## wp1763 - Annex A). Estevao & Samake (2013) also present estimates of multipliers for government

## Source details

**Canonical URL:** [wp1763 - Annex A). Estevao & Samake (2013) also present estimates of multipliers for government](https://www.imf.org/-/media/files/publications/wp/2017/wp1763.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2017/wp1763.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2017/wp1763.pdf.json)

---

### Government spending multipliers: VAR evidence and cumulative estimates
- Impact multipliers:
  - Current expenditure: around 0 on impact.
  - Capital expenditure: impact multiplier around 0.1 for total GDP and 0.0 for non-agro non-energy GDP (Table 1, t=0).
- Medium- to long-run cumulative multipliers in Paraguay (Table 1):
  - Current Expenditure — Total GDP:
    - t=0: 0.0
    - t=8: 0.2*
    - t=20: 0.5**
  - Current Expenditure — Non-Agro Non-Energy GDP:
    - t=0: 0.0
    - t=8: 0.2**
    - t=20: 0.5**
  - Capital Expenditure — Total GDP:
    - t=0: 0.1
    - t=8: 1.4**
    - t=20: 2.1**
  - Capital Expenditure — Non-Agro Non-Energy GDP:
    - t=0: 0.0
    - t=8: 0.5
    - t=20: 0.8
- Robustness with Ministry of Finance quarterly fiscal data (2003:Q1-2016:Q1, GFS 2001 presentation):
  - Impact multipliers remain around zero.
  - Point estimates over longer horizons generally lower than baseline VAR results.
  - Multipliers estimated as:
    - Real current primary expenditure: around 0.1
    - Capital expenditure: between 0.4 to 0.9
- Statistical significance notation:
  - ** indicates statistical significance at the 5 percent level.
  - * indicates significance at the 10 percent level.

### Tax multipliers: VAR evidence
- VAR setup and identification choices:
  - Models include four endogenous variables: real total government expenditures; real GDP (or real non-agriculture non-energy GDP); real tax revenues (or the change in cyclically adjusted tax revenues); and changes in the real effective exchange rate.
  - Deterministic linear time trend included as an exogenous variable.
  - For models with cyclically-adjusted revenues, tax policy assumed not to respond contemporaneously to output shocks; for unadjusted revenues, contemporaneous response allowed to capture automatic stabilizers.
- VAR implied responses and cumulative multipliers (Table 2):
  - Real Tax Revenue — Total GDP:
    - t=0: 0.0
    - t=8: -0.1
    - t=20: -0.2
  - Real Tax Revenue — Non-Agro Non-Energy GDP:
    - t=0: 0.0
    - t=8: 0.0
    - t=20: 0.1
  - Cyclically Adjusted Tax Revenue — Total GDP:
    - t=0: 0.0
    - t=8: -0.1
    - t=20: -0.2
  - Cyclically Adjusted Tax Revenue — Non-Agro Non-Energy GDP:
    - t=0: 0.0
    - t=8: -0.2*
    - t=20: -0.4**
- Interpretation:
  - Multipliers implied by VAR impulse responses are close to zero in magnitude and not statistically significant in most cases.
  - Notable exception: cumulative multiplier of cyclically adjusted tax revenues on non-agro non-energy GDP estimated at -0.4 after 5 years (-0.2 after two years) and significant at conventional levels.
  - Tax multipliers larger (in magnitude and significance) for non-agro non-energy GDP — consistent with low taxation in agriculture and large agricultural GDP share.
- Comparisons:
  - Estimates differ from Gunter, Riera-Crichton, Vegh & Vuletin (2016) and Matheson & Pereira (2016), who find cumulative multipliers around -2 for developing countries and Brazil.
  - Colman, Franco, & Tello (2016) using a structural VAR approach also find tax multipliers in Paraguay close to zero.
- Robustness with Ministry of Finance quarterly fiscal data (2003:Q1-2016:Q1):
  - Estimates of the tax multiplier over this period are also close to zero.

