Ecological, Psychological, and Cognitive Components of Reading Difficulties: Testing the Component Model of Reading in Fourth Graders Across 38 Countries

LDXspldxJ Learn DisabilJournal of Learning Disabilities0022-21941538-4780SAGE PublicationsSage CA: Los Angeles, CA10.1177/002221941143124110.1177_0022219411431241Ecological, Psychological, and Cognitive Components of Reading DifficultiesTesting the Component Model of Reading in Fourth Graders Across 38 CountriesChiuMing Ming1McBride-ChangCatherine2LinDan31State University of New York at Buffalo, Buffalo, NY, USA2Chinese University of Hong Kong, Hong Kong SAR, China3Hong Kong Institute of Education, Hong Kong SAR, ChinaMing Ming Chiu, Department of Learning and Instruction, University at Buffalo-State University of New York, Buffalo, NY, USA Email: mingmingchiu@gmail.com92012455391405© Hammill Institute on Disabilities 20122012Hammill Institute on DisabilitiesThe authors tested the component model of reading (CMR) among 186,725 fourth grade students from 38 countries (45 regions) on five continents by analyzing the 2006 Progress in International Reading Literacy Study data using measures of ecological (country, family, school, teacher), psychological, and cognitive components. More than 91% of the differences in student difficulty occurred at the country (61%) and classroom (30%) levels (ecological), with less than 9% at the student level (cognitive and psychological). All three components were negatively associated with reading difficulties: cognitive (student’s early literacy skills), ecological (family characteristics [socioeconomic status, number of books at home, and attitudes about reading], school characteristics [school climate and resources]), and psychological (students’ attitudes about reading, reading self-concept, and being a girl). These results extend the CMR by demonstrating the importance of multiple levels of factors for reading deficits across diverse cultures.reading attitudeshome literacyschool resourceshome resourcesgender differencescover-dateSeptember/October 2012The component model of reading (CMR; Aaron, Joshi, Gooden, & Bentum, 2008) was proposed to highlight the importance of the cognitive, ecological, and psychological domains of a given child for explaining his or her reading difficulty. Aaron et al. (2008) argue that difficulties in any of these domains can result in reading deficits. Indeed, targeting specific and varied cognitive difficulties (e.g., word recognition vs. comprehension training) has, in some studies, been shown to facilitate children’s literacy achievement over more generalized literacy instruction (e.g., Aaron et al., 2008; Connor, Morrison, & Katch, 2004). Certain ecological aspects of reading difficulties have also been highlighted in past research (e.g., Berninger, Dunn, Lin, & Shimada, 2004; Chiu & McBride-Chang, 2010; Dudley-Marling, 2004). Finally, psychological aspects of learning in the CMR, including motivation and interest in learning, individual difference variables such as gender, and learning styles (e.g., Aaron et al., 2008; Berninger et al., 2004; Chiu & Klassen, 2009), are also associated with reading achievement in some studies. However, little research has explicitly tested features of this CMR across all three domains thus far. In the present study, we focused particularly on the ecological and psychological domains related to low reading achievement in fourth graders across 38 countries (45 regions). These domains were explored while statistically controlling a measure of children’s past cognitive performance. In clustering the variables used to tap the three domains of the CMR, we referred specifically to the model from Aaron et al. (2008) to highlight the characteristics of each. Each of these domains within the context of the present study is briefly considered below.Cognitive DomainIn the present study, we used a rough composite of early literacy skills to represent the domain of cognitive ability in the CMR. Admittedly, this domain was by far the weakest in terms of measurement because it consisted of only five items that had been estimated by children’s parents rather than actual tests of these skills by the children themselves. The estimations included students’ abilities to recognize and write letters of the alphabet, their skills in reading and writing words, and their skills in reading sentences. Early literacy skills tend to be strongly related to subsequent literacy skills across cultures (for a review, see Joshi & Aaron, 2006; McBride-Chang, 2004). Although this scale was a rather crude indicator of children’s cognitive ability overall because it did not directly tap children’s actual cognitive skills, we nevertheless included it as a potentially important indicator of early cognitive skills relevant for reading achievement. It was also important to include this measure as one aspect of the cognitive domain so as to compare the strengths of the ecological and psychological domains with it, given the centrality of the cognitive domain in previous work on the CMR (e.g., Aaron et al., 2008).Ecological DomainApart from the widely recognized cognitive domain within the CMR, the ecological domain represents the interaction of multiple environmental influences. At the broadest level of the ecological domain is the country in which one lives. A country’s gross domestic product (GDP) per capita is associated with students’ school achievement across various domains (e.g., D. P. Baker, Goesling, & Letendre, 2002; Heyneman & Loxley, 1982, 1983), including reading difficulty itself (Chiu & McBride-Chang, 2006). Thus, GDP per capita was included in the present study as a central variable capturing countrywide variability.Beyond this country-level variable, our exploration of the ecological systems domain extended to the three main areas of aspects of family, school, and individual classroom teachers. Family socioeconomic status (SES) has consistently been shown to be associated with students’ literacy skills (e.g., Chiu & McBride-Chang, 2010; Willms, 1999). SES broadly represents access to not only material resources such as money but also additional resources such as education and status (e.g., Chiu & McBride-Chang, 2010).Apart from SES, six family-related ecological variables were included in our CMR of reading difficulty in fourth graders around the world. The first two were parents’ attitudes about reading, focused on the value and enjoyment of reading, as well as parents’ attitudes about school as inclusive and supportive. Parents’ own beliefs about reading may be important for children’s reading behaviors as well as their motivations for reading (L. Baker, Scher, & Mackler, 1997); the present study tested the association between reading attitudes and reading achievement for older children. Two additional family variables included in our analyses were two measures of books in the home. A previous cross-cultural study of 15-year-olds demonstrated that even with parents’ SES and educational levels (and many other variables) statistically controlled, total number of books in the home was a unique marker of reading achievement (Chiu & McBride-Chang, 2010). Therefore, in the present study, both total number of books in the home and, more specifically, number of children’s books in the home were included as unique correlates of low reading achievement. The final two family-related variables of the present study were parents’ reports of their literacy-related behaviors with their children at first grade and their own reading behaviors (i.e., actual time spent reading). First-grade