## wp17198

## Source details

**Canonical URL:** [wp17198](https://www.imf.org/-/media/files/publications/wp/2017/wp17198.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2017/wp17198.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2017/wp17198.pdf.json)

---

### I. Introduction and research design
- Context and motivation:
  - Since the Global Financial Crisis medium-term growth has been declining in both advanced and developing economies (IMF, 2017).
  - Fiscal policy has become increasingly constrained due to high debt-to-GDP ratios.
  - Fiscal stabilization can enhance medium-term growth by reducing aggregate volatility; theoretical and empirical literature documents a negative relationship between volatility and growth.
- Theoretical benchmark and mechanism:
  - Mechanism from Aghion et al. (2010): firms choose between short-term and long-term projects; credit constraints raise liquidation risk in bad states so reducing aggregate volatility increases survival of long-term projects. Predicts larger positive effects of fiscal stabilization in more credit-constrained industries and especially in bad times.
- Empirical extensions and contributions:
  - Considers additional channels beyond external financial dependence (e.g., asset fixity, labor intensity, investment lumpiness, input specificity, ISTC, R&D).
  - Uses a time-varying, country-year-varying measure of fiscal stabilization (three-dimensional country-sector-year panel) and an alternative cyclically-adjusted fiscal balance measure to isolate discretionary policy.
  - Extends analysis to developing economies and decomposes industry growth into labor, capital, and productivity channels.
- Data and identification:
  - Unbalanced panel: 22 manufacturing industries; 55 advanced and developing economies; period 1970-2014.
  - Industry technological characteristics measured from U.S. firm-level data (EFD, DEP, ISTC, RND, FIX, LAB, LMP, SPEC).
  - Econometric specification: Y_i,c,t = α_i,c + γ_i,t + δ_c,t + β X_i FS_c,t + ε_i,c,t with industry-country, industry-time, and country-time fixed effects; OLS with standard errors clustered at industry-country level.
  - Main independent variable: interaction between time-varying fiscal stabilization (FS_c,t) and industry characteristics X_i.

### II. Fiscal stabilization measure (FS): construction and stylized facts
- Concept:
  - FS is the responsiveness of budget balance-to-GDP to changes in economic activity; larger FS indicates more countercyclical (stabilizing) fiscal policy.
- Estimation approach:
  - Baseline regression b_i,t = α_i,t + FS_i,t Δy_i,t + ε_i,t with FS_i,t evolving slowly (FS_i,t = FS_i,t−1 + v_i,t) estimated via Schlicht varying-coefficient method (method-of-moments for σ_i^2).
  - Advantages: uses all observations per year, innovations in same year drive changes, accounts for slow policy changes, reduces reverse causality concerns.
- Empirical observations and statistics:
  - Time-average fiscal stabilization coefficient: about 0.25-0.30.
  - Additional country sample for FS estimates: 69 countries with at least 23 years of fiscal stabilization estimates (1994-2016).
  - Degree of fiscal stabilization has increased over time, particularly in advanced and emerging market economies; less change in low-income countries or oil exporters.
  - Greater fiscal stabilization is associated with lower output volatility on average (strong negative relationship between time-average FS and standard deviation of real GDP growth; negative relation also in 5-year non-overlapping averages purged by fixed effects).

### III. Industry characteristics and theoretical expectations
- Industry characteristics studied and expected interaction signs with fiscal stabilization:
  - External financial dependence (EFD): measured as median across U.S. firms of (total capital expenditures − current cash flow) / total capital expenditures (Rajan and Zingales, 1998). Expect positive interaction.
  - Capital depreciation (DEP): higher depreciation implies lower resale values and higher liquidity risk. Expect positive interaction.
  - Investment-specific technological change (ISTC): captures technological obsolescence via quality-adjusted price declines. Expect positive interaction.
  - R&D intensity (RND): R&D expenditures / total capital expenditure. Interaction sign ambiguous (theory yields either sign).
  - Asset fixity (FIX): fixed assets / total assets. Lower asset fixity (more intangible assets) implies harder collateralization; expect fiscal stabilization to increase growth more in industries with lower asset fixity (i.e., negative coefficient on FIX interaction).
  - Labor intensity (LAB): wages and salaries / value added. Expect positive interaction if labor-intensive firms benefit from stabilization.
  - Investment lumpiness (LMP): average number of investment spikes per firm per decade. Expect positive interaction.
  - Relationship-specific input usage (SPEC): share of inputs not sold on organized exchange. Expect positive interaction.

### IV. Baseline empirical findings (value-added growth interactions)
- Sample and estimation notes:
  - Full sample N=20786 (all economies); advanced economies N=11744; developing economies N=9042.
  - Differential effect defined as change in industry growth when moving industry from 25th to 75th percentile of characteristic while FS moves from 25th to 75th percentile.
- Main interaction coefficients and differential effects (selected, preserving numeric values exactly):
  - All economies (N=20786):
    - EFD: Coef 6.80; s.e 4.67; Differential effect 0.79
    - DEP: Coef 1.89*; s.e 1.14; Differential effect 2.64
    - ISTC: Coef 3.39; s.e 2.93; Differential effect 2.32
    - RND: Coef 1.16; s.e 4.88; Differential effect 0.09
    - FIX: Coef -52.77***; s.e 14.69; Differential effect -2.15
    - LAB: Coef 44.80**; s.e 19.25; Differential effect 2.98
    - LMP: Coef 6.08**; s.e 2.78; Differential effect 1.29
    - SPEC: Coef 22.69**; s.e 11.57; Differential effect 3.14
  - Advanced economies (N=11744):
    - EFD: Coef 5.45**; s.e 2.63; Differential effect 0.64
    - FIX: Coef -19.46**; s.e 9.78; Differential effect -0.79
    - LAB: Coef 31.90**; s.e 13.21; Differential effect 2.13
    - LMP: Coef 2.77*; s.e 1.47; Differential effect 0.59
    - SPEC: Coef 11.24*; s.e 6.52; Differential effect 1.56
  - Developing economies (N=9042):
    - FIX: Coef -73.07***; s.e 25.67; Differential effect -2.97
    - LAB: Coef 54.00*; s.e 32.09; Differential effect 3.59
    - LMP: Coef 7.83*; s.e 4.77; Differential effect 1.68
    - SPEC: Coef 34.74*; s.e 20.92; Differential effect 4.80
    - EFD: Coef 6.43; s.e 8.53; Differential effect 0.75
- Key interpretative points:
  - Asset fixity (FIX) shows large negative interaction coefficients consistently significant across samples, implying industries with higher FIX fare worse with greater fiscal stabilization movements.
  - Labor intensity (LAB), investment lumpiness (LMP), and relationship-specific investment (SPEC) commonly show positive and often statistically significant interaction coefficients.
  - External financial dependence (EFD) has positive coefficients in most specifications; statistical significance varies by subgroup.

