## wpiea2021118-print-pdf

## Source details

**Canonical URL:** [wpiea2021118-print-pdf](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021118-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021118-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021118-print-pdf.pdf.json)

---

### Key findings and abstract-level results
- Positive shocks to ESI Funds are followed by an increase in output ranging from 1.2 percent on impact to 1.8 percent after 1 year.
- A 1 percent increase in ESI Funds’ investments increases private investment by around 0.7-0.8 percent of GDP (EU average).
- For the CEE group that includes Slovenia:
  - A 1 percent of GDP positive shock in ESI Funds increases private investment by 1.2-1.3 percent of GDP on impact and after 1 year.
  - It increases GDP by 1.3 percent of GDP on impact and by 1.6 percent after 1 year.
- Aggregate ESI Funds → investment elasticities (EU average):
  - When ESI funds disbursements increase by 1 percent of GDP:
    - GDP increases by 1.2 percent on impact and cumulatively by 1.8 percent after 1 year.
    - Total investment increases by 1.5 percent of GDP on impact and cumulatively by 1.7 percent of GDP after 1 year.
    - Private investment increases by 0.8 percent of GDP on impact and cumulatively by 0.7 percent of GDP after 1 year.
    - Employment increases by 0.1 percent on impact and by 0.1 percent after 1 year (coefficients statistically insignificant).

### Methodology and identification
- Data:
  - Panel of 28 countries (27 EU plus the UK), annual data approximately from 1994 to 2018.
- Main econometric framework:
  - Baseline specification (Eq. 1): ΔY_{i,t+h} / GDP_{i,t-1} = α_i + τ_t + β ΔX_{i,t} / GDP_{i,t-1} + δ′V_{i,t} + ε_{i,t+h}
  - Regional interaction specification (Eq. 4): adds γ (ΔX/GDP × D_i^c); region-specific multiplier inferred as (β + γ).
- Identification strategy:
  - Instrumental Variables (2SLS) following Kraay (2014): instrument actual ESI disbursements with a predicted disbursement series constructed from program and recipient-region characteristics (predicted disbursements are by construction exogenous to domestic shocks).
  - Predicted disbursement series constructed distinguishing ESI disbursements by program and status of recipient region; technical steps in Appendix B.1.
- Controls and estimation notes:
  - Country and time fixed effects included; p-values computed using robust standard errors clustered at the country level.
  - Variables expressed in constant euros using overall GDP deflator.
  - Employment dependent variable uses ΔY_{i,t+h} / Y_{i,t-1} so estimated β captures percentage change in employment associated with a 1 percent of GDP ESI disbursement increase.

### Aggregate multipliers (EU average and econometric results)
- 2SLS estimates (instrument: predicted ESI disbursement profile):
  - GDP: impact multiplier = 1.2 percent; 1-year cumulative = 1.8 percent (significant at 1% and 5% respectively).
  - Total investment: impact = 1.5 percent of GDP; 1-year = 1.7 percent of GDP (highly significant).
  - Private investment: impact = 0.8 percent of GDP; 1-year = 0.7 percent of GDP (highly significant).
  - Employment: impact = 0.1 percent; 1-year = 0.1 percent (statistically insignificant).
- OLS estimates show downward bias relative to 2SLS, possibly due to countercyclicality of ESI payments.
- Headline 2SLS coefficients reported (Table 4, Panel 1):
  - Impact Multiplier (2SLS) for GDP: 1.207*** (0.425)
  - Impact Multiplier (2SLS) for Total Investment: 1.507*** (0.339)
  - Impact Multiplier (2SLS) for Private Investment: 0.790** (0.361)
  - Impact Multiplier (2SLS) for Employment: 0.106 (0.216)
- 1-year 2SLS (Table 4, Panel 2):
  - GDP: 1.765** (0.693)
  - Total Investment: 1.688*** (0.418)
  - Private Investment: 0.724** (0.355)
  - Employment: 0.0598 (0.459)

### Sectoral multipliers and crowding-in effects
- Sectoral outcomes analyzed: Gross Value Added (GVA), total investment, and employment at NACE Rev.2 sectoral breakdown; instrumented with predicted ESI disbursement series in 2SLS.
- Contemporaneous significant sectoral total-investment crowding-in (selected):
  - Manufacturing (C): investment increases by 0.3 percent of GDP on impact.
  - Utilities (D and E): strong contemporaneous effects.
  - Professional, Scientific and Technical Activities (M): significant.
  - Administrative and support services (N): significant contemporaneous effect.
  - Public Administration and Defense (O): investment increases by 0.4 percent of GDP on impact.
  - Education (P): investment increases by 0.1 percent of GDP on impact.
- Aggregate from sectoral sum:
  - Contemporaneous sum of sectoral total-investment multipliers = 1.1.
  - One-year sum = 1.6 (consistent with aggregate estimates).
- Semi-elasticities (impact relative to sector average investment; Table 6 Panel 1 examples):
  - Manufacturing: contemporaneous increase equals 8.4 percent of average manufacturing investment for a 1 percent of GDP ESI shock.
  - Electricity investments: 21.0 percent of average electricity investment.
  - Public Administration and Defense investments: 26.4 percent of average investments in that sector.
- Sectoral GVA multipliers (selected):
  - Agriculture (A): contemporaneous multiplier = 0.1 and 1-year = 0.2 (both significant at 10%).
  - Manufacturing: contemporaneous GVA multiplier = 0.2 and 1-year = 0.4 (both significant at 10%).
  - Construction (F): contemporaneous = 0.4 and 1-year = 0.5 (both significant at 1%).
  - Wholesale and Retail trade (G): impact multiplier = 0.2 (significant at 10%).
  - Sum of sectoral GVA multipliers = 0.9 on impact and 1.0 after 1 year.
- Sectoral employment:
  - Generally statistically insignificant at aggregate sectoral breakdowns.
  - Exceptions: sectors K-N show contemporaneous employment increase = 1.3 percent and 1-year = 2.8 percent (both significant at 5%).
  - Sector O-U shows contemporaneous negative impact = -0.6 percent (10% significance).

