Skilled Students and Effective Schools: Reading Achievement in Denmark, Sweden, and France

ABSTRACT

This study investigates how reading achievement relates to student and school characteristics in countries with different reading scores at the fourth grade level. Data comes from the Progress in International Reading Literacy Study (PIRLS) 2011 for Denmark, Sweden, and France and the multilevel analysis includes two levels: student/home and schools. The school effectiveness and the home literacy models informed the selection of the independent variables. Results show that students’ early literacy skills, home literacy practices and resources, and reading behavior are associated with reading scores in all countries. Furthermore, across different countries there are student/home universals and school particulars that explain variation in reading achievement. Educational policies should address home and school literacy skills and practices, school climate, and school composition to improve students’ reading ability.

KEYWORDS:

• Reading achievement
• PIRLS
• home literacy model
• school effectiveness

Introduction

Recent research with data from the Progress in International Reading Literacy Study (PIRLS) has shown that effective schools contribute to boosting students’ achievement in literacy (Martin, Foy, Mullis, & O’Dwyer, 2013). Nonetheless, student and home characteristics explain a significant source of variance in the achievement of students within schools (Martin et al., 2013). Large-scale surveys such as PIRLS, which counts many participating countries and has run on a 5-year cycle since 2001, offer the possibility to investigate how the variability in the school environment and in the home background of students is related to reading achievement (Lenkeit, Chan, Hopfenbeck, & Baird, 2016).

Mullis, Kennedy, Martin, and Sainsbury (2006) claimed that PIRLS provides “a wealth of information that can be used not only to improve the reading curriculum and instruction for younger students, but also help in interpreting the results for 15-year-olds in Programme for International Student Assessment (PISA)” (p. 102). Although some researchers have pointed out that there is little evidence that students’ achievement in PIRLS is related to literacy instruction (Shiel & Eivers, 2009), others have found links between teacher/school instruction and students’ reading outcomes. For example, Cheung, Kam Tse, Lam, and Ka Yee Loh (2009) found that in Hong Kong teachers’ instructional strategies and choice of reading materials were related to students’ reading achievement.

Conversely, Stancel-Piatak, Mirazchiyski, and Desa (2013) found that in three PIRLS participating European countries—Denmark, Germany, and France—a school’s emphasis on supporting understanding of a text was not significantly associated with achievement. Importantly, the findings of this study showed that schools’ cultural capital was one of the most relevant variables in explaining variation in reading scores. For example, the higher the socioeconomic composition of a school in Denmark, France, and Germany the higher the students’ achievement in reading. The same ubiquitous effect was not found with respect to the emphasis a school places on academic success, with a higher emphasis corresponding to higher achievement only in France, which suggests that school effects may be country specific. Research by Myrberg and Rosén (2006) also found that in PIRLS 2001 students’ cultural capital accounted for a great part of reading score differences between independent and public schools. Taken together, these findings suggest that any investigation of the relation between home and school characteristics and reading achievement needs to take into account schools’ cultural capital (Caro, Sandoval-Hernández, & Lüdtke, 2014).

Regarding home cultural capital, findings from studies that use PIRLS have reached similar conclusions. Myrberg and Rosén (2009) found that parental book reading and storytelling prior to school entry made a positive contribution to reading achievement and that book reading was mediated by cultural capital, as measured by the number of books at home. Additionally, this study showed that Swedish students’ early literacy skills positively impacted reading attainment, without any mediating effect of cultural capital, or number of books at home.

The studies by Martin et al. (2013) and Stancel-Piatak et al. (2013) confirm that early literacy skills in conjunction with other characteristics of effective schools have a positive influence on achievement. Thus, studies that use a school effectiveness framework suggest that early literacy skills positively impact the PIRLS reading score, but have not addressed the contribution of early home literacy practices or, specifically, of parental book reading. Although evidence indicates that both early literacy skills and early literacy practices are strong predictors of reading ability (Sénéchal, 2012), in research that uses PIRLS data only Myrberg and Rosén’s studies (2008, 2009) and Alivernini (2013) considered the influence of these factors on PIRLS 2001 reading achievement. Their findings confirm that both early skills and practices are associated with reading achievement in different ways, or that only one of these is associated. However, these studies did not include parental book reading as a specific early literacy predictor and, as such, evidence is lacking on whether this practice is associated with reading achievement when school variables are also considered.

The present study uses the PIRLS 2011 reading score to investigate how student background variables and school factors may be related to achievement. Specifically, it explores how the two levels of analysis—students and schools—may be associated with achievement and whether similar associations are present in different school systems. For this purpose, this study considers parental book reading and students’ ability to name letters of the alphabet as predictors of reading achievement in three European countries with different achievement levels. These two student/home background factors have been shown to explain more variance in students’ achievement in PIRLS than aggregate measures of more general early literacy practices and skills (Araújo & Costa, 2012, 2015). The multilevel model used also includes students’ engagement in reading, as research indicates that motivation for reading is associated with reading achievement (Guthrie & Humenick, 2004; OECD, 2010). At the school level, this study incorporates school effectiveness factors to understand how the school “has an effect on student achievement over and above” (Martin et al., 2013, p. 111) student/home predictors.