home literacy behaviors were those centered on reading and writing with their children. The home literacy model (Senechal, 2006; Senechal & LeFevre, 2002) suggests that those activities in which parents engage specifically to teach word recognition and writing skills do tend to facilitate early literacy skills in children, whereas storybook reading, which is a more informal literacy experience, may not directly facilitate subsequent reading skills per se. In the present study, however, both formal and informal literacy activities were included within individual questions by the Progress in International Reading Literacy Study (PIRLS) developers, so these could not be entirely disentangled.Reading scores also vary across schools, as demonstrated with data from 2001 from the PIRLS (Ogle et al., 2003). Indeed, a variety of school-related resources are associated with children’s school achievement (Greenwald, Hedges, & Laine, 1996; Taylor, Pearson, Clark, & Walpole, 2000). In the present study, we tested the association of peers’ achievement on students’ own achievement, following previous work demonstrating that peer culture (specifically in relation to reading enjoyment in that study) is associated with reading skill (Chiu & McBride-Chang, 2006). In addition, general aspects of school resources, including specialists to support learning needs, as well as overall school satisfaction and communication between the school and parents were examined in relation to reading achievement to tap the dimension of school in the ecological domain of CMR.A final aspect of this ecological domain included in the present study was teacher characteristics. Findings on characteristics of teachers in relation to literacy outcomes are somewhat complex (e.g., Connor, Son, Hindman, & Morrison, 2005; Darling-Hammond & Youngs, 2002; Podhajski, Mather, Nathan, & Sammons, 2009). Some teacher qualifications may be associated with student achievement outcomes, but these can differ across studies, areas of achievement tested, and developmental levels of the students. In the present study, we included four measures of teacher training and satisfaction. These were specific training in how to teach reading, professional development in terms of current readings in professional journals or books on education, teachers’ approaches to how literacy is taught to students, and overall career satisfaction.Psychological DomainIn addition to the cognitive and ecological domains in the CMR, Aaron et al. (2008) highlight the importance of the psychological domain for understanding reading difficulties. A central focus of the psychological domain is the issue of gender differences. Previous studies across cultures have demonstrated that girls tend to outscore boys on some tests of reading achievement (Chiu, Chow & McBride-Chang, 2007; Wagemaker, Taube, Munck, Kontogiannopoulou-Polydorides, & Martin, 1996). There are multiple reasons for this difference, many of which are related to socialization practices. For example, Chiu and McBride-Chang (2006) found that enjoyment of reading mediated the effect of higher rates of reading difficulty among male adolescents across cultures. In the present study, therefore, we included a measure of reading attitudes and tested the extent to which gender was associated with reading difficulties.To be as specific as possible in understanding correlates of reading achievement from the psychological domain, we also tapped three other aspects of learning style and interest in reading from the students themselves. The first was perceived actual reading activities at home, the second was reported time spent in various reading activities, and the third was each student’s reading self-concept. Students’ self-concepts about reading skills tend to be at least moderately associated with their literacy skills as they progress through primary school (e.g., Chapman & Tunmer, 1995). We also included a measure of students’ own perceptions of school. With all of these scales included, we covered the psychological domain of the CMR, at least with respect to self-perceptions, fairly comprehensively.Study FocusTo summarize, this study examined whether characteristics of countries, families, schools, teachers, and students themselves from 45 different regions and countries were linked to low reading achievement in fourth graders via multilevel analysis. Furthermore, we analyzed whether these links differed across contexts via random effects. We were most interested in viewing these data within the framework of the CMR, in which reading difficulties can be associated with cognitive, ecological, and psychological domains. The cognitive domain was roughly represented by estimates of children’s past achievement in various literacy skills before they entered first grade. Both because others (e.g., Aaron et al., 2008) have extensively explored aspects of the cognitive domain within this model and because our measure of cognitive skills was relatively crude in the present study, this domain was not our main focus. Rather, the ecological and psychological domains were of central interest. Thus, multiple measures of the ecological domain, including country-, family-, and school-level variables, were included in the present study. In addition, six variables related to students’ own psychological characteristics were analyzed. By including aspects of each of these three domains, we were able to (a) examine the extent to which aspects of each of these domains would independently emerge as correlates of reading difficulty and (b) isolate those individual variables from each domain that appeared to be most strongly associated with poor reading among fourth graders.MethodDataThe International Association for the Evaluation of Education Achievement–Progress in International Reading Literacy Study (IEA-PIRLS) assessed 186,725 fourth grade students’ reading achievement and asked students and principals to complete questionnaires related to their perceptions of themselves and their immediate environments as well as knowledge of the children. International experts defined reading achievement, built assessment frameworks, created test items and questionnaire items, forward- and backward-translated them, and pilot tested them to check their validity and reliability (for details and sample items, see Martin, Mullis, & Kennedy, 2007, and www.pirls.org). Participating students completed an 80-min assessment booklet and then a 15- to 30-min questionnaire. We also used economic data from the World Bank (2007). See Table 1 for variable descriptions and summary statistics. The regions and countries that participated were Austria, Belgium (both Flemish and French regions), Bulgaria, Canada (across Alberta, British Columbia, Nova Scotia, Ontario, and Quebec), Chinese Taipei, Denmark, England, France, Georgia, Germany, Hong Kong, Hungary, Iceland, Indonesia, Iran, Israel, Italy, Kuwait, Latvia, Lithuania, Luxembourg, Macedonia, Moldova, Morocco, the Netherlands, New Zealand, Norway, Poland, Qatar, Romania, the Russian Federation, Scotland, Singapore, Slovak Republic, Slovenia, South Africa, Spain, Sweden, Trinidad and Tobago, and the United States.Table 1.Means, Standard Deviations, and Other Basic Descriptors for All VariablesVariableMSDMinMaxReliabilityDescriptionLow reading0.1958131 = correctly answering less than 10% of the 165 questions on the reading test. Equivalent to a reading score below 400. The student reading scores estimated by the Rasch models were calibrated to a mean of 500 and a standard deviation of 100 (Martin, Mullis, & Kennedy, 