Faith on the Move - The Religious Affiliation of International Migrants
Appendix B: Methodology and the Construction of the Global Religion and Migration Database (GRMD)
The religion of international migrants has been investigated in many parts of the world (for research reviews see Cadge and Ecklund 2007, Ebaugh 2003, and Koenig 2005).21 But most studies have focused on particular religious groups in specific destination countries, such as Catholic immigrants in the U.S. or Muslim immigrants in the United Kingdom. And, although some studies have looked at specific immigrant religious groups across several countries (examples include Fetzer and Soper 2005, Foner and Alba 2008, Mooney 2009), no previous research has attempted to provide a baseline set of estimates of ALL migrant groups by origin, destination and religion – an essential step for comparing migrant religious groups around the world.
Fortunately, some previous research projects have estimated the origins and destinations of international migrants (see Parsons et al 2007) or used origin-and-destination information to estimate the movement of migrants by gender (see Ozden et al. 2011) and education level (see Docquier and Marfouk 2006). Origin-and-destination grids also have been used to study the monetary remittances that many migrants send to their home countries (see Ratha and Shaw 2007). So, in short, there is some recent experience and collective know-how among researchers in compiling and harmonizing global data on migrants’ origins and destinations.
The Global Religion and Migration Database (GRMD) adds a layer of complication by including data for migrants to and from every country by religious group. But Pew Forum researchers were able to consult with migration experts who have constructed similar migrant databases. Although the Global Religion and Migration Database is new in many respects, the method for constructing it is similar to previous studies that have attempted to estimate other characteristics of international migrants, such as gender and education.
The GRMD does not attempt to measure degrees of religiosity among migrants. Scholars including Peggy Levitt (2007) and Jacqueline Hagan (2008) have argued that just as migrants’ other circumstances (for example, employment and family composition) may change as a result of moving across international borders, their religious beliefs and practices also may change. So, although a Muslim emigrant from Morocco may still self-identify as a Muslim in France, his or her religious beliefs and practices may be much different in France than they were in Morocco. The Global Religion and Migration Database does not seek to measure these qualitative changes, which are perhaps best captured by surveys and ethnographic approaches. Instead, the aim of the GRMD is to provide a basic demographic picture of the religious affiliation of international migrants.
Since there has been no single, worldwide census or survey of international migrants and their religious affiliation, certain assumptions inevitably have to be made in order to estimate the size of each religious group by origin and destination countries. Due to data limitations, some of these assumptions are less than ideal. Pew Forum researchers have attempted to assess the reliability of each data point in the database, as well as to conduct a series of tests to determine the robustness of the key assumptions. For the great majority of data points, the GRMD passes these tests. The assumptions and various robustness tests are explained in this methodology. A discussion of data limitations and potential resulting biases can be found here. Data sources for each destination country are available in Appendix C (PDF).
Overview of Procedures
In constructing the Global Religion and Migration Database, the Pew Forum first sought all census and survey data available on the origins of immigrants living in each destination country. Next, Pew Forum staff looked for all available information on the religious makeup of these immigrant groups, estimating the religious breakdown of migrants from each origin country to each destination country. Once all these data were collected, they were merged to create the Global Religion and Migration Database, which generates a count of the number of people in each of the world’s seven major religious groups (including the unaffiliated) who have moved from every origin country to every destination country.
Migrant Origins and Destinations
The Global Religion and Migration Database is based primarily on data from destination countries. The reason is simple: Many countries collect data on where their new residents come from, but relatively few countries keep records on where their former residents have gone. With these data in hand, three steps were taken in harmonizing all the data for every destination country.
First, destination information on newcomers often is incomplete. Many destination countries group immigrants into catch-all categories – reporting, for example, on migrants from the “rest of Africa” and other broadly defined areas. In addition, some countries do not release information on their residents’ countries of birth. Pew Forum researchers borrowed missing values from an earlier research project on migrant origins and destinations conducted by the University of Sussex’s (U.K.) Development Research Centre using data from the 2000 round of censuses. (See Detailed View of Procedures for a description of this project.)