### V. Robustness checks and alternative specifications (highlights)
- Lagged FS interactions (Table 6, N=20786):
  - EFD: Coef 7.28*; s.e 4.42; Differential effect 0.85
  - FIX: Coef -32.59**; s.e 15.44; Differential effect -1.33
  - LAB: Coef 48.31**; s.e 19.21; Differential effect 3.22
- Gross output vs value added (Table 7):
  - Value added FIX -52.77***; s.e 14.69; differential -2.15
  - Gross output FIX -35.23***; s.e 13.92; differential -2.35
  - SPEC positive and significant in both measures.
- Weighted least squares to account for FS estimation uncertainty (Table 8, N=20786):
  - FIX: Coef -66.08***; s.e 17.74; Differential effect -2.69
  - SPEC: Coef 36.32**; s.e 17.52; Differential effect 5.03
- Alternative FS measures (Table 9):
  - Primary balance to GDP (N=19158): FIX -13.53*; s.e 8.32; Differential effect -0.55; LAB 19.38**; s.e 9.97; Differential effect 1.29
  - Cyclically-adjusted balance to GDP (N=12001): FIX -34.01**; s.e 17.30; Differential effect -1.38; LAB 43.61**; s.e 19.11; Differential effect 2.91
- Omitted-variable interactions (Table 10):
  - Interacting industry characteristics with country-specific macro variables (Real GDP growth; Government expenditure/GDP; Private credit/GDP; Inflation; Uncertainty) largely preserves sign pattern; FIX and LAB remain significant in many specifications.
- Horse-race (joint inclusion of channels, Table 13, N=20786):
  - FIX: Coef -75.61***; s.e 29.36; Differential effect -3.08 (emerges as the most robust determinant when channels are jointly included).

### VI. Mechanisms: labor, capital, and productivity decomposition
- Dependent variables and measurement:
  - Employment growth, gross fixed capital formation growth (deflated by country CPI), and labor productivity (real value added per employee).
  - Labor productivity used due to TFP measurement concerns.
- Findings (Table 11):
  - Employment growth (N=20650):
    - FIX -27.47***; s.e 9.41; differential -1.12
    - LAB 36.29***; s.e 12.47; differential 2.42
    - SPEC 17.43***; s.e 5.85; differential 2.41
  - Capital growth (N=15675):
    - EFD 18.04*; s.e 10.60; differential 2.10
    - SPEC 72.92**; s.e 30.55; differential 10.09
    - Notes on measurement error: capital estimates suffer from deflation using CPI due to lack of capital-good price deflators.
  - Labor productivity growth (N=20650):
    - ISTC 3.16*; s.e 1.87; differential 2.16
    - FIX -25.72***; s.e 9.90; differential -1.05
- Main interpretation:
  - Differential effects on employment growth are larger and more statistically significant than on capital growth and labor productivity, suggesting fiscal stabilization affects industry growth mainly through employment increases.

### VII. Nonlinear regimes: recessions versus expansions
- Smooth transition estimation using contemporaneous GDP growth to define regimes (F(d_c,t) with θ = 1.5):
  - Equation: Y_i,c,t = α_i,c + γ_i,t + δ_c,t + β_L X_i F(d_c,t) FS_c,t + β_H X_i (1−F(d_c,t)) FS_c,t + ε_i,c,t
- Results (Table 12, N=20779):
  - During recessions:
    - EFD Coef 13.48**; s.e 6.98; differential 1.57
    - FIX -77.90***; s.e 21.42; differential -3.17
    - LMP 9.07***; s.e 3.72; differential 1.92
  - During expansions:
    - RND Coef -9.48*; s.e 5.76; differential -0.74
    - LAB Coef 83.35****; s.e 25.77; differential 5.55 (note: table displays stars beyond standard legend for this entry)
- Interpretation:
  - Fiscal stabilization has larger effects in recessions for industries that are more credit constrained (higher EFD, lower FIX) and with higher investment adjustment costs (LMP).
  - Some channels (e.g., labor intensity) show larger effects during expansions.

### VIII. Synthesis versus theoretical predictions
- Table 5 alignment summary (qualitative):
  - EFD: Theory +; Findings + overall, ++ in advanced economies, + in developing economies.
  - DEP: Theory +; Findings + overall and in subgroups.
  - ISTC: Theory +; Findings + overall and in developing economies but − in advanced economies.
  - RND: Theory +; Findings + overall and in advanced economies but − in developing economies.
  - FIX: Theory −; Findings consistently − across samples.
  - LAB: Theory ambiguous; Findings generally + and often significant.
  - LMP: Theory +; Findings generally + and often significant.
  - SPEC: Theory +; Findings ++ in several specifications.

### IX. Main quantitative magnitudes and sample statistics (preserved exactly)
- Country-industry-time panel: 22 manufacturing industries; 55 advanced and developing economies; period 1970-2014.
- Additional country sample for FS estimates: 69 countries with at least 23 years of fiscal stabilization estimates (1994-2016).
- Time-average fiscal stabilization coefficient: about 0.25-0.30.
- Industry classification: 22 manufacturing industries from INDSTAT2 2016, ISIC Revision 3 (original INDSTAT2 includes 23 industries; one excluded).
- Baseline sample sizes and key regression N reported in tables: N=20786 (all economies), N=11744 (advanced), N=9042 (developing), N=20650 (employment and labor productivity regressions), N=15675 (capital growth regression).

