Pupil has iPhone (1). Yes (12%), No (84%), Missing (4%). Pupil has TV in bedroom (1). Yes (51%), No (45%), Missing (3%). Parent report of expectations for ...
RUNNING HEAD: READING ACHIEVEMENT AND PARENTAL INVOLVEMENT
1
1
Parental involvement and pupil reading achievement in Ireland: Findings
2
from PIRLS 2011
3
Introduction
4
Research has shown that parental involvement is positively associated with a range of
5
student outcomes such as achievement, level of educational attainment and behaviour
6
(Byrne & Smyth, 2010; Desforges & Abouchaar, 2003; Fan & Chen, 2001; Klemenčič,
7
Mirazchiyski & Sandoval-Hernández, 2014; McNeal, 1999; OECD, 2012; Postlethwaite
8
& Ross, 1992). Furthermore, parental involvement has been shown to mitigate the
9
negative effects of low socioeconomic status (Dearing, Kreider, Simpkins & Weiss,
10
2006; DENI, 2006). Not surprisingly therefore, education systems focus on involving
11
parents in their child’s education (DES, 2011; DfE, 2010; U.S. Department of
12
Education, 2002). One way in which parental involvement is encouraged is by ensuring
13
that parents are provided with information about their child’s school. In many countries,
14
schools are legally required to provide this information.
15
In the USA, the No Child Left Behind Act (NCLB; U.S. DoE 2002) requires
16
parents to be provided with information on the performance of students, schools, and
17
school districts. Reforms to the NCLB (U.S. DoE, 2008) further emphasised
18
communication between schools and parents, requiring schools to give timely and clear
19
notification to parents of school choice and supplementary educational services. Since
20
2006, parents in Northern Ireland (2006) are legally entitled to receive an annual report
21
about their child’s progress at school. Although the role of parents in education has long
22
been officially recognised in the Republic of Ireland, the 2011 Strategy for Literacy and
23
Numeracy (DES, 2011) highlights the need for improved information for parents about
24
their children’s learning. In England, the 2010 Schools White Paper (DfE, 2010)
25
requires schools to be accountable to parents and to provide sufficient information on
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT
2
26
student and school performance. It has been shown however, that it is not always easy
27
for parents to interpret information about schools, in particular performance data, so
28
increasing quantities of information available to parents does not necessarily support
29
greater levels of involvement in children’s education (De Wolf & Janssens, 2007).
30
Parental involvement is also promoted by ensuring that parents have reasonable
31
access to school staff and sufficient opportunities to volunteer (U.S. DoE, 2002). Again,
32
these may not necessarily lead to the type of parental engagement desired by policy
33
makers. Having access to school staff does not necessarily support a partnership
34
between teachers and parents if the access is provided though formal parent-teacher
35
meetings. The parent may be a passive recipient of information and meetings can serve
36
to reinforce existing power relations (Henderson & Mapp, 2002; Power & Clark, 2000).
37
Involvement through volunteering has a negligible association with achievement,
38
although it may have other benefits such as fostering community spirit (Sui-Chu &
39
Willms, 1996; Okpala, Okpala & Smith, 2001). It is likely that in order to influence
40
student achievement, deeper engagement in learning is required on the part of parents. It
41
is therefore useful to distinguish between parental involvement in school-based
42
activities, which may include activities such as volunteering, and parental engagement
43
with their child’s learning (Harris & Goodall, 2007).
44
This paper examines two questions. Firstly, is parental involvement across a
45
broad spectrum of activities related to pupil reading achievement in Ireland? Secondly,
46
what aspects of parental involvement are most important in explaining variation in
47
reading achievement?
48
Dimensions of parental involvement and association with pupil outcomes
49
Bourdieu’s (1977) concepts of the different types of capital – economic, cultural, and
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT
3
50
social – can be used to provide a theoretical underpinning for the importance of parental
51
involvement in education. Parental involvement is inextricably linked to the concept of
52
cultural capital in particular. Kellaghan (2001) operationalises cultural capital in an
53
educational context by outlining the structures and activities of homes that are
54
associated with it. The six components he describes that are associated with cultural
55
capital are: (i) modelling (in the use of complex language and in planning and
56
organisation); (ii) stimulation to explore and discuss ideas and events; (iii) providing
57
motivation for, guidance in, and reinforcement of school-related activities and
58
independence in decision-making; (iv) holding and communicating high academic
59
aspirations and expectations; (v) providing guidance on school matters and direct
60
instruction, and monitoring/helping with homework; and (vi) ensuring that activities
61
engaged in are developmentally appropriate.
62
Parental involvement also relates to social capital. McNeal (1999) suggests that
63
parental involvement in school organisations is an example of social capital as it gives
64
parents access to shared information through their extended network. He argues that
65
such involvement is less likely to be associated with academic achievement and more
66
likely to be protective against behaviours such as truancy, as parents in the network take
67
greater collective responsibility for each other’s children. Parental involvement is less
68
likely to represent a form of economic capital, although Lareau (1989) posits that
69
parental involvement is greater amongst those high in economic capital.