Second, destination countries define who is and is not an immigrant differently. Some countries count foreigners by their nationality or citizenship rather than by their country of birth, while other countries define migrants by ethnicity regardless of where they were born. Additionally, the age of some of the data presented difficulties. Most of the information from censuses and surveys in the Global Religion and Migration Database dates from between 2000 and 2010, but there are countries whose data are older. To help standardize different types of data, as well as to update data from different years, the Pew Forum harmonized the various sources using the United Nation’s 2010 total immigrant population estimates for each destination country.
Third, some countries do not count refugees as immigrants. To ensure that refugees were not overlooked or undercounted, data for each country were compared with estimates from the United Nations High Commissioner for Refugees (UNHCR) and the United Nations Relief and Works Agency for Palestinian Refugees (UNRWA). If the UNHCR/UNRWA estimate was higher than the previously calculated estimate, the UNHCR/UNRWA estimate was used in place of the destination country’s estimate.
Religious Breakdown of Migrants
The first step in trying to determine the religious makeup of migrants was to decide which religious groups would be included in the database. Should there be categories for smaller religious groups? What about subdivisions within each major religious tradition? To a considerable extent, the options were limited by the religious categories in the data sources. For example, although censuses and surveys in many countries divide Christian immigrants into subgroups such as Catholics, Protestants and Orthodox, there are many countries for which data are available only on Christians as a whole. Similarly, the data sources do not consistently make distinctions within other major faith traditions, such as between Sunni and Shia Muslims or among various schools of Buddhism. And although the Pew Forum sought to collect migration statistics on several additional religious groups (such as Sikhs, Jains and traditional Chinese religions), this proved impossible because censuses and surveys in many countries do not provide separate counts of these groups. Based on such considerations, Pew Forum researchers chose to divide international migrants into seven major religious categories: Christian, Muslim, Hindu, Buddhist, Jewish, all other religions and unaffiliated (which includes atheists, agnostics and those who have no particular religion).23
To determine the religious breakdown of migrants, the Pew Forum used two techniques – original data and data proxies. First, Pew Forum staff sought information on the religious affiliation of immigrants in each destination country. The best data come from religion questions in censuses or surveys of immigrants that also include information about the immigrants’ country of birth. By cross-tabulating this information, it is possible to see what percentage of immigrants who have moved to Country A and were born in Country B identify themselves as belonging to each major religious group. Very often, the religious makeup of this immigrant population is different from the religious makeup of the general population in the origin country. For example, survey data from the New Immigrant Survey show that the percentage of Christians among immigrants to the United States from Egypt (62%) is higher than the percentage of Christians in the Egyptian population as a whole (<10%).24 Similarly, the percentage of Indian migrants to the U.S. who are Christian (9%) is higher than the percentage of Christians in India’s overall population (<5%).25
When detailed survey or census data on immigrants and religion were unavailable, the Global Religion and Migration Database used proxy measures. In some cases, Pew Forum researchers assumed that the religious makeup of migrants going to a certain country is similar to the religious distribution of migrants going to a religiously similar country for which census or survey data do exist. An example of this type of “destination proxy” is found in the migration estimates for the Gulf Cooperation Council countries. Bahrain does not release detailed data on the religious affiliation of its foreign-born population. But Egypt – which, like Bahrain, has a Muslim majority – does have census data on its immigrants. In this case, the religious distribution of migrants from some origin countries (for example, India and the Philippines) to Bahrain was assigned the religious distribution of migrants from the same origin countries to Egypt.
