Getting Data for Business Statistics: A Response Model for Business Surveys

1
Getting Data for Business StatisticsA
Response Model for Business Surveys

• Ger Snijkers
• Statistics Netherlands
• Utrecht University

2
Getting Data for Business Statistics
Statistical picture of a country

• How do we get the data we needfor business
  statistics?
• Yesterday, today, tomorrow

• Data
• In time
• Complete
• Correct

Survey Parameters in and out of control
Respondent Parameters
NSI
3
Getting Data for Business Statistics

• Over the years
• The day before yesterday ICES-I 1993
• Yesterday ICES-II 2000
• Today ICES-III 2007
• Tomorrow
• International Conference on Establishment
  Surveys

4
Getting Data for Business Statistics The day
before yesterday

• ICES-I (1993)
• 1. Surveying various branches of
  industryagriculture, energy, health care,
  trade, finance, education, manufacturing
  industry
• 2. Quality of business frames sampling
• 3. Data analysis Estimation
• Data collection methodologydata quality,
  registers, non-response, Q-design
• ? Stove-pipe approach
• ? The one-size-fits-all survey design

5

• Internal business factors
• Policy
• Data
• Resources
• Market position

• External business factors
• Econ. climate
• Regulatory requirements
• Political climate

• The survey
• Topic
• Population and sample
• Sponsor / Survey organisation
• Resources
• Planning
• Authority/confidentiality

• Informant
• Mandate
• Data knowledge
• Job priority

The survey design
One-size-fits-all
Motivation
Contact strategy
Modes of data collection
Paper
Letters Mandatory

• Response
• In time
• Complete
• Correct

Questionnaire
Data WE want
Respondentburden De facto Perception
Decision to participate
Answering behaviour
Survey designs not coordinated Stove-pipe
approach
NSI
A business
NSI
6
Getting Data for Business Statistics Yesterday

• ICES-II (2000)
• Issues in government surveys
• Data collection modes non-response
• The response process
• Use of register data
• Sampling
• Editing and Data Quality
• Data analysis, estimation and dissemination

7
Getting Data for Business Statistics Today

• ICES-III (2007)
• Survey data collection methodology
  questionnaire design pre-testing survey
  participation non-response reduction, response
  burden, bias mixed-mode designs e-data
  collection understanding the response process
  in buss
• Using administrative data
• 3. Business frames Sampling
• 4. Weighting, Outlier detection, Estimation
  Data analysis

8
Getting Data for Business Statistics Today

• International Workshop on Business Data
• Collection Methodology
• London, 2006 ONS
• Ottawa, 2008 Statistics Canada
• Organising Committee Ger Snijkers (Stats
  Netherlands) Gustav Haraldsen (Stats Norway)
  Jacqui Jones (ONS) Diane Willimack (US
  Census Bureau)
• Practices, developments, research issues

9
Getting Data for Business Statistics Today

• International Workshop on Business Data
• Collection Methodology
• Primary data collection
• questionnaire design pre-testing survey
  participation non-response reduction, response
  burden bias, contact strategies mixed-mode
  designs e-data collection understanding the
  response process in buss
• 2. Secondary data collection
• use of registers
• 3. Multi-source designs
• combining survey and administrative data

10
Getting Data for Business Statistics Over the
years

• General picture
• 1993
• 2007

Stove-pipe approach One-size-fits-all
Survey organisation is central

• 2000 Transition

Systematisation and standardisation of
methods Mixed-mode, multi-source Respondent
is central tailoring
11
Getting Data for Business StatisticsThe data
collection design today

• Challenge
• Good statistics
• relevant
• more integrated information
• faster
• Less money
• Less compliance costs
• providing data only once to government
• Consequences for the data collection

12
Getting Data for Business Statistics
Consequences for data collection

• 1. Using more and more register data
• Definitions of variables
• Definitions of units
• Timeliness of register
• Quality of register data
• Combining register and survey data
• ? Managing integrated sets of statistics using
  various data sources Not Managing
  stove-pipes (a survey and related statistics)

13
Getting Data for Business Statistics
Consequences for data collection

• 2. Additional data collection
• When register data are not available
• ? Not in time
• ? Additional information needed
• - variables
• - target population
• ? Quality is not good

14
Getting Data for Business Statistics Consequences
for data collection

• 3. Sample design
• Controlling for overlap across surveys
• Controlling for rotation over time
• To avoid this

15
Getting Data for Business Statistics
Consequences for data collection

• 4. Survey design
• Mode of data collection
• ? EDI XBRL
• ? Mixed-mode designs
• Internet, paper, telephone (CATI)
• Questionnaire design
• ? Tailored to information buss have in their
  records
• ? Controlling for overlap across questionnaires
• ? Pre-tested for Q-A process and usability
• Contact strategy
• ? When data are available (not when we need
  them)
• ? Motivating and stimulating respondents
• - Compliance principles

? To avoid these reactions
16
Getting Data for Business Statistics
Consequences for data collection

• Reactions by businesses
• What is the use of this survey?It is
  pointless!
• There is no connection with my business
  activities.
• It only costs money and time!The costs
  outweigh the added value.There is no added
  value.
• Pick someone else. Although you say it is a
  sample, I am in it every time.