2007).CognitiveStudent’s early literacy skillsa0.011.01−2.491.82.95Index of students’ ability before 1st grade: recognize most of the letters of the alphabet; read some words; read sentences; write letters of the alphabet; write some words. Choices = not at all, not very well, moderately well, very well.EcologicalCountry variableGDP per capita15,8357,6512,88436,316Gross domestic product per person in U.S. dollars (World Bank, 2007).Log GDP per capita9.670.667.9710.83Linear GDP per capita did not fit the data as well as Log GDP per capita (World Bank, 2007).Family variablesParent SESa−0.051.00−2.952.84.65Index of father’s education; mother’s education; father’s occupation; mother’s occupation; family’s financial situation.Parents’ attitudes toward readinga−0.020.99−2.941.69.82Index of I read only if I have to; I like talking about books with other people; I like to spend my spare time reading; I read only if I need information; reading is an important activity in my home. Choices: disagree a lot, disagree a little, agree a little, agree a lot.Parents’ view of schoola−0.021.00−3.321.05.84Index of my child’s school includes me in my child’s education; my child’s school cares about my child’s progress in school; my child’s school does a good job in helping my child become better in reading. Choices: disagree a lot, disagree a little, agree a little, agree a lot.Number of books at homeb2.921.240–10 (15%); 11–25 (20%); 26–100 (34%); 101–200 (16%); > 200 (15%).Number of children’s books at homeb2.781.240–10 (18%); 11–25 (20%); 26–50 (31%); 51–100 (19%); > 100 (13%).Early home literacy activitiesa−0.040.99−3.231.69.86Index of students’ home literacy activities at 1st grade: read books; tell stories; sing songs; play with alphabet toys; talk about what parents had read; play word games; write letters or words; read aloud signs and labels; visit a library. Choices: never or almost never, sometimes, often.Parents’ reading activities at homea−0.010.78−1.831.36.60Index of time spent reading at home per week (choices: < 1 hr, 1–5 hr, 6–10 hr, > 10 hr); frequency of reading for enjoyment (choices: never or almost never, once or twice a month, once or twice a week, every day or almost every day).School variablesSchoolmates’ early literacy skillsa0.011.01−1.271.07.97Index of schoolmates’ abilities before 1st grade: recognize most of the letters of the alphabet; read some words; read sentences; write letters of the alphabet; write some words. Choices = not at all, not very well, moderately well, very well.Shortage of school resourcesa0.051.00−1.092.71.94Index of shortage of qualified teaching staff; teachers with a specialization in reading; instructional materials; supplies; school buildings and grounds; heating/cooling and lighting systems; instructional space; computers for instructional purposes; computer software for instructional purposes; library books; audiovisual resources. Choices: not at all, a little, some, a lot.School climatea−0.050.99−3.192.65.88Index of school characterization of teachers’ job satisfaction; teachers’ expectations for student achievement; parental support for students’ achievement; students’ regard for school property; students’ desire to do well in school; students’ regard for one another’s welfare.Home–school involvementa−0.090.99−3.671.58.64Index of frequency of letters, calendars, newsletters, etc., sent home to provide parents with information about the school; events at school to which parents are invited (choices: never, once a year, 2–3 times a year, 4–6 times a year, 7 or more times a year); and percentage of parents: attend teacher-parent conferences; attend cultural, sporting, or social events at school (choices: 0–10%, 11–25%, 26–50%, > 50%).Teacher variablesAvailability of specialistsa−0.010.99−0.982.75.87Index of a reading specialist is available to work in my classroom with students who have difficulty with reading; a reading specialist is available to work in a remedial reading classroom with those students; other professionals are available to work with those students. Choices: never, sometimes, always.Teacher reading traininga−0.010.87−1.981.43.87Index of teachers’ trainings in pedagogy/teaching reading; remedial reading; reading theory. Choices: not at all, overview or introduction to topic, it was an area of emphasis.Teachers’ professional developmenta0.061.02−3.421.84.93Index of reading for professional development: books or professional journals related to teaching in general; books or professional journals related to teaching reading; children’s books. Choices: never or almost never, once or twice a year, once or twice a month, at least once a week.Teacher develops skillsa−0.011.01−3.112.04.9Index of teacher asking students to: identify the main ideas of what they have read; explain or support their understanding of what they have read; compare what they have read with experiences they have had; compare what they have read with other things they have read; make predictions about what will happen next in the text they reading; make generalizations and draw inferences based on what they have read; describe the style or structure of the test they have read. Choices: never or almost never, once or twice a month, once or twice a week, every day or almost every day.Teacher career satisfactiona−0.031.00−3.831.30.82Index of I am content with my profession as a teacher; I am satisfied with being a teacher at this school; I would describe the teachers at this school as a satisfied group; I had more enthusiasm when I began teaching than I have now; I do important work as a teacher. Choices: disagree a lot, disagree a little, agree a little, agree a lot.Psychological variables at the student levelGirl0.491 = girl; 0 = boy.Students’ reading activities at homea−0.011.01−3.082.02.85Index of parents: listen to my child read a loud; talk with my child about things we have done; talk with my child about what he/she is reading on his/her own; go to the library or a bookstore with my child; help my child with reading for school. Choices: never or almost never, once or twice a month, once or twice a week, every day or almost every day.Students’ reading time/frequencya0.031.00−3.493.46.9Index of frequency of: I read aloud to someone at home; I listen to someone at home read aloud to me; I talk with my friends about what I am reading; I talk with my family about what I am reading; I read for fun outside of school; I read to find out about things I want to learn; I read comic books; I read stories or novels; I read books that explain things; I read magazines; I read newspapers; I read directions or instructions; I read brochures and catalogs; My teacher reads aloud to the class; I read aloud to the whole class; I read aloud to a small group; I answer questions in a workbook or on a worksheet about what I have read; I write something about what I have read; I answer questions aloud about what I have read; I talk with other students about what I have read (choices: never or almost never, once or twice a month, once or twice a week, every day or almost every day); and time of: reading stories and articles on the Internet; reading stories and articles in books or magazines (choices: no time, up to 1 hr, 1–3 hr, 3–5 hr, 5 hr or more).Students’ reading attitudesa0.011.00−3.001.59.79Index of I read only if I have to; I like talking about books with other people; I would be happy if someone gave me a book as a present; I think reading is boring; I enjoy reading. Choices = disagree a lot, disagree a little, agree a little, agree a lot.Students’ view of schoola−0.011.00−3.061.69.77Index