The role of fear in psychological costs: an
experiment of domestic violence reporting
Marylis Fantoni
PhD Candidate
Indiana University
[Link]
And yet...chronic underreporting is a reality worldwide.
47%
(National Crime Victimization Survey, 2023)
29%
(Brazilian Safety Forum, 2023)
(European Agency for Fundamental Rights, 2014)
And yet...chronic underreporting is a reality worldwide.
47%
(National Crime Victimization Survey, 2023)
29%
(Brazilian Safety Forum, 2023)
(European Agency for Fundamental Rights, 2014)
Reporting domestic violence is an extremely
burdensome task, with survivors enduring many
psychological costs along the way.
Are there policy changes that can foster domestic
violence reporting?
Are there policy changes that can foster domestic
violence reporting?
Yes! Online reporting instead of the traditional police approach
increases requests of protective orders in 30%
Ex-ante cost-benefit analysis
“Before a woman considers contacting police for help, situational
and personal factors already in place affect her decision to consider
or seek police help.”
(Wolf et al, 2003, p. 123)
A key barrier for reporting domestic violence:
FEAR
Of retaliation Of police interaction Of consequences
1 2 3
by the abuser and/or response for oneself/family
(Wolf et al, 2003; Birdsey and Snowball, 2013; Evans and Feder, 2014; Goodman-Delahunty and Crehan, 2016; Decker et al, 2019; Heron and
Eisma, 2020;)
“I think some people do want the help, but they’re
terrified to call. I probably wouldn’t be alive if he’d
ever seen that I had called 911 off from my phone,
before the cops got there. That’s a reality. It’s a fear.”
(Decker et al, 2019)
“Many police did not take domestic violence reports seriously (...) Citizens feared
retaliatory victimization by police and/or perceived that complaining was futile”
(Goodman-Delahunty and Crehan, 2015)
“So it was like every time it (violence) would happen, I’m like oh
no, if I call the police, then they’re going to report me to
CPS(...) I sat there and took the beating instead of calling the
police because I was afraid they would take my child.”
(Wolf et al, 2003)
(Delahunty and Crehan, 2016)
(Decker et al, 2019)
Other
didn’t want to involve police
fear of retaliation
it wasn’t important
lack of proofs
didn’t think police could help
solved it on their own
(Brazilian Safety Forum, 2023) (National Crime Victimization Survey, 2023)
(Delahunty and Crehan, 2016)
(Decker et al, 2019)
Other
didn’t want to involve police
fear of retaliation
it wasn’t important
lack of proofs
didn’t think police could help
solved it on their own
(Brazilian Safety Forum, 2023) (National Crime Victimization Survey, 2023)
If we want to increase domestic violence reporting, policy
implementation processes - including public service
announcements - must consider and address the role of fear
in this process.
FEAR and Administrative Burdens
the onerous interactions between individuals and policy implementation
(Burden et al., 2012)
Learning costs
Compliance costs
Psychological costs
FEAR and Administrative Burdens
the onerous interactions between individuals and policy implementation
(Burden et al., 2012)
Learning costs
Compliance costs
STIGMA
Psychological costs
loss of autonomy? stress? uncertainty?
(Baekgaard and Tankink, 2021; Döring and Madsen, 2022)
FEAR and Administrative Burdens
the onerous interactions between individuals and policy implementation
(Burden et al., 2012)
Learning costs
Compliance costs “Further development of the cost categories
may help researchers become aware of
important phenomena not currently included
Psychological costs in the cost categories”
(Baekgaard and Tankink, 2021)
FEAR and Administrative Burdens
the onerous interactions between individuals and policy implementation
(Burden et al., 2012)
Learning costs
Compliance costs
STIGMA
Psychological costs
FEAR
A wave of behavioral intervention “nudging” studies:
(Bhargava and Manoli, 2015; Finkelstein and Notowidigdo, 2019; Bird, 2021; De La Rosa et al, 2021; Linos et al, 2022; Lasky-Fink and Linos, 2023)
Psychological ownership framing increased interest in claiming government
benefits by 20% to 128% in 3 experiments (De La Rosa et al, 2021).
Addressing stigma in psychological costs for rental assistance program
communication increased application requests by about 18%, with larger
effects for renters of color (Lasky-Fink and Linos, 2023).
