• No results found

Non-Aggregate Regression Analysis

Appendix to Chapter 2

A.2. ADDITIONAL ANALYSIS 131 characteristics (aggregate) PWB is significant in three out of four expected situations

A.2.3 Non-Aggregate Regression Analysis

In Appendix A.2.2, an aggregate regression analysis of the relationship between well-being and virtuous behavior has been provided. In this section, I try to extend this analysis by a similar non-aggregate regression analysis. The major message of this analysis is the following: Significance of the aggregate PWB measure in the aggregate regression analysis (Table A.3a-A.3c) is mainly driven by the SAI and SoWB, not by the PWBI. For SWB,

some measures occasionally become significant in the non-aggregate analysis for the DG and the SPD (especially when only a limited number of controls is used), but there seems to be no measure that is robustly significant across this two games reflecting results of the aggregate analysis. Only for my punishment settings, measures of positive and negative affect seem to be related across games, as will be discussed in more detail below.

Following KE, Table A.4a-A.4f present a summary of regression results for each game.

The regressions performed are very similar to those done in Table A.3a-A.3c. The only difference is that now non-aggregate well-being measures (as in Table 2.2a-2.2b) are used.

In each regression, one well-being measure is the dependent variable that is explained by measures of virtuous behavior (giver dummy, trustor dummy, cooperator dummy and punisher dummy) and controls. For each well-being measure, six different ordered logit regressions are run (specification 1-6) and each time only the coefficient of the virtuous behavior dummy and the dummy’s p-value are reported. For all Tables a-f, specification 1)All includes the same set of controls as Table A.3a and hence replicates KE’s analysis most closely. Specification 2)Conf6, restricts the number of observations to those participants who expressed at least a confidence level of six regarding the truthfulness of their answers to the well-being questions (and skips subjects’ level of confidence). The reason to look only at these observations is the following: During the experiment, I realized that some non-German subjects had difficulties to fully understand the meaning of all well-being questions. This might bias the results of the well-being questions for these subjects.6

6Overall, 9 out of 102 subjects stated a confidence level below 6. One subject chose a level of 2, another subject a level of 3, three subjects chose a level of 4 and four subjects chose a level of 5. 13 subjects chose a confidence level of 6, 26 a level of 7, 40 a level of 8, and 14 a level of 9. Evidence in favor of a language problem comes from the following observation: Overall, 18 out of 102 subjects (17%) were not Germans,

Specification 3)All - Demo. uses the full set of available demographic controls as in Table A.3b. Specification 4)Conf6 - Demo. again restricts the regressions with additional controls to those subjects with a confidence level equal or above six. Specification 5) All -Big 5 adds additionnally the -Big Five Inventory (as in Table A.3c) and specification 6) Conf6 - Big 5 restricts these regressions to those subjects with a confidence level equal or

above six.

What is the overall message of Table A.4a-A.4f? First, let’s focus on giving, trusting and cooperating: For SWB, KE found in their analysis a weak connection with giving.

In my setting, only some long-run SWB measures become occasionally significant in some situations and some specifications. However, there is no single SWB measure which is robustly related across specifications and games, which reflects that the relationship between SWB and virtuous behavior is already quite weak without any controls, as in Table 2.2a. For PWB, the situation is different. KE found a strong connection between PWB and giving. In my results, two out of three PWB measures seem to be more or less related with virtuous behavior across games and across different specifications: the Self-Actualization Index (SAI) and Social Well-Being (SoWB). Both measures are more or less significant in three out of four expected situations (generosity in the DG, trust and cooperation in the SPD, punishment in the DG-P). Including the Big Five Inventory (specifications 5 and 6) seems only to worsen significance levels in the DG, reflecting the results of the aggregate regression analysis of Table A.3c. The Index of PWB (PWBI), however, is not robustly related with virtuous behavior which reflects the smaller significance level of this measure in Table 2.2a-2.2b.

