Competition and Welfare Gains from Trade: A Quantitative Analysis of China Between 1995 and 2004

Transcription

1 Competition and Welfare Gains from Trade: A Quantitative Analysis of China Between 1995 and 2004 Wen-Tai Hsu Yi Lu Guiying Laura Wu SMU NUS NTU At NUS January 15, 2016 Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

2 Competition and Gains from Trade Trade liberalization may bring at least four different gains from trade: Better productivity (lower costs and hence lower price) Ricardian effect Lower markups (domestic firms may lower their prices to deter foreign entry and successful foreign entry may also bring lower markup) Higher profits there are more chances for firms to earn profits from foreign countries Less dispersion of markup distribution allocative effi ciency Competition and markups matter in the latter three above-mentioned forces pro-competitive effects. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

3 Allocative Effi ciency When some goods are monopolized and others are not, the resource allocation across goods is distorted. When mark-ups are the same across all goods, first-best allocative effi ciency is attained. The condition that the price ratio equals the marginal cost ratio, for any pair of goods, holds because of constant mark-ups. In other words, net social benefits are equalized across goods. With markup dispersion, firms with low markups may produce/employ more than optimal whereas those with high markups may produce/employ less than optimal. Trade liberalization tends to depress markups both in mean and its dispersion (Holmes, Hsu, and Lee 2014). Despite the economics of pro-competitive effects of trade being well-understood, its quantitative magnitudes are still under-studied. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

4 This Paper This paper utilizes China s Economic Census Data at 1995 and 2004 to quantitatively examine welfare gains from trade. Decomposition: a Ricardian component and two pro-competitive components Separate the effect of tariff reduction from non-tariff trade costs. Using Economic Census Data at 1995 and 2004 avoids the trunction problem in the commonly used annual survey of manufacturing firms ( 5 million RMB). China s entry into WTO is mainly about tariff reduction, although it may also involve some deregulations. The exercise on the impact of tariff reduction provides a lower bound of the the effect of the entry to WTO. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

5 China s Gains from Trade Between 1995 and 2004 China entered WTO in As a condition to enter WTO (or its earlier form, GATT), it was required to lower tariffs. China s tariffs were reduced substantially between 1992 and There was another round of drop in tariff after 2001 to carry out its promise to WTO members. During , (weighted) average import tariff drops from 25.5% to 6.3%, whereas average export tariff drops from 6.4% to 3.2%. Reforms on State Owned Enterprises (SOE) reduces some entry barriers and make markets more competitive. Improvement in infrastructure both within-the-country and across the borders. Import share increases from 0.13 to 0.22, and export intensity also grows at a similar rate. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

6 Quantitative Framework Our quantitative framework is a variant of the canonical trade model in Bernard, Eaton, Jensen, and Kortum (2003; henceforth BEJK). Productivity (Frechét distribution) differ across firms and countries Firms compete in Bertrand fashion Markups are generated by productivity differences lowest cost firm charges the price at the second lowest marginal cost The distribution of markup is invariant to changes in trade costs...no pro-competitive effect of trade. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

7 Change in the Distribution of Markups (of Firms) Figure: Markup Distributions (1995 v.s. 2004) Mean markup decreases from 1.43 to 1.37, and the standard deviation decreases from 0.50 to Such a pattern holds true when we break the sample into exporters and non-exporters. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

8 Quantitative Framework Holmes, Hsu, and Lee (2014) develop a model that deviate from BEJK in allowing a general distribution of productivity and assuming there are finite number of firms per product to draw from the distribution. Their result is that pro-competitive effects of trade will emerge when the tail of the productivity distribution is not too fat. We adopt the model framework in Holmes, Hsu, and Lee (2014) and assume Productivity draws is from log-normal Number of firms per product is drawn from Poisson Log-normal distribution has a kind-of fat-tail, but is less fat than Pareto/Frechét, and it matches the entire distribution better than Pareto, except at the very upper tail (Head, Mayer, and Thoenig 2014). Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

