Carnegie Middle East Governance and Islam Dataset, 1988-2010

Data Archive > International Surveys and Data > Multiple Nation Surveys > Summary

The Carnegie Middle East Governance and Islam Dataset was created by Mark Tessler at the University of Michigan. The data set includes both individual-level and country-level variables. Data on individual-level variables are drawn from 35 surveys carried out in 12 Arab countries, Turkey and Iran. Most of the surveys were carried out either as the first wave of the Arab Barometer, the third, fourth and fifth waves of the World Values Survey, or a project on attitudes related to governance carried out by Mark Tessler with funding from the National Science Foundation.

Data File
Cases: 54,894
Variables: 268
Weight Variable: None
Data Collection
Date Collected: 1988-2010
Funded By
Islamic Scholars Program of the Carnegie Corporation of New York
Collection Procedures
Data on individual-level variables are drawn from 35 surveys carried out in 12 Arab countries, Turkey and Iran. Taken together, a total of 56,178 men and women were surveyed. Almost all surveys involved face-to-face interviews.

Most of the surveys were carried out either as the first wave of the Arab Barometer, the third, fourth and fifth waves of the World Values Survey, or a project on attitudes related to governance carried out by Mark Tessler with funding from the National Science Foundation. The dataset also includes four earlier surveys. Methodological information is provided below. Not all of the questions asked in each survey are included in the Carnegie Dataset that has been constructed by merging the 35 individual data files. Questions pertaining to governance and political life, to Islam and its political role, and to a number of policy issues have been included, however. All of the surveys contain a large number of relevant questions, and the Carnegie Data Set thus includes almost 200 individual level variables pertaining to politically relevant attitudes, values and behavior. There are also many individual-level variables pertaining to the personal attributes of respondents, such as age, sex and educational level.

The Carnegie Dataset also includes variables based on 34 time-specific country-level characteristics. Examples include Freedom House political freedom ratings; labor force attributes compiled by the World Bank, the United Nations and others; natural resource rents, also complied by the World Bank; ratings on the United Nations Human Development Index; freedom of religion indices developed by the U.S. State Department; and linguistic, ethnic and religious fractionalization indices compiled in a private research project. In most cases, country-level data pertain to the year in which the survey in the country was conducted. In a few cases, available data did not make this possible, and these cases the data are from the year before or the year after, or in rare instances two years before or after, the survey. The data set also includes variables based on five- and/or 10-year lags of many of these characteristics.
Principal Investigators
Mark Tessler, University of Michigan
Related Publications
Tessler, Mark. 2011. Public Opinion in the Middle East: Survey Research and the Political Orientations of Ordinary Citizens. Bloomington, IN: Indiana University Press.
Question wording
With few exceptions, question wording is the same in the surveys that are part of a common project, such as the Arab Barometer or the World Values Survey. In many instances the wording is also the same, or almost the same, in surveys carried out as parts of different projects. In other instances, however, questions with common content are worded differently in the different sets of surveys. In these cases, the variable label in the data set is based on the wording most frequently used. Recoding was occasionally necessary in order to merge the different surveys and create a single data file.
Survey Methodology Information
World Values Survey

Please see the World Values Survey website at:

Arab Barometer: First Wave

Please see the Arab Barometer website. Survey methodology for Jordan 2008 is identical to Jordan 2006 and for Palestine 2008 is identical to Palestine 2006. Survey methodology for Bahrain 2009 is described below.

Egypt and Kuwait 1988

Public opinion surveys carried out in Egypt and Kuwait in mid-1988 deal with issues of religion and politics and may accordingly shed additional light on the origins of popular support for Islamist movements. Based on stratified samples of 295 Egyptian adults in Cairo and 300 adult Kuwaiti citizens in Kuwait City, respondents reflect the heterogeneous nature of the general population. Each sample includes both men and women, and each is heterogeneous with respect to age, education, socioeconomic status, and neighborhood. Although better-educated individuals are somewhat overrepresented, the samples are generally representative of the active, adult, urban population.

Egyptian Christians and Kuwaiti Shi‘I were excluded from the samples for analytical purposes to
facilitate comparison of Sunni Muslim populations in other Arab countries. The same survey instrument was employed in each country.

The surveys were carried out under the direction of Professor Jamal al-Suwaidi of the United Arab Emirates University and the Emirates Center for Strategic Studies and Research. Interviews were conducted by teams of research assistants, or “intermediaries,” who were selected on the basis of previous experience in survey research administration. Intermediaries were also given a four-day orientation, and the survey instrument was pre-tested in both countries.

Palestine 1995

The Palestinian survey was conducted in August 1995 by the Center for Palestine Research and Studies (CPRS) in Nablus under the supervision of its director, Khalil Shikaki, and the head of its polling unit, Nadir Said. Multistage area probability sampling techniques were employed to select respondents, and the interview schedule was administered to 1,184 adults residing in the West Bank and Gaza.

CPRS has since become the Ramallah-based Palestine Center for Policy and Survey Research (PCPSR).

