Economic Sanction Games among US, EU and Russia: Solutions and Potential Effects *

Transcription

1 Workng Paper No October 18, 2016 东艳 李春顶 Economc Sancton Games among US, EU and Russa: Solutons and Potental Effects * Abstract Economc sancton of the US and EU to Russa because of Ukrane crss n 2014 s a hot topc. Ths paper uses a 16-country numercal general equlbrum model wth trade cost and exogenous trade mbalance to explore ths three-country economc sancton game solutons, and smulate the effects of sanctons on ndvdual countres. Our analyss fnd that all sancton nvolved countres wll be hurt, but comparatvely Russa wll be hurt more, and the US and EU wll be hurt less. Sancton measures of EU have larger mpacts to Russa than the US measures, and meanwhle Russan counter-sancton measures wll generate larger mpacts on the EU than on the US. From the economc perspectve, the optmal choce for US and EU s to gve up sancton measures to Russa, and retalaton s Russa s optmal choce when faced wth sancton measures. Countres out of the sancton game wll gan because of trade dverson effects. Keywords: Optmal Tarffs, General Equlbrum, odel Structure, Trade Lberalzaton JEL Classfcaton: F11, C63, F13 * Yan Dong, Insttute of World Economcs and Poltcs, Chnese Academy of Socal Scences; Chundng L, Insttute of World Economcs and Poltcs, Chnese Academy of Socal Scences. 1 / 28

2 1. Introducton The Russan mltary nterventon n Ukrane, whch began n late February 2014, prompted a number of governments to apply sanctons aganst ndvduals, busnesses and offcals from Russa. These sanctons were manly from the Unted States (US) and the European Unon (EU). Russa has responded wth counter sanctons aganst them, ncludng a total ban on food mports from the EU, the US, Norway, Canada and Australa (Wkpeda, 2015). In ths paper, our nterests on the economc sanctons are focusng on how these dfferent sancton measures nfluence nvolved countres and non-nvolved countres, and what are the solutons to the sancton game. Exstng lteratures on economc sanctons among US, EU and Russa are manly analytcal, few of them has ever used numercal methods to explore the sancton game solutons and ts nfluence to ndvdual countres. Galbert (2015) assesses the outcome and future of Russa sanctons from a European perspectve. Dreyer and Popescu (2014) analyzes the effects and possble mpacts of sanctons aganst Russan. Oxensterna and Olsson (2015) comprehensvely studes the mpacts and prospects of the economc sanctons aganst Russa wth analytcal methodology. ICC (2015) studes the potental mpact of the EU sanctons aganst Russa on nternatonal arbtraton admnstered by EU-based nsttutons. Nelson (2015) generally analyzes economc mplcatons of the US sanctons on Russa. Some lteratures analyze economc sancton from theoretcal perspectve. Kaempfer and Lowenberg (1988) uses a publc choce approach to study the theory of nternatonal economc sancton. Ths paper uses numercal general equlbrum modellng and smulaton methodology to compute sancton game solutons and then to explore the nfluences of economc sanctons. The methodology s new n sancton lteratures and the results are mportant for polcy. The US and EU sanctons aganst Russa and Russa s counter sanctons are cycle, ncremental, step-by step, and from soft sanctons to hard sanctons. From arch 2014 to now, there are about three rounds of sanctons and counter sanctons among the US, the EU and Russa. The frst round s the threat stage, the US and the EU use restrctve measures, manly asset freezes and vsa bans on selected ndvduals, to send a strong message to the Russan government that there are consequences for ther actons that volate the soveregnty and terrtoral ntegrty of Ukrane s prepared to take addtonal steps to mpose further poltcal and economc costs. The second round s that the US and EU use concrete measures to ncrease Russa s poltcal solaton as well as the economc costs to Russa, especally n areas of mportance to Russa economy, the measures nclude mport bans on Russa s energy and defense sectors, embargo on the mport and export of arms, exports bans on certan energy-related equpment and technology to Russa and fnancal sanctons. The thrd round began from July 2014 to now. In response to the escalatng War n Donbass, on 17 July 2014 the US extended ts transactons ban to two major Russan energy frms and two banks. After that a seres of EU countres take more tghten sanctons to Russa. Even though n June 2015 the G7 collectvely extended sanctons already n place for an addtonal sx months (Wkpeda, 2015). Russa s counter actons nclude travel bans, mport ban on food from western countres, and mport bans on used cars, clothes and consumer products n the frst two stages. The economc sancton and counter sancton game among the US, the EU and Russa s n developng and wll last a long tme. So t s valuable to explore the mpacts of ths sancton game numercally. We set up a three-round sequental game to analogue ths sancton process and explore ts possble nfluence. 2 / 28

