Once an Enemy, Forever an Enemy? The Long-run Impact of Japanese Invasion of China on Trade and Investment

Transcription

1 Once an Enemy, Forever an Enemy? The Long-run Impact of Japanese Invasion of China on Trade and Investment Yi Che a, Julan Du b, Yi Lu c, and Zhigang Tao a a University of Hong Kong, b Chinese University of Hong Kong, c National University of Singapore December 2011 Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

2 Introduction A at world with endless con icts Whether con icts among countries have long-run, adverse impact on trade and investment This paper explores an important historical event, the second Sino-Japanese war ( ), to investigate whether the war has lasting-e ect on the bilateral trade between China s regions and Japan and also direct investment in China from Japan Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

3 Introduction Sino-Japanese War In 1931, shortly after Mukden Incident, Japan s Kwantung Army invaded Manchuria of China. The Marco Polo Bridge Incident of 1937 marked the beginning of a total war between Japan and China. From 1931 to 1945, China su ered more than 35 million military and non-military casualties and a total of 383 billion US dollars property losses Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

4 Introduction Antagonism against the Japanese Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

5 Introduction Antagonism against the Japanese Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

6 Introduction Antagonism against the Japanese Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

7 Introduction Antagonism against the Japanese Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

8 Roadmap Introduction Literature Data and Descriptive Estimation Speci cations Trade Investment Empirical Findings Conclusion Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

9 Related Literature Con icts and trade Blomberg and Hess (2006), Martin, Mayer, and Thoenig (2008), Glick and Taylor (2010) History and Trade Head, Mayer and Ries (2010) Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

10 Data China Customs Data in 2001 A survey of foreign-invested enterprises conducted by the National Bureau of Statistics of China in 2001 China s Damages During World War II and Statistical Abstract of the Republic of China Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

11 Data Information about Sino-Japanese trade and Japanese direct investment in China Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

12 Data Civilian Casualties The percentage of civilians who su ered minor wounds, severe wounds, or deaths for China regions Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

13 Trade Estimation Model, Setup Demand ) Production U l = J l Z 1 Nlj µ lj ln y lj (i) α αj j di (1) j=1 0 y l (i) = µ l E l p l (i) ε (2) Pl 1 ε Firm i in place l can produce one unit of output at a constant marginal cost c l a i a i is drawn from a cumulative distribution function G (.) with support h a L l, ah l i, where a H l > a L l > 0 Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

14 Trade Estimation Model, Setup Selling Domestic market, no additional cost Foreign market Fixed cost of exporting: c l f ll 0 Per unit transport cost: τ ll 0 > 1 Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

15 Trade Estimation Model, Equilibrium The optimal price rm i from place l charges in place l 0 is p ll 0(i) = c la i α τ ll 0 (3) The pro t for rm i from place l to serve in place l 0 is τll 0c π ll 0(i) = (1 α) l a 1 ε i µ αp l 0 l 0E l 0 c l f ll 0 (4) ) the cuto level of productivity 1 µl 0E a i l 0(1 α) ε 1 αp l 0 a ll 0 (5) c l f ll 0 τ ll 0c l Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

16 Trade Estimation Model, Equilibrium An individual rm s export volume τll 0c R ll 0(i) = l a 1 i αp l 0 ε µ l 0E l 0. (6) ) the total export volume from l to l 0 is M ll 0 = Z Nl 0 R ll 0(i)di = N l Z a H l a L l R ll 0(a)dG (a) = αpl 0 τ ll 0c l ε 1 µ l 0E l 0N l V ll 0, (7) where ( R all 0 a 1 ε dg V ll 0 al L l (a) 8a ll 0 al L 0 otherwise. (8) Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

17 Trade Estimation Speci cation The export volume from l to l 0 can be estimated in a log-linear form, i.e., m ll 0 = (ε 1) ln α + [(ε 1) p l 0 + ln µ l 0 + e l 0] + [n l (ε 1) ln c l ] (ε 1) ln τ ll 0 + v ll 0, (9) Assume τ ε 1 ll 0 D γ ll 0 exp βz ll 0 + X 0 ll 0 δ u ll 0, then we have the following estimation equation m ll 0 = θ + λ l + ξ l 0 γd ll 0 βz ll 0 X 0 ll 0δ + v ll 0+u ll 0, (10) A consistent estimation requires a control for v ll 0 ( rm heterogeneity) and the possible correlation between u l 0 l and independent variables ( rm self-selection) Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

