The research design combined two sources of variation to create four treatment groups in a factorial design (Table 1). First, the soap opera was broadcast via a community loudspeaker only reaching a portion of the community due to topographical conditions (the natural experiment component). Households within the loudspeaker’s reach were also randomly invited by the regional NGO to listen to the soap opera program at a community meeting, with the remaining households able to hear the broadcast at their homes regardless.
The authors conducted a survey experiment on 1,257 Turkish citizens who resided in southeastern and central Turkey in June 2014. The survey experiment presented different messages about the possible effects of hosting refugees—increased economic burden, disruption of ethnic balance, and ties with rebels, as well as a positive message of saving innocent women and children. These messages were tailored to resemble elite cues as they appear in the media.
Using a survey experiment, the authors randomize indirect and direct questions about support for state forces. The indirect measure is a list experiment, a method used in other contexts to reduce pressure on respondents to falsify preferences, due to fear of coercion or social sanctioning. Under a list experiment, one group receives a list of control items, and another group receives the same list with an item of interest added, and respondents in each sample are asked how many items they support.
In 2006, the UK government and the International Rescue Committee funded and implemented Community-Driven Reconstruction (CDR) projects across Liberia. Under these projects, a number of local councils throughout Liberia were given the chance to choose infrastructure projects for their villages and to implement their construction.
8:30 - Coffee and Light Breakfast
9:00-9:15 - Welcome and Introductions
The researchers study 300 respondents in 35 communities that experienced a wide range of violent events. To identify past violence, the authors measure how many deaths communities experienced during the war, and also interview individual households about their experiences during war.
This study discerns the type and extent of anti-American biases by measuring behavioral responses: whether or not subjects refuse to take the survey depending on who they believe sponsored it. In a mass attitude survey in Lebanon, subjects are randomly exposed to different survey sponsors, and their refusal rates are compared between treatment conditions.
In order to elicit honest levels of support, the researchers used a series of four endorsement experiments. Rather than asking civilians whether they supported the Taliban or ISAF, the survey asked whether they support four policies, which they prefaced with “The Taliban supports,” “The ISAF supports,” or “Some people support.” (The policies are in fact all endorsed by both sides.) This does not measure the individual’s support for each side, but provides an estimate of support for a group within a region’s population.
An experimental design was used in which the NSP program was partially extended to ten non-NSP districts in 2007; in each of these, 50 villages were selected and then randomly assigned to take part in NSP or not. Selected villages were grouped into matched pairs based on background characteristics and a requirement that they not be within 1km of each other; one unit in each pair was then randomly assigned to treatment.