## wpiea2022216-print-pdf

## Source details

**Canonical URL:** [wpiea2022216-print-pdf](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022216-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022216-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022216-print-pdf.pdf.json)

---

### Context and motivation
- Post-pandemic fiscal stimuli and increases in public debt have made mobilizing revenues a policy priority, with tax reforms an essential part of the process.
- The paper studies political pressure from upcoming elections as a potential determinant of the timing of tax reforms.
- Monthly data and announcement dates are used to avoid fiscal foresight issues and to capture short-lived electoral cycles.

### Data and identification strategy
- Sample and data sources:
  - Monthly data from 22 advanced economies (AEs) and emerging markets (EMs) over 1990-2018.
  - Tax Policy Reform Database (TPRD) provides direction of tax changes and precise month of announcement; announcement dates used.
  - Election months manually collected for the 22-country sample.
- Key measurement choices:
  - Main definition: dummy = 1 whenever at least one major tax increase announcement occurs in a month (major = rate change > 1 pp or base change affecting a large group/large potential mobilization).
  - Alternative definitions used in robustness checks: net change dummy, include “minor” increases, count of announcements.
- Sample counts and coverage:
  - total announcements in database: 3462;
  - announcements during 1990-2018: 2260;
  - of those 2260 events, 953 are related to tax increases, with 772 identified as “major”;
  - out of 772 major announcements, 707 events have information regarding the month of announcement (about 98 percent of which were eventually implemented).
- Other variables:
  - Institutional quality proxies: law and order (ICRG, range 0 to 6), bureaucratic quality (ICRG, range 0 to 4), civil liberties (Freedom House, inverse used).
  - Political system: DPI 2020 dummy = 1 if presidential, 0 otherwise.
  - IMF lending arrangements: monthly dummy = 1 during months with an IMF lending arrangement.
  - Fiscal rules: annual dummy = 1 for years with some fiscal rule (Davoodi et al. (2022)).
  - Economic controls: industrial production index (monthly) and inflation (CPI yoy, monthly) from IFS.
- Sample size:
  - number of country-month observations: 7,392.
  - total number of months with reform announcements in sample: 381.
  - overall probability of reforms in sample: 5.2 percent.
  - number of elections in sample: 170.

### Stylized facts (aggregate patterns)
- Aggregate probabilities:
  - overall probability of tax reform announcements in sample: 5.2 percent;
  - average monthly probability during the 6-month period running up to elections: about 2.5 percent;
  - probability in the months of elections: 2.4 percent;
  - average monthly probability during the 6-month aftermath of elections: 8.3 percent.
- Monthly pattern:
  - likelihood of reform announcements in the 6 months before and during elections stays below the sample average (5.2 percent);
  - probability becomes much larger in post-election months, particularly during the third month after elections.
- Interpretation: governments are less likely to announce tax reforms before and during elections and more likely to announce them following elections.

### Empirical strategy (main specification)
- Panel linear probability model with fixed effects:
  - Dependent variable: Prob(Reform_c,t) = 1 if a tax reform announcement occurs in country c, month t; 0 otherwise.
  - Election dummies: Election_c,t+p and Election_c,t−r for p=1,...,6 and r=1,...,6 to capture 6-month pre- and post-election windows and the election month.
  - Fixed effects: country-year (θ_c,y) and month (θ_t).
  - Standard errors clustered at the country-year level.
- Identification advantages emphasized: monthly announcement dates capture short-lived cycles and allow absorption of country-year and month shocks.

### Main empirical findings
- Core result: probability of tax reform announcements declines during months running up to elections and in election months; increases in some months after elections.
- Selected coefficient estimates (Table 1, column (3); coefficients with standard errors in parentheses):
  - Election t+6: -0.047*** (0.012)
  - Election t+5: -0.015 (0.019)
  - Election t+4: -0.036** (0.015)
  - Election t+3: -0.038*** (0.014)
  - Election t+2: -0.006 (0.021)
  - Election t+1: -0.047*** (0.015)
  - Election t: -0.051*** (0.016) [when included]
  - Lagged post-election notable effect: Election t−3: 0.087*** (0.029) in specifications with lags.
- Aggregate interpretation:
  - Cumulative decline during the 6-month window before elections: 16.9 pp (corresponding to a 2.8 pp lower probability per month over that period on average).
  - Election month effect: probability declines by 4.7 pp.
  - These effects are large relative to the sample average monthly probability of tax reform announcements: 5.2 percent.
- Model fit (Table 1, column (3)):
  - Country-year and month F.E.: Yes
  - R-squared: 0.143
  - Observations: 7,392
- Results robust when extending the window to 12 months before and after elections; avoidance before elections persists while post-election window remains short-lived.

