## 1.   Correlations of Economic Policies

## Source details

**Canonical URL:** [1.   Correlations of Economic Policies](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05118.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05118.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05118.pdf.json)

---

### Introduction and approach
- Broad agreement: sound macroeconomic management and structural reforms positively affect economic performance.
- Complementarity: macro policies address imbalances; structural reforms improve transmission of fiscal and monetary policies and strengthen supply response.
- Policy design requires:
  - Integrating initial conditions and diagnosing supply-side bottlenecks.
  - Paying attention to political factors that may strengthen support for sound economic policies.
  - Focusing on a high-payoff reform agenda to enhance credibility and provide early evidence of benefits.
- Identification challenges:
  - High correlation of policy indicators and diversity of indicators (scale differences and colinearity).
  - Need to tailor pace and sequence of reforms to policy distortions and political constraints.
- Empirical approach:
  - Apply factor analysis (FA) to create fewer, uncorrelated variables (“clusters of economic policy”).
  - Examine pace of reforms (deviations from trend) and sequencing effects on growth using panel data of four 5-year periods (1981–2000).

### Factor analysis: purpose, properties, and implementation
- Purpose:
  - Detect structure among correlated indicators by creating latent factors representing independent sources of variation while minimizing information loss.
- Properties and distinctions:
  - Derived factors normalized with mean equal to zero and variance equal to one.
  - Factors are uncorrelated and preserve much of the original variation.
  - FA differs from PCA: FA detects data structure; PCA minimizes total residual variance.
- Rules and caveats:
  - Number of factors decided using Kaiser criterion, scree test, percentage of variation explained, and interpretability.
  - Factors are hypothetical; FA is heuristic and not unique.
- Algebraic representation (verbatim style in source):
  - jjmjmjjj
    YuFaFaFaz++++=K
    2211
    where   j=1, 2, ..., n
  - Interpretation: each observed variable described linearly by m common factors and one unique factor; “loadings” and “factor loadings” terminology preserved.

### Design of economic policies: data, variables, and econometric methodology
- Data and horizon:
  - Panel dataset of 5-year periods (four 5-year periods between 1981 and 2000).
  - Rationale: medium-term horizon appropriate to assess permanent effects of policies; some transitional factors not captured by 5-year averages.
- Growth equation specified:
  - G = f (IC , PC , SH , POL)
    - IC = initial conditions
    - PC = country-specific choices
    - SH = shocks (domestic or external)
    - POL = policy environment (divided into MP, SP, IF)
  - Dependent variable: growth rate of per capita output (G).
- Non-policy regressors and instruments:
  - Initial income instrumented by lagged values of Penn World data.
  - Private choices proxied by fertility rates.
  - Environmental variables include internal shocks (e.g., weather) and external shocks (e.g., terms of trade).
- Policy categories and indicators:
  - Macroeconomic policy (MP): inflation (CPI), fiscal balance (FB), inflation threshold effect (ITE).
  - Structural policies (SP): openness to external trade (OPEN), exchange rate premium/index (ERP; interval (0, 1] where a number close to 1 implies a low premium), share of credit to private sector in total credit (PSC).
  - Institutional factors (IF): simple average of political risk category in ICRG index.
- Estimation sample and methods:
  - Estimation limited to 1981-2000 (four 5-year periods).
  - Iterative regressor selection retaining variables significant at the 10 percent level.
  - Instrumental variables for non-policy regressors; aggregation over 5-year periods to reduce endogeneity concerns.
  - Two-stage least squares estimation with GLS correction for heteroscedasticity.
  - Departures from Batista and Zalduendo (BZ): different ERP index, exclusion of human capital from initial conditions, application of FA to all policy indicators.

### FA application and the five policy clusters (F1–F5)
- FA implementation:
  - Varimax rotation used; decision to retain five factors largely based on interpretability.
  - Unique variance of each original variable small (0 to 27 percent), supporting five-cluster representation.
- Five policy clusters (interpretation from factor loadings):
  - Lack of Economic Stability (inflation-related)
    - Positively associated with CPI and ITE; interpreted as negatively correlated with growth.
  - Lack of Fiscal Sustainability
    - Related to fiscal position (FB); increases reduce per capita growth.
  - Enabling Business Environment
    - Includes ERP and ICRG institutional index; factor loadings show closer correlation with price liberalization (0.93) than institutional factors (0.72).
  - Trade Liberalization
    - Reflects openness to external trade (OPEN).
  - Financial Sector Development
    - Represented by PSC (credit to private sector share), reflecting banking intermediation role and need for financial/institutional reforms.
- Expected correlations with growth:
  - Business Environment, Trade Liberalization, Financial Sector Development: expected positive correlation with growth.
  - Economic Stabilization and Fiscal Sustainability: expected negative correlation with growth (inflation and lack of fiscal consolidation hurt per capita growth).