3 We compute payoffs for all game tree ponts, and explore the game equlbrums and smulate the sancton game mpacts. We assume the three-round game are separately soft sancton round, hard sancton round and forbdden sancton round. Each round sancton has seven choce of game ponts. Sancton and counter-sancton measures nclude tarff, non-tarff, and captal flow control. In ths paper, we construct a 16-country global general equlbrum model ncludng fxed trade mbalance, controlled captal flow and trade cost. Controlled captal flow assumpton can help to explore the captal control effects, and trade cost structure s helpful to explore the effects of tarff and non-tarff barrers. Some sancton measures are hard to modellng n the general equlbrum structure, so our analyss focus on trade sancton measures. We use the numercal general equlbrum model calbraton and smulaton methodology to compute payoffs for game tree ponts, and then to explore the sancton game nfluences to ndvdual countres. Our numercal model has 16 countres, whch are the US, the EU, Russa, Chna, Japan, Korea, Canada, exco, Inda, AN (Australa and New Zealand), CP (Chle and Peru), BSV (Brune, alaysa, Sngapore, and Vetnam ), CILPT (Camboda, Indonesa, Laos, yanmar, Phlppne, and Thaland), ODDC (other developed countres, ncludng Swtzerland, Norway, Israel, and Iceland), ODC (other developng countres, ncludng Brazl, Egypt, Argentna, and South Afrca), and ROW (the rest of the world). Each country produces two goods whch are tradable manufacturng goods and non-tradable non-manufacturng goods wth two factors (labor and captal). Our numercal analyss fnd that sancton and counter-sancton measures among the US, the EU and Russa wll defntely hurt all sancton nvolved countres, but beneft all non-nvolved countres. For the US and EU, ther optmal choces from the economc perspectve are gvng up sancton measures. The optmal choce for Russa s retalatng when faced wth sancton measures. The mpacts of sanctons to dfferent countres are dfferent. Negatve mpacts to the US and EU are comparatvely small compared wth ther economc scale, so they are not afrad of Russa s counter sancton threat. But negatve mpacts of sancton to Russa s large compared wth her economc scale, whch means that Russa wll be heavly hurt by economc sancton from the US and EU. The EU sancton measures wll generate more effects to Russa than the US measures, meanwhle Russa s counter sancton measures wll hurt the EU more than the US. Addtonally, soft sanctons have less nfluence than hard sanctons, and hard sanctons have less mpact than forbdden sanctons. The negatve nfluences to nvolved countres under optmal sancton are less than under arbtrary sancton. 2. A Game of Sequental Economc Sanctons We assume and buld the sancton game accordng to the economc sancton facts among the US, the EU and Russa. In ths three-country economc sancton game, the senders are the US and EU, and the responder s Russa. The US and EU punsh Russa, and Russa choose to retalate. We set up a three-round sancton-counter game, the sanctons are ncrementally ntensfed. The frst round s the soft sancton game, the second round s the hard sancton game, and the thrd round s the forbdden sancton game. Each round of the games nvolves three players, the US, the EU and Russa. We assume that the US and EU are sancton ntaton countres and they decde whether or not to take sancton measures to Russa, and Russa s the counter country and she decdes whether or not to take counter retalaton measures. We also assume that Russa wll take symmetrc counter actons, whch means Russa wll retalate the country who take sancton 3 / 28

4 measures to her at the same sancton level but wll not retalate the countres wthout sancton measures to her. We further assume that ths three-round sancton-counter game s a sequental game. Only at the stuaton that the US and EU take sancton measures smultaneously to Russa and meanwhle Russa retalate, then the three players enter the second round sancton game, and the same assumpton to the thrd round sancton game. In each round of the games, each player only has two acton choces. For the US and EU, ther two actons are Sancton (we denote t as S) and Non-Sancton (we denote t as NS). Russa s two acton choces are Counter-Sancton (we denote t as CS) and Non-Counter-Sancton (we denote t as NCS). Specfcally, n the frst round of the game, we assume a three-step acton process. The frst step s for the US to decde whether to sancton or not, the second step s for the EU to decde whether to sancton or not, the thrd stage s for Russa to decde whether to counter-sancton or not. Therefore, n each round of the game, there are 8 dfferent decson choces (see Fgure 1). We defne them as follows: The Frst Round Game: Soft Sancton O11 = (US: S; EU: S; Russa: CS); O12 = (US: S; EU: S; Russa: NCS); O13 = (US: S; EU: NS; Russa: CS); O14 = (US: S; EU: NS; Russa: NCS); O15 = (US: NS; EU: S; Russa: CS); O16 = (US: NS; EU: S; Russa: NCS); O17 = (US: NS; EU: NS; Russa: NCS); The Second Round Game: Hard Sancton O21 = (US: S; EU: S; Russa: CS); O22 = (US: S; EU: S; Russa: NCS); O23 = (US: S; EU: NS; Russa: CS); O24 = (US: S; EU: NS; Russa: NCS); O25 = (US: NS; EU: S; Russa: CS); O26 = (US: NS; EU: S; Russa: NCS); O27 = (US: NS; EU: NS; Russan: NCS); The Thrd Round Game: Forbdden Sancton O31 = (US: S; EU: S; Russa: CS); O32 = (US: S; EU: S; Russa: NCS); O33 = (US: S; EU: NS; Russa: CS); O34 = (US: S; EU: NS; Russa: NCS); O35 = (US: NS; EU: S; Russa: CS); O36 = (US: NS; EU: S; Russa: NCS); O37 = (US: NS; EU: NS; Russa: NCS); 4 / 28

5 Fgure 1: The US, the EU and Russa Sancton-Counter Game US S NS EU EU S NS S NS Russa Russa Russa Russa CS NCS CS NCS CS NCS NCS O11 O12 O13 O14 O15 O16 O17 Frst Round Game: Soft Sancton US S NS EU EU S NS S NS Russa Russa Russa Russa CS NCS CS NCS CS NCS NCS O21 O22 O23 O24 O25 O26 O27 Second Round Game: Hard Sancton Thrd Round Game: Forbdden Sancton O31 O32 O33 O34 O35 O36 O37 Source: by authors. 3. A GE odel wth Economc Sancton 3.1 Basc Set-up of the odel We use a global general equlbrum model wth an exogenous trade mbalance and trade barrers based on trade costs to compute game solutons and effects of economc sanctons. 5 / 28

