COLLECTING AND ANALYZING SURVEY DATA
Survey research has played a less prominent role in the study of conflict processes. Two possible reasons are that survey data are often difficult to collect and sample surveys can be expensive to conduct.
For many conflict zones, population registries are not available, respondents are difficult to contact, and attributions about the researcher’s intent lead to suspicions about how the data will be used. But, even when lists are available and respondents are eager to be of assistance, the costs can be prohibitive. Most probability samples require large numbers of respondents, many of whom must speak through interpreters. Without substantial support, surveys - unlike many of the other approaches discussed in this chapter - are difficult to implement. With proper support, surveys can provide valuable insights into the way people think about their group identities and the implications of those attachments for the course of a conflict and its settlement. A good example of this kind of survey is the recent study by Burn (2006).Burn implemented a survey of citizens in two countries, Kyrgyzstan and Kazakhstan. Her team conducted face-to-face interviews with 753 citizens drawn from the most recent (1999) census in each country. Although these were not probability samples - in the sense that each citizen had the same chance of being chosen - a careful attempt was made to represent the demographic breakdown of groups. This was done by matching sample percentages with those in the population. For example, the census and sample breakdowns for ethnic groups in Kazakhstan was Kazakh (58% vs. 53%); Russian (23% vs. 22%); Uzbek (6 % vs. 13%); and other (13% vs. 11%). Representative samples have the advantage of ensuring that the key categories are included in the survey in proportion to the population; they have the disadvantage of over- or under-representing respondents from certain categories.
The proportional representation approach to sampling used by Burn was a strength of the study. Not only did this sampling approach enable her to examine impacts of the varied group identities in each society, it also shed light on important theoretical questions about factors that facilitate or impede transitions to democracy.The survey methodology used by Burn enabled her to compare groups and to assess relationships between the theoretically relevant concepts. One set of hypotheses compared the strength of clan identities among various demographic categories. For example, she hypothesized that clan identity will be stronger in rural than in urban areas. People living in urban areas have more opportunities to identify with other groups. Another set of hypotheses examined relationships among variables. For example, strong clan identity is correlated with greater tolerance for authoritarian leaders; it is also correlated with less support for democratic institutions. Each of the variables was defined by indicators that could easily be represented by survey questions. For example, salience of clan identity was measured by answers to three questions - willingness to financially support clan members, voting preferences for clan members, and preferential treatment of clan members (compared to non-clan members). The survey questions on clan identity captured several aspects of identity including commitment, spread, durability, and salience. These were the intervening variables in a framework that connected clan expectations, roles, and behavior to political institutions. The survey data provided evidence about the constraining and facilitating roles played by group identities in the transition of political systems.
Surveys are useful for gathering information about defined populations. For some issues, all members of the population can be interviewed. An example is Birkhoff’s (2001) study of mediators’ perspectives on power. She mailed surveys to each of the members of the Society for Professionals in Dispute Resolution (SPIDR), which were considered to be a population of professional mediators.
Although her return rate was only 37%, she was able to show that this “sample” represented the membership on key demographic variables. In this study, the “sample” was defined by the returns rather than by random selection from population lists. The responses to her questions suggested an interesting taxonomy of the varied meanings of power in practice. Other research questions that would benefit from surveys include the extent of polarization on values within societies, the spread and durability of group or national identity, and changes in perceptions of relationships with other societies following such interventions as problem-solving workshops. Panel (repeated) surveys would be particularly useful for gauging changes in attitudes through time. Conducting surveys in conjunction with other methodologies may be useful for linking micro- (small-group interactions) with macrolevel (societal trends) processes.