of I like being in school; I think that teachers in my school care about me; I feel safe when I am at school; students in my school show respect to each other; students in my school care about each other; students in my school help each other with their work. Choices = disagree a lot, disagree a little, agree a little,agree a lot.Students’ reading self-concepta0.001.00−2.771.54.80Index of reading is very easy for me; I do not read as well as other students in my class; When I am reading by myself, I understand almost everything I read; I read slower than other students in my class. Choices = disagree a lot, disagree a little, agree a little, agree a lot.Note: N = 186,725. Data are from the Progress in International Reading Literacy Study, unless otherwise noted.aAll indices were standardized at the country level to M = 0 and SD = 1. Differences in the populations of each country cause the overall student means and standard deviations to be slightly different from 0 and 1, respectively. Aside from the test score index, the authors created all the indices.bAnalyses with dummy variables showed roughly linear results, so an ordered variable was used to aid interpretation.The PIRLS 2006 framework defines the two major aspects of students’ reading literacy—reading purposes and comprehension processes. Reading for literary experience and reading to acquire and use information are the two major purposes that account for the majority of reading experiences of young children. Thus, the questions are divided equally so that 50% address each purpose. Readers make meaning of texts in many ways, depending not only on the purpose for reading but also on the difficulty of the text and the readers’ prior knowledge. PIRLS looks at four processes of comprehension: to focus on and retrieve explicitly stated information (20% of the questions); make straightforward inferences (30%); interpret and integrate ideas and information (30%); and examine and evaluate content, language, and textual elements (20%). Low-achieving readers correctly answered less than 10% of the 165 test items, resulting in scores of less than 400.Methodological DesignInvestigating the cognitive, ecological, and psychological correlates of reading difficulty across countries requires representative sampling, precise tests and questionnaires, and suitable statistical models. In each country, IEA chose at least 150 representative schools based on neighborhood SES and student intake and sampled one or two fourth grade classes from each school (stratified sampling), resulting in a sample size of about 4,000 students per country or region (Martin et al., 2007). With suitable weights, IEA created representative samples of each country’s schools and fourth grade students.Students received subtests (overlapping subsets of all multiple choice and open-ended questions) for wider coverage of reading skills while reducing student fatigue and test-learning effects (a balanced incomplete block test; F. B. Baker & Kim, 2004). A graded response Rasch model of these subtests measured the difficulty of each test item to estimate each student’s reading competence more precisely (F. B. Baker & Kim, 2004).To reduce measurement error, several questionnaire items for each theoretical construct (e.g., SES) were used to create an index via a Rasch model (Warm, 1989). The multigroup Rasch models for each item in each country yielded similar parameters, indicating measurement equivalence across countries (May, 2006). (Unlike factor analysis, a multigroup Rasch model has two advantages: It requires only one invariant anchor item across countries and models heterogeneous use of the ordinal rating scale [Rossi, Gilula, & Allenby, 2001].) Other studies also show consistent questionnaire responses and participant understandings across countries (Brown, Micklewright, Schnepf, & Waldmann, 2005; Martin et al., 2007; Schulz, 2003). To estimate reliability, the Rasch models included computations of the information function (F. B. Baker & Kim, 2004); when the information function is greater, there is more information, smaller standard errors, more precision, and greater reliability (see Table 1 for the reliability of each index). All indices based on questionnaire items were standardized at the country level to a mean of 0 and a standard deviation of 1. Differences in the populations of each country cause the overall student means and standard deviations to be slightly different from 0 and 1, respectively.Multilevel logit analysis of plausible values yields more precise standard errors than does ordinary least squares (Goldstein, 1995; Rust & Rao, 1996). Missing questionnaire response data (9%) can reduce estimation efficiency, complicate data analyses, and bias results. Markov chain Monte Carlo multiple imputation estimates these missing data, which is more effective than deletion, mean substitution, or simple imputation (Peugh & Enders, 2004).AnalysisA multilevel logit model tests if each level’s variance component is significant.Low_Readijk={1+exp[−(β000+f0jk+g00k)]}−1+eijkThe dichotomous outcome variable Low_Readijk of student i in class j in country k has a grand mean intercept β000, with student-, class-, and country-level residuals (eijk, f0jk, g00k). Explanatory variables were entered in sequential sets to estimate the variance explained by each set (Kennedy, 2003). The index of a student’s early literacy skills before enrolling in elementary school (see Table 1 for details) reflects the cognitive component of reading and is entered first. Next, we consider ecological variables, namely, country, family, and school characteristics. Country variables might affect family variables. As families might choose their children’s schools, family variables might affect school variables. Schoolmate characteristics might affect teacher behaviors. All of these, broadly defined as the ecological domain in which reading takes place, might affect the psychological variables across students. Hence, we entered the variables as follows: cognitive, ecological (country, family, school, schoolmates, teacher), and psychological (see variable descriptions in Table 1). All continuous variables were centered on their country mean.Low_Readijk={1+exp[−(β000+f0jk+g00k+β1jkCognitive+β001GDPpc+βfjkFamilyijk+β0skSchool0jk+β0tkTeacher0jk+βpjkPsychijk)]}−1+eijkFirst, we entered each student’s index of literacy skills (cognitive component) before entering school (Cognitive; see Table 1). Then, we tested whether sets of predictors were significant with a nested hypothesis test (χ2 log likelihood; Kennedy, 2003). Nonsignificant variables were removed. Then, we applied this procedure for Cognitive to the ecological components, starting with the country variable, log GDP per capita (GDPpc).Next, we applied the procedure for GDPpc to family variables: SES, parents’ attitudes toward reading, parents’ views of school, number of books at home, number of children’s books at home, early home literacy activities, and parents’ reading activities (Familyf). Applying a random effects model (Goldstein, 1995), we tested if the f regression coefficients (βfjk = βv00 + gf0k) differed across countries (gf0k ≠ 0?) or correlated with GDPpc.Then, we applied the procedure for Family to school variables: schoolmates’ early literacy skills, shortage of school resources, school climate, availability of specialists, and home–school involvement (Schools). Next, we applied this procedure to teacher variables: reading training, professional development, skill development, and career satisfaction (Teachert). Note that the