9 Preview of Main Results Counter-factual analysis reveals that pro-competitive effects account for 25.4% of the total welfare gains during Comparison with symmetric-country estimation indicates that the asymmetry between countries may be an important source of gains from trade, whereas the relative contribution of pro-competitive effects remains similar. Tariff reduction accounts for 32% of the reduction in trade cost, and in terms of welfare, the relative contribution of tariff reduction is 39.6%. (lower bound of the effect of WTO entry). We also ask: Did China trade-liberalize the right sectors? Answer: Yes, because there is a tendency that when a sector has a higher markup, there is a larger degree of trade liberalization. Welfare improves because the markup dispersion across sectors is reduced. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

10 Literature Review: Constant Markup and ACR CES + monopolistic competition/perfect competition: constant markup and hence no pro-competitive effects (Krugman 1980 and countless models following this) Arkolakis, Costinot, and Rodriguez-Clare (2012; henceforth ACR): Despite more than 10 years of development in trade theories in firm heterogeneity, the combination of CES and monopolistic competition still entails no pro-competitive effects. In a class of influential trade model, welfare can be simply determined by domestic consumption share and trade elasticity: Ŵ = ˆλ 1/ɛ. BEJK is in this class, as well as Melitz with Pareto. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

11 Variable Markup and Oligopoly Approach Oligopoly Approach by Edmond, Midrigan, and Xu (2015): quantitatively evaluate pro-competitive effects using the model of Atkeson and Burstein (2008) - heterogeneous-product Cournot competition. Got estimate from 11% to 38%. Markups are directly linked to market shares of firms. Taiwanese data work well for their oligopoly environment because they can go down to very fine product level to look at with a few firms. In large countries, due to numerous firms in a given industry, market share per firm is typically small, and the market structure looks less concentrated and pro-competitive effects dissipate. This is not to say pro-competitive effects do not exist for large countries. Rather, it may be that there are several markets within a large country, but we simply don t know how to separate them. Our approach allows an oligopoly structure to work even in the presense of hundreds or thousands of firms in an industry, as market structure facing a firm is inferred rather than linked with industrial/product classifications. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

12 Variable Markups in Monopolistic Competition Arkolakis, Costinot, Donaldson and Rodriguez-Clare (2015) deviate from CES by allowing a flexible specification of non-homothetic preference with monopolistic competition and a fat-tailed distribution (Pareto). Result: Welfare is lower with markup dispersion, but there is little pro-competitive effect of trade. Feentra (2014) proposes a monopolistic competitive model with a non-ces preference and a bounded productivity distribution that allows pro-competitive effects. In monopolistic competition, a change in the trade cost only affects a domestic firm through general equilibrium effects that might shift or rotate the firm s demand curve. In contrast, in a Bertrand environment, the pro-competitive force of trade operates at the level of the particular good, not through general equilibrium. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

13 Model: Consumption A continuum of goods, and each good indexed by ω [0, γ]: γ is exogenous Only ω γ measure of goods will be actually produced, ω is endogenous. Assume CES utility. U = ( ω 0 ) σ q σ 1 σ 1 σ ω dω, for σ > 1. There are two countries, i = 1 (China), 2 (ROW). Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

14 Model: Consumption The standard CES price index in country j: P j ( ω 0 ) 1 p 1 σ 1 σ jω dω. Country j s total consumption expenditure of good ω is given by ( ) 1 σ pjω E jω = R j, P j ( ) 1 σ pjω where P j is expenditure share on good ω, and Rj is the total revenue, as well as total income, of country j. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

