Attention and Allocations

To examine Hypothesis 3, the relationship between dictators’ attention to merit and luck information and allocation decisions is analyzed. Specifically, it is tested whether an increase in time spent looking at merit and luck information relatively to outcome information leads to higher shares of the surplus given to the receiver, in the case of ‘advantaged’, or lower shares of the surplus given to the receiver, in the case of ‘disadvantaged’, therefore reducing the self-serving bias. It is relevant to mention that the “other” participant, the receiver, always corresponds to an ‘advantaged’ subject for ‘disadvantaged’ subjects, and vice-versa.

Figure 7 helps to visually observe the effect of looking at merit and luck information as opposed to outcome information on the percentage of the surplus allocated to the other participant.

Figure 7: Attention to merit and luck information versus outcome information and the percentage of the surplus given to the receiver, by dictator status.

Attention O-M (s): Difference in dwell time between outcome (O) and merit (M) information, in seconds. Attention O-L (s): Difference in dwell time between outcome (O) and luck (L) information,

in seconds. % given to the other: Share of the surplus given by the dictator to the receiver.

Looking relatively more at outcome information may bias dictators’ allocations towards

‘advantaged’ keeping more and ‘disadvantaged’ keeping less. When the difference in dwell times between outcome and merit information decreases, the relative time spent looking at outcome information decreases as well. The relationships in figure 7 seem to corroborate that hypothesis, as ‘advantaged’ dictators looked relatively more at merit information, their allocations to their ‘disadvantaged’ receivers increased. Inversely, as ‘disadvantaged’ dictators look relatively more at merit information, their allocations to their ‘advantaged’ receivers decreased. Likewise, when the difference in dwell times between outcome and luck information decreases, the relative time spent looking at outcome information decreases. Once more, the relationship highlighted in figure 7 shows that, as ‘advantaged’ dictators looked

relatively more at luck information, their allocations to their ‘disadvantaged’ receivers increased. Moreover, as ‘disadvantaged’ dictators looked relatively more at luck information, their allocations to their ‘advantaged’ receivers decreased.

To better quantify the relationship between a dictator's relative attention to merit and luck information and division decisions, Tables 5 and 6 show the results of linear regressions with standard errors clustered at the individual level and controlling for subject characteristics.

In the models in Tables 5 and 6, the dependent variable is the percentage of the surplus given to the receiver, so that the influence of attention on division decisions is examined.

Table 5: The Effect of Attention to Merit versus Outcome Information on the Share Given to Other.

All models are linear regressions, with the standard errors clustered at the individual level in parentheses and controlling for individual characteristics. List of controls: age (5 categories), gender

(3 categories), nationality (12 categories), education (5 categories), income (6 categories), political leaning (8 categories), and total dwell time of individuals in Columns (3), (4), (5), and (6).

Attention (O-M): difference in dwell time between outcome information (O) and merit information (M), in seconds. Adv: advantaged. Attention All: total dwell time of individuals.

The results in Column (1) of Table 5 show that, aggregated over all treatments, increasing the difference in dwell time between outcome and merit information by one second positively but not significantly affects the share given to the receiver by 0.58%. Column (2) of the same table clarifies that results by adding an interaction term between the difference in dwell times and advantage. It shows that when ‘advantaged’ dictators look at outcome

information relative to merit information for a second more, their allocations to the receiver decrease by 4.72% of the surplus, a result that is highly significant (p < 0.001). It is also worth mentioning that they give a lot less to the receiver than do ‘disadvantaged’ dictators (17.69%

of the surplus), a result that is also highly significant (p < 0.001). Thus, relatively increasing time spent looking at merit information does not eliminate the difference between ‘advantaged’

and ‘disadvantaged’ subjects, even if it does reduce it. In Column (3), the total dwell time of subjects is controlled for, and the results are similar and remain highly significant. In this model, this regressor is certainly not endogenous and it likely does not bias the estimated effects. Additionally, ‘disadvantaged’ dictators give 2.63% more of the total surplus to their

‘advantaged’ receiver if their time looking at outcome information relative to merit information increases by one second, which perhaps suggests that ‘disadvantaged’ subjects also contribute to the self-serving bias by giving more to the ‘advantaged’, even if that means keeping less to themselves. Columns (4), (5), and (6) present the same model as in (3), with regards to multipliers 3, 5, and 7, respectively. They show that being ‘advantaged’ and increasing the difference in dwell time between outcome and merit information by one second leads to a reduction in the share of the surplus given to the receiver by 3.58% in multiplier 3, 5.17% in multiplier 5, and 6.49% in multiplier 7. The results relative to each multiplier are also highly statistically significant (p < 0.001).