### X. Policy implications and suggested research avenues
- Policy-relevant implications directly inferred from empirical patterns:
  - Fiscal counter-cyclicality (fiscal stabilization) supports medium-term industrial value-added growth disproportionately in industries that are:
    - more dependent on external finance (EFD),
    - less asset-fixity intensive (lower FIX),
    - more labor intensive (LAB),
    - have lumpier investment (LMP), or
    - rely more on relation-specific inputs (SPEC).
  - Asset fixity (FIX) is the most robust channel: industries with higher FIX show significantly worse outcomes when fiscal stabilization increases from the 25th to the 75th percentile.
  - Effects of fiscal stabilization tend to be larger during recessions for credit-constrained industries.
  - Effects are typically larger in developing economies but more precisely estimated in advanced economies.
- Policy recommendations and research directions:
  - In addition to structural reforms, fiscal counter-cyclicality can play an important role in lifting medium-term growth by reducing aggregate volatility and supporting productive investment and employment in targeted industries.
  - Further research should investigate determinants of fiscal counter-cyclicality, which fiscal policy components deliver higher stabilization, and extend analysis to firm-level data across advanced and developing economies.

*Source: wp17198 - Chapter 2 and References, International Monetary Fund.*

### References .............................................................................................................

### wp17198 - References ....................................................................................................................

### I. INTRODUCTION
- Since the Global Financial Crisis medium-term growth has been declining in both advanced and developing economies (IMF, 2017).
- Fiscal policy has become increasingly constrained due to high debt-to-GDP ratios.
- Fiscal stabilization can enhance medium-term growth by reducing the volatility of the aggregate economy (empirical literature suggests a negative relationship between volatility and growth; see Ramey and Ramey, 1995; Martin and Rogers, 2000).
- Mechanism from Aghion et al. (2010):
  - Firms choose between short-term projects (exposed to aggregate productivity shocks) and long-term productivity-enhancing projects (subject to liquidity risk).
  - If credit constraints bind during contractions, reducing aggregate volatility increases the likelihood long-term projects survive liquidity shocks in bad states without affecting good states.
  - The higher the fraction of credit-constrained firms, the larger the positive effect of reducing aggregate volatility.
  - Predicts that countercyclical fiscal policy that reduces aggregate volatility will have larger effects on high-productive investment in more credit-constrained industries, particularly in bad times.
- Prior empirical work:
  - Aghion et al. (2014) and Furceri and Jalles (2017) find fiscal stabilization increases growth and productivity-enhancing investment in industries more credit-constrained using Rajan and Zingales (1998) difference-in-difference methodology.
  - Choi et al. (2017) show aggregate uncertainty reduces TFP growth more in industries that depend heavily on external finance.
- This paper’s extensions relative to prior literature:
  - Considers additional channels through which fiscal stabilization affects industrial growth (e.g., interactions of technological characteristics with output volatility).
  - Uses a time-varying measure of fiscal stabilization that varies over time for each country (three-dimensional country-sector-year panel) instead of a unique time-invariant parameter.
  - Provides an alternative measure based on cyclically-adjusted fiscal balance to isolate discretionary fiscal policy from automatic stabilizers.
  - Extends analysis to developing economies where fiscal policy has recently become more countercyclical.
  - Examines mechanisms by investigating effects on labor, capital, and productivity growth.
- Empirical design:
  - Applies Rajan and Zingales (1998) difference-in-difference methodology to an unbalanced panel of 22 manufacturing industries for 55 advanced and developing economies over the period 1970-2014.
  - Advantages of the three-dimensional (i industries, c countries, t time) dataset:
    - Controls for aggregate and country-sector shocks via country-time (c, t), industry-country (i, c), and industry-time (i, t) fixed effects.
    - Mitigates reverse causality concerns (industry-level outcomes less likely to drive aggregate fiscal stabilization once country-time fixed effects included).
    - Main independent variable is interaction between fiscal stabilization and industry-specific technological characteristics obtained from U.S. firm-level data, reducing plausibility of reverse causality.
- Main findings:
  - Fiscal stabilization fosters industrial growth through channels including: external financial dependence, asset fixity, labor intensity, investment lumpiness, and relationship-specific input usage.
  - Industries more dependent on external finance or with less tangible assets as collateral (or higher labor share) benefit more from fiscal stabilization.
  - Industries dependent more on specific types of investment grow faster under fiscal stabilization, suggesting fiscal stabilization reduces firms’ adjustment costs.
  - Most robust channels: asset fixity and labor intensity.
  - Effects are typically larger in developing economies but more precisely estimated in advanced economies.
  - Differential effect of fiscal stabilization tends to be larger during recessions—particularly for industries with higher credit constraints.
  - Results robust to different measures of industrial growth and fiscal stabilization and to inclusion of interactions between macroeconomic variables and industrial characteristics.
- Paper organization:
  - Section II: Channels through which fiscal stabilization can affect growth.
  - Section III: Underlying data used.
  - Section IV: Econometric methodology.
  - Section V: Main results and robustness exercises.
  - Last section: Conclusions and policy implications.

