Effects of FTA Provisions on the Market Structure of the Korean Automobile Industry

Rev Ind Organ(2014)45:39–58DOI10.1007/s11151-014-9417-0Effects of FTA Provisions on the Market Structureof the Korean Automobile IndustryMinsoo Park·Hongjai RheeReceived:24August2012/Accepted:7January2014/Published online:20January2014©Springer Science+Business Media New York2014Abstract The Korean passenger car market is dominated by a few domestic producers. In this paper,we investigate whether this market structure is the result of import tariffs and taxes that penalize foreign manufacturers.Our counterfactual simulation under two FTA scenarios shows that the market shares of imported and domestic cars show minimal adjustments,suggesting that import tariffs and taxes may not explain the lack of penetration of import cars in Korea.We predict changes in consumer surplus and discuss the policy implications of ourfindings.Keywords FTA·Tariff elimination·Automobile industry·Nested logitJEL Classification F14·F17·L621IntroductionThe Korean passenger car market has long been criticized as one of the most closed markets in the world.Although the overall share of imported cars in the Korean market has grown,in2010it remained low at approximately7%.It is striking that no globally competitive car manufacturers from the US,the EU,or Japan accounts for more than a1%share in a market in which as many as1.5million new cars are sold per year. M.ParkDepartment of Economics,Sungkyunkwan University,25-2,Sungkyunkwan-ro,Jongno-gu,Seoul,Koreae-mail:minsoopark@H.Rhee(B)College of Business Administration,Ajou University,San-5,Wonchun-dong,Suwon,Koreae-mail:hrhee@ajou.ac.kr40M.Park,H.Rhee The recent free trade agreements(FTAs)between Korea and other auto-exporting countries such as the US(effective March2012)and the EU(effective July2011) are expected to have substantial economic implications for the Korean automobile market,because tariff-and tax-related barriers are scheduled to be eliminated or greatly reduced by the FTA provisions.In this paper,we predict the impacts of the FTA provisions on the competitive structure of the Korean passenger car market.Our analysis proceeds with the fol-lowing steps:First,we estimate demand for passenger cars using data for all new car registrations between2006and2009.Second,we simulate the optimum price of each product under two counterfactual scenarios of automobile tariffs and taxes. Third,we predict the counterfactual market shares under the two scenarios by apply-ing the estimated demand equations with some optimality assumptions with regard to supply-side behavior.Finally,we identify several managerial and policy implications by comparing the actual and counterfactual shares of imported cars in aggregate or by their countries of origin.The remainder of this paper is organized as follows:Sect.2summarizes the existing literature and the contributions of the present study.Section3introduces the Korean passenger car market with a brief summary of the current status of and scheduled changes in the Korean automobile tariff and taxation system.In Sect.4,we introduce the estimation model and data that are used in our analyses.We present and discuss the estimation and simulation results in Sect.5.Section6concludes the paper.2Literature ReviewThere is a vast amount of theoretical and empirical literature evaluating the economic effects of trade policy,which has long been an important but inconclusive topic in growth and international economics.1The impacts of FTAs have also been investi-gated in many studies(e.g.,Baier and Bergstrand2007);however,few studies have quantitatively predicted or assessed the consequences of FTAs in the basis of product-level data with the use of a structural supply and demand estimation model.2Among these,Feenstra(1988),Goldberg(1995),and Berry et al.(1999)investigated the effects of voluntary export restraints on Japanese car imports to the US by using the estimation strategies that we have implemented in the present study.Other previous studies of FTAs did not consider full equilibrium.For example,Berry et al.