Finally, in some cases, immigrants to Country A who were born in Country B were assigned the religious makeup of the general population in Country B. Although this assumption is less than ideal, it is the best alternative when other reliable data are lacking. Fortunately, tests for these “origin proxies” indicate few problems in estimating the religious affiliation of international migrants.26
Phase One: Country Origins
The first phase in assembling the Global Religion and Migration Database involved the construction of a two-way table of 231 origin and destination countries and territories for which the United Nations Population Division provides general population estimates.27 Each data point represents a different origin-destination combination. This process of data collection and data harmonization is similar to methods used by researchers both at Sussex University’s Development Research Centre on Migration, Globalisation & Poverty and the World Bank in constructing previous origin-destination grids for 2000 and 2005.28
During 2010, Pew Forum staff sought census and survey data from destination countries, downloaded information from statistical databases and made numerous email and telephone queries to statistical agencies around the world in an effort to assemble the most complete picture of the world’s migrant (foreign-born) population. After examining the various data sources available, Pew Forum researchers selected the most recent data that best fit the Global Religion and Migration Database’s definition of an international migrant: a person who currently resides in a country other than his or her country of birth and has lived in that new country for one year or longer. For most destination countries, the best data came from the most recent census figures for the foreign-born population. But in some cases, data came from population registers and large-scale surveys, mostly gathered through secondary sources (e.g. the World Bank, United Nations and Eurostat). For a full list of origin data by destination country, see Appendix C (PDF).
Fortunately, most migrant-origin figures (representing 86% of the international migrant population) are based on actual census or other population data. However, since a small percentage of data were still missing, a second step in determining migrant origins involved imputing missing data. The table for Step 1 provides an example of the database in its first phase of construction. The rows (Bulgaria and the United States) are destinations, while the countries in the columns (Afghanistan, etc.) are origins. As the table for Step 1 shows, data for many smaller-sized origin countries are missing. In Bulgaria, for example, Albania is the only origin country listed with data. In the United States, Andorra as an origin country is also missing. This does not mean that no one from these missing countries lives in either Bulgaria or the United States; rather, it suggests that migrants may have been labeled in a category other than their specific country of origin. For example, many destination countries group smaller origin countries into categories like “rest of Africa” or “Europe, not included elsewhere.”
In dealing with these missing data, Sussex University’s bilateral migration grid proved invaluable.29 The Sussex Global Migrant Origin Database, which took a number of years to construct, imputed missing data by assuming that the origin distribution for missing data cells is similar to the origin distribution of destination countries in the same geographic region.30 It is important to keep in mind that in most cases, these imputations are for missing cells within destination countries where data exist for the majority of the immigrant population. However, there are several destination countries (Afghanistan, Algeria, China, Eritrea, Lebanon, Maldives, Morocco, North Korea and Somalia) where all immigrant origins are imputed. Nonetheless, only 14% of GRMD’s population is based on imputations of missing data, including destination countries where all immigrant origin data are drawn from Sussex’s imputations.31
The red font numbers in the table for Step 2 represent missing origin countries that received an immigrant count based on a remainder or leftover country category provided in the original data. As the table for Step 2 indicates, most imputed values are quite small numbers. In many respects, then, they are mere placeholders that allow for immigrants to be assigned somewhere. Yet, due to their small population sizes, these imputations do not result in substantial differences in global and national totals.
The third major step in determining the final origin numbers involved standardizing them with the United Nations Population Division’s estimates for the number of immigrants in each destination country. Based on various data sources and projection techniques, the U.N. estimates the total immigrant population of every country in the world. These U.N. estimates were used to harmonize the various types of origin data (e.g., census and population register figures) as well as to smooth out problems caused by the fact that data from different sources were collected in different years. The origin data were redistributed to equal the U.N. immigrant total for each destination country.
In most instances, this meant scaling upward or inflating the numbers, since most countries’ immigrant populations are rising. For both the Bulgaria and U.S. examples, the U.N. total immigrant estimates were higher than the original data – mostly as a consequence of the original data being collected a few years earlier than 2010.32
For the fourth and final step in putting together a complete picture of migrant origins, the Pew Forum consulted refugee data to be sure the origin estimates were in line with this important slice of the migrant population. For a variety of political and economic reasons, many destination countries do not include refugees in their foreign-born estimates; however, refugees are included in the total immigrant count provided by the U.N. Previous migration databases, such as those constructed by researchers at the World Bank, largely removed refugees from U.N. totals because the researchers were more concerned about “economic” migrants. However, since religion is often an important variable in understanding the complexities surrounding the movement of refugees across international boundaries, it did not seem appropriate to exclude refugees from the Global Religion and Migration Database.
