Muslim Population US

8/2/2019 Muslim Population US

1/13

ESPACE. POPULATIONS. SOCIETES. 2003-1 pp. X9-/0/

John R. WEEKS International Population CenterDepartment of GeographySan Diego State UniversitySan Diego. C \ 921S2--I-I9.\[email protected]

Estimating the MuslimPopulation in the United StatesUsing Census 2000 DataAcknowledgments: JenniferPaluch provided valuable research assistance on thisproject.

INTRODUCTIONThere has been a Muslim presence in theUnited States for centuries. It is virtuallycertain that many of the slaves brought tothe Americas from Africa were Muslimbecause western Africa. from which mostslaves came. has a long history of Muslimcivilization (Nyang, 1992), dating back tothe Illh and 12th centuries (Levtzion, 1968).For example. the northern part of Nigeriahas been largely Muslim since at least the1300s. and Nigeria was frequented by slavetraders in the seventeenth and eighteenthcenturies. According to Haddad (1986a),there is evidence that as early as 1717 therewere Arabic-speaking slaves in Americawho reportedly ate no pork and believed inAllah and Muhammad. There is some evidence that as many as ten percent of slavesbrought to North America were Muslim(Austin. 1984), but Christianity wasimposed upon the slave population, andslaves who refused to convert were persecuted or killed (Nyang, 1999).

In the late nineteenth and early twentiethcenturies, Muslim (as well as Christian)migrants entered the United States fromvarious middle eastern nations, includingwhat are now Lebanon, Syria, and Iraq(Haddad, 1986b). Although these immigrants established a clear presence for Islamin American society (Haddad, 1986b;Rashid, 1999), it was not easy to be otherthan a Christian in the United States. Thepassage of the highly restrictive national ori-;gins quota system in the US in the 1920seffectively cut off immigration from all butnorthwestern European and Latin Americancountries. until that law was replaced by theless restrictive Immigration Act of 1965.The post-WorldWar II partition of Palestine,and subsequent political and economicunrest in the region led to refugee migrationto the United States, but the volume ofmigrants and refugees from a number of predominantly Muslim nations has increased


2/13

largely because of the liberalization ofimmigration laws in the mid-1960s.Efforts began within the African-Americanpopulation to build a community of Islamduring the Reconstruction period followingthe American Civil War. Initially these activities were outside the mainstream of Islam(McCloud, 1995), but since the 1970s thereappears to have been a steady (albeit largelyunmeasured) increase in the number ofAfrican-Americans who adhere to mainstream Islam (American Muslim Counci l,1991; Rashid, 1999).Although the number of Muslims in theUnited States is almost certainly large andgrowing, it is not certain how large or at

90what rate the population is increasing. Forthis reason, the geographic distribution ofthe Muslim populat ion also is somewhatuncertain, although anecdotal evidence canbe used to discern the basic patterns thatexist . My purpose in this paper is to reviewestimates that have been made of the size ofthe Muslim population in the United States,and compare them with new estimates thatI derive from proxy measures based onCensus 2000 data. I then use the censusbased measures to estimate the geographicdistribution of the Muslim population in theUnited States in 1990 and :WOO, and fromthis I am able to calculate rates of growth bydifferent areas of the country.

1. ESTIMATING THE NUMBEROFMUSLIMS IN THE UNITED STATES

In the United States, unlike in Canada andseveral other countries in the world, religionhas never been asked as part of the regulargovernment-funded decennial censuses. TheCensus Bureau did collect information in itsCensus of Religious Bodies from 1906-1936,but Public Law 94-521 prohibits the CensusBureau from asking a question on religiousaffiliation on a mandatory basis and so itcannot be included as part of the decennialcensus. Questions on religion were asked aspart of the March 1957 Current PopulationSurvey, in anticipation that a question onreligion might be included in the 1960decennial census, but ultimately that planwas dropped by the Census Bureau.Surveys can fill in gaps in census data, hutonly recently in the United States have surveys begun routinely to include "Islam" or"Muslim" as a category of response when aquestion about religion is asked. Since mostresidents of the U.S. are at least nominallyChristian, even small samples are able toprovide reasonable estimates of the numherof such individuals and many Christianchurches keep membership lists which arecompiled hy various groups to estimate thetotal population of Christians by branch ofChristianity (see, for example, the wehsitehttp://www.adherents.com). For less populous groups, estimation is more problematiceven at the national level and, of course.

