The Welfare Effects of Pfiesteria- Related Fish Kills: A Contingent Behavior Analysis of Seafood Consumers

Transcription

1 University of Delaware From te SelectedWorks of George R. Parsons October, 2006 Te Welfare Effects of Pfiesteria- Related Fis Kills: A Contingent Beavior Analysis of Seafood Consumers George R Parsons, University of Delaware As O Morgan, Appalacian State University Jon C Witeead, Appalacian State University Tim C Haab, Oio State University Available at: ttps://works.bepress.com/george_parsons/3/

2 Te Welfare Effects of Pfiesteria-Related Fis Kills: A Contingent Beavior Analysis of Seafood Consumers George R. Parsons, As Morgan, Jon C. Witeead, and Timoty C. Haab We use contingent beavior analysis to study te effects of pfiesteria-related fis kills on te demand for seafood in te Mid-Atlantic region. We estimate a set of demand difference models based on individual responses to questions about seafood consumption in te presence of fis kills and wit different amounts of information provided about ealt risks. We use a random-effects Tobit model to control for correlation across eac observation and to account for censoring. We find tat (i) pfiesteria-related fis kills ave a significant negative effect on te demand for seafood even toug te fis kills pose no known treat to consumers troug seafood consumption, (ii) seafood consumers are not responsive to expert risk information designed to reassure tem tat seafood is safe in te presence of a fis kill, and (iii) a mandatory seafood inspection program largely eliminates te welfare loss incurred due to misinformation. Key Words: pfiesteria, seafood demand, non-market valuation Pfiesteria piscicida is a single-celled microorganism, a toxic dinoflagellate, found in te sediments of many estuaries in te Mid-Atlantic region of te United States. It as been identified as te cause of many fis kills in tis region. Tousands, even millions, of fis can die in a single kill. During periods of warm weater and ig nutrient concentrations, pfiesteria becomes a toxic predator to certain species of fis. Wile te scientific evidence suggests tat tese outbreaks are letal to te fis, it also suggests tat tey pose no ealt risk to umans in te seafood market. 1 George R. Parsons is Professor in te College of Marine Studies and Department of Economics at te University of Delaware in Newark, Delaware. As Morgan is Visiting Assistant Professor in te Department of Marketing and Economics at te University of West Florida in Pensacola, Florida. Jon C. Witeead is Associate Professor in te Department of Economics at Appalacian State University in Boone, Nort Carolina. Timoty C. Haab is Associate Professor in te Department of Agricultural, Environmental, and Development Economics at Oio State University in Columbus, Oio. Tis work was (partially) supported by Grant No. NA86RG0036 from te National Sea Grant College Program, National Oceanic and Atmosperic Administration, to te Nort Carolina Sea Grant College Program. Te autors tank Bob Cristian, Isaac Kwakye, Emily Boyd, and Paul Hindsley for teir contributions. 1 For more information on Pfiesteria and its uman ealt risks, see Kleindinst and Anderson (2001). A good website wit links to bibliograpies, background information, and public response is usda.gov/wqic/pfiest.tml. Neverteless, media coverage of pfiesteria-related fis kills as led to rater large reductions in seafood consumption during periods of an outbreak. Te associated loss in economic welfare is potentially quite large and is seemingly due to misinformation. In tis paper we measure te welfare effects of a ypotetical pfiesteria outbreak using contingent beavior analysis in a seafood demand model. We also consider te effects of different forms of information provision on attenuating te losses due to misinformation. Identifying te forms of information provision tat ave te largest positive impact on consumer beavior will provide important policy-based information for related government agencies and industry representatives seeking to reassure consumers of product safety and ealt concerns. Our researc follows a framework developed by Sulstad and Stoevener (1978), wo measured te welfare losses incurred by Oregon s peasant unters in reaction to news of mercury contamination in peasants. Since ten, researcers ave considered te impact of newsinduced ealt scares on te demand for a variety of goods. See, for example, Swartz and Strand (1981), Smit, van Ravenswaay, and Tompson (1988), Brown and Scrader (1990), Wessells Agricultural and Resource Economics Review 35/2 (October 2006) Copyrigt 2006 Norteastern Agricultural and Resource Economics Association