Making adjustments to include refugees involved a lengthy comparison process for each destination country. Taking the 2009 United Nations High Commissioner for Refugees’ (UNHCR) origin-destination grid for the world’s current refugee population as well as United Nations Relief and Works Agency (UNRWA) data for Palestinian refugees, each destination country was studied in detail to determine if the estimated refugee in-bound population exceeded the Pew Forum’s earlier calculated estimates. When the UNCHR/UNRWA estimate for refugees living in a specific country was more than 1,000 individuals and exceeded the Pew Forum’s previous estimates, the UNHCR/UNRWA estimate replaced the previous origin country’s estimate.
Most of the adjustments to refugee estimates involved figures for less-developed regions of the world. For example, no changes were made to the United States, since the U.S. Census Bureau records refugees just like other foreign-born persons. Adjustments to refugee figures were most frequent for three origins: Afghanistan, Iraq and the Palestinian territories. In fact, about 8.5 million migrants, or more than 60% of a total of 13.6 million refugees counted in UNHCR and UNRWA data, were from these three origins; most of these refugees have gone to Iraq, Iran, Pakistan, Jordan, Syria, Egypt and Lebanon. For the rest of the less-developed world, refugee adjustments were minor.
Continuing with Bulgaria as an example, since the UNHCR indicated there to be more than 2,000 refugees from Afghanistan living in Bulgaria (see table for Step 4), this number replaced the previous estimate of 252 Afghanis (see table for Step 3). When a change of this nature was made, the sum total of refugee replacements were subtracted from the total Bulgaria migrant count. Following this, the remaining origin countries were redistributed to the new total for Bulgaria, not including refugees. All specific changes by destination country are listed in detail in Appendix C (PDF).33
Phase Two: Religious Distribution
The second phase of constructing the Global Religion and Migration Database involved estimating the religious distribution of migrants by their origin countries. Before immigrants could be assigned a religion, however, a consistent list of religious groups across the world was needed. After analyzing the various data sources, the following seven religious affiliation categories were deemed as the most manageable while at the same time offering a sufficient level of detail for this stage of the project: Christian, Muslim, Jewish, Buddhist, Hindu, other religions and unaffiliated.34 The unaffiliated category refers to migrants who are atheist, agnostic, humanist or claim no particular religion.35 The “other religion” category includes African traditional religions; Chinese traditional religions, Shintoism and other Asian religions; Sikhism, Jainism, Zoroastrianism and Baha’i, among other faiths.36
With these religious categories in hand, the Pew Forum sought out the best data describing the religious breakdown of immigrant populations in each destination country. Religious affiliation is asked in some country censuses, and this can be broken down by country of birth. In all, about 15% of the GRMD’s population uses such census data to estimate the religious distribution of international migrants.
The second best data source for estimating the religious distribution of migrants is general population and immigrant surveys. Since religion is not asked in the U.S. census, and about one-fifth of international migrants reside in the United States, survey data became crucially important in estimating the size of immigrant religious groups. Drawing from the New Immigrant Survey and a variety of Pew Forum surveys, the religious distribution of most origin groups in the United States was estimated.37 Additional surveys for other destination countries besides the U.S. were also used. In total, the religious distribution of 18% of the international migrant population was calculated on this basis.
