Estimating the welfare effects of the Transatlantic Trade and
Investment Partnership
Thanh Nam Le
ID: S2258684Master Thesis - Supervisor: Petros Milionis Research Master in Economics and Business
University of Groningen
Abstract
This paper assesses the impact of the Transatlantic Trade and Investment Partnership on each member of European Union, the United States as well as other major economies of the world. I focus on one particular aspect of a trade agreement, which is the elimination of bilateral tariffs. By using the framework of Caliendo and Parro (2015), I estimate welfare effects as the relative changes of real income in the equilibrium against a baseline model calibrated with actual data of 2014. The results show welfare gains for majority of EU economies and the US. The decomposition of welfare changes show that increases in volume of trade improve welfare across all but one EU member. In some economies, the positive contribution of volume of trade is counterweighted by deterioration in ’terms of trade’, which results in welfare losses. There is evidence that countries with higher diversification in their baskets of exports and imports tend to have larger welfare gains. In addition, countries that are overly open to trade tend to have smaller welfare gains (i.e. more open economies are more prone to declines in ’terms of trade’). Given the already low tariffs between EU and the US, the welfare gains due to tariff reduction are significant but not very large.
1
Introduction
The Transatlantic Trade and Investment Partnership (TTIP) is a bilateral trade agreement being negotiated by European Union and the United States, which together account for nearly 50% of global GDP and around one third of world trade. Its main objectives are to give European and American firms better access to their partner markets by eliminating trade barriers (i.e. tariffs and duties on imports, and non-tariff measures), improve regulatory environment by cutting red tape that complicates efforts to export, and set new rules to facilitate and promote fairness in trading and investing 1. With these objectives, the governments of EU member states and the US hope that TTIP could help to generate jobs and economic growth. With negotiations having started since 2013, it is the biggest bilateral trade partnership that has ever got negotiated. TTIP has the potential to have a large impact on the world economy, and in this paper, I aim to estimate its effects on international trade and welfare of the world.
The sheer scale of TTIP and the increasing complexity of global production linkages require any attempt of analysing the consequences of new trade policies to take into account the complication from cross-sectoral and cross-border trading activities. Fragmentation of global supply chains has become a key feature of international trade. Nowadays, we see products being produced not by just one but many countries. A classic example would be a smartphone being assembled in China but with components from Japan, South Korea or the US. Many studies have tried to put forward theories to characterise this fragmentation of global production (e.g. Antr`as, Garicano and Rossi-Hansberg, 2006; Costinot, Vogel and Wang, 2013; Grossman and Rossi-Hansberg, 2008; Rodr´ıguez-Clare, 2010; Yi, 2003). In the case of a bilateral tariff reduction, not only the industries with reduced tariffs are affected but prices in sectors that purchase materials from those industries will also be influenced. Moreover, new tariffs will also affect prices in non-tradable sectors that are purchasing production inputs from tradeable sectors. The magnitudes of these direct and indirect effects depend on the extent of sectoral linkages.
In this paper, by using a model with cross-sectoral and cross-border production linkages, I will estimate the trade and welfare effects of the Transatlantic Trade and Investment Partnership. Potentially, TTIP is a comprehensible trade pact that will cover many aspects, including market
1
access (concerning trades in goods and services), regulation (cutting red tape and costs), rules (con-cerning investment, competition, intellectual property, etc.) and institutional provisions. Despite this multifaceted complexity of the deal, I aim to examine the impact of one traditional aspect of a trade agreement, which is the reduction of tariffs. TTIP negotiations seek total elimination of tariffs and other duties and charges, with substantial elimination upon entry into force of the agreement and a transitional period for remaining tariffs imposed on certain products2.
I quantify the trade and welfare effects of a tariff reduction between EU and the US in the manner of Caliendo and Parro (2015) ( abbreviated as C&P henceforth), who estimate the effects of NAFTA on its members (Canada, Mexico, the US). Specifically, taking 2014 as the data reference point, I calibrate the model of Caliendo and Parro (2015), numerically project the effects of a hypothetical situation where import tariffs between EU members and the US are reduced to zero (which is the final goal of the agreement), and highlight which economies would benefit the most from a tariff reduction between EU and the US.
A number of studies have tried to examine the potential impact of TTIP, but with limitations. For instance, Francois et al. (2013) study different aspects of the trade agreement from a limited case of only tariff liberalization to a full-fledged free trade agreement that involves tariffs, non-tariff barriers and spill-over effects to third countries (e.g. falling trade-related costs to EU and the US, or adoption of similar regulations and standards). In the case of only tariff liberalization, the authors find positive impact on both EU and the US but negative effects on other regions of the world. Felbermayr et al. (2015) also project that TTIP will positively affect EU members (with stronger effects in geographically peripheral economies) and the US but negatively in the majority of third-party countries. The study of Francois et al. (2013) aggregates EU as a whole and highlights effects on different sectors of the economies. On the other hand, Felbermayr et al. (2015) analyse the EU members separately but use a multi-country one-sector model.
As a complement to these studies, my paper assesses the impact of a tariff reduction by using the model of Caliendo and Parro (2015), which features multiple countries and multiple sectors with inter-sectoral and global linkages through the trading of intermediate inputs and materi-als. Caliendo and Parro (2015) integrate cross-country inter-sectoral linkages into the influential
2
one-sector, multi-product and multi-country model by Eaton and Kortum (2002). Cross-sectoral linkages exist in the form of intermediate-good producers purchasing materials from domestic sup-pliers of all sectors. In turn, international trade exists in the purchases of globally lowest-priced intermediate goods to produce materials. The model further assumes different efficiencies in the production of intermediate goods across countries, which means that a country has comparative advantages to export only certain products. Together with tariffs on imports, geographical barriers incur additional costs on traded goods.
While Felbermayr et al. (2015) use a one-sector model, which does not reflect the role of fragmentation in the global production, the model of C&P features multiple sectors with inter-connection. Compared to the study of Francois et al. (2013) who use a more complex computable general equilibrium model, the C&P model offers simplicity in terms of data requirements and esti-mated parameters required for the estimaton. Furthermore, C&P propose to solve the equilibrium in relative changes against a baseline model, which reduces the reliance on estimates of unobserved parameters. The tractability of C&P model also allows for the decomposition of welfare effects after a tariff reduction into effects due to terms of trade and due to volume of trade. This is convenient for an analysis with a large number of countries, and I am able to estimate the effects for each member of EU instead of EU as one entity as in Francois et al. (2013).
To estimate the effects of a hypothetical tariff reduction between EU and the US, I first calibrate a baseline model with bilateral trade data, bilateral tariffs, gross outputs and gross value added, and input-output data. In particular, input-output data are obtained from the World Input-Output Database (WIOD), which covers 40 countries. Specifically, they are members of EU-27 and 12 major economies (Australia, Brazil, Canada, China, India, Indonesia, Japan, Korea, Mexico, Russia, Turkey, United States), as well as one representative for the rest of the world (RoW) 3. Therefore, I am able to estimate the effects on not only EU-27 members and the US but also 12 third-party economies. Given the calibrated baseline model, welfare gains (or losses) as relative changes of real income in the equilibrium are estimated following the approach of Caliendo and Parro (2015).
The main results show welfare gains for majority of EU members and the US following the 3
hypothetical elimination of bilateral tariffs. The decomposition of welfare changes shows that increases in volume of trade improve welfare across all countries in TTIP (except only for Lithuania). Terms of trade effects are positive for many countries. However, in some countries, negative terms of trade effects counterbalance the positive contribution by gains in volume of trade, and the net effect could even be a welfare loss. Overall, the hypothetical reduction in tariffs does not affect sectoral specialisation (i.e. shares of exports by sector remain more or less unchanged). There is a positive relationship between a country’s openness to trade before the tariff reduction and gains in volume of trade afterwards. However, countries that are more open to trade tend to have worse terms of trade effects and lower gains (or more losses) in welfare. In addition, there is evidence that countries with exports and imports more concentrated in certain sectors and certain trade partners tend to gain less. In other words, more diversified baskets of exports and imports (in terms of sectors and trade partners) are associated with larger welfare gains after the tariff reduction. Extensive robustness checking shows that the results of welfare gains, particularly the gains from volume of trade, for most economies negotiating TTIP remain robust.