3 Parsons et al. Te Welfare Effects of Pfiesteria-Related Fis Kills 349 and Anderson (1995), and Wessells, Miller, and Brooks (1995). Ours is te first to consider pfiesteria-related fis kills and te first to use contingent beavior tecniques to elicit consumers stated preferences in tis context. We begin wit a brief discussion of our survey and study design before turning to te model. Survey and Study Design Contingent beavior or stated preference tecniques are often used to measure consumer preferences. Individuals are asked to respond to survey questions pertaining to a market or non-market good. Te provision of te good is altered in some fasion and te individuals are asked ow tey migt respond to tat cange. In our case, respondents are asked ow teir seafood consumption migt cange in te presence of a pfiesteria-related fis kill. We conducted a pon -pone survey of seafood consumers over te age of 18 in Delaware, Maryland, Wasington, D.C., Virginia, and Nort Carolina in Te sample frame was stratified based on a split between urban and rural areas and a split between Nort Carolina and te oter four areas. Pfiesteria outbreaks are common in te Mid-Atlantic, and we ad a particular interest in Nort Carolina in our project. Te goal was to conduct te survey during te fis kill season: June troug November. Te first pone survey was conducted from August to October. Te second pone survey was conducted from October to November. Te mail portion of te survey was mailed out to individuals between te pone surveys and contained information about pfiesteria. Two focus groups were conducted to develop te pfiesteria information packet for mailing. Te first focus group was conducted in Wasington, Nort Carolina, and included five members of a local environmental organization. Te second focus group was conducted in Baltimore, Maryland, wit ten members of a curc group. During eac session, te facilitators presented sections of te information mailout and asked participants for teir tougts on wat information tey tougt te text and visual aids conveyed. Overall, participants found te information straigtforward. Were appropriate, suggestions received during tese sessions were incorporated into te final version of te mailed information. Te survey questions were developed wit input from participants in an East Carolina University undergraduate environmental economics course and during 15 one-on-one (telepone and in-person) interviews. Participants in te one-onone interviews were cosen based on convenience. Tese sessions focused on question wording, organization, and skip patterns. Suggestions received during tese sessions were incorporated into te final version of te questionnaires. A pretest of 160 seafood consumers in Delaware, Maryland, Nort Carolina, and Virginia was also conducted during June July Frequency and statistical analyses of te pretest data revealed no major flaws in te questionnaire. Only minor canges were made to te questions. Te first pone survey used random digit dialing and screened people based on weter or not tey ate seafood. 2 Te survey was designed to collect information on seafood consumption patterns, costs, knowledge of pfiesteria, and socioeconomic caracteristics of respondents. In addition, eac respondent was asked ow is or er number of seafood meals consumed (montly) would cange if te price of seafood were to rise and to fall. Te actual questions appear in Table 1 as Questions 1 and 2. Individuals were recruited in te initial pone survey to participate in a follow-up pone survey. Between pone calls, individuals were sent an information mailout 3 wic included te following: a ypotetical press release describing a pfiesteria-related fis kill information describing pfiesteria and its ealt risks a two-sided color pamplet describing a new seafood inspection program. Te press release described eiter a major or a minor kill. A major kill involved undreds of tousands of fis over a large area of a river. A minor kill involved fewer fis over a smaller area. Eac respondent received one or te oter of tese press releases split about equally across our sample. Te fis kill was on te Neuse River in 2 Seafood meals are defined as finfis or sellfis meals consumed at ome or in a restaurant. Not included are canned seafood meals or seafood meals consumed at oter people s omes. Frozen seafood meals were eligible. 3 Please contact te autors for copies of all te materials sent to respondents.

