THE EFFECTS OF STARLINK CORN FOOD SAFETY EVENTS ON RETURNS AND RISK OF AGRIBUSINESS FIRMS

Transcription

1 THE EFFECTS OF STARLINK CORN FOOD SAFETY EVENTS ON RETURNS AND RISK OF AGRIBUSINESS FIRMS by Alla Golub, Chrstne A. Wlson and Allen M. Featherstone Staff Paper #05-08 July 005 Dept. of Agrcultural Economcs Purdue Unversty Purdue Unversty s commtted to the polcy that all persons shall have equal access to ts programs and employment wthout regard to race, color, creed, relgon, natonal orgn, sex, age, martal status, dsablty, publc assstance status, veteran status, or sexual orentaton.

2 THE EFFECTS OF STARLINK CORN FOOD SAFETY EVENTS ON RETURNS AND RISK OF AGRIBUSINESS FIRMS by Alla Golub, Chrstne A. Wlson and Allen M. Featherstone Dept. of Agrcultural Economcs, Purdue Unversty West Lafayette, Indana Staff Paper #05-08 July 005 Abstract Usng event study methodology, we examne the effects of three Starlnk corn food safety events on the rsk and returns of drectly nvolved agrbusness frms and the dffuson effects on other frms n the ndustry. We test the hypothess that Starlnk food safety events do not have an mpact on the magntude and volatlty of the returns and the nondversfable rsk of the frms. Based on the results of ths study, we conclude that Wall Street s not senstve to the Starlnk corn food safety events. Keywords: event study, food safety, GMO, Starlnk corn. Copyrght by Alla Golub, Chrtsne A. Wlson, Allen Featherstone. All rghts reserved. Readers may make verbatm copes of ths document for non-commercal purposes by any means, provded that ths copyrght notce appears on all such copes.

3 THE EFFECTS OF STARLINK CORN FOOD SAFETY EVENTS ON RETURNS AND RISK OF AGRIBUSINESS FIRMS by Alla Golub, Chrstne A. Wlson and Allen M. Featherstone Introducton Publc concern about food safety has ncreased rapdly durng the last decade. Food safety concerns are at the forefront of much of the work n the food and agrbusness ndustry. Food safety nvolves a wde array of ssues ncludng but not lmted to mcrobologcal food borne pathogens, chemcal toxns, labelng, boterrorsm, and technologcal and new product developments. Food safety events nclude the Starlnk corn controversy, the alar ssue wthn the apple ndustry, and the e-col ssues n the meat packng ndustry. Prevous research has generally explored consumer preferences and consumer reactons to food safety ssues (Baker; Henneberry, Pewthongngam and Qang; Msra, Huang and Ott). However, consumers are not the only group mpacted by food safety; frms and the ndustry, nvolved n food recalls and other food safety ssues, also bear sgnfcant costs. These expenses can relate to closng producton facltes and dsposng or recoverng recalled products. More mportant, the frm ncurs a reducton of demand for ts products, and f a frm nvested n ts mage, these nvestments may turn out to be a net loss. To estmate wealth losses due to nvolvement n food safety ssues, one may use eventstudy methodology and measure the changes n market valuatons of the frms nvolved n the event. Recent work has expanded the area of nqury to nclude the examnaton of the reacton of stock returns of publcly traded frms to specfc food recall events (Saln and Hooker; Thomsen and McKenze), thereby provdng a startng pont for understandng return and rsk responses to food safety ssues. However, addtonal research n ths area s necessary to more completely quantfy shareholder reacton to food recalls and other food safety events. One such food safety ssue s that of human consumpton of genetcally modfed organsms (GMOs). Several events have focused around GMO Starlnk corn and the controversy over ts use n human foods durng the last years. Avents nvented Starlnk corn by ncorporatng Cry9C, a proten solated from common sol bactera, nto corn. The Cry9C proten s effectve aganst caterpllars. StarLnk corn was approved by the U.S. Envronmental Protecton Agency (EPA) for anmal feed but not for human food untl addtonal testng was completed because the proten Cry9C s consdered a medum rsk potental human allergen. The controversy began when traces of DNA from StarLnk corn were found n taco shells and other corn related products. These events were followed by the seres of food recalls and closng of producton facltes for checkng and cleanng. A number of food and agrbusness frms have ether been nvolved n or mpacted by Starlnk corn food recalls. Most of the recalls were conducted by local manufacturers, whch are not publcly traded. For ths study, three events were dentfed that drectly nvolved three publcly traded frms. Two of the events are the Kraft Foods recall of Taco Bell taco shell products from grocery stores on September, 000, and the Kellogg recall of Mornngstar 1

