## Appendix 5. First-Stage Regressions

## Source details

**Canonical URL:** [Appendix 5. First-Stage Regressions](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019260-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wpiea2019260-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019260-print-pdf.pdf.json)

---

### Theoretical framework (summary)
- Model: dual-economy choice between tax compliant (formal) and fully non-compliant (informal) firms.
- Firm output: e_i = alpha_i A.
- Formal profits: G^F_i = alpha_i A (1−T) − C.
- Informal expected profits: E(G^I_i) = alpha_i b A (1−e), where 0 < b < 1 and e is probability of being caught.
- Cutoff productivity s* solves s* A (1−T) − C = s* b A (1−e), so
  - s* = 1 / { A [1−T − b (1−e)] − C }.
- Comparative implications:
  - Cutoff productivity (and share of non-compliant firms) decreases in A and e; increases in T and C.
  - Responsiveness of cutoff to T, C, e varies with A (less rapid changes in more productive economies).

### Data and measurement
- Micro data: World Bank Enterprise Surveys for 47 emerging and developing economies (manufacturing and services).
- Main dependent variable: percent of sales reported for tax purposes (survey question: “what percentage of total sales would you estimate the typical establishment in your area of activity reports for tax purposes?”).
- Productivity measures: sales per worker (labor productivity, log), value added per worker (log), TFP (Levinsohn and Petrin, 2003 estimator).
- Sample sizes:
  - OLS (Table 1): Observations 31,584; # of firms 31,241; # of countries 87; # of industries 21.
  - IV baseline: Observations 11,865; # of firms 11,741; # of countries 47; # of industries 18.
- Descriptive statistics (IV baseline sample, selected):
  - Sales reported: MIN 0, p25 70, p75 100, MAX 100, Average 81.95, Std. Dev. 28.07, Number 11,865
  - Labor productivity: MIN -2.33, p25 7.49, p75 10.45, MAX 19.34, Average 8.73, Std. Dev. 2.47, Number 11,865
  - Age: MIN 0, p25 1.95, p75 3.14, MAX 4.61, Average 2.55, Std. Dev. 0.81, Number 11,865
  - Exporter: MIN 0, p25 0, p75 0, MAX 1, Average 0.20, Std. Dev. 0.40, Number 11,865
  - Perception of tax admin.: MIN 0, p25 0, p75 1.33, MAX 4, Average 0.93, Std. Dev. 0.76, Number 11,865

### Instrument strategy (first-stage construction)
- Two instruments used jointly:
  1. Regional IV (IV-1): log of average productivity of same industry in other subnational region(s) of the same country (other-region industry-year average).
     - Rationale: industry-level shocks common across regions while tax-evasion enforcement is region-specific.
     - Cross-regional variation in percent of sales reported: range around 8 to 100 percent; standard deviation 19 percent (full sample sd 28 percent). Average within-country region difference ≈ 7 percent.
  2. International IV (IV-2): log of average productivity in same industry in up to two much larger neighboring countries in same World Bank region (omitting the two largest economies).
     - Matching rules: instrument countries must have GDP at least five times larger; commodity-exporter status matched (commodity exporters defined as exports of oil, gas, metals representing 20 percent or more of total exports or fiscal revenues).
     - Robustness alternatives: relax size threshold to 4 times larger; pair irrespective of commodity-exporter status; exclude advanced economies.

### First-stage evidence and statistics
- First-stage relevance: industry-level average productivity correlated across paired countries/regions.
- Kleibergen-Paap F-statistics (reported in IV regressions) exceed weak-instrument thresholds:
  - Table 2: 18.18, 16.15, 12.93, 19.33 (across specifications).
  - Table 3: 17.39, 17.65, 16.90, 19.20 (robustness specifications).
- Hansen J. p-values (overidentification tests) reported:
  - Table 2: 0.29, 0.39, 0.10, 0.29.
  - Table 3: 0.86, 0.29, 0.30, 0.33.

