Optimal choice of Voluntary traceability as a food risk management tool

Size: px
Start display at page:

Download "Optimal choice of Voluntary traceability as a food risk management tool"

Transcription

1 Optmal choce of Voluntary traceablty as a food rsk management tool Souza Montero, D.M., Caswell J.A. Paper prepared for presentaton at the th EAAE Congress People, Food and Envronments: Global Trends and European Strateges, Gent (Belgum), 6-9 August 008 Copyrght 008 by [Souza Montero, D.M., Caswell J.A. ]. All rghts reserved. Readers may make verbatm copes of ths document for noncommercal purposes by any means, provded that ths copyrght notce appears on all such copes.

2 Optmal choce of Voluntary traceablty as a food rsk management tool Souza Montero, D. M., Caswell J. A. Kent Busness School, Unversty of Kent, Wye, Ashford, UK Department of Resource Economcs, Unversty of Massachusetts, Amherst (MA), USA Abstract Traceablty systems are nformaton tools mplemented wthn and between frms n food chans to mprove logstcs and transparency or to reduce total food safety damage costs. Informaton about locaton and condton of products s crtcal when food safety ncdents arse. Ths paper uses a prncpal-agent model to nvestgate the optmal choce of voluntary traceablty n terms of precson of nformaton on a gven attrbute at each lnk of a food chan. The results suggest that four scenaros may emerge for the supply chan dependng on the costs of a system and whether or not the ndustry can nternalze total food safety damages: no traceablty, traceablty for one lnk, equal traceablty for all lnks, or dfferent postve traceablty levels across all lnks. Keywords Traceablty, food safety, prncpal-agent model I. INTRODUCTION Traceablty systems were frst and voluntarly ntroduced n food supply chans n the late 980s when the nternatonal traceablty standard NF EN ISO 840 was ssued []. Intally traceablty was conceved as part of a qualty assurance system and ntended to facltate effectve nformaton management. Traceablty can also be envsoned as a food safety rsk management tool [], as t can record nformaton on attrbutes of a product and may establsh a relaton between nputs and outputs, dfferent agents n the food chan, and events n producton processes. A crtcal ssue s the approprate level of nformaton n traceablty, who determnes t, and who governs t. Ths paper nvestgates optmal levels of traceablty at dfferent stages of food chans as endogenous choces. Prevous lterature has analyzed how exogenous levels of traceablty mpact lablty and ncentves for food qualty and safety [3], ncentves for anonymty [4], nspecton polces [5], or total costs of a food recall [6]. Traceablty s but one of many tools frms have to manage food safety. In ts essence a traceablty system s merely an nformaton management tool and s only useful f data theren s relevant, relable, and readly accessble. To develop and mplement traceablty requres leadershp and coordnaton among partners n a supply chan. Specal types of governance structures may have to be created to assure that the level of traceablty provded by each frm corresponds to the optmal level of traceablty for the whole supply chan. Falure to coordnate the amount of nformaton n traceablty systems may lead to dsruptons and mpede an effectve response to a food safety ncdent. Traceablty systems can potentally be used as a tool to prevent food safety ncdents (.e., as an ex-ante food safety system). In practce, most systems only react to an exstng occurrence and thus are ex-post means of mtgatng the total potental damages of an accdental or ntentonal safety falure. Ths paper provdes nsght nto the mpact of traceablty by analysng the choce of traceable nformaton through a prncpal-agent model. The paper s organzed as follows: the next secton revews prevous work on traceablty systems and voluntary and mandatory approaches to the mtgaton of food safety rsks. The followng secton models the choce of traceablty. The fourth secton analyss and dscusses the model. The fnal secton concludes and suggests future research. II. TRACEABILITY AND FOOD SAFETY Informaton s a key element of compettveness n food markets; t s also an element of food qualty and ts avalablty s vtal to manage food crses. However nformaton s an elusve concept, t must be related to somethng (for example, an attrbute of a product, a producton process, or a cost). Moreover t must be defned, t has to be dentfed, collected, analyzed and