Following an imputation procedure similar to the one used for missing origin data in phase one, the religious distribution of about 7% of the GRMD’s population was based on origin-destination pairs for which reliable census data were available. These “destination proxies” were used for some of the largest migrant population groups. For example, it would be expected that migrants from Country X to the Gulf Cooperation Council countries would be proportionally more Muslim than the general population of Country X. This may be especially true for origin countries such as India and the Philippines that have substantial Muslim minority populations. Therefore, the religious distribution of migrants to Gulf Cooperation Council countries from India and the Philippines was based on the known religious distribution of migrants from these same countries to a similar destination country, in this case Egypt. Another example of a destination proxy involves imputing the religious distribution of Pakistani migrants to India (a Hindu-majority country) based on the known religious distribution of Pakistanis to Nepal (also a Hindu-majority country, but where religion census data for immigrant populations are available). All destination proxies are listed in Appendix C (PDF).38
More than 40% of the religious distribution of international migrants was estimated using census/survey data or a reasonable destination proxy that was also based on census data. But for more than half of the international migrant population, no such religious affiliation data by country of origin or reliable origin-destination replacements exist; therefore, an “origin-proxy” method was used. The origin proxy method assumes, for lack of better information, that the religious profile of a country’s emigrants is the same as the religious profile of that country’s population as a whole. Relying on the 2010 religious affiliation estimates in the World Religion Database (WRD) by origin country, the origin-proxy method was used when no better data were available. As illustrated in the table for Step 3, using the origin-proxy method fills in all remaining cells for the religious distribution of international migrants.
Merging Origin and Religious Distribution Data
With the completion of the origin-destination grid (phase one) and the religious distribution of international migrants (phase two), a simple multiplication procedure was performed to obtain a count for each religious group by each country of birth within each destination country.
The table for the final phase displays the final counts of migrants from Afghanistan, Albania, Algeria, American Samoa and Andorra to both Bulgaria and the United States. This is only a very small portion of a very large dataset, as the complete database contains nearly half a million records.39 Given that there is a value for every cell in the database, the rows and columns of the dataset can be reversed to become an origin-to-destination database. In this way, both emigration and immigration can be examined.
In constructing any new cross-national database, a number of limitations exist, mostly due to the absence of reliable data. First, migrant populations are often undercounted even in census data. Even though scaling to the U.N. total migrant counts helps alleviate this concern, potential census under-counting of some immigrant groups may still result in groups that are routinely underestimated. It is also more difficult to determine whether a religious under- or over-counting of migrant populations within a destination country occurs when surveys (and some censuses) are voluntary or may be completed by an individual other than the respondent.
Second, it is also difficult to assess whether migrants within a given destination country arrived with their currently stated religious affiliations or changed religious affiliations once settled in the destination country.40 Although no nationally representative data finds mass religious conversion among immigrants across broad religious categories (e.g., from Christianity to Islam or from Hinduism to Christianity), some studies have found that sizable numbers of immigrants who had no religious affiliation in their home country eventually adopt some kind of religious affiliation after living in the United States (see Chen 2008 and Skirbekk et al. forthcoming). The aim of the Global Religion and Migration Database is to estimate the current religious affiliation of international migrants in 2010, including those who may have changed religious affiliations since migration. Estimates relying on census and survey data account for religious change, but other estimates relying on the origin-proxy method do not.
The third and perhaps most important area of potential bias is the origin-proxy method itself – the assumption that the religious composition of emigrants is the same as the religious composition of the general population in their country of origin. For example, recent media reports have described the disproportionately large number of Christians who have exited some Middle Eastern countries.41 This type of religious migration would be undercounted by the origin-proxy method, since the great majority of residents in most of these countries are Muslim, not Christian. Substantial scholarship has pointed to the importance of age, gender, education and other variables in the self-selection of emigrants, and there is also strong evidence of selective migration on the basis of religion. For example, the WRD estimates India’s overall population to be 73% Hindu, while census data suggest that only 26% of Indian immigrants in Canada are Hindus, and only 38% of Indian immigrants in the United Kingdom are Hindus. Without question, in some countries, certain religious groups are more likely to leave than others, and they are also more likely to choose certain destination countries over others.