6 Let N 1,2, n L be the set of n countresn 3, each of whch s populated by dentcal consumers who consume a manufacturng goods and a non-manufacturng goods. We assume manufacturng goods s tradable and non-manufacturng goods s non-tradable. The two factors n each country are labor and captal, whch are moble between but mmoble among countres. Producton Output of each goods l produced by country s gven by l l l 1 1 l l l l l l l l l 1 Q [ ( L ) (1 )( K ) ], country, l goods (1) where l L and l K denote labor and captal respectvely, used n the producton of goods l. Frst order condtons mply the factor nput demand equatons. Consumpton On the consumpton sde, we use the Armngton assumpton of product heterogenety across countres, and use a nested CES utlty functon for each country (, N ) [ 1 ( ) 2 ( N ) ] U X X X X, country (2) where N X denotes the consumpton of non-manufacturng goods n country, X denotes the consumpton of composte Armngton manufacturng goods n country. Addtonally 1 and 2 are share parameters and s the top level elastcty of substtuton n consumpton. The composte manufacturng goods s defned as another reflectng the country from whch goods come. We assume ths level 2 composte consumpton s also of CES form and represented as 1 ' 1 ' ' ' ' 1 j j j X [ x ], j country (3) where x j s the consumpton of manufacturng goods from country j n country. If j ths mples that ths country consumes ts domestcally produced tradable goods. j s the share parameter for country ' js ' manufacturng goods consumed n country. s the elastcty of substtuton n level 2 preferences n 6 / 28

7 country. The utlty optmzaton problem above yelds X E ( ) [ ( ) ( ) ] 1 1 N 1 P 1 P 2 pc (4) X E ( ) [ ( ) ( ) ] N 2 N 1 N 1 pc 1 P 2 pc (5) where P and N pc are the separate consumpton prces of composte manufacturng goods and non-manufacturng goods n country. E s the total consumpton expendture of country. For the composte manufacturng goods whch enter the second level preferences and come from dfferent countres, the country specfc demands are x ( X P ) j j ' pcj j pcj j ' (1 ) ( ) [ ( ) ] (6) where pc s the consumpton prce n country of manufacturng goods produced n country j, X P j s the total expendture on manufacturng goods n country. Consumpton prce of the composte manufacturng goods s N j pcj j1 P 1 ' ' (1 ) 1 [ ( ) ] (7) and the total consumpton expendture of country s E P X pc X (8) N N N Trade Cost We add trade cost nto our model, whch can be dvded nto mport tarff and non-tarff barrer. We denote the mport tarff n country as t, and non-tarff trade barrer as barrer for country N j (ad volume tarff-equvalent non-tarff mported from country j ). Ths yelds the followng relatonshp between consumpton prce and producton prce n country for country ' js export, 7 / 28

8 pc (1 t N ) p (9) j j j Where p j s the producton prce of manufacturng goods n country j. We assume trade costs are covered by mportng country. Import tarffs generate revenue R, whch are gven by R p x t (10) j j j, j Non-tarff barrers generate no revenue, and mporters need to use real resources to cover the costs nvolved. In the model, we assume that these resource costs are denomnated n terms of domestc non-manufacturng goods. We ncorporate ths resource usng feature through use of non-manufacturng goods equal n value terms to the cost of barrers. We assume reduced non-tarff barrer (ncludng transportaton cost) wll thus occur under trade lberalzaton as an ncrease n non-manufacturng goods consumpton NR by the representatve consumer n mportng countres, where NR p x N (11) j j j j, j arket Clearng Condtons that Equlbrum n the model characterzed by market clearng prces for goods and factors n each country such Q x j j (12) Q N NR N X (13) p N l K K, L L l l l (14) For the trade clearance, we assume an exogenously determned fxed trade mbalance, denoted as S, whch wll be postve when n trade surplus and negatve when n trade defct. Trade equlbrum wll nfluence ndvdual country s budget constrant. In the equlbrum, we have I E S (15) whch means that one country s total ncome equals ts total consumpton expendture plus ts surplus (trade mbalance), f one country has trade surplus then ts ncome wll more than consumpton expendture, but f one 8 / 28

9 country has trade defct then ts ncome wll be less than consumpton expendture. In the equlbrum, the representatve consumer s ncome n country s gven by w K w L R I (16) K L A zero proft condton must also be satsfed n each ndustry n each country, such that l l K l L l p Q w K w L l, N (17) 3.2 Economc Sanctons Accordng to the real economc sanctons among the US, the EU and Russa, two man knds of measures are taken by these three countres. One s the tarff and non-tarff barrer sancton, and the other s fnancal sanctons lke controllng captal flow. In order to capture the effects of both trade barrer sancton and fnancal sancton, our mult-country global general equlbrum model has ncluded relevant structure to explore the both sancton measures. Frstly, we have ncluded trade cost structure nto the model whch can be dvded nto tarff and non-tarff barrer, and so that can explore the tarff and non-tarff barrer sancton effects. Secondly, we have ncluded a fxed trade mbalance assumpton n the model, and the model does not permt freely captal flow so as to explore the fnancal sancton. Under ths assumpton, captal flow s forbdden n the model whch ft for the captal flow control sancton between the US, the EU and Russa. In our mult-country global general equlbrum model, the assumpton of captal control have been ncluded. So we focus on the trade related economc sanctons ncludng both tarff and non-tarff. In our model, they are separately t and N j. In the equlbrum, the total utlty of representatve consumer n country equals to [ 1( ) N 2( ) ] U I P pc (18) We take the equaton of I and P nto the utlty equaton and get 1 ' 1 ' ' ' ' 1 ( ) ' N j j xj K L N ' ' j (1 ) 1 j ' j j j, j ( pcj p j ) j1 U { w K w L [ p [ ( pc ) ] t } 1 N 1 ' ' (1 ) 1 1 N j pcj 2 pc j1 [ ([ ( ) ] ) ( ) ] (19) 9 / 28

