Do emotions matter? Coherent preferences under anchoring and ...

More documents

Recommendations

Info

ECOLOGICAL ECONOMICS 66 (2008) 700– 711707Table 2 – Mean WTP (€) for different sizes of the networkby starting bids (standard deviation in parenthesis)Starting bidSize of the walking path30 km 100 km 300 kmLowest (6.01€) 14.88 (2.38) 16.37 (2.51) 21.36 (2.49)Total sample 19.94 (2.44) 25.83 (3.86) 28.31 (3.76)Highest (48.04 €) 21.94 (3.69) 37.93 (4.05) 40.82 (4.01)not linear but U-shaped. The degree of anchoring declines asemotional intensity increases, reaching a minimum for anaverage value of EIS. At this point, anchoring effects are notsignificantly different from zero at the 99% level. Thus, eventhough anchoring effects are significant across the sample,those subjects with average EIS are not influenced by the firstbid in the valuation process.These results can be related to the evidence that purport aU-shaped relationship between human performance andemotional intensity (Ashcraft and Faust, 1994; Idzikowski andBaddeley, 1983). The “Yerkes–Dodson law” (Yerkes and Dodson,1908) states that performance requires an intermediatelevel of emotional intensity (Leibenstein, 1987; Kauffman,1999). When emotional intensity is too low there is insufficientattention and mental arousal, and short-term memory isblocked (Kahneman, 1973). When emotional intensity is toohigh thinking becomes disorganized, and there is difficulty torationally evaluate the benefits and costs of alternatives(Eysenck, 1982; Yates, 1990; Lazarus, 1991; Oatley, 1992).4.3. Emotional load and scope effectThe emotional state of the individual can also have an influenceon his ability to perform according to coherent preferences, i.e.successfully passing the scope test. This hypothesis can beappreciated by looking at the relationships between the EIS andthe size of the walking paths to be rehabilitated in the policyproposal. Table 4 shows the mean WTP for the different levels ofthe walking paths network according to three groups ofindividuals as bunched by their level of emotional state (low,average, and high). Fig. 3 depicts the general relationshipsbetween the mean WTP and the variable KIL for the three typesof emotional profiles considered.It can be seen that WTP is less sensitive for initial changes inscope (from 0 to 30 km) when the emotional scale is low. Thissensitivity rises as the emotional scale increases, from 15.20 € to31.21 €. In addition, for further increases in the size of thewalking paths (beyond 30 km) WTP remains invariant for thegroups of high and low emotional scales. Thus, the scope test isfailed for subjects with extreme emotional scales (low and high).These subjects also posed a large degree of anchoring effects.However, the group of individuals with average emotional stateshowed a steeper valuation function in relation to scope (i.e. toTable 3 – Anchoring effects parameter by EIS levelsLow EIS Avg EIS High EISAnchoring effect (η i ) 0.42 (0.10) 0.15 (0.07) 0.43 (0.11)Confidence interval (99%) [0.16, 0.67] [−0.01, 0.30] [0.17, 0.68]Fig. 2 – Estimated relationship between anchoring effectsand EIS.the level of km). Thus, only those subjects with an averageemotional state are likely to behave according to coherentpreferences with no anchoring effect.These results can be seen as giving some support to thehypothesis claimed by Hsee and Rottenstreich (2004) thatunder “valuation by feelings” preferences tend to be verysensitive to changes in the initial values of the good and veryinsensitive to changes for higher values; and under “valuationby calculations” preferences are less influenced by changes ininitial values, and more sensitive to further values.5. ConclusionsEmotions are present everyday in human life. Several theoriesin neurosciences, psychology and sociology acknowledge itscentral role in explaining human behavior. Although someeconomists have recognized the role of emotions in explaininghuman behavior (e.g. Smith, 1759, Commons, 1934), the overallsignificance of emotions has been virtually ignored in theeconomics literature in general, and in environmental valuationin particular, until recent times. In this paper we have tried toempirically test the role of emotions in non-market valuationand the potential trade-offs between cognitive and emotionalintensity. The study utilizes a reduced emotional intensity scalescore which has been successfully used in other sciences tomeasure the emotional intensity of the individual.The cognitive dimension of the formation of individual'spreferences is studied by considering the anchoring effects in theDBDC elicitation process in the contingent valuation method. Asimultaneous equation Bayesian approach has been used toestimate the individual's anchor to the first bid price. The resultsshow that the answers to the second question are anchored bythe bid price offered in the first question. Thus, anchoring is aTable 4 – Mean WTP (€) of the size of the network by EISlevels (standard deviation in parenthesis)30 km 100 km 300 kmLow EIS 15.29 (1.19) 16.64 (2.55) 16.87 (2.18)Avg EIS 19.83 (3.73) 25.80 (2.46) 28.29 (4.07)High EIS 31.21 (5.97) 38.77 (2.01) 37.68 (4.72)See Li (1998) for details.
