## _wp05146 — Input Price Indices in Indonesia, 1991–2001

## Source details

**Canonical URL:** [_wp05146 — Input Price Indices in Indonesia, 1991–2001](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05146.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05146.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05146.pdf.json)

---

### Introduction: scope, data, and key empirical patterns
- Objective: disentangle productivity gains from reductions in final goods tariffs versus intermediate input tariffs using Indonesian plant-level data.
- Data and sample:
  - Manufacturing census covering 1991 to 2001 for all plants with 20 or more employees.
  - Unbalanced panel of around 10,000 firms per year with a total of 116,121 observations.
  - Plant-level information on output, employment, ownership, exports, and imports; imported inputs observed at the plant level.
  - Imported-input weights constructed from a 1998 firm-level detailed input list; input mix assumed constant over the sample.
- Tariff patterns:
  - Final goods tariffs fell from an average of 21 percent in 1991 to an average of 8 percent in 2001; some tariffs remain as high as 170 percent.
  - Example of escalating structure (2001, selected ISIC codes): motor vehicle bodies 0 percent; motor vehicle components 11 percent; motor vehicles 31.6 percent.
  - WTO commitment starting 1995 to reduce all bound tariffs to 40 percent or less (over a 10-year period, subject to an exclusion list).
- Key empirical preview:
  - A 10 percentage point fall in input tariffs → 3 percent productivity gain on average; 11 percent productivity gain for importing firms.
  - A 10 percentage point fall in output tariffs → about 1 percent productivity gain.
  - Excluding input tariffs more than doubles the estimated effect of final-goods tariffs (omitted variable bias).

### Estimation strategy and methodology
- Production function and productivity estimation:
  - Cobb-Douglas: Y_it = A_it(τ) L_it^β_l K_it^β_k M_it^β_m (logs used for estimation).
  - Use Olley-Pakes (1996) semi-parametric estimator to address simultaneity and endogenous exit; modified to account for simultaneity between the decision to import intermediate inputs (d_it) and productivity shocks (Kasahara and Rodrigue (2004) approach).
  - First-stage: y_it = β^l l_it + β^m m_it + φ_it(I_it, k_it, d_it) + η_it; φ modeled with a fourth-order polynomial in investment, capital, and the import decision; time-dependent to allow interactions with Asian crisis years.
  - Separate production functions estimated at the three-digit ISIC level (29 industries).
- Deflators and price measures:
  - Three-digit price deflators used for all variables.
  - Imported inputs deflated with an import price deflator; domestic materials deflated with a three-digit domestic material price deflator.
- Second-stage (TFP regressions on trade policy):
  - Main specification: tfp_kit = γ_0 + α_i + α_t + γ_1 tariff_kt + γ_2 inputtariff_kt + γ_3 inputtariff_kt * FM_it + γ_4 FM_it + ε_it.
    - tariff_kt: final goods tariff at 5-digit ISIC level (simple average of HS nine-digit codes).
    - inputtariff_kt: input tariff constructed as weighted average of all tariffs with weights from cost shares at 5-digit level.
    - FM_it: firm-level indicator = 1 if imports account for more than 10 percent of total intermediate inputs (alternative specifications interact with continuous imported-input share).
  - Alternative measure: effective protection (erp_kit = (tariff_kt − a_ki,t inputtariff_kt) / (1 − a_ki,t), where a_ki,t is ratio of inputs to outputs).
- Identification and robustness:
  - OLS with firm fixed effects (α_i) and time fixed effects (α_t).
  - Olley-Pakes TFP compared with OLS and labor productivity; sample correlation between OLS and Olley-Pakes estimates = 0.78.
  - Instrumental variables (two-stage least squares) use instruments including share of unskilled labor and 1991 levels of tariffs; additional IV specifications use lagged tariffs and industry 1991 controls.
  - Fixed effects mitigate time-invariant political economy biases; experiments with NTB indicators and crisis interactions performed.