### Multipliers by type of public investment (COFOG classification)
- Public investment share by sector (average EU country):
  - Economic Affairs: 33 percent
  - General Public Services: 14 percent
  - Education: 12 percent
  - Defense: 10 percent
  - Health: 8 percent
  - Environment: 6 percent
  - Housing: 6 percent
  - Culture: 5 percent
  - Public Order: 3 percent
  - Social Protection: 2 percent
- Estimation caveat: OLS used (no IV) for public investment by COFOG area; results interpreted with caution.
- GDP responses to 1 percent of GDP public investment shock by COFOG area (selected contemporaneous multipliers, Table 8a Panel 1):
  - Public Order: GDP increases by 4.0 percent on impact (largest effect).
  - Recreation: GDP increases by 2.8 percent on impact (insignificant after 1 year).
  - Environmental protection: GDP increases by 2.7 percent on impact.
  - Education: GDP increases by 1.9 percent on impact.
  - Defense: GDP increases by 1.1 percent on impact.
  - Public services: GDP increases by 1.1 percent on impact.
  - Aggregated contemporaneous sectoral multiplier (Eq. 3) = 1.3; 1-year aggregated = 2.0.
- Aggregate investment responses to public investment shocks:
  - Contemporaneous aggregate investment multiplier = 2.4 and 1-year = 2.6.
  - Private investment aggregated contemporaneous multiplier = 0.3 and 1-year = 0.8.
- Employment responses to public investment:
  - Significant positive reactions for public services, housing, health and recreation.
- Policy-relevant note:
  - Environmental protection and health (research and health R&D) show strong multipliers for GDP, investment and employment.

### Regional heterogeneity: CEE focus and Slovenia
- Motivation: heterogeneity driven by exchange rate regime, trade openness, labor market rigidities, public debt, and public investment governance.
- ESI relevance: ESI funds amount to 38 percent of public investments in CEE vs. 11 percent in Western Europe.
- Regional econometric interaction: adds interaction term to baseline to estimate region-specific multipliers; region chosen: Slovenia, Croatia, Czech Republic, Estonia and Latvia.
- Aggregate multipliers in CEE sub-sample (Table 9 summary):
  - GDP impact multiplier = 1.3; 1-year multiplier = 1.6.
  - Private investment response: 1.3 percent of GDP on impact for a 1 percent of GDP ESI shock (vs. 0.8 for EU average).
  - Employment: impact = 0.1 percent; 1-year = 0.2 percent (vs. 0.1 percent EU average).
  - Sum of sectoral crowding-in coefficients: aggregate impact = 1.4 of GDP; 1-year aggregate = 2.6 of GDP (CEE).
- Sectoral multipliers in CEE sub-sample (selected, Table 10 Panel 1):
  - Utilities (D): investment increases by 0.2 percent of GDP on impact and by 0.9 percent of GDP after 1 year (EU average 0.1 and 0.3 respectively).
  - Real estate activities (L): investment increases by 0.3 percent of GDP on impact and by 0.4 percent after 1 year.
  - Public administration (O): investment increases by 0.5 percent of GDP on impact and by 0.4 percent after 1 year.
  - Education (P): investment increases by 0.2 percent of GDP on impact and after 1 year (EU average 0.1 for both).
- Comparison to prior literature on Slovenia and CEE:
  - Appendix C bucket-approach application to Slovenia suggests a multiplier range 0.5 to 1.2; accounting fully for Covid-19 output gap could raise the Slovenian multiplier range to between 0.8 and 2.1.

### Robustness, limitations and caveats
- IV strategy aims to address endogeneity; predicted disbursement profiles are constructed excluding the loan/operation in question to ensure exogeneity.
- For public investment by COFOG area, OLS is used (no instrument); results may be biased.
- Sectoral employment data limited in disaggregation (consistent NACE Rev.2 employment available from 2008 onward).
- Outlier treatment:
  - Judgmental exclusion of specific austerity years for certain countries (e.g., Cyprus 2012–2014; Portugal 2011; Spain 2011; Ireland 2011, 2015, 2016; Finland 2015, 2016).
  - Automated exclusion of standardized residuals exceeding 2.5 standard deviations.
- Point estimates subject to statistical uncertainty; significance notation: ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.
- Aggregation of sectoral multipliers via summation does not account for inter-sectoral spillovers.

### Robustness checks (selected numerical results)
- Table 14 (Aggregate multipliers robustness, 2SLS):
  - Impact Multipliers:
    - ESI Funds → GDP: 1.293*** (0.474)
    - ESI Funds → Tot. Inv.: 1.403*** (0.338)
    - ESI Funds → Priv. Inv: 0.680* (0.356)
    - ESI Funds → Empl.: 0.0393 (0.212)
  - 1Y Multipliers:
    - GDP: 1.607** (0.652)
    - Tot. Inv.: 1.689*** (0.430)
    - Priv. Inv: 0.754** (0.383)
    - Empl.: 0.0390 (0.413)
- Table 15 (CEE sub-sample robustness, 2SLS):
  - Impact Multipliers:
    - ESI Funds → GDP: 1.332*** (0.469)
    - ESI Funds → Tot. Inv.: 1.729*** (0.562)
    - ESI Funds → Priv. Inv: 1.078* (0.553)
    - ESI Funds → Empl.: 0.108 (0.195)
  - 1Y Multipliers:
    - GDP: 1.505** (0.760)
    - Tot. Inv.: 2.086*** (0.638)
    - Priv. Inv: 1.326** (0.548)
    - Empl.: 0.128 (0.783)