Inversely, the impact of raising the difference in dwell time between outcome and merit information by a second for ‘disadvantaged’ dictators significantly increases the share allocated to the receiver by 2.63%, considering all data (p < 0.001). Column (4), (5), and (6) demonstrate that that effect remains significant and increases throughout multipliers (p < 0.01 in multiplier 3, p < 0.001 in multipliers 5 and 7).

Correspondingly, Table 6 displays the results for the relationship between attention to luck information versus outcome information and allocations to the receiver.

Table 6: The Effect of Attention to Luck versus Outcome Information on the Share Given to Other.

All models are linear regressions, with the standard errors clustered at the individual level in parentheses and controlling for individual characteristics. List of controls: age (5 categories), gender

(3 categories), nationality (12 categories), education (5 categories), income (6 categories), political leaning (8 categories), and total dwell time of individuals in Columns (3), (4), (5) and (6).

Attention (O-L): difference in dwell time between outcome information (O) and luck information (L), in seconds. Adv: advantaged. Attention All: total dwell time of individuals.

Column (1) of Table 6 shows that, aggregated over all treatments, raising the difference in dwell time between outcome and luck information by one second results in giving 0.22% of the surplus to the receiver, although it is not significant. When controlling for the total dwell time of subjects in Column (3), the interaction term between the difference in dwell times and advantage is highly significant and reflects a discount in the share given to the other by 3.88%

(p < 0.001). Columns (4), (5), and (6) show the analogue results for multiplier 3, 5, and 7, respectively. It is observed that, for ‘advantaged’, the influence of looking one second more at outcome relative to luck information is a discount by 3.30% in multiplier 3 (p < 0.01), 4.03%

in multiplier 5 (p < 0.001), and 4.49% in multiplier 7 (p < 0.01) in the share given to the other.

Thus, looking relatively less at luck information leads to an increase in the share given to their

‘disadvantaged’ receivers, as inequality in payment rates increases. Finally, it is interesting to observe that, as inequality increases and the relative time spent looking at luck information

increases, ‘disadvantaged’ dictators give less to their ‘advantaged’ receivers, a result that is significant for multipliers 5 (p < 0.01) and 7 (p < 0.001). They give more to the other when looking at luck in multiplier 3 perhaps suggesting that, because it is the lowest level of inequality, it is more legitimate for them to give more to ‘advantaged’ and, somehow, contribute to the self-serving bias themselves. That bias is mitigated as inequality rises and

‘disadvantaged’ look relatively more at luck information, showing the possible importance of acknowledging the role that luck plays in success to mitigate the bias, even for those who are randomly assigned a ‘disadvantaged’ position to start with.

When analyzing the effect of absolute dwell time to merit information, luck information, and their sum on the share given to the receiver, it seems at first glance that attention to merit information might have a much greater impact on reducing the bias than looking at luck information (Appendix A.1.). Nevertheless, it is more relevant to study their dwell times relatively to dwell time on outcome information, as a means of reducing the bias driven by individual differences in overall dwell time on the studied effects.

Result 3.1. As ‘advantaged’ subjects spend relatively more time looking at merit information than at outcome information, allocations to ‘disadvantaged’ subjects increase, a result that is highly significant overall and across multipliers (p < 0.001). As ‘advantaged’ subjects spend relatively more time looking at luck information than at outcome information, allocations to

‘disadvantaged’ subjects increase, a result that is highly significant overall (p < 0.001) and holds across multipliers (p < 0.01 for multipliers 3 and 7, p < 0.001 for multiplier 5).

Result 3.2. As ‘disadvantaged’ subjects spend relatively more time looking at merit information than at outcome information, allocations to ‘advantaged’ subjects decrease, a result that is highly significant overall (p < 0.001) and holds across multipliers (p < 0.01 for multiplier 3, p < 0.001 for multipliers 5 and 7). As ‘disadvantaged’ subjects spend relatively more time looking at luck information than at outcome information, allocations to ‘advantaged’ subjects decrease, a result that is highly significant overall (p < 0.001) and remains significant for multipliers 5 (p < 0.01) and 7 (p < 0.001).