  - Column (2) 3-year High-Low gap: -0.0443** (std. error 0.0219).
  - Column (3) 5-year High-Low gap: -0.0129 (std. error 0.0215).
  - Column (4) 1-year High-Low gap: 0.0689* (std. error 0.0384).
  - Column (5) 3-year High-Low gap: -0.0498** (std. error 0.0194).
  - Column (6) 5-year High-Low gap: -0.0047 (std. error 0.0235).
- Observations and fit:
  - Observations: 1,792 (1-year), 1,116 (3-year), 1,115 (5-year).
  - R-squared: 0.169, 0.126, 0.103 respectively.
- Note: The high-low gap is the difference between the highest and lowest inflation forecast at each horizon, normalized by the median forecast.

### Composition of consolidations: tax vs expenditure effects (Table 5)
- Tax based consolidations coefficients (inflation expectations):
  - 3-year inflation expect. (column (2)): -0.1627*** (std. error 0.0532).
  - 5-year inflation expect. (column (3)): -0.1739*** (std. error 0.0367).
- Tax based consolidations coefficients (dispersion):
  - 3-year Dispersion (column (5)): -0.1529*** (std. error 0.0426).
  - 5-year Dispersion (column (6)): -0.0811*** (std. error 0.0268).
- Expenditure based consolidations coefficients:
  - 3-year inflation expect. (column (2)): -0.1794* (std. error 0.0937).
  - No significant negative effects on dispersion reported at conventional levels.
- Combined tax and expenditure consolidations: coefficients imprecise (standard errors large; e.g., 0.1347 with std. error 0.1542).
- Observations and fit:
  - Observations: 3,084 (1-year), 1,411 (3-year), 1,410 (5-year) for expectations; 1,786, 1,116, 1,115 for dispersions.
  - R-squared: 0.938, 0.853, 0.748 for expectations; 0.844, 0.624, 0.396 for dispersions.

### Role of government debt and fiscal prudence (Table 6)
- Consolid. announcement dummy:
  - Column (2) 3-year: -0.3208** (std. error 0.1245).
  - Column (3) 5-year: -0.3034*** (std. error 0.1086).
- Government debt (percent of GDP) contemporaneous:
  - Column (1): -0.0051** (std. error 0.0021).
  - Column (2): -0.0068* (std. error 0.0036).
- Government debt x Consolid. ann.:
  - Column (2): 0.0037** (std. error 0.0016).
  - Column (3): 0.0033** (std. error 0.0013).
- Development of fiscal balance relative to WEO forecast, t-3 and its interaction:
  - Consolid. ann. x Dev.of fiscal balance relative to WEO forecast, t-3:
    - Column (3): -0.0255** (std. error 0.0120).
- Observations and fit:
  - Observations: 3,073 (1-year), 1,411 (3-year), 1,410 (5-year) in baseline; alternative sample with fewer observations also reported.
  - R-squared: up to 0.938 (1-year), 0.854 (3-year), 0.749 (5-year).

### Role of inflation targeting (Tables 7 and 8)
- Table 7: Full sample split by Inflation Targeting Countries vs Non-Inflation Targeting Countries.
  - Inflation Targeting Countries (columns (1)–(3)):
    - Consolid. announcement dummy:
      - Column (2) 3-year: -0.0483** (std. error 0.0234).
      - Column (3) 5-year: -0.0453 (std. error 0.0362).
  - Non-Inflation Targeting Countries (columns (4)–(6)):
    - Consolid. announcement dummy:
      - Column (4) 1-year: 0.0593 (std. error 0.1906).
      - Column (5) 3-year: -0.4900 (std. error 0.6160).
      - Column (6) 5-year: -0.4343 (std. error 0.4269).
  - Observations and fit:
    - Inflation Targeting Countries: Observations 968, 545, 545; R-squared 0.932, 0.819, 0.657.
    - Non-Inflation Targeting Countries: Observations 609, 200, 199; R-squared 0.951, 0.940, 0.891.
- Table 8: Pre-2020 sample split by inflation-targeting status shows similar structure; example coefficients:
  - Pre-2020 Inflation Targeting Countries:
    - Consolid. announcement dummy Column (2) 3-year: -0.0593** (std. error 0.0238).
    - Column (3) 5-year: -0.0792*** (std. error 0.0278).
  - Pre-2020 Non-Inflation Targeting Countries have imprecise large standard errors in some columns.