### Benchmark econometric results and fit diagnostics
- Estimation details:
  - Two-stage least squares with GLS correction.
  - Combined time and cross-section variation: 253 observations for 81 countries across four 5-year periods (1981–2000).
  - Non-policy regressors instrumented by lagged values.
- Main findings:
  - Coefficients on initial conditions, shocks, and private choices generally have expected signs and are statistically and economically significant.
  - Low inflation and strong fiscal positions support growth; evidence of a non-linear relationship between inflation and growth.
  - Reforms that liberalize trade, improve financial intermediation, and liberalize prices support growth.
  - Coefficient estimates on the ICRG index and terms of trade have expected sign but are not statistically significant.
  - Estimation results using FA-derived clusters are equivalent to those using the original indicators.
- Fit comparisons:
  - Standard error of regression: 0.0212 (original dataset) versus 0.0220 (FA-derived factors).
  - R-squared: 0.47 (without FA) versus 0.43 (with FA).
  - Number of observations: 253.
  - Number of different countries: 81.
  - Qualitative finding: enabling business environment (F1), intermediation role of banks (F2), and trade liberalization (F4) support per capita income growth; high inflation (F3) and lack of fiscal sustainability (F5) hurt growth. All cited coefficients statistically significant at the 1 percent level (unless otherwise noted).

### Growth and level effects of reforms (NEW F1–NEW F5)
- Mapping factors into [0, 1] and transforming so positively correlated with growth (NEW F1–NEW F5).
- Estimation sample for NEW factors: 172 observations; number of different countries: 61.
- Regression summary (Equation 4: combined contemporaneous and lagged effects):
  - Standard error of regression: 0.0208
  - R-squared: 0.45
  - Number of observations: 172
  - Number of different countries: 61
- Findings on contemporaneous and lagged effects:
  - Contemporaneous policy coefficients in pooled specifications are positive and statistically significant.
  - Lagged policy indicators often have negative signs, but the sum of contemporaneous and lagged coefficients is positive.
  - Interpretation: sound policies produce an initial spurt in growth (level effect) that weakens over time; improvements can generate one-time investment or threshold-driven gains rather than permanently higher growth rates. Policy design and growth projections should account for level effects, not only steady-state shifts.

### Magnitudes: effects on annual growth rates (based on Equation 2 in Table 5)
- Business environment: coefficient 0.04; standard deviation 0.14; annual growth effect 0.51
- Financial sector development: coefficient 0.02; standard deviation 0.16; annual growth effect 0.28
- Economic stabilization: coefficient 0.06; standard deviation 0.07; annual growth effect 0.42
- Trade liberalization: coefficient 0.07; standard deviation 0.07; annual growth effect 0.49
- Fiscal sustainability: coefficient 0.07; standard deviation 0.08; annual growth effect 0.58
- Summary: improving each policy cluster by one standard deviation leads to improvements in growth that range from 0.3 to 0.6 percent per year (reported range in text).

### Pace of policy implementation: deviations and speed (Table 7)
- Two pace measures:
  - Cross-country deviation (deviation from world average policies).
  - Speed relative to a country’s own trend (deviation from trend).
- Selected coefficient estimates (Equation 5: cross-country deviation; Equation 6: speed relative to trend)
  - Cross-country deviation of F1: 0.0580  ***
  - Cross-country deviation of F2: 0.0206  ***
  - Cross-country deviation of F3: 0.0803  ***
  - Cross-country deviation of F4: 0.0868  ***
  - Cross-country deviation of F5: 0.0808  ***
  - Speed F1: 0.0264 (t-statistic 1.15; not significant at conventional levels)
  - Speed F2: 0.0313  *** (t-statistic 2.93)
  - Speed F3: 0.0692  * (t-statistic 1.86)
  - Speed F4: 0.2242  *** (t-statistic 3.27)
  - Speed F5: 0.0890  *** (t-statistic 4.01)
- Interpretation:
  - Faster implementation, especially relative to a country’s own trend, is associated with higher growth rates; many speed coefficients are positive and statistically significant.
  - Early reformers tend to perform better; rapid implementation tends to raise growth, reflecting transitional factors and shifts toward a higher steady state.

