Overcoming Free Riding A Cross Country Analysis of Firm Participation in Antidumping Petitions Kara M. Reynolds American University
Introduction • The growth in the imposition of antidumping (AD) protection over the past 10 years has been well-documented and studied. • Much of the previous research has studied the degree to which macroeconomic and political factors have caused this increase. • How do firms make the decision to request AD protection?
Free-Riding • All firms within an industry benefit from most forms of trade protection. • Only those who actively lobby for protection, or submit the AD petition, have to pay the costs. • The industry’s ability to seek protection suffers from the free-riding problem. ▫ Firms prefer to let their competitors file the antidumping petition. If all firms choose to free- ride, the industry will forgo requesting AD protection.
Firm Participation in AD, 1995-2005
Heterogeneity Across Industries
Heterogeneity Across Countries
Research Goals • Empirically estimate the determinants of the firm decision to file for AD protection. • Explore potential heterogeneity in the severity of the free-riding problem across industries and countries. • Preliminary evidence suggests that more firms participate the lower the cost of filing for AD protection. • Little evidence that firms perceive that their participation will increase the expected benefits associated with AD petitions.
Literature Review • Demand for AD protection ▫ Knetter and Prusa (2003), Feinberg (2005) ▫ Feinberg and Reynolds (2006,2007) • Theoretical Models of Rent-Seeking and Trade ▫ Grossman and Helpman (1996) ▫ Magee (2002)
Free Riding in AD • Herander and Pupp (1991) ▫ Representative firm contributes to petition if expected benefits exceed the costs. ▫ Benefits– contributing may increase the probability that petition will be successful. ▫ Empirical results… � Number of firms participating decreases with expected costs per firm. � Free-riding less severe when benefits concentrated among few firms and when firms represented by trade association. � Probability of protection increases with participating firms. • Olson (2004)
A S imple Model of Firm Behavior • Firms contribute to AD petition if the expected benefits of contributing exceed the costs. • Firms derive benefits from contributing if the expected level of protection associated with petition increases from their participation. = π − π + π − − * ( ) ( ) B B B B Costs C NC C C C NC π : Probabilit y that petition w ill be successful . B : Increase in profits from successful petition
Dependent Variable • Number of firms filing the AD petitions considered by 10 countries between 1995 and 2005. ▫ Argentina, Australia, Canada, China, EU, India, Mexico, New Zealand, U.S. and South Africa • Global AD Database ▫ Includes domestic firms filing each petition and HS codes associated with each petition. ▫ Data supplemented with case information from the country.
Explanatory Variables = π − π + π − − * ( ) ( ) B B B B Costs C NC C C C NC • Number of participating firms should increase with the expected benefits and decrease with the expected costs. • I do not observe the firm’s perceived impact on the probability of success or the expected level of protection. • I include variables that have been found to impact the probability of success and variables I expect to impact the value of benefits.
Explanatory Variables = π − π + π − − * ( ) ( ) B B B B Costs C NC C C C NC • Insignificant coefficients could have two interpretations 1. Variables do not impact the probability of success and expected benefits. 2. Variables do impact the probability success or benefits, but firms do not believe that their contribution matters.
Probability of S uccess • UN’s COMTRADE Database ▫ One-Year Growth in Imports from the Country Under Investigation ▫ Targeted country’s share of the total imports of the product. • World Bank’s World Development Indicators (WDI) ▫ Two-year appreciation of country’s real exchange rate. ▫ Two-year growth in country’s real GDP.
Probability of S uccess • World Bank’s Trade, Production and Protection Database (TPP) ▫ Size of the Industry ▫ Average Value Added of three digit ISIC industry between 1995 and 2004
Benefits of Protection • Total Value of Imports from Country Targeted in Petition (COMTRADE) • Number of firms in the industry ▫ Number of establishments in the three-digit ISIC industry (TPP).
Expected Costs • Number of petitions filed simultaneously by the industry. • Trade Association • Union • Country-Specific Factors ▫ Political Regime. Polity variable from the Polity IV Project. ▫ Level of Development (WDI).
