INTRODUCTION
The religion and state (RAS) project is intended to create a set of measures that systematically gauge the intersection between government and religion. Specifically, it develops and collects multiple variables that measure government religion policy. This endeavor is intended to accomplish three goals: (1) to provide an accurate description of government religion policies worldwide, (2) to create a tool that will lead to greater understanding of the factors that influence government religion policy, and (3) to provide the means to examine how government religion policy influences other political, social, and economic factors, as well as how those factors influence government religion policy.
The first round of the RAS project collected 62 variables on government involvement in religion for 175 countries.Footnote 1 These variables were coded yearly for 1990 to 2002. Round 2 will collect 147 variables for 176 statesFootnote 2 for 1990 to 2008. This article is intended to present the project, its methodology, and the variables included in the new version of RAS, the genesis of these variables, and how these variables compare with other religion databases.
THE RAS PROJECT
The RAS dataset is intended to provide variables measuring government involvement in religion (GIR). For the purposes of the project, this includes any government support for religion and any interference or restrictions by governments on any religion within the state's borders. Thus, RAS specifically measures state religion policy.
This is certainly only one of many aspects of religion's role in politics, economics, and society — otherwise known as the religious economy — that can potentially be measured. The RAS project chose to focus on this limited topic for a number of reasons. First, while limited to one section of the larger religious economy, the topic is sufficiently broad and complicated that it is a major undertaking to codify and measure.Footnote 3 Second, the country-level coding and the explicitly governmental focus make the RAS data compatible with a number of existing datasets on other topics including regime,Footnote 4 conflict,Footnote 5 and economics.Footnote 6 Thus, among the many potential topics for religion data collection, this topic is arguably among the broadest in potential applicability. Third, it is more straightforward to code and collect than many other potential topics, such as religious political parties and social movements in respect to both the universe of cases, and types of activities to be coded. Finally, it is a topic for which sufficient information is available to code the variables.
The coding is based on a wide variety of sources. For each country, a research assistant writes a report based on general sources such as the U.S. State Department Religious Freedom ReportsFootnote 7 and Barrett, Kurian and Johnson (Reference Barret, Kurian and Johnson2001), among others, as well as country-specific academic writings, and media sources from the Lexis/Nexis database.Footnote 8 While the amount of information available varies from state to state, there is clearly sufficient information for all states to meet the minimum threshold for coding the variables. In fact, the general sources and country-specific sources tend to provide high levels of mutual confirmation. That is, when general sources identify high levels of various types of GIR, this tends to be confirmed by the country-specific media, and academic sources that usually provide significant amounts of additional information beyond what is provided by the general sources. Conversely, when the general sources show low levels of GIR, country specific sources tend to also reveal low levels of GIR — often in the absence of any reports of GIR. This is what we would expect because media articles are rarely written to say, for example, that there is no religious discrimination. However, the presence of such discrimination tends to be reported, even if it takes place in remote regions. This is partially due to the fact that this is a topic of interest in the media. It is also partially due to the presence of numerous advocacy groups that seek to publicize the presence of religious discrimination. Furthermore, such groups tend to use a broader definition of religious discrimination than does the RAS project in constructing its variables, and include favoritism for some religions over others including through religious legislation.
These reports are then used as the basis for coding the RAS variables. They also provide the basis for more traditional comparative assessment of the topic, as well as a basis for clarifying the coding, and the sources upon which the coding are based in cases where the coding are questioned at a later date.
