Monday, April 25, 2016

In "By the Numbers" Film, Clarion Project Misunderstands Pew's (2013) "The World's Muslims" Survey Report

A section of Clarion Project’s (2015) “By the Numbers” video on “radical Islam” deals with Pew’s (2013) survey report on Muslims’ opinions. That section of the video contains some erroneous numbers. Examination of Clarion Project’s explanation of how they obtained their numbers indicates that they missed the relevant percentages shown in Pew’s report for the three harsh sharia punishments (i.e., for apostasy, theft, and adultery, pp. 219-221). Proceeding from that mistake, Clarion Project made two major factual errors, namely, assuming that the three harsh punishment questions were (1) asked only of the subset of Muslims who favored making sharia the official law of the land, and (2) not asked in the sub-Saharan countries surveyed. These two errors led Clarion Project to make incorrect estimates of the overall percentages of Muslims who favored these harsh punishments. In addition, their estimate of the average percentage of Muslims who favored sharia as the law of the land is questionable. They used less-than-appropriate methods to estimate the average percentages of Muslims who supported the harsh punishments. Finally, Clarion Project implies that support for sharia as the official law of the land is not an indicator of radicalism. Empethop addresses these claims and corrects the factual errors by drawing upon evidence from the Pew (2013) survey, its primary researcher, and its accompanying microdata files.


By now, Clarion Project's 2015 video "By the Numbers: The Untold Story of Muslim Opinions and Demographics" about the troubling rise of “radical” Islam, has probably been viewed by many people. The video, narrated by Raheel Raza, raises some important issues for discussion. Unfortunately, a section of the video (from about 10:47) dealing with Muslims’ opinions is built upon some false assumptions about Pew’s (2013) "The World's Muslims" survey, leading to some false claims about the numbers. The author(s) seem to have misread, or to have missed, important parts of the survey’s full report. For example, they seem to have missed entirely the tables in the Topline section showing the percentages of Muslims who favor or oppose hardline sharia punishments (pp. 219-221) including the death penalty for leaving Islam, whipping and cutting off of hands for crimes like theft, and stoning of adulterers. As I’ve documented, Pew’s famous 2013 report on Muslims’ opinions has been widely misread or not read by many of those who discuss it or write about it.

Clarion Project explains how they arrived at their numbers based on the Pew survey’s full report, to which they link, in their pdf titled “By the Numbers Movie -- How We Measured the Stats.” That explanatory document is accessible through clicking on a tab, below the video, labelled “Find out How We Reached the Numbers.” Readers will probably find it helpful to open and keep available that explanatory pdf, as well as Pew’s (2013) full report pdf, for reference while reading the present article.

False Claim #1: Only Muslims who favored sharia as the official law of their country were asked the three questions about harsh sharia punishments.

In their pdf, Clarion Project explains how they calculated the percentages for the harsh punishment questions that they number as 3 (death for apostasy), 4 (whippings and cutting off of hands), and 5 (stoning adulterers):
“For statistical purposes concerning the topics listed below in numbers 3, 4 and 5, the question was asked first if he/she believes that sharia law should be the law of the land. Only those answering “yes” were then asked questions 3, 4 and 5. Therefore, percentages for those beliefs were computed within the percentage of those who believe that sharia law should be the law of the land.”
That’s major error number one. The harsh punishment questions were asked of the general sample of Muslims, not just the sharia-favoring subset. In October of 2014, I sought confirmation from the primary researcher of the Pew (2013) survey report in relation to this issue. In his reply, he wrote that the questions about the harsh punishments were asked "...of all Muslims, not just the ones who want sharia to be law.” Figure 1, below, shows a screen capture of that reply:

Figure 1.

(Dr. Bell's response mentioned Egypt because I’d cited it in an example of a popular media error). In most countries there was a small percentage of respondents who “opposed sharia” (as presented in Q79a), but who favored at least one of the three harsh sharia punishments (Q92b, c, d). Some people are puzzled by these results, but as I noted briefly on this point in my fact-check article, the subset
“…that supports the death penalty for apostasy but not sharia (as presented in Q79a) may consist of respondents of at least a few different kinds, including those who support some sharia but not a comprehensive implementation of it; or support sharia as a source of law, but not as the only source of law; or who support sharia but regard the nation state as un-Islamic and want it replaced by a purely Islamic system under a Caliphate.” 
(I suggested those possibilities based on other survey results, and plan to discuss them in a future article). In addition, some who responded to Q79a with a “don’t know” or a “refusal” to answer favored at least one of the harsh sharia punishments. In sum, there is a significant minority of respondents who for various reasons did not give a favor response to Q79a but did give a favor response to at least one of the harsh punishment items.