### Local Projection (LP) method results and multipliers
- Method and identification:
  - Use Jorda’s (2005) local projection method; multipliers directly computed as cumulative response of GDP divided by cumulative government spending over a given horizon (Ramey & Zubairy, 2016 approach).
  - Blanchard-Perotti identification for shocks: residuals from regression of real government expenditures on lagged expenditures, lagged real GDP, and lagged changes in the real effective exchange rate.
  - GDP and government expenditures normalized by trend GDP using the HP filter.
  - Ninety-five percent heteroscedasticity and autocorrelation consistent (HAC) standard errors reported.
- LP method findings (Figure 3 summary):
  - Capital expenditure multipliers:
    - Impact multiplier for capital expenditure on total GDP: zero on impact.
    - Reaches 2 after four quarters and is statistically significant at the 5 percent level.
    - Multipliers for capital expenditure on non-agro non-energy GDP are statistically significant under LP (contrary to some VAR results).
  - Current expenditure multipliers:
    - Cumulative multiplier for total GDP becomes statistically significant after seven quarters and reaches 0.9.
    - This 0.9 is substantially larger than VAR-based estimates.
  - Overall, LP confirms higher multipliers for capital expenditure relative to current expenditure.
- Robustness checks (Annex E):
  - Alternative specifications: removing real effective exchange rate from regressions; including real effective exchange rate in levels instead of first differences.
  - Main results hold: medium-term multipliers substantially higher for capital expenditure; impact multipliers remain statistically insignificant at the 5 percent level.
- Cyclical behavior:
  - Correlation between cyclical component of capital expenditure and cyclical component of output:
    - For 2003-2015: negative and statistically significant at around -0.3.
    - For full sample (1998-2015): negative correlation not statistically significant.
  - Interpretation: capital expenditure has behaved in a countercyclical manner in the more recent period.

### LP estimates of tax multipliers
- LP tax shock construction:
  - Shock 1: residuals from regression of real tax revenue on lagged tax revenue; contemporaneous and lagged government expenditures; contemporaneous and lagged GDP; lagged changes in the real effective exchange rate; and deterministic trend.
  - Shock 2: same as Shock 1 but using cyclically adjusted revenues.
- LP results:
  - Estimated tax multipliers in Paraguay are not statistically significant at conventional levels in all LP cases.
  - Contrasts with VAR results for non-agriculture non-energy GDP where negative and significant multipliers at longer horizons were found; aligns with VAR results for total GDP.
  - Robustness checks in Annex E confirm tax multipliers in Paraguay are not significant both economically and statistically.

### Further identification attempts
- Kraay (2014) style identification using predicted disbursements by official creditors:
  - Rationale: predicted disbursements by multilateral and bilateral creditors can isolate predetermined component of public spending uncorrelated with contemporaneous macro shocks.
  - Paraguay context:
    - Kraay’s dataset (annual, 1972-2010): disbursements by official creditors represent 13.7 percent of total government spending in Paraguay over the period (sample average 13.4 percent).
  - Replication results (Table 3 summary):
    - OLS and Two-Stage Least Squares point estimates of multiplier range from 0.2 to 0.6.
    - None of the coefficients statistically significant at conventional levels (even after weak-instrument-consistent confidence intervals).
    - First-stage regressions indicate a weak link between predicted disbursements and government expenditures in Paraguay.
- Narrative approach for tax shocks:
  - Challenges in Paraguay: small number of statutory tax rate changes over sample period.
  - Vegh and Vuletin (2015) dataset (1979-2013) includes only four changes in tax rates in Paraguay: introduction of VAT in 1991; corporate tax reductions from 30 to 20 percent in 2005 and from 20 to 10 percent in 2006; introduction of the personal income tax of 10 percent in 2007.