(1992)predicted the effect of the Mexico-US FTA on the Mexican automobile industry simply by making the assumption that the Mexican auto prices would fall to North American levels.Similarly,Tovar(2012)examined the consequences of a tariff reduction on the Colombian automobile industry,and Brambilla(2005)investigated the effect of a customs union agreement between Brazil and Argentina.Most recently,Hong(2011)analyzed the Korean passenger car market by predict-ing the changes that would be expected after the removal of Korea’s tariff barriers,1For a review of the openness-growth controversy,see Edwards(1998),Rodriguez and Rodrik(2000),and Singh(2010).2See Asian Development Bank(2010)for a summary of estimation methods.FTA Provisions on the Market Structure41 Table1Korean vehicle production and sales(in thousands of units)20032004200520062007200820092010Production3,1783,4693,6993,8404,0863,8273,5134,272 Domestic sales1,3181,0941,1431,1641,2191,1541,3941,465 Exports1,8152,3802,5862,6482,8472,6842,1492,772 Source:Korea automobile manufacturers association(KAMA2011)but employed an oversimplified assumption that the prices of domestic cars would remain unchanged before and after the implementation of the FTA.3This assumption is unrealistic because Korean car manufacturers have options to adjust their prices whenever it pays to do so.In contrast,we provide an equilibrium prediction that con-siders the best responses of all market players in the market to the proposed tariff and tax changes.3An Overview of the Market and FTA Provisions3.1The MarketThe automobile industry in Korea has been a key component of the country’s economic growth.The industry is currently thefifth largest in the world in production and is among the top ten largest in domestic market size,with more than one million registrations per year.Korean car manufacturers have been expanding their shares in markets across the world.For example,two Korean companies,Hyundai and Kia,sold 1.1million cars in the US market in2010,accounting for approximately8.9%of the share in unit sales(Automotive News2011).Table1reports the vehicle production and sales in Korea from2003to2010.The Korean market has long been perceived as preferring domestic over imported passenger cars.According to the Korea Automobile Importers and Distributors Asso-ciation(KAIDA)only90,000units of imported passenger cars were sold in2010, amounting to approximately6.9%of the entire passenger car market in Korea.Fig.1 shows the overall market share of imported cars in Korea from2001to2010.4 As noted earlier,the market share of imported cars in Korea,although it is increasing gradually,remains low compared to other countries.Ford(2010)reported that the average import share among OECD member countries is approximately40%of the market.Of the many possible reasons for the poor performance of imported cars in Korea,the Korean tariff and tax system for automobiles is specifically addressed in this paper.3While Hong(2011)considered only the effects of tariff elimination,we take into account domestic tax reform as an additional change in non-tariff barriers after FTA implementation.4In terms of imported cars alone,EU car manufacturers(63.4%and mostly German)constituted the largest share in2010,followed by Japanese(26.9%)and US manufacturers(9.0%)(KAMA2011).The best-selling imported car brand is BMW,followed by Mercedes-Benz.42M.Park,H.RheeFig.1The market share(%)of imported passenger cars in Korea.Source:KAMA(2011)3.2The Korean Tariff and Taxation System Before the FTABefore we investigate the impact of the FTA provisions on the Korean automobile market,it may help to briefly mention the pre-FTA tariff and taxation system for imported cars in Korea.First,every imported car carries an8%tariff that is based on cost,insurance,and freight(CIF)prices.Second,an individual consumption tax is levied,either as10%(for large cars)or5%(for small cars)of the tariff-included base prices for imported cars or the wholesale price for domestically produced cars.Third, other taxes(e.g.,education tax,V AT,and acquisition tax)are applied at the same rates for every domestic and imported car.In particular,foreign car manufacturers have complained about the substantial dis-crepancies in the individual consumption tax according to vehicle engine capacity (either5or10%).Although the tax rate itself is independent of production origin, they claim that the tax discriminates against imported cars,because most of the cars that are imported into Korea tend to have larger engine capacities than do domestic vehicles.It is interesting to note that the price discrepancy between domestic and imported cars is even larger than the tariff rate itself,because the tariff disadvantage is cascaded into other taxes.3.3Proposed Changes After the FTAAccording to the Korea-US FTA,Korea promised immediately to remove its10% tariff on US-made trucks and reduce the8%tariff on US-made passenger cars to4% and then phase out the auto tariff overfive years.Likewise,the Korea-EU FTA also contains a similar schedule for tariff removal over a span offive years.Table2summarizes the scheduled changes to automobile import tariffs in Korea. In addition,Korea has also agreed to address the allegedly discriminatory nature of its automobile taxation policy,which has been criticized by foreign car manufacturers as one of the most significant non-tariff