Transcript Counting and Locating

2A. Counting and Locating
Constructing a sampling frame
Wednesday 1st October
Basic Demographics
• Fifteen year olds are a very small proportion
of total population
• Average number of households to be visited
to find one 15 year old = 10 (except Senegal)
• Average density of 15 year olds per square
kilometre = <2 (except 3 in Guatemala)
• DHS estimates of same order of magnitude,
except for Cambodia, Tanzania and Zambia
Low Birth Registration and Age Heaping
• In LA countries birth registration are above
90% (judged to be complete registration)
• Among other countries, low in Tanzania (16%)
and Zambia (14%)
• Demographers have shown that both
caregivers and respondents tend to give ages
to nearest 5 or 10 years to interviewer
• The most pronounced deviations are at age 10
for Cambodia, Tanzania and Zambia and age
20 for Senegal. The effect at age 15 is small
but should be considered at analysis stage
Current Definitions of Out of School
• CREATE defined 6 zones of exclusion from
Primary and Lower Secondary
• UIS and UNICEF defined 5 Dimensions of
Exclusion based on age rather than grade
• Neither approach addresses problem of
defining current status vis-à-vis school
• Both CREATE and UIS/UNICEF are based on
rights of children to a basic education and so
are ‘school based’ in sense that school defines
whether or not child is registered is enrolled
and currently attending sufficiently to be
counting as not having dropped out
Survey Definitions
• Current questions in DHS and MICS surveys
leave judgement as to whether 15 year old is
currently attending to the respondent
themselves
• Obviously subjective and contextual
questionnaire needs to robe more deeply
within constraints of questionnaire length
• Also surveys responses are not based on
actual current attendance but on whether or
not the 15 year old has attended school at
some time in the year
Alternative Education Programme
• Need to decide how to treat Accelerated
Learning Programmes including NGO-provided
NFE programmes and some apprenticeships
• How should religious education be treated
• How can contextual questionnaire be detailed
and comprehensive so as to capture the whole
range of NFE and religious education within
the time constraints
In school: variation across countries
Age 15
Total
Urban
Rural
Brazil
92
93
92
Cambodia
66
79
64
Ecuador
90
93
81
Guatemala
63
76
53
Senegal
52
67
38
Tanzania
61
71
58
Zambia
88
91
85
Source: UNESCO, EFA GMR
7
In school by grade:
variation across countries
By grade
Below 7
Brazil
33
Cambodia
24
Ecuador
3
Guatemala
40
Senegal
25
Tanzania
66
Zambia
59
Source: UNESCO, EFA GMR
7
19
24
4
21
20
22
14
8
35
28
8
30
24
5
17
9
12
18
21
9
21
6
8
10
1.4
5.7
63.3
0.1
8.2
0.8
1.7
Above 10
0.0
0.3
0.7
0.0
2.2
0.0
0.7
8
Out of school
variation across countries
Never been to
Out of school:
school
Brazil
Cambodia
Ecuador
Guatemala
Senegal
Tanzania
Zambia
Source: UNESCO, EFA GMR
6
7
9
16
70
12
19
Incomplete
primary
31
47
8
39
26
17
61
Complete
primary or
higher
63
47
83
45
5
71
20
9
Groups omitted or systematically
under-represented in survey data
• Sampling Frames based on sometimes very
out-of-date censuses
Household surveys exclude, by design:
• the homeless
• those in institutions:
– hospitals, prisons, refugee camps
• the mobile:
– Nomads, pastoralists, travellers.
Groups Omitted or Systematically
Uncounted in Survey Data (B)
• In practice, they tend to exclude those in fragile,
disjointed households, those in urban slums and
those in insecure areas
– This will also be a problem for in-school assessment
• Those six categories constitute a pretty good
ostensive definition of the poorest of the poor
• There are two other groups of adults and children
who may be ‘hidden’ in households:
Those who are illegally resident
Those who suffer discrimination e.g. disabled
Approach to Counting
• To complement the approach of a standard
household survey, we need to search actively
for marginalised or vulnerable groups likely to
include OOS 15 year olds
• Given unknown errors in age-specific
population & unreliability of enrolment data,
we cannot count OOS numbers by subtraction
of enrolment figures form population figures
• Most marginalised will be found among
(isolated) rural poor and informal settlements
Using household surveys to count and
locate OOS 15 year olds
• Estimates can be made of percentages likely to
be missing from sampling frames of standard
household surveys in rural and urban areas
• This can be combined with assumptions about
how additional numbers would be divided
between out-of-school groups
Difficulties in counting OOS 15 year olds
• Several data sources, e.g.:
• International administrative routine data with
issues of coverage, accuracy, timeliness
• Standardised Household Surveys, with
problems of coverage of marginalised groups
• Used standardised household surveys because
they are comparable
Questions for discussion
•
•
•
•
•
What is our cohort? 15.3-16.2 year olds?
What is an out of school 15 year old?
Is our picture correct?
Where are out-of-school 15 year olds?
How do we construct the sample frame?
• What is an appropriate ambition for PISA-D?