China | Elites | Central Committee | Hu Jintao

Dissertation Project

 

Networks of Power. Using Social Network Analysis to understand who will rule and who is really in chargein the Chinese Communist Party

Presented at APSA 2014, received John Sprague Award of APSA's Political Network section

 

This paper proposes a new theoretical and empirical approach to the understudied informal institutions of authoritarian regimes, relying only on publicly available information. It thus suggests a more principled, but also more nuanced way of measuring one such institution - informal networks - by inferring all possible ties among a set of elites. Social Network Analysis demonstrates that we can not only measure such a network among Chinese Communist elites - exponential random graph models confirm qualitative accounts of tie formation based on provincial origin, pre-P.R of China experience, and alleged factional alliance - but also that members of the future inner circle (the Politburo five or ten years later) can be identified solely by their specific network position capturing popularity as a coalition partner, and that
stable coalition leaders hold positions that prevent opponents from forming alternative coalitions. This finding is consistent with the results from an agent-based model, which associates the former with closeness centrality, and the latter with betweenness centrality. The paper thus demonstrates the importance of extending the empirical analysis of promotions in China beyond simple leader-follower dyads, by showing the effect of having multiple ties to power holders and an independent power base, and connects it with theories of authoritarian regimes, and coalition formation in particular.

 

 

 

 

Network Experiments

Turnout | Kazakhstan

Degrees of Separation: How Many Hands Must a Kazakh Citizen Shake until She Reaches the President? (with Adele Del Sordi)

Working paper available upon request

 

This paper explores how “close” residents of the two largest Kazakh cities (Almaty and Astana) are to the government - represented by a parliamentarian, a high-level official of the ruling party, and a member of the local administration in Almaty, respectively. The paper reports on two parts of a pilot: the results of an inquiry into the ego networks of a random sample of residents in the capital finds that respondents who have more government employees among their relatives are more likely to vote.

They also display a higher level of  diffuse support for the government, but are not more satisfied with concrete public goods provision. The effect is weaker if the relatives only work in menial positions.
The second part of the survey attempts to trace the informal chains between ordinary citizens and government representatives using the method employed by Milgram's “six degrees of separation” experiment: I ask volunteers and randomly sampled residents to relay a envelope with a greeting card to one of the government representatives by passing it to an acquaintance more likely to know the target official on a personal basis, and asking him or her to pass it on in a similar manner. Respondents apparently navigate the social network successfully: the messages reach more connected individuals the further they travel. Individuals closer to the government on those chains are also more likely to support the regime and turn out to vote.

 

 

 

Authoritarian Stability

 

(Why) Do Revolutions Spread?

Presented at APSA 2012

 

The research of revolutions - attempts to overthrow or change the government relying on mass movements - has been mostly limited to qualitative case studies. Big-N, quantitative studies have focused on events that are similar, but not identical to revolutions, like civil wars, political violence, democratization or general instability.

Revolutions | Map | Contagion

This paper introduces a new dataset of revolutionary events, covering independent states between 1946 and 2007. It tests if such events are "contagious", because opposition leaders and citizens use the situation in neighboring, similar countries as a heuristic to gauge hidden support for a popular uprising in their own country. The spatial model employed provides evidence that revolutions do indeed spread between autocracies, but less so among democracies. This is consistent with the mechanism suggested, as preference falsi fication is more common in autocracies. Other factors that increase the chance of a revolutionary event occuring are the size of the population, having a neither highly authoritarian nor very democratic regime, democratic neighbors and the absence of free resources like oil. Sharing regime type and geographical region with a country undergoing a revolution makes it more likely that the unrest will spread, while sharing the same cultural zone does not.

 

Protest | Contagion | Threshold model | Diffusion

How to Spot a Successful Revolution in Advance

Received Best Poster Award at PolNet 2013, currently under review, complete working paper available upon request

 

Despite conventional wisdom that a dictator is better o ff facing an opposition divided into many groups and with few connections between them, this paper finds that under some circumstances, a strategy of "divide and conquer" could actually help the emergence of a popular protest movements. Using network simulations, I model how individuals join protest movements based on their friend's decisions to protest and predict in what kind of societies a small group of protesters can set off a "protest cascade", mobilizing a large part of society.

In this examination of the spread of complex contagion in a network with homophilic clustering based on group membership, I find that both the level of segregation and the number of subgroups display a inverse U-shaped relationship with the probability of a cascade occurring. Furthermore, in a society with two groups, a smaller group can be more e ffective in starting a cascade than a larger one. This fi nding can be explained in terms of clustering induced by the size and number of groups. A low level of clustering prevents cascades from setting off , while a high level of clustering increases the probability that a cascade fails to spread to other subgroups. Unlike the findings of other theoretical studies on cascade capacity, the latter result can be validated empirically, because we do have information on the size and number of subgroups in di fferent societies. I therefore test the model's prediction using data on the size and number of ethnic groups and protests in Kyrgyzstan's 56 districts (2000-2007).

 

 

© 2021 Franziska Keller

Revolutionary Events worldwide 1946-2007