Modeling Supply and Demand in the Chinese Automobile Industry. Yuan Chen, C.-Y. Cynthia Lin Lawell, and Erich J. Muehlegger

Transcription

1 Modeling Supply and Demand in the Chinese Automobile Industry Yuan Chen, C.-Y. Cynthia Lin Lawell, and Erich J. Muehlegger University of California at Davis 34th USAEE North American Conference

2 GX7 Produced by Geely

3 Qin-100 Plug-in Electric Hybrid Produced by BYD

4 Hangzhou, Zhejiang Province

5 Hangzhou, Zhejiang Province

6 Agenda Introduction Related Literature Econometric Model Data Preliminary Results Conclusion

7 Agenda Introduction Related Literature Econometric Model Data Preliminary Results Conclusion

8 Motivation China is experiencing rapid economic growth and, along with it, rapid growth in vehicle ownership Evidence from Chinese cities suggests average annual growth rates in per capita vehicle ownership of 10% to 25% Vehicle ownership increased by nearly 56 times between 1990 and 2011 Rapid growth in vehicle ownership and vehicle usage is linked to increasing global warming, emissions, air pollution, and other problems

9 Number of firms and models in Chinese automobile industry Source: Li, Xiao and Liu (2015)

10 Passenger Vehicle Output and Sales Number of Vehicles BPV(output) BPV(Sales) MPV(output) MPV(Sales) SUV(output) SUV(Sales) Total Output Total Sales Note: Data for SUVs and MPVs starts in 2008 Source: China Auto Market Almanac; China Automobile Industry Year Book

11 Chinese Automobile Policy 1994: China s National Development and Reform Commission (NDRC) implemented a policy encouraging state-owned firms to partner with international car makers to form joint ventures 2001: China entered the WTO and gradually cut the import tariffs on automobiles from 100% to 25% during the 5-year transitional period 2010: Energy Saving Projects - subsidy to car makers for low displacement cars to encourage energy saving

12 Market Structure Source: Hu, Xiao and Zhou (2014)

13 Geographic Location FAW Group Beijing Group Chery SAIC Group Geely Chang an Automotive Group Dongfeng Group Guangdong Auto Group BYD

14 Research Objectives Analyze supply and demand in the Chinese automobile industry Analyze what factors affect thedemand and cost of vehicles Analyze how consumers in China trade off various vehicle characteristics (such as fuel efficiency, whether the vehicle is an electric vehicle, etc.) with each other and with price Analyze the demand and cost for new vehicles Analyze the effects of various government policies, including those regarding alternative vehicles, on demand and cost

15 Methodology Develop and estimate a structural econometric model of demand and supply in the Chinese automobile market Estimate demand and cost parameters for all vehicles in China Estimate demand- and cost-side parameters, own- and cross-price elasticities, markups, and variable profits for alternative vehicles Use the model to simulate the demand and cost for new vehicles, and also the effects of various government policies on demand and cost

16 Advantages of structural econometric models Use actual data on vehicle purchase decisions, instead of self-reported survey data Comprehensive data set on the vehicle purchase decisions of all vehicle owners in China, not just those of the consumers that are surveyed Econometric model will enable me to statistically control for multiple factors that may affect vehicle purchase decisions Parameters can be used to calculate consumer utility, firm profits, and welfare Estimate standard errors and confidence intervals for my parameters, which enables me ascertain whether my parameters are statistically significant Simulate demand, supply, and welfare under counterfactual policy scenarios

17 Contributions to existing literature Model the Chinese automobile market using a structural econometric model Estimate cost and demand parameters relating to alternative vehicles Model the behavior of the state-owned automobile companies Analyze the effects of counterfactual policies on supply, demand, and welfare

