The Effects of WhatsApp on Politics: Evidence From a Deactivation Experiment

class: center, middle, inverse, title-slide

.title[
# The Effects of WhatsApp on Politics: Evidence From a Deactivation Experiment
]
.author[
### Tiago Ventura, Rajeshwari Majumdar, Jonathan Nagler, and Joshua A. Tucker <br> <br><span style="font-size: 80%;">Center for Social Media and Politics, NYU</span>
]
.date[
### <span style="font-size: 60%;"> <br> 80th Annual MPSA Conference <br> 04/15/2023 </span>
]

---

layout: true

<div class="my-footer"><span>Tiago Ventura (CSMaP) &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbsp &nbspMPSA Annual Meeting</span></div>

---
class:middle
## Motivation

---
### Research Goal

.content-box-gray[Provide causal estimates for the effects of exposure to misinformation on WhatsApp on political outcomes.]

---
class: inverse, center, middle

## WhatsApp Multimedia Deactivation Field Experiment during the weeks leading up to the 2022 Presidential Election in Brazil

---
## Deactivation Studies

---
class: middle

- Disable their automatic download of media on WhatsApp 
- Do not purposefully click and watch any media on WhatsApp

- **Start:** Three weeks before general election in Brazil
- **End:** three days after the voting day

---
## Overview of the Experiment

.panelset.sideways[

.panel[.panel-name[Recruitment]

- Facebook Ads targeting adults Brazil

- Started on September 8th

- Number of respondents: 1,947

- Participants willing to join the deactivation: 1,8861

]

.panel[.panel-name[Treatment Assignment]

- **Started the experiment on September 15**

- **Treatment**:
  
  - Turn off the automatic download of all multimedia
  - Three .red[weeks] without consuming images, videos, and audio on WhatsApp
  
- **Control**:

- Three .red[days] without consuming images, videos, and audio on WhatsApp

- **Invited**: 1,135 respondents

- **Completed**: 773 respondents
]

.panel[.panel-name[Deactivation]

.center[
<img src="./output/iphone_deact.jpeg" width="35%" />

]
]

.panel[.panel-name[Compliance Checks]

.pull-left[

**Compliance Tasks:**

- Screenshots of WhatsApp storage intormation

- Every week + final survey

- After the  first compliance check, the control group was activated

]

.pull-right[
.center[
<img src="./output/iphone_compliance.jpeg" width="70%" />
]
]

]

.panel[.panel-name[Post-Treatment Survey]

#### .darkblue[General Info]

- October 5th, three days after the election

- Complete Responses: 732

- Attrition Rate: 5.3%

]
]
---
class: middle

## Outcomes

**Self-Reported Exposure and Accuracy Judgment**

- False Rumors published on fact-checking webpages during the month of the experiment + True News stories headlines from mainstream media.

**Polarization**

- Affective polarization (Feeling Thermometer)
  - Social Polarization (Willingness to engage with outgroup voter)
  - Issue Polarization (Abortion, gun control, etc..)
  - Perceptions about ideological polarization ('Where do I observe the two main candidates?')

**Subjective Well-being**

- How did you feel for the past three weeks? (Happy, Anxious, etc...)

---
class:middle

## Hypotheses

H1: Deactivated users will report lower levels of previous exposure to FALSE rumors.

H2a: Deactivated  will display a higher ability to accurately identify FALSE rumors.

H2b Deactivated users will display lower levels of mainstream news knowledge.

H3: Deactivated will display lower levels of outgroup political polarization.

H4: Deactivated will display higher levels of the aggregated index of subjective well-being.

---
class: center, middle, inverse

# Results

---

### First-Stage Results

.center[
<img src="./output/models_compliance.png" width="90%" />
]

---
### Effects on Exposure to Misinformation Rumors (H1)

.center[
<img src="./output/exposure_h1.png" width="90%" />
]

---
### Reduction in exposure is large for false news

.center[
<img src="./output/difference_true_false_exposure.png" width="90%" />
]

---
### Effects on Belief Accuracy (H2)

.center[
<img src="./output/belief_accuracy_h2.png" width="90%" />
]

---
### Effects on Polarization & Subjective Well-Being (H3 & H4)

.center[
<img src="./output/pol_sw_index.png" width="90%" />
]

---
class: middle

## Discussion

---
class:middle

## Limitations

- Scope conditions: focus on direct effects of WhatsApp.

- Larger effect might exist exactly in hard to reach populations.

- Small effect size vs Power

---
class: middle

## Next steps.

### Replication, Replication, Replication

---
class:middle, center

## Thank you!

---
### Belief Accuracy Conditional on WhatsApp usage for Politics

.center[
<img src="./output/sif_het_use_whatsapp_fac.png" width="90%" />
]

---
### Polarization Outcomes

.center[
<img src="./output/pol_outcomes.png" width="90%" />
]

---
class:middle

## Examples of Misinformation Outcomes

### False Item:

_Only complete votes are counted by the electoral justice. If the voter only votes for the President, and votes blank for all the other races, the vote is considered a partial vote, and will be annulled_

### True Item:

_After consecutive reductions, the current oil price in Brazil is below the global average value_