Local Protectionism, Market Structure, and Social Welfare: China s Automobile Market

Transcription

1 Local Protectionism, Market Structure, and Social Welfare: China s Automobile Market Panle Jia Barwick Shengmao Cao Shanjun Li Cornell University June 2016 (Cornell) Local Protectionism June / 50

2 China s Economic Growth Remarkable growth during its transformation to a market economy GDP growth in 2015 was 6.9%, the lowest in 20 years (Cornell) Local Protectionism June / 50

3 Economic Growth and Market Frictions Export has been an important driver of economic growth Recent discussions on strengthening domestic demand as the engine for future growth Examine domestic market frictions that could hinder efficient resource allocation and growth: Intra-country trade barriers policies and practices that protect local firms against competition from non-local firms (Cornell) Local Protectionism June / 50

4 Local vs. National Market Shares: Automobiles 35% 30% 25% 20% 15% 10% 5% 0% Local market shares National market shares The first five are state-owned enterprises (SOE), next five are joint ventures (JV), and last two are private automakers Local market shares refer to shares in their respective headquarter provinces (Cornell) Local Protectionism June / 50

5 Local vs. National Market Shares: Cigarettes Market Share of Cigarette Products from Major Manufacturers ( ) Local Market Share vs. National Market Share Anhui Chongqing Fujian Guangdong Guangxi Local Market Share Guizhou Hebei Henan Hubei Hunan Jiangsu Jiangxi National Market Share Shaanxi Shandong Shanghai Sichuan Yunnan Zhejiang (Cornell) Local Protectionism June / 50

6 Factors that Contribute to Home Bias Transportation costs: About 1-2% of vehicle prices. Controlled using distance Dealer network: Local brands might have more dealers in the region Control the number of dealers, though it is affected by policies Local preference: Consumers prefer local brands (information and reputation) Local protectionism: Policies that protect local firms against competition from nonlocal firms (Cornell) Local Protectionism June / 50

7 Local Protectionism: the Auto Industry Trade war between Shanghai and Hubei province in the 1990s Changchun, Jilin province: First Auto Works (FAW) is a SOE headquartered in Changchun,Jilin Local governments in Jilin heavily promote FAW brands From Require government procurement to give priority to FAW group Waive all fees (including registration fees and tolls) for individual purchase of FAW brands From Subsidize FAW indigenous brands (not JV brands) by Yuan Guangxi province (2015): Subsidy up to 2000 Yuan for local brands (Cornell) Local Protectionism June / 50

8 Local Protectionism: Institution China s political personnel system is top-down: Local officials are evaluated and promoted based on GDP growth SOEs are important for local GDP and fiscal revenue Management: Top executives of SOEs and JVs (their domestic partner) are appointed and their management decisions influenced by the government Rent-seeking: SOEs often provide private benefits for local government officials Lack of federal regulations: No effective central policies that prohibit interregional trade barriers U.S. Constitution s Commerce Clause prohibits state regulations that interfere with or discriminate against interstate commerce (Cornell) Local Protectionism June / 50

9 Research Questions This paper focuses on: To what extent are the differences between local and national shares driven by local protectionism? How does local protectionism affect prices and demand? What is the welfare impact? Questions to explore in future research: The impacts on firm entry and exit, market structure, and capacity utilization The impacts on resource allocation across regions and their welfare consequences (Cornell) Local Protectionism June / 50

10 Literature Existing literature that examines intra-national trade barriers in China: Young (2000), Bai et al. (2004), Holz (2009), Eberhardt et al. (2015) They identify protectionism by looking for changes in industry specialization and price convergence at the aggregate level We use disaggregate data for one industry that covers all products in all regions Ours is the first study to quantify the welfare impacts Literature on resource allocation and TFP growth: Hsieh and Klenow (2009) examines misallocation of inputs We focus on the misallocation in the product market, which could lead to misallocation of inputs (Cornell) Local Protectionism June / 50