### Robustness checks (selected)
- Dependent variable and model variants (Table 2) produced consistent negative pre-election coefficients for key months. Selected 퐸푙푒푐푡푖표푛푡+6 estimates across specifications:
  - (1) net change dummy: -0.028*** (0.010)
  - (2) number of tax increase announcements: -0.049*** (0.017)
  - (3) minor increases included: -0.056*** (0.013)
  - (4) Probit with analytical bias correction: -0.147*** (0.021)
  - (5) controls for lagged tax increases: -0.056*** (0.012)
  - (6) controls for lagged tax decreases: -0.049*** (0.012)
  - (7) controls for regional reform share: -0.047*** (0.012)
  - (8) adds lagged industrial production: -0.052*** (0.013)
  - (9) adds lagged inflation: -0.047*** (0.013)
- Lagged positive post-election effects at Election t−3 persist across specifications (examples: 0.059** (0.024) in column 1; 0.094*** (0.030) in column 3; 0.236*** (0.074) in column 4).
- Additional robustness exercises performed (results available upon request): leave-one-country-out, exclusion of countries by reform counts, subperiod analyses (pre/post-2008), excluding snap elections, different clustering, different fixed-effect combinations, weighted regressions, including all control variables.

### Predictive power (ROC/AUC)
- Probit model performance:
  - Probit model with fixed effects but without election dummies: AUC = 0.616 with a standard error of 0.015.
  - AUC for the full model (including election dummies): 0.696 with a standard error of 0.014.
  - The AUC of the full model is statistically different from 0.5 and from the model excluding election dummies at the 1 percent significance level.
  - Conclusion: inclusion of election dummies increases predictive power significantly; electoral cycles are informative but not perfect predictors.

### Heterogeneity and interaction results
- Institutional quality:
  - Pre-election effect on probability of tax reforms is weaker in countries with relatively high institutional quality (law and order, bureaucratic quality, civil liberties), consistent with institutions shielding policy from electoral pressure.
  - Selected interaction patterns (Table 3; coefficients with standard errors in parentheses; significance markers preserved):
    - Election t+6 examples: -0.103** (0.044); -0.039*** (0.014); -0.060*** (0.022)
    - Election t+4 examples: -0.124** (0.051); -0.020 (0.018); -0.062*** (0.023)
    - Election t+1 examples: -0.150*** (0.053); -0.039** (0.018); -0.065*** (0.024)
    - Election t examples: -0.145*** (0.047); -0.036** (0.017); -0.076*** (0.024)
    - Election t−3: 0.088*** (0.029) (consistent across columns)
    - Example interaction (Election t × X_t): 0.020** (0.010); 0.049** (0.020); 0.020* (0.011); -0.052* (0.028)
  - Table 3 model statistics (columns 1–7):
    - Country-year and month F.E.: Yes
    - R-squared: 0.144, 0.145, 0.144, 0.144, 0.143, 0.147, 0.145
    - Observations: 7,362; 7,362; 7,362; 7,392; 7,392; 7,056; 7,392
- Country groups and political system:
  - Emerging Markets (EMs): electoral cycles are somewhat more pronounced in EMs relative to AEs.
  - Presidential systems: no substantial differential effect found.
- Fiscal rules and IMF programs:
  - Neither fiscal rules nor IMF-supported programs have a direct impact on the likelihood of tax reforms in the sample, nor differential effects on electoral cycles in tax reforms in the sample.
  - Caveat: IMF programs limited in the sample (seven countries with at least one program; around 6 percent of country-month observations, 469 observations).

### Tax measure and tax-type evidence
- Tax measures:
  - Both tax base changes and tax rate changes exhibit electoral cycles: governments avoid announcing increases in rate and/or base before elections and are more likely to announce such reforms in the first few months following elections.
  - Table A.2 counts for measures: Base 201; Rate 139.
- Tax types (counts and patterns):
  - Dependent variables for increases in CIT, PIT, VAT, SSC, EXE, PRO show most tax types exhibit electoral cycles.
  - Exceptions and interpretability limits:
    - Social security contributions (SSC) — electoral cycle effects less pronounced; number of reform announcements relatively low for SSC, EXE, and PRO.
  - Table A.2 counts for types: CIT 91; PIT 103; VAT 70; SSC 28; EXE 31; PRO 12.

### Conclusions and policy implications
- Empirical conclusions:
  - Tax reforms exhibit electoral cycles: governments tend to avoid announcing reforms that increase tax rates and/or broaden the tax base in months before elections and are more likely to announce them in the first few months after elections.
  - Post-election increase in reform likelihood suggests political capital or perceived political opportunity influences timing.
  - Results robust to alternative dependent variables, econometric specifications, and additional controls.
  - Electoral cycles in tax reforms are weaker in countries with stronger institutions (law and order, bureaucratic quality, civil liberties).
  - Fiscal rules do not appear to mitigate electoral cycles in tax reforms in this sample.
  - IMF-supported programs do not appear to affect strength of electoral cycles in the countries in the sample, but sample limitations warrant caution.
- Policy implications and avenues:
  - Challenge: provide appropriate incentives for election-oriented governments to design tax policy optimally given fiscal rules alone do not appear sufficient to shield tax policy from election-driven timing.
  - Electoral cycles may complicate cross-border coordination of tax policy by affecting timing of reforms.
  - Suggested future research: explore role of IMF-supported programs in electoral cycles in tax reforms in developing countries as more tax reform data become available.

*Source: IMF Working Paper — 1. Introduction (wpiea2022216-print-pdf).*

### 1. Introduction ........................................................................................................