### First-stage regression coefficients (selected exact estimates)
- First Stage of Table 2 — Dependent variable: Labor productivity [log] (regional and international IV coefficients with standard errors)
  - Regional IV for labor prod. [log]: 0.179*** [0.067], 0.175** [0.073], 0.179*** [0.043], 0.174*** [0.066] (columns (1)–(4))
  - International IV for labor prod. [log]: 0.250*** [0.042], 0.151*** [0.030], 0.250*** [0.050], 0.249*** [0.040] (columns (1)–(4))
  - Covariates (examples):
    - Age of firm [log]: 0.053** [0.024], 0.054** [0.024], 0.053*** [0.019]
    - Exporter [dummy]: 0.340*** [0.073], 0.345*** [0.070], 0.340*** [0.046], 0.341*** [0.070]
    - Foreign-owned [dummy]: 0.428*** [0.086], 0.401*** [0.082], 0.428*** [0.061], 0.419*** [0.082]
    - Perception of tax admin.: 0.048*** [0.015], 0.048*** [0.014], 0.048*** [0.015]
  - Observations and sample counts (columns (1)–(4)):
    - Observations: 11,865; 13,208; 11,865; 11,869
    - # of firms: 11,741; 13,056; 11,741; 11,745
    - # of countries: 47; 47; 47; 47
    - # of industries: 18; 18; 18; 18
  - Note: Errors in brackets clustered at the region and industry-year level except spec. 3 where clustering is at the country-industry level.

- First Stage of Table 3 — Dependent variable: Labor productivity [log] (alternate samples/specifications)
  - Regional IV for labor prod. [log]: 0.185*** [0.031], 0.181*** [0.067], 0.187*** [0.070], 0.208*** [0.078] (columns (1)–(4))
  - International IV for labor prod. [log]: 0.130 [0.091], 0.251*** [0.042], 0.255*** [0.044], 0.254*** [0.042] (columns (1)–(4))
  - Covariates (examples):
    - Age of firm [log]: 0.005 [0.049], 0.057** [0.025], 0.062** [0.025], 0.050** [0.025]
    - Exporter [dummy]: 0.408*** [0.058], 0.365*** [0.081], 0.366*** [0.076], 0.365*** [0.075]
    - Foreign-owned [dummy]: 0.446*** [0.073], 0.486*** [0.083], 0.540*** [0.089], 0.465*** [0.097]
    - Perception of tax admin.: -0.064 [0.084], 0.048*** [0.017], 0.068*** [0.020], 0.053*** [0.015]
  - Observations and sample counts (columns (1)–(4)):
    - Observations: 13,618; 9,980; 13,937; 10,198
    - # of firms: 13,464; 9,856; 13,725; 10,079
    - # of countries: 50; 38; 54; 46
    - # of industries: 18; 18; 20; 17
  - Note: Errors in brackets clustered at the region and industry-year level.

### Key empirical findings (OLS and IV estimates preserved exactly)
- OLS (selected coefficients from Table 1):
  - Labor productivity [log]:
    - Spec (1): 1.443*** [0.242]
    - Spec (2): 1.144*** [0.368]
    - Spec (3, value-added per worker): 2.011*** [0.346]
    - Spec (5, TFP): 1.789*** [0.588]
  - Other OLS:
    - Exporter [dummy]: 2.406*** [0.540] (Spec (1) example)
    - Small [dummy]: -3.001*** [0.690] (Spec (1) example)
  - OLS textual baseline: a 1 percent increase in firm productivity results in 0.01 percentage points increase in share of sales reported (baseline OLS statement in text).