Home and School Factors and Reading Achievement

According to the home literacy model proposed by Sénéchal (2012), when parents engage in book reading with their kindergarten children they informally teach them vocabulary because interactions during storybook reading revolve around discussing the meaning of print. When parents call their children’s attention to the printed words (e.g., find all the A’s on the page), formal learning of the written code takes place. As Sénéchal and LeFréve (2002) found, it is not that all parents engage in both practices; some may focus on the formal and informal aspects while others may focus on only one or the other. Children whose parents teach them literacy skills and read to them frequently have higher reading achievement in fourth grade (Araújo & Costa, 2015; Sénéchal, 2012). In contrast, children whose parents report low shared reading and low formal teaching of the written language have lower reading scores and those whose parents engage in only one of the practices exhibit comparable medium achievement (Sénéchal, 2012).

The findings of Myrberg and Rosén’s (2009) study with PIRLS data lend support to the home literacy model. However, whereas they found a mediating effect of early literacy practices and of early literacy skills on achievement, the latter were not associated with the number of books at home. Thus, early literacy skills acquired before the start of compulsory education mediated reading achievement independently of the presence of books at home. In a similar cross-country comparative study, these authors also found that in six European countries—Bulgaria, France, Hungary, Italy, Norway, and Sweden—the direct effects of books at home on students’ achievement were insignificant (Myrberg & Rosén, 2008). Nonetheless, these studies indicate that parents’ educational level is a significant predictor of, and has a direct effect on, reading achievement. The extent to which its impact is mediated by books at home, early literacy practices, such as book reading and storytelling, and early literacy skills, such as the ability to name letters of the alphabet, is what differs among countries.

These findings suggest that there is great variation among countries regarding the direct effect of socioeconomic background on achievement and variation also with regard to the positive influence high socioeconomic parents exert through the provision of supporting literacy environments (Park, 2008). This influence of home literacy environments on achievement has been found to be stronger in more developed countries that participate in PIRLS (Park, 2008). Nonetheless, in the great majority of European countries early literacy practices are associated with a gain in reading scores larger than that associated with the number of books at home or with parental attitudes toward reading (e.g., I like to read). In a similar vein, but for OECD countries, Alivernini (2013) found that knowing the letters of the alphabet very well before the start of compulsory education was associated with a higher percentage of students from less economically affluent homes reaching higher PIRLS 2006 scores.

Regarding early literacy skills, but following a school effectiveness framework, Martin et al. (2013) combined both early literacy and early numerical¹ skills in a PIRLS study as a predictor of reading achievement. Results indicate that Swedish students improved their reading score as a result of having good literacy and numerical skills. Araújo and Costa (2012) also report a positive relationship linking children’s ability to recognize most of the alphabet and frequency of parental book reading with reading achievement. Importantly, these studies show that such gains occurred in conjunction with the inclusion of other home and school background variables.

In PIRLS, children’s early literacy skills are measured in The Could Do Early Literacy Tasks When Began Primary School (ELT) scale, which includes the ability to recognize most of the alphabet, read some words, read sentences, and write letters and words before the start of compulsory education. Research indicates that, among these, children’s ability to name letters of the alphabet before formal reading instruction begins is one of the strongest predictors of children’s reading ability (Bond & Dykstra, 1967; Piasta & Wagner, 2010; Riley, 1996). This is the case because alphabet knowledge shares a reciprocal relation with phonological awareness (Adams, 1990; Verhoeven, van Leeuwe, Irausquin, & Segers, 2016), which is a universal predictor of reading across different alphabetical languages (Seymour, Aro, & Erskine, 2003; Ziegler et al., 2010). In other words, some letter names also include their sounds (Ehri, 1983) and knowledge about letter–sound relationships facilitates reading acquisition in alphabetic languages (Seymour et al., 2003; Vaessen et al., 2010).

Several studies show that children acquire reading faster in transparent orthographies such as Finnish and Spanish which have writing systems with a one-to-one mapping between phonemes and graphemes. As Seymour et al. (2003) found, English and Danish first graders take longer than Spanish and Finnish children to master word decoding. A consistent finding across languages is that this happens because in opaque languages like English different graphemes can correspond to the same phoneme and vice versa (Vaessen et al., 2010). Ziegler et al. (2010) show that phonological awareness tested in different alphabetic languages is a predictor of reading ability in grade 2, but that its influence is weaker in transparent orthographies. A more recent study with Finnish kindergartners corroborates this; phonological awareness and letter knowledge were found to be the best predictors of reading fluency development in grade 2 (Mägi et al., 2013). However, when comparing similarly transparent languages—Estonian and Finnish—there seems to be no advantage in teaching Estonian kindergarten children the letters of the alphabet in terms of reading attainment at the end of first grade (Soodla et al., 2015).