In sum, the fairly strong connection of PWB and virtuous behavior seems to be fairly but not perfectly robust against including additional controls (when abstracting from the weakness of the PWBI and when abstracting from the weakness of the SAI (SoWB) in the trust (DG-P punishment) setting). So overall, the significance of the aggregate PWB measure in Table A.3a-A.3c seems to be driven mainly by the SAI and SoWB, not by the PWBI. The fairly weak connection of SWB and virtuous behavior disappears when controlling for additional variables. The only robust exception for SWB measures is the change in mood (MID) when cooperating or not in the SPD. The change in current happiness is not significant, but the relevant Trustor and Cooperator dummies at least have the expected sign. Overall, the change in mood may be the more reliable measure.

For the change in happiness, only a single question is asked again whereas for the change in mood several questions are asked. Subjects may remember the happiness question and try to induce some artificial consistency.

Table A.4a-A.4f provide the same summary of regressions for the punishment settings:

but 5 out of 9 subjects (56%) with a confidence level below 6 were foreigners. This proportion falls to 23% (15%) foreingers for subjects with confidence level 6 (7). For reasons of consistency, all subjects with a confidence level below six are excluded and not only foreigners.

A.2. ADDITIONAL ANALYSIS 133 Mini-UG, SPD-P and DG-P. Already Table 2.2b showed a less clear relationship compared to Table 2.2a. When controlling for various factors, this relationship becomes even weaker.

Nonetheless, we can observe the following: If there is any significant relationship for the Mini-UG (Table A.4d) at all, positive affect (PAS and partly PA) seems to be negatively related with rejecting in the DG (as expected). Those who have experienced a lower degree of good emotions in the past seem to reject more. Additionally, for Lowest Happiness (LH) the Rejector dummy is not significant but has a negative sign (as expected). For the SPD-P, we observe a shift: The signs of these two measures changes. In SPD-P, positive affect (at least the PAS) is positively related with punishing whereas LH is negatively related. Those who have experienced a higher degree of good emotions in the past seem to punish more. For the DG-P, this observed shift is even stronger to the degree that not only the PAS seems to be positively correlated with punishing but also one measure of PWB: the SAI. Hence, only for the most altruistic form of punishment, we observe that (at least) one measure of PWB is significant.

Another interesting feature of the results is that the Negative Affect Schedule (NAS) is positively correlated with punishment in all three settings. This correlation is however only significant for the punisher dummy in the SPD-P. For the SPD-P, both the Positive Affect Schedule (PAS) and the Negative Affect Schedule (NAS) are positively related with punishment free-riders. This might be an indication that different behavioral forces, one related to experiencing good feelings and one related to experiencing bad feelings, are important in this game. In sum, although results are not very robust, even these regression summaries suggest that there seems to be at least a shift from a weak negative correlation between rejecting and well-being in the Mini-UG to a weak positive correlation of punishing and well-being in the DG-P. Additionally, measures of positive and negative affect seem to play a prominent role in punishment. The aggregate analysis of Table A.3a-A.3c is not able to detect this because of aggregating the affective and the cognitive-evaluative component of SWB.

Overall, the non-aggregate analysis extends the aggregate analysis of Table A.3a-A.3c.

Virtuous behavior seems to be fairly robustly correlated with the SAI and SoWB. For punishment, the affective component of SWB seems to play a crucial role and we observe an interesting shift between SP- and TP-punishment. People with high positive affect (measured by the PAS) are less likely to reject an unfair offer in the ultimatum game, but

they are more likely to punish unfair behavior in the third-party punishment setting.