relatively small national surveys providelittle information about the geographic distribution of a population. For these reasons,it is necessary to employ indirect methods inthe estimation of the Muslim population.People have been trying to figure out howmany Muslims reside in the United Statessince at least 1973, when Lovell suggestedthat there might be 900,000 Muslims in theUS and Canada, based on "preliminary tabulations from a religious census being conducted by a committee through the IslamicCenter of Washington, D.C." (reprinted asLovell, 1992:60). The "census" consisted ofa questionnaire sent to Muslim communityleaders throughout the US and Canadaasking for their assessment of the localMuslim popula tion. Since Canada hasapproximately one-tenth the population ofthe US, if we assume that Muslims weredistributed proportionately between the twocountries, it would imply that approximatelyR20,OOO were in the US circa 1970. In 19S0Thomas Phillipp argued that "there are perhaps 200,000 to 300,000 Muslims in theUnited States today; it is impossible toobtain more accurate figures ... This estimate however does not include :2 millionAfro-Americans claimed by the Nation ofIslam... Nor does this es timate includeMuslim students in the United States"(Phillipp, 19XO, p. 7:12). If we include those


3/13

persons. the number of Muslims in the USby his estimate would have been about 2.3million in 19XO. In a more systematic analysis. Ghayur ( 19X I ) estimated that there were1.2 million Muslims in the United States in19XO. His method was to list the immigrantethnic groups that were composed predominantly of Muslims and then to estimate thenumher of persons in each group using census data and immigration data. He thenadded his estimate of 75.000 AfricanAmericans to reach his total of 1.2 million.A similar method was used by Weekes(19X4) to estimate a total of 1.4 millionMuslims in the United States as of approximately 19XO. The principal differencebetween the estimates of Ghayur andWeekes is that Weekes estimated a largernumber of African-Americans to be Muslimthan did Ghayur.Stone (1991) estimated that in 1980 theimmigrant Muslim population in the U.S.was 2.3 million. To this she added, somewhat arbitrarily, one million AfricanAmerican Musl ims. for a total Muslim population in 1980 of 3.3 million. She beganwith the 1980 Census of Population andused place of birth and ancestry to estimatethe number of people who were of probableMuslim origin. She then used Immigrationand Naturalization Serv ice data on immigrants by country of origin to estimate thenumber of immigrant Muslims added to theUS population after the census, applying toeach set of immigrants a fraction equal tothe proportion of persons in each country oforigin who were estimated to be Muslim.She then applied a birth rate of approximate Iy 16 births per thousand per year,added in new immigrants since 1980 andproduced an estimate of 4.0 millionMuslims in 1986. Further updating using thesame methodology generated a 4.6 millionfigure for 1988 which appeared in Timemagazine and was widely quoted at the time(Gatling. 1988).A much lower. and also highly publicizednumber was published by Kosmin and hisassociates as part of the 1989-90 Nat ionalSurvey of Religious Identif icat ion (NSRI)(Goldstein and Kosmin, 1991; Kosmin.1991: Kosmin and Lachman. 1993). Thesedata were from a national probability sampleof households designed especially, although

91not exclusively, to provide estimates of theJewish populat ion in the United States.Based on telephone interviews with 113,000households in the United States. the NSRIinitially concluded that there were 527.000Muslim adults in the US, representing0.3 percent of the total US adult population.Applying that same percentage across allages produced an estimate of 750,000Muslims as of 1990. The researchers weresubjected to considerable cri ticism for thisfinding and although they defended theirresults (see the Appendix of Kosmin andLachman, 1993), they also acknowledgedthat the overall response rate to the telephone interviews was only 50 percent, evenafter four attempts to make contact. Theyalso acknowledged problems with languageand they acknowledged that immigrantsfrom countries like Iran. "with their experience of persecution" (Kosrnin andLachman, 1993, p. 287) might have beenreluctant to reveal their religion, even if theycooperated with the rest of the survey.Ultimately, the weighting for Muslim households was adjus ted upward to increase theiroverall estimate of the US Muslim population to 1.2 million as of 1988 - a numberwell below Stone's estimate of 4.6 million.Despite its shortcomings, the NSRI studydemonstrated the potential utility of deriving estimates of the population of allreligious groups, including Musl ims, fromsurvey data. From the mid-1990s through200 I there were several surveys from whichestimates can be drawn of the Muslim population, including an updated version of theNSRI called ARIS - the American ReligiousIdentification Survey of 200 I. This surveywas based on a random digit-dialed telephone survey of 50,281 American residential households in the 48 states of thecontinental U.S. Among these respondinghouseholds, 219 were identified as Muslim.This produced a weighted number ofI, 104,000 Muslim adults. "Allowing for asampling error of +/-0.5 percent. the ARIS200 I figure maybe adjusted upwards to itsmaximum range of 1.0 percent of all 208million American adults. With such anadjustment, the total national figure for USMuslims is 2.2 million, giving a total national population (including children) of justunder 3 million" (Kosmin and Mayer, 200 I,p. I) .


