Middlemen, fair traders, and poverty

Transcription

1 J Eon Inequal (2016) 14: DOI /s Middlemen, fair traders, and poverty Nany H. Chau 1 Hideaki Goto 2 Ravi Kanbur 3 Reeived: 18 January 2013 / Aepted: 7 July 2015 / Published online: 7 August 2015 Springer Siene+Business Media New York 2015 Abstrat We propose a spatial model of produer market aess where loal middlemen reap market power due to math frition, and fair traders enter to present an alternative. The model features loation as a key determinant of the impat of fair trader entry on the market share of fair traders, the distribution of onsumer willingness to pay between middlemen and produers, and intra- / inter-regional poverty inidene. For governments who wish to minimize the poverty gap, our results support direting resoures to subsidize fair trade organizations, and/or to produers with no aess to markets, rather than to loal middlemen intermediaries. Keywords Middlemen Fair traders Poverty Market aess Eletroni supplementary material The online version of this artile (doi: /s ) ontains supplementary material, whih is available to authorized users. Nany H. Chau hy3@ornell.edu Hideaki Goto h-goto@iuj.a.jp Ravi Kanbur sk145@ornell.edu 1 Charles H. Dyson Shool of Applied Eonomis and Management, Cornell University, Ithaa, NY 14853, USA 2 Graduate Shool of International Relations, International University of Japan, Niigata , Japan 3 Charles H. Dyson Shool of Applied Eonomis and Management and Department of Eonomis, Cornell University, Ithaa, NY 14853, USA

2 82 N. H. Chau et al. 1 Introdution Middlemen, trading entrepreneurs who link the bakwaters of developing ountries to emerging markets nationally and espeially globally, seem to be universally reviled despite the eonomi servie they provide. Without their apital, speialized knowledge and information, onsumer demand in growing markets might be outside the reah of the small holder in the rural area, or of the home-based artisan in the urban slum. By bridging this gap, albeit for profit, surely they help to alleviate poverty? And yet it is this profit motive, and the laim that these middlemen make eessive profits beause of market power, that is at the root of muh of the onern. Thus, for eample, MMillan et al. (2004) study the ase of ashews in Mozambique, and report that ashew growers only reeive 40 % to 50 % of the border prie, even after border taes are allowed for. They go on to note: it is lear that the marketing hannels for raw ashew nuts remain imperfetly ompetitive. Farmers inomes are depressed not only by transport and marketing osts, but also by the market power eerised by the traders. (p 120). Many soial movements in developing ountries address themselves to providing an alternative hannel to the market for poor produers. For eample, the Self Employed Women s Assoiation (SEWA) in India has set up the SEWA Trade Failitation Center (STFC):...the artisan women though skilled in hand embroidery had to foribly migrate in searh of work, or undertake earth digging work. The trade of their old valuable embroidered trouso would typially our during distressed time. This would our with traders and middlemen sine the artisans had aess to a very limited market. Moreover they only had their traditional skills but still required produts whih they ould sell in the market. Lak of market information hindered artisans from building a strong relationship with the buyers. With the ineption of STFC, integrated supply hain mehanism was reated and the prodution beame organized. (SEWA Trade Failitation Center 2015). Some organizations in the fair trade business promise a number of outomes, inluding an improved outome for the impoverished produers of the produts being sold. 1 Aording to the Fair Trade Federation:...Fair trade organizations work diretly with produers, utting out middlemen, so they an keep produts affordable for onsumers and return a greater perentage of the prie to the produers. (Fair Trade Federation 2015). The jutaposition of middlemen trading for profit and fair traders with epliit goals to assist poor produers raises interesting and important analytial and poliy issues. What determines equilibrium share of produers served by middlemen and fair traders? Does fair trader entry reinfore the poverty impat of market aess by hanging the middlemen markup as well as the etent of market prie pass-through? What preisely is the impat of 1 Over 1.2 million individual farmers and workers in 58 produing ountries are involved in produing Fairtrade ertified produts in 2009 aording to the Fairtrade Labelling Organizations International. Fairtrade produts are sold in over 70 ountries, and in some national markets, Fairtrade aounts for as high as 20 % 50 % of total market share (Fairtrade Labelling Organizations International 2010a, 2010b).

3 Middlemen, fair traders, and poverty 83 fair traders on poverty when the full round of market reperussions, inluding entry and eit of middlemen and of fair traders, is taken into aount? Should a government interested in poverty redution subsidize the entry of fair traders, or should it perhaps subsidize the entry of middlemen? These are the questions to whih this paper is direted, and they are questions to whih we believe the literature provides only a partial answer. To do so, we work with a model that ehibits three features. First, we fous on situations where produers rely on middlemen to deliver their produts to market. As suh, prie pass-throughs depend ritially on the market power of the middleman, and the assoiated markup. Seond, we posit that math frition eists between produers and middlemen. The resulting information asymmetry about prie offers reeived by other produers indues endogenous middlemen market power and prie dispersion at the farm gate level even with free entry. Third, we eamine the equilibrium at eah point along a spatial ontinuum by allowing middlemen market power to vary endogenously aross loations. The resulting framework yields middlemen market power, dispersion in farm gate pries, and an endogenous prie pass-through that jointly depend on a produer s loation along the spatial ontinuum. Within this setting, we introdue fair traders motivated by a onern for poverty either on behalf of the final onsumers they serve via a loation speifi prie premium, or the preferenes of the fair traders themselves. These fair traders at as an alternative intermediary other than traditional middlemen linking markets and individual produers. 2 Like many models of voluntary ontribution and of altruisti behavior, this onern for poverty will be epressed in the form of prie shifter via a warm glow effet (Andreoni 1990), whih applies whenever the fair trader operates in loations where poverty is known to be pervasive. 3 Thus, this paper brings together two so far distintive areas of researh inquiry: (i) the role of middlemen and market power in determining prie behavior and as suh intraand interregional poverty in a spatial setting, and (ii) the role of fair traders in failitating prie pass-throughs, and in providing a market based mehanism for epressing onsumer onerns for the plight of marginalized produers along the loational ontinuum. The role of middlemen and market power in determining prie pass-throughs has been widely ommented upon (Arndt et al. 2000; HertelandWinters2005; Thapa and Pokhrel 2007). This paper ontributes to the literature by epliitly modeling math frition àla Mortensen (2003) due to farmers imperfet knowledge about prie offers by middlemen to produers in the same loation. In doing so, we depart from the multi-tiered marginalization approah used for eample in MMillan et al. (2004), and onsider instead a setting that an aommodate heterogeneity in the distribution of the eport surpluses between middlemen and produers within a given region. In addition, we also depart from the familiar math frition setting by introduing a spatial dimension to the model. Consistent with a number of reent empirial empirial studies, market equilibrium is haraterized by prie dispersion among otherwise idential produers given loation, and heterogeneity in prie distribution aross loations (Svensson and Yanagizawa 2009; Goyal2010; Mookherjee et al. 2013). 2 Lewin et al. (2004) finds that only about 20 % of global fair trade prodution apaity is sold at fair trade pries. The prevailing world prie plays a key role in determining farmers hoie between fair traders and loal intemediaries (Renard and Pèrez-Grovas 2007). 3 Eamples abound: The Fairtrade Foundation ooperates with growers in remote villages where the poorest offee farmers are typially loated (Global Ehange 2015). Eonomi Development Imports is another organization that works with loal artisans in the most remote parts of East and West Afria by providing aess to world markets (World Inquiry 2015).