4 350 October 2006 Agricultural and Resource Economics Review Table 1. Five Contingent Beavior Questions Question # Question 1: Price up Question 2: Price down Question 3: Fis kill Question 4: Fis kill wit inspection Question 5: Fis kill wit inspection and price increase Wording Seafood prices cange over time. For example, if a lot of fis are caugt, prices go down. Wen fewer fis are caugt, prices go up. Suppose te price of your portion of your average seafood meal goes up by $X but te price of all oter foods stays te same. Compared to te [NUMBER] meals you ate last mont, do you tink you would eat more, less, or te same number of meals next mont wit te iger price? (X is randomly assigned $1, $3, $5, or $7) Ten, About ow many more/less seafood meals do you tink you will eat next mont? Now suppose te price of your average seafood meal goes down by $X, but te price of all oter foods stays te same. Compared to te [NUMBER] meals you ate last mont, do you tink you would eat more, less, or te same number of meals next mont wit te lower price? (X is randomly assigned $1, $2, $3, or $4) Ten, About ow many more/less seafood meals do you tink you would eat next mont wit te lower price? Tinking about seafood meals again, suppose tat te average price of your seafood meals stays te same. Compared to te [NUMBER] meals you ate last mont, do you tink you would eat more, less, or te same number next mont after te fis kill? Ten, About ow many more/less seafood meals do you tink you would eat next mont after te fis kill? Now suppose te average price of your seafood meals stays te same. Compared to te [NUMBER] meals you ate last mont, do you tink you would eat more, less, or te same number next mont after te fis kill and wit te mandatory seafood inspection program? Ten, About ow many more/less seafood meals do you tink you would eat next mont? Suppose tat wit te mandatory seafood inspection program te price of your portion of your average seafood meal goes up by $X, but te price of all oter food stays te same. Compared to te [NUMBER] meals you ate last mont, do you tink tat you would eat more, less, or te same number next mont after te fis kill? (X is randomly assigned $1, $3, $5, or $7) Ten, About ow many more/less seafood meals do you tink you would eat next mont? Nort Carolina for Nort Carolina residents and on te Pocomoke River in Maryland for all oters. Te respondents also received a map pinpointing te location of te event. Te information sent to respondents describing pfiesteria and its ealt risks came in tree different forms: (i) no information, (ii) a brocure, or (iii) a brocure and insert. Eac respondent received one or te oter of tese packets split about equally across our sample. Te brocure explains wat pfiesteria is and notes tat te risks of eating seafood are not canged as a result of te fis kills related to pfiesteria outbreaks. 4 Te insert is more direct and empasizes tat tere is 4 Te brocure is based on a brocure publised by te U.S. Environmental Protection Agency s Office of Water titled Wat you sould know about Pfiesteria Piscicida. Te brocure and insert information was simplified, sortened, and revised based on comments received from focus groups and from a review by an ecologist familiar wit te pfiesteria scientific literature. no scientific evidence linking pfiesteria outbreaks to increased ealt risks in seafood consumption. Finally, eac respondent was sent a sort description of te National Oceanic and Atmosperic Administration (NOAA) voluntary seafood inspection program. Te second pone survey ten focused on our next tree contingent beavior questions: Questions 3, 4, and 5 in Table 1. Question 3 asked individuals ow tey would cange teir seafood consumption if te pfiesteria-related fis kill reported in te press release were to occur. 5 Question 4 asked te same question, but told respondents to assume tat te government safety inspection program described in te pamplet was 5 Respondents were asked prior to questioning in te second survey if tey ad read te brocure and/or te insert. If tey ad not, ten tey were asked if tey were prepared to do so, and told tat tey would be called back at a later date, allowing tem time to read te information.