### Policy implications emphasized by the analysis
- ESI Funds and public investment have meaningful positive macroeconomic effects on GDP and investment, and can crowd-in private investment—especially in labor-intensive sectors.
- Targeting public investment toward environmental protection and health (including R&D) yields large multipliers for GDP, investment and employment, supporting the allocation priorities of the European Budget toward climate and health.
- Country and regional heterogeneity matter: in CEE countries (including Slovenia), estimated multipliers are larger for private investment and total investment than EU averages.
- Fiscal stimulus design should account for sectoral composition: labor-intensive sectors (manufacturing, education, public services) generate stronger responses in GVA and private investment.
- When using multipliers for macro forecasting and policy design, region-specific multipliers and institutional characteristics (exchange rate regime, public investment governance, labor market rigidity, trade openness) should be considered.

### Appendix D — Bucket approach applied to Slovenia (selected quantitative outcomes)
- Structural characteristics scores for Slovenia (Table17): total Slovenia score = 4.
- Mapping scores to multiplier ranges (Table16):
  - Low Multiplier: 0.1–0.3
  - Medium Multiplier: 0.4–0.6
  - High Multiplier: 0.7–1.0
- Adjustments applied:
  - Business cycle interpolation (Slovenia 2019 output gap: 1.8 percent of potential GDP).
  - Monetary policy adjustment: Slovenia in Euro Area → increase range by 30 percent.
- Resulting ranges for Slovenia (normal times):
  - Medium multiplier final range: M_medium = [0.47,0.71].
  - High multiplier final range: M_high = [0.83,1.18].
- COVID-19 adjustment (economy at lowest point; Slovenia GDP decline 2020: 5.5 percent):
  - M_medium_Covid = [0.83,1.25]
  - M_high_Covid = [1.46,2.08]

*Source — Selected content from IMF Working Paper (wpiea2021118-print-pdf).*

### 0.7 and 0.8 percent of GDP. We address country heterogeneity by dividing countries according to

### wpiea2021118-print-pdf - 0.7 and 0.8 percent of GDP. We address country heterogeneity by dividing countries according to

### Key findings and abstract-level results
- Positive shocks to ESI Funds are followed by an increase in output ranging from 1.2 percent on impact to 1.8 percent after 1 year.
- A 1 percent increase in ESI Funds’ investments increases private investment by around 0.7-0.8 percent of GDP (EU average).
- For the CEE group that includes Slovenia:
  - A 1 percent of GDP positive shock in ESI Funds increases private investment by 1.2-1.3 percent of GDP on impact and after 1 year.
  - It increases GDP by 1.3 percent of GDP on impact and by 1.6 percent after 1 year.
- Aggregate ESI Funds → investment elasticities (EU average):
  - When ESI funds disbursements increase by 1 percent of GDP:
    - GDP increases by 1.2 percent on impact and cumulatively by 1.8 percent after 1 year.
    - Total investment increases by 1.5 percent of GDP on impact and cumulatively by 1.7 percent of GDP after 1 year.
    - Private investment increases by 0.8 percent of GDP on impact and cumulatively by 0.7 percent of GDP after 1 year.
    - Employment increases by 0.1 percent on impact and by 0.1 percent after 1 year (coefficients statistically insignificant).

### Methodology and identification
- Data: Panel of 28 countries (27 EU plus the UK), annual data approximately from 1994 to 2018.
- Main econometric framework:
  - Baseline specification (Eq. 1): ΔY_{i,t+h} / GDP_{i,t-1} = α_i + τ_t + β ΔX_{i,t} / GDP_{i,t-1} + δ′V_{i,t} + ε_{i,t+h}
  - Regional interaction specification (Eq. 4): adds γ (ΔX/GDP × D_i^c) to allow region-specific multipliers; region-specific multiplier inferred as (β + γ).
- Identification strategy:
  - Instrumental Variables (2SLS) approach following Kraay (2014) logic: instrument actual ESI disbursements with a predicted disbursement series constructed from program and recipient-region characteristics (predicted disbursements are by construction exogenous to domestic shocks).
  - Predicted disbursement series constructed distinguishing ESI disbursements by program and status of recipient region; technical steps in Appendix B.1.
- Controls and estimation notes:
  - Country and time fixed effects included; p-values computed using robust standard errors clustered at the country level.
  - Variables expressed in constant euros using overall GDP deflator.
  - Employment dependent variable uses ΔY_{i,t+h} / Y_{i,t-1} so estimated β captures percentage change in employment associated with a 1 percent of GDP ESI disbursement increase.

### Aggregate multipliers (EU average and econometric results)
- 2SLS estimates (instrument: predicted ESI disbursement profile):
  - GDP: impact multiplier = 1.2 percent; 1-year cumulative = 1.8 percent (significant at 1% and 5% respectively).
  - Total investment: impact = 1.5 percent of GDP; 1-year = 1.7 percent of GDP (highly significant).
  - Private investment: impact = 0.8 percent of GDP; 1-year = 0.7 percent of GDP (highly significant).
  - Employment: impact = 0.1 percent; 1-year = 0.1 percent (statistically insignificant).
- OLS estimates show downward bias relative to 2SLS, possibly due to countercyclicality of ESI payments.