4 Farms Corn Dogs on March 14, 001. The thrd event happened on October 17, 000, when the Wall Street Journal (WSJ) announced that ConAgra Foods had closed ts only corn mll located n Atchson, Kansas, because t mght have receved Starlnk corn. The general objectve of ths research s to quanttatvely examne the effects of the Kraft Foods recall of Taco Bell taco shell products, the Kellogg recall of Mornngstar Farms Corn Dogs and the announcement about ConAgra Foods closed plant on the rsk and the returns of food and agrbusness frms. Not only stocks of drectly nvolved frms, but also stocks of ndrectly affected frms are examned n order to assess the dffuson of the mpact of the events to other frms n the ndustry. Specfcally, a partal event analyss s used to measure and analyze changes n the magntude and volatlty of the returns and the nondversfable rsks of agrbusness frms due to the three food safety events. We test the null hypothess that such events have no mpact on the amount of rsk assocated wth the frm and on the frm s returns. As noted by Sara Lee s Corp. CEO n an ntervew for Reuters (Factva), U.S. consumers do not seem concerned about the use of boengneered ngredents n food, despte recalls related to a certan stran of genetcally modfed corn. The paper s organzed nto 6 sectons. Secton two provdes an overvew of the relevant research. Secton three s devoted to the methodology used; secton four descrbes the data; secton fve provdes estmaton results, and secton sx s the concluson. Relevant Lterature One method to measure the effects of food safety events on a frm s to apply event study methodology (Fama et al.; MacKnlay). Ths method does not requre measurement of drect costs that may result for the frm due to the event; rather, t gves the measure of the economc mpact on the frm nvolved based on the changes n equty market or product market valuatons. The relevant event studes can be dvded nto two groups. The frst group of studes ncludes research devoted to quantfyng the effects of food safety events on the prces of contamnated products. The second group ncludes lterature nvestgatng the effects of food safety events on the stock prces of the nvolved frms. Lusk and Schroeder, and McKenze and Thomsen represent the frst group of studes n ths lterature revew. Lusk and Schroeder examned the effect of meat recalls on lve cattle and lean hog futures prces. They found that the announcement of recalls dd not have a strong effect on ether prce seres. Usng event study methodology, McKenze and Thomsen examned the mpact of beef recalls due to E. Col O157:H7 contamnaton on wholesale and farm-level beef prces. They found that prces of boneless beef reacted negatvely to recalls; however there was no reacton n the lve cattle prces and very lttle reacton n the boxed beef prces. They concluded that there are ncentves for packng frms to control food contamnaton and nvest n food safety technology, but there are no such ncentves at the farm level. Thomsen and McKenze, Saln and Hooker, and Henson and Mazzocch represent the second group of studes, nvestgatng the effects of food safety events on the stock prces of affected frms. Thomsen and McKenze analyzed federally supervsed meat and poultry recalls

5 from 198 to They used standard event study methodology to quantfy the effects of the announcements of the class 1, and 3 recalls on the stock returns of the recallng frm. Thomsen and McKenze found evdence that class 1 recalls nvolvng serous threats to consumer health, translated to shareholder losses whle recalls nvolvng less serous volatons had no negatve mpact. Saln and Hooker used the partal event analyss method to quantfy frm-specfc effects of ncdents of mcrobologcal (arsng from bactera, vruses and parastes) contamnaton of food, whch lead to mmedate health concerns for consumers. They nvestgated four recalls conducted by three agrbusness frms. These recalls were dfferent n scope and severty, and the frms dffered by sze and dversfcaton. Wth partal event study analyss, the specfc abnormal performance of each frm was quantfed and statstcally tested. Saln and Hooker found that returns fell after the recall only for the smallest frm n the study. Rskness of the stocks, measured by volatlty, ncreased after the recalls, but none of the consdered frms experenced an ncrease n nondversfable rsk followng the recall. Saln and Hooker concluded that fnancal markets reacted n a lmted way to certan food recalls. Henson and Mazzocch examned the mpact on agrbusnesses of the Unted Kngdom government s announcement of a possble lnk between Bovne Spongform Encephalopthy and human health. They found that sgnfcant negatve abnormal returns n the beef, pet food, anmal feed, and dary sectors and postve abnormal returns n other meat sectors. Prevous studes have focused only on the frms drectly nvolved n food recalls, or they have examned ndustry effects n aggregate. In ths study, we nvestgate the effects of Starlnk food safety events on the rsk and the returns of food and agrbusness frms that were not only drectly nvolved, but also ndrectly affected by the ncdents n order to assess the dffuson mpact n the ndustry. Methodology Perhaps the most comprehensve analyss of the event study methodology was publshed by MacKnlay. The event study methodology rests on the noton of market effcency. Gven the sem-strong form of market effcency any new nformaton wll be quckly reflected n the market value of the affected frms and the event s economc mpact can be measured usng changes n the stock prces around the event. After the event date s determned, the event wndow, the perod over whch the securty prces of the frms nvolved n the event wll be examned, and market model estmaton perod are dentfed. When an event s an announcement on a gven day, the length of the event wndow s one or two days. If leakages of nformaton may occur, t s approprate to expand the event wndow. In ths study, abnormal returns are measured to assess event mpacts. The abnormal return s the actual post event return of the securty over the event wndow mnus the normal return of the frm over the event wndow. The normal return s the return whch would be expected f the event dd not take place. MacKnlay dscusses the fact that there are three common choces for 3