### Narrative identification and narrative-shock LP tax multipliers
- Methodology for Identifying Fiscal Shocks:
  - Adopted the Romer and Romer (2010) “narrative” measure of exogenous fiscal shocks based on the estimated revenue impact of policy changes.
  - Motivation categories considered:
    - “Spending Driven” (excluded).
    - “Countercyclical Action” (excluded).
    - “Deficit Driven” (included as exogenous).
    - Exogenous changes motivated by raising long-run growth (included as exogenous).
  - Timing convention: if a tax change occurs before the mid-point of the quarter, it is assigned to that quarter; otherwise it is assigned to the next quarter.
  - Size of tax shocks: estimated revenue impact at implementation expressed in annual terms and prevailing level of GDP; converted to local currency units in real terms (deflated by the GDP deflator).
  - Identified 11 instances of “tax shocks” over the period 2001-2015, with most shocks being positive.
  - Footnote: the personal income tax was only effectively implemented in 2012.
- Narrative-Shock Estimated Tax Multipliers (Local Projections - LP):
  - Sample period for LP estimates: 1998:Q1-2015:Q2.
  - Impact multipliers:
    - Total GDP: -0.1
    - Non-agro non-energy GDP: -0.2 (marginally significant)
  - Medium-term cumulative multipliers:
    - In the order of -2 (consistent with Gunter, Riera-Crichton, Vegh & Vuletin (2016) narrative estimates and Ramey (2016) results for advanced economies).
- Conventional VAR Results and Robustness:
  - When including the narrative tax shock in conventional VAR models, impulse responses point to multipliers that are not statistically significant.
  - Alternative identification approaches (conventional VARs) suggest:
    - Tax multipliers are low relative to other countries and seem to affect non-agro non-energy GDP more prominently.
    - Estimates of tax multipliers range from 0 to -0.4 over the medium-term.
  - Diagnostics:
    - Bootstrapped standard errors (500 repetitions) used for VAR impulse responses.
    - Lag length selection based on AIC, SBC, and LR test; higher order VAR used when conflicting results were obtained.

### Fiscal expenditure multipliers and comparative findings
- Reported ranges and key figures:
  - Capital expenditure medium-term multipliers range from 0.4 to 2.
  - Current expenditure multipliers range from 0.1 to 0.9.
- Annex Table (government spending multipliers, cumulative):
  - Current Primary Expenditure (Total GDP): t=0 0.0; t=8 0.1; t=20 0.1
  - Current Primary Expenditure (Non-Agro Non-Energy GDP): t=0 0.0; t=8 0.1**; t=20 0.1**
  - Capital Expenditure (Total GDP): t=0 0.0; t=8 0.9**; t=20 0.9**
  - Capital Expenditure (Non-Agro Non-Energy GDP): t=0 0.0; t=8 0.4*; t=20 0.4*
- Possible explanation:
  - Capital expenditure is to a large extent financed by external borrowing in Paraguay, mitigating crowding-out effects.

### Policy implications and recommendations
- Evidence supports IMF policy advice to contain growth in current primary expenditure while preserving capital expenditure, given substantially higher multipliers for capital expenditure.
- Given the (uncertain) estimates of tax multipliers, the “balanced budget” multiplier for Paraguay — i.e., the effect on output of an increase in expenditures (in particular capital expenditure) financed by taxes — is likely to be positive.
- Capital expenditure appears to have behaved in a countercyclical manner over 2003-2015 and might have been used effectively as a stabilization tool by accelerating execution during downturns and restraining it during expansionary phases.

### Caveats and remaining challenges
- Estimates of expenditure and tax multipliers should be taken with caution due to:
  - Sensitivity to identification strategy (LP narrative approach vs. conventional VARs).
  - Lack of robustness of larger narrative-based tax multiplier estimates to different econometric techniques.
- The challenge to identify fiscal shocks in a satisfactory way remains open.

### Annex E. Additional results: variations in model specification (summary)
- LP cumulative multipliers reported for alternative specifications:
  - Real exchange rate excluded and real exchange rate in levels both examined.
  - Sample period 1998:Q1-2015:Q2.
  - Shaded area represents 95 percent confidence interval (HAC standard error bands).
  - Horizontal axis horizons displayed as 0 2 4 6 8.
  - Vertical axis ranges depicted in figures:
    - Capital Expenditures panels: from -1 to 4; 0 to 4.
    - Current Expenditures panels: from -1.5 to 1; -1 to 1; or -2 to 2; -0.5 to 1.5.
    - Tax multiplier panels: from -2 to 2; -1 to 1.5; or -0.5 to 1.5.