18 Trade Estimation Speci cation De ne a variable Y ll 0 Y ll 0 as αpl ε 1 0 (1 α) τ ll 0 c µl l 0E l 0 al L 1 c l f ll 0 ε, (11) We only observe positive exports when Y l 0 l > 1 Assume f ll 0 exp ˆβz ll 0 + W 0 ll 0 ϕ υ ll 0 and then latent variable y l 0 l ln Y l 0 l can be written as y ll 0 = { + ς l + φ l 0 γd ll 0 βz ll 0 X 0 ll 0δ W0 ll 0ϕ+η ll 0, (12) Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

19 Trade Estimation Speci cation De ne T ll 0 not to equal 1 when place l exports to l 0 and 0 when it does ρ ll 0 = Pr (T l 0 l = 1jobserved variables) = F { + ς l + φ l 0 γd ll 0 βz ll 0 X 0 ll 0δ W0 ll 0ϕ, (13) Let ˆρ ll 0 be the predicted value and ŷ ll 0 = F 1 (ˆρ ll 0) As v ll 0 is a monotonic function of y ll 0, a consistent estimate of v ll 0 can be derived as ˆv ll 0 = H(ˆρ ll 0). (14) Meanwhile, E [u ll 0j., T ll 0 = 1] = corr (u ll 0, η ll 0) σ u /σ η ηll 0, where a consistent estimate of η ll 0 is obtained from the inverse Mills ratio, i.e., ˆη ll 0 = f (ŷ ll 0)/F (ŷ ll 0) = I (ˆρ ll 0). (15) Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

20 Trade Estimation Speci cation With ˆv ll 0 and ˆη ll 0, we can consistently estimate equation (10) m ll 0 = θ + λ l + ξ l 0 γd ll 0 βz ll 0 X 0 ll 0δ + B(ˆρ ll 0 )+e ll 0, (16) where B(ˆρ ll 0) H(ˆρ ll 0) + corr (u ll 0, η ll 0) σ u /σ η I (ˆρll 0); and e ll 0 is an i.i.d. error term satisfying E [e ll 0j., T ll 0 = 1] = 0 Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

21 Trade Estimation Issues, Reference Country Two-stage estimation, with the rst-stage investigating the extensive-margin e ect while the second-stage the intensive-margin e ect A place-pair fl, l 0 g, one has to be a China s province-level region (r), and the other being Japan and a reference country (f ) The inclusion of the reference country (in which z ll 0 = 0) can allow us to control for all China s province-level region e ects The United States (U.S.) is used in the benchmark analysis In robustness checks, we experiment with the inclusion of German, Singapore, and the United Kingdom (U.K.) Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

22 Trade Estimation Issues, Product-level Analysis Argumented with product dummy ρ ll 0 j = F { + ς l + φ l 0 + ϱ j γd ll 0 βz ll 0 X 0 ll 0δ W0 ll 0ϕ (17) m ll 0 j = θ + λ l + ξ l 0 + ϖ j γd ll 0 βz ll 0 X 0 ll 0δ + B(ˆρ ll 0 )+e ll 0 j. (18) In the benchmark analysis, we de ne a product at the HS-4 digit level (with 1, 253 product dummies) To investigate whether there is any aggregation bias, in a robustness check, we conduct a linear estimation with product de ned at the HS-6 digit level (with 5, 453 product dummies) Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

23 Trade Estimation Issues, First-stage Estimation In the benchmark analysis, we rst assume that η ll 0 follows a normal distribution and hence use the Probit estimation as in Helpman, Melitz, and Rubinstein (2008) We then experiment with three alternative estimation approaches Logit estimation: a non-linear estimation assuming a logistic distribution of η ll 0 OLS estimation: a linear estimation with no prior distribution assumption of η ll 0 A semi-nonparametric estimation developed by Gallant and Nychka (1987): no prior distribution assumption of η ll 0 Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

24 Trade Estimation Issues, Second-stage Estimation To approximate as exibly as possible an arbitrary functional form of B(.), we use a semi-nonparametric estimation First, we partition the predicated value ˆρ ll 0 into a number of bins with equal observations and assign a bin indicator to every bin Then, we replace B(ˆρ ll 0) with a number of bin indicator dummies The compromise of using this exible estimation is that we are not able to distinguish the rm-heterogeneity and rm self-selection e ects As it is a linear equation, we use the OLS estimation with 50 or 100 bins to ensure a large degree of exibility as in Helpman, Melitz, and Rubinstein (2008) Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

