Supplementary Materials/Online Appendices

Supplementary Materials/Online Appendices

Here we provide further details concerning our study, supplementary analyses, and robustness checks. This document is organized as follows, with tables and graphs listed by section.

Appendix OA: Sample Characteristics, Randomization Checks, Measures, & Pre-Test Results
Tables OA1-OA4: Demographics and Political Characteristics of Samples
Tables OA5-6: Randomization Tests
Tables OA7-OA9: Manipulation Checks
Table OA10-OA11: Descriptive Statistics for DVs by Condition
Table OA12: Pre-Test Results for Study 1
Appendix OB: Analyses For In-Text Figures and Supplementary Analyses
Tables OB1-OB3: Analyses in relation to Figure 1
Tables OB4-OB5 & Figure OB1: Issue Agreement, Co-Partisanship, and Reactions to the Treatments
Table OB6: Analyses in relation to Figure 2
Tables OB7-8 Analyses in Relation to Figure 3
Tables OB9-10: Motive Attributions and Overall Evaluations
Tables OB11-12: Replication of Figure 1 Analyses with Individual Evaluative Items
Table OB13: Replication of Study 1 Results with Consistent + Justification Conditions
Appendix OC: Matching and Account Satisfaction
Table OC1: Predicting Account Satisfaction
Table OC2 & Figure OC1: Account Satisfaction and Persuasion
Table OC3-OC4: Account Satisfaction and Motive Attributions
Table OC5: CEM Matching Results
Tables OC6-OC7, Figure OC2: Matched Account Satisfaction and Persuasion
Tables OC8-OC10, Figure OC3: Matched Account Satisfaction and Motive Attributions
Appendix OD: Robustness Checks for Sample Characteristics
Tables OD1-OD2: Treatment Differences are Robust to the Inclusion of Demographic Controls
Tables OD3-OD4 Treatment Differences Do Not Significantly Differ By Education (Study 1, Study2), Age (Study 2), PID (Study 1, Study 2), or Ideology (Study 2)

Online Appendix A

Sample Demographics

Table OA1: Demographic Characteristics of Sample for Study 1

Variable: / Sample Data: / Population Data:
Gender
% Female / 50.06% / 51.32%
Race
% White / 72.96% / 66.98%
% African American / 13.49% / 11.65%
% Asian American / 3.24% / 4.81%
% Hispanic / 7.52% / 14.22%
% Other / 2.79% / 2.35%
Age
% 18-24 / 9.53% / 17.63%
% 25-34 / 17.77% / 16.59%
% 35-50 / 36.25% / 25.77%
% 51-64 / 27.69% / 23.75%
% 65+ / 8.75% / 16.27%
Median Value/Age / 35-50 / 37.2
Education
% < HS / 1.48% / 13.65%
% HS Graduate / 19.37% / 28.02%
% Some College or Associates Degree / 39.32% / 29.23%
% Bachelor's + / 39.83% / 29.10%
Income
Median Value/Income / $40,000-49,999 / $51,371

Note: Population data for gender, race, and age is derived from 2010 US Census. Race and gender percentages are for the 18+ population. For age, the 18-24 population data cell actually captures the 15-24 population due to the way the census breaks down age categories. Age and Gender data are taken from Census table QT-P1 while Race data comes from table P11. Education and income population data comes from the 2012 American Community Survey 1-year estimate. Education data covers the over 25 population and comes from Table DP02. Income data uses 2012 inflation adjusted dollars and is found in Table S1901.