### Monetary policy stance and initial inflation interactions (Table 9)
- Consolid. announcement dummy across specifications:
  - Full sample columns (1)–(3): not statistically significant for 1-year, 3-year, 5-year except:
    - Column (5) 3-year with interaction: -0.1141** (std. error 0.0572) in one specification.
  - Consolid. ann. x Tight monetary policy dummy:
    - Column (5): -0.2248** (std. error 0.1065).
  - Consolid. ann. x High Inflation dummy:
    - Columns show strong negative interactions:
      - Column (2): -0.8400** (std. error 0.4030).
      - Column (3): -0.9064*** (std. error 0.2699).
      - Column (6): -0.9421*** (std. error 0.2322).
      - Other pre-2020 and sample variations also show large negative coefficients (e.g., -0.7957**, -0.6653***, -0.7695***).
- Pre-2020 subsample reported (columns (7)–(9)) with larger coefficients for output-gap lags in some specifications (e.g., output gap (t-2) coefficients 2.7843***, 5.2303**, 5.9223**).
- Observations and fit:
  - Full sample Observations: 3,084; 1,411; 1,410. R-squared: 0.938, 0.853, 0.749.
  - Pre-2020 sample Observations: 2,830; 1,157; 1,156. R-squared: 0.918, 0.731, 0.563.

### Robustness checks and additional specifications (Tables 10–12)
- Table 10: Controlling for time-varying fixed effects:
  - Consolid. announcement dummy:
    - Column (2) 3-year: -0.0988* (std. error 0.0542).
    - Column (3) 5-year: -0.0896** (std. error 0.0358).
    - Column (4) 1-year Dispersion: -0.0380* (std. error 0.0202).
    - Column (5) 3-year Dispersion: -0.0991*** (std. error 0.0302).
  - Observations and fit: Observations 2,892; 1,313; 1,312 (expectations) and 1,678; 1,036; 1,035 (dispersions). R-squared up to 0.887 and 0.825.
- Table 11: Controlling for leads and lags of fiscal consolidation (fiscal consolidation t, t+1, t-1):
  - fiscal consolidation t:
    - Column (2) 3-year: -0.1074* (std. error 0.0573).
    - Column (3) 5-year: -0.1130** (std. error 0.0448).
    - Column (5) 3-year Dispersion: -0.0817*** (std. error 0.0303).
    - Column (6) 5-year Dispersion: -0.0515** (std. error 0.0239).
  - fiscal consolidation t+1 and t-1 coefficients reported and generally less precisely estimated.
  - Observations and fit: expectations sample Observations 3,064; 1,404; 1,403. R-squared 0.937, 0.852, 0.748.
- Table 12: Pre-2020 sample regressions:
  - Consolid. announcement dummy:
    - Column (2) 3-year: -0.1389** (std. error 0.0640).
    - Column (3) 5-year: -0.1446*** (std. error 0.0454).
    - Dispersion effects: Column (5) 3-year Dispersion: -0.1011*** (std. error 0.0332); Column (6) 5-year Dispersion: -0.0509* (std. error 0.0299).
  - Observations and fit: expectations Observations 2,830; 1,157; 1,156. R-squared 0.917, 0.731, 0.562.

*Source: Annex 1: Figures and Tables, Can Fiscal Consolidation Announcements Help Anchor Inflation Expectations? Working Paper No. WP/2024/060*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024060-print-pdf.pdf_