18 Agenda Introduction Related Literature Econometric Model Data Preliminary Results Conclusion

19 Related Literature Structural econometric models of demand and supply in differentiated product markets Berry (1994); Berry, Levinsohn and Pakes (1995); Petrin (2002); Berry, Levinsohn and Pakes (2004); Dube, Fox and Su (2012); Knittel and Metaxoglou (2014); Berry and Haile (2014); Berry and Haile (2015); Bajari et al. (2015) Vehicle markets and policy Gallagher and Muehlegger (2011); Beresteanu and Li (2011); Sallee (2011); Sallee and Slemrod (2012); Jacobsen (2013); Heutel and Muehlegger (2015); Jacobsen and van Benthem (2015); Li and Zhou (2015); Aghion et al. (2016); Sallee, West and Fan (2016); Anderson and Sallee (forthcoming)

20 Related Literature Vehicle markets and policy in China Huo et al. (2007); Lin and Zeng (2013), Lin and Zeng (2014); Hu, Xiao and Zhou (2014); Li, Xiao and Liu (2015); Liu and Lin Lawell (2016) Mixed oligopoly De Fraja and Delbono (1989); De Fraja and Delbono (1990); Fjell and Pal (1996); White (1996); Poyago-Theotoky (2001); De Fraja (2009); Lutz and Pezzino (2010); Bennett and La Manna (2012); Haraguchi and Matsumura (2016)

21 Agenda Introduction Related Literature Econometric Model Data Preliminary Results Conclusion

22 Econometric Model A traditional logit model of vehicle demand assumes the independence of irrelevant alternatives, and can therefore generate unrealistic substitution patterns. In a logit model, if you take away a car model from the choice set, then consumers of that car will buy other cars according to their market shares. However, in reality, if you remove, say, a luxury car, the consumers of that luxury car are probably more likely to buy another luxury car than a random consumer would, even if luxury cars have low market share. In contrast, a Berry, Levinsohn and Pakes (1995) random coefficients model of vehicle demand addresses this problem by allowing for interactions between unobserved consumer characteristics and observed product characteristics, thus allowing different consumers to vary in how much they like different car characteristics.

23 Our innovations to Berry, Levinsohn and Pakes (1995) Model Chinese automobile market using a structural econometric model Estimate cost and demand parameters relating to alternative vehicles Model the behavior of the state-owned automobile companies Analyze the effects of counterfactual policies on supply, demand, and welfare

24 Model Assumptions Consumers are price-takers Bertrand (Nash-in-prices) equilibrium among multiproduct firms

25 Traditional logit model Utility of consumer i for vehicle model j is given by: u "# x #, p #, ξ #, θ = δ # + ε "# where: δ # = x # β αp # + ξ # x # = x #6 6 is a vector of observable vehicle characteristics k for vehicle model j p # is the price of vehicle model j β 6 is the mean taste parameter for vehicle characteristic k ξ # is a vector of unobservable vehicle characteristics for vehicle model j

26 Traditional logit model Utility of consumer i for vehicle model j is given by: u "# x #, p #, ξ #, θ = δ # + ε "# It is usually assumed that consumer characteristics only enter as ε "#, and E ε = 0 => ξ # is the mean (across consumers) of the unobserved component of utility => δ # is the mean of the utility from good j

27 Traditional logit model Utility of consumer i for vehicle model j is given by: u "# x #, p #, ξ #, θ = δ # + ε "# All substitution effects depend only on the mean utilities δ #. Conditional on market shares, substitution patterns do not depend on the observable characteristics of the product. Additive separability plus i.i.d. ε ; s generates unreasonable substitution patterns. Independence of Irrelevant Alternatives

28 Random Coefficients Model of Demand Random coefficients specification for the utility of consumer i: μ "# = δ # + ν "# where δ # is the common component of the utility for vehicle model j and is given by: δ # = x # β αp # + ξ # and the first two terms in the idiosyncratic component ν "# product characteristics: interact consumer and ν "# = >x #6 σ 6 ζ "6 1 p y # + ε "# 6 " x # = x #6 : a vector of observable vehicle characteristics k for vehicle model j 6 ξ # is a vector of unobservable vehicle characteristics for vehicle model j p # is the price of vehicle model j β 6 is the mean taste parameter for vehicle characteristic k ζ "6 is a characteristic of consumer i that affects i s taste for vehicle characteristic k