3 communcated f t s to have any mpact. A number of tools (for example the nternet) enable easy and almost nstantaneous assess to all sorts of nformaton, yet vtal peces of nformaton may not be freely accessble. Ths ncludes nformaton on safety attrbutes of foods such as the level of pathogens n mlk enterng a cheese manufacturng plant or the type of pestcdes used to spray a vegetable crop. Traceablty systems are nformaton management tools wth a partcular feature: they enable the dentfcaton of the path of a product along each stage of the supply chan [7],[8]. In recent years a number of studes analysed the supply and demand for traceablty n food and feedstuffs. Golan et al. (004) studed food traceablty systems n the US, developng a framework to analyse whether exstng systems delver an effcent level of traceablty and how a regulator may nduce the socally optmum level [8]. They classfy systems n terms of depth (how far back or forward the system tracks relevant nformaton), breadth (how much nformaton s avalable) and precson (the detal and accuracy of the nformaton). Golan et al. (004) found that each ndustry had a dfferent effcent level of traceablty. However, they could not assure that the system n place provded the socally optmal level of traceablty n terms of quck response to food safety hazards. Ths paper draws on the framework proposed by Golan s et al. (004) to classfy traceablty systems proposng a model for the choce of traceablty n terms of depth and precson. In a recent paper, Starbrd and Amanor-Boadu (007) propose a prncpal-agent model where traceablty s an exogenous varable mpactng the nature of contractual relatons between agents n the supply chan [9]. More specfcally a monopsonstc buyer, wth mperfect nformaton on ts nput safety levels, has to desgn a set of contracts for ts heterogeneous potental supplers. The contracts offer a bd prce related to the contamnaton rate of food. Traceablty s an exogenous factor that decreases the levels of nformaton asymmetry and permts a shft of the costs of food safety damages to the source of contamnaton [9]. Ths paper relates to our approach n that t uses agency theory to model the governance of a food chan. However, rather than usng traceablty as a parameter we model t as an endogenous varable. Poulot and Sumner (008) nvestgate traceablty n the context of food safety [3]. They offer a stylzed model of a supply chan composed of farmers, marketers and consumers, where traceablty s not a choce varable but s lnked to food safety and lablty as t enables the dentfcaton of the source of system falures and mproves lablty attrbuton [3]. They conclude that when traceablty s not avalable frms are anonymous and may free-rde on the producers of safer food. Ths work proposes a formal model of the supply of safe food by dfferent players n a supply chan. In [3] the model treats traceablty as an exogenous probablty of dentfyng a source. Overall, prevous research on the economcs of food traceablty has not analyzed the choce of optmal traceablty levels by frms and/or regulators n food chans. III. MODELLING VOLUNTARY ADOPTION OF TRACEABILITY Traceablty can be defned as a flow of nformaton on product attrbutes and processes between players n a supply chan. One must dstngush a traceablty system from traceable nformaton; the former refers to the process and structure (for example computer hardware and software) through whch nformaton s shared along lnks between frms n a food chan. Traceable nformaton s what s flowng through these lnks. Dfferent factors determne the level of complexty of a traceablty system: the number of exstng nodes and lnks, traceablty levels, traceable unts, and the governance structures. Followng [8], we defne traceable nformaton n three dmensons: depth (the number of lnks for whch t s avalable), breadth (the number of attrbutes covered), and precson (the detal and accuracy of the nformaton). Denote j as the level of precse traceable nformaton for attrbute j on lnk. Assume that precson vares between zero and one, the later beng maxmum precson. For example, a traceablty system for pork may track nformaton on Salmonella and E. col (j=), between farms, feedlots, slaughterhouses, and retalers (3 lnks, =3), wth a th Congress of the European Assocaton of Agrcultural Economsts EAAE 008

