Patterns of Trade and Tax Interventions and Firm Performance

Transcription

1 Patterns of Trade and Tax Interventions and Firm Performance Ann Harrison (World Bank and University of California, Berkeley) Jing Cai (University of California, Berkeley) Justin Yifu Lin (World Bank)

2 Questions Which types of sectors do governments target for trade and tax policies? 1) Sectors which are more labor intensive 2) Sectors which are more export-oriented 3) Sectors which are more skill-intensive 4) Sectors which are associated with a higher R&D intensity How do patterns of tariff and tax policies to promote industry affect productivity levels and growth within China?

3 Motivation There is great debate about the role of industrial policies in economic growth. Recent studies show industrial policies such as trade protection are more likely to work if targeted at certain sectors: Nunn and Trefler (2006): Skill-intensive sectors Lin (2009): Sectors with comparative advantages Harrison and Rodriguez-Clare (2009): Sectors with latent comparative advantages Clemens and Williamson (2001): Emerging sectors Helpman and Grossman (1991): R&D intensive sectors However, empirical studies of structures of interventions are very limited.

4 Measures of Patterns of Interventions Define a series of correlations to identify what sectors are targeted by governments. Focus on four types of sectors: Labor intensive (L/K): total workforce/fixed assets for production Export intensive (EXP/SALES): export procurement/industrial sales Skill intensive (S/UNSK): #skilled workers /#unskilled workers, skilled workers are those with at least high school education R&D intensive: #high-level technicians and engineers/total workforce Use initial value of industry-city level endowments and current period policy variables to measure patterns of policies.

5 Measures of Patterns of Interventions Patterns of tariff policies are measured by: Ω =Corr( Tariff, L / K ) rt jt jr0 jr0 ω = Corr( Tariff, EXP / SALES ) rt jt jr0 jr0 ρ = Corr( Tariff, S / UNSK ) rt jt jr0 jr0 υ = Corr( Tariff, RDIntensity ) rt jt jr0 For example, the first correlation, which is between the industry-city level initial of period labor intensity and current period tariffs for sector j in city r, measures whether tariff protection is biased towards labor intensive sectors in city r in year t.

6 Estimation Strategy Estimate the effect of patterns of policies on firm performance. Measures of performance: Total factor productivity estimated using four methods: AW et al 2001 (AW), OLS, OLS with fixed effects, and Olley & Pakes 1996 (OP). Three sets of estimations using different measures of performance: Levels of firm productivity Growth of firm productivity Levels of industry-city aggregate productivity

7 Estimation Strategy Firm-level estimation : lntfp = β + β lntariff + β Ω + β X + f + D + ε ijrt 0 1 jt 1 2 rt 3 jrt i t ijrt lntfp = δ + δ lntariff + δω + δx + f + D + ε ijrt 0 1 jt 1 2 rt 3 jrt i t ijrt lntfp = λ + λlntariff + λρ + λx + f + D + ε ijrt 0 1 jt 1 2 rt 3 jrt i t ijrt lntfp = τ + τ lntariff + τυ + τ X + f + D + ε ijrt 0 1 jt 1 2 rt 3 jrt i t ijrt lntfpgrowth = β + β lntariff + β Ω + β X + D + ε ijrt 0 1 jt 1 2 rt 3 jrt t ijrt lntfpgrowth = δ + δ lntariff + δω + δx + D + ε ijrt 0 1 jt 1 2 rt 3 jrt t ijrt lntfpgrowth = λ + λlntariff + λρ + λx + D + ε ijrt 0 1 jt 1 2 rt 3 jrt t ijrt lntfpgrowth = τ + τ lntariff + τυ + τ X + D + ε ijrt 0 1 jt 1 2 rt 3 jrt t ijrt

8 Estimation Strategy Industry-level estimation: lntfp = β + β lntariff + β Ω + β X + η + h + D + ε jrt 0 1 jt 1 2 rt 3 jrt j r t ijrt lntfp = δ + δ lntariff + δω + δx + η + h + D + ε jrt 0 1 jt 1 2 rt 3 jrt j r t ijrt lntfp = λ + λlntariff + λρ + λx + η + h + D + ε jrt 0 1 jt 1 2 rt 3 jrt j r t ijrt lntfp = τ + τ lntariff + τυ + τ X + η + h + D + ε jrt 0 1 jt 1 2 rt 3 jrt j r t ijrt The aggregate industry productivity is measured by: lntfp = s lntfp jrt it ijrt i