29 Random Coefficients Model of Demand Random coefficients specification for the utility of consumer i: μ "# = δ # + ν "# where δ # is the common component of the utility for vehicle model j and is given by: δ # = x # β αp # + ξ # and the first two terms in the idiosyncratic component ν "# product characteristics: interact consumer and ν "# = >x #6 σ 6 ζ "6 1 p y # + ε "# 6 " x # = x #6 : a vector of observable vehicle characteristics k for vehicle model j 6 ξ # is a vector of unobservable vehicle characteristics for vehicle model j p # is the price of vehicle model j β 6 is the mean taste parameter for vehicle characteristic k ζ "6 is a characteristic of consumer i that affects i s taste for vehicle characteristic k ε "# is distributed type I extreme value ζ "6 has a standard normal distribution => mean and variance of the marginal utilities associated with characteristic k are β 6 and σ 6 E y " is log normally distributed.

30 Estimation Equation for Demand δ # s # = x # β αp # + ξ # where: δ # is the common component of the utility for vehicle model j s # is the share of consumers who purchase vehicle model j δ # s # is calculated from inverting the expected market share equation, where the expectation is taken over the distribution of consumer characteristics x # = x #6 is a vector of observable vehicle characteristics k for vehicle 6 model j ξ # is a vector of unobservable vehicle characteristics for vehicle model j p # is the price of vehicle model j β 6 is the mean taste parameter for vehicle characteristic k

31 Model of Supply Profits π H for firm f are: π H p, z, ξ, ω #, θ = > p # Ms # x, ξ, p, θ L C # (q #, w #, ω #, γ) where: #SFH M is the market size s # is the market share C # is the total cost

32 Pricing Equation Assuming the existence of a Bertrand (Nash-in-prices) equilibrium, the pricing equation is: where p Δ VW s = c Δ is a matrix in which Δ #6 = YZ [ Y\ ] if j and k are produced by the same firm and Δ #6 = 0 otherwise marginal cost c # is given by: c # = x # γ + ηq # + ω # where ω # are the unobservable cost variables.

33 International Joint Ventures If a domestic firm forms a joint venture with an international firm, then its profits may be affected. Estimation equation on the DEMAND side is the same: δ # s # = x # β αp # + ξ # Estimation equation on the SUPPLY side: c # = x # γ + wγ + ηq # + ω # where: w is a joint venture dummy vector, of which the f cd element equals one if the maker that produces good j forms a joint venture with company f ω # are the unobservable cost variables

34 Different Objectives for State-Owned Firms On the SUPPLY side, we allow state-owned firms to have different objectives from private firms. We assume the objective function (utility function) for state-owned firms is given by: ( ) U() = ρπ() + ρ CS() + ρtar() + 1 ρ ρ HYB() where: π f is profit CS f is consumer surplus TAR f is the squared deviation from a target number of hybrid vehicles HYB f is the number of hybrids produced by that firm The ρ ; s are weights put on the possible terms in the state-owned company s utilityfunction.

35 Estimation Equation for Supply Estimation equation for both types of firms on the SUPPLY side is: c # = x # γ + wγ + ηq # + ω # where: w is a joint venture dummy vector, of which the f cd element equals one if the maker that produces good j forms a joint venture with company f ω # are the unobservable cost variables. However, the pricing equations are different for firms with different ownerships For private firms: p Δ VW s = c For state-owned firms: p Δ VW s ρ E ρ W e + 1 ρ W (2ρ s f M > s t t h)) + 1 ρ E ρ s Δ t f 1 = c

36 Effects of International Joint Ventures To examine whether forming joint ventures with international firms would help the domestic automobile technology development, we create a has international joint venture (HIJV) dummy variable and interact it with some of the technology-related car characteristics. HIJV = y 1, has formed joint venture s with any international firms 0, otherwise The technology-related car characteristics that we interact with the HIJV dummy are: the dummy for the car j being alternative vehicle fuel efficiency horsepower Hence, the estimation equation on the supply side are: c # = x # γ + wγ + ηq # + ψ f tech " c ˆ c"š + ω # where tech " c ˆ c"š is the HIJV dummy interacted with the chosen characteristics.