If it were true that only those respondents who favored sharia were asked about the three harsh punishments, then that should be reflected in the raw sample sizes for those questions. Fortunately, we can check those sample sizes because Pew has made freely-available to the public the microdata files for the surveys. In a previous article I described how to access those data files and how to do a basic analysis of them using the free statistical software PSPP. Practically anyone with internet access who is familiar with how to use a spreadsheet can use PSPP to open Pew’s data files to explore them or to fact-check claims about them. A simple crosstab analysis of the main data file for The World’s Muslims survey shows that the raw (unweighted) total numbers of those who were asked the apostasy question are more than the numbers of those who favored sharia (compare Figures 2 and 3, below). This demonstrates that Clarion Project’s claim that only those who favored sharia were then asked the apostasy question is false. Also, the PSPP outputs show that the unweighted total numbers of respondents per country for the sharia question (Figure 2, Q79a) and the apostasy question (Figure 3, Q92b) are the same. That fact confirms the point stated by the Pew survey’s primary researcher, quoted above. As a corroborating point of reference from a different file, Figure 4 is a screenshot of the Muslim sample sizes given on pages 39 and 150 of Pew’s (2013) report. Again, note that the sample sizes on pages 39 and 150 are the same as the totals for Q79a and Q92b.

Figures 2 and 3. Data Source: Pew’s “The World’s Muslims” Dataset (2012). Analysis by Empethop.

Figure 4. Source: Pew’s (2013) “The World’s Muslims” Survey Report (pp. 39 and 150).

Notes about the above Figures 2, 3, and 4: Sample sizes shown are unweighted, whereas Pew’s reported percentages for responses are based on weighted data. Algeria, though included in Pew’s data file, did not appear in their report because, due to an administrative error, the data only represented 75% of the Muslim population there. In Morocco the sharia question was asked but not the three harsh punishment questions, while in Uzbekistan none of those four questions were asked. Thailand and Russia were asked an alternative version of the sharia question, which is listed separately, though, again, all Muslim respondents there were asked (a non-standard version of) the sharia question and the standard three harsh punishment questions. Niger appears here because it was surveyed in the same 2011-2012 period as the above countries and appears in the same data set. 

I used the apostasy question in the example above, but the total sample sizes for the other harsh punishment questions (whipping/cutting, adultery) are the same as those shown above and can be verified the same way. In the sub-Saharan African countries, all respondents in Pew’s survey who were asked the sharia question were also asked the three harsh punishment questions (see below).

The fact that the three harsh punishment questions were asked of the general samples of Muslims is reasonably clear from a careful reading of the Pew report, as I show in Appendix I of my fact check article. This is further confirmed by checking the percentages for these questions in Pew’s Global Religious Futures interactive resource; these are not described as for a subset but are for the general samples of Muslims. For example, when you find the summary data for the apostasy question, click on the preview chart icon to see the table of data, as I show in Figure 5, below. One can see that these numbers match* those in the table in Pew’s (2013) full report, page 219. *(There was one minor discrepancy for Nigeria--listed with 28% favoring death for apostasy--perhaps due to an error. The sub-Saharan study’s data file has Nigeria at 29.48% in favor of death for apostasy, and both the “Tolerance and Tension” and “The World’s Muslims” reports have it at 29%).

Figure 5.

Notes for Figure 5. “Age Group: all” means all age groups among those surveyed, i.e., age groups among those 18 years of age or older. The sub-Saharan countries were surveyed between 2008 and 2009. Their results are listed in the above table as for 2010, and they were published in that year in the “Tolerance and Tension” report and in Pew’s (2013) report.

