Political Discourse and Civic Protests

There are 6 treatment arms: A, B, C , D, E and F, that are directly read from different scripts. The objective is to learn first how people react to the high gas revenues expectations presented by many actors in Tanzanian political discourse. To this end, we compare the survey answers of one group (A) who is not read any text, with group (B) who is read the script below (high expectations script):
Script 1
Tanzania has discovered large amounts of natural gas offshore Mtwara and Lindi. Some politicians have previously stated that these gas resources could generate very large revenues and many new jobs. The gas resources, could transform Tanzania into a much wealthier country. In that respect, one leading politician recently stated that “the government have embarked on a grand plan, and Tanzanians in general should expect economic revolution in few years to come.”

Today, the expected gas revenues are much lower than what they have been in the past. Some experts fear that the new moderate discourses influence citizens trust in the government and propensity to protest. To test this, group C is read the same script as group B, plus (moderate expectations script):
Script 2
Today, however, Tanzanian future gas benefits and revenues are in reality very uncertain. According to some recent estimates, if things continue the way they are, there could in fact be very modest revenues, if any, from gas extraction.
One of the main reasons why the revenues could be low, is the resource curse and the rentier state. People may react strongly to this possibility of gas revenues weakening the public sector. To analyze resource curse effects, group D will also be read this script (resource curse script)
Script 3
One reason is that gas revenues are often poorly managed, weaken the bureaucracy and lead to a more authoritarian state. Potential incomes and benefits then become much smaller than what they could have been.
Another potential adverse effect of petroleum revenues is to increase income inequalities. Again, that could be an important source of factor changing citizens’ views. Group E will then also be read this script (inequality script):
Script 4
Moreover, gas revenues are often unequally distributed. Some few, often those already rich, typically become even richer, while the majority does not benefit much, causing inequality to rise.
Finally, group F is read all scripts.

2016-10-01

2016-12-01

The main outcomes, Y1, Y2 and Y3 are our measures of propensity to protest and resort to violence. They come from the following questions:
Y1
Which of the following statements is closest to your personal view?
Statement 1: People should not take part in protest marches against the government
Statement 2: People should take part in protest marches to hold the government to account.

Y2
Which of the following statements is closest to your personal view?
Statement 1: The use of violence can never be justified
Statement 2: It can be necessary to use violence in support of a just cause

Y1 and Y2 are equal to 1 if the respondent agrees, or strongly agrees with statement 2, and equal to 0 otherwise.

Y3 is equal to 1 if Y1 = 1 or Y2 = 1, and equal to zero otherwise.

To investigate the mechanisms behind an impact on Y1, Y2 and Y3, we will use the following additional outcomes:
Y4 - trust in the government - is equal to 1 if the respondent agrees, or strongly agrees with the statement “In the years to come, I trust the government to do what is right for Tanzania.”
Y5 – national economic expectations - the respondent is asked “Looking ahead, do you expect in the next twelve months better or worse economic conditions in this country”
Y6 – individual economic expectations - the respondent is asked “Looking ahead, do you expect in the next twelve months better or worse economic conditions for you”
For Y5 and Y6, the respondent must answer on a 5 points scale (from very bad to very good). Y5, and Y6, are equal to 1 if the respondent’s answer is good or very good. The middle category “neither good nor bad” will be included either with “very bad, bad” if that group is smaller than the groups that answered “very good, good”.
Y7 – tax compliance – is equal to one if the respondent answers “I will definitely pay” or
“I will most likely pay” to the question “How likely are you to pay your taxes in the future?”

We draw a random sample of households in three regions of Tanzania: the largest city, Dar Es Salaam, and Lindi and Mtwara where important reserves of gas have been recently discovered. In each region we sample ten wards and 34 villages (or their urban equivalent sub wards). In each village, we sample 30 respondents (15 men and 15 women). The total sample size is therefore 3060.
There are 6 treatment arms: A, B, C , D, E and F. They are randomized at the respondent level and blocked by gender and village.

randomization done in office

individual

No

not clustered

3060

510