The Courts Service: Building a New Organisaiton

1
Conference of European Statisticians Seminar
Session 2
Theme Efficiency of Statistical
Offices Discussant Donal Garvey
(Ireland)Geneva, 11 June 2007
2

• Four Invited Papers
• Finland, Iceland, Israel, Netherlands
• Two Contributed Papers
• Hungary, Israel
• Many common themes - I will deal with some of the
  unique themes first and then focus most of the
  discussion on some common threads

3

• Resource-based planning in Hungary

• Understanding the resources really needed to
  carry out public sector tasks was a challenge for
  the HCSO
• Resources were dissipated across dispersed
  regional offices
• HCSO reorganised from 19 county offices to six
  regional offices. Other countries perhaps went
  even further.
• Q Is six still even too many for coherent staff
  mobility and development?
• 2 years developing a framework of program
  elements (550 elements for 2007) to support
  resource-based prioritisation.
• Q Is the framework changing the way middle and
  senior managers are approaching their work and
  resource utilisation?

4

• Tracking productivity - Finland

• An overt total quality approach to
  inputs/systems/outputs to meet stakeholders
  needs. Inputs and outputs must be clearly
  defined and measured
• Total productivity 27 (2000-2006)
• Q Can you incorporate in the calculations the
  investment by other authorities in their
  registers which contribute towards better quality
  statistical outputs?

5

• Interviewing costs - Israel

• Hard to find populations different
  characteristics than those who are easy-to-find
  but repeated attempts to make face-to-face
  contact are expensive.
• Limiting the number of attempts will help control
  costs but will result in lower response rates.
  Ireland four attempts for the first wave of the
  QNHS and two attempts for subsequent waves -
  unless the Interviewer is passing the door.
• Israeli Finding Survey estimates robust for main
  survey variables even when a cap on the number of
  attempts could reduce response rates from 88 to
  77 and no statistically significant bias is
  introduced. This would save 14 of interviewers
  time.

6

• Managing/training Interviewers - Questions

• Q Israeli findings based on a post-hoc analysis.
  Could a cap affect interviewer behaviour even
  before that cap is reached?
• Q Will a cap encourage more no change answers
  accepted over the telephone?
• Q If interviewers are paid a bonus for success
  rates above a certain target level will capping
  the number of attempts influence their behaviour?
• Q Do any countries use a model to determine
  number of attempts to interview different
  subgroups?

7

• Functional centralisation

• The paper from Iceland mentioned the synergies
  and efficiencies to be gained from functional
  centralisation
• co-operation rather than misguided competition
• less duplication of resources
• statistics the main rather than a secondary role
• gains in coverage, efficiency and quality
• improved methods, processes and practices
• common web-based dissemination approach
• raised competence of new and existing staff
• Q Will this lower the analytical capability
  within other Ministries? and pose a risk to the
  status of statistics in the future?
• In Ireland we have a very centralised statistical
  system and this lack of analytical capability
  within Ministries is a challenge.

8

• Common themes - administrative data

• Strong focus to reduce administrative burden for
  20 years or more unique identification systems
  for Enterprises, Buildings, Persons (Finland)
• Large efficiency gains if the national culture
  accepts that the normal way is to maximise the
  value of administrative records and public
  information systems (Iceland) - must be
  legislated for properly
• Clause in 2004 Act requires use of all available
  administrative data before surveying
  (Netherlands)
• Increasing use of identity numbers a rich and
  cheap source of administrative data a gold
  mine of data of a census type (Israel) - use can
  depend on culture/legislation in a country

9

• Downsides of administrative data - Israel

• May present a dangerous path for a Democracy
• Q Why? The use of administrative data for
  statistical purposes seems to create no extra
  dangers?
• May be of little statistical value
• tend to be biased (owner agencys priorities)
• definitional changes without prior notice
• discontinuities make change evaluation difficult
• NSI only a residual user Agency sole decision
  maker
• Agency interested in the present NSI in history
  also
• Q How are these issues dealt with in Finland,
  Iceland ?

10

• Downsides - Continued

• Security risk Israel suggests a higher security
  risk for administrative data but Iceland says
  that managing confidentiality is easier.
• Q Who is right?
• Late availability, only when agency has completed
  its task (Israel)
• Comment Not necessarily so. Even in Ireland we
  have online access to certain important
  administrative data.
• Do not substitute for everything a modern society
  needs - must co-exist with survey data (Israel)
  increasing need for survey data also (Iceland)
  but over 95 of Statistics Finlands basic data
  is derived from administration registers
  collected by other authorities (Finland)
• Q Is there any risk that a system built almost
  entirely on administrative data could become less
  than fully responsive to the changing information
  needs of society?

