Deliberative Democracy and Resource Rents in Tanzania
Last registered on April 12, 2015

Pre-Trial

Trial Information
General Information
Title
Deliberative Democracy and Resource Rents in Tanzania
RCT ID
AEARCTR-0000689
Initial registration date
April 12, 2015
Last updated
April 12, 2015 10:51 AM EDT
Location(s)
Primary Investigator
Affiliation
Center for Global Development
Other Primary Investigator(s)
PI Affiliation
Stanford University
PI Affiliation
Center for Global Development
PI Affiliation
Center for Global Development
Additional Trial Information
Status
On going
Start date
2015-01-27
End date
2015-10-25
Secondary IDs
Abstract
In 2010, Tanzania discovered natural gas reserves off its southern coast worth roughly fifteen times its annual GDP at prevailing prices. The large literature on the resource curse suggests this new gas discovery brings considerable risks to a low-income country with already weak governance institutions. How should Tanzania manage this new revenue to avoid the traditional resource curse? We will report on multiple rounds of a nationally representative opinion poll soliciting ordinary Tanzanians' views on how to manage and spend gas revenues. Between polling rounds, a random subset of poll respondents will receive detailed information about the natural gas discovery, and the pros and cons of various gas policy options. A randomly drawn subset of individuals receiving the information treatment will be invited to a national deliberative event, where they will debate gas policy options and participate in question and answer sessions with experts. We will report treatment effects of information and deliberation on support for alternative uses of gas revenue, including saving versus spending, and direct distribution of resource rents. We will also test a number of hypotheses about the process of deliberation itself, related to peer effects, leader effects, and polarization.
External Link(s)
Registration Citation
Citation
Birdsall, Nancy et al. 2015. "Deliberative Democracy and Resource Rents in Tanzania." AEA RCT Registry. April 12. https://www.socialscienceregistry.org/trials/689/history/4100
Sponsors & Partners

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information
Experimental Details
Interventions
Intervention(s)
The experiment includes two interventions. The deliberation treatment will happen after registration of this trial. The information treatment is of secondary importance (primarily intended to measure mechanisms underlying effects of the deliberation treatment) but comes first chronologically, and has already taken place at the time of registration.

1. Information treatment

A random subset of poll respondents will receive detailed information about the natural gas discovery, and the pros and cons of various gas policy options. The information will be provided in the form of a roughly thirty-minute video, screened in the field by survey teams after the baseline poll. The video aims to provide a balanced view of controversial alternatives, and is based on a script written with input and approval from a panel of researchers, Tanzanian industry representatives, civil society leaders, and politicians spanning all three major political parties as well as relevant government agencies.

2. Deliberation treatment

A randomly drawn subset of individuals receiving the information treatment will be invited to a national deliberative event. The event will be held over two days in April 2015, in Dar es Salaam, with travel and accommodation provided by the project. At the event, poll respondents will participate in small group deliberations about specific gas policy options, followed by question and answer sessions with experts.
Intervention Start Date
2015-04-13
Intervention End Date
2015-04-14
Primary Outcomes
Primary Outcomes (end points)
The main outcome variables are based on poll responses about preferences for how to manage or spend natural gas revenues. These include:
1. Support for extracting and exporting natural gas
2. Support for saving rather than spending gas revenue
3. Support for government spending versus direct distribution of rents
4. Support for government spending on social services versus infrastructure, transport, and industry
5. Support for transparency and oversight of gas revenues, or additional restrictions on their use

We will also examine treatment effects on survey responses measuring:
6. Knowledge of the natural gas discovery

Primary Outcomes (explanation)
Each of the six numbered outcomes above will be measured as a mean effects index combining several survey questions. The components of each index are described in brief below. Full wording of survey questions is included in the questionnaire also uploaded to this registry. Codes listed after the dash ("--") correspond to question numbers in the survey.

[See also the pre-analysis plan uploaded to this registry.]