15 Number of Firms and Measure of Goods The number of firms for each good ω [0, γ] in country i is a random realization from Poisson distribution with parameter λ i : f i (n) = e λi λ n i n! The actual number of goods produced ω < γ is the complement set of the event (n 1, n 2 ) = (0, 0). Formally, [ ω = γ (1 f 1 (0) f 2 (0)) = γ 1 e (λ1+λ2)]. (1) In the event (n 1, n 2 ) = (0, 1) or (1, 0): monopoly. If n 1 + n 2 2, then firms engages in Bertrand competition.. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

16 Productivity, Marginal Cost, and Pricing Production features constant returns to scale. Labor is the only input. Each potential competitor draws a productivity ϕ ω,ik from a log-normal distribution so that for k {1, 2,..., n ω,i }, ln ( ϕ ω,ik ) N (µi, η i ). In n ω,i draws, let ϕ ω,i and ϕ ω,i be the first and second highest. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

17 Productivity, Marginal Cost, and Pricing Shipping domestically is free, but shipping to foreign countries requires an iceberg cost τ > 1. τ ij = 1 if i = j, and τ ij = τ > 1 if i j. For each ω, the relevant marginal costs to deliver to country j are { } τ 1j w 1 ϕ, τ 1j w 1 sω,1 ϕ, τ 2j w 2 sω,1 ϕ, τ 2j w 2 sω,2 ϕ. ω,2 Let a jω and a jω be the lowest and second lowest elements of this set. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

18 Productivity, Marginal Cost, and Pricing Monopoly pricing: p jω = Pricing: σ σ 1 a jω. p jω = min ( p { } jω,ajω ) σ = min σ 1 a jω,ajω. The markup of good ω at j is therefore { } m jω = p jω σ a = min σ 1,a jω a, jω jω Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

19 General Equilibrium There are three conditions that pins down total revenues R 1, R 2, and equilibrium wage ratio w w 2 /w 1 = w 2 : The labor market clearing condition in both countries. Balanced trade condition, which is implied by the market clearing of commodities. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

20 Aggregate Markups From the firms viewpoint, if it is a nonexporter, then its markup is m f jω = m ω,j. For an exporter, constant returns to scale implies that its markup is a revenue-weighted average of markups in both countries. Welfare depends on the markups from both consumers and producers points of view, but the former is not directly observable. What we observe is firms markups, which are useful for inferring structural parameters. Consumers aggregate markup is the revenue-weighted harmonic mean across goods with destination at i. ( ω 1 M buy i = m 1 ω,i φ ω,i dω). 0 Producers aggregate markup M sell i is defined in a similar way. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

21 Welfare Decomposition Let A i be the price index under marginal cost pricing, A i = ω 0 a iω q m iωdω, where ~q m i = { q ω,i : ω [0, 1]} is the expenditure-minimizing consumption bundle that delivers one unit of utility. We can then write W Total i = R i P i = w i L i M sell i 1 P i (2) = w i L i 1 Msell i A i M buy i w i L i W Prod i Ai M buy i P i Msell i M buy Wi A, (3) i where W Prod i is productive effi ciency index, and W A i is allocative effi ciency index. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

22 Welfare Decomposition W Total i = w i L i 1 Msell i A i M buy Wi A. i We focus on country 1, and we will set w 1 = 1. As the labor supply L i is fixed, the first term can be ignored. W Prod 1 A i is what the welfare index would be with constant mark-up. The index varies with technological changes (µ i, η) or when trade cost (τ) changes declines. Terms of trade effects also show up in W Prod, as wages also enter the marginal costs. It can be shown that this term traces the ACR statistics very closely. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

23 Welfare Decomposition W Total i = w i L i 1 Msell i A i M buy Wi A. i When markups are a constant, the third and fourth terms drop out. The third term is a terms of trade effect on mark-ups. The higher the producers aggregate markups, or the lower the consumers aggregate markup, the higher the welfare. This term drops out under autarky. It also drops out under symmetric countries. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