4/13

Several other national surveys have collected information on religious identification ofrespondents and included the category ofMuslim in the coded responses. Theseresults are shown in Table 1. drawing upondata made available by the American~ R e l i g i o u s Data Archive (http://www.thearda.corn). For each survey listed in the table.I have downloaded the data files and calculated the number of Muslim respondents. Allof these surveys are of adults (people aged18 and older) and so the assumption has tobe made that the population under age 18has the same representation of Muslims asdocs the adult population. Accepting this

92assumption. I have calculated the percentageof respondents in each survey who indicatedthat they were Muslim and then applied thatpercentage to the total U.S. population forthe year of the study. using the populationestimates of the U.S. Census Bureau. It canbe seen that all numbers hover close to onemillion. They are all within the same rangeas the ARIS. so if we accept the reasoningthat each survey may somewhat underestimate Muslims. we can accept the Kosminand Mayer suggestion of an upper limit ofapproximately 3 million Muslims as of theyear :WOO.

Table I. Estimates of the Size of the Muslim Population in the United StatesRased on National SurveyDataImplied S; ofMuslims Muslims in Hlackas fraction US Blacks who areSurvey Total Muslims of total population Muslims Muslim

Religion & Politics survey 1994-1995(weighted) 26.726 XX 0.00.3 X72.72.3 59 ') ')Religion & Politics Survey 1996(weighted) 4.150 21 0.005 1357.764 U .3.0

God and Society in North AmericaSurvey. 1996 .3.002 10 0.00.3 X9.3.X04 4 2.0Civic Involvement Survey 1997(run with weight I) .3.267 I I o.om lJ14,444 I 0.2GSS 1998 2.8.32 U 0.005 1.261.716 6 1.5GSS 2000 2.817 12 0.004 I.llJX.722 8 1.9American Religious IdentificationSurvey 2001 50.281 219 0.004 1.2.3lJ.XX6 59 1.4

AVERAGE I.7Source . Datacourtesy ofAmerican ReligIOUS DataArchive Ihttp://www.thearda.coml.Another study conducted in 200 I of theMuslim population was completed as partof a larger study of American congregationscalled "Faith Communities Today." whichwas coordinated by the Hartford Seminary'sInstitute for Religious Research. The projectinvolved surveying a congregational leaderat each of more than 30.000 congregationsof all major religious groups across thecountry. In this process the study identified1.20l) mosques in the United States and63 I of these were randomly selected to beincluded in the survey. Responses werereceived from 416 (66 percent) of those 631(Bagby. Perl. and Froehle. 200 I ). A leaderfrom each responding mosque provided esti-

mates of the number of people attendingeach Friday's Jum'ah prayer. The averageattendance was reported to be 292 peopleper mosque. which would imply that353.000 Muslims pray at a mosque in theU.S. each Friday. A multiplier of 5.56 wasthen somewhat arbitrarily applied to thisnumber to estimate the total number of people associated with a mosque (an average of1.(25). This implies that 2 million Muslimsare associated with a mosque even if onlyj X percent of those attend Friday prayer.Another somewhat arbitrary multiplier wasapplied to that number to estimate the totalnumber of Muslims. whether or not associated with a mosque. This number was


5/13

estimated to be 6-7 million. which theauthors called "reasonable" (Bagby. Perl.and Frochle. 200 I. p. 3) although it is ofcourse highly dependent upon the multipliers used.Although the estimate from the mosque studywas twice the highest value estimated fromthe ARIS. the mosque study found that about30 percent of people associated with mosques were converts and that most of theseindividuals were African-American. TheARIS estimated that 27 percent of Muslimswere black. so both of these studies implythat about 30 percent of Muslims are AfricanAmerican. Table I shows that the nationwidesurveys implied that an average of 1.7 percent of the African-American population wasMuslim. If we round that to 2 percent andcombine it with the estimate that 30 percentof the Muslim population is African-American. then once we know the number ofAfrican-Americans. we can solve the equation for the number of Muslims. Census 2000counted 34.7 million African-Americans, and

932 percent of that number would be 694.000.If that number is 30 percent of the Muslimpopulation. then the number of Muslimswould have to be 2.3 million. In order for theMuslim population to be larger than this, theneither African-Americans must account for asmaller fraction of all Muslims. or else amuch larger fraction of African-Americansmust be Muslim. The data currently available. as shown in Table I. do not provideevidence of either one of those possibilities.The highest percent Muslim among blacks asshown in Table I is 3 percent. If that were thecorrect number. and 30 percent of Muslimsare black, then the resulting Muslim population is 3.5 million. In order for the number ofMuslims to be 7 million (the top estimatederived from the mosque study), if weassume that 30 percent are black. then 6 percent of blacks must be Muslim. Alternatively.if it is true that 2 percent of blacks areMuslim. then blacks could represent no morethan 10 percent of a population of 7 millionMuslims.