1. Support for extracting and exporting natural gas

* Support for extracting gas -- [no baseline data]
* Support for using gas for domestic energy consumption -- [no baseline data]
* Support for using gas to subsidize domestic energy versus exporting to earn government revenue -- [no baseline data]
* Support for exporting gas and paying the full price for gas and using earnings from selling gas to spend on public goods -- s2q5
* Support for exporting the gas at the highest possible price -- s2q6

2. Support for saving rather than spending gas revenue

* Support for strict limits on spending -- s2q3
* Aversion to spending money on building things for the people, such as roads and the electricity system -- s2q7
* Aversion to spending money on public services, such as health care and education -- s2q8
* Save money for future generations -- s2q11
* Support for saving the majority of money for the future as opposed to spending more money now -- s2q20
* Support for saving as opposed to using expected money from gas to borrow money and spend on domestic needs sooner -- s2q27
* Building public goods as opposed to providing social services -- s2q21

3. Support for government spending versus direct distribution of rents

* Give money directly to households -- s2q9
* Give money to households with children or the elderly -- s2q10
* Giving money to households will fight poverty and hunger -- s2q14
* Giving money directly to people will help children in poor families have better nutrition and a greater chance of doing well in school -- s2q15
* Giving money directly to households will make the government more accountable for the amount people expect each year -- s2q16
* Giving money directly to people as opposed to giving it to government to spend through the national budget -- s2q22
* Support for a child wealth account -- s2q24
* Support for an adult wealth account -- s2q25

4. Support for government spending on social services versus infrastructure, transport, and industry

* Using the money to build public goods will create jobs -- s2q12
* Using the money to build public goods will help the economy grow faster s2q13
* Spending money on social services will help poor people -- s2q17
* Spending money on social services will help the economy grow faster -- s2q18
* Belief that when government increases spending on infrastructure, a lot of money is wasted -- s2q26
* Building public goods as opposed to providing social services -- s2q21

5. Support for transparency and oversight of gas revenues, or additional restrictions on their use

* Support for publishing oil and gas contracts -- s2q2
* Support for money being managed by an independent and international group of experts -- s2q4
* Belief that government will spend money on that which is most needed -- s2q19
* Knowledge of Tanzania's natural gas discoveries -- s2q1
* Where the natural gas was discovered -- s2q28

6. Knowledge of the natural gas discovery

* Knowledge of under whose presidency recent offshore natural gas discoveries were made -- s2q29
* Knowledge of whether the government has started receiving revenues from natural gas -- s2q30
* Knowledge of expected revenue amount compared to the cost of a new hospital -- s2q31a
* Knowledge of expected revenue amount compared to the cost of 10 new regional hospitals -- s2q31b
* Knowledge of expected revenue amount compared to the entire national government budget -- s2q31c
* Knowledge of expected revenue amount compared to the size of the whole national economy -- s2q31d
* Knowledge of expected annual revenue divided equally among Tanzanian citizens -- s2q32


Secondary Outcomes
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
Experimental Design
Experimental Design
A nationally representative baseline poll asked a target sample of 2,000 adult Tanzanians in 200 rural and urban clusters about their knowledge and policy preferences regarding Tanzania's recent natural gas discovery and the use of any future revenues. All respondents received a free mobile phone to enable follow-up polling by phone.

In each of 100 randomly selected survey clusters, 7 of 10 respondents were invited to view a documentary video presenting pros and cons of various gas policy options (the information treatment). Of the respondents who were invited to the information treatment, 4 individuals were invited to a democratic deliberation about the use of the natural gas.

The 400 invitees to the deliberative event will be divided into 25 randomly assigned small groups to discuss a sequence of gas policy options. The discussions will be broken into four rounds, and groups will be randomly reassigned after each round.

Follow-up polling will measure the medium term (i.e., weeks rather than hours after treatment) impact of information and deliberation on respondents' knowledge and policy preferences. Follow-up polling will be conducted by phone and broken into multiple rounds, due to limitations on respondents' attention span on the phone.
Experimental Design Details
Randomization Method
Both clusters and individuals were randomly assigned to treatment arms by one of the PIs on a computer in the office using Stata.
Randomization Unit
1. Cluster level randomization: The survey spans 200 primary sampling units (PSUs). We use the terms cluster and PSU interchangeably here. Of these 100 clusters, individuals chosen to participate in the information and deliberation treatments will be drawn exclusively from 100 treatment clusters. In rural areas, clusters are defined as a sub-village or hamlet (kitongoji). In urban areas, clusters are defined as a block or sub-ward (mtaa). Ten individual poll respondents were sampled from each PSU.