### Annex F. Narrative exogenous tax shocks (selected entries and details)
- Identification note:
  - DD = A deficit-driven tax change is a tax increase designed to reduce an inherited budget deficit. Exogenous to output.
  - LR = A long-run tax change is one aimed at raising long-run growth. Exogenous to cyclical considerations.
  - Methodology follows Romer and Romer (2010).
- Selected tabulated tax-shock entries (Year | Description | Timing | Size | Motivation):
  - 2001 | Increase in excise tax on diesel; Inclusion of transfers and personal services in VAT tax base. | Q1 | 0.5 percent of GDP | DD
  - 2003 | Overhaul of tax and customs administration. | Q1 | 0.9 percent of GDP | DD
  - 2003 | Increase in the excise tax on diesel fuel (6 percentage point hike). | Q3 (August). | 0.5 percent of GDP | DD
  - 2004 | Increase in the excise tax on diesel fuel (6 percentage point hike). | Q1 (January) | 0.25 percent of GDP | DD
  - 2004 | Temporary export tax on soy. | Q1 (February) | 0.2 percent of GDP | DD
  - 2004 | Tax reform (broad measures including instituting a new personal income tax). | Q3 (approved on June 25). | 0.1 percent on GDP | DD/LR
  - 2005 | Elimination of temporary export taxes soy. | Q1 (February) | -0.2 per cent of GDP | LR
  - 2005 | Reduction in corporate income tax rate. From 30 to 20 percent. | Q3 (August). | -0.4 percent of GDP | LR
  - 2006 | Reduction in corporate income tax rate. From 20 to 10 percent. | Q3 (August). | -0.7 or -1.2 percent of GDP in 2006 | LR
  - 2012 | Introduction of personal income tax (10 percent rate). | Q3 (August). | Original estimate between 0.2 to 0.3 percent of GDP. 2014 estimate is 0.1 percent of GDP. | DD/LR
  - 2014 | IRAGRO and Agricultural VAT. Revamped tax on agricultural income. | Q1 (January) | 0.23 percent of GDP | DD/LR
- Key dataset and identification details:
  - Sample period for multiplier estimations: 1998:Q1-2015:Q2.
  - Confidence intervals reported: 95 percent (HAC standard error bands).
  - Horizon ticks as shown in figures: 0, 2, 4, 6, 8.

*Source: wp1763 - Annex A). Estevao & Samake (2013) also present estimates of multipliers for government; introduction of the personal income tax of 10 percent in 2007; Annex E and Annex F excerpts.*

### Annex A). Estevao & Samake (2013) also present estimates of multipliers for government

### wp1763 - Annex A). Estevao & Samake (2013) also present estimates of multipliers for government

### Government spending multipliers: VAR evidence and cumulative estimates
- Impact multipliers:
  - Current expenditure: around 0 on impact (general finding).
  - Capital expenditure: impact multiplier around 0.1 for total GDP and 0.0 for non-agro non-energy GDP (Table 1, t=0).
- Medium- to long-run cumulative multipliers in Paraguay (Table 1):
  - Current Expenditure — Total GDP:
    - t=0: 0.0
    - t=8: 0.2*
    - t=20: 0.5**
  - Current Expenditure — Non-Agro Non-Energy GDP:
    - t=0: 0.0
    - t=8: 0.2**
    - t=20: 0.5**
  - Capital Expenditure — Total GDP:
    - t=0: 0.1
    - t=8: 1.4**
    - t=20: 2.1**
  - Capital Expenditure — Non-Agro Non-Energy GDP:
    - t=0: 0.0
    - t=8: 0.5
    - t=20: 0.8
- Statistical significance notation:
  - ** indicates statistical significance at the 5 percent level.
  - * indicates significance at the 10 percent level.
- Robustness with Ministry of Finance quarterly fiscal data (2003:Q1-2016:Q1, GFS 2001 presentation):
  - Impact multipliers remain around zero.
  - Point estimates over longer horizons generally lower than baseline VAR results.
  - Multipliers estimated as:
    - Real current primary expenditure: around 0.1
    - Capital expenditure: between 0.4 to 0.9