4 84 N. H. Chau et al. Furthermore, we show that aross loations and aounting for produers operating in inreasingly remote areas, middlemen market power intensifies. 4 This endogenous variation in middlemen market power aross loation further ditates the etent of imperfet prie transmission. Indeed, we find that produers in more remote loations are more suseptible to an unequal division of the size of the market surplus. More generally, the funtioning of middlemen intermediaries has been shown to span a number of distintive ategories (Spulber 1996a), inluding (i) prie setting and market learing (MMillan et al. 2004), (ii) inventory holding and liquidity provision (Clower and Leijonhufvud 1975; Spulber 1985), (iii) suppliers and buyers mathing (Rubinstein and Wolinsky 1987; Spulber 1996b; Rauh and Watson 2004; Antràs and Costinot 2011), and (iv) performane monitoring and quality guarantees (Biglaiser 1993; Biglaiser Friedman 1994; Bardhan et al. 2013). The introdution of fair traders in our model ontributes to this literature by introduing heterogeneity in motivations between ompeting middlemen as an additional line of researh, allowing for differential poliy presription targeting middlemen depending on motivations. Furthermore, we show that the introdution of fair traders impat middlemen market share as well as priing behavior, and the strength of suh impats depends in partiular on produer loations. The plan of the paper is as follows. Setion 2 sets up the framework and equilibrium with middlemen only. Setion 3 introdues fair traders and haraterizes equilibrium when both types of intermediaries an enter. Setion 4 takes up the poliy questions who eatly should the government subsidize? Setion 5 offers a number of etensions of the model presented. Setion 6 onludes. 2 The model and equilibrium with middlemen 2.1 The basi setup We onsider a spatially dispersed eonomy, in whih prodution takes plae at a range of distanes [, ] away from a transport hub. At eah loation there is a large number of idential produers, N. Eah produer has a unit of output for sale. Any output bound for market must be transported to the hub. For eah unit sold, let p be the unit market prie, and p p the farm gate prie. Output not bound for market an be used for ownonsumption, and the revenue equivalent of any output not marketed is. We assume that 0 <p for there to be a potential for gains from trade. 2.2 Middlemen: the Bertrand benhmark Transportation is arried out by middlemen, who inur a loation-speifi transportation ost t per unit output. 5 If free entry and fritionless Bertrand prie ompetition prevail among middlemen, the equilibrium produer prie p t is stritly loation-speifi at 4 Where markets are ompetitive with minimal entry osts, etortionary middlemen markups have not been found (Fafhamps et al. 2005; Osborne 2005; Minten et al. 2013;deJanvryetal.2015). But in remote areas, studies show rural households onfront higher transport ost and middlemen with market power (Goetz 1992; Seton et al. 1991; Niita 2004; Hertel and Winters 2005; Hertel and Winters 2005; Fafhamp and Minten 2012). 5 A Samuelson ieberg type transportation ost an similarly be imposed, by assuming for eample that t = τp,τ>0, without affeting the qualitative impliations of the model.