### Sequencing of economic policies: interaction-term evidence (Table 8)
- Three sequencing links identified:
  - Economic stabilization and fiscal sustainability:
    - Lag F5 * NEW F3 coefficient: -0.7610; t-statistic: -3.06; statistically significant and large (equation 8).
    - Lag F3 * NEW F5 coefficient: -0.0579; t-statistic: -0.15; not significant (equation 7).
    - Implication: stabilization (control of inflation) should be prioritized over fiscal sustainability; price signals matter strongly for growth.
  - Stabilization before trade liberalization:
    - Lag F3 * NEW F4 coefficient: 0.6278  **; t-statistic: 2.25 (equation 9).
    - Implication: securing stabilization prior to trade liberalization yields stronger growth benefits.
  - Fiscal sustainability before financial liberalization:
    - Lag F5 * NEW F2 coefficient: 0.2559  ***; t-statistic: 3.27 (equation 9).
    - Implication: fiscal sustainability should be addressed before liberalizing financial sector activities to avoid amplifying fiscal vulnerabilities.
- Caveat: interaction-term results identify important pre-conditions and complementarities rather than absolute, one-size-fits-all sequences.

### Policy design implications and recommendations
- Use FA-derived policy clusters (F1–F5) to reduce multicollinearity and identify coherent policy packages.
- Anticipate both level and growth effects from reforms:
  - Expect an initial growth spurt (level effect) after reform; temper long-term growth projections accordingly.
- Pace matters:
  - Faster implementation of sound policies generally produces higher growth—early reforms are beneficial.
- Sequence matters (priority ordering suggested by results):
  - Prioritize economic stabilization (control inflation) as a first-order condition for growth.
  - Secure fiscal sustainability before undertaking financial liberalization.
  - Achieve stabilization before undertaking trade liberalization.
- Recognize second-best considerations: when distortions are entrenched, optimal reform sequences and pace can materially affect growth outcomes.

*Source: _wp05118 - 1. Correlations of Economic Policies (IMF Working Paper).*

### 1.   Correlations of Economic Policies..................................................................................

### _wp05118 - 1.   Correlations of Economic Policies

### Introduction
- Broad agreement that sound macroeconomic management (primarily policies aimed at securing economic stability and fiscal sustainability) and structural reforms have a positive effect on economic performance.
- Structural reforms emphasis traced to the early 1980s when a number of developing countries were having difficulties in adjusting to the oil shocks of the preceding decade.
- Macroeconomic management and structural reforms are complementary: macro policies address economic imbalances; structural reforms improve the transmission mechanism of fiscal and monetary policies and strengthen supply response.
- Policy design requires:
  - Integrating initial conditions and a correct diagnosis of supply-side bottlenecks.
  - Paying close attention to political factors that may strengthen support for sound economic policies (de Melo et al. (1995) identify a positive correlation between reforms and political liberalization).
  - Focusing on a high-payoff reform agenda to enhance government credibility and provide early evidence of benefits from policy changes.
- Identification of empirical regularities is hampered by:
  - High correlation of policy indicators (Staehr (2003): “one of the reasons why numerous studies have failed in pinpointing economic policies that matter for growth is the high correlation of policy indicators.”).
  - Diversity of indicators (scale differences and colinearity) that complicate identifying “second best regularities.”
  - Need to tailor pace and sequence of reforms to existing policy distortions and political constraints.
- Approach of the paper:
  - Apply factor analysis (FA) to limit effects of high correlation and indicator diversity.
  - Use FA to create new, fewer variables representing independent sources of variation (“clusters of economic policy”).
  - Examine pace of reforms (deviations from trend values of policy) and evidence on optimal reform sequences affecting growth impact.
- Paper organization (as provided):
  - Section II describes factor analysis techniques.
  - Section III examines design aspects of economic policies, focusing on speed and sequencing of macroeconomic and structural reforms.