- IV (Table 2 baseline and Table 3 robustness) — labor productivity effects on percent of sales reported:
  - Table 2 (Labor productivity [log]):
    - Spec (1): 12.74** [6.330]
    - Spec (2): 20.96** [10.50]
    - Spec (3): 12.74*** [1.456]
    - Spec (4): 12.12* [6.286]
  - Table 2 Kleibergen-Paap F stat: 18.18, 16.15, 12.93, 19.33 (corresponding specs).
  - Table 2 Hansen J. p-val.: 0.29, 0.39, 0.10, 0.29.
  - Table 3 robustness (Labor productivity [log]):
    - Spec (1): 17.70* [10.00] (IV-2 uses countries at least 4 times larger)
    - Spec (2): 12.74** [6.377] (excluding advanced economies)
    - Spec (3): 12.64* [6.530] (pair irrespective of commodity-exporter status)
    - Spec (4): 12.42* [7.211] (IV-1 constructed only if ≥ 5 firms in other region)
  - Table 3 Kleibergen-Paap F stat: 17.39, 17.65, 16.90, 19.20.
  - Table 3 Hansen J. p-val.: 0.86, 0.29, 0.30, 0.33.
- Textual IV interpretation preserved exactly: “Our baseline estimates suggest that an increase in productivity of 1 percent increases the sales reported to tax administrations by around 0.12 percent, indicating a response elasticity of more than ten percent.”
  - Note: paper reports both scaled elasticity (“0.12 percent” increase in sales reported per 1 percent productivity increase) and raw IV coefficients (e.g., 12.74).

### Interpretation, mechanisms, and robustness
- Causal claim: increases in firm-level productivity causally increase the share of sales firms report for tax purposes.
- Estimated magnitude: baseline IV point estimates imply a sizable economic effect; textual statement: “an increase in productivity of 1 percent increases the sales reported to tax administrations by around 0.12 percent.”
- Robustness: results hold across alternative productivity measures (value-added per worker, TFP), clustering approaches, and multiple IV construction choices.
- Instrument quality:
  - First-stage relevance: Kleibergen-Paap F-statistics well above 10 (Staiger and Stock, 1997 criterion).
  - Overidentification: Hansen J tests yield p-values ≥ 0.1 in main specs, failing to reject instrument validity.

### Policy implications and caveats
- Policy implications:
  - Productivity-enhancing policies can yield tax-compliance gains: improving firm productivity can increase tax revenues by raising the share of sales reported.
  - Productivity policies can complement tax-policy and tax-administration reforms aimed at reducing tax evasion.
- Caveats and limitations:
  - IV strategy relies on industry-level productivity changes which may capture entry/exit and reallocation rather than pure productivity shocks.
  - Limited panel dimension and reliance on self-reported percent of sales reported as the tax-evasion measure.
  - Authors recommend further research to examine enforcement narratives and mechanisms linking productivity to compliance.

*Source: wpiea2019260-print-pdf - Appendix 5. First-Stage Regressions*

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

### References

### Appendix

- Appendix 1. Macroeconomic Stylized Facts: Details of the Estimation and Robustness Checks ....    25

- Appendix 2. Derivations and Model Extension ................................................................................................... 29

- Appendix 3. Variable Definitions and Descriptive Statistics .......................................................................... 30

- Appendix 4. Regional and International Instrument Construction ............................................................. 31

*Source: wpiea2019260-print-pdf - References .............................................................................................................*

### Appendix 5. First-Stage Regressions ....................................................................................

### wpiea2019260-print-pdf - Appendix 5. First-Stage Regressions

### Theoretical framework (summary)
- Dual-economy model where firms choose between being tax compliant (formal) or fully non-compliant (informal).
- Firm output: e_i = alpha_i A.
- Formal profits: G^F_i = alpha_i A (1−T) − C.
- Informal expected profits: E(G^I_i) = alpha_i b A (1−e), where 0 < b < 1 and e is the probability of being caught.
- Cutoff productivity s* solves s* A (1−T) − C = s* b A (1−e), so
  - s* = 1 / { A [1−T − b (1−e)] − C }.
- Implications:
  - The cutoff productivity value (and share of non-compliant firms) decreases in A and e; increases in T and C.
  - The responsiveness of the cutoff to changes in T, C, and e varies with A (less rapid changes in a more productive economy).