5 Middlemen, fair traders, and poverty 85 eah, as long as p t is greater than the reservation value. 6 All middlemen thus earn zero profit: they sell output at prie p at the hub, after having inurred t as transportation ost, and p t as payment to produers. This simple framework yields a stark set of preditions regarding the volume of marketed output, the distribution of inome between middlemen and produers, as well as the inidene of poverty among produers. In terms of marketed volume, all produers loated at (p )/t o + market all of their output sine p t whenever o +. In what follows, we assume that there eist at least some loations where bringing output to market generates stritly positive surplus p t >0. Equivalently, we assume that < o +. To furthermore aommodate the possibility of inomplete overage and loation as a binding onstraint on sales for at least some produers, we assume in addition that o +. In terms of produer pries, Bertrand ompetition implies a perfet one-to-one passthrough of the net market prie p t to produers, and the full market surplus p t at eah loation, aounting for opportunity ost, is aptured by the produer. Consequently, inequality in the distribution of revenue among produers is a stritly interregional phenomenon. At any given loation o +, all produers earn identially p t. Aross loations, produer prie is dereasing in. For > o +, market prie p is insuffiient to over the full ost t +, and produers reeive the own onsumption value per unit. Let p denote the minimal produer prie required to sustain private onsumption at a level no less than an eogenously given poverty line, with p > so poverty among at least some produers is a real possibility. With Bertrand ompetition among middlemen, poverty inidene as measured by the share of produers living under the poverty line is disontinuous along the loational ontinuum: for (p p)/t o p (< o + ), no one is poor, but immediately thereafter, every produer is poor. In summary, there are two ritial threshold loations, o p and o +.For p o, aess to market at prie p t ompletely eradiates poverty. For (o p, o + ), unit revenue lies stritly between and the poverty line p. Produers at loations even more distant than o + from the hub have effetively no aess to markets. Figure 1 summarizes, and shows the equilibrium produer prie along the loation ontinuum. 2.3 Middlemen market power We depart now from the Bertrand assumption, where every produer is perfetly aware of eah and every middlemen prie offer, and onsider instead an arguably more realisti senario in whih there is math frition between produers and middlemen (Mortensen 2003). We are interested in eploring how middlemen market power impats prie information arrival rates, middlemen profit margins, and the distribution of realized produer pries. In our spatial setup, we an also asertain differenes in prie information arrival, middlemen profit margin, and the distribution of realized produer prie aross the spatial ontinuum as middlemen market power hanges endogenously. Speifially, math frition arises whenever produers are aware of only a partial set of prie offers made by middlemen in a given loation. Let the likelihood that a produer omes aross z = 0, 1, 2,... offers be given by a Poisson distribution with parameter λ = M /N, 6 This is onsistent with Fafhamps et al. (2005) who find evidene supporting onstant returns to sale in agriultural trade inluding transportation ost.

6 86 N. H. Chau et al. p * t p * t p p o o Fig. 1 Produer prie and loation with Bertrand ompetition or, Pr(z; λ ) = e λ λ z /z!,7 where M denotes the endogenous number of middlemen in loation with a prie offer p. This formulation allows the probability distribution of prie information arrival to be loation-speifi, and the average number of prie offers a produer in loation reeives is equal to λ. λ measures the loation-speifi middleman market power. Whether produers in more remote loations tend to reeive more or less prie information is thus endogenously determined via λ as a funtion of in our setup. In this setting, eah of the M middlemen proposes an offer to one of the N produers hosen at random. Let F (p) be the umulative distribution funtion of middlemen prie offers at loation. Eah produer ranks any and all offers reeived this way, aepts the highest offer, and rejets the rest. Sine the distribution of eah suh prie offer is F (p), the umulative distribution of the maimal offer reeived is: e λ λ z H (p) F (p) z z! z=0 = e λ (1 F (p)). (1) With math frition, produers are only aware of a subset of prie offers in eah loation, and thus the middleman margin p t p need not be driven to zero in equilibrium. Indeed, the likelihood that any prie offer p out-ompetes all other offers reeived by a randomly seleted produer is given by H (p) in Eq. 1. For any given prie offer p, 7 As is well-known, the Poisson distribution is the limit of the binomial distribution with parameter λ as the number of trials approahes infinity.

7 Middlemen, fair traders, and poverty 87 therefore, middlemen epeted profit is H (p)(p t p) K,whereK 0isafied searh ost per produer ontated. Eah middleman s problem is simply: = ma p H (p)(p t p) K (2) The solution to Eq. 2 gives the following equilibrium prie offer distribution: F (p) = 1 λ ln ( p ) t 1 p, f (p) = t p λ (p t p). (3) Naturally, F () = 0 for no produer will aept a prie offer less than. Meanwhile, the highest prie offer given aross all middlemen p + at loation an be found by setting F (p + ) = 1, or equivalently: p + = (1 e λ )(p t) + e λ, (4) a weighted average of the net market prie (p t), and the domesti value (). The relative weights depend only on the etent of middlemen market power, λ.asλ, F (p) puts unit mass on lim λ p + = p t.however,asλ 0, p + tends to the domesti prie. Let there be free entry of middlemen in eah loation. Doing so drives epeted profits to zero, and implies that the realized distribution of produer pries for all p, is given by { } K H (p) = min p t p, 1. (5) In partiular, H () = min{k/(p t ), 1} gives the fration of produers who are left out of markets, and H (p) H () the fration for whom output is bound for market, and who feth a prie at p or less from their middlemen. The maimal prie offer is p + = ma{p t K,}. From Eqs. 1 and 5 evaluated at p =, endogenous entry implies equilibrium market power at: λ = M /N = ma{ln((p t )/K), 0}. (6) Equation 6 gives the equilibrium market power as parameterized by λ as determined by the interplay between searh ost K, and the market surplus p t. Thus,ifK 0, middlemen market power vanishes as λ and H (p) puts positive mass only on p = p t.nowfork > 0, equilibrium λ = M /N inreases with market prie p refleting more intense ompetition among middlemen. Furthermore, λ dereases with distane refleting a deepening of middlemen power along the loation ontinuum. In the end, with positive searh ost, λ is positive only in the range (p K)/t +. In omparison with the Bertrand outome with no math frition, o + = (p )/t > +, produers in loations ( +, o + ] are no longer servied by middlemen, and as suh they do not have aess to market. 2.4 Farm gate prie and pass-through Reall that in the absene of math frition, all farmers in loation reeive the same prie p t. By ontrast, from Eq. 5, math frition additionally gives rise to loation-speifi