### Data construction and cleaning
- Deflators and price mapping:
  - Output deflators: wholesale price indices (WPI) from Buletin Statistik Bulanan Indikator Ekonomi (BPS); unpublished concordance maps 192 WPI commodity codes into five-digit ISIC codes.
  - Domestic input deflators: constructed by weighting final goods wholesale prices with their cost shares as intermediate inputs (cost shares from 1998 input list aggregated to five-digit ISIC; 277 manufacturing input/output table).
  - Imported input deflators: officially published import price deflators.
  - Capital deflator: weighted aggregate of imported machinery and transport goods and WPI of manufactured construction materials using SI shares at four-digit ISIC level.
- Tariff construction:
  - Input tariff constructed using HS nine-digit tariffs matched to five-digit ISIC via a partly manual concordance.
  - Correlation between final goods tariff and input tariff = 0.66 after merging with firm data (industry-level correlation 0.46).
- Panel cleaning operations:
  - Interpolation for isolated missing year reports (applies to <1% of sample).
  - Drop extreme spikes (employment growth >200%, output growth >500%).
  - Capital stock interpolation for 1996 using 1995 and 1997 values.
  - Post-TFP estimation removal of unrealistic TFP outliers; results robust to these exclusions.

### Empirical results — productivity and tariffs (benchmarks and interactions)
- Benchmark (final goods tariff only, Table 4, column 1):
  - "A fall in output tariffs of 10 percentage points increases productivity by 2.1 percent."
  - Table 4 coefficient: Output Tariff = -0.212*** (standard error 0.040).
- Including input tariffs (Table 4, column 2):
  - Output tariff effect more than halved: a 10 percentage point fall in output tariffs increases productivity by 0.8 percent (point estimate).
  - Input tariff large effect: "a 10 percentage point fall in input tariffs increases productivity by 4 percent."
  - Table 4 coefficient: Input tariff = -0.420*** (standard error 0.075) in the specification without interactions; -0.349*** when both included (standard error 0.075).
- Importer interaction (Table 4, column 3 and following):
  - Interaction Input tariff * FM = -0.831*** (standard error 0.127).
  - FM (importing firms indicator) = 0.086*** (standard error 0.018).
  - Interpretation: a 10 percentage point fall in input tariffs → 11 percent productivity gain for importing firms; 3 percent for non-importing firms (text calculation).
  - Continuous import share interaction (column 6): Input tariff * import share = -1.548*** (standard error 0.223); import share coefficient = 0.170*** (standard error 0.033). Mean import share for importing firms = 0.46 → effect consistent with interaction specification.
- Robustness and diagnostics:
  - Herfindahl index negative but insignificant; Exit dummy = -0.025*** indicating exiting plants are on average 2.5% less productive.
  - Baltagi and Wu (1999) correction for first-order serial correlation yields similar estimates.
  - Five-period and 10-period differencing produce broadly similar patterns (10-period difference: 10 percentage point fall in output tariffs → 4 percent productivity increase; 10 percentage point fall in input tariffs → 4.7 percent increase on average and 13 percent for importing firms).
  - OLS TFP and labor productivity regressions show similar qualitative patterns; interaction with importers remains important.

### Results — effective protection, endogeneity, and IV estimates
- Effective protection (Table 5):
  - Erp coefficient negative and significant: Erp = -0.150*** (column 1; standard error 0.011).
  - Using mean input share = 0.52, combined direct and indirect effects of output tariff reductions yield total effect close to -0.17 (text), similar to -0.21 from Table 4.
  - Total effect of reducing input tariffs on importing firms (direct + indirect via erp) equals -1.1, close to Table 4 results.
- Endogeneity diagnostics:
  - Regressing tariff change (1991–2001) on initial tariffs in 1991: coefficient = −0.58 (t-stat=8.26, R-squared=0.2). Excluding four product groups with tariff increases: coefficient = −0.69 (t-stat=42.7, R-squared=0.88).
  - Instruments: 1991 levels of output tariff, input tariff, input tariff*FM, exclusion dummy; lagged tariffs and industry 1991 controls used in alternative specifications.
- IV results (Table 9, five-period difference specification):
  - Shea partial R-squared reported (example column 1): Output tariff 0.43, Input tariff 0.44, Input tariff*import share 0.46.
  - Overidentification tests: Hansen J p-values range from 0.58 to 0.69 (text; Table 9 reports χ2(1)=0.70, χ2(1)=0.68, χ2(3)=0.58).
  - IV estimates larger for output tariffs and interaction effects: Output tariff coefficients in ln(TFP i,t)-ln(TFP i,t-5) IVs ≈ -0.545***, -0.511***, -0.495*** (columns 1–3 reported in Table 9); interaction effects for importing firms higher (text: 1.4 to 1.5 range).
  - Under IV, input tariff effect on non-importing firms is lower or insignificant in some specifications.

