Transcript Presentation title - Danmarks Statistik

Implementing cawi
into the data collection
process
Kees van Berkel
Mariëtte Vosmer
Jerusalem, 21-24 July 2013
Why
•
•
•
•
Meet customers’ demands;
Meet respondents’ expectations;
Budget cuts. It’s all about the money;
Increase response rates, reach more groups.
Be flexible
Implementing cawi into data collection process
1
History
• 2005: start web data collection research;
• 2005 – 2007: experiments
 ICT-survey
 Crime survey
 Survey measuring Underground Economy;
• 2008 – now: surveys
 Dutch Housing and Living survey
 Crime survey
 National Mobility survey
 Health Survey
 Labour Force Survey;
Implementing cawi into data collection process
2
Sampling frame
Municipal basic registration of population data.
This registration contains information of all
people who are registered at a municipality.
The municipalities own the data, but the Ministry
of the Interior and Kingdom Relations is
responsible for the regulation.
Statistics Netherlands is allowed to use it,
under strict conditions.
Implementing cawi into data collection process
3
Sampling design
First stage: stratified systematic pps sample
of municipalities.
Second stage: random sample of addresses.
Stratification: the country is split into 40 corop areas.
The sample is self-weighting:
Every address has the same inclusion probability.
Implementing cawi into data collection process
4
Sampling design
In each stratum h :
the number of addresses to be drawn nh
is proportional to the number of addresses Nh ,
𝑛ℎ ~ 𝑁ℎ
𝑔ℎ = 𝑛ℎ 𝑚
m : desired number of addresses to be selected
in each selected municipality  cluster size.
gh : number of municipalities to be selected
in stratum h.
Implementing cawi into data collection process
5
Sampling design
Small m  many municipalities
 large travelling expences
Large m  few municipalities
 cluster effect
 large standard errors
Face to face: cluster size m = 12
Mixed-mode: cluster size m = 1
Implementing cawi into data collection process
6
Fixed amounts
Capi
Cawi
respons
•
Cawi
non-respons
Cati
Increase of variance
1.16
1.14
1.12
1.1
1.08
1.06
1.04
1.02
1
0.98
0
20
40
60
80
100
120
Implementing cawi into data collection process
7
Weighting procedure
Mode included as weighting term
Increase of inclusion weights due to partially approaching
cawi non respondents by telephone or face to face
Implementing cawi into data collection process
8
Approach strategy
Approach letter
+
Reminder 1
+
Reminder 2
+
Telephone
Face to face
Implementing cawi into data collection process
9
LFS adaptations
1)
> 2 or > 3 in the age of 15 years or more to capi?
% unknown sample unit / telephone number closed
18.0
2)
16.0
14.0
Quality of
phone
numbers
12.0
%
10.0
Mobile
Landline
8.0
6.0
4.0
2.0
0.0
Sept
Okt
Nov
Month
Dec
Implementing cawi into data collection process
Jan
10
Distribution of responses in time
Crime Survey
160
140
120
R
e 100
s
p 80
o
n 60
s
40
20
0
1
2
3
4
5
6
7
8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Day
Implementing cawi into data collection process
11
Response rates
Implementing cawi into data collection process
12
Response results LFS (1)
Implementing cawi into data collection process
13
Response results LFS (3)
Implementing cawi into data collection process
14