Qualitatively, the mechanism underlying the welfare gains is that a tariff reduction makes im-ports of production inputs cheaper, which boosts the demand and lowers costs of production. Even non-tradable sectors enjoy lower production costs because they purchase materials from producers that use cheaper imported inputs. Subsequently, countries enjoy surging flows of trade. Together with higher demand for goods, wages increase and real income grows.
In terms of modelling, the C&P model is an extention of Eaton and Kortum (2002)’s Ricardian-type model, which features multiple countries and multiple goods but a single sector. Other studies that extend Eaton and Kortum (2002) model to multiple sectors include Alvarez and Lucas (2007), Arkolakis et al. (2012), and Costinot, Donaldson and Komunjer (2012). Nevertheless, as Caliendo and Parro (2015) emphasise, their model differs in the explicit formulation of sectoral linkages between tradable and non-tradable sectors. The model also differs in the assumption of different productivity levels for both tradable and non-tradable good producers. Furthermore, solving the equilibrium in relative changes minimises the data requirements and does not rely on estimates of some unobserved structural parameters.
while gravity models are used in ex post analyses (Plummer, Cheong and Hamanaka, 2010). Fran-cois et al. (2013) is an example of using a CGE model to assess the impact of a free trade agreement. Some studies that use CGE models to estimate the effects of NAFTA include Brown, Deardorff, and Stern (1995), Brown and Stern (1989), and Kehoe and Kehoe (1994). A number of papers, such as Fox (1999), Kehoe (2003) and Shikher (2012), evaluate the performance of CGE models in predicting effects of free trade agreements. Aside from general equilibrium models, gravity model (Tinbergen, 1962; Anderson, 1979) has been an empirical workhorse for analysing the patterns of trade flows and the effectiveness of free trade agreements (Anderson and van Wincoop, 2003; Baier and Bergstrand, 2007, 2009; Head and Mayer, 2014). A basic gravity model seeks to explain bilat-eral trade flows by regressing the trade flows on sizes of the economies and the geographic distance between them. One way to assess the impact of a free trade agreement is to include a binary variable in the regression to indicate whether an agreement exists between two certain countries or in a certain period (Plummer, Cheong and Hamanaka, 2010).
In terms of concepts, the C&P model and my analysis are related to the literature on the fragmentation of global supply chains and cross-sectoral linkages. For example, Costinot, Vogel and Wang (2013) theorise that the production of final goods requires a continuum of intermediate stages that are subject to mistakes. When a mistake occurs at some stage, the intermediate good is entirely lost. Therefore, countries with less skilled workers or worse infrastructure are more likely to experience costly defects and delays. In equilibrium, countries with lower probabilities of making mistakes at all stages specialise in later stages of production. Grossman and Rossi-Hansberg (2008) characterise the fragmentation of global production by proposing a theory on offshoring of tradable tasks. Offshoring may be attractive if certain factors of production can be hired more cheaply abroad than at home. Reductions in cost of trading tasks can generate gains for all domestic factors. Such theoretical studies provide complementary and broader views on the matter of cross-sectoral linkages aside from the simple notion of purchasing lowest-priced goods as in C&P model and other Ricardian trade models.
The paper is concluded in section 5.
2
Theoretical framework
In this paper, I follow strictly the theoretical model of Caliendo and Parro (2015). First, I describe the multi-country multi-sector model, which is an extension of the Ricardian trade model of Eaton and Kortum (2002) to incorporate cross-border sectoral linkages. It continues with explanations on the decomposition of welfare effects due to tariff reduction and the estimation approach.
2.1 The Model
The model features households and two types of producers, one that produces intermediate goods and another one that produces composite intermediate goods. Producers of composite intermediate goods also produce final goods for the consumption of households. Each producer makes goods within a certain sector, and each sector has two types, tradable and non-tradable.
2.1.1 Households
In each country n ∈ N , there are Ln households who maximize utility by consuming goods Cnj according to the optimization problem:
max Cnj u(Cn) = J Y j=1 Cjα j n n (1) , subject to wnLn+ Rn+ Dn= PJ j=1C j n, where PJj=1αjn= 1.
Denote In= wnLn+Rn+Dnas households’ income, which includes labour wages wnLnas well as lump-sum transfers Rn (tariff revenues) and Dn (sectoral trade deficits, i.e. transfers from the rest of the world). Parameters αnj (with j = 1, . . . ., J ) indicate utility elasticities of comsumption goods. Consumption goods Cnj are produced by domestic producers of composite intermediate goods in sector j. In equilibrium, the households spend a fixed portion of income on each consumption good, or that Cnj = αjnIn.
Therefore, higher demand for consumption goods boosts trading indirectly because producers in tradable sectors import globally lowest-priced intermediate goods for the production.
2.1.2 Intermediate goods
Each sector j produces a continuum if intermediate goods ωj ∈ [0, 1]. Producers employ labour and use composite intermediate goods (materials) from all domestic sectors (both tradable and non-tradable). Value added is generated in this sector. The producers of tradable goods in this sector sell there products to the international market but only purchase materials from the domestic market. That is, there are exports but no imports for this type of producers. Producers of non-tradable goods sell products to the domestic market only. Each good ωjis produced with a technical efficiency of zjn(ωj). Specifically, the production technology of a good ωj in country n is
qjn(ωj) = zjn(ωj)[ljn(ωj)]γnj
J Y
k=1
[mk,jn (ωj)]γk,jn
where qnj(ωj) is the output, ljn(ωj) is labour being employed, and mk,jn (ωj) are the composite intermediate goods from domestic sector k being used in the production of good ωj. Regardless of whether sector j is tradable or non-tradable, the producers purchase composite goods from all sectors of the domestic economy. This is different from the composite good sector where producers purchase inputs only from the same sector.
Parameters γnk,j are output elesticities of composite intermediate goods, and in equilibrium, they also indicate the shares of materials from sector k in the production. Parameter γnj indicates the output elasticity of labour and also the share of value added in the output. These parameters vary across sectors and countries, but it always holds that PJ
k=1γ k,j
n = 1 − γnj. In other words, there is constant return to scale.
A firm’s optimisation problem is to maximise profit
pjn(ωj)qjn(ωj) − wnljn(ωj) − J X
k=1
Pnkmk,jn (ωj).
Since markets are perfectly competitive and production is at constant returns to scale, it is optimal for firms to set prices pjn(ωj) at unit costs, which equal cjn/zjn(ωj). Variable cjn denotes the cost of an input bundle, or
cjn= Υjnwγnj n J Y k=1 (Pnk)γnk,j (2) where constant Υjn=QJk=1(γ k,j n )−γ k,j n (γj n)−γ j
n. Intuitively, unit cost depends on the level of wages and prices of materials, and the exact cost is determined based on the shares of labour and materials in the production.
2.1.3 Composite intermediate goods
To cater to the domestic production of intermediate goods, producers of composite intermediate goods in sector j of country n purchase intermediate goods ωj and supply Qj
n goods according to production function Qjn= Z rjn(ωj)1−1/σjdωj σj/(σj−1)
where σj is the elasticity of substitution. Here, producers purchase inputs only within the same sector. In other words, there is cross-border trading but no cross-sectoral linkage. In the case of tradable goods, rjn(ωj) is the demand of goods ωj from the lowest-cost supplier in the international market. In the case of non-tradable goods, rjn(ωj) is the demand of goods ωj from domestic suppliers.
Producers maximise profits
PnjQjn− Z
pjn(ωj)rjn(ωj)dωj
optimisation problem gives the demand rnj(ωj) = p j n(ωj) Pnj !−σj Qjn
and the price
Pnj = Z pjn(ωj)1−σjdωj 1 1−σj .
Intuitively, the demand of a certain good depends on how cheap (or expensive) it is compared to other goods. The price of a composite good depends on prices of the intermediate goods used in its production.
The composite intermediate good producers supply to not only the production of intermediate goods but also the final consumption of local households. In other words,
Qjn= Cnj + J X k=1 Z mj,kn (ωk)dωk.
