The Theoretical Analysis of the Allocation of Family Control Rights B Oc17 2014 | Page 5
J. Cai, T. Loo
In the absence of an existing scale for beliefs and attitudes towards education, measurement scales and extant
literature on education, place branding, marketing and country of origin were consulted. The items were then
confirmed in interviews with 8 experts involved either professionally or academically with the marketing of international tertiary education.
Belief about education. To measure the image of education, an 18-item, 5-point Likert-type scale was developed from extant literature (e.g. Srikatanyoo & Gnoth, 2002; Bourke, 2000), e.g. “the UK has high standards in
selecting students”. The scale was anchored at 1 = strongly disagree and 5 = strongly agree.
Attitude towards education. Attitude towards education was measured by a 3-item scale which aimed to
measure the cognitive, affective and conative components of attitude: “Please indicate your overall satisfaction
with the UK as provider of tertiary education”, “I would recommend the UK as a provider of tertiary education
to my colleagues or friends” and “I have an affinity with the UK as provider of tertiary education”. A 5-point,
Likert-type scale was used, with the first question ranging from 1 = very dissatisfied to 5 = highly satisfied and
the last two questions anchored as per the measurements of the image of the nation.
Purchase intention. A single-item measure, with a 5-point, Likert-type scale, ranging from 1 = highly unlikely
to 5 = highly likely, was used: “How likely are you to further your tertiary education in the UK in the next 5
years?”
5. Model Testing & Results
Structural equation modeling was used to develop and refine the conceptual model. In such circumstances,
Jöreskog (1993) pointed out that models obtained via modifications might have been derived to some extent by
“capitalising on chance” (i.e., data mining). Therefore, he recommended that the final model be cross-validated
on independent data. This was also to make sure that data idiosyncrasies were not misleading the research results. Furthermore, the extent to which the model fits the data with an independent data set provided some evidence of the generalisability of the model (Bullock et al., 1994). The total sample size of 864 was therefore randomly split up into a calibration sample (n = 433), consisting of half Chinese and half %