OTHER MEASURES OF RELIGION
As discussed in more detail (Fox Reference Fox2008a, 31–46), most social science measures of religion used in cross-national analyses fall into one of four categories. The most common are identity variables that measure the presence or differences between various religious traditions in a state.Footnote 9 These variables are relatively simple to collect, but have limited use in that they can only measure whether there is a difference in behavior between groups belonging to different religious traditions, but can not directly tie these differences to the influence of religious factors. Religious diversity variables — the second type of religion variable measure the diversity of the religious population of a state — have similar limitations.Footnote 10
Third, sociological survey-based variables measure the extent to which individuals are religious with questions such as do you believe in God? How often do you attend your place of worship? and do you consider yourself a religious person? While most such studies are not useful for cross-country analysis because they are taken only in a single state or a portion of that state, there exist international surveys such as the World Values Survey (WVS) that cover a large proportion of the world's states.Footnote 11 This individual-level data can be translated into country-level data by taking national averages.Footnote 12 However these variables are limited in that the WVS — the largest such program — only covers 76 states and not all of them in the same year. There may also be issues of selection bias in the states included and issues of compatibility between country surveys within the project (Fox Reference Fox2008a, 38–39).
The final approach to cross-sectional religion variables is the type used by the RAS project — variables that quantify some aspect of groups, governmental, or institutional behavior. A popular example of this type of variable is a measure of whether a state has an official religion. Versions of this type of variable were collected by Round 1 of RAS and Barrett, Kurian and Johnson (Reference Barret, Kurian and Johnson2001), among others. A more sophisticated version of this type of variable provides a list of related variables that can be combined into a single larger measure. Norris and Inglehart (Reference Norris and Inglehart2004), Chaves and Cann (Reference Chaves and Cann1992), Grim and Finke (Reference Grim and Finke2006; Reference Grim and Finke2007), Minkenberg (Reference Minkenberg2002), and Price (Reference Price2002) as well as Round 1 of the RAS project all collected this type of variable.
I posit that this type of variable has the most potential for productive cross-country studies of religion for at least two reasons. First, it can potentially be collected for all states or in cases where the unit of analysis is not the state all relevant cases. This avoids selection bias issues and allows the widest possible applicability of the findings. Second, these variables, if constructed correctly, can allow us to directly measure the presence of explicitly religious behavior and phenomena, and their impact on other social, political, and economic factors.
All of the previous collections of the more sophisticated version of this type of variable noted above, other than Grim and Finke (Reference Grim and Finke2006; Reference Grim and Finke2007) do not realize the first of these advantages in that they only include a portion of the world's states. Also, none of them include more than 20 related variables. As noted above, Round 1 of RAS includes 62 and Round 2 will include 147.
BUILDING BETTER VARIABLES
The philosophy of the RAS project is to collect information in as much detail and quantity as possible, within the existing practical limitations including the resources available to collect the data — primarily available man-hours — and the availability of information. Variables included in the dataset must meet two criteria. First, they must measure a distinct and clearly defined aspect of government policy, laws, or structure that relates to religion. Second, there must be sufficient information available to code this variable for all states included in the dataset.
The 62-item variable list for Round 1 was based on a list developed by the project's founder and comments on by multiple colleagues. Due to resource limitations, this list was fixed early in the data collection process for Round 1.Footnote 13 However, in the process of collecting data for Round 1 and analyzing that data, a number of additional factors that meet the two criteria listed above were identified. In addition, many factors that were included in the original set of variables were discovered to represent sufficiently diverse and complex phenomena that more detailed variables were deemed appropriate. This process, along with additional comments by colleagues, resulted in the 147 variables included in the codebook for Round 2.
This new set of variables covers the same ground as the set from Round 1 of RAS but adds additional detail in two ways. First, existing variables were expanded. In the case of variables with multiple possible coding, this involved adding more categories. For example, the official support variable from Round 1 — that measures the formal relationship between religion and state — did not differentiate between different types of official religions. Round 2 now includes five categories of official religions.Footnote 14 This is important because countries like Saudi Arabia, Greece, and the United Kingdom all have official religions, yet the relationship between religion and state in these countries is clearly very different. In countries with no official religion, research revealed many states with complex policies that include multi-tiered levels of recognition for religious organization. Round 2 official support variable added several categories to account for this.
