Slowing the Internet might slow Democracy

Our team Klaus Ackermann (University of Chicago), Simon Angus (Monash University), Roland Hodler (University of St Gallen) and Paul Raschky (Monash) is currently working on an empirical research project that analysis the effect of the Internet on political mobilization world-wide.

 

One particular focus of this project is to investigate the role of Internet latency on the number of weekly demonstrations around the world. Latency relates to the travel time that a data package requires to get from its source point to its destination point. In the context of political mobilization, Internet latency is essential because it defines the time it takes to upload or download a video or a photo, access particular sites or communicate via voice-over-IP.

 

Here we present some stylized facts based on a first look at our data.
[NOTE: The numbers and figures below are correlations and should not be interpreted in a causal way.]

The Data:

We have constructed a unique dataset that combines geo-referenced data on Internet access and Internet quality and weekly demonstrations for almost 20,000 subnational regions (i.e. US counties, Indian districts) from over 200 countries & territories over the years 2006-2012.

 

The data on weekly demonstrations worldwide was sourced from Global Database of Events, Language, and Tone (GDELT). We exclude all demonstrations that have been labelled “Violent Protests” as well as “Strikes & Boycotts”.

 

Data on weekly Internet Latency was constructed as part of a our ongoing research work. We have compiled a unified Internet activity and location dataset of an unparalleled scope and accuracy drawn from over a trillion (1.5×1012) observations of end-user internet connections, with temporal resolution of just 15 min over 2006-2012. First results have been published in Ackermann et al. (2017) and a description of the project can be found on one-trillion.org.

 

Some Stylized Facts:

Figure 1. presents the average number of weekly demonstration in a country if ping is below the 75% percentile (“High speed”, left bar) and above the 75% percentile (“Low speed”, right bar) of the COUNTRY’s median speed. The average number of demonstrations in weeks where the ping response time is below the 75% percentile (labelled as faster Internet weeks) is 8.62, while the number of demonstrations in weeks where the ping response time is above the country’s 75% (slow Internet weeks) is 6.64.

 

In other words, a country observes around 22% (1.98) fewer demonstrations during weeks with slower Internet (the difference is statistically significant at the 1%-level based on a simple t-test).
Note: Comparison of average number of weekly demonstrations in a sample of 236 countries & territories between 2006 and 2012. The left bar shows the mean of demonstrations in weeks when the country’s average ping speed was below the country’s 75 percentile of ping speed. The right bar shows the mean of demonstrations in weeks when the country’s average ping speed was above the country’s 75 percentile of ping speed. Data on weekly protests comes from gdelt.org. Data on countries’ weekly Internet latency comes from Ackermann et al. 2017).

 

Figure 2 shows the difference in the weekly number of demonstrations between faster and slower Internet weeks for a subset of 15 countries (China, Iran, India, Iraq, South Korea, North Korea, Russia, Pakistan, Vietnam, UAR, Tajikistan, Uzbekistan, Turkey, Egypt, Syria) with known government censorship of the Internet. In those countries, the number of weekly demonstrations over this time period is in general higher. The results show that in countries with Internet censorship, the average number of demonstrations is around 14% lower during weeks with worse Internet latency.

 

Note: Comparison of average number of weekly demonstrations in a sample of 15 countries with known government censorship of the Internet between 2006 and 2012. The left bar shows the mean of demonstrations in weeks when the country’s average ping speed was below the country’s 75 percentile of ping speed. The right bar shows the mean of demonstrations in weeks when the country’s average ping speed was above the country’s 75 percentile of ping speed. Data on weekly protests comes from gdelt.org. Data on countries’ weekly Internet latency comes from Ackermann et al. 2017).

 

In Figure 3, we compare the average number of weekly demonstrations in weeks with good and bad Internet Latency but exclude the 15 countries with known government sponsored Internet censorship. Once again, the number of demonstrations is around 23% lower in weeks with low Internet latency.

Note: Comparison of average number of weekly demonstrations in a sample of 221 countries & territories between 2006 and 2012. The left bar shows the mean of demonstrations in weeks when the country’s average ping speed was below the country’s 75 percentile of ping speed. The right bar shows the mean of demonstrations in weeks when the country’s average ping speed was above the country’s 75 percentile of ping speed. Data on weekly protests comes from gdelt.org. Data on countries’ weekly Internet latency comes from Ackermann et al. 2017).

 

Our project can have important implications for the ongoing policy debate about the importance of net neutrality. If Internet speed matters for political mobilization, any systematic discrimination in the handling of different Internet packets, could potentially have consequences for political participation and election outcomes.

 

Press and Media Contact Information

For press information and media resources please get in touch at: [email protected]



Leave a Reply

Your email address will not be published. Required fields are marked *