### Tax multipliers: VAR evidence
- General VAR setup and identification choices:
  - Models include four endogenous variables: real total government expenditures; real GDP (or real non-agriculture non-energy GDP); real tax revenues (or the change in cyclically adjusted tax revenues); and changes in the real effective exchange rate.
  - Deterministic linear time trend included as an exogenous variable.
  - For models with cyclically-adjusted revenues, tax policy assumed not to respond contemporaneously to output shocks; for unadjusted revenues, contemporaneous response allowed to capture automatic stabilizers.
- VAR implied responses and cumulative multipliers (Table 2):
  - Real Tax Revenue — Total GDP:
    - t=0: 0.0
    - t=8: -0.1
    - t=20: -0.2
  - Real Tax Revenue — Non-Agro Non-Energy GDP:
    - t=0: 0.0
    - t=8: 0.0
    - t=20: 0.1
  - Cyclically Adjusted Tax Revenue — Total GDP:
    - t=0: 0.0
    - t=8: -0.1
    - t=20: -0.2
  - Cyclically Adjusted Tax Revenue — Non-Agro Non-Energy GDP:
    - t=0: 0.0
    - t=8: -0.2*
    - t=20: -0.4**
- Interpretation:
  - Multipliers implied by VAR impulse responses are close to zero in magnitude and not statistically significant in most cases.
  - Notable exception: cumulative multiplier of cyclically adjusted tax revenues on non-agro non-energy GDP estimated at -0.4 after 5 years (-0.2 after two years) and significant at conventional levels.
  - Tax multipliers larger (in magnitude and significance) for non-agro non-energy GDP — consistent with low taxation in agriculture and large agricultural GDP share.
- Comparisons:
  - Estimates differ from Gunter, Riera-Crichton, Vegh & Vuletin (2016) and Matheson & Pereira (2016), who find cumulative multipliers around -2 for developing countries and Brazil.
  - Colman, Franco, & Tello (2016) using a structural VAR approach also find tax multipliers in Paraguay close to zero.
- Robustness with Ministry of Finance quarterly fiscal data (2003:Q1-2016:Q1):
  - Estimates of the tax multiplier over this period are also close to zero.

### Local Projection (LP) method results and multipliers
- Method and identification:
  - Use Jorda’s (2005) local projection method; multipliers directly computed as cumulative response of GDP divided by cumulative government spending over a given horizon (Ramey & Zubairy, 2016 approach).
  - Blanchard-Perotti identification for shocks: residuals from regression of real government expenditures on lagged expenditures, lagged real GDP, and lagged changes in the real effective exchange rate.
  - GDP and government expenditures normalized by trend GDP using the HP filter.
  - Ninety-five percent heteroscedasticity and autocorrelation consistent (HAC) standard errors reported.
- LP method findings (Figure 3 summary):
  - Capital expenditure multipliers:
    - Impact multiplier for capital expenditure on total GDP: zero on impact.
    - Reaches 2 after four quarters and is statistically significant at the 5 percent level.
    - Multipliers for capital expenditure on non-agro non-energy GDP are statistically significant under LP (contrary to some VAR results).
  - Current expenditure multipliers:
    - Cumulative multiplier for total GDP becomes statistically significant after seven quarters and reaches 0.9.
    - This 0.9 is substantially larger than VAR-based estimates.
  - Overall, LP confirms higher multipliers for capital expenditure relative to current expenditure.
- Robustness checks (Annex E):
  - Alternative specifications: removing real effective exchange rate from regressions; including real effective exchange rate in levels instead of first differences.
  - Main results hold: medium-term multipliers substantially higher for capital expenditure; impact multipliers remain statistically insignificant at the 5 percent level.
- Cyclical behavior:
  - Correlation between cyclical component of capital expenditure and cyclical component of output:
    - For 2003-2015: negative and statistically significant at around -0.3.
    - For full sample (1998-2015): negative correlation not statistically significant.
  - Interpretation: capital expenditure has behaved in a countercyclical manner in the more recent period (consistent with Correa, Colman, and Tello (2016)).