70
Various typologies of parental involvement have been proposed, including
71
influential work by Epstein (1992) and Hoover-Dempsey and Sandler (1995, 1997).
72
Epstein considers the overlapping spheres of influence of the family, the school and the
73
community while Hoover-Dempsey and Sandler examine the psychological
74
underpinnings of parents’ decisions to become involved in their children’s education. In
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT
4
75
reviewing empirical studies of parental involvement, Sui-Chu and Willms (1996)
76
distinguish between two types of home involvement and two types of school
77
involvement. The first type of home involvement relates to discussing school activities;
78
the second relates to monitoring the child’s out-of-school activities. The first component
79
of school involvement refers to contacts between parents and school personnel; the
80
second relates to involvement in activities such as volunteering in the school and
81
attending parent-teacher meetings.
82
McNeal’s (1999) dimensions of parental involvement can also be separated into
83
those that relate to parenting activities that take place in the home and those that involve
84
parents more directly in school activities. The four dimensions he identifies are parent-
85
child discussion; parent involvement in school organisation; monitoring; and direct
86
involvement in educational processes. McNeal posits the importance of discussion as a
87
theoretical mechanism which allows the importance of education to be conveyed to the
88
child. In addition, discussion allows parents to identify disengagement at an early stage.
89
McNeal suggests that monitoring activities, such as setting limits on time spent
90
watching television, reflect a more structured home environment. However, as greater
91
monitoring may be adopted as a consequence of children underperforming, McNeal
92
hypothesizes that associations between achievement and monitoring practices would be
93
inconsistent or negative. Empirical research has indeed reported contradictory findings
94
on the association between monitoring practices and student outcomes. In particular,
95
studies have found negative associations between achievement and parental help with
96
homework (e.g., Eivers et al., 2010; Madigan, 1994; Muller, 1993). In Ireland, findings
97
from the National Assessments of Reading and Mathematics show that pupils who
98
receive more help from parents with homework had lower mean scores in reading and
99
mathematics (Eivers et al., 2010). Furthermore, a moderate negative correlation is
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT
5
100
reported between time spent on homework and achievement in reading and mathematics
101
(but cf Keith, Reimers, Fehrman, Pottebaum, & Aubey, 1986). Eivers et al. conclude
102
that weaker pupils may need more help and more time for homework (see also Epstein,
103
1988). A positive association was reported in an earlier Irish national assessment
104
between achievement and parental rules about time spent watching TV or playing
105
computer games, and between achievement and parental rules about the content of
106
permissible TV programmes and computer games (Eivers, Shiel, Perkins & Cosgrove,
107
2005).
108
Turning to direct involvement in educational processes, this refers to activities
109
such as attending parent-teacher meetings or speaking to the child’s teacher. It is likely
110
that these forms of parental involvement may increase when a child is experiencing
111
difficulties. In Ireland, 2004 National Assessment findings showed that an increased
112
number of home-school contacts was associated with a significant decline in pupil
113
achievement, although direction of causality cannot be determined as it was a cross-
114
sectional study (Eivers et al, 2005).
115
Previous research at secondary-level in Ireland distinguished between formal
116
and informal parental involvement (Byrne & Smyth, 2010). Formal involvement
117
includes activities such as attending meetings, talking to teachers, involvement with the
118
parents’ association and being a member of the board of management. Activities such as
119
helping with homework and study, discussing progress in school and giving advice
120
about programme and study options are classified as informal parental involvement.
121
Both formal and informal parental involvement measures were associated with student
122
outcomes. Parent-child discussion and attendance at parent-teacher meetings were
123
positively associated with achievement, while checking homework was negatively
124
associated with achievement (Byrne & Smyth, 2010).
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT 125
6
Despite the general recognition of the key role played by parents in their
126
children’s education, research findings in the area are often inconsistent (e.g. Epstein,
127
1988; Madigan, 1994; McNeal, 1999; Muller 1993). One possible explanation is that, as
128
previously noted, parents may become involved in school activities only when students
129
are struggling (OECD, 2012). While the converse has also been shown, i.e., that parents
130
are more involved when children have higher levels of attainment (Harris & Goodall,
131
2007), the possible confounding effect of parental socioeconomic status cannot be
132
overlooked (Lareau, 1989).
133
Other reasons for inconsistencies in the associations between achievement and
134
parental involvement relate to differences in the achievement measures used and
135
differences in the interpretation of what constitutes parental involvement. As noted
136
earlier, it is relevant to distinguish between parental involvement in school activities and
137
deeper engagement in student learning on the part of parents. A review by Desforges
138
and Abouchaar (2003) found that the aspects of parenting that relate to activities in the
139
home are more strongly associated with achievement and adjustment than other aspects
140
of parental involvement, such as those in the school. A meta-analysis by Fan and Chen
141
(2001) found that parental expectations and aspirations have the strongest association
142
with academic achievement, while supervision at home has a much weaker association.
