Geographic politics, loss aversion, and trade policy: The case of cotton and China

Transcription

1 Geographic politics, loss aversio, ad trade policy: The case of cotto ad Chia (JOB MARKET PAPER) Weshou Ya School of Ecoomics The Uiversity of Adelaide November 26, 2016 Abstract This paper seeks to explai how govermets respose to world market price fluctuatios. It develops a theoretical model of trade policy icorporatig loss aversio ad referece depedece. Like Freud ad Özde (2008), this paper assumes oly trade policy istrumets are available to the govermet, but it goes beyod their model by addig a spatial dimesio to iterest-group politics. The model suggests that: (1) politically sesitive products receive more trade protectio; (2) the govermet s chagig trade distortios isulate the domestic market from iteratioal price fluctuatios by settig trade protectio lower (higher) whe the world price is higher (lower) tha a targeted domestic referece price; ad (3) variatios i market itervetio help producers at the expese of cosumers i periods whe the iteratioal price is well below tred, ad help cosumers at the expese of producers i high-price periods. These predictios from theory are show to still hold whe the model is exteded to a large coutry case ivolvig terms of trade effects. The model is tested empirically ad foud to offer a plausible explaatio of the puzzlig chages i cotto protectio i Chia. JEL classificatio: F13, F14, F59, Q17, Q18 Key words: Political ecoomy, Geographic politics, Loss aversio, Referece depedecy, Political sesitive product, Price volatility A special thaks to my supervisors Professor Kym Aderso, Dr Floria Ploeckl ad Dr Nicholas Sim for their guidace, isightful discussio ad commets. I have also beefited from discussios with ad commets from Professor James R. Markuse ad Dr Beedikt Heid. Ay errors are my ow.

2 1 Itroductio Chia s cotto policy attracts attetio iteratioally due to its cotto idustry s substatial role i the world cotto market, ad domestically because of where it is produced i Chia. Sice 2005, the share of Chia s cotto productio i the politically sesitive Xijiag area has bee cotiuously icreasig ad ow exceeds 50%. Chia s cotto trade protectio, measured by Nomial Rates of Assistace (NRA) 1, fluctuates at a higher degree betwee 2005 ad 2015 (Figure 1). This paper seeks to explai that chage, drawig o the approach used by Freud ad Özde (2008) who exted the Grossma ad Helpma model (1994) (G- H model hereafter) icorporatig agets prefereces characterisig loss aversio ad referece depedece i a small ope ecoomy. More specifically, we seek to uderstad the Chiese govermet s trade policy resposes to iteratioal price fluctuatios by takig cotto as a case study. I additio to sheddig light o the specific case of cotto i Chia, this paper cotributes to the literature i two other ways. Firstly, it specifies the govermet s objective fuctio as the sum of political support ad the aggregrate social welfare for a o-democracy characterised with sesitive geographic dimesios to iterest-group politics. It thus goes beyod the moetary cotributio model of Grossma ad Helpma (1994). Ad secodly, it cosiders two cases (whe the world price is higher, as well as whe it is lower, tha a referece price), which therefore goes beyod Freud ad Özde (2008) who oly cosider a world price slump. 1 It is expressed as NRA = P t P t P t where P t is the domestic price ad P t is the border price at time t. 1

3 Figure 1: Cotto trade protectio i Chia, Jauary 2005 to Jauary 2015 The structure of this paper is as follows. Sectio 2 summaries the pertiet literature. Factual backgroud iformatio o Chia s cotto productio, Xijiag s geographic positio, ad cotto trade policies are summarized i sectio 3. Sectio 4 develops the theoretical model ad exteds it from a small coutry to a large coutry. Chia s cotto trade policy is used to empirically test the model i sectio 5, ad sectio 6 cocludes. 2 Literature review 2.1 Political icetives drivig iefficiet ad suboptimal policies The perspective of political ecoomy provides a framework for politicias ad ecoomists to ucover the formatio of ad variatios over time i policy itervetios. Various braches of thought, datig back to the 1960s, have give isight ito the iteractios of ecoomic ad political forces amog differet iterest groups affectig the policy equilibrium. Amog the importat cotributios, Olso (1965) pioeered the role of collective actios to overcome 2

4 free-rider problem to ifluece policy outcomes of govermet. Numerous other traditioal political models, icludig regulatio theory (Stigler, 1971), pressure group theory (Becker, 1983 ad 1985), policy preferece fuctios (Rausser ad Freebair, 1974), political support fuctios (Hillma, 1982), political preferece fuctios (Bullock, 1994) ad the coservative social welfare fuctio (Corde, 1997) seek to explai the reasos why govermets implemet iefficiet distorted policies i differet sectors. I the case of agricultural policies, the arable lad edowmet per worker, the employmet share i the agricultural sector, terms of trade for agriculture, the share of agriculture i GNP, ad the share of food i total expediture are discussed based o collective actio by differet iterest groups (Aderso & Hayami, 1986; Rausser, 1982). Other factors icludig low farm icomes, slow farm productivity growth, ad low supply ad demad elasticities are also emphasized (Garder, 1987). Grossma ad Helpma (1994) improved the iterest group model by providig microecoomic foudatios, such that it became the workhorse tool to explai trade policy formatio. Based o the G-H model, a preferece for iequality aversio is itroduced ito idividual s utility fuctio (lü et al., 2012). This comparative static model was followed by a dyamic political ecoomy model with overlappig geeratios, heterogeeous agets, edogeous huma capital ivestmet, ad costly worker adjustmet (Blachard ad Willma, 2013), ad used to aalyse the protectioist overshootig pheomeo. Specifically, whe politically iflueced workers are stuck i adversely affected importcompetig sectors, they are more likely to get short-term policy remediatio i the form of higher tariffs. The more uequal the iitial distributio of gais ad losses from the magitude of potetial overshootig will be, the loger the iduced policy distortio will persist. 3

5 I parallel with the political cotributio model, the tariff-formatio-fuctio model (Fidlay & Wellisz, 1982), campaig-cotributio model (Magee et al., 1989), political support model (Rodrick, 1995) ad media-voter model (Mayer, 1984) were developed ad adopted to aalyse agricultural policy formatio. Other cotributios to policy makig that have bee emphasized more recetly are istitutios (Acemoglu ad Robiso, 2006 ad 2012), limited access orders (North, Wallis ad Weigast, 2009), the role of costitutios (Persso ad Tabellii, 2000), ad electoral istitutios (Besley ad Persso, 2011). 2.2 Loss aversio ad trade policy itervetios The G-H model hypothesizes that a idividual's utility oly depeds o his or her cosumptio budle, which meat it could ot explai behavioral elemets associated with political ecoomy dyamics behid trade protectio (Dissaayake, 2014). Agets prefereces toward loss aversio ad referece depedece are ow beig built ito political cotributio models (Freud ad Özde, 2008; Tovar, 2009). Loss aversio refers to people's tedecy to feel stroger about avoidig losses tha acquirig gais, ad losses reflect particular referece poits. Freud ad Özde (2008) explai why trade protectio is give whe the world price falls below a give referece price. Tovar (2009) icorporates idividual prefereces exhibitig loss aversio ito the political objective fuctio, ad poits out that a idustry is more likely to orgaize ad lobby the govermet if it suffers a loss. Durig recet years, loss aversio has bee built ito aalyses of govermet resposes to market shocks. Aderso ad Nelge (2012) set up loss aversio i quadratic rather tha liear form, which is cosistet with the coservative social welfare fuctio i Corde (1997). They show that durig price upward spike periods, developig coutries alter their agricultural trade policies more tha high-icome coutries, ad vice versa durig dowward agricultural price shocks. Giordai et al. (2016) aalyze the multiplier effect of foodexportig coutries seekig to isulate the domestic market from the world market. 4