### II. FISCAL STABILIZATION AND GROWTH: CHANNELS
- General premise:
  - Fiscal stabilization’s effect on industry growth can operate through the external financial dependence (credit constraint) channel and through interactions between industry technological characteristics and output volatility.
  - “Uncertainty” and “volatility” are used interchangeably.
- Theoretical mechanisms and implications for sign of interaction with fiscal stabilization:
  - Growth options and Oi-Hartman-Abel effect:
    - If firms can costlessly rescale existing capital and labor, higher volatility can enhance growth; fiscal stabilization could reduce medium-term growth for industries with larger adjustment costs.
  - Real options theory and risk aversion channel:
    - Irreversibility of investment/hiring and precautionary wait-and-see behavior imply negative relationship between uncertainty and growth.
    - Fiscal stabilization, by reducing uncertainty, encourages timely investment/hiring and can increase growth, especially in financially constrained industries where risk premiums rise with uncertainty.
- Empirical benchmark:
  - Credit constraint (external financial dependence) channel is a natural benchmark due to existing supportive empirical evidence and relevance for both advanced and developing economies.
  - Consistency check: effects on industries with less collateral should align with external finance channel predictions.
- Industry characteristics studied (measures adopted):
  - External financial dependence (EFD)
    - Measured as median across U.S. firms in an industry of: (total capital expenditures minus current cash flow) / total capital expenditures (Rajan and Zingales, 1998; updated data by Hui Tong; Tong and Wei, 2011).
    - Expect positive interaction between EFD and time-varying fiscal stabilization.
  - Capital depreciation (DEP)
    - Higher depreciation implies lower resale values and higher liquidity risk; expect positive interaction with fiscal stabilization.
    - Also aligns with real options theory: faster depreciation makes postponing investment costlier.
    - Industry-level values from Samaniego and Sun (2016) using BEA capital flow tables (physical and economic depreciation considered).
  - Investment-specific technological change (ISTC)
    - Captures economic depreciation via technological obsolescence; measured by rate of decline in quality-adjusted price of capital goods used by each industry relative to price of consumption and services (industry-level indices from Samaniego and Sun, 2016).
    - Expect positive interaction with fiscal stabilization.
  - R&D intensity (RND)
    - R&D intensity measured as R&D expenditures / total capital expenditure using Compustat (Samaniego and Sun, 2016).
    - Interaction sign ambiguous: growth options imply negative sign; credit-constraint channel implies positive sign as uncertainty reduces R&D investment.
  - Asset fixity (FIX)
    - Measured as fixed assets / total assets (Compustat; values from Samaniego and Sun, 2016 extending Braun and Larrain, 2005; Ilyina and Samaniego, 2011).
    - Lower asset fixity (more intangible assets) implies harder to use as collateral; expect fiscal stabilization to increase growth more in industries with lower asset fixity.
  - Labor intensity (LAB)
    - Measured as total wages and salaries / total value added in the U.S. using UNIDO data (Samaniego and Sun, 2016).
    - Fiscal stabilization may be more beneficial for labor-intensive industries if credit constraints bind in recessions and learning-by-doing matters; conversely, if labor is more variable input, volatility could increase optimal investment—empirical question.
  - Input specificity: investment lumpiness (LMP) and relationship-specific input usage (SPEC)
    - Relationship-specific input usage measured by proportion of inputs not sold on organized exchange or reference-priced in trade publications (Nunn, 2007).
    - Lumpiness of investment defined as average number of investment spikes per firm per decade in each industry (Compustat; Samaniego and Sun, 2016).
    - Expect positive interactions: industries with greater relationship-specific inputs or higher adjustment costs gain more from fiscal stabilization.

### III. DATA
A. Fiscal Stabilization
- Conceptual measure:
  - Budget balance-to-GDP ratio is used as proxy for fiscal policy’s effect on aggregate demand (Blanchard, 1993).
  - Degree of fiscal stabilization (FS) inferred from responsiveness of budget balance to changes in economic activity: larger FS means higher countercyclicality/stabilization.
- Baseline regression specification:
  - Equation (1): b_i = α_i + FS_i Δy_i + ε_i
    - b is budget balance-to-GDP ratio; Δy is GDP growth or output gap; FS measures fiscal stabilization degree.
- Time-varying generalization:
  - Equation (2): b_i,t = α_i,t + FS_i,t Δy_i,t + ε_i,t
  - FS_i,t evolves slowly: FS_i,t = FS_i,t−1 + v_i,t where v_i,t ~ N(0, σ_i^2) (Equation (3)).
  - FS_i,t estimated via Varying-Coefficient model (Schlicht, 1985, 1988) using method-of-moments estimator for variances σ_i^2.
- Advantages of the Schlicht varying-coefficient approach:
  - Uses all observations in sample for each year (unlike rolling windows).
  - Changes in FS in each year come from innovations in the same year rather than neighboring years.
  - Accounts for slow policy changes that depend on immediate past.
  - Reduces reverse causality concerns since FS depends on the past.
- Empirical observations:
  - Figure 1: Average and time path of FS estimated for sample of 69 countries with at least 23 years of estimates (1994-2016).
  - Time-average fiscal stabilization coefficient is positive (about 0.25-0.30), consistent with generally counter-cyclical budget balances (Lane, 2003; Aghion and Marinescu, 2008).
  - Degree of fiscal stabilization has increased over time, particularly in advanced and emerging market economies; less change in low-income countries or oil exporters (mid-1990s level roughly same as today).
  - Greater fiscal stabilization is associated with lower output volatility on average (Figure 3 shows strong negative relationship between time-average FS and standard deviation of real GDP growth).
  - Over time, 5-year non-overlapping averages of FS and 5-year non-overlapping standard deviation of real GDP growth (both purged by country- and time-fixed effects) are negatively related (Figure 4), indicating fiscal stabilization is negatively related to output volatility across countries and over time.
  - Note: details on country-specific FS charts are available from the authors upon request.
- Footnotes and caveats:
  - Baseline analysis does not differentiate discretionary fiscal policy from automatic stabilizers; benchmark FS measure encompasses both. Robustness checks use cyclically-adjusted fiscal balance to control for automatic stabilizers (see Eichenbaum, 1997; Taylor, 2009).
  - In principle, budget balance and taxes should be adjusted by marginal propensity to consume and consider anticipated future deficits and public debt stock for full dynamic assessment (Blanchard and Summers, 1984; Blanchard, 1985).

