Background

In a 2 (taboo vs. standard) cell design, participants read about one of five taboo transactions, or one of five standard economic transactions, depending on condition. They were then asked, in a random order, about the actors’ benefit from the transaction and the power balance in the transaction.

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

cond failcheck passcheck
nontaboo 16 184
taboo 17 185

Alright, that leaves us with 369, pretty evenly distributed between conditions.

Demographics

Race

race N Perc
asian 21 5.69
black 39 10.57
hispanic 20 5.42
multiracial 20 5.42
white 260 70.46
NA 9 2.44

Gender

gender N Perc
man 173 46.88
woman 193 52.30
NA 3 0.81

Age

age_mean age_sd
40.71003 12.51781

Education

edu N Perc
noHS 3 0.81
GED 101 27.37
2yearColl 46 12.47
4yearColl 152 41.19
MA 46 12.47
PHD 17 4.61
NA 4 1.08

Income

Analysis

Condition -> Benefit

Descriptives

cond benefit_A_m benefit_A_sd benefit_B_m benefit_B_sd
nontaboo 1.554348 1.333586 1.1413043 1.575645
taboo 1.924324 1.172466 -0.5243243 1.681324

Two-way Repeated Measures ANOVA

Effect DFn DFd F p p<.05 ges
cond 1 367 43.606 0
0.047
person 1 367 153.826 0
0.196
cond:person 1 367 77.835 0
0.110

Bonferroni-corrected post-hoc comparisons: Party

person Effect DFn DFd F p p<.05 ges p.adj
benefit_A cond 1 367 8.012 0.005
0.021 0.01
benefit_B cond 1 367 96.387 0.000
0.208 0.00

Bonferroni-corrected post-hoc comparisons: Condition

cond Effect DFn DFd F p p<.05 ges p.adj
nontaboo person 1 183 6.883 0.009
0.020 0.018
taboo person 1 184 210.899 0.000
0.418 0.000

One-sample t-tests

Buyers in Taboo Condition: t(184) = 22.32, p < .001, d = 1.64

Buyers in Standard Condition: t(183) = 9.83, p < .001, d = 0.72

Sellers in Taboo Condition: t(184) = -4.24, p < .001, d = -0.31

Sellers in Standard Condition: t(183) = 9.83, p < .001, d = 0.72

Condition -> Power

Let’s take a look at the effect on power. Power was rated from -3 (Buyer has much more power) to 3 (Seller has much more power).

Descriptives

cond power_M power_SD
nontaboo 0.54 1.56
taboo -0.04 2.17

Power is pretty much balanced in the taboo condition. In the non-taboo condition, the seller has more power.

One-sample t-tests

Taboo Condition:

t(184) = -0.24, p = .813, d = -0.02
Non-Taboo Condition:

t(183) = 4.73, p < .001, d = 0.35

Plot: Condition -> Power


Mediation model: condition -> power -> seller benefit

0 = Standard; 1 = Taboo

Call: psych::mediate(y = benefit_B ~ condition + (power), data = formed)

Direct effect estimates (traditional regression) (c’) X + M on Y benefit_B se t df Prob Intercept 1.00 0.12 8.55 366 3.39e-16 condition -1.52 0.16 -9.24 366 2.05e-18 power 0.26 0.04 5.92 366 7.29e-09

R = 0.53 R2 = 0.28 F = 70.21 on 2 and 366 DF p-value: 1.55e-26

Total effect estimates (c) (X on Y) benefit_B se t df Prob Intercept 1.14 0.12 9.50 367 2.77e-19 condition -1.67 0.17 -9.82 367 2.35e-20

‘a’ effect estimates (X on M) power se t df Prob Intercept 0.54 0.14 3.90 367 0.000114 condition -0.58 0.20 -2.95 367 0.003330

‘b’ effect estimates (M on Y controlling for X) benefit_B se t df Prob power 0.26 0.04 5.92 366 7.29e-09

‘ab’ effect estimates (through all mediators) benefit_B boot sd lower upper condition -0.15 -0.15 0.06 -0.28 -0.05

a = -0.58 (p = 0.003); b = 0.26 (p = 0); direct = -1.67 (p = 0); indirect = -1.52 (p = 0).

Ok.. a partial mediation. Not bad.

Order effects: Benefit first

Condition -> Benefit

Descriptives

cond benefit_A_m benefit_A_sd benefit_B_m benefit_B_sd
nontaboo 1.588785 1.386956 1.1962617 1.645114
taboo 1.961039 1.105507 -0.4415584 1.568552

Two-way Repeated Measures ANOVA

Effect DFn DFd F p p<.05 ges
cond 1 182 18.773 2.43e-05
0.044
person 1 182 75.256 0.00e+00
0.185
cond:person 1 182 38.919 0.00e+00
0.105

Bonferroni-corrected post-hoc comparisons: Party

person Effect DFn DFd F p p<.05 ges p.adj
benefit_A cond 1 182 3.805 0.053 0.020 0.106
benefit_B cond 1 182 46.132 0.000
0.202 0.000

Bonferroni-corrected post-hoc comparisons: Condition

cond Effect DFn DFd F p p<.05 ges p.adj
nontaboo person 1 106 3.227 0.075 0.017 0.15
taboo person 1 76 110.931 0.000
0.443 0.00

One-sample t-tests

Buyers in Taboo Condition: t(76) = 15.57, p < .001, d = 1.77

Buyers in Standard Condition: t(106) = 7.52, p < .001, d = 0.73

Sellers in Taboo Condition: t(76) = -2.47, p = .008, d = -0.28

Sellers in Standard Condition: t(106) = 7.52, p < .001, d = 0.73

Condition -> Power

Let’s take a look at the effect on power. Power was rated from -3 (Buyer has much more power) to 3 (Seller has much more power).

