Page 38 - JICE Volume 6 Issue 2 FULL FINAL

Chang Da Wan anD BeneDiCt Weerasena
Table 1. Descriptive Statistics and Definition of Variables

Variable Definition Mean Std Dev Min Max
Urban School Dummy Variable for Urban School 0.70 0.50 0 1
Hours spent in Continuous variable for hours spent at 4.53 5.90 0 30
internal tuition internal tuition
Father’s Categorical variable for Father’s level of 3.39 1.07 1 6
educational level education
Academic Categorical variable for no. of A’s in Lower 3.44 2.75 0 9
Excellence Secondary Assessment
Chinese Dummy Variable for ethnic Chinese 0.35 0.43 0 1
students

Indian & Others Dummy Variable for students of Indian 0.08 0.23 0 1
and other ethnicities (excluding Malay,
Chinese and East Malaysian Bumiputera)

East Malaysian Dummy Variable for students of 0.11 0.41 0 1
Bumiputera ethnicities from East Malaysia
Spending on Continuous variable for spending on 181.43 163.17 5 1100
external tuition external tuition
Hours spent in Continuous variable for hours spent at 6.00 4.00 1 30
external tuition external tuition

Table 1 presents the descriptive statistics and definition of variables. Out of 343 respondents,
70 % were in urban schools and the remaining in rural schools. The respondents also comprised of
46 % Malay, 35 % Chinese, 8% Indian and Other, and 11% Bumiputera from East Malaysia. In terms
of the respondents’ father’s level of education, 3%were primary school leavers, 13 % were lower
secondary school, 47 % completed upper secondary school, 21 % with a diploma, 11 % with a bachelor
degree, and 5% with a postgraduate degree. In terms of the previous academic achievement, 13%
did not score any A in their Lower Secondary Assessment and 1.5 % scored nine As.
The main variable examined in this paper is the amount of spending on supplementary tutoring
outside of school. The average spending for supplementary tutoring is RM 181.43 per month with
a standard deviation of RM163.17. The minimum value is RM5.00 per month and the maximum
value is RM1100.00. Consistent with most expenditure data, the spending amount is not normally
distributed. This was confirmed using the Kolmogorov-Smirnov Normality Test, which indicates that
the spending variable was not normally distributed. Hence, as a way to normalise the data, the
spending variable was transposed with logarithm into ‘log spending’.
In addition, the amount of hours spent attending external tuition is also used as a dependent
variable to examine the extent of participation. The average number of hours a week spent is 6 hours
with a standard deviation of 4. The minimum value is one hour and the maximum value is 30 hours.
Similar to spending, the number of hours spent on tuition is not normally distributed and therefore
normalised by transposing the time variable with logarithm into ‘log hours outside’.
The selection of the independent variables was guided by the review of literature on
determinants of shadow education as well as availability in this dataset. Primary analysis using a
stepwise additive regression was used to determine the variables for the full specification multiple
regression model. The independent variables identified were: (i) urban-rural school to represent
the geographical differences; (ii) ethnicity, given that Malaysia is a multi-ethic country and ethnic
differences have been a major determinants on income and spending; (iii) hours spent on internal

96 Journal of International and Comparative Education, 2017, Volume 6, Issue 2