4.1. An overview
As we have already emphasized, parents’ involvement is an active area of
The recent widespread development of initiatives by schools to involve
parents is rooted in the belief that what parents think and do is significant to
educational outcomes: as Hoover-Dempsey & Sandler [1997, p. 8] write,
“while schools cannot realistically hope to alter a student’s family status,
schools may hope to influence selected parental process variables in the
direction of increased parental involvement”.
Existing programmes can be classified along a variety of dimensions and
differ in many ways from each other. An influential classification distinguishes programmes according to the type of involvement that schools try
to foster. Joyce L. Epstein distinguishes six types of involvement (Epstein
; Epstein & Dauber ):
Type I Involvement in basic obligations at home (the provision of school
supplies, general support and supervision at home).
Type II School to home and home to school communications.
Type III Assistance at the school (volunteering).
Type IV Assistance in learning activities at home.
Type V Involvement in school decision-making, governance and advocacy.
Type VI Collaboration and exchange with community organizations.
Parental involvement programmes typically address more than one type
of involvement; an additional component to many existing school-based
parenting programmes is parent academic education (e.g. language training
for non-native speakers), in an attempt to increase their skills.
In their review of 41 US parent involvement programmes, Mattingly et al.
 note that most programmes are multidimensional and include on
average three to four components (defined as above). A majority of programmes include components to increase parental involvement in home
learning (75 %), to improve parenting skills (61 %), or to improve parent/
school communications (54 %).
More recently Desforges & Abouchaar  have suggested that
attempts to promote parental involvement in school can be classified into
three categories: first, programmes which focus on the immediate connectivity between schools and parents; next, programmes which cast parental
involvement more broadly in the context of family and community education programmes; thirdly, parent training programmes aimed at promoting
parental psychosocial health and/or relationship skills which are known to
be foundational to parental involvement. This classification distinguishes
programmes with a more narrow focus on promoting children’s levels of
achievement from programmes that have broader objectives.
4.2. Evaluations of parental involvement programmes
The perception that parental involvement has a positive effect on students’
academic success has become almost common sense, and has influenced
the development of parental involvement programmes. Attempts to increase
parent involvement are a regular feature of national, state, and local education policies in the US and the UK.
Research on interventions to promote parental involvement has, however,
mostly failed to deliver convincing measures of their impact; in the summary of their review, Desforges & Abouchaar [2003, p. 5] noted that “evaluations of interventions are so technically weak that it is impossible on the
basis of publicly available evidence to describe the scale of the impact on
In a similar vein, a review of evaluations of US programmes by Mattingly
et al.  concludes:
“The exponential development of parent involvement programs, many
funded by federal and state grants, is both promising and troublesome [...]
the effectiveness of various program designs and components remains
unknown (p. 553). The majority of studies we assessed had weak evaluation
designs (p. 568). A majority of the measured outcomes did not show a
significant improvement in studies that used the most stringent criteria [...]
suggesting that the purported effectiveness of many parent involvement
programs is an artifact of weak evaluation methods” (p. 571).
Indeed, research into parental involvement has devoted much effort to
capture contextual issues – determinants, barriers, conditioning factors –
while intervention studies, that would help understand what programmes
are most effective, are rare or of low quality.
Any statement on the impact of family involvement programmes rests on
some form of comparison: impact is defined as the change delivered by the
programme, and its measure supposes the comparison of the observed
end-point with some measure of what would have happened in the absence
of the programme – a counter-factual situation. In randomized control trials,
the design of the intervention delivers a credible measure of this counterfactual because before intervention, test and control groups are – by construction, given the properties of random assignment – two representative
samples of the population targeted by the programme.