5 Parsons et al. Te Welfare Effects of Pfiesteria-Related Fis Kills 351 in operation. Question 5 asked te same question but told respondents tat te safety program was in operation and tat te price of seafood would increase as a result. Tese questions were designed to ascertain weter te seafood demand function sifted in te presence of a fis kill and if te inspection program attenuated tat sift. Te different treatments also allowed us to examine te extent to wic demand sifts differ wit different size fis kills and different information provided about ealt risks. Te first pone survey generated a sample of 1,790 respondents. Te response rate was 61 percent completed interviews divided by contacts, were contacts include refusals and completed interviews. Of tese 1,790 respondents, 845 completed te second pone interview a response rate of 47 percent. Te mean annual income of our respondents was about $50,000, te mean age was 47, and te mean education level was 2 years beyond ig scool. Tirty-six percent were male and 71 percent were wite. All statistics are weigted to account for te sample stratification. Model We treat a pfiesteria-related fis kill as a factor affecting an individual s perception of te ealt risks associated wit consuming fis. Tat perception, in turn, affects te individual s demand for seafood meals. In our analysis a seafood consumer as an indirect utility function over a fixed time period of te form v = v(p,q,y,(s); c), were p is te price of a seafood meal, q is te price of a composite of all oter goods, y is income for te relevant time period, is te perceived quality of seafood, s is a vector of attributes tat govern an individual s perception of quality, and c is a vector of individual caracteristics accounting for eterogeneity of te population. Following conventional consumer teory, we expect ( v/ p) < 0, ( v/ q) < 0, ( v/ y) > 0, and ( v/ ) ( / s i ) or 0. Te term si is one of i elements in te vector s. Te elements can affect perceived ealt risks positively or negatively, and in our application will pertain to te ypotetical pfiesteriarelated fis kill and information on te ealt risks associated wit a kill presented in our contingent beavior question. Roy s Identity implies an uncompensated demand function for seafood meals of te form ( v/ p)/( v/ y) = x( p, q, y, ( s); c ). In our application we use linear forms for (s) and x(p,q,y,(s);c) to estimate seafood demand and te impact of fis kills on demand. First, consider te contingent beavior questions for a cange in te price of seafood. Individuals are asked ow muc teir quantity demanded would cange wit a ypotetical cange in price. Let x be te reported cange in te quantity demanded and p be te size of te ypotetical price cange. In our demand model, (1) x =β p+β q + β y+β 0 p q y α s+ βcc is te demand at te current price p, and (s) = α s. Similarly, (2) x =β ( p+ p) +β q β y+β α s+ β c 1 + p q y c is te demand at te new price p+ p. Subtracting equation (1) from equation (2) gives a demand difference (3) x =β p, p were x = x 1 x 0 is te reported cange in te quantity consumed in response to te ypotetical price increase. Te term βq( q q) +β y( y y) + β α (s - s) + α c(c - c) drops out of te demand difference by design. In te contingent beavior question tere is no variation in income, oter prices, risk factors, or individual caracteristics between te current state and te ypotetical state. In our application we estimate β p using equation (3). Variation in price comes from te survey design individuals receive different p s in te contingent beavior questions. For a price increase, p takes on a value of $1, $3, $5, or $7. For a price decrease, it takes on a value of -$1, -$2, -$3, or -$4 (see Questions 1 and 2 in Table 1). We estimated separate equations for price-up and price-down. Tese are (4) x =β p +ε Q1 pu up Q1 x =β +ε p. Q2 pd down Q2

