Wednesday, 28 November 2012

How to Identify Who Suffers Most from Riots and Why Does it Matter for Policy?

Jean Pierre Tranchant is a Research Fellow with the Vulnerability and Poverty Reduction TeamBy Jean-Pierre Tranchant

Rioting has received a great deal of attention from policymakers, scholars and the general public in the recent past notably due to the presence of high profile riots in Europe (Greece since the Euro Crisis, the UK in 2011 or France in 2005) and to considerably more deadly instances of communal rioting that occurred in places such as Nigeria and India (Assam riots this year, Khandamal riots in 2009). These are a reminder that riots can cause tremendous human suffering and threaten peace and stability in diverse societies. 

Yet we know very little as to who is primarily affected by instances of rioting. It is an important question if we want to understand the dynamics unleashed by and responsible for these episodes of violence. I discuss here how the methods we used in a recent research paper published with IDS colleagues Jaideep Gupte and Patricia Justino can be helpful to shed light on this issue.We surveyed close to 1100 households in urban Maharashtra as part of a long-standing study of civil violence and welfare. In our sample 12.5% of the respondents reported they were impacted by riots. Overwhelming proportions of them have not directly suffered from injuries or physical damages but has been adversely affected by the indirect consequences of riots. We further discussed these findings on the MICROCON blog. Indeed riots generate fear, strain community relations, disrupt markets and institutions, and are generally accompanied by heavy-handed measures to restore law and order such as curfew.

Ecological studies

Riots are localised events in that they unfold within some particular spaces. Thus by looking at a map of the events of violence, one can start noticing patterns. The Guardian displayed such a map at the occasion of the UK riots where the level of deprivation of very small geographical areas was matched with the location of events of rioting[1]. The visual correlation between poverty and violence was obvious. The quantitative academic literature on riots follows a fundamentally similar strategy, albeit in a more systematic fashion: researchers use a unit of analysis (census tracts, cities, states), collect relevant information on causes and consequences of rioting for each of these units and match it with geographical information on violence events. Regarding India, such studies suggest that riots are caused by e.g. low growth, population pressure, weak social spending, low level of social capital, proximity to elections and political configurations.

The problem with this approach is that it conflates space and people. For instance, even if riots tend to happen in poor places, it does not automatically follow that the poorest individuals are the most exposed. Similarly, if one finds an association in the data between social capital and low levels of violence, is it because places with high social capital are able to prevent violence[2] or because people with good social connections can survive and cope more easily amidst episodes of violence.[3]  By design, ecological studies of the kind described above cannot disentangle these effects. They are prone to the ecological fallacy whenever the relationships between two variables differ at various levels of analysis. Economics, sociology, criminology and education studies are abound with examples where such is the case, which cast doubt on the interpretation we can give to results of the riot studies.

Micro-level studies:

Another branch of empirical research on violence takes the opposite direction and stresses the individual level of analysis. The IDS plays a leading role in that stream of research, notably through the MICROCON and other research programmes hosted at the Conflict, Violence and Development Cluster. However, to eliminate various sources of bias, researchers usually feel the need to look at variations between households but within defined geographical areas. This can tell us whether individuals with good social connections are less vulnerable than socially isolated individuals, but it cannot say anything on whether areas with more social capital are more peaceful.

Multilevel analysis of riot victimisation in Maharashtra:

The methodology we used in our study of riots in Maharashtra allows us to combine micro and macro-level analysis. We sampled 1089 households in 45 (predominantly poor) neighbourhoods within 10 districts of Maharashtra. As such the structure of the data is nested so that households are nested within neighbourhoods which in turn are nested within districts. We used a multilevel statistical analysis to model riot victimisation as depending on household, neighbourhood and district effects.


Among other results we found that:

1) Within neighbourhoods, households with weak coping capacity, that are socially isolated, reliant on informal economic groups, and living close to crime areas are those who disproportionately suffer from riots.

2) Community characteristics matter: As argued by
  • Varshney, neighbourhoods with higher level of social capital are characterised by lesser victimisation rate. However, we don't find that these neighbourhoods see fewer riots, but rather that when hit by a riot; they are better able to shield their most vulnerable residents from its detrimental impact.
  • Victimisation tends to be higher in neighbourhoods with large caste fragmentation.
  • The effect of poverty is unclear: although the sampling design revealed that disenfranchised communities are over-represented among riot-affected sites, our results suggest that riot victimisation is higher among the least disenfranchised ones.
These results are useful for policy in that they emphasise the extent of indirect victimisation and identify potential channels to mitigate it. First and foremost, the results show that development policies aimed at reducing vulnerability should acknowledge vulnerability to violence as an important dimension. Indeed, economic, social, physical and violence vulnerabilities overlap and reinforce each others. Secondly, initiatives aimed at fostering community relations are useful in that they reduce the indirect sufferings caused by riots. Thirdly, these initiatives are all the more important in socially diverse environments (caste) which are the most fragile to the deleterious impact of riots. Tackling this victimisation is an important task to break the cycle of deprivation and violence.