37 Parameters to be estimated Means β of the marginal utility associated with each particular characteristic Standard deviations σ of the marginal utility associated with each particular characteristic Parameter α in the marginal disutility of price Cost parameters γ Cost parameter Γ associated with jointventurecompanies Coefficient η on quantity in marginal cost Coefficient ψ on the has international joint venture dummy interacted with technology variables Weights ρ on different terms in state-owned firms utilityfunction Target number h of hybrids to produce

38 Instruments Instruments for endogenous price, market share, and vehicle characteristics: The number of cars with similar values of the characteristic Car j and k are similar in characteristic r if x # x 6 E < W W max x min x E The value of the characteristic for the car k closest to car j in the value of the characteristic These are good instruments for price in the demand equation because characteristics of other cars k of other firms are independent of the utility for a particular car j, and because they are correlated with price via the markup in the supply-side first-order conditions. Characteristics of other cars k of other firms also serve as good instruments for the market share of car j in the supply-side pricing equation.

39 Calculating the common component of utility δ # δ # s # is calculated from inverting the expected market share equation, where the expectation is taken over the distribution of consumer characteristics Step 1: Calculate market share function conditional on consumer characteristics (y ", ζ "# ) Step 2: Use simulation method to get the market shares Step 3: δ = δ + ln s ln [s p, x, δ, P Z ; θ solve for δ recursively

40 Econometric Model Compute the demand-side unobservable ξ as the residual in the common component of the demand-side utility estimation equation. Compute the cost-side unobservable ω as the residual in the supply-side marginal cost estimation equation, where the marginal cost is given by the pricing equation. Interact the instruments with the computed demandand cost-side unobservables to form the moment conditions

41 Econometric Estimation The demand and supply side equations are jointly estimated using instruments for endogenous price, market share, and vehicle characteristics via generalized method of moments (GMM)

42 Agenda Introduction Related Literature Econometric Model Data Preliminary Results Conclusion

43 Data Price and Quantity sales: Quantity sales data for from China Auto Market Almanac Price data from China Automotive Industry Year book Manufacturer s Suggested Retail Price (MSRP) data from Vehicle characteristics: State-owned dummy variable baike.baidu.com China Industry Business Performance Data for 2013 Adult population: Adult population (age 15-64) for from World Development Indicators Urban Income: urban income across all provinces for from China Statistical Year Book.

44 Data on Vehicle Characteristics (1) Basic information: year; dummy for vehicle manufacturers; vehicle type (2) Information about vehicle engines: cylinder layout types; number of cylinders, etc. (3) Information about powertrain: top speed (km/h); acceleration from 0 to 100km/h (in seconds); horsepower (PS); transmission types; number of transmission speeds; types of drivetrain; types of four wheel drive; types of power steering (4) Information about fuel: dummy variable fuel sources, fuel efficiency (100km/L); displacement (in ml and L); ways of air intake (5) Dimensions: length, width, height, wheelbase; number of doors, passenger capacity (number of seats); (6) Safety equipment: frontal driver air bag; side airbag; brake ABS; front radar; rear radar; back up camera; remote control key; keyless active feature; keyless entry feature; (7) Exterior features: electronic sunroof; panorama sunroof. (8) Interior features: heated front seats; heated rear seats, ventilated front seats; ventilated rear seats; GPS; bluetooth interface; build-in TV; Airconditioner. (9) Advanced technologies: park assist; side assist. (10)Electric engine: total power of electric engine (kw); the torque of the electric engine (Newton-metre); energy density (kwh) and charge-depleting range (km).

45 Dimensions and capacities Engine Drivetrain Chassis

46 Summary Statistics for Chosen Vehicle Characteristics Variable # Obs Mean Std. Dev. Min Max Alternative vehicle (dummy) Fuel efficiency (100km/L) Length (mm) Weight (kg) Capacity (number of seats) Horsepower (PS)

47 State-owned dummy variable Since the majority of car companies in China operate under the shareholding system, in a strict sense, there are few state-owned car companies. However, governments do hold a majority of the stocks of some of thecompanies. Throughout the paper, a state-owned firm is defined as a car manufacturer of which a majority of stock of its parent company is held by governments (either the central government or local governments), although some of its stock might be held by foreign companies such as those joint venture firms.