8 88 N. H. Chau et al. dispersion of realized farm gate pries H (p). The epeted farm gate prie onditional on reeiving at least one aeptable prie offer (Ep ) an be epressed as: Ep p + pdh (p) 1 H () = (1 γ )(p t) + γ <p t. (7) Ep is thus a weighted average of the net market prie (p t) and the domesti value, where γ λ e λ /(1 e λ ).Sinep t represents net market surplus, the share that goes, on average, to middlemen is thus: p t Ep p t = γ λ e λ 1 e λ γ thus aptures the etent of unequal division of the market surplus between middlemen and produers. Evidently, this division depends systematially on middlemen market power λ, and as suh the division of surplus varies endogenously along the loational ontinuum. From Eq. 6, it an be verified that γ rises with distane for middlemen market power strengthens with distane. Produers at farther away loations are aordingly left with a smaller slie of the market surplus. Using Eqs. 6 and 7, the average farm gate prie with endogenous entry an also be epressed as: ( p Ep = p ) t p t t K ln K p t K. (8) The epeted middlemen markup ( p p ) t p t t Ep = K ln K p t K > 0 is stritly positive whenever the searh ost is stritly positive K > 0. The size of this markup, aounting for endogenous entry, is stritly inreasing in the net market prie p t. Intuitively, a higher net market prie enourages entry (λ ) and simultaneously inreases the number of middlemen who fail to strike a suessful math M N(1 H ()) = N(λ (1 e λ )) as they are outbid by other middlemen. To justify this risk, the markup p t Ep for those who sueed in striking a math must also rise in tandem. Using Eq. 8, the average revenue of all produers at, regardless of whether they sueed in marketing their produts, an be epressed simply as the weighted average: ER = (1 H ())Ep + H () = p t (1 + λ )K. (9) where λ = ln(p t ) ln K from Eq. 6. The responsiveness as measured by ER (p t) = 1 K p t = 1 e λ < 1 (10) gives the etent of prie pass-through along the spatial ontinuum. From Eq. 10, the revenue of the average produer at loation rises less than one for one with net market prie preisely beause middlemen market power as aptured by λ. The etent of this imperfet pass-through due to middlemen market power worsens endogenously with distane from

9 Middlemen, fair traders, and poverty 89 * ER, p, p t * ER p t (1 ) K * p p t K p * t p p p o o Fig. 2 Average produer revenue with endogenous middlemen market power Eq. 6, all the way up until the point is reahed where + and thus λ = 0. Produers loated here and beyond are by definition untouhed by fores of markets. Figure 2 illustrates, and shows the range of middlemen prie offer [, p + ] at eah loation, as well as the average revenue of all loal produers along the loation ontinuum. 2.5 Intra- and inter-regional poverty With prie dispersion both within region through H (p), and aross regions as H (p) varies with, the inidene of poverty has both an intra- and an inter-regional dimension. These are shown in Fig. 3, for three suessively more remote regions (from H 1 to H p and then to H 2 ), and aordingly three produer prie distributions that an be rank ordered in the sense of first order stohasti dominane. Define P,α m and P,α as the poverty of produers who gain aess to markets through middlemen, and the poverty of all other produers at loation. We adoptthe Foster-Greer- Thorbeke (1984) poverty indiator: 8 P,α = ( p p ) α min{ p,p + } ( p p p ) α dh (p) 1 H () = P m,α (11) where ( p p)/ p is the poverty gap ratio, and α 0 parameterizes the etent of poverty aversion. As shown in Eq. 11, of the two groups of produers living under the poverty 8 As a poverty indiator, the Foster-Greer-Thorbeke measure represents a lass of poverty indiators that satisfies the subgroup onsisteny aiom. This aiom guarantees that if poverty of some groups have hanged and others have not, then poverty of the hanging subgroup and total poverty move in the same diretion. This property is partiularly relevant in our ontet where total eonomy-wide poverty is both inter regional and intra-regional.

10 90 N. H. Chau et al. H n ( p) 1 3 H p H 1 H Hinterland Hub H ( p) K /( p * t 1 p) H ( p) K /( p * t p p) Fig. 3 Produer prie distribution p p line, the first group is poor beause they fail to bring output to market beause no offers greater than the reservation value arrives. The number of these produers is NH (). A seond group remains poor despite aess to market for they reeive a prie lower than p. The number of this seond group of produer is N(H (min{ p, p + }) H ()). Between the two, produers who fail to bring output to market are poorer sine P,α P m,α whenever α 0. Naturally, overall poverty at loation an be epressed as the weighted sum: P,α = H ()P,α + (1 H ())P m,α. (12) As long as some produers are non-poor, or p + > p, a small inrease in the market prie p, all else equal, dereases the share of produers with no aess to markets H () = K/(p t ) and thus alleviates poverty due to lak of market aess from Eq. 5. The same inrease in market prie p also dereases poverty among produers who do have aess to markets, for the distribution of prie offers H (p) = min{k/(p t p), 1} improves with p from Eq. 5. Furthermore, as ompetition heightens between middlemen (an inrease in λ ) either beause of inreases in p, or a redution in K, λ as K/(p t ) 0 approahing the Bertrand outome. At the limit, therefore, the first soure of poverty H () vanishes with the disappearane of math frition as long as the net market prie p t eeeds the poverty line p, while the measure of the seond H ( p) H () likewise approahes zero, sine no middlemen an offer less than the full p t and get away with it in the absene of math frition. Inter-regional differenes in poverty arise for two reasons in our setup. First, remote loations are naturally disadvantaged for the net market prie (p t) is lower there. Seond, remote loations are in fat made doubly worse off for middlemen market power