6 Dissaayake (2014) presets a geeral equilibrium model that projects chages i trade restrictios irrespective of the lobbyig behaviors of iterested groups who make moetary cotributios to the democratic govermet. Theakoo (2015) follows Baldwi (1987) with a partial equilibrium model i which the govermet objective fuctio is the weighted summatio of cosumer surplus, producer surplus, ad tariff reveue, ad uses loss aversio as i Freud ad Özde (2008) ad Tovar (2009) to aalyze govermet resposes to dowward spikes i iteratioal prices. Loss aversio is also used by Fulto ad Reyolds (2015) i cosiderig the rice export system i a o-democratic coutry, Vietam. They coclude that i such a settig, the elite could icrease their political ad ecoomic power from restrictig exports. I this paper, we documet the effects of sesitive political groups o the govermet s trade policy formatio process i a oe-party coutry characterised with geographic dimesios of iterest-group politics. The govermet s objective fuctio is set with behavior features icludig referece depedece ad loss aversio ot oly from a producers perspective but also from that of cosumers. Chia s cotto policy is show to be cosistet with the predictios of that theoretical model. 3 Geography, politically sesitive products, ad preferece 3.1 Geography ad politically sesitive products Policy pressure arises from policy prefereces of self-iterested agets. Ecoomic actors ca orgaize to ifluece govermet policy to their advatage because of the spatial distributio of ecoomic edowmets (Chase, 2015). Geography ca sometimes shape idividual s prefereces, collective actio ad aggregate prefereces of the govermet if the edowmet factor is located geographically i particular ways. Self-iterest ca be pursued by creatig 5

7 social urest, sedig petitios to the cetral govermet, or otherwise fightig for their rights. Regios with a high proportio of miorities i the total populatio ca be highly sesitive politically, as ca oes i which a product is cocetrated i just oe politically sesitive regio. A formal defiitio of a politically sesitive product, drawig o Jea et al. (2011), could be: A politically sesitive product is oe whose output is produced usig a specific edowmet factor geographically located i a politically sesitive regio, ad the producers are vulerable to chages i govermet policy affectig that product. 3.2 Politically sesitive regios: Xijiag The geographic locatio ad the large share of Muslims i Xijiag make it a politically sesitive regio. The largest of Chia's admiistrative regios, Xijiag borders eight coutries - Mogolia, Russia, Kazakhsta, Kyrgyzsta, Tajikista, Afghaista, Pakista ad Idia. It is located i the far Northwest of Chia, ad trasportatio liks to the east through the cetral area of mailad Chia are weak. The shares of the total populatio of each provice that is a miority are listed for 2014 i Table 1. Xijiag raks secod oly to Tibet out of the 26 provices whose statistics are available, with 60% of its total populatio beig Uyghur. The higher the share of miorities i the provice, the more they share commo iterests ad prefereces. The miorities are more likely to orgaize political groups to fight agaist local or cetral govermets, or create social urest to force the govermet to allocate beefits to them. Table A i the appedix depicts the cases of social urest (coflicts betwee Ha ad Uyghur) from 2007 to I 2009, the biggest coflicts betwee Ha ad Uighur people occurred. I that social urest, almost 200 people were killed, 1721 people were ijured ad 6

8 1000 people were arrested by the govermet. I 2014, there are 9 social urests related to Xijiag Uighur group whose umber is much higher tha other years. Table 1: Share of miority i total populatio i each provice i 2014 Rakig Provice % Rakig Provice % 1 Tibet Hubei Xijiag Hebei Qighai Beijig Guagxi Tiaji Guizhou Fujia Nigxia Guagdog Yua Hea Neimeggu Zhejiag Haia Shadog Liaoig Ahui Hua Shaghai Jili Shaaxi Gasu Jiagsu Chogqig Shaxi Heilogjiag Jiagxi Sichua 4.98 Notes: The uit of the value is percetage. Data source 2 : The role of cotto i Xijiag Xijiag s cotto sector i plays a importat role i Chia. The cotto yield i Xijiag i 2009 was 123 kilogram per acre (M), which is oe-quarter above the atioal average. Xijiag s share of total productio of cotto 3 i Chia was 30% i 2002, but the it sharply 2 Origial data sources: The sixth cesus of Chiese govermet ad reports of the local govermet. 3 The average aual cotto productio i Chia betwee 1995 to 2004 was 22,319 [uit: thousad 480-poud bales], but it icreased to a peak of 32,332 betwee 2005 ad

9 icreased to 62.5% by Figure 2 illustrates the cotto productio geography i Chia i 2012, whe Xijiag s share was 52%. Cotto productio has become a importat part of Xijiag s ecoomy. Cotto accouts for 65% of its crop sector ad 1/3 of its total agricultural sector. More tha 50% percet of people i Xijiag are egaged i cotto productio, ad 35% of their icome is from cotto o average but that share was up to 60% i the cotto-itesive areas i For the local govermet, 15% of their fiscal icome is from cotto productio ad related sectors. For some cotto-itesive productio couties, the proportio of fiscal icome peaks at more tha 50%. The cotto sector also accouted for more tha 17% of Xijiag's GDP i Figure 2: Geographic distributio of Chia s cotto productio i 2012 Source: Author s calculatio "Cotto is itimately associated with lad usage, owership, employmet, ad Ha immigratio. It's all tied up". ---Tom Cliff, a scholar at the Australia Natioal Uiversity (20 February 2015) 4 4 See /?r=AU&IR=T 8

10 The Chiese govermet is more likely to protect cotto platers due to the importat role of cotto i employmet ad icome i Xijiag. Social urest ad agricultural price shocks have a positive relatioship which has bee tested recetly by Bellemare (2014) ad Arezki ad Brucker (2011). If a product is geographically cocetrated i its productio, 5 the Chiese govermet teds to protect the sector whe cosiderig major employmet. Besides, those workig as cotto platers are relatively uskilled. If the govermet does ot protect the cotto sector, a higher uemploymet rate may result ad potetially lead to social ad political urest i Xijiag. Maitaiig social stability is a objective of Chia s cotto policies: Chia s cotto policy is cogizat of social stability. They wat to cotrol riotig i the Xijiag provice, where most of the cotto is grow Elto Robiso (15. March 2013) 6 I short, cotto is a politically sesitive product whose productio is geographically cocetrated i Xijiag provice -- a politically sesitive regio. 3.4 Cotto trade policy i Chia Chia is the world s largest cotto producer, cosumer ad importer i the world. Table 2 shows Chia s et trade volumes betwee 2005 ad Table 2: Cotto et import volume, Year Value 19,212 10,500 11,468 6,912 10,880 11,857 24,478 20,280 14,096 8,213 4,800 Notes: The uit of the value is (000) 480-poud bales. Data source: USDA-Foreig Agriculture Service 5 The coal sector i some Europea coutries receives higher protectio ad govermet subsidy. The geographically cocetrated idustry is ofte a major employer i a tow or city ad ivolves a small umber of tows or cities (Aderso, 1995b). 6 See Chiese cotto policy- Social stability, ot trade. 9