9 Data Firm-level unbalanced survey data in China, Tariff dataset from the World Integrated Trading Solution (WITS)

10 Summary Statistics: Patterns of Interventions Year Correlation (Tariff, L/K) Table1. Patterns of Tariffs and Tax by Year, excluding alcohol and tobacco sectors Correlation(Tariff, Export/Sales) Correlation(Tariff, Skilled/Unskilled) Correlation(Tariff, RD intensity) Correlation (Tax, L/K) Correlation(Tax, Export/Sales) Correlation(Tax, Skilled/Unskilled) Correlation(Tax, RD intensity) Total Notes: All endowments are measured by industry-city level beginning of period values (1998 for labor and export intensity, 2004 for skill intensity and RD intensity). Labor intensity is defined as the ratio of total workforce to fixed assets for production; export intensity equals export procurement divided by industrial sales; the ratio between skilled workers (educations equivalent to or higher than senior high school) and unskilled workers is defined as skill intensity; RD intensity is defined as the share of high level technicians and engineers in total number of employees. Alcoholic beverages and tobacco sectors have been excluded because tariff and tax have been imposed for revenue-raising reasons.

11 Results Summary: Effects of Patterns of Tariff on Firm Performance Firm Level lntfp Tariff Correlation (Tariff, L/K) Correlation(Tariff, Export/Sales) Correlation(Tariff, Skilled/Unskilled) Correlation(Tariff, RD intensity) AW - / + - / OLS FE / OP / Industry-City Level Aggregate lntfp AW OLS - / FE - / OP - / Annual Growth of Firm Level lntfp FE / / Notes: Summary of effects of structures of tariff policy on firm productivity, industry-city level aggregate productivity, and the annual growth of firm level productivity. TFP is estimated using four methods: AW et al 2001 (AW), OLS, OLS with fixed effects, and Olley & Pakes (OP). Skilled workers are defined as workers with educations equivalent to or higher than senior high school. RD intensity is defined as the share of high level technicians and engineers in total number of employees. + indicates significantly positive, - indicates significantly negative, and / indicates insignificant.

12 Estimation Results: Effects of Patterns of Tariff on Firm-level Productivity VARIABLES Firm Level lntfp AW OLS FE OP Panel A. Effect of tariff policies biased towards labor-intensive sectors lntariff *** *** *** *** ( ) ( ) ( ) ( ) Correlation (Tariff, L/K) ** ** ** ( ) ( ) ( ) ( ) lntariff_input *** *** *** ( ) ( ) ( ) ( ) Panel B. Effect of export-oriented biased tariff policies lntariff *** *** *** *** ( ) ( ) ( ) ( ) Correlation (Tariff, Export/Sales) *** *** *** *** ( ) ( ) ( ) ( ) lntariff_input *** * *** *** ( ) ( ) ( ) ( ) Panel C. Effect of skill- biased tariff policies lntariff *** *** *** *** ( ) ( ) ( ) ( ) Correlation(Tariff, Skilled/Unskilled) *** *** *** *** ( ) ( ) ( ) ( ) lntariff_input *** ** *** *** ( ) ( ) ( ) ( ) Panel D. Effect of tariff policies targeting on R&D intensive sectors lntariff *** *** *** *** ( ) ( ) ( ) ( ) Correlation(Tariff, RD intensity) *** ( ) ( ) ( ) ( ) lntariff_input *** ** *** *** ( ) ( ) ( ) ( ) Observations 1,146,938 1,292,877 1,292,877 1,292,877 Firm Fixed Effects Yes Yes Yes Yes Year Dummies Yes Yes Yes Yes Notes: Robust clustered standard errors are presented in parentheses. TFP are estimated using four methods: AW et al 2001 (AW), OLS, OLS with fixed effects, and Olley & Pakes (OP). Skilled workers are defined as workers with educations equivalent to or higher than senior high school. RD intensity is defined as the share of high level technicians and engineers in total number of employees. *Significant at 10% level, **significant at 5% level, and ***significant at 1% level.

13 Estimation Results: Effects of Patterns of Tariff on Firm-level Productivity Growth VARIABLES Annual Growth of Firm Level lntfp (FE Estimation) (1) (2) (3) (4) (5) lntariff ** * Correlation(Tariff, L/K) *** ( ) ( ) ( ) ( ) ( ) ( ) Correlation(Tariff, Export/Sales) *** ( ) Correlation(Tariff, Skilled/Unskilled) *** ( ) Correlation(Tariff, RD intensity) ( ) Foreign Share *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) HKTM Share *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) State Share *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) Export Share *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) lntariff_input *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) Observations 1,022, , ,543 1,019,870 1,019,833 Year Dummies Yes Yes Yes Yes Yes R-squared Notes: Robust clustered standard errors are presented in parentheses. TFP are estimated using four methods: AW et al 2001 (AW), OLS, OLS with fixed effects, and Olley & Pakes (OP). Skilled workers are defined as workers with educations equivalent to or higher than senior high school. RD intensity is defined as the share of high level technicians and engineers in total number of employees. *Significant at 10% level, **significant at 5% level, and ***significant at 1% level.