11 Middlemen, fair traders, and poverty 91 also deepens along the loational ontinuum from Eq. 5. Taken together, poverty worsens with distane, and indeed from Eq. 13, the poverty indiator an be re-epressed to reflet diretly the impat of loation on poverty: P,α = ᾱ p p ( p p p ) α 1 K p t p dp. for loations where there are at least some produers who are non-poor. The threshold distane, all it p, beyond whih produers are universally living below the poverty line is reahed when the maimal prie offer p + an no longer over p, or equivalently min{ p, p + }=p+.fromeqs.4and 6, p ( p p e λ (p t ) ) /t = (p p K)/t (13) p p = o p. t Consequently, middlemen market power deepens the inidene of poverty along both the intensive (P,α > 0) and etensive ( p < p o ) margins (Fig. 2). For > p, all produers are poor, and thus, P,α = ᾱ p p + ( ) p p α 1 ( K p p + ) α p p t p dp + p and P,α ontinues to worsen with distane (say from H p to H 3 in Fig. 3). Heneforth, we assume that p eeeds at least the lower bound, in order to eamine the set of fators that effet hanges along the etensive margin. We have so far demonstrated loation as a key determinant of middlemen market power. In addition, we have demonstrated the impliations of suh a link in terms of markups, pass-through, and intra-and inter-regional poverty. In this ontet, how does the introdution of fair traders impat middlemen market power along the loational ontinuum? And what about the related issues of markup and pass-through, as well as intra- and inter-regional poverty? 3 Fair traders There are many different ways in whih fair traders ontrat with produers on the ground. Some organizations ontrat diretly with produer ooperatives (Develtere and Pollet 2005). Others alternative trading organizations ontrat diretly with individual produers (Abufarha 2013; FairTradeUSA2012). As the ase may be, the N produers at eah loation an be interpreted as N produer ooperatives or N individual produers in a similar loation. Their ability to ommand a higher prie depend on the prie offers they reeive. As before, math frition ditates the likelihood of prie offer arrival to eah produer / produer ooperative, depending on middlemen market power whih will now need to be adjusted to reflet the presene of fair traders. Speifially, like profit-maimizing middlemen, fair traders serve produers by bringing output to markets at transportation ost t. But as the various eamples in the introdution demonstrate, unlike profit-maimizing middlemen, fair traders are additionally motivated by loation-speifi fators that impat poverty and aess to the world market among

12 92 N. H. Chau et al. produers. We inorporate these onerns by a loation speifi prie premium p n p > 0, where p n is the valuation that fair traders put on striking a ontrat with a produer at loation. 9 Our restrition on p n p is mild, and our objetive is simply to apture a onern by fair traders or final onsumers for produers in loations otherwise haraterized by high markups and / or isolation from world markets. Speifially, we assume p n p = a n + a where a 0 aptures the overall strength of the poverty and loational onerns of the fair trader, while a n aptures loationally invariant onsumer preferene for partiular attributes assoiated with fair trade produts. 10 In what follows, we further assume that the dependene of the prie premium on is not overly strong, so that the net market prie aounting for transportation osts, t, ontinues to indiate remoteness as a deterrene to trade, or equivalently p n t = p + a n (t a) is dereasing in. 11 In terms of fair trader fied osts, ontrating with produers an be more epensive relative to loal middlemen due for eample to fair traders ost disadvantage relative to longstanding middlemen, and other osts of monitoring and ertifiation required to justify the final onsumer prie premium. 12 We reflet these onsiderations by assuming that the fied ost appliable to fair traders K n is stritly greater than K. For produers, ontrating with fair traders an be ostlier relative to ontrating with loal middlemen either beause of the fied ost required to join and / or maintain a ooperative, to set up ontrats, or to abide by internationally reognized produt or environmental standards (Loureiro and Lotade 2005). Thus, let d n 0 represent the fied ost of ontrating with fair traders faing eah produer. 13 The ost of any produt standards required by middlemen will be normalized at zero here. Furthermore, let ρ be the prie offer, and p the prie offer net of the ost of ompliane, or in other words, p = ρ d n. Due to the fied ost of ontrating d n, the atual ost to the fair trader based on a net prie offer p is equal to ρ = p + d n. 14 For notational eonomy, we heneforth denote the differene between the additional ost of fair trade prodution and onsumer preferene for fair trade attributes d n a n as δ n. δ n may be positive or negative depending on the relative importane of ost and onsumer preferene onsiderations assoiated with fair trade prodution for eah produer. The ombined (money equivalent) gains to a fair trader who serves a produer loated at, and offers a net prie offer p (aounting for the ost of ontrating), thus ontinues to be loation-speifi p n dn t p = p + a δ n t p. 9 Empirial work on the willingness to pay for fair trade produts is a nasent field of researh. The importane of the poverty impat of the program on onsumer s willingness to pay, as we have assumed here, has been demonstrated by Basu and Hiks (2008, 2009) in a diverse set of eperimental settings, where test subjets ompare different versions of the fair trade offee label with differing degrees of promised impat on grower wellbeing. 10 There are often other provisions suh as in the fair trade inluding the adoption of sustainable praties or organi farming. 11 The linearity of the prie premium in assumed here an be easily relaed to aommodate any monotonially inreasing funtion a() without affeting the results to follow, as long as t a is inreasing in and distane ontinues to deter trade. 12 The opposite senario where K n <Kan be easily inferred from the analysis to follow. 13 Meanwhile, ontrating with fair traders an also lead the way to skill aquisition, and other tehnology transfers (Abufarha 2013; Fair Trade USA 2012). In these ases, the parameter d n introdued here will then take on a negative value. 14 Reall that total output is normalized at unity, and thus p gives the total revenue of a produer.