11 The Chiese govermet s trade policy has bee largely focused o maagig import flows to competig iterests of cosumers 7 ad cotto farmers. A Slidig Scale Duty (SSD) system has bee i place sice Source: Wag et al. (2014) Figure 3: Chia s actual Tariff Rate Quota system I Chia, its i-quota import volume icludes regular quotas ad additioal quotas permitted by the Slidig Scale Duty system. As illustrated i Figure 3, withi the regular import quota, the import tariff is very low at 1%. If the import exceeds the sum of the regular quota ad additioal quota, the tariff is take to the highest level of 40%. If the import volume belogs to additioal quota, the govermet will implemet a Slidig Scale Duty to calculate the tariff rate uder the Slidig Scale Duty system, which is ot allowed to be higher tha 40%. Chia s actual Tariff Rate Quota (TRQ) system shows the tariff rate is fixed withi the regular quotas. The fluctuatios of tariff rates deped o the additioal quotas cotext. 7 Cotto cosumers are mills i textile idustry rather tha citizes, because the raw cotto is the itermediate iput to produce clothes. 10

12 Figure 4 gives the compositio of cotto imports. The primary istrumets determiig Chia s cotto imports are import size, timig, ad coditioality of quotas. Most of Chia s cotto imports are uder the Slidig Scale quota (SSQ). Figure 4: Chia s cotto import compositio durig 2008 to 2013 Notes: Policy e.g. imports by Chia Natioal Cotto Reserves Corporatio (CNCRC); Other: imported at full 40% WTO boud tariff Source: MacDoald et al. (2015) This paper mostly focuses o the variatio of the import tariff. It aalyses how the tariff rate is calculated based o the Slidig Scale Duty withi the additioal quotas rage. The Chiese govermet itroduced the Slidig Scale Duty system i 2005 ad has sice adjusted it cosiderably. The Slidig Scale Duty system is categorized ito two stages (periods): durig period 1 (May 1, 2005 December 31, 2006), if the import price icludig Cost Isurace ad Freight (CIF) was at or above a referece price (P t ) 8 set by the govermet, the tariff rate is 5% 8 The Chiese govermet sets the referece price (cost, isurace, ad freight) for the lowest tariff o cotto imported uder a slidig scale quota (caledar year) (MacDoald et al., 2015). The referece price was Yua/To i 2005, Yua/To i 2006, Yua/To betwee 2007 ad 2011, ad yua/to from 2012 to

13 for imported cotto withi the additioal quotas. If the CIF price is higher tha the referece price, the followig formula was adopted to calculate cotto tariff rate. T I = mi(int((p t /CIF 1) )/1000, 0.4) (1) where T I is the import tariff rate i period oe; CIF is the import price; 9 P t is the value of P t ( ); ad mi is a miimum fuctio idicatig that the maximum import tariff value is 0.4. INT is a INT fuctio to get the iteger part of the value i the outer paretheses. The secod period covers a loger time (Jauary 1, 2007 December 31, 2015) ad the Slidig Scale Duty for additioal quota was maaged by the govermet as follows. If the import price is at or above govermet s referece price P t, the tariff rate for import cotto was 0.57 Yua per kilogram. If the referece price is below the referece price, the import tariff is calculated usig the followig formula. T II = mi (INT (P t CIF + α CIF ) 1000, 0.4) 10 (2) where T II is the import tariff rates i period two; P t = ( )Pt α(p t ) 2 ; α is a costat umber whose value takes 2.526% from , 3.235% i 2012, ad i 2013, respectively. From the Slidig Scale Duty system, we ca see that the import tariffs ragig from 1% to 40% are determied by the category of the imports ad the year the cotto is imported. Wag et al. (2014) rewrite the above two formulas ito equivalet ad valorem tariff format, but it does ot chage the import tariff rules. I short, the tariff rate is iversely related to the iteratioal price. 9 Uder the Slidig Scale Duty system, the Chiese govermet ca import from ay other cotto exportig coutries. These coutries are competitive. I the empirical part, the CIF price is fuctioed as the world price. 10 See Wag et al. (2014). 12

14 4 Theoretical framework Cotto producers are almost always et sellers i the short-term, which makes them differet from staple food producers. The icome effects due to a product price chage are ot ambiguous for cotto platers: they gai whe facig domestic market price icreases, ad vice versa. This sectio presets a theoretical framework to be used i the applied empirical aalysis i sectio 4.5. Two cases are cosidered: a small ope ecoomy, ad a large ope ecoomy. 4.1 Model assumptios 11 Cosider a small ope ecoomy populated by idividuals with idetical prefereces. Idividuals ow differet types of specific factors ad labor edowmets. All the agets have the followig cosumptio preferece characterized by loss aversio ad referece depedece: U = x 0 + u i (x i ) + mi(h(x u i (x i ) U ), 0) (3) where x 0 is umeraire good produced oly by labor with costat retur to scale, ad the iput-output coefficiet equals 1 (x 0 = L 0 ). The umeraire good could be defied as the import good or the export good. By defiitio, its domestic price ad world price are equal to 1. Uder a competitive labor market, the wage rate is equal to 1. x i is cosumptio of good i, i = 1, 2,,. All the ormal goods require labor- ad sector-specific iputs with fixed supply i the ecoomy exhibitig costat returs to scale. While the specific factors are immobile across sectors, laborers have free mobility i the ecoomy. With the wage rate equal to oe, the returs to the specific factor owers deped oly o the domestic market 11 The author borrows some of the basic assumptios from Grossma ad Helpma (1994) i the small ope ecoomy, ad assumptios from Grossma ad Helpma (1995) i the cotext of a large coutry. The differece betwee the two cases is whether the coutry ca affect the iteratioal market price. 13

15 price p i deoted by π i (p i ). The supply of good i is deoted by y i (p i ) = π (pi i ), which is a applicatio of Hotelig s lemma. Followig Freud ad Özde (2008), we itroduce behavior features ito cosumer utility through a h( ) fuctio. The h( ) fuctio is called gai-loss utility 12 (Dissaayake, 2014), ad its first derivative is positive 13 ad secod derivative is egative. 14 I other words, the gai-loss term is icreasig i the differece betwee the actual utility level ad the referece utility level idicated by U. U is a idividual s referece utility derived from cosumig a referece cosumptio budle. The fuctio takes a egative value whe the actual utility is lower tha the referece level, ad zero otherwise. With the above prefereces a idividual cosumes x i = d i (x i ) ormal goods, i = 1,2,,, where demad is the iverse of U i (xi ) ad x 0 = E 1 p i x i. The related idirect utility fuctio is expressed as: W i (p i ) = E i p i d i (p i ) + i=1 u i (d i (p i )) + mi (h(e p i d i (p i ) i + i=1 u i (d i (p i )) H ), 0) (4) The utility equatio could be rewritte as: W i (p i ) = E + s(p i ) + mi(h(e + s(p i ) U ), 0) (5) where s(p i ) = i=1 u i (d i (p i )) i p i d i (p i ) idicates the cosumer surplus. If we deote the referece level of utility as U = E + S(p), the the above fuctio (5) could be rearraged as: W i (p i ) = E + s(p i ) + mi(h(e + s(p i ) E + S(p) ), 0) (6) 12 The price of the umeraire goods is costat. Therefore, the utility fuctio is liear i x 0 but ot other ormal goods. 13 Idicatig as h ( ) > 0, which meas the extet of loss a idividual feels for havig less tha they are accustomed to. 14 Idicatig as h ( ) < 0, which meas the margial icrease is decliig i the size of loss due to dimiishig sesitivity to losses. 14