6 modelng the normal return: the constant mean return model, the market model, and the factor model. In ths study, the market model s used, where the normal return s defned as: R = X θ + ε, (1) where R s a vector of estmaton-wndow returns on securty, and X = [l R m ] s a matrx wth a vector of ones n the frst columns and vector of market returns R m n the second column, ε s the zero mean dsturbance term wth varanceσ ε, and θ are parameters of market model. The estmaton wndow s a subset of data pror to the event wndow. The event wndow tself s not ncluded n the estmaton perod to prevent the event from nfluencng the normal performance model parameters estmates (Campbell, Lo, and MacKnlay). Abnormal returns are aggregated over the event wndow and post event wndows for each frm n order to draw overall nferences for the event of nterest. Both parametrc and nonparametrc tests have been used for testng whether the abnormal returns are statstcally dfferent from zero. To conduct a parametrc test, we adopted the method descrbed n Campbell, Lo, and MacKnlay. 1 Ths approach s based on specfc assumptons about the dstrbuton of abnormal returns. The alternatve nonparametrc tests allow the robustness of conclusons based on the parametrc tests to be checked. The Wlcoxon rank-sum test examnes whether two samples came from populatons wth the same dstrbuton. Ths test s used n ths study to test the null hypothess that the actual and predcted returns by the normal model (1) over [-5, 0], [0, 5] and [0, 10] wndows came from the same dstrbuton, or that the event does not have an mpact on the returns. Statstcal analyss of the volatlty of returns before and after the events can gve some estmate of the effects of the recalls on the stock prce rsk. To analyze the volatlty of returns, we use a form of the Goldfeld Quandt test. We denote the volatltes of returns of frm before and after the recall as σ 1 and σ respectvely. Then, under the assumpton that returns are normally dstrbuted, ˆ σ / σ s dstrbuted as F(n -1, n 1-1), where σ >σ 1 and n and n 1 ˆ1 are days after and before the event over whch varances σ and σ 1 are estmated. Ths allows the null hypothess that σ =σ 1, or the event has no mpact on the stock prce rsk, to be tested. The alternatve hypothess s that the stock prce rsk after the event s greater than t was before the event. We conducted ths test for 30, 50, 100, 10 and 150 days before and after the event. It s mportant to notce that ths test does not control for overall varablty of the market envronment. The last test used n ths study s devoted to examnng the changes n nondversfable rsk of the securtes due to the events. Rskness of returns s an mportant characterstc of frms stocks because nvestors requre a hgher expected return n exchange for bearng hgher nondversfable rsk. Followng Saln and Hooker, we use a Chow test to statstcally test the change n rsk, estmated wth beta, the Captal Asset Prcng Model (CAPM) measure of nondversfable rsk of the frm, before and after the event. 4

7 Data Data n ths study consst of daly rates of return to common stock for seventeen food and agrbusness frms that were drectly or potentally ndrectly affected by the three Starlnk corn events durng and the rates of return for the Center for Research and Securty Prces (CRSP) Value Weghted market ndex. All return data were obtaned from the CRSP. Starlnk corn food safety events were dentfed usng the Food and Drug Admnstraton (FDA) web page for nformaton on food recalls. Thrteen recalls occurred durng the last years due to contamnaton wth Starlnk corn. However, most of these recalls were conducted by local manufacturers, whch are not publcly traded. Three events drectly nvolvng three publcly traded frms, Kraft Foods, Kellogg and ConAgra, were dentfed. Factva and LexsNexs were used to dentfy the Starlnk Corn food safety event dates. Kraft Foods announced a voluntary recall of all Taco Bell taco shells contamnated wth Starlnk Corn from grocery stores shelves on September, 000; however, news dscussng the possble ncdent ht the papers on September 18, 000, n the Washngton Post. Kellogg announced a recall of Mornngstar Farms Corn Dogs on March 14, 001; however, news of ths ncdent was publshed on March 8, 001, n the Los Angeles Tmes. The event dates used n ths study are those correspondng to the frst news of the food contamnaton (September 18, 000 and March 8, 001) snce those are the dates when the market frst potentally reacted to the news. The market model for these two events was estmated usng 10 calendar days, whch s equvalent to 78 busness days for the Kraft recall and 76 busness days for the Kellogg recall, wth the 5 days pror to the event day omtted. Ths choce allows any overlappng of market model perods for the two events to be avoded. The abnormal returns were estmated over [-5, 10] event wndow to examne the effects of the recalls on the frms returns. The abnormal returns were further aggregated over 5, 10, 30, 50 and 100 days after the event to examne the cumulatve mpact of the recalls over tme. The FDA Enforcement Report on November 15, 000, ndcates that ConAgra Foods recalled cornmeal, corn flour, snack meal and flakng and polenta grts because these products appeared to contan Starlnk corn. The products where dstrbuted n 11 states. The nterestng thng about ths event s that ConAgra started to recall ts products on October 4, 000, but the FDA dd not publsh an announcement untl November 15, 000. To determne the announcement day, Factva and LexsNexs were searched for any news about ConAgra durng the Fall 000. The frst news related to the recall appeared n the WSJ and other publcatons smultaneously on October 17, 000. Accordng to the news (Factva), on October 11, 000, ConAgra Foods closed ts only corn mll located n Atchson, Kansas, because t mght have receved Starlnk corn. Interestngly, the company very quetly started the recall on October 4, quetly closed the plant on October 11, allowed the news to go out on October 17, and then the offcal FDA report dd not appear untl November 15. Gven ths stuaton, t s very dffcult to dentfy the event date. Potentally, October 17, 000, was the frst day that the market could have receved new nformaton about the relatedness of ConAgra to Starlnk corn and the possble food contamnaton. Snce ths was the date of the frst publc announcement, ths date was used as the event date n ths study. The market model estmaton perod and event wndow 5