There are no exports for this type of producers but only imports. Furthermore, there are no direct imports of final consumption goods. Changes in the demand of households will affect trade values through the purchase of imported goods by producers. The supplies and cross-purchases between intermediate good producers and composite good producers create cross-sectoral and cross-border linkages.
2.1.4 International trade costs and prices
The model assumes two types of trade costs: iceberg costs and ad-valorem flat-rate tariffs. To ship 1 unit of tradable intermediate goods in sector j to country n, the tradable sector in country i needs to produce djni > 1 due to ice-berg costs. Domestic shipment is assumed to be cost-free, or djnn = 1. In addition, goods imported into country n from country i have to pay an ad-valorem flat-rate tariff τnij over each unit of goods. Total trade costs to ship 1 unit of good is then
The model further assumes triangular inequality κjnhκjhi > κjni. That is, it is more costly to ship goods from country i to h and then the same goods from h to n than to ship goods directly from country i to country n.
Intermediate good producers of sector j in country i set prices at unit costs, cji/zij(ωj). However, given trade costs, unit prices of imported intermediate good ωj shipped from country i to country n are cjiκjni/zji(ωj). Therefore, the lowest-cost price of good ωj in country n is
pjn(ωj) = min i ( cjiκjni zji(ωj) ) .
In non-tradable sectors, trade costs are assumed to be infinitely high, κjni= ∞, and hence, goods are purchased from domestic suppliers. In these sectors, the price is pjn(ωj) = cjn/znj(ωj) and the demand of good ωj is also the amount being produced domestically and given by rj
n(ωj) = qnj(ωj). Following Eaton and Kortum (2002), Caliendo and Parro (2015) assume that the efficiency zjn(ωj) of producing an intermediate good ωj in country n is the realisation of a Fr´echet distribution
Fnj(z) = e−λjnz−θj
where location parameter λjn> 0 varies across sectors and countries, and shape parameter θj varies across sectors. A higher value of λjn makes the average productivity of the sector j higher, which is a notion of absolute advantage, while a smaller value of θj implies higher dispersion of productivity across goods ωj, which determines comparative advantage. For the model to be solved, Caliendo and Parro (2015) assume independent distributions across goods, sectors and countries, and that 1 + θj > σj.
The price of composite intermediate good in a tradable sector j of country n depends on the productivity of the sector and the costs of globally purchased inputs. It is given by
Pnj = Aj " N X i=1 λji(cjiκjni)−θj #−1/θj (4)
, where Aj is a constant. And the price in a non-tradable sector, where κjni = ∞, is Pnj = Aj(λj
n)−1/θ
We see that changes in trade costs κjni, inclusive of changes in tariffs, have direct effects on prices of tradable composite intermediate goods Pnj. Subsequently, through the cross-sectoral and cross-border linkages, changes in trade costs influence the costs of input bundles cjn indirectly.
Since households purchase consumption goods from composite intermediate good producers, the prices of consumption goods are also Pnj. Given households’ optimal behaviour, the consumption price index is given by
Pn= J Y
j=1
(Pnj/αjn)αjn. (5)
It is intuitive that the price index depends on prices of consumption goods Pnj and proportions of goods being consumed αjn.
2.1.5 Expenditure shares
Denote Xnj = PnjQjn as the total expenditure on sector j goods by country n. And Xnij is the expenditure by country n on sector j goods from country i. Then, we have Xnj =PNi=1Xnij , and the share of expenditure in country n on sector j goods from country i is πnij = Xnij /Xnj.
Given the Fr´echet distribution, expenditure share can be derived as a function of technologies, prices and trade costs
πnij = λ j i[c j iκ j ni]−θ j PN h=1λ j h[c j hκ j nh]−θ j. (6)
It can be seen that changes in tariffs, which leads to changes in trade costs κjni, have direct effects on expenditure shares. When tariffs change, producers adjust their purchases of imports, and the composition of their purchases changes accordingly.
2.1.6 Total expenditure, trade balance and wage
where γnj,k is the share of goods j as inputs in the production of goods k, andPNi=1Xik π k in 1+τk in is the total expenditure of all countries on goods k produced by country n (excluding payment of tariffs). Hence, the whole term PJ
k=1γ j,k n PNi=1Xik π k in 1+τk in
indicates the demand for goods of sector j by the producers in all sectors of country n to meet the global demand for their products. As mentioned earlier, αjnIn is the consumption by households on sector j goods in country n.
Households’ income In in country n is the sum of labour wages and lump-sum transfers
In= wnLn+ Rn+ Dn (8)
, where total tariff revenues being imposed on imports Rn = PJj=1 PN i=1τ j niM j ni. Variable M j ni = Xnj πjni
1+τnij indicates country n’s imports of sector j goods from country i.
National trade deficits Dnis the sum of all sectoral trade deficits Dn=PJk=1Dkn. On the global scale, the summation of all countries’ deficits is zero, or thatPN
n=1Dn= 0. A sector’s trade deficits are the difference between the sector’s total imports and total exports, Dnj =PNi=1Mnij −PNi=1Enij, where Enij = Xij π
j in
1+τinj is country n’s exports of sector j goods to country i.
Given the definition of national deficits, sectoral decifits and expenditure, we derive the condi-tion for the balance of trade in country n,
J X j=1 N X i=1,i6=n Xnj π j ni 1 + τnij − Dn= J X j=1 N X i=1,i6=n Xij π j in 1 + τinj . (9)
The left-hand side is country n’s total imports minus national deficits, and the right-hand side is the country’s total exports. By definition, trade deficits equal total imports minus total exports. To determine wage rates, notice that revenues of intermediate-good producers in sector j of country n is equal to the total expenditure of all countries on the intermediate goods of this sector, and labour earns a share of γnj out of revenues. That is, labour in sector j of country n earns γnj PNi=1Xij π
j in
1+τinj . Aggregating these earnings across all sectors give the total wage payments of country n. Therefore, the wage rate in country n is
2.2 Equilibrium
2.2.1 Equilibrum in absolute changes Definition 1. (Caliendo and Parro, 2015, p. 11)
“Given Ln, Dn, λjn and djni, an equilibrium under tariff structure τ is a wage vector w ∈ RN++ and prices {Pnj}J,Nj=1,n=1 that satisfy equilibrium conditions (2), (4), (6), (7)and (9) for all j, n.”
2.2.2 Equilibrium in relative changes
Solving an equilibrium in relative changes offer some advantages: (i) matching data to a base year and analyse changes with respect to the base year, (ii) identifying effects on equilibrium outcomes due to pure changes in tariffs, and (iii) no need to estimate paramaters that are hard to identify in the data, i.e. productivities λjn and ice-berg costs djni (since these are cancelled out during the derivation in relative changes).
Definition 2. (Caliendo and Parro, 2015, p.11)
“Let (w, P ) be an equilibrium under tariff structure τ and let (w0, P0) be an equilibrium under tariff structure τ0. Define (w, ˆb P ) as an equilibrium under τ
0 relative to τ , where a variable with a
hat ”ˆx” represents the relative change of the variable, namely ˆx = x0/x. The equilibrium conditions in relative changes satisfy:
Cost of input bundles
ˆ cjn= ˆωγni n J Y k=1 ( ˆPnk)γnk,j (10) Price index ˆ Pnj = " N X i=1 πnij [ˆκjnicbji]−θ j #−1/θj (11) Bilateral trade shares
ˆ πnij = " ˆ cjiκˆjni ˆ Pnj #−θj (12) Total expenditure in each country n and sector j
Trade balance J X j=1 N X i=1 πj0ni 1 + τnij0X j0 n − Dn= J X j=1 N X i=1 πjin0 1 + τinj0X j0 i , (14) where ˆ κjni= (1 + τnij0)/(1 + τnij), In0 =wcnwnLn+PJj=1PNi=1τj 0 ni πnij0 1+τnij0X j0 n + Dn,” and w0n= L1 n PJ j=1γ j nPNi=1X j0 i πinj0 1+τinj0.