10 As pc (1 t N ) p, so tarff and non-tarff barrer wll undoubtedly nfluence the welfare of j j j country. Ths means that one country can use tarff and non-tarff barrer as economc sancton weapons to force other countres to concede n negotaton. 4. Numercal odel Data and Parameters Calbraton We buld a 16-coountry numercal general equlbrum model usng above model structure. These sxteen countres are the US, the EU, Russa, Chna, Japan, Korea, Canada, exco, Inda, AN (Australa and New Zealand), CP (Chle and Peru), BSV (Brune, alaysa, Sngapore, and Vetnam ), CILPT (Camboda, Indonesa, Laos, yanmar, Phlppne, and Thaland), ODDC (other developed countres, ncludng Swtzerland, Norway, Israel, and Iceland), ODC (other developng countres, ncludng Brazl, Egypt, Argentna, and South Afrca), and ROW (the rest of the world). 4.1 Data and Calbraton We use 2011 as our base year n buldng a benchmark general equlbrum dataset for use n calbraton and smulaton followng the methods set out n Shoven and Whalley (1992). We use world values mnus all ndvdual countres to generate ROW values. For the two goods, we assume secondary ndustry (manufacturng) reflects manufacturng goods, and prmary and tertary ndustres (agrculture, extractve ndustres, and servces) yeld non-manufacturng goods. For the two factor nputs, captal and labor, we use total labor ncome (wage) to denote labor values for nputs by sectors. All data are n bllon US dollars. We adjust some of the data values for mutual consstency for calbraton purposes. Table 1: Base Year Data Used For Calbraton and Smulaton (Unt: Bllon US$) Country GDP T-G NT-G Balance Captal Labor T-G NT-G T-G NT-G USA EU Russan Chna Japan Korea Canada exco Inda AN CP BSV CILPT ODDC / 28

11 ODC ROW Note: (1) T-G denotes tradable goods producton; NT-G denotes non-tradable goods producton. (2) We add countres together to generate AN, CP, BSV values. (3) We use world values mnus all ndvdual countres to generate ROW values. Sources: EU data from EU statstcs, and the currency unt s Euro, we use annual average exchange rate to change them nto US dollar currency unt; Other countres data are all calculated from WDI of World Bank database. EU data s from EU statstcs, and the currency unt s Euro, we use annual average exchange rate to change them nto US dollar; Other countres data are all calculated from WDI of World Bank database. We use agrculture and servce share of GDP data and GDP data to yeld producton data of manufacturng goods and non-manufacturng goods, and use captal/gdp rato to yeld captal and labor nput n producton. Trade data between each par of countres are from the UN Comtrade database. We use ndvdual country total export and mport values to ndrectly yeld exports to and mports from the ROW. Usng producton and trade data, we can then calculate each country s consumpton values. We dvde trade costs nto two parts, mport tarffs and non-tarff barrers. We obtan each country s mport tarff data from WTO Statstcs Database. For ROW, we use EU s tarff rate to denote these values. We calculate non-tarff barrers by usng trade costs mnus mport tarffs. Table 2: Import Tarffs for Countres (Unt: %) Country USA EU Russa Chna Japan Korea Canada exco Tarff Country Inda AN CP BSV CILPT ODDC ODC ROW Tarff Notes: (1) Import tarffs here are smple average FN appled tarff rates. (2) We use the average ndvdual country s mport tarff to get country groups mport tarff. Source: WTO Statstcs Database. There are no avalable estmates of elastctes for ndvdual countres on the demand and producton sdes of the model. We set all these elastctes n our model to 2 followng Whalley and Wang (2010). We change these elastctes later n senstvty analyss to check ther nfluence on smulaton results. Wth these data, we calbrate the model parameters. When used n model soluton these wll regenerate the benchmark data as an equlbrum for the model. Then, usng these parameters we can smulate the game equlbrum and explore effects of sancton games among the US, the EU and Russa. 11 / 28

12 Exporter Countres Table 3:Trade between Countres (Unt: Bllon USD) Importer USA EU Russan Chna Japan Korea Canada exco Inda AN CP BSV CILPT ODDC ODC ROW USA EU Russan Chna Japan Korea Canada exco Inda AN CP BSV CILPT ODDC ODC ROW Notes: (1) We get trade data of AN, CP, and BSV by addng separate country s trade data together, and these do not nclude nner trade between these group countres. (2)We get the ROW trade data by deductng from each country s total export, total mport and total world trade value. Sources: Unted Natons (UN) Comtrade database and WTO Statstcs. 12 / 28

13 4.2 Trade Cost Calculaton We report our calculatons of trade cost n ths part whch provdes trade cost estmates for use n our general equlbrum model. The methodology we use s from Novy (2013) and Wong (2012). We calculate and report ad valorem tarff-equvalent trade costs between countres. The measure of trade barrers used here s based on the gravty equaton derved from Chaney s (2008) model of heterogeneous frms wth blateral fxed costs of exportng. Trade barrers can take two forms n the model, a varable trade barrer r and a fxed cost of exportng F r. The varable trade barrer r s an ceberg cost. In order to delver one unt of good to from r, 1 unt of good has to be delvered. The gravty equaton supported by ths model s: r X r Y ( 1) Yr wr r 1 ( ) Fr (20) Y Where X r s mport of country from country r. Y, Y r and Y are the economc szes of both countres and the total world, w r s labor costs, r s varable trade costs and F r s the fxed cost of exportng. The Pareto parameter governs the dstrbuton of frm productvtes. s the elastcty of substtuton n preferences. s a remoteness measure for the mportng country whch captures trade dverson effects. The mechansm s that the further away s from the rest of the world, the more lkely that r could export more to due to less competton from thrd party countres n the mporter country. Ths has a smlar nterpretaton to the multlateral resstance term n Anderson and Wncoop (2003). We can relate data on trade flows to unobservable trade barrers by takng ratos of blateral trade flows of two regons over local purchases of each of two countres: X X X X F F ( 1) 1 ( ) ( ) (21) F F r r r r r r rr rr rr Ths equaton reveals the relatonshp between observable trade data and unobservable trade barrers and elmnates the need to worry about the omsson of unspecfed or unobserved trade barrers. If the fxed costs of exportng are not blaterally dfferentated ( Fr Fr ) or s they are constant across locatons ( Fr F ), the fxed costs drop out of ths measure and the measured trade costs would smply be nterpreted as varable trade costs, as n models wthout fxed export costs such as Eaton and Kortum (2002) and Anderson and Wncoop (2003). For smplcty of exposton, we normalze domestc trade costs to 1,.e. 1 and F 1. Defnng the 13 / 28