8 for ths event were chosen to be, respectvely, 10 busness days from t=-15 to t=-6 and from t=-5 to t=10 relatve to announcement date t=0. Note, the market model estmaton perod ncludes the Kraft Foods recall, but the event wndow does not. Also, October 11, 000, the date when the mll was actually closed, s ncluded n the event wndow. The news (Factva) on October 17, 000, underlned that the mll could have receved the same type of genetcally modfed corn that sparked a natonwde recall of some taco-shell brands. It s not clear whether the effect on the stock returns of ths announcement s postve or negatve. Gven that n September 000 Kraft Foods announced a Starlnk food recall, and gven that ConAgra s announcement on October 17, 000, dd not say that contamnaton was found, the fact that the mll was closed may actually be perceved by the market as a preventve acton. An announcement of such preventve acton may have a postve mpact on stock returns. In addton to the companes drectly nvolved n the events examned, ths study also examnes varous frms wthn the ndustry n order to determne the dffuson of the mpact of the events to other frms n the ndustry. Hoover s Onlne was used to dentfy the compettors of Kraft Foods and Kellogg n the breakfast cereals, snacks, bakery, frozen foods and other markets nvolvng corn products, and the compettors of ConAgra n the agrcultural products, ngredents and corn flour markets. Possble compettors for Kraft Foods and Kellogg are Dannon, Sara Lee, General Mlls, Henz, Quaker Oats, Frto-Lay, Keebler, Lance, Inc., Unlever (UK), and Unlever (Netherlands, NV). ConAgra s compettors are Archer Danels Mdland (ADM) and Corn Products Internatonal. Our expectaton about changes n the stock prces of these compettors s mxed. The reacton may be postve because these frms are compettors of the affected frms, and durng the event wndows, representatves of these compettor companes made announcements that they do not use and have never used Starlnk corn (Factva). ADM announced that they tested ther delveres at the elevators for Starlnk corn durng the ConAgra event wndow (Factva). Wth respect to the Kellogg and Kraft events, announcements contaned nformaton that Unlever and Frto-Lay use only GMO-free ngredents. On the other hand, Kellogg, General Mlls, Quaker Oats and other companes are mentoned n the press durng the event wndows as companes usng genetcally modfed corn, not Starlnk corn but other types of GMOs (Factva). So, the market reacton may be negatve as well. Taco Bell, a company dealng consderably wth corn products and whose name s exactly the same as the name of the product contamnated wth Starlnk corn, was also ncluded n ths analyss to see f the events had any effects on ts value. Also, accordng to the news, Taco Bell tested ts corn based products for Starlnk corn because t mght have receved contamnated products (Factva). Avents, the frm that nvented Starlnk corn, was blamed n the news by government offcals as a company that was supposed to ensure that farmers kept Starlnk corn separate from other varetes, but faled to do so (Factva, Assocated Press Wrter, 10/6/000). Avents was ncluded n the sample to see f negatve publcty around the events had any negatve mpact on ts stock prces. Thus, seventeen total food and agrbusness frms that were drectly or potentally ndrectly affected by the three Starlnk corn events durng were chosen for the analyss. 3 6