Different results across studies might be due to both the nature of the measures used and the time of testing. For example, whereas speed in letter naming is more predictive of reading ability in transparent writing systems, accuracy in naming letters is a better predictor in deep or opaque languages (Caravolas, Volin, & Hulme, 2005). Similarly, while some studies test students’ reading ability at the end of first grade, others do so at the end of second grade or later. Moreover, some studies use cloze, or fill in the blank tests, and others use reading comprehension questions to assess reading ability (Araújo, Costa, & Morais, 2013). These differences can produce different results because the automaticity in word reading that results from phonological awareness will boost students’ ability to acquire vocabulary knowledge through reading in later grades (Perfetti, 1992; Ziegler et al., 2010). Reading comprehension relies on, among other aspects, knowledge of the words in a text and vocabulary knowledge takes many years to acquire (Catts, 2009).

In short, even though overreliance on the importance of phonological awareness in English (Share, 2008) can produce an overestimation of its predicting power in more transparent languages, it remains a universal predictor of reading ability in alphabetic scripts. As Ziegler et al. (2010) state, its “precise weight varies systematically as a function of script transparency” (p. 557).

In the PIRLS data collection it is not possible to ascertain whether students acquired early literacy skills, and specifically letter knowledge, in pre-primary school or at home. Parents are merely asked about how well their children mastered the different early literacy skills before the start of compulsory education. In contrast, when parents report on early literacy practices, including singing songs, telling stories, playing word games, playing with alphabet toys, talking about things they had done, talking about what they had read, writing letters or words, reading aloud signs and labels, and book reading, they are asked about their frequency. Specifically, they are asked how frequently they or someone in the household engaged children in these literacy practices prior to the start of compulsory education.

Convergent research results clearly indicate that home book reading contributes to boosting primary students’ reading achievement (Kalb & van Ours, 2014) and that code instruction may have more impact when done at school than at home (Mol & Bus, 2011; Piasta & Wagner, 2010). Across different studies in different countries, frequent exposure to home shared reading during preschool and kindergarten has been found to be positively related to children’s oral expressive vocabulary skills. Importantly, home book reading explains about 7% of the variation in their reading comprehension in later grades (Sénéchal, Pagan, Lever, & Ouellette, 2008). When preschool-age children are exposed to book reading, they develop their understanding of vocabulary that is not commonly used in daily oral interactions (Sénéchal, 2012). This is a distinctive characteristic of the verbal interactions that evolve around books.

Research conducted with Australian children (Kalb & van Ours, 2014) shows that children who are read to in the home frequently, 3–5 days a week and 6–7 days a week, obtain higher scores in literacy and numeracy by the equivalent of 6 to 12 months at 8/9 years of age. When interactions between vocabulary skills and school instruction have been explored, children with high vocabulary skills at the beginning of first grade benefited from implicit classroom instruction on decoding. That is, strong vocabulary skills seem to support children’s encounters of new words when engaging in child-initiated silent reading (Connor, Morrison, & Katch, 2004). This finding suggests that strong vocabulary supports end of first grade reading abilities. More specifically, vocabulary knowledge facilitates the recognition of unknown words (Stanovich, 2000).

In sum, learning to crack the alphabetical code facilitates decoding ability in alphabetical languages with different degrees of transparency and vocabulary knowledge sustains reading comprehension (Whitehurst, 2001). Moreover, these two skills complement each other in the development of reading ability (Perfetti, 1992; Stanovich, 2000).

As children progress through primary education, the positive relationship between the frequency of independent reading for pleasure and reading outcomes is also clear. Students who read independently for recreational purposes score higher on achievement tests, have better vocabulary, and have greater world knowledge (Cullinan, 1992; Krashen, 2004). This happens because in the process of reading for pleasure outside of school, children infer the meaning of new words and integrate the information read (Sénéchal, 2012, p. 47). Indeed, reading for enjoyment has been found to be significantly associated with reading comprehension from infancy through adulthood (Mol & Bus, 2011). Moreover, reading for enjoyment is related to children’s early literacy skills. For example, Silinskas et al. (2013) observed that Finnish mothers whose children had average and high reading skills reported that their first grade children read independently more often than low-skilled children. These findings are in accord with Perfetti’s (1992) verbal efficiency theory and with Stanovitch’s (2000) findings, namely that children with good beginning reading skills are more likely to become independent readers and thus reinforce their reading skills. Less is known about how being read to in the home during infancy may be related to children’s engagement in independent reading, but parental book reading is thought to positively influence children’s reading habits during subsequent school years (Cullinan, 1992).

The school environment is also thought to influence children’s motivation for and engagement in reading. School resources, such as well-equipped libraries, may promote interest in reading and help bridge the gap between more and less advantaged peers (European Commission, 2012). Children that come from high socioeconomic backgrounds have more access to reading material not only at home but also in their community and at school (Krashen, 2004). School effectiveness research further suggests that besides school resources, school climate is a factor that explains variance in students’ achievement in reading (Mullis, Martin, Foy, & Drucker, 2012; OECD, 2013). More specifically, not only are material resources and adequacy of facilities thought to affect attainment (Dompnier, Patisu, & Bressoux, 2006; Teddlie, Stringfield, & Reynolds, 2000), but also teachers’ instruction and their expectations of students’ performance (Creemers & Kyriakides, 2008).