11 Overview 1 Introduction 2 Theory 3 Industry Background and Data 4 Empirical Strategy and Results 5 Simulations and Welfare Analysis 6 Conclusion (Cornell) Local Protectionism June / 50

12 Hotelling Model A linear city from interval [0,1], uniform distribution of consumers. Consumers have unit demand Two firms: one local (A), one non-local (B). Locating at two extremes and selling identical goods Marginal cost of production: 2cq Transportation cost: td 2 where d is the distance Consumer surplus (consumer i choosing j): u ij = ū P j td 2 ij (Cornell) Local Protectionism June / 50

13 Hotelling Model At point x = 1 2 : CS and MC from two choices equalized Equilibrium prices: P A =P B =t(1+2c) (Cornell) Local Protectionism June / 50

14 Hotelling with Subsidy to Local Firm Consumers who buy from local firm (A) receive subsidy S DWL: distortions in consumer choices and production (Cornell) Local Protectionism June / 50

15 Overview 1 Introduction 2 Theory 3 Industry Background and Data 4 Empirical Strategy and Results 5 Simulations and Welfare Analysis 6 Conclusion (Cornell) Local Protectionism June / 50

16 China s Auto Market: Growth and Size China overtook the U.S. as the world s largest auto market in 2009 (Cornell) Local Protectionism June / 50

17 Industry Background Market shares for different types of firms: Joint ventures (JVs, 71%), State owned enterprises (SOEs, 12%), domestic private firms (10%), and imports (7%) All major international automakers are present in China as JVs: By law, foreign automakers cannot have their own production VW and GM have the largest presence in China The auto industry is targeted by many government policies: Important for GDP, employment, and spillovers to other industries Strategic industry in 26 provinces (out of 31) from 2005 to 2010 Fragmented industry with production capacity in 26 provinces (Cornell) Local Protectionism June / 50

18 U.S. and Chinese Auto Markets U.S. and Chinese Auto Markets in 2015 U.S. China PV Sales in 2015 (million) Number of Firms Capacity Utilization (%) Top 6 Market Share (%) Top 12 Market Share (%) Top 20 Market Share (%) (Cornell) Local Protectionism June / 50

19 Production Capacity in China in 2013 All except five provinces have auto production Average annual capacity is 980,000 per province in 2013 (Cornell) Local Protectionism June / 50

20 Production Capacity in the US in states have production Among them, average annual capacity one million per state (Cornell) Local Protectionism June / 50

21 Data Registration data at the individual vehicle level from 2009 to 2011: About 30 million observations Vehicles purchased by individuals: 90% Vehicles purchased by governments and commercial entities: 10% Registration county, model name, engine size, vehicle type: Aggregate to province, model, year: observations Other data sets: Demographic information (income distribution) from 2005 Census Prices and attributes from Polk and other sources Auxiliary data sets for gasoline prices, dealership network, etc (Cornell) Local Protectionism June / 50

22 Summary Statistics Excluding small and very expensive models, 286 models in 31 provinces during Summary Statistics of Key Attributes Variable Mean Std. Dev. Min Max Sales Real price (1000 Yuan) Fuel cost (Yuan/100km) Engine size (liter) Vehicle size(m 2 ) Auto transimission SUV Minivan Number of dealers Distance to headquarter (1000km) Number of observations: (Cornell) Local Protectionism June / 50

23 Higher Income HH More Likely to Buy Cars Fraction of Households by Monthly Income (Yuan) Year 4k 4k-8k 8k-12k 12k Among All Households Among Vehicle Buyers Source: Ford Automobile Buyer Survey and Annual Statistical Yearbook Segments: mini/small sedan, compact sedan, medium/large sedan, SUV, MPV Household income distribution for each segment micro moments (Cornell) Local Protectionism June / 50

24 Overview 2 Introduction 3 Theory 4 Industry Background and Data 5 Estimation and Results 6 Simulations and Welfare Analysis 7 Conclusion (Cornell) Local Protectionism June / 50