However, it is important to note that of the nearly 130 million migrants whose religious affiliation is estimated using the origin-proxy method, nearly 75 million (about 60%) have moved within geographic regions where the majority religion for the origin and destination countries is the same. For example, more than 30 million migrants whose estimated affiliation relies on the origin-proxy method have moved within Christian-majority countries in Europe. Religious selection would not be expected to be a major factor within these migration corridors. Another migrant corridor where religious selection does not appear to be a major issue is from majority-Muslim countries to continental Europe (about 15 million migrants from such countries as Turkey, Morocco and Pakistan, or nearly 12% of all migrant data relying on the origin-proxy method). These migrants to Europe are from countries in which the population is almost entirely Muslim, ruling out the possibility of large numbers of non-Muslim emigrants. In all, it is expected that religious selection is not occurring for more than two-thirds of migrants where affiliation is estimated using the origin-proxy method.42
To assess the data quality of its estimates, the Pew Forum performed two further comparisons. The first comparison looked at the GRMD’s origin estimates and compared them with other published estimates of the size of immigrant communities. These estimates are mostly found online and usually focus on a particular ethnic diaspora in a particular country. Generally, GRMD foreign-born estimates by destination country should be lower than these estimated diaspora populations, since the latter estimates tend not only to include immigrants but also to include their native-born offspring. The second comparison involved lining up GRMD estimates with the 2010 World Religion Database. As with the diaspora estimates, all WRD estimates for total populations by religious group should be higher than the GRMD migrant estimates because WRD estimates include not only migrants but also all people of a particular religious group in a particular country. Thus, whenever GRMD estimates were higher than diaspora or WRD estimates, Pew Forum researchers re-examined the data to see whether mistakes may have been made.
Finally, Pew Forum staff coded the quality of all data in the GRMD.43 Data priorities were determined for each of the two phases (migrant origins and religious distribution) in constructing the GRMD. For the migrant origins phase, data from original sources (e.g., census, population register, UNCHR and other survey estimates) were considered to be of the highest quality (86% of the migrant population). If census or similar data were unavailable, imputations based on the population share of the source country or the Sussex Global Migrant Origin Database were used to fill remaining gaps in the data (14% of the migrant population).
For the religious distribution phase, census data were considered to be the best, followed by survey data (both immigrant surveys and immigrant sub-samples of general social surveys). This type of data is considered the most reliable and represents 33% of the world’s total migrant population.
When no other data for the religious distribution of immigrants were available, the destination-proxy was used, accounting for 7% of the international migrant population. Origin-proxies where tests indicate a high level of confidence in the data were used for 35% of the international migrant population. These origin-proxies mostly represent migrants moving within or between regions where the majority religion is the same (for example, Christian migrants moving within the Americas) or where migrants originate from a country whose population is composed almost entirely of one religious group (for example, Muslim migrants from Turkey to Europe). The second part (25% of migrants) also uses an origin-proxy but represents migration between countries in which there may be more selection on the basis of religion.
In all, the estimated religious affiliation of more than three-fourths of the migrant population relies on high quality data. But because countries measure their immigrant populations in various ways, and solid information on the religious affiliation of migrants is not always available, the Pew Forum advises readers of this report and accompanying interactive graphics to treat all the figures as estimates and to take the sources of information and methods of estimation into account. (See Appendix C (PDF) for a complete list of data sources and adjustments for each destination country.)
22 This is the United Nations Population Division’s general definition of an international migrant. In addition, the U.N. considers refugees and, in some cases, their descendants (such as Palestinians born in refugee camps) to be international migrants. For the U.N.’s detailed definition of an international migrant, see United Nations, Department of Economic and Social Affairs, Population Division, “International Migration Report 2006: A Global Assessment (PDF),” page 337, 2009. For further discussion on defining international migrant stock, see United Nations, Department of Economic and Social Affairs, Statistics Division, “Recommendations on Statistics of International Migration: Revision 1 (PDF),” Statistical Paper Series M No. 58, pages 84-91, 1998. (return to text)