Although PIRLS does not follow a specific theoretical model, its assessment framework includes many of the factors that operate at the classroom and school levels that Creemers and Kyriakides’ (2008) dynamic model of school effectiveness considers. In particular, the PIRLS framework includes aspects related to: (1) Resources for teaching reading, (2) Climate—school discipline and safety, (3) Emphasis on academic success, and (4) Characteristics of instructional practices for teaching reading. Research using PIRLS data indicates that a school’s emphasis on academic success, its emphasis on reading skills, and adequate resources contribute positively to boosting students’ achievement (Martin et al., 2013). Moreover, the studies with PIRLS data that are based on a school effectiveness model also suggest that motivational factors, such as students’ enjoyment of reading, are positively related to achievement (Stancel-Piatak et al., 2013). Nonetheless, students’ lack of interest in their teachers’ choice of reading materials, as measured in PIRLS, has been found to be both negatively and positively associated with their reading score (Martin et al., 2013; Stancel-Piatak et al., 2013).

Regarding the school environment, the socioeconomic composition of schools has been repeatedly shown to be related to students’ achievement and included as a school variable in PIRLS studies (Martin et al., 2013; Stancel-Piatak et al., 2013). In particular, in what regards compositional effects, students’ reading attainment is lower in schools with a high percentage of students with few economic resources or with a large percentage of students from economically disadvantaged backgrounds (Cortina, Carlisle, & Zeng, 2008). Taken together, these findings suggest that resources for teaching reading, school climate, a school’s emphasis on academic achievement, and the socioeconomic composition of schools are related to achievement.

Objectives of the Study

The purpose of this study is to investigate how reading achievement relates to home and school characteristics. It builds on the findings of previous studies with PIRLS data that were informed by a school effectiveness framework while also accounting for student background factors. An important feature of our study is that the school effectiveness factors are included with student/home predictors according to the home literacy model. The main hypothesis is that these predictors are associated with students’ achievement across different countries. Given that their predicting quality has been found to be universal across languages, we expect their effect to be present across countries, irrespective of country variations in reading performance levels in PIRLS. This assumption is tested with a multilevel model and educational policy is discussed in light of the results. Particular focus is on how early literacy skills and practices need to be taken into account to improve students’ learning.

Educational System Characteristics and Reading Curricula

In order to investigate the home factors and the school variables that have an effect on students’ achievement in high, medium, and low achieving countries we selected Denmark (high-performing, 554 reading score), Sweden (medium-performer, 542 reading score), and France (low-performer, 520 reading score) (Mullis, Martin, Foy, & Drucker, 2012). These countries share similarities and differences in terms of system-level variables, such as school curricula and autonomy (Eurydice, 2013). All of them have compulsory schooling from the age of 6 to 16 and a centralized school curriculum. However, in Denmark local schools have partial autonomy (schools develop their own curricula) and in Sweden the system is decentralized and allows for school choice (parents’ and students’ own choices). In Denmark recruitment of teachers is done by local authorities, in Sweden the responsibility for employing teachers varies depending on the category of school, and in France the central government is responsible for hiring (Eurydice, 2013).

With respect to reading curricula goals, in Denmark students should be able to use simple reading comprehension strategies and demonstrate an understanding of what they read. In Sweden, students are expected to be able to discuss their experiences from reading and to reflect on texts by the end of fifth grade. In France, students should understand explicitly stated information, find answers to simple questions, and find the subject of a literary text (Mullis, Martin, Minnich, Drucker, & Ragan, 2012).

Methods

We used the PIRLS 2011 data-set. PIRLS is an international large-scale assessment conducted by the International Association for the Evaluation of Educational Achievement (IEA) and designed to measure trends in reading achievement at the fourth grade level. Accordingly, its target population consists of students enrolled in the fourth grade of compulsory primary education. The survey includes the student achievement data as well as the student, parent, teacher, school, and curricular background data. Specifically, PIRLS collects information about the home literacy environment, the school curriculum and curriculum implementation, instructional practices, and school resources in each participating country. Regarding the PIRLS sample design, first schools are randomly selected (with a probability proportional to the estimated number of students enrolled in the target grade) and then one or two classrooms are randomly selected within each school. PIRLS is administered as a pencil-and-paper assessment and includes both multiple choice and constructed response test items. The PIRLS scaling of achievement data is based on item response theory (IRT) with the scores scaled to have an international average of 500 and a standard deviation of 100 points. The first cycle of PIRLS was carried out in 2001 and it has been administered every 5 years since then. In 2011, 48 countries participated in PIRLS.

Participants

For the selected countries in this study, the PIRLS 2011 sample was composed of 4,594 students in Denmark, 4,707 in Sweden, and 4,438 students in France with 232, 152, and 174 schools in Denmark, Sweden, and France, respectively.

