|
Field
Trial Status
|
Before
completed
|
After
on_going
|
|
Field
Abstract
|
Before
Our study will conduct a number of survey experiments to investigate the accuracy of different approaches widely used to estimate the prevalence of sensitive behavior in the social sciences.The context of the study is tax compliance in a sample of small- to mid-sized enterprises in Kampala, Uganda. By combining survey data with administrative tax records for the same firms, we assess the performance of different measurement approaches, benchmarked against the administrative data, within the same survey. This allows us to quantify measurement error in different self-reported measures of tax compliance and make recommendations for policy and research based on our findings.
|
After
Our study will conduct a number of survey experiments to investigate the accuracy of different approaches widely used to estimate the prevalence of sensitive behavior in the social sciences.The context of the study is tax compliance in a sample of small- to mid-sized enterprises in Kampala, Uganda. By combining survey data with administrative tax records for the same firms, we assess the performance of different measurement approaches, benchmarked against the administrative data, within the same survey. This allows us to quantify measurement error in different self-reported measures of tax compliance and make recommendations for policy and research based on our findings.
Based on the findings of the first survey data collection, we are running a second survey round to replicate the findings and study key mechanisms behind our results.
|
|
Field
Trial End Date
|
Before
October 21, 2023
|
After
March 31, 2025
|
|
Field
Last Published
|
Before
November 01, 2023 04:10 PM
|
After
February 09, 2025 02:29 PM
|
|
Field
Intervention (Public)
|
Before
We conduct a number of survey experiments to collect self reported data on tax compliance among the firms in our sample.
|
After
We conduct a number of survey experiments to collect self reported data on tax compliance among the firms in our sample. We then conduct a second survey to understand mechanisms behind respondents' reporting behavior.
|
|
Field
Intervention End Date
|
Before
October 21, 2023
|
After
December 11, 2024
|
|
Field
Planned Number of Clusters
|
Before
1500 firms
|
After
First survey: 1500 firms
Second survey: 2000 firms
|
|
Field
Planned Number of Observations
|
Before
1500 firms
|
After
First survey: 1500 firms
Second survey: 2000 firms
|
|
Field
Sample size (or number of clusters) by treatment arms
|
Before
Even split for each survey experiment and tax level. Expected sample sizes per experiment:
Experiment 1: List experiment
750 (URA), 750 (KCCA)
Experiment 2: Randomized Response
750 (URA), 750 (KCCA)
Experiment 3: Bayesian Truth Serum
750 (BTS), 750 (Control / Direct Questioning)
|
After
Even split for each survey experiment and tax level. Expected sample sizes per experiment:
** Survey 1 **
Experiment 1: List experiment
750 (URA), 750 (KCCA)
Experiment 2: Randomized Response
750 (URA), 750 (KCCA)
Experiment 3: Bayesian Truth Serum
750 (BTS), 750 (Control / Direct Questioning)
** Survey 2 **
Respondents were randomized into the following treatment arms
1) Municipal tax (trade license) compliance: 1000
2) National business tax compliance: 1000
For the question phrasing experiment, respondents in each treatment arm were randomized into
1) Original (negated) question format: 1800 (90%)
2) Non-negated question format: 200 (10%)
we conducted this randomization stratified by the abovementioned two treatment arms.
|
|
Field
Intervention (Hidden)
|
Before
We conduct a number of survey experiments to collect self reported data on tax compliance among the firms in our sample. Specifically, we compare estimates from 1) direct questioning, 2) a list experiment, 3) a randomized response, 4) incentivized response (Bayesian Truth Serum, Prelec 2004).
|
After
We conduct a number of survey experiments to collect self reported data on tax compliance among the firms in our sample. Specifically, we compare estimates from 1) direct questioning, 2) a list experiment, 3) a randomized response, 4) incentivized response (Bayesian Truth Serum, Prelec 2004). In a second survey, we study (i) whether our main findings replicate in a new sample of households from the same population, (ii) how differences in question phrasing affect our main findings, (iii) the mechanisms behind the response patterns we find (insufficient understanding, inattention/satisficing), (iv) alternative ways to obtain unbiased estimates of sensitive behavior in survey data (error-correction methods for existing techniques, enumerator guesses, second-order beliefs, ensemble/meta-analytical techniques that combine various estimates into one).
|