6 352 October 2006 Agricultural and Resource Economics Review Te metod is te same for estimating sifts in demand due to te fis kill analyzed in te last tree contingent beavior questions. In tis case, (5) x =β p+β q + β y+β 1 p q y α ( s+ s) + βcc is te demand wit te ypotetical fis kill and s is a vector of te cange in te factors tat affect perceptions of risk. Subtracting equation (1) from equation (5) gives (6) x =β α s. x = x 1 x 0 is te reported cange in te quantity consumed in response to te ypotetical fis kill, and β p (p p)+β q (q q)+β y (y y)+β c (c c) drops out of te demand difference since tere is no cange in p, q, y, and c between te current and ypotetical states in te contingent beavior question. Some elements in s, owever, do cange by design, wic gives rise to te specification in equation (6). Now, consider Question 3 in Table 1. Individuals face eiter a major or a minor fis kill and are given one of tree levels of information: (i) no information, (ii) a brocure, or (iii) a brocure and an insert. Tis gives te following form of our demand difference (7) x = β Q3 α major-kill + 1 βα minor-kill + 2 βα brocure + 3 βα brocure insert +ε, 3 4 & were te rigt-and side variables are our s s. We ave major-kill (= 1 if kill is major), minorkill (= 1 if te kill is minor), brocure (= 1 if respondent received brocure), and brocure & insert (= 1 if respondent received brocure and an insert). Te coefficients on major-kill and minor-kill are expected to be negative. Te ypotesis is tat individuals ave misperceptions about te dangers of seafood consumption believing it is dangerous to eat after a pfiesteria-related fis kill wen in fact te danger is sligt. Te coefficients on brocure and brocure & insert are expected to be positive information on risk sifts demand back to te rigt. Te ypotesis is tat te safety information counters te misperception of seafood ealt risks and reduces te extent of te Q leftward sift. Te latter is a recovery of lost welfare due to poor information. In Question 4, everyone is asked ow is or er response to Question 3 would differ if a seafood inspection program ad been in place. Question 5 is te same as 4 except tat individuals are told tat te inspection program will increase te price. Te price increase was $1, $3, $5, or $7. Te equations for Questions 4 and 5 ten are (8) x = β α major-kill + Q4 1 βα minor-kill + 2 βα brocure + 3 βα brocure & insert + 4 βα inspection +ε 5 Q 4 (9) x = β α major-kill + Q5 1 βα minor-kill + 2 βα brocure + 3 βα brocure & insert + 4 βα inspection + 5 price for inspection + ε, 6 Q5 β α were inspection equals 1 if te inspection program is in place, and price for inspection equals te price increase per seafood meal due to program. Introduction of a seafood inspection program, inspection, would presumably work to sift demand back to te rigt we expect a positive coefficient. Te price for inspection sould dampen te extent of te rigtward sift since consumers realize tey ave to pay for te program we expect a negative coefficient. We estimate equations (4), (7), (8), and (9) simultaneously as a linear model wit eigt parameters. 6 Simultaneous estimation allows us to constrain parameters across equations to be constant and to estimate te model wit random effects. Random effects allow te error terms in te model to be correlated across equations for eac observation. It stands to reason tat te same unobserved elements tat influence an individual s sift in demand due to a fis kill witout an inspection program will also influence tat indi- 6 Te eigt parameters are β pu, β pd, and β α 1 troug β α 6. Since te individual parameters β and α i are not identified in our model, we estimate β α i as a single parameter for eac i. Tis as no bearing on our final welfare calculations.

7 Parsons et al. Te Welfare Effects of Pfiesteria-Related Fis Kills 353 vidual s sift wit an inspection program in place. Since all observations in te sample do not make it to te second survey and since tere is some attrition due to simple cleaning of te data, an unbalanced version of a random effects model is estimated. ave a brocure and an insert, (iv) an inspection program is in place, and (v) an inspection program is in place and tere is a price rise. 9 Tere are several noteworty findings. First, te effects of a price increase and a price decrease differ te slope of te demand function is larger for a decrease tan for an increase. Te coefficient on p down is -.346, and te coefficient on p up is At te mean, tis is a price elasticity of demand of.78 for a price decrease and.49 for a price increase. In effect, tere is a kink in te demand function at te point of current consumption. Quantity demanded seems to be more responsive to a price decrease tan a price increase. Tis finding appears to be consistent wit teories of loss aversion tat individuals value losses more igly tan gains of equivalent magnitude. One may be inclined to argue tat tis is due to individuals inability to reduce consumption beyond teir current level, tereby capping te response to price increases. However, keep in mind tat we ave estimated a version of te model tat accounts for truncation at current consumption. Te model is also estimated as a Tobit regression wit censoring at x, te negative of te quantity consumed. Tis is because individuals cannot reduce teir consumption of fis by more tan te quantity consumed. Since individuals consume different quantities, te censoring point varies across observations. 