Article_Closing the Digital Divide

8/3/2019 Article_Closing the Digital Divide

1/21

LONG PAPER

Paul F. Cleary Glenn Pierce Eileen M. Trauth

Closing the digital divide: understanding racial, ethnic, social class, gender and geographic disparities in Internet use among school agechildren in the United States

Published online: 8 December 2005 Springer-Verlag 2005

Abstract The purpose of this study is to explore thedynamics underlying disparities in Internet use amongschool age children in the US. The analysis found that abroad range of demographic, geographic and economicfactors signicantly inuence Internet use among chil-dren. Signicantly, the availability of household com-puting resources and adult Internet users in thehousehold were most important in explaining disparitiesin use among children. To expand universal Internetaccess, future public policy should focus on providingsupport for in-home access; continued support forpublic access at out-of-home locations such as schools,and providing technical support, training and expertiseto school age children.

Keywords Universal Internet access Workplacecapital Household computing resources andexpertise School age children

1 Introduction

If a society is to continue to grow and remain viable forcontinuing socioeconomic change in the Information

age, its citizens must be conversant in computers andcommunication technologies. Children, in particular,on whose future most societies rest, must be able toaccess, understand, and use the computer and Internetin order to develop the skills necessary to succeedindividually and to contribute to future growth andprosperity of the Information society. More than ever,children need to acquire the basic skills of reading,writing and basic mathematics that comprise thebuilding blocks necessary in order to acquire good jobs.As the number of unskilled jobs in society continues toshrink, there is a growing urgency to make sure that allchildren have these core skills [38, p 18]. Many believethat computing technologies can play a signicant rolein helping society achieve this end in new ways, byaddressing the unique aspects of individual learners,although there are constraints such as the costs of

developing and distributing good educational software[38, p 18].

However, in spite of tremendous progress in the rateof Internet access across the USA, a digital divide stillexists, and in spite of a good deal of existing research onthe magnitude of the digital divide in society, much lessis known about the reasons for existing gaps in Internetuse, particularly among our children. Often, barriers toaccess exist that prevent many in society from accessingcomputers and the Internet. To address the questionabout how to provide access to all those unable to affordit, particularly among children, the nature of the dis-crepancy in Internet use must rst be understood more

clearly. It is critical that a society be able to identify andunderstand the factors that contribute to the disparitiesin Internet use among children, in order to formappropriate strategies to close them. The focus of thispaper is on gaining a better understanding of the mag-nitude of the digital divide among school age children, aswell as of the underlying factors that account for inter-group differences.

Understanding the dynamics of disparities in accessto computer technology and the Internet among chil-dren is important for three principal reasons: (1), the

This is an updated and expanded version of a paper presented bythe authors at the 30th Annual Telecommunications Policy Re-search Conference (TPRC) on September 30, 2002 in Alexandria,VA.

P. F. Cleary ( & )University College, Northeastern University,Boston, MA, USAE-mail: [email protected].: +1-617-788-0153Fax: +1-617-788-0125

G. PierceCollege of Criminal Justice, Northeastern University,Boston, MA, USA

E. M. TrauthThe Pennsylvania State University,University Park, Park, PA, USA

Univ Access Inf Soc (2006) 4: 354373DOI 10.1007/s10209-005-0001-0


2/21

effects on the education of children in the informationsociety; (2), the ability to secure future careers in theinformation economy; and (3), the central role inhelping the information society remain economicallycompetitive. In addition, because education is crucialto master the technology of modern communicationand information, access to new educational opportu-nities that teach the necessary skills required to be-come procient in the operational technologies of theinformation society must be made available to youngpeople very early in the education process. Researchhas shown that 3- and 4-year-old children who usecomputers with supporting activities that reinforce themajor objectives of the programs have signicantlygreater developmental gains when compared to chil-dren without computer experiences in similar class-rooms - gains in intelligence, nonverbal skills,structural knowledge, long-term memory, manualdexterity, verbal skills, problem solving, abstraction,and conceptual skills [ 21]. Moreover, as Mansellpoints out, in general, organizations around the worldbelieve that education is critically important to eco-nomic development. Ultimately, organizations tend tosuggest that knowledge gaps are the greatest barriersto economic development and they can be narrowedthrough increased attention to education provisionand skills development [29, p 414].

Disparities in access to the Internet and computertechnology may also have profound effects on careeroptions. Currently, computer skills are important formany jobs in the service and trade sectors and most jobsin the nance, technology, and manufacturing sectors.Computer, engineering, and information technology jobs were among the fastest growing occupations priorto the most recent recession and the rapid decline of thedot com industry, according to the occupational projections of the USA. Department of Labors Bureau of Labor Statistics (BLS, Employment Projections 1998 2008, 2000, March 14). Furthermore, according to theNational Telecommunications and InformationAdministration (NTIA) and the US Department of Education, the level of education and Internet usagecontinue to be highly correlated (NTIA, 1999, July,pp 68).

This study examines the issue of Internet use amongschool age children from a more expanded perspective.In addition to reviewing the inuence of socioeconomicfactors, the contribution of household computing re-sources and expertise and the potential impact of school-based Internet resources are also considered. The paperbuilds upon work conducted by NTIA and other agen-cies.

Specically, the analysis presented in the paperexamines the independent and combined inuences of individual and family socioeconomic characteristics,family resources, household computing resources andexpertise, and school resources on Internet use amongschool age children. The principal questions that thispaper attempts to address, are the following:

1. How are the demographic characteristics of children(i.e., gender and age) and the characteristics of theirhouseholds/families (i.e., gender, race, ethnicity andcitizenship status of household heads, geographiclocation, and household type) related to Internet useamong school age children? This question involvesexamining the inuence of an expanded set of indi-vidual and household characteristics, beyond thoseused in previous research, on Internet use amongchildren. Specically, the analysis is extended to in-clude demographic characteristics of the head of achilds household, as well as characteristics of thechild.

2. How are family characteristics and resources such aseducation, marital status, and family income, relatedto Internet use among school age children?

3. How does household computing resources andexpertise, in the form of computers available in thehome and in the form of adults in the home withcomputing expertise/experience (in the form of adultexpertise), relate to Internet use among school agechildren? Prior research did not examine the potentialrole of household computing resources and expertise.An essential question addressed in this study, iswhether these types of resources and expertise affectthe Internet use of school age children beyond morestandard measures.

4. How are public school computing resources relatedto Internet use among school age children? Thisquestion involves examining the inuence of identi-ed public school computing resources on Internetuse among children. In this regard, a central questionis whether school computing resources, provided tothe public through the states, can help to close thedigital divide for Internet use among school agechildren.

2 Literature review

Since the mid-1990s, the role of Internet as a principalinformation conduit and communications medium hasgrown rapidly in the United States. The rate of growth of household Internet access across the USA has increasedat a rate of 20% per year since 1998 (NTIA, 2002,February, p 10). Schement argues that full-edged accessto a democratic society in the Information Age of thetwenty rst century requires every home to be equippedwith at least a telephone and a computer with Internetaccess [36, p 8]. Although the rate of growth of Internetaccess in the US continues to rapidly increase, noteveryone shares this access to the same degree [24 - p.2].Current data show that Internet growth continues tooccur at an uneven rate throughout society (NTIA, 2002,February, pp 17). There are still signicant disparities insociety, including children, between those with access tothe Internet and those without access [ 2, 10, 14].

A view held by many is that Internet access is a basichuman right, and writers and decision makers have been

355


3/21

wrestling for some time with the question of whethereveryone in society should have access to the Internet.Sawhney, for example, believes that every person in theUSA has a right to these services by the mere virtue of being a citizen [35, p.378]. Blackman [ 6] claims that ac-cess to telecommunications services is a basic right of allcitizens, and is a basic element of the right to freedom of expression.

However, observed instances of inter-group dispari-ties throughout society run counter to this belief.Therefore, we begin by describing the principal factorsof, and the dynamics associated to, existing disparities inInternet access in society, particularly among school agechildren. This is followed by a discussion of underlyingbarriers to Internet access and use by children, includingthe cost of computing technology and education.

2.1 Existing disparities in Internet use in society

What are the major discrepancies in Internet use in theAmerican society and are they similar to those experi-enced among children? It is believed that disparities inInternet use among school age children follow a patternsimilar to that of adults, although Internet use across allage groups in society is highest among the older youth.It is well known that broad factors such as economics,fairness, and cost along with more specic factors suchas age, race, geographic location, education, income,and home computer access inuence the nature andscope of observed gaps in Internet access and use in theUSA (NTIA, 2002, February).

A 2002 NTIA study found that age is an importantfactor affecting Internet use. A signicant inverse rela-tionship between Internet use and age exists amongadults. A sharp decline in Internet use by age from 18 to49 years of age (64.5%), to only 37.1% for persons over50 was reported (NTIA, 2002, February, p. 28). Bycontrast, the study found very little difference in Internetuse among males and females, with 53.9% of males inthe study using the Internet compared with 53.8% of females (NTIA, 2002, February, p. 28).