143
A limitation of the analysis by Fan and Chen was the lack of control for parental
144
socioeconomic status. Sui-Chu and Willms (1996) also show that high expectations for
145
children’s performance and general monitoring of performance are positively associated
146
with achievement. Conversely, helping with homework and attending parent-teacher
147
meetings do not have strong effects and may be negatively associated with achievement.
148
A final reason that may account for inconsistencies in research findings relates to a
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT
7
149
failure to take into account variations in the national context in which parental
150
involvement is taking place.
151
Parental involvement in the Progress in International Reading Literacy Survey
152
The Progress in International Reading Literacy Survey (PIRLS; Mullis, Martin, Foy &
153
Drucker, 2012) offers an opportunity to examine parental involvement in education and
154
its association with reading achievement at Fourth grade. An advantage of the PIRLS
155
dataset is that it allows detailed consideration to be given to the national context when
156
considering the associations between parental involvement and student outcomes.
157
Taking, for example, parental support for homework as an indicator of parental
158
involvement, clearly PIRLS data can be used to gain an insight into the variation across
159
participating countries. PIRLS 2011 revealed a strong reliance on homework for home-
160
school communication in Ireland (Eivers & Creaven, 2013). Almost all parents in
161
Ireland (95%) reported setting aside time for homework on a daily or almost daily basis.
162
This is a higher percentage than in most other high performing PIRLS countries such as
163
Hong Kong SAR (68%), Russian Federation (87%), Finland (77%), and Singapore
164
(72%). Similar to parents in the Russian Federation (71%), almost 70% of parents in
165
Ireland indicated that they help with homework daily or almost daily. This compares to
166
just 26% of parents in Finland, 50% in Singapore and 56% in Hong Kong SAR. While
167
homework activities are supported by parents in Ireland, the prevalence of private
168
tutoring in some other countries suggests that elsewhere, help with homework may be
169
supported by tutors (Bray, 2007).
170
Compared to the average levels across PIRLS countries, Irish schools were far
171
less likely to give parents regular updates on the behaviour and well-being of their child,
172
and less likely to discuss parents’ concerns or wishes about their child’s learning (Eivers
173
& Creaven, 2013). In spite of this, Irish parents were very positive about their child’s
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT 174
school and almost all respondents indicated that the school both provided a safe
175
environment and cared about their child’s education.
176
8
Irish parents were less likely than those in many other countries to receive
177
information about the school in general, such as updates about general academic
178
achievement or about the school’s educational goals. One quarter of parents in Ireland
179
never received updates about the overall academic achievement of the school. Among
180
the highest performing countries in PIRLS 2011, i.e., those with significantly higher
181
mean reading scores than Ireland, only Finland (32%) was similar in this regard (Eivers
182
& Creaven, 2013).
183
PIRLS also revealed similarities across countries regarding parental
184
involvement. In particular, parental support for academic achievement was a universally
185
positive correlate of reading achievement (Eivers & Creaven, 2013). Also, a majority of
186
participating countries had policies for the involvement of parents in their children’s
187
education (Mullis, Martin, Minnich, Drucker & Ragan, 2012).
188
Previous research in the area of parental involvement using the PIRLS 2011
189
dataset has used a single measure of parental involvement derived from school principal
190
responses (Klemenčič et al., 2014). It also used a limited number of explanatory
191
variables across a large number of countries. In the current paper, data from the School,
192
Teacher, Parent and Pupil questionnaires are used to give a more detailed insight into
193
parental involvement in children’s educational activities. Also, this paper focuses on the
194
Irish context as this allows a more detailed look at parental involvement in one
195
education system. Much previous work on formal parental involvement in Ireland was
196
conducted in secondarylevel schools (Byrne & Smyth 2010; Cosgrove, Gilleece &
197
Shiel, 2011). As parental involvement declines and reduces in effectiveness as children
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT
9
198
get older (Crosnoe, 2001; Domina 2005), it is pertinent to use the PIRLS dataset which
199
provides data on primary school children (aged about 10).
200
Data and Method
201
Data for this study are drawn from the combined 2011 PIRLS/TIMSS (Trends in
202
International Mathematics and Science Study) database (Mullis, Martin, Foy & Arora,
203
2012; Mullis, Martin, Foy & Drucker, 2012; for Irish results, see Eivers & Clerkin,
204
2012), focussing on reading achievement as the outcome measure. Reading was selected
205
as the dependent variable as measures of home background have been shown to be
206
somewhat more effective at explaining variance in reading achievement than
207
achievement in other domains (Cosgrove & Creaven, 2013; Sui-Chu & Willms, 1996).
208
Pupils in Ireland achieved a mean reading score in PIRLS 2011 that was significantly
209
above the PIRLS scale centrepoint (i.e., a reference point determined in the first cycle of
210
PIRLS in 2001) and were ranked tenth of participating countries.
