Determinants of growth: Should we protest?

What information do protests give us about future growth? One can imagine that protests occur when the population isn’t content with the present situation. Motivations can be rooted in the poor state of the economy, lack of civil rights, religious discontent etc. In this paper I am looking at the relationship between protests and growth as well as other economic variables like unemployment, inflation, national debt, savings etc.

While the direct relationship between growth and protest activity appears to be negative it doesn’t have to be so. After all, a protest can serve as a signal to the government and impact growth indirectly though government enacting reforms rather than directly through protesters not showing up for work. I look at the growth pattern before and after the protest event and find structural breaks in growth patterns that are associated with protests. I use various specifications that differ by the frequency of the data, formulation of protest intensity variable and estimation method.

I am using GDELT (Global Database of Events, Language, and Tone) database that covers the majority of countries during 1979-2014 period to construct the variable on the intensity of protests. The dataset provides daily information on protests, but I aggregate it to quarterly and yearly levels. The data in GDELT comes from the protest coverage of major news outlets worldwide. The economic variables come from Penn World Tables 8.0 and OECD Stats.

The variable that serves as a proxy for protest intensity is compiled in three different ways. The simplest way to define it is to count the number of protests that occur in a certain country during a designated time period. This method doesn’t account for the magnitude of the protest thereby lumping small and large protest events together. Second specification counts the number of times a certain protest was mentioned in the media and gives me a proxy for intensity of the protest event. Lastly, I find a cut-off point for a large protest and make the variable binary rather than continuous. This allows me to interpret the results easier, although some of the information is lost due to this simplification.

I have found that on average there is a negative correlation between number and intensity of protests and subsequent growth controlling for other factors.  I don’t have full results yet, but expect them to not all show a negative relationship between economic variables and protests.