barriers to market penetration.5Following the 5For detailed tariff-and non-tariff provisions on the automobile industry,see the summary reports by the Congressional Research Service(2011a,b)and the US International Trade Commission(2011).FTA Provisions on the Market Structure43Table2Schedule for automobile tariff reduction in KoreaVehicle type Current tariff ratein Korea(%)Korea-US FTA Korea-EU FTAPassenger cars8Reduce tariff from8to4%immediately andfully eliminate the tariffin year5Tariff eliminated over three orfive years depending on engine sizeElectric/plug-in hybrid cars8Tariff eliminated overfiveyearsTrucks10Eliminate tariffimmediately Tariff eliminated immediately or in three-to-five years depending on truck sizeThis table was adapted from the Congressional Research Service(2011b,p.9)agreement,the individual consumption tax on automobiles with engines larger than 2,000cc will drop from10to5%by2015.Given the scheduled implementation of FTA provisions,we posit the two that are scenarios described below for our investigation of the effect of the agreements on Korea’s passenger car market:Scenario#1.The import tariffs of8%are completely removed regardless of pro-duction origins(EU,the US,and even Japan),while the other taxation schedules remain intact.The reason we include Japanese cars in the tariff-exemption lists is that Japanese car manufacturers have substantial off-shore production bases so that they could possibly enjoy the benefits of Korea’s FTAs,as do other countries.Further-more,Korea is already in the process of negotiating FTAs with neighboring countries, including China and Japan.6Scenario#2.In addition to the abolition of import tariffs,every car in the market pays a5%individual consumption tax regardless of engine displacement.If foreign cars perform better in the large-sized category as foreign manufacturers claim,we expect this change to increase the shares of imported cars in the Korean market.In both scenarios,we hold constant all other exogenous variables,such as changes in consumer preference,technology,and automobile regulations and policies.74Model and Data4.1The Discrete Choice ModelThe model used in this study follows the equilibrium analysis of Berry(1994)and Berry et al.(1995,1999).In our paper,a logit demand structure nested by class and import status is used to investigate the impacts of FTAs on different classes and origins 6For recent trends in Asian FTAs,see Kawai and Wignaraja(2011).7Although changes in non-tariff measures,such as safety and emissions standards,are very important components of the FTA provisions,the schedules for these components has not yet been specified for quantification.44M.Park,H.Rhee of cars.Throughout the analysis,we assume that vehicle characteristics such as size, design,and engine capacity arefixed and that thefirms choose the optimal price(s)of their product(s)solely strategically.Because the vehicles are subject to import tariffs(if applicable)and purchase taxes, the net producer price(ˆp j)is discounted from the consumer price(p j)by tariff rates (t j),tax rates(τj),and any other transaction costs(λj)such thatˆp j=p jex j(1+t j)(1+τj)(1+λj).(1) By optimizing prices,we obtain the following supply condition(see Berry et al.1995,p.853):s j+k∈B fˆp k−mc k∂s k∂ˆp j=0,(2)where s j is the market share,mc j is the marginal cost of each car alternative j=1,2,...,J,and B f is the set of car models produced by eachfirm in the market, f=1,2,...,F.8Note that equilibrium market shares and prices are jointly deter-mined to account for costs,tariffs,taxes,and the own-and cross-price elasticities of demand.To simulate the effects of the FTA,we assume that the marginal costs are given at the current equilibrium level while the levels of tariffs and taxes are provided by the corresponding scenarios.Therefore,we are able to predict the post-FTA equilibrium shares and prices for a given tariff and tax scenario,once we know how the market shares depend on prices and other attributes by estimating a demand equation.Following Berry et al.(1995,1999),we employ a random utility discrete choice model for demand.In the model,the utility of every vehicle model is specified as9U i j=X jβ−αP j+ξj+εi j≡δj+εi j,(3) where X j is the vector of observable vehicle characteristics,P j is the consumer price,ξj is the random variable for unobservable(to researchers)vehicle characteristics,εi jis an unobservable idiosyncratic error term,andβandαare the parameters of interest to be estimated.10We assume that consumers select the maximum-utility alternative according to Eq.