In the case of religious discrimination, religious regulation, and religious legislation — all lists of multiple specific types of GIR that fit into these categories — this meant adding items to these lists. For example, Round 1 included a single variable for restrictions on access to religious publications in the religious discrimination category. Round 2's list of religious discrimination includes three variables that differentiate between (1) publishing religious publications, (2) importing them, and (3) the confiscation of personal religious publications from individuals. A brief description of the extent of changes of the variables between Round 1 and Round 2 is provided in Table 1. Each category of variable can be combined to make a composite scale.
Round 2 Variables Compared to Round 1 Variables
a In addition to being coded for each state in general, 29 of these items will be coded separately for each religious minority in the state which constitutes at leas 0.25% of the population.
Second, the research from Round 1 revealed that some types of GIR were important but too complex for the Round 1 variables to fully capture the diversity of policies from state to state. In these cases, more detailed variables were added to quantify these policy areas. These include the following:
• Three variables focusing on religious education in public schools including whether such education is mandatory, in what religions it is available, and who is responsible for teaching the religion classes.
• One variable measuring the presence of prayer in public schools.
• Eight variables measuring the presence, nature, and consequences of requirements for religious organizations to register.
• Nine categories of restrictions on abortion. These categories mostly focus on in what instances is abortion allowed or restricted.
• One variable measuring whether government officials are required to be members of the majority religion and whether oaths of office include mandatory references to religion or God.
• 12 variables measuring restrictions on proselytizing including the nature of the restrictions and to whom they apply.
• One variable measuring whether citizenship requirements are linked to religion.
In addition to the complexity and diversity of government policies on these issues, all of these categories of variables meet one or both of two criteria. First, they are present in a large number of states. For instance, based on the Round 1 data for 2002, 102 of 175 states had religious education in public schools, 118 states had policies of registering religious organizations, and 77 restricted proselytizing. Second, the issue is of sufficient gravity or sufficiently controversial that more detail is warranted. This includes the issues of abortion, linking citizenship to religion, and linking religion to the right to hold political office.Footnote 15
CONCLUSIONS
The potential significance and range of this project can be assessed through the results based on the analysis of Round 1 that include: (1) GIR increased between 1990 and 2002. This contradicts modernization and secularization theory that predict that religion is in decline (Fox Reference Fox2006, Reference Fox2008a).Footnote 16 (2) GIR influences human rights violations by states (Fox Reference Fox2007b, Reference Fox2008b). (3) Democracies tend to have more separation of religon and state than do non-democracies (Fox Reference Fox2007a, Reference Fox2008a; Fox and Rynhold Reference Fox and Rynhold2008). (4) Most states, including democracies, do not have separation of religon and state, no matter how the term is defined (Fox Reference Fox2007a, Reference Fox2008a; Fox and Rynhold Reference Fox and Rynhold2008).Footnote 17 (5) Different world regions and different religious traditions (Catholicism, Protestant Christianity, Islam, etc.) have different patterns of GIR and patterns of change in GIR over time (Fox Reference Fox2008a). (6) Findings support the supply-side theory of religion, a major sociological theory that is under considerable debate. This theory argues that for a number of reasons that state religious monopolies result in a less religious population with regard to religious practices but not to belief (Fox Reference Fox2008a; Fox and Tabory Reference Fox and Tabory2008).Footnote 18 (7) Religious freedom does not significantly impact the economic well-being of a country (Wong Reference Wong2006).
As noted above, the RAS data can potentially be used to find correlations with any state-level data. This applies both to the more global variables and to each of the 147 more specific variables that will allow research into more narrow issues. For instance, the detailed variables on religious education will allow for an examination of the impact of religious education on individual religiosity. This is an interesting question in and of itself, as well as in the context of the supply-side theory of religion, and it is only one example among many of potential uses of this data. The data is also of interest beyond political science and sociology researchers — the core audience for the dataset. Copies of Round 1 have been acquired by researchers in diverse areas including economics, business administration, public policy, psychology, theology, religion studies, law, and social work. Based on this I expect Round 2 of RAS to be an asset to the research community.
