Sabri Ciftci, Kansas State University, firstname.lastname@example.org
Michael Robbins, Princeton University, Arab Barometer Project, email@example.com
Sofya Zaytseva, University of Georgia, firstname.lastname@example.org
Religion informs political attitudes and hence is consequential for understanding political beliefs and preferences. This is especially true for the Middle East and North Africa (MENA) where religion is a formidable social force, ranging from its influence on the Arab Spring, various civil wars, the failure of democratization, and responses to COVID-19. Islam, especially, factors prominently in social, political, and economic life in MENA societies.
Typically, personal religiosity is taken as the central marker of religion in empirical studies. Students of religion and politics in MENA have increasingly exploited the burgeoning public opinion data and used survey items tapping self-reported devotion, belief in various aspects of doctrine, and religious attendance, among others, to measure this concept. Arab Barometer surveys, for example, are valuable resources that allow researchers to measure various dimensions of Muslim religiosity and test religion’s effects on political attitudes. Using these data, researchers have developed sophisticated explanations concerning the interplay of religiosity and political attitudes, including those related to democracy, Islamism, economic development, and international actors. Despite these advances, no agreement exists about measurement of religiosity – the conceptual boundaries and statistical validity of religiosity as a central variable remains ambiguous. It is necessary to cross-validate survey-based measurement strategies with external data other than the public opinion surveys to increase clarity in the field.
In our recent article, we provide some clarity by constructing an external measure of religiosity and by cross-validating the survey-based approaches in operationalization of this key variable. Our strategy relies on aggregated constructs of religiosity since it is more realistic to find objective measures at higher levels of analysis (e.g., nations). We used individual-level data from Arab Barometer, census data, and satellite imagery to obtain multiple measures of religiosity at the sub-national level. The study focused on the Egyptian governorates (akin to American states), but our approach could be easily applied to any country in MENA or Muslim-majority societies.
We compared aggregated measures of religiosity gathered from surveys to a unique objective measure of sub-national religiosity calculated from the nighttime lights data in Egypt’s governorates during the holy month of Ramadan. We expected to observe an increase in the volume of nighttime lights during Ramadan because several Ramadan rituals take place at night including the dinner (iftar), preparatory night meal (sahur), and communal evening prayer. Therefore, if religiosity indicated in surveys is reliable and valid, it should be correlated with a greater volume of nighttime lights during Ramadan.
Arab Barometer surveys interview a sufficiently large number of respondents in Egyptian governorates in waves 2, 3, and 4 (combined) to produce reliable estimates at the governorate level. We used measures of self-reported religiosity, prayer frequency, and engagement (reading or listening) with the Qur’an. Figure 1 provides the difference between a governorate-level religiosity index and global mean of religiosity at the Egyptian national level. There is significant variation across Egyptian governorates. A comparison of vote patterns across governorates in 2012, when the Muslim-Brotherhood backed candidate Mohammed Morsi ran for president can provide external validation of these initial results. Some of these governorates carried by Mohammed Morsi in the 2012 Presidential elections by significant margins were either most religious (Menya and Fayoum) or least religious (Suez). However, in less religious governorates including Luxor, Kalyubia, and Cairo, Morsi did not take a majority of the votes.
Figure 1: Governorate Level Religiosity Relative to National Religiosity in Egypt. Source: Arab Barometer.
We then compared the aggregated measure of religiosity at the governorate level to the estimates obtained from MrP allowing for the estimation of sub-national opinion using a single nationally representative sample of as small as 1,000 respondents. The crux of MrP is to predict the variable of interest using multi-level regression and then to weight these estimates with true values of certain demographic variables from census data, which are also included in the first stage estimation. We see significant overlap between MrP estimates of religiosity and averages obtained from our method of aggregation (Figure 2). There is a degree of skepticism about the survey measures of religiosity in the Arab region due to social desirability or misreporting. For Egyptian governorates, disaggregation index of religiosity performs well when compared to the MrP estimates. We conclude that the survey data can accurately distinguish aggregate levels of religiosity at the provincial level.
