Access denied: exploring Muslim American representation and exclusion by state legislators

ABSTRACT

Post 9/11, evidence that Muslim Americans are experiencing discrimination is pervasive. This paper tests what has previously been only anecdotal evidence that discrimination extends to Muslims' treatment by legislators. I conduct two audit experiments on state legislators to test for discrimination. I first compare how state legislators in all fifty states assist low versus high socioeconomic Muslim American individuals applying for an internship and compare their treatment to that of whites. Not only does socioeconomic status not matter for Muslims – whites receive more responses regardless of SES – but party affiliation does not affect response rates either. I then run a similar experiment testing responses to religious leaders requesting a legislative visit. Imams are significantly less likely than their Pastor counterparts to receive a response. Across both studies, the results are consistent: the American Muslim community and its members experience widespread discrimination at the hands of elected representatives.

KEYWORDS:

• Muslim Americans
• racial politics
• religion and politics
• minority representation

Acknowledgments

I thank seminar participants at the Uppsala University Political Economy Workshop at the Department of Economics, Marisa Abrajano, Dan Butler, Mia Costa, Matz Dahlberg, Zoltan Hajnal, Seth Hill, Andreas Kotsadam, John Kuk, and Deborah Seligsohn for their helpful conversations and comments. A special thanks is extended to Seulgi Lee who provided thoughtful and thorough research assistance, and without whom this project would not have come into fruition.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1 “Why Muslims are the world's fastest growing religious group.” Pew Research Center. http://www.pewresearch.org/fact-tank/2017/04/06/why-muslims-are-the-worlds-fastest-growing-religious-group/. April 6, 2017.

2 Dow v. United States, 226 F. 145, (4th Cir. 1915).

3 The Census does not track religious identification information, and the “white” category obfuscates any variation in national origin that Muslims may want to highlight. Individuals are ask to mark and identify as white if they are “[a] person having origins in any of the original peoples of Europe, the Middle East, or North Africa.” See https://www.census.gov/topics/population/race/about.html. Meanwhile, in January 2018, it was announced that efforts at developing a MENA category in the 2020 Census has failed. See https://www.npr.org/2018/01/29/581541111/no-middle-eastern-or-north-african-category-on-2020-census-bureau-says.

4 About Half of Americans Say Trump Moving too Fast. http://www.gallup.com/poll/203264/half-americans-say-trump-moving-fast.aspx. Accessed February 5, 2017.

5 “The radical right was more successful in entering the political mainstream last year than in half a century. How did it happen?” Southern Poverty Law Center. Mark Potok. https://www.splcenter.org/fighting-hate/intelligence-report/2017/year-hate-and-extremism. Accessed May 29, 2017.

6 In fact, Michigan – a key battleground state in the most recent presidential election – is home to one of the oldest and largest concentrations of Muslims in America.

7 “How Many Muslims Will Vote for Donald Trump?” https://www.theatlantic.com/politics/archive/2016/02/muslim-voters-2016/458691/. Accessed May 22, 2017.

8 “Muslim Americans: No Signs of Growth in Alienation or Support for Extremism.” Pew Research Center. http://www.people-press.org/2011/08/30/section-1-a-demographic-portrait-of-muslim-americans/. Accessed March 16, 2017.

9 The Pew “Muslim Americans: No Signs of Growth in Alienation or Support for Extremism” report indicates that those with higher family incomes are more likely than those at the lower end of the scale to say they are registered. It also finds that three-quarters of those with family incomes of 75,000 or more (78%) say they are certain they are registered, compared with 60% among those with incomes of less than 30,000. See http://www.people-press.org/2011/08/30/section-5-political-opinions-and-social-values/.

10 This occurred for several reasons. As Findley (2001) and Barreto and Bozonelos (2009) note, the foreign policy issue of Jerusalem as the “undivided and undisputed” capital of Israel and their dissatisfaction with Joseph Lieberman, an Orthodox Jew, as the Democratic running mate put the Muslim American vote in the Republican Party's reach.

11 https://www.hillaryclinton.com/briefing/factsheets/2016/09/21/stronger-together-hillary-clintons-vision-for-muslim-americans/. Accessed February 8, 2017.

12 For instance, many state legislatures go so far as to insist that potential applicants for internships contact their representatives directly and ask for applications by email. See Figure A1 or the Massachusetts State Assembly page for an example: https://malegislature.gov/StateHouse/EducationalOpportunities/Internships.

13 It is also entirely possible that this name can be construed as “Arab” and could signal an Arab Christian or a non-Middle Eastern Muslim name. Thus, if I find any evidence of discrimination, it could be interpreted as legislators exhibiting anti-Arab and not necessarily anti-Muslim bias.

14 Figures A2 and A3 display screenshots of the Google images for each of the names typed into the search box in February 2018, three years after the study was conducted. However, the Google image results are revealing; the “Abdul Al-Nawad” images are of Middle Eastern looking men and the “Jake Thompson” images are mostly of white men, though most of them are of a baseball player.

