Page 12 - IJPS-3-2

Household structure and child education in Cambodia

analytical strategy.
2.1 Study Population

The target population of the baseline survey consisted of residents from six new Population-Registration Areas (PRA), in
addition to all residents from the original MIPopLab site.
All seven PRA are located along the Mekong River, which flows through Eastern Cambodia and the capital city,
Phnom Penh. To be enumerated as a resident, a seasonal migrant needs to have spent less than three months away from
the household in the preceding six months and a circular migrant, such as a child attending school, needs to have spent
less than 4 nights a week away. The project was only designed to provide representative data for the population of the
contiguous districts along the Mekong River, where 20% of the rural households in the country resided at the time of
the 1998 General Population Census (National Institute of Statistics, 1999). However, comparisons between nationally-
representative and MIPopLab or MIPRAoC data have repeatedly shown little differences in marital, reproductive, or
household composition levels and trends (Heuveline and Hong, 2016).
The combined population—at initial registration in the six new PRA and after demographic update in the original
MIPopLab site—was close to 60,000. Due to the recent establishment of several garment factories, the population of one
of the PRA, located at the outskirts of the Phnom Penh, was much larger at the time of the baseline survey (Round 1,
2008) than in the sampling frame from the 1998 General Population Census. The response rate was 96.3% in this PRA
and 99.1% elsewhere (averaging 98.4%). The seemingly high response rate is in fact typical of social surveys conducted
in the country (e.g., National Institute of Statistics, 2015).
Overall, the MIPRAoC is representative of the Mekong River Valley households in Cambodia and follows the national
trends. Unlike existing nationally representative surveys, the MIPRAoC is designed to be a longitudinal data collection
project, where the ongoing data collection will enable researchers to examine the long-term impact of household
compositions. This paper examines the initial census of the seven PRA, which contained 14,989 children under the age of
18 across 7,205 households.
2.2 Analytical Strategy

As we are concerned with the academic achievements of children, the analytic sample has been restricted to school
age children (ages 6 to 17 years). Because the sample might thus include several siblings from the same household, we
use Hierarchical Linear Models (HLM) with each child’s characteristics as level-1 independent variables and (possibly
shared) household characteristics as level-2 independent variables. Among these variables are the parental characteristics
described below. Because parental characteristics are only available when the parent co-resides with her child, models
with parental characteristics are estimated on two sub-samples: a sub-sample of children co-residing with their mother (with
or without their father) and a sub-sample of children co-residing with their father (with or without their mother).

2.3 Measurements
At the time of initial registration into the MIPopLab/MIPRAoC HDSS (benchmark census or subsequent in-migration),
each household head provides, for each resident household member, their name (later replaced by a unique identifier),
gender, birth date, relationship to the head, parental information (on survival, current residence, or timing of death),
marital status, literacy, education, and occupation.

2.3.1 Outcomes
For dependent variables, we consider three measures of academic achievements reported by the head of the household or
another adult household resident: 1) literacy (self-reported ability to read and write in any language), 2) school attendance
(attending any formal educational institution), and 3) child’s grade level relative to the expected grade for age at the time
of the baseline survey.
2.3.2 Individual-Level Covariates

Level-1 independent variables include age groups, gender, and parental survival. We use the age groups 6 to 8, 9 to 11,
12 to 14, and 15 to 17 years, corresponding to the first and second half of primary school, lower secondary, and upper
secondary school respectively. The proportion of MIPRAoC children (17 years and under) who are of school age (6 years
and over) is 68%, and among those a majority (87%) indeed attended school, but only 73% were literate (see Table 1).
2.3.3 Household-Level Covariates

6 International Journal of Population Studies | 2017, Volume 3, Issue 2