### Sectoral multipliers and crowding-in effects
- Approach:
  - Sectoral outcomes: Gross Value Added (GVA), total investment, and employment at NACE Rev.2 sectoral breakdown.
  - Instrument: predicted ESI disbursement series used in 2SLS for sectoral estimations.
  - Aggregate check: sum of sectoral multipliers (Eq. 2) used to reconstruct aggregate multiplier.
- Crowding-in of total investment (selected significant contemporaneous sectoral multipliers):
  - Manufacturing (C): investment increases by 0.3 percent of GDP on impact.
  - Utilities (D and E): strong contemporaneous effects.
  - Professional, Scientific and Technical Activities (M): significant.
  - Administrative and support services (N): significant contemporaneous effect.
  - Public Administration and Defense (O): investment increases by 0.4 percent of GDP on impact.
  - Education (P): investment increases by 0.1 percent of GDP on impact.
- Aggregate from sectoral sum:
  - Contemporaneous sum of sectoral total-investment multipliers = 1.1.
  - One-year sum = 1.6 (consistent with aggregate estimates).
- Semi-elasticities (impact relative to sector average investment; Table 6 Panel 1 examples):
  - Manufacturing: contemporaneous increase equals 8.4 percent of average manufacturing investment for a 1 percent of GDP ESI shock.
  - Electricity investments: 21.0 percent of average electricity investment.
  - Public Administration and Defense investments: 26.4 percent of average investments in that sector.
  - Water supply and Education also show significant relative increases despite lower absolute multipliers.
- Relationship with labor intensity:
  - Sectors that are more labor intensive exhibit larger responses in GVA and private investment; scatterplots indicate labor-intensity correlates with multiplier magnitude.
- Sectoral GVA multipliers (selected):
  - Agriculture (A): contemporaneous multiplier = 0.1 and 1-year = 0.2 (both significant at 10%).
  - Manufacturing: contemporaneous GVA multiplier = 0.2 and 1-year = 0.4 (both significant at 10%).
  - Construction (F): contemporaneous = 0.4 and 1-year = 0.5 (both significant at 1%).
  - Wholesale and Retail trade (G): impact multiplier = 0.2 (significant at 10%).
  - Sum of sectoral GVA multipliers = 0.9 on impact and 1.0 after 1 year.
- Sectoral employment:
  - Generally statistically insignificant responses at aggregate sectoral breakdowns.
  - Exceptions: sectors K-N (Financial and insurance; Real estate; Professional, scientific, technical; Administration and support) show contemporaneous employment increase = 1.3 percent and 1-year = 2.8 percent (both significant at 5%).
  - Sector O-U (Public administration, defense, education, human health, other services) shows contemporaneous negative impact = -0.6 percent (10% significance).

### Multipliers by type of public investment (COFOG classification)
- Public investment share by sector (average EU country):
  - Economic Affairs: 33 percent
  - General Public Services: 14 percent
  - Education: 12 percent
  - Defense: 10 percent
  - Health: 8 percent
  - Environment: 6 percent
  - Housing: 6 percent
  - Culture: 5 percent
  - Public Order: 3 percent
  - Social Protection: 2 percent
- Estimation caveat: OLS used (no IV) for public investment by COFOG area; results interpreted with caution.
- GDP responses to 1 percent of GDP public investment shock by COFOG area (selected contemporaneous multipliers, Table 8a Panel 1):
  - Public Order: GDP increases by 4.0 percent on impact (largest effect).
  - Recreation: GDP increases by 2.8 percent on impact (insignificant after 1 year).
  - Environmental protection: GDP increases by 2.7 percent on impact.
  - Education: GDP increases by 1.9 percent on impact.
  - Defense: GDP increases by 1.1 percent on impact.
  - Public services: GDP increases by 1.1 percent on impact.
  - Aggregated contemporaneous sectoral multiplier (Eq. 3) = 1.3; 1-year aggregated = 2.0.
- Aggregate investment responses to public investment shocks:
  - Broadly large and significant multipliers across most COFOG sectors; contemporaneous aggregate investment multiplier = 2.4 and 1-year = 2.6.
  - Private investment responses significant mainly for Public Services, Defense and Education; aggregated contemporaneous private-investment multiplier = 0.3 and 1-year = 0.8.
- Employment responses to public investment:
  - Significant positive reactions for public services, housing, health and recreation.
- Policy-relevant note:
  - Environmental protection and health (research and health R&D) show strong multipliers for GDP, investment and employment, highlighting macroeconomic benefits of EU funding targeted at climate and health research areas.