### Quantitative interpretation and incidence
- Average changes and implied productivity gains (text calculations):
  - Output tariffs fell on average by 16.6 percentage points between 1991 and 2001 → translates to a productivity gain of 1.5 percent based on column 6 of Table 4.
  - Input tariffs fell on average by 8.3 percentage points → translates to a productivity gain of 2.9 percent for all firms and 8.8 percent for importing firms.
  - Some output tariffs fell by as much as 46 percentage points → associated total productivity gain of 4.1 percent.
  - Some input tariffs fell by as much as 29 percentage points → leading to a gain of 10 percent for all firms, and 31 percent for importing firms over the sample period (annual average increase of 3.1 percent).
- Sample incidence:
  - Importing firms constitute a minority: FM mean = 0.145 (19 percent in text); mean import share for importing firms = 0.46.
  - Mean input share used in erp calculations = 0.52.

### Interpretation of channels and additional findings
- Mechanisms:
  - Input-tariff reductions increase productivity primarily through importing firms via access to foreign technology embodied in inputs, higher-quality inputs, and variety/learning effects.
  - Output tariff reductions operate mainly through import competition, trimming inefficiency and producing modest productivity gains.
  - Indirect spillovers: importing firms can pass benefits along vertical chains to non-importing firms; indirect effects expected smaller than direct importer gains.
- Omitted-variable concern:
  - Excluding input tariffs from regressions leads to overstatement of the competition effect of final-goods tariff reductions (coefficients more than double when input tariffs omitted).
- Crisis and other controls:
  - Interactions with Asian crisis (1997–1998) leave key tariff coefficients unchanged; some specifications show Input tariff * crisis dummy effects indicating non-importing firms performed relatively better during the crisis.
  - Export share generally negative and weakly significant in some specs (e.g., Share of exports = -0.015* with standard error 0.008); foreign share insignificant.

### Policy implications and conclusions
- Core policy-relevant findings:
  - "A 10 percentage point fall in input tariffs increases productivity by 3 percent, whereas an equivalent fall in output tariffs increases productivity by a little less than 1 percent."
  - "A 10 percentage point fall in input tariffs leads to a productivity gain of 11 percent for importing firms."
  - Including input tariffs is essential to avoid omitted-variable bias when estimating productivity effects of trade liberalization.
- Policy recommendations and considerations:
  - Trade liberalization that reduces input tariffs can yield substantial productivity gains, disproportionately benefiting firms that import intermediate inputs.
  - Policy assessments focusing only on final-goods tariff reductions likely understate total productivity gains and may misattribute gains to import competition instead of improved access to foreign intermediates.
  - Disaggregated tariff analysis is essential due to wide variation across and within industries and escalating tariff structures along production chains.
  - Facilitating firms’ access to imported intermediate inputs (reducing input tariffs and import barriers) is a direct channel to enhance firm-level productivity and foster diffusion of foreign technology.
  - Caution: welfare implications depend on broader factors (entry, location of firms, agglomeration); complementary reforms to increase resource mobility could amplify gains from reduced output tariffs.

### Appendix I — methodology and data details
- Plant-level data cleaning:
  - Interpolation for isolated missing values (<1% of sample); drop extreme spikes (employment growth >200%, output growth >500%); capital stock interpolation for 1996 using 1995 and 1997; removal of extreme TFP outliers post-estimation.
- Deflators:
  - Output deflators from WPI mapped to five-digit ISIC via unpublished concordance.
  - Domestic material deflators constructed from weighted final goods WPI using 1998 cost shares aggregated to five-digit ISIC; input mix assumed constant 1991–2001.
  - Imported input deflators from official import price deflators.
  - Capital deflator constructed from aggregate imported machinery and transport goods indices and WPI of manufactured construction materials, weighted by SI shares at four-digit ISIC.