211
In 2011, the administration of PIRLS coincided with the administration of
212
TIMSS and Ireland, as well as many other participating countries, participated in both
213
assessments. In addition to the achievement tests, the studies gathered extensive
214
demographic and background information from pupils, parents, teachers and school
215
principals. For the current study, measures of parental involvement in their children’s
216
education have been derived from questions on the pupil, parent and school
217
questionnaires. Relevant questions on the pupil questionnaire asked about the frequency
218
with which parents ask the child what he/she is learning in school, the frequency of
219
talking about schoolwork with parents, the frequency of parents ensuring that time is set
220
aside for homework and the frequency of parents checking homework. The parent
221
questionnaire asked parents of participating pupils for comparable information. Each
222
country participating in PIRLS added up to six country-specific wealth indicators in the
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT
10
223
pupil questionnaire. In Ireland, pupils were asked if they had their own iPhone and
224
whether or not they had a television in their bedroom. Thus, quite detailed information
225
is available at the pupil level about pupils’ home background and the informal
226
involvement of parents in educational activities in the home. While direct instruction on
227
the part of parents at home is likely to be associated with achievement, parent questions
228
on the frequency of activities such as reading books with the child and telling stories
229
refer to the period before the child began primary school. Given the potential for
230
changed circumstances and the length of time elapsed since the child started school,
231
these variables were not included in this analysis.
232
The school questionnaire asked for information about more formal involvement
233
of parents in activities such as volunteering for school trips or serving on school
234
committees. Thus, measures of formal or school involvement are available at school
235
level only.
236
It is particularly useful to examine the PIRLS 2011 data for Ireland as PIRLS
237
took place prior to the implementation of the National Strategy for Literacy and
238
Numeracy (DES, 2011). Therefore, it allows an examination of the role of parental
239
involvement prior to any increase that may be expected to take place in response to the
240
strategy.
241
Sample
242
TIMSS and PIRLS employ a two-stage random sampling process, first selecting a
243
stratified sample of schools and then one or more classes within each school (Martin &
244
Mullis, 2012). In Ireland, 4524 pupils from 221 classes in 151 schools completed the
245
PIRLS test, representing a 93.8% pupil response rate (Eivers & Clerkin, 2012). Parent
246
questionnaires were returned for all participating pupils. The response rates for the
247
Teacher and Principal questionnaires were 99.5% and 96.0%, respectively. After
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT
11
248
deletion of cases missing data on multiple variables, the current analysis is based on
249
4231 pupils from 209 classes, representing 94% of the dataset. The mean reading
250
achievement of the full Irish sample and the sample used in the current analysis were
251
551.6 (SE = 2.26) and 552.4 (SE = 2.35), respectively.
252
Variables
253
The dependent variable in the current analysis is reading achievement in PIRLS 2011.
254
Reading achievement is represented by five plausible values. This methodology uses
255
background data and pupils’ responses to test items to impute multiple achievement
256
scores. This leads to more accurate estimation of population parameters (Beaton, 1987;
257
Martin & Mullis, 2012; Mislevy, 1991). It is necessary to employ this approach in
258
PIRLS as the rotated booklet design means that not all pupils have answered all
259
questions; therefore, it is necessary to infer missing values to get overall estimates of
260
achievement.
261
Independent variables were divided into those that could be considered to
262
represent background or control variables, informal parental involvement, formal
263
parental involvement and school-home communication variables. One reason for
264
categorising variables in this way is that the formal/informal distinction was used in
265
previous research of parental involvement in second-level education in Ireland (Byrne
266
& Smyth, 2010). Also, informal involvement could be considered an overarching term
267
for two of McNeal’s dimensions (monitoring and discussion), while formal involvement
268
covers involvement in parent-teacher organisations and educational support strategies.
269
The school-home communication group of variables is included as they represent
270
activities initiated by the school rather than the parent and appear to represent the
271
school’s attempts to encourage parental involvement rather than acting as measures of
272
parental involvement per se.
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT 273
12
Variables are classified as either Level 2 (school- or class-level) or Level 1
274
(pupil-level). The rationale for using multilevel modelling and dividing variables into
275
two levels is explained in the next section. The specific variables examined in this
276
analysis are outlined in Table 1 (percentages shown are the percentages of the 4231
277
cases used in the current analysis). All variables (apart from total jobs in the household)
278
were categorical variables, entered in the models using dummy coding (see Table 1).
279
The total jobs variable was not centered in order to facilitate the interpretation of the
280
parameter estimate associated with this variable; the parameter estimate for total jobs
281
corresponds to the expected change in reading achievement associated with one
282
additional job in a pupil’s household.
283
The variables representing pupil ownership of an iPhone or bedroom television
284
are taken as indicators of the degree to which parents monitor their child’s activities. It
285
is expected that having either or both of these gives the child largely unsupervised
286
access to various media. The frequency of the teacher talking to parents about the
287
child’s learning progress is included as a measure of informal parental involvement as
288
this measure allows for phone calls or short chats to be counted. Also, the frequency of
289
teachers sending home progress reports is included as an informal measure as progress
290
reports could encompass an informal note written by the teacher in a pupil’s homework
291
notebook.