Table OA2: Political Characteristics of Sample in Study 1

Variable / Sample Statistic
Ideology:
Mean (Range: 1-7; Higher = More Conservative) / 4.12 (1.55)
% Ext. Liberal / 5.37%
% Liberal / 12.80%
% Slightly Liberal / 9.57%
% Moderate/Middle of the Road / 36.20%
% Slightly Conservative / 14.93%
% Conservative / 14.80%
% Extremely Conservative / 6.33%
Partisanship
Mean (1-7; Higher = More Republican) / 3.56 (2.14)
% Str. Dem / 25.22%
% Not Very Strong Dem / 15.78%
% Lean Dem / 9.07%
% Independent / 17.33%
% Lean Rep / 7.60%
% Not Very Strong Rep / 10.55%
% Strong Rep / 14.55%
Dream Act: Issue Importance (Range: 1-5; Higher = More Importance) / 3.03 (1.19)
% Not at All Important / 12.13%
% Slightly Important / 19.39%
% Moderately Important / 35.47%
%Very Important / 19.46%
% Extremely Important / 13.55%
Political Interest
Mean (1-4; Higher = More Importance) / 2.74 (0.94)
% Not At All Interested / 9.66%
% Somewhat Interested / 30.94%
% Interested / 34.73%
% Very Interested / 24.67%
Political Cynicism
Mean (Range: 1-5; Higher = More Cynical) / 3.49 (0.84)

Table OA3: Sample Demographics in Study 2

Variable: / Experiment 2: / Population Data:
Gender
% Female / 48.97% / 51.32%
Race
% White / 80.15% / 66.98%
% African American / 6.34% / 11.65%
% Asian American / 7.55% / 4.81%
% Hispanic / 4.10% / 14.22%
% Other / 1.86% / 2.35%
Age
% 18-24 / 16.48 / 17.63%
% 25-34 / 42.64 / 16.59%
% 35-50 / See note / 25.77%
% 51-64 / See note / 23.75%
% 65+ / 6.98 / 16.27%
Median Value/Age / 25-34 / 37.2
Education
% < HS / 0.28% / 13.65%
% HS Graduate / 8.78% / 28.02%
% Some College or Associates Degree / 35.87% / 29.23%
% Bachelor's + / 55% / 29.10%
Income
Median Value/Income / $30-39,999 / $51,371

Note: Sources for population data are the same as Table OA1. The age variable on Study 2 had categories for 35-44 (22.53%), 45-54 (9.96%) and 55-64 (6.98%). Thus, 39.47% of the sample is between 35-64, compared to 49.53% of the population.

Table OA3: Political Characteristics of Experiment 2 Sample

Variable / Experiment 2
Ideology:
Mean (Overall Scale; Range: 1-7; Higher = More Conservative) / 3.43 (1.71)
% Ext. Liberal / 12.87%
% Moderately Liberal / 25.09%
% Slightly Liberal / 15.21%
% Moderate/Middle of the Road / 18.38%
% Slightly Conservative / 13.43%
% Moderately Conservative / 11.01%
% Extremely Conservative / 4.01%
Mean Ideology: Economic Issues (See note) / 3.91 (1.80)
Mean Ideology: Social Issues / 2.95 (1.97)
Partisanship
Mean (1-7; Higher = More Republican) / 3.38 (1.74)
% Str. Dem / 16.71%
% Weak Dem / 19.05%
% Lean Dem / 18.58%
% Independent / 20.35%
% Lean Rep / 11.20%
% Weak Rep / 8.59%
% Strong Rep / 5.51%
Mean Identity Importance (Range: 1-4; Higher = More Important) / 2.26 (0.95)
Political Interest
Mean (Range: 1-7; Higher = More Interest) / 4.24 (1.57)
Political Cynicism
Mean (Range: 1-5; Higher = More Cynical) / 3.85 (0.61)

Note: Respondents were asked to place themselves on three ideology scales: an overall scale, one for economic issues, and one for social issues (see Klar 2014, “A Multidimensional Study of Ideological Preferences and Priorities among the American Public.” Public Opinion Quarterly 78: 344–59). All three scales range from 1-7 with higher scores indicating greater conservatism. The two sub-domain scales show very high correlations with general ideology (r = 0.80 for social ideology and 0.80 for economic), but a weaker, albeit still strong, relationship with each other (r = 0.53).