*Source: _wp05146 — Input Price Indices in Indonesia, 1991–2001 (IMF Working Paper).*

### 1. Input Price Indices in Indonesia, 1991–2001............................................................20

### 1. Input Price Indices in Indonesia, 1991–2001............................................................20

### Major content headings in this unit
- 1. Input Price Indices in Indonesia, 1991–2001............................................................20
- 2. Average Total Factor Productivity in Indonesia, 1991–2001 ...................................20
- 3. Change in Tariffs (1991–2001) Relative to Initial Levels ........................................21

### Tables listed in this unit
- 1. Coefficients of the Production Function ...................................................................22
- 2. Tariffs in Indonesia, 1991–2001 ...............................................................................23
- 3.      Summary      Statistics....................................................................................................23
- 4.      Basic      Results .............................................................................................................24
- 5. Effective Protection Rate ..........................................................................................25
- 6.      Additional      Controls...................................................................................................26
- 7.      Alternative      Productivity      Measures ...........................................................................27
- 8.      Alternative      Econometric      Specifications ...................................................................28
- 9.      Endogeneity ..............................................................................................................29      

### Data appendix contents
- 1.      Plant-Level      Data      and cleaning operations ................................................................30
- 2.      Deflators....................................................................................................................30

*Source: _wp05146 - 1. Input Price Indices in Indonesia, 1991–2001............................................................20*

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

### _wp05146 - References

### Introduction: scope, data, and key empirical patterns
- Objective: disentangle productivity gains from reductions in final goods tariffs versus intermediate input tariffs using Indonesian plant-level data.
- Data:
  - Manufacturing census covering 1991 to 2001 for all plants with 20 or more employees.
  - Plant-level information on output, employment, ownership, exports, and imports.
  - Imported inputs observed at the plant level (1998 detailed input list used to construct input weights; assume constant technology over sample).
- Tariff patterns and reform:
  - Escalating tariff structure observed (example ISIC codes for 2001): motor vehicle bodies 0 percent; motor vehicle components 11 percent; motor vehicles 31.6 percent.
  - Final goods tariffs fell from an average of 21 percent in 1991 to an average of 8 percent in 2001; some tariffs remain as high as 170 percent.
  - WTO commitment starting 1995 to reduce all bound tariffs to 40 percent or less.
- Key empirical finding preview:
  - A 10 percentage point fall in input tariffs → 3 percent productivity gain on average; 11 percent productivity gain for importing firms.
  - A 10 percentage point fall in output tariffs → about 1 percent productivity gain.
  - Excluding input tariffs in regressions more than doubles the estimated effect of final-goods tariffs (omitted variable bias).