### Regional heterogeneity: CEE focus and Slovenia
- Motivation: cross-country heterogeneity driven by exchange rate regime, trade openness, labor market rigidities, public debt, and public investment governance (Batini, Eyraud and Weber bucket approach).
- ESI relevance: Central and Eastern European (CEE) countries rely more on ESI Funds on average (ESI funds amount to 38 percent of public investments in CEE vs. 11 percent in Western Europe).
- Regional econometric interaction:
  - Specification adds interaction term to baseline model to estimate region-specific multipliers; region chosen: Slovenia, Croatia, Czech Republic, Estonia and Latvia.
  - Regional multiplier computed as (β + γ); standard errors via Delta Method.
- Aggregate multipliers in CEE sub-sample (Table 9, summary):
  - GDP impact multiplier = 1.3; 1-year multiplier = 1.6.
  - Private investment response: 1.3 percent of GDP on impact for a 1 percent of GDP ESI shock (vs. 0.8 for EU average).
  - Employment: impact = 0.1 percent (similar to EU average); 1-year = 0.2 percent (vs. 0.1 percent EU average).
  - Sum of sectoral crowding-in coefficients: aggregate impact = 1.4 of GDP; 1-year aggregate = 2.6 of GDP (CEE).
- Sectoral multipliers in CEE sub-sample (selected, Table 10 Panel 1):
  - Utilities (D): investment increases by 0.2 percent of GDP on impact and by 0.9 percent of GDP after 1 year (EU average 0.1 and 0.3 respectively).
  - Real estate activities (L): investment increases by 0.3 percent of GDP on impact and by 0.4 percent after 1 year (EU average insignificant or lower).
  - Public administration (O): investment increases by 0.5 percent of GDP on impact and by 0.4 percent after 1 year (EU average 0.4 for both horizons).
  - Education (P): investment increases by 0.2 percent of GDP on impact and after 1 year (EU average 0.1 for both).
- Comparison to prior literature on Slovenia and CEE:
  - SVAR-based estimates for Slovenia in previous literature often find small or insignificant multipliers (e.g., multipliers in range 0.1-0.4), while this study’s CEE-focused estimates are larger and align more with model-based predictions that allow for higher multipliers under fixed-exchange or limited openness conditions.
  - Appendix C provides a bucket-approach application to Slovenia suggesting a multiplier range 0.5 to 1.2; accounting fully for Covid-19 output gap could raise the Slovenian multiplier range to between 0.8 and 2.1.

### Robustness, limitations and caveats
- IV strategy aims to address endogeneity of ESI disbursements; predicted disbursement profiles are constructed excluding the loan/operation in question to ensure exogeneity.
- For public investment by COFOG area, OLS is used (no instrument); results may be biased and should be interpreted cautiously.
- Sectoral employment data is limited in disaggregation (consistent NACE Rev.2 employment available from 2008 onward); some sectoral employment results are based on partial breakdowns.
- Point estimates are subject to statistical uncertainty; significance assessed at conventional p-values (∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01).
- Aggregation of sectoral multipliers via summation does not account for inter-sectoral spillovers arising from shocks in one sector affecting public investment in others.

### Policy implications emphasized by the analysis
- ESI Funds and public investment have meaningful positive macroeconomic effects on GDP and investment, and can crowd-in private investment—especially in labor-intensive sectors.
- Targeting public investment toward environmental protection and health (including R&D) yields large multipliers for GDP, investment and employment, supporting the allocation priorities of the European Budget toward climate and health.
- Country and regional heterogeneity matter: in CEE countries (including Slovenia), estimated multipliers are larger for private investment and total investment than EU averages, implying stronger macro-relevance of ESI disbursements in those economies.
- Fiscal stimulus design should account for sectoral composition: labor-intensive sectors (manufacturing, education, public services) generate stronger responses in GVA and private investment.
- When using multipliers for macro forecasting and policy design, region-specific multipliers and institutional characteristics (exchange rate regime, public investment governance, labor market rigidity, trade openness) should be considered.

*Italic: Source — Selected content from IMF Working Paper (wpiea2021118-print-pdf).*

### 0.1 for the EU average), a contemporaneous multiplier equal to 0.3 for manufacturing (compared to

### wpiea2021118-print-pdf - 0.1 for the EU average), a contemporaneous multiplier equal to 0.3 for manufacturing (compared to

### Key finding
- A contemporaneous multiplier equal to 0.3 for manufacturing is reported.
- A value of 0.1 for the EU average is referenced.

### Contextual comparison
- Manufacturing contemporaneous multiplier: 0.3
- EU average contemporaneous multiplier: 0.1

*Source: wpiea2021118-print-pdf - 0.1 for the EU average), a contemporaneous multiplier equal to 0.3 for manufacturing (compared to*

### 0.2 for the EU average) and a contemporaneous and 1-year multipliers equal to 0.4 for construction

### wpiea2021118-print-pdf - 0.2 for the EU average) and a contemporaneous and 1-year multipliers equal to 0.4 for construction

### Sectoral multipliers (EU average and signs)
- Construction (F): contemporaneous multiplier equal to 0.4 on impact and 0.5 after 1 year for the EU average; in one analysis reported a contemporaneous and 1-year multipliers equal to 0.4 for construction (other line: 0.2 for the EU average).
- Retail (G category): contemporaneous multiplier of 0.1 (compared to 0.2 for the EU average).
- Health care (Q): negative multiplier in one specification, with GVA falling by only -0.05 percent contemporaneously (the multiplier is insignificant for the EU average).
- Other sector coefficients: insignificant, consistent with results found for the EU.
- Aggregate sector sum: aggregate impact multiplier equal to 1.0, and 1-year aggregate multiplier equal to 0.7.

### Employment effects (sectoral patterns)
- Agricultural sector (A): employment negatively affected, -3.8 percent.
- O-U activities (public administration, defense, education, human health and social work activities, other services): employment -0.9 percent.
- Construction (F): strong positive effect of ESI Funds on employment, 2.8 percent on impact, and 6.6 percent after 1 year.
- K-N sectors (finance, real estate, professional and administrative services): contemporaneous increase in employment of 2.3 percent and 4.6 percent after 1 year.
- Regional sensitivity: the sensitivity of employment in these sectors is much stronger in the CEE sub-region than in the EU.