B. UNIDO data
- Industry-level variables from UNIDO database (INDSTAT2 2016, ISIC Revision 3) to allow inclusion of developing economies and more disaggregated manufacturing industries compared to KLEMS.
- Industry growth measured by value added growth (results similar with gross output growth; robustness checks discussed).
- Also study growth in labor, capital, and productivity at industry-level.
- Nominal variables deflated by country-level CPI of local currency from World Economic Outlook.
- Sample covers 22 manufacturing industries (INDSTAT2 2016); original INDSTAT2 includes 23 but “manufacture of recycling” excluded due to insufficient observations.
- Industry classification aligned with INDSTAT3 used in earlier literature (Braun and Larrain, 2005; Ilyina and Samaniego, 2011) with minor exceptions handled by averaging where needed.

C. Industry-level characteristics
- Industry characteristics (22 manufacturing industries) constructed from U.S. firm-level data (Compustat and other sources) as detailed in Section II:
  - EFD, DEP, ISTC, RND, FIX, LAB, LMP, SPEC.
- INDSTAT2 classification similarity to INDSTAT3 noted; adjustments described (e.g., averaging ISIC 311 and ISIC 313 to obtain ISIC 16 values).

### Key statistics and sample details (explicit numeric values preserved)
- Country-industry-time panel: 22 manufacturing industries; 55 advanced and developing economies; period 1970-2014.
- Additional country sample for FS estimates: 69 countries with at least 23 years of fiscal stabilization estimates (1994-2016).
- Time-average fiscal stabilization coefficient: about 0.25-0.30.
- Industry classification: 22 manufacturing industries from INDSTAT2 2016, ISIC Revision 3 (original INDSTAT2 includes 23 industries; one excluded).

*Source: wp17198 - References. PDF: wp17198 - References.*

### Appendix compares the industry classification between INDSTAT2 and INDSTAT3.

### Appendix compares the industry classification between INDSTAT2 and INDSTAT3.

### Data and sample
- Final sample: unbalanced panel of 55 countries (21 advanced, and 34 developing countries).
- Time coverage: period 1970-2014. Advanced country samples typically start between the late 1970s and the mid-1980s; developing country samples mostly start between the late 1980s and the early 1990s.
- Industries: analysis uses 22 manufacturing industries in the empirical specification. (Footnote: There are 28 manufacturing industries in INDSTAT3.)
- U.S. exclusion: United States not included in the final sample because industry characteristics are measured using U.S. firm-level data and U.S. fiscal policies could bias results.
- Industry characteristics: measured primarily from U.S. firm-level data; correlations across measures (reported in Table 2) are intuitive (e.g., higher external finance dependence correlates with higher depreciation and lower asset fixity; higher R&D intensity correlates with lower asset fixity).

### Methodology
- Estimation equation (4) for industry i, country c, year t:
  - Y_i,c,t = α_i,c + γ_i,t + δ_c,t + β X_i FS_c,t + ε_i,c,t
  - Fixed effects: industry-country (α_i,c), industry-time (γ_i,t), and country-time (δ_c,t).
  - Y = measure of industry growth; X = industry characteristic; FS = time-varying fiscal stabilization.
- Identification strategy:
  - Three-way fixed effects control for industry-specific, global industry-time, and country-time shocks.
  - Industry characteristics from U.S. firm data mitigate reverse causality concerns for external influence.
  - OLS estimation with standard errors clustered at the industry-country level.
- Concerns addressed:
  - Potential omitted macroeconomic interactions (credit-to-GDP, inflation, uncertainty) examined in robustness checks.
  - Generated-regressor uncertainty in FS handled via Weighted Least Squares using inverse of standard deviation of FS estimates.

### Baseline results (summary)
- Interaction findings (full sample of advanced + developing economies):
  - Fiscal stabilization increases growth for industries with:
    - higher external financial dependence and lower asset fixity;
    - higher investment adjustment costs (capital depreciation, investment lumpiness, and input-specific technological change);
    - higher labor intensity.
- Magnitudes:
  - Differential growth gains from increasing fiscal stabilization from the 25th to the 75th percentile range from 0.8 to 3.0 percentage points.
  - Example: differential growth gain of 0.8 percentage point for external financial dependence channel when moving FS from the 25th to the 75th percentile for low vs high external financial dependence industries.
- Heterogeneity by country group:
  - Interaction effects typically larger for developing economies but more precisely estimated for advanced economies.
  - For advanced economies, external financial dependence and asset fixity channels are highly statistically significant when interacted with fiscal stabilization.
  - For developing economies, asset fixity channel is highly statistically significant; external financial dependence channel is not statistically significant.
  - Other robust channels across groups: labor intensity, investment lumpiness, and input specificity.
- Comparison to Aghion et al. (2014):
  - Aghion et al. find about 1.5 percentage points differential via external financial dependence; this paper finds for advanced economies a range between 0.5 and 0.9 percentage point (consistent with within-variation from fixed effects).

### Robustness checks
- Lagged specification:
  - Using a lag of the fiscal stabilization variable in the interaction term (Table 6) yields findings that hardly change.
- Alternative growth measure:
  - Repeating analysis using gross output (deflated by CPI) produces similar sign, size, and significance for interactions; capital depreciation channel loses statistical significance using gross output (Table 7).
- Uncertainty in FS estimates:
  - WLS with weights = inverse of standard deviation of FS estimates (Table 8) yields parameters similar to OLS.
- Alternative FS measures:
  - Re-estimated using primary balance-to-GDP and cyclically-adjusted balance-to-GDP (IMF FAD data). Average correlation with baseline measure: 0.74 (primary balance), 0.36 (cyclically-adjusted); some countries show negative correlations.
  - Results with these alternative FS measures (Table 9) confirm statistically significant impacts through asset fixity and labor intensity channels.
- Controls for omitted-variable bias:
  - Augmented specification (equation (5)) interacts industry characteristics with country-specific Z_c,t variables: GDP growth, government size (government expenditure to GDP), financial development (bank credit to private sector to GDP), inflation, and stock market volatility (Baker and Bloom, 2013).
  - Table 10: interaction effects for asset fixity and labor intensity remain significant. External finance channel remains significant in advanced-economy-only analysis. Loss of significance when interacting uncertainty driven by sample reduction due to limited stock market data.