### LP estimates of tax multipliers
- LP tax shock construction:
  - Shock 1: residuals from regression of real tax revenue on lagged tax revenue; contemporaneous and lagged government expenditures; contemporaneous and lagged GDP; lagged changes in the real effective exchange rate; and deterministic trend.
  - Shock 2: same as Shock 1 but using cyclically adjusted revenues.
- LP results:
  - Estimated tax multipliers in Paraguay are not statistically significant at conventional levels in all LP cases.
  - Contrasts with VAR results for non-agriculture non-energy GDP where negative and significant multipliers at longer horizons were found; aligns with VAR results for total GDP.
  - Robustness checks in Annex E confirm tax multipliers in Paraguay are not significant both economically and statistically.

### Further identification attempts
- Kraay (2014) style identification using predicted disbursements by official creditors:
  - Rationale: predicted disbursements by multilateral and bilateral creditors can isolate predetermined component of public spending uncorrelated with contemporaneous macro shocks.
  - Paraguay context:
    - Kraay’s dataset (annual, 1972-2010): disbursements by official creditors represent 13.7 percent of total government spending in Paraguay over the period (sample average 13.4 percent).
  - Replication results (Table 3 summary):
    - OLS and Two-Stage Least Squares point estimates of multiplier range from 0.2 to 0.6.
    - None of the coefficients statistically significant at conventional levels (even after weak-instrument-consistent confidence intervals).
    - First-stage regressions indicate a weak link between predicted disbursements and government expenditures in Paraguay.
- Narrative approach for tax shocks:
  - Challenges in Paraguay: small number of statutory tax rate changes over sample period.
  - Vegh and Vuletin (2015) dataset (1979-2013) includes only four changes in tax rates in Paraguay: introduction of VAT in 1991; corporate tax reductions from 30 to 20 percent in 2005 and from 20 to 10 percent in 2006; (text cuts off thereafter).

*Source: wp1763 - Annex A). Estevao & Samake (2013) also present estimates of multipliers for government (excerpt).*

### introduction of the personal income tax of 10 percent in 2007.

### wp1763 - introduction of the personal income tax of 10 percent in 2007.

### Methodology for Identifying Fiscal Shocks
- Adopted the Romer and Romer (2010) “narrative” measure of exogenous fiscal shocks based on the estimated revenue impact of policy changes.
- Considered four motivation categories for tax changes:
  - “Spending Driven” (excluded from regressions as typically endogenous to output).
  - “Countercyclical Action” (endogenous to the cycle; excluded).
  - “Deficit Driven” (discretionary, not linked to the cycle; included as exogenous).
  - Exogenous changes motivated by raising long-run growth (included as exogenous).
- Timing convention: if a tax change occurs before the mid-point of the quarter, it is assigned to that quarter; otherwise it is assigned to the next quarter (measuring the effect of “anticipated” tax changes).
- Size of tax shocks: estimated revenue impact at implementation expressed in annual terms and prevailing level of GDP; converted to local currency units in real terms (deflated by the GDP deflator) for regressions.
- Identified 11 instances of “tax shocks” over the period 2001-2015, with most shocks being positive (aiming at increasing revenues).
- Footnote: As shown in Annex F, the personal income tax was only effectively implemented in 2012.

### Narrative-Shock Estimated Tax Multipliers (Local Projections - LP)
- Sample period for LP estimates: 1998:Q1-2015:Q2.
- Impact multipliers:
  - Total GDP: -0.1
  - Non-agro non-energy GDP: -0.2 (marginally significant)
- Medium-term cumulative multipliers:
  - In the order of -2 (consistent with Gunter, Riera-Crichton, Vegh & Vuletin (2016) narrative estimates and Ramey (2016) results for advanced economies).

### Conventional VAR Results and Robustness
- When including the narrative tax shock in conventional VAR models, impulse responses point to multipliers that are not statistically significant, undermining robustness of LP-based estimates.
- Alternative identification approaches (conventional VARs) suggest:
  - Tax multipliers are low relative to other countries and seem to affect non-agro non-energy GDP more prominently.
  - Estimates of tax multipliers range from 0 to -0.4 over the medium-term.
- Summary of method diagnostics:
  - Bootstrapped standard errors (500 repetitions) used for VAR impulse responses.
  - Lag length selection based on AIC, SBC, and LR test; higher order VAR used when conflicting results were obtained.