### Factor Analysis (FA): purpose and properties
- Main objective: detect structure among a set of different indicators by creating new variables that represent independent sources of variation while minimizing information loss.
- Underlying assumption: unobserved latent variables account for the pattern of relationships among observed (and frequently correlated) variables; inferred independent variables are called factors.
- FA characteristics and distinctions:
  - Extracts proportion of variance due to common factors (shared among several variables) and generates new variables representing independent sources of variation.
  - Derived factors are normalized with mean equal to zero and variance equal to one.
  - Derived factors are uncorrelated and preserve much of original variation.
  - FA differs from principal components analysis (PCA):
    - PCA generates new variables that minimize total residual variance; the first principal component extracts the maximum variance, subsequent components extract maximum remaining variance.
    - FA is preferred to detect data structure; PCA is preferred to reduce data into fewer variables.
- Rules for deciding number of factors (as discussed):
  - Kaiser criterion: drop all factors with eigenvalues less than (or close to) one.
  - Scree test: plot factors (x axis) against eigenvalues (y axis); drop all factors that follow the cease in the plot’s decline.
  - Percentage of variation explained rule: keep factors until, say, 90 percent of all the variation is explained.
  - Interpretability criterion: keep as many factors as required to interpret the underlying structure.
- Caveats:
  - Factors are hypothetical and FA is a heuristic approach to data representation—convenient but not unique.
  - Many different factor combinations can represent the original set of variables; simpler theoretical representations increase discrepancy with data, while more accurate data depiction reduces interpretability of underlying theory.
- Algebraic representation (verbatim from source):
  - jjmjmjjj
    YuFaFaFaz++++=K
    2211
    where   j=1, 2, ..., n
  - Interpretation: each of the n observed variables is described linearly in terms of m common factors and one unique factor; common factors (F) account for correlations among variables; each unique factor (Y) accounts for remaining variance (including the error). Coefficients of the factors are referred to as “loadings.” The correlation coefficients between the variables and the factors are called factor loadings. The squared factor loading is the percent of variance explained by the factor. The eigenvalue for a given factor measures the variance in all the variables that are accounted by that factor.

### Design of Economic Policies (empirical approach)
- Empirical focus: panel dataset of 5-year period—i.e., four 5-year periods between 1981 and 2000.
- Rationale for 5-year (medium-term) horizon:
  - Advantages:
    - Appropriate for assessing the permanent effects of economic policies on growth.
    - Serves to identify appropriate pace and sequence of economic policies.
  - Disadvantages:
    - Some transitional factors cannot be covered with data based on 5-year averages.
- FA use in this context:
  - Create “clusters of economic policy” (new normalized variables) to address high correlation and colinearity among policy indicators.
  - Reduce instability of coefficient estimates in regression analysis by using fewer, uncorrelated factors.
  - Facilitate examination of reform pace (deviations from trend) and sequencing effects on growth.

*Source: _wp05118 - 1.   Correlations of Economic Policies*

### conclusions in this paper are in terms of what is critical for growth and the combinations of

### _wp05118 - conclusions in this paper are in terms of what is critical for growth and the combinations of policies that help strengthen a country’s growth prospects, but do not allow an assessment of policy interactions over shorter time periods.

### A. Econometric Methodology
- Growth equation specified as: G = f (IC , PC , SH , POL), where
  - IC = initial conditions,
  - PC = country-specific choices,
  - SH = shocks (domestic or external),
  - POL = policy environment (divided into MP, SP, IF).
- Dependent variable: growth rate of per capita output (G).
- Regressors not contemporaneously related to policies:
  - Initial income (instrumented by lagged values of the Penn World data).
  - Private choices proxied by fertility rates.
  - Environmental variables (internal shocks, e.g., weather; external shocks, e.g., terms of trade).
- Policy categories and indicators:
  - Macroeconomic policy (MP): inflation (CPI), fiscal balance (FB), inflation threshold effect (ITE).
  - Structural policies (SP): openness to external trade (OPEN), price liberalization / exchange rate premium (ERP; index defined over the interval (0, 1] where a number close to 1 implies a low premium), share of credit to the private sector in total credit (PSC).
  - Institutional factors (IF): simple average of the political risk category in the ICRG index.
- Estimation sample and period focus:
  - Estimation limited to 1981-2000 (4 five-year periods).
- Methodological choices to mitigate common reduced-form problems:
  - Iterative regressor selection retaining variables significant at the 10 percent level (following Batista and Zalduendo (BZ) paper).
  - Instrumental variables for non-policy regressors and aggregation over 5-year periods to reduce endogeneity concerns.
  - Two-stage least squares estimation with GLS correction for heteroscedasticity (see Section C).
- Departures from the BZ paper:
  - Different index on black market exchange rate premium (interval (0, 1]).
  - Human capital excluded from initial conditions (not statistically significant after 1970s oil shocks).
  - Application of factor analysis to all economic policy indicators.