9 Results Frst, we present results for two recalls, and then we dscuss the results for the ConAgra closed plant event. Fgure 1 and show Kraft Foods daly stock prce movement around the recall of Taco Bell taco shell products and Kellogg daly stock prce around the Kellogg recall of Mornngstar Farms Corn Dogs, respectvely. It seems that the stock market value of Kellogg s not senstve to the recall of ts product, when the value of the Kraft Foods was negatvely mpacted by the recall. The fgure 1 ndcates possble leakage of nformaton before the news. However, to make a concluson about the mpact of the event on the frm, t s mportant to take nto account overall stock market movement. Fgures 4 and 5 depct the abnormal returns of some of 17 companes durng [-5, +10] wndow around the Kraft Foods and Kellogg events, respectvely. For the Kraft Foods event, 13 of the 17 frms experenced negatve excess returns on the event date when the news was frst released about the potental ssue. However, four days later, on September, 000, when Kraft announced the actual recall, only two frms, Keebler and Taco Bell, experenced negatve excess returns. Kraft experenced negatve excess returns on the event date, two days after the event date, and one through four days pror to the event date. Somewhat smlarly, Taco Bell experenced negatve excess returns on the event date, one and four days after the event date, and two, four, and fve days pror to the event date. The parametrc t-test results ndcated that Kraft dd not experence any sgnfcant abnormal returns durng the event wndow. Sara Lee, Frto-lay, ADM and Corn Products Internatonal experenced sgnfcant negatve abnormal returns on September 0, 000, two tradng days after the announcement. Dannon experenced very large postve abnormal returns on September, 000, when Kraft announced the actual recall. The Wlcoxon nonparametrc rank-sum tests for -5, +5, and +10 days, however, ndcate that only Sara Lee and Keebler experenced actual returns that were statstcally dfferent at the α=0.05 level from the expected normal returns. Interestngly, only Corn Products Internatonal experenced large negatve abnormal returns on the event day and two days after wth large postve returns on the thrd day after the event. Addtonal nvestgaton of the news about Corn Products Internatonal revealed that from September 15 to September 19, 000, there were several announcements n the press about a reducton of yearly earnngs by 40% compared to the prevous year (Factva). On September 1, 000, Reuters announced that Corn Products Internatonal had ther nvestment ratng rased. Thus, the observed prce movement for Corn Products Internatonal around September 18, 000, s more lkely due to earnngs and nvestment ratng announcements, nstead of the Kraft Foods recall event. Wth respect to the Kellogg recall, only 4 of the 17 companes experenced negatve excess returns on the event date of March 8, 001, and Kellogg was not one of these frms. However, four tradng days later, on March 14, 001, when Kellogg offcally announced the recall, 14 of the 17 companes experenced negatve excess returns; Dannon, Keebler, and Corn Products Internatonal experenced postve excess returns. Kellogg experenced negatve excess returns on one and four days pror to the event, and on one, two, four, and sx through nne days after the event date, but not on the event date. The parametrc test ndcated that none of the frms experenced sgnfcant negatve abnormal returns on the event day. Kraft and Unlever, NV experenced sgnfcant negatve abnormal returns on the second day after the 7

10 announcement, and Dannon experenced sgnfcant negatve abnormal returns on the thrd day after the announcement. Unlever, LTD and Unlever, NV experenced sgnfcant negatve abnormal returns for days 9 and 10 of the event wndow. Kellogg dd not experence any sgnfcant negatve returns durng the event wndow. Results of the Wlcoxon nonparametrc rank-sum tests ndcate that Kraft, Dannon, and Lance, Inc. experenced actual returns that were statstcally dfferent from the expected normal returns for 5 days pror to the event date. Kraft, General Mlls, Quaker Oats, Frto-Lay, Taco Bell, Unlever (LTD), and Unlever (NV) experenced sgnfcant returns for 5 days after the event, whch s consstent wth the results of the parametrc test. Kraft, General Mlls, Henz, Quaker Oats, Frto-Lay, ADM, Unlever (LTD), and Unlever (NV) experenced statstcally dfferent returns for 10 days after the event, and Kellogg dd not experence returns sgnfcantly dfferent from normal returns based on the Wlcoxon test. The abnormal returns were further aggregated over ntervals of 5, 10, 30, 50, and 100 days after the event date to examne the cumulatve mpact of the recall over tme. At most, 5 of 17 frms experenced negatve cumulatve abnormal returns for any of the post Kraft Foods event tme perods examned, and ths result occurred for the [0,5] wndow. The number of negatve cumulatve abnormal returns for the Kellogg Starlnk corn recall s much greater. Negatve cumulatve abnormal returns were experenced n 80% of the companes and post event tme perods examned. Ths fndng may suggest that the Kellogg recall had a larger negatve mpact than the Kraft Foods recall, or that the food ndustry experenced a down movement durng that tme perod when compared wth the overall market. To address ths queston, we reestmated the market model wth the S&P food ndustry ndex, obtaned through Datastream, as an explanatory varable. Results usng the S&P food ndustry ndex are smlar to the results descrbed above for both events n terms of sgn and magntude of the returns. The Kellogg recall leads to a slghtly less persstent reacton when measured wth respect to the S&P food ndustry ndex; the number of negatve cumulatve abnormal returns s fewer, whch can be explaned as some down movement experenced by the food ndustry around the Kellogg recall. After accountng for the food ndustry down movement, we conclude that the Kellogg recall had a larger total negatve mpact n the market than the Kraft Foods recall. However, the parametrc test ndcated that none of the cumulatve abnormal returns for any event are statstcally sgnfcant. The cumulatve abnormal returns over [-5, t] event wndows, where t belongs to [-5, 10] nterval were also estmated for both events. Strkngly, many of the cumulatve abnormal returns are negatve, but none are statstcally sgnfcant. Next, we compare stock prce rsk, measured by standard devatons of the stock returns, before and after the recalls for dfferent wndows around the event days. For the 30 day wndow, sx frms experenced a statstcally sgnfcant ncrease n the volatlty of the returns after the Kraft Foods recall. Kraft Foods s not among the affected frms. Wth respect to the Kellogg recall, for the 30 day wndow, fve frms experenced a statstcally sgnfcant ncrease n volatlty after the event. Agan, Kellogg s not among the affected frms. A statstcally sgnfcant ncrease n volatlty s also detected for larger wndows for several frms. No statstcally sgnfcant ncreases n the volatlty are found for the 150 day wndow for the Kraft Foods recall, and only Quaker Oats and Keebler show sgnfcant ncreases n volatlty for the 8