2. GEOGRAPHIC DISTRIBUTION OF THE MUSLIM POPULATIONThe available survey data reviewed above(see also Smith. 20(1) suggest that theMuslim population of the United States isprobably around 3 million. None of theseestimates. however, provides enough information to tell us about the spatial patterningof the Muslim population within the UnitedStates. We need a much bigger database toaccomplish that task, and so I have turned tothe census data to provide proxies for theMuslim population at sufficient geographicdetail so that a spatial pattern can be discerned.Although others have used census data forthe purpose of estimating the number ofMuslims in the U.S. (see. for example.Stone. 1991). my analysis builds on thePublic Use Microdata Sample (PUMS) fromthe 1990 census. which no one previouslyhas utilized for this purpose. These data provide us with the opportunity to generate estimates of the spatial distribution of theMuslim population. The 1990 estimates atthe state level then provide a way to use

regression analysis to estimate the Muslimpopulation by state for 2000, using information that has just recently become availablefrom Census 2000.2.1. How can Census Data Be Used toEstimate the Muslim Population?We are in a the midst of a brief historicalwindow of opportunity when census data canbe used to help identify the Muslim population in the United States because it is stilltrue that most Muslims in the U.S. are eitherimmigrants or are residing in households ofimmigrants. Thus, despite the lack of a question about religion. we can make inferencesabout the "possibly Muslim" population by using information that is derived aboutancestry, country of birth, and language. Inanother generation. when most Muslims willhave been born in the United States, it willbecome more difficult to identify them fromthese kinds of census questions.Other researchers have used census data forthis purpose. as I have already noted, and my


6/13

use involves the same caveats offered by previous researchers: Not all people from predominantly Muslim countries are Muslim;not all people who speak the language spokenin predominantly countries are Muslim; noall people who share the ancestry of thosewho are Muslim are themselves Muslim; andsome people who are Muslim will not shareany of the characteristics of ancestry, Ianguage, or place of birth that are being used asproxies for being Muslim. With respect to theconcern that not all people who might seem tobe Muslim are necessarily Muslim, the use ofthe census data works on the "where there issmoke there is probably fire" theory. That isto say, the existence of a large Arab community, for example, in a particular part of theUnited States probably signals the existenceof a Muslim community even if we acknowledge that many Arabs in the United States arenot Muslim. The presence in a region of people who are Arab, along with people who areIndonesian, along with people who are fromIran, probably increases the likelihood thatthere witl be a substantial Muslim population,even if not all such people are Muslim.With respect to the existence of Muslimswho are not immigrants and do not share anyof the language. place of birth or ancestrycharacteristics with other Muslims, it turnsout that in the United States most such individuals are African-American, so the taskbecomes one of estimating the percentage ofa local African-American population thatmight be Muslim. The estimating percentagecomes from outside the census data and isbased initially on a national average, butthen is applied to the census data to derive anumber that will be added to the "possiblyMuslim" population identified through thecombination of language, ancestry, and placeof birth. The following sections describe themethodology in more detail.

t

--

2.2. The "Possibly Muslim" Population in1990 Derived from the PUMS dataThe estimating process begins with the 5 percent Public Use Microdata Sample (PUMS)from the 1990 census. This file contains all ofthe information collected for each member ofthe household from a 5 percent sample of allhouseholds enumerated in the 1990 census.These are responses to the "long-form" ques-

94tionnaire which was administered to a one insix sample (17 percent) of all households. sothe data represent nearly one in three of alllong-form questionnaires from the 1990 census. The geographic scale goes down to thelevel of the Public Use Microdata Area(PUMA) which is smaller than a state. butlarger than a census tract. and is designed tobe sufficiently large in area so that privacy ismaintained. but sufficiently small in area soas to provide the possibility of spatial analysis of the data. For this analysis. I have aggregated data at the state level and the data areweighted to reflect the total population fromwhich they were drawn.People were assigned to the "possiblyMuslim" category based on ancestry if theiranswer to the first or second ancestry question indicated a category that is typicallyassociated with Muslims. The 1990 census(long-form) asked "What is this person'sancestry or ethnic origin'?" I coded as havingpossible Muslim ancestry those personswhose ancestry (either Ancestry I or 2) wasfrom a predominately Muslim country (asdelimited inWeeks 19HH: and updated in Belt20(2). The ancestries included as "possiblyMuslim" included (in Census Bureau numbering order): Turkish Cypriot (019). Albanian (100). Azerbaijani (10 I), Turkestan](168), Bosnian (177), most North Africanand Southwest Asian ancestries (400 through499. with the exception of Israelis. Chaldeans. Armenians. Coptics. and a few othernon-Muslim ancestr ies). Nigerian Fulani(554). Nigerian Hausa (555). Somali (56H).Afghani (600). Bangladeshi (603). Pakistani(680), Indonesian (730), and Malaysian(770). In some instances. people respondedthat their ancestry was "Muslim" or "Islam."but the Census Bureau did not code thoseresponses separately. Instead. they weregiven a code of "998" which we coded inconjunction with the language question.Language was then examined as a potentialindex of Muslim identification. especially forthe immigrant population. The census asked"Does this person speak a language other thanEnglish at home'?" and if the answer was yes,a follow-up question asked "What is this language'?" The languages coded as being typically spoken by immigrants from predominantly Muslim countries included (in Census