### B. Application of Factor Analysis to Policy Indicators
- Motivation and method:
  - Factor analysis applied to all proxies of economic policies to identify individual policy clusters and derive indicators representing speed and sequencing issues.
  - Varimax rotation used to facilitate interpretation of factor loadings.
  - Decision taken to retain five factors largely based on interpretability of the data.
- Diagnostic on uniqueness and variance:
  - Unique variance of each original variable is small (0 to 27 percent), supporting representation of policies through five clusters.
- Five policy clusters (descriptions drawn from factor loadings and interpretation):
  - Lack of Economic Stability
    - Reflects lack of success in keeping low and stable inflation rates.
    - Positively associated with CPI and the inflation threshold effect (ITE).
    - Interpreted as negatively correlated with growth (positive factor loadings on inflation variables imply negative growth correlation).
  - Lack of Fiscal Sustainability
    - Relates to fiscal position (FB) and represents lack of sustainability given negative factor loadings for fiscal balance.
    - An increase in this factor should reduce per capita growth.
  - Enabling Business Environment
    - Includes exchange rate risk premium (ERP) and ICRG institutional index.
    - Reflects market-determined prices and institutional support for private sector activity; factor loadings suggest closer correlation with price liberalization than institutional factors (0.93 for price liberalization and 0.72 for ICRG as reported).
  - Trade Liberalization
    - Reflects policies opening the economy to competitive pressures (OPEN).
  - Financial Sector Development
    - Represented by PSC (credit to the private sector as a share of total credit), reflecting the banking sector’s role in channeling resources to the private sector and the need for financial and institutional reforms.
- Expected correlations with growth:
  - Business Environment, Trade Liberalization, and Financial Sector Development factors are expected to be positively correlated with growth due to large positive factor loadings on ERP/ICRG, OPEN, and PSC respectively.
  - Economic Stabilization and Fiscal Sustainability factors are expected to be negatively correlated with growth (inflation and lack of fiscal consolidation reduce per capita growth).

### C. Benchmark Econometric Results
- Estimation details:
  - Two-stage least squares with GLS correction for heteroscedasticity.
  - Combined time and cross-section variation: 253 observations for 81 countries in the four 5-year periods between 1981 and 2000.
  - Non-policy regressors instrumented by their lagged values.
- Main findings (mirrors BZ paper results with some departures):
  - Most coefficient estimates on initial conditions, shocks, and private choices have expected signs and are statistically and economically significant.
  - Low inflation and strong fiscal positions support growth; evidence of a non-linear relationship between inflation and growth.
  - Growth is supported by reforms that liberalize trade, improve the financial intermediation role of banks, and liberalize prices.
  - Departures from the BZ paper: coefficient estimates on the ICRG index and on the terms of trade regressor have the expected sign but are not statistically significant.
- Equivalence of policy-cluster approach:
  - Estimation results using clusters of economic policy derived through factor analysis are equivalent to those using the original dataset of economic indicators.

*Source: Conclusions and methodology section of _wp05118 (PDF).*

### conclusions reached remain broadly in line with those discussed in the BZ paper.

### _wp05118 - conclusions reached remain broadly in line with those discussed in the BZ paper.

### Factor analysis and regression fit
- Factor analysis (FA) groups policy indicators into clusters (F1–F5) and produces derived factors with limited information loss.
- Standard error of regression: 0.0212 (original dataset) versus 0.0220 (factors derived by FA).
- R-squared: 0.47 (without FA) versus 0.43 (with FA).
- Number of observations: 253.
- Number of different countries: 81.
- Key qualitative finding: an enabling business environment (F1), the intermediation role of banks (F2), and trade liberalization (F4) support per capita income growth; high inflation (F3) and lack of fiscal sustainability (F5) hurt growth. All cited coefficients are statistically significant at the 1 percent level (unless otherwise noted).