Readers can verify these facts using Pew’s published reports and data sets. Those who still feel unsure after having considered the evidence I presented above and in my previous fact-check article (especially in Appendix I) can easily contact Pew for clarification (see “For Further Information Contact” on the cover page of Pew’s 2013 full report).

False Claim # 2: Muslims in the sub-Saharan countries weren’t asked the three questions about the harsh sharia punishments.

The Clarion Project author(s) do not seem to be aware of pages 219-221 of Pew’s (2013) full report. They wrote:
“Although Muslims in Sub-Saharan Africa were asked if they believe Sharia should be the law of the land, they were not asked the questions about apostasy, honor killings, corporal punishment and stoning for an unfaithful spouse. Therefore, we have not computed those countries in our statistics.”
That’s major error number two. At this point, the reader is probably aware that on pages 219-221 of Pew’s (2013) full report, the data for Muslims in the sixteen sub-Saharan countries are shown, indicating that they were asked the three harsh punishment questions. (This was also confirmed incidentally in Figure 5, above, for the apostasy question). The sub-Saharan countries do not appear, however, in the bar plot showing the percentages favoring the harsh punishments among the sharia-favoring subset (pages 52-55 of the full report pdf). If one only looked at the bar plots and did not read the full results in the Topline, one might get the wrong impression that the sub-Saharan respondents weren’t asked the harsh punishment questions. On pages 15 and 46, the sub-Saharan countries are listed with the others where the sharia question was asked.

Clarion Project invites comments via a box below the By the Numbers video link on their site. On April 2nd of 2016, I sent a message via the comment box notifying Clarion Project of the two major errors above, though as of the present (April 25, 2016) there has been no indication that a correction has been or will be made. [Update: On April 27, I also sent them a direct message via twitter asking them to check and correct their numbers, and provided a link to this article]. If it comes to my attention that they’ve issued a correction, I’ll update the present article accordingly.

What about Clarion Project’s estimated percentages and numbers of Muslims favoring sharia, death penalty for apostasy, etc.?

As one can imagine, the calculations based on Clarion Project’s erroneous assumptions led to some incorrect results. The video's section dealing with Pew's (2013) findings begins at about 10:47. Clarion Project’s explanatory pdf gives some clues as to how they calculated their percentages. They do not state with sufficient clarity how they calculated their averages, but my attempts to reproduce their average percentages suggest that they took the unweighted mean of the country percentages, at least for the suicide bombing and honor killing items (see Mean of Countries, in Table 1, below). Using that same method, however, I could not reproduce their across-countries average of 53% in favor of sharia. As shown in Table 1, below, I obtained 57%, whether Thailand and Russia were included (n = 38) or excluded (n = 36). I don’t know how they arrived at 53%. For the sharia-favoring subset, I did reproduce their means of 38%, 52%, and 51% in favor of the punishments for apostasy, crimes like theft/robbery, and adultery, respectively, by finding the means of the percentages for the 20 countries shown in the bar plots on pp. 52-55 of Pew’s (2013) full report.

For each of the harsh punishment items, due to wrong assumptions, Clarion Project appears to have multiplied, for each of the aforementioned 20 countries, the proportion favoring sharia by the proportion favoring the harsh punishment, then taken the mean of those 20 products to be the average of those countries. Using their false-assumption-based method, I reproduced their erroneous across-countries averages of 27%, 32%, and 33% in favor for the apostasy, whipping/cutting, and adultery items, respectively.

To estimate the numbers of Muslims who favored death for apostasy, Clarion Project used
“…the sum of the Muslim populations of each country surveyed by Pew multiplied by the average for that particular country.” 
Using that method, they found that “302 million people” favored death for apostasy. I was unable to reproduce their numbers using Pew’s published 2010 estimates of the countries’ (all-ages) Muslim population sizes. For example, when I followed their above description for finding the sum of those who favored death for apostasy, I got 289,368,860, not 302 million. That led me to speculate that Clarion Project may have used other estimates of the populations, perhaps for years that overlap more closely with the survey periods (i.e., 2011-2012, for the 20 countries in Clarion Project’s apostasy calculation). Still, I don’t think that fully accounts for the difference between their estimated numbers and mine (see below). They state in their pdf that “…27% of the total sample surveyed support the death penalty for leaving Islam, which translates to 237 million people.” That implies a total sample (sum of the 20 countries’ Muslim populations) of approximately 878 million, plus or minus several million due to the use of rounded figures. That’s much more than the 809,640,000 that I found while trying to reproduce Clarion Project’s numbers from the sum of the Muslim populations of those 20 countries using Pew’s 2010 estimates. It would be helpful for Clarion Project to cite exactly where, or explain exactly how, they obtained the original population numbers used in their estimates.