### Data and measurement
- Micro data: World Bank Enterprise Surveys for 47 emerging and developing economies (manufacturing and services).
- Main dependent variable: percent of sales reported for tax purposes (survey question: “what percentage of total sales would you estimate the typical establishment in your area of activity reports for tax purposes?”).
- Productivity measures used: sales per worker (labor productivity, log), value added per worker (log), and TFP (Levinsohn and Petrin, 2003 estimator).
- Sample specifics:
  - Table 1 (OLS) sample sizes: Observations 31,584; # of firms 31,241; # of countries 87; # of industries 21.
  - IV baseline sample sizes: Observations 11,865; # of firms 11,741; # of countries 47; # of industries 18.

### Instrument strategy (first-stage construction)
- Two instruments exploited jointly:
  1. Regional IV (IV-1): log of average productivity of the same industry in other subnational region(s) of the same country (other-region industry-year average).
     - Rationale: industry-level shocks common across regions but tax-evasion enforcement is region-specific.
     - Cross-regional variation in percent of sales reported: range around 8 to 100 percent; standard deviation 19 percent (full sample sd 28 percent). Average within-country region difference ≈ 7 percent.
  2. International IV (IV-2): log of average productivity in the same industry in up to two much larger neighboring countries in the same World Bank region (omitting the two largest economies from the sample).
     - Matching rules: instrument countries must have GDP at least five times larger; commodity-exporter status matched (commodity exporters defined as exports of oil, gas, metals representing 20 percent or more of total exports or fiscal revenues).
     - Alternative robustness: relax size threshold to 4 times larger, pair irrespective of commodity-exporter status, exclude advanced economies.
- First-stage evidence:
  - Industry-level average productivity is correlated across paired countries/regions (see Appendix 5 tables referenced in source).
  - First-stage Kleibergen-Paap F-statistics (reported in IV regressions) exceed weak-instrument thresholds:
    - Table 2 reported Kleibergen-Paap F stat: 18.18, 16.15, 12.93, 19.33 (across specifications).
    - Table 3 reported Kleibergen-Paap F stat: 17.39, 17.65, 16.90, 19.20 (robustness specs).

### Key empirical findings (OLS and IV)
- OLS (selected coefficients from Table 1; errors in brackets clustered at region and industry-year level):
  - Labor productivity [log]:
    - Spec (1): 1.443*** [0.242]
    - Spec (2): 1.144*** [0.368]
    - Spec (3, value-added per worker): 2.011*** [0.346]
    - Spec (5, TFP): 1.789*** [0.588]
  - Other notable OLS coefficients:
    - Exporter [dummy]: positive and significant (e.g., 2.406*** [0.540] in Spec (1)).
    - Small [dummy]: negative and significant (e.g., -3.001*** [0.690] in Spec (1)).
  - OLS interpretation: a 1 percent increase in firm productivity results in 0.01 percentage points increase in share of sales reported (baseline OLS statement in text).

- IV (baseline and robustness; Table 2 and Table 3) — labor productivity effects on percent of sales reported:
  - Table 2 (IV baseline and variants; dependent variable: Sales reported):
    - Labor productivity [log]:
      - Spec (1): 12.74** [6.330]
      - Spec (2): 20.96** [10.50]
      - Spec (3): 12.74*** [1.456]
      - Spec (4): 12.12* [6.286]
    - Kleibergen-Paap F stat: 18.18, 16.15, 12.93, 19.33 (corresponding specs).
    - Hansen J. p-val.: 0.29, 0.39, 0.10, 0.29 (cannot reject instrument validity).
    - Notes: IV-1 = average productivity in other regions of same country; IV-2 = average productivity in same industry/year in up to two countries at least 5 times larger in same WB region.
  - Textual interpretation in source: “Our baseline estimates suggest that an increase in productivity of 1 percent increases the sales reported to tax administrations by around 0.12 percent, indicating a response elasticity of more than ten percent.”
    - The paper reports both the scaled textual elasticity (“0.12 percent” increase in sales reported per 1 percent productivity increase) and the raw IV coefficients shown in Table 2 (e.g., 12.74).
  - Robustness (Table 3):
    - Labor productivity [log]:
      - Spec (1): 17.70* [10.00] (IV-2 uses countries at least 4 times larger)
      - Spec (2): 12.74** [6.377] (excluding advanced economies)
      - Spec (3): 12.64* [6.530] (pair irrespective of commodity-exporter status)
      - Spec (4): 12.42* [7.211] (IV-1 constructed only if ≥ 5 firms in other region)
    - Kleibergen-Paap F stat: 17.39, 17.65, 16.90, 19.20.
    - Hansen J. p-val.: 0.86, 0.29, 0.30, 0.33.