11 150 day wndow for the Kellogg event. Such results for a large post event wndow are expected because the potental effect of the recall should dsappear after a larger perod. 4 Next, the dfferences n rsk, estmated wth the CAPM-based measure of nondversfable rsk, beta, before and after the events are statstcally tested. If the values of the beta coeffcent have ncreased for companes affected by the recalls, one can nfer that recalls have ncreased rsk, whch nvestors have to bear. A Chow F-statstc s used to test shfts n beta. Table 1 reports the results of the tests. General Mlls and Frto-Lay experenced sgnfcant changes n nondversfable rsk mmedately after the Kraft Foods recall. Kraft s not among the affected frms. The pre-event beta and post-event beta of General Mlls, estmated over a 30 day wndow, are 0.16 and 0.0 respectvely. That s, the beta changed n sgn and the absolute magntude of the beta decreased. The Chow test shows ths change n sgn. However, the nondversfable rsk of General Mlls decreased, not ncreased as was expected. For the 30 day wndow, Frto-Lay s pre-event beta s 0.16 and the post-event beta s That s, the sgn on the beta changed, whch reflects the change of the drecton of the stock movement relatve to the market, and beta ncreased n absolute magntude reflectng the ncreased nondversfable rsk. ADM, Unlever, LTD and Unlever, NV show sgnfcant changes n nondversfable rsk, but these changes happened for larger post event tme perods, whch suggests that these changes are not due to the consdered recalls. Kellogg recall had a greater mpact on the nondversfable rsk of the frms. Table 1 ndcates Kraft, Frto-Lay, ADM, Unlever, LTD and Unlever, NV experenced sgnfcant changes n the nondversfable rsk mmedately after the recall. However, Kellogg s not among the affected frms. Betas of these frms were negatve before the event, but postve and greater by absolute magntude after the event. That s, there s some evdence that the Kellogg recall ncreased the rskness of these frms. Fgure 3 shows down movement of ConAgra daly stock prce after the ConAgra s announcement that t had closed ts only corn mll. Seven frms, ncludng ConAgra, experenced negatve abnormal returns on the event day, but none of these abnormal returns were statstcally dfferent from zero. None of the cumulatve abnormal returns over the [-5, t] event wndows, where t belongs to [-5, 10] nterval, were statstcally sgnfcant. 5 Agan, no mpact of the event and no persstence of negatve abnormal returns, as t was wth the Kellogg recall, were detected. The volatltes of the returns of four of the consdered stocks, ncludng ConAgra, dd change mmedately after the event. For ConAgra, the results are sgnfcant for all perods consdered. For nne frms, volatltes changed not mmedately after the recall, but wthn the 50 day perod, whch would seem to ndcate the nfluence of other factors not related to the event. Chow tests (Table 1) ndcate changes n the market regme for ADM, Unlever, LTD and Unlever, NV. Conclusons Interest n food safety ssues s at a hgh level n the U.S., and food safety events affect busness and ndustry, consumers, and the general publc. Prevous research has found shareholder losses occur when companes are mplcated n serous food safety hazards and that stock market reacton to food safety ssues has dffered by ncdent and by frm. 9

12 The objectve of ths research was to quanttatvely examne the effects of Starlnk corn food safety events on the rsks and the returns of food and agrbusness frms, drectly and ndrectly affected by the events. In ths study, abnormal returns and cumulatve abnormal returns of 17 agrbusness frms around Kraft Foods and Kellogg s Starlnk corn food recalls and ConAgra s closed plant event were estmated and tested. Results suggest that the company enactng the recall does not necessarly experence the largest or even sgnfcant mpacts on ts stock returns. Only a few frms experenced sgnfcant abnormal returns on some of the days durng the event-wndow perod. The results of parametrc tests and Wlcoxon rank-sum tests are consstent only for some of these few frms. The levels of returns of the frms enactng the recalls were not affected by the events. Based on parametrc tests, none of the cumulatve abnormal returns over the [-5, t] event wndows, where t belongs to the [-5, 10] nterval, were statstcally sgnfcant. None of the cumulatve abnormal returns over ntervals of 5, 10, 30, 50 and 100 days after the event date were statstcally sgnfcant. Analyss of the overall rskness of returns before and after the events revealed that at most sx frms experenced greater volatlty after the events, and the changes n volatlty were sgnfcant for short post event wndows. The analyss of the nondversfable rsk before and after the events showed that only Frto-Lay experenced an ncrease n nondversfable rsk after the Kraft Foods recall, whch s consstent wth the results of the volatlty analyss. Kellogg recall had a greater mpact on frms rsk and returns and affected a greater number of frms: Kraft, Frto-Lay, ADM, Unlever, LTD and Unlever, NV experenced sgnfcant changes n the nondversfable rsk after the event. Kellogg recall occurred nearly 6 months after the Kraft Foods recall, however had greater effect. Ths result may suggest that as new food safety events occur that relate to pror events and addtonal nformaton s ncorporated n the market, the mpacts of food safety ssues such as food recalls may ncrease. The ConAgra event seems to have had no mpact on the level of the stock returns and some mpact on the volatlty of the returns. Probably, the nformaton about ConAgra was revealed too gradually to the market over more than one month, whch can explan the absence of a reacton on the level of returns but the presence of some reacton on volatlty. Results of ths analyss ndcate that t s not possble to reject the hypothess that Wall Street s not senstve (n terms of levels of returns) to the Starlnk corn food recalls. Overall rskness of the stocks, measured by volatlty, and nondversfable rsk of the stocks appeared to ncrease after the events for some of the frms. However, t s not clear whether these effects are only due to the Starlnk corn events or due to somethng else also, snce durng the event and post event wndows other announcements about the consdered frms, but unrelated to Starlnk corn, took place. We conclude that fnancal market reacton to the consdered Starlnk corn food safety events s very lmted. One of the possble explanatons s that management had n place proper tools to mnmze the mpact of the event to preserve the mage of the frm n the stock market. It could also be that the sales of contamnated product were a mnor component of the corporate revenue. Fnally, prevous research has found wealth losses occur when companes are nvolved n serous food safety hazard stuatons. However, the consumpton of foods contamnated wth 10