Figure 2: MrP Estimates and Survey-Based Religiosity in Egyptian Governorates. Source: Arab Barometer, Waves 2-4 (2011-2016).
We seek further validation by calculating a unique measure of religiosity from the satellite imagery of nighttime lights data in Egypt’s governorates during the holy month of Ramadan. Ramadan is a time when religious activity increases through personal (fasting and spiritual cleansing) and communal (participation in prayers and Ramadan meals) activities. Particularly, there are mandatory rituals performed after sunset that would increase activity at nighttime during Ramadan. These activities include the iftar (Ramadan dinner), sahur (night meal), and taraweeh (additional prayer following the isha (night) prayer). Our expectation is to observe an increase in the intensity of nighttime lights during Ramadan compared to the preceding month (and a decrease in the succeeding month) controlling for overall socio-economic development. To obtain this measure, we used Suomi NPP satellite data made available through the Visible Infrared Imaging Radiometer Suite (VIIRS) sensor. The VIIRS data is available from December 2011 until the present time and contains annual and monthly nighttime light composites. We used the nighttime lights data for 2014, 2016, and 2017, the three years where there was a correspondence between the Muslim and Gregorian calendars. This strategy is necessary, because VIIRS data are freely available by month according to the Gregorian calendar. To estimate governorate-level religiosity using nighttime lights data, we corrected for extreme or unusual values caused by lights from fires, oil and gas wells, and landforms, as outlined in our article. Figure 3 shows the raw (panel a) and corrected (panel b) lights for Egyptian governorates.
Figure 3: Nighttime lights during Ramadan (2014)
Note: Map of Average Nighttime Radiance Values for Egypt for July 2014 featuring a) raw data and b) corrected data.
The expectation is that the nighttime light intensity would spike for the month of Ramadan, as compared to pre- and post-Ramadan months. To demonstrate this pattern, we derive a sum of lights (SOL) measure of light activity for each month, as shown in Figure 4. The dashed grey line indicates the holy month of Ramadan that year. Notice that, in general, the data exhibits a seasonal, cyclical pattern with higher SOL values occurring during the winter months.
Figure 4: Change in the sum of lights index (SOL) pre- and post- Ramadan for year 2014.
Despite a relatively small change in nighttime lights during Ramadan (<10%), a strong correlation is observed between the relative change in night lights during Ramadan and the average religiosity in governorates. We depict this correlation visually for Ramadan 2017 in Figure 5. In general, there is a positive and curvilinear trend, but at higher levels of religious activity in Ramadan, the correlation with survey-based estimates of religiosity takes a negative turn. We suspect this could be related to a number of factors including the outlier status, size of the governorate, tourism activity, the Christian population, differing rates of poverty, or other factors that may add noise to the nightlights data.
Figure 5: Ramadan Lights and Survey Based Estimates of Religiosity. Note: The distribution of governorates is presented with Loess smoothing.
The results of our study deliver some good news to the students of religion and politics, in and beyond MENA, especially those who rely on surveys in their research. For this particular study, the results provide external validation to the questions contained in the Arab Barometer to measure religiosity. Although the measures are not perfect, they do broadly correspond with the differences in nightlights during Ramadan, suggesting that students of religion and politics in MENA can have increased confidence in them.
More broadly, most survey data sets used in comparative politics, including Arab Barometer, are designed to be representative at the national level but not at smaller geographic units such as governorates. This paper shows that disaggregation can be utilized to fill this gap. And when disaggregation is not possible, and we think this will be the case for most surveys, the MrP approach can be a substitute for understanding regional differences. Additionally, we propose that scholars can use nighttime lights innovatively during holy times, such as Ramadan or Christmas, as a non-survey-based measure of religiosity even in geographical units smaller than provinces. This opens up many opportunities for the study of religion on a variety of sub-national levels.
*This essay is based on the following article: Ciftci, Sabri, Michael Robbins, and Sofya Zaytseva. 2020. “Devotion at Sub-National Level: Ramadan, Nighttime Lights, and Religiosity in the Egyptian Governorates.” International Journal of Public Opinion Research. https://doi.org/10.1093/ijpor/edaa019