### Decomposition of industry growth (mechanisms)
- Dependent variables analyzed: growth rates of industry-level employment, gross fixed capital formation, and labor productivity (real value added per employee).
- Measurement note: gross fixed capital formation deflated by country CPI; labor productivity used due to TFP measurement errors.
- Findings (Table 11):
  - Interaction signs consistent with value-added growth results.
  - Differential effects on employment growth are larger and more statistically significant than on capital growth and labor productivity.
  - Suggests fiscal stabilization affects industry growth mainly through employment increases.
  - Caveat: capital growth estimates may suffer from measurement error because of lack of capital-good price deflators; deflating by CPI is more problematic.

### Recessions vs. expansions (nonlinear regime analysis)
- Smooth transition specification (equation (6)):
  - Y_i,c,t = α_i,c + γ_i,t + δ_c,t + β_L X_i F(d_c,t) FS_c,t + β_H X_i (1−F(d_c,t)) FS_c,t + ε_i,c,t
  - F(d_c,t) = exp(−θ d_c,t) / (1 + exp(−d_c,t)), θ > 0; analysis uses θ = 1.5 following Auerbach and Gorodnichenko (2012).
  - d = contemporaneous GDP growth normalized to zero mean and unit variance; F(d) provides smooth transition between regimes.
- Results (Table 12):
  - Fiscal stabilization has larger effects on industrial growth in recessionary periods for industries that are more credit constrained (higher external financial dependence and lower asset fixity) and with higher investment adjustment costs (investment lumpiness).
  - Effects are larger during expansions for industries that are more labor intensive.

### Robust channels (joint inclusion)
- Horse-race exercise (Table 13) including all significant variables simultaneously:
  - Asset fixity emerges as the most robust determinant for both advanced and developing economies.
  - Interpreted as supporting the importance of financial constraints (ability to raise external funds) in amplifying fiscal stabilization effects.

### Conclusion and policy implications
- Summary of main findings:
  - Fiscal stabilization fosters medium-term industrial growth through multiple channels: external financial dependence and asset fixity; labor intensity; investment lumpiness; relationship-specific input usage.
  - Most robust channels: asset fixity and labor intensity.
  - Effects generally larger in developing economies but more precisely estimated in advanced economies.
  - Effects of fiscal stabilization tend to be larger during recessions, especially for credit-constrained industries.
  - Results robust to alternative growth and FS measures and inclusion of interactions with macro variables.
- Policy implications and research avenues:
  - In addition to structural reforms, fiscal counter-cyclicality can play an important role in lifting medium-term growth.
  - Future research: investigate determinants of fiscal counter-cyclicality and which fiscal policy components deliver higher stabilization; extend analysis to firm-level data across advanced and developing economies.

*Source: wp17198 - Appendix compares the industry classification between INDSTAT2 and INDSTAT3.*

### References

### References

### Fiscal policy, government size, and stabilization
- Afonso, Antonio, and Davide Furceri. “Government size, composition, volatility and economic growth.” European Journal of Political Economy 26.4 (2010): 517-532. 
- Debrun, Xavier, Jean Pisani-Ferry, and Andre Sapir. “Government Size and Output Volatility: Should We Forsake Automatic Stabilization?” IMF Working Paper (2008): 1-53. 
- Fatás, Antonio, and Ilian Mihov. “Government size and automatic stabilizers: international and intranational evidence.” Journal of International Economics 55.1 (2001): 3-28. 
- Frankel, Jeffrey A., Carlos A. Vegh, and Guillermo Vuletin. “On graduation from fiscal procyclicality.” Journal of Development Economics 100.1 (2013): 32-47 
- Furceri, D. and Jalles, J. T. (2017), “Fiscal Stabilization and Productive Investment: Evidence from Advanced Economies.”, IMF Working Paper (forthcoming) 
- Gemmell, N. (2004), “Fiscal Policy in a Growth Framework”, In: T. Addison and A. Roe, eds. Fiscal Policy for Development: Poverty, Reconstruction and Growth. Great Britain: Palgrave Macmillan, 149-176 
- Blanchard, O. J. (1993), “Suggestions for a New Set of Fiscal Indicators.” In the Political Economy of Government Debt, edited by H. A. A. Verbon and F. A. A. M. Van Winden. Amsterdam: Elsevier Science. 
- Blanchard, O. J. (1985), “Debt, Deficits and Finite Horizons”, Journal of Political Economy, 93(2), 223–47. 
- Blanchard, O. J., and L. H. Summers (1984), “Perspectives on High World Real Interest Rates.” Brookings Papers on Economic Activity (2): 273–334. 

### Uncertainty, volatility, and growth
- Aghion, P. and I. Marinescu (2008), “Cyclical Budgetary Policy and Economic Growth: What Do We Learn from OECD Panel Data?”, NBER Macroeconomics Annual, 22 
- Aghion, Philippe, George-Marios Angeletos, Abhijit Banerjee, and Kalina Manova. “Volatility and growth: Credit constraints and the composition of investment.” Journal of Monetary Economics 57, no. 3 (2010): 246-265. 
- Aghion, Philippe, David Hemous, and Enisse Kharroubi (2014), “Cyclical fiscal policy, credit constraints, and industry growth.” Journal of Monetary Economics 62, 41-58. 
- Baker, Scott R., and Nicholas Bloom (2013), “Does uncertainty reduce growth? Using disasters as natural experiments.” NBER Working Paper 
- Bloom, Nicholas (2009), “The impact of uncertainty shocks.” Econometrica 77(3), 623-685. 
- Choi, Sangyup. “The impact of US financial uncertainty shocks on emerging market economies: an international credit channel.” Working Paper, (2016). 
- Choi, Sangyup, Davide Furceri, Yi Huang, and Prakash Loungani (2017), “Aggregate Uncertainty and Sectoral Productivity Growth: The Role of Credit Constraints.” Journal of International Money and Finance (forthcoming). 
- Christiano, Lawrence J., Roberto Motto, and Massimo Rostagno. “Risk shocks.” American Economic Review 104.1 (2014): 27-65. 
- Czarnitzki, Dirk, and Andrew A. Toole (2011), “Patent protection, market uncertainty, and R&D investment.” Review of Economics and Statistics 93(1), 147-159. 
- Baker, Scott R., and Nicholas Bloom (2013), “Does uncertainty reduce growth? Using disasters as natural experiments.” NBER Working Paper