(3)when they choose to purchase vehicles.To avoid the well-known problem of independence of irrelevant alternatives(IIA)in a discrete choice model(McFadden1974),we apply a three-stage nested logit model as illustrated in Fig.2.First,consumers choose either a certain class of cars to buy,or no car.Passenger cars in Korea are classified into six categories:compact(below1,000cc8When afirm sells multiple car models,it maximizes the joint profit.9This specification of the indirect utility function is adopted from Berry(1994).10An indirect utility function is often used for this kind of demand estimation,but we do not have income in the function.Data on individual purchasing decision and income is needed to include the term(income-price).However,we do not have such data since only market-level aggregate sales data is available for our research.FTA Provisions on the Market Structure 45Fig.2Three-stage decision hierarchy for a vehicle purchasein engine capacity),small (1,000–1,600cc),medium (1,600–2,000cc),large (above 2,000cc),sports,and SUV .11Second,they decide whether to purchase a domestic vehicle or an imported vehicle from the class in the first stage.Finally,they choose a specific vehicle model in the selected categories (by class and production origin).In this setting,the market share of a specific vehicle model can be represented as the product of three stage-contingent shares in the hierarchy.For example,the market share (s j )of Sonata (the most popular mid-size domestic brand in Korea)in Fig.2is the product of (i)the share of mid-size cars (group)out of all cars (the first-stage share,s g ),(ii)the share of domestic cars (subgroup)out of mid-size cars (the second stage share,s sg |g ),and (iii)the share of Sonata out of domestic mid-size cars (the third stage share,s j |sg ):s j =s j |sg ∗s sg |g ∗s g .(4)4.2Market Share and Price ElasticityFollowing Berry (1994),we derive the following reduced-form equation for the market share of each product:ln (s j )−ln (s 0)=X j β−αP j +σ1ln (s j |sg )+σ2ln (s sg |g )+ξj +εi j ,(5)for j =1,2,...,J.The variable X j represents characteristics of the vehicle j .We also include manufacturer dummy variables ( B b =1D b j )to account for brand-specific unobservable characteristics such as maintenance costs,which will be discussed in Sect.4.3.Since our market is defined monthly and the study period is 48months,there can be time-specific demand shifters.To control for them,time dummy variables ( 2009y =2006T y j )are also included in X.For the three-stage nested logit model to be consistent with random utility maximization,the following order condition must be satisfied (Goldberg 1995;Verboven 1996):1211Medium-and small-class cars are the most popular in Korea,constituting about 25and 43%of the passenger car market in 2011,respectively.12When σ1=σ2=1,the model reduces to the multinomial logit model.46M.Park,H.Rhee0<σ2<σ1<1.(6) Finally,the price elasticities of market shares are derived as follows:1)Own-price elasticity of product j13j j=∂s j∂p jp js j=αp j11−σ1−11−σ1−11−σ2s j|sg−σ21−σ2s j|g−s j;(7)2)Cross-price elasticity when products j and k are of the same class and productionorigin(domestic versus imported)jk=∂s j∂p kp ks j=αp k−11−σ1−11−σ2s k|sg−σ21−σ2s k|g−s k;(8)3)Cross-price elasticity when products j and k are of the same class but have adifferent production originsjk=∂s j∂p kp ks j=αp k−σ21−σ2s k|g−s k;(9)4)Cross-price elasticity when products j and k are of different classes and productionoriginsjk=∂s j∂p kp ks j=−αp k s k.(10)4.3DataTo implement the model,we require data on sales,market shares,prices,and observ-able characteristics of passenger cars that have been sold in Korea.Our unit of obser-vation is a model-month pair,such as Taurus-December2006.We collected data over four years,from2006to2009.For domestic cars,the number of newly reg-istered cars is obtained from<Automobile Statistics>,which is published by the Korea Automobile Manufacturers Association(KAMA).Each newly purchased car must be registered,and registration records are managed by the Korean government. Therefore,the number of newly registered cars is a reliable proxy for the number of sales.The listed prices and specifications of domestic vehicle models were obtained from the monthly magazine<Car Life>.Data on the number of registrations,prices,and other attributes of imported vehicle models were provided by the Korea Automobile Importers and Distributors Association(KAIDA).Consumer prices are list prices and are deflated by the consumer price index(CPI)for which the base year is2005.We also include indexes for power,size,fuel efficiency,fuel type,and number of seats as non-price attributes of cars.The power index is measured as the reported13In Eqs.