14 geometrc average of trade costs between the country par and r as X ( ) X rr r ( ) ( r r ) ( Fr Fr ) Xr Xr t (22) Data for ths equaton s relatvely easy to obtan, and so we have a comprehensve measure of trade barrers, and the ad valorem tarff-equvalent blateral average trade cost between country and r can be wrtten as t r 1 2 X Xrr tr 1 ( ) 1 (23) X X r r Usng the trade costs equaton above, we can calculate actual trade costs between countres n our general equlbrum model, whch are needed n buldng a benchmark data set for use n calbraton and smulaton. We need to calculate trade costs between each country par for Chna. For trade costs, n equaton (23), X r and X r are separately exports and mports between countres and r. Ths trade data s from the UN Comtrade database, and total world trade data s from WTO Internatonal Trade Statstcs. Due to market clearng, ntranatonal trade X or X rr can be rewrtten as total ncome mnus total exports X y X (24) where X s the total exports, defned as the sum of all exports from country, whch s X X (25) r, r r Ths data s from the UN Comtrade database. For y, GDP data are not sutable because they are based on value added, whereas the trade data are reported as gross shpments. In addton, GDP data nclude servces that are not covered by the trade data (Novy, 2013). It s hard to get ths ncome data accordng to such a defnton, so here we use GDP data mnus total servce value added. We get GDP data from World Bank database, and the servce share of GDP data from World Development Indcators (WDI) of World Bank database, we then calculate results for GDP mnus servces. We take the value of to be 8.3 as n Eaton and Kortum (2002). We only use trade cost data for 2011 n our numercal general equlbrum model, calculaton results are shown n Table / 28

15 Table 4: Ad Valorem Tarff-Equvalent Trade Costs Between Countres (Unt: %) Country US EU Russan Chna Japan Korea Canada exco Inda AN CP BSV CILPT ODDC ODC ROW US EU Russan Chna Japan Korea Canada exco Inda AN CP BSV CILPT ODDC ODC ROW Notes: We see group countres as a whole to calculate trade costs. Source: Calculated by authors. 15 / 28

16 5. Numercal Analyss of Sancton Game Solutons and Effects We solve the three-stage sancton game among the US, the EU and Russa numercally n ths part. We focus our analyss on two aspects. One s game solutons, we compute payoff matrx for each game tree pont. The other s the effects of the three-stage economc sanctons. For the game solutons, we smulate payoff matrxes for all game ponts. We use both equvalent varaton (EV) and compensaton varaton (CV) to denote payoffs of sanctons. Equvalent varaton refers to the amount of addtonal money an agent would need to reach ts ntal utlty after a change n prces, or a change n product qualty, or the ntroducton of new products. Compensaton varaton can be used to fnd the effect of a prce change on an agent s net welfare. CV reflects new prces and the old utlty level. EV s a closely related measure that uses old prces and the new utlty level. It measures the amount of money a consumer would pay to avod a prce change, before t happens. They have the followng equatons EV e( p, v( p, m )) e( p, v( p, m )) e( p, v( p, m )) m CV e p v p m e p v p m m e p v p m (, (, )) (, (, )) (, (, )) (26) where 0 denotes former stuaton, 1 denotes the stuaton after change. We use the benchmark scenaro as former stuaton and sancton scenaro as the stuaton after change. Then EV and CV can show the welfare changes affected by economc sanctons. For the effects of economc sanctons, we pay attenton to the nfluence on utlty, export and mport. We separately explore the effects of soft sancton, hard sancton and forbdden sancton to man countres. As to the sequental three stage sanctons. We defne soft sancton as a 20% trade barrer rase and controlled captal flow, hard sancton as a 50% trade barrer rase and controlled captal flow, and forbdden sancton as a 900% trade barrer ncrease and controlled captal flow. The forbdden sancton s assumed to be no trade, we use a very hgh trade barrer of 900% to denote ths stuaton for smplcty. 5.1 Three-Stage Game Solutons wth EV and CV We have assumed a three-stage sequental game among the US, the EU and Russa. For each stage game, there are seven game tree ponts whch we have already denoted them n the prevous part. We smulate and compute payoffs for each pont n the game tree. EV s used to be the man ndcator of payoffs, and we use CV to compare the results for robustness. Payoff matrx for all game ponts are lsted n Table 5. We fnd that mutual economc sanctons wll hurt all nvolved countres, comparatvely the negatve mpacts of forbdden sanctons are larger than hard sanctons, and the negatve effects of hard sanctons are larger than soft sanctons. On the country sde, f the US and EU both take sancton measures to Russa, and Russa retalate to both the US and EU, all three nvolved countres wll hurt. Specfcally, Russa wll be hurt the most, then wll be the EU, and the US wll receve the least mpact. eanwhle, the effects of the EU s sancton measures to Russa s sgnfcant and severe, comparatvely the negatve mpacts of the US sancton to Russa s weak. The effects of Russa s retalaton to the EU are much more 16 / 28

