Background

In a 3 (extent to which seller is informed: high vs. low vs. control) cell design, participants read about one of three taboo transactions (cancerous cell-phone tower; storing hazardous chemicals; testing beauty products for side effects) and were asked to indicate to what extent the seller benefited from each of them and to what extent the buyer benefited from each of them. They also indicated who they believe had more power in the transaction.

Attention check

This time, we also added an attention check: What are the roles of Person A and Person B in the transaction that took place?
The correct answer is: Person A paid money and Person B received money

## `summarise()` has grouped output by 'cond'. You can override using the
## `.groups` argument.
cond pass_check N
low 0 1
low 1 65
control 0 4
control 1 63
high 0 2
high 1 65

Alright, 7 people didn’t pass the attention check, and it looks like it’s more or less even across conditions. That leaves us with 193 participants.

Demographics

Race

race N Perc
asian 20 10.36
black 16 8.29
hispanic 6 3.11
multiracial 8 4.15
white 139 72.02
NA 4 2.07

Gender

gender N Perc
man 98 50.78
woman 91 47.15
NA 4 2.07

Age

age_mean age_sd
42.34715 11.94557

Education

edu N Perc
GED 42 21.76
2yearColl 25 12.95
4yearColl 89 46.11
MA 27 13.99
PHD 7 3.63
NA 3 1.55

Income

Analysis

Condition -> Benefit

DV operationalized as mean benefit score

Descriptives

party cond benefit_M benefit_SD
buyer low 1.85 1.31
buyer control 1.43 1.29
buyer high 1.45 1.29
seller low -1.34 1.55
seller control -0.06 1.77
seller high -0.52 1.72

Two-way Repeated Measures ANOVA

Effect DFn DFd F p p<.05 ges
cond 2 197 3.138 0.046000
0.010
party 1 197 151.462 0.000000
0.340
cond:party 2 197 8.176 0.000388
0.053

Bonferroni-corrected post-hoc comparisons: Condition

party Effect DFn DFd F p p<.05 ges p.adj
buyer cond 2 197 2.42 0.092000 0.024 0.184000
seller cond 2 197 9.07 0.000171
0.084 0.000342

Bonferroni-corrected post-hoc comparisons: Party

cond Effect DFn DFd F p p<.05 ges p.adj
low party 1 65 106.090 0.0e+00
0.546 0.00e+00
control party 1 66 25.387 3.9e-06
0.179 1.16e-05
high party 1 66 35.299 1.0e-07
0.279 4.00e-07

No real difference between control and high. If anything, control is higher, which we wouldn’t really expect. Low is way down.

Categorical DV

Let’s look at the share of people who believed that: (1) both buyer and seller benefited; (2) buyer benefited and seller was harmed; (3) seller benefited and buyer was harmed; and (4) both buyer and seller were harmed. I’ll categorize 0 as benefit. So really, 1 means unharmed and 0 means harmed.

cond both benefit both harmed buyer benefit seller harmed buyer harmed seller benefit
low 14 3 45 3
control 37 1 24 1
high 26 1 35 3

Hmm, quite a lot of win-win in the control condition. And in the high condition, much more win-lose. Looks like high explicitly mentioning that sellers are informed is doing something, but not quite lending them more benefit. It looks like there might be an even greater penalty to their “dignity” if they know the risks but decide to do it anyway. That’s very interesting.

Chi-Square

chi^2(6) = 19.47, p = .003

Chi-Square post-hoc test: condition (bonferroni-corrected)

Dimension Value low control high
both benefit Residuals -3.71138981948556 3.71961361552863 0.0209501
both benefit p values 0.0025* 0.0024* 1.0000000
both harmed Residuals 1.26178912834871 -0.610847228129145 -0.6557329
both harmed p values 1 1 1.0000000
buyer benefit seller harmed Residuals 3.04744623482343 -3.06355907535857 -0.0079154
buyer benefit seller harmed p values 0.0277* 0.0262* 1.0000000
buyer harmed seller benefit Residuals 0.523399356559488 -1.05507393053534 0.5233994
buyer harmed seller benefit p values 1 1 1.0000000

Condition -> Power

We asked participants: In your opinion, what is the balance of power in this transaction between Person A (the buyer) and Person B (the seller)? (-3 = Person A has much more power to 3 = Person B has much more power)

Importantly, this appeared right after the benefit question.

cond power_M power_SD
low -1.42 1.48
control -0.43 1.60
high -0.80 1.39


One-way ANOVA:


diff lwr upr p adj
control-low 0.987 0.363 1.610 0.001
high-low 0.615 -0.003 1.234 0.052
high-control -0.371 -0.995 0.252 0.339


Lower information means lower power. High information and control are pretty similar.

Mediation: cond -> power -> seller benefit

0 = control; 1 = low

a = -0.99 (p = 0); b = 0.16 (p = 0.088); direct = -1.27 (p = 0); indirect = -1.11 (p = 0).

Supplementary analysis

Let’s break these down by transaction (stats aren’t really necessary here because we’d be underpowered anyway)

Cancerous cell-phone tower

Condition -> Benefit

party cond benefit_M benefit_SD
buyer low 2.37 1.07
buyer control 1.57 1.27
buyer high 2.04 1.07
seller low -2.47 0.90
seller control -0.78 1.88
seller high -1.52 1.41

Condition -> Power balance

cond balance_M balance_SD
low -1.42 1.39
control -0.22 1.91
high -0.48 1.50

Storing hazardous chemicals

Condition -> Benefit

party cond benefit_M benefit_SD
buyer low 1.75 1.42
buyer control 1.50 1.47
buyer high 1.26 1.48
seller low -1.54 1.59
seller control 0.23 1.74
seller high 0.16 1.98

Condition -> Power balance

cond balance_M balance_SD
low -1.04 1.68
control -0.18 1.33
high -0.47 1.17

Testing beauty products for dangerous side effects

Condition -> Benefit

party cond benefit_M benefit_SD
buyer low 1.50 1.30
buyer control 1.17 1.10
buyer high 1.00 1.13
seller low -0.14 1.08
seller control 0.50 1.38
seller high -0.09 1.35

Condition -> Power balance

cond balance_M balance_SD
low -1.82 1.26
control -1.00 1.41
high -1.39 1.31