Yet, even if we assume that Clarion Project’s “total sample” is correct in this case, their “27%” and the accompanying number of 237 million is almost certainly an incorrect estimate of the (all-ages) Muslim population favoring death for apostasy. It does not make sense to multiply (a) an average taken across countries treated as equal units by (b) the total population that is comprised of countries that are highly variable in size (i.e., unequal units) and expect to get a valid result. But that appears to be what Clarion Project did when they multiplied their mean of country proportions by the “total sample” for each item. Instead, to obtain an average that accounts for the different population sizes of the different countries surveyed, we can divide the estimated number of people favoring the death penalty for apostasy by the estimated total number of people for that item. When we divide 289,368,860 by 809,640,000, we get about 36% of Muslims of the 20-country population favoring death for apostasy. That calculation is an improvement, but is still not valid, though, because of Clarion Project’s exclusion of numerous relevant countries and respondents.

The fact that Clarion Project provides two different numbers for those who favor or justify the various items is a major problem. There should be only one population estimate for each item. Allowing a reasonable amount of error or uncertainty, multiple methods of estimation should produce results that converge closely; results should not differ by proportionately large amounts. For the numbers favoring stoning of adulterers, Clarion Project’s two estimates—289 million and 387 million—differ by 98 million! There is a difference of 151 million between their two estimates for the numbers favoring sharia.

One of the problems with Clarion Project’s method of calculating averages in this case is that it uses countries as the units, not persons. While there are instances where it’s appropriate to use countries as the units in calculating averages, in this case we are attempting to estimate what the overall relevant population of persons believes about various extreme or fundamentalist practices. If we want to estimate what percentage of Muslims favor the death penalty for apostasy, it is much more appropriate to use persons as the units. To do this, we need to take into account the sizes of Muslim populations in the various surveyed countries. This is important because, obviously, between countries there are vast differences in numbers and thus vast differences in the shares of the world’s Muslim population. For example, Indonesia’s general Muslim population is about 126 times the size of that of Kosovo. But taking the average of countries, as Clarion Project does, gives Kosovo equal weight to Indonesia in the estimation of Muslims’ opinions.

In addition, the relevant Muslim populations should be of those age 18 or older (Muslim adults) because this is the age group of the surveyed population. In the method I’ve used to estimate the percentage of Muslim adults who favor death for apostasy, for example, one (a) sums the adult Muslim populations of the relevant 36 countries surveyed by Pew (2013) to find the relevant overall adult Muslim population size, and one (b) multiplies each country’s adult Muslim population by the proportion who favored death for apostasy and sums these 36 products. One then divides sum b by sum a to obtain the overall proportion of adult Muslims who favor death for apostasy. (Note: For the apostasy item I include the countries on page 219 of Pew’s 2013 full report, except for Thailand because its sample is not nationally representative. That leaves 36 countries). Using this method, I estimate that about 39.42% of Muslim adults favor death for apostasy, as shown in Table 1. I used the same method to estimate the other percentages in the Mean of Muslims ≥ 18 years of age column. I go into detail about this method in a forthcoming article.[1]

Table 1.

Estimates of Muslims’ Support for Sharia, Fundamentalism, and Militancy, based on Pew’s Survey (2013)

Mean of
Clarion P.
Clarion P.
Mean of
Number of
Muslims ≥ 18y
Muslims ≥ 18y
Favor sharia as the law
Favor death for apostasy
Favor whippings, amputations
Favor stoning of adulterers
Suicide bombing justified (o s)
Honor killing justified (o s r)

Notes. Displayed numbers are rounded.                                                                                                               
Thailand is excluded from Empethop’s estimations because its sample is not nationally representative.                                                                                                     
*Indicates number is erroneous and/or obtained based on erroneous assumption(s) or inappropriate method.      
^Indicates number is of unknown origin and probably not valid.