48 Agenda Introduction Related Literature Econometric Model Data Preliminary Results Conclusion

49 First-stage F-statistics Endogenous variables Instruments Chosen First-Stage F-Statistic Alternative Fuel Vehicle n_sim_fueleff Fuel Efficiency n_sim_length n_sim_weight Length n_sim_hp n_sim_capacity Weight fueleff_sim1 length_sim1 weight_sim Capacity hp_sim1 fueleff_sim1_alter Horsepower weight_sim1_alter other_altervehicle_frac Price has_inter_other_altervehicle_fra has_inter_n_sim_fueleff has_inter_n_sim_hp Quantity

50 Model with full set of interactions between consumer preferences and product characteristics Characteristics chosen: alternative vehicle, fuel efficiency, length, weight, capacity (number of seats), and horsepower Compare my estimation results with and without including the state-owned dummy variable in the supply-side pricing equation.

51 Results Mean β of marginal utility of: Alternative vehicle 0.190*** (dummy) (0.025) Fuel efficiency 0.050* (100km/L) (0.021) Length (m) 0.160*** (0.033) Weight (metric ton) 0.170*** (0.032) Capacity (0.029) Horsepower (PS) 0.210*** (0.023) Constant 0.08* (0.032) (1) (2) 0.700*** (0.004) 0.600*** (0.012) 0.150*** (0.007) 0.380*** (0.006) *** (0.004) 0.370*** (0.003) 0.150*** (0.001) (1) (2) Standard deviation σ of marginal utility of: Alternative vehicle (dummy) Fuel efficiency (100km/L) 0.160*** (0.042) 0.450*** (0.055) Length (m) 0.510*** (0.054) Weight (metric ton) 0.230*** (0.028) Capacity (0.021) Horsepower (PS) 0.350*** (0.044) Constant 0.080*** (0.027) 0.20*** (0.004) 0.420*** (0.005) 0.020*** (0.004) 0.250*** (0.005) 0.480*** (0.010) 0.330*** (0.007) 0.400*** (0.005) Parameter α in marginal disutility of price (10,000 yuan) 0.480*** (0.057) 0.330*** (0.005)

52 Results (1) (2) Coefficient γ in marginal cost on: Alternative vehicle (dummy) 0.230*** (0.027) Fuel efficiency (100km/L) 0.290*** (0.046) Length (m) 0.070*** (0.019) Weight (metric ton) 0.170*** (0.034) Capacity 0.360*** (0.041) Horsepower (PS) 0.600*** (0.061) 0.450*** (0.003) 0.280*** (0.002) 0.190*** (0.006) 0.020*** (0.004) 0.700*** (0.004) 0.150*** (0.006) State-owned (dummy) 0.110*** (0.012) Constant 0.160*** (0.026) 0.130*** (0.006) Coefficient η on quantity in marginal cost 0.010*** (0.002) 0.001** (0.000) # Observations Notes: Standard errors are reported in parentheses. Significance codes: * p<0.05; ** p<0.01; ***p<0.001.

53 Model with international joint venture dummies, possibly allowing statedowned firms to have different objectives, For the model allowing state-owned firms to have different objectives, we are estimating four more parameters: ρ W, ρ E, ρ s, h.

54 Results Mean β of marginal utility of: Alternative 0.160*** vehicle (dummy) (0.012) Fuel efficiency 0.090*** (100km/L) (0.009) Length (m) 0.120*** (0.006) Weight (metric 0.200*** ton) (0.007) Capacity 0.520*** (0.008) Horsepower (PS) 1.440*** (0.004) Constant 0.300*** (0.007) (1) (2) 0.270*** (0.065) 0.220*** (0.061) (0.076) (0.059) 0.110** (0.042) (0.085) (0.064) (1) (2) Standard deviation σ of marginal utility of: Alternative vehicle (dummy) 0.120*** (0.008) 0.220*** (0.054) Fuel efficiency (100km/L) 0.300*** (0.004) (0.07) Length (m) 0.110*** (0.009) (0.046) Weight (metric ton) 0.460*** (0.005) 0.130* (0.051) Capacity 0.230*** (0.005) 0.250*** (0.072) Horsepower (PS) 0.470*** (0.012) 0.270*** (0.051) Constant 0.310*** (0.008) 0.090*** (0.058) Parameter α in marginal disutility of price (10,000 yuan) 0.230*** (0.002) 0.150*** (0.060)

