Tax Rates and Tax Evasion: Evidence from “Missing Imports” in China Raymond Fisman Columbia University Shang-Jin Wei IMF and NBER
Tax Rates and Tax Evasion: Evidence from “Missing Imports” in China
Raymond FismanColumbia University
Shang-Jin WeiIMF and NBER
Plan of the Presentation
• Motivation & Basic Methodology
• Data
• Basic Results
• Economic Interpretation
• Robustness Checks and Extensions
• Conclusions
Public Finance Theory: The Role of Evasion
• Often, collection assumed to be perfect
• Correlates of evasion:-punishment (cost)-tax rate (benefit)
• How important is this effect?
Theory: Rate-Evasion Relationship may be Counterintuitive
Model: Allingham & Sandmo (1972)Predicted Relationship: Depends on U’’’(w)
Model: Yitzhaki (1972)Predicted Relationship: Positive if punishmentis a function of evaded taxes
How much evasion is there? Measuring what we cannot observe
1. Discrepancy in Macro-variables-Difference between national income and product accounts-Ratio of currency to M2
2. U.S. Taxpayer Compliance Measurement Program (TCMP)
Measuring Evasion: Previous Work
Method Approach/
Examples
Problems
Inference from obs. quantities
1. M Demand
2. I – C
-based on dubious assumptions
-measures evasion level (not sensitivity)
U.S. TCMP Intensive Audit
-Rate-Evasion relationship not separable from income effect
Our Approach
Without Evasion (and measurement error):
Exports from Hong Kong to China
Imports from China to Hong Kong
In reality…
Noise
Evasion
HK-Reported Exports to China
China-reported Imports from HK
Does Tax Rate Affect Evasion?
HK-Reported Exportsi – China-Reported Importsi
= ‘Evasion Gap’i + Noisei
= i + *(Tax Rate)i + I
H0: > 0
Predictions: Types of Evasion
• Smuggling that bypasses Hong Kong: Not detected by analyses
• Under-reporting prices
• Under-reporting quantities
• ‘Mislabeling’ goods
Summary of PredictionsType of Evasion Predictions
Under-reporting per unit value
Tax rate positively correlated with evasion gap in values
Under-reporting quantities
Tax rate positively correlated with evasion gap in values and quantities
‘Mislabeling’ goods Tax rate of closely related goods negatively correlated with evasion gap in values and quantities
Plan of the Presentation
• Motivation & Basic Methodology
• Data
• Basic Results
• Economic Interpretation
• Robustness Checks and Extensions
• Conclusions
Trade Flow Data: WITS(World Integrated Trade Solutions)
• Derived from UN Comtrade
• 6-digit Harmonized Commodity Description and Coding System (HS)
• Available since 1996
• Data Utilized: 1998 (little year-to-year variation in tax rates) (and later also 1997)
• Both Value and Quantity data available
Industry-level data: Examples540210 High tenacity yarn of nylon or other polyamides540220 High tenacity yarn of polyesters540231 Textured yarn :-- Of nylon or other polyamides, measuring per single yarn not more than 50 tex540232 Textured yarn :-- Of nylon or other polyamides, measuring per single yarn more than 50 tex540233 Textured yarn :-- Of polyesters540239 Textured yarn :-- Other
291211 Acyclic aldehydes without other oxygen function :-- Methanal (formaldehyde)291212 Acyclic aldehydes without other oxygen function :-- Ethanal (acetaldehyde)291213 Acyclic aldehydes without other oxygen function :-- Butanal (butyraldehyde, normal isomer)291219 Acyclic aldehydes without other oxygen function :-- Other291221 Cyclic aldehydes without other oxygen function :-- Benzaldehyde291229 Cyclic aldehydes without other oxygen function :-- Other
010111 Horses :-- Pure-bred breeding animals010119 Horses :-- Other010120 Asses, mules and hinnies
Tax Rate on Product k
• = Tariff rate on Product k + VAT on k
• In 1998, across 8 digit categories
• Mean = 36, min=13 and max =135
• standard deviation = 10.34
Direct vs. Indirect Imports/Exports
• Entry point for many 3rd country exports
• Indirect exports reported separately
HK Reported Exports
• In theory, Chinese reported imports reflect country of origin
• Not always successfully separated (esp. Taiwan)
China Reported Imports
Key Fact: Tax Rates are Identical for Direct & Indirect Imports
Evasion Gap Defined
Gap_Value = log(Export_Value) - log(Import_Value)
Gap_Qty = log(Export_Qty) - log(Import_Qty)
*All Export Figures reflect Direct Exports only
Tax Rate Data: WITS
• Derived from UNCTAD TRAINS (Trade Analysis and Information System) Database
• Tariff and VAT rates at the 8-digit level• No within-variation for almost all 6-digit
industries: analysis restricted to this subsample• TAX = Tariff + VAT• Taxes on similar goods: Tax rate of all other
goods in 4-digit class
Variable Mean Std. Dev.