Palestine 1999

The Jerusalem Media and Communications Centre (JMCC) carried out this survey in May 1999. It is based on a random sample of 1,200 Palestinian adults over age 18. Interviews were conducted face-to-face throughout the West Bank and the Gaza Strip. The interviews were conducted in randomly selected homes, and the subjects inside each home were also selected randomly according to Kish tables. The interviews were conducted in 60 sampling points chosen randomly according to population. For additional details on JMCC’s sampling process, please visit their website at

Jordan and Palestine 2003; Algeria 2004

The surveys in Jordan and the Palestinian territories were funded by a grant from the National Science Foundation and the survey in Algeria was funded by the American Institute for Maghrib Studies. All are based on representative national samples. The three surveys were conducted, respectively, by the Center for Strategic Studies at the University of Jordan, the Palestinian Center for Policy and Survey Research in Ramallah, and a team at the University of Algiers. All three surveys involved three-stage cluster sampling based on the most recent national census; 1994 in Jordan, 1997 in the West Bank and Gaza, and 1998 in Algeria. In Jordan and the Palestinian Territories, districts or clusters were first randomly selected from the list of all clusters in the national census, followed by a random selection of households within each of the selected clusters, and a Kish table was finally used to choose the person in each household to be interviewed. In Algeria, governorates and then communes were selected randomly, and quota sampling based on age, sex, and education was employed at the commune level. Census data were used to establish the quotas. The same interview schedule was administered in Jordan, the Palestinian territories, and Algeria, although in Algeria it was supplemented by a number of items dealing with domestic policy issues in that country.

Kuwait 2005

A sample of 750 was selected on the basis of a group of 10,000 Kuwaiti citizens chosen randomly from the register of the Civil Information Authority, which issues civil IDs and maintains information on all who live in Kuwait. Attempts were made to contact these individuals, but this proved successful for only 20% of respondents, thus falling short of the target of 750. The rest of the interviews were conducted at or around the main supermarkets of the cooperative societies. Cooperative societies in Kuwait control over 80% of the retail sales all over Kuwait and all coops are centrally located in each Kuwaiti residential area. Each area was chosen area according to its respective percentage in the actual population. Each researcher was supplied with a letter requesting the cooperation of the coops. The reaction was generally positive and researchers collected a large number of interviews. In some cases the administration of the coop was helpful and provided surveyers with an office in which to conduct the interviews, although some coops were not as helpful. In these cases, a large number of interviews were conducted at the home of the respondent and some were done in a place near by store. The higher number of respondents recruited by this method led to the removal of incomplete interview schedules from the sample while recording 750 complete surveys.

Morocco 2005

The survey was conducted by a team from the Hassan II University, Mohammedia. The sample was drawn using area probability sampling with quotas provided by the National Bureau of Statistics in Morocco. They selected 100 zones (60 urban and 40 rural) and took quotas for type of living situation, gender, age, whether married or not, socio-economic level and level of education. Interviews were conducted face-to-face.

Yemen 2006

In Yemen, the Sanaa-based Yemen Polling Center, the country’s first licensed independent polling agency, drew the sample and administered the interviews, with a sample size of 1,440. Given the greater difficulties of acquiring a reasonable sample frame, a combination of area and cluster sampling techniques was employed, drawing a stratified sample from half of the country’s provinces.

The provinces sampled were Aden, Amran, Dhamar, Hadramawt, Hajja, Hudayda, Ibb, Marib, Sanaa (the capital), and Taizz. Selection of particular provinces into the sample was semirandom, with their probability of inclusion based on their weighted population shares, subject to the constraint that both the capital (Sanaa) and at least two provinces from the former southern republic (in the sample, Aden and Hadramawt) be represented. Clustering and stratification were based on preliminary figures from the 2004 census down to the village or city neighborhood level, with individuals sampled via random walk patterns.

Jordan and Palestine 2008

The methodology is the same as that employed in the Jordan and Palestine 2006 surveys. The interview schedules for these surveys contained a subset of the items on the full Arab Barometer survey instrument.

Bahrain 2009

The Bahrain survey, carried out by Justin Gengler in collaboration with the Arab Barometer, utilized a random sample of 500 Bahraini households provided by the Bahrain Center for Studies and Research. The sample was prepared by the Central Informatics Organization, which administers the nation’s census and maintains this and other electronic population databases. As a rule, Shi‘I fieldworkers conducted interviews in Shi‘a-dominated areas and Sunni fieldworkers conducted interviews in Sunni areas, although in urban centers some cross-sectarian interviewing was inevitable. The survey instrument was the Arab Barometer interview schedule, to which a small number of items were added.

Qatar 2010

The sample for the Qatar World Values Survey is designed to ensure the reliability and representativeness of statistical results derived from the data. The sampling frame for the survey comes from the Electricity and Water Company (Kahramaa). Since Kahramaa is the sole provider of water and electricity service in Qatar, this frame covers almost all Qatari households.

The State of Qatar is divided into six administrative municipalities, each further divided into many zones. This information is used to carry out proportionate stratified sampling, which ensures representativeness in the sample and increases the accuracy of statistical estimates. Stratification does not imply any departure from probability sampling; rather, it simply requires that the population be divided into subpopulations, or strata, and that probability sampling be conducted independently within each stratum. In this survey, we rely on zones for stratification. Within each zone (stratum), a respondent is randomly selected via two-stage sampling. In the first stage, households are randomly selected with proportionate stratification. That is, a stratum containing a given percentage of households in the population is represented by the same proportion of the total number of sampled households. In the second stage, an adult (18 years or older) within each household is randomly selected. At this stage, all adults in the household have the same chance of being selected. A total of 1,455 households were sampled and 1,060 interviews completed.

Bookmark and Share