Mean of Countries = sum of countries’ favor/justify response percentages, divided by number of countries.
Clarion Project Estimate %: See text for details. Estimates are likely based on Mean of Countries calculation.  
Clarion Project Est. (Millions): average proportion from Mean of Countries, multiplied by total Muslim population.
Mean of Muslims ≥ 18y = sum of adult Muslims favor/justify, divided by sum of adult Muslims. By Empethop.
Number of Muslims ≥ 18y = sum of adult Muslims. Estimation by Empethop based on Pew numbers.
Questions from Pew (2013) The World’s Muslims survey report: Favor sharia as the law Q79a (and variants). Favor death for apostasy Q92b. Favor whippings, amputations Q92c. Favor stoning of adulterers Q92d. Suicide bombing and other forms of violence justified often or sometimes Q89. Honor killing of female justified often, sometimes, or rarely Q54 (and variant).

Is Clarion Project correct that the majority of Muslims are not “radicalized”? 

In this section I express disagreement with some of Clarion Project’s conclusions. In their explanatory pdf, they wrote:
“Our conclusion, according to the facts and public-opinion polls, is that a majority of Muslims are not radicalized. However, according to these same facts and polls, there is no way to deny there is a significant and serious problem of radicalism in the Muslim world.”
Most people can probably agree with the latter sentence. As to the former, with respect to Pew’s data, the conclusion based on a correct assessment is less clear than Clarion Project suggests. They included Islamic fundamentalism as one of the spheres of “radicalization” (see 10:21 of the video). It appears to me that they didn’t end up classifying the majority of Muslims as radicalized mainly for two reasons:

(1) They made major errors in their understanding of Pew’s report and less-than-appropriate analyses that grossly underestimated the mean percentages of Muslims who support harsh sharia punishments, and

(2) They characterize sharia as so diverse in its interpretation and application that in their view favoring it in Pew’s survey is not indicative of radicalism.

I’ve laid out the case for reason #1 in the previous sections. With a correct understanding of the data and appropriate analyses, the percentages of support for the harsh punishments are much higher than those claimed by Clarion Project. According to my estimate, the majority of adult Muslims (51%) in the population sampled by Pew favor the stoning of adulterers, as indicated in Table 1 above and as I will present further in a forthcoming article. My estimate of the percentage of support for whippings and amputations as punishments for crimes like theft rounds off to 50% in Table 1, but more precisely it is a plurality at 49.63% (the percentage opposed is 45.58%, the remainder don’t know/refused), also as shown in the forthcoming article.

The percentages of support for the death penalty for apostasy and honor killing, each, are less than majorities at about 39%—literally consistent with Clarion Project’s claim that the majority is not radicalized. With minorities as high as 39% for each of those items openly supporting murder, though, there is much cause for concern, as no doubt the people at Clarion Project would agree.

There is much more to consider, such as the questions not asked and the countries not surveyed by Pew, surveys by other organizations, and the extent to which the findings about beliefs correspond with actual practices. Those issues are important, but discussions of them are beyond the scope of the present article.

The attempt to classify respondents as either radical or not radical (or “moderate”) based on their survey answers raises an important question: How many different kinds of murder or extreme violence would someone have to favor or justify to qualify as a “radical”? At least one? At least two? All of them? Given the nature of the harsh punishments and honor killing, for example, it seems reasonable to suggest that support for at least one of them would suffice to make someone a radical, at least in the sense of supporting some hardline fundamentalist practices. In another forthcoming article, I will document results of analyses of the Pew data that examine how many such items respondents tend to support, how many they tend to oppose, and the percentages that support at least one, at least two, etc.

In explaining reason #2, Clarion Project states: 
“It is very important to note that just because someone supports Sharia as the law of the land does not mean he is radical. What the Pew found was no one could really agree on how Sharia is defined and how it is to be implemented. Some see it as mainly spiritual, some see it as primarily connected to civil laws like marriage, divorce, and money, and some see it as a political worldview.” 
I disagree with Clarion Project here. In my opinion, if someone wants sharia to be the official law of the land in their country—which covers criminal laws and a comprehensive regulation of society—that in itself is a pretty strong indication that he or she is “radical.” Keep in mind that Clarion Project includes support for the harsh sharia punishments and honor killings as indicative of radicalism.[2]  Overall, the sharia-favoring subset is more likely to support the harsh punishments and honor killings than is the sharia-opposing subset. Even if some who favor sharia as the national law don’t want a complete implementation, favoring it implies favoring at least a pretty extensive implementation.