SES, family, and school variables were all included as aspects of the ecological domain within the CMR because all were environment-related measures at some level. Last, we applied this procedure to psychological variables: gender, reading activities at home, reading time/frequency, reading attitudes, view of school, and reading self-concept (Psychp).We report how a 10% increase in each continuous variable above its mean is linked to low reading achievement (result = b × SD × [10% / 34%]; 1 SD ≈ 34%). As percentage increase is not linearly related to standard deviation, scaling is not warranted. The odds ratio of each variable’s total effect (E = direct plus indirect) was reported as the percentage increase or decrease (+E% or –E%) in the outcome variable in the tables (Kennedy, 2003).We used an alpha level of .05. To minimize false positives, we controlled for the false discovery rate with the two-stage linear step-up procedure, which outperformed 13 other methods in computer simulations (Benjamini, Krieger, & Yekutieli, 2006). The small sample of countries (N = 38) limits identification of nonsignificant country-level results (Konstantopoulos, 2008; see Table 2 for details). We also analyzed residuals for influential outliers.Table 2.Statistical Power at Each Level of Analysis for Each Effect Size Given a Sample of 186,725 Fourth Grade Students From 10,103 Classes in 38 CountriesEffect SizeLevel.1.2.3.43. Country.07.12.21.342. Class.40.921.001.001. Student.40.931.001.00We used multilevel mediation tests across the above vectors (Krull & MacKinnon, 2001). For significant mediators, the proportional change was 1 − (b′/b), where b′ and b were the regression coefficients of the explanatory variable, with and without the mediator in the model, respectively. Then, we did a three-level path analysis (Goldstein, 1995). To test the robustness of these results, we (a) did two-level regressions (class, student) for each country, (b) used standardized scores for both three- and two-level analyses, and (c) ran the analyses without the early literacy skills explanatory variable.ResultsSummary StatisticsThis sample included a variety of countries, ranging from poor (e.g., South Africa) to rich (e.g., Netherlands), across five continents. See Table 1 for overall summary statistics and Table 3 for correlation–variance–covariance matrices). Percentages of students with low reading achievement within a country ranged from nearly zero (0.6%) in the Netherlands to a few percent (6%) in the median country Slovenia to a high of 81% in South Africa. These three countries also showed progressively smaller scores on cognitive, ecological and psychological components. Cognitively, students’ early literacy scores were lower (0.2, 0.2, −0.4). Ecological components were also progressively lower: GDP per capita (US$26,479, US$20,659, US$9,146), parent SES index (0.2, 0.1, −0.3), parent attitude toward reading (0.1, 0.0, −0.1), school resources (0.7, 0.6, −0.6), and school climate (0.0, −0.2, −0.7). Last, students’ psychological measures were lower: reading attitude (0.1, 0.1, −0.1) and reading self-concept (0.2, 0.2, −0.4). The statistical analyses below reflect these patterns.Table 3.12345678910111213141. Low Read0.15.00−.06−.08−.04−.10−.15.05.11−.06−.09−.01−.03−.082. Early L S−.011.02.01.09.11.07.07−.21.09.03−.04.05.10.153. GDP−.22.020.44.14.02.18.29−.12−.18.16.17.00−.09.044. SES−.22.08.210.99.27.44.57−.11−.14.14.17.00.02.155. P Attitude−.10.11.03.280.98.27.37−.01−.06.06.07.01.10.136. Books−.21.06.22.35.221.55.70−.12−.17.15.18.02.11.227. Child Books−.31.06.35.46.31.451.54−.15−.30.25.27.02.02.188. Schmate ELS−.12.21.17.11.01.10.121.03.07.21.04.00−.01.039. Lack Sch Res.27.09−.28−.14−.06−.14−.24−.071.00−.20−.26.00.07−.0410. Sch climate−.17.03.25.14.06.12.20−.21−.200.97.35.01.01.0511. Home-Sch−.23−.04.26.17.07.14.22−.04−.26.360.97.01−.03.0612. Girl−.04.09.00.00.01.03.03.00.00.02.010.25.09.0313. S attitude−.07.10−.14.02.10.09.02.01.07.01−.03.190.99.2014. Self-concept−.21.15.06.16.13.18.14−.03−.04.05.06.06.201.00Note: Correlations, variances, and covariances are in the lower-left triangle, on the diagonal, and in the upper-right triangle of the matrix, respectively. Low Read = low reading achievement; Early L S = early literacy skills; GDP = log GDP per capita; SES = parent socioeconomic status; P Attitude = parents’ attitudes toward reading; Books = number of books at home; Child Books = number of children’s books at home; Schmate ELS = schoolmate early literacy skills; Lack Sch Res = shortage of school resources; Sch climate = school climate; Home-Sch = home–school involvement; S attitude = students’ reading attitudes; self-concept = students’ reading self-concept.Explanatory ModelCognitive, ecological, and psychological variables all accounted for differences in students’ low reading achievement (see Figure 1 and Table 4). Differences across countries accounted for most of the variance (61%), followed by differences across classes (30%) and differences across students (9%). This result shows the overwhelming importance of ecological differences (more than 91%) compared to cognitive or psychological differences (less than 9%; because of the structure of the data, family differences are also included in the 9%). All results discussed below describe first entry into the regression, controlling for all previously included variables. Ancillary regressions and statistical tests are available on request.Figure 1.Path diagram of low reading achievement.Note: Thicker lines indicate larger effect sizes.Table 4.Summaries of Seven Multilevel Logit Regressions Predicting Low Reading Achievement, with Standardized Coefficients (and Standard Errors)Seven Multilevel Logit Regressions Predicting Low Reading AchievementExplanatory VariableModel 1Model 2Model 3Model 4Model 5Model 6Model 7Students’ early literacy skills−0.417***−0.320***−0.320***−0.323***−0.323***−0.281***−0.253***(0.008)(0.008)(0.008)(0.008)(0.008)(0.007)(0.007)Log GDP per capita−1.170*−0.926−0.721−0.678−0.573−0.761(0.464)(0.476)(0.475)(0.475)(0.494)(0.500)SES−0.258***−0.255***−0.255***−0.228***−0.224***(0.009)(0.008)(0.008)(0.009)(0.008)Parents’ attitudes toward reading−0.116***−0.116***−0.116***−0.088***−0.090***(0.008)(0.008)(0.008)(0.008)(0.006)Parents’ views of school−0.035−0.034−0.035−0.041(0.046)(0.046)(0.046)(0.046)Number of books at home−0.093***−0.092***−0.092***−0.058***−0.046***(0.005)(0.005)(0.005)(0.005)(0.005)Number of children’s books at home−0.099***−0.098***−0.098***−0.093***−0.101***(0.006)(0.006)(0.006)(0.006)(0.007)Early home literacy activities−0.030−0.030−0.030−0.073(0.098)(0.097)(0.097)(0.098)Parents’ reading activities at home−0.014−0.013−0.013−0.008(0.010)(0.010)(0.010)(0.009)Schoolmate early literacy skills−0.124***−0.122***−0.125***−0.128***(0.022)(0.022)(0.023)(0.023)Shortage of school resources0.117***0.117***0.133***0.132***(0.022)(0.022)(0.024)(0.023)School climate−0.243***−0.243***−0.247***−0.241***(0.021)(0.021)(0.022)(0.021)Availability of specialists0.0150.0130.009(0.019)(0.019)(0.020)Home–school involvement−0.127***−0.126***−0.135***−0.129***(0.020)(0.020)(0.022)(0.021)Teacher reading training0.0460.041(0.031)(0.032)Teachers’ professional development0.0880.099(0.052)(0.051)Teacher develops skills−0.027−0.032(0.020)(0.021)Teacher career satisfaction−0.028−0.031(0.018)(0.019)Girl−0.322***−0.282***(0.014)(0.014)Student’s reading activities at home0.110(0.070)Student’s reading time/frequency0.232(0.269)Student’s reading attitudes−0.379***−0.280***(0.008)(0.007)Student’s view of school0.058(0.070)Student’s reading self-concept−0.564***−0.551***(0.008)(0.008)Variance at Each LevelVariance ExplainedCountry (61%).000.107.112.115.116.144.141Class (30%).000.253.296.324.331.304.294Student (9%).118.237.255.247.244.331.257Total variance explained.011.163.180.190.192.209.197Note: Each regression included a constant term.