In the main analysis in section 4, deficits Dn are assumed to be exogenous. Therefore, given parameter values and exogenous changes in trade costs ˆκjni, solving for the equilibrium post-tariff reduction requires finding the endogenous variables: relative changes in costs of input bundles ˆcjn, wages ˆωn, prices of composite goods ˆPnj, expenditure shares ˆπnij and new expenditure X
j0
n for all sectors and countries.
2.3 Welfare effects from tariff changes
Welfare of households in a country n is Wn= In/Pn, or the value of real income. Total differentiat-ing Wn together with using equilibrium conditions gives the change in welfare due to tariff changes as d ln Wn= 1 In J X j=1 N X i=1 (Enij d ln cjn− Mnij d ln cji) + 1 In J X j=1 N X i=1 τnij Mnij (d ln Mnij − d ln cji) (15)
where the first term on the right hand side indicates the multi-lateral and multi-sectoral terms of trade effects, and the second term measures the multi-lateral and multi-sectoral volume of trade effects.
The contribution of a sector j to the welfare gains of country n due to changes in terms of trade, or PN
i=1(E j nid ln c
j
n− Mnijd ln cji) depends on sectoral trade deficits (changes in exports Enij and imports Mnij) and sectoral changes in export and import prices (d ln cjn and d ln cji).
trade, orPN i=1τ j niM j ni(d ln M j ni− d ln c j
i), depends on how much more trade is created (as measured by changes in imports deflated by changes in import prices, d ln Mnij − d ln cji). The initial tariffs and import values τnijMnij weight how important the change (d ln Mnij − d ln cji) is to the welfare gains.
Based on the equation (15) of aggregate welfare changes, we can decompose welfare gains into contributions by bilateral and sectoral changes. Specifically, we can decompose the total welfare effect of country n by trading partner into changes in bilateral terms of trade between country n and any country i
dln totni= J X
j =1
(Enij dlncjn− Mnijdlncji)
and changes in bilateral volume of trade between country n and country i
dln votni= J X
j =1
τnij Mnij(d ln Mnij − d ln cij).
Alternatively, we can decompose the welfare change of country n by sector into changes in sectoral terms of trade of sector j
dln totjn = N X
i =1
(Enij d ln cjn− Mnij d ln cji),
and changes in sectoral volume of trade of sector j
dln votjn = N X
i =1
τnijMnij(d ln Mnij − d ln cij).
By aggregating bilateral changes across trade partners or by aggregating sectoral changes across sectors, we should arrive at the total welfare changes as in equation (15). That is
d ln Wn= 1 In N X i=1 (d ln totni+ d ln votni) = 1 In J X j=1 (d ln totjn + d ln votjn).
2.4 Approach to find numerical solution
tariffs is equivalent to a decrease in total trade costs. Specifically, a change of tariff rate τ to rate τ0 on sector j imports by country n from country i is captured by the relative change of trade cost
b
κjni. In addition, it should be remembered that aggregate trade deficits Dn are fixed at the values in the base year 2014.
To numerically solve the equilibrium in relative changes,
i. start with a guess on the vector of wage changes ˆw = ( ˆw1, . . . ., ˆwN), e.g. ˆw=1 (which means that wages are unchanged as compared to the base year).
ii. Given the vector of wages, solve equations (10) and (11) for the vectors of changes in composite intermediate good prices bPjn( ˆw) and input-bundle costs ˆcjn( ˆw).
iii. Given prices bPjn( ˆw), input-bundle costs ˆcjn( ˆw), old tariffs πnij together with tariff changesκb j ni and parameter θj, solve equation (12) for the new expenditure shares πnij0( ˆw).
iv. Given new expenditure shares πjni0( ˆw), new tariffs τ0, parameters γj, γj,kn and αjn, solve equation (13) for the new total expenditures Xnj0( ˆw).
v. Given new total expenditures Xnj0( ˆw), new expenditure shares πj
0
ni( ˆw), new tariffs τ
0 and trade
deficits Dn, check if the trade balance holds in equation (14).
vi. If the trade balance in (14) does not hold, adjust the guess for wage vectors ˆw until the condition (14) is obtained.
2.5 Estimate trade elasticities
To estimate trade elasticities, Caliendo and Parro (2015) propose a convenient approach that does not require heavy data collection. Consider three countries indexed by n, i and h. Given the expression for expenditure shares πjni= Xnij/Xnj and equation (6) of
πjni= λ j i[c j iκ j ni] −θj PN h=1λ j h[c j hκ j nh]−θ j
, we can derive an expression that involves only trade flows, trade costs and trade elasticities,
XnijXihj Xhnj Xnhj XhijXinj = κjni κjin κjih κjhi κjhn κjnh !−θj . (16)
Also recall that total trade costs include both ice-berg costs and ad-valorem tariff payments, κjni = (1 + τnij)dnij =˜τnijdjni. Further modelling ice-berg cost as a linear function of cross-country characteristics gives
ln κjni= ln ˜τnij + ln djni= ln ˜τnij + νnij + µjn+ δji + εjni (17)
where νnij = νinj are symmetric bilateral trade costs (i.e. distance, language, common border, belonging to same an FTA or not), µjn captures country n (exporter)’s sectoral fixed effects, δji captures country i (importer)’s fixed effects, and εjni are external random disturbances.
Combining equations (16) and (17) gives
ln X j niX j ihX j hn Xnhj XhijXinj ! = −θjln τ˜ j ni ˜ τinj ˜ τihj ˜ τhij ˜ τhnj ˜ τnhj ! + ˜εj (18) where ˜εj = εjin− εjni+ εhij − εjih+ εjnh− εjhn.
3
Data
The required data for model calibration include bilateral trade flows, bilateral tariffs, value added and gross outputs by sector, and input-output tables. Since I use input-output data from the World Input-Output Database (WIOD), the numbers of countries and sectors are in accordance with WIOD. Specifically, the input-output data cover 40 countries, including members of EU-27 and 12 major economies (Australia, Brazil, Canada, China, India, Indonesia, Japan, Korea, Mexico, Russia, Turkey, United States), as well as data for one representative for the rest of the world (RoW)4. Non-WIOD sectoral data will be aggregated into 35 ISIC Rev. 2 industries as in WIOD. Since the results could be sensitive to the magnitudes of tariff reduction, I collect the latest available data to reflect the most up-to-date values of tariffs. For the countries in the sample, latest and complete tariff data are of 2014. The data sources are as follows
1. Bilateral trade flows: 2014 data from UN-COMTRADE database. Commodities are defined according to HS1992 system. I re-classify and aggregate them into 35 ISIC Rev. 2 industries as in WIOD.
2. Tariffs: 2014 data from UNCTAD-TRAINS database (accessed from the World Integrated Trade Solution). I re-classify average tariffs (weighted by values of imports) on imported com-modities from HS2007 system into 35 ISIC Rev. 2 industries. To construct the hypothetical tariff reduction between EU and the US, I replace tariffs between these 27 EU countries and the USA by 0% (which is the final target of the trade agreement) while keeping tariffs with other countries unchanged at 2014 values.
3. Value Added and Gross Output: from WIOD and United Nations Statistics Division. Since WIOD only has data up to 2011 and UN Statistics Division does not provide 2014 data of value added by sector as required, I first use 2011 WIOD data to calculate the ratio of value added over gross output for each sector. I assume that these ratios hold constant over the years. I further assume that the share of each sector in total value added (computed from WIOD’s 2011 data) also remains unchanged between 2011 and 2014. Using these ratios and the 2014 total value added by country from UN Statistics Division, I construct the value added and gross outputs by sector for each country in 2014.
4
4. Input-Output Tables and Intermediate Consumption: from WIOD. Since WIOD has data up to only 2011, I assume that the input-output structure of all economies remain unchanged between 2011 and 20145.
Table 1 provides shortened names of the 35 sectors and their full description in accordance with WIOD. The first 16 sectors are tradable, and the remaining 19 ones are non-tradable. Tradable sectors are determined by whether large bilateral trade flows across countries exist in the data. For some non-tradable sectors such as Other Services, trade flows occur between few sectors and countries, which are negligible and cannot be defined as tradable.