13 Starlnk corn may result n allerges, but does not lead to mmedate health concerns. Investors may perceve that nvolvement n the Starlnk corn food safety events does not destroy the frm s mage and does not sgnfcantly mpar the long-run value of the frm whch may explan the absence of statstcally sgnfcant stock market reacton to these events. 11

14 Endnotes 1 See Appendx for the detaled descrpton of the test. Here t was not possble to avod overlappng the market model perod for the ConAgra event wth market model perod for the Kraft Foods event. So, the market model estmaton perod s chosen to be 10 busness days. 3 Durng the market model estmaton perod, Kraft was owned by Phlp-Morrs; Quaker Oats was acqured by Pepsco n August 001; Frto-Lay was owned by Pepsco; Keebler was acqured by Kellogg n March 001; Taco Bell was owned by Trcon Global Restaurants; Unlever (UK) and Unlever (NV) trade separately. These ssues are accounted for n the stock returns used. 4 We also examned the standard devatons of the stock returns before and after the recalls for dfferent wndows around the [-5, 5] perod, but excludng the [-5, 5] wndow. Ths should leave out the ncreased volatlty of returns around the event, whch s reflected n the smaller number of sgnfcant changes of the volatltes of the stock returns. In general, the excluson of [-5,5] perod does not change the results. 5 Here and for the volatlty and nondversfable rsk analyses, the post event wndows are restrcted by 90 days to avod overlappng wth the Kellogg event. 1

15 References Baker, G.A. Consumer Preferences for Food Safety Attrbutes n Fresh Apples: Market Segments, Consumer Characterstcs, and Marketng Opportuntes. J. Agr. Resour. Econ., 4(July 1999): Campbell, John Y., Andrew W. Lo and Crag A. MacKnlay. The econometrcs of fnancal market. nd edton, Prnceton Unversty Press, New Jersey, Center for Research n Securty Prces (CRSP) Database. Unversty of Chcago, Chcago, Illnos. Source of stock returns. Datastream Factva Fama. E.F., L. Fsher, M.C. Jensen, and R. Roll. The Adjustment of Stock Prces to New Informaton. Int. Econ. Rev., 10 (February 1969): Food and Drug Admnstraton Henneberry, S.R., K. Pewthongngam, and H. Qang. Consumer Food Safety Concerns and Fresh Produce Consumpton. J. Agr. Resour. Econ., 4(July 1999): Henson, Spencer, and Maro Mazzocch. Impact of Bovne Spongform Encephalopthy on Agrbusness n the Unted Kngdom: Results of an Event Study of Equty Prces. Amer. J. Agrc. Econ., 84(May 00): Hoover s Onlne LexsNexs Lusk, J.L., and T.C. Schroeder. Effects of Meat Recalls on Futures Markets Prces. Agrcultural and Resource Economcs Revew, 31(Aprl 00):

16 MacKnlay, A.C. Event Studes n Economcs and Fnance. J. Econ. Lt. 35(March 1997): McKenze, A.M. and M.R. Thomsen. The Effect of E. Col O157:H7 on Beef Prces. J. Agr. Resour. Econ., 6(December 001): Msra, S.K., C.L. Huang, and S.L. Ott. Consumer Wllngness to Pay for Pestcde-Free Fresh Produce. West. J. Agr. Econ., 16(December 1991):18-7. Saln, V. and N.H. Hooker. Stock Market Reacton to Food Recalls. Rev. of Agr. Econ., 3(Sprng/Summer 001): Thomsen, M.R. and A.M. McKenze. Market Incentves for Safe Foods: An Examnaton of Shareholder Losses from Meat and Poultry Recalls. Amer. J. Agrc. Econ., 83(August 001):