### Interpretation and mechanism
- Main causal claim: increases in firm-level productivity causally increase the share of sales firms report for tax purposes.
- Estimated magnitudes:
  - Baseline IV point estimates imply a notable economic effect; textual statement: “an increase in productivity of 1 percent increases the sales reported to tax administrations by around 0.12 percent.”
  - The authors interpret this as a response elasticity of more than ten percent.
- Robustness: results hold across alternative productivity measures (value-added per worker, TFP), clustering approaches, and multiple IV construction choices.
- Tests of instrument quality:
  - First-stage relevance: Kleibergen-Paap F-statistics well above 10, rejecting weak instrument concerns (Staiger and Stock, 1997 criterion).
  - Overidentification: Hansen J tests yield p-values ≥ 0.1 in main specs, not rejecting instrument validity.

### Policy implications
- Productivity-enhancing policies can yield tax-compliance gains:
  - Broader efforts to improve firm productivity can increase tax revenues by increasing the share of sales reported.
  - These productivity policies can complement tax-policy and tax-administration reforms aimed at reducing tax evasion.
- Caveats noted by authors:
  - IV strategy relies on industry-level productivity changes which may reflect entry/exit and reallocation rather than pure productivity shocks.
  - Limited panel dimension and reliance on self-reported percent of sales reported as the tax-evasion measure.
  - Further research recommended to examine enforcement narratives and the mechanisms linking productivity to compliance.

*Source: wpiea2019260-print-pdf - Appendix 5. First-Stage Regressions*

### References

### wpiea2019260-print-pdf - References

### Cited literature: thematic overview
- Core topics in the references: tax evasion and compliance, informality and firm-level formality decisions, corruption and firm outcomes, productivity estimation and impacts, enforcement of regulation, and instruments for identifying productivity.
- Representative cited works (authors and topics as listed): Allingham & Sandmo (income tax evasion theory); Alm, Martinez-Vazquez & McClellan (corruption and firm tax evasion); Andreoni, Erard & Feinstein (tax compliance review); Dabla-Norris & collaborators (informality, underground economy, and determinants of informality); Feldman & Slemrod (estimating tax noncompliance); Kleven et al. (tax audit experiment and agency model); Levinsohn & Petrin (production function estimation); Kanbur & Keen (thresholds and informality); and multiple empirical firm-level studies on formality, productivity, and enforcement.

### Appendix 1 — Macroeconomic stylized facts: variable definitions and baseline IV estimates
- Variable definitions (labels and units preserved)
  - GDP: Real GDP per capita (WEO) log
  - Self-employment: Self-employed, total (% of total employed) (Key Indicators of the Labour Market 2015, ILO) log
  - oil: The interaction between the change in the international oil price and countries' average oil net-export GDP shares number
  - twwi: The weighted sum of world income for each country with time-invariant weights varying across countries depending on their trade patterns number
  - LIC: 1 for low income countries, 0 otherwise binary
  - APD: 1 for countries in the Asia-Pacific region, 0 otherwise binary
  - EUR: 1 for countries in the European region, 0 otherwise binary
  - WHD: 1 for countries in the Western Hemisphere, 0 otherwise binary
  - post-95: 1 for observations after 1995, 0 otherwise binary
  - Rural population: Rural population expressed in percent of total population (WB) percent
  - Tax revenue: Tax revenue expressed in percent of GDP (WEO) percent
  - Rule of law: Degree of a legal system strength from 0 [min] to 6 [max] (World Government Indicators, WB) number
  - Schooling: Average number of years a person (15+) attended a school or received any formal education (Barro, 2013) number