Considerable progress in the rate of computer andInternet use by race has been reported to take place since1997. However, differences in computer and Internet usecontinue to exist across racial categories. Whites, AsianAmericans and Pacic Islanders continue to exhibithigher rates of Internet use than African Americans andHispanics (NTIA, 2002, February, p 21, 2627). Nearlytwo-thirds of Asian Americans, Pacic Islanders andWhites (about 60.0% each) reported that they use theInternet, while Internet use rates for both AfricanAmericans (39.8%) and Hispanics (31.6%) were signif-icantly lower, although their growth rates were faster(NTIA, 2002, February, p 21, 2627).

Household residence continues to be a signicantindicator of Internet use. Internet use rates amongpeople who live in non-central city urban locations werehighest (57.4%) while strong growth rates were reported

among those living in rural households (24%, annual-ized since 1998) and central city urban households (19%,annualized since 1998) (NTIA, 2002, February, p 20, 28 29).

Education and income continue to be major deter-minants of Internet use among individuals. Internet useremains strongly correlated with the level of educationand a huge difference in rates of Internet use by level of educational attainment is still observed. Those with abachelors degree or higher are more likely to be Internetusers than those with less than a high school education(83.7 and 12.8%, respectively).

Income as well, remains a critical factor of whetheran individual (or household) will use the Internet. Therecontinues to be a strong direct relationship between in-come and Internet use even though Internet use hasgrown considerably among people who live in lowerincome households. Among those living in householdsearning less than $15,000 per year, Internet use hasgrown from 9.2% in 1997 to 25% in 2001 (NTIA, 2002,February, p 11). Internet use growth is also faster amongthose with lower family incomes. Among those living inlower income households, where family income is lessthan $15,000 per year, Internet use grew by 25% from1998 to 2001, while Internet use among those living inhouseholds with family incomes of $75,000 per yeargrew at a more modest 11% per year growth rate(NTIA, 2002, February, p 12).

Finally, although children show increasing rates of Internet use (NTIA, 2002, February, p 1), the questionof Internet access remains an important issue. Childrenwithout any Internet access or greatly limited accessrelative to their peers may be at a serious disadvantagein American society. Children without Internet accesscould be at immediate risk in terms of their educationalprogress as well as at future risk in terms of potentialfuture job opportunities. This suggests that apart fromthe potential cost, wider public access to the Internetmay be necessary, especially for children, if everyone isto share in the countrys future prosperity. In addition towiring schools, libraries, and community centers, asmandated by the Telecommunications Act of 1996, thecontributions of individual state technology policies andthe availability of Community Access Centers (CACs)across the USA may signicantly contribute toimproving the current situation. Kling asserts thatcommunity access centers may help ll the gap in accessby providing technical assistance that increases socialaccess to computers and the Internet [27, p5]. Further-more, because the impact of information technology oneducation is so signicant, the role that individual statepolicies can have in helping to promote equal access andnarrow the digital gap becomes critical [8, pp 226228].

The good news is that the problem is somewhatmitigated because many children in society have becomequite accustomed to computers and Internet use over thepast several years. Lloyd asserts that children of theI-Generation (Peach 1997) or Nintendo Genera-tion(Kenway 1995) have grown up in the information

356


4/21

age and are becoming regular and accomplished users of the Internet and other communication technologies.Some authors assert that childhood, as we have knownit, is currently in transition and has been dramaticallychanged in the twenty rst century because of the con-stant barrage of media information [28, p 1].

2.2 Internet access and education

The use of the Internet in education is now widely be-lieved to be of value to school age children and, there-fore, access and use of Internet technology is becomingincreasingly more important to education. Solowayet al. [39, p 19] contends that computer technology andWeb-based educational programs can cost-effectivelyhelp children with reading and math in ways that reectthe growing diversity of students in society [38, p 11, 15].He also asserts that the Web is supportive of learning,especially among middle age teens in areas such asteamwork [39, p 21]. Moschella [ 32] strongly argues thatunless all students are sufficiently exposed to the Internettoday, those without exposure will be at a serious dis-advantage in the job market of the future. Igbaria [ 23]discusses the need to expose individuals to technology atan early age in order to provide the opportunity to buildknowledge and to be adaptive to change. Mengel [ 30]cites other advantages of Internet exposure for children,such as increased interaction with a wide segment of students and the availability of wide ranging informa-tion on a multitude of topics. Still others argue that theInternet, along with computers, has the capability of revitalizing American education by transforming thetypical classroom from a teacher-directed to a student-centered environment [ 32].

As Internet use has become an increasingly moreimportant educational tool, educational programing hasbecome increasingly more adaptable to both the leveland developmental capabilities of a child, regardless of age. Studies have shown that Internet use can be suc-cessfully integrated into a variety of educational settings[5, p 9]. According to Baumgarten, Internet software isnow available for young children 69 years of age and1014 years of age, and can be integrated into educa-tional curricula regardless of the physical, cognitive andpsychosocial capabilities of the child. As their readingability increases, children become more able and there-fore, more likely to take advantage of existing Internetprograms targeted to their specic level. As children age,they can begin to manipulate the content of Internetprograms in ways that express creativity, independence,social awareness and programs that require logic, stra-tegic and abstract thinking [5, p 9]. School curriculafrom elementary school through high school now regu-larly incorporate the use of computers and the Internetfor educational instruction in science, social studies andother subject areas.

In addition, as the penetration rate of the Internet inthe household continues to increase, data are beginning

to appear that more clearly demonstrate that close linksbetween Internet use and the level of academic perfor-mance in school exist. Attewell and Battle stress theimportance to children of helping them raise the level of performance in school. They make a strong case for aclose positive relationship between access to homecomputing and the academic performance of children inschool and argue that lack of access could worsen cur-rent social inequality and widen the gap between theaffluent and the poor [4, p 1]. Attewell further citesGiacquinta et al. (1993), who early on examined the useof home computing resources and found that the pri-mary use was for playing games and, in the rare case fortheir use in education, that signicant parental hands onuse was required.

Internet access in the schools has become a criticaltechnological requirement to the educational system.However, a key issue that remains to be addressed ishow to deliver that access. One approach has centeredon the federal governments efforts to subsidize theconnection of all schools and libraries to the Internetthrough the e-learning program. Another has focused onthe perceived need for more computers in the classroom,and the educational value of having home-based accessto both PCs and the Internet [ 32]. It appears that at thistime, however, there is no longer any real debate on themerits of Internet access and use in the educationalsystem. Wiring of the nations public schools continuesand there is wide recognition of the value of Internetaccess in education [ 17]. However, it is still unclear howmuch training and expertise are required to help stu-dents and teachers achieve an optimal educational use of computer technology and the Internet.

2.3 Internet access and costs

Cost is another important aspect of Internet access anduse for children. To begin with, Internet use is a functionof the cost of the required computer equipment, but it ismuch more complex and costly than earlier communi-cation technologies such as the telephone. For mostInternet users, these expenses generally include the costof computer hardware, software, Internet connectivity,selection of an Internet service provider (ISP) and nallythe cost of training. These different aspects of Internetcosts closely conform to the concept of the total cost of ownership (TCO) of technology, an idea that wasdeveloped to provide a more accurate estimate of theactual cost of computer technology for business enter-prises.

According to the Gartner Group, the total cost of ownership (TCO) [ 20] is signicantly higher than thetotal cost of simply purchasing technology hardware orservices. Gartners denition of the total cost of own-ership for the K-12 environment consists of direct andindirect costs. Direct costs includes the cost of thecomputer hardware and software, hardware and soft-ware upgrades, maintenance, technical support, and

357


5/21

training (formal and informal) while indirect costs con-sist of peer support, le and data management, anddowntime [34, pp 45]. According to this denition, theprincipal determinants of the TCO of Internet use wouldnot fall in the area of service (i.e., connectivity), butrather in the areas of computer hardware and softwareownership, computer expertise, and nally the cost of training. Yet, as equipment costs decline, these costsbecome a proportionately smaller share of the total costthan other equally critical support costs such as exper-tise, training and maintenance. Kling points out,through industry surveys conducted in the late 1990s,that equipment costs were really a small part of TCO. Inaddition, the costs of acquiring the necessary expertiseand training to use computer technology can be veryhigh. He suggested that over time, support costs, whichinclude training, maintaining equipment, upgrading andconguration, generally become the major costs in theutilization of the technology [27, pp 56]. Estimates varybut many indicate that the TCO is about three to fourtimes more expensive than the cost of computer hard-ware alone (http://www.webopedia.com/TERM/T/TCO.html).

