THE DIGITAL DIVIDE: COMPUTER USE,
BASIC SKILLS AND EMPLOYMENT
THE THREE DIVIDES IN PORTLAND
AND LONDON
Structural equation modelling (SEM) was used to
investigate the extent to which literacy proficiency,
computer use and time spent in employment are
related to each other across time and hence can be
considered to influence each other. Other socio-
economic measures - gender, parents’ education,
number of children and ethnicity (the latter in the
US only) - were also taken into account to serve as
statistical controls.
The time sequencing of measures in such a model
can be used to support causal interpretations of the
relationships between the measures involved. For
example, if literacy proficiency in 2000 is more
likely to lead to increased computer use in 2004,
rather than vice versa, then the strength of the
relationship between literacy proficiency in 2000
and computer use in 2004 will be stronger than
between computer use in 2000 and literacy
proficiency in 2004.
The strength of the relationship between any two
measures (e.g. literacy proficiency and
employment) are estimated by statistically
significant standardised regression coefficients
which hold constant the effects of all other
variables included in the SEM model. Such a
regression coefficient has a range from +1 (strong
positive relationship) to -1 (strong negative
relationship).
The first step in the modelling strategy was to set
out all the possible causal relationships between
the three ‘divide’ variables. Estimating the strength
of the relationships between the measures
encompassed by the model was then carried out in
three stages.
1. Portland model
In this we estimated the strength of the
relationships between literacy proficiency, ICT
competence and time spent in employment at both
time points in Portland.
Findings
The model showed clear effects of time spent in
employment and ICT use in 2000 on literacy
proficiency in 2004, and a much more modest effect
of ICT use in 2000 on employment in 2004. More
specifically literacy proficiency showed no direct
effect on employment or ICT use in 2004.
Interpretation
These results indicate that both exposure to ICT
and employment experience probably boost literacy
proficiency rather than the other way round.
Extending the time spent on ICT at work to total
time spent using ICT at home or at work produced
stronger effects, suggesting that in the more
depressed Portland labour market, all ICT
experience, rather than just that gained at work,
was likely to be helpful in getting a job.
2. Comparative model
We then estimated a reduced model with literacy
proficiency only included in 2004 (age 34). This
model does not estimate a change in literacy
proficiency but enables comparison between
Portland and London to be drawn.
Findings
Even without the measure of literacy in 2000, much
the same pattern of relationships between the
three divides was evident in Portland and London,
but there were also differences. First, although the
strength of the path between ICT use in 2000 and
literacy proficiency in 2004 in the Portland model
remained much the same, the path between
employment in 2000 and literacy proficiency in 2004
strengthened to twice its original size. Secondly, in
London there was also a strong effect of time spent
in employment in 2000 on ICT use in 2004 not seen
in Portland. Finally, in the London model there was
no significant relationship between ICT use in 2000
and amount of employment in 2004.