- Baseline specifications — Dependent variable: Self-employment (selected coefficients and statistics preserved exactly)
  - GDP coefficients across columns (1)–(7):
    - (1) -0.197** [0.085]
    - (2) -0.696*** [0.227]
    - (3) -0.698*** [0.227]
    - (4) -0.760*** [0.217]
    - (5) -0.610*** [0.186]
    - (6) -0.724*** [0.190]
    - (7) -0.483*** [0.134]
  - Observations: 461, 461, 461, 290, 349, 271, 252 (for columns (1)–(7) respectively)
  - # of countries: 101, 101, 101, 86, 101, 86, 78 (for columns (1)–(7) respectively)
  - F stat (columns (1)–(7)): 6.474, 6.387, 5.575, 5.839, 6.932, 4.527, 5.282
  - Kleibergen-Paap F stat (columns (2)–(7)): 13.45, 13.65, 27.98, 57.15, 43.28, 21.71
  - Hansen J stat p-value (columns (2)–(7)): 0.242, 0.231, 0.179, 0.226, 0.280, 0.201
  - Notes preserved: Robust standard errors, clustered at the country level, are in brackets. 5-year averages used. Log of self-employment is the Y variable. Outliers in growth (< -60% or > 60%) dropped. Outliers in self-employment (changes < 1st percentile and > 99th percentile) dropped. Instruments: trade-weighted world income (using average trade shares from 2000-2009) and oil price shocks.

- Robustness checks — Dependent variable: Self-employment (selected entries)
  - GDP coefficients (columns (1)–(4)):
    - (1) -0.664*** [0.239]
    - (2) -0.642*** [0.211]
    - (3) -0.715*** [0.261]
    - (4) -0.648*** [0.174]
  - Observations: 440, 401, 437, 295 (for columns (1)–(4))
  - # of countries: 97, 89, 96, 97 (for columns (1)–(4))
  - F stat (columns (1)–(4)): 5.937, 6.385, 5.620, 11.96
  - Kleibergen-Paap F stat (columns (1)–(4)): 12.64, 17.81, 10.04, 33.03
  - Hansen J stat p-value (columns (1)–(4)): 0.283, 0.172, 0.259, 0.219
  - Notes preserved: same as baseline (robust SEs clustered, 5-year averages, outlier treatment, instruments).

- Robustness checks with interaction terms — Dependent variable: Self-employment (selected coefficients)
  - GDP: -0.698*** [0.226], -0.693*** [0.200], -0.667*** [0.218], -0.698*** [0.218], -0.705*** [0.254] (across columns shown)
  - Interaction terms (examples):
    - GDP * APD: 0.223* [0.129]
    - GDP * WHD: 0.321*** [0.109]
  - Observations: 461 (all columns)
  - # of countries: 101 (all columns)
  - F stat (columns shown): 8.357, 6.094, 5.750, 5.881, 5.565
  - Kleibergen-Paap F stat (columns shown): 13.21, 15.74, 16.56, 13.08, 13.04
  - Hansen J stat p-value (columns shown): 0.255, 0.494, 0.200, 0.200, 0.236

### Appendix 2 — Derivations and model extension (comparative statics and partial evasion)
- Comparative statics signs (preserved expressions)
  - ∂s* / ∂A = −1 / A^2 [1 − T − b(1 − e) − C] < 0
  - ∂s* / ∂e = −b / A^2 [1 − T − b(1 − e) − C]^2 < 0
  - ∂s* / ∂T = ∂s* / ∂C = 1 / A [1 − T − b(1 − e) − C]^2 > 0
  - Further: ∂s* / ∂x = sF(s*) sss* ... > 0 with first bracketed term positive and where x = T, C, or p.
  - Second derivatives examples:
    - ∂^2 s* / ∂A ∂T = ∂^2 s* / ∂A ∂C = −1 / A^2 [1 − T − b(1 − e) − C]^2 < 0
    - ∂^2 s* / ∂A ∂T = ∂^2 s* / ∂A ∂C = b / A^2 [1 − T − b(1 − e) − C]^2 > 0