24 Welfare Decomposition W A i A i M buy ω i 0 = a iω qa iω dω P ω i 0 a iω q iωdω 1. Under marginal cost pricing, q ω,i a is the optimal bundle, and hence ω 0 a iω qa iω dω ω 0 a iω q iωdω. Under constant markups, for any pair of goods, the ratio of actual prices equals the ratio of marginal cost, and hence, W A i = 1. Those with higher markups produce/employ less than optimal, and those with low markups produce/employ more than optimal. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

25 Data Economic census from China s NBS (National Bureau of Statistics) data (1995 and 2004). From World Bank s WDI (World Development Indicators), obtain world manufacturing GDP and GDP per capita. The aggregate Chinese trade data is from UN comtrade. Combine data of GDP per capita and labor income share to calculate w = w 2 /w 1. Tariff data from WITS. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

26 Estimation of Markups We estimate markups using De Loecker and Warzynski s (2012) approach, which calculate markups as m DLW ω = θx ω α X, ω where θ X ω is the input elasticity of output of input X, and α X ω is the share of expenditure on input X in total sales. An alternative is to use CRS assumption and simply calculate markups by m raw ω = revenue ω total costs ω, which we call raw markups. We use this measure as a robustness check. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

27 Estimated Markup Distribution Table 1. The (unweighted) mean markups all decrease between years 1995 and 2004 for all firms, exporters and non-exporters. The (unweighted) standard deviation of markups decreases for non-exporters, but it increases for exporters. As there are more non-exporters than exporters, the overall standard deviation decreases. For the percentiles, almost all of them decreases between 1995 and This is consistent with the pattern visually seen in Figure 1-3 that the entire distribution becomes more condensed. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

28 Inference of the Elasticity of Substitution The model implies that m [ ] σ 1,. σ 1 When the second marginal cost is high, the markup is bounded by the monopoly one because the firm profits is still subject to the substitutibility between products. The higher the substitutibility (σ), the lower the bound. Considering the possibility of measurement error and outliers, we equate to the 99-percentile of estimated markup distribution. Result: σ = 1.4. This is quite different estimates from the literature which typically estimate/calibrate σ under monopolistic competition models, which often feature constant markup. Suppose we take σ = 4, then it implies that m [1, 1.33], which would cut off 35% of the markups (at the right tail). σ σ 1 Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

29 Simulated Method of Moments We use SMM to estimate the remaining parameters τ : trade cost γ : measure of goods λ i : mean number of firms per product µ i : mean parameter of log-normal productivity draw η i : standard deviation parameter of log-normal productivity draw For productivity, we normalize µ 2 = 0 (when ln ϕ is zero, ϕ = 1). This is because only the relative magnitude of µ 1 to µ 2 matters. Choosing µ 2 amounts to choosing unit. Given data moments of R 1, R 2, w and the inferred σ from markup distribution, we simulate 12 moments for each set of parameter (τ, γ, λ 1, λ 2, µ 1, η 1, η 2 ). Table 2. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

30 Counter-Factual Analysis We conduct counter-factuals under 2004 parameter values, and take τ to the 1995 value and to autarky (Table 3). The welfare gains from 1995 s openness to 2004 s level is 9.4%, in which the pro-competitive effect account for 25.4%. Allocative effi ciency W A alone accounts for 22.3% of these gains. W TOT = M sell M buy accounts for 3.1% decreases in both M sell and M buy, but that of M buy is larger. More trade openness benefits Chinese consumers by lowering the markups, but it also hurts Chinese firms profits. Gains from trade from autarky is 33.4%, but the relative contributions of pro-competitive effects are similar (23.3% overall, 22.4% for W A alone) Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

31 Comparing with the Literature In Edmond et al., diminishing returns in allocative effi ciency % in Edmond et al. Import share Total Welfare Ricardian W A importance of W A 0 to 10% % 10% to 20% % We do a similar exercise, and our corresponding table is % in this paper Import share Total Welfare Ricardian W A importance of W A 0 to 10% % 10% to 20% % Not only allocative effi ciency exhibits a similar diminishing returns pattern, but also the Ricardian component and hence total welfare. However, the relative importance of allocative effi ciency remains stable around a quarter in our model. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