13 Middlemen, fair traders, and poverty 93 In what follows, we assume that a δ n > 0forall [, ], in order to reflet a stritly positive prie premium assoiated with fair trade produts. Otherwise, the baseline solution of the model with no fair traders trivially applies. 3.1 Equilibrium with middlemen and fair traders Like the baseline senario with no fair traders, produers / ooperatives have information on only a subset of the net prie offers made by middlemen and fair traders. They make ontrat deisions by omparing all net prie offers p reeived inlusive of those from fair traders and middlemen, selet the best, and rejet the rest. Denote F n (p), p, as the umulative distribution of prie offers inlusive of both fair traders and middlemen. Let H n(p) = e λn (1 F n(p)) be the umulative distribution of the realized produer prie distribution, where λ n is now the ratio (mn + Mn )/N, andmn and Mn respetively denote the endogenous number of fair traders and middlemen. The problem of fair traders is otherwise similar to that of a middlemen: they aount for the net (prie premium augmented) gains from serving a produer (p n dn t p), adjusted appropriately to reflet the likelihood of suessfully striking a math, H n (p), in the fae of ompeting prie offers from other middlemen: π n = ma p H n (p)(pn dn t p) K n (14) n = ma p H n (p)(p t p) K. (15) Sine the premium adjusted gains from a math is higher for a fair trader p n dn t p = p + a δ n t p>p t p for any given prie offer, it is straightforward to show that the equilibrium range of fair trader prie offers, if non-empty, is always higher than that of profit-maimizing middlemen. 15 Put another way, the entry of fair traders effetively onfines profit maimizing middlemen to servie produers who are not mathed with fair traders. Aordingly, let ˆp be an endogenous prie offer threshold, dividing the range of middlemen and fair trader prie offers. F n( ˆp ) thus gives the share of prie offers from private profit maimizing middlemen (M n/(mn +Mn )), and 1 F n( ˆp ) the share from fair traders. For any p ˆp,Eq.15 implies F n (p) = 1 ( p ) t λ n ln p (16) t p 15 To see this, suppose that p n and p are solutions to Eqs. 14 and 15 respetively. By virtue of profit maimization, it must be the ase that p n dn t p n p n dn t p H n ( p) H n ( p n ) p t p n p t p, with It follows that p n p. a δ n 0.

14 94 N. H. Chau et al. and otherwise with p> ˆp for fair traders, Eqs. 15 and 16 give F n (p) = F n ( ˆp ) + 1 ( p n λ n ln t ˆp d n ) p n. (17) t p dn The threshold ˆp is determined via Eq. 17 as soon as the share of middlemen at loation, F n ( ˆp), is known. This important share is endogenized here by way of simultaneous endogenous entry of profit maimizing middlemen and fair traders respetively: e λn (p t ) = e λn (1 F n ( ˆp )) (p t ˆp ) = K, (18) e λn (1 F n ( ˆp )) (p n dn t ˆp ) = e λn (1 F n (p)) (p n dn t p) = K n (19) where the marginal fair trader λ n (1 F n ( ˆp )) = m n /N equates the prie premium augmented epeted market surplus e λn (1 F n( ˆp )) (p n dn t ˆp ) and the ost of entry K n from Eq. 20, and the marginal middlemen λ n = (mn + Mn )/N then equates the epeted market surplus (e λ (p t ))withtheostofentryk from Eq. 19. Thus for interior solutions F n( ˆp ) (0, 1), the threshold prie offer dividing middlemen and fair traders is: ˆp = p t (a δn )k 1 k where k denotes the ratio of fied osts K/K n < 1. Clearly, the higher the prie premium a, the narrower the range of middlemen prie offers [, ˆp ]. Meanwhile, the greater the fair trader s (fied) ost disadvantage K n /K = 1/k, the higher the middlemen s maimal prie offer. Furthermore, the higher the fied ost assoiated with fair trade prodution relative to onsumer gains δ n, the wider the range of middleman prie offers. These show interestingly the tendeny for middlemen to optimally harge a higher markup on average to produers they ontinue to serve, the higher the fair traders prie premium. At the other end of the prie offer spetrum, the maimal fair trader prie offer is given by: (20) p n+ = p + a δ n t K n (21) from Eq. 20 evaluated at F n(p+ ) = 1. The larger the prie premium, the higher of ourse will be the maimal fair trader offer in equilibrium. From Eqs. 18 and 19, the number of fair traders (m n ) and the number of middlemen (M n ) at an interior equilibrium are respetively, ( a δ m n = Nλn (1 F n n ) ( ˆp )) = N ln K n (22) K ( p M n ) + mn = Nλn = N ln t. (23) K Thus, the equilibrium number of middlemen rises with market prie p. This is onsistent with evidene in Renard and Pèrez-Grovas (2007, pp. 143), who find that there are opportunisti growers who swith in favor of loal intermediaries in periods of high world pries. Making use of Eqs. 1, 16 and 17 as well as Eqs. 22 and 23, the realized prie distribution assoiated with middlemen mediated trade p ˆp is { } H n (p) = min K p t p, 1

15 Middlemen, fair traders, and poverty 95 and for p> ˆp assoiated with fair trader mediated trade, { H n (p) = min K n } p n dn t p, 1. It remains to be shown whether the threshold prie ˆp is an interior solution in the range of pries so that both middlemen and fair traders o-eist, or if in fat equilibrium aommodates either middlemen or fair traders. We turn to these questions net. 3.2 Farm gate prie and pass-through Depending on the prie premium, fied osts, as well as the market surplus, it follows from Eqs. 22 and 23 that there are three possible equilibrium onfigurations, respetively when there eist only middlemen, only fair traders, and when the two oeist: Proposition 1 By introduing fair traders with prie premium a t(k n + δ n K)/(p K), produers are served 1. entirely by middlemen at loations losest to the hub <(K n + δ n K)/a 2. by a mi of both middlemen and fair traders, for [(K n + δ n K)/a,[(1 k)(p ) + δ n k]/(ka + (1 k)t)) 3. entirely by fair traders, for [[(1 k)(p ) + δ n k]/(ka + (1 k)t), (p K n δ n )/(t a)]. Produers at even farther away loations are not served by either middlemen or fair traders. We relegate the details of the proof to the Online Appendi, and desribe here the intuitions behind the lassifiation shown in Proposition 1. The proposition begins with a requirement that fair trader profit (p + a t p δ n K n ) is higher than that of middlemen profit (p t p K) in at least some loations where middlemen operate + (p K)/t. This gives the ondition a t(k n + δ n K)/(p K) as shown. Now, at loations suffiiently lose to the hub, fair traders fail to out-ompete middlemen if the prie premium a is too small to over the added fied ost K n K. Region I of Fig. 4 illustrates all (, a) ombinations onsistent with this equilibrium, and 1 (a) (K n +δ n K)/a denotes the threshold distane, depending on a, beyond whih at least some produers will be served by fair traders in equilibrium. At the other etreme, a no-middlemen equilibrium applies at suffiiently remote loations ( > [(1 k)(p ) + δ n k]/(ka + (1 k)t) 2 (a)), but not too remote (< (p K n δ n )/(t a) n+ (a)) for even fair traders annot make ends meet in these loations. These two restritions are illustrated in Fig. 4 by region III inluding all ombinations of (, a) bounded respetively to the left and right by shedules 2 (a) and n+ (a). Reall from Eq. 6 that + denotes the threshold loation beyond whih there will be no more middleman in an equilibrium in the absene of fair traders. Thus, as shown in Fig. 4, region III an be further divided into two parts. The first part [ 2 (a), + ) represents all loations wherein fair traders take over and middlemen are ompletely displaed. The seond inlude all other loations where fair traders now serve as brand new links to markets where previously none eisted.