7 Finally, we use te estimated model to report te cange in consumer surplus due to ypotetical major and minor fis kills. Tis surplus loss is sometimes called avoidance cost. It is te loss associated wit avoiding fis consumption wen in reality fis is safe to eat. It is te difference in an individual s consumer surplus wit and witout a fis kill. In te linear demand model, an individual s loss is {(x+ x) 2 x 2 }/ 2β p, were x is reported montly consumption, x is te reported cange due to a fis kill, and β p is from te estimated model. We report tese losses under different assumptions about information provision. 8 Second, te coefficients on major-kill and minor-kill are negative and significant as expected. Tis general result is supported by oter Results Te regression results appear in Table 2. Tese are random effects Tobit regressions wit censoring at te negative of te number of meals consumed. Table 3 sows te cange in surplus or avoidance cost due to minor and major fis kills per seafood meal. We report surplus losses assuming tat (i) individuals ave no information, (ii) individuals ave a brocure, (iii) individuals 7 Tere are a number of ways te model could ave been more complex econometrically. In principle, we ave a difference of two count data variables for our dependent variable. Tis introduced a number of complications tat make a simple count model (our first coice) for te demand differences infeasible: some of our differences are negative, te distribution of te difference of two count variables is not a simple count variable [see for example Consul (1989)], and we really ave a difference of two censored variables at two points. Te econometrics gets complicated and is not really sorted out in te literature as far as we can tell. Our purpose in tis paper is to present a simple slice of te data using some basic econometric tecniques. We tink tere are some interesting findings to sare in tis regard alone. 8 We report welfare canges using te price-up coefficient, β pu, in equation (4) since all te measures of surplus we consider are integrated over te portion of te demand curve corresponding to a price increase. studies [see Anderson and Anderson (1991) or Aluwalia, Burnkrant, and Unnava (2000)]. Wat is unexpected is tat te effect of a major kill and a minor kill are about te same. Tere is no statistical difference in teir coefficients. Te implication is tat te size and scope of a fis kill is not particularly important. Hundreds of tousands 9 Two caveats are wort noting ere. First, since we question people about te number of seafood meals (and cange in te number of seafood meals) consumed in a mont, in a sense tey ave recent information about te kill for eac meal in a mont. Following an actual kill, an individual will ave recent information for only a day or so, not a full mont. If an individual s reaction to information canges as time passes, our estimates will be biased. For example, a press release may ave a large impact in only te first few days, its impact tereafter diminising. Weter or not people mimic tat type of beavior in our survey is uncertain. To te extent tat tere is decay in te effect of a fis kill on consumption of seafood over a mont, we may be overstating te impact. Second, since we use a composite measure of seafood, we miss substitution across types of seafood tat may occur as a result of a pfiesteria outbreak. For example, if a person canges te type of fis e or se eats in response to te outbreak but does not alter is or er total seafood consumption, tere is a welfare loss. In our model, we would observe no cange in seafood consumption and no welfare loss. In tis respect, our analysis will understate welfare losses.

