|
Field
Trial End Date
|
Before
February 27, 2023
|
After
March 31, 2023
|
|
Field
Last Published
|
Before
January 23, 2023 08:02 AM
|
After
February 20, 2023 11:11 AM
|
|
Field
Intervention End Date
|
Before
February 27, 2023
|
After
March 31, 2023
|
|
Field
Planned Number of Clusters
|
Before
at least 800 individuals
|
After
about 1500 individuals
|
|
Field
Planned Number of Observations
|
Before
800 individuals (same as clusters); they will be recontacted four weeks later for a second endline.
|
After
1500 individuals (same as clusters); they will be recontacted four weeks later for a second endline.
|
|
Field
Sample size (or number of clusters) by treatment arms
|
Before
The sample will be equally allocated to treatment and control - 400 treated, 400 control.
|
After
Respondents will be allocated to treatments and control in equal shares - 500 T1, 500 T2, 500 C.
|
|
Field
Power calculation: Minimum Detectable Effect Size for Main Outcomes
|
Before
Based on the outcome I have already measured (outcome 2, personal support for legal reform), the minimum detectable effect sizes of this experiment are reasonable: According to Haaland et al. (2020), the typical effect size in an informational experiment is about 15 percent of a standard deviation. For testing the null hypothesis that the treatment effect is 0, I obtain a minimum detectable effect size of 9.4 percentage points of a standard deviation (mean 0.345, SD 0.476), assuming N=800 equally allocated to T and C, ICC=0, alpha=0.05 and 80% power.
Power for detecting heterogeneity in treatment effects according to misperceived second-order beliefs will be lower. Based on the distribution of second-order beliefs from my previous survey, I expect 25% of respondents to underestimate, 10% to guess correctly and 65% to overestimate support for reform. Assuming N=400 allocated to T and a ratio of 2.6 of respondents who overestimate vs. underestimate support, ICC=0, alpha=0.05 and beta=0.8, I would be able to detect a 14.9 percentage points of a standard deviation increase in support for legal reform.
|
After
Based on the outcome I have already measured (outcome 2, personal support for legal reform), the minimum detectable effect sizes of this experiment are reasonable: According to Haaland et al. (2020), the typical effect size in an informational experiment is about 15 percent of a standard deviation. For testing the null hypothesis that the treatment effect is 0, I obtain a minimum detectable effect size of 8.4 percentage points of a standard deviation (mean 0.345, SD 0.476), assuming N=1000 equally allocated to T1 and C, ICC=0, alpha=0.05 and 80% power.
Power for detecting heterogeneity in treatment effects according to misperceived second-order beliefs will be lower. Based on the distribution of second-order beliefs from my previous survey, I expect 25% of respondents to underestimate, 10% to guess correctly and 65% to overestimate support for reform. Assuming N=500 allocated to T and a ratio of 2.6 of respondents who overestimate vs. underestimate support, ICC=0, alpha=0.05 and beta=0.8, I would be able to detect a 13.4 percentage points of a standard deviation increase in support for legal reform.
|
|
Field
Intervention (Hidden)
|
Before
All respondents are asked to guess how many Tunisians support reform of the current gender discriminatory inheritance law regime. After recording their guess, the enumerator will reveal the true percentage only to the respondents who are allocated to treatment (about 35%). Treated respondents are then given information on gifting as it is defined in the Tunisian legal code. They are told that gifting can be used for transferring assets to one’s daughter: By “gifting” their daughter, parents can arrange a more gender equitable inheritance split.
|
After
All respondents are asked to guess how many Tunisians support reform of the current gender discriminatory inheritance law regime. After recording their guess, the enumerator will reveal the true percentage (about 35%) only to the respondents who are allocated to treatment 1. Respondents in treatment group 1 and 2 are then given information on gifting as it is defined in the Tunisian legal code. They are told that gifting can be used for transferring assets to one’s daughter: By “gifting” their daughter, parents can arrange a more gender equitable inheritance split.
|