Methods of Analysis

Due to the hierarchical structure of the data (students within schools), multilevel regression analysis (Goldstein, 2003) including information from the student, home, and school questionnaires was used. A two-level analysis was performed using MLWIN version 2.3 (Rasbash, Charlton, Browne, Healy, & Cameron, 2009) with students at level 1 and schools at level 2. The missing values were excluded from the analysis. The variance components model was used and the model was then estimated using iterative generalized least squares (Goldstein, 1986). The regression coefficients of the multilevel models indicate the estimated effect of each student/home and school variable (predictor) on the outcome variable (students’ reading achievement). In particular, the magnitude and the direction of the coefficients as well as the significance of the difference from zero show the relationship between the predictor and achievement, keeping all the other predictors in the model constant.

Variables

Our dependent variable is the students’ reading scores—with an international mean of 500 and a standard deviation of 100. This standardized PIRLS measure reflects the reading ability of fourth grade students. The independent variables are those associated with the home literacy and with the school effectiveness models. The two-level multilevel regression models for Denmark, France, and Sweden included the student/home variables in level 1 and the school variables in level 2. At the student/home level, the variables were: (1) Home resources for learning, (2) Parental book reading, (3) Knowledge of the letters of the alphabet, (4) Students like reading scale, and (5) Students’ engagement in reading lessons. At the school level the variables were: (1) School average of home resources for learning, (2) School average of students’ engagement in reading lessons, (3) School emphasis on academic success, (4) School discipline and safety, (5) Emphasis in early grades on reading skills, and (6) Shortage of reading resources. The five plausible values in reading were used in the analysis, as well as student and school weights.

Table 1 presents a description of the variables included in the multilevel model as well as the number of response categories for each item and the source/questionnaire where the information was collected. The model includes single items and some scales constructed by IEA. The scales were constructed using IRT scaling methods, namely the Rasch partial credit model (Masters and Wright, 1997).

Table 1. Scales and Variables Included in the Student/Home and School Levels.

Name of the variable	Items	No. of categories	Questionnaire
Student/home-level variables
Home resources for learning scale (HRL)	Number of books in the home	5	Student
	Number of children’s books in the home	5	Home
	Number of home study supports	3	Student
	Highest level of education of either parent	5	Home
	Highest level of occupation of either parent	4	Home
Students like reading scale (SLR)^a	– What do you think about reading? Tell how much you agree with each of these statements
	I read only if I have to^1	4	Student
	I like talking about what I read with other people	4	Student
	I would be happy if someone gave me a book as a present	4	Student
	I think reading is boring^1	4	Student
	I would like to have more time for reading	4	Student
	I enjoy reading^1	4	Student
	– How often do you do these things outside of school?
	I read for fun	4	Student
	I read things that I choose myself	4	Student
Letters of the alphabet^a		3	Home
Parental book reading^a		4	Home
Students engaged in reading lessons scale (ERL)^a	– Think about the reading you do for school. How much do you agree with these statements about your reading lessons?
	I like what I read about in school	4	Student
	My teacher gives me interesting things to read	4	Student
	I know what my teacher expects me to do	4	Student
	I think of things not related to the lesson	4	Student
	My teacher is easy to understand	4	Student
	I am interested in what my teacher says	4	Student
	My teacher gives me interesting things to do	4	Student
School characteristics
Instruction affected by reading resource shortages scale (RRS)^a	How much is your school’s capacity to provide instruction affected by a shortage or inadequacy of the following?		Principal
	A. General school resources
	(1) Instructional materials (e.g., textbooks)	4	Principal
	(2) Supplies (e.g., papers, pencils)	4	Principal
	(3) School buildings and grounds	4	Principal
	(4) Heating/cooling and lighting systems	4	Principal
	(5) Instructional space (e.g., classrooms)	4	Principal
	(6) Technologically competent staff –	4	Principal
	(7) Computers for instruction	4	Principal
	B. Resources for reading instruction	4	Principal
	(1) Teachers with a specialization in reading	4	Principal
	(2) Computer software for reading instruction	4	Principal
	(3) Library books	4	Principal
	(4) Audio-visual resources for reading instruction	4	Principal
School discipline and safety scale (DAS)^a
	To what degree is each of the following a problem among fourth grade students in your school?		Principal
	(1) Arriving late at school	4	Principal
	(2) Absenteeism (i.e., unjustified absences)	4	Principal
	(3) Classroom disturbance	4	Principal
	(4) Cheating	4	Principal
	(5) Profanity	4	Principal
	(6) Vandalism	4	Principal
	(7) Theft	4	Principal
	(8) Intimidation or verbal abuse among students (including texting, emailing, etc.)	4	Principal
	(9) Physical fights among students	4	Principal
	(10) Intimidation or verbal abuse of teachers or staff (including texting, emailing, etc.)	4	Principal
School emphasis on academic success scale (EAS–T)^a
	How would you characterize each of the following within your school?		Teacher
	(1) Teachers’ understanding of the school’s curricular goals	5	Teacher
	(2) Teachers’ degree of success in implementing the school’s curriculum	5	Teacher
	(3) Teachers’ expectations for student achievement –	5	Teacher
	(4) Parental support for student achievement	5	Teacher
	(5) Students’ desire to do well in school	5	Teacher
Emphasis in early grades on reading skills scale (RSS)^a
	At which grade do the following reading skills and strategies first receive a major emphasis in instruction in your school?		Principal
	(1) Reading isolated sentences	5	Principal
	(2) Reading connected text –	5	Principal
	(3) Locating information within the text	5	Principal
	(4) Identifying the main idea of a text –	5	Principal
	(5) Explaining or supporting understanding of a text –	5	Principal
	(6) Comparing a text with personal experience	5	Principal
	(7) Comparing different texts	5	Principal
	(8) Making predictions about what will happen next in a text	5	Principal
	(9) Making generalizations and drawing inferences based on a text	5	Principal
	(10) Describing the style or structure of a text	5	Principal
	(11) Determining the author’s perspective or intention	5	Principal
School average of HRL		5	Principal