Gap_Value -0.62 2.43
Gap_Qty -1.06 2.56
Tax Rate (Tariff+VAT)
36.09 10.34
Direct Export Ratio
0.17 0.23
Summary Statistics
Specification
log(Exportk ) - log(Importk )
= α + β Taxk + εk
Import*k = Importk + Misclassified Indirect Importk
But observed imports (Import*) are imperfectly recorded:
Modeling Assumption:
Misclassified Indirect Importk = k k Importk
Recall: Tax Rates are Identical for Direct & Indirect Imports
Tax rate uncorrelated with Misclassified Imports
Import*k = Importk + Misclassified Indirect Importk
= (1+ k k ) Importk
log(Exportk ) - log(Import*k )= α* + β Taxk + ek
Hence:
Combining with previous equations, we estimate:
Evasion Gap
Summary of Predictions
Sign In Value Regression
Sign In Qty Regression
Tax Avg(Tax) Tax Avg(Tax)
Underreported Prices
+ 0 0 0
Underreported Quantities
+ 0 + 0
Relabeling + – + –
Typ
e of
Eva
sion
Tax Rates and Tax Evasion
• Motivation & Basic Methodology
• Data
• Specification
• Results and Interpretations
• Extensions
Table 2: Tax Rate and Evasion
Dep.var.=evasion (1) (7)
Tax rate 2.82 (0.78) 3.44 (1.02)
Excluding outliers no yes
Excluding obs.Lacking quantities
no yes
Excluding obs.Lacking similar products.
no yes
#obs./ adj. R2 2043 / 0.015 1165 / 0.025
Table 3: Aggregation by Tax Rates
Dep.var.=Evasion (1) (2)
Tax Rate 3.10 3.47
(0.99) (0.98)
Method of aggregation mean median
Weighted OLS yes yes
#obs./ adj. R2 44/0.26 44/0.34
Economic Interpretation of the Estimate
• Let M = reported imports; X= true imports.
• M = X - Evasion
• Evasion = constant + Tax rate + noise
• dlog(M)/d Tax = dlog(X)/dTax -
• | dlog(M) / d Tax | > 3
Economic Interpretation (continued)
• Any tax rate above 33.3% is on the wrong side of the Laffer curve: a reduction in rate can result in an increase, rather than decrease, in revenue collection.
• The avg. tax rate in China was 36% (in 1998).• So many rates were too high even from a revenue
point of view.
Effect of mis-labeling: increasing taxes on similar goods
reduces evasion
Gap_Valuek = α + β1*Taxk
+ β2*Avg(Tax_o) + υk
If mis-labeling exists: β2 < 0
Table 4: Evidence of Mis-labeling
Dep. Var.=Evasion (2) (7)
Own Tax rate 6.32 (1.77) 7.20(1.39)
Tax Rate on ‘Similar’Products
-3.57 (1.70) -4.09 (1.47)
Excluding outliers? no yes
Excluding productsW/o Quan. Info.?
no yes
#obs. / adj. R2 1760 / 0.016 1165 / 0.029
Table 5: Evidence on Quantities
Dep. Var.=Evasion in Quantity
(5) (6)
Own Tax rate 7.59 (2.08) 8.11(1.62)
Tax Rate on ‘Similar’Products
-7.40 (2.07) -8.13 (3.50)
Excluding outliers? no yes
Excluding productsW/o Quan. Info.?
yes yes
#obs. / adj. R2 1191 / 0.011 1145 / 0.017
Robustness Checks
1. Controlling for Exemptions
2. Exclude ‘indirect export dominated’ industries
Results: Qualitatively unchanged
Table 6: Controlling for Exemptions
Evasion in value Evasion in Quantity
Own Tax Rate 6.03 (1.84) 6.97 (2.06)
Tax Rate onSimilar Products
-3.04 (1.77) -6.57 (2.09)
Exemption Ratio -0.72 (0.26) -1.02 (0.32)
# obs. / adj. R2 1639 / 0.028 1104 / 0.026
Table 7: First Differences 1997-98
Change inevasion invalue
Change inevasion inquantity
Change in own tax 3.90 (1.92) 5.43 (2.77)
Change in tax on similarproducts
-1.89 (2.01) 3.12 (3.08)
#obs. / adj R2 1574 / 0.006 1045 / 0.004
Flexible Functional Form
Tax –1st quartile -0.97 (3.56) -0.10 (4.96)
Tax –2nd quartile 0.77(4.47) -5.26 (4.72)
Tax –3rd quartile 7.41(3.04) 10.50 (3.98)
Tax –4th quartile 1.59 (1.20) 5.51 (2.65)
Tax on “similar” -3.69 (1.79)
#obs / adj. R2 2043 / 0.017 1760 / 0.022
Conclusions
• Evasion is highly correlated with tax rate• ‘Above average’ tax rates cause most
evasion• Evasion correlated with tax rates is
primarily:- under-reporting the unit value
• - mis-reporting goods as lower-taxed types
Future Work
• Apply the same technique to a broader set of countries
• -How does tariff dispersion affect evasion?
• -Is this an objective measure of corruption?