17 sgnfcant than to the US. Therefore, economc sancton and retalaton measures wll hurt all nvolved countres. The optmal game tree pont for the US s that the EU take sancton measures to Russa and Russa retalate to the EU but the US does not take sancton measures. The optmal game tree pont for the EU s that the US take sancton measures and Russa retalate to the US but the EU does not take sancton measures. The optmal game tree pont for Russa s that all three country groups take free trade, but f the US or EU or both took sancton measures, Russa s optmal choce s retalaton. Table 5: Game Solutons wth EV (Unt: Bllon US$) Country O11 O12 O13 O14 O15 O16 O17 US EU Russan Country O21 O22 O23 O24 O25 O26 O27 US EU Russan Country O31 O32 O33 O34 O35 O36 O37 US EU Russan Source: compled by authors. We use CV to recalculate payoffs for the game and to check the robustness of game solutons wth EV. All results are reported n Table 6. We fnd that all results are close to the results of EV. So above results are relable. Table 6: Game Solutons wth CV (Unt: Bllon US$) Country O11 O12 O13 O14 O15 O16 O17 US EU Russan Country O21 O22 O23 O24 O25 O26 O27 US EU Russan Country O31 O32 O33 O34 O35 O36 O37 US EU Russan Source: compled by authors. In general, economc sanctons among the US, the EU and Russa wll hurt all nvolved countres. The US s sancton threats to Russa are weaker than the EU s, and the Russa s retalaton to the EU has much larger 17 / 28

18 negatve nfluence than the retalaton to the US. When Russa receves economc sancton measures, her optmal choce s retalaton. 5.2 Senstvty Analyss to Elastctes for Game Solutons We perform senstvty analyss for game payoffs to elastctes n ths part. In our man numercal smulaton model, we choose elastctes to equal 2. Here we assume elastctes to separately equal 1.5 and 2.5 to compare game soluton results. For smplcty, we just perform senstvty analyss to the second round game (hard sancton). Results are shown n Table 7. Senstvty analyss results reveal that as the elastctes of substtuton ncrease, negatve effects of economc sancton and retalaton wll decrease for the US and Russa, and wll ncrease for the EU. The nfluence drecton of the mpacts are the same. 18 / 28

19 Table 7: Senstvty Analyss of Hard Sancton Game Solutons to Elastctes (Unt: % Change) Countres O21 O22 O23 O24 O25 O26 O27 O21 O22 O23 O24 O25 O26 O27 EV, E=1.5 CV, E=1.5 US EU Russan EV, E=2.0 CV, E=2.0 US EU Russan EV, E=2.5 CV, E=2.5 US EU Russan Source: Compled by authors. 19 / 28

20 5.3 Three-Stage Game Solutons wth EV and CV as a Share of GDP EV and CV show an absolute value change of welfare but cannot reveal the comparatve burden of the welfare change. We use EV and CV as a share of GDP to show the comparatve nfluences of sancton games to nvolved countres. These ndcators can reveal whether the nfluence s huge to the country or not. We set EV as a share of GDP to ndcate our man results, and use CV as a share of GDP as a robustness check ndcator. Table 8 reports payoffs for game solutons. We fnd that mpacts to both the US and EU are small compared wth ther economc scale, especally for the nfluence to the US. Ths result can explan why the US and EU are not afrad of retalatons and negatve mpacts. But the nfluences to Russa are sgnfcantly large and wll hurt Russa heavly. The nfluences of forbdden sanctons are larger than hard sanctons, and the mpacts of hard sanctons are larger than soft sanctons. For ndvdual countres, the optmal choce for both the US and EU are free trade and have no sancton measures to Russa, the optmal choce for Russa s retalaton when the US and EU take sancton measures. Table 8: Game Solutons wth EV as a Share of GDP (Unt: %) Country O11 O12 O13 O14 O15 O16 O17 US EU Russan Country O21 O22 O23 O24 O25 O26 O27 US EU Russan Country O31 O32 O33 O34 O35 O36 O37 US EU Russan Source: compled by authors. We use CV as a share of GDP to check the robustness of our results (see Table 9). We fnd results are close. Negatve mpacts of sancton and retalaton to the US and EU are comparatvely weak, but are strong to Russa. Table 9: Game Solutons wth CV as a Share of GDP (Unt: %) Country O11 O12 O13 O14 O15 O16 O17 US EU Russan Country O21 O22 O23 O24 O25 O26 O27 US EU Russan / 28

21 Country O31 O32 O33 O34 O35 O36 O37 US EU Russan Source: compled by authors. In general, the economc sancton losses of the US and EU are small compared wth ther bg economc scale, but the loss of Russa are large compared wth her own economc scale. Therefore, the economc sancton and retalaton wll hurt Russa much more than the US and EU. But faced wth sancton, the optmal choce for Russa s retalaton. 5.4 Effects of Soft Sanctons among US, EU and Russa We explore the mpacts of economc sanctons and retalatons among the US, the EU and Russa from ths part, and we analyze three dfferent stages of sanctons (soft sancton, hard sancton, and forbdden sancton) one by one. Three aspects of nfluences are chosen whch are utlty, export and mport. We explore the nfluences on both sancton nvolved countres and some other large countres. Ths part explores the effects of soft sanctons. Table 10 reports the smulaton results. We fnd that all three sancton nvolved countres hurt, but almost all other countres gan from the sancton and retalaton measures. For the sancton nvolved countres, Russa wll be hurt the most. The mpacts to the US and EU are small and comparatvely the EU wll be hurt more than the US. For other man countres out of the sancton game, most of them wll gan. On the utlty sde, Chna and Korea wll comparatvely gan more than other countres. On the export and mport sde, Chna, Canada and exco can gan comparatvely more than other countres. We take the game pont of both the US and EU take sancton measures and Russa retalate as an example to specfcally compare dfferent effects. On the utlty sde, the US wll lose %, the EU wll lose -0.08%, the Russa wll lose %; but Chna wll gan 0.012%, Japan wll gan 0.005%, Korea wll gan 0.022%, Canada wll gan 0.003%, exco wll gan 0.007%, and Inda wll gan 0.007%. Table 10: Effects of Soft Sanctons to ajor Countres (Unt: % change) Countres O11 O12 O13 O14 O15 O16 O17 Utlty US EU Russan Chna Japan Korea Canada exco Inda Export US / 28