8 354 October 2006 Agricultural and Resource Economics Review Table 2. Regression Results a Parameter Estimates for Equations 4, 7 9 Variable Coefficient t-statistic p up Amount of price increase -.218* p down Amount of price decrease -.346* major-kill Dummy variable for major fis kill -1.19* -8.0 minor-kill Dummy variable for minor fis kill -1.27* -9.2 brocure Dummy variable for brocure included brocure & insert Dummy variable for information insert included inspection Dummy variable for inspection program in place 1.06* 8.0 price for inspection Amount of price increase due to seafood program -.183* -6.8 Sigma(v) Sigma(u) a Random-effects Tobit model wit censoring at te negative of te number of meals purcased per mont and allowing for correlation across 5 contingent beavior questions. Note: Asterisk means statistically significantly different from zero at te 99 percent level of confidence. Table 3. Avoidance Cost Estimates Due to Fis Kill a Average Cange in Consumer Surplus per Meal Information Scenario Major Kill Minor Kill Wit no information -$4.17 -$4.34 Wit brocure -$4.38 -$4.54 Wit brocure & insert -$4.20 -$4.37 Wit inspection program -$0.60 -$0.92 Wit inspection program and $1 price increase in meals -$1.37 -$1.65 a Average cange in consumer surplus per meal per person for a major and minor fis kill assuming different levels of information provision. of dead fis signal an increase in ealt risk comparable to tens of tousands of dead fis. 10 Te welfare loss associated wit te fis kills, ignoring for te moment te cases wit information provision and inspection programs, is on te order of $4 per meal. 10 One referee noted an alternative interpretation: peraps our contingent beavior survey failed to pass a scope test [see Hanemann (1994), p. 34]. Tird, information provision in te form of a brocure or a brocure along wit an insert appears to ave limited sway on consumers. Te coefficients on brocure and brocure & insert are statistically insignificant. It follows tat te welfare loss associated wit te fis kills assuming individuals ave a brocure or ave a brocure and te insert is about te same as te cost wit no information. Tis finding seems to suggest tat simply providing information based on experts judgments carries little weigt in altering individuals perceptions. It is also possible tat te manner in wic te information was packaged and presented was te cause for te limited impact people ignored it or found tat it lacked credibility. For example, te brocure is rater long and may simply be disregarded. Or, wat was intended to make consumers feel safe may ave inadvertently raised an issue tey ad not really considered before. For a discussion of te credibility of te sources of information, see Hovland and Weiss (1951), Sterntal, Pillips, and Dolakia (1978), Smit, Young, and Gibson (1999), Tse (1999), and Frewer, Scoldere, and Bredal (2003). Tese coefficients are consistent wit te argument tat positive information as less of an effect on consumer beavior tan negative media

9 Parsons et al. Te Welfare Effects of Pfiesteria-Related Fis Kills 355 coverage. Te negative press releases sifted demand significantly; te positive brocures sifted it only sligtly. Kroloff (1988) found tat te impact of media exposure gives negative news quadruple weigt compared wit positive news. Serrell et al. (1985) calculated tat it takes five times more positive information to offset te effects of any negative information. Fourt, te presence of an inspection program, unlike information provision, sifts te demand function significantly rigtward returning it close to its pre-fis kill position. Te coefficient on inspection nearly perfectly offsets te initial sift due to te ypotetical fis kill. Te coefficient is also statistically significant. Tis result is consistent wit Wessells and Anderson (1995), wo considered te role of a variety of measures of providing seafood safety assurances and found tat consumers placed a ig value on seafood inspection programs. So, te cost of te kill, wit an inspection program in place, drops dramatically, as sown in Table 3. Te curious ting ere is tat inspection programs would discover noting related to pfiesteria tat one could actually act upon to reduce risk, since pfiesteria poses no ealt treat to umans wo consume infected seafood to begin wit. In tis regard a program may comfort consumers, but it would be a somewat peculiar government response. Fift, te impact of a rise in seafood prices due to an inspection program is about te same as a general price rise a sensible result. Te coefficient on p up is and on price of inspection is Tis as te potential of offsetting some of te recaptured losses due to te inspection program. In Table 3 we present te welfare loss for a fis kill assuming an inspection program is in place and raising te price of fis by $1. Conclusions As expected, individuals react to fis kills by reducing consumption of fis, even