### Multipliers by type of public investment (Slovenia and CEE sub-region)
- GDP (Panel 1, CEE sub-sample, selected COFOG categories):
  - Public services: contemporaneous increase of 5.6 percent, 1-year increase of 6.7 percent.
  - Public order: contemporaneous increase of 10.2 percent, 1-year increase of 15.5 percent.
  - Housing: contemporaneous increase of 4.2 percent, 1-year increase of 6.8 percent (albeit insignificant).
  - Health: contemporaneous multiplier for GDP equal to 6.1 (EU average health GDP multiplier estimated to be insignificant).
  - Education: contemporaneous multiplier for GDP equal to 2.5.
  - Environmental protection and social protection: no statistically significant effects on GDP found.
- Total investments (Panel 2, CEE sub-sample, impact crowding-in effects):
  - Public services: impact multiplier 5.7 (tends to increase after 1 year).
  - Public order: impact multiplier 9.8 (tends to increase after 1 year).
  - Housing: impact multiplier 5.6 (tends to increase after 1 year).
  - Environmental protection and social protection: no statistically significant effects.
  - Recreation, culture and religion: no statistically significant effects.
- Private investments (Panel 3, CEE sub-sample):
  - Private investments generally do not react significantly following shocks to public investment, with exceptions for public services, public order, housing and education.
- Employment (Panel 4, CEE sub-sample):
  - Significant impacts for public investments in public services, public order, health and recreation (similar to EU sample).
- Aggregated weighted multipliers (weights = share of public investment expenditures in each category, CEE sub-sample reported for Slovenia and the region):
  - GDP aggregate contemporaneous multiplier equal to 2.7 and 1-year equal to 3.1.
  - Total investments aggregate contemporaneous multiplier equal to 3.2 and 1-year equal to 3.6.
  - Private investments aggregate contemporaneous multiplier equal to 1.1 and 1-year equal to 1.5.

### Detailed subsectors (selected findings)
- General Public Services (adjusted: Basic Research, R&D General public services, General public services n.e.c.):
  - Point estimates mostly consistent with Table 11 but tend to display lower significance.
- R&D (aggregated across all sectors):
  - No significant multipliers found for aggregated R&D.
- Basic Research (sub-sector of General Public Services):
  - Positive effects on both GDP and private investment detected.
- Economic Affairs subcategories:
  - Fuel and Energy and Transport: significant multipliers for GDP and total investments.
- Water supply (sub-sector of Housing and community amenities):
  - Multipliers insignificant for all variables.
- Employment for these detailed subsectors:
  - No statistically significant results found.

### Aggregate results (EU-wide; headline figures)
- Using 2SLS (Table 4, Panel 1): a 1 percent of GDP increase in ESI disbursements increases GDP contemporaneously by 1.207 percent on average.
  - Impact Multiplier (2SLS) for GDP: 1.207*** (standard error (0.425))
  - Impact Multiplier (2SLS) for Total Investment: 1.507*** (0.339)
  - Impact Multiplier (2SLS) for Private Investment: 0.790** (0.361)
  - Impact Multiplier (2SLS) for Employment: 0.106 (0.216)
- 1-year multipliers (2SLS, Table 4, Panel 2):
  - GDP: 1.765** (0.693)
  - Total Investment: 1.688*** (0.418)
  - Private Investment: 0.724** (0.355)
  - Employment: 0.0598 (0.459)
- OLS comparisons (Table 4) also reported (examples preserved exactly):
  - Impact Multiplier (OLS) for GDP: 0.877*** (0.192)
  - 1Y Multiplier (OLS) for GDP: 1.478*** (0.302)
- CEE sub-sample aggregate multipliers (Table 9):
  - Impact M. (2SLS) GDP: 1.257*** (0.450)
  - Impact M. (2SLS) Tot. Inv.: 1.809*** (0.561)
  - Impact M. (2SLS) Priv. Inv.: 1.258** (0.601)
  - 1Y M. (2SLS) GDP: 1.629** (0.797)
  - 1Y M. (2SLS) Tot. Inv.: 2.058*** (0.659)
  - 1Y M. (2SLS) Priv. Inv.: 1.247** (0.529)

### Methodological and implementation notes (identification and ESI features)
- Identification strategy: compute an exogenous (“predicted”) series of ESI Funds disbursements from 1994 to 2018 to estimate aggregate and sectoral EU-wide multipliers.
- ESI Funds overview (Appendix A):
  - ESI disbursed through five funds: ERDF, ESF, CF, EAFRD, EMFF, and also Youth Employment Initiative (YEI).
  - Programming periods: 1989-1993, 1994-1999, 2000-2006, 2007-2013, 2014-2020.
  - For the 2014-2020 program, funds totaled approximately EUR450 billion.
  - Distinction between commitments and payments emphasized; decommitment rules (“N+2” and “N+3”) reduce delays.
- Instrumental variable approach: predicted disbursement series computed at country level following Kraay (2014) (identification steps summarized in Appendix B).

### Policy implications and conclusion
- Overall findings: large and significant effects of ESI Funds disbursements and public investment on GDP, total investment and private investment.
- Sectoral heterogeneity: largest crowding-in effects in labor intensive industries such as construction, defense, manufacturing, public order and education.
- Targeting implications: channeling public investments in research, public services and public order has significant implications for macroeconomic aggregates.
- Cross-country heterogeneity: regional (CEE) estimates reveal larger fiscal multipliers compared to EU average; importance of distinguishing country groups with similar features for precise estimates.
- Policy recommendations:
  - Fiscal policy plays a critical role in supporting the economy in the near term.
  - Prioritize economic sectors where Keynesian effects are strongest when deploying recovery funds.
  - Prioritize the quality and efficient deployment of public investments to preserve fiscal sustainability and support long-term growth and resilience.

*Italic: Source — Excerpts from the provided content unit.*

### 1. Assign  each  fund  disbursement  to  a  specific  “program period-objective  1”  bin.  The

### 1. Assign each fund disbursement to a specific “program period-objective 1” bin

### Disbursement assignment procedure
- Program periods of ESI Funds: 1994–1999, 2000–2006, 2007–2013, 2014–2020.
- Funds included: ERDF, ESF, CF, EAFRD, EMFF.
- Objective 1 criterion: distinguishes regions covered by the convergence criterion and those that are not.
- Overall number of bins: 4×2 = 8 bins. Call the set of bins ב.