(7)through(9),sj|g denotes the share of a product j in its native class.FTA Provisions on the Market Structure47 Table3Repair costs for a test collision(ten thousand KRW)Mercedes C200HondaAccordV olkswagenGolfGrandeurHGHyundaii40Production origin Imported Imported Imported Domestic Domestic Repair cost for the samecollision16771394826330256 Sales price46204120331035102640 Cost/price36.3%33.8%25.0%9.4%9.7% KRW is the Korean currency unit,and1,100KRW is approximately US$1,as of December2011horsepower per10kilograms of vehicle weight.The size index is determined by each car’s length×width in square meters.For the fuel efficiency index,we use the reported mileage(in kilometers)per liter.For fuel type,we include the dummy variable for gasoline-powered cars to adjust for differences in consumer preference and fuel efficiency between gasoline and diesel vehicles.There should be other important value attributes,such as the costs of parts and repair. In fact,the limited number of after-sale maintenance sites and high parts costs have been reported in many consumer surveys as significant factors that prevent Korean consumers from buying imported cars.For example,the Korea Insurance Develop-ment Institute(KIDI2013)reported substantial differences in the costs of repairs for identically simulated collisions(at15km/h)between imported and domestic cars. Table3shows the repair costs forfive sample cars.In terms of the cost/price ratio,the two domestic cars showed substantially lowerfigures(less than10%)than the three imported cars(about30%).Unfortunately,data are not available for every sample observation.Instead,to reduce the possible omitted variable bias,we included car-manufacturer dummy variables as additional explanatory variables in Eq.(5).14After excluding some incomplete or extreme15observations,we are left with5,426monthly model-level observations. Table4shows the summary statistics of variables.The mean value of the dependent variable,0.000042in terms of natural num-bers,is very small since the number of sales of each model is small compared to the potential purchasers who did not buy any car during a given period.For convenience, we used the log of the dependent variable in Table4in the following estimation. In our sample,the average list price of car models that were sold in the market is 56.26million won(about US$51,145)and63percent of all models are gasoline vehi-cles.While most domestic SUVs are diesel models,recently many domestic and foreign brands have begun selling both gasoline and diesel models in the Korean market.14We included25manufacturer dummy variables(e.g.,Audi,Ford,Nissan,and Kia).As a reference manufacturer,the dummy coefficient for Hyundai was normalized to be zero.15We excluded models with more than seven seats and those for which the price was higher than US$200,00048M.Park,H.RheeTable4Statistical summaries of variablesVariable Observation Mean S.D.Min MaxDependent variable Sales of a product/nopurchase5,4260.0000420.000106 1.81e−070.0013IndependentvariablesPrice(million KRW)5,42656.2633.487.61199.02Power(horsepowerper10kilograms)5,426 1.200.320.63 3.00 Size(square meters)5,4268.600.92 5.2310.99Fuel efficiency(kilometers per liter)5,42610.28 2.47 5.5019.50 Gasoline=15,4260.630.480.00 1.00No.of seats5,426 5.000.85 2.007.00Table5shows the aggregate annual sales(based on registrations)and the market shares of domestic and imported cars infive class segments.16Again,the shares of imported cars in Korea remain small across class segments except for the sports seg-ment,in which Korean car manufacturers have no more than two brands that compete against the premium sports vehicle brands such as Porsche,Ferrari,and Lamborgh-ini.17However,in terms of the absolute quantities sold,the largest segment is the large class segment,with23,853units,which totals about half of all imported vehicles that were registered in2009.5Results5.1EstimationWe estimated the demand function in Eq.(5).To control for the possible endogeneity of price and within-group shares,we applied a two-stage regression analysis with instrument variables(IV).As in Berry et al.(1995),the functions of characteristics of other products were used as instruments because hardware attributes are assumed to be exogenous and the existence of similar products negatively affects price.18We also used the consumption tax rate as an additional instrument because it is exogenously determined by the government,but it varies across classes and affects price and within-group shares.16Compact and small cars are merged into a single category because no compact imported cars were sold in Korea through2009.17The annual shares of imported cars are slightly different from those reported in Fig.1because some incomplete observations were excluded from our data.The reduction in import shares in2009,except for the large car segment,is probably due to the globalfinancial crisis during that period.18Specifically,we used the same instruments as were used in(Berry et al.,1995,p.861):a)the sums of the values of the same specifications(power,size,fuel efficiency,etc.)of other car models that were offered by a manufacturer,and b)the sums of the values of the same specifications of car models that were offered by other manufacturers.As observed in Berry et al.