Randomization Checks

We follow Levendusky and Horowitz (2012) and perform an ordered logistic regression on experimental condition assignment for both studies. The predictors for the Study 1 model are: gender, race, education, age, political interest, income, ideology, partisanship, issue importance, and political cynicism. The results in Table OA5 below show that these variables are jointly insignificant.

For Study 2, the set of predictors is broadly the same with two changes. First, we use a measure of party identity importance rather than issue importance. Second, we offer the results of two models. In the first we use our general measure of ideology, while in the second we instead include our separate social and economic issues ideology measure (see Table OA4 for details). The results are below in Table OA6. As above, the predictors are jointly insignificant.

Table OA5 : Randomization Check Study 1

(1)
Exp. Condition
Female / 0.0816
(0.83)
African American [Base = White] / 0.0373
(0.24)
Asian American / 0.431
(1.62)
Hispanic / 0.242
(1.27)
Other / -0.0877
(-0.28)
R's Education / -0.0986
(-1.43)
R's Age / 0.0354
(0.74)
R's Political Interest / 0.00876
(0.15)
R's Annual Income / 0.0171
(0.87)
R's PID / 0.0316
(1.11)
R's Ideology / -0.0387
(-1.03)
Issue Importance / 0.0449
(1.06)
Political Cynicism / 0.0839
(1.44)
Observations / 1317
Pseudo R2 / 0.002
chi2 / 12.12
Prob > chi2 / 0.518

Note: Results from OLOGIT model. Cut points omitted for space. t statistics in parentheses

+p < 0.10, *p < 0.05, **p < 0.01

Table OA6: Randomization Check (Study 2)

(1) / (2)
Full 10 Condition / Full 10 Condition
Female / -0.143
(0.112) / -0.140
(0.112)
Black / 0.177
(0.221) / 0.194
(0.224)
Hispanic / 0.0540
(0.262) / 0.0220
(0.266)
Asian / 0.0870
(0.212) / 0.0942
(0.211)
Other / -0.0471
(0.388) / -0.0431
(0.387)
Age / -0.00364
(0.0495) / -0.00666
(0.0501)
Education / 0.0455
(0.0471) / 0.0470
(0.0473)
Income / -0.0158
(0.0235) / -0.0199
(0.0235)
Political Interest / -0.0383
(0.0389) / -0.0385
(0.0390)
Cynicism (Mean) / -0.0755
(0.0898) / -0.0801
(0.0893)
Party ID / 0.0312
(0.0511) / 0.00631
(0.0475)
PID Imp / 0.103
(0.0628) / 0.100
(0.0629)
Ideology (General) / -0.00378
(0.0519)
Ideology (Economic Issues) / 0.0359
(0.0433)
Ideology (Social Issues) / -0.00527
(0.0369)
Observations / 1044 / 1045
Pseudo R2 / 0.002 / 0.002
ll_0 / -2403.3 / -2405.6
ll / -2399.4 / -2401.2
chi2 / 7.819 / 8.854
p / 0.8552 / 0.8403

Standard errors in parentheses

+p < 0.10, *p < 0.05, **p < 0.01

Manipulation Checks

Both studies featured manipulation checks in which the respondents were asked knowledge questions regarding the vignette elite. In Study 1, respondents read about a candidate who took two positions (which were either the same or not); respondents were asked to indicate whether the elite supported or opposed the Dream Act most recently (i.e. the second position). Table OA7 provides a descriptive breakdown of responses both overall and by condition. Overall, 77% of respondents correctly answered this question. Interestingly, correct responses were more likely to occur when the elite offered a positive position last (regardless of position change). As Table OA8 demonstrates, this effect is particularly strong among those who also said they support the Dream Act. Holding other variables in Model 3 at mean values, the probability of a correct response for someone who said they support the Dream Act and who read about a candidate who most recently took this position (regardless of consistency) is 0.93; for individuals who read about the same type of elite but who said neither or oppose have probabilities of correct response equaling 0.72 and 0.79. Ultimately, however, these differences do not substantially affect our conclusions about Study 1. For instance, including a dummy variable for correctness alongside indicators for treatment condition in models predicting our dependent variables reveals the same patterns as seen in Figure 1 in text. Moreover, interactions between treatment assignment and correctness are all statistically insignificant.