### Prediction of disbursement profiles (leave-one-out)
- For a given bin 퐼∈ב and selected funding 푖:
  - Compute the average disbursement profile across all other EU funding belonging to 퐼, excluding funding 푖.
  - Apply this average predicted disbursement profile to the original commitment associated with 푖 (by taking the ratio) to obtain the series of predicted funding-level disbursements.
- Repeat for each 푖∈퐼 and for each 퐼∈ב.
- Aggregate predicted funding-level disbursement series at the country level.

### Key assumption for completed past programs
- Benchmark assumption: for each completed past program, the measured total sum of expenditures matches the initially planned amount (full absorption).

---

### Imputation for 2014–2020 ongoing program
- Separate dataset contains both expenditures and commitments for 2014 to 2020 (“The EU 2014-2020 payments” dataset).
- Compute disbursement rates at the country-year-fund level from that dataset.
- Impute country-fund-year disbursement rates homogeneously to NUTS2 regions within that country-fund-year category.
- Combine NUTS2 payments with disbursement rates to impute an initial commitment value for each region-fund-program in the payments dataset.

### Assessment of full absorption assumption for completed programs
- Use datasets on commitments by country-fund-year and theme to compute disbursement rates at country-fund level, despite lack of consistent regional breakdown.
- Figure 7 (density of final disbursement rates for 2000–2006 and 2007–2013) shows that, in virtually all cases, final observed absorption rate exceeded 90% of initial commitments (with a significant peak at 1), supporting the full absorption assumption for past completed programs.

---

### B.2. Econometric model and estimation

### Controls included in vector 푿
- Lag of GDP growth (to account for effects of economic activity).
- Lag of the outcome variable (to capture persistence/predictability of shocks).
- One measure of year-on-year changes in (lagged) institutional quality (proxied by ICRG ratings).
- One measure of year-on-year changes in (lagged) financial risk (proxied by ICRG ratings).
  - Higher values of ICRG ratings signify better institutional quality and less financial risk.
- Dummy variable for 2009, interacted with the fiscal shock.
- Time fixed effects (no explicit controls for monetary and nominal interest rates).
- For OLS regressions on sectoral shocks to public investments: additionally include (lagged) change in sum of public investments excluding the sector under scrutiny.

### Outlier treatment
- Two-step approach:
  1. Judgmental exclusion of specific years associated with austerity programs:
     - Cyprus (2012, 2013, 2014)
     - Portugal (2011)
     - Spain (2011)
     - Ireland (2011, 2015, 2016)
     - Finland (2015, 2016)
  2. Automated exclusion of regression (standardized) residuals exceeding 2.5 standard deviations (as in Acemoglu et al. (2019)).

### Delta Method for standard errors
- Standard errors associated with 훽+훾 computed using the Delta Method (first-order Taylor approximation).
- For a vector of random variables 푍 and transformation 퐺(푍):
  - 푉푎푟(퐺(푍)) ≈ 훻퐺(푍)ᵀ × 퐶표푣(푍) × 훻퐺(푍).

---

### B.3. CEE sub-sample selection procedure

### Theoretical basis
- Fiscal multipliers generally:
  - Higher in countries with either flexible exchange rate & low capital mobility or fixed exchange rate & high capital mobility.
  - Higher in countries that are less open to trade.

### Variables used
- Exchange rate regime: Shambaugh (2004) peg classification (0-1 dummy).
- Capital account openness: Quinn-Toyoda Index (0 to 100).
- Trade integration: ratio of exports plus imports to GDP.
- For each CEE country, compute sample average of each variable.

### Classification and scoring
- Exchange rate: flexible if average peg < 0.5; fixed if average peg > 0.5.
- Capital account openness and trade integration: classify into 3 buckets based on distribution quantiles.
- Scoring (Batini, Eyraud and Weber (2014) style):
  - Trade integration: upper bucket = 2, intermediate = 1, lower = 0.
  - Capital mobility conditional on peg:
    - If fixed peg: upper = 2, intermediate = 1, lower = 0.
    - If flexible peg: upper = 0, intermediate = 1, lower = 2.
- Sum scores for each country; higher total score implies larger multiplier.
- Grouping for regional dummy 퐷_c: Slovenia grouped with Czech Republic, Croatia, Estonia and Latvia.

### CEE sample (countries listed)
- Slovenia, Czech Republic, Poland, Slovakia, Croatia, Lithuania, Latvia, Estonia, Bulgaria, Hungary, Romania.

### Table 13 (selected values and scores)
- Table columns shown in source: iso, Cap. Mob., Peg, Trade, Score.
- Example entries:
  - PL: 73.61, 0.00, 0.83, 4
  - RO: 91.67, 0.33, 0.70, 4
  - HR: 98.86, 0.82, 0.86, 3
  - LT: 100.00, 0.94, 1.26, 3
  - CZ: 99.31, 0.22, 1.31, 2
  - EE: 100.00, 1.00, 1.42, 2
  - LV: 99.31, 0.67, 1.07, 2
  - SI: 100.00, 0.83, 1.31, 2
  - BG: 90.97, 0.94, 1.11, 1
  - SK: 95.83, 0.56, 1.60, 1
  - HU: 100.00, 0.11, 1.51, 0

---

### XII. Appendix C — Robustness checks

### Additional controls
- Add (lagged) change in “other” gross inflows to GDP.
- Add (lagged) change in the real effective exchange rate.
- Data availability reduces sample size; coefficients remain closely aligned with main aggregate multipliers.