17
Getting Data for Business Statistics The data
collection design today

• More complex than yesterday
• More data sources
• Dependent on providers of registers
• Mixed-mode designs
• Coordinated development over modes Tailored
  to mode
• Tailoring to subgroups
• Tailored to target populations - opening the
  black box the response process
• Coordinated over surveys (ask only once)
• Tailored multi-source/mixed-mode design

18

• Internal business factors
• Policy
• Data
• Resources
• Market position

• External business factors
• Econ. climate
• Regulatory requirements
• Political climate

• The survey
• Topic
• Population and sample
• Sponsor / Survey organisation
• Resources
• Planning
• Authority/confidentiality

• Informant
• Mandate
• Data knowledge
• Job priority

Image
The survey design
Motivation
Contact strategy
Modes of data collection

• Response
• In time
• Complete
• Correct

Questionnaire
Respondentburden De facto Perception
Decision to participate
Answering behaviour

• More than one survey
• More than once
• In other ways
• ? Registers
• ? EDI

NSI
NSI
A business
19
Getting Data for Business Statistics The data
collection design today

• Tailored multi-source/mixed-mode design
• Small businesses
• register data ( survey data)
• Middle-sized businesses
• register data survey data
• Large businesses
• consistent data collection for
• - all businesses
• - all variables
• It is our job to make statistics out of these data

20
Getting Data for Business Statistics Tomorrow

• Improving the
• tailored multi-source/mixed-mode design
• Advanced statistical modelling
• Estimations based on multiple sources
  and mixed-mode surveys Managing integrated
  sets of statistics (not stove-pipes)
• Opening the businesses
• Insight in the response process Tailored
  surveys to the internal businesss processes
• Opening the survey process
• Improved relationships with businesses
• - What surveys, when, feedback, involving
  buss Systematisation and standardisation of
  survey designs - Survey parameters in control

21

• Internal business factors
• Policy
• Data
• Resources
• Market position

• External business factors
• Econ. climate
• Regulatory requirements
• Political climate

• The survey
• Topic
• Population and sample
• Sponsor / Survey organisation
• Resources
• Planning
• Authority/confidentiality

Statistical picture of a country

• Informant
• Mandate
• Data knowledge
• Job priority

Image
The survey design
Motivation
Contact strategy
Modes of data collection

• Response
• In time
• Complete
• Correct

Questionnaire
Respondentburden De facto Perception
Decision to participate
Answering behaviour

• More than one survey
• More than once
• In other ways
• ? Registers
• ? EDI

Registerdata
NSI
NSI
A business
22
References

• American Statistical Association, Proceedings of
  ICES-I (1993), ICES-II (2000) and ICES-III
  (2007). Alexandria (Virginia).
• Groves, R.M., and M.P. Couper (1998), Nonresponse
  in Household Interview Surveys. Wiley, New York.
• Hedlin, D., T. Dale, G. Haraldsen, and J. Jones
  (2005), Developing Methods for Assessing
  Perceived Response Burden. Statistics Sweden,
  Stockholm, Statistics Norway, Oslo, and UK Office
  for National Statistics, London.
• Snijkers, G. (2007), Between Chaos and Creation.
  Inaugural lecture Utrecht University (in Dutch).
  Statistics Netherlands, Heerlen.
• Snijkers, G. (2007), Collecting Data for Business
  Statistics Yesterday, Today, Tomorrow.
  Presentation at 56th Meeting of the ISI, 22-29
  August 2007, Lisbon, Portugal.
• Snijkers, G. (2007), Collecting Data for Business
  Statistics A Response Model. Proceedings of the
  56th Meeting of the ISI (CD-rom), 22-29 August
  2007, Lisbon, Portugal.
• Willimack, D.K., E. Nichols, and S. Sudman
  (2002), Understanding Unit and Item Nonresponse
  in Business Surveys. In Groves, R., D. Dillman,
  J. Eltinge, and R. Little (eds.), Survey
  Nonresponse, pp. 213-227. Wiley, New York.