According to industry analysis cost models developedin 1997 independently by the Gartner Group and For-ester Research, researchers were able to document theimpact of expertise and training on overall costs. Gart-ner estimated that 45% of the annual industry cost of aPC was attributable to end-users. Over one-third (34%)of the cost to end-users was due to formal and informallearning and to peer support. Forrester found that PCtraining accounts for 12% of the overall annual cost of PCs (http://web.ics.purdue.edu/ $ kimfong/TCO/mod-els.html). Furthermore, when applied to K-12 education,Gartner estimates that when direct and indirect costs areconsidered, 60% of the overall TCO can be attributed toindirect costs [34, pp 45]. Consequently, when con-nectivity and both formal and informal training costsare taken into account, it is estimated that the TCO of computer and Internet hardware and software is morethan 12 times more expensive than the telephone.

Access to computer resources and the Internet from alocation outside the home, e.g., from public school, alibrary, or a CAC, is generally less costly. Computerresources at such locations also generally involve aminimal fee or are free for individuals. Sometimes thereis expertise and peer support available to assist at nocost as well. However, access to telecommunicationsservices is only one aspect of access to this technology.When the total cost of using computer technology isconsidered, it becomes apparent that the expertise andtraining required to use this technology is equallyimportant as simply purchasing or having access tocomputer and Internet service.

Although private rms and institutions can absorbthe additional costs of both formal and informal Inter-net expertise and make it available within theirorganizations, individual households are generally less

able to do so. Consequently, unless a household memberhas the available expertise to assist a child with aneducational task on the Internet, it is unlikely that thechild will be able to make maximum use of the tech-nology for educational purposes.

This study explores the notions presented in thissection and builds upon work begun by Attewell, Battle,Birenbaum, Giacquinta and others. Specically, thepaper examines Internet use among school age childrenanywhere, both at home and outside the home.

3 Methodologic approach

The present study draws on data from the November2002 Current Population Survey (CPS) to examineInternet use among school age children. The currentresearch expands beyond the work of previous studiesby focusing on school age children, by incorporatingadditional demographic factors not examined in previ-ous research (i.e., citizenship status), and by incorpo-rating factors related to training and expertise resourcespotentially available to school age children. Specically,the degree to which Internet use among school agechildren varies by individual and family is examined,along with the extent to which social and economiccharacteristics can be explained by the availability of household computing resources and expertise. It hasalso examined the degree to which school computingresources and expertise affect variation in Internet useamong school age children by individual and family,social and economic characteristics, and by the avail-ability of household computing resources and expertise.The methodology used in this paper is structured inthree parts: sources of data, variables used for theanalysis, and the statistical approach used for analysis,which are described in the subsequent subsections.

3.1 Sources of data

The primary data used in this study is the CPS forSeptember of 2001 administered through the USADepartment of Commerce, Bureau of the Census on amonthly basis. In the September 2001 CPS survey, theCensus Bureau interviewed approximately 57,000 ran-domly selected sample households and administered aComputer and Internet Use Supplement. As with eachmonthly CPS survey, the national sample is selectedfrom the 1990 decennial census les, which cover all ftystates and the District of Columbia. The sample iscontinually updated to account for new residentialconstruction (NTIA, 1999, July, pp xvxvi). The dataused to examine Internet use among school age childrenconsisted of 25,960 reported cases of school age childrenbetween the ages of 6 and 17. In addition to the CPSsurvey, a second source of data on available publicschool Internet resources was obtained from the 2001

358


6/21

nationwide school survey conducted by Education Weekin collaboration with the Milken Exchange on educationpolicy and the U.S.A. Department of Education. Thissurvey provided a measure of the average level of Internet resources available to school age children. Thissurvey unfortunately provides a somewhat imprecisemeasure of public school resources because it is a statelevel indicator (i.e., average number of students perInternet connections for a state). Obviously, it would bepreferable to have a more precise measure (e.g., thenumber students per Internet connection for eachstudents own school), but this was not available.

3.2 Analytic approach

The analysis examines the potential effect of ve majorcategories of explanatory variables on Internet use bychildren aged 6 to 17. The specic categories included inthe analysis are:(1) family and child background char-acteristics (e.g., gender, age, race, ethnicity, citizenshipstatus, geographic location, living quarters), (2) familystructure and educational resources (i.e., marital statusand education level of the reference person for thehousehold 1), (3) family nancial resources,(4) householdcomputing resources and expertise (i.e., the presence of acomputer at home, Internet use among adults at homeand Internet use among adults outside the home), and(5) public school Internet resources (i.e., the number of students per Internet connection). The last two catego-ries of explanatory variables were not examined in pre-vious research on Internet use among children.

The analysis rst examines the bivariate relationshipbetween explanatory variables and Internet use amongschool age children. In the second phase of the analysis,multivariate logistic regression analysis is used toexamine the potential independent effect of each of theexplanatory variables controlling for the other variables.The variables enter the analysis in terms of theirapproximate temporal relation to the dependent variable(i.e. Internet use by school age children). Thus familyand child background characteristics, such as race andage, enter the multivariate analysis rst, and this allowsthe effect of these variables to be assessed before othercausally subsequent explanatory variables enter into theanalysis. In an effort to be conservative in the assessmentof the potential effect of the impact of computing re-sources and expertise on Internet use by school agechildren, these factors are examined only after family

and child background characteristics, family structure,educational resources, and nancial resources enter intothe analysis.

3.3 Variables used for the analysis

3.3.1 Family and child background characteristics

The rst set of exogenous/explanatory variables usedin this analysis includes a variable for the gender of each child in the study. Also included is a dichoto-mous variable for the gender household referenceperson. The age of children in the study is a variable,which is coded as 1=611 years of age, 2=12 13 years of age, and 3=1417 years of age. The raceand ethnicity of the household reference person isclassied into four categories, 1=white, non-Hispanic,2=Black, non-Hispanic, 3=Hispanic, and 4=othernon-Hispanic racial categories. We also include thecitizenship status of the household reference person as

an explanatory variable, which was classied as1=natural born citizen, 2= naturalized citizen, and3= not a citizen. Geographic location of the house-hold residence is included as an explanatory variable.For this purpose, the USA Census classication(available in the CPS survey data) is used, whichgroups locations into four geographic categories:1=central city, 2=balance of central city, 3=non-metropolitan areas and 4=not identied. Finally, avariable is also included that classies the residentialliving quarters of survey respondents into three cate-gories, 1=owned by a household member, 2=rentedfor cash, 3=occupied without payment of cash rent.

3.3.2 Family marital status and education

The second set of exogenous/explanatory variables in-cludes the martial status and the education level of thehousehold reference person, coded as 1.0=married,2.0=once married, and 3.0=never married. The edu-cational level of the household reference person is codedinto the categories 0=16 years of education, 1.0=7 9 years of education, 2.0=1012 years of educationwith no high school degree, 3.0=high school degree,4.0=some college, 5.0=a college degree, and6.0=masters degree or higher.

3.3.3 Family nancial resourcesfamily income

The total income reported by families in the CPS surveyis used to measure family nancial resources. The familyincome variable is coded as 1.0=under $5,000 per year,2.0=$59,999 per year, 3.0=$1019,999 per year,4.0=$2034,999 per year, 5.0=$3549,999 per year,6.0=$5075,999 per year and 7.0=$75,000 and aboveper year.

1The US Census Bureaus denition of household reference personis household person number 1 in Census surveys and refers tothe person (or one of the people) in whose name the housing unit isowned or rented (maintained) or, if there is no such person, anyadult member, excluding roomers, boarders, or paid employees. If the house is owned or rented jointly by a married couple, thehouseholder may be either the husband or the wife. The persondesignated as the householder is the reference person to whomthe relationship of all other household members, if any, is recordedreference person. (http://www.census.gov/population/www/cps/cpsdef.html)

359


7/21

3.3.4 Household computing resources and expertise

Measures of household computing resources andexpertise are included as explanatory variables to studythe potential importance of available household exper-tise (by adults in a household) on Internet use by schoolage children. A measure of computing resources is pro-vided by information on computers in the householdand is coded into a dichotomous variable with 0=nocomputers in the household and 1.0= one or morecomputers in the household. Measures of available adultcomputing and Internet expertise in the household weredeveloped by aggregating the number of adults (i.e., anyperson 18 years of age and older) in the household thatused the Internet at home or the number of adults thatused the Internet outside the home. From this infor-mation two measures of potential adult Internet exper-tise are developed. The rst is a count of the number of adults in the household that used the Internet at homecoded as 0 = no adults, 1= one adult, and 2= two ormore adults. The second is a count of the number of adults in the household that used the Internet outsidethe home (e.g., at work, school, etc.) coded as 0 = noadults, 1= one adult, and 2= two or more adults.

3.3.5 School Internet resources

The nal resource related variable included in theanalysis is the average number of students per Internetconnection in public schools within a given state. Al-though this attribute is measured at a level of aggrega-tion that is too high, the variable provides a grossmeasure of the level of Internet resources available topublic school students within a given state. A muchbetter measure of available school resource would havebeen the average number of students per Internet con-nection within a given school, but such data were notavailable. Nevertheless, the state level variable availableto this study provides a rst cut estimate of the potentialimpact differences in school computing resources onstudents use of the Internet. The average number of students per Internet connection in public schools withina state ranged from a high state resource level of 3.4students per Internet connection to a low level of 10.1students per Internet connection.