Table 1: Tradable and non-tradable sectors
Sector ID Sector Description
Tradable
1 Agriculture Agriculture, Hunting, Forestry and Fishing
2 Mining Mining and Quarrying
3 Food Food, Beverages and Tobacco
4 Textile Textiles and Textile Products
5 Footwear Leather, Leather and Footwear
6 Wood Wood and Products of Wood and Cork
7 Paper Pulp, Paper, Paper , Printing and Publishing
8 Petroleum Coke, Refined Petroleum and Nuclear Fuel
9 Chemicals Chemicals and Chemical Products
10 Plastics Rubber and Plastics
11 Minerals Other Non-Metallic Mineral
12 Basic metals Basic Metals and Fabricated Metal 13 Machinery, nec Machinery, Nec
14 Electrical Electrical and Optical Equipment
15 Transport Transport Equipment
16 Other Manufacturing, Nec; Recycling
Non-tradable 17 Electricty Electricity, Gas and Water Supply
18 Construction Construction
19 Auto-related Sale, Maintenance and Repair of Motor Vehicles and Motorcycles; Retail Sale of Fuel 20 Wholesale Wholesale Trade and Commission Trade, Except of Motor Vehicles and Motorcycles 21 Retail Retail Trade, Except of Motor Vehicles and Motorcycles; Repair of Household Goods
22 Hotels Hotels and Restaurants
23 Land transport Inland Transport 24 Water transport Water Transport
25 Air transport Air Transport
26 Aux. transport Other Supporting and Auxiliary Transport Activities; Activities of Travel Agencies
27 Post Post and Telecommunications
28 Finance Financial Intermediation
29 Real estate Real Estate Activities
30 Renting Mach Renting of M&Eq and Other Business Activities 31 Public Public Admin and Defence; Compulsory Social Security
32 Education Education
33 Health Health and Social Work
34 Other services Other Community, Social and Personal Services
35 Private Private Households with Employed Persons
5
Table 2 shows the tariffs being imposed on imports between EU and the US in 2014. Overall, the tariffs are already quite low in 2014. EU countries impose higher tariffs on goods from the US than the rates imposed by the US on EU products. The rates are highest in Food, Textile and Agriculture. Tariffs on Paper, Petroleum, Mining and Machinery, nec are particularly low.
Table 2: The hypothetical reduction of tariffs between EU and the US, in % Import tariffs of Sector ID Sector EU on US US on EU 1 Agriculture 7.04 1.91 2 Mining 0.05 0.32 3 Food 18.90 4.83 4 Textile 8.21 8.92 5 Footwear 6.49 7.11 6 Wood 2.90 1.90 7 Paper 0.10 0.07 8 Petroleum 0.99 0.68 9 Chemicals 4.65 2.76 10 Plastics 4.58 2.87 11 Minerals 3.49 3.13 12 Basic metals 2.01 1.51 13 Machinery, nec 1.85 1.20 14 Electrical 2.37 1.83 15 Transport 4.04 1.84 16 Other 2.70 2.62
4
Results
4.1 Estimates of trade elasticities
As mentioned above, the estimation of trade elasticities only needs bilateral trade flows and bilateral import tariffs. The sources and the aggregation of these data are similar to the description in section 3.1. However, I collect 2014 data for 16 countries and regions: Argentina, Australia, Canada, Chile, China, Colombia, EU25 6, India, Indonesia, Japan, Korea, Norway, New Zealand, Switzerland, Thailand and the USA. These 16 countries represent all major economic regions across the world and should produce estimates of trade elasticities that are applicable to the welfare estimation in 6Here, I should use data of EU25 as a whole and not separate data of each member state. Since trade elasticities
this paper.
Following the approach of Caliendo and Parro (2015), I do the estimation twice, one time with the full sample and the other time with a 99% sample in which countries with trade flows less than 1% of total trade values in each sector are removed. The 99% is to prevent the influence of outliers on the estimates. This influence could be seen from equation (18), whose left-hand side ln
XnijXihjXhnj Xnhj XhijXinj
involves bilateral trade flows between any three countries. Suppose that country n has trade flows less than 1% of total values in sector j, the ratios Xnij /Xinj and Xhnj /Xnhj will tend to be too small or too large (e.g. for small numbers, Xnij /Xinj = 3/1 = 3, but for large numbers, Xihj /Xhij = 102/100 = 1.02). Hence, in the presence of outliers, the estimates of trade elasticities will tend to be either inflated or deflated. A change in tariffs is unlikely to cause large adjustments in very small flows of trade, but inflated or deflated elasticities of trade will significantly affect the estimated adjustments of large trade flows in the counterfactual exercise. Therefore, it is reasonable to exclude outliers.
Table 3: Trade elasticities, or the dispersion of productivity
100% sample 99% sample Sector θj s.e. N θj s.e. N Agriculture 0.302 0.437 452 0.302 0.437 452 Mining 4.233 3.975 373 1.679 5.019 229 Manufacturing Food 1.015 0.396 452 1.273 0.418 361 Textile 7.331 1.625 452 8.532 1.971 283 Footwear 5.983 1.755 426 5.625 2.377 207 Wood 21.868 3.781 394 15.781 4.223 349 Paper 9.197 4.988 439 13.368 3.493 361 Petroleum 30.917 7.714 314 16.787 8.221 249 Chemicals -0.959 0.857 452 2.690 2.550 361 Plastics 10.836 2.697 452 10.836 2.697 452 Mineraln 1.897 2.363 439 -0.216 2.592 283 Basic Metals 32.224 4.663 452 32.897 5.217 361 Machinery, nec 11.759 2.509 452 13.981 2.964 283 Electrical 2.636 2.936 452 0.771 3.157 218 Transport 2.834 1.970 439 9.328 1.837 361 Other 3.122 2.641 452 10.683 2.181 218 Aggregate 2.132 0.313 6,892 2.194 0.329 5,028
estimation results show that, given a change in tariffs, we should expect small adjustments in half of Manufacturing sectors, least strongly in Basic Metals and relatively more strongly in Petroleum, Wood, Paper and Machinery. On the other hand, large increases or decreases in trade flows could be expected in Agriculture, Food, Chemicals, Minerals and Electrical following new policies on tariffs.
The estimated elasticities are consistent with the shares of exports by sector in tables 9, 9 and 9 in the next section. We see that sectors with low elasticities (e.g. Chemicals and Electrical) have high shares in exports across countries, which is consistent with the notion that goods of these sectors are more substitutable. At the same time, sectors with high elasticities (e.g. Petroleum and Basic Metals) have very low shares in exports of most countries and high shares only in certain countries (for instance, Basic Metals has low shares in most countries but high in Australia and Russia), meaning that goods of these sectors are less substitutable.
We see that the estimates are robust for most sectors between 100% and 99% samples. In the cases of Chemicals and Minerals, the signs of estimates change across samples. Therefore, I will use the estimates with 99% sample but the elasticity of Minerals will be replaced by the value from the 100% sample so that the elasticities of all sectors have positive values. Despite collecting data for the same set of 16 countries, my estimates are markedly different from the results in Caliendo and Parro (2015). On one hand, my aggregation into sectors in accordance with WIOD is different from C&P. On the other hand, C&P use 1993 data while mine are of 2014, and changes in trade elasticities reflect the changes in comparative advantages having occurred during that period of 21 years. In sectors like Agriculture, the low value of 0.302 means that these countries can now produce and export Agriculture products. This is contrary to the value of 9.11 in C&P, which means that the production of Agriculture goods was less widespread and goods were less substitutable across countries in 1993.
flows.
4.2 Main results with 2014 trade deficits
The welfare effects of tariff reduction are obtained by estimating the relative changes of welfare (i.e. real income) in the equilibrium against a benchmark model calibrated with 2014 data. As mentioned in section 3, 2014 is the latest year with full available data. The results should be sensitive to the data that are used to calibrate the model, and I choose 2014 as the base year to reflect the most up-to-date policy on tariffs.