### Fiscal Expenditure Multipliers and Key Comparative Findings
- Current versus capital expenditure multipliers in Paraguay:
  - Capital expenditure medium-term multipliers range from 0.4 to 2.
  - Current expenditure multipliers range from 0.1 to 0.9.
- Possible explanation: capital expenditure is to a large extent financed by external borrowing in Paraguay, mitigating crowding-out effects.
- Annex Table (government spending multipliers, cumulative):
  - Current Primary Expenditure (Total GDP): t=0 0.0; t=8 0.1; t=20 0.1
  - Current Primary Expenditure (Non-Agro Non-Energy GDP): t=0 0.0; t=8 0.1**; t=20 0.1**
  - Capital Expenditure (Total GDP): t=0 0.0; t=8 0.9**; t=20 0.9**
  - Capital Expenditure (Non-Agro Non-Energy GDP): t=0 0.0; t=8 0.4*; t=20 0.4*
  - Note: ** indicates statistical significance at the 5 percent level. * indicates significance at the 10 percent level.

### Policy Implications and Recommendations
- Evidence supports IMF policy advice to contain growth in current primary expenditure while preserving capital expenditure, given substantially higher multipliers for capital expenditure.
- Given the (uncertain) estimates of tax multipliers, the “balanced budget” multiplier for Paraguay — i.e., the effect on output of an increase in expenditures (in particular capital expenditure) financed by taxes — is likely to be positive.
- Capital expenditure appears to have behaved in a countercyclical manner over 2003-2015 and might have been used effectively as a stabilization tool by accelerating execution during downturns and restraining it during expansionary phases.

### Caveats and Remaining Challenges
- Estimates of expenditure and tax multipliers should be taken with caution due to:
  - Sensitivity to identification strategy (LP narrative approach vs. conventional VARs).
  - Lack of robustness of larger narrative-based tax multiplier estimates to different econometric techniques.
- The challenge to identify fiscal shocks in a satisfactory way remains open.

*Source: wp1763 - introduction of the personal income tax of 10 percent in 2007.*

### Annex E. Additional Results: Variations in Model Specification

### Annex E. Additional Results: Variations in Model Specification

### Cumulative Multipliers — LP Method (Real Exchange Rate Excluded)
- Methodology and sample:
  - Shaded area represents 95 percent confidence interval (HAC standard error bands).
  - Sample period 1998:Q1-2015:Q2.
  - Multipliers estimated correspond to the cumulative response of GDP divided by the cumulative government expenditure or tax revenue over a given horizon.
- Reported series (panel labels as in figure):
  - Cumulative Multiplier: Capital Expenditures
  - Cumulative Multiplier Non-Agro Non-Energy: Capital Expenditures
  - Cumulative Multiplier: Current Expenditures
  - Cumulative Multiplier Non-Agro Non-Energy: Current Expenditures
  - Cumulative Tax Multiplier: Adjusted Revenue Total GDP
  - Cumulative Tax Multiplier: Adjusted Revenue Non-Agro Non-Energy GDP
- Horizontal axis shown as horizons: 0 2 4 6 8 (plotted as "02468" in figure).
- Vertical axis ranges (as depicted in panels):
  - Capital Expenditures panels: from -1 to 4 (left panel) and 0 to 4 (right panel).
  - Current Expenditures panels: from -1.5 to 1 and -1 to 1.
  - Tax multiplier panels: from -2 to 2 and -1 to 1.5.

### Cumulative Multipliers — LP Method (Real Exchange Rate in Levels)
- Methodology and sample:
  - Shaded area represents 95 percent confidence interval (HAC standard error bands).
  - Sample period 1998:Q1-2015:Q2.
  - Multipliers estimated correspond to the cumulative response of GDP divided by the cumulative government expenditure or tax revenue over a given horizon.
- Reported series (panel labels as in figure):
  - Cumulative Multiplier: Capital Expenditures
  - Cumulative Multiplier Non-Agro Non-Energy: Capital Expenditures
  - Cumulative Multiplier: Current Expenditures
  - Cumulative Multiplier Non-Agro Non-Energy: Current Expenditures
  - Cumulative Tax Multiplier: Adjusted Revenue Total GDP
  - Cumulative Tax Multiplier: Adjusted Revenue Non-Agro Non-Energy GDP
- Horizontal axis shown as horizons: 0 2 4 6 8 (plotted as "02468" in figure).
- Vertical axis ranges (as depicted in panels):
  - Capital Expenditures panels: from -1 to 4 and 0 to 4.
  - Current Expenditures panels: from -2 to 2 and -0.5 to 1.5.
  - Tax multiplier panels: from -2 to 2 and -0.5 to 1.5.