4 3 level of precson such that each cut of meat can be traced to an ndvdual anmal. Consder the development of a voluntary traceablty system for a food supply chan where three representatve frms are lnked vertcally. Further suppose ths stylzed supply chan represents an entre ndustry, composed of farmers, processors and dstrbutors. Followng [0] and [], retalers are supply chan leaders n the provson of food safety. We present a model n whch a downstream prncpal (the retaler) defnes the level of traceablty each agent upstream n the chan wll have to provde. An approprate framework to analyse the desgn of voluntary or prvate traceablty systems (and mandatory or regulatory systems, as well) s the prncpal-agent model. A. Model Suppose a monopsonst retaler ams to desgn a voluntary traceablty system requrng ts supplers to provde a traceablty level to mtgate ex post food safety damage costs. Assume further that only one of the product s attrbute s traceable (say orgn), thus t s not necessary to use the subscrpt (j) dentfyng the attrbute. Both the prncpal (the retaler) and the agents (the farmer and the processor) are rsk neutral. The problem of the decson maker s to nduce the optmal level of traceablty from each lnk n the food chan ( from farm to processor and from processor to retaler). Fgure shows a representaton of the supply chan. Where the dash arrows denote product flow and bold ones the nformaton flow. Assume that f traceablty s mplemented nformaton can flow up and downstream at the same cost, allowng the dentfcaton of orgn and destny of products to whch t relates. Only the farm and processng plant provde nformaton to the system. The retaler demands traceablty to mtgate safety rsks. Farm Retaler Processor Fgure : A stylzed food chan traceablty system The retaler has perfect nsght nto the costs and actons of agents and contracts the levels of traceablty from the farm or the processor ndependently. The task of the retaler s to desgn the least costly compensaton scheme (b (=, )) to nduce the farmer and the processor to provde a level of traceablty ( ) that decreases the total damages (D) caused by a food contamnant (e.g., Salmonella), occurrng wth exogenous probablty (ψ). The total damage cost of a food ncdent s a decreasng and convex functon of upstream traceablty levels. The subscrpt s used throughout to dentfy the lnk between the farmer and the processor (=) or between the processor and the retaler (=). Assume that as precson decreases to zero, nformaton becomes useless,.e., when =0 there s no traceablty. The problem of the retaler s to mnmze ts total expected costs (E[TCr]) wrtten as: r = ψd + b D, [0,]; b (), b () 0 = E[TC] (, ) ( (, )) s.t.: b( D(, )) c 0 b ( D(, )) c 0 ( ) ( ) ( ) ( ) ( ) ( ) b( D(, )) c b( D(, )) c, [0,] b ( D(, )) c b ( D(, )) c, [0,] Where c (=,) are the ncreasng and convex costs of traceablty ncurred by farmer and retaler. The frst two constrants are the ndvdual ratonalty or partcpaton constrants: the farmer and processor accept a contract to provde traceablty nsofar as ts payoff s t least as large as ther respectve reservaton utltes, assumed to be zero. The thrd and fourth nequaltes are the ncentve compatblty constrants; they guarantee that both farmer and processor offer the level of traceablty requred by the retaler to mnmze ts total costs. The retaler s better off the larger the savngs from the compensaton pad for havng traceablty. The partcpaton constrants are bndng as the prncpal has no motve to offer more compensaton than s necessary for the agents to accept a contract. Focusng for now on the condtons to accept a contract and recallng the assumpton of full nformaton, we substtute the ndvdual ratonalty constrants nto the objectve functon n () to yeld: () th Congress of the European Assocaton of Agrcultural Economsts EAAE 008