### 1. Introduction

### Context and motivation
- Post-pandemic fiscal stimuli and increases in public debt have made mobilizing revenues a policy priority, with tax reforms an essential part of the process.
- Understanding factors affecting the timing of tax reforms is important for tax policy design and for international tax policy coordination.
- This paper focuses on political pressure from upcoming elections as a potential determinant of the timing of tax reforms.

### Data and identification strategy highlights
- Monthly data from 22 advanced economies (AEs) and emerging markets (EMs) over the period of 1990-2018 are used.
- The Tax Policy Reform Database (TPRD) constructed by Amaglobeli et al. (2018) provides direction of tax policy changes and the precise month of announcement of tax reforms; announcement dates are used to avoid fiscal foresight issues.
- Election dates (months) were manually collected for the 22-country sample.
- Monthly frequency allows:
  - Analysis of short-lived electoral cycles that could be cancelled out in annual data.
  - Controlling for economic and political factors at country level for a given year and absorbing common monthly shocks.
- Identification relies on election dates being mostly predetermined relative to the announcement of tax reforms.

### Main empirical findings
- Governments tend to avoid announcing tax reforms during months running up to elections.
- Estimated effects on tax reform announcement probability:
  - Probability of tax reform announcements decreases cumulatively by around 17 percentage points (pp) over the six-month window before elections.
  - Probability becomes an additional 4.7 pp lower in the months of elections.
  - Average monthly probability of tax reform announcements in the sample is 5.2 percent.
- Post-election behavior:
  - Governments become more likely to announce tax reforms in the months following elections, but only within a short three-month period.
  - This indicates a short-lived window of opportunity for newly elected governments to leverage political capital to reform the tax system.
- Breadth of the effect:
  - Electoral cycles are broad-based regarding changes in tax base and rate, and across various types of tax.

### Heterogeneity and robustness of results
- Electoral cycles in tax reforms are stronger in EMs relative to AEs.
- Electoral cycles are weaker in countries with stronger institutions, consistent with institutions shielding policy from electoral pressure.
- Neither fiscal rules nor IMF lending arrangements appear to have much impact on the strength of electoral cycles in tax reforms; however, the authors caution that the latter finding should be treated with caution given the limited presence of IMF programs in the sample.
- Use of monthly data yields results that differ from some previous studies using annual data (e.g., Fuest et al. 2021), which found increases in taxes in the post-election year but no pronounced pre-election effect.

### Contribution to the literature
- Adds to literature on opportunistic political business cycles (PBCs) by documenting electoral cycles in tax reforms using high-frequency (monthly) data.
- Extends literature on drivers of tax reforms by demonstrating the role of political economy determinants and interaction with institutional strength.
- One of the first cross-country studies of electoral cycles in policy at monthly frequency, complementing studies like Aidt et al. (2020) that examine other policy areas at monthly frequency.

*Source: IMF Working Paper — 1. Introduction (wpiea2022216-print-pdf).*

### 2. Data and Stylized Facts

### 2. Data and Stylized Facts

### 2.1. Tax policy reforms
- Data source: the most recent version of TPRD (Amaglobeli et al. (2018)), compiled using more than 950 OECD Economic Surveys and 53,000 tax related news from IBFD archives.
- TPRD reports:
  - direction of tax changes (increase or decrease);
  - whether change relates to tax base or rate;
  - tax types: corporate income tax (CIT), personal income tax (PIT), value-added and sale taxes (VAT), social security contributions (SSC), excise (EXE), and property tax (PRO);
  - dates for announcements and implementation (including month) and significance (“major” or “minor”).
- Advantages of monthly announcement-date data:
  - captures short-lived electoral cycles that annual data may hide;
  - allows absorption of country-level economic and political effects via country-year fixed effects;
  - controls for global monthly shocks via month fixed effects.
- Sample coverage in TPRD for this study:
  - countries: Australia, Austria, Brazil, Canada, China, Czechia, Denmark, France, Greece, Germany, India, Ireland, Italy, Japan, Korea, Luxembourg, Mexico, Poland, Portugal, Spain, Turkey, United Kingdom, United States;
  - analysis period: 1990-2018.
- Definition of "tax reform" in main analysis:
  - focus on tax increase announcements identified as “major”;
  - TPRD classifies rate changes as major if change > 1 pp; base changes are major if they affect a large group or have large potential to mobilize resources;
  - dummy variable = 1 whenever at least one major tax increase announcement occurs in a month.
- Alternative definitions used in robustness checks:
  - net change: dummy = 1 if number of increases in a month > number of decreases;
  - include “minor” increases (instead of only “major”);
  - use count of tax reform announcements in a month rather than a dummy.
- Counts and coverage:
  - total announcements in database: 3462;
  - announcements during 1990-2018: 2260;
  - of those 2260 events, 953 are related to tax increases, with 772 identified as “major”;
  - out of 772 major announcements, 707 events have information regarding the month of announcement (about 98 percent of which were eventually implemented).

### 2.2. Elections
- Election-month data:
  - manually collected months of elections for 22 countries over 1990-2018 from online sources;
  - China excluded due to peculiarities of its electoral system.
- Selection rule for election type:
  - presidential systems (e.g., US): use presidential elections;
  - parliamentary systems (e.g., Germany): use parliamentary elections.
- Cross-checks: where available, data cross-checked with Comparative Political Data Set (CPDS).
- Table A.1 in Appendix provides dates over sample period (not reproduced here).