7/13

Bureau numbering order): Yugoslav (649).Persian (6.'ib r, Afghan (657). Kurdish (65H).Kirghiz (bS7). Uzbek (6Sl). Azerbaijani(bl)O). Turkish (blll). Indonesian and Malaylanguages (n2 through 7.+ I). Arabic (777).Hausa (7S2). Somali (7HJ). Sudanese (7H.+).Swahili (701) and Fulani (794).The third characteristic that might identify aperson who is potentially Muslim is place ofbirth outside the United States. We chosethose countries estimated to have a Muslimmajority t Weeks. 19SH) for inclusion in thiscategory. The countries are predominantly inthe north of Africa. western Asia. and southeast Asia.Overall. this process identified 1.891.959people who might have beenMuslim. based ontheir possession of one or more these characteristics according to the 1990 PUMS data. Ofthis number. only 23 percent fit all threecategories of place of birth. ancestry, and language. while another 20 percent fit into two ofthe three categories. and 57 percent fit intoonly one of three categories. of which ancestrywas the most common (3H percent of all of the"possibly Muslim"). In the interest of settingan upper bound on the number of Muslims, Ihave included all of these individuals in thecategory of "Possibly Muslim." To this groupwe must now add the estimated number ofAfrican-Americans who are Muslim.2.J. Adding African-Americans to thePossibly Muslim PopulationThe only way to determine the likely numberof African-Americans who are Muslim is tousc survey results that cross-tabulate race byreligion. As already shown in Table I, thedata suggest that approximately 2 percent ofAfrican-Americans are Muslim. If weassume that this percentage has remainedstcadv over time. then in 1990 this wouldrepresent a total of 599,729of the 29,986,060blacks enumerated in the census that year.Only a small fraction of those people wouldhave already been included in our estimate ofthe possibly Muslim population since in thatgroup there were only 29.328 blacks whowere born in the United States and, of these.only 16.572 indicated that their ancestry was"Afro-American." If we assume no overlap,then adding 599.729 African-Americans tothe already estimated possibly Muslim popu-

95lation yields a total of 2A91.6HO Muslims inthe United States in 1990. Given the methodsof estimation that I have employed. thisalmost certainly represents an upper limit onthe number of Muslims in that year.While it may be relatively easy to estimate thetotal number of African-Americans who areMuslim. it is more complex to estimate thegeographic distribution of those individuals.It is unlikely that in every community twopercent of the African-American communityis Muslim. It is more likely that the presenceof a larger immigrant Muslim population willencourage conversion (or reversion as it isusually called within Islam), whereas a smaller immigrant population of Muslims willprobably be associated with a smaller numberof African-Americans who are Muslim. evenin the presence of an otherwise large AfricanAmerican community. I have used these twoconstraints - the size of the African-Americancommunity as enumerated in the census. andthe number of possibly Muslim people as estimated by the PUMS data - to estimate thestate-by-state distribution of African-Americans who are Muslim in 1990. If we assumethat 2 percent of African Americans are Muslim and that the total Muslim population is2,491,680, then African-Americans represented 24 percent of all Muslims in the UnitedStates in 1990. This percentage is on the lowend of the estimates assembled by Smith(200I), but that is largely because Smithassumes that there are fewer Muslims than theabove total would suggest. As the estimate ofthe total population of Muslims goes down,then the percentage of those people that areAfrican-American obviously increases in theabsence of any change in the assumptionabout the number of African-Americans whoare Muslim.Given the above percentages, the populationof African-American Muslims was constrained initially to be the smaller of either2 percent of the total state African-Americanpopulation or 24 percent of the total Muslimpopulation in the state (which involved multiplying the PUMS possibly Muslim total forthat state by 0.317).However. we wanted thetotal African-American Muslim populationto sum to 599,729.so the totals for each stateare controlled to that value. These numbersare then added to the PUMS estimate of the


8/13

96possibly Muslim population to produce thestate-level estimate of the number ofMuslims residing in that state in 1990. These

data. along with the implied percentage ofthe total population that IS Muslim. areshown in Table 2.T ~ b k 2 Estirnarcs of the Muslim Population by State: L'nitcd Stare-, 1l)l)O and 2000

