Understanding the resource curse: A survey experiment in Tanzania
Last registered on January 22, 2020

Pre-Trial

Trial Information
General Information
Title
Understanding the resource curse: A survey experiment in Tanzania
RCT ID
AEARCTR-0000768
Initial registration date
July 10, 2015
Last updated
January 22, 2020 3:44 AM EST
Location(s)
Primary Investigator
Affiliation
Centre for Applied Research at NHH (SNF) / Stockholm School of Economics
Other Primary Investigator(s)
PI Affiliation
Norwegian School of Economics
PI Affiliation
Norwegian School of Economics
PI Affiliation
REPOA
PI Affiliation
Chr. Michelsen Institute
Additional Trial Information
Status
Completed
Start date
2015-07-09
End date
2015-08-15
Secondary IDs
Abstract
Tanzania has discovered large reservoirs of off shore natural gas, and might become a large exporter of liquefied natural gas (LNG) during the next decades. With this comes the promise of significant petro-revenues to the Tanzanian government, and prospects of improved economic and social conditions for the country's population. However, the empirical literature suggests that natural resource wealth constitutes a considerable challenge. Natural resource wealth is associated with lower economic growth (Sachs and Warner 1995, Sachs and Warner 1999, Mehlum et al. 2006); prolonged authoritarian regimes (Aslaksen 2010, Ross 2001, 2009, Tsui 2011, Andersen and Aslaksen 2013); corruption (Vicente 2010, Leite and Weidmann 2002, Treismann 2007, Busse and Gröning 2013); and violent conflict (Collier and Hoeffler 2004).

While there seems to be a consensus that such a thing as a "resource curse" exists, there is large heterogeneity in the economic, social and political outcomes of natural resource wealth. Researchers have proposed different channels that can explain these diverging experiences. The present study will draw on the political economy literature that focuses on the quality of institutions, governance and democracy (Ross 2015, Busse and Gröning 2013). This research project aims to contribute to the scarce body of micro-level empirical investigations of the resource curse by investigating the effect of natural resource rents the political attitudes and behavior of citizens. In addition to look at citizens willingness to pay tax and engage in corrupt activities, we will also examine people's trust in the government. While trust has been shown to be an important determinant of economic growth, we have not found any studies investigating the effect of natural resource rents on the level of trust in a society.

The research questions we seek to answer are the following: How do expectations about resource revenues in a country affect
- people's willingness to pay taxes?
- people's willingness to engage in corrupt/rent-seeking activities?
- people's trust in the government?

To answer these questions, we will conduct a survey experiment that will also include incentivised games. In the experiment, respondents will first answer some key background questions. They are then randomly assigned to a control or one of four treatment groups. In all treatments, we will provide information about the estimated total value of the gas revenues, but how the totals are illustrated will be varied. We use a 2x2 design to study if the results are sensitive to whether the revenues are illustrated (a) in per capita or population terms and (b) in totals or in terms of the interest that can be generated in infinity if the money is saved.

The information manipulation is important both from a policy perspective, since it may inform policy-makers about the impact of how different informational responses may affect the public, but also because it may shed light on how the different informational approaches affect the mindset of people.

The purpose of the treatments is to generate exogenous variation in expectations about gas revenues, by increasing the salience of existing expectations among those who already know about them, and by creating expectations among those who do not know about gas revenues.
After receiving the treatment or control group information, respondents answer a series of questions intended to capture willingness to pay taxes, willingness to engage in corruption and trust in the government. The treatment manipulations allow us to study the causal effect of increased expectations about gas revenues on willingness to pay tax and to engage in corrupt activities, as well as trust in the government by simply reminding the participants in the treatment conditions about the fact that Tanzania has discovered large reservoirs of natural gas.

In the last part of the experiment, respondents an incentivised game. We have developed two games. The first game intends to measure trust in other citizens in economic decisions. The second game intends to measure cheating behavior (proxy for willingness to engage in corrupt activities). The incentivised games provide us with behavioral measures of how information about gas revenues, and the way in which it is presented, affects trust and cheating behavior in society.
External Link(s)
Registration Citation
Citation
Cappelen, Alexander et al. 2020. "Understanding the resource curse: A survey experiment in Tanzania." AEA RCT Registry. January 22. https://doi.org/10.1257/rct.768-2.1.
Former Citation
Cappelen, Alexander et al. 2020. "Understanding the resource curse: A survey experiment in Tanzania." AEA RCT Registry. January 22. https://www.socialscienceregistry.org/trials/768/history/61252.
Experimental Details
Interventions
Intervention(s)
Respondents are randomly assigned to one of four treatments or a control group. All groups will be presented with an information video with illustrations, soundtrack and text in Kiswahili. It will be displayed on the enumerator's tablet and the respondent will be provided with a headset to hear the sound.

The purpose of the treatment is to generate exogenous variation in the mindset of respondents when thinking about gas revenues. The idea is that providing information about gas to the respondents increases the salience of their expectations / cause them to form new expectations. We want to test whether the expectations causally affect attitudes towards tax and corruption, and trust in the government. As a robustness check we also want to investigate whether the different informational approaches have psychological effects on attitudes, beliefs and behavior of people.

