Andrey Veykher National Research University «Higher School оf Economics» in St-Petersburg, Russia The representative study of households based on the data of individual interviewing
Andrey Veykher National Research University
«Higher School оf Economics»in St-Petersburg, Russia
The representative study of households based on the data
of individual interviewing
Questions raised by clients of surveys
1) How many households with two children of school or preschool age are there in the city?
2) What is the percentage of single-livings person and couple households?
3) How are the households distributed by amount of total income?
4) What part of the households with total month income between $ 2,000 and $ 4,000 have real estate in the countryside and haven't car?
2ESRA - Lausanne - 2011- 07 - 22
Methodological problemIf the units of representative survey were individuals, how to recalculate the percentages, means, another
parameters of distributions of individuals into the same
parameters of distributions for units – households?
How many children are in your family?
Survey of households
Survey of individuals
the number of households with 2 or 3 children – possible
object of social assistance
the number of people who live in the
households with 2 or 3 children – data with
no practical value
4ESRA - Lausanne - 2011- 07 - 22
Differences of individuals and households distributions
single-living couple (2 adult)
in population (individuals) 20% 42%
in households
Data source: survey company “Crona Corsinto Ltd” (St-Petersburg, 2009)
5ESRA - Lausanne - 2011- 07 - 22
Differences of individuals and households distributions
single-living couple (2 adult)
in population 20% 42%
in
households
38% 40%
Data source: survey company “Crona Corsinto Ltd” (St-Petersburg, 2009)
6ESRA - Lausanne - 2011- 07 - 22
Algorithm as a part of complexComplex was created by survey company
“Crona Corsinto” (© 2001):• 4 steps stratified sample with quota
sampling “age + gender” at the end stage• Interview “face-to-face by houses
adjoining survey”• Algorithm of recalculating the data of
population into the data of households• Method of external validity of sampling and
implementing data collection ESRA - Lausanne - 2011- 07 - 22 7
Basic assumption
Households of M members of the target populations age are represented in the probability sample at M times more than households of single-living.
,%100
:
:
,..2,1 ,...2,1
,..2,1
n
j
n
iij
n
jij
j
Mn
Mn
X
where: • Xj – percentage of households with feature J;• I – number of adult household members, I = 1,2,3…;• nij – number of respondents in the survey, who have
feature J and belong to the household with I adult people
9ESRA - Lausanne - 2011- 07 - 22
Steps of calculation - start
Scale values in different
distributions
Adult person in household Sum
1 2 3 4100
No children (individuals) 12 18 12 8 50No children
(households)
One child (individuals) 2 14 18 4 38
One child (households)
Two children (individuals) 1 4 3 4 12Two children (households)
10ESRA - Lausanne - 2011- 07 - 22
1st steps of calculation – weight = 1
Scale values in different
distributions
Adult person in household
Sum1 2 3 4
No children (individuals) 12 18 12 8 50No children
(households) 12One child
(individuals) 2 14 18 4 38One child
(households) 2Two children (individuals) 1 4 3 4 12Two children (households) 1 11ESRA - Lausanne - 2011- 07 - 22
2nd steps – weight = 0.5
Scale values in different
distributions
Adult person in household
Sum1 2 3 4
No children (individuals) 12 18 12 8 50No children
(households) 12 9One child
(individuals) 2 14 18 4 38One child
(households) 2 7Two children (individuals) 1 4 3 4 12Two children (households) 1 2 12ESRA - Lausanne - 2011- 07 - 22
3rd steps – weight = 0.33
Scale values in different
distributions
Adult person in household
Sum1 2 3 4
No children (individuals) 12 18 12 8 50No children
(households) 12 9 4One child
(individuals) 2 14 18 4 38One child
(households) 2 7 6Two children (individuals) 1 4 3 4 12Two children (households) 1 2 1 13ESRA - Lausanne - 2011- 07 - 22
4th steps – weight = 0.25
Scale values in different
distributions
Adult person in household
Sum1 2 3 4
No children (individuals) 12 18 12 8 50No children
(households) 12 9 4 2One child
(individuals) 2 14 18 4 38One child
(households) 2 7 6 1Two children (individuals) 1 4 3 4 12Two children (households) 1 2 1 1 14ESRA - Lausanne - 2011- 07 - 22
Steps of calculation - finish
Scale values in different
distributions
Adult person in household Sum
1 2 3 4 48
No children (individuals) 12 18 12 8 50No children
(households) 12 9 4 2 27One child
(individuals) 2 14 18 4 38One child
(households) 2 7 6 1 16Two children (individuals) 1 4 3 4 12Two children (households) 1 2 1 1 5 15ESRA - Lausanne - 2011- 07 - 22
What is the household limits?
ESRA - Lausanne - 2011- 07 - 22 16
In St-Petersburg:- more than 10% citizens live in another place than they are registered as electors;- 40% persons older after 70 live separately of adult children, but they have common family budget;- more often we observe how families of high middle class have two or three flats, where live older children without economical independence.
Condition and consequence of algorithm implement
ESRA - Lausanne - 2011- 07 - 22 17
Condition – special question in survey questionnaires "How many people in your household who is over 18 years /or another limit of target population/?" Consequence – obtaining additional data for external validity for estimation of sample and data collection quality.
Result of real project
ESRA - Lausanne - 2011- 07 - 22 18
In Saint-Petersburg (2007) the deviation of calculated from survey data number of children - 752 thousands – from statistics – 742 thousands (error - 1,3%) allows us to say that proportions of households distributions by number of children correspond to reality.
Conclusions:
ESRA - Lausanne - 2011- 07 - 22 19
•There are possibilities to obtain many answers about household by more cheap survey of individuals.•Effective and attractive for clients control quality of sampling and data collection implementing is possible by estimation external data and the result of matrix sequential calculations on survey data about households.