16 The wedge betwee the domestic market price (p i ) ad the iteratioal market price (p i w ) is t i s, created by the govermet s price-distortig policy. The relatioship betwee the domestic market price ad the world price is simply expressed as: p i = p i w + t i s (7) If t s i > 0, it meas that the domestic market price is higher tha that i the iteratioal market, idicatig that the govermet imposes a tariff o imports or a export subsidy o exports. Whe t s i < 0, it meas the domestic market price is lower tha the world price, i which case imports are subsidized or exports are taxed. 15 As per the above assumptio, the govermet oly imposes trade distortios to maage the variatios i the domestic price. The assumed aim of the govermet is to maximize its objective by implemetig pricedistortig policies, with the ultimate objective of beig to stay i office ad cotrol the coutry s power. I the cotext of Chia, there is o formal lobby group to make moey cotributios to the govermet. However, iterest groups ca express their uwilligess or ager through, for example, creatig social urest. We model the govermet s political objective fuctio as the summatio of total political support from politically sesitive groups, ad the aggregrate welfare of the ecoomy as the followig liear fuctio: OFG = i g PS i + φ i=1 W(t s i ) φ 0 (8) where OFG is the objective fuctio of the govermet; i g PS i is the political support from politically sesitive groups idicated by g; W(t s i ) is the aggregrate social welfare; ad φ represets the weight that the govermet puts o aggregate social welfare. The value of φ is a positive value. We propose that political support is a strictly mootoic icreasig fuctio 15 The protected product could be import goods or export goods. I Chia, cotto is a type of et import agricultural product. The Chiese govermet imposes tariffs o import cotto to maage domestic market price. 15

17 with respect to the welfare of the politically sesitive group. Equivaletly, the govermet s objective fuctio could be rewritte as: Ω = i g H(t s i ) + φ i=1 W(t s i ), φ 0 (9) I this model, the govermet of Chia cosiders politically sesitive areas which are geographically related to producig a specific product. The govermet would like to cosider that regio s welfare more tha the welfare of other groups, which is expresses as follows: i g H(t s i ) = a g l + a g t s i M i (p i ) + i g π i (p i ) +a g i=1 i=1 s i (p i ) (10) where g i the third term is a set of politically sesitive groups which have the higher power to argue with the govermet, ad a g is the proportio of idividuals i the total populatio who belog to the politically sesitive groups. For the secod term, the aggregate social welfare cosists of four terms: i=1 W(t s i ) = l + t s i M i (p i ) + π i (p i ) + i=1 i=1 i=1 s i (p i ) (11) where l is the total labor icome (wage rate is oe ad total labor supply is l); i=1 t s i M i (p i ) deotes total tariff reveue ad M i (p i ) is the trade value for product i; i=1 π i (p i ) is the total retur for specific factors; ad i=1 s i (p i ) is the total cosumer surplus. The equilibrium optimal tariff rate ca be solved by maximizig the govermet s objective fuctio (equatio (9)) with respect to the trade protectio level (t i s ): t s i = arg max ( i g H(t s i ) + φ i=1 W(t s i )) (12) 4.2 A small coutry model: Three scearios Regardig the model assumptios, the idividuals prefereces deped o the differece betwee the actual cosumptio ad the referece cosumptio levels. Because of this, the 16

18 form of the govermet objective fuctio depeds o the differece betwee the equilibrium domestic market price ad the referece price set by the govermet authority. Therefore, three scearios are cosidered i tur i aalysig the optimal trade policy for the govermet to maximize its object fuctio: whe the equilibrium price exactly equals, is lower tha, or is higher tha the referece price. The equilibrium domestic price equals the referece price Whe the domestic equilibrium market price equals the referece price, the idividuals will have a utility fuctio excludig the loss-gai term. The welfare of the politically sesitive groups ad the aggregate social welfare are the same as equatios (10) ad (11) respectively. Substitutig the two equatios ito the govermet objective fuctio (equatio (9)), we get: Ω = [a g l + a g t s i M i (p i ) + i g π i (p i ) +a g i=1 i=1 s i (p i ) ] + φ[l + i=1 t s i M i (p i ) + π i (p i ) + i=1 i=1 s i (p i )] (13) Simplifyig the above equatio (13): Ω = (φ + a g )l + (φ + a g ) i=1(t s i M i (p i ) + s i (p i )) + i=1 (φ + g i )π i (p i (14) Tryig to choose the optimal trade protectio vector (based o political support schedules) is equivalet to maximizig the objective fuctio of the govermet with respect to protectio level t i s, which is followig the idea of equatio (12). The first-order coditio is give as the followig equatio (15) by usig Roy s idetity ( s(p i) = d i (p i ) ) ad Hotellig's lemma p i (π i (pi ) = y i (p i )), where d i (p i ) is domestic demad ad y i (p i ) is the domestic supply for product i. Besides, the relatioship betwee domestic market price ad iteratioal market price (equatio (5)) is applied here. (φ + a g ) [ d i (p i ) + t i s M i (p i ) + M i (p i )] + (φ + g i )y i (p i ) = 0 (15) 17

19 The relatioship betwee domestic demad, supply ad import is expressed as: M i (p i ) d i (p i ) = y i (p i ) (16) The equatio (15) ca be expressed as: (φ + a g ) [ y i (p i ) + t i s M i (p i )] + (φ + g i )y i (p i ) = 0 (17) Rearragig the above equatio, the optimal trade protectio level is give by: t i s = [ g i a g φ+a g ] y i(p i ) M i (pi ) (18) The solutio ca be rewritte as: t i s p i = [ g i a g φ+a g ] z i e i (19) where e = ( M i (p i ) M i (p i ) )/( p i pi ) = M i(p i ) p i = M (pi ) M i (p i ) i p i p i M i (p i ) is the import demad or export supply elasticity of good i; ad z i = y i(p i ) is a equilibrium ratio of domestic output M i (p i ) to imports (egative for exports). I the followig, we chage the form of the tariff to become ad valorem: 16 t i 1+t i = [ g i a g φ+a g ] z i e i (20) This is the solutio for the bechmark situatio whe the domestic equilibrium price equals the referece price. No loss aversio is created by a upward spike i the agricultural price 16 Because the domestic price (p i = p i w + t i s ) is kow, we ca get the trade distortio as t i s = p i p i w, the t i s p i w t i s = p w i p i p i p i w = p i p i = p i p i. I the fial step, the umerator ad deomiator are simutaeously divided by p i p i w p i w p i p i w = p i p i w p i w p i p i w + p i w p i w = p i p i w p i w 1 + p i p i w p i w = t i 1 + t i : 18