Notes: – The Cronbach’s alpha reliability of all the scales used for Denmark, France, and Sweden is equal to or greater than 0.7 for all scales, except for France in the EAS-T scale (0.57). More detailed information about the scales used in this paper can be found in the PIRLS 2011 report at http://timssandpirls.bc.edu/methods/t-context-q-scales.html

^astands for reverse coding.

Analysis

First, we estimated the fully unconditional model (or null model). The null model doesn’t include covariates other than a constant and allows us to obtain the proportion of variability, calculated using the variances estimated for the errors between students and between schools. We obtained the amount of variance explained at each level in the model for the three countries: Denmark, France, and Sweden. Comparing these estimates with the final model determines the amount of variance explained by adding the variables related with the home literacy model and the school effectiveness framework.

Results

Table 2 presents the estimates of the multilevel regressions for each country and for the three different models. In model (1) we present the estimates for the null model. The null model shows that the percentage of the variance in the dependent variable at the student level is 87.5% in Denmark, 83.7% in Sweden, and 83% in France. The between-school differences are 12.5% in Denmark, 16.3% in Sweden, and 17% in France of the total variance in reading. These results show that there are considerable differences among schools in Denmark, Sweden, and France in the achievement of their students and they can be explained by school differences. Model (2) includes control variables at the student/home level and at the school level and model (3) incorporates also the variables related with the home literacy model, namely, “parental book reading” and “knowledge of letters of the alphabet before primary school”. For the coefficients presented, reverse coding was converted. For example, a high frequency of parental book reading corresponds to the highest value of the category.

Table 2. Multilevel Coefficients for the Relationship Between Student/Home and School Characteristics and Students’ Reading Achievement.

Reading	(1) Null			(2)			(3)
	Denmark	France	Sweden	Denmark	France	Sweden	Denmark	France	Sweden
Fixed part (constant)	554.14	521.64	538.23	471.54	466.78	460.25	530.71	490.92	502.73
Student/home characteristics
Home resources for learning scale (HRL)				29.35 * (1.90)	35.38 * (2.19)	31.85 * (2.31)	21.20 * (1.88)	27.33* (2.12)	23.18* (2.24)
Students like reading scale				18.74* (1.63)	23.35* (1.70)	23.12* (1.66)	15.63* (1.58)	20.16* (1.59)	18.84* (1.62)
Letters of the alphabet							21.04* (1.42)	19.26* (1.54)	19.80* (1.50)
Parental book reading							13.28* (1.83)	14.98* (1.86)	17.31* (1.94)
Students engaged in reading lessons (ERL)				−1.24 (1.90)	−1.73 (1.95)	−6.37* (2.14)	−2.54 (1.89)	−1.75 (1.84)	−6.35* (2.04)
School characteristics
Instruction affected by reading resource shortages				−0.54 (1.33)	3.64 (4.29)	−4.51 (4.00)	−0.13 (1.34)	2.91 (4.05)	−4.65 (3.79)
School discipline and safety				3.60 (2.73)	2.13 (3.17)	0.15 (3.17)	3.34 (2.66)	1.26 (3.07)	0.29 (3.04)
School emphasis on academic success				−0.27 (2.68)	7.66 (2.38)	1.19 (2.84)	1.05 (2.65)	6.13* (2.40)	−0.21 (2.73)
Emphasis in early grades on reading skills				0.85 (3.27)	1.71 (3.78)	3.97 (3.29)	0.28 (3.10)	1.44 (3.67)	2.64 (3.17)
School average of HRL				14.63* (1.98)	15.77* (1.88)	15.83* (1.75)	12.51* (2.01)	15.04* (1.87)	14.10* (1.75)
School average of ERL				1.15 (2.29)	1.78 (2.46)	−0.81 (2.87)	0.51 (2.19)	2.64 (2.46)	−0.39 (2.43)
Random part (school level)	540.62	800.11	790.40	222.56	186.83	173.37	210.04	171.96	157.62
Random part (student/home level)	3,769.32	3,912.29	4,061.73	3,327.83	3,360.12	3,446.28	3,085.57	3,126.19	3,159.70
Deviance	50,944.76	49,485.58	52,319.55	44,826.75	41,984.51	34637.36	42,935.61	40,902.56	33,893.78
AIC	50,944	49,486	52,320	44,836	41,994	34,646	42,937	40,904	33,895
School variance expl. %	12.5	17	16.3	–	–	–	–	–	–
Student variance expl. %	87.5	83	83.7	–	–	–	–	–	–

Notes: Multilevel regression of students’ performance in reading. Model (1) is the null model; model (2) includes student/home- and school-level control variables; model (3) adds the home literacy model variables at the student level. Standard errors in parenthesis p-value: *p ≤ 0.05. The bold values indicate statistically significant coefficients.