### Table 14: Aggregate multipliers (robustness check) — key results
- Note: Using 2SLS method, a 1 percent of GDP increase in ESI disbursements increases GDP, contemporaneously, by 1.293 percent on average.
- Panel 1 — Impact Multiplier (2SLS) (standard errors in parentheses):
  - ESI Funds → GDP: 1.293*** (0.474)
  - ESI Funds → Tot. Inv.: 1.403*** (0.338)
  - ESI Funds → Priv. Inv: 0.680* (0.356)
  - ESI Funds → Empl.: 0.0393 (0.212)
- Panel 1 R-squared and Observations:
  - GDP: R-squared 0.828, Observations 460
  - Tot. Inv.: R-squared 0.688, Observations 456
  - Priv. Inv: R-squared 0.679, Observations 456
  - Empl.: R-squared 0.779, Observations 459
- Panel 2 — 1Y Multiplier (2SLS):
  - GDP: 1.607** (0.652), R-squared 0.796, Observations 427
  - Tot. Inv.: 1.689*** (0.430), R-squared 0.628, Observations 426
  - Priv. Inv: 0.754** (0.383), R-squared 0.643, Observations 425
  - Empl.: 0.0390 (0.413), R-squared 0.733, Observations 428

### Table 15: Aggregate multipliers for CEE sub-sample (robustness check) — key results
- Note: Using 2SLS method, a 1 percent of GDP increase in ESI disbursements increases GDP, contemporaneously, by 1.332 percent on average.
- Panel 1 — Impact Multiplier (2SLS):
  - ESI Funds → GDP: 1.332*** (0.469)
  - ESI Funds → Tot. Inv.: 1.729*** (0.562)
  - ESI Funds → Priv. Inv: 1.078* (0.553)
  - ESI Funds → Empl.: 0.108 (0.195)
- Panel 1 R-squared and Observations:
  - GDP: R-squared 0.825, Observations 461
  - Tot. Inv.: R-squared 0.682, Observations 457
  - Priv. Inv: R-squared 0.678, Observations 456
  - Empl.: R-squared 0.779, Observations 459
- Panel 2 — 1Y Multiplier (2SLS):
  - GDP: 1.505** (0.760), R-squared 0.796, Observations 427
  - Tot. Inv.: 2.086*** (0.638), R-squared 0.6286, Observations 426
  - Priv. Inv: 1.326** (0.548), R-squared 0.643, Observations 425
  - Empl.: 0.128 (0.783), R-squared 0.7333, Observations 428

---

### XIII. Appendix D — The bucket approach of Batini, Eyraud and Weber (2014) and application to Slovenia

### Method overview
- Bunch countries into three groups (Low, Medium, High multipliers) based on structural characteristics associated with fiscal multipliers.
- Implementation steps for a country:
  a) Assign value 1 if a characteristic associated with a large multiplier is satisfied, 0 otherwise.
  b) Sum scores to determine first-year “normal times” multiplier category.

### Structural characteristics used (for Slovenia) and scores
- a) Low trade openness (exports+imports)/GDP:
  - Slovenia average (2015–2019): approximately 1.54.
  - Threshold ≈ 0.59 → score = 0.
- b) Small automatic stabilizers (public spending/GDP):
  - Slovenia average (2014–2018): 0.45 of GDP.
  - Threshold ≈ 0.34 → score = 0.
- c) Fixed or Quasi-fixed exchange rate regime:
  - Slovenia is in the Eurozone → score = 1.
- d) High degree of labor market rigidity:
  - Slovenia less flexible than world median (as of 2017) → score = 1.
- e) Low level of public debt (gross government debt to GDP):
  - Slovenia gross government debt = 66 percent of GDP in 2019 → score = 1.
- f) Effective public expenditure management and revenue administration:
  - Judged as effective relative to peers → score = 1.
- Total Slovenia score = 4 (see Table17 in source).

### Mapping scores to multiplier ranges (Table16)
- Low Multiplier: 0.1–0.3
- Medium Multiplier: 0.4–0.6
- High Multiplier: 0.7–1.0

### Adjustments for cyclical position and monetary stance
- Business cycle adjustment:
  - At lowest point of cycle: increase range by 60 percent (both bounds).
  - At peak of cycle: decrease range by 40 percent (both bounds).
  - Interpolate otherwise.
  - Slovenia 2019 output gap: 1.8 percent of potential GDP; historical peak 7.6 percent in 2008 → by interpolation shrink range by approximately 9 percent on both bounds.
- Monetary policy adjustment:
  - If at effective lower bound and fully constrained: increase range by 30 percent.
  - Slovenia in Euro Area → increase range by 30 percent on both bounds.

### Resulting ranges for Slovenia (normal times)
- Medium multiplier final range: M_medium = [0.47,0.71].
- High multiplier final range: M_high = [0.83,1.18].

### COVID-19 adjustment (economy at lowest point in cycle)
- Slovenia GDP decline in 2020: 5.5 percent.
- Under assumption economy at lowest point in cycle:
  - M_medium_Covid = [0.83,1.25]
  - M_high_Covid = [1.46,2.08]

### Practical notes
- Bucket approach useful when data are limited or as a diagnostic before formal models.
- Yields wide ranges to allow judgment based on priors, size of government-controlled economy, and fiscal program credibility.
- Caveat: COVID-19 contraction is unique; standard fiscal stimulus can be less effective if sectors are shut down (see Guerrieri, Lorenzoni and Straub (2020) discussion in source).

*Source: wpiea2021118-print-pdf.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021118-print-pdf.pdf_