Field
Experimental Design (Public)
|
Before
We randomized the rollout of an information campaign at the neighborhood level. In addition, neighborhoods were randomly assigned to a second treatment arm in which voters were exposed to campaigns with only persuasive content (``slogans"). We also cross-randomized the information source of the two campaigns across three types of sources. Randomization was stratified by six quantiles of the baseline outcome variable, 2018 vote share for the National Alliance, to estimate heterogeneous treatment effects and to increase statistical power. We selected a well-balanced draw from among 100000 re-randomizations to avoid other chance imbalances for each treatment arm.
|
After
We randomized the rollout of an information campaign at the neighborhood level. In addition, neighborhoods were randomly assigned to a second treatment arm in which voters were exposed to campaigns with only persuasive content (``slogans"). We also cross-randomized the information source of the two campaigns across three types of sources. Randomization was stratified by six quantiles of the 2018 vote share for the National Alliance to estimate heterogeneous treatment effects and to increase statistical power. We selected a well-balanced draw from among 100000 re-randomizations to avoid other chance imbalances for each treatment arm.
|