### 2.3. Other variables
- Institutional quality proxies:
  - law and order index (ICRG): range 0 to 6;
  - bureaucratic quality index (ICRG): range 0 to 4;
  - civil liberties index (Freedom House): original range 0 to 7; authors use inverse (7 minus index) so higher values indicate greater institutional quality;
  - law and order and bureaucratic quality available at monthly frequency; civil liberties at annual frequency.
- Political system:
  - Database for Political Institutions (DPI) 2020 used;
  - dummy = 1 if political system is presidential, 0 otherwise (annual frequency).
- IMF programs and fiscal rules:
  - monthly data on IMF lending arrangements manually collected from IMF website; dummy = 1 during months with an IMF lending arrangement;
  - fiscal rules (Davoodi et al. (2022)) used at annual frequency; dummy = 1 for years with some fiscal rule (expenditure, revenue, budget, or debt).
- Economic controls:
  - industrial production index (monthly) from IMF’s International Financial Statistics (IFS) as proxy for economic activity;
  - inflation (consumer prices, year-on-year) from IFS (monthly).

### 2.4. Sample
- Sample: monthly data from 22 AEs and EMs over 1990-2018.
- number of country-month observations: 7392.
- total number of months with reform announcements in sample: 381.
- overall probability of reforms in sample: 5.2 percent.
- number of elections in sample: 170.
- cross-country variation: total months with reform announcements varies across countries; robustness checks address influence of countries with low or high event counts.

### 2.5. Stylized Facts
- Aggregate probabilities (Figure 2 summary):
  - overall probability of tax reform announcements in sample: 5.2 percent;
  - average monthly probability during the 6-month period running up to elections: about 2.5 percent;
  - probability in the months of elections: 2.4 percent;
  - average monthly probability during the 6-month aftermath of elections: 8.3 percent.
- Monthly pattern around elections:
  - likelihood of reform announcements during the 6 months running up to elections and in election months stays below the sample average (5.2 percent);
  - probability becomes much larger in post-election months, particularly during the third month after elections.
- Interpretation:
  - tax reforms exhibit electoral cycles: governments are less likely to announce tax reforms before and during elections, and more likely to announce them following elections.

### 3. Empirical Strategy (summary of main specifications)
- Main specification: panel linear probability model with fixed effects:
  - dependent variable: Prob(Reform_c,t) = probability that a tax reform announcement occurs in country c, month t (equals 1 in months with announcement, 0 otherwise).
  - election dummies: Election_c,t and forward and lagged values Election_c,t+p and Election_c,t−r for p=1,...,6 and r=1,...,6 to capture 6-month pre- and post-election windows and the election month effect.
  - coefficients α_p (p=1,...,6) capture change in likelihood during 6-month period before elections; α_0 is effect of election month; α_−r (r=1,...,6) capture change in 6-month aftermath.
  - fixed effects: country-year fixed effects (θ_c,y) and month fixed effects (θ_t).
  - standard errors clustered at the country-year level.
- Identification advantages of monthly data:
  - captures short-lived electoral cycles without cancellation from annual aggregation;
  - allows absorption of country-year and month effects to mitigate omitted variable bias.
- Expectations:
  - α_p and α_0 expected to be negative if popularity-concerned governments avoid tax increases before and during elections;
  - α_−r expected to be nonnegative or positive if reduced pre-election announcements reflect re-election concerns.
- Heterogeneity analysis:
  - second-step specification includes interaction terms between pre-election variables and country-specific factors X_c,t:
    - interaction coefficients β_p (p=1,...,6) and β_0 capture how those factors affect the strength of pre-election policy manipulation;
    - β captures direct effect of X_c,t on likelihood of reforms;
    - country-year and month fixed effects included as in main specification.

*Source: IMF working paper — 2. Data and Stylized Facts (section text).*

### 4. Results

### 4. Results

### 4.1. Main results
- Regression framework: linear probability model based on equation 1. Columns reported: (1) forward election dummy for 6-month period; (2) adds months of elections; (3) adds lagged values of the election dummy.
- Core empirical finding: the probability of a tax reform announcement declines during months running up to elections and declines in election months; increases in some months after elections.
- Key magnitudes and statistical significance (based on column (3) in Table 1):
  - 퐸푙푒푐푡푖표푛푡+6: -0.047*** (0.012)
  - 퐸푙푒푐푡푖표푛푡+5: -0.015 (0.019)
  - 퐸푙푒푐푡푖표푛푡+4: -0.036** (0.015)
  - 퐸푙푒푐푡푖표푛푡+3: -0.038*** (0.014)
  - 퐸푙푒푐푡푖표푛푡+2: -0.006 (0.021)
  - 퐸푙푒푐푡푖표푛푡+1: -0.047*** (0.015)
  - 퐸푙푒푐푡푖표푛푡: -0.051*** (0.016) [when included in specification]
  - Lagged positive post-election effect notably at 퐸푙푒푐푡푖표푛푡−3: 0.087*** (0.029) in specifications that include lags (Table 1, column 3 shows this pattern in lags).
- Aggregate interpretation:
  - During the 6-month window before elections, the cumulative decline in the probability of reform announcements is 16.9 pp (corresponding to a 2.8 pp lower probability per month over that period on average).
  - Election months: probability declines by 4.7 pp (example magnitude reported).
  - These effects are large relative to the sample average monthly probability of tax reform announcements: 5.2 percent.
- Interpretation: governments avoid announcing tax increases before and during elections; there is a short post-election "window of opportunity" (statistically significant primarily at the third month after elections) when governments are more likely to announce tax increases, consistent with a political-opportunity explanation.