In Study 2, respondents were asked two questions: one concerning which party the Representative belonged to and one concerning which committee the Representative serves on in Congress. As Table OA9 shows, correct responses were high and varied little across conditions. As with Study1, including indicators for correctness in models predicting the DVs alongside treatment condition does not change results and interaction coefficients are null.

Table OA7: Study 1, Manipulation Check

% Correct
Total / 77.07%
No Account Received
C1: Pro, Pro / 87.93%
C2: Con, Con / 70.15%
C3: Pro, Con / 69.75%
C4: Con, Pro / 84.06%
Societal Fairness Account
C5: Pro, Pro; Soc. Fairness / 81.15%
C6: Con, Con; Soc. Fairness / 68.55%
C7: Pro, Con; Soc Fairness / 64.75%
C8: Con, Pro; Soc Fairness / 82.61%
Comparison Account
C9: Pro, Pro; Comparison / 81.82%
C10: Con, Con; Comparison / 75.94%
C11: Pro, Con; Comparison / 76.47%
C12: Con, Pro; Comparison / 81.20%
By Position Proximity
Last Position for Candidate = Respondent’s / 82.78%
Last Position for Candidate ≠ Respondent’s / 77.40%
By Last Position
Most Recent Position = Pro / 83.05%
Most Recent Position = Con / 71.09%

Table OA8: Manipulation Check (Study 1)

(1) / (2) / (3)
Correct / Correct / Correct
Repositioned [Base: Consistent] / -0.108
(0.125) / -0.0731
(0.136) / -0.0907
(0.136)
Soc. Fairness [Base: No Account] / -0.206
(0.153) / -0.350*
(0.167) / -0.341*
(0.168)
Comparison / 0.0478
(0.154) / -0.0586
(0.170) / -0.0486
(0.170)
Recent Candidate Position: Support / 0.689**
(0.127) / 0.657**
(0.138) / 1.406**
(0.276)
Respondent Position: Neither
[Base = Support] / -1.015**
(0.160) / -1.015**
(0.184) / -0.608*
(0.237)
Respondent Position: Oppose / -0.612**
(0.151) / -0.617**
(0.172) / -0.243
(0.210)
R's Ideology / 0.0209
(0.0539) / 0.0283
(0.0538)
R's PID / 0.0280
(0.0383) / 0.0278
(0.0381)
Issue Importance / -0.122*
(0.0618) / -0.119+
(0.0621)
R's Political Interest / 0.103
(0.0802) / 0.108
(0.0805)
Recent Position: Support # Neither / -1.049**
(0.375)
Recent Position: Support # Oppose / -1.041**
(0.348)
Constant / 1.512**
(0.168) / 1.507**
(0.351) / 1.197**
(0.361)
Observations / 1530 / 1320 / 1320
Pseudo R2 / 0.046 / 0.050 / 0.058
chi2 / 76.40 / 69.91 / 81.16
p / 1.98e-14 / 4.62e-11 / 2.47e-12

Standard errors in parentheses

+p < 0.10, *p < 0.05, **p < 0.01

Table OA9: Manipulation Check (Study 2)

% Correct
PID / Committee
Total / 90.25 / 87.35
Representative PID
Rep is Democrat / 90.02 / 85.64
Rep is Republican / 90.49 / 89.10
Position History (Rep = Democrat)
Baseline / 90.99 / 89.19
Consistent / 87.61 / 85.85
Repositioned (no Justification) / 91.51 / 85.05
Repositioned (New Info) / 88.68 / 85.98
Repositioned (P. Fairness) / 91.43 / 81.90
Position History (Rep = Republican)
Baseline / 90.10 / 91.18
Consistent / 91.59 / 89.81
Repositioned (no Justification) / 90.91 / 89.19
Repositioned (New Info) / 91.59 / 89.72
Repositioned (P. Fairness) / 88.12 / 85.58
Partisan Congruence
Rep is Same PID as Respondent / 91.20 / 87.19
Rep is Different PID / 91.56 / 88.97
R is Independent / 85.71 / 84.40