20 for the cosumers or by a dowward spike for the producers. From the above optimal protectio, politically sesitive groups receive positive protectio. This is because g i is a idicator variable: if the group who ow a specific factor to produce a politically sesitive product, the value equals oe, ad zero otherwise. The other effect of oe specific product is the output to import ratio. If that oe specific product accouts for a large share, the specific group has more power to gai from price distortios. The protectio level is egatively related to the import demad elasticity. The other two variables are the weight o the aggregrate social welfare, ad the share of the populatio that belogs to the politically sesitive groups. I short, the predictios of the above optimal trade protectio are: Bechmark results: Politically sesitive groups receive positive protectio. The protectio level is positively related to the output-import ratio; egatively proportioal to the share of the total populatio i the politically sesitive regios, the import demad elasticity, ad the govermet s weight o the aggregrate social welfare. The equilibrium domestic price is below the referece price What should be the trade protectio level whe the equilibrium price is lower tha the referece price? I this situatio, the retur of specific factors will be low due the decrease i the output price. Therefore, the egative deviatio of price from its referece price will result i further welfare loss for the producers through the loss aversio term if they produce that product. Followig the same argumet as Freud ad Özde (2008) ad Dissaayake (2014), the producers pay more attetio to the retur of factor icome tha to chages i tariff reveue ad cosumer surplus. The other idividuals, whose specific factors are ot used to produce this product whose price decreases, are et buyers. The price decrease of this product will 19

21 cotribute to the positive gai of et idirect utility to cosumers. However, the positive gai i the loss-gai fuctio does ot add additioal utility gai. Based o these argumets, the stadard aggregate social welfare (equatio (11)) becomes the followig form: i=1 W(t s i ) = l + t s i M i (p i ) + π i (p i ) + i=1 i=1 i=1 s i (p i ) + π i (p i ) ) α i N mi ( a i Nh ( π l i=1, 0) (21) The first four terms are the same as equatio (11) idicatig total labor icome, tariff reveue, total specific factor icome, ad cosumer surplus. The last term i the above equatio is the loss aversio part from producers whose specific factors experiece retur decreases, leadig to egative social welfare. I the loss aversio term, α i deotes the share of the populatio who ows oe specific factor i, ad N is the total populatio. Followig the same logic, the welfare of the politically sesitive groups becomes: i g H(t s i ) = α g l + α g t s i M i (p i ) + α g i g π i (p i ) + α g s i (p i ) i=1 i=1 + π i (p i ) ) α i N mi ( a i Nh ( π l i g, 0) (22) The oly differece betwee equatio (22) ad equatio (10) is that the loss aversio term eters the welfare fuctio, which is expressed as the last term i equatio (22). Therefore, we substitute euqatios (21) ad (22) ito equatio (9) ad rearrage the equatio as: Ω = (φ + α g )l + (φ + α g ) (t s i M i (p i ) + S i (p i )) + i=1 (φ + g i )π i (p i ) i=1 + π i (p i ) mi ( (φ + g i )a i Nh ( π l i=1 ), 0) (23) α i N 20

22 We maximize Ω with respect to t i s followig the idea of equatio (12), which yields the followig first-order coditio by usig Roy s idetity, Hotellig's lemma ad equatio (7) agai: (φ + α g ) [ d i (p i ) + t i s M i (pi ) + M i (p i )] + (φ + g i )y i (p i ) (φ + g i )a i Nh ( π l i (p i ) ) 1 ( y α i N α i N i (p i )) = 0 (24) Solvig this equatio with respect to the optimal trade distortio ad writig it i ad valorem form o good i gives: t i = [ g i a g +(φ+g i )h ( ) ] z i 1+t i φ+a g e i (25) where e is import demad elasticity of good i or export supply; ad z i is a equilibrium ratio of domestic output to imports (egative for exports). Comparig the optimal protectio level with equatio (20), the oly chage is the term from the umerator (φ + g i )h ( ). Accordig to the characteristics of the loss aversio fuctio, the first derivative is positive, illustrated as h ( ) > 0, ad the (φ + g i )h ( ) > 0. Thus the optimal protectio level is higher compared with the protectio level whe the equilibrium price equals the referece price. Whe trade protectio is higher, the domestic market price must be higher tha the world price. If the equilibrium domestic price goes lower tha the referece price, the world price is lower tha the referece price. Hece the followig Propositio: Propositio 1: Whe the world price is below its referece price i.e. p w i < p i, the govermet itroduces a higher distortio tha the level of distortio whe the world price is at the referece level. 21

23 [ g i a g +(φ+g i )h ( ) ] z i > [ g i a g ] z i (26) φ+a g e i φ+a g e i The equilibrium domestic price is above the referece price If the equilibrium price goes above the referece price, producers gai. However, et buyers whose specific factors do ot experiece a price icrease will lose. The loss aversio term eters the objective fuctio of the govermet due to this loss of cosumers surplus. I for the special case of cotto, the gai for producers domiats the situatio. The differece betwee the gais i factor icome ad the loss i cosumer surplus is positive for producers who are et sellers. I this sceario, the loss aversio term from cosumers perspective is added to the stadard aggregate social welfare (equatio (11)). i=1 W(t s i ) = l + t s i M i (p i ) + i=1 s i (p i ) + i=1 i=1 π i (p i ) a i )Nh ( t i s s l=1 M l l=1 ti Ml (p i ) + l=1 s l i=1 si (p i ) N + mi ( (1 ), 0) (27) where a i is the share of idividuals that experiece a price icrease i the good they produce; 1 a i = β i represets the share of idividuals who are et buyers of the good that experiece a world price icrease. I this case, followig the same logic, the welfare of the politically sesitive groups is: i g H(t s i ) = α g l + α g t s i M i (p i ) + α g i=1 s i (p i ) + i g π i (p i ) mi ( β i g Nh ( i=1 + t i s s l=1 M l l=1 ti Ml (p i ) + l=1 s l i=1 si (p i ) N ), 0) (28) where β i g is the share of idividuals, who are et buyers of the good that experieces a world price icrease i the politically sesitive groups. β i g is smaller or equal to β i i the ecoomy. Substitute equatio (27) ad (28) ito equatio (9), we get: 22

24 Ω = (φ + α g )l + (φ + a g ) (t s i M i (p i ) + s i (p i )) + i=1 (φ + g i )π i (p i ) mi ( (φβ i + β i g )Nh ( i=1 + t i s s l=1 M l l=1 ti Ml (p i )+ l=1 s l i=1 si (p i ) N ), 0) (29) Applyig Roy s idetity, Hotellig's lemma ad equatio (7), the first-order coditio of equatio (29) with respect to t i s is: (φ + a g ) [ d i (p i ) + t i s M i (pi ) + M i (p i )] + (φ + g i )y i (p i ) + (φβ i + β i g )Nh ( )[ y i (p i ) + t i s M i (p i ) + M i (p i )] 1 N = 0 (30) Rearragig the above equatio, we ca solve the politically optimal trade protectio: t i s = [g i a g (φβ i +β i g )h ( )] [φ+a g +(φβ i +β i g )h ( )] y i (p i ) M i (pi ) (31) Fially, we write the protectio i ad valorem form o good i as: t i g )h ( ) ] z i g φ+a g +(φβ i +β i )h ( ) 1+t i = [ g i a g (φβ i +β i e i (32) Compared with the bechmark protectio level (equatio (20)), the oly differet term eterig the politically optimal solutio is (φβ i + β g i )h ( ), which takes a positive value accordig to the characteristics of the loss aversio fuctio. The decrease of the umerator ad the icrease of the deomiator lead to the ratio value decrease. Thus the protectio level is lower tha i the sceario where the equilibrium domestic price equals the referece price. I additio, whe the protectio level is lower ad domestic market is lower tha the referece price, the iteratioal market price must be lower tha the referece price. Propositio 2 summarizes this coclusio as follows: Propositio 2: whe the world price goes higher tha the referece price i.e. p w i > p i, the govermet itroduces lower distortios tha the level of distortio itroduced whe the world price is at its referece price. 23