P o s s l h l ~Muxlimtrom, Possihl) African- r L ' ~ r t . ' S S l o n Afncun-

Muvhm American TOlal cquauon AIllL'rI(Jn TOlal Uungc PercentSIJtl'

(rpm ]()40Pl:\IS MuslimsIQQO Muvhm-,!QQ(1 applied 10Cen-u-, 21)(j(1 Muxlim,21XXI Mu-lims21XXI from 19'-)0\0 ~ O O ( l Percentl h J n ~ t . ' I )! allMuxlirn- Llll ' J11UnOuotnm

New York 2:;2.704 X:;.4XO :nX.IX4 410..117 7')':;\1 49:;.X:;0 1:;7.oon 40.02 11.)Y; 210 ICalifornia 407YIO 00.D)9 47.1.629 40.1.122 :;').7.11 402.X:;3 - 10.77:; -2.2H 13.60',; 1.1.1(1Texav XX.:;98 41.9X:; 1.10':;8.1 21X.905 H44.1 282..14X 1:;I.7M 1/6.22 OW; 1120Michig.m 114.0XS .lX.619 IS2.704 IX5.


9/13

lA. Est imat ing the Muslim Populat ion byState for .WOOThe PL'\lS data for Census 2000 were notavailable at the time of this writing. but wecan use the SFJ (detailed long-form) data toderive est imates consistent with the 1990derivation. We do that by generating an ordinary least-squares regression model fromthe 1990 census data in which we predict the19l)O state estimates of the Muslim populanon from 1990 census variables on ancestry.language. and place of birth measured at thestate level. We then apply that regressionmodel to the Census 2000 variables onancestry. language. and place of birth measured at the state level in order to generateestimates of the "possibly Muslim" population in each state in the year 2000. Toaccomplish this task. we need comparablevariables available from both Census 2000and the 1990 at the state level. We also mustuse only a small number of predictor variables because the relatively small number ofstates (4R continental states) means that theregression model has to be parsimonious inits choice of variables. We also seek to haveat least one variable from each of the threecategories of characteristics - ancestry, language. and place of birth. Finally, we seek aset of variables that captures some of theregional diversity in the origin of Muslimimmigrants to the United States.Given the above considerations, threedependent variables were chosen as the variables used to predict the number of possiblyMuslim people in each s ta te in 1990:(I) the number of people who indicatedthat they were of Arab ancestry (meaningthat they responded 'Arab' or indicatedthat their ancestry was an Arab country)(:-\RAB_ANC):(2) the number of people who indicatedthat they spoke the Persian language athome (PERS), and(3) the number of people born in Pakistan(PAKCPOB).These three variables combined to predictthe number of possibly Muslim people ineach state with considerable precision.Actually. only two of the three variablesremained in the model . Because of the highcollinearity between the number of peopleof Arab ancestry and the number of persons

97speaking Persian (r .897), the latter variable dropped out of the model. leaving thetwo variables of people of Arab ancestry andthe number of people born in Pakis tan as thetwo predictors of the number of "possiblyMuslim" people.The overall R2 was .996. and there was onlyone outlier beyond 3 standard deviationunits in terms of the standard res iduals - thepossibly Muslim population of Texas wasoverpredicted by the regression model. Twostates. New York and Ohio. were slightlyunderpredicted, and each had a standardresidual that was greater than 2. but less than3. It was not clear from the data why thesestates were not more accurately predicted.The regression model that was then used topredict the number of possibly Muslim people in each state in 2000 was as follows:N_2000 = -339.849 + (ARAB_ANC 1.35)+ (pAKCPOB * 4.94). *Applying this regression to the same predictor variables drawn from Census 2000 generated an estimate of the number of possiblyMuslim persons by state for the year 2000,and these results are shown in Table 2. Thetotal number of "possibly Muslim" (exclusive of African-Americans) for 2000 was2,709,852. The estimates by state then provide input for the calculation of the numberof African-American Muslims in 2000 andultimately the estimate of the total Muslimpopulation by state in 2000.African-American Muslims in 2000 wereestimated in the same way as they had beenfor the year 1990. In 2000 the total U.S.African-American population was enumerated in the census to be 34,658,190. Thepopulation of African-American Muslimswas constrained ini tially to be the smaller ofeither 2 percent of the total state AfricanAmerican population or 20 percent of thetotal Muslim population in the state (whichinvolved multiplying the PUMS possiblyMuslim total for that state by 0.256). However. we want the total African-AmericanMuslim population to sum to 693,164, so thetotals for each state were controlled to thatvalue. These numbers were then added tothe regression-based estimate of the possibly Muslim population in 2000 to producethe state-level estimate of the number ofMuslims residing in that state in 2000. These

=


10/13

data. are shown in Table 2 where it can beseen that the total Muslim population in2000 was estimated by this method to be3.403.016. which is higher than. but stillgenerally in line with most other estimates

98of the Muslim population in the UnitedStates. Once again. given the methodologythat I employed. this number almost certainly represents the upper bound of the numberof Muslims.