- Table 1 summary statistics (column (3)):
  - Country-year and month F.E.: Yes
  - R-squared: 0.143
  - Observations: 7,392
- Additional specification notes: results robust when extending the window to 12 months before and after elections; pattern of avoidance before elections persists over 12 months while post-election window remains short-lived, motivating focus on 6-month windows for main analysis.

### 4.2. Some robustness checks
- Dependent variable alternatives and model variants tested (Table 2):
  - Column 1: net changes in tax policy in a month (dummy = 1 if increases > decreases).
  - Column 2: number of tax increase announcements in a month.
  - Column 3: dummy accounting for minor tax increases.
  - Column 4: Probit model with analytical bias correction for incidental parameter problem.
  - Column 5: controls for lagged values of tax increase announcements.
  - Column 6: controls for lagged values of tax decrease announcements.
  - Column 7: controls for lagged values of the percentage of countries in the same region that announced reforms (excluding the country itself).
  - Column 8: adds lagged values of the industrial production index.
  - Column 9: adds lagged values of inflation.
- Selected robustness coefficients for 퐸푙푒푐푡푖표푛푡+6 across Table 2 specifications:
  - (1): -0.028*** (0.010)
  - (2): -0.049*** (0.017)
  - (3): -0.056*** (0.013)
  - (4) Probit: -0.147*** (0.021)
  - (5): -0.056*** (0.012)
  - (6): -0.049*** (0.012)
  - (7): -0.047*** (0.012)
  - (8): -0.052*** (0.013)
  - (9): -0.047*** (0.013)
- Additional robustness findings:
  - Coefficients for pre-election months (t+4, t+3, t+1, t) generally remain negative and in many specifications statistically significant.
  - Lagged post-election positive effects at 퐸푙푒푐푡푖표푛푡−3 persist across specifications (e.g., 0.059** (0.024) in column 1; 0.094*** (0.030) in column 3; 0.236*** (0.074) in column 4).
- Robustness exercises reported as performed (results available upon request):
  - (i) leave-one-country-out regressions for 21 remaining countries;
  - (ii) exclusion of countries above 75th or below 25th percentile in total number of reforms;
  - (iii) analyses for various subperiods (e.g., before/after the Global Financial Crisis of 2008), dropping early years, excluding 2018;
  - (iv) different clustering choices or no clustering;
  - (v) different combinations or absence of fixed effects;
  - (vi) excluding snap elections to address reverse causality concerns;
  - (vii) weighted regressions (weights: GDP, GDP per capita averages, or inverses) to test sensitivity to large/small or rich/poor economies;
  - (viii) including all control variables from Table 2 together.
- Table 2 summary statistics:
  - Country-year and month F.E.: Yes for all columns
  - R-squared range across columns: 0.125 to 0.173
  - Observations range across columns: 2,294 to 7,392

### 4.3. Explanatory power
- Evaluation metric: receiver operating characteristic (ROC) curve and area under the curve (AUC) for Probit models.
- Baseline comparison:
  - Probit model with fixed effects but without election dummies: AUC = 0.616 with a standard error of 0.015.
  - Interpretation: AUC = 0.616 is statistically different than 0.5 at the 1 percent significance level, implying the fixed effects model performs significantly better than random guessing (coin flip) in predicting tax reform announcements.
- Note: Figure 5 compares the goodness of fit of the Probit model with and without election dummies; the text reports the AUC for the fixed-effects-only model as 0.616 (s.e. 0.015). The increase in AUC when adding election dummies is discussed but the exact numeric increase is not provided in the supplied text.

*Source: 4. Results (from wpiea2022216-print-pdf)*

### 0.696 with a standard error of 0.014. This means that the AUC value of the full model it is statistically different

### wpiea2022216-print-pdf - 0.696 with a standard error of 0.014. This means that the AUC value of the full model it is statistically different

### Model predictive power and ROC/AUC
- AUC for the full model: 0.696 with a standard error of 0.014.
- The AUC is statistically different (i) from 0.5, and (ii) from the model which excludes election dummies at the 1 percent significance level.
- Inclusion of election dummies increases predictive power significantly; electoral cycles are informative of tax reform announcements, though not a perfect predictor.
- Figure description: Grey line = Probit model with election variables (fourth column in Table 2); Black line = Probit model excluding election dummies but keeping two sets of fixed effects. AUC ranges between 0 and 1.

