• Keine Ergebnisse gefunden

RESULTS CHAPTER 3. EMBODIED TRUST With respect to perceived trustworthiness we observe the following: the coecient

On the consequences of not matching other's ideals

CHAPTER 3. EMBODIED TRUST 3.3. RESULTS

3.3. RESULTS CHAPTER 3. EMBODIED TRUST With respect to perceived trustworthiness we observe the following: the coecient

of LargeRespondent in Table 3.8 (and Table 53 in appendix C) is positive and signif-icant and indicates that, at mean distance, large respondents were signsignif-icantly more likely to be perceived trustworthy than skinny ones. Further, all interaction terms of LargeRespondent with the distance measures are signicant. Figure 3.2 shows how the likelihood that the respondent is perceived trustworthy decreases when the distance between the respondent's size and the proposer's female ideal increases. A centered distance of zero corresponds to the mean absolute distance and roughly to the societal female ideal of around 4 on the FRS. The gure shows that at a centered distance the large proposer is more likely to be perceived trustworthy than a skinny one. Similar observations are shown for each distance measure in Figures 10, 11, and 12 in appendix C.

1 2 3 4

Large Respondent 15.60*** 33.48*** 10.68*** 12.10**

15.16 35.37 9.58 12.1

Large Respondent x Centered Distance to Female Ideal 0.08**

Large Respondent x Centered Distance to Male Ideal 0.09 0.17*

Large Respondent x Centered Distance to Actual Size 0.17 0.20**

Large Respondent x Centered Distance to Preferred Size 0.16 0.19*

0.17

No. of Obs. 62 61 64 63

Standard errors in italics, * p<0.10 ** p<0.05 *** p<0.01

Ordered Logit Model, dependent variable: perceived trustworthiness of respondent; Session xed eects

Table 3.8: Trustworthiness and direction of distance

When we interact Af ro−T riniRespondent with the centered distance measures (see Table 42 in appendix C) we nd that a greater distance between the proposer's own, preferred and ideal body shapes and the respondent's body shape was on average associated with lower investments when the respondent was Afro-Trini, but this relation was not signicant, except with respect to the distance between the proposer's preferred size and the Afro-Trini respondent's size: An increase in distance was on average associated with a decrease in investment size of 5.7 TTD. With respect to the Indo-Trinis, the eect of a greater distance between the proposer's own, preferred and ideal

CHAPTER 3. EMBODIED TRUST 3.3. RESULTS

Figure 3.2: AME of large respondent for distance to female ideal

body shapes and the respondent's body shape on investment size was consistently positive but insignicant (represented by the coecients of CenteredDistanceto... in Table 42 in appendix C).

In Table 54 in appendix C the eects of the interaction ofAf ro−T riniRespondent with the centered distance measures on perceived trustworthiness are signicant. Fig-ure 3.3 shows the average marginal eect of Af ro−T riniRespondent at dierent values of the centered distance to the proposers female ideal: As long as the dierence between the respondent's body size and the proposer's female ideal is below the mean distance (when centered distance is zero), Afro-Trini Respondent's are more likely to be perceived trustworthy than Indo-Trini respondents. As the distance increases, the trust bonus decreases and eventually turns negative, meaning that Afro-Trini respon-dent's are less likely to be perceived trustworthy than Indo-Trini respondents. Figures 13, 14, and 15 in appendix C show similar results of the average marginal eect of being an Afro-Trini respondent for the other distance measures. On the other hand, in general large Indo-Trini respondents were more likely to be perceived trustworthy than skinny Indo-Trinis (represented by the coecient ofLargeRespondent in column 1, Table 54 in appendix C).

When instead of Af ro − T riniRespondent we use the indicator variable P + RShareEthnicityand interact it with the distance measures, we nd that a deviation

3.3. RESULTS CHAPTER 3. EMBODIED TRUST

Figure 3.3: AME of Afro-Trini respondent for distance to female ideal

from the proposer's own, preferred and ideal shapes is associated with a signicant decrease in the average investment size (see Table 43 in appendix C) and perceived trustworthiness (see Table 55 and Figures 16, 17, 18, and 19 in appendix C). However, when we use community xed eects the signicance is lost, but the direction of the eects remains (see Tables 49 & 61 in appendix C). These ndings hint that deviations from ethnicity specic ideals lead to a reduction of trust within the same ethnic group, but less so between groups.

With respect to the impact of SES, i.e. education and age, we would have assumed that people of higher education and older age and hence, higher income, would have been more likely to invest more, simply because they have to care less about the amount they earn or lose in the game. However, we do not nd an eect of either variable on investment size (see Tables 38 - 49 appendix C). On the other hand, we nd that an increase in age is associated with an increase in the likelihood to perceive the respondent trustworthy (see Tables 50 - 55 in appendix C). When we use community xed eects, the direction of the age eect remains, but signicance is partly lost (see Tables 56 -61 in appendix C).

3.3.2 Are larger people assumed to be more trusting?

In the previous section we analyzed decisions of proposers in round one: they made investments and after the investment decision they were asked how trustworthy they

CHAPTER 3. EMBODIED TRUST 3.4. DISCUSSION AND CONCLUSION thought the respondent they had just played was. We now turn to round two, where participants played the part of the respondent and received an investment. They were then asked to respond to this investment. After the response decision they were questioned about their assumptions concerning the characteristics of the person who might have sent the investment they had just received. If larger people were thought to be more trusting, we would expect people to assume that larger investments were sent by larger people.

1 2 3

Size of investment received in Round 2 1.03 1.06 1.04 0.03 0.05 0.04

No. of Obs. 62 58 61

Standard errors in italics, Logistic Regression, Odds Ratios reported. Model 2, CV:

Session FE and Enumerator knew participant personally; Model 3, Community FE and Enumerator knew participant personally

Table 3.9: Likelihood to assume proposer in round 2 was large

Table 3.9 (see also Table 62 in appendix C) shows that although the direction of the coecient is as expected, we do not nd a signicant eect of the size of the investment that respondents received on the likelihood that the respondent assumed that the investment was sent by a large person.