RUNNING HEAD: READING ACHIEVEMENT AND PARENTAL INVOLVEMENT
292
13
Table 1: Variables used in modelling: Coding and descriptive information Variable (Level)a and description Outcome: Reading achievement (5 Plausible values)
Coding (weighted % of pupils per category)b Mean = 552, SE= 2.4, SD = 73.9
Background variables DEIS statusc1 (2): Indicator of socioeconomic disadvantage
Band 1 (7%), Band 2 (7%), Rural (4%) & Non-DEIS (82%)
Enrolment size (2): Stratifying variable indicating likely number of fourth grade pupils at test time
Small (≤ 20 pupils, 31%), Med (21 – 34 pupils, 27%) & Large (35+, 42%)
Language of instructionc2 (2)
English (94%) or Irish (6%)
School gender mix (2)
Boys (12%); Girls (9%); Mixed (73%); Infant mixed/girls only (5%)
Pupil gender (1)
Girl (49.9%), Boy (50.1%)
Parental education (1)
≤ Lower sec (11%), Upper sec (27%), Short third-level (ISCED 5b, 23%), ≥ Degree (31%), Missing (8%).
Parental employmentd (1): Total jobs in the household (Range 0 – 2)
Mean=1.2; sd=0.59
Number of books at home (1)
≤10 books (10%); 11 – 25 (22%); 26 – 100 (33%); >100 (32%), Missing (4%).
Frequency of speaking English at homeb (1)
Always (81%), Sometimes (13%); Never (2%), Missing (4%).
Informal parental involvement Teacher report of parental support for pupil achievement (2): if missing, data from principal teacher are substituted
Low (6%), Medium (31%), High (46%), Very high (17%)
Freq – teacher meeting or talking to parents individually to discuss child’s learning progress (2):
≤ 3 times per yr (80%), More often (13%), Missing (7%).
Freq – teacher sending home a progress report on the pupil’s learning (2)
≤ 3 times per yr (83%), More often (11%), Missing (7%).
Pupil has iPhone (1)
Yes (12%), No (84%), Missing (4%)
Pupil has TV in bedroom (1)
Yes (51%), No (45%), Missing (3%)
Parent report of expectations for pupil educational attainmentc3 (1)
≤ PLC (10%), Diploma (19%), ≥Degree (64%), Missing (8%).
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT
14
Freq – pupil talks about schoolwork to parentse (1)
Never or almost never (12%), Once or twice a month (8%), Once or twice a week (29%), Daily (52%).
Freq - parent asking pupil about learning in school (1)
Never or almost never (6%), At least monthly (31%), Daily (63%).
Freq – parents ensuring pupil sets aside time for homeworke (1)
≤ 2 a month (10%), Once or twice a week (7%), Daily (83%).
Freq – parents checking if pupil does homework e (1)
Daily (86%), 1 or 2 per week (9%), Monthly or less (6%), Missing (1%)
Formal parental involvement (principal reports) Freq – school asks parents to volunteer for school projects (2)
Never (9%), Once per yr (19%), 2-3 times (41%), > 3 per yr (31%).
Freq –parents asked to serve on sch committees (2)
Once a year or less (48%), 2-3 times (24%), > 3 per yr (28%)
Freq – workshops or seminars for parents on learning or pedagogical issues (2)
Never (43%), 1-3 times per yr (49%), > 3 (8%).
Level of parental involvement in school activities (2)
Low or very low (17%), Medium (37%), High or very high (46%).
Classroom has adult volunteer to work with pupils who have reading difficulties (2)
Always (2%), Sometimes (13%), Never (77%), Missing (7%).
School-home communication (based on principal reports) Freq – inform parents – child’s learning progress (2)
Once a year or less (17%), 2-3 times (71%), > 3 times per yr (13%)
Freq – informing parents about behaviour and wellbeing of child at school (2)
Once a year or less (11%), 2-3 times (69%), > 3 per yr (21%)
Freq – discussing parents’ concerns or wishes about child’s learning (2)
Once a year (17%), 2-3 times (50%), > 3 per yr (33%).
Freq – informing parents – overall academic achievement (2)
Never (26%), Once a year (55%), 23 times per yr or more (20%)
Freq – informing parents – sch accomplishments (2)
Once a year or less (8%), 2-3 times (30%), > 3 per yr (63%).
Freq – informing parents – educational goals & pedagogic principles of school (2)
Never (19%), At least yearly (78%), Missing (3%)
Freq – inform parents – school rules (2)
Once a year or less (64%), 2-3 times (21%), > 3 per yr (14%), Missing (0.7%)
Freq – discuss parents’ concerns or wishes about school’s organisation (2)
≤Yearly (46%), 2-3 times (34%), >3 per yr (20%), Missing (0.5%)
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT 293 294 295
a
15
Level 2: school or class, Level 1: pupil; bBold: reference category cNFootnote N has further detail.dVariable has missing indicator to preserve cases in the dataset; eIf pupil data are missing, parent data are substituted where possible.