25 Trade Estimation Issues, Estimation of Standard Errors The robust standard error controlling for clustering at the place-pair fl, l 0 g in the rst-stage The bootstrapped standard errors in the second-stage to correct for the tted regressors in the estimation Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

26 Trade Estimation Issues, Excluded Variable To identify the e ects in the second-stage, we need some variables only belong to the rst-stage Some variables a ect only the xed but not variable trade costs Helpman, Melitz, and Rubinstein (2008) use regulation costs of rm entry and religion We instead propose to use an indicator for an embassy of the foreign country (Japan and the reference country) in each China s region Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

27 Investment Estimation Model Let the pro t function of Japanese multinational f from investing in region r of China in year t be approximated by π frt = α + βz r + X 0 rt 1δ + ɛ frt, (19) n o Observing the information z r, X 0 rt 1 for all regions, Japanese multinational f at time t picks up a region to invest when its pro t obtained from that region is largest among all China s regions De ne T frt = 1 if Japanese multinational f invests in region r at time t, p frt = PrfT frt = 1jobserved variablesg = Prfπ frt π fkt 8k 2 Rg = Prf(ɛ frt ɛ fkt ) β(z k z r ) + (X 0 kt 1 X 0 rt 1)δ 8k 2 Rg, Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

28 Investment Estimation Model Assuming ɛ frt follows a Type I Extreme distribution, McFadden (1974) derives the following explicit solution for p frt, p frt = exp(βz r + X 0 rt 1 δ) (20) k2r exp(βz k + X 0 kt 1δ). Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

29 Investment Estimation Issue, IIA Condition of the Independence of Irrelevant Alternatives (IIA) the choice between two alternative regions in China not being a ected by the inclusion of other alternative regions in the choice set As robustness checks, we conduct three tests on the satisfaction of the IIA assumption Investigate whether our results are a ected by the exclusion of some in uential regions (Head, Ries, and Swenson, 1995) Use the Poisson regression (Guimarães, Figueiredo, and Woodward, 2004) Relax the assumption of IIA and assume a nested discrete choice structure Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

30 Investment Estimation Issue, Counterfactual Analysis Without region dummy, it is possible Counterfactual analysis E [ɛ frt z r jobserved variables] 6= 0 (21) Use the sample of multinationals from China s major FDI countries like the U.S., Germany, Singapore and the U.K. Conceptually, if there are some regional unobservables (not captured by the long-list of X kt 1 ) that are correlated with z r and persist to adversely a ect the contemporary FDI location choice, we may pick up such adverse e ects in at least one of these China s major FDI countries Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

31 Investment Estimation Issue, Sample of 2001 As the Survey was conducted in 2001, we observe the location of each foreign-invested enterprise in 2001 and its establishment year To back out the location when these foreign-invested enterprises invested in China, we implicitly assume there is no location change after the establishment However, if there are substantial changes of location over years and/or entry and exit, our estimation could be biased As a robustness check, we focus on a sub-sample of Japanese multinationals invested in 2001, in which we have a precise location at the time of investment Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

32 Investment Estimation Issue, JV versus Wholly-owned Two common types of foreign investment: joint-venture and fully-owned subsidiaries Conditional on forming a joint-venture with local partners, Japanese multinationals should be less a ected by the disaster caused by the Japanese invasion An experiment Two sub-samples, a sub-sample of rms fully-owned by Japanese multinationals and a sub-sample of joint-ventures Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

33 Trade Benchmark Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

34 Trade Robustness I, Alternative Estimation of the First-stage Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

35 Trade Robustness II, Alternative Reference Countries Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

36 Trade Robustness III, HS 6-digit analysis Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

37 Investment Benchmark Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

38 Investment Robustness I, IIA Checks Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

39 Investment Robustness II, Counterfactual Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

40 Investment Robustness III, Sub-sample of Year 2001 Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

41 Investment Robustness IV, JV versus Wholly-owned Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42

42 Conclusion Our research contributes to the emerging literature on the role of historical events in explaining contemporary developments, and the literature on international trade and investment by examining the long run impacts of the Japanese invasion of China on trade and direct investment from Japan. Our study shows that historical animosity still matters for international trade and investment despite the trend toward a at world. Che, Du, Lu, Tao (HKU, CUHK, NUS) Once an Enemy, Forever an Enemy? December / 42