Descriptives

cond power_M power_SD
nontaboo 0.64 1.43
taboo -0.08 2.11

One-sample t-tests

Taboo Condition:

t(76) = -0.32, p = .747, d = -0.04
Non-Taboo Condition:

t(106) = 4.67, p < .001, d = 0.45

Plot: Condition -> Power


Ok, same patterns for “benefit first” participants.

Order effects: Power first

Condition -> Benefit

Descriptives

cond benefit_A_m benefit_A_sd benefit_B_m benefit_B_sd
nontaboo 1.506493 1.263073 1.0649351 1.480911
taboo 1.898148 1.222376 -0.5833333 1.762141

Two-way Repeated Measures ANOVA

Effect DFn DFd F p p<.05 ges
cond 1 183 21.501 6.7e-06
0.044
person 1 183 73.524 0.0e+00
0.198
cond:person 1 183 35.809 0.0e+00
0.107

Bonferroni-corrected post-hoc comparisons: Party

person Effect DFn DFd F p p<.05 ges p.adj
benefit_A cond 1 183 4.488 0.035
0.024 0.07
benefit_B cond 1 183 44.793 0.000
0.197 0.00

Bonferroni-corrected post-hoc comparisons: Condition

cond Effect DFn DFd F p p<.05 ges p.adj
nontaboo person 1 76 3.894 0.052 0.025 0.104
taboo person 1 107 107.335 0.000
0.403 0.000

One-sample t-tests

Buyers in Taboo Condition: t(107) = 16.14, p < .001, d = 1.55

Buyers in Standard Condition: t(76) = 6.31, p < .001, d = 0.72

Sellers in Taboo Condition: t(107) = -3.44, p < .001, d = -0.33

Sellers in Standard Condition: t(76) = 6.31, p < .001, d = 0.72

Condition -> Power

Let’s take a look at the effect on power. Power was rated from -3 (Buyer has much more power) to 3 (Seller has much more power).

Descriptives

cond power_M power_SD
nontaboo 0.40 1.72
taboo -0.01 2.22

One-sample t-tests

Taboo Condition:

t(107) = -0.04, p = .965, d < 0.01
Non-Taboo Condition:

t(76) = 2.06, p = .043, d = 0.23

Plot: Condition -> Power




Yeah, there’s basically no order effects.

By Scenario

Power by scenario

Scenario 1: Cancerous cell-phone tower vs. noisy street
Scenario 2: Kidney vs. car
Scenario 3: Hazardous chemicals vs. furniture
Scenario 4: Beauty product side effects vs. bugs and glitches
Scenario 5: Doctor’s appointment vs. concert tickets

transaction cond n power_M power_SD
1 nontaboo 38 0.18 1.71
1 taboo 32 0.06 2.26
2 nontaboo 41 1.00 1.24
2 taboo 36 1.94 1.53
3 nontaboo 38 0.89 1.01
3 taboo 41 -0.32 1.81
4 nontaboo 37 -0.73 1.48
4 taboo 36 -1.83 1.48
5 nontaboo 30 1.50 1.33
5 taboo 40 0.00 2.01

hmm, yeah, power is still a bit all over the place. Every participant only did one of these scenarios so we can’t do a fixed effects model to account for that variance, but we might be fine by just saying that this is what we preregistered, leave it at that, and strengthen it by taking only scenario 3 (where we see a clear difference in power) for the informed/rational interaction.

Seller benefit by scenario

Scenario 1: Cancerous cell-phone tower vs. noisy street
Scenario 2: Kidney vs. car
Scenario 3: Hazardous chemicals vs. furniture
Scenario 4: Beauty product side effects vs. bugs and glitches
Scenario 5: Doctor’s appointment vs. concert tickets

transaction cond n benefit_B_M benefit_B_SD
1 nontaboo 38 -0.92 1.53
1 taboo 32 -1.34 1.96
2 nontaboo 41 1.95 1.09
2 taboo 36 0.03 1.63
3 nontaboo 38 1.68 0.84
3 taboo 41 -0.66 1.80
4 nontaboo 37 1.46 0.93
4 taboo 36 -0.61 1.36
5 nontaboo 30 1.57 1.36
5 taboo 40 -0.15 1.39

Pretty remarkable that people don’t think kidney sellers are harmed. That’s the one scenario where we don’t replicate the effect.