### Estimation strategy and methodology
- Production function:
  - Cobb-Douglas production function: Y_it = A_it(τ) L_it^β_l K_it^β_k M_it^β_m (logs used for estimation).
- Productivity estimation:
  - Use Olley-Pakes (1996) semi-parametric estimator to address simultaneity between inputs and unobserved productivity shocks and to control for endogenous exit.
  - Modification: account for simultaneity between the decision to import intermediate inputs (d_it) and productivity shocks as in Kasahara and Rodrigue (2004).
  - First-stage specification: partial linear model y_it = β^l l_it + β^m m_it + φ_it(I_it, k_it, d_it) + η_it; φ modeled with a fourth-order polynomial in investment, capital, and the import decision; function made time-dependent to allow interactions with the Asian crisis years.
  - Separate production functions estimated at the three-digit ISIC level (29 industries).
  - Deflators: three-digit price deflators used for all variables; imported inputs deflated with an import price deflator; domestic materials deflated with a three-digit domestic material price deflator.
- Robustness:
  - Compare Olley-Pakes TFP with OLS and labor productivity estimates; sample correlation between OLS and Olley-Pakes estimates = 0.78.
  - Note divergence post-1997: OLS TFP shows larger increases after 1997 relative to Olley-Pakes, attributed to OLS not modelling exit or crisis effects.
  - Average TFP (Olley-Pakes measure) increased from 1991 and peaks first in 1998 and later in 2001, with average TFP 18.7 percentage points higher relative to 1991.
- Second stage (TFP regressions on trade policy):
  - Main specification: tfp_kit = γ_0 + α_i + α_t + γ_1 tariff_kt + γ_2 inputtariff_kt + γ_3 inputtariff_kt * FM_it + γ_4 FM_it + ε_it.
    - tariff_kt: final goods tariff at 5-digit ISIC level.
    - inputtariff_kt: input tariff constructed as weighted average of all tariffs with weights from cost shares at 5-digit level.
    - FM_it: firm-level indicator = 1 if imports account for more than 10 percent of total intermediate inputs (also alternative specifications interacting with actual imported-input share).
  - Hypotheses (signs):
    - γ_1 expected negative (lower final goods tariffs → increased import competition → productivity improvement).
    - γ_2 and γ_3 expected negative (lower input tariffs → higher productivity, especially for importing firms).
    - γ_4 expected positive (imported inputs generate technological externalities).
  - Alternative measure: effective protection (erp_kit = (tariff_kt − a_ki,t inputtariff_kt) / (1 − a_ki,t), where a_ki,t is ratio of inputs to outputs for firm i in industry k at time t).
- Estimation details:
  - Equation 6 estimated with OLS including firm fixed effects (α_i) and time fixed effects (α_t).
  - Two-stage least squares robustness checks use instruments including share of unskilled labor and 1991 levels of tariffs.

### Empirical results and interpretation
- Main quantitative results:
  - Input tariffs:
    - 10 percentage point fall in input tariffs → 3 percent productivity gain on average.
    - 10 percentage point fall in input tariffs → 11 percent productivity gain for importing firms (FM = 1).
  - Output (final goods) tariffs:
    - 10 percentage point fall in output tariffs → about 1 percent productivity gain (interpreted as import-competition effect).
  - Omitted variable issue:
    - Regressing productivity only on final goods tariffs (excluding input tariffs) more than doubles the estimated effect of final-goods tariffs, indicating possible omitted variable bias and underestimation of total effects when input tariffs are excluded.
- Interpretation of channels:
  - Input-tariff reductions increase productivity primarily through importing firms via access to foreign technology embodied in inputs, higher-quality inputs, and variety/learning effects.
  - Lower output tariffs increase import competition, trimming inefficiency and producing modest productivity gains.
  - Possible indirect spillovers: importing firms can pass benefits along vertical production chains to non-importing firms; these indirect effects expected smaller than direct benefits to importers.
  - Effective protection framework considered: lower effective protection hypothesized to increase productivity, but empirically lower input-tariff benefits are expected to dominate potential negative competition effects.
- Additional empirical diagnostics:
  - Instrument relevance: regressing tariff change (1991–2001) on initial tariffs in 1991 yields coefficient −0.58 (t-stat=8.26, R-squared=0.2). Excluding four product groups with tariff increases, coefficient = −0.69 (t-stat=42.7, R-squared=0.88), suggesting initial tariff levels predict tariff changes.

### Trade policy context and identification considerations
- Indonesia WTO accession and commitments:
  - WTO member on January 1, 1995, with a commitment (starting 1995) to reduce all bound tariffs to 40 percent or less over a 10-year period (subject to an exclusion list).
  - 73 five-digit ISIC codes included at least one excluded HS code; 9 ISIC codes contained 10 or more excluded HS codes; exclusions concentrated in motor vehicles and components and iron and steel basic industries.
- Endogeneity concerns and mitigation:
  - Political economy concerns: literature suggests industries may lobby for protection; however, Mobarak and Purbasari (2005) find political connections do not affect tariff rates in Indonesia; politically connected firms receive import rights, but such firms account for about 1 percent of firms.
  - Fixed effects mitigate bias from time-invariant political economy factors.
  - Instruments and robustness checks (unskilled labor share; initial tariff levels) used to address potential time-varying endogeneity.
- Non-tariff barriers (NTBs):
  - NTBs are acknowledged as an important but hard-to-measure form of protection; an NTB indicator post-1995 for five-digit product codes where at least one HS nine-digit product was listed as having an NTB removal commitment was experimented with.