25 [ g i a g (φβ i +β i g )h ( ) φ+a g +(φβ i +β i ps )h ( ) ] z i e i < [ g i a g φ+a g ] z i e i (33) 4.3 Do terms of trade effects matter? From the above geeral equilibrium model, we ca predict the politically optimal tariff respose i a small ope ecoomy to chages i the iteratioal market price. However, the politically optimal policies for a large ope ecoomy take ito accout a coutry s ability to ifluece its iteratioal terms of trade (Feestra, 2016, p. 213). Broda et al. (2008) argue that market power explais more of the tariff variatio tha a commoly used political ecoomy variable. Freud ad Özde (20) ad Dissaayake (2014) igore terms of trade. This sub-sectio explores whether the above theoretical predictios for a small ope ecoomy are still relevat if terms of trade matter to the govermet. We assume two coutries exist ad both have the power to affect the world price, but otherwise keep the same assumptios as i the small coutry case. The foreig coutry is w idicated by. The world price is expressed as p i for product i. I order to simplify the calculatio process, the relatioship betwee the domestic market ad iteratioal market is assumed to be p i = t i p w i i the home coutry ad p i = t w i p i for the foreig coutry. If t i ad t i are bigger tha oe, it meas the govermets implemet a import tariff or export subsidy. If t i ad t i are less tha oe, it idicates a import subsidy or export tax. The political support fuctios are the same as i the small coutry case. The welfare of the politically sesitive groups is: i g H(t s i ) = a g l + a g i=1 r(t i, p w i ) + i g Π i (t i, p w i ) +a g i=1 S(t i, p w i ) (34) where l cotiues to represet the total labor icome. The remaiig three terms are fuctios of trade protectio ad the world price. r(t i, p w i ), Π i (t i, p w i ) ad S(t i, p w i ) idicate trade reveue, retur for specific factors ad cosumer surplus, respectively. 24

26 The aggregate welfare of the ecoomy is expressed as the followig equatio whose four terms have the same meaig as equatio (11). However, terms of trade effects are cosidered i this case. i=1 W(t s i ) = l + i=1 Π i (t i, p w i ) + i=1 r(t i, p w i ) + i=1 S(t i, p w i ) (35) The objective fuctio of the govermet does ot chage, which is the same as equatio (9), ad we cotiue to solve the politically optimal trade protectio followig the idea of equatio (12). Equilibrium domestic price equals the referece price Whe the equilibrium domestic market price equals the referece price, the welfare chage will ot lead to ay egative deviatio from the target value. Because of this, the loss aversio term does ot eter the govermet objective fuctio, which is the same as sceario oe of a small coutry case. Substitutig the welfare of politically sesitive groups (equatio (34)) ad the aggregate social welfare (equatio (35)) ito govermet objective fuctio (equatio (9)) yields: Ω = a g l + a g i=1 r(t i, p w i ) + i g Π i (t i, p w i ) +a g i=1 S(t i, p w i ) + φl + φ i=1 Π i (t i, p w i ) + φ i=1 r(t i, p w i ) + φ i=1 S(t i, p w i ) (36) The trade reveue fuctio (r(t i, p i w )) is kow as the product of trade quatity ad tariff rate: r(t i, p i w ) = (t i 1)p i w [d i (t i, p i w ) 1 N y i(t i, p i w )] i=1 (37) Substitutig equatio (37) ito equatio (36) ad rearragig the fuctio gives: Ω = (a g + φ)l + (φ + g i ) 1 Π i (t i, p w i ) + (a g + φ){ i=1(t i 1)p w i [d i (t i, p w i ) 1 y N i (t i, p w i )] + i=1 S(t i, p w i )} (38) 25

27 Maximizig the objective fuctio of the govermet with respect to t i provides the first order coditio: (φ + g i )y i p i w + (φ + g i )y i t i p i1 w + (a g + φ) [g i M i (t i, p i w ) + (t i 1)p i1 w M i (t i, p i w ) + (t i 1)p i w M i pi w + (t i 1)p i w t i M i pi1 w d i (t i, p i w )p i w d i (t i, p i w )t i p i1 w ] = 0 (39) Equatio (39) is simplified as: (g i a g )(p i w + t i p i1 w )y i + (φ + a g )[(t i 1)p i w (p i w + t i p i1 w )M i pi1 w M i (t i, p i w )] = 0 (40) I the two large coutries case, the terms of trade effects matter for the variace of trade distortios. Based o the world market clearig coditio, oe obtais the followig equatio: M i (t i, p i w ) + M i (t i, p i w ) = 0 i = 1, 2,......,. (41) The partial derivative of the world price with respect to home coutry s trade policy is expressed as: p w i1 = p i w = M w pi i t i M i ti +M i ti (42) The politically optimal trade protectio is solved by substitutig equatios (41) ad (42) ito equatio (40) to get: (t i 1) = (g i a g ) (φ+a g ) y i p i w ( M i ) + 1 e i for i = 1, 2,...., (43) The above solutio is the maximized political target fuctio regardig the protectio level for the home coutry. 17 e i = π im i t i M i is the elasticity of foreig coutry s export supply. 17 The uilateral trade policy for the foreig coutry could be derived through the same method expressed as: (t i 1) = (g i a g ) couty. (φ +a g ) y i p i w M i + 1 e i. The resultig structure is the same as for the home 26

28 Compared with the small coutry case, the oly differece is the terms of trade effect expressed as 1 e i. The equilibrium domestic price is lower tha the referece price Whe the domestic equilibrium price goes below the referece price, producers will experiece a loss. Followig the same idea as i sceario two i the small coutry case, the loss aversio part for the producers will eter the objective fuctio of the govermet. The other idividuals whose specific factors are ot used to produce this product are et buyers. The price decrease of this product will cotribute to a positive gai of et idirect utility. However, the positive gai i the loss-gai fuctio does ot add additioal utility gai to cosumers. I this case, the aggregate social welfare becomes: φ i=1 W(t s i ) = φl + φ i=1 Π i (t i, p w i ) + φ i=1 r(t i, p w i ) + φ S(t i, p w i ) Π i (t i,p i w ) i=1 + mi ( a i Nh ( π l i=1 ), 0) (44) a i N The welfare of the politically sesitive groups will be expressed as follows: i g H(t s i ) = a g l + a g i=1 r(t i, p w i ) + i g Π i (t i, p w i ) + a g i=1 S(t i, p w i ) + Π i (t i,p i w ) mi ( a i Nh ( π l i g ), 0) (45) a i N Pluggig equatios (44) ad (45) ito the govermet objective fuctio (equatio (9)) gives: Ω = a g l + a g i=1 r(t i, p w i ) + i g Π i (t i, p w i ) +a g S(t i, p w i ) Π i (t i,p i w ) i=1 + mi ( a i Nh ( π l i g ), 0) + φl + φ i=1 Π i (t i, p w i ) + φ r(t i, p w i ) a i N Π i (t i,p i w ) i=1 + φ i=1 S(t i, p w i ) + mi ( a i Nh ( π l i=1 ), 0) (46) a i N Substitutig the tariff reveue equatio (37) ito (46) yields: 27