(1995),we found that including higher-order functions of characteristics did not significantly alter the results.Table5Sales and market shares of domestic and imported cars in KoreaClass segments Year Number of models Number of newly registered vehicles Market shareof imported Domestic Imported Domestic Imported Totalcars(%) Sports20061151,7791,1902,96940.08 20072162,6561,7434,39939.6220082142,3621,7394,10142.4120092157,1431,4288,57116.66SUV20061322205,8805,191211,071 2.46 20071026207,5298,955216,484 4.1420081232179,15910,041189,200 5.3120091232194,7899,564204,353 4.68 Compact/small200600101,2950101,2950.00 200771117,125163117,2880.14200861174,979260175,2390.15200961161,538134161,6720.08 Medium2006617220,3426,313226,655 2.79 2007718254,6827,193261,875 2.752008921244,7889,756254,544 3.8320091022369,23111,944381,175 3.13 Large2006939317,47417,203334,677 5.14 2007839323,26021,915345,175 6.3520081138351,80425,475377,279 6.7520091035282,80723,853306,6607.78Table6shows the estimated parameters.Overall,our modelfits well(pseudo R2= 0.91),and the estimated signs of parameters are all intuitive,being negative for price and positive for all other value attributes.19For brevity,Table6excludes estimates for manufacturer dummy -pared to the reference domestic brand(Hyundai),the dummy coefficients for most imported car manufacturers are estimated to be negative(−0.81on average),repre-senting perceptual handicaps.By production origin,the average dummy estimates for EU-,Japan-,and US-brand cars are−0.67,−0.76,and−1.26,respectively.The num-bers suggest that imported cars,particularly US-made cars,have substantial perceptual disadvantages in value compared to otherwise equivalent domestic cars,probably due to the expensive after-sale maintenance costs that were mentioned previously.To assess the validity of our hierarchical specification,wefirst tested the order condition in Eq.(6).Becauseσ1=0.692.σ2=0.490from Table6,the order condition is well satisfied.In addition,we tested whether the IIA property holds for the last segments of the decision tree by excluding one alternative from the choice set19The coefficients in Table6are the derivatives of the log of the dependent variable with respect to each explanatory variable(i.e.,d ln y/dx).Table6Summary of parameter estimationNewey–West heteroscedasticity and autocorrelation consistent (HAC)standard errors are in parentheses.Coefficient estimates on brand dummy variables are not reported.***p<0.01,**p<0.05,*p<0.1IV nested logit estimate(s.e.)Price(million KRW)−0.015(0.007)**Power0.545(0.406)Size0.471(0.102)***Fuel efficiency0.083(0.020)***Gasoline=10.333(0.073)***No.of seats0.413(0.078)*** Constant−16.90(0.897)***log(s sg|g)0.490(0.105)***log(s j|sg)0.692(0.148)***Year2006−0.081(0.084)Year20070.038(0.069)Year2008−0.024(0.064)#of observations5,426R-squared0.911and comparing the coefficients(Hausman and McFadden1984;Hausman and Leonard 2005).In most cases,removing one vehicle model did not greatly alter the estimation results,and therefore the within-nest IIA assumption was supported.Finally,we performed a similar analysis using another order of hierarchy in which the production origin is selected before the class.However,the order of the heterogene-ity parameters in this case did not comply with the order condition in Eq.(6),which suggests a violation of the utility maximization principle.Overall,the specification tests suggest that our three-stage nesting structure(Fig.2)is an acceptable model of actual purchase behavior in the Korean automobile market.We estimated regression models for each year separately,and calculated the F-test statistic as follows to justify our pooling regression analysis:F3k,n−4k=(RSS of pooled regression−sum of RSS of separate regressions)/3k sum of RSS of separate regressions/(n−4k),where k(=38)is the number of regressors and n(=5426)is the total number of obser-vations.The RSS(regression sum of squares)of the pooled regression is2,506,and the sum of the RSS values of the separate regressions by year is2,456.The F-statistic is therefore1.03,which is less than the critical value at the5%(and1%)significance level.Therefore,these results suggest that the four-year data can be aggregated.5.2Price ElasticityUsing the estimates reported in Table6,we can calculate the own-and cross-price elasticities of market shares for every vehicle model that is included in the sample。