### Policy implications and conclusions drawn in the text
- Main policy message:
  - Trade liberalization that reduces input tariffs can yield substantial productivity gains, especially for firms that import intermediate inputs.
  - Policy assessments focusing only on final-goods tariff reductions likely understate total productivity gains from liberalization and may misattribute gains to import competition instead of improved access to foreign intermediates.
- Practical policy considerations:
  - Disaggregated tariff analysis is essential due to wide variation across and within industries and escalating tariff structures along production chains.
  - Facilitating firms’ access to imported intermediate inputs (reducing input tariffs and import barriers) constitutes a direct channel to enhance firm-level productivity and foster diffusion of foreign technology into domestic industries.

*Source: _wp05146 - References.*

### 1995. There were 17 such 5-digit product codes. However, we found that this had an insignificant

### _wp05146 - 1995. There were 17 such 5-digit product codes. However, we found that this had an insignificant

### Data
- Main data source: Manufacturing Survey of Large and Medium-sized firms (Survei Industri, SI) for 1991 to 2001.
- Sample: unbalanced panel of around 10,000 firms per year with a total of 116,121 observations.
- Key measured variables and construction:
  - Outputs and inputs deflated by wholesale price indices to obtain productivity estimates.
  - Domestic input deflators constructed by weighting final goods wholesale prices with their cost shares as intermediate inputs.
  - Officially published import price deflators used for imported inputs.
  - 1998 firm-level listing of individual intermediate inputs aggregated within five-digit industry categories to provide a 277 manufacturing input/output table (assumed input mix constant over sample).
  - Input tariff constructed using HS nine-digit tariffs matched to five-digit ISIC via an (partly manual) concordance.
  - Simple average of HS nine-digit codes used to construct final goods tariff for each five-digit industry.
- Correlation: correlation between final goods tariffs and input tariff is 0.66 after merging with firm data (industry-level correlation 0.46).
- Notes on data cleaning and coverage: start year 1991 (avoid reclassification issues), capital stock completeness improved after 1991; cleaning described in appendix.

### Results — Productivity and Tariffs
- Benchmark (final goods tariff only, column 1 of Table 4):
  - "A fall in output tariffs of 10 percentage points increases productivity by 2.1 percent."
  - Table 4 coefficient: Output Tariff = -0.212*** (standard error 0.040).
- Including input tariffs (column 2 of Table 4):
  - Output tariff effect more than halved and less statistically significant: point estimate suggests a 10 percentage point fall in output tariffs increases productivity by 0.8 percent.
  - Input tariffs much larger: "a 10 percentage point fall in input tariffs increases productivity by 4 percent."
  - Table 4 coefficient: Input tariff = -0.420*** (standard error 0.075) in specification without interactions; -0.349*** when both included (standard error 0.075).
- Importer interaction (column 3 and following):
  - Interaction Input tariff * FM (FM=1 if import share>0.1) = -0.831*** (standard error 0.127).
  - FM (importing firms indicator) = 0.086*** (standard error 0.018).
  - Interpretation: adding interaction to input tariff effect implies a 10 percentage point fall in input tariffs improves productivity for importing firms by 11 percent, whereas non-importing firms benefit by 3 percent (text calculation).
  - Using continuous import share interaction (column 6): Input tariff * import share = -1.548*** (standard error 0.223); import share coefficient = 0.170*** (standard error 0.033). Multiplying interaction by mean import share for importing firms (equal to 0.46) yields effect ≈0.71 (text).
- Robustness checks in Table 4:
  - Herfindahl index inclusion: negative but insignificant; Exit dummy (exit in next period) = -0.025*** indicating least productive plants exit (firms that exit are on average 2.5% less productive).
  - Column 7 uses Baltagi and Wu (1999) correction for first-order serial correlation; estimates similar to column 6.