*p < .05. ***p < .001.CognitiveStudents with 10% better early literacy skills before attending school were 0.4% less likely to have low reading achievement (see Table 4, Model 1, odds ratio computation; Kennedy, 2003). This cognitive component accounted for only 1% of the variance of low reading achievement (see Table 4, Model 1, last row: total variance explained = 0.011).EcologicalEcological variables (country, family, school, and teacher) accounted for most of the variance in low reading achievement.GDP per capitaWhen a country’s GDP per capita exceeded the mean by 10%, its students were 1% less likely to have low reading achievement. (Regressions with log GDP per capita fit the data better than linear GDP per capita, explaining more of the variance in low reading achievement.) GDP per capita accounted for 16% of the variance (= [total variance explained in Model 2] − [total variance explained in Model 1]). As seen in Table 4, GDP per capita accounts for the large majority of the variance explained (83%) and is the most important explanatory variable in this model.Family variablesFamily SES, parents’ attitudes toward reading, and books at home were linked to low reading achievement. Students whose family SES exceeded the mean by 10% were 0.3% less likely to be low achievers in reading. If their parents’ attitude toward reading exceeded the mean by 10%, students were 0.2% more likely to have low reading achievement. Also, students in homes with more books or more children’s books were less likely to have low reading achievement. Controlling for family SES and the book variables, the GDP per capita regression coefficient was no longer significant (see Table 5 for mediation tests).Table 5.Mediation TestsInitial VariableMediator% ChangezLog GDP per capitaParent SES22a−4.891***Number of books at home−6.273***Number of children’s books at home−6.578***Schoolmates’ early literacy skillsHome–school involvement8−4.419***Shortage of school resourcesHome–school involvement74.604***School climateHome–school involvement13−6.144***Parent SESStudents’ attitude toward reading11−7.695***Students’ reading self-concept−34.000***Parents’ attitude toward readingStudents’ attitude toward reading31−26.789***Students’ reading self-concept−27.718***Number of books at homeStudents’ attitude toward reading49−30.608***Students’ reading self-concept−40.621***aThe effect of the initial variable was no longer significant after entering the mediator(s).***p < .001.School and teacher variablesSchoolmates’ early literacy skills, school resources, school climate, and home–school involvement were linked to low reading achievement, but teacher characteristics were not. Students whose schoolmates’ early literacy skills exceeded the mean by 10% were 0.2% less likely to have low reading achievement. In schools with 10% more shortage of school resources, students were 0.1% less likely to have low reading achievement. If school climate or home–school involvement exceeded the mean by 10%, students were 0.3% or 0.1% less likely to have low reading achievement, respectively. Controlling for home–school involvement, the regression coefficients of schoolmates’ early literacy skills, shortage of school resources, and school climate fell by 8%, 7%, and 13%, respectively.PsychologicalPsychological variables (gender, reading attitude, and self-concept) were also linked to low reading achievement. Boys were 1% more likely than girls to have low reading achievement. Students whose reading attitude or reading self-concept exceeded the mean by 10% were 0.3% or 0.6% less likely to have low reading achievement, respectively. Controlling for students’ reading attitude and self-concept, the regression coefficients of family SES, parents’ attitudes toward reading, and number of books at home fell by 11%, 31%, and 49%, respectively (Table 3).Other variables were not significant. The above effects did not show substantial differences across countries, and the cumulative effect of all country interaction terms accounted for less than 1% of the variance. Robustness tests yielded similar results.DiscussionThe present study has underscored the importance of understanding reading difficulties within the CMR (Aaron et al., 2008) approach. Overall, the ecological domain accounted for more than 91% of the variance in reading difficulty in the present study. These findings are generally in line with those from Chiu and McBride-Chang (2006), who found that reading difficulties were correlated most strongly with country-level (32%) and school-level (56%) effects, rather than individual student-level effects (less than 12%). In both large-scale studies across many countries, student-level variables, as compared to broader ecological variables, were relatively modestly associated with reading difficulties. Collectively, these results suggest that low reading achievement is largely a societal phenomenon, rather than an individual one. In addition, in the present study, only 1% of the variance in reading difficulty could be attributed to the cognitive domain of early literacy skills; the other 8% of the variance at the individual level was accounted for by variables under the broad psychological domain and family variables (ecological domain) following the CMR (Aaron et al., 2008).In the final model of reading difficulty, there were four unique correlates at the family level of the ecological domain. These were family SES, parents’ attitudes toward reading, number of total books and number of children’s books at home. Family SES is a well-known correlate of students’ academic achievement, and results of the present study are in line with those found in previous cross-cultural studies of reading difficulty (e.g., Chiu & McBride-Chang, 2006). Family poverty has been associated with a plethora of environmental and psychosocial difficulties that are often cumulative in their effects (for a review, see Evans, 2004). Children with fewer economic resources often lack a variety of supports as compared to those with richer parents, and these resources are linked directly to literacy development (e.g., Willms, 1999).Parents’ own attitudes about reading were also uniquely associated with reading difficulty in the present study. These findings are in line with previous ones (e.g., L. Baker et al., 1997) underscoring the importance of parents’ attitudes about reading. Basically, the more parents reported endorsing statements about the importance and value of reading, the less likely their children were to have reading difficulties. It is possible that such results reflect, in part, the passive epigenetic program, in which the parents’ genetic makeup is correlated with the environment they provide for their children (e.g., Scarr, 1991). In this case, parents who have more difficulties in reading and/or dislike reading likely read less, so they provide fewer positive examples of literacy skills for their children. Assuming skills and attitudes are somewhat genetically determined, in this case both the genetic makeup and the environment of the child whose parents do not read interact to promote a lack of reading practice and/or enjoyment. This is admittedly quite speculative. However, the fact that both total books in the home and total children’s books in the home were the other two family-related