3. DISCUSSION OF FINDINGSNew York had more Muslims than any otherstate as of 2000. reversing places withCalifornia which had been the home of moreMuslims than New York in 1990, but whichslipped to second in 2000. Nearly half a million Muslims were estimated to be living inNew York state. accounting for almost IS percent of the nation's Muslim population. TheARIS survey also found that New York wasthe state with the greatest number of Muslims(Kosrnin and Mayer, 200 I). although in thatstudy 24 percent of all Muslims were estimated to be in New York. That number seemsanecdotally to be too high. and the estimatesshown in Table I seem more reasonable interms of geographic distribution. On the otherhand. if we combine the states of New York.New Jersey and Connecticut, we haveaccounted for 22 percent of the Muslim population in the United States.New York's increase between 1990 and 2000could have come at California's expense.since the estimates show a slightly smallerpopulation in California in 2000 than in1990. whereas New York's populationincreased more than any other state.Nonetheless. California is estimated to have463.000 Muslims. nearly as many as in NewYork. Texas was a close second in terms ofthe growth in the Muslim population between1990 and :WOO and that increase of more than150.000 pushed it from fifth place in 1990 tothird place in 2000. ahead of Michigan andNew Jersey. which had been third and fourth,respectively. in 1990. These results have tobe tempered by the caution that Texas wasthe only state for which in 1990 the regression model significantly overstated the estimated Muslim population. so it is possiblethat growth in Texas was not quite as rapid asthese numbers show. However. if we controlfor that effect hy comparing the predictednumber of Muxlimx in 1990 in Texas with the

predicted number in :WOO. the difference isstill an increase of 130.000 and Texas is stillthe second fastest growing state with respectto the number of Muslims,Michigan and Illinois round out the top fivemost populous Muslim states which whencombined account for 49 percent ofMuslims in the United States. The states thatcomprise the remainder of the top ten are. inorder. Florida. New Jersey. Virginia. Ohioand Pennsylvania. These then states combine to account for 73 percent of Muslims.Notably. however. there are four states Massachusetts, Maryland. Georgia, andNorth Carolina - that experienced significant absolute increases in the number ofMuslims between 1990 and 2000 eventhough they are not (yet) among the top tenin terms of the total population of Muslims.The latter of these two states. Georgia andNorth Carolina. were in the top five in termsof the percentage change in populationbetween 1990 and :WOO. Along with Texas.Utah and Nevada the Muslim populationalso increased by more than 80 percent inthat intercensal period. This was substantially above the 37 percent increase in theentire Muslim population. which in tum wassubstantially higher than the 13 percentincrease in population size of the whole U.S.population during that period of time.The Muslim population was thus growingalmost three times as fast as the U.S. population. and that growth was geographicallyuneven. Figure I maps the population ofMuslims by state in 2000. showing proportionately the numbers by each state.Muslims are concentrated especially alongthe cast coast (42 percent are in states thatare bordered by the Atlantic ocean). and inthe Great Lakes region. with Texas andCalifornia looking somewhat like geographic outliers on the map.


11/13

Figure I. Number of Muslims by Slate: :WOO

.


12/13

Figure 2. Location Quotients of Muslims by State: 2000

4 . ~ .;-i.- . / :. .- _--- --.. ,.. " - -I

:----- --i\..__ , r.r-

, ] __ 1 \ .rr.. , L. . . . .i ; . ..,.- - .A ' . . . , . . J '

Location Quotient in 2000:' J 0.000 - 0.799 (Cold spots).0 .800 1 .199 (Expected) N 1.200 - 2.161 (Hot spots) +

100

the first part of the decade of the 1990s thatdid not affect Texas, and housing prices inTexa s have been consistently lower than inCalifornia, making Texas a more attractivelocation if salaries are otherwise comrnensu-

rate. Finally, the map shows that most states,especially those in the middle of the country,were "cold spot s," indicating that there werefewer Muslims than would be expected onthe basis of total population.