55 Results (1) (2) Coefficient γ in marginal cost on: Alternative vehicle (dummy) 0.560*** (0.001) 0.280*** (0.082) Fuel efficiency (100km/L) 0.170*** (0.013) (0.089) Length (m) 0.910*** (0.012) 0.340*** (0.050) Weight (metric ton) 0.300*** (0.001) 0.170*** (0.033) Capacity 0.340*** (0.008) 0.120* (0.049) Horsepower (PS) 0.240*** (0.021) 0.300*** (0.060) State-owned (dummy) 0.240*** (0.002) (0.067) Constant 0.470*** (0.009) 0.170** (0.058) Coefficient η on quantity in marginal cost 0.001*** (0.000) 0.001** (0.000) Notes: Standard errors are reported in parentheses. Significance codes: * p<0.05; ** p<0.01; ***p<0.001.

56 Results (1) (2) Coefficient Γ in marginal cost on joint venture with: Mazda 0.090*** (0.007) Honda 0.130*** (0.006) Kaihatsu 0.630*** (0.003) Toyota 0.380*** (0.003) Suzuki 0.260*** (0.003) Nissan 0.020* (0.010) Mitsubishi 0.340*** (0.008) Isuzu 0.970*** (0.009) Daimler 0.210*** (0.012) BMW 0.270*** (0.002) Volkswagen 0.300*** (0.004) 0.160*** (0.045) (0.072) (0.069) (0.057) (0.056) (0.065) (0.061) 0.200*** (0.046) 0.160* (0.063) (0.053) 0.380*** (0.073) Audi 0.230*** (0.006) Lotus 0.040*** (0.002) Rover 0.160*** (0.001) GM 0.070*** (0.010) Ford 0.840*** (0.008) Chrysler 0.240*** (0.014) Hyundai 0.190*** (0.001) Kia 0.920*** (0.014) Volvo 0.750*** (0.012) Saab 0.500*** (0.008) PSA 0.040*** (0.011) Fiat 0.440*** (0.009) (1) (2) (0.060) (0.058) 0.280*** (0.067) 0.220*** (0.060) 0.180* (0.065) (0.072) 0.160* (0.069) 0.360*** (0.057) (0.070) (0.056) 0.290*** (0.063) 0.26*** (0.059)

57 Results (1) (2) Estimated parameters in state-owned firms utility ρ W weight on profit ρ E weight on consumer surplus ρ s weight on squared deviation from target number of hybrid vehicles h target number of hybrid vehicles 0.260*** (0.054) 0.250*** (0.069) 0.160*** (0.046) *** (0.000) # Observations Notes: Standard errors are reported in parentheses. Significance codes: * p<0.05; ** p<0.01; ***p<0.001.

58 Model with international joint venture dummies with interactions between has international joint venture dummy and technology variables, possibly allowing stated-owned firms to have different objectives,

59 Results Mean β of marginal utility of: Alternative vehicle 0.312*** (dummy) (0.0006) Fuel efficiency 0.114*** (100km/L) (0.0011) Length (m) *** (0.0006) Weight (metric ton) 0.483*** (0.0009) Capacity 0.528*** (0.0006) Horsepower (0.01PS) 0.438*** (0.0006) Constant *** (0.0009) (1) (2) 0.030** (0.0111) 0.210*** (0.0071) *** (0.0090) 0.170*** (0.0073) *** (0.0113) 0.090*** (0.0114) 0.080*** (0.0112) (1) (2) Standard deviation σ of marginal utility of: Alternative vehicle (dummy) 0.521*** (0.0003) *** (0.0088) Fuel efficiency (100km/L) 0.395*** (0.0014) 0.180*** (0.0102) Length (m) 0.534*** (0.0011) 0.120*** (0.0068) Weight (metric ton) 0.334*** (0.0011) 0.120*** (0.0071) Capacity 0.356*** (0.0011) 0.150*** (0.0150) Horsepower (0.01PS) 0.135*** (0.0014) 0.040*** (0.0085) Constant 0.758*** (0.0006) 0.090*** (0.0040) Parameter α in marginal disutility of price (10,000 yuan) 0.242*** (0.0006) 0.120*** (0.0117)