22 EU Russan Chna Japan Korea Canada exco Inda Import US EU Russan Chna Japan Korea Canada exco Inda Source: compled by authors. 5.5 Effects of Hard Sanctons among US, EU and Russa We then explore the effects of hard sanctons. Table 11 reports all results. It s obvous that hard sancton measures wll hurt nvolved countres more than soft sancton measures, and wll beneft non-nvolved countres more than soft sancton measures. Impacts on ndvdual countres of hard sanctons are the same as the nfluences of soft sanctons. All sancton nvolved countres wll be hurt, comparatvely Russa wll receve heaver mpacts than both the US and EU, and the EU wll receve more nfluence to the US. The economc sancton among the US, the EU and Russa wll heavly hurt Russa but just slghtly hurt the US and EU. All non-nvolved countres can gan from ths sancton game. Chna and Japan can gan more on utlty. exco, Canada and Chna can gan more on both export and mport. We take the export mpacts under the stuaton of US and EU both take sancton measures and Russa retalate as an example to specfcally compare the nfluence. The exports of sancton nvolved countres of the US, the EU and Russa are separately %, % and %. Chna s export ncreases 0.054%, Japan s export ncreases 0.005%, Korea s export decreases %, Canada s export ncreases about 0.081%, exco s export ncreases 0.101% and Inda s export ncreases 0.051%. Table 11: Effects of Hard Sanctons to ajor Countres (Unt: % change) Countres O11 O12 O13 O14 O15 O16 O17 Utlty US EU / 28

23 Russan Chna Japan Korea Canada exco Inda Export US EU Russan Chna Japan Korea Canada exco Inda Import US EU Russan Chna Japan Korea Canada exco Inda Source: compled by authors. 5.6 Effects of Forbdden Sanctons among US, EU and Russa Forbdden sancton s the heavest sancton and retalaton. Table 12 reports smulaton results. The effects of forbdden sanctons are much stronger than hard sanctons. All nvolved countres lose and all non-nvolved countres gan. Russa wll be hurt heavly, the US and EU are hurt slghtly. All non-nvolved sancton countres wll gan, and comparatvely Chna and Korean wll gan more than other countres on utlty; exco, Canada and Chna can gan more than other countres on both export and mport. We take the utlty effects under mutual sancton and retalaton stuaton as an example to compare mpacts on dfferent countres. Effects on the US, the EU and Russa are separately %, % and %. Influences on Chna, Japan and Korea are separately 0.217%, 0.089% and 0.404%. Impacts on Canada, exco and Inda are separately 0.040%, 0.102% and 0.108%. Table 12: Effects of Forbdden Sanctons to ajor Countres (Unt: % change) Countres O11 O12 O13 O14 O15 O16 O17 23 / 28

24 Utlty US EU Russan Chna Japan Korea Canada exco Inda Export US EU Russan Chna Japan Korea Canada exco Inda Import US EU Russan Chna Japan Korea Canada exco Inda Source: compled by authors. From the above analyss n ths part, we fnd that all sancton nvolved countres wll lose and all non-nvolved countres wll gan. For three nvolved countres, Russa wll be hurt heavly, but the US and EU wll be hurt just slghtly. The EU s sancton threat to Russa are much stronger than the US s, and Russa s retalaton measures wll hurt the EU more than the US. Optmal game choce for the US and EU s to avod sancton measures, and the optmal choce for Russa s to choose retalaton measures. 24 / 28

25 6. Optmal Sanctons for US, EU and Russa We explore the optmal sancton tarffs for the US, the EU and Russa and then study the potental effects of the optmal sancton tarffs n ths part. In our numercal mult-country general equlbrum model, economc sanctons nclude both tarff and non-tarff barrers. We assume that the sancton measures are manly tarff, therefore the optmal sanctons wll be the optmal tarff. Here, the optmal tarff s not a whole optmal tarff that targeted to all countres n our numercal model, but t s a partal optmal tarff that targeted to economc sancton related countres of the US, the EU and Russa, whch means that the US and the EU choose the optmal tarffs to levy on Russa, and the Russan choose the optmal sancton tarffs to levy on both the US and the EU. There are two knds of optmal sancton tarffs. One s optmal sancton tarff wthout retalaton, whch means all three sancton countres choose ther own optmal sancton tarffs at the same tme. The other one s optmal sancton tarff wth retalaton, whch means that all three sancton countres frstly choose ther own optmal tarff but then three sancton countres change ther optmal sancton tarff accordng to ther partner s choce and terate untl all three countres wll not change and convergence to an equlbrum. The optmal sancton tarff wth retalaton s actually a non-cooperate Nash equlbra. We frstly calculate the optmal tarffs wthout retalaton. In the computaton, we maxmze one country s utlty subject to the global general equlbrum and assume all other countres do not change ther tarff and non-tarff barrers, and then we get the optmal sancton tarff wthout retalaton for ths country. Ths process can be descrbed as axu ( t ) s.. t GE country (27) Then we calculate the optmal tarff wth retalaton, whch are non-cooperatve Nash equlbra. In order to compute non-cooperatve Nash equlbra, we terate over calculatons of optmal tarff polcy responses by ndvdual regon to tarff settngs of other regons; subject to the constrant of full general equlbrum wthn the perod. We then terate across country tarffs and then countres untl convergence to a non-cooperatve Nash equlbrum s acheved. In computng non-cooperatve equlbra, we adopt Nash s (1951) non-cooperatve soluton concept. We use (29) to obtan convergence to a Nash equlbrum. Optmal sancton tarffs are reported n Table 13. Table 13: Optmal Sancton Tarffs for the US, the EU and Russan (Unt: %) Countres Wthout Retalaton Wth Retalaton US EU Russan Source: Calculated and compled by authors. Usng the optmal sancton tarffs, we can explore the mpacts of optmal sanctons to some major countres. Table 14 reports these results. Under the optmal sancton equlbrum, all three sancton nvolved countres wll 25 / 28