Measurement of Core Variables

Study 1:

1)Dependent Variables

a)Feeling Thermometer:

i)How would you rate Candidate A on the scale below, where ratings between 50 and 100 degrees mean that you feel favorable and warm toward him/her, ratings between 0 and 50 degrees mean that you don't feel favorable toward Candidate A and that you don't care too much for him/her. You would rate Candidate A at the 50 degree mark if you don't feel particularly warm or cold toward Candidate A.

b)Traits

i)Introduction:

(1)Listed here are several characteristics used to describe politicians. Please indicate the extent to which you agree or disagree that each describes Candidate A based on what you know about him/her.

ii)Individual Measures:

(1)Candidate A is…

(a)Honest, intelligent, a strong leader, open-minded, compassionate.

iii)Response options: (1) Strongly disagree, (2) disagree, (3) neither agree nor disagree, (4) agree, (5) strongly agree

iv)A factor analysis shows just one dimension with an eigenvalue over 1 (EV = 3.29, proportion explained: 0.66) with all five variables loading on this dimension; un-rotated factor loadings range from 0.79 (openness) to 0.83 (leader). The alpha coefficient for these items is 0.87.

2)Issue Attitudes

a)Dream Act Attitude:

i)How strongly do you support or oppose the Dream Act, which would provide undocumented immigrants brought to the U.S. as children the ability to gain legal resident status if they join the military or go to college?

ii)Response options: (1) strongly support, (2) moderately…, (3) moderately…, (4) neither support nor oppose, (5) somewhat oppose, (6) moderately…, (7) strongly….

b)Importance of Attitude

i)How important is this issue (the Dream Act/immigration reform) to you personally?

ii)Response options: (1) not at all important, (2) slightly important, (3) moderately important, (4) very important, (5) extremely important

3)Account Satisfaction

a)How satisfied or dissatisfied are you with the explanation offered by Candidate A for their change in position?(1) Very Dissatisfied, (2) Moderately…, (3) Somewhat…, (4) Neither dissatisfied Nor Satisfied, (5) Somewhat Satisfied…, (6) Moderately…, (7) Very Satisfied

4)Motives

a)[Intro:] The following are a list of potential reasons for why Candidate A took his/her final position. Please rate how important you think each is in explaining why Candidate A took his/her final position.

i)A desire to help all Americans

ii)A desire to help his/her constituents.

iii)To pander to voters

iv)To increase political influence

v)To win office/re-election

b)Response Options: (1) Extremely unimportant, (2) moderately unimportant, (3) somewhat unimportant, (4) neither important nor unimportant, (5) somewhat important, (6) moderately important, (7) extremely important

c)A factor analysis reveals two dimensions with eigenvalues over 1 (before rotation): Dimension 1: Office, Influence, and Pandering all load on this dimension (0.84-0.93), EV = 2.44, proportion explained = 0.49; Dimension 2: the two ‘help’ variables load on this dimension (0.86-0.87), EV = 1.51, proportion explained 0.30.

5)Political Cynicism

a)Cynicism is a scale formed by averaging two items asking respondents how much they agree/disagree that politicians are honorable people and most politicians are untrustworthy (Range = 1-5; Higher = more cynical; Mean = 3.49). The two variables are moderately correlated (r = 0.36).