Table A.4a: Summary of Ordered Logit Results Coef. for the Giver Dummy in the DG

PAS 0.41 0.27 0.61 0.11 0.20 0.60 0.40 0.32 0.03 0.95 0.16 0.71

NAS 0.72 0.06 * 0.63 0.12 0.42 0.31 0.28 0.51 0.31 0.48 0.10 0.83

NHD 0.19 0.63 0.14 0.71 0.19 0.66 0.36 0.42 -0.01 0.98 0.21 0.66

MID 0.05 0.90 0.14 0.71 0.02 0.96 0.18 0.67 -0.09 0.83 0.17 0.70

Life satisfaction

SWL 0.02 0.95 0.10 0.81 0.13 0.74 0.25 0.55 -0.02 0.97 0.11 0.80

PWB

PWBI 0.33 0.38 0.23 0.55 0.20 0.62 0.01 0.97 -0.28 0.52 -0.64 0.16

SAI 0.84 0.02 ** 0.80 0.03 ** 0.87 0.03 ** 0.88 0.03 ** 0.83 0.06 * 0.81 0.06 *

SoWB 0.76 0.05 ** 0.70 0.08 * 0.73 0.08 * 0.73 0.09 * 0.63 0.14 0.52 0.24

Notes: */** indicate significane at 10/5% level; All includes all observations, Conf6 includes only those subjects who indicated a confidence level of at least 6; Specifications 3) and 4) include additional demographic controls (age, age squared, sex, health dummy, student’s subject dummy, partnership dummy, religion dummy, ethnicity); Specificantions 5) and 6) additionally include the Big Five Inventory as a control.

Table A.4b: Summary of Ordered Logit Results Coef. for the Trust Dummy in the SPD

Dep. var. Trustor Dummy

PAS 0.53 0.24 0.88 0.08 0.31 0.51 0.65 0.20 0.11 0.83 0.39 0.46

NAS -0.43 0.18 -0.25 0.58 -0.66 0.15 -0.27 0.57 -1.00 0.04 ** -0.79 0.13

NHD 0.12 0.69 0.29 0.50 0.08 0.86 0.22 0.64 0.03 0.95 0.02 0.97

MID 0.35 0.16 0.29 0.50 0.26 0.54 0.22 0.62 0.43 0.35 0.37 0.45

Life satisfaction

SWL -0.22 0.92 -0.17 0.70 -0.27 0.55 -0.32 0.50 -1.02 0.04 ** -0.96 0.07 *

PWB

PWBI 0.72 0.21 1.08 0.03 ** 0.68 0.14 1.04 0.04 ** 0.32 0.53 0.71 0.18

SAI 0.57 0.12 0.72 0.10 0.48 0.26 0.63 0.17 0.48 0.30 0.79 0.11

SoWB 1.04 0.00 ** 1.34 0.00 ** 1.05 0.02 ** 1.34 0.01 ** 1.17 0.02 ** 1.33 0.01 **

Notes: */** indicate significane at 10/5% level; All includes all observations, Conf6 includes only those subjects who indicated a confidence level of at least 6; Specifications 3) and 4) include additional demographic controls (age, age squared, sex, health dummy, student’s subject dummy, partnership dummy, religion dummy, ethnicity); Specificantions 5) and 6) additionally include the Big Five Inventory as a control.

Table A.4c: Summary of Ordered Logit Results Coef. for the Coop. Dummy in the SPD

Dep. var. Cooperator Dummy

NAS -0.12 0.76 -0.16 0.71 -0.11 0.81 -0.20 0.66 -0.44 0.35 -0.71 0.16

Highest/ lowest

NHD 0.59 0.19 0.45 0.33 0.64 0.18 0.64 0.21 0.53 0.29 0.48 0.37

MID 1.21 0.01 ** 1.08 0.02 ** 1.32 0.01 ** 1.28 0.01 ** 1.48 0.00 ** 1.56 0.00 **

Life satisfaction

SWL -0.06 0.89 -0.01 0.98 -0.08 0.87 0.07 0.89 -0.19 0.70 -0.16 0.77

PWB

PWBI 0.26 0.54 0.30 0.49 0.21 0.64 0.22 0.63 -0.18 0.70 -0.45 0.37

SAI 1.12 0.01 ** 1.15 0.01 ** 1.08 0.02 ** 1.21 0.01 ** 1.16 0.02 ** 1.37 0.01 **

SoWB 1.61 0.00 ** 1.50 0.00 ** 1.53 0.00 ** 1.43 0.00 ** 1.67 0.00 ** 1.32 0.02 **

Notes: */** indicate significane at 10/5% level; All includes all observations, Conf6 includes only those subjects who indicated a confidence level of at least 6; Specifications 3) and 4) include additional demographic controls (age, age squared, sex, health dummy, student’s subject dummy, partnership dummy, religion dummy, ethnicity); Specificantions 5) and 6) additionally include the Big Five Inventory as a control.