### Growth and level effects of reforms
- Mapping factors into [0, 1] and transforming so they are positively correlated with growth facilitates interpretation (NEW F1–NEW F5).
- Estimation sample for NEW factors: 172 observations; number of different countries: 61.
- Regression summary (selected):
  - Equation 4 (combined contemporaneous and lagged effects):
    - Standard error of regression: 0.0208
    - R-squared: 0.45
    - Number of observations: 172
    - Number of different countries: 61
  - Contemporaneous policy coefficients in pooled specifications are positive and statistically significant; lagged policy indicators often have negative signs, but the sum of contemporaneous and lagged coefficients is positive.
- Interpretation:
  - Sound economic policies produce an initial spurt in growth (level effect) that weakens over time; thus policy improvements can generate one-time investment or threshold-driven gains rather than permanently higher growth rates.
  - Policy design and growth projections should account for level effects, not only steady-state shifts.

### Magnitudes: effects on annual growth rates (Table 6; based on Equation 2 in Table 5)
- Business environment: coefficient 0.04; standard deviation 0.14; annual growth effect 0.51
- Financial sector development: coefficient 0.02; standard deviation 0.16; annual growth effect 0.28
- Economic stabilization: coefficient 0.06; standard deviation 0.07; annual growth effect 0.42
- Trade liberalization: coefficient 0.07; standard deviation 0.07; annual growth effect 0.49
- Fiscal sustainability: coefficient 0.07; standard deviation 0.08; annual growth effect 0.58
- Summary: improving each policy cluster by one standard deviation leads to improvements in growth that range from 0.3 to 0.6 percent per year (reported range in text).

### Pace of policy implementation (Table 7)
- Two pace measures examined:
  - Deviation from world average policies (cross-country deviation).
  - Deviation from a country’s own trend path (speed).
- Selected coefficient estimates (Equation 5: cross-country deviation; Equation 6: speed relative to trend)
  - Cross-country deviation of F1: 0.0580  ***
  - Cross-country deviation of F2: 0.0206  ***
  - Cross-country deviation of F3: 0.0803  ***
  - Cross-country deviation of F4: 0.0868  ***
  - Cross-country deviation of F5: 0.0808  ***
  - Speed F1: 0.0264 (t-statistic 1.15; not significant at conventional levels)
  - Speed F2: 0.0313  *** (t-statistic 2.93)
  - Speed F3: 0.0692  * (t-statistic 1.86)
  - Speed F4: 0.2242  *** (t-statistic 3.27)
  - Speed F5: 0.0890  *** (t-statistic 4.01)
- Interpretation:
  - Faster implementation (especially relative to a country’s own trend) is associated with higher growth rates; many speed coefficients are positive and statistically significant.
  - Early reformers tend to perform better; rapid implementation tends to raise growth, reflecting transitional factors and shifts toward a higher steady state.

### Sequencing of economic policies (Table 8; interactive terms)
- Three sequencing links identified (interaction-term evidence):
  - Economic stabilization and fiscal sustainability:
    - Lag F5 * NEW F3 coefficient: -0.7610; t-statistic: -3.06; statistically significant and large (equation 8).
    - Lag F3 * NEW F5 coefficient: -0.0579; t-statistic: -0.15; not significant (equation 7).
    - Implication: stabilization (control of inflation) should be prioritized over fiscal sustainability; price signals matter strongly for growth.
  - Stabilization before trade liberalization:
    - Lag F3 * NEW F4 coefficient: 0.6278  **; t-statistic: 2.25 (equation 9).
    - Implication: securing stabilization prior to trade liberalization yields stronger growth benefits.
  - Fiscal sustainability before financial liberalization:
    - Lag F5 * NEW F2 coefficient: 0.2559  ***; t-statistic: 3.27 (equation 9).
    - Implication: fiscal sustainability should be addressed before liberalizing financial sector activities to avoid amplifying fiscal vulnerabilities.
- Caveat: interactive-term results identify important pre-conditions and complementarities rather than absolute, one-size-fits-all sequences.

### Policy design implications and recommendations
- Use FA-derived policy clusters (F1–F5) to reduce multicollinearity and identify coherent policy packages.
- Expect both level and growth effects from reforms:
  - Anticipate an initial growth spurt (level effect) after reform; temper long-term growth projections accordingly.
- Pace matters:
  - Faster implementation of sound policies generally produces higher growth—early reforms are beneficial.
- Sequence matters (priority ordering suggested by results):
  - Prioritize economic stabilization (control inflation) as a first-order condition for growth.
  - Secure fiscal sustainability before undertaking financial liberalization.
  - Achieve stabilization before undertaking trade liberalization.
- Recognize second-best considerations: when distortions are entrenched, optimal reform sequences and pace can materially affect growth outcomes.

*Italic: IMF Working Paper conclusions as presented in the source content.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05118.pdf_