15 Block random assignment is a procedure at the randomization stage whereby subjects are partitioned into subgroups (called blocks) and complete random assignment of each treatment and control occurs within each block (Gerber and Green 2008). This technique ensures that equal numbers of covariates will be assigned to each experimental condition. It also has the advantage of reducing sample variability and allows the research to eliminate entirely the possibility of rogue randomizations (Gerber and Green 2008)

16 I waited six months to collect the data to account for the fact that the legislative sessions of state legislatures vary from state to state and that some states are in session as little as 50% of the time. See NCSL Full and Part Time Legislatures Map on Gold Legislatures for more details. http://www.ncsl.org/research/about-state-legislatures/full-and-part-time-legislatures.aspx. Accessed May 29, 2017

17 See Costa (2017a) for information on average response rates across a wide range of audit studies.

18 A Pew Research Center poll, for instance, found that Muslim Americans overwhelmingly approved of the way that President Obama handled his job as president, with 76% approving of his job performance in June 2011 compared to 46% of the general public.“Muslim Americans: No Signs of Growth in Alienation or Support for Extremism.” Pew Research Center. http://www.people-press.org/2011/08/30/section-5-political-opinions-and-social-values/. Accessed May 26, 2017.

19 In 2014, Pew reported that 38% of Americans say they know someone who is Muslim, compared to 61% who know a Jewish person or 59% who know an atheist.“How Americans Feel About Religious Groups. Pew Research Center.” http://www.pewforum.org/2014/07/16/how-americans-feel-about-religious-groups/. Accessed May 29, 2017.

20 “How Americans Feel About Religious Groups. Pew Research Center.” http://www.pewforum.org/2014/07/16/how-americans-feel-about-religious-groups/. Accessed May 29, 2017.

21 Table A2 shows the corresponding models, but instead conditions on the response.

22 Of those who responded, the most helpful responses came from legislators in Florida (82.3%), Illinois (86.6%), Indiana (76.9%), Massachusetts (71.4%), Maine (90.9%), New Jersey (86.9%), New York (89.2%), Texas (92.8%), Virginia (85.7%), Wisconsin (88.9%). What is particularly noteworthy is that these states are not necessarily among the most professionalized state legislatures according to the NSCL. With the exception of New York, these states are “full-time lite,” “hybrid,” and “part-time lite” state legislatures. See http://www.ncsl.org/research/about-state-legislatures/full-and-part-time-legislatures.aspx.

23 In Table A2, I display models that introduce bias (see Coppock 2018), but which follow the lead of prior scholarship (e.g., White, Nathan, and Faller 2015). The models in Table A2 code legislators who replied to the fictional alias but did not assist them in procuring an application for the political internship as 0. But it is important to highlight two important limitations to these analyses. First, they introduce bias in favor of finding less discrimination against Muslims insofar as examining only those responses from those legislators who responded to email queries conditions on a post-treatment outcome, which “de-randomizes” the experiment. Second, the analyses in Table A2 also raise a “truncation by death” problem. Table 7 remedies this limitation: 1 corresponds to a helpful response and 0 indicates that either a response was not received or that it was not helpful. As Table 7 demonstrates, the results are more even more powerful and consistent when I examine the RH dependent variable. While the magnitude of the coefficients for the Muslim alias treatment is smaller in Models 1 and 2 than the corresponding models in Table A2, the standard errors are much smaller. Moreover, the disaggregated Muslim low and high education treatments in Models 3 and 4 in Table 7 display similar findings: Muslims receive less helpful responses than their White high education counterparts. Unlike the corresponding models in Table A2, the magnitude of the effects are smaller, which one would expect given that the observations have grown tenfold.

24 Only the Muslim, low education treatment is significant in Table A2. But this is likely due to the small number of observations. There are only 87 Muslim, high education treatment observations who received a reply from all legislators in Study 1 and the negative coefficient borders on significance at p-value = 0.101.

25 In addition, Figures A4 and A5 display screenshots of the Google images for each of the names for the Pastor and Imam typed into the search box.

26 As in Study 1, I block randomized in Study 2 to reduce sample variability and to eliminate entirely the possibility of rogue randomizations.

27 As noted before, I used the same email addresses for state legislators in Study 1 as in Study 2. Differences in response rates may be due in part to differences in the nature of the requests. Moreover, while the aggregated response rate in Study 1 was 7.8%, in states with large Muslim populations, it was 12.23% (which of course is still much lower than 35.5%).

28 There are several ways of interpreting the 26.75% response rate to the Imam treatment. On the one hand, the 26.75% response rate in Study 2 and in states where Muslims comprise large numbers is significantly higher than the 5.11% response rate to the Muslim treatments in Study 1. On the other hand, however, Republicans were almost twice as likely to respond to the non-Muslim treatments as to the Muslim treatments. Future studies should exploit this discrepancy and observe whether responsiveness to Muslim treatments is similar to other racialized minorities, such as Hispanics, Asian Americans, and African Americans.

29 A helpful response in this experiment is coded as 1 if the legislator's reply assists the fictional alias in scheduling a legislative visit and a 0 if the legislator's reply does not.

30 Coder 1 evaluated the helpfulness of these 442 responses.

31 https://twitter.com/MohamedMOSalih/status/928110455851065345