25 Two Key Steps 1 Identify local protection: separate it from other factors 2 Estimate its welfare impacts: a structural model (Cornell) Local Protectionism June / 50

26 Two Key Steps 1 Identify local protection: separate it from other factors 2 Estimate its welfare impacts: a structural model (Cornell) Local Protectionism June / 50

27 Berry-Logit Model ln( s jmt ) s 0mt = β 1 LCL jm + β 2 LCL jm SOE j + β 3 LCL jm JV j +X j θ αp j + τb j + δ m + η t + ξ jmt s jmt : quantity share of model j in province m and month t LCL: a dummy for local products. Two types of local products: HQ: 1 if j is produced by a firm headquartered in market m Plant: 1 if j is produced by a firm with a plant in market m B j, δ m, η t : brand, province, and year fixed effects Separate regressions for individual and institution purchases Parameters of interest: β 1, β 2, β 3, which capture home bias (Cornell) Local Protectionism June / 50

28 Local Protectionism: Modeling considerations Besides official documents discussed above, government protection against non-local firms can take many forms and is often hidden: Direct or indirect subsidies Entry barriers for dealers of nonlocal brands For modelling purposes, we model local protectionism as affecting consumer preference for local brands: Private brands enjoy little protection from the government. Its coefficient captures consumer preference that is not driven by policy The home bias for SOEs and JVs above that of private firms is considered as driven by local protectionism (Cornell) Local Protectionism June / 50

29 Instrumental Variables Prices are instrumented by BLP instruments and consumption tax. BLP IVs: Number of products within the same segment by the same firm, or by rival firms Consumption tax rate: Engine Size Before 9/1/08 After 9/1/08 1.0L or below 3% 1% L 3% 3% L 5% 5% L 9% 9% L 12% 12% L 15% 25% above 4.0L 20% 40% (Cornell) Local Protectionism June / 50

30 Logit Parameter Estimates OLS IV I IV II IV III Variables Est. S.E. Est. S.E. Est. S.E. Est. S.E. HQ*Private HQ*JV HQ*SOE Plant*Private Plant*JV Plant*SOE Ln(price) Distance No. of Dealers No. of observations Two comparisons: Local vs. non-local products: consumer preference Local private vs. local SOEs/JVs products: local protection Control for vehicle attributes, brand, year, and vehicle-type-by-province fixed effects (Cornell) Local Protectionism June / 50

31 Logit Parameter Estimates OLS IV I IV II IV III Variables Est. S.E. Est. S.E. Est. S.E. Est. S.E. HQ*Private HQ*JV HQ*SOE Plant*Private Plant*JV Plant*SOE Ln(price) Distance No. of Dealers No. of observations Two comparisons: Local vs. non-local products: consumer preference Local private vs. local SOEs/JVs products: local protection Control for vehicle attributes, brand, year, and vehicle-type-by-province fixed effects (Cornell) Local Protectionism June / 50

32 Logit Parameter Estimates OLS IV I IV II IV III Variables Est. S.E. Est. S.E. Est. S.E. Est. S.E. HQ*Private HQ*JV HQ*SOE Plant*Private Plant*JV Plant*SOE Ln(price) Distance No. of Dealers No. of observations Two comparisons: Local vs. non-local products: consumer preference Local private vs. local SOEs/JVs products: local protection Control for vehicle attributes, brand, year, and vehicle-type-by-province fixed effects (Cornell) Local Protectionism June / 50

33 Sales Impacts of Home Bias HQ*Private HQ*JV HQ*SOE Individual 4.08% 93.48% % Institution 24.61% % % Plant*Private Plant*JV Plant*SOE Individual 0.02% -7.69% 75.07% Institution 13.88% 0.72% 15.03% SOE s sales in the home market is higher by 232% on average for individual purchases, and by 560% for institutional purchases Bai et al. (2009): industries with more SOEs exhibit stronger protectionism: Local governments have more to gain from protecting SOEs As local government officials hold the right to appoint the chief executives of SOEs, they have many more ways of milking the SOEs as compared with other enterprises. (Cornell) Local Protectionism June / 50