### Results — Effective Protection Rates
- Inclusion of effective rate of protection (erp) (Table 5):
  - Erp coefficient negative and significant: Erp = -0.150*** (column 1; standard error 0.011).
  - A fall in effective protection (which could stem from lower output tariffs, higher input tariffs, or changes in input intensity) leads to an increase in productivity; effect persists after controlling for output or input tariffs.
  - Calculations in text: using coefficients in column 2 and mean input share equal to 0.52, total effect of reducing output tariffs via direct plus indirect (through erp) equals -0.17, close to -0.21 from Table 4.
  - Total effect of reducing input tariffs on importing firms (direct plus indirect via erp) equals –1.1, close to Table 4 results.

### Results — Additional Controls
- Table 6: adding export share and foreign ownership:
  - Share of exports generally negative and significant only at 10 percent in some specs (example: Share of exports = -0.015* with standard error 0.008).
  - Foreign share insignificant.
- Asian crisis (crisis dummy = 1 for 1997 and 1998):
  - Interactions of tariffs with crisis dummy leave key coefficients on final goods and input tariffs unchanged — results robust to crisis.
  - Interaction: Input tariff * crisis dummy shows additional effect (negative and occasionally significant), indicating non-importing firms performed relatively better during the Asian crisis (text discussion).
  - Some specifications show Input tariff * crisis dummy * importing firms positive and significant (e.g., 0.392*** and 0.632*** in Table 6 columns).

### Results — Alternative Specifications
- Alternative productivity measures (Table 7):
  - Using OLS TFP: output tariff effect becomes insignificant once input tariffs controlled, input tariff effect magnitude similar.
  - Labor productivity regressions (ln value added per worker): reducing output tariffs increases labor productivity; inclusion of capital per worker does not change these results materially.
  - Interaction patterns with importing firms remain consistent: importing firms enjoy larger gains.
- Econometric variants (Table 8):
  - Five-period differencing and 10-period differencing produce broadly similar patterns.
  - 10-period difference (2001 less 1991) results: a 10 percentage point fall in output tariffs increases productivity by 4 percent; a 10 percentage point fall in input tariffs associated with a 4.7 percent increase in productivity on average and a 13 percent increase for importing firms (text).
  - First-difference specifications reduce coefficients and significance (possible measurement error induced by differencing).

### Results — Endogeneity (Instrumental Variables)
- Table 9: Instrumenting for output tariffs, input tariffs, and input tariff interacted with import shares in the five-period difference model.
- Instruments used include: 1991 levels of output tariff, input tariff, input tariff*FM, and an exclusion dummy; or t-5 lagged tariffs; and additional industry 1991 controls in column 3.
- First-stage fit: Shea partial R-squared reported (e.g., Output tariff 0.43, Input tariff 0.44, Input tariff*import share 0.46 in column 1).
- Overidentification tests: p-values (Hansen J statistic) range from 0.58 to 0.69 in text (Table 9 reports χ2(1)=0.70, χ2(1)=0.68, χ2(3)=0.58).
- IV results:
  - Output tariff coefficients larger (range 0.49 to 0.55 in text summary; Table 9 reports Output tariff = -0.545***, -0.511***, -0.495*** in columns 1–3 for ln(TFP i,t)-ln(TFP i,t-5) specification).
  - Interaction effects for importing firms higher (1.4 to 1.5 in text summary compared with OLS 1.1).
  - Input tariff effect on non-importing firms lower or insignificant under IV (text).

### Interpretation of Results (numerical summary)
- Average tariff changes and associated productivity effects (text):
  - Output tariffs fell on average by 16.6 percentage points between 1991 and 2001 → translates to a productivity gain of 1.5 percent based on column 6 of Table 4.
  - Input tariffs fell on average by 8.3 percentage points → translates to a productivity gain of 2.9 percent for all firms and 8.8 percent for importing firms.
  - Some output tariffs fell by as much as 46 percentage points → associated total productivity gain of 4.1 percent.
  - Some input tariffs fell by as much as 29 percentage points → leading to a gain of 10 percent for all firms, and by 31 percent for importing firms over the sample period (annual average increase of 3.1 percent).
- Sample and incidence:
  - Importing firms constitute only 19 percent of the sample (text); summary statistics show FM mean = 0.145 and FM=1 if imported inputs>0.1 mean 0.145 in Table 3.
  - Mean input share used in erp calculations equal to 0.52 (text).
  - Mean import share for importing firms equal to 0.46 (text and Table 3: Share of imported inputs if >0 mean 0.458).
- Key interpretation points:
  - Reducing input tariffs yields larger productivity gains than reducing output tariffs, especially for importing firms (consistent across specifications).
  - Excluding input tariffs from analysis can produce omitted-variable bias, overstating the competition effect of lower output tariffs.
  - Gains from lower output tariffs (import competition) appear muted relative to expected because resource mobility across sectors in a regulated economy like Indonesia may be limited.

