The Elements of Political Persuasion: Content, Charisma, and Cue
Torun Dewan, Macartan Humphreys, Daniel Rubenson

Forthcoming, Economic Journal

Replication Data

Our goal is to identify the extent to which effective political communication depends on the content of the message being delivered or the type of people supporting the messages (endorsers). We estimate these effects by partnering with the British Columbians for BC-STV Campaign to undertake rigorous message testing of arguments in favor of the Single Transferable Vote (STV).

The core strategy is for canvassers to engage in standard door to door canvassing in the BC area. Our task will be to separate message from messenger effects by working with the campaign to vary which campaigners canvas in which areas, which messages they deliver as they canvas, and on which of the campaign endorsers they provide information (a list of the campaigns endorsers is available on the campaign website:

Twenty four campaigners visit approximately 2500 households, speaking to eligible voters about the STV vote. In addition they leave campaign material at each household. [Below there is a detailed description of the activities of the campaigners and the material that they provide]. A team of twenty four enumerators then follows up with a survey to households that are visited in order to collect information on understandings of the issues and voting intentions.

Variation in voting intentions is then related to exposure to different messages, messengers, and endorsers. In addition we ask respondents if they are willing to participate in a follow up telephone survey which would take place after the referendum in order to measure actual voting. More detail on the design is given in the canvasser's manual.

Ultimately this strategy will let us work out:

- Which messages are most effective?
- Which endorsements are most effective?
- Which messages/endorsements are most effective in which different areas?
- Which types of canvassers are most effective in which different areas?

In addition, data is collected on general intentions to vote and reasons for voting one way or another.

First Report (Non-Academic)

Campaign Website (Prev:
MOU with campaign (unsigned copy)

Past Referendum Data
Power Analysis Code

Treatments Summary
Zip File of Flyers [Large]
Treatment Allocation Code is contained in the replication archive.

**Canvasser's Manual
Canvasser Webform for Data Entry
Canvasser Daily Sheet (216 pages)
Canvasser Training

Enumerator's Manual
Survey Instrument
Enumerator Daily Sheets (216 pages)
Enumerator Recruitment
Survey Web Form (Prev at:
Enumerator Training