In the theoretical model, national trade deficits are assumed to be exogenous, while sectoral trade balances are endogenously determined. Therefore, exogenous values of trade balances of 2014 will remain in the counterfactual equilibrium with tariff changes. Nevertheless, it can be expected that the estimates of welfare effects will be sensitive to the values of trade balances being assumed in the equilibrium. In this section, I leave trade balances of 2014 to hold in the post-tariff-reduction equilibrium. In section 4.3, I will check the robustness of quantitative results by assuming different values for national trade balances.
Table 4 reports the welfare effects, the contributions due to changes in terms of trade and in volume of trade. The table also shows the effects on wage levels across countries after the reduction of tariffs between the US and European Union. I separate the table into two sections of countries. The first section shows the results for the EU members and the US. The second section displays the effects for other countries not in the TTIP.
The results show that 20 (out of 27) EU countries together with the US can expect welfare improvement following the bilateral tariff reduction. The gains are particularly strong in Lithuania (0.453%), Luxembourg (0.162%), Romania (0.147%) and Ireland (0.121%). In other countries, gains are ranging from 0.004% to 0.034%. Estonia and Malta expect no changes, while Czech Republic, Hungary, Latvia, Poland and Slovenia experience welfare decreases. Overall, all larger economies (e.g. the US, UK, Germany and France) can expect moderate welfare gains. However, while some smaller economies have large gains, some others have welfare losses.
Table 4: Welfare effects from the tariff reduction between EU and the US Total Effects Terms of Trade Volumes of Trade Real wages
Table 5: Indicators of trade, pre-tariff reduction
Trade Openness HHI exports HHI imports Austria lower 0.015 0.021 Belgium higher 0.015 0.018 Bulgaria higher 0.014 0.023 Cyprus lower 0.023 0.024 Czech higher 0.019 0.020 Denmark lower 0.013 0.012 Estonia higher 0.026 0.014 Finland lower 0.011 0.019 France lower 0.014 0.015 Germany lower 0.011 0.013 Greece lower 0.061 0.053 Hungary higher 0.022 0.021 Ireland lower 0.056 0.021 Italy lower 0.011 0.017 Latvia lower 0.015 0.013 Lithuania lower 0.013 0.041 Luxembourg higher 0.020 0.023 Malta lower 0.096 0.031 Netherlands higher 0.016 0.018 Poland higher 0.011 0.018 Portugal lower 0.011 0.025 Romania lower 0.014 0.012 Slovakia higher 0.018 0.023 Slovenia higher 0.017 0.012 Spain lower 0.012 0.028 Sweden lower 0.011 0.013 UK lower 0.018 0.015 USA lower 0.016 0.024 Australia lower 0.114 0.024 Brazil lower 0.036 0.028 Canada lower 0.088 0.040 China lower 0.041 0.053 India lower 0.037 0.138 Indonesia lower 0.029 0.042 Japan lower 0.036 0.066 Korea higher 0.054 0.066 Mexico lower 0.099 0.041 Russia lower 0.047 0.015 Turkey lower 0.021 0.016 RoW lower 0.022 0.022
Table 6: Coefficients of correlation between welfare-related effects and pre-tariff reduction trade indicators
Total effects Terms of Trade Volume of Trade Real Wages HHI exports HHI imports Trade openness Total effects 1.00 Terms of Trade 0.97 1.00 Volume of Trade 0.25 0.01 1.00 Real Wages 0.90 0.83 0.40 1.00 HHI exports -0.21 -0.15 -0.25 -0.24 1.00 HHI imports -0.09 -0.02 -0.31 -0.07 0.31 1.00 Trade openness -0.26 -0.32 0.20 0.11 -0.02 0.12 1.00
for welfare reduction in the five countries mentioned above. Furthermore, the model predicts strong positive effects on wages across all countries negotiating TTIP, except for Latvia.
Recall from the decomposition of welfare effects in equation (15) that terms of trade gains stem from increases in export prices (weighted by 2014 export values) relative to changes in import prices (weighted by 2014 import values); and volume of trade effects are influenced by the changes in import values deflated by import prices. Therefore, negative terms of trade effects in 12 countries of the trade agreement mean that they have weighted export prices improve less than prices of their trade partners. Interestingly, 8 out of these 12 economies are among the most open to international trade, as shown by the first column of table 5. This column indicates whether a country has a more-than-average ratio of total trade values (imports + exports) over total value added before the tariff reduction. Countries with more than average trade openness tend to have negative terms of trade effects. Since terms of trade effects depend crucially on the structure of an economy (i.e. input-output coefficients, which influence what and how much of a product each country would export or import), it is possible that these more open countries rely more on imports from countries with fast rising wages and costs.
The model also explains why other economies outside the trade pact might expect welfare reduction. It is possible that goods from these countries become relatively expensive after the EU-US tariff deal, which leads to decreasing trade flows and lower wages. Lower wages also influence terms of trade negatively. Table 4 shows that the net effect on welfare is negative for all countries outside TTIP, except for India.
However, there could be cases where countries outside of the trade agreement would also enjoy some benefits. That is when these countries also purchase cheaper imports for their production, or when reduced wages make their exports more competitive, which creates a boost to exports (see table 13). Moreover, rising income in EU and the US could also create more demand for products from non-TTIP countries and help to increase these countries’ exports.
To show if there is a relationship between welfare gains, a country’s openness to trade and the diversification of its trade portfolios, table 5 shows whether a country has a more-than-average ratio of total trade values over total value added in the first column, Herfindahl-Hirschman Index (HHI) to show the diversification of exports in the second column, and HHI to show the diversification of imports in the third column 7. A lower value of HHI means that the exports (or imports) of a country is more diversified and more widespread to sectors across countries. In contrast, a higher value of HHI means that exports (or imports) of this country are more concentrated in a smaller number of sectors and trade partners. These trade indicators are computed with the benchmark 2014 data, and they are meant to indicate if there is a relationship between a country’s characteristics before the tariff reduction and its welfare changes afterwards. Table 6 shows the coefficients of correlation between the three trade indicators and the welfare-related effects.
As mentioned before, 8 out of 12 countries with negative terms of trade effects have more-than-average openness to trade. This observation is consistent with the negative correlation (-0.32) between openness and terms of trade effects in table 6. Nevertheless, we see that openness is positively correlated to the gains in volume of trade, which is commonly expected. The HHIs of exports and imports are negatively correlated with all welfare-related effects, particularly with the volume of trade effects8. That is, countries with less diversified portfolios of trade (exporting and
7
In order to compute these Herfindahl-Hirschman indices, I first compute the shares of exports (or imports) by sector and country. There are 16 tradable sectors and each country can have up to 39 trade partners. Hence, Herfindahl-Hirschman index of a country is the sum of squared values of 16 × 39 sectoral shares of exports (or imports).
importing in a smaller number of sectors and trade partners) before the tariff reduction tend to have smaller gains in wages and welfare. An explanation is that, since tariffs are unanimously reduced in all sectors, countries with more diversified trade portfolios can reap benefits from a larger number of sectors.
Tables 7 reports the contributions by sector to the terms of trade effects. It should be noted that, in cases of negative total effects, positive sectoral contributions mean welfare reduction. Sectoral contributions to terms of trade effects vary widely across sectors and countries. There is no clear pattern nor any apparent connection from the magnitude of tariff reduction or the dispersion of productivities to these sectoral contributions. A caveat is that while the percentage changes of Denmark in table 7 are particularly large, the terms of trade change is only 0.001%, and hence, the nominal contributions by sector are not unusually large.
We see clearer trends in the sectoral contributions to volume of trade effects, which are reported in table 8. Some sectors (e.g. Food, Chemicals and Electrical) show relatively significant contri-butions, while some others (e.g. Mining, Paper and Minerals) contribute modestly in all TTIP countries. Noticeably, sectors with most significant contributions are also the ones with highest shares in total exports. Table 9 shows the export shares by sector of each country before and after the tariff reduction. Given the Ricardian nature of the model, sectors with highest shares are the ones with comparative advantage in a country. Across all countries, the shares of export do not substantially change after the new tariff rates. Therefore, the increases of trade flows occur consis-tently across all sectors according to their shares in exports, regardless of comparative advantages. As in Caliendo and Parro (2015), I present normalised Herfindahl index values (HH index) as a measure of sectoral specialisation. We see that the index values remain more or less unchanged in all countries. The tariff reduction has no impact on the sectoral specialisation.