16 96 N. H. Chau et al. a 1 ( a) 2 a ( ) (a) n I II III t ( K n n * K ) /( p K ) Fig. 4 Middlemen and fair traders I: middlemen only II: oeistene III: fair traders only In between these two etremes, middleman and fair traders oeist. Region II in Fig. 4 illustrates this intermediate range. Depending on the equilibrium ombination of middlemen and fair traders in a partiular loation as shown in Proposition 1, we an use Eqs to determine the impat of the introdution of fair traders on the farm gate prie distribution, the orresponding average farm gate prie, and the responsiveness of the average loal produer revenue to market prie hanges. In partiular Proposition 2 For all n+ (a), the introdution of fair traders with prie premium a t(k n + δ n K)/(p K) 1. gives rise to a first order stohastially dominating shift in the produer prie distribution as H n(p) H (p), 2. raises the average farm gate prie sine E n p Ep, and 3. improves the responsiveness of average loal produer revenue, inlusive of those who have aess to markets and those who do not, to market prie hanges sine E n R / (p t) ER / (p t). Proof See the Online Appendi. To see what drives these results, we will utilize the region by region summary of the impat of fair trader entry in Fig. 4. We will also referene Fig. 5, in whih a family of realized prie distributions is shown going from distanes nearest to the hub, to more remote loations deeper into the hinterland, based on the realized prie distribution derived in Setion 3.1. Starting with Region I of Fig. 4, where there eists no fair traders in equilibrium, the distribution of realized pries is naturally unaffeted by the entry of fair traders. Consequently, the average produer prie, the responsiveness of average produer revenue to world prie

17 Middlemen, fair traders, and poverty 97 H n ( p) 1 4 H 3 H H 2 H 1 H n H ( p) n * 2 2 n K /( p a t p ) n H ( p) n * p p n K /( p a t p ) n H ( p) K /( p * t p p) H n ( p) K /( p * t 1 p) p p Fig. 5 Produer prie distribution with fair traders and middlemen hanges, along with intra-regional poverty, all remain untouhed by the possibility of fair trader entry. In Fig. 5, H 1 is eample of suh a prie distribution. In region III of Fig. 4, there are only fair traders here where a is relatively higher in loations suffiiently far away from the hub. In terms of prie distribution, H 4 in Fig. 5 is one suh eample. We know from earlier disussion in Eqs. 6 and 8 that higher market pries lead to a first order stohastially dominating shift in the farm gate prie distribution onsistent with an improvement in the average farm gate prie. Furthermore, we also know from earlier disussion following (6) thatasλ rises with higher pries, the responsiveness of produer revenue with respet to hanges in p t rises as well. Thus, the entry of fair traders shifts the distribution of farm gate pries, raises the average farm gate prie, and improves the responsiveness of farm gate prie to market prie hanges. Finally in region II of Fig. 4, fair traders and middlemen o-eist. For eah loation, the prie distribution at the range of pries that middlemen ontinues to offer remain stritly unhanged. However, the prie distribution at the range of (higher) pries the fair traders offer has hanged to represent a first order stohastially dominating shift onsistent with higher prie offers by fair traders. H 2 and H 3 in Fig. 5 are eamples of this mi. These hanges implies that the entry of fair traders raises the assoiated average farm gate prie. What is partiularly interesting is that in this region where fair traders and middlemen ompete, fair trader entry displaes eisting middlemen. In fat, from Eq. 6 λ n = mn + Mn N = ln(p t ) ln(k) = M N = λ. All else onstant, therefore, the entry of one more fair trader has the effet of diretly displaing a middleman. While this result is of interest in its own right, it also suggests that the responsiveness of the average farm gate prie to market prie p t remains unhanged despite the introdution of fair traders sine market power as parameterized by λ n remains unhanged. In summary, Region I in Fig. 4 ehibits invariane of farm gate prie distribution despite the possibility of fair trader entry. In the intermediate Region II overall produer revenue