5 4 ψ r E [ T C ] = D (, ) + c ( ) + c ( ) ( ), [ 0, ] Dfferentatng wth respect to and the retaler determnes the optmal levels of traceablty requred from the farmer and processor. E [ T C ] = 0 D ( ) = c ( ) ψ E [ T C ] = 0 D ( ) = c ( ) ψ ( 3 ) Where D y s the partal frst dervatve of the damage functon wth respect to the traceablty level of agent, and c denotes the margnal costs of traceablty. Gven the assumptons on food safety damage costs and traceablty cost functons, the determnant of the Hessan matrx below s postve and therefore the suffcent condton s met: ψ D ( ) + c ( ) ψ D ( ) H = (4) ψ D ( ) D ( ) c ψ + ( ) By the mplct functon theorem, the system of equatons (3) defnes the optmal levels of voluntary traceablty requred from the farmer and the processor. These are found where the margnal reducton of expected ex post food safety damages from traceablty equals the margnal cost level. Denote the optmal levels of traceablty by y * and the correspondng expected damages as D* = ψd(y *, y * ). The retaler must desgn a compensaton scheme that guarantees the provson of these optmal levels of traceablty by upstream agents. From the ncentve compatblty constrants, we know that both farmer and processor wll choose y * f and only f t provdes them more utlty than they receve choosng any alternatve traceablty level. The payment scheme below s suffcent to assure both partcpaton and that the optmal level of traceablty requested by the retaler s chosen by both farmer and processor: * * * c ( ) f = b ( D ) = (=,) (5) 0 otherw se We note that there are many other alternatve payment schemes, for example one wth the resdual clamancy soluton, where the prncpal would assume a fxed amount of the damages and shft the remanng part upstream. Ths was the payment suggested n [9] though n a slghtly dfferent context. IV. RESULTS AND DISCUSSION A number of cases emerge when determnng whch level of traceablty each agent upstream has to provde. We follow a beneft-cost analyss framework. Traceablty wll not be mposed f ts costs are larger than the benefts from damage mtgaton. Traceablty s feasble f benefts outwegh costs for at least one lnk n the supply chan. Before proceedng wth a detaled analyss, fgure provdes ntal ntuton on the determnants of the decson. The fgure s constructed fxng the level of traceablty from the processor and focusng on the choce of voluntary traceablty from the farm. Furthermore we assume a lnear traceablty and convex food safety damage functons. Snce the objectve of the prncpal downstream s to desgn a traceablty system that reduces totals costs, f these are larger than the expected reducton of prvate damages the prncpal s better off wthout traceablty. Fgure : Illustraton of the choce of the optmal levels of traceablty The analyss s further complcated when one consders other lnks n the supply chan. The more lnks nvolved and attrbutes that are traced, the more complex s the decson. Should the traceablty level be the same regardless of the lnk or attrbute to whch t refers? Table compares four dfferent cases and provdes nsght nto ths queston. The results are th Congress of the European Assocaton of Agrcultural Economsts EAAE 008

6 5 based on the frst order condtons of the voluntary model of traceablty. Case Case Table : Cases of voluntary traceablty Farm to Processor ψ D < c ψ D < c Case 3 D c ψ Case 4 ψ D c Processor to Retaler ψ D < c ψ D c ψ D c ψ D c The levels of traceablty wll be postve and equal across lnks when margnal damage mtgaton and cost of traceablty are equal across lnks The levels of traceablty wll be postve but dfferent across lnks when both the margnal damages and costs are dfferent across lnks. Case results n no traceablty. It s based on analyss of the Kuhn-Tucker condtons and corresponds to the corner soluton scenaro n fgure. In the voluntary case, f the margnal mpact of extra levels of traceablty n terms of expected damage mtgaton s smaller than the margnal costs, then a ratonal prncpal should not mplement traceablty. For example, ths could be the case where more detaled nformaton on the presence of a contamnant does not contrbute to the reducton of the food safety damages t causes. In case, traceablty s feasble for one of the lnks n the supply chan. Ths s the lnk for whch the margnal costs of traceablty equal the margnal partal expected damage mtgaton costs. Suppose ths s the lnk between the processor and the retaler. In ths stuaton, the margnal costs of traceablty for the farmer are larger than ts margnal effect on the mtgaton of expected food safety damages and t s not worth havng traceablty on ths lnk. In case 3, traceablty s feasble at every lnk of the supply chan and the same levels of traceablty emerge. Ths presumes that nformaton s equally mportant, as t has the same costs and contrbutes n the same manner to the mtgaton of damage costs, regardless of the lnk to whch t refers. An example s traceablty for a contamnant that perssts along the supply chan (say a chemcal component of food) and cannot be removed; the only way to assure ts absence s by detectng t and keepng a record throughout the supply chan. Fnally, case 4 s perhaps the most realstc. It llustrates the case of a voluntary system that traces nformaton for every lnk n the chan, but wth dfferences n levels of traceablty across lnks. For nstance n beef supply chans each head of cattle may be traced from the feedlot to the slaughter house. In the next lnk, as anmals are processed nto beef cuts, precse traceablty by the head may be replaced wth lot traceablty. V. CONCLUSIONS AND FUTURE RESEARCH Ths paper nvestgates the choce of voluntary traceablty by frms n a food chan usng a prncpalagent model. The prncpal s the downstream retaler that desgns two ndependent contracts for traceablty from a farm and a processor upstream. Traceablty reduces ex post damages of a food safety ncdent. Our results suggest that four cases may emerge. Frst no traceablty wll emerge f ts margnal costs are larger than the benefts to the retaler. Partal traceablty occurs when only one lnk of the chan s chosen to be traceable,.e., only n one lnk are the margnal costs of traceablty equal to the margnal benefts to the retaler. Thrd the same level of traceablty wll occur at each lnk of the food chan f the margnal costs of traceablty and the partal margnal benefts are the same to each lnk of the chan. Fnally, there wll be dfferent levels of traceablty n each lnk f the margnal costs and partal margnal benefts are dfferent for each upstream agent. Understandng the condtons under whch voluntary traceablty wll develop s mportant to companes and government agences n makng decsons on managng food safety rsks. Traceablty requrements by retalers and other downstream partcpants n food supply chans may be ressted by some supplers as overly burdensome. In other cases, governmental authortes may conclude that prvate ncentves to nsttute traceablty are nadequate, for example where damages n the case of food safety falures do not fall fully on responsble companes. Future research focusng on ex post analyss of the th Congress of the European Assocaton of Agrcultural Economsts EAAE 008