34 Boundary Discontinuity Analysis To address unobserved preference, focus on counties that share provincial boarders Assume the same unobserved preference for different models in each county-cluster (Cornell) Local Protectionism June / 50

35 Boundary Discontinuities: Counties All Clusters Clusters Clusters Est. SE Est. SE Est SE Est SE HQ*Private HQ*JV HQ*SOE County FE Yes Yes Yes Yes Year FE Yes Yes Yes Yes Model FE Yes Yes Yes Yes Cluster-model FE No No Yes Yes No. of obs No. of clusters The last column restricts to clusters where the ratio between the highest and lowest GDP per capita is no larger than 1.6 (Cornell) Local Protectionism June / 50

36 Structural Estimation: BLP A random coefficient discrete choice model of vehicle demand with rich consumer heterogeneity: 1 Define consumers choice set: all available vehicle models; each choice is defined as a bundle of attributes 2 Define household utility function over product attribute space with preference heterogeneity Go to utility function 3 Generate household choice probabilities and aggregate to market shares 4 Match predicted market shares of different vehicle models with observed market shares 5 Estimate preference parameters and conduct welfare analysis (Cornell) Local Protectionism June / 50

37 Utility Function Each year, a household decides whether to buy a new vehicle and what to buy among J models. Choice set: {0, 1, 2,...J} We model local protection as price discounts for local products Utility from vehicle model j: K u mtij = α mti p mtj + X mtjk β mtik + δ m + η t + ξ mjt + ɛ mtji k=1 α mti = eᾱmti +α 1 lny mti +σν mti p mtj = p 0 tj (1 γ 1 LCL mj JV j γ 2 LCL mj SOE j ) β mtik = β k + σ k ν mtik Arranging terms, we have: u mtij = δ mtj (θ 1 ) + µ mtij (θ 2 ) (Cornell) Local Protectionism June / 50

38 Choice Probabilities and Market Shares Utility from not buying: u mti0 = ɛ mti0. ɛ mtij, j = {0, 1,..., J}: i.i.d. type I extreme value distribution Define the probability of household i choosing vehicle model j: exp(δ mtj (θ 1 ) + µ mtij (θ 2 )) Pr mtij = 1 + h [exp(δ mth(θ 1 ) + µ mtih (θ 2 ))] Aggregate individual choice probabilities to get market shares: S mtj = Pr mtij df (µ mtij (θ 2 )) (Cornell) Local Protectionism June / 50

39 Estimation Strategy Simulated GMM with nested contraction mapping Inner loop: BLP contraction mapping to recover the mean utility δ mtj conditional on individual specific parameters θ 2 Outer loop: Form aggregate moment conditions based on the recovered mean utility: observed vehicle attributes and instruments are uncorrelated with unobserved product attributes Form micro-moment conditions based on simulated household choices Stack both sets of moment conditions, properly weighted, to form GMM objective function (Cornell) Local Protectionism June / 50

40 Supply Side Firms maximize static profit by choosing prices: π f = FOC: M [ ] (pj 0 t j (pj 0, X j ) mc j )q jm (p 0, θ) m=1 j F h (p 0 h t h mc h ) Q h p 0 j The equilibrium price in matrix notation: p 0 = mc + 1 q(p 0, θ) + Q j (1 t j pj 0 ) = 0 jh = { Q h if models j and h produced by same firm pj 0 0 otherwise (Cornell) Local Protectionism June / 50

41 Parameter Estimates: Part 1 (1) (2) Est. S.E. Est. S.E. LCL*JV, γ LCL*SOE, γ eᾱ(eᾱ1 ) eᾱ eᾱ eᾱ ln(income), α The first specification requires taste for price to vary continuously with income. The second specification allows each income group to have a distinct base taste for price Both control for vehicle attributes, brand, year, and vehicle-type-by-province fixed effects (Cornell) Local Protectionism June / 50