3.4 Statistical approach

The analysis rst examines the bivariate relationshipsbetween Internet use among school age children any-where and the independent explanatory factors that arepotential determinants of Internet use. Internet use isdened as school age child using the Internet eitherwithin or outside the home, and is coded as 0 = no useof the Internet or 1 = use of the Internet at home,outside the home or both. Tables 1, 2, 3, 4, 5, 6, 7, 8, 9,10, 11, 12, 13 and 14 present the basic cross-tabular

analyses, and Chi square tests provide tests of the sta-tistical signicance of the bivariate relationships.

Following the bivariate analysis, the unique contri-butions of each of the explanatory variables on Internetuse among school age children is then assessedemploying logistic regression analysis. Logistic regres-sion models estimate the average effect of each inde-pendent explanatory variable on the odds that adependent variable outcome will occur (school age childwill use the Internet at any location).

In the present study, the odds ratio is the ratio of theprobability of Internet use to the probability of non-Internet use. It is the probability that a child uses theInternet anywhere divided by the probability that a childdoes not use it. For example, if a childs likelihood of using the Internet anywhere is 0.75 (P), then the prob-ability of non-use is 0.25 (1-P). The odds ratio in thisexample is 0.75/.25 or 3 to 1. The odds of Internet useanywhere in this case are 3 to 1.

In the model adopted in this study, the dependentvariable is the natural logarithm of the odds ratio, y, of a school age child using the Internet anywhere. Thus, y= p / 1 p and;

ln y ^

a 0 X ^

a i ni

where^

a 0 is an intercept,^

a i are the i coefficients for the iindependent variables, X is the matrix of observationson the independent variables, and n i is the error term.

To make the B coefficients more understandable, theyare typically converted into adjusted odds ratios, cal-culated by exponentiating the coefficient of interest. Anodds ratio measures the change in the odds that an eventwill occur for each unit change in a given independentvariable. When the variable is dichotomous, the oddsratio measures the change in the odds that is due tobelonging to the selected category versus the comparisoncategory. The adjusted odds ratio is an estimate of theodds ratio independent of (or after accounting for) othervariables [13, p 64].

When interpreting odds ratios, an odds ratio of 1,means that someone with that specic characteristic is just as likely to exhibit the characteristic as not. An oddsratio of exactly 1.0 indicates that the likelihood of achild using the Internet anywhere changed by a factor of 1.0, or not at all. Odds ratios greater than one indicate ahigher likelihood of Internet use among school agechildren that have a positive value for that particularindependent variable. When the independent variable iscontinuous, the odds ratio indicates the increase in theodds of Internet use for each unitary increase in thepredictor.

Finally, in order to insure variable integrity, a seriesof multicollinearity diagnostic tests were run on theexogenous variable attributes in this model to determinewhether there was a presence of multicollinearity. Testresults indicated that there was virtually no multicol-linearity found among the exogenous variables in themodel. The variance ination factor (VIF) for each of

360


8/21

the exogenous variables appearing in the model wasfound to be well within acceptable limits. Various textsprovide slightly different thresholds to the presence of multicollinearity in a regression model, but in general, if the VIF is 5.0 or greater, some degree of mulitcollin-earity exists [1, p 573]. Among the exogenous variablesuses in this model, the calculated VIFs attached to eachranged from a low of 1.001 for the variable gender of child (PESEX2) to a high of 3.285 for the variable familyincome with missing values (RFAM1_NM). All werewell within the threshold, indicating that multicollin-earity was not present.

4 Analysis

The analysis is organized into two parts. The rst partconcerns the bivariate analysis of internet use amongschool age children, and the second is a multivariatelogistic regression analysis of Internet use. In both parts,the analysis proceeds in terms of the presumed relativetemporal order of the explanatory variables relative tothe dependent variable, Internet use anywhere by schoolage children.

4.1 Bivariate analysis of Internet use among schoolage children

4.1.1 Internet use by family and child Background characteristics

Examination of family and child background charac-teristics found that Internet use among school age chil-dren was essentially the same for boys and girls, 62.2%for males vs. 63.7% for females (see Table 1). Thus,slightly more girls than boys indicated that they use theInternet anywhere. In contrast and in terms of the gen-der of the household reference person, Internet use

among school age children anywhere was found to besignicantly higher in households where the householdreference person was a male. As shown in Table 2,Internet use among school age children is higher in male-headed households than in female-headed households(66.6 vs. 59.1%, respectively). The fact that householdswith male reference persons are more likely to be twoparent households where nancial resources areprovided by two adults rather than one as in manyfemale-headed households may help account for thisassociation.

The age of school age children is an importantdeterminant of Internet use. As Table 3 shows, Internetuse among school age children increases directly withage, with less than half (47.1%) of children 611 years of age using the Internet to nearly four-fths (77.8%) usingthe Internet among children 1417 years of age. Thisappears reasonable since, as children age, their literacyskills and knowledge tend to increase as well.

Attributes of the reference persons of householdssuch as race and ethnicity are important predictors of disparities in Internet use among school age children. AsTable 4 shows, Internet use among school age childrenvaries widely by race and ethnicity. Nearly three-fourths

(70.3%) of the children residing in households withwhite non-Hispanic reference persons use the Internetanywhere while less than half (44.5%) of the childrenresiding in households with Black non-Hispanic refer-ence persons use the Internet anywhere. For householdswith Hispanic reference persons only 40.9% of theschool age children use the Internet. Finally, as shown inTable 4, nearly two-thirds (61.0%) of the children inhouseholds with other non-Hispanic reference personsuse the Internet.

The citizenship status of the household referenceperson of school age children is also a signicantpredictor of Internet use among school age children.

Table 5 shows that Internet use among school age chil-dren was highest among natural born citizens wherenearly two-thirds (65.3%) of the school age children inthese households reported that they used the Internet. Incontrast, only 40.5% of children from households wherethe reference person was not a citizen reported using theInternet. Over half (56.8%) of children from householdswhose reference person is a naturalized citizen use theInternet.

Geographic location and residential ownership arealso associated with Internet use among school agechildren. Table 6 shows that slightly over one-half (53.3%) of children who reside in central cities use the

Internet while nearly two-thirds of children who residein other locations outside the central city use the Internetanywhere. Specically, 65.7% of children in metropoli-tan non-central city areas, 65.1% of children innon-metropolitan areas, and 65.2% of the children inundened areas use the Internet.

Finally, Table 7 shows that home ownership is alsoassociated with Internet use among school age children.Internet use is highest among children who live inhouseholds owned by a household member with overtwo-thirds (68.0%) of the children in these households

Table 1 Internet use anywhereby gender of the child

Pearson Chi Square 6.584,P =0.010

Internet use any where Gender of Child Total

male female

Total % yes 62.2 63.7 62.9Count (13,346) (12,614) (25,960)% 100.0 100.0 100.0

361


9/21

using the Internet while only about one-half (48.4%) of children living in housing that is rented, report using theInternet.

4.1.2 Internet use and family marital status and education

The marital status of the household reference person is asignicant predictor of Internet use among school agechildren. As Table 8 shows that Internet use is highest

among school age children residing in households wherethe household reference person is married (66.8%),while signicantly less than half (41.5%) of the childrenreported using the Internet in households where thereference person was never married. Reported Internetuse among school age children from households whosereference person is either separated or divorced is58.2%.

Internet use among school age children also variesdramatically with the education of the household refer-ence person. As the level of education of the household

reference person increases, Internet use among schoolage children increases as well. As Table 9 shows, inhouseholds where the highest level of education isgrammar school, slightly more than one-quarter (27.8%)of the children use the Internet, while over four-fths(80.9%) of the children who reside in households wherethe highest level of education is a masters degree orhigher, report using the Internet anywhere.

4.1.3 Internet use and nancial resourcesfamily incomeFamily resources as measured by family income arestrong determinants of Internet use among school agechildren. Internet use among school age children variesdirectly with the income of the family. As shown inTable 10, Internet use among school age children in-creases steadily from 34.0% in households where thehousehold income is less than $5,000 per year, to four-fths (80.9%) of the children who reside in householdswhere family income is $75,000 per year or more.

