|
|
Methodological information on the
Methodology of the GPS 2002 survey
The survey was conducted in April-May 2002 (interviews were carried
out between 4th April and 18th May) in all 15 EU Member States plus Switzerland
and the US, using computer-aided telephone interviews. The survey was
co-ordinated and executed by INRA Deutschland GmbH, Mölln. The
population for this study is all persons aged 15 and over living in
private households in the respective countries and speaking the respective
national language(s). 11,832 interviews were successfully completed.
The average interview length per country varied between 10 (Greece)
and 20 minutes (Sweden).
Sampling: Target households were selected at random
in all countries, either by random dialling techniques such as permutation
of final digits
or by drawing a random sample from official sources. Mostly a geographical
stratification was implemented beforehand. For the selection of the
target person common random keys were applied in all countries except
for the
UK where quota was used. In two cases (Spain, the US), screening had
to be directed towards male respondents towards the very end of the
field in order to gain gender representativeness.
There were three adjustments necessary in order to
provide reliable data:
- Transformation from household sample to person
sample. As only one person per household is interviewed, the described
sample procedure provides a household sample, i.e. each household of
the base population has the same likelihood of being in the sample but
not each person. With the weighting stage of the transformation the
equal likelihood of households is replaced mathematically by the equal
likelihood of the individuals. To this end, each data set is multiplied
by the amount of people in the household aged 15 or over. This number
is subsequently divided by the average household size in order to obtain
the actual case number.
- Adjustment of unweighted sample structure to the
official statistic. Because random samples are not evenly distributed
across all population strata, the distribution of unweighted samples
regularly and systematically deviate from the population distribution
from official statistics. Through the mathematical weighting the sample
distribution was adjusted to the official statistics. The national weighting
factor, which results from the iterative weighting, was included in
the data material.
- Adjustment of weighted sample structure to the
EU-15 Member States population. This weighting factor was necessary
to calculate total figures according to the whole population of the
European Union Member States. Furthermore it is useful to compare the
EU with the US. Population sizes of each Member State are weighted to
reduce the distortion based on the sample sizes in each country.
Methodology of the GPS-NAS 2003 survey
The survey was conducted in January 2003 (interviews
were carried out between 1st January and 31st January) in the 10 Newly
Associated
States Bulgaria, Czech Republic, Estonia, Hungary, Lithuania, Latvia,
Poland,
Romania, Slovenia and Slovakia, using personal aided personal interviews
(PAPI). The survey was co-ordinated and executed by NFO AISA Czech
Republic, Prague. The population for this study is all persons aged
15 and over living in private households in the respective countries
and speaking the respective national language(s). 10,379 interviews
were successfully completed. The average interview length per country
varied between 20 (Romania) and 40 minutes (Lithuania).
Sampling: Target households were selected at random
in all countries, either by multistage stratified random-route sampling
or by drawing
a random sample from official sources. Mostly a geographical stratification
was implemented beforehand. For the selection of the target person
common
random keys were applied in all countries, i.e. the next birthday
method and the Kish method, except for Bulgaria where quota was used.
There were three adjustments necessary in order
to provide reliable data:
- Transformation from household sample to
person sample in Poland and Slovenia. As only one person per household
is interviewed, the described sample procedure provides a household
sample, i.e. each household of the base population has the same likelihood
of being in the sample but not each person. With the weighting stage
of the transformation the equal likelihood of households is replaced
mathematically by the equal likelihood of the individuals. To this end,
each data set is multiplied by the amount of people in the household
aged 15 or over. This number is subsequently divided by the average
household size in order to obtain the actual case number.
- Adjustment of unweighted sample structure to the
official statistic. Because random samples are not evenly distributed
across all population strata, the distribution of unweighted samples
regularly and systematically deviate from the population distribution
from official statistics. Through the mathematical weighting the sample
distribution was adjusted to the official statistics. The national weighting
factor, which results from the iterative weighting, was included in
the data material.
- Adjustment of weighted sample structure to the
NAS-10 countries population. This weighting factor was necessary to
calculate total figures according to the whole population of the Newly
Associated States. Furthermore it is useful to compare the NAS with
the EU. Population sizes of each of the ten states are weighted to reduce
the distortion based on the sample sizes in each country.
Methodology of the DMS 2002 survey
The survey was conducted in March-May 2002 (interviews
were carried out between 21st March and 15th May) in seven EU Member
States using computer-aided telephone interviews. The survey was co-ordinated
and executed by INRA Deutschland GmbH, Mölln. The population for
this study is defined as all establishments belonging to four aggregated
industry sectors in the seven Member States Germany, Finland, France,
Greece, the UK, Italy and Spain. The interview was conducted with IT
responsible persons in companies across all sectors of the economy. 3,139
interviews were successfully completed. The average interview length
per country varied between 14 (France) and 18 minutes (Italy).
Sampling: The sample was set up according to given
industry and size class quota. Accordingly a stratified random sample
was drawn from the universe, allowing for the relevant industries within
four aggregated sectors (manufacturing, construction, primary sector;
distribution, catering, transport & communication; financial & business
services; public administration, education, health, other personal and
social services). Drawing the sample was organised locally by the national
executing institutes.
Weighting: For the SIBIS DMS a sample stratified
by sector/ size cells was used which ensured that in each sector, establishments
from all size classes (1 to 9, 10 to 49, 50 to 199, 200 to 499 and 500+)
were sampled. In order to be able to raise figures to national level,
some form of weighting is required which adequately reflects the structure
and distribution of establishments (or related variables) in the universe
of the respective country (and, by implication, EU-15).
- Original weight: Within each country, the
interviews were split according to a quota plan which guaranteed
that the sample is not dominated by micro and small companies. The
quotas roughly reflect the distribution of employment over sector
and establishment size bands in the EU, and derive from research
into establishment sampling frames undertaken for previous studies
by Infratest and GfK in the course of ECaTT. They represent best
estimates, but do not take account of country differences. Weighting
was used in cases where the quotas could not be reached exactly in
line with this quota plan (mostly due to the limited absolute number
of establishments in the two biggest size classes). Note that because
of the use of a single quota plan for all countries, country differences
in the distribution of employment over establishment size bands which
occur in reality are not reflected in the data. This is due the lack
of available data on the distribution of employment across establishments
size bands in almost all EU Member States, and constitutes a considerable
problem. This weight is therefore not used for presenting SIBIS results.
- Weighting by employment: The data available on
the distribution of employment over establishment size bands is very
limited for most EU Member States. SIBIS used data from a variety of
sources, including BT database (United Kingdom), ISTAT Industry and
Services Intermediate Census (Italy), National Statistical Service
of Greece (Greece), SIREN (France), Tilstokeskus Official Statistics
(Finland), Heins + Partner B-Pool (Germany) and Schober Business Pool
(Spain) and adjusted using data from the DG Enterprise/ Eurostat SME
Database (latest available, 1997), to estimate the establishment/ employment
structure for each country in the sample. Using this weight, the weighted
sample for each country therefore reflects employee distribution between
the five establishment size bands within that country. This means that
a data reference of, for example, ”20% of all establishments
in country A” should be understood to mean ”establishments
accounting for 20% of all employees in country A”.
- Weighting by employment for EU-7 averages: Additionally
another weighting factor was created to calculate average figures for
all countries in the sample (which together represent roughly 82% percentage
of total EU employment). Each country is represented in this weight
according to its share in the total employment of the 7 EU countries
in which the survey was conducted.
|