### Finance, credit constraints, and productive investment
- Caballero, Ricardo J., and Arvind Krishnamurthy (2004). “Fiscal policy and financial depth.” NBER Working Paper. 
- Braun, Matias, and Borja Larrain (2005), “Finance and the business cycle: international, inter‐industry evidence.” Journal of Finance 60(3), 1097-1128.  
- Ilyina, Anna, and Roberto Samaniego. “Technology and financial development.” Journal of Money, Credit and Banking 43.5 (2011): 899-921. 
- Hart, Oliver, and John Moore (1994), “A theory of debt based on the inalienability of human capital.” Quarterly Journal of Economics 109(4), 841-879.  
- Caballero, Ricardo J., and Arvind Krishnamurthy (2004). “Fiscal policy and financial depth.” NBER Working Paper. 

### Methodology, modelling, and theory
- Granger, Clive, and Timo Teräsvirta. “Modelling Non-Linear Economic Relationships.” Oxford University Press, 1993. 
- Eichenbaum, Martin. “Some thoughts on practical stabilization policy.” American Economic Review 87.2 (1997): 236-239. 
- Fischer, Stanley, and Franco Modigliani. “Towards an understanding of the real effects and costs of inflation.” Review of World Economics 114.4 (1978): 810-833. 
- Auerbach, Alan J., and Yuriy Gorodnichenko. “Measuring the output responses to fiscal policy.” American Economic Journal: Economic Policy 4.2 (2012): 1-27. 
- Choi, Sangyup. “The impact of US financial uncertainty shocks on emerging market economies: an international credit channel.” Working Paper, (2016). 

*Source: wp17198 - References*

### Chapter 2, April 2015, International Monetary Fund.

### wp17198 - Chapter 2, April 2015, International Monetary Fund

### Fiscal stabilization: time profile and cross-country patterns
- Figure 1: Time profile of the TVC coefficient estimates for the entire sample includes 69 countries with at least 23 observations. (Visual shows mean and percentiles from 1995 to 2016.)
- Figure 2: Interquartile and mean evolution of the TVC coefficient estimates presented for groups:
  - Advanced Economies: panel a) includes 12 countries with at least 36 observations; panel b) contains 21 countries with at least 23 observations.
  - Emerging Market Economies: panel c) contains 33 countries with at least 23 observations.
  - Low Income Countries: panel d) contains 15 countries with at least 23 observations.
- Figure 3: Displays the correlation between the average fiscal stabilization measure and the standard deviation of real GDP growth across countries.
- Figure 4: Displays the correlation between the 5-year non-overlapping average of the fiscal stabilization measure and the 5-year non-overlapping standard deviation of real GDP growth; both measures are purged by country- and time-fixed effects.

### Industry-specific characteristics (Tables 1 and 2)
- Table 1: Industry-level indicators per ISIC code include EFD, DEP, ISTC, R&D, FIX, LAB, LMP, SPEC. Example entries:
  - ISIC 15 Food products and beverages: EFD 0.11, DEP 7.09, ISTC 3.96, R&D 0.06, FIX 0.37, LAB 0.26, LMP 1.24, SPEC 0.75.
  - ISIC 30 Office, accounting and computing machinery: EFD 0.96, DEP 9.21, ISTC 4.46, R&D 1.19, FIX 0.18, LAB 0.38, LMP 2.79, SPEC 0.98.
- Table 2: Correlation matrix (note: * indicates statistical significance at the 5 percent level):
  - DEP correlated with EFD 0.61*.
  - R&D correlated with EFD 0.78* and ISTC 0.57*.
  - LMP correlated with EFD 0.76* and R&D 0.79*.
  - SPEC correlated with EFD 0.60* and DEP 0.64*.

### Main empirical findings: effect of fiscal stabilization on industry value-added growth (Table 4, baseline)
Estimates based on equation (4). T-statistics use clustered standard errors at the industry-country level. *, **, *** denote significance at 10, 5 and 1 percent, respectively. Differential effect (%) computed for an industry increasing from the 25th to the 75th percentile of the financial dependence distribution when fiscal stabilization increases from the 25th to the 75th percentile.

All economies (N=20786)
- EFD: Coef 6.80; s.e 4.67; Differential effect 0.79
- DEP: Coef 1.89*; s.e 1.14; Differential effect 2.64
- ISTC: Coef 3.39; s.e 2.93; Differential effect 2.32
- RND: Coef 1.16; s.e 4.88; Differential effect 0.09
- FIX: Coef -52.77***; s.e 14.69; Differential effect -2.15
- LAB: Coef 44.80**; s.e 19.25; Differential effect 2.98
- LMP: Coef 6.08**; s.e 2.78; Differential effect 1.29
- SPEC: Coef 22.69**; s.e 11.57; Differential effect 3.14

Advanced economies (N=11744)
- EFD: Coef 5.45**; s.e 2.63; Differential effect 0.64
- DEP: Coef 0.64; s.e 0.65; Differential effect 0.90
- ISTC: Coef -0.88; s.e 1.30; Differential effect -0.61
- RND: Coef 1.80; s.e 1.92; Differential effect 0.14
- FIX: Coef -19.46**; s.e 9.78; Differential effect -0.79
- LAB: Coef 31.90**; s.e 13.21; Differential effect 2.13
- LMP: Coef 2.77*; s.e 1.47; Differential effect 0.59
- SPEC: Coef 11.24*; s.e 6.52; Differential effect 1.56