### Conclusions and Policy-relevant Findings
- Core empirical findings:
  - "A 10 percentage point fall in input tariffs increases productivity by 3 percent, whereas an equivalent fall in output tariffs increases productivity by a little less than 1 percent." (text; headline numerical result).
  - "A 10 percentage point fall in input tariffs leads to a productivity gain of 11 percent for importing firms." (text).
  - Including input tariffs is essential to avoid omitted-variable bias when estimating the productivity effect of output tariff reductions.
- Robustness:
  - Results robust to: Olley-Pakes TFP estimation, OLS TFP, labor productivity measures, five-period differencing, 10-period differencing, instrumental variables, additional controls (Herfindahl, export share, foreign ownership), and accounting for the Asian crisis.
- Policy implications highlighted in text:
  - Trade reform that reduces input tariffs can deliver substantial productivity gains, disproportionately benefiting firms that import inputs (importing firms are a minority of firms).
  - Caution: higher productivity gains from reducing input tariffs should not be taken as an endorsement of escalating tariff structures; welfare effects depend on wider factors (entry, location of firms, agglomeration), and alternative tariff structures (e.g., uniform lower final goods tariffs) could yield different welfare and agglomeration outcomes.
  - In regulated economies, constraints on resource reallocation may limit the potential productivity gains from import competition; complementary reforms to increase resource mobility could unlock further gains.

*Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05146.pdf*

### APPENDIX                                                             I

### APPENDIX I

### Methodology and Data

- Plant-Level Data and Cleaning Operations
  - Data cleaning objectives: minimize noise from nonreporting, misreporting, and typing mistakes.
  - Three key adjustments applied to plant-level panel:
    - Interpolation: filled gaps when a plant reported no value in a given year but reported values in the year prior and the year after the missing one. This applies to less than 1% of the sample.
    - Drop extreme spikes: plants with unrealistically large spikes were dropped (examples given: employment growth of more than 200%, growth in output of more than 500%).
    - Capital stock interpolation: capital stock measured by replacement value of fixed assets was missing for the year 1996; interpolated the capital stock for 1996 using the 1997 and 1995 values.
  - Post-TFP estimation cleaning: a small number of plants with unrealistically high or unrealistically low TFP levels were dropped; results were not sensitive to dropping these observations.

- Deflators
  - Output deflators:
    - Used wholesale price indices (WPI) published monthly in the Buletin Statistik Bulanan Indikator Ekonomi of the Indonesian Statistical Agency (Badan Pusat Statistic, BPS), the Monthly Statistical Bulletin of Economic Indicators.
    - Used an unpublished concordance from the BPS to map the 192 WPI commodity codes into the five-digit ISIC product codes.
    - Where some ISIC codes mapped into more than one commodity code, took the average price of those to obtain a price index at the five-digit ISIC product code.
  - Material input deflators:
    - Domestic input deflators constructed by weighting final goods wholesale prices with their cost shares as intermediate inputs.
    - Cost shares obtained from the list of firm self-reported inputs (only available in 1998), aggregated within five-digit ISIC industry categories to provide a 277 manufacturing input/output table.
    - Assumed the mix of inputs used by industries does not change over the sample period, 1991 to 2001.
    - Imported input deflators are the officially published import price deflators for imported inputs.
  - Capital deflator:
    - Constructed using the aggregate price index of imported electrical and non-electrical machinery and equipment, imported transport goods, and the wholesale price index of manufactured construction materials.
    - Used information from the SI to compute the shares of vehicles, buildings and equipment at the four-digit ISIC level.
    - Those shares are used to weight each individual aggregate deflator to obtain a capital deflator at the sector level.

*Source: APPENDIX I, _wp05146 - APPENDIX*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05146.pdf_