We actually asked people this question before and **after** treatment (at the very start and end of the survey)
![[20240924175429.png]]
# pre vs post treatment candidate choice
```r
# no. of harris/trump votes pre/post treatment
condition harris_vote1 harris_vote2 trump_vote1 trump_vote2
<fctr> <int> <int> <int> <int>
1: persuadeHarris 570 580 477 461
2: persuadeTrump 565 547 428 431
condition topic harris_vote1 harris_vote2 trump_vote1 trump_vote2
<fctr> <char> <int> <int> <int> <int>
1: persuadeHarris personality 279 287 241 230
2: persuadeHarris policy 291 293 236 231
3: persuadeTrump personality 277 270 211 211
4: persuadeTrump policy 288 277 217 220
# pre
# 249 non harris/trump votes
vote_1 N
<char> <int>
1: Kamala Harris 1135
2: Donald Trump 905
3: I would not vote, but I could have 79
4: I would not vote out of protest 65
5: Prefer not to say 39
6: I would not vote for reasons outside of my control 13
7: Jill Stein 8
8: Chase Oliver 7
9: Kennedy 4
10: Cornel West 4
11: Robert Kennedy Jr 2
12: RFK 2
13: Robert F. Kennedy 2
14: RFK Jr. 2
15: Peter Sonski 2
16: Robert Kennedy JR 1
17: Robert F. Kennedy, Jr. 1
18: uncertain 1
19: Kenndey 1
20: Bernie Sanders 1
21: RFK, Jr 1
22: anyone but those two 1
23: Libertarian Candidate 1
24: I don't know yet 1
25: RFK Jr 1
26: Uncertain right now 1
27: I'm still shopping, now that Robert F. Kennedy, Jr. has dropped out. 1
28: Manchin 1
29: Robert Kennedy 1
30: Absolutely anyone else. 1
31: none of the above 1
32: Robert F. Kennedy Jr. 1
33: Independent 1
34: Libertarian candidate 1
35: Mitt Romney 1
vote_1 N
# post
# 230 non harris/trump votes
vote_2 N
<char> <int>
1: Kamala Harris 1148
2: Donald Trump 911
3: I would not vote, but I could have 73
4: I would not vote out of protest 53
5: Prefer not to say 44
6: Jill Stein 10
7: I would not vote for reasons outside of my control 7
8: Chase Oliver 7
9: Robert Kennedy Jr 3
10: Cornel West 3
11: Kennedy 2
12: RFK 2
13: Robert F. Kennedy 2
14: RFK Jr 2
15: Peter Sonski 2
16: Whomever is the candidate for the Green Party 1
17: Robert F. Kennedy, Jr. 1
18: kenndy 1
19: Bernie Sanders 1
20: RFK, Jr 1
21: cornel west 1
22: unsure 1
23: Libertarian Candidate 1
24: Not sure 1
25: no 1
26: Libertarian candidate Chase Oliver 1
27: Jasmine Sherman 1
28: Manchin 1
29: Robert Kennedy 1
30: Robert F Kennedy Jr 1
31: Anyone else. Maybe my neighbor's dog. 1
32: none of the above 1
33: Robert F. Kennedy Jr. 1
34: 1
35: Mitt Romney 1
vote_2 N
```
# Trump vote
Recode responses such that if response is 'trump', 1, else 0. `vote_trump_2` is post, `vote_trump_1` is pre
```r
# logistic regression
> feglm(vote_trump_2 ~ condition * scale(vote_trump_1), data = d3, family = binomial)
GLM estimation, family = binomial, Dep. Var.: vote_trump_2
Observations: 2,289
Standard-errors: IID
Estimate Std. Error z value Pr(>|z|)
(Intercept) -1.403017 0.235233 -5.964375 2.4557e-09 ***
conditionpersuadeTrump 1.115540 0.337157 3.308664 9.3742e-04 ***
scale(vote_trump_1) 3.754290 0.210031 17.874911 < 2.2e-16 ***
conditionpersuadeTrump:scale(vote_trump_1) 0.250103 0.341863 0.731589 4.6442e-01
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Log-Likelihood: -233.7 Adj. Pseudo R2: 0.846189
BIC: 498.3 Squared Cor.: 0.911351
# linear probability model
> feols(vote_trump_2 ~ condition * scale(vote_trump_1), data = d3)
OLS estimation, Dep. Var.: vote_trump_2
Observations: 2,289
Standard-errors: IID
Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.387511 0.004281 90.516477 < 2.2e-16 ***
conditionpersuadeTrump 0.021463 0.006101 3.517747 0.00044372 ***
scale(vote_trump_1) 0.464871 0.004254 109.279433 < 2.2e-16 ***
conditionpersuadeTrump:scale(vote_trump_1) 0.005603 0.006105 0.917796 0.35882268
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
RMSE: 0.145738 Adj. R2: 0.911235
```
# Harris vote
Recode responses such that if response is 'harris', 1, else 0. `vote_harris_2` is post, `vote_harris_1` is pre
```r
# logistic regression
> feglm(vote_harris_2 ~ condition * scale(vote_harris_1), data = d3, family = binomial)
GLM estimation, family = binomial, Dep. Var.: vote_harris_2
Observations: 2,289
Standard-errors: IID
Estimate Std. Error z value Pr(>|z|)
(Intercept) -0.408422 0.195159 -2.09276 0.03637037 *
conditionpersuadeHarris 1.656430 0.414548 3.99575 0.00006449 ***
scale(vote_harris_1) 3.633363 0.194630 18.66806 < 2.2e-16 ***
conditionpersuadeHarris:scale(vote_harris_1) 0.732176 0.416656 1.75727 0.07887246 .
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Log-Likelihood: -259.9 Adj. Pseudo R2: 0.834284
BIC: 550.8 Squared Cor.: 0.90058
# linear probability > feols(vote_harris_2 ~ condition * scale(vote_harris_1), data = d3)
OLS estimation, Dep. Var.: vote_harris_2
Observations: 2,289
Standard-errors: IID
Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.486391 0.004701 103.475161 < 2.2e-16 ***
conditionpersuadeHarris 0.029861 0.006597 4.526238 6.3123e-06 ***
scale(vote_harris_1) 0.472606 0.004701 100.529828 < 2.2e-16 ***
conditionpersuadeHarris:scale(vote_harris_1) 0.003768 0.006599 0.570974 5.6807e-01
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
RMSE: 0.157654 Adj. R2: 0.900449
```