The first school-based parenting programme to have been submitted to
impact assessment using rules that randomly assign subjects to test and
control groups is the SPOKES programme (Supporting Parents on Kids’
Education in School) (Scott et al. ). This is an intervention which combines an adult literacy programme, focused on reading readiness, with
parenting support (the “Incredible Years” videotape programme) delivered
to families in a disadvantaged area in South London, with most eligible
families from ethnic minority groups. Parents were recruited at the primary
school of their children. The evaluation found significant changes in parenting attitudes (increased parental sensitivity, more child-centered parenting,
increased use of calm discipline in response to unwanted behavior, reduction in criticism); the trial was also associated with an increase in the child’s
attention on task, while no significant change could be measured in antisocial behavior nor in child reading ability.
Balli et al.  conduct and analyse a very small-scale experiment – one
involving three classes, that assigns them randomly to two treatments and
one control – where the two possible treatments consist in different levels of
invitations to parents for involving with their childs’ mathematics homework. Due to the small number of randomization units (3), this study qualifies for a pilot study more than for a full-scale experiment, and any inference
must be regarded cautiously. The authors find that prompting parents to get
involved is an effective way of increasing their level of involvement; they do
not, however, detect differences in the performance of students from the
A somewhat more systematic attempt at evaluating rigourously parental
involvement programmes exists in the context of early childhood interventions. The U. S. Even Start programme (see, e.g. Ricciuti et al. ) for
instance includes an experimental sample. This is an ongoing family literacy
programme targeted at low-income families (see U. S. Department of Education, 2008, p. 152): under this scheme, families are provided with interactive
parent-child literacy activities, parenting education, as well as more general
adult education and early childhood education. The evaluation sample consists of 463 families (309 test and 154 control); no study on this evaluation
sample finds significant improvements associated with the programme,
although the design has adequate statistical power (ability to detect effects
of a certain magnitude, given the sample size) only if the programme was
expected to deliver large effects.
When randomization is not included in the intervention design, evaluation
rests on assumptions that make some non-randomly chosen control group
(for quasi-experimental studies) or the past (for pre-post studies) representative of the situation of the programme beneficiaries in the absence of the
programme (the counterfactual).
Among quasi-experimental studies, the most rigorous construct the control group by matching on observable characteristics. Sheldon  evaluates the impact on student attendance of the “National Network of Partnership Schools” programme in 69 participating school, by ex-post
constructing a comparison group of 69 non-participating schools with similar demographic characteristics, and finds a positive difference in favor of
Other matched control groups evaluations of parental involvement programmes are reviewed in Mattingly et al. ; these fail generally to
provide evidence of the effectiveness of the programme with respect to
Most studies however claim to identify an impact of parental involvement
from pure cross-sectional correlations between outcomes and inputs. This is
equivalent to saying that any surveyed group that did not experience the
programme is a valid control group. The most sophisticated among these
studies use longitudinal data to define outcomes in terms of change, but do
not take first differences for inputs (Sheldon & Epstein [2002, 2005]) 
A similar approach is used by (Hallam et al. ).... This
may improve precision in the estimated coefficients, if outcomes are serially
correlated, but does not alter identification hypotheses with respect to
simple cross-sectional correlation. The vast majority of “evaluations” are, in
fact, correlational; the typical method is a regression which includes controls
for confounding factors (when available), in which the dependent variable is
a measure of parental involvement or of student achievement, and the independent variable of interest is a scale of the schools’ effort to improve
parental involvement (see, e.g. Sanders et al. ).
Finally, a number of studies more modestly limits itself to qualitative statements about impact (see, e.g. Harris & Goodall ). The interested reader
may be referred to Mattingly et al. , a meta-analysis of 41 evaluations
of parental involvement programmes at US schools, and Desforges &
Abouchaar  for the UK.
To summarize the sparse evidence on the causal impact of parental
involvement programmes, a common and undisputed finding is that levels
of involvement can be raised. However, as Desforges & Abouchaar [2003,
p. 70] write, “the jury is out on whether this makes a difference to pupil
achievement”. Available evidence suggests that effects are more important
on non-cognitive abilities (patience, self-control, ...) than on pure cognitive