Table 2 Internet use anywhereby gender of householdreference person

Pearson Chi Square 154.769,P = 0.000

Internet use any where Gender of householdreference person

Total

Male Female

Total % yes 66.6 59.1 63.1Count (13,484) (12,172) (25,656)% 100.0 100.0 100.0

Table 3 Internet use anywhereby age of child


Internet use any where Age of Child (years) Total

611 1213 1417

Total % yes 47.1 69.6 77.8 62.9Count (10,776) (6,665) (8,519) (25,960)% 100.0 100.0 100.0 100.0

Table 4 Internet use anywhereby race/ethnicity of householdreference person


Internet use any where Race/ethnicity of household reference person Total

Whitenon-Hispanic

Blacknon-Hispanic

Hispanic Othernon-Hispanic

Total % yes 70.3 44.5 40.9 61.0 63.1Count (18,016) (3,052) (3,141) (1,447) (25,656)% 100.0 100.0 100.0 100.0 100.0

Table 5 Internet use anywhereby citizenship of householdreference person


Internet use any where Citizenship status of household reference person Total

Natural born citizen Naturalized citizen Not a citizen

Total % yes 65.3 56.8 40.5 63.1Count (22,445) (1,346) (1,865) (25,650)% 100.0 100.0 100.0 100.0

362


10/21

4.1.4 Internet use and household computing resourcesand expertise

Household computing resources and expertise arepotentially among the most important determinants of Internet use among school age children. As Table 11shows, the availability of a computer in the home iscritical to the use of Internet among children. Nearlythree-quarters (74.6%) of all school age children residingin households with computers reported using the Inter-net, while only slightly more than one-fourth (28.4%) of children residing in households without a home com-puter reported using the Internet anywhere.

The presence of adult(s) that use Internet in thehousehold is also an important predictor of Internet useamong school age children. Household computing

expertise is measured by counting the number of adultsthat use the Internet within their household and bycounting the number of adults that use the Internetoutside of their household (e.g., at work, school, etc.).As Table 12 shows, Internet use among school agechildren anywhere increases as the number of adults inthe household who the Internet at home increases. Overfour-fths (81.8%) of children in households with two ormore adults report using the Internet anywhere, while inhouseholds with no adult, uses the Internet at home,only slightly over one-third (36.0%) of the children re-port using the Internet anywhere.

Finally, the presence of adult(s) that use the Internet

outside the household is also an important predictor of Internet use among school age children. As Table 13

shows, Internet use among school age children anywhereincreases from less than one-half (47.9%) in householdswhere no adults use Internet outside the home, to overthree-fourths (81.9%) in households where two or moreadults use it outside the home. These ndings suggestthat use of the Internet by adults regardless of locationmay be a signicant predictor of Internet use amongchildren. This may occur because, as more adults in ahousehold use the Internet, the more exposed childrenliving in these households get the knowledge of how touse the Internet and to the potential value of the Internetto their success.

4.1.5 Internet use and public school internet resources

Finally, public computing resources may be a critical

resource for children, especially in cases where there areno family or household Internet resources in thehousehold. The average number of students per Internetconnection per state provides a measure of the relativeavailability of public school computing resources forschool age children. For the purposes of the tabularanalysis, the variable was recoded into ve quintilecategories. The values for the recoded version rangefrom 1=3.45.6 students per Internet connection to2=5.76.3 students, 3=6.46.9 students, 4=7.18.3students, and 5=8.410.1 students. Table 14 indicatesthat public school Internet resources show a relativelymodest effect on Internet use among children any-

where, although, as noted, this affect may be somewhatattenuated in part because this measure is based on

Table 7 Internet use anywhereby living quarters status


Internet use any where Living quarters status Total

Owned byhh member

Rented forcash

Occup. w/lcash rent

Total % yes 68.0 48.4 59.6 62.9Count (19,031) (6,585) (344) (25,960)% 100.0 100.0 100.0 100.0

Table 8 Internet use anywhereby marital status of householdreference person


Internet use any where Mari tal status of household reference person Total

Married Divorced-separated Never married

Total % yes 66.8 58.2 41.5 63.1Count (18,674) (4,887) (2,095) (25,656)% 100.0 100.0 100.0 100.0

Table 6 Internet use anywhereby central city status


Internet use any where Central city status Total

Centralcity

Balance of metro area

Non-metroarea

Not identied

Total % yes 53.3 65.7 65.1 65.2 62.9Count (5,322) (9,854) (6,468) (4,316) (25,960)% 100.0 100.0 100.0 100.0 100.0

363


11/21

aggregated state level data. As Table 14 shows, Internetuse among school age children is slightly higher (67.0%)in states with the lowest of the average number of stu-dents per Internet connection in public schools (i.e.,quintile 1) vs states with the highest average number of students per Internet connection (i.e., quintile 5) where56.3% of school age children report using the Internet.This measure should be explored in greater detail infuture studies. Data on students per Internet connectioncollected at the school district or individual school levelwould provide better estimates of the potential effect of school resources on student Internet use.

4.2 Multivariate analysis of Internet use amongschool age children

To examine whether the attributes examined in thebivariate analysis remained signicant predictors con-trolling for the effects of the other factors, a multivariateanalysis of the independent effect of each of theexplanatory variables on Internet use by school agechildren was conducted. As with the bivariate analysis,the potential effect on Internet use among each of theve sets or blocks of explanatory variables is assessed.The ve blocks are (1) family and child backgroundcharacteristics, (2) family marital status and education,

(3) nancial resources family income, (4) householdcomputing resources and expertise, and (5) schoolInternet resources.

Each block of explanatory variables enter the analysisaccording to its presumed temporal position relative tothe dependent variable and the other independent vari-ables. The independent effects of individual explanatoryvariables are rst assessed at the temporal stage whenthey enter the analysis. Importantly, this enables theresearchers to examine the mediating effects of individ-ual blocks of explanatory variables introduced into theanalysis at earlier stages.

Table 15 lists the specic explanatory variables in-cluded in the logistic analysis, along with their variabledenitions. The explanatory variables are coded toconform to the requirements of the logistic regressionanalysis. With the exception of family income, most of the explanatory variables in the study have very little(1.2% or less) missing data. In case of family income,however, 12.4% of the observations have missinginformation. There are a variety of techniques availableto researchers that deal with the problem of missing datain multivariate analyses (see Little and Rubin 1987 for afull discussion). For the purpose of the logistic regres-sion, a dummy variable (see Table 15, variable 16) wasincluded in the analysis to control the missing infor-mation in the family income variable.

Table 9 Internet use anywhere by education of household reference person

Internet use any where Education of household reference person Total

Grades 16 79 1011 a HS degree Some college College degree MA or higher

Total % yes 27.8 39.3 44.8 58.0 67.5 75.3 80.9 63.1Count (719) (966) (1,890) (8,092) (7,595) (4,223) (2,171) (25,656)% 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

Pearson Chi Square 1611.212, P =0.000a

This category can include students who reached grade 12 but did not graduate from high school

Table 10 Internet use anywhere by family income

Internet use any where Family income Total

$4999 or less $59999 $1019999 $2034999 $3549999 $5074999 $75000 plus

Total % yes 34.0 37.9 42.9 54.4 66.7 71.0 79.3 64.1Count (553) (916) (2,415) (4,276) (3,724) (5,022) (5,824) (22,730)% 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

Pearson Chi Square 1831.212, P =0.000

Table 11 Internet use anywhere by presence of ahousehold computer(s)


Internet use any where Household computer(s) Total

No computer One + computers

Total % yes 28.4 74.6 62.9Count (6,584) (19,376) (25,960)% 100.0 100.0 100.0

364


12/21

Table 16 presents the results of the logistic regressionanalysis. Eleven variables representing family and childbackground characteristics enter the logistic regressionanalysis in the rst stage/block. These include dichoto-

mous representing the gender and age of the child, thegender, race/ethnicity and citizenship status of thehousehold respondent (variables 1 9), the geographiclocation of the household residence and the ownership/rental status of the household residence.

In the basic model (block 1) the Exp (B) coefficientfor the gender of the child (variable 1 in Table 16) showsthat the odds of female school age children using theInternet anywhere are 1.075 times greater than maleschool age children, controlling for the other variables inthe analysis. As other variables (i.e., variables 2 20),enter the analysis in later blocks, this effect becomesnearly statistically insignicant and the Exp (B) effect

remains modest.The Exp (B) coefficient in the basic model for thegender of the household reference person (variable 2,Table 16) indicates that the odds of children from ahousehold with a female reference person using theInternet is 0.854, controlling for the other variables inblock 1 of the analysis. The Exp (B) value of 0.854indicates that children from female reference personhouseholds, show a 14.6% lower use of the Internet thantheir counterparts from male reference person house-holds. The effect of the Exp (B) coefficient for femalehousehold reference person decreases steadily from acoefficient of 0.854 controlling only for the block 1 basic

model explanatory variable to 0.920, 0.944, 0.949, and0.954 (a coefficient of 1 equals no or zero effect), as

blocks 2 5 explanatory variables enter the analysis.Importantly, the statistical effect of a female householdreference person on childrens Internet use becomesstatistically insignicant as later explanatory variables

enter the analysis.In contrast to a childs gender, the age of children in

this study is a strong predictor of Internet use. Forschool age children 1417 years of age (variable 4,Table 16), the Exp (B) coefficient in the basic model is4.041, controlling for the other variables in the analysis.The results indicate that the odds of a child using theInternet anywhere are, on average, much higher(304.1%) for students in this age group than it is forchildren aged 6 to 11 in the reference category. Impor-tantly, the effect of age on Internet use is largely unaf-fected by the explanatory variables entered insubsequent steps. In fact, the effect of age actually

increases modestly as successive blocks of explanatoryvariables enter the analysis. Thus, as family resourcesand household computing resources and expertise, andpublic school resources are added stepwise to the model,the Exp (B) coefficient for the dichotomous variablerepresenting the age group 1417 increases from 4.041 inthe rst block to 4.491, 4.519, 5.310, and 5.304, respec-tively.