17 Fgure 1. Kraft Foods daly stock prce around the Kraft Foods recall of Taco Bell taco shell products Stock Prce ($) News Recall /1/00 9/8/00 9/15/00 9//00 9/9/00 10/6/00 Date Fgure. Kellogg daly stock prce around the Kellogg recall of Mornngstar Farms Corn Dogs Stock Prce ($) News Recall 5.50 //01 3/1/01 3/8/01 3/15/01 3//01 3/9/01 Date 15

18 Fgure 3. ConAgra Foods daly stock prce durng the perod when ConAgra closed ts only corn mll Closed plant Stock Prce ($) News //00 10/9/00 10/16/00 10/3/00 Date Fgure 4. Abnormal returns around Kraft Foods event for some of 17 consdered frms, fve days before and 10 days after the event. Abnormal returns are calculated usng market model usng CRSP value weghted market ndex Abnormal Returns Event Days Kraft Dannon Sara Lee Keebler ADM Corn Products Internatonal Taco Bell 16

19 Fgure 5. Abnormal returns around Kellogg event for some of 17 consdered frms, fve days before and 10 days after the event. Abnormal returns are calculated usng market model usng CRSP value weghted market ndex Abnormal Returns Event Days Kellogg Kraft Unlever, NV General Mlls Quaker Oats Frto-lay Taco Bell 17

20 Table 1. Results of Chow tests for changes n CAPM-based measure nondversfable rsk, beta, before and after the events. Company 10/17/000 09/18/000 event 03/08/001 event event days days days days days days days days days days days days days Kraft *** *** *** *** *** ** ** Dannon Sara Lee General Mlls * *** ** * * Henz * ** *** *** Kellogg * * ** ConAgra Quaker Oats * Frto-Lay * ** * ** ** *** *** *** Keebler ADM * ** ** ** ** *** ** ** ** * * Corn Products Int'l * Lance Taco Bell Avents * ** ** Unlever. LTD ** *** ** ** *** *** *** *** *** *** ** *** Unlever. NV ** ** * *** *** *** *** *** ** ** ** *** Sgnfcant at 1% level. ** Sgnfcant at 5% level. * Sgnfcant at 10% level. 18

21 Appendx To conduct a parametrc test, we adopted the method descrbed n Campbell, Lo, and MacKnlay (1997). If we denote the length of the estmaton wndow as L 1, the length of vectors R and X s L 1. Then, the ordnary least squares (OLS) estmators of the market-model parameters θˆ and ˆ σ ε, usng an estmaton wndow of L 1 observatons, are: ˆ 1 θ = ( X X ) X R () 1 = ˆ ε (3) ˆ σ ε ˆ ε L1 ˆ ε = R X ˆ θ (4) Var ˆ 1 [ θ ] = ( X X ) σ (5) ε * L denotes the length of the event wndow. The vector of abnormal returns, ˆε, over the event wndow for securty can be found usng market model estmates of the normal returns: = ˆ θ, (6) * * * ˆ R X ε where R* s a vector of event-wndow returns on securty and X * = [ι R* m ] s a matrx wth a vector of ones n the frst column and vector of event-wndow market returns R* m n the second column. We make the assumpton that, condtonal on the market return over the event wndow, the abnormal returns are jontly normally dstrbuted wth zero condtonal mean and condtonal covarance matrx V : ˆ* ε ~ N(0, ) (7) V V 1 = Iσ + X ( X X ) X ' σ, (8) ε * ' * ε where I s a square dentty matrx of sze L. Dstrbutonal propertes (7) can be used to test the null hypothess that the event has no mpact on the mean or varance of the returns. Abnormal returns are aggregated nto cumulatve abnormal returns. Let CAR (t 1, t ) defne the cumulatve abnormal return for securty for nterval [t 1, t ], where t 1 t and t 1 and t belong to event wndow of length L. Note, when t 1 =t the CAR (t 1, t ) s smply AR (t 1 ). That means that a statstcal procedure to test CARs can be used to test ARs over the event wndow as well. Let γ be a vector of length L wth ones n postons [t 1, t ] and zeros elsewhere. The estmated CAR (t 1, t ) and ts varance are calculated as: 19

22 CAR (, ) ' ˆ t1 t = γ ε * (9) Var[ CAR ( t1, t)] = γ ' Vγ = σ ( t1, t). (10) ˆ The estmate of the varance of CAR (t 1, t ), σ ( t1, t ), are obtaned by substtutng (3) nto the expresson (8). Under the null hypothess, CAR t, t ) ~ N(0, σ ( t, )). (11) ( 1 1 t The standardzed cumulatve abnormal return s: SCAR ˆ ( t1, t) = CAR ( t1, t) / σ ( t1, t), (1) whch s dstrbuted as the Student t-statstc wth L 1 - degrees of freedom. Ths provdes for a test of null hypothess that CAR (t 1, t ) = 0, or the event has no mpact on the returns to be tested. 0