708 ECOLOGICAL ECONOMICS 66 (2008) 700– 711Fig. 3 – Estimated relationships between sensitivity to scope(km) and WTP for three EIS levels.relevant empirical effect that influences the elicitation of thevalue of non-market goods using DBDC.However, in our empirical application we found that preferenceswere sensitive to the scope of the good, suggesting thatthe embedding effect was not relevant in this context, since thesubject reacted significantly to the different dimensions of thegood to be valued. Further, the sensitivity of WTP to the scope ofthe good was also found for the various levels of the anchorgiven by the first bid price; in other words, the “coherent” relationshipbetween WTP and the dimension of the good was notaffected by the bid price offered in the first binary question.Thus, we can conclude that anchoring effects do not seem tohave a relevant influence on the individual's ability to discernamong different dimensions of a non-market good in a valuationscenario. A useful implication might be that relative WTP couldbe successfully elicited. Nevertheless, anchoring effects are stillpresent in our application and raise serious concerns about theunderlying nature of human values and the capability ofpreference elicitation techniques to capture them. If individualchoices are influenced by external anchors, it can be questionedhow can preferences be defined and how can elicitationtechniques be able to seize them. As pointed out by Ariely et al.(2003), “even if there are no clear violations of the transitivityaxiom, the researcher cannot ascertain whether elicited choicesreveal a set of unique and well-defined preferences”.In order to shed light on the conditions under which bidanchors could have an influence on the formation of individualpreferences, we looked at the potential role of individual'semotions. In particular, we focused on the relationship betweenthe EIS and the degree of anchor raised by the bid offered in thefirst binary question. The results show that anchoring effectsdecline as emotional intensity increases, reaching a minimumfor an average value of EIS. After this point, anchoring effects areagain significant. The major implication of these findings is thatindividuals tend to improve their ability for the non-marketvaluation task when their emotional intensity is moderate.On the other hand, the omission of EIS in the valuationfunction could bias non-market valuation results. The correlationbetween cognitive and emotional intensity implies thatsome index of the latter needs to be considered in the systematicpart of the utility or expenditure functions. To our knowledge,previous work has generally assumed independence betweenthe cognitive and emotional dimensions. The common incorporationof unexplained emotional conditions as part of thestochastic term introduces correlation between the systematicand the stochastic parts, leading to biased results.Some researchers have claimed that the finding of insensitivityto scope in some applications is a clear evidence of theinability of stated preference methods to capture preferences forpublic goods (Kahneman et al., 1999; Diamond and Hausman,1994) Our results suggest that the degree of sensitivity to scopecan be also related to the emotional load involved in thevaluation task. This might prompt a need for further evidenceon the role of emotions in the valuation of private and public orenvironmental goods.Our results concur with the notion that there might be a tradeoffbetween the emotional and cognitive dimensions in nonmarketvaluation tasks. That is, some degree of emotionalintensity might help reduce the cognitive load and enhanceperformance in human decision making. Nevertheless, it shouldbe acknowledged that this relationship is complex because of themultivariate factors that can influence individual's emotions andthe cognitive aspects involved in the survey instrument. Furtherresearch should explore the relationships of other emotional andcognitive factors that might play a role in the decision makingtask and the formation of individual's preferences.Appendix A. Estimation of the Bayesian model forDBDCIn this appendix we illustrate the application of the modeloutlined in Section 3 for the double-bounded dichotomous choicedata. For simplicity, let us decompose the joint bivariate normaldistribution for (ε i 1 , ε i 2 ) into the product of the marginal distributionof ε i 1 and the conditional distribution ε i 1 /ε i 1 ,thatis,WTP 1 i¼ x 1 i b 1 þ e 1 iðA1:1ÞWTP 2 i¼ x 2 i b 2 þ B 1 i g 21 þ e 1 i r 21 þ m i ðA1:2Þwhere ε 1 i =WTP 1 i −x 1 i β 1 , σ 2 =σ 2 2 −σ 2 21 , and v i ∼N(0, σ 2 ), ε 1 ∼N(0,1)are independents. Thus, the set of unknown parameters is θ={α,σ 21 , σ 2 }, where α=(η 21 , β 1 , β 2 ). The following independent priors areassumed:fðaÞfMVN a 0 ; w 1 0ðA1:3Þfðr 21 ÞfN r 0 ; b 1 ðA1:4Þ0f r 2 fIG a 02 ; c 0ðA1:5Þ2where MVN and N are a multivariate and univariate normaldistribution respectively, and IG is the inverted gamma distribution.Since we have no prior information on model parameters,very non-informative diffuse priors are assumed by consideringα 0 =c 0 =r 0 =α 0 =0, and large values for the parameters collecting thevariance (b 0 , ψ 0 ). Therefore, the joint posterior distribution takesthe following form,p P WTP 1 ; P WTP 2 ; r 12 ; r 2 ; ajY 1 ; Y 2¼ j ni¼1P nni 1 y 1ið Þð1 y 2 iÞ yy y 1Pi y2 iiP nyi 1 y 1ið Þy 2 iP yni y 1 1 y 2 iðiÞ fðaÞfðr 21 Þf r 2 ðA1:6Þ
Page 3 and 4: 702 ECOLOGICAL ECONOMICS 66 (2008)
Page 6 and 7: ECOLOGICAL ECONOMICS 66 (2008) 700-
Page 11 and 12: 710 ECOLOGICAL ECONOMICS 66 (2008)

Do emotions matter? Coherent preferences under anchoring and ...

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?