RUNNING HEAD: READING ACHIEVEMENT AND PARENTAL INVOLVEMENT
16
296
Method
297
Given the hierarchical nature of the data (pupils within classes within schools),
298
multilevel modelling using MLwiN (Version 2.30, Rasbash, Browne, Healy, Cameron,
299
& Charlton, 2014) was employed. This approach allows the contribution of various
300
explanatory variables to be examined simultaneously. For each independent variable in
301
the model, the parameter estimate describes the linear relation between this variable and
302
the outcome (reading achievement), having taken into account the effects of all other
303
variables in the model.
304
Earlier analyses of the Irish PIRLS data examined the use of two-level models
305
(pupils within schools) compared to three-level models (pupils, classes and schools) and
306
concluded that a two-level model using class as the grouping variable is most
307
appropriate (Cosgrove & Creaven, 2013). This work showed that even in schools with
308
two classes, there was little variation in reading achievement between the classes (4%),
309
but using class as the Level 2 grouping variable facilitated the inclusion of teacher
310
variables. Therefore, for the current study, the approach advocated by Cosgrove and
311
Craven is adopted, using a two-level structure with class as the grouping variable.
312
Between-class variance is computed on the basis of the null model using the first
313
plausible value as the outcome variable. As many schools had only a single class
314
participating in PIRLS, and given the low between-class variance in schools with two
315
participating classes, between-class variance can be considered as an approximation of
316
between-school variance in this paper.
317
The analytical approach involved testing each explanatory variable compared to
318
the null model for statistical significance (p < .05). The first plausible value only was
319
used for this step. Variables that were statistically significant individually (p < .05) were
320
retained in their block (i.e. background, informal, formal or communication) and blocks
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT
17
321
of variables were then added one at a time to the null model. This was done five times,
322
once for each plausible value. When blocks were added to the null model, variables that
323
were not statistically significant in their blocks were dropped. Once each block was
324
finalised, all blocks were added to the null model together to give the final model. The
325
final model was run five times to get parameter estimates, standard errors, and statistical
326
tests (either chi-square or t) for each plausible value. Variance explained was computed
327
on the basis of the fitted model compared to the null model.
328
The plausible value methodology employed in PIRLS requires certain additional
329
steps to be taken when conducting multilevel analyses. Parameter estimates and tests of
330
statistical significance such as chi-square were first computed for each plausible value
331
separately; i.e., the model was run five times using each plausible value as the outcome
332
variable. The final parameter estimate or test statistic is equal to the average of the five
333
individual estimates (OECD, 2009). Averages should not be computed at the student
334
level as this will lead to a biased estimate (for details, see OECD 2009). Standard errors
335
for the final estimates were computed using the standard errors associated with the five
336
individual estimates as well as the sampling variance, the imputation variance and the
337
number of plausible values.
338
A further complexity associated with large scale survey data relates to the non-
339
straightforward issue of sampling weights (see e.g., Aitkin & Aitkin, 2011;
340
Rabe_Hesketh & Skrondal, 2006; Rutkowski, Gonzalez, Joncas & von Davier, 2010).
341
Relevant literature has examined calculation methods (e.g., Rutkowski et al., 2010),
342
their application in different software packages (Centre for Multilevel Modelling,
343
2011), and alternative approaches (e.g., Aitkin, Francis & Hinde, 2005). Using the
344
weighting documentation of the Irish PIRLS national study centre and following the
345
methods outlined in Rutkowski et al. (2010), class weights and pupil within class
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT
18
346
weights were computed. Given the very high response rates on PIRLS in Ireland,
347
adjustments for non-response were minimal. Using the first plausible value as the
348
outcome measure, an examination of models computed with and without weights found
349
little difference in the parameter estimates and standard errors between the two.
350
Therefore, following Aitkin et al. (2005), stratifying variables were examined as part of
351
the modelling process and weights were not included. The remainder of this paper refers
352
to unweighted models.
353 354
Effect sizes for the model were calculated as follows (Strand, 2004, p.51):
The effect size of a dichotomous variable was calculated as the regression
355
coefficient divided by the outcome standard deviation (SD), i.e., the student
356
level SD in reading achievement associated with the first plausible value;
357
the effect size of a continuous variable was calculated as the regression
358
coefficient multiplied by twice the variable’s SD, divided by the outcome SD.
359
The effect size gives the difference (in terms of standard deviation units)
360
between the predicted reading scores of pupils who are one SD above the mean
361
and those who are one SD below the mean on the predictor variable.
362
Effect sizes are interpreted as small, if less than 0.20; small but noteworthy if between
363
0.20 and 0.50; medium if between 0.51 and 0.80; and strong if above 0.80 (Cohen,
364
1977). The What Works Clearing House (WWC), an initiative of the Institute for
365
Education Sciences at the U.S. Department of Education, define effect sizes of 0.25
366
standard deviations or larger to be “substantively important”. Even if not statistically
367
significant, effect sizes of this magnitude “are interpreted as a qualified positive (or
368
negative) effect” (WWC, 2014, p.23).
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT
19
369
Results
370
When background variables were added one-by-one to the null model, all except school
371
enrolment size were significantly associated with reading achievement. When the
372
significant variables were examined together, school gender composition was not
373
statistically significant. Therefore, subsequent analyses excluded school enrolment size
374
and school gender composition.