It should be noted that the tariff reduction has no effect on sectoral specialisation, but table 6 has shown that the diversification of trade, which is related to sectoral specialisation, is important to welfare gains. Overall, the estimates show welfare improvement in EU and the US. Especially, countries that are not over-exposed to trade (i.e. moderately open to trade) and more diversified in their baskets of trade products tend to have higher gains in welfare.
Table 7: Shares by sector in the terms of trade effects, in % Tradable sectors
Table 8: Shares by sector in the volume of trade effects, in % Tradable sectors
Table 9: Export shares by sector before and after tariff reduction between the US and European Union, in %
Table 9: Export shares by sector before and after tariff reduction between the US and European Union, in % (continued)
Luxembourg Malta Netherlands Poland Portugal Romania Slovakia Slovenia Sector Before After Before After Before After Before After Before After Before After Before After Before After Agriculture 4.80 4.80 4.98 4.98 0.85 0.84 1.49 1.48 0.13 0.13 0.20 0.20 8.14 8.14 8.38 8.91 Mining 1.74 1.74 26.43 26.45 0.49 0.48 0.32 0.32 0.07 0.07 0.09 0.09 8.13 8.16 0.95 0.98 Food 7.94 7.95 16.13 16.15 10.32 10.26 7.26 7.26 0.55 0.55 1.05 1.05 10.99 11.02 16.77 17.44 Textile 11.09 11.02 7.47 7.40 0.75 0.77 6.34 6.76 0.98 0.98 1.93 1.92 3.15 3.28 4.49 4.59 Footwear 2.51 2.51 3.64 3.61 0.17 0.17 5.73 5.89 0.06 0.06 0.60 0.60 0.21 0.21 0.38 0.40 Wood 0.10 0.10 2.13 2.14 0.64 0.60 0.37 0.37 0.01 0.01 0.01 0.01 15.30 15.41 2.92 2.47 Paper 0.44 0.45 3.03 3.04 2.36 2.27 1.87 1.84 0.68 0.69 0.38 0.38 2.60 2.62 1.65 1.36 Petroleum 16.53 16.63 2.23 2.24 1.07 1.07 3.51 3.44 1.68 1.69 4.82 4.83 9.96 9.82 21.76 23.48 Chemicals 12.80 12.78 7.28 7.28 53.15 53.80 11.78 11.66 10.73 10.71 13.46 13.45 7.68 7.60 15.06 15.29 Plastics 1.66 1.66 2.13 2.13 1.43 1.44 3.52 3.49 3.76 3.76 2.08 2.09 2.22 2.21 2.77 2.47 Minerals 1.03 1.03 0.68 0.68 0.24 0.24 2.09 2.08 1.20 1.20 0.57 0.57 2.00 1.98 1.15 1.18 Basic metals 7.96 7.98 5.47 5.49 4.06 3.78 10.44 10.51 9.65 9.69 3.33 3.34 8.25 8.43 1.78 1.11 Machinery, nec 4.17 4.17 2.72 2.72 4.73 4.64 19.71 19.58 17.65 17.65 5.92 5.92 4.12 4.09 3.92 3.24 Electrical 3.99 4.00 7.92 7.93 17.46 17.41 8.03 7.96 24.79 24.83 44.59 44.63 9.99 9.87 7.34 7.73 Transport 8.03 7.98 4.05 4.04 1.27 1.21 13.12 12.90 26.68 26.60 17.43 17.39 4.63 4.47 4.42 3.83 Other 15.20 15.22 3.71 3.72 1.00 1.01 4.43 4.46 1.39 1.39 3.53 3.54 2.64 2.69 6.27 5.53 HH index 0.08 0.08 0.10 0.10 0.31 0.32 0.08 0.08 0.16 0.16 0.23 0.23 0.06 0.06 0.10 0.11
Table 9: Export shares by sector before and after tariff reduction between the US and European Union, in % (continued)
India Indonesia Japan Korea Mexico Russia Turkey RoW
Sector Before After Before After Before After Before After Before After Before After Before After Before After
4.3 Robustness checks
4.3.1 Baseline models with different national trade deficits
The main results in section 4.2 are based on a baseline model where national trade balances are assumed to hold constant at 2014 values. Nevertheless, as in equation (14), different values of trade balances will influence the estimates of trade flows and welfare effects post tariff reduction.
In this section, I check for the robustness of welfare effects by doing the analysis with four cases where trade balances are 80%, 90%, 110% and 120% of 2014 values. That is, in each scenario, the trade balance of each country would be 80%, 90%, 110% or 120% of that country’s balance in 2014. On a global scale, aggregate balance is still zero. To do this robustness check, I first calibrate a baseline model with new values of trade balances but with pre-reduction 2014 tariffs. Then, I use this baseline model to estimate the relative changes of welfare in the equilibrium.
Table 10 reports the welfare effects of these four cases, together with the results on welfare from section 4.2. We see that in the cases where trade balances are 80% or 90% of their values in 2014, many TTIP countries have negative welfare effects. However, as in the case of section 4.2, this welfare reduction is due to the terms of trade effects. Volume of trade effects are still positive for TTIP countries. The welfare gains seem more robust in the cases where trade balances are 110% and 120% of 2014 values, and most TTIP countries see welfare improvement. However, an interesting case is Germany, which expects welfare losses in the cases of 110% and 120%. Since Germany has trade surplus in 2014, these two scenarios mean that the country will run larger trade surpluses. Nevertheless, the estimation shows that the wage and the welfare of Germany are expected to decrease. A possible explanation is that for a country that has large trade surpluses like Germany, an equilibrium where it has an even larger surplus is where the wage rate is reduced to make exports more competitive and boost export volume. The lower wage rate affects terms of trade and real income negatively.
4.3.2 Baselines models with different input-output coefficients
Besides values of trade balances, structures of the economies in the form of input-output coefficients also have a decisive role in determining the equilibrium values of trade flows and welfare effects (see equation 14). Therefore, I provide robustness check with different values for the input-output structure. The main results in section 4.2 use the input-output coefficients of 2011. In this section, I redo the estimation with the input-output coefficients of 2007 and 2009. Besides the difference in input-output data, baseline models are still calibrated with 2014 data of bilateral trade, tariffs, gross outputs and gross value added 9.
Table 11 reports the results. Between 2007, 2009 and 2011, the input-output coefficients change drastically (e.g. many coefficients of same sectors in same countries are more than ten times larger in 2011 than in 2007 or 2009). However, we see that the results for TTIP countries remain robust across different I-O coefficients. The analysis is not very sensitive to the choice of input-output coefficients.
4.3.3 Baseline models differing in the elasticities of trade
In the model, the dispersion of productivity θj, which is related to the elasticity of trade, influ-ences how responsive trade flows are to a change in tariffs. Therefore, it is important to check if different values of θj will drastically change the estimates of trade flows and welfare gains in the counterfactual exercise or not.
In order to see how sensitive the results are to the values of trade elasticities θj, I redo the analysis with the elasticities estimated using the 100% sample in 4.1 (inclusive of outliers, which account for less than 1% of total trade values in each sector). Furthermore, I also re-run the analysis with the trade elasticities from Caliendo and Parro (2015), which differ more starkly from my estimates and should show whether welfare estimates are truly sensitive to the values of θj.
Table 12 show the welfare effects. Overall, the differences across the scenarios are not large. Despite large discrepancies between my estimated elasticities and C&P’s estimates, the results are robust. Countries such as Austria, Italy and the US have the same welfare gains in all three scenarios. The conclusion remains that a tariff reduction improves welfare in EU and the US.
9
Table 11: Total welfare effects given baseline models with input-output coefficients of different years 2011 I-O 2007 I-O 2009 I-O
Table 12: Total welfare effects given baseline models differing in elasticities of trade 99% Sample 100% Sample C&P estimates
4.3.4 Baseline model with 2011 data
In the paper so far, I have been using data of 2014 with the assumption that input-output structure remains the same between 2011 and 2014. The objective is to have the latest data on tariffs. In this sub-section, I check the robustness by redoing the main analysis with gross outputs, value added, bilateral trade and tariff data of 2011. The values of trade elasticities θj are the same as in the 2014 baseline model.