Children aged 1213 also show signicantly higherlevels of Internet use, although somewhat lower levelsthan their 1417 year old counterparts. The Exp (B)coefficient for this age group when it rst enters into theanalysis is 2.719 (variable 3, Table 16). As family re-

sources, household computing resources and expertise,and public school resources are added stepwise to the

Table 14 Internet use anywhereby level of school internetresources (state averageInternet connections perstudent)


Internet use any where Internet connections per student Total

(1) 3.45.6 (2) 5.76.3 (3) 6.46.9 (4) 7.18.3 (5) 8.4 +

Total % yes 67.0 63.7 65.2 61.6 56.3 62.9Count (6,004) (4,526) (5,246) (5,187) (4,997) (25,650)% 100.0 100.0 100.0 100.0 100.0 100.0

Table 12 Internet Use AnyWhere by Internet Use Adult(s)(18+) at Home


Internet use any where Internet use by adults (18+)at home Total

No adults One adult Two + adults

Total % yes 36.0% 75.3% 81.8% 62.9%Count (9,983) (5,113) (1,086) (25,960)% 100.0 100.0 100.0 100.0

Table 13 Internet use anywhere by internet use adult(s)(18+) outside the home


Internet use any where Internet use by adults (18+) outside home Total

No adults One adult Two + adults

Total % yes 47.9% 71.6% 81.9% 62.9%Count (11,678) (9,245) (503) (25,960)% 100.0 100.0 100.0 100.0

365


13/21

Table 15 Logistic analysis explanatory variables

Sequence Variable name Variable denition

Block 1 1. PESEX2 Dichotomous variable for gender of child (0 = male, 1 = female)2. PESEX12 Dichotomous variable for gender of HH ref. person (0 = male, 1 = female)3. PEAG1213 Dichotomous variable for childs age 12 to 13 a

4. PEAG1417 Dichotomous variable for childs age 14 to 17 b

5. PREHHBNH Dichotomous variable for HH ref. person Black non-Hispanic b

6. PREHHHIS Dichotomous variable for HH ref. person Hispanic b

7. PREHHONH Dichotomous variable for HH ref. person other non-Hispanic b

8. PRCITNAT Dichotomous variable for HH ref. person naturalized citizen c9. PRCITNOT Dichotomous variable for HH ref. person not a citizen c

10. GEMSTAT1 Dichotomous variable for household located in central city d

11. HETENUR1 Dichotomous variable for residence rented e

Block 2 12. PEMARNEV Dichotomous variable for HH ref. person never married f

13. PEMARDIV Dichotomous variable for HH ref. person divorced/separated f

14. PEEDUCHH Education level of the HH ref person coded 0 to 6 g

Block 3 15. RFAMI_MD Family income coded 0 to 6 h

16. RFAMI_NM Dichotomous variable for missing data on family income i

Block 4 17. HESC11 Dichotomous variable for computer(s) in the household j

18. PRNET2A2 Number is adults in the HH that use Internet at home k

19. PRNET3A2 Number is adults in the HH that use Internet outside the home k

Block 5 20. SPIC Average number of students per Internet connectionstate level

a Reference category is children aged 6 to 11b Reference category is white non-Hispanic childrenc Reference category is natural born citizen childrend Reference category is children in residences outside of centralcitiese Reference category is children in owner occupied housesf Reference category is children in households with a marriedreference persong Category 0 = grades 1 to 6, 1= 7 to 9, 2 = 10 to 12 (no degree),3 = high school grad., 4 = some college, 5 = college grad., and6 = Masters or higher

h Category 0 = under $5,000 or missing, 1 = $5,000 to $9,999, 2 =$10,000 to 19,999, 3 = $20,000 to 34,999, 4 = $35,000 to 49,999, 5= 50,000 to 74,999, and 6 = $75,000 or higheri Dichotomous variable for missing data on family income where 0= data reported on family income and 1 = no data reported onfamily income j Category 0 = no computer in the household and 1 = one or morecomputers in the householdk Category 0 = no adults in the household use the Internet, 1 =one adult, and 2 = two or more adults

Table 16 Logistic regression of the effects of background attributes, family characteristics and nancial resources, household computing

resources and expertise, and public school internet resources on internet use among school age childrenAttribute Block 1 Block 2 Block 3 Block 4 Block 5

B Sig Exp(B) B Sig Exp(B) B Sig Exp(B) B Sig Exp(B) B Sig Exp(B)

1. PESEX2 0.072 0.010 1.075 0.083 0.004 1.086 0.082 0.005 1.085 0.069 0.025 1.072 0.069 0.026 1.0712. PESEX12 .158 0.000 0.854 0.083 0.009 0.920 0.058 0.071 0.944 0.052 0.130 0.949 0.047 0.170 0.9543. PEAG1213 1.000 0.000 2.719 1.056 0.000 2.874 1.071 0.000 2.920 1.188 0.000 3.282 1.187 0.000 3.2784. PEAG1417 1.396 0.000 4.041 1.502 0.000 4.491 1.508 0.000 4.519 1.670 0.000 5.310 1.668 0.000 5.3045. PREHHBNH 0.919 0.000 0.399 0.747 0.000 0.474 0.670 0.000 0.512 0.284 0.000 0.753 0.250 0.000 0.7786. PREHHHIS 0.990 0.000 0.372 0.599 0.000 0.549 0.561 0.000 0.570 0.281 0.000 0.755 0.246 0.000 0.7827. PREHHONH 0.252 0.000 0.778 0.243 0.000 0.784 0.190 0.004 0.827 0.033 0.638 0.967 0.022 0.754 0.9788. PRCITNAT 0.008 0.902 0.992 0.039 0.568 0.961 0.029 0.672 0.971 0.001 0.990 1.001 0.044 0.558 1.0459. PRCITNOT 0.410 0.000 0.663 0.306 0.000 0.736 0.287 0.000 0.751 0.182 0.008 0.833 0.148 0.031 0.86210. GEMSTAT1 0.085 0.019 0.919 0.106 0.004 0.900 0.119 0.001 0.888 0.121 0.003 0.886 0.107 0.008 0.89911. HETENUR1

0.448 0.000 0.639

0.220 0.000 0.803

0.068 0.069 0.935 0.075 0.062 1.078 0.084 0.039 1.087

12. PEMARNEV 0.389 0.000 0.678 0.216 0.000 0.806 0.120 0.056 1.128 0.123 0.052 1.13013. PEMARDIV 0.201 0.000 0.818 0.033 0.423 0.967 0.289 0.000 1.335 0.297 0.000 1.34514. PEEDUCHH 0.343 0.000 1.410 0.271 0.000 1.312 0.093 0.000 1.097 0.093 0.000 1.09715. RFAMI_MD 0.325 0.000 1.384 0.329 0.000 0.720 0.302 0.000 0.73916. RFAMI_NM 0.200 0.000 1.221 0.009 0.509 1.009 0.016 0.248 1.01617. HESC11 0.917 0.000 2.503 0.922 0.000 2.51518. PRNET2A2 0.668 0.000 1.950 0.670 0.000 1.95519. PRNET3A2 0.380 0.000 1.462 0.374 0.000 1.45320. SPIC 0.053 0.000 0.949Constant 0.342 0.000 1.408 1.036 0.000 0.355 1.639 0.000 0.194 2.110 0.000 0.121 1.800 0.000 0.165Nagelkerke R Square 0.193 0.236 0.253 0.381 0.382Cox& Snell R Square 0.141 0.173 0.185 0.279 0.280

366


14/21

model, the Exp (B) coefficient increase from 2.719 to2.874, 2.920, 3.282, and 3.278, as successive blocks of explanatory variables enter the analysis, and they alwaysremain statistically signicant.