Study 2

6)Dependent Variables

a)Feeling Thermometer:

i)How would you rate Representative A on the scale below, where ratings between 50 and 100 degrees mean that you feel favorable and warm toward him/her, ratings between 0 and 50 degrees mean that you don't feel favorable toward Representative A and that you don't care too much for him/her. You would rate Representative A at the 50 degree mark if you don't feel particularly warm or cold toward Representative A.

b)Traits

i)Introduction:

(1)Listed here are several characteristics used to describe politicians. Please indicate the extent to which you agree or disagree that each describes Representative A based on what you know about him/her.

ii)Individual Measures:

(1)Representative A is…

(a)Strong leader, intelligent, and open-minded

(i)Note: the order of these items was randomized

iii)Response options: (1) Strongly disagree, (2) disagree, (3) neither agree nor disagree, (4) agree, (5) strongly agree

iv)A single dimension emerges in a factor analysis (EV = 1.99, proportion explained = 0.66), with unrotated factor loadings ranging from 0.75 (open-minded) to 0.85 (intelligent). The alpha coefficient for these items is 0.73.

7)Account Satisfaction

8)Motives

a)[Intro:] The following are a list of potential reasons for why Representative A changed positions. Please rate how important you think each is in explaining why Representative A changed positions

i)Help all Americans

ii)Help constituents

iii)Make good public policy

iv)Ideological reasons

v)Own policy preferences

vi)Influence of special interests

vii)Winning re-election

viii)Pandering to voters

ix)Own personal values

(1)Note: the order of these items was randomized

c)A factor analysis reveals three dimensions with EV > 1. Dimension 1 (EV = 3.21, proportion explained = 0.36), Dimension 2 (EV = 1.89, proportion explained = 0.21), and Dimension 3 (EV = 1.13, proportion explained = 0.13). Loading on D1 were the political motives (with rotated loadings between 0.79 and 0.82); D2 were helping and make good policy motives (rotated loadings ranging from 0.78 to 0.85); D3 captures the remaining three motives (loadings from 0.74 to 0.78).

d)Alpha coefficients: political values (0.78); representation (0.81); policy (0.66).

9)Political Cynicism

a)Respondents were asked how strongly they agreed/disagreed with the statements: what politicians say depends on who is listening, politicians are willing to do whatever it takes to win, politicians often lie about their goals, and politicians generally have good intentions. The order of the four statements was randomly varied.

b)All four items load on a single dimension via a factor rotation with varimax rotation (EV = 2.25, proportion explained = 0.56). Alpha coefficient = 0.73.

Descriptive Statistics for Outcome Variables

Table OA10: Mean Ratings on DVs by Condition for Study 1 with 95% Confidence Intervals

Combined Evaluation Scale / Thermometer / Traits Index
No Justification
Support, Support / 0.20 (0.01, 0.38) / 54.62 (49.83, 59.41) / 3.35 (3.22, 3.47)
Support, Oppose / 0.15 (-0.02, 0.33) / 44.44 (40.15, 48.73) / 2.95 (2.80, 3.10)
Oppose, Support / -0.27 (-0.49, -0.05) / 50.70 (47.01, 54.39) / 3.15 (3.02, 3.29)
Oppose, Oppose / -0.01 (-0.20, 0.19) / 54.29 (50.55, 58.03) / 3.21 (3.09, 3.34)
Societal Fairness Justification
Support, Support / -0.02 (-0.19, 0.15) / 49.39 (45.22, 53.56) / 3.18 (3.06, 3.31)
Support, Oppose / -0.07 (-0.27, 0.12) / 52.67 (48.79, 56.55) / 3.17 (3.04, 3.30)
Oppose, Support / -0.02 (-0.20, 0.16) / 51.49 (47.54, 55.43) / 3.28 (3.15, 3.41)
Oppose, Oppose / 0.07 (-0.12, 0.26) / 53.93 (49.95, 57.90) / 3.11 (2.97, 3.26)
Comparison of Ends Justification
Support, Support / 0.03 (-0.15, 0.21) / 47.65 (43.24, 52.05) / 3.21 (3.08, 3.34)
Support, Oppose / 0.10 (-0.09, 0.29) / 51.23 (47.34, 55.13) / 3.09 (2.96, 3.21)
Oppose, Support / -0.05 (-0.22, 0.12) / 47.81 (43.70, 51.93) / 3.15 (3.01, 3.28)
Oppose, Oppose / -0.13 (-0.32, 0.07) / 55.17 (50.64, 59.70) / 3.26 (3.12, 3.40)

