Feature Stories - 2006

Surveying rare populations for international studies

    Release Date: 9/19/2016

    By Benjamin Phillips, PhD

    Surveying rare populations pose budgetary, methodological and often times cultural challenges, particularly in the international context.  The challenges can vary by the target population.  We focus on surveying Muslims in this post. 

    Abt SRBI has conducted a number of surveys of Muslims: the 2007 Pew Muslim Survey, the 2011 Pew Muslim Survey, and surveys of Muslims in France, Germany, Indonesia, Malaysia, Niger, Nigeria, Senegal, and the United Kingdom.

    The smaller the Muslim population, the more of the study budget must be spent on speaking to (mostly) non-Muslims in order to find Muslims to survey. To make the most efficient use of research funds, we try to find estimates of the size of the Muslim population in different areas of a country. Based on these estimates, regions with a similar Muslim percentage of the population are grouped together into strata.  For pre-dominantly Muslim countries, the sample design must address the geographic dispersion of Muslims within the country.

    Identifying where Muslims are located is easiest in countries that collect information on religion in the national census, like the U.K. It is more challenging in countries that do not, such as France, Germany, and the U.S. As a large percentage of the Muslim population in these countries has migrated relatively recently, we can use proxy indicators that are recorded by government surveys and censuses like language (e.g., Arabic, Bengali, Pashto, Turkish, and Urdu) and national origin of migrants (e.g., Muslim-majority countries) to help identify where the Muslim population is to be found. In some cases, we can use other large surveys that asked about religion to help us calibrate our estimates. Our work in Germany was aided by a previous survey that estimated the Muslim population size of German states. In the future, we will build on a technique we first used in the 2013 Pew Jewish American survey  that used small area estimates (statistical models that blend together different types of data) to make more accurate estimates of the population, particularly in geographic areas with lower incidence. These estimates inform sample stratification and don’t have to be perfect, because we are still conducting screening interviews to find Muslims.

    In order for the survey to be affordable, it is usually necessary to make proportionally more efforts in strata with a higher percentage of Muslim residents. Increasingly, we are using what is termed “optimal allocation,” mathematical procedures that determine the distribution of interviews across strata that will result in the lowest margins of error for a given budget. Where lists of possible Muslims are available from commercial sources, we can also build these into a sample design to improve efficiency. Although these lists don’t come close to covering the entire Muslim population and don’t always get who is a Muslim right, they can help us to be more efficient. In some cases, areas with few, if any, Muslims may not be eligible for inclusion at all. Lists are very helpful here, allowing us to include at least some Muslims from these areas. The cultural diversity of Muslims across countries demand sensitivity to country-specific question wordings. This topic will be covered in future post.