### Role of political and institutional factors (results from interaction models)
- Specification: interaction between country-level political/economic variables and election dummies for pre-election period and election month (as in equation 2).
- Institutional quality:
  - Proxies used: (i) law and order (ICRG), (ii) bureaucratic quality (ICRG), (iii) civil liberties (Freedom House).
  - Finding: pre-election effect on probability of tax reforms is weaker in countries with relatively high institutional quality — strong institutions can shield tax policy from electoral interference.
- Table 3 model summary statistics (exact values preserved):
  - Country-year and month F.E.: Yes (in all columns).
  - R-squared: 0.144, 0.145, 0.144, 0.144, 0.143, 0.147, 0.145 (columns 1–7 respectively).
  - Observations: 7362, 7362, 7362, 7392, 7392, 7056, 7392 (columns 1–7 respectively).
- Selected coefficient patterns reported in Table 3 (coefficients with standard errors in parentheses; statistical significance markers preserved where shown):
  - Election t+6: -0.103** (0.044); -0.103** (0.051); -0.091* (0.048); -0.039*** (0.014); -0.041*** (0.014); -0.060*** (0.022); -0.044*** (0.012).
  - Election t+4: -0.124** (0.051); -0.152** (0.062); -0.180*** (0.057); -0.020 (0.018); -0.038** (0.016); -0.062*** (0.023); -0.034** (0.016).
  - Election t+1: -0.150*** (0.053); -0.203*** (0.063); -0.136** (0.057); -0.039** (0.018); -0.048*** (0.016); -0.065*** (0.024); -0.047*** (0.016).
  - Election t (election month): -0.145*** (0.047); -0.214*** (0.062); -0.151** (0.061); -0.036** (0.017); -0.045*** (0.017); -0.076*** (0.024); -0.045*** (0.017).
  - Election t−3: 0.088*** (0.029) (consistent across columns).
  - Interaction examples (Election × X_t):
    - Election t+4 × X_t: 0.018* (0.010); 0.035* (0.019); 0.027** (0.011); -0.075*** (0.028); 0.009 (0.040); 0.026 (0.027); -0.027 (0.048).
    - Election t+1 × X_t: 0.021* (0.011); 0.046** (0.020); 0.017 (0.011); -0.042 (0.028); 0.004 (0.004); 0.028 (0.028); -0.000 (0.054).
    - Election t × X_t: 0.020** (0.010); 0.049** (0.020); 0.020* (0.011); -0.052* (0.028); -0.008 (0.040); 0.041 (0.029); -0.023 (0.057).
- Notes on estimation:
  - Results based on equation 2.
  - Columns 1–3 examine institutional quality; columns 4–5 examine EMs and presidential systems; columns 6–7 examine fiscal rules and IMF programs.
  - Standard errors clustered at the country-year level. Significance: *** p<0.01, ** p<0.05, * p<0.1.
  - Footnote on annual-frequency variables: whenever X_c,t is available at annual frequency it is dropped due to multicollinearity from country-year fixed effects; otherwise X_c,t is added.

### Differences by country group, political system, fiscal rules, and IMF programs
- Emerging Markets (EMs):
  - Interaction coefficients indicate electoral cycles are somewhat more pronounced in EMs relative to AEs.
  - Consistent with literature that electoral cycles tend to be stronger where economic development and institutional strength are lower.
- Political system:
  - Presidential systems tested; no substantial differential effect found — political system does not have much impact on strength of electoral cycles in tax reforms.
- Fiscal rules and IMF-supported programs:
  - Columns 6–7: neither fiscal rules nor IMF-supported programs have a direct impact on likelihood of tax reforms in the sample.
  - Neither variable appears to have differential effects on electoral cycles in tax reforms.
  - Robustness: result on fiscal rules holds across types (debt, expenditure, budget, revenue) when tested separately.
  - Caveat on IMF programs: sample contains only seven countries with at least one IMF program during period considered, corresponding to around 6 percent of the sample regarding country-month observations (469 observations).

### Different tax measures and tax types
- Tax measures:
  - Both tax base changes and tax rate changes exhibit electoral cycles — governments avoid announcing increases in rate and/or base in months running up to elections and are more likely to announce such reforms in the first few months following elections.
  - Figure 6: coefficients presented as percentage points with 90 percent confidence intervals (vertical lines).
- Tax types:
  - Dependent variables replaced with dummies for increases in CIT, PIT, VAT, SSC, EXE, and PRO.
  - Findings: reforms in most tax types exhibit electoral cycles.
  - Exception: social security contributions (SSC) — electoral cycle effects are less pronounced for SSC.
  - Note: number of reform announcements relatively low for SSC, EXE, and PRO, which limits interpretability for those tax types.

### Conclusions and policy implications
- Main empirical conclusions:
  - Tax reforms exhibit electoral cycles: governments tend to avoid announcing reforms that increase tax rates and/or broaden the tax base in months before elections and are more likely to announce them in the first few months after elections.
  - Post-election increase in reform likelihood suggests political capital or perceived political opportunity influences timing.
  - Results robust to alternative dependent variable definitions, econometric specifications, and additional controls.
  - Electoral cycles in tax reforms are weaker in countries with stronger institutions (law and order, bureaucratic quality, civil liberties).
  - Fiscal rules do not appear to mitigate electoral cycles in tax reforms in this sample.
  - IMF-supported programs do not appear to affect strength of electoral cycles in the countries in the sample, but sample limitations (few EMs, absence of lower income developing countries) warrant caution.
- Policy implications and avenues:
  - Challenge: provide appropriate incentives for election-oriented governments to design tax policy optimally given fiscal rules alone do not appear sufficient to shield tax policy from election-driven timing.
  - Electoral cycles may complicate cross-border coordination of tax policy by affecting timing of reforms.
  - Suggested future research: explore role of IMF-supported programs in electoral cycles in tax reforms in developing countries as more tax reform data become available.