60 Results Coefficient γ in marginal cost on: Alternative vehicle (dummy) 0.049*** (0.0003) Fuel efficiency (100km/L) 0.335*** (0.0014) Length (m) 0.424*** (0.0011) Weight (metric ton) 0.685*** (0.0014) Capacity 0.203*** (0.0009) Horsepower (PS) 0.526*** (0.0011) State-owned (dummy) 0.518*** (0.0006) Constant 0.531*** (0.0003) Coefficient η on quantity in marginal cost *** (0.0003) (1) (2) 0.090*** (0.0138) 0.100*** (0.0142) (0.0124) 0.220*** (0.0076) 0.060*** (0.0104) 0.120*** (0.0178) 0.160*** (0.0050) 0.060*** (0.0097) 0.040*** (0.0112)

61 Results (1) (2) (1) (2) Coefficient Γ in marginal cost on joint venture with: Mazda *** 0.090*** Honda (0.0009) (0.0068) 0.099*** 0.180*** Kaihatsu (0.0003) (0.0065) 0.258*** 0.190*** Toyota (0.0006) (0.0105) 0.508*** 0.090*** Suzuki (0.0006) (0.0074) 0.149*** *** Nissan (0.0003) (0.0088) 0.138*** 0.330*** Mitsubishi (0.0006) (0.0064) *** 0.160*** Isuzu (0.0003) (0.0115) 0.371*** 0.210*** Daimler (0.0003) (0.0104) 0.707*** 0.050*** BMW (0.0009) (0.0131) 0.357*** *** (0.0009) (0.0039) Volkswagen 0.268*** (0.0006) 0.230*** (0.0089) Audi Lotus 0.235*** (0.0014) 0.177*** Rover (0.0009) *** GM (0.0009) 0.01*** Ford (0.000) 0.674*** Chrysler (0.0011) 0.368*** Hyundai (0.0006) 0.416*** Kia (0.0011) 0.523*** Volvo (0.0009) 0.656*** Saab (0.0011) *** PSA (0.000) 0.452*** Fiat (0.0006) 0.034*** (0.0011) *** (0.0099) 0.260*** (0.0128) 0.310*** (0.0095) 0.260*** (0.0149) 0.180*** (0.0147) *** (0.0108) *** (0.0081) (0.0136) 0.080*** (0.006) 0.080*** (0.0118) 0.230*** (0.0151) *** (0.0101)

62 Results Estimated parameters in state-owned firms utility on different objectives (1) (2) ρ W weight on profit 0.850*** (0.0100) ρ E weight on consumer surplus 0.090*** (0.0141) ρ s weight on squared deviation from target number of hybrid vehicles 0.050*** (0.0131) h target number of hybrid vehicles *** (0.0000) Estimated parameters in Has international joint venture dummies interaction terms ψ W on HIJV * alternative vehicle 0.522*** (0.0006) ψ E on HIJV * fuel efficiency 0.440*** (0.0006) ψ s on HIJV * horsepower 0.200*** (0.0000) 0.070*** (0.0077) 0.230*** (0.0042) 0.170*** (0.0096) # Observations Notes: Standard errors are reported in parentheses. Significance codes: * p<0.05; ** p<0.01; ***p<0.001.

63 Agenda Introduction Related Literature Econometric Model Data Preliminary Results Conclusion

64 Ongoing work Calculate welfare Simulate the effects of counterfactual scenarios on supply, demand, and welfare

65 Conclusions It is important to allow for consumers to vary in how much they like different car characteristics State-owned firms may have different objectives

66 Thank you!

67 Projected Highway Vehicle (HWV) Stock Source: Huo (2007)