26 hurt but the negatve effects are much less than arbtrary sanctons. Comparatvely, Russa hurts the most, then goes the EU, and the US lose the least. ost of non-nvolved countres n the economc sancton wll gan except Canada and exco, whch means that not all non-nvolved sancton game countres gan under optmal sancton scenaro. Influence results under optmal tarff wthout retalaton and optmal tarff wth retalaton are close. Table 14: Effects of Optmal Sancton among the US, the EU and Russan (Unt: % Change) Countres US EU Russan Chna Japan Korea Canada exco Inda Optmal Tarff Wthout Retalaton Welfare Export Import Trade Optmal Tarff Wth Retalaton Welfare Export Import Trade Source: Calculated and compled by authors. The nfluence drecton of optmal sancton are the same as arbtrary soft sancton, hard sancton and forbdden sancton. The dfference s that the negatve effects on all nvolved countres are much smaller than arbtrary sancton measures. Optmal sancton s the sancton equlbrum n whch no countres want to move. 7. Conclusons The economc sancton and counter-sancton game among the US, the EU and Russa has attracted extensve attenton n the world. Ths paper uses a numercal general equlbrum model to explore the sancton game solutons and to smulate the effects of sanctons. We set up a three-stage sancton game ncludng soft sancton, hard sancton and forbdden sancton. Each stage game has seven game tree ponts and we compute payoffs for each pont. Our general equlbrum model has 16 countres who produce 2 goods (manufacturng goods and non-manufacturng goods) wth 2 factors (captal and labor). We nclude trade cost structure nto the model whch can be dvded nto tarff and non-tarff barrer, to explore the trade sancton polcy effects. We also use a fxed trade mbalance assumpton n the model to form an endogenously determned trade mbalance structure. We use both EV and CV to denote payoffs for ndvdual countres. The three-stage game soluton analyss fnd that strcter sancton measures have stronger effects to all nvolved countres. All three countres of the US, the EU and Russa wll be hurt by sancton and counter-sancton measures. Comparatvely Russa wll be heavly hurt, and both the US and EU wll receve small loss. The EU sancton measure wll generate stronger negatve mpact to Russa than the US measures, and Russa counter-sancton measure wll has more mpact to the EU than to the US. The optmal choce for the US and EU s 26 / 28

27 to avod takng sancton measures, and the optmal choce for Russa when faced wth sancton measures s retalaton. The sancton effects analyss show that all sancton and counter-sancton nvolved countres wll lose but all non-nvolved countres wll gan. Our emprcal analyss prove that the US and EU may contnue to use sancton measures to gve pressure to Russa for the comparatve mpacts to them are small compared wth ther economc scale. Faced wth sancton measures, Russa has no better choce than retalaton, but the ultmate results s Russa be hurt heavly. References 1 Anderson, J. and E.V. Wncoop Borders, Trade and Welfare. Brookngs Trade Forum 2001, Susan Collns and Dan Rodrk, eds., Washngton: The Brookngs Insttute, Chaney, T Dstorted Gravty: The Intensve and Extensve argns of Internatonal Trade. Amercan Economc Revew, 98(4), pp Dreyer, I. and N. Popescu Do Sanctons Aganst Russa Work. European Unon Insttute for Securty Studes, December. 4 Eaton, J. and S. Kortum Technology, Geography, and Trade. Econometrca, 70(5), pp Galbert, S.D A Year of Sanctons aganst Russa---Now What?. Center for Strategc & Internatonal Studes Report, October. 6 ICC The Potental Impact of the EU Sanctons aganst Russa on Internatonal Arbtraton Admnstered by EU-based Insttutons. Internatonal Chamber of Commerce Report, June Kaempfer, W.H. and A.D. Lowenberg The Theory of Internatonal Economc Sanctons: A Publc Choce Approach. The Amercan Economc Revew, 78(4), pp Nash, J.F Non-cooperatve Games. Annals of athematcs, 54(2), pp Nelson, R US Sanctons on Russa: Economc Implcatons. Congressonal Research Servce Report No.43895, February Novy, D Gravty Redux: easurng Internatonal Trade Costs wth Panel Data. Economc Inqury, 51(1), pp Oxensterna, S. and P. Olsson The Economc Sanctons aganst Russa: Impact and prospects of Success. FOI Report No. FOI-R-4097-SE, September. 12 Shoven, J.B. and J. Whalley Applyng General Equlbrum. Cambrdge Unversty Press. 13 Whalley, J and L. Wang The Impact of Renmnb Apprecaton on Trade Flows and Reserve Accumulaton on a onetary Trade odel. Economc odellng, 28, pp Wkpeda Internatonal Sanctons durng the Ukranan Crss. Avalable at: 15 Wong, A easurng Trade Barrers: An Applcaton to Chna s Domestc Trade. Job arket Paper, Unversty of Chcago, January / 28

28 IGI 简介 : 国际问题研究系列 (Insde Global Issues) 是由中国社会科学院世界经 济与政治研究所国际贸易研究室组织和发布的 该系列涉及的研究领域主要为国际经济与贸易 ; 主要成员包括余永定研究员 宋泓研究员 姚枝仲研究员 倪月菊研究员 田丰研究员 东艳研究员 李春顶副研究员 高凌云副研究员 马涛副研究员 张琳博士和苏庆义副研究员 声明 : 本报告为非成熟稿件, 仅供内部讨论 报告版权为中国社会科学院世界经济与 政治研究所国际贸易研究室所有, 未经许可, 不得以任何形式翻版 复制 上网和刊登 本报告仅代表作者的个人观点, 并不代表所在单位的观点 欢迎通过扫描下面的二维码订阅和关注我们的微信公众平台 ( 微信号 :wep_te, 名称 :IWEP 国际经济贸易研究 ) 28 / 28