SELEKT in the International Trade in Services Survey

SELEKT in the International
Trade in Services
International Trade in Services
Statistics on international trade in services, wages
and transfers are based on a quarterly sample
survey involving about 5 000 enterprises and
organizations.
The variable of interest is value
and is reported for three levels
Expenditure and income
1. for each company
The variable of interest is value
and is reported for three levels
Expenditure and income
1. for each company
2. for each company and service
The variable of interest is value
and is reported for three levels
Expenditure and income
1. for each company
2. for each company and service
3. for each company and service and country
The variable of interest is value
and is reported for three levels
Expenditure and income
1. for each company
2. for each company and service
3. for each company and service and country
The second level is the primary input in the statistical output.
The data structure
Corporate ID
Year
Quarter
Service Code Direction
Country
Value
165500000000 2009
4
090
1
A1
500 000
165500000000 2009
4
090
5
A1
30 000
165500000000 2009
4
410
1
A1
200 000
165500000000 2009
4
410
1
GB
100 000
165500000000 2009
4
410
1
PO
100 000
165500000000 2009
4
411
1
A1
300 000
165500000000 2009
4
411
1
NO
200 000
165500000000 2009
4
411
1
US
100 000
165500000000 2009
4
432
5
A1
30 000
165500000000 2009
4
432
5
BE
30 000
165500000000 2010
1
090
5
A1
40 000
165500000000 2010
1
432
5
A1
40 000
165500000000 2010
1
432
5
NL
40 000
Service Code
410 Computer services
411 Information services
432 Accounting
Direction
1 Expenditure
5 Income
Input, throughput, output (1)
Throughput
Input
Sampled unit (k)
Observed Background variables
unit (l)
Direction Service aggregate
1
Expenditure
2
Expenditure
14E
3
4
5
Income
Expenditure
Expenditure
14E
14E
14E
Service code
090
411
Country
A1
A1
411
411
411
A1
SL
PO
Value
yk2
-Coding
-Editing
-Imputation
-Estimation
Output
Sum of value by Direction
Direction
Expenditure Income
Total
Use
-Decision making
-Information
Sum of value by Service aggregate and Direction
Service
Direction
aggregate Expenditure Income
14C
14D
14E
14F
14G
Sum of value by Service code and Direction
Service
Direction
Code
Expenditure Income
410
411
Input, throughput, output (2)
Throughput
Input
Sampled unit (k)
Observed Background variables
unit (l)
Direction Service aggregate
1
Expenditure
2
Expenditure
14E
3
4
5
Income
Expenditure
Expenditure
14E
14E
14E
Service code
090
411
Country
A1
A1
411
411
411
A1
SL
PO
Value
yk1
yk2
-Coding
-Editing
-Imputation
-Estimation
Output
Pseudo
total
090
Direction
Expenditure
Use
-Decision making
-Information
Income
Sum of value by Direction
Direction
Expenditure Income
Total
Sum of value by Service aggregate and Direction
Service
Direction
aggregate Expenditure Income
14C
14D
14E
14F
14G
Sum of value by Service code and Direction
Service
Direction
Code
Expenditure Income
410
411
Potential impact
The potential impacts of a variable value on the estimates
are calculated as the weighted difference between the
expected value and the observed value relative to the
estimated standard errors of the estimated sums.
Each type of output table has a coefficient of importance
multiplied to the potential impact which creates a local score
for selective editing.
These coefficients are set to achieve a balanced editing of
data on the three levels simultaneously.
This minimizes the number of re-contacts with each
company.
Expected values
Selekt offers two possibilities to calculate the expected values:
• Time series data for company, service and direction have
been used when available, even if it consists of only one
observation.
• Cross sectional data for similar observations based on edit
groups. The cross sectional data are a poor source of
information to use when calculating the expected values.
The edit groups are not homogenous.
How to find good expected values for event-based data?