375
When measures of informal parental involvement were added one-by-one to the
376
null model, only one of the three class/school variables was statistically significant. The
377
only significant variable was the teacher-reported measure of parental support. No
378
significant effects were found for the frequency with which teachers talked to parents
379
individually to discuss their child’s learning progress nor the frequency with which
380
teachers sent home a progress report on the pupil’s learning. At pupil-level, all but one
381
of the variables were statistically significant. The only variable that was not statistically
382
significant at the pupil-level in the informal block was the frequency of having
383
homework checked by parents.
384
When measures of formal involvement were added one-by-one to the null
385
model, two out of five were statistically significant. The two significant variables were:
386
frequency of workshops or seminars for parents, and, principals’ reports of parental
387
involvement in school activities. There were no pupil-level variables in this block.
388
Of the communications variables examined, only one was statistically
389
significant when added one-by-one to the null model. This was the frequency of the
390
school discussing parents’ concerns or wishes about their child’s learning. This variable
391
was not statistically significant in subsequent analyses that included the background
392
variables. No significant associations were found for the frequency with which schools
393
informed parents about: their child’s learning progress; their child’s behaviour and
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT
20
394
wellbeing at school; the school’s overall academic achievement; school
395
accomplishments; the school’s educational goals and pedagogic principles; the rules of
396
the school; or, the school’s organisation (e.g. rules and regulations, timetables, or safety
397
measures). Given the non-significance of the communications variables, no further
398
reference is made to them.
399
Table 2 presents three models: background variables only; background variables
400
and measures of informal parental participation; and, background variables and formal
401
parental participation. (Effect sizes will be presented later for the final model only.) In
402
interpreting these models, it should be noted that the parameter estimate for a variable
403
represents the association between this variable and outcome, having controlled for all
404
other variables in the model. Looking at Level 2 variables in Model 1 (i.e., the model
405
with background variables only), it can be seen that having controlled for the other
406
variables in the model, pupils in schools where the language of instruction is Irish have
407
significantly higher reading scores than those in schools where the language of
408
instruction is English. Also, all else being equal, pupils in DEIS schools have
409
significantly lower achievement than those in Non-DEIS schools. There is a large
410
standard error associated with the parameter estimate for Rural DEIS, thus the statistical
411
significance of the DEIS variable is largely attributable to the differences between
412
pupils in non-DEIS schools and those in Band 1 and Band 2 schools.
413
Looking at pupil-level variables, Model 1 shows that girls significantly
414
outperform boys, all else being equal (Table 2). Parental education is also statistically
415
significant, with lower achievement among pupils whose parents have completed lower
416
secondary education or below compared to those whose parents have completed upper
417
secondary. Higher levels of achievement are associated with higher levels of parental
418
education such as short third-level courses (e.g. Diploma or non-degree level courses)
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT
21
419
or university qualifications. Having controlled for the various other variables in the
420
model, the numbers of books in a pupil’s home is also relevant: children with 10 or
421
fewer books have significantly lower achievement than those with 26 to 100 books
422
(reference category). Children with 11 to 25 books have significantly lower
423
achievement than those in the reference category, i.e., 26 to 100 books, while children
424
with more than 100 books have significantly higher levels of achievement than those
425
with 26 to 100 books. Children who reported ‘sometimes’ or ‘never’ speaking English
426
at home had significantly lower scores than those who reported ‘always’ speaking
427
English at home. Model 1 explains 73% of between-class variance, 18% of within-class
428
variance and 27% of the total variance in reading achievement.
429
Model 2 presents the results of adding the background variables and the
430
informal participation variables to the null model (Table 2). Looking firstly at the
431
background variables, the parameter estimates for DEIS are seen change substantially
432
between Models 1 and 2, suggesting that informal parental involvement may account
433
for some of the differences found between pupils in DEIS and non-DEIS schools. The
434
effects of parental education are also reduced between Model 1 and Model 2, suggesting
435
that differing levels of informal parental involvement may be associated with parental
436
education. In particular, the parameter estimates associated with short third level
437
courses and university degree are reduced in Model 2 compared to Model 1 so it
438
appears that informal parental involvement may be one mechanism by which parental
439
education plays a role in pupil achievement.
440
All informal measures of parental involvement added to Model 2 are statistically
441
significant after controlling for background characteristics. At school-level, parental
442
support for pupil achievement is statistically significant. Pupils whose teachers report
443
low levels of parental support for pupil achievement have significantly lower reading
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT
22
444
achievement than those whose teachers report medium levels. High and very high levels
445
of parental support are associated with significantly higher reading achievement than
446
medium levels of support (Table 2).
447
All pupil-level measures of informal parental involvement are also statistically
448
significant (Table 2). Having controlled for the other variables in the model, pupils who
449
have an iPhone have significantly lower levels of reading achievement than pupils who
450
reported not having an iPhone. There is quite a large difference of 26 points (roughly
451
one third of a standard deviation) between pupils with and without an iPhone. This is
452
the the net effect of having an iPhone, given the inclusion of all other variables in the
453
model. Also, pupils who have a TV in their bedroom have significantly lower
454
achievement than those without.