variables that were both uniquely associated with reading difficulty in children adds weight to this idea. Even with family SES statistically controlled, these crude measures of valuing of books were independently associated with reading difficulty. Chiu and McBride-Chang (2006) obtained similar results, though the present study is the first to demonstrate unique effects on literacy difficulties across cultures separately for both children’s books at home and all books at home.At another level of the ecological domain of the CMR, no independent effects of teachers were obtained, but four aspects of the school environment were significant. The first of these was schoolmates’ early literacy skills. As found previously (e.g., Chiu & McBride-Chang, 2006; Ogle et al., 2003), the influence of peer achievement and interest in reading tends to be fairly strongly associated with an individual student’s own achievement in literacy learning. In addition, measures of both school resources and school climate were also uniquely correlated with reading difficulties. These results are in line with previous findings on the importance of school resources for student achievement (e.g., Greenwald et al., 1996; Taylor et al., 2000). Finally, home–school involvement was independently associated with reading difficulties. When schools and families reinforce one another in positive ways, one aspect of the mesosystem in Bronfenbrenner’s (1993) ecological systems theory, fewer reading difficulties emerge in students.It is important that there were three aspects of the psychological domain of the CMR that emerged as uniquely associated with reading difficulties in the present study. The first was gender, with fourth grade boys being more likely than girls to be low-achieving readers, as found previously by Chiu and Chow (2010) in 15-year-olds across cultures. This gender variable itself accounted for 1% unique variance in explaining reading difficulty. In addition, lack of reading enjoyment was uniquely associated with reading difficulty, as found in previous studies (e.g., Chiu & McBride-Chang, 2006). Reading enjoyment was moderately correlated (r = .37) with time spent reading as well. It is impossible to determine whether reading experience influenced attitudes or vice versa in this sample, but this association between reading behaviors and reading attitudes is likely bidirectional by fourth grade. The last psychological variable to show a unique association with reading difficulty was students’ reading self-concept. The self-image that a student has built up as a competent reader appears to be negatively associated with reading difficulty as well. Again, the extent to which this feeling of competence or otherwise as a reader is a result of cognitive skills or emotional or motivational aspects of the reading process remains unclear with correlational data. Nevertheless, reading self-concept was an important factor in explaining reading difficulties in the present study, as it has been in past work (e.g., Chapman & Tunmer, 1995).One particularly strong limitation of the present study was the fact that the cognitive domain of the CMR was not an ideal measure of cognitive skills in these students. This was a retrospective variable that essentially represented parents’ past impressions of students’ early literacy skills. Perhaps a better measure of cognitive skill for this study would have been current word reading or spelling ability, based on the conceptualization of the CRM (e.g., Aaron et al., 2008). However, separate tests of general cognitive and/or word recognition skills were not included in the IEA-PIRLS data collection process. Nevertheless, we felt compelled to include the literacy skills measure, despite its relative crudeness, in the present study for the sake of completeness, and the measure held up well across analyses. For example, this literacy skills measure had an internal consistency reliability of .95 and was moderately correlated with current reading comprehension status. Indeed, this measure of the cognitive domain was uniquely associated with reading skill across all models tested in the present study. At the same time, however, it is unclear whether, had a more proximal measure of cognitive ability been included in the model, more of the variance in reading differences might have been attributed to this cognitive domain, relative to the ecological and psychological domains.Despite this issue, as well as the problem of distinguishing causations from mere correlations throughout the article, the present study has been important as a test of the CMR (Aaron et al., 2008). Although perhaps the majority of past work in reading difficulties has focused on various aspects of the cognitive domain of reading development and impairment (for a review, see McBride-Chang, 2004), CMR equally emphasizes the ecological and psychological domains in which learning takes place. The present study demonstrated that all three of these domains were uniquely associated with reading difficulty across a wide-ranging sample of fourth graders. The result that more than 91% of the variance in reading difficulty in the study could be attributed to various levels of the ecological domain suggests the need for greater attention to the context in which reading development takes place in future research.In terms of implications for practice, results from the present study particularly highlight the importance of attitudes about reading in both parents and classmates at school. Parents who themselves enjoy reading and demonstrate this with the number of books they have at home tend to have children who read better. Perhaps public service announcements emphasizing the importance of reading, both as a way of modeling literacy skills development for children and as a demonstration of parents’ valuing of reading overall, might be useful across countries as a way to help alleviate reading difficulties. At the school level, these results underscore the difficulties of schools that lack basic attributes such as qualified staff and working facilities. School quality and morale of staff and parents are all interrelated and associated with students’ achievement across cultures. Although causality cannot be determined here, one practical implication of this work may be that school morale, including parent–school communication, is essential to reduce the number of students with reading difficulties within a school. Finally, included within this issue of school morale should be an emphasis on students’ views of themselves as competent readers. Those who feel as though they are poor readers may be reflecting the reality of their performance. At the same time, it is possible that these data partly also suggest that both enjoyment of, or motivation for, reading as well as competency in it may ultimately facilitate reading skill, as indicated in a 12-year longitudinal study of American children (Archambault, Eccles, & Vida, 2010). Issues of causality in particular as raised by the present study should be addressed in future work guided by the CMR.We appreciate the research assistance of Yik Ting Choi.Declaration of Conflicting InterestsThe author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.FundingThe author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was partially funded by a grant from the Spencer Foundation to Ming Ming Chiu and by GRF (Hong Kong Government) Grant 448608 to Catherine McBride-Chang.About the AuthorsMing Ming Chiu, PhD, is a professor in the Learning and Instruction Department at the State University of New