18 98 N. H. Chau et al. rises thanks to the entry of fair traders, but the degree of market prie pass-through remains untouhed. The final region III enompasses loations where fair traders ompletely displae middlemen, and loations in whih fair traders failitate trade where isolation from world markets was the norm. At these loations, both overall produer revenue, and the etent of market prie pass-through improve with fair traders. Proposition 2 presents a summary of these points. 3.3 Intra- and inter-regional poverty inidene The poverty impliations of fair trader entry are illustrated in Fig. 5. Of partiular interest is the utoff distribution H. For loations loser to the hub relative to H, the poor are served by middlemen only, and fair traders offer a prie higher than the poverty line p, or ˆp p p p kδ n /(1 k) n (a). (24) t + ak/(1 k) Sine the poor remains untouhed by fair traders, the introdution of fair traders at these loations naturally leave poverty unhanged at: P,α n = ᾱ p ( ) p p α 1 K p p p t p dp = P,α. By ontrast, for loations further into the hinterland than H, either some (for [ n (a), 2 (a)), or all of the poor ( [ 2 (a), n+ (a))) will be served by fair traders. At all loations where there are at least some non-poor, p p n+, poverty delines with the entry of fair traders: P n,α = ᾱ p ᾱ p p p ( ) p p α 1 H n p (p)dp ( ) p p α 1 H (p)dp. = P,α p whih follows sine H n (p) first order stohastially dominate H (p) from Proposition 1. In the Online Appendi, we disuss all of the remaining ases, depending on (i) whether all produers are poor, and (ii) whether middlemen and fair traders o-eist. But in all: Proposition 3 At the national level, averaging aross all loations [, ], the introdution of fair traders 1. redues overall average poverty (P,α n P,α)ifa t(k n + δ n K)/(p K), and remains unhanged otherwise, 2. pushes bak towards the hinterland the threshold loation beyond whih all produers are poor if in addition a t(k n + δ n K)/(p p K); otherwise the threshold loation remains unhanged. Proof See the Online Appendi. To see the intuition behind the seond part of the proposition, note that the maimal fair trader prie offer p + a t δ n K n from Eq. 21 is below the poverty line p p δ n K n t a np (a) p = p p K t (25)

19 Middlemen, fair traders, and poverty 99 a 1 ( a) 2 a ( ) (a) n (a) (a) np n n nonepoor II, m, somepoor n somepoor II, m, allpoor n somepoor III, n allpoor III, I m, somepoor t ( K n n * K ) /( p p K ) n allpoor II, m, allpoor t( K n n * K ) /( p K ) I m, allpoor p Fig. 6 Intra- and Inter- regional poverty I: middlemen only II: oeistene III: fair traders only if and only if a t(k n + δ n K)/(p p K) as stated based on Eq. 13. With the help of Eqs , Fig.6 illustrates the full array of possibilities in our model of middlemen and fair traders. Depending on the size of the premium a, there are two main lasses of outomes. 16 The first lass involves a at relatively low levels, between t(k n +δ n K)/(p K) and t(k n + δ n K)/(p p K). In this range of prie premia, fair trader an emerge and do so in regions II n,allpoor m,allpoor where they o-eist with middlemen, and IIIn,allpoor where fair traders only serve as produers link to markets. Importantly, with prie premia this low, in no loation are fair traders able to prie above the poverty line. The seond lass of ases involves relatively high prie premia a t(k n + δ n K)/(p p K). For eah a in this range, there are as many as five distintive equilibrium middlemen-fair trader onfigurations, depending on whether middlemen and fair traders oeist (regions I, II or III), and whether at least some of the produers served are non-poor. Interestingly, therefore, at a given prie premium, a, the mean poverty of produers served by fair traders aross all loations: P n α = 2 (a) 1 (a) 1 (a) Pα m = min{ p,p n+ } ˆp [( p p)/ p] α dh n (p)d+ n+ (a) 2 (a) 2 (a) 1 (a) 1 H n( ˆp )d + n+ (a) 2 (a) 1 (a) min{ p,p n+ } [( p p)/ p] α dh n (p)d 1 H n()d (26) may well be higher or lower than the mean poverty of produers served by middlemen: min{ p,p + } [( p p)/ p] α dh n(p)d+ 2 (a) min{ p, ˆp } [( p p)/ p] α dh n(p)d 1 (a) 1 H n()d + 2 (a) 1 (a) H n( ˆp ) H n()d (27) 16 An additional lass has a smaller than t(k n + δ n K)/(p K) and as shown in Fig. 5, a is too small for fair traders to ever emerge in equilibrium.

20 100 N. H. Chau et al. sine there are regions (say II n,allpoor m,allpoor ) where fair traders offer stritly higher pries ompared to middlemen in the same loation, but there are also loations (say, III n,allpoor )where some fair traders offer stritly lower pries ompared to middlemen in other loations (say, region I n,somepoor ). These an be ontrasted against the average poverty of produers who ontinue to have no aess to markets, Pα = [( p )/ p]α H n()d H n()d = In what follows, we offer an observation omparing the three. ( ) p α. (28) p Lemma 1 For all a t(k n + δ n K)/(p K) where either middlemen, fair traders, or both an emerge in equilibrium depending on loation, poverty among produers who do not have aess to market is the greatest but the poverty ranking between produers served by middlemen and fair traders is in general ambiguous: P α ma{p m α,pn α }. In the speial ase of poverty head ount with α = 0, and prie premia in the range a [t(k n + δ n K)/(p K),t(K n + δ n K)/(p p K)] 1 = P 0 = P n 0 >Pm 0 all produers who have no aess to market and all produers served by fair traders are poor, but some produers served by middlemen are not. There are two opposing fores at work here. With a prie premium, fair traders do indeed offer higher pries (Propositions 2 and 3). But with poverty aversion, fair traders tend to work in remote loations where produers are poorer (Proposition 1). Consequently, the relative poverty of produers served by middlemen and fair traders is indeed ambiguous. As we will see, however, this partiular ranking turns out to be key in the determination of poverty reduing strategies. 4 Poliy for poverty redution We have by now seen the full range of possibilities in the absene of government interventions, in terms of the equilibrium share of middlemen and fair traders, the impat of fair traders on markup and pass-through, as well as the equilibrium poverty impliation of fair traders. We now address a final question: Should a government interested in poverty redution subsidize fair traders, or should it perhaps subsidize middlemen? This is partiularly important for in many ases suh fair traders reeive diret support from governments and other donors. DFID (2009, pp ), for eample, reports for one of the most reent government ommitments to support fair trade through diret investment and prourements. Suppose that the objetive of the government is the mitigation of overall poverty. Should government use the same budget to subsidize indigenous middlemen instead? With three groups of produers served respetively by middlemen, fair traders, and no intermediaries at all, there are three orresponding diret mehanisms of intervention: (i) a subsidy on trade mediated by middlemen s m ; (ii) a subsidy on trade mediated by fair traders