42 Parameter Estimates: Part 2 (1) (2) Est. S.E. Est. S.E. Ln(fuel cost) Ln(engine size) Ln(vehicle size) Auto Transmission Dealer Distance(1000km) Random Coefficients σ for price σ for constant σ for ln(fuel cost) σ for ln(engine size) There is a large dispersion of price sensitivity even at a specific income level (Cornell) Local Protectionism June / 50

43 Price Elasticities The range of price elasticity is to -6.43, with a mean of (Cornell) Local Protectionism June / 50

44 Net Profit Margins Net profit margin varies from 16.3% to 36.0%, with a mean of 22.2% (Cornell) Local Protectionism June / 50

45 Overview 2 Introduction 3 Theory 4 Industry Background and Data 5 Structural Model and Results 6 Simulations and Welfare Analysis 7 Conclusion (Cornell) Local Protectionism June / 50

46 Impact of Local Protection on Prices (2009) 6% 5% 4% 3% 2% 1% 0% 1% 2% 3% 4% All private firms cut prices after local protection is introduced Price response by SOEs and joint ventures are mixed (Cornell) Local Protectionism June / 50

47 Impact on Home-market Sales (2009) Sales with local protection Sales without local protection Local protection more than doubles home market sales for most SOEs. Home-market sales by move JVs increase marginally (Cornell) Local Protectionism June / 50

48 Impact on Total Sales (2009) 18% 16% 14% 12% 10% 8% 6% 4% 2% 0% 2% Total sales increase by about 40,000, or 0.6% of existing sales SOEs and JVs gain sales at the expense of private firms (Cornell) Local Protectionism June / 50

49 Welfare Impacts Loss of consumer surplus: Choice distortion local protectionism leads to suboptimal choices Price effect firms raise prices due to increased market power in their home markets Loss of producer surplus: Production distortion production shifts to high-cost producers Other welfare losses: Welfare loss from taxation to finance local protection (i.e., through subsidies) Dynamic impacts on entry and exit, capacity utilization, and resource allocation across regions (Cornell) Local Protectionism June / 50

50 Consumer Welfare Loss by Province (2009) Welfare loss(bn yuan) Original Price New Equilibrium Price We measure welfare loss by the difference between total subsidies and the total lump-sum transfers that would make consumers as well-off Total welfare loss is 3.5 billion yuan in 2009, out of which 2.0 billion is driven directly by local protection, while the other 1.5 billion is driven by price changes induced by local protection (Cornell) Local Protectionism June / 50

51 Impact on Firm Profits (2009) 14% 12% 10% 8% 6% 4% 2% 0% 2% Total profits increase by 2.5 billion yuan, or 1.37% percent of current profits Percentage profit changes for private firms, SOEs and JVs are -1.76%, 4.17%, and 1.35%, respectively. (Cornell) Local Protectionism June / 50

52 Conclusion Strong local protectionism exists in the auto market: More than doubles the sales of local SOEs Increases the sales of local JVs by 70% Leads to choice distortions and welfare loss: About 12 billion Yuan ($1.9 bn) from 2009 to 2011 Taxation and subsidies are considered transfers and not included in the welfare analysis. Losses from taxation is excluded from our current analysis The policy has benefited rich car buyers at the expense of the average taxpayer (Cornell) Local Protectionism June / 50

53 Conclusion Dynamic inefficiencies of local protection: Diverts production from low-cost producers (JVs and private firms) to high-cost producers (SOEs) Leads to a large number of inefficient firms Induces excess capacity and low capacity utilization Although China has made great stride in integrating with the world economy, intra-country trade barriers still exist Future reform that aims at eliminating local barriers and facilitating a national market would help economic growth (Cornell) Local Protectionism June / 50