Social Statistics Integrated Use of Survey and Administrative Data at Statistics Finland

1
Social Statistics - Integrated Use of Survey and
Administrative Data at Statistics Finland

• IAOS Conference
• Shanghai 14-16 October 2008
• Veli-Matti Törmälehto
• Statistics Finland

2
Outline of the presentation

• The register-based statistics and sample surveys
  on households and persons
• Preconditions for combined use
• Exploitation of registers in the survey process
  key points
• Concluding remarks

3
Population and housing census entirely from
registers since 1990
4
The sample surveys and the register-based
population and social statistics

• Register-based annual subject matter statistics
• - population, employment, dwelling-units and
  housing conditions, families, education, income
• Sample surveys to meet the user needs that cannot
  be satisified with register-based statistics
• - complete or partial lack of register data
• - the needs of the European Statistical System
• - other user needs, timeliness

5
(No Transcript)
6
Preconditions for exploitation of registers in
survey-based official statistics

• A register system to be exploited
• Informed consent from the respondents
• Identifiers also for the survey units for record
  linkage
• Statistical units
• - household as a consumption unit in surveys
• - household as a dwelling unit in the registers
• - dwelling-unit not a sufficient proxy for
  household in surveys
• Reference periods
• - current moving (in continuous surveys such as
  the LFS and the HBS)
• - current fixed (in registers and in some
  surveys such as SILC)
• - usual calendar year, 12 months preceding the
  interview etc.

7
The survey process and exploitation of registers
8
The survey process and exploitation of registers

• At Statistics Finland, similar exploitation of
  registers in all surveys in frame creation,
  sampling, estimation, (quality control,
  dissemination)
• More variation in the data collection and
  processing phases
• - timeliness restricts use in the monthly surveys
• - Survey on Income and Living Conditions
  explicitly built on data fusion of register- and
  interview-based data
• - more degrees of freedom in surveys without
  EU-regulation

9
The design phase

• Most of the surveys are able to use data on basic
  personal demographics (e.g. year of birth,
  country of birth, sex, citizenship, de jure
  marital status), completed education and annual
  incomes from the registers as statistical
  variables
• These cover 10 of the 16 core variables of social
  surveys as defined by Eurostat
• Labour variables and variables on the
  relationships between dwelling-unit members
  mainly useful as auxiliary data
• - definitions ( e.g. ILO definition, part-time
  work, reference times), coverage, timeliness of
  the registers, punctuality of the surveys, two
  definitions of a household

10
The design phase lessons learned

• Electronic questionnaires are heavily routed and
  filtering questions asked (e.g. housing
  conditions conditioned by tenure status in SILC,
  HBS job characteristics by labour status in LFS,
  SILC etc.)
• --gt The replacement data from the registers best
  considered at the level of topics, not at the
  level of variables, and taking into account
  questionnaire design
• Maximal use of registers may lead to more
  fragmented questionnaires, more complicated
  design and programming

11
The design phase

• The mode of collection
• Shorter questionnaires, especially thanks to
  register-based incomes
• --gt CATI mode feasible
• --gt lower costs, lower unit non-response

12
Key points from the other survey phases

• Samples of persons are drawn instead of samples
  of addresses /households
• - a valid PIN is the necessary precondition for a
  respondent to be selected into the sample, and in
  household surveys for any member to be included
  in the sample
• Data from registers may be fed to the electronic
  questionnaires
• - to be used as such (personal information)
• - to be verified in the interview (location,
  household composition)
• - missing register data may be pre-filled, and
  questions asked only when register data is
  missing
• Use of registers --gt more checking, more editing
  (consistency editing), less imputations

13
Key points from the other survey phases

• Existence of dwelling-register and pre-filling to
  questionnaire greatly facilitates record
  linkage identifiers need to searched only for
  the members added to household roster in the
  interview
• Use of calibration estimators a very important
  tool
• - to improve accuracy of the estimates and
  coherence with other statistics
• - calibration of sampling weights to marginal
  distributions of exactly matched register
  variables
• - basic demographic distributions always used
• - register-based job seeker status in the LFS,
  income in EU-SILC etc.

14
Key points from the other survey phases

• Quality control of surveys and registers
• - errors of measurement in surveys and registers
• - errors of estimation in surveys
• - monitoring register quality with surveys
• Anonymized micro-data a key output of surveys
• --gt use of data for analytical as well as
  descriptive purposes
• --gt internal consistency in multi-dimensional
  surveys, relationships between phenomena
  important

15
Summary

• The combined use of registers and interviews
• decreases direct data collection costs
• decreases total survey error by reducing errors
  of measurement and errors of estimation
• increases capacity to meet old and new user
  needs
• - may increase processing costs and production
  time
• - may lead to more fragmented questionnaires

16
Summary

• Efficient integration of surveys and registers
  requires very high quality from the statistical
  register variables
• - definitions, the level of detail, coverage
• - timeliness, stability
• - reactivity to changes in definitions (usually
  defined at European level)
• - international comparability surveys only
  relevant if they are comparable within the
  European Statistical System
• - flexibility with regard to data sources may
  minimize intra-country total survey error but
  introduce bias to across-countries comparisons

17
(No Transcript)
18
(No Transcript)
19
At-risk-of poverty rates, EU-SILC 2006 (60 of
median)
20
At-risk-of poverty rates, EU-SILC 2006The
register countries
21
Thank you for your attention!