29 Ω = (a g + φ)l + (φ + g i ) 1 Π i (t i, p w i ) + (a g + φ) { i=1(t i 1)p w i [d i (t i, p w i ) 1 y N i (t i, p w i )] + i=1 S(t i, p w i )} + mi ( i=1(φ + g i ) a i Nh ( π l Π i (t i,p w i ) i a i N ), 0 ) (47) Maximizig the above fuctio with respect to t i gives the first order coditio as follows: (φ + ps i )y i p i w + (φ + ps i )y i t i p i1 w + (a g + φ) [p i w M i (t i, p i w ) + (t i 1)p i1 w M i (t i, p i w ) + (t i 1)p i w M i pi w + (t i 1)p i w t i M i pi1 w d i (t i, p i w )p i w d i (t i, p i w )t i p i1 w ] (φ + ps i )a i Nh ( ) 1 a i N ( y ip i w y i t i p i1 w ) = 0 (48) Solvig the optimal trade protectio level by applyig equatios (41) ad (42) gives: (t i 1) = [g i a g +(φ+g i )h ( )] [a g +φ] y i p i w ( M i ) + 1 e i (49) Comparig equatio (49) with (43), the oly chage is the term i the umerator (φ + g i )h ( ), which is positive accordig to the characteristics of the loss aversio fuctio. This meas the protectio level is higher compared with the protectio level whe the equilibrium price equals the referece price. Whe trade protectio is higher, the domestic market price must be higher tha the world price. If the equilibrium domestic price goes lower tha the referece price, the world price is lower tha the referece price. Therefore, we ca coclude that the terms of trade do ot chage the result of Propositio 1, which could be expressed as follows: { [g i a g +(φ+g i )h ( )] [a g +φ] y i p i w ( M i ) + 1 e i } > { (g i a g) (φ+a g ) y i p i w ( M i ) + 1 e i } (50) The equilibrium domestic price is higher tha the referece price If the equilibrium price is higher tha the referece price, producers gai. Followig the argumets of Freud ad Özde (2008), whe the producers ow oe specific factor ad the 28

30 product experieces a icrease of price, the retur domiates the loss aversio term for producers. However, the et buyers whose specific factors do ot experiece price icreases will have loss aversio effects. The loss aversio term eters the govermet objective fuctio due to loss of cosumers surplus. I this sceario, addig the loss aversio term from cosumers perspective meas the stadard aggregate social welfare (equatio (11)) becomes: φ i=1 W(t s i ) = φl + φ i=1 Π i (t i, p w i ) + φ i=1 r(t i, p w i ) + φ S(t i, p w i ) mi ( φβ i Nh ( l=1 r(t i,p w i ) w l=1 r(ti,p i ) + l=1 S(t i,p w i ) i=1 S(ti,p w i ) N i=1 + ), 0) (51) Followig the same idea, the welfare of the politically sesitive groups becomes: i g H(t s i ) = a g l + a g i=1 r(t i, p w i ) + i g Π i (t i, p w i ) +a g S(t i, p w i ) mi ( β g i Nh ( l=1 r(t i,p w i ) w l=1 r(ti,p i ) + l=1 S(t i,p w i ) i=1 S(ti,p w i ) N i=1 + ), 0) (52) Substitutig the above two equatios (51) ad (52) ito the govermet objective fuctio (equatio (9)) yields: Ω = φl + φ i=1 Π i (t i, p w i ) + φ i=1 r(t i, p w i ) + φ S(t i, p w i ) mi ( φβ i Nh ( l=1 r(t i,p w i ) i=1 + w l=1 r(ti,p i ) + l=1 S(t i,p w i ) i=1 S(ti,p w i ) N a g i=1 r(t i, p w i ) + i g Π i (t i, p w i ) +a g S(t i, p w i ) mi ( β g i Nh ( l=1 r(t i,p w i ) i=1 + w l=1 r(ti,p i ) + l=1 S(t i,p w i ) i=1 S(ti,p w i ) N ), 0) + a g l + ), 0) (53) From tariff reveue equatio (37), the equatio (53) could be rewritte as: 29

31 Ω = (a g + φ)l + (φ + g i ) 1 Π i (t i, p w i ) + (a ps + φ) { i=1(t i 1)p w i [d i (t i, p w i ) 1 y N i (t i, p w i )] + i=1 S(t i, p w i )} + mi ( (φβ i + (ti w w 1 1)p i [di (t i,p i ) N y w i=1 i(t i,p i )] + l=1 S(t i,p w i ) i=1 S(ti,p w i ) N β g i )Nh ( l=1 r(t i,p w i ) ), 0) (54) We ca get the followig first-order coditio by usig Roy s idetity, Hotellig's lemma, ad relatioship betwee the domestic market price ad the iteratioal market price: (φ + g i )y i p i w + (φ + g i )y i t i p i1 w + (a g + φ) [p i w M i (t i, p i w ) + (t i 1)p i1 w M i (t i, p i w ) + (t i 1)p i w M i pi w + (t i 1)p i w t i M i pi1 w d i (t i, p i w )p i w d i (t i p i w )t i p i1 w ] (φβ i + β i g )Nh ( ) [ p i w M i (t i, p i w ) (t i 1)p i1 w M i (t i, p i w ) (t i 1)p i w M i pi w (t i 1)p i w t i M i pi1 w + d i (t i, p i w )p i w + d i (t i, p i w )t i p i1 w ] 1 N = 0 (55) Rearragig the above fuctio, we ca get the optimal protectio level by applyig equatios (41) ad (42): (t i 1) = [g i a g (φβ i +β i g )h ( )] [a g +φ+(φβ i +β i g )h ( )] y i p i w ( M i ) + 1 e i (56) Compared with case oe i a large coutry, the term (φβ i + β i g )h ( ), which takes a positive value accordig to the characteristics of the loss aversio fuctio, eters the umerator ad deomiator of the politically optimal trade policy. The decrease of the umerator ad the icrease of the deomiator lead to the ratio value decreasig. Thus the protectio level is lower tha i the sceario whe the equilibrium domestic price equals the referece price. I additio, whe the protectio level is lower ad the domestic market price is lower tha the referece price, the iteratioal market price must be lower tha the referece price. Fially, we ca coclude that the large coutry model does ot chage the 30

32 theoretical predictio of Propositio 2 whe cosiderig the terms of trade effect. This is illustrated i the followig iequality equatio: { [g g i a g (φβ i +β i )h ( )] g [a g +φ+(φβ i +β i )h ( )] y i p i w ( M i ) + 1 e i } < { (g i a g) (φ+a g ) y i p i w ( M i ) + 1 e i } (57) 5 Empirical test I this sectio, we ivestigate the cotto protectio patter i Chia for cosistecy with our theoretical model predictios. The cotto sector provides a ideal experimet to aalyze trade policy formatio motives of the govermet. Firstly, as illustrated i sectio 4.3.4, Chia s imports accout for virtually all of Chia s cotto trade activity. Secodly, Cotto is a special agricultural product, ad it makes up a relatively smaller share of expediture for poor producers, ad i most cases they are et sellers (Marti, 2009). This is exactly the same as our model assumptios. Durig price spikes periods, the icome of the cotto producer domiates the producers loss-gai utility. The Chiese govermet has set a cotto referece price 18 i each year from 2005 to For this reaso, we do ot have to estimate the referece price. This helps us test the effects of loss aversio ad referece depedece o trade policy formatio more accurately. Fially, the cotto producers i Xijiag beefit whe the govermet icreases the tariff (helped by the fact that the Chiese govermet subsidizes domestic trasportatio of cotto delivered from Xijiag to Southeast regios of the coutry). 5.1 Data sources I the empirical part, we apply motly, seasoal ad aual frequecy data to ru the empirical model. The mothly data cover 132 moths from May 2005 to December Freud ad Özde (2008) set the referece price as the average of the world price durig a particular period. 31