Comparing the null (1) with the full model (3), there is a clear reduction of the deviance. Also, the inclusion of the school effectiveness variables in model (2) results in a decrease in the variance component representing variation between countries. The variance component between schools diminished significantly in all countries. Finally, there is also a reduction of the deviance when comparing model (2) and model (3). In addition, the Akaike Information Criterion (AIC) index confirms that model (3) is the best-fit model to explain students’ performance.

The fixed effects estimates of model (3) show that the student/home level variables associated with the home literacy model are statistically significant in the three countries, ranging from a 13- to 21-point increase in reading scores. The difference between the level 1 variance components of model (3) and the previous model indicates that the proportion of variance in students’ achievement explained by these variables is 7.3% in Denmark, 7% in France, and 8.3% in Sweden. These results show that primary school students whose parents read more often to them and students who knew the letters of the alphabet before primary school present better results in reading. Additionally, students’ literacy knowledge (alphabet) and practices (book reading) acquired and experienced prior to school entry present similar coefficients in the three countries. In what regards the coefficients of other individual and school variables included in the regression that are statistically significant, we found that home resources for learning and reading for enjoyment are positively related with achievement. In all countries there is a very strong relationship between home resources for learning and students’ achievement in reading, with regression coefficients ranging from 21 to 27 points.

Additionally, students that report they like reading perform better in reading (increase from 16 to 20 score points). At the school level, the average of home resources for learning is also a significant predictor of students’ achievement, with regression coefficients ranging from a 13- to 15-point increase in reading scores. Students’ engagement in reading lessons was statistically significant only in Sweden and school emphasis on academic success is positively related with students’ reading achievement only in France. Unexpectedly, in Sweden being in a school where students report higher engagement in reading lessons has a negative relationship with students’ reading achievement. In France being in a school where there is a higher emphasis on academic success benefits students’ achievement. All other school-level variables were not significantly associated with reading achievement.

Overall, our model accounts for 24–32% of the total variability in reading scores. More specifically, we found that 24% of variance in Denmark, 30% in France, and 32% in Sweden was explained by school, home literacy, and home background control variables.

Discussion

Results indicate that factors associated with the home literacy model—book reading and alphabet knowledge—are strongly associated with students’ achievement in the three countries. Reading for enjoyment, as measured by the students like reading scale, is also related with reading attainment in all countries. At the school level, there is a compositional effect whereby students in schools with a higher socioeconomic intake, measured by home resources for learning, have higher reading achievement. In Sweden, students’ report of classroom engagement in reading lessons, capturing whether they like what they read in school and think their teachers give them interesting materials to read, is negatively related to achievement; lower interest relates to higher achievement. This is in line with findings by Stancel-Piatak et al. (2013) for Germany, France, and Denmark. In their study in all of these countries there was a negative association between the students’ report of classroom instruction and students’ achievement. This finding is surprising as one would expect more interested students to have better attainment, especially in a country like Sweden where parents can choose the school their children attend.

Our study also shows that socioeconomic status both at the student level and at the school level is similarly related to achievement in Denmark, France, and Sweden. As such, although equity may be compromised, results suggest that it is no more compromised in systems that have school choice than in the ones without school choice, at least at the fourth grade level. This seems to contrast with the finding that social selection, explored with PIRLS 2006 data, characterizes independent schools in Sweden (Myrberg & Rosén, 2006). However, our study uses a more recent data-set—PIRLS 2011—and addresses the relationship between socioeconomic status and achievement across countries, not within the same country. In this sense, it is reasonable to assume that cross-country comparisons suggest, as analyses of PISA 2012 show, that school choice does not improve students’ learning outcomes and may compromise equity (OECD, 2013). The other statistically significant effect related to school effectiveness was found in France. More specifically, the emphasis a school places on academic success, with a higher emphasis corresponding to higher achievement, was significant only in France. Taken together, these findings suggest that school effects are country specific, whereas student/home variables have a similar relationship with achievement in all three countries.

Using PIRLS 2011 data, this study identifies home background and school variables that are related to reading achievement at a time when students have moved from learning to read to reading to learn (Chall, 1996). The effect of student/home variables is ubiquitous and significant in the countries under analysis whereas the effect of school level factors differs by country. This study is, to our knowledge, the first to consider how letter knowledge and parental book reading are related to PIRLS achievement. Previous studies used only letter knowledge (Alivernini, 2013) or letter knowledge and book reading and storytelling together as a latent construct (Myrberg & Rosén, 2009). Thus, and because PIRLS measures the comprehension of authentic texts, this study adds new evidence to previous findings. Existing studies have investigated the relationship between reading predictors and the ability to read isolated word lists as outcome measure (Ziegler et al., 2010). As such, this study adds to our understanding of the relationship between reading predictors—letter knowledge and parental book reading before compulsory school starts—and reading comprehension of authentic texts.