Examination of the Exp (B) coefficients for the raceand ethnicity of household reference persons in Table 16indicates that the odds of school age children in house-holds with a Black non-Hispanic or Hispanic referenceperson are signicantly less likely to use the Internetthan children from households with a white referenceperson. For children from households with Black non-Hispanic households (i.e., households with a Black ref-erence person), the Exp (B) value of 0.399 indicates thatthe odds of a child using the Internet anywhere are, onaverage, 60.1% lower, from these households than forchildren from white non-Hispanic households. As thesubsequent blocks of explanatory variables enter theanalysis (i.e., blocks 1 through 5), the Exp (B) coefficientfor the effect of race of householder on Internet usesteadily improve from 0.399 for only block 1 explana-tory variables to 0.474, 0.512, 0.753, and 0.778, respec-tively, as successive set of variables enter the analysis(see variable 5, Table 16). Thus, the disparities in In-ternet use by children from Black households improvefrom 60.1% below to 22.2% below their white non-Hispanic counterparts, controlling for all the subsequentexplanatory variables. Thus, much of the disparity inInternet use for children from Black non-Hispanichouseholds can be accounted by family educational,nancial, and computing resources. Equally importantamong these factors, the greatest impact on Internet useamong school age children from Black non-Hispanichouseholds occurs with the addition of householdcomputing resources and expertise to the analysis.Therefore, as family resources, household computingresources and expertise, and school computing resourcesare added to the basic model, disparities in Internet useamong children in Black non-Hispanic households de-cline. Signicantly, the greatest decrease occurs whenhousehold computing and expertise are added to theanalysis. Specically the Exp (B) coefficient improvesfrom 0.512 to 0.753 (see variable 5, Table 5), and chil-dren from Black non-Hispanic households fall only24.7% below their counterparts from white non-His-panic households Internet use compared with 48.8%below before household computing resources andexpertise enter the analysis.

The results of the multivariate analysis for childrenfrom Hispanic households (i.e. households with a His-panic reference person) are very similar to those forchildren from Black households. The results of themultivariate analysis for children from 2.719 to 2.874,2.920, 3.282, and 3.278, as successive blocks of explan-atory variables, and indicates that the odds of a childusing the Internet anywhere are, on average, 62.8%lower for Hispanic households compared with the chil-dren from white non-Hispanic households. As familyresources, household computing resources and expertiseand public school resources enter the analysis, the Exp

(B) coefficient improves dramatically from 0.372 to0.549, 0.570, 0.755, and 0.782, for explanatory variableblocks 15 respectively (see variable 6, Table 16).Therefore, like children from Black non-Hispanichouseholds, as family resources, household computingresources and expertise, and school computing resourcesare added to the basic model, disparities in Internet useamong children in Hispanic households decline. Signif-icantly, the greatest decrease occurs when householdcomputing and expertise enters the analysis. Specicallythe Exp (B) coefficient improves from .570 to .755 (seevariable 6, Table 5), and children from Hispanichouseholds fall only 24.5% below their counterpartsfrom white non-Hispanic households Internet usecompared with 43.0% below before household com-puting resources and expertise enter the analysis.

The Exp (B) coefficient in the basic model for othernon-Hispanic races households (households with a ref-erence person that is an other non Hispanic race,including Asian, pacic islander and Native American)is 0.778 (see variable 7, Table 16). The Exp (B) value of 0.778 indicates that the odds of a child from thesehouseholds using the Internet are on average, 22.2%lower than for children from non-Hispanic whitehouseholds controlling for the explanatory variables inthe basic model in block 1. Importantly, with theintroduction of the subsequent blocks of explanatoryvariables in block 25, the Exp (B) coefficient improvessteadily from 0.778 to 0.784, 0.827, 0.967, and 0.978,respectively. Thus, when subsequent explanatory vari-ables enter the analysis, and in particular, when house-hold computing resources and expertise enter theanalysis, disparities in Internet use among children innon-Hispanic other race households become statisticallyinsignicant.

The citizenship status of households also has an im-pact on Internet use by school age children. The Exp (B)coefficient for households with a household referenceperson that is not a citizen (variable 9, Table 16) showsthat the odds ratio of Internet use among school agechildren from non-citizen households is 0.663 control-ling for the other explanatory variables in block 1, thebasic model. This indicates that the odds of a child usingthe Internet are 33.7% lower for children from non-citizen households compared with households where thereference persons are citizens. As in the case of Blacknon-Hispanic and Hispanic households, as subsequentexplanatory factors enter the analysis, the Exp (B)coefficient steadily improves from 0.663 to 0.736, 0.751,0.833, and 0.862 (as blocks 15 are added). Therefore, ascomputing resources and expertise are added to the basicmodel, the disparities in Internet use among childrenliving in households where the household referenceperson is not a citizen (as is the case for Black non-Hispanic and Hispanic households), show the greatestdecline.

In contrast to households where the reference per-son is not a citizen, children from households wherethe reference person is a naturalized citizen show no

367


15/21

signicant differences with other children when theexplanatory variables in the basic model are controlled.The Exp (B) coefficient for households with a natural-ized citizen reference person is 0.992, controlling for theother block 1 variables, and is statistically insignicant(variable 9, Table 16). As the subsequent explanatoryvariables enter the analysis of the Exp (B) coefficient,these households remains statistically insignicant.

Geographic location appears to have a modest effecton Internet use by school age children, independent of the other explanatory variables in the analysis. The Exp(B) coefficient for central city status (variable 10,Table 16) for children residing in central city householdsis 0.919, controlling for the other variables in the basicanalysis. The Exp (B) value of 0.919 is the odds ratio of Internet use among school age children in householdslocated in central cities, and indicates that, the odds of achild using the Internet is, on average, 8.1% lower forhouseholds located in central cities. As the subsequentexplanatory variables enter the analysis, the Exp (B)coefficient for central city households remains stable,going from 0.900 to 0.888, 0.886, and 0.899 for blocks 2 5 respectively. Therefore, independent of family char-acteristics and resources, the modest disparities in In-ternet use among children living in central cityhouseholds versus those outside the central cities remainunchanged and statistically signicant. This nding isconsistent with census data, which show clearly that alarger proportion of lower income households are lo-cated in central cities across the US.

Examination of the Exp (B) coefficient for ownershipstatus of households (see variable 11, Table 16) is 0.639,controlling for the other variables in the block 1 analy-sis. These results indicate that the odds of a child livingin rented household quarters are 36.1% lower than forchildren living in owner occupied households. Asexplanatory variables from blocks 1 to 5 enter theanalysis, the Exp (B) coefficient improves from 0.639 to0.803, 0.935, 1.078, and 1.087, respectively. Therefore, asthe analysis controls for family education, familynancial resources, household computing resources andexpertise, and public school resources, disparities in In-ternet use among children living in rented householdsdecline sharply and become statistically insignicant.

The second stage of the analysis (block 2) introducesthe marital status and education level of the householdreference person into the analysis. The Exp (B) for levelof education of the reference person is 1.410 (variable14, Table 15). Thus, the odds of a child using the In-ternet anywhere increase by an average of 41.2% as thelevel of education of the household reference personincreases. Importantly, the effect of education is nearlyfully mediated with the introduction into the analysis of household computing resources and expertise. Whenthese resource variables enter the analysis in block 4, theExp (B) for education drops to 1.097.

Marital status is introduced into the analysis throughtwo dichotomous variables; one that identies house-hold reference persons that were never married (variable

12, Table 16) and a second that identies referencepersons that are divorced or separated (variable 13,Table 16). The Exp (B) coefficient for never marriedhouseholds is 0.678, when it enters the analysis in block2 and controlling for the other variables in the analysisat that point. The Exp (B) value of 0.678 indicates thatthe odds of a child using the Internet are, on average,32.2% lower for households with never married house-hold reference person compared with households withmarried reference persons. As family income andhousehold computing resources and expertise (block 2 5), enter the analysis, the Exp (B) coefficient improvessteadily from 0.678 to 0.806, 1.128, and 1.130, respec-tively. Although the effects of never married householdreference persons on Internet use actually become posi-tive, they do not achieve statistical signicance. Never-theless, the analysis indicates that the availability of family nancial resources and household computingresources and expertise appear to account for disparitiesin Internet use of children living in households withreference persons who were never married.

Examination of the Exp (B) coefficient for divorcedor separated household reference persons (variable 13,Table 16) indicates that the odds of a child from one of these households using the Internet anywhere are, onaverage, 18.2% lower for households with marriedhousehold reference persons. As family income, house-hold computing resources and expertise, and schoolcomputing resources enter the analysis; the Exp (B)coefficient for these households improves steadily from0.818 to 0.967, 1.335, and 1.345, for blocks 25,respectively. With the introduction of these additionalresources, the impact of Internet use among children inhouseholds where the household reference person is di-vorced becomes positive and statistically signicant.This indicates that with an infusion of resources, Inter-net use increases and is actually greater than in house-holds with married reference persons.

Family nancial resources are measured using re-ported family income (variable 15, Table 16). Whenfamily income enters the analysis in block 3, the Exp (B)coefficient for this variable is 1.384. Importantly,household computing resources and expertise, whichenter the analysis in block 4, appear to mediate most of the effect of family income. When the variables forhousehold computing resources and expertise enter theanalysis, the Exp (B) coefficient for family income dropsfrom 1.386 to 1.097 and become statistically insigni-cant.