Note: 95% confidence interval in parentheses

Table OA11: Mean Ratings on DVs by Condition for Study 2 with 95% Confidence Intervals

Evaluation Scale / Thermometer / Traits Index
Representative is a Democrat
Baseline / 0.44 (0.28, 0.60) / 66.01 (62.11, 69.90) / 3.87 (3.76, 3.98)
Consistent / 0.21 (0.02, 0.40) / 62.17 (58.12, 66.22) / 3.71 (3.58, 3.84)
Repositioned (No J) / 0.05 (-0.12, 0.22) / 56.11 (52.60, 59.63) / 3.66 (3.54, 3.78)
Repositioned (New Info) / 0.42 (0.26, 0.58) / 63.70 (60.53, 66.87) / 3.91 (3.80, 4.02)
Repositioned (P. Fairness) / 0.27 (0.10, 0.44) / 59.71 (56.09, 63.34) / 3.83 (3.71, 3.95)
Representative is a Republican
Baseline / -0.09 (-0.28, 0.09) / 54.11 (49.79, 58.45) / 3.50 (3.37, 3.63)
Consistent / -0.41 (-0.60, -0.22) / 47.02 (42.34, 51.70) / 3.34 (3.21, 3.47)
Repositioned (No J) / -0.51 (-0.70, -0.32) / 41.27 (37.31, 45.23) / 3.32 (3.19, 3.45)
Repositioned (New Info) / -0.16 (-0.36, 0.04) / 50.62 (46.75, 54.50) / 3.55 (3.41, 3.70)
Repositioned (P. Fairness) / -0.16 (-0.36, 0.04) / 49.26 (45.27, 53.25) / 3.51 (3.37, 3.64)

Note: 95% confidence interval in parentheses

Pre-Test Analyses

We pretested the two justifications used in Study 1, along with three others, using a sample from Amazon’s Mechanical Turk; this sample was independent of the one used in Study 2 (and indeed the pre-test was run well over a year before the second study). A total of 446 individuals participated and were randomly exposed to the same stimuli as the experimental subjects but only asked to indicate how satisfied they were with the account and whether they gave the candidate credit/blame/or both for the change in position.

In addition to the comparison and societal fairness accounts, we also fielded justifications wherein the candidate indicated they felt compelled to change positions to satisfy their conscience, indicating that they changed positions because they had come to believe that the alternative would be a worse outcome, or highlighted the potential economic benefits (if con to pro) or costs (if pro to con) that would come from passing the Dream Act.

Table O12 provides the results from this exercise. We were particularly interested in seeing if there were any noticeable differences between justifications, between the directions within a justification (i.e. pro to con vs. con to pro), and, very important, whether there was sufficient variation on the outcome measures so that we could explore individuals at both high and low levels of satisfaction. The ‘conscience’ and ‘could have been worse’ justifications were quickly dropped because of their poor performance on the account satisfaction measure compared to the other justifications (although the ‘could have been worse’ justification performed admirably on the credit giving variable). Knowing that we would have room on our survey for two justifications, our decision came down to the comparison of ends and economic costs/benefits justifications given that the societal fairness measure performed well overall and because of a substantive interest in this type of account. We were initially intrigued by the high level of performance of the economic benefits justification (i.e. the con to pro category), but ultimately decided to use the comparison of ends justification as (1) there ostensibly no difference in the pro-to-con/con-to-pro sub-arguments which could not be said for the economic justification and (2) we feared that the economic justifications were conceptually distinct enough (given that one focused on benefits and the others costs, which could plausibly make salient issues of loss aversion and negativity bias) that we felt this justification should be disqualified.