Importantly, this study shows that comparable literacy knowledge and skills in different European countries are positively associated with reading achievement and this has implications for practice. The findings corroborate those of Sénéchal (2012) with Canadian children and support the notion that book reading in the home should be encouraged prior to the start of compulsory education. The positive difference this practice can make in reading achievement offsets the influence of family social background, as it was also found in Australia (Kalb & van Ours, 2014). Additionally, the findings of this study support the notion that pre-primary school curricula should encompass the teaching of the alphabet. As was the case in the study by Alivernini (2013) with PIRLS data, knowing the letters of the alphabet is clearly associated with higher reading scores. These can be taught at school; alphabet teaching during kindergarten has been found to boost reading comprehension in later grades (Piasta & Wagner, 2010) and can be part of a balanced reading curriculum that promotes children’s interest in letters and written texts during pre-primary education (Soodla et al., 2015).

In short and according to Shager et al. (2013), “achievement-based skills such as early reading, early math, and letter recognition skills appear to be more sensitive to Head Start attendance than cognitive skills such as IQ, vocabulary, and attention which are less sensitive to classroom instruction” (p. 90). This study cannot determine if children learned the alphabet at home or at school, but it clearly suggests that both alphabet skills and the knowledge acquired from home book reading are positively related to reading achievement. As evidence from other studies suggests, in the case of book reading the knowledge accrued is associated with vocabulary knowledge (Perfetti, Landi, & Oakhill, 2005).

Finally, this study offers compelling evidence of the ubiquitous importance, in different European countries, of specific student/home predictors and of the influence of particular school factors in the development of reading ability. Moreover, the results for the predictors hold in countries with different levels of reading achievement and with diverse system-level characteristics. Therefore, it adds converging evidence that initiatives to encourage home book reading, alphabet knowledge, and recreational reading should be considered in the different European countries analyzed. This is one advantage of using PIRLS: the opportunity to provide comparative indicators on the basis of the theoretical constructs integrated in the assessment framework (Lenkeit et al., 2016).

Although the PIRLS survey follows a cross-sectional design, other longitudinal studies point in the same direction, namely those investigating the influence of book reading and alphabet knowledge on reading achievement (Kalb, & van Ours, 2013; Sénéchal & LeFréve, 2002). Importantly, Sénéchal’s (2012) longitudinal research confirms that early teaching of the alphabet may facilitate basic skills and reading fluency (Catts, 2009), but this advantage may not extend to eventual success in reading comprehension without the additional support provided by early home reading. The early development of reading skills and the exposure to home reading may also result in a greater desire to read for pleasure. In essence, this encapsulates the Mathew effect (Stanovich, 2000) and the reciprocal causality between reading ability and reading for pleasure. Those that are more able read more and, as a result, become more and more able readers. Future studies might want to focus on how children develop different reading profiles. For example, are children who have been read to frequently more likely to read more for enjoyment once they acquire reading skills? Research clearly shows that good readers read more and thus become even better readers (Stanovich, 2000). However, empirical evidence on how frequency of home book reading during the preschool years might be related to future frequency of reading for enjoyment is lacking.

Some possible limitations of the present analysis need, however, to be mentioned. As previously noted, PIRLS follows a cross-sectional study and this raises questions about causal inferences. Although this remains a limitation, when considering the theoretical background that informs this study we also considered empirical convergent evidence (Stanovich, 2000) from longitudinal studies. Another limitation concerns the retrospective nature of the parental responses to the home questionnaire. It is possible that social desirability might have influenced parents’ indication of the frequency of early literacy practices and of their children’s ability to recognize the letters of the alphabet. Still, the advantage of using PIRLS is that the data collected for different countries allows for cross-country comparisons across different time waves. Importantly, previous studies using the different waves of PIRLS (2001, 2006, 2011) data have used the same (Alivernini, 2013) or similar measures from the home questionnaire (Myrberg & Rosén, 2009) as predicting variables of students’ achievement and have obtained similar results (Araújo & Costa, 2012). The stability of results over time reduces potential social desirability, or at least keeps that factor constant over time as a country-specific characteristic.

In the case of the present study, the comprehensive nature of the data-set allowed for the investigation of which student/home characteristics and school factors are related to reading achievement. More specifically, this study identifies what is unique in different school systems and what are the common home background factors associated with reading attainment that are observed in varying contexts. In this sense, this study adds to our understanding of the contribution of home and school environments to reading literacy achievement in different countries.

Disclaimer

The views expressed are purely those of the writers and may not in any circumstances be regarded as stating an official position of the European Commission.

Disclosure Statement

No potential conflict of interest was reported by the authors.

ORCID

Patrícia Costa http://orcid.org/0000-0002-2943-4270

Luísa Araújo http://orcid.org/0000-0001-6491-5777

Notes

1 Fourth grade TIMSS—The Third International Mathematics and Science Study 2011.