7 6 development of voluntary traceablty systems and on the ex ante condtons under whch governments ntervene to mandate traceablty wll lead to a better understandng of the economcs of ts adopton. ACKNOWLEDGMENT Ths research was supported by a doctoral fellowshp from the Portuguese Foundaton for Scence and Technology (FCT) and by a USDA Cooperatve State Research, Educaton, and Extenson Servce (CSREES) Specal Grant to the Food Marketng Polcy Center, Unversty of Connectcut and by subcontract at the Unversty of Massachusetts Amherst. REFERENCES. Gencod EAN France. 00. Traceablty n the Supply Chan From Strategy to Practce, Aprl.. Standng Commttee on the Food Chan and Anmal Health (004) Gudance on the Implementaton of Artcles,, 6, 7, 8, 9 and 0 of Regulaton (EC) N 78/00 on General Food Law: Conclusons. at df 3. Poulot S, Sumner D (008) Traceablty, Lablty, and Incentves for Food Safety and Qualty. Amer. J. Agr. Econ 90(): Amanor-Boadu V, Starbrd S A (005) In Search of Anonymty n Supply Chans. Journal of Chan and Network Scences 5: Starbrd S A, Amanor-Boadu V (005) Do Inspecton and Traceablty Provde Incentves for Food Safety? Journal of Agrcultural and Resource Economcs 3: Resende-Flho M A, Buhr B L A Prncpal-Agent Model for Evaluatng the Economc Value of a Traceablty System: A Case Study wth Injecton-ste Leson Control n Fed Cattle Amer. J. Agr. Econ. Forthcomng 7. European Parlament and of the Councl (00) Regulaton (EC) No 78/00, layng down the general prncples and requrements of food law, establshng the European Food Safety Authorty and layng down procedures n matters of food safety. Offcal Journal of the European Communtes, L 3: Golan E, Krssoff, Kuchler F, Nelson K, Prce G, Calvn L (004) Traceablty n the U.S. Food Supply: Economc Theory and Industry Studes. Economc Research Servce, U.S. Department of Agrculture, Agrcultural Economc Report No March Starbrd S A, Amanor-Boadu, V (007) Contract selectvty, Food Safety and Traceablty. Journal of Agrcultural & Food Industral Organzaton 5() at 0. Henessy D A, Roosen J, Mranowsk J A (00) Leadershp and the Provson of Safe Food. Amer. J. Agr. Econ. 83(4): Henson S J, Reardon T (005) Prvate Agr-Food Standards: Implcatons for Food Polcy and the Agr-Food System. Food Pol. 30(3): 4-53 Correspondng author: Dogo M. Souza-Montero (PhD), Lecturer, Kent Busness School, Wye Campus, Unversty of Kent, Wye, Ashford, Kent TN5 5AH, UK (D.M.Souza-Montero@kent.ac.uk) th Congress of the European Assocaton of Agrcultural Economsts EAAE 008