32 Symmetric Countries For the purpose of comparison, we also estimate a symmetric country case. Table 4 & 5. W TOT = 1 always in this case, and W A accounts for % of the total effects. The gains from trade, as well as its components, are smaller in symmetric-country case. As productivity draws between the two countries become the same, the Ricardian gains are reduced. When two countries are the same in their productivity draws and entries, not only the distribution of markups becomes more similar, but the dispersion of markups becomes smaller. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

33 Robustness The relative contribution of pro-competitive effects remains similar to the following robustness checks Based on 1995 estimates and taking τ to 2004 level or autarky. Using a raw measure of markups: m = revenue total costs. Using 97.5%-tile to infer σ. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

34 The Effect of Tariff Following the literature, we decompose trade cost by or equivalently, τ = (1 + t) τ nontariff, ln τ = ln (1 + t) + ln τ nontariff Using tariff data from WITS and calculate weighted average using weights by trade volumes, Average import tariff drops from 25.5% to 6.3% Average export tariff drops from 6.4% to 3.2%. Relative importance of tariff reduction in overall reduction of trade cost ln (1 + t) ln τ = 31.6% The relative contribution of tariff reduction in total welfare gains is 39.6%. This provides a lower bound of the WTO effect. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

35 Multi-Sector Economy For robustness check, we also extend the model to a multiple-sector one to take into account the cross-sector heterogeneity in trade costs, as well as in productivity distribution, entry pressures, and preference parameters. The inferred/estimated parameters show a substantial variation across sectors. But, the general pattern of the changes in parameters between 1995 and 2004 still hold true. (Table 8A and 8B) The welfare analysis results are similar to the one-sector economy case, although the magnitude is somewhat smaller (Table 9). The total gains reduces from 9.4% to 7.2%, and the relative contribution of the pro-competitive effect reduces from 25.4% to 20.0%. The relative contribution of tariff reduction is 34.8%. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

36 Did China Trade-Liberalize the Right Sectors? If a sector has a higher M buy 1s at 1995, do we also actually see a larger degree of trade liberalization between 1995 and 2004? Most pro-competitive gains from trade are due to allocative effi ciency. If a sector s has higher M buy 1s initially, then allocative effi ciency would improve more if the government targets its trade liberalization more in these higher-markup sectors because this reduces the dispersion of markups across sectors. Rank the 29 sectors by their values of M buy 1s groups (15 and 14). at 1995 and divide them into two Average M buy 1s are 1.21 and Average changes in trade costs τ s (i.e, τ s = τ s,2004 τ s,1995) are and If we measure trade liberalization by sectoral import tariffs, the corresponding changes are and 0.215, respectively. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

37 Did China Trade-Liberalize the Right Sectors? Table 10. We are not asking a causal question. We don t know how this happened exactly. Perhaps a benevolent government or it could be mechanical because Before entering WTO, there was a large degree of variation in the import tariff in China. The conditions of WTO entry generally requires larger tariff reductions in those industries with higher initial tariffs (see, e.g., Lu and Yu 2015). Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38

38 Conclusion The quantitative framework is useful for examining pro-competitive effects and applicable to countries of any size. The key is that the assumed market structure allows the number of competitors of a firm to be inferred rather than to be tied with the number of firms of an industry or product category. Pro-competitive effects account for 25.4% of the total gains, and allocative effi ciency alone account for 22.3%. Tariff reduction account for 31.6% of the reduction in trade cost, and it contributes to welfare gains about 40%. China on average trade-liberalized the right sectors. Hsu (SMU), Lu (NUS), and Wu (NTU) Welfare (SMU Gains from Trade NUS - China NTU) At NUS January 15, / 38