14 Estimation Results: Effects of Patterns of Tariff on Industry-City Aggregate Productivity VARIABLES Industry-City Level Aggregate lntfp AW OLS FE OP Panel A. Effect of tariff policies biased towards labor-intensive sectors lntariff *** *** *** *** (0.0404) (0.0416) (0.0402) (0.0402) Correlation(Tariff, L/K) ** ( ) ( ) ( ) ( ) lntariff_input ** ** ** ** (0.112) (0.123) (0.115) (0.115) Panel B. Effect of export-oriented biased tariff policies lntariff *** ** *** *** (0.0413) (0.0424) (0.0412) (0.0412) Correlation (Tariff, Export/Sales) *** *** *** *** (0.0102) ( ) (0.0119) (0.0111) lntariff_input ** ** ** ** (0.112) (0.122) (0.115) (0.115) Panel C. Effect of skill- biased tariff policies lntariff *** *** *** *** (0.0402) (0.0415) (0.0404) (0.0403) Correlation(Tariff, Skilled/Unskilled) *** *** *** *** ( ) ( ) ( ) ( ) lntariff_input ** ** ** ** (0.111) (0.122) (0.114) (0.114) Panel D. Effect of tariff policies targeting on R&D intensive sectors lntariff *** *** *** *** (0.0402) (0.0415) (0.0403) (0.0403) Correlation(Tariff, RD intensity) ** ** *** *** ( ) ( ) (0.0108) (0.0102) lntariff_input ** ** ** ** (0.111) (0.122) (0.115) (0.114) Observations 110, , , ,261 Industry Fixed Effects YES YES YES YES Region Dummies YES YES YES YES Notes: Robust clustered standard errors are presented in parentheses. TFP are estimated using four methods: AW et al 2001 (AW), OLS, OLS Year with Dummies fixed effects, and Olley & Pakes (OP). For all estimations, YES dependent variable YES is the industry-city YES level weighted YES aggregate lntfp. Skilled workers are defined as workers with educations equivalent to or higher than senior high school. RD intensity is defined as the share of high level technicians and engineers in total number of employees. *Significant at 10% level, **significant at 5% level, and ***significant at 1% level.

15 Estimation Results: Effects of Patterns of Tax Policies on Industry-City Aggregate Productivity VARIABLES Industry-City Level Aggregate lntfp AW OLS FE OP Panel A. Effect of tax policies biased towards labor-intensive sectors lntax *** *** ** ( ) ( ) ( ) ( ) Correlation(Tax, L/K) *** *** *** *** ( ) ( ) ( ) ( ) Panel B. Effect of export-oriented biased tax policies lntax *** *** ** ( ) ( ) ( ) ( ) Correlation (Tax, Export/Sales) ** *** *** *** ( ) ( ) ( ) ( ) Panel C. Effect of skill- biased tax policies lntax *** *** ** ( ) ( ) ( ) ( ) Correlation(Tax, Skilled/Unskilled) *** ( ) ( ) ( ) ( ) Panel D. Effect of tax policies targeting on R&D intensive sectors lntax *** *** ** ( ) ( ) ( ) ( ) Correlation(Tax, RD intensity) *** *** *** *** ( ) ( ) ( ) ( ) Observations 104, , , ,895 Industry Fixed Effects YES YES YES YES Region Dummies YES YES YES YES Year Dummies YES YES YES YES Notes: Robust clustered standard errors are presented in parentheses. TFP are estimated using four methods: AW et al 2001 (AW), OLS, OLS with fixed effects, and Olley & Pakes (OP). Tax is the sum of income tax payable and value added tax payable and is on the citysector level. For all estimations, dependent variable is the industry-city level weighted aggregate lntfp. Skilled workers are defined as workers with educations equivalent to or higher than senior high school. RD intensity is defined as the share of high level technicians and engineers in total number of employees. *Significant at 10% level, **significant at 5% level, and ***significant at 1% level.

16 Conclusions Final goods and input tariffs have significantly negative effects on both firm and industry level productivity and growth; Tariff protection and tax reduction policies are most successful when targeted at laborintensive, export-oriented, unskilled-intensive, and less R&D intensive sectors in China. Tariff and tax settings in China have been systematically biased in favor of sectors where China has a comparative advantage.

17 Discussion How could we identify sectors with latent comparative advantage? Option 1: By looking at the evolution of correlations between tariff/tax growth and growth in industrial sales. Correlation Sector28: Measuring, cultural and office eqiupment Year Corr(Tariff growth, Sales growth) Corr(Tax growth,sales growth)

18 Discussion Option 2: 1) Identify sectors with comparative advantage 2) Use methods developed in Hausmann and Klinger (2006) to measure the relatedness of other sectors to comparative advantage sectors 3) Use the relatedness as a measure of latent comparative advantage.

19 Thank You!