455
Parental expectations of academic attainment are significantly associated with
456
reading achievement; i.e. ceteris paribus, pupils whose parents expect them to complete
457
a third-level certificate or diploma have higher reading scores than those pupils whose
458
parents expect them to finish formal education at or below the level of a Post-Leaving
459
Certificate (PLC) course or apprenticeship. Pupils whose parents expect them to study
460
to degree level or beyond have higher average achievement than those expected to study
461
to PLC level or below.
462
Parental frequency of asking about what pupils are learning in school is
463
negatively associated with achievement; i.e., all else being equal, pupils whose parents
464
ask on a daily basis about what they are learning have significantly lower achievement
465
than those whose parents ask on a monthly basis. This suggests that parents may
466
question lower achieving pupils on a more frequent basis about what they are learning
467
in school; i.e., that asking about learning is a reaction to low achievement.
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT 468
23
Talking about schoolwork is found to be positively associated with achievement,
469
although the main advantage is found for pupils who talk to their parents about
470
schoolwork on a monthly basis compared to those who never do. The parameter
471
estimate associated with daily talking about school work is smaller than its associated
472
standard error so there is no significant difference in reading achievement between
473
pupils who never talk about school work (reference category) and those who talk about
474
it on a daily basis. This non-significant difference could be a consequence of the daily
475
discussion group including a mix of pupils, some of whom have regular discussion in
476
response to low achievement and others for whom daily discussion of educational issues
477
is encouraged, regardless of achievement.
478
Having time set aside for homework on a regular basis is associated with higher
479
average achievement, all else being equal. Thus pupils who report that their parents
480
ensure that they make time for homework on a daily basis have significantly higher
481
reading achievement than those who indicated that this took place on a monthly basis or
482
less often. Model 2 explains 81% of between-class variance, 24% of within-class
483
variance and 33% of the total variance in reading achievement.
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT
24
Table 2. Models of reading achievement – background variables only, background and informal parental involvement, background and formal parental involvement. Model 1: Background PE
SE
538.28
7.337
13.99
6.745
Band 1
-20.31
8.434
Band 2
-25.58
8.655
5.46
12.621
p
Model 2: Bkgrd & informal PE
SE
511.66
7.105
11.44
6.073
-1.09
6.075
-12.03
5.699
13.80
8.611
Low
-11.88
6.335
High
8.13
3.347
14.90
4.472
Intercept School level Language of instruction is Irish
*
p
^
Model 3: Bkgrd & formal PE
SE
533.43
4.834
9.03
6.445
-14.09
6.365
-16.96
7.138
9.08
8.992
2.63
4.743
11.65
3.244
p
NS
Participation in DEIS (Ref: Not in DEIS)
Rural
***
^
*
Parental support (Ref: medium)
Very high
***
Parental involvement in school activities (Ref: medium) Low or very low High or very high
***
Frequency of workshops for parents (Ref: 1-3 per yr) Never More than 3 Pupil level Gender (Ref: Boy)
7.43
2.207
-14.65
4.094
Short third lev
11.04
3.147
Degree or above
23.61
3.339
10.73
1.922
Parental
Educationa
4.85
2.119
-9.72
3.925
2.86
3.225
10.58
3.236
8.66
1.851
*
3.258 6.421
7.52
2.185
-14.32
4.091
10.89
3.140
23.93
2.981
10.89
1.910
*
***
(Ref: Upper sec)
Lower sec or below
Parental
***
-5.84 -11.86
employmenta
(Total jobs in household)
*** ***
*** ***
*** ***
READING ACHIEVEMENT AND PARENTAL INVOLVEMENT
25
Books at homea (Ref: 26-100) 10 or fewer
-45.00
3.929
-38.98
3.868
11 to 25
-17.31
2.944
-13.69
2.848
20.78
2.827
18.08
2.735
Sometimes
-13.08
3.079
-12.57
3.009
Never
-43.84
7.501
-40.56
7.308
-26.42
2.977
***
-11.03
2.364
***
Diploma
14.06
4.139
University degree or above
32.99
3.843
Never or almost never
-2.25
4.701
Daily
-7.59
2.341
Once or twice a month
10.93
5.080
Once or twice a week
6.25
4.339
Daily
2.23
3.863
Once or twice a week
-6.50
5.102
Daily
13.60
3.559
More than 100 Pupil speaks English at
homea
Pupil has TV in
***
3.938
-17.21
2.935
21.02
2.817
-12.92
3.085
-43.75
7.505
***
(Ref: Always)
Pupil has iPhonea (Ref: No) bedrooma
***
-44.25
(Ref: No)
***
***
Expectations for pupil's educationa (Ref: PLC or below) ***
Parent asks pupil what he is learning in school (Ref: monthly) **
Parent and pupil talk about schoolwork (Ref: Never) *
Parents ensure time for homework (Ref: monthly or less)
Note: aVariable has missing indicator, *** p