toug te fis kill is unlikely to pose increased ealt risks. Tis result as been documented elsewere in te literature and suggests tat tere may be a role for government in providing information to consumers about risks. Wen individuals reduce seafood consumption, tey are said to incur avoidance costs welfare loss associated wit unnecessarily avoiding a desirable meal. Te benefit of a government information program ten is te avoidance cost saved by informing consumers. Te avoidance costs in question appear to be rater large. Using our model, te aggregate cost over te four-state region is on te order of $60 million per mont, depending on te amount of risk information provided to individuals. We found tat consumers were not responsive to expert risk information sent in a mail packet in te form of a brocure. Te brocure empasized tat eating fis after a kill was safe. For te most part, individuals beaved as tey would ave witout te information. Te savings in avoidance cost were small. Peraps experts ave little sway in ow individuals form perceptions of risk. Or, peraps our information packets and metod of dissemination failed to communicate te risk meaningfully or individuals simply ignore te information. On te oter and, we found tat consumers were quite responsive to seafood inspection programs. Avoidance costs are nearly eliminated by te ypotetical inspection program used in our experiment. Tis suggests tat consumers ave confidence in suc programs and tat concrete action by government autorities can affect consumer decisions. Tese results old even toug inspections programs, in principle, could discover noting related to pfiesteria tat one could actually act upon to reduce risk since pfiesteria poses no ealt treat in seafood. We also found tat te gain in surplus realized by suc programs can be easily dissipated if individuals believe te programs will lead to a rise, even a small rise, in te price of fis. Tere were a number of oter interesting findings. Individuals did not seem to differentiate between major and minor size fis kills. We surmised tat tere is some tresold level tat triggers a response by consumers and tat our kills surpassed tat tresold. We also found tat te people responded asymmetrically to price increases and price decreases people were more responsive to price decreases. References Aluwalia, R., R.E. Burnkrant, and H.R. Unnava Consumer Response to Negative Publicity: Te Moderating Role of Commitment. Journal of Marketing Researc 37(2):

10 356 October 2006 Agricultural and Resource Economics Review Anderson, J.G., and J.L. Anderson Seafood Quality: Issues for Consumer Researcers. Te Journal of Consumer Affairs 25(1): Brown, D.J., and L.F. Scrader Colesterol Information and Sell Egg Consumption. American Journal of Agricultural Economics 72(3): Consul, P.C On te Differences of Two Generalized Negative Binomial Variates. Communications in Statistics: Teory and Metods 18(2): Frewer, L.J., J. Scoldere, and L. Bredal Communicating about te Risks and Benefits of Genetically Modified Foods: Te Mediating Role of Trust. Risk Analysis 23(6): Hanemann, W.M Valuing te Environment Troug Contingent Valuation. Te Journal of Economic Perspectives 8(4): Hovland, C.I., and W. Weiss Te Influence of Source Credibility on Communication Effectiveness. Public Opinion Quarterly 15(4): Kleindinst, J., and D. Anderson Pfiesteria-Related Educational Products and Information Resources Available to te Public, Healt Officials, and Researcers. Environmental Healt Perspectives 109(5): Kroloff, G At Home and Abroad: Weiging In. Public Relations Journal 44(October): Serrell, D., R.E. Reidenbac, E. Moore, J. Wagle, and T. Spratlin Exploring Consumer Response to Negative Publicity. Public Relations Review 11(1): Sulstad, R.N., and H.H. Stoevener Te Effects of Mercury Contamination in Peasants on te Value of Peasant Hunting in Oregon. Land Economics 54(1): Smit, M.E., E.I. van Ravenswaay, and S.R. Tompson Sales Loss Determination in Food Contamination Incidents: An Application to Milk Bans in Hawaii. American Journal of Agricultural Economics 70(3): Smit, A.P., J.A. Young, and J. Gibson Hey Now, Mad Cow? Consumer Confidence and Source Credibility During te 1996 BSE Scare. European Journal of Marketing 33(11/12): Sterntal, B., L.W. Pillips, and R. Dolakia Te Persuasive Effect of Source Credibility: A Situational Analysis. Public Opinion Quarterly 42(3): Swartz, D.G., and I.E. Strand Avoidance Costs Associated wit Imperfect Information: Te Case of Kepone. Land Economics 57(2): Tse, A.C.B Factors Affecting Consumer Perceptions on Product Safety. European Journal of Marketing 33(9/10): Wessells, C.R., and J.G. Anderson Consumer Willingness to Pay for Seafood Safety Assurance. Journal of Consumer Affairs 29(1): Wessells, C.R., C.J. Miller, and P.M. Brooks Toxic Contamination and Demand for Sellfis: A Case Study of Demand for Mussels in Montreal. Marine Resource Economics 10(2):