- Model extension: partial tax evasion (preserved notation and key conditions)
  - Let z_i denote the fraction of firm i’s output declared for tax purposes; 1 − z_i is undeclared fraction.
  - Simplifying assumptions: b = 1; T denotes the tax rate.
  - Declared-profit component: s_i A z_i (1 − T).
  - Probability of detection increases with scope: e(s_i A (1 − z_i)), with e′, e′′ > 0.
  - Expected profit of noncompliant activity: s_i A (1 − z_i) (1 − e).
  - Total expected profit: s_i A z_i (1 − T) + s_i A (1 − z_i) (1 − e).
  - First-order condition at interior solution (preserved equation form):
    - s_i A (1 − T) − s_i A (1 − e) + s_i A e′ s_i A = s_i A (e − s) + [s_i A]^2 e′ = 0
  - Rearranged condition: [s_i A] e′ + e = T.
  - Total differentiation result preserved: ∂z_i / ∂s > 0 (larger technological opportunity set s increases declared share z_i).

### Appendix 3 — Variable definitions and descriptive statistics (IV baseline sample)
- Descriptive statistics (Table Appendix 3.2) — Selected variables with exact numeric summaries
  - Sales reported: MIN 0, p25 70, p75 100, MAX 100, Average 81.95, Std. Dev. 28.07, Number 11,865
  - Labor productivity: MIN -2.33, p25 7.49, p75 10.45, MAX 19.34, Average 8.73, Std. Dev. 2.47, Number 11,865
  - Age: MIN 0, p25 1.95, p75 3.14, MAX 4.61, Average 2.55, Std. Dev. 0.81, Number 11,865
  - Exporter: MIN 0, p25 0, p75 0, MAX 1, Average 0.20, Std. Dev. 0.40, Number 11,865
  - Foreign-owned: MIN 0, p25 0, p75 0, MAX 1, Average 0.11, Std. Dev. 0.31, Number 11,865
  - Government-owned: MIN 0, p25 0, p75 0, MAX 1, Average 0.03, Std. Dev. 0.18, Number 11,865
  - Small: MIN 0, p25 0, p75 0, MAX 1, Average 0.11, Std. Dev. 0.32, Number 11,865
  - Perception of tax admin.: MIN 0, p25 0, p75 1.33, MAX 4, Average 0.93, Std. Dev. 0.76, Number 11,865

- Variable definitions (preserved)
  - Sales reported: Percent of sales reported by an enterprise for tax purposes
  - Labor productivity: Sales in 2002 USD per employee (in logs)
  - Regional IV for labor prod.: IV constructed as the simple average of productivity in other regions of the same country, industry and year
  - International IV for labor prod.: IV constructed as the average of productivity in the same industry and year in two other countries in the sample that have a GDP that is at least 5 times as large and are located in the same World Bank region (commodity exporters not matched with non-commodity exporters)
  - TFP, LP: Total factor productivity based on Levinsohn and Petrin (2003) (in logs)
  - Age: Age of the establishment (in logs)
  - Exporter: 1 if a firm exports goods or services, 0 otherwise
  - Foreign-owned: 1 if the primary owner is a foreign individual or entity, 0 otherwise
  - Government-owned: 1 if the establishment is government-owned or operates under control of the government, 0 otherwise
  - Small: 1 if a firm employs fewer than 5 workers, 0 otherwise
  - Perception of tax admin.: Perception of tax administration as a constraint to doing business (on a scale from 1 (least binding) to 4 (most binding) and normalized by average constraint perception)