The estimates of welfare effects given 2011 data are reported in table 13. Overall, the 2011 welfare effects are considerably higher than those with 2014 data, most visibly in the case of Lithuania (2.199% vs. 0.453%).
The conclusion does not change that the US and most of EU members have welfare gains. Increases in volume of trade contribute positively to the welfare. In the 2014 scenario, 12 countries in EU expect negative terms of trade effects. However, in this case with 2011 data, only 3 countries have negative terms of trade effects. In fact, terms of trade is the main contributor to welfare gains across the countries.
Another difference is that non-TTIP economies also have high welfare gains, not losses as in the scenario with 2014 data. An economy not in TTIP can also enjoy a boost in trading activities and gains in welfare through its linkages in global supply chains. It can benefit from cheaper imports and higher demand (due to rising income) from TTIP countries. In some cases, reduced wages make exports cheaper and boost trade volume.
Quantitavely, we see that the estimates are quite sensitive to the set of data being used, but the new results strengthen the conclusion of welfare gains due to tariff reduction. Nonetheless, the welfare gains are still not large. Only 4 EU countries have gains larger than 0.1%. As in the main analysis with 2014 data, it is because bilateral tariffs between EU and the US are already low, and the elimination of low tariffs will not result in large changes.
5
Conclusion
Table 13: Welfare effects given 2011 data
of using easily available data (trade flows, tariffs, value added & gross output, and input-output coefficients).
Sectoral linkages exist in the form of composite intermediate goods from all domestic sectors being used in the production of intermediate goods. Cross-border activities occur in the form of globally lowest-priced intermediate goods being the inputs in the assembling of composite interme-diate goods in tradable sectors. Labour is only employed in the production of intermeinterme-diate goods, and hence, it is the production phase that generates value added. The costs of production are influenced by production efficiencies and international trade costs (ice-berg trade costs and tariffs). Qualitatively, a reduction in tariffs would initially lower the prices of imports and create more demand for tradable goods. Cheaper imports go into the production of cheaper composite goods. These cheaper composite goods in tradable sectors are then purchased for the production of interme-diate goods in both tradable and non-tradable sectors. Lower prices boost exports of intermeinterme-diate goods in tradable sectors. Lower prices of intermediate goods also influence the prices of composite goods in non-tradable sectors. Through cross-sectoral and cross-border linkages, both tradable and non-tradable sectors are influenced by the reduction of tariff. Potentially, the economies could expect more trade flows and value added. The estimation of relative changes of welfare in the equilibrium against a baseline model calibrated with 2014 data confirms this notion.
Robustness checking shows that the results are most sensitive to the assumed values of trade balances. The results are also sensitive to the data of bilateral trade, tariffs, gross outputs and value added. They are robust against different input-output coefficients and different values of trade elasticities. All in all, it remains robust that EU and the US can expect welfare gains following a reduction of tariffs.
REFERENCES
1. Allen, T., and Arkolakis, C. (2016). Elements of Advanced International Trade, mimeo, Northwestern University and Yale University.
2. Alvarez, F., and Lucas, R. E. (2007). General equilibrium analysis of the Eaton-Kortum model of international trade. Journal of Monetary Economics, 54: 1726-1768.
3. Anderson, J. E. (1979). A Theoretical Foundation for the Gravity Equation. American Economic Review, 69(1): 106-116.
4. Anderson, J. E., and van Wincoop, E. (2002). Borders, Trade and Welfare. In Collins, S., and Rodrik, D. (eds.), Brookings Trade Policy Forum 2001, 207-230.
5. Anderson, J. E., and van Wincoop, E. (2003). Gravity with Gravitas: A Solution to the Border Puzzle. American Economic Review, 93(1): 170-192.
6. Anderson, J. E., and van Wincoop, E. (2004). Trade Costs. Journal of Economic Literature, 42(3): 691-751.
7. Antr`as, P., Garicano, L., and Rossi-Hansberg, E. (2006). Offshoring in a knowledge economy. Quarterly Journal of Economics, 121(1): 31-77.
8. Arkolakis, C., Costinot, A., and Rodr´ıguez-Clare, A. (2012). New Theories, Same Old Gains? American Economic Review, 102(1): 94-130.
10. Baier, S., and Bergstrand, J. (2007). Do Free Trade Agreements Actually Increase Members Inter- national Trade?. Journal of International Economics, 71(1): 72-95.
11. Baier, S., and Bergstrand, J. (2009). Bonus vetus OLS: A simple method for approximating international trade-cost effects using the gravity equation. Journal of International Eco-nomics, 77(1): 77-85.
12. Brown, D. K., Deardorff, A. V., and Stern, R. M. (1995). Estimates of a North American Free Trade Agreement. In Kehoe, P. J., and Kehoe, T. J. (eds.), Modeling North American Economic Integration, 59-74. Kluwer Academic Publishers.
13. Brown, D. K., and Stern, R. M. (1989). U.S.-Canada Bilateral Tariff Elimination: The Role of Product Differentiation and Market Structure. In Feenstra, R. C. (ed.), Trade policies for International Competitiveness, 217-245. Chicago: University of Chicago Press.
14. Caliendo, L., and Parro, F. (2015). Estimates of the Trade and Welfare Effects of NAFTA. The Review of Economics Studies, 82(1): 1-44.
15. Chaney, T. (2008). Distorted Gravity: The Intensive and Extensive Margins of International Trade. American Economic Review, 98(4): 1707-1721.
16. Costinot, A., Donaldson, D., and Komunjer, I. (2012). What Goods Do Countries Trade? A Quantitative Exploration of Ricardo’s Ideas? Review of Economic Studies, 79(2): 581-608. 17. Costinot, A., and Rodr´ıguez-Clare, A. (2014). Trade Theory with Numbers: Quantifying
the Consequences of Globalization. In Gopinath, G., Helpman, E., and Rogoff, K., (eds.), Handbook of International Economics, vol. 4, chapter 4, 197-261. Amsterdam: Elsevier. 18. Costinot, A., Vogel, J., and Wang, S. (2013). An Elementary Theory of Global Supply Chains.
Review of Economic Studies, 80: 109-144.
20. Eaton, J, and Kortum, S. (2002). Technology, Geography, and Trade. Econometrica, 70(5): 1741-1779.
21. Eaton, J., and Kortum, S. (2012). Putting Ricardo to Work. Journal of Economic Perspec-tives, 26(2): 65-90.
22. Feenstra, R. C. (1994). New Product Varieties and the Measurement of International Prices. American Economic Review, 84(1): 137-177.
23. Feenstra, R. C., and Hanson, G (1996). Globalization, Outsourcing and Wage Inequality. American Economic Review, 86(2): 240-245.
24. Felbermayr, G., Heid, B., Larch, M., and Yalcin, E. (2015). Macroeconomic potentials of transatlantic free trade: a high resolution perspective for Europe and the world. Economic Policy, 30(83): 491-537.
25. Fox, A. K. (1999). Evaluating the Success of a CGE Model of the Canada-U.S. Free Trade Agreement. Unpublished working paper, Michigan University.
26. Francois, J., Manchin, M., Norberg, H., Pindyuk, O., and Tomberger, P. (2013). Reduc-ing Transatlantic Barriers to Trade and Investment: An Economic Assessment. Report TRADE10/A2/A16 for the European Commission.
27. Grossman, G. M., and Rossi-Hamsberg, E. (2008). Trading Tasks: A Simple Theory of Offshoring. American Economic Review, 98(5): 1978-1997.
28. Head, K., and Mayer, T. (2014). Gravity equations: Workhorse, Toolkit, and Cookbook. In Gopinath, G., Helpman, E., and Rogoff, K., (eds.), Handbook of International Economics, volume 4, chapter 3, 131-195. Elsevier.
29. Hummels, D., Ishii, J., and Yi, K-M (2001). The Nature and Growth of Vertical Specialization in World Trade. Journal of International Economics, 54(1): 75-96.