### Appendix — election list and reform counts
- Table A.1: Elections in the sample (country and month/year listings provided in the Appendix).
- Table A.2: Number of observations with reforms in each tax measure and type (exact counts):
  - Tax measure: Base 201; Rate 139.
  - Tax type: CIT 91; PIT 103; VAT 70; SSC 28; EXE 31; PRO 12.

*Source: wpiea2022216-print-pdf (IMF Working Paper content provided).*

### References

### References

### Political business cycles, elections, and macroeconomic policy
- Aidt, Toke, Zareh Asatryan, Lusine Badalyan, and Friedrich Heinemann. 2020. "Vote buying or (political) business (cycles) as usual?" Review of Economics and Statistics, 102: 409-425.
- Akhmedov, Akhmed, and Ekaterina Zhuravskaya, 2004. "Opportunistic political cycles: test in a young democracy setting." The Quarterly Journal of Economics, 119: 1301-1338.
- Alesina, Alberto, Gerald D. Cohen, and Nouriel Roubini., 1992. "Macroeconomic policy and elections in OECD democracies." Economics & Politics, 4: 1-30.
- Alesina, A., N. Roubini and G.D. Cohen, 1997, “Political Cycles and the Macroeconomy” (Cambridge: MIT Press).
- Alesina, Alberto, and Matteo Paradisi. 2017. "Political budget cycles: Evidence from Italian cities." Economics & Politics, 29: 157-177.
- Brender, Adi, and Allan Drazen. "Political budget cycles in new versus established democracies." Journal of Monetary Economics 52, no. 7 (2005): 1271-1295.
- Chauvet, Lisa, and Paul Collier. "Elections and economic policy in developing countries." Economic Policy 24, no. 59 (2009): 509-550.
- Drazen, Allan. "The political business cycle after 25 years." NBER Macroeconomics Annual 15 (2000): 75-117.
- Dubois, Eric. "Political business cycles 40 years after Nordhaus." Public Choice 166, no. 1 (2016): 235-259.
- Fair, Ray C. "The effect of economic events on votes for president." The review of economics and statistics (1978): 159-173.
- Nordhaus, W. D., 1975. “The political business cycle” Review of Economic Studies, 42: 169-190.
- Rogoff K., Sibert A. “Elections and macroeconomic policy cycles” Review of Economic Studies, 55 (1) (1988), pp. 1-16
- Rogoff K. “Equilibrium Political Budget Cycles” American Economic Review, 80 (1), pp. 21-36 (1990).
- Shi, Min, and Jakob Svensson. "Political budget cycles: Do they differ across countries and why?" Journal of Public Economics 90, no. 8-9 (2006): 1367-1389.

### Tax policy, tax reforms, and revenue
- Amaglobeli, David, Valerio Crispolti, Era Dabla-Norris, Pooja Karnane, and Florian Misch. “Tax policy measures in advanced and emerging economies: a novel database”. International Monetary Fund Working Paper No 110, 2018.
- Amaglobeli, David, Valerio Crispolti, and Xuguang Simon Sheng. “Cross-Country Evidence on the Revenue Impact of Tax Reforms.” International Monetary Fund Working Paper No 199, 2022.
- Berry, Frances Stokes, and William D. Berry. "Tax innovation in the states: Capitalizing on political opportunity." American Journal of Political Science (1992): 715-742.
- Berry, Frances Stokes, and William D. Berry. "The politics of tax increases in the states." American Journal of Political Science (1994): 855-859.
- Biswas, Siddhartha, Indraneel Chakraborty, and Rong Hai. "Income inequality, tax policy, and economic growth." The Economic Journal 127, no. 601 (2017): 688-727.
- Clausing, Kimberly A. "Corporate tax revenues in OECD countries." International Tax and Public Finance 14, no. 2 (2007): 115-133.
- Ehrhart, Helene. "Elections and the structure of taxation in developing countries." Public Choice 156, no. 1 (2013): 195-211.
- Fuest, Clemens, Klaus Gründler, Niklas Potrafke, and Fabian Ruthardt. "Read My Lips? Taxes and Elections." CESifo Working Paper (2021).
- Gunter, Samara, Daniel Riera-Crichton, Carlos A. Vegh, and Guillermo Vuletin. "Non-linear effects of tax changes on output: The role of the initial level of taxation." Journal of International Economics 131 (2021): 103450
- Gupta, Sanjeev, and João Tovar Jalles. "On the political economy determinants of tax reforms: evidence from developing countries." (2020).
- Lami, Endrit, and Drini Imami., 2019 "Electoral cycles of tax performance in advanced democracies." CESifo Economic Studies, 65: 275-295.
- Mertens, Karel, and Morten O. Ravn. 2012. "Empirical Evidence on the Aggregate Effects of Anticipated and Unanticipated US Tax Policy Shocks." American Economic Journal: Economic Policy, 4 (2): 145-81.
- Mertens, Karel, and Morten O. Ravn. 2013. "The Dynamic Effects of Personal and Corporate Income Tax Changes in the United States." American Economic Review, 103 (4): 1212-47.
- Vegh, Carlos A., and Guillermo Vuletin. 2015. "How Is Tax Policy Conducted over the Business Cycle?" American Economic Journal: Economic Policy, 7 (3): 327-70.