### Appendix 4 — Regional and international instrument construction (notes and metadata)
- Regional instruments (Main Sample): country–region observations listed with percent of observations (examples preserved as in source)
  - 2003 Cambodia KHM Phnom Penh 62.75
  - 2003 El Salvador SLV San Salvador 71.51
  - 2003 Honduras HND Tegucigalpa 24.40
  - 2003 Kyrgyz Republic KGZ Bishkek 44.87
  - 2003 Lesotho LSO Maseru 66.67
  - ... (full listing preserved in source)
- Notes preserved:
  - The column “Region” either displays the identified region or the largest among two identified regions.
  - Enterprise Surveys from specific countries are either excluded because a regional instrument could not be constructed or because no data from a country could be used as an instrument.
  - Commodity exporters defined as countries that export oil, gas, and metals where these commodities represent 20 percent or more of total exports or fiscal revenues (IMF, 2015b).
- Commodity Country Pairs (IV Baseline Sample) — selection criteria preserved
  - Note: GDP of the country used as an instrument must be at least 5 times larger than that of an instrumented country.
  - Example entries preserved:
    - Europe and Central Asia 2003: Instruments Poland, Uzbekistan; Instrumented countries Kyrgyz Republic, Moldova, Tajikistan; Excluded countries -; Commodity Exporter No
    - Latin America and Caribbean 2003: Instruments Brazil; Instrumented countries Honduras, Nicaragua, El Salvador; Excluded countries Guatemala; Commodity Exporter No
    - Sub-Saharan Africa 2006: Instruments Tanzania, Uganda; Instrumented country The Gambia; Excluded countries -; Commodity Exporter No
    - ... (full table preserved in source)

### Appendix 5 — First-stage regressions (selected coefficients and statistics)
- First Stage of Table 2 — Dependent variable: Labor productivity [log] (regional and international IV coefficients with standard errors)
  - Regional IV for labor prod. [log]:
    - 0.179*** [0.067], 0.175** [0.073], 0.179*** [0.043], 0.174*** [0.066] (columns (1)–(4))
  - International IV for labor prod. [log]:
    - 0.250*** [0.042], 0.151*** [0.030], 0.250*** [0.050], 0.249*** [0.040] (columns (1)–(4))
  - Covariates (examples with coefficients and standard errors):
    - Age of firm [log]: 0.053** [0.024], 0.054** [0.024], 0.053*** [0.019] (where included)
    - Exporter [dummy]: 0.340*** [0.073], 0.345*** [0.070], 0.340*** [0.046], 0.341*** [0.070]
    - Foreign-owned [dummy]: 0.428*** [0.086], 0.401*** [0.082], 0.428*** [0.061], 0.419*** [0.082]
    - Perception of tax admin.: 0.048*** [0.015], 0.048*** [0.014], 0.048*** [0.015]
  - Observations and sample counts (columns (1)–(4)):
    - Observations: 11,865; 13,208; 11,865; 11,869
    - # of firms: 11,741; 13,056; 11,741; 11,745
    - # of countries: 47; 47; 47; 47
    - # of industries: 18; 18; 18; 18
  - Notes preserved: Errors in brackets clustered at the region and industry-year level except in spec. 3 where clustering is at the country-industry level.

- First Stage of Table 3 — Dependent variable: Labor productivity [log] (alternate sample/specifications)
  - Regional IV for labor prod. [log]:
    - 0.185*** [0.031], 0.181*** [0.067], 0.187*** [0.070], 0.208*** [0.078] (columns (1)–(4))
  - International IV for labor prod. [log]:
    - 0.130 [0.091], 0.251*** [0.042], 0.255*** [0.044], 0.254*** [0.042] (columns (1)–(4))
  - Covariates (examples preserved):
    - Age of firm [log]: 0.005 [0.049], 0.057** [0.025], 0.062** [0.025], 0.050** [0.025]
    - Exporter [dummy]: 0.408*** [0.058], 0.365*** [0.081], 0.366*** [0.076], 0.365*** [0.075]
    - Foreign-owned [dummy]: 0.446*** [0.073], 0.486*** [0.083], 0.540*** [0.089], 0.465*** [0.097]
    - Perception of tax admin.: -0.064 [0.084], 0.048*** [0.017], 0.068*** [0.020], 0.053*** [0.015]
  - Observations and sample counts (columns (1)–(4)):
    - Observations: 13,618; 9,980; 13,937; 10,198
    - # of firms: 13,464; 9,856; 13,725; 10,079
    - # of countries: 50; 38; 54; 46
    - # of industries: 18; 18; 20; 17
  - Notes preserved: Errors in brackets clustered at the region and industry-year level.

*Content unit: wpiea2019260-print-pdf - References*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019260-print-pdf.pdf_