1. Michael Biddington, UN ESCAP Statistics Division, Biddington@un.org
Administrative Data
Michael Biddington, UN ESCAP Statistics Division,

2. Introduction Global context Australian history Lessons from Australia
Data sources
Data providers
Best practice

3. Global Context
Demand from government, businesses, people & international community for data
NSO’s need unit level data and economic surveys
Quality and quantity of admin data is growing
Thus SBRs are key
Admin data can be made better

4. Australian SBR Admin Data (History)
Australian Bureau of Statistics (ABS) First Integrated Register of Businesses ( )
Statistics Law – admin data
Australian Taxation Office (ATO)
One Form – Group Employer Registration Form
All were sent a Mail profiling form
Activity (ISIC) & locations
Turnover
Contact details
Used as survey frame for business surveys

5. Australian SBR Admin Data (History)
Thirty years of collaboration led to:
The New Taxation System (2000)
Australia introduced GST (or VAT)
Australian Business Register (ABR)
Separate agency (ABS & ATO staff)
Unique ID – all government
Registration form designed jointly (ABS & ATO)
All businesses, charities, schools, etc have to register
Reduced provider burden
Enabled data intergration

6. Admin Data in Australian SBR (Now)
Australian Taxation Office
VAT (or GST) Tax
Company Tax
Wages and Salary Reporting
R&D + Other tax concessions
Australian Business Register – Australian Business Number
Australian Securities and Investments Commission
Australian Prudential Regulation Authority
Australian Charities
Tourist Accommodation
Type of establishment and star rating
Listing is a private (non government) list.
99% of relevant data is from ATO and ABR. All use Unique ID from Australian Business Register. Very low cost to acquire and link other data.

7. Data Selection – Australia
Country context is essential
Issues to consider;
Coverage
Coherence
Bias
Cost
Quality
Purpose
Lets consider this given Australia
Coverage – Does it coverage a meaningful proportion of your economy?
Coherence – Can the data sets be combined on a conceptual level? Or will duplicates occur? Will is make your systems too complex?
Bias – Are certain businesses tax exempt? Fraud? (I have examples for Australia…)?
Cost – Administrative data is cheap – when compared to surveys and censuses. It is not free. Relationship management of providers, data cleaning, systems and data linking need to be considered. Consider these factors before adding / using data.
Quality – Consider how the data is collected / updated. Who is checking? Actual Tax Paid to office usually very accurate.
Purpose – What are you trying to do with the data. What issue is it addressing?

8. Key Lessons – Australia
What we learnt from 30 years of admin data
Coverage – all employers
Coherence – one primary source / form
Bias – employers / formal
Cost – One source / unique ID
Quality – Second form
Purpose – design the form
Coverage – One form for all of Australia. Legal Entity. No establishment data.
Coherence – one primary source / form
Bias – employers / formal. Many more Bias’s. But only including employers is worst. People working for themselves.
Cost – unique ID
Quality – We used to send a survey form to EVERY BUSINESS. Ignored other data.
Purpose – design the form

9. Selecting Data Sources
Types of Administrative Sources
Business Registration Systems
Social security registers
Labour and employment registers
Tax Records/VAT
Sectoral lists

10. Selecting Data Sources
Types of Other Sources
Utility Company Records
Private Sector data poviders
Directories
Big data

11. Selecting Data Sources
Data Quality Framework (ABS)
Relevance, Accuracy, Timeliness, Interpretability, Coherence, Accessibility
Don’t try to link them all
No one does
High cost

12. Communication with Data Providers
Empathise
You are a risk
You aren’t their job or purpose
Persuade or convince
Understand their process
Limit your communications / requests
No one has perfected this
Get an Memorandum of Understanding (MOU)
This is true if new or existing

13. Standard MOUs Conditions for the supply of administrative data
Agreed timetable for supply of data
How data will be supplied, its format and quality
Confidentiality and security of data and its use
Engagement protocols around changes to data supply or systems changes.

14. Combining Multiple Data Sources
Unique ID
Coherence (legal entity or establishment)
Unique ID matching can still be wrong
If no Unique ID
Name matching
Fuzzy matching
Survey forms
Need a human to check
Really consider if it is worth it.

15. Best Practice Statistical Law Data provided by MOU
Design data collection processes
Shared Unique ID’s across government
Achieve these and admin data becomes exponentially more useful

16. Q&A
Thank you!