### Capital flows, monetary policy, and macroprudential measures
- Clark, William Roberts, and Mark Hallerberg. "Mobile capital, domestic institutions, and electorally induced monetary and fiscal policy." American Political Science Review 94, no. 2 (2000): 323-346.
- Gavoille, Nicolas, and Katharina Hofer. "Capital Controls and Electoral Cycles." IMF Economic Review 69, no. 2 (2021): 275-324.
- Julio, Brandon, and Youngsuk Yook. "Political uncertainty and corporate investment cycles." The Journal of Finance 67, no. 1 (2012): 45-83.
- Sever, Can, and Emekcan Yücel, 2022. "The effects of elections on macroprudential policy." Journal of Comparative Economics, 50 (2): 507-533.

### Fiscal institutions, fiscal rules, and fiscal councils
- Davoodi, Hamid R., Paul Elger, Alexandra Fotiou, Daniel Garcia-Macia, Xuehui Han, Andresa Lagerborg, W. Raphael Lam, and Paulo A. Medas. "Fiscal Rules and Fiscal Councils: Recent Trends and Performance during the COVID-19 Pandemic." International Monetary Fund Working Paper No 11, 2022.
- Persson T., Tabellini G. “Macroeconomic Policy, Credibility and Politics” Hardwood Academics Publishers, Chur, Switzerland (1990)
- Schuknecht, Ludger., 2000 "Fiscal policy cycles and public expenditure in developing countries." Public Choice 102: 113-128.

### Structural reforms and macroeconomic effects
- Da Silva, A., Givone, A., and Sondermann, D., 2017. “When Do Countries Implement Structural Reforms” ECB Working Paper No. 2078, European Central Bank.
- David, A. C., Komatsuzaki, T., and Pienknagura, S.. “The Macroeconomic Effects of Structural Reforms in Latin America and the Caribbean” International Monetary Fund Working Paper No 195, 2020.
- Leeper, Eric M., Todd B. Walker, and Shu‐Chun Susan Yang. "Fiscal foresight and information flows." Econometrica 81, no. 3 (2013): 1115-1145.

### Databases, empirical methods, and technical references
- Amaglobeli, David, Valerio Crispolti, Era Dabla-Norris, Pooja Karnane, and Florian Misch. “Tax policy measures in advanced and emerging economies: a novel database”. International Monetary Fund Working Paper No 110, 2018.
- Armingeon, Klaus, Sarah Engler and Lucas Leemann. 2022. Comparative Political Data Set 1960-2020. Zurich: Department of Political Science, University of Zurich.
- Cruz, Cesi, Philip Keefer, and Carlos Scartascini. 2021. “Database of Political Institutions 2020”. Washington, DC: Inter-American Development Bank Research Department.
- Political Risk Services Group. The International Country Risk Guide (ICRG) Database.
- International Monetary Fund. International Financial Statistics (IFS) Database.
- Freedom House. Freedom in the World Database.
- Cruz-Gonzalez, Mario, Iván Fernández-Val, and Martin Weidner. 2017. "Bias corrections for probit and logit models with two-way fixed effects." The Stata Journal 17: 517-545.
- Lancaster, Tony, 2000. "The incidental parameter problem since 1948." Journal of Econometrics 95: 391-413.
- Lami, Endrit, and Drini Imami., 2019 "Electoral cycles of tax performance in advanced democracies." CESifo Economic Studies, 65: 275-295.

### Additional theoretical and empirical contributions
- Clausing, Kimberly A. "Corporate tax revenues in OECD countries." International Tax and Public Finance 14, no. 2 (2007): 115-133.
- Chauvet, Lisa, and Paul Collier. "Elections and economic policy in developing countries." Economic Policy 24, no. 59 (2009): 509-550.
- Gunter, Samara, Daniel Riera-Crichton, Carlos A. Vegh, and Guillermo Vuletin. "Non-linear effects of tax changes on output: The role of the initial level of taxation." Journal of International Economics 131 (2021): 103450
- Mertens, Karel, and Morten O. Ravn. 2012. "Empirical Evidence on the Aggregate Effects of Anticipated and Unanticipated US Tax Policy Shocks." American Economic Journal: Economic Policy, 4 (2): 145-81.
- Mertens, Karel, and Morten O. Ravn. 2013. "The Dynamic Effects of Personal and Corporate Income Tax Changes in the United States." American Economic Review, 103 (4): 1212-47.
- Vegh, Carlos A., and Guillermo Vuletin. 2015. "How Is Tax Policy Conducted over the Business Cycle?" American Economic Journal: Economic Policy, 7 (3): 327-70.

*References as listed in the content unit.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022216-print-pdf.pdf_