One may also consider the fact that someone can support some sharia, perhaps even more than is implied by Q79a, but still not favor Q79a. As I’ve pointed out, there is a small but significant subset that opposes sharia as worded in Q79a but favors at least one of the specific harsh sharia punishments. Clarion Project, due to their error, excluded that subset from their consideration of the three harsh punishment items, thus underestimating the support for these aspects of sharia.

I think it’s important to distinguish between what Pew found empirically and some of their looser or more general comments about sharia in parts of the text of the 2013 report. Hence, I don’t think Clarion Project has it quite right when they say that Pew found that “…no one could really agree on how Sharia is defined and how it is to be implemented.” That would imply that knowing someone’s response on Q79a has no significant correlative or predictive value in guessing how they responded on other items pertaining to sharia. It is probably fairer to say that Pew found both differences and commonalities in the understanding of sharia. I think they found enough commonalities that knowing whether a respondent favored or opposed sharia can give us some indication as to what else that respondent believes about the specifics, including the harsh punishments. While there are some in the sharia-favoring subset who opposed the harsh punishments, this does not necessarily mean that they oppose all punishments for apostasy, theft, or adultery. They may well favor some punishments, even harsh ones, just not those that Pew used in their few questions covering this subject.


Pew Research is not responsible for my analyses or my interpretation of their data. From the Pew instructions for downloading data sets: “All manuscripts, articles, books, and other papers and publications using our data should reference the Pew Research Center’s Religion & Public Life Project as the source of the data and should acknowledge that the Pew Research Center bears no responsibility for the interpretations presented or conclusions reached based on analysis of the data.”


[1] The following is adapted from a draft of my forthcoming article:
“…the 36-country population-weighted percentage of Muslims ≥ 18 years of age who favored the death penalty for apostasy is 39.42. (For points of comparison, the figure for the 36-country all-ages Muslim population is 40.57%, and for the 36-country age ≥ 15 years Muslim population is 39.64%). …An alternative though slightly less accurate method for estimating the 2010 age ≥ 18 Muslim populations uses (a) the U.N.’s published estimates for the general (not religion-specific) age ≥ 18 population for each country, multiplied by (b) Pew’s published estimates of the proportions of Muslims (not age-specific) in each country. Using this alternative estimate, the 36-country population-weighted percentage of Muslim adults who favor death for apostasy is 39.11.” 
The fact that those various numbers converge fairly closely, despite differences in methods, source estimates, and age groups, suggests that 39.42% is probably a valid estimate of the relevant population’s support for the death penalty for apostasy, among the countries surveyed by Pew, as compiled in the 2013 report. On the other hand, as I also show in that same forthcoming article, with a set of seven countries surveyed by Pew Global in 2010, a plurality of about 48% of that seven-country adult Muslim population supported making the death penalty for apostasy the law in their country.

[2] Terms such as “radicalism,” “extremism,” “fundamentalism,” “traditionalism,” etc., are potentially misleading because the troubling views in question are, according to the survey results, mainstream among Muslims worldwide. Terms such as “liberal” and “conservative” as applied to the views of Muslims across various countries and cultures can also be somewhat misleading due to political connotations. “Islamist,” to describe someone who wants to make sharia a significant part of the political or social order (or to make it more so), is also potentially misleading in this context, because that would categorize the majority of Muslims as Islamists. Despite the problems with such terms, we will probably have to make do with them, with caveats and clarifications, until better alternatives are established in popular usage.

Main References

Clarion Project (2015) “By the Numbers: The Untold Story of Muslim Opinions and Demographics,” Full film (video). The segment from 10:23-13:18 deals with the Pew findings.

Clarion Project. “By the Numbers Movie: How We Measured the Stats.”

Pew Research Center (April 30, 2013). The World’s Muslims: Religion, Politics and Society. Full report, also called the Complete report.


No comments:

Post a Comment