6. Willingness to Pay for Pro-Environmental Farming
6.5. Existence of Starting Point Bias and Information Effects
When information is perceived in a broad sense, the starting point bias can also be regarded as a certain type of information effect. If the respondent believes for some reason that the starting bid contains information about the "correct" value level of the valuation object, then this information will influence his WTP response. The same applies to virtually ali CVM design issues: any change in any element of the valuation instrument can be considered to alter the informa-tion content. This should be kept in mind when the effects of informainforma-tion are analyzed, because it may be difficult to identify how much different changes in questionnaire design have influenced the information content and further the individual WTP. However, although the exact sources of information effects could not be localized, the overall information sensitiveness of the mean WTP would indicate that people's valuations are highly dependent on the information provided. This should somehow be reflected in the interpretation of results.
As explained previously, six slightly different questionnaire formats were used in order to detect the possible starting point bias and information effects.
There were three different starting bids and two different information levels, which were created by changing the sequence of questions and including a very
brief package of additional information. The mean WTPs of each questionnaire category are printed in Table 6.4. The reported mean WTPs were calculated by using the combined bidding game-payment card elicitation method.
The type of the questionnaire has a recognizable influence on the mean. The higher the starting bid, the higher the average WTP. In addition, there appears to be a connection between additional information and higher mean WTPs. Statis-tical significance of the differences among the mean WTPs is examined in Table 6.5 by using a t-test, which is an appropriate method to compare group means when variables with continuous values are in question.
The t-test results give support to the existence of the starting point bias.
Especially, if a closer look is taken to the pairwise comparisons between Q1 and Q3 or between Q4 and Q6, this becomes obvious. In both cases, the difference between means is statistically significant, the risk of falsely rejecting the null hypothesis being much less than 5%. Thus, a large increase in the starting bid (from FIM 100 to FIM 1,300) raises WTP responses. There is also some indica-tion of the existence of the informaindica-tion effect, but the evidence is not statisti-cally significant enough. If the pairwise comparisons Q1-Q4, Q2-Q5, and Q3-Q6 are taken under inspection, we can see that the risk of falsely rejecting the null hypothesis is in each case higher than 5%. This is also true in the pairwise comparison between no-additional-info and additional-info. Of course, in the latter case not even a t-value less than 0.05 would be a convincing piece of evidence because of the joint-effect of information and starting bids. Corre-spondingly, comparisons between SB 100, SB 500, and SB 1,300 do not pro- Table 6.4. The Mean WTPs of Different Questionnaires and Their Combina-tions in Respect of the Starting Bid and Additional Information.
Type of the questionnaire Mean WTP, FIM
Q1 (SB=100 / no add. info) 116 17.5 272
Q2 (SB=500 / no add. info) 109 16.4 333
Q3 (SB=1300 / no add. info) 107 16.2 474
Q4 (SB=100 / add. info) 118 17.8 316
Q5 (SB=500 / add. info) 109 16.4 495
Q6 (SB=1300 / add. info) 104 15.7 544
No add. info (Q1+Q2+Q3) 332 50.0 357
Add. info (Q4+Q5+Q6) 331 50.0 446
SB100 (Q1+Q4) 234 35.3 294
SB500 (Q2+Q5) 218 32.9 414
SB1300 (Q3+Q6) 211 31.8 509
Ali 663 100.0 402
Table 6.5. T-test Values of Pairwise Comparisons of Different Questionnaires and Their Combinations.
Pairs under comparison Prob>IT1 Pairs under comparison Prob>ITI
Q1 -Q2 0.2968 Q3 - Q6 0.4845
Q1 -Q3 0.0145 Q4 - Q5 0.0545
Q1 - Q4 0.5383 Q4 - Q6 0.0098
Q1 - Q5 0.0110 Q5 - Q6 0.6372
Q1 -Q6 0.0009
Q2 - Q3 0.0779 No add. info - Add. info 0.0719
Q2 - Q4 0.7924
Q2 - Q5 0.0585 SB100 - SB500 0.0299
Q2 - Q6 0.0076 SB100 - SB1300 0.0004
Q3 - Q4 0.0742 SB500 - SB1300 0.1486
Q3 - Q5 0.8431
duce unambiguous evidence for the support or rejection of the null hypothesis because it is not possible to separate the influence of additional information from the starting bid. Thus, this kind of examination does not necessarily confirm or falsify the hypothesis about information effects.
The same ambiguous conclusion about the possible existence of information effect can be made when the logit standard model is applied. The rather big difference in mean WTP between those who received additional information and those who did not (FIM 384 — FIM 207 = FIM 177) seems to suggest that information has an identifiable effect. However, if the standard deviations of the mean WTPs (87 and 103, respectively) are taken into account, the approximate
"confidence intervals" (384+87 —> [297..471]) and (207±103 —> [104..310]) slightly overlap, indicating that the difference between mean WTPs is not statistically significant.
However, the combined effect of additional information and starting bids indicates that the changes in the information content do affect people's willing-ness to pay. For instance, the pairwise comparison between Q1 and Q6 shows this very clearly (t-value = 0.0009). Although we cannot really explain the influencing mechanism, we can assume that the information provided has had an impact through two channels. On the one hand, additional information has increased people's knowledge about positive ramifications of the conversion from conventional agriculture to pro-environmental farming. It is quite conceiv-able that when people learn more about the advantages of a policy proposal, they are also more willing to contribute to it. On the other hand, people may feel uneasy when confronted by the interviewer because they are not previously familiar with the context of the valuation method or situation. Consequently,
they do not really know how they are expected to act. Then, it is natural that any information given by the interviewer is considered valuable. In this kind of situation, the starting bid easily transforms into a benchmark that becomes an indication of the "correct" answer.
Because there exists preliminary although not properly confirmed evidence that information may have a considerable effect on WTP responses in certain occasions and conditions, it is meaningful to examine in more detail the rela-tionship between respondents' WTP and the information content delivered through the survey instrument. This is done in the next chapter by introducing an attitudinal dimension.