Developing economies (N=9042)
- EFD: Coef 6.43; s.e 8.53; Differential effect 0.75
- DEP: Coef 2.73; s.e 2.05; Differential effect 3.80
- ISTC: Coef 5.98; s.e 4.91; Differential effect 4.09
- RND: Coef -1.49; s.e 8.61; Differential effect -0.12
- FIX: Coef -73.07***; s.e 25.67; Differential effect -2.97
- LAB: Coef 54.00*; s.e 32.09; Differential effect 3.59
- LMP: Coef 7.83*; s.e 4.77; Differential effect 1.68
- SPEC: Coef 34.74*; s.e 20.92; Differential effect 4.80

Key interpretation from baseline:
- Asset fixity (FIX) shows large negative interaction coefficients robustly significant, implying industries with more asset fixity fare worse when fiscal stabilization increases from the 25th to the 75th percentile.
- Labor intensity (LAB), investment lumpiness (LMP), and relation-specific investment (SPEC) commonly show positive and often statistically significant interaction coefficients.
- External financial dependence (EFD) has positive coefficients in most specifications; significance varies by sample and robustness checks.

### Robustness and alternative specifications (selected results)
- Lagged specification (Table 6, N=20786):
  - EFD: Coef 7.28*; s.e 4.42; Differential effect 0.85
  - FIX: Coef -32.59**; s.e 15.44; Differential effect -1.33
  - LAB: Coef 48.31**; s.e 19.21; Differential effect 3.22
  - SPEC: Coef 6.41; s.e 9.24; Differential effect 0.89
- Gross output vs value added (Table 7):
  - Value added (N=20786) FIX -52.77***; s.e 14.69; differential -2.15
  - Gross output (N=20816) FIX -35.23***; s.e 13.92; differential -2.35
  - SPEC: Value added Coef 22.69**; s.e 11.57; differential 3.14; Gross output Coef 20.66**; s.e 9.58; differential 2.86
- Weighted least squares (WLS) (Table 8, N=20786):
  - FIX: Coef -66.08***; s.e 17.74; Differential effect -2.69
  - SPEC: Coef 36.32**; s.e 17.52; Differential effect 5.03
- Alternative fiscal stabilization measures (Table 9):
  - Primary balance to GDP (N=19158): FIX -13.53*; s.e 8.32; Differential effect -0.55; LAB 19.38**; s.e 9.97; Differential effect 1.29
  - Cyclically-adjusted balance to GDP (N=12001): FIX -34.01**; s.e 17.30; Differential effect -1.38; LAB 43.61**; s.e 19.11; Differential effect 2.91
- Omitted variable checks (Table 10):
  - Including Real GDP growth, Government expenditure/GDP, Private credit/GDP, Inflation, or Uncertainty generally leaves the sign pattern intact; FIX remains strongly negative and LAB strongly positive in many specifications.
- Labor, capital, and productivity channels (Table 11):
  - Employment growth (N=20650): FIX -27.47***; s.e 9.41; differential -1.12; LAB 36.29***; s.e 12.47; differential 2.42; SPEC 17.43***; s.e 5.85; differential 2.41
  - Capital growth (N=15675): EFD 18.04*; s.e 10.60; differential 2.10; FIX -52.32; s.e 41.22; differential -2.13; SPEC 72.92**; s.e 30.55; differential 10.09
  - Labor productivity growth (N=20650): ISTC 3.16*; s.e 1.87; differential 2.16; FIX -25.72***; s.e 9.90; differential -1.05
- Recessions vs expansions (Table 12, N=20779):
  - During recessions: EFD Coef 13.48**; s.e 6.98; differential 1.57; FIX -77.90***; s.e 21.42; differential -3.17; LMP 9.07***; s.e 3.72; differential 1.92
  - During expansions: RND Coef -9.48*; s.e 5.76; differential -0.74; FIX -19.83; s.e 18.96; differential -0.81; LAB Coef 83.35****; s.e 25.77; differential 5.55 (note: table displays stars beyond standard legend for this entry)
- Horse race (Table 13, all channels jointly, N=20786):
  - EFD Coef 5.99; s.e 8.34; differential 0.70
  - FIX Coef -75.61***; s.e 29.36; differential -3.08
  - LAB Coef 31.45; s.e 21.19; differential 2.09
  - LMP Coef -8.55; s.e 7.18; differential -1.81
  - SPEC Coef 12.25; s.e 14.72; differential 1.70

### Synthesis of findings versus theoretical expectations (Table 5)
- Summary of alignment between theory and empirical findings:
  - EFD (external financial dependence): Theory +; Findings show + in full sample, ++ in advanced economies, + in developing economies.
  - DEP (capital depreciation): Theory +; Findings show + overall and in subgroups.
  - ISTC (investment specific technological change): Theory +; Findings show + overall and in developing economies but – in advanced economies.
  - RND (R&D intensity): Theory +; Findings show + overall and in advanced economies but – in developing economies.
  - FIX (asset fixity): Theory –; Findings consistently – across samples.
  - LAB (labor intensity): Theory ambiguous (+ or –); Findings generally + and often significant.
  - LMP (investment lumpiness): Theory +; Findings generally + and often significant.
  - SPEC (relation-specific investment): Theory +; Findings ++ in several specifications.

### Policy-relevant implications (inferred directly from empirical patterns)
- Fiscal stabilization interacts heterogeneously with industry characteristics:
  - Policies that promote fiscal stabilization are associated with stronger value-added growth in industries that are labor-intensive (LAB), have high investment lumpiness (LMP), or high relation-specific investment (SPEC).
  - Industries with high asset fixity (FIX) experience significantly worse outcomes under greater fiscal stabilization; caution is warranted for sectors with high asset fixity when designing stabilization policies.
  - External financial dependence (EFD) generally enhances the positive effect of fiscal stabilization on industry growth, notably during recessions.
- Robustness checks and alternative measures support the persistence of the main patterns across specifications (lagged effects, gross output, WLS, alternative fiscal metrics, and controlling for macro variables).

*Source: Chapter 2, April 2015, International Monetary Fund (wp17198).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2017/wp17198.pdf_