York University at Buffalo-State University of New York. His current research interests focus on international comparisons of students and statistical discourse analyses of classroom conversations.Catherine McBride-Chang, PhD, is a professor in the Psychology Department at the Chinese University of Hong Kong. Her current research interests focus particularly on cognitive development and parenting, including reading and vocabulary development and impairment in children.Dan Lin, PhD, is an assistant professor in the Department of Psychology Studies at the Hong Kong Institute of Education. Her current research interests focus particularly on literacy development and exogenous parent–child interactions and endogenous cognitive skills in shaping this progress.ReferencesAaronP. G.JoshiR. M.GoodenR.BentumK. E. (2008). Diagnosis and treatment of reading disabilities based on the component model of reading. Journal of Learning Disabilities, 41, 67–84.ArchambaultI.EcclesJ. S.VidaM. N. (2010). Ability self-concepts and subjective value in literacy: Joint trajectories from grades 1 through 12. Journal of Educational Psychology, 102, 804–816.BakerD. P.GoeslingB.LetendreG. K. (2002). Socioeconomic status, school quality, and national economic development. Comparative Education Review, 46, 291–312.BakerF. B.KimS.-H. (2004). Item response theory. Boca Raton, FL: CRC.BakerL.ScherD.MacklerK. (1997). Home and family influences on motivations for reading. Educational Psychologist, 32, 69–82.BenjaminiY.KriegerA. M.YekutieliD. (2006). Adaptive linear step-up procedures that control the false discovery rate. Biometrika, 93, 491–507.BerningerV. W.DunnA.LinS.-J. C.ShimadaS. (2004). School evolution. Journal of Learning disabilities, 37, 500–508.BronfenbrennerU. (1993). The ecology of cognitive development: Research models and fugitive findings. In WozniakR. H.FisherK. W. (Eds.), Development in context (pp. 3–44). Hillsdale, NJ: Lawrence Erlbaum.BrownG.MicklewrightJ.SchnepfS. V.WaldmannR. (2005). Cross-national surveys of learning achievement—How robust are the findings?Journal of the Royal Statistical Society (Series A), 170, 623–646.ChapmanJ. W.TunmerW. E. (1995). Development of young children’s reading self-concepts. Journal of Educational Psychology, 87, 154–167.ChiuM. M.ChowB. W.-Y. (2010). Culture, motivation, and reading achievement. Learning and Individual Differences, 20, 579–592.ChiuM. M.ChowB. W.-Y.McBride-ChangC. (2007). Universals and specifics in learning strategies: Explaining adolescent mathematics, science, and reading achievement across 34 countries. Learning and Individual Differences, 17, 344–365.ChiuM. M.KlassenR. M. (2009). Calibration of reading self-concept and reading achievement among 15-year-olds. Learning and Individual Differences, 19, 372–386.ChiuM. M.McBride-ChangC. (2006). Gender, context, and reading: A comparison of students in 43 countries. Scientific Studies of Reading, 10, 331–362.ChiuM. M.McBride-ChangC. (2010). Family and reading in 41 countries: Differences across cultures and students. Scientific Studies of Reading, 14, 514–543.ConnorC. M.MorrisonF. R.KatchL. E. (2004). Beyond the reading wars: Exploring the effect of child-instruction interactions on growth in early reading. Scientific Studies of Reading, 8, 305–336.ConnorC. M.SonS.-H.HindmanA. H.MorrisonF. J. (2005). Teacher qualifications, classroom practices, family characteristics, and preschool experience. Journal of School Psychology, 43, 343–375.Darling-HammondL.YoungsP. (2002). Defining “highly qualified teachers”: What does “scientifically based research” actually tell us?Educational Researcher, 31, 13–25.Dudley-MarlingC. (2004). The social construction of learning disabilities. Journal of Learning disabilities, 37, 482–489.EvansG. (2004). The environment of childhood poverty. American Psychologist, 59, 77–92.GoldsteinH. (1995). Multilevel statistical models. Sydney, Australia: Edward Arnold.GreenwaldR.HedgesL. V.LaineR. D. (1996). The effect of school resources on student achievement. Review of Educational Research, 66, 361–396.HeynemanS. P.LoxleyW. A. (1982). Influences on academic achievement across high and low income countries. Sociology of Education, 55, 13–21.HeynemanS. P.LoxleyW. A. (1983). The effect of primary-school quality on academic performance across twenty-nine high- and low-income countries. American Journal of Sociology, 88, 1162–1194.JoshiR. M.AaronP. G. (2006). Handbook of orthography and literacy. Mahwah, NJ: Lawrence Erlbaum.KennedyP. (2003). A guide to econometrics. Cambridge, MA: MIT Press.KonstantopoulosS. (2008). The power of the test in three-level cluster randomized designs. Journal of Research on Educational Effectiveness, 1, 66–88.KrullJ. L.MacKinnonD. P. (2001). Multilevel modeling of individual and group level mediated effects. Multivariate Behavioral Research, 36, 249–277.MartinM. O.MullisI. V. S.KennedyA. M. (2007). PIRLS 2006 technical report. Chestnut Hill, MA: Boston College, TIMSS & PIRLS International Center.MayH. (2006). A multilevel Bayesian IRT method for scaling socioeconomic status in international studies of education. Journal of Educational and Behavioral Statistics, 31, 63–79.McBride-ChangC. (2004). Children’s literacy development. Oxford, UK: Oxford University Press.OgleL.SenA.PahlkeE.JocelynL.KastbergD.RoeyS.. . . WilliamsT. (2003). International comparisons in fourth-grade reading literacy (U.S. Department of Education, NCES 2003-073). Washington, DC: Government Printing Office.PeughJ. L.EndersC. K. (2004). Missing data in educational research. Review of Educational Research, 74, 525–556.PodhajskiB.MatherN.NathanJ.SammonsJ. (2009). Professional development in scientifically based reading instruction: Teacher knowledge and reading outcomes. Journal of Learning Disabilities, 42, 403–417.RossiP. E.GilulaZ.AllenbyG. M. (2001). Overcoming scale usage heterogeneity. Journal of the American Statistical Association, 96, 20–31.RustK. F.RaoJ. N. K. (1996). Variance estimation for complex estimators in sample surveys. Statistics in Medical Research, 5, 381–397.ScarrS. (1991). Theoretical issues in investigating intellectual plasticity. In BrauthS. E.HallW. S.DoolingR. J. (Eds.), Plasticity of development (pp. 57–72). Cambridge, MA: MIT Press.SchulzW. (2003). Validating questionnaire constructs in international studies. Camberwell: Australian Council for Educational Research.SenechalM. (2006). Testing the home literacy model: Parent involvement in kindergarten is differentially related to Grade 4 reading comprehension, fluency, spelling, and reading for pleasure. Scientific Studies of Reading, 10, 59–87.SenechalM.LeFevreJ. (2002). Parental involvement in the development of children’s reading skill: A 5-year longitudinal study. Child Development, 73, 445–460.TaylorB. M.PearsonP. D.ClarkK.WalpoleS. (2000). Effective schools and accomplished teachers: Lessons about primary-grade reading instruction in low-income schools. Elementary School Journal, 101, 121–165.WagemakerH.TaubeK.MunckI.Kontogiannopoulou-PolydoridesG.MartinM. (1996). Are girls better readers?Amsterdam, Netherlands: International Association for the Evaluation of Education Achievement.WarmT. A. (1989). Weighted likelihood estimation of ability in item response theory. Psychometrika, 54, 427–450.WillmsJ. D. (1999). Quality and inequality in children’s literacy: The effects of families, schools, and communities. In KeatingD. P.HertzmanC. (Eds.), Developmental health and the wealth of nations (pp. 72–93). New York, NY: Guilford.World Bank. (2007). The world development report 2006. New York, NY: Oxford University Press.