Hypotheses:
1. By addressing fear in public service announcements related to
domestic violence, intent to report will increase.
2. There might be a heterogeneous treatment effect based on
age or socioeconomic status.
Social Media Experiment
Testing whether an ad addressing
fear in reporting domestic violence
receives more attention than one
that does not mention fear at all
*pre-registration: 10.17605/[Link]/JMQXB
Step 1. Ad construction
Target survivors
Remove gore depictions
Women as empowered
There’s life after DV
What laws/support are there
Importance of representation
(West, 2013; Martin et al., 2020; Reis et al., 2023;
Nascimento, 2024)
20% rule/Meta’s specifications
Targeting and relevance
On-brand association
Step 1. Ad construction
Target survivors
Remove gore depictions
Women as empowered
There’s life after DV
What laws/support are there
Importance of representation
(West, 2013; Martin et al., 2020; Reis et al., 2023;
Nascimento, 2024)
20% rule/Meta’s specifications
Targeting and relevance
On-brand association
Step 2. Validation
Mturk Validation Survey:
67 women in Brazil who have a Facebook account.
36% had some type of domestic violence experience.
69% between the ages of 26 and 40.
86% validated that the proposed ad addressed fear.
63% thought the ad provided a solution for fear.
35% thought the ad maybe provided a solution for fear.
Step 3. Pilot
Ran the ads for a week (August 20th to August 28th) in the
5th most populous city of the state of Rio de Janeiro.
City divided in two regions: subdistrict 1 and subdistrict 2.
N = 61,932
Learned how to work the ads.
Found a negative effect of 20%.
Social Media Experiment
Divided the city of Rio de Janeiro in 15 distinct regions. Randomization
occurred at the individual level within the regions. Ran both arms
(treatment and control) in each region.
Social Media Experiment
Ads were displayed only for
18-60 year old females. Ads
were live from Sep 24th to
October 20. Ads were shown
on Facebook and Instagram
mainly on feeds and stories.
Link on ad redirected towards
the online tool for requesting
protective orders.
Variables of interest
1. Clicks on the ad
2. Age of women (5 categories)
3. Number of women who saw the ad
4. Number of times the ad was seen on average
5. Dummy for low HDI regions
6. Dummy for treatment/control arm
7. Other engagement forms (likes, comments, saving, etc)
Main regression:
where = clicks/number of people in region
and = dummy for treatment indicator.
Main regression:
where = clicks/number of people in region
and = dummy for treatment indicator.
Heterogeneous treatment effects:
Descriptive statistics
Sample size: 206,503
Treatment group: 109,305
Control group: 97,198
Descriptive statistics
Average number of times ad was seen:
Treatment group: 5.8
Control group: 6.2
Treatment group click rate: 0.3
Control group click rate: 0.4
Overall number of clicks:
Treatment group: 2,510
Control group: 2,728
Results
The click rate is 18% lower
in the treatment group.
The ad that mentioned fear
actually discouraged
clicking.
Results - Heterogeneous treatment effects
There is no significant
difference on click rates
between control and
treatment groups on
regions with low HDI.
Results - Heterogeneous treatment effects
1. Click rates increase with age
25-34: 36.56% higher than 18-24
35-44: 129.01% higher than 18-24
45-54: 283.29% higher than 18-24
55-64: 446.86% higher than 18-24
Results - Heterogeneous treatment effects
[Link] treatment effect for 18-24;
nor significant difference for
25-44 age groups against 18-24
in treatment vs control
Results - Heterogeneous treatment effects
[Link] 45-64 age groups,
treatment significantly reduces
click rate by 20.5%
Conclusions
Addressing fear when advertising domestic violence reporting
REDUCES intent to report
This result seems to be largely driven by older women (45+).
There is no effect on socioeconomic status.
Although older women (45+) were the minority of the sample (19 to
11%), their overall click rates were 283-446% higher than 18-24 women.
Remaining Questions:
Why are the results in the opposite direction?
Is it that directing their attention to words of fear actually deters
them from reporting?
Is it because the text is longer?
Why are there age disparities specially in the older population?
Contributions
Given that domestic violence is heavily underreported worldwide,
understanding what might improve or hinder reporting has high salience.
The study also provides some more evidence to the collection the
behavioral intervention literature, and it adds two important new
contexts: domestic violence and a setting outside of the US.
Thank you!