11

• Reputational Challenges

• Administrative data is of a census type so
  there is a general tendency to see them as the
  truth (Israel)
• poor analysis by inexperienced researchers of
  micro-data from administrative sources can cause
  problems
• different figures can unjustifiably reduce the
  trust in survey data and tarnish the reputation
  of NSIs.
• The Israeli paper gives an example of different
  estimates of the annual wage bill which allowed
  politicians to undermine politically challenging
  survey data. I will use a simpler Irish example.

12

• 1996 Irish study

13

• LFS Unemployment Registered Unemployment
  (Ireland, 1996)

• LFS unemployment down 15 1994-1996 Registered
  unemployment no change. Huge debate on the
  real level of unemployment with reputation of
  CSO LFS at risk.
• Full administrative file requested by CSO and a
  1 sample selected. Addresses of these
  registrants included in the LFS (only 4 or 5
  people in CSO knew - unknown to field staff) and
  analysed separately.
• Over 11 of the respondents indicated they had
  full-time jobs and 4 out of 5 of these said the
  job started more than one month previously (many
  at least a year previously!). About 3 out of 4
  indicated the job was a permanent job.
• About 10 said they were working part-time and
  nearly 3 out of 4 of these were Not
  underemployed

14

• 1996 Irish study - continued

• Conclusion many people who are not
  statistically classified as unemployed in the LFS
  are included in the Live Register also there
  are unemployed persons (according to LFS
  definitions) who are not on the Live Register
• Changed the debate in Ireland. Put pressure on
  the Registration Authority to eliminate fraud
• Prime Minister asked us to repeat the survey -
  CSO said No
• Data Protection Commissioner agreed that the CSO
  acted within its rights and within data
  protection guidelines
• Q Are there any quality issues around
  administrative registers in Finland, Iceland,
  Netherlands ? How do you guard against it?

15

• Some answers (Finland)

• Extensive, continuous co-operation with other
  authorities
• Nominated individuals link between statistics
  units and register changes
• Continuous quality monitoring - example LFS a
  quality check on the population register
• Q Does the population register track emigrants
  well? Any interesting insights from the LFS
  quality checks?

16

• Data matching/linking

• Improved record linkage methodology.
  Administrative data complimentary to, combined
  with survey data (Iceland, Israel) EU-SILC
  (Iceland)
• Ireland increasingly ask for personal public
  service number (PPSN) in surveys to reduce the
  numbers of questions and to increase quality by
  pulling administrative data. EU SILC an example
• Protocol on our website explaining what we do
  agreed with Data Protection Commissioner
  maintain a public log of data linking activity
• Q A tricky issue for statistics. Strong
  statistics legislation needed. What pitfalls
  should be avoided?

17

• Methodology

• Standardisation of corporate methods (also
  supports IT standardisation). Only validated
  statistical methods within corporate
  methodological framework may be used
  (Netherlands)
• Neglect of some distinct activities in the value
  chain has caused quality issues in the past.
  Statistical audit of key processes, rolling
  5-year basis (Netherlands)
• IT and methods reorganised to support practical
  implementation (Finland)
• Q How do methods people respond to
  implementation roles?
• Q How do business owners respond to
  standardisation in Netherlands?
• Comment In Ireland many different approaches
  to editing and imputation - we are imposing a
  narrower corporate framework

18

• Redesign of business processes

• Objective better quality and efficiency through
  processes which are coherent, flexible,
  consistent, reproducible with fewer ICT
  applications (Netherlands)
• uniform IT platform with software all from same
  family and central software deployment
  decision-making (Iceland)
• software like a bowl of spaghetti with poor
  documentation and unauthorised automation. Aim
  to re-use applications with lower development and
  maintenance costs (Netherlands)
• align technical development projects into
  mutually supporting entities more uniform
  framework for developing and maintaining data
  systems (Finland)

19

• Business processes - continued

• Business Architecture, 4 database design
  (Netherlands)
• re-use of data and metadata
• distinguish design from implementation
• design driven by output requirements within
  frameworks for methodology, data collection,
  finance, organisation
• no production without standardised metadata
• Imposed on every development project through
  standardised documentation
• Q If migration will take several more years
  will SN be supporting two entirely different
  approaches for several years? What problems are
  foreseen?

20

• Business processes - continued

• Program does not aim at building generic data
  processing solutions. In our view that is one
  bridge too far (Netherlands)
• Q Does anyone disagree that it is a bridge too
  far?
• Substantial productivity gains to be made in new
  data editing methods (Finland) - looking at
  editing/imputation software developed by
  Statistics Canada
• Q Any useful information for the meeting?
• Electronic collection of data from enterprises
  (Finland, Iceland)
• Comment Finland raised the possibility of more
  international co-operation (also with IT
  service providers) in the field of integrated
  data collection

21
Questions and Discussion