ABCDEFGHIJKLMNOPQRSTUVWXYZAAABACADAEAFAGAHAI
1
on-platform behavioroff-platform survey results
2
paperplatform/
estimand		daily active userstime spentpct feed political
(FB 13.5%, IG 5.3%)		pct feed political news
(FB 6.2%)		pct feed cross-cutting
(FB 20.7%)		pct feed untrustworthy (FB 2.6%, IG 1.3%)on-platform political engagementnews knowledgefactual discernmentaffective polarizationpolitical polarization compositeissue polarizationpartisan news clicksperceived legitimacysubjective well-beingself-reported
turnout		self-reported party line presidential vote
3
Guess et al. (Science) Chronological Feed
- (supplement, PAP)		FB, SATE-20.81%+1.99pp
[1.82,2.15]		-2.04pp
[-2.26,-1.83]		+1.88
[1.79,1.97]		-0.209sd
[-0.220, -0.198]		-0.035sd
[-0.061, -0.008]		0.002sd
[-0.032, 0.029]
-0.002sd
[-0.019, 0.016]
0.003sd
[-0.022. 0.0291
4
FB, PATE+1.68pp
[1.41,1.95]		-2.48pp
[-2.96,-2.00]		+1.59pp
[1.45,1.73]		-0.118sd
[-0.127,-0.108]		-0.01sd
[-.07,+.03]		0sd
[-0.03, +0.03]		0.02sd
[-.01,.04]		-0.02
[-0.06,+0.01]		-0.009sd
[-0.061,0.043]		+0.008
[-0.027,0.042]
5
IG, SATE-11.11%+0.23pp
[0.17,0.29]		+0.30pp
[0.17, 0.29]
-0.130sd
[-0.144. -0.1161
-0.000sd
[-0.025. 0.024]
0.007sd
[-0.009, 0.0231
-0.000sd
[-0.016, 0.015]
-0.002sd
[-0.027, 0.024]
6
IG, PATE+0.39pp
[0.26,0.52]		+0.30pp
[0.15, 0.44]		-0.9sd
[-.11,-.09]		0.02sd
[-0.03,+0.06]		0sd
[-0.02, +0.02]		+0.01sd
[-0.02,+0.03]		+0.02
[-0.01,+0.06]
7
Guess et al. (Science) Remove all reshares
- (supplement)		SATE-5.20%-2.69pp
[-2.82,-2.57]		-3.70
[-3.79, -3.61]		-0.54pp
[-0.69,-0.39]		-0.86pp
[-0.91 -0.81]		-0.069
[-0.095, -0.043]
–0.027
[–0.048, –0.005]
0.012
[–0.005, 0.030]
0.007
[–0.010, 0.025]
8
PATE-1.91pp
[-2.15,-1.67]		-2.70
[-2.84, -2.57]		-0.75pp
[-1.15,-0.34]		-0.83pp
[-0.91, -0.81]		-0.053
[-0.101, -0.004]		-0.002
[-0.032,0.029]
0.012
[–0.018, 0.042]
-0.019
[-0.038, 0.001]
9
Nyhan et al. (Nature) Downrank like-minded content
- (supplement)		PATE-0.68pp+7.3pp0.000sd
[-0.037, 0.038]		-0.00sd
[-0.029, 0.022]
0.02sd
[-0.027,0.0751]
0.01sd
[-0.044. 0.0621]
10
SATE
-0.02sd
[-0.050, 0.004]
-0.009sd
[-0.031, 0.013]
-0.017
[-0.031, -0.003]
-0.003
[-0.030, 0.024]
-0.009sd
[-0.037, 0.020]		-3.0pp
[-7.2, 1.1]
11
12
Experiments on social media
13
Bond et al. (2012), get-out-the-vote messages in 2010
14
Jones et al. (2017), get-out-the-vote messages in 2012+0.2pp
15
Allcott et al. (2020), 1 month FB deactivation in 2018
(supplement)		-90%?-0.2sd
(±0.1)		-0.05sd
[-0.13,+0.03]		-0.16sd
[-0.24,-0.08]		+0.09sd
[0.01,0.17]
16
Asimovic (2021), 1 week FB deactivation in Bosnia & Herzogivina
+0.24sd
[0.02,0.45]		+0.18sd
[-0.03,0.39]
17
Allcott et al. (2022) bonus: 3 week pay to reduce usage of a group of apps
+0.09sd
[0.01,0.17]
18
Allcott et al. (2022) limit: 3 week self-imposed limits on a group of apps
+0.04sd
[-0.02,0.10]
19
20
Facebook Papers
21
- Chronological FB timeline (2018), 1 week, all trending down
-3%
22
23
Experiments on political advertising
24
Aggarwal et al. (2023), "A 2 million-person, campaign-wide field experiment shows how digital advertising affects voter turnout"
8 month ad campaign for Biden to targed population
(appendix C says they constituted 0.5% of total election ad spend on FB, but argue that their ads were incremental because targeted to a small subgroup)
-0.06pp
(±0.12pp)		+0.7pp
(±0.4pp)
25
Kalla et al. (2018) "The minimal persuasive effects of campaign contact in general elections"
(mostly short ad campaigns)
+0.13pp
(±0.4pp)
26
Spenkuch and Toniatti (2016)
county spillovers, exposed voters to "one additional television campaign ad"
0pp+0.02pp
27
28
Studies of Media
29
Broockman et al. (2023): Pay Fox viewers to watch CNN for 1 month
-.15sd
(±0.2)
30
DellaVigna and Kaplan (2007): Natural experiment on Fox News exposure
+0.4-0.7pp
31
Martin & Yurukoglu (20127): Natural experiment on Fox News exposure
+0.3pp
(from 3 mins/week)
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100