2. Individual level randomization: Within the 100 treatment clusters, individual participants for the information treatment and deliberation treatment will be drawn at random. (For the information treatment, this will include 7 of 10 individuals in each treatment cluster. For the deliberation treatment, this will include 4 of 10 individuals in each treatment cluster, a strict subset of the information treatment group.)

3. Random formation of deliberative groups. The 400 individuals to be invited to the deliberative event will be assigned to small groups to deliberate on policy options. Assuming perfect compliance (i.e., all invitees attend the event), participants will be assigned to 25 groups of 16 people. The event will span four rounds of deliberation. Groups will be randomly reassigned for each of the four rounds. The random assignment of moderators to deliberative groups is implicit in this design.
Was the treatment clustered?
Yes
Experiment Characteristics
Sample size: planned number of clusters
200 primary sampling units (sub-villages or "vitongoji" in rural areas; blocks or "mitaa" in urban areas)
Sample size: planned number of observations
2,000 individuals
Sample size (or number of clusters) by treatment arms
Pure control clusters = 100
Treatment clusters = 100

Within treatment clusters:
Information treatment = 700 individuals, 7 per treatment cluster
Deliberation treatment = 400 individuals, 4 per treatment cluster (subset of info treatment arm)
Control group within treatment clusters = 300 individuals, 3 per treatment cluster
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
We present the minimum detectable effect (MDE) for each treatment arm on a hypothetical measure of public opinion. We calculate the MDE for two parameters: 1. The intent-to-treat effect of deliberation plus information. 2. The intent-to-treat effect of information alone. Our design is intended to measure spillover effects (i.e., information flows) within clusters, and we focus here on the MDE for treatment effects that are not contaminated by spillovers, comparing individuals invited to treatment with control individuals in pure control villages. We calculate MDEs based on a power (κ) of 80% and a significance level (α) of 5%. A key unknown parameter is the intraclass correlation of responses within our clusters. We estimate this correlation using data from Tanzania’s National Panel Survey (NPS), rounds 1 and 2. The NPS is uniquely suited to our purposes here, in that it (a) has a clustered sample design, (b) collects information on public opinion, in this case support for the respondent’s member of parliament, and (c) follows the same respondent over time to enable us to calculate variances and intraclass correlations in terms of both levels and changes. Using the NPS data, we conducted power calculations using a variety of possible outcome measures: MDEs in terms of a binary response variable, and levels and changes of both variables. Both variables show an intraclass correlation of approximately 0.16. All calculations below use this parameter value as a conservative estimate of our anticipated MDE. Because the experimental design proposed above involves unequal divisions of individuals between treatment and control groups (and between various treatment arms), we calculate MDEs by simulation in Stata based on randomly generated numbers with the intraclass correlation found in the NPS and the sample design described above. Each repetition of the simulation produces slightly different standard errors and thus MDEs. We repeated the simulation 20 times and averaged the MDEs over all iterations. Results show that we will be able to detect impacts of the information treatment on public opinion of roughly 7 percentage points in either direction on a binary outcome, and approximately 8 percentage points for the deliberation treatment.
IRB
INSTITUTIONAL REVIEW BOARDS (IRBs)
IRB Name
Council on Science and Technology (COSTECH) Tanzania
IRB Approval Date
2014-09-09
IRB Approval Number
2014269NA2014133
Analysis Plan

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information
Post-Trial
Post Trial Information
Study Withdrawal
Intervention
Is the intervention completed?
No
Is data collection complete?
Data Publication
Data Publication
Is public data available?
No
Program Files
Program Files
Reports and Papers
Preliminary Reports
Relevant Papers