The major difference between the control and treatment group videos is that the control condition does not provide any information about gas or gas revenues. In all treatments, we will provide information about the estimated total value of the gas revenues, but how the totals are illustrated will be varied. We use a 2x2 design to study if the results are sensitive to whether the revenues are illustrated (a) in per capita or population terms and (b) in totals or in terms of the interest that can be generated in infinity if the money is saved.

- Treatment 1: Annual real return if gas revenues are saved, per capita
- Treatment 2: Annual real return if gas revenues are saved, for all Tanzanians
- Treatment 3: Total value, per capita
- Treatment 4: Total value, for all Tanzanians


Intervention Start Date
2015-07-09
Intervention End Date
2015-08-15
Primary Outcomes
Primary Outcomes (end points)
1) Willingness to pay tax
2) Willingness to engage in corrupt activities
3) Trust in the government
Primary Outcomes (explanation)
1) Willingness to pay tax is measured by the survey question: "In order for Tanzania to achieve a desired social and economic development in the years to come, the taxes paid by ordinary Tanzanians, like the VAT, should"
- Decrease a lot (1)
- Decrease (2)
- Stay the same (3)
- Increase (4)
- Increase a lot (5)

2) Willingness to engage in corrupt activities is measured by the survey question: "In the years to come, I expect the extent of corrupt activities to"
- Decrease a lot (1)
- Decrease (2)
- Stay the same (3)
- Increase (4)
- Increase a lot (5)

This question is intended as a proxy for the respondent's own willingness to engage in corrupt activities.

3) Trust in the government is measured by the survey question: "In the years to come, I trust the government to do what is right for Tanzania"
- Strongly disagree (1)
- Disagree (2)
- Stay the same (3)
- Agree (4)
- Strongly disagree (5)

In addition to the main outcome variables, we pose questions on other aspects of tax, corruption and trust, as well as on expectations about provision of public services in the future. These questions will aid our understanding of the respondents' answers to the main outcome questions.
Secondary Outcomes
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
Experimental Design
Experimental Design
We have sampled three regions: Mtwara, Lindi and Dar es Salaam.

In Dar es Salaam we will include all three municipalities: Ilala, Kinondoni and Temeke. In Lindi, Lindi Municipal is the only urban district and it was therefore selected without randomization. In addition, the two rural districts Lindi (rural) and Nachingwea were randomly selected. In Mtwara, Mtwara Municipal was intentionally selected from the two urban districts to ensure that we include a costal district. Masasi and Newala were randomly selected among the rural districts. Within each of the 9 districts, 3 wards have been randomly selected using data from the 2012 household census. Within each of the 27 wards, three villages/streets will be randomly selected in the field. Within each of these 81 villages/streets, between 35 and 40 households will be randomly selected in the field.
Experimental Design Details
Randomization Method
Randomization of information videos and incentivised games will be conducted using the Qualtrics Offline Survey Application.
Randomization Unit
We randomize at the individual level.
Was the treatment clustered?
No
Experiment Characteristics
Sample size: planned number of clusters
3000 respondents
Sample size: planned number of observations
3000 respondents
Sample size (or number of clusters) by treatment arms
We plan to have 1000 respondents in the control group, and 500 in each of the five treatment groups.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
The sample size is powered to detect differences between all four treatment groups seen as a whole and the control group, for all the three main outcomes defined above. We adjust for multiple hypothesis testing using the Benferroni correction. With the planned sample of 3000 observations (distributed as 1000 observations in the control group, and 500 observations in each of the four treatment groups) we can detect effect sizes of 0.2 SD between treatment and control outcomes with a power of 99% and a level of significance of 5%. The planned sample will also allow us to detect effect sizes of 0.2 SD with a power of 78% and a level of significance of 5%.
IRB
INSTITUTIONAL REVIEW BOARDS (IRBs)
IRB Name
IRB Approval Date
IRB Approval Number
Analysis Plan
Analysis Plan Documents
Pre-analysis plan_ Understanding the resource curse_ a survey experiment in Tanzania.pdf

MD5: ddabe9a36e5a8d310e1b401047a0bc1a

SHA1: a32b460daa34f3d436709e6d80e13494e14063b9

Uploaded At: July 10, 2015

Post-Trial
Post Trial Information
Study Withdrawal
Intervention
Is the intervention completed?
Yes
Intervention Completion Date
August 15, 2015, 12:00 AM +00:00
Is data collection complete?
Yes
Data Collection Completion Date
August 15, 2015, 12:00 AM +00:00
Final Sample Size: Number of Clusters (Unit of Randomization)
3004 individuals
Was attrition correlated with treatment status?
Final Sample Size: Total Number of Observations
Final Sample Size (or Number of Clusters) by Treatment Arms
Data Publication
Data Publication
Is public data available?
No
Program Files
Program Files
Reports and Papers
Preliminary Reports
Relevant Papers
Abstract
Corruption is considered to be an important driver of the resource curse in developing countries. Based on a large-scale field experiment in Tanzania, this paper studies whether expectations about future natural resource revenues shape expectations about corruption and the willingness to engage in corrupt behavior. We find some evidence that information about the discovery of natural gas in Tanzania causes people to expect more corruption in the future, but no evidence of a corresponding effect on peoples’ willingness to engage in corrupt activities.
Citation
Cappelen, Alexander W., Odd-Helge Fjeldstad, Donald Mmari, Ingrid Hoem Sjursen, and Bertil Tungodden. "Understanding the resource curse: A large-scale experiment on corruption in Tanzania." CMI Working Paper (2018).