CONCLUSIONNo one can know for certain how manyMuslims there are in the United States, butall but one recent study suggest that thenumber is not currently very far above 3million. None of the previous studies wasable to provide estimates of the populationofMuslims at the state level. but I have doneso in this paper by combining results fromthe 1990 Public Use Microdata Sample withdetailed (SF3 ) data released in Jate 2002 bythe U.S. Census Bureau. [ have combineddata on ancestry, language, and 'place ofbirth to estimate the "possibl y Muslim"population. which is largely exclusive of theAfrican-American Muslim population. Thelatter group has been estimated from cen susdata. but based on survey data from which

one can derive the percentage of the blackpopulation that is Muslim, con strained bysurvey data suggesting the percentage of theMuslim population that is estimated to beAfrican-American . The overall numbers ofMuslims estimated by this method - 2.5 million in 1990 and 3.4 million in 2000 - areslightly higher than the results from surveydata, and suggest that the numbers for eachstate are reasonable. albeit probably maximum , representations of the actual numbersof Muslims in those states.These estimates provide the most quantifiable data thu s far produced of the geographic distribution of the Muslim population. Those researchers familiar with theU.S. Muslim population may not be sur-


13/13

prised to see the clusters of Muslims in theNew York-New Jersey area, the Washington,D.C., area and the upper Midwest-GreatLakes. Nor will they be surprised by thelarge numbers of Muslims in California andTexas. However. the numbers and details areof considerable interest. In particular, thesedata suggest that California has a very largeMuslim population. but it is probably

declining in size rather than growing. Issuesof interfaith relations can perhaps be projected from the fact that the Muslim population is growing in percentage terms mostquickly in parts of the "Bible Belt" (especially the states of Texas, Georgia, andNorth Carolina) and in the western mountainstates that are increasingly dominated byMormons, including Utah and Nevada.

REFERENCES

101

American Muslim Council (1991), Fact Sheet: U.S.Muslim Population, Washington , D.C .. AmericanMuslim Council.AUSTIN A.D. (1984) , African Muslims in AntebellumAmerica: A Sourcebook, New York, Garland PublishingCo.BAGBY I.. P. PERL, and B. FROEHLE (2001), TheMosque in America: A National Portrait. Washington,D.C., Council on American-Islamic Relations.BELT D. (2002), The World of Islam, NationalGeographic, (January).GHAYUR M.A . (1981), Muslims in the United States:Settlers and Visitors . The Annals of the AmericanAcademy of Political and Social Science, n? 454,pp. 150-163.GOLDSTEIN S. and B. KOSMIN (1991), Religiousand Ethnic Self-Identification in the United States1989-90: A Case Study of the Jewish Population.Presented at Population As sociation of America.Washington, D.C.HADDAD Y.Y. (1986a) , A Century o f Islam inAmerica. The American University, Washington . D.C.HADDAD Y,Y. (l986b) , Muslims in America, TheMuslim World. n 76, pp . 93-122 .KOSMIN B. (1991) , Research Report: The NationalSurvcv of Religious Identification 1989-90, CityUniversity of New York. Graduate School and Graduate Center. New York.KOSMIN B. and S. LACHMAN (1993), One Nation'Under God: Rel ig ion in Contemporary AmericanSociety. New York, Crown Trade Paperbacks.KOSMIN B. and E. MAYER (2001), Profile of the U.S.Muslim Population, Graduate Research Center of theCity University of New York. New York .

LEVTZION N. (1968), Muslims and Chiefs in WestAfrica, Oxford, Clarendon Press .LOVELL E.K. (1992), "A survey of the Arab-Muslimsin the United Slates and Canada", in Islam in NorthAmerica: A Sourcebook , edited by M. A. Koszegi andJ. G. Melton, New York, Garland Publishing, Inc .McCLOUD A.B. (1995), African American Islam, NewYork, Routledge.,NYANG S. (1992) , "Islam in the United States ofAmerica: A review of sources", in Islam in NorthAmerica: A Sourcebook, edited by M. A. Koszegi andJ. G. Mel ton, New York, Garland Publishing, Inc.NYANG S.S. (1999), Islam in the United States ofAmerica, Chicago, ABC International Group, Inc.OATLING R.N . (1988), Americans facing towardMecca, Time, May, 23 .PHILLIPP T. (1980), "Muslims", in The HarvardEncyclopedia of American Ethnic Groups, edited byS. Thernstrom . Cambridge , MA , The Belknap Press ofHarvard University Press.RASHID S. (1999), Islamic Inf luence in America:Struggle, Flight, Community, Journal o f MuslimMinority Affairs, vol . 19, pp, 7-31.SMITH T.W. (2001), Estimating the MuslimPopulation in the United States, The American JewishCommittee. New York.STONE C. (1991), "Estimate of Muslims Living inAmerica", in The Muslims of America, edited byY. Y. Haddad. New York, Oxford University Press.WEEKES, R.A. (1984), Muslim Peoples: A WorldEthnographic Survey, Westport , CT. Greenwood Press.WEEKS J.R. (1988) , The Demography of IslamicNations, Population Bulletin, vol. 43, pp. I-53. .-

Muslim Population US

Documents

Transcript of Muslim Population US