23 The final database does contain some estimates for Christian subgroups (Catholic, Protestant and Orthodox), but the figures for these Christian subgroups are considered less reliable and thus not presented in the Faith on the Move report. Users of GRMD data are advised to be cautious in citing subgroup differences within Christian estimates. (return to text)
24 The New Immigrant Survey (NIS) is a nationally representative study of new legal immigrants to the United States and their children. The first full wave of the NIS was conducted in 2003 and 2004, involving nearly 10,000 respondents. Interviews were conducted face-to-face and by telephone in the respondent’s preferred language. The NIS was designed by Guillermina Jasso, Douglas S. Massey, Mark R. Rosenzweig and James P. Smith and funded by the National Institutes of Health, National Science Foundation and the U.S. Citizenship and Immigration Services. Additional support was provided by the U.S. Department of Health and Human Services, Office of the Assistant Secretary for Planning and Evaluation, and The Pew Charitable Trusts. The Pew Forum had access to restricted-use data, which was retrieved in August 2007. For further information, see The New Immigrant Survey. (return to text)
25 The Pew Forum’s December 2011 report Global Christianity: A Report on the Size and Distribution of the World’s Christian Population estimates that Christians make up 5.3% of Egypt’s population and 2.6% of India’s population as of 2010. (return to text)
26 For the religious composition of the general population in countries of origin, the GRMD primarily relies on the 2010 World Religion Database. The Pew Forum is preparing its own estimates for the religious composition of each country in the world. Until this data is available, the 2010 World Religion Database is used. Besides census and survey reports, WRD estimates take into account other sources of information on religious affiliation, including anthropological and ethnographic studies as well as statistical reports from religious groups. The WRD is an outgrowth of the international religious demography project at Boston University’s Institute on Culture, Religion, and World Affairs. (return to text)
27 Previous Pew Forum reports have covered 232 countries – the United Nations’ list of 230 countries and territories plus Kosovo and Taiwan. All these countries except for Pitcairn Islands are included in the GRMD, resulting in 231 countries and territories. Since South Sudan became independent in 2011, it is not included for 2010 migration estimates. No migration data for Kosovo as a destination country are available, but it is included as an origin country nonetheless. Migration data for Taiwan were drawn from Taiwan’s 2009 population register. (return to text)
28 For methodological descriptions of previous origin-destination grids, see the Sussex Global Migrant Origin Database methodological paper (Parsons et al. 2007), World Bank Global Bilateral Migration Database methodological paper (Ozden et al. 2011) and World Bank Bilateral Migration and Remittances methodological paper (Ratha and Shaw 2007). (return to text)
29 See parsons et al. 2007. (return to text)
30 Using Sussex imputations in the GRMD assumes that the origins of migrants within the same geographic region have not changed considerably since 2000. An updated version of the Sussex database, the World Bank’s Global Bilateral Migration Database (see Ozden et al. 2011), was released in late 2011; however, this data was unavailable when the Global Religion and Migration Database was constructed. (return to text)
31 Sussex University’s origin-destination grid has 226 countries, compared with 231 countries in the GRMD. Therefore, there are instances in which data for country groupings in the original data (e.g. former Yugoslavia, Hong Kong and Macau, Israeli-Palestinian territories) could not be subdivided according to Sussex University’s origin distribution because no data in Sussex University’s grid existed for these countries and territories. In these isolated cases, country groupings were redistributed according to the relative size of the origin country’s population in 2010. Additionally, origin estimates for a few destination countries (e.g., Channel Islands, Isle of Man, Kosovo, Vatican City and Western Sahara) were entirely based on scholarly publications. Again, this is because these countries were unavailable in the Sussex University origin-destination grid. All decisions involving the editing of origin data are noted in Appendix C (PDF). (return to text)
32 Most population totals from original data sources closely matched the U.N. totals. However, some destination country data were old (collected prior to 2000) while other countries defined immigrants differently than the U.N. Most of the population differences between source data and the U.N.’s 2010 estimates were resolved by adding in refugees (see Step 4) or were the result of increases in immigrants since data was originally collected. In all, these population differences by destination country represent only 10% of the globe’s migrant population. (return to text)