33 Trade protectio is measured by NRA 19 calculated usig the domestic cotto price ad the iteratioal market price. Chia cotto mothly prices are maily collected from the Chia Cotto Almaac from Jauary 2005 to October 2014, ad the data from November 2014 to December 2015 are compiled from the website of Chia Cotto. 20 The iteratioal cotto price is from the Natioal Cotto Coucil of America (NCCA) betwee Jauary 2005 ad December To measure the world cotto price i Remibi (RMB), the mothly exchage rate data are collected from the Board of Goverors of the Federal Reserve System ( ). 21 The referece price is the aual value set by the Chiese govermet at the begiig of each year. The referece price does ot stick to a fixed value, ad it icreased from Yua/To i 2005 to Yua/To i Cocerig seasoal fluctuatios, we aggregate the mothly data to the seasoal level ad add the harvest cycle as the cotrol variable measured by Si( ) ad Cos( ) fuctios. These data are created by combiig the value of π ad code umbers of the domestic price series. As a robustess check, a seasoal dummy variable is also applied to cotrol for the productio cycles. Meawhile, i order to test for it beig a politically sesitive product, the aual pael data method is implemeted by coverig 11 agricultural products ad the NRA data are from Aderso ad Nelge (2013). 5.2 The effects of political sesitivity o cotto trade protectio Oe predictio is that a politically sesitive product receives a higher protectio level relative to o-politically sesitive products. Trade protectio for differet types of products are measured by NRA. The products iclude cotto, poultry, fruits, rice, sugar, soybea ad wheat. Figure 5 illustrates the extet of trade protectio for six agricultural products excludig sugar. Cotto trade protectio is higher comparig with other agricultural products, 19 See Measurig distortios to agricultural icetives, Revisited Aderso et al. (2008) for more details of NRA idicator. 20 See 21 See 32

34 accordig to Figure The policy trasfers from taxig to subsidizig the cotto sector could be explaied by the theoretical model because of cotto s role as a politically sesitive product. Figure 5: Cotto compared with other agricultural products trade protectio I order to cofirm the theoretical predictio ad the role of politically sesitive products, we add two dummy variables to represet cotto ad sugar. Table 3 reports the empirical regressio results betwee politically sesitive product dummies ad the chage of border trade protectio by applyig pael data. From colum 1 to 5, the effect of politically sesitive cotto o the chages of its trade protectio is positive ad statistically sigificat at 1% level. Year dummy ad product dummy variables are all cotrolled i the five models. Cotto receives poit higher trade protectio measured by NRA whe cotrollig the iteratioal price as reported i colum 1. Whe addig the self-sufficiecy ratio (colum 2), 22 Sugar receives higher protectio tha cotto or other agricultural products. However, sugar is ot geographically produced i oe or a few regios. I the empirical part, we test whether sugar is a politically sesitive product. However, that hypothesis is ot supported by the ecoometrics (see Table 3). 33

35 productio value i GDP (colum 3), cosumptio value i total agriculture (colum 4) ad productio value i total agriculture (colum 5) ito the regressio as cotrol variables, the effect size is still aroud 0.01 ad statistically sigificat. Ulike for cotto, the empirical results do ot show ay sigificat effect of the dummy variable for sugar from colums 1 to 5 reported i Table 3. Table 3: Effect of politically sesitive products o the chages of protectio level Variables Protectio level (1) Protectio level (2) Protectio level (3) Protectio level (4) Protectio level (5) Dummy variable 0.023*** 0.011** 0.011** 0.010* 0.010** for cotto (0.003) (0.005) (0.005) (0.005) (0.005) Dummy variable for sugar (0.091) (0.005) (0.007) (0.007) (0.006) l(world price) *** *** *** *** *** (0.091) (0.069) (0.083) (0.087) (0.080) (0.290) (0.286) (0.306) (0.291) self-sufficiecy ratio productio value i GDP * (0.677) (0.587) (0.791) cosumptio value i total agriculture (0.477) productio value i total agriculture (0.662) Costat 0.034*** 0.037*** 0.035*** 0.034*** 0.032** (0.012) (0.012) (0.012) (0.013) (0.013) Year dummy Yes Yes Yes Yes Yes Product dummy Yes Yes Yes Yes Yes Observatios No. of products Adj. R Notes: 1) Robust stadard errors are reported i the paretheses; 2) * sigificat at 10%; ** sigificat at 5%; *** sigificat at 1%. 34

36 5.3 The effects of loss aversio o cotto trade protectio This sectio empirically tests the effects of loss aversio ad referece depedece o the variatios of trade restrictio i the Chia cotto sector, both whe the world price is lower tha the referece price ad whe it is higher tha referece price. Visual figure illustratio The relatioship betwee cotto trade protectio, the iteratioal market price, ad the referece price is illustrated i Figure 6. The author divides the period ito six, depedig o the level of world price compared with the referece price level. The blue dotted lie idicates the iteratioal market price ad the gree dashed lie represets the level of Chia cotto trade protectio. The red horizotal lie is the domestic referece price set by the govermet. Whe the world price is lower tha the referece price, the trade protectio level is higher, ad whe the world price is higher tha that of the referece price, the cotto protectio level is lower. This is cosistet with the theoretical model predictios summarized as Propositios 1 ad 2. Whe world price is lower, ad trade protectio is higher, compesatig protectio is triggered for producers (Freud ad Özde, 2008). Oe iterestig fidig is i the last period. Whe the world price cotiues to decrease, cotto protectio sharply decreases despite beig outside the compesatio period. That govermet behavior could be explaied by dimiishig sesitivity to loss. 23 The cotiuous decrease i the world price leads producers to adjust their platig ad ivestig behaviors. Fially, they accept the loss pheomeo due to the further decrease of the world price. Accordig to the productio data, total productio of cotto i Chia decreased from 30 millio 480-poud bales i 2014 to 23.8 millio a year later. 23 Dimiishig sesitivity meas that the margial value of gais ad loss decreases with their size. 35

37 Figure 6: Relatioships betwee world price, referece price ad trade protectio Visually, Figure 6 illustrates the theoretical predictios quite well betwee the trade protectio level, the referece price ad the world price due to loss aversio effects. However, that does ot tell us if the effects of loss aversio ad referece depedece are statistically sigificat. I the followig, the correlatio is tested usig the whole sample ad subsamples. Correlatio cofirmatio Table 4 reports the correlatio ad sigificace level betwee trade protectio ad the world price cotiget o the referece ad the magitude of loss aversio effects from cosumers ad producers durig world price upward spike ad dowward spike periods, respectively. The relatioship betwee the iteratioal market price ad cotto trade protectio level should be egatively correlated. The correlatio is calculated from the full sample ad subsample separately. Based o the calculatio of the full sample, the correlatio betwee Chia cotto protectio ad the iteratioal market price is which is sigificatly differet from zero at 1% sigificat level. Whe applyig the first half sample ad the other later half sample to the robustess checks for correlatio, the absolute value of the coefficiet 36