Household computing resources and expertise enterthe analysis in block 4. The Exp (B) coefficient for thepresence of a computer(s) in the household (variable 17,Table 16) is 2.503, controlling for the other variables inblocks 14, indicates that the odds of a child using theInternet anywhere are, on average, 150% higher forhouseholds in which there is a computer present. How-ever, as school computing resources enter the analysis,the Exp (B) coefficient remains largely unchanged,increasingly slightly from 2.502 to 2.515. This indicates

368


16/21

that children living in households where a computer ispresent are far more likely to access the Internet at homethan at school.

In addition to household computing resources, thepresence of adult computing expertise in the householdalso appears to have a signicant effect on the use of theInternet by children. Examination of the Exp (B) coef-cient for Internet use among adults at home (variable18, Table 16), shows an odd ratio of 1.950, controllingfor the other variables in blocks 14. The Exp (B) valueof 1.950 is the odds ratio of Internet use among schoolage children when adults use the Internet at home. Theresults indicate that the odds of a child using the Internetanywhere are, on average, 95.0% higher in householdsfor each adult (up to two adults) that uses the Internet inthe home. This indicates that children living in house-holds where adults use it at home are more likely toaccess the Internet at home than their counterparts fromhouseholds where no adults use the Internet at home.

Similarly, the Exp (B) coefficient for Internet useamong adults at home (variable 19, Table 16) shows anodds ratio of 1.462, controlling for the other variables inblocks 1 through 4. The Exp (B) value of 1.462 indicatesthat the odds of a child using the Internet are, onaverage, 46.2% higher in households where each adult(up to two adults) uses it outside the home. This indi-cates that children living in households where adults usethe Internet outside the home, e.g., at work, are far morelikely to access the Internet at home rather than atschool.

Finally, the potential impact of the availability of public school Internet resources has been examined.School Internet resources are measured as the averagenumber of Internet connections in public schools perstudent for each state (variable 20, Table 16), and areintroduced in the nal step of the analysis, block 5. TheExp (B) coefficient for school computing resources is0.949, controlling for the other variables in the analysis.The results indicate that the odds of a child using theInternet anywhere are, on average, 5.1% lower forhouseholds in which children use it at school. The effectof available school Internet resources on the use of theInternet by children, though statistically signicant, isrelatively modest when compared to a number of theother explanatory variables. However, as noted, thisattribute is only available for the purposes of thepresent study at a level of aggregation that is too high,and, as a result, this variable provides only a roughmeasure of Internet resources available to public schoolstudents. A much better measure of available schoolresources would be the average number of students perInternet connection within a given school, but datawere not available. An analysis that incorporated amore rened measure of available school resourcesmight also nd a larger effect due to these resourcesthan are found here.

Although logistic regression techniques do not pro-duce a genuine measure of explained variance (i.e., R 2),there are estimated measures of R 2 that are useful for

providing an indication of the relative contribution of avariable or set of variables to a logistic analysis. Overall,the explanatory variables in block 1 account for nearly20% of all variation in Internet use among school agechildren using the Nagelkerke estimation for a logisticregression for an R 2 (R n =0.193, see Table 16). Whenfamily characteristics (i.e., marital status and educationlevel) enter the analysis in block 2, the Nagelkerke R 2

rises to 0.236, and when family nancial resources (block3) enter the analysis, the Nagelkerke R 2 rises to 0.253.

With the entry of household computing resources andexpertise into the analysis in block 4, the largest increasein the Nagelkerke R 2, which rises from 0.253 to 0.381, isobserved. Thus, an analysis of the Exp (B) coefficientsindicates that the availability of household computerresources and expertise appear to be one of the majordeterminants of Internet use among school age children.Importantly, as noted above, when household comput-ing resources and expertise enter the analysis, thegreatest decline occurs in disparities in Internet use forchildren from Black non-Hispanic, Hispanic, and non-citizen households.

The importance of household computing resourcesand expertise in the performed analysis highlights thecritical role of access to persons with technical experi-ence for the acquisition of computing related skills. Amajor source of technical expertise for members of families undoubtedly comes from the occupationalexperience of adults. The current analysis suggests thattechnical experience, some of which is obtained in theworkplace, can be transferred to other non-workingmembers of the household. In this way, some of thefrequent signicant amounts of technical training givento US employees may be transferred to children inhouseholds where adults have beneted from suchtraining. This human capital developed in the work-place, or workplace capital, might then be transferredto other members of the household.

Of course, this type of transfer is not available to allhouseholds with children, because not all jobs requirethe same level of computer related expertise. Table 17examines this question with regard to Internet use byadults across different types of occupations. Not sur-prisingly, there is a strong positive relationship betweenthe level of occupational skill complexity and Internetuse among adults. As Table 17 shows, over four-fths(85.8%) of adults in managerial or professional occu-pations report they use the Internet, while only slightlyover one-third (38.6%) of adults who are employed asmachine operators or laborers report any use. Further-more, only one-fourth (27.9%) of adults who areunemployed or not in the labor force report they use theInternet. Thus, Internet use among adults appears highlycorrelated with the procedural and technical require-ments of their job. Professional and managerial occu-pations require signicant amounts of these technicalskills which, when learned, provide workers with in-creases in human capital that can also be transferred tothe children in their households.

369


17/21

4.3 Summary of ndings

The analysis found that a broad range of backgroundcharacteristics, including the age of a child, the race andethnicity of the household reference person, the genderof the household reference person, the geographiclocation of the household, and the ownership status of the familys residence were important determinants of Internet use among school age children. In terms of these factors, the most critical disparities in Internet useamong school age children occurred among childrenfrom Black (non-Hispanic), Hispanic, and non-citizenhouseholds.

The analysis also showed that the level of familyeducation, marital status and family income are impor-tant independent determinants of Internet use amongschool age children. In addition, the introduction of these factors into the analysis accounts for some of thedisparity in Internet use among children from Black,non-Hispanic and Hispanic households and for childrenfrom non-U.S. citizen households. This indicates thatchildren from these families in part, show lower levels of Internet use because they come from families with lessincome and education.

The most important nding of this analysis is that theavailability of household computing resources andexpertise showed the greatest impact of any set of attributes in reducing observed disparities in Internet useamong school age children in Black (non-Hispanic),Hispanic and non-citizen households. Apart from familyand child background characteristics entered in the rststep of the analysis, the introduction of householdcomputing resources and expertise appears to representthe single most important set of determinants of Internetuse among school age children. Each of the three factorsincluded in this step of the analysis, specically, thepresence of a computer(s) in the household, an adult(s)that uses the Internet at home, and/or an adult(s) thatuses Internet outside the home are statistically signi-cant and independent determinants of Internet useamong school age children. The entry of householdcomputing resources and expertise into the analysis alsomediates much of the effect of family education, maritalstatus, and income.

These ndings indicate that children from Black(non-Hispanic) Hispanic and non-citizen families in partshow lower levels of Internet use because they comefrom households that lack a computer(s) in the house-hold, an adult(s) that uses the Internet in the household,and/or an adult(s) that uses the Internet outside thehousehold. The effects are larger and independent of theeffects of family income, education, and marital status.

This suggests that households with computing re-sources and expertise are much more able to support andencourage children to learn on the Internet. Some of these households may benet in part from a type of transfer of capital skills that adults often learn in theworkplace to children residing in their homes. This canbe seen as a form of a transfer of workplace humancapital or human capital acquired in the workplace toother members of the household. In addition, in homeexpertise is a form of in kind peer support contribu-tion to children on a continual basis, similar to peer andco-worker support in the workplace. The technologicalcomplexity of certain occupations can provide a com-parative advantage in terms of computer support andexpertise to households beyond those obtained fromincome and benets alone. Therefore, the availability of such expertise and support (independent of where adultsacquire it) greatly increases the likelihood that childrenresiding in households with such support will useInternet anywhere.

Finally, there is some evidence that schools are afactor in providing Internet access to school age childrenwho do not have the resources to access the Internet athome. Evidence was found that school computing re-sources do make a difference in Internet access whenschool age children have little or no home based com-puting resources or expertise. School-based resourcescan mitigate mismatches in access depending on theavailability of family resources. Unfortunately, thenumber of students per Internet connection is a poorlymeasured variable indicator. School level data are notcaptured at an appropriate level for adequate analysisthat is consistent with the CPS personal and householdlevel data used in this paper. Further analysis needs tobe conducted with more recent school use data at thesub-state level, preferably at the school district level.

Table 17 Internet use of adults(18 years and over) byoccupation

Occupation Internet UseAnywhere byAdults

Total % Total#

Yes No

1. in school - not in labor force(%) % 82.1 17.9 100.0 (2,028)2. managerial-professional(%) % 85.8 14.2 100.0 (22,597)3. Tech, sales, admin support(%) % 73.9 26.1 100.0 (20,686)4. Service occupations(%) % 45.8 54.2 100.0 (9,707)5. Precision prod, craft and repair(%) % 48.3 51.7 100.0 (8,083)6. Operators, laborers, etc.(%) % 38.6 61.4 1

Documents

Article_Closing the Digital Divide