### Annex F. Narrative Exogenous Tax Shocks
- Identification note:
  - DD = A deficit-driven tax change is a tax increase designed to reduce an inherited budget deficit. Exogenous to output.
  - LR = A long-run tax change is one aimed at raising long-run growth. Exogenous to cyclical considerations.
  - Methodology to identify tax shocks follows Romer and Romer (2010).
- Tabulated tax-shock entries (Year | Description | Timing | Size | Motivation | Source):
  - 2001 | Increase in excise tax on diesel. Inclusion of transfers and personal services in VAT tax base. | Q1 (report refers to measures of late 2000 early 2001). | 0.5 percent of GDP | DD | IMF Country report 01/87. Page 10
  - 2003 | Overhaul of tax and customs administration. | Q1 (report mentions improvements in the first 3 months). | 0.9 percent of GDP | DD | IMF Country Report 04/66 . Page 8 and Box 1.
  - 2003 | Increase in the excise tax on diesel fuel (6 percentage point hike). | Q3 (August). | 0.5 percent of GDP | DD | IMF Country Report 04/66 . Page 8.
  - 2004 | Increase in the excise tax on diesel fuel (6 percentage point hike). | Q1 (January) | 0.25 percent of GDP | DD | IMF Country Report 04/294, page 7
  - 2004 | Temporary export tax on soy. | Q1 (February) | 0.2 percent of GDP | DD | IMF Country Report 04/294, page 7
  - 2004 | Tax reform. The law eliminates most exemptions to the corporate income tax while reducing rates, broadens the base of the VAT, institutes a new personal income tax, and institutes a new agricultural income tax to replace the IMAGRO tax. | Q3 (approved on June 25). | 0.1 percent on GDP, estimated effect in 2004. | DD/LR | IMF Country Report 05/59, page 13, page 25, page 43. Table 16.
  - 2005 | Elimination of temporary export taxes soy. | Q1 (February) | -0.2 per cent of GDP. | LR | IMF Country Report 06/30, page 11, footnote 7.
  - 2005 | Reduction in corporate income tax rate. From 30 to 20 percent. | Q3 (August). | -0.4 percent of GDP | LR | IMF Country Report 06/100, page 5. Annual impacts estimated in IMF Country Report 05/59 Table 16.
  - 2006 | Reduction in corporate income tax rate. From 20 to 10 percent. | Q3 (August). | -0.7 or -1.2 percent of GDP in 2006 | LR | IMF Country Report 05/59 Table 16. Footnote 2 of IMF Country Report 06/100, page 5, estimates higher impact at 1.2 percent of GDP.
  - 2012 | Introduction of personal income tax (10 percent rate). | Q3 (August). | Original estimate between 0.2 to 0.3 percent of GDP. 2014 estimate is 0.1 percent of GDP. | DD/LR | IMF Country Report 10/170, page 19. IMF Country Report 14/60, Page 7 and Table 1.
  - 2014 | IRAGRO and Agricultural VAT. Revamped tax on agricultural income. | Q1 (January) | 0.23 percent of GDP | DD/LR | IMF Country Report 14/60, Page 7 and IMF staff estimate in country framework.

### Key dataset and identification details reported
- Sample period for multiplier estimations: 1998:Q1-2015:Q2.
- Confidence intervals reported: 95 percent (HAC standard error bands).
- Horizon ticks as shown in figures: 0, 2, 4, 6, 8.

*Source: wp1763 - Annex E. Additional Results: Variations in Model Specification*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2017/wp1763.pdf_