A.2. ADDITIONAL ANALYSIS 135 Table A.4d: Summary of Ordered Logit Results coef. for the Rejection Dummy in the Mini-UG

NAS 0.30 0.39 0.36 0.33 0.29 0.44 0.33 0.40 0.10 0.80 0.14 0.74

Highest/ lowest

SWL -0.28 0.44 -0.41 0.29 -0.33 0.38 -0.45 0.26 -0.17 0.65 -0.21 0.60

PWB

PWBI -0.47 0.19 -0.54 0.16 -0.51 0.16 -0.60 0.13 -0.24 0.52 -0.13 0.75

SAI 0.06 0.86 -0.16 0.66 0.00 0.99 -0.20 0.60 0.14 0.70 -0.04 0.92

SoWB -0.19 0.59 0.03 0.94 -0.21 0.57 0.06 0.88 -0.08 0.84 0.25 0.52

Notes: */** indicate significane at 10/5% level; All includes all observations, Conf6 includes only those subjects who indicated a confidence level of at least 6; Specifications 3) and 4) include additional demographic controls (age, age squared, sex, health dummy, student’s subject dummy, partnership dummy, religion dummy, ethnicity); Specificantions 5) and 6) additionally include the Big Five Inventory as a control.

Table A.4e: Summary of Ordered Logit Results coef. for the Punishment Dummy in the SPD-P

SWL 0.11 0.76 0.22 0.55 0.07 0.85 0.23 0.55 -0.23 0.55 -0.16 0.69

PWB

PWBI -0.05 0.89 0.06 0.86 -0.18 0.63 -0.04 0.91 -0.51 0.19 -0.79 0.06 *

SAI 0.40 0.27 0.34 0.36 0.39 0.30 0.32 0.41 0.39 0.31 0.27 0.50

SoWB 0.41 0.26 0.31 0.41 0.29 0.44 0.18 0.65 0.11 0.78 -0.23 0.59

Notes: */** indicate significane at 10/5% level; All includes all observations, Conf6 includes only those subjects who indicated a confidence level of at least 6; Specifications 3) and 4) include additional demographic controls (age, age squared, sex, health dummy, student’s subject dummy, partnership dummy, religion dummy, ethnicity); Specificantions 5) and 6) additionally include the Big Five Inventory as a control.

Table A.4f: Summary of Ordered Logit Results Coef. for the Punishment Dummy in the DG-P

NAS 0.28 0.49 0.41 0.32 0.54 0.20 0.62 0.15 0.49 0.25 0.71 0.11

Highest/ lowest

SWL 0.32 0.45 0.47 0.28 0.21 0.63 0.35 0.45 -0.08 0.85 0.02 0.97

PWB

PWBI 0.44 0.30 0.61 0.16 0.24 0.58 0.37 0.41 -0.06 0.90 -0.20 0.67

SAI 1.10 0.01 ** 1.18 0.01 ** 1.07 0.01 ** 1.15 0.01 ** 1.09 0.01 ** 1.18 0.01 **

SoWB 0.55 0.20 0.72 0.10 0.44 0.31 0.61 0.18 0.35 0.45 0.32 0.50

Notes: */** indicate significane at 10/5% level; All includes all observations, Conf6 includes only those subjects who indicated a confidence level of at least 6; Specifications 3) and 4) include additional demographic controls (age, age squared, sex, health dummy, student’s subject dummy, partnership dummy, religion dummy, ethnicity); Specificantions 5) and 6) additionally include the Big Five Inventory as a control.

Outline

Related documents