Empirical Methodology • Parameters of the model estimated using a panel count model. ▫ Explanatory variable a discrete count of firms. µ E(y ijp = ) ijp µ δ γ β ln( ) = + w + x ijp ij ij ijp y µ µ exp(- ) ijp ijp ijp Pr(y | w , x ) = ijp ij ijp Γ (y + 1) ijp
Overdisperion • Poisson model assumes that the mean number of firms is equal to the variance. • Summary statistics and GOF tests suggest the data exhibits over dispersion—or the variance is greater than the mean. ▫ Mean is 2.9, while variance is 7.25 • Solution—Negative Binomial Model
Unobserved Heterogeneity • Unobserved heterogeneity in panels can be measured as fixed effect or random effect. • If unobserved component is correlated with the explanatory variable, random effects will result in biased coefficients. • Hausman tests suggest that random effects are inappropriate in this case.
Unconditional FE NB Model µ E(y | w , x ) = ijp ij ijp ijp µ δ γ β + ε ln( ) = + w + x ijp ij ij ijp ijp = + κµ ) µ var( | , ) ( 1 y w x ijp ij ijp ijp ijp θ Γ + θ y ( )r (1 - r ) ijp y ijp ijp Pr(y | w , x ) = ijp , ijp ij ijp Γ Γ θ (y + 1) ( ) ijp κ 1 θ = = , r κ κ + µ ijp ijp
Empirical Problems • Fixed effects models do not allow for the estimation of the impact of petition invariant factors. • Omitted variable bias ▫ Total number of firms in the industry • Sample selection bias
Results from the FE NB Model (1) b Variable Benefits Target's Share of Imports 0.748** (2.42) Target's One-Year Import Growth 0.991* (1.71) GDP Growth 3.508 (1.54) Exchange Rate Appreciation 1.408* (1.75) Log(Imports) 1.007 (1.64) Costs Association 1.630** (9.15) Union 1.070 (0.98) Cumulated Cases 1.027** (5 23)
FE versus RE Variable FE RE Benefits Target's Share of Imports 0.748** 0.665** (2.42) (3.48) Target's One-Year Import Growth 0.991* 0.990** (1.71) (1.75) GDP Growth 3.508 2.485 (1.54) (1.32) Exchange Rate Appreciation 1.408* 1.166 (1.75) (0.83) Log(Imports) 1.007 1.011** (1.64) (2.48) Costs Association 1.630** 1.674** (9.15) (10.04) Union 1.070 1.100 (0.98) (1.41) Cumulated Cases 1.027** 1.028** (5.23) (5.55)
Addition of Petition Invariant Characteristics (Benefits) (1) b (2) c (3) c (4) c Variable Benefits Target's Share of Imports 0.748** 0.665** 0.676** 0.807 (2.42) (3.48) (3.35) (1.42) Target's One-Year Import Growth 0.991* 0.990** 0.991 0.991 (1.71) (1.75) (1.64) (1.26) GDP Growth 3.508 2.485 3.637* 0.302 (1.54) (1.32) (1.76) (0.96) Exchange Rate Appreciation 1.408* 1.166 1.277 0.887 (1.75) (0.83) (1.31) (0.40) Log(Imports) 1.007 1.011** 1.010** 1.003 (1.64) (2.48) (2.30) (0.52) Log(Value Added) 1.153** (2.97) Log(Number of Establishments) 1.155** (2.45)
Addition of Petition Invariant Characteristics (Costs) (1) b (2) c (3) c (4) c Variable Costs Association 1.630** 1.674** 1.677** 2.388** (9.15) (10.04) (10.07) (12.19) Union 1.070 1.100 1.089 (0.98) (1.41) (1.25) Cumulated Cases 1.027** 1.028** 1.028** 1.001 (5.23) (5.55) (5.54) (0.11) Log(GDP per Capita) 1.126** 1.025 (2.91) (0.46) Polity 0.981 0.972 (1.15) (1.56) θ 0.000 0.367** 0.330** 0.174** (0.00) (7.27) (7.02) (4.97)
S ummary • Very strong evidence that more firms participate when the costs of participation are lower. • Other parameters estimates suggest that fewer firms participate when the probability of success is higher. ▫ Share/Growth in Imports, GDP Growth • The results provide only weak evidence that firms believe their contribution will increase the probability that the petition will be successful and/or the level of protection. ▫ Exchange Rate, Total Value of Imports
Recommend
More recommend
