Upload
clifford-wells
View
215
Download
0
Tags:
Embed Size (px)
Citation preview
1
Measuring Quality Issues Associated with Internal
Migration EstimatesJoanne Clements, Amir Islam, Ruth Fulton & Jane Naylor
Demographics Methods Centreand Centre for Demography
2
Outline
• Background
• Internal Migration Quality Issues
• Research methods
• Findings
• Issues arising
• Next Steps
3
Project
Improve understanding,
measurement and reporting of
the quality of population
estimates
4
Context
• Debate about amount of uncertainty in
population estimates
• Improving Migration and Population Statistics
(IMPS) Project – Quality strand
• ‘ONS should flag the level of reliability of
individual local authority population estimates’
(UK Statistics Authority)
• Leading new international research
5
Key Methodology Points
• Map out the procedures and data sources used to derive population estimates
• Identify associated quality issues
• Attempt to quantify uncertainty using statistical theory & empirical evidence instead of expert opinion
• Combine individual measures of uncertainty by simulating potential errors in the data
6
Key Methodology Points
• Map out the procedures and data sources used to derive population estimates
• Identify associated quality issues
• Attempt to quantify uncertainty using statistical theory & empirical evidence instead of expert opinion
• Combine individual measures of uncertainty by simulating potential errors in the data
7
Key Methodology Points
• Map out the procedures and data sources used to derive population estimates
• Identify associated quality issues
• Attempt to quantify uncertainty using statistical theory & empirical evidence instead of expert opinion
• Combine individual measures of uncertainty by simulating potential errors in the data
8
Key Methodology Points
• Map out the procedures and data sources used to derive population estimates
• Identify associated quality issues
• Attempt to quantify uncertainty using statistical theory & empirical evidence instead of expert opinion
• Combine individual measures of uncertainty by simulating potential errors in the data
9
Progress
• Initial work proved feasibility of simulation methodology
• Focus now on sources of error with greatest impact; internal and international migration
• Currently focussing on internal migration
10
Internal Migration Methodology
• Individual moves captured from GP re-registration data
• Annual (end July) download of patient registers
• Moves identified from changes