33 Because refugee replacements occurred in countries where origin data on immigrants are less reliable, it is difficult to know whether the refugee replacement is still underestimating the actual immigrant population for a particular origin country. For example, it is reasonable to assume that more immigrants than the UNHCR/UNRWA estimates of registered refugees are living in these destination countries, some of whom may have migrated principally for economic reasons. To avoid double counting, the GRMD takes a conservative approach and assumes that the UNHCR/UNRWA estimate is the minimum population size of immigrants living in the destination country. (return to text)
34 These categories are consistent with the religious affiliations analyzed in previous Pew Forum reports, though some previous reports have included further breakdowns, such as dividing Christians into Catholics, Protestants, Orthodox and Other Christian. See, for example, Global Christianity: A Report on the Size and Distribution of the World’s Christian Population (December 2011). (return to text)
35 Although some individuals may have multiple religious affiliations, few censuses or surveys around the world report data on overlapping affiliations (e.g., on Christians who also consider themselves to be Jews). For this reason, the seven categories are treated as mutually exclusive. Because of the inclusion of the “other religion” and “unaffiliated” categories, the seven categories are also treated as exhaustive. Many censuses and other data sources include a missing or “not stated” category for religious affiliation. This category was removed from the denominator, which assumes that all religious groups are equally likely to refuse to answer religious affiliation questions. Given the variety of data sources, this was the only consistent way of dealing with non-response in censuses and surveys. (return to text)
36 While the “other religion” category includes a disparate mix of religious groups, knowing the general religious makeup of an origin country does permit users of the GRMD in many cases to speculate about which religious groups are most likely to be present within this category. For example, among migrants from India, people of “other religions” (besides Hinduism, Islam, Christianity and Judaism) are likely to include many Sikhs and Jains. (return to text)
37 The New Immigrant Survey (NIS) is a nationally representative study of new legal immigrants to the United States and their children. The first full wave of the NIS was conducted in 2003 and 2004 and involved nearly 10,000 respondents. It could be argued that the religious distribution of new immigrants is not reflective of the total immigrant population from a given country within the United States – both the immigrants and the immigrant flows have changed over several years. Fortunately, however, about half of the NIS sample represents immigrants adjusting their status to permanent residents, many of whom have lived in the United States for five years or more. Mostly using NIS data, the GRMD estimates that there were 2.1 million Muslim immigrants living in the U.S. in 2010. As a point of comparison, using the Pew Research Center’s 2011 survey of Muslim Americans, Pew Research demographers estimated that there were 2.0 million Muslim immigrants in the U.S. one year later, which is quite consistent with the 2010 GRMD estimate. (See “Muslim Americans: No Sign of Growth in Alienation or Support for Extremism (PDF),” Pew Research Center, 2011.) Therefore, it appears that using NIS data does not severely bias immigrant religion estimates, at least for Muslim immigrants in the U.S. (return to text)
38 As an example of a robustness check for the destination-proxy method, Bahrain’s 2010 census estimates 55% of immigrants to be non-Muslim. Using the destination-proxy method, the GRMD estimates 47% of immigrants in Bahrain are non-Muslim. (return to text)
39 The complete GRMD includes data on some Christian sub groups (Catholics, Protestants and Orthodox) which results in 231 origin countries x 230 destination countries x nine religious groups (Catholics, Protestants, Orthodox, Muslims, Jews, Buddhists, Hindus, adherents of other religions and the unaffiliated), totaling nearly half a million records. (return to text)
40 An analysis by the Pew Forum of the first wave of the New Immigrant Survey (2003) suggests that religious switching across major religious groups (for example, Hinduism to Christianity, Buddhism to Islam) is relatively rare among first-generation migrants. (return to text)
41 For example, the proportion of Christians among migrants leaving Iraq is believed to be higher than in the Iraqi population as a whole. See Jon Pedersen and Kristin Dalen, Iraqis in Jordan: Their Number and Characteristics (PDF), Fafo, 2007. (return to text)
42 A thorough assessment of religious selection was conducted by Pew Forum researchers for data points relying on the origin-proxy method. Migrant population using the origin-proxy method were subdivided into regions of origin and destination and then within majority religion groups by origin-destination countries. The origin country’s religious distribution for the largest origins of migrants by destination country was compared to similar destinations where reliable census data were available. For example, using the Canadian census, migrants from France to Canada are 73% Christian whereas France itself is 68% Christian (2010 WRD). Additionally, several census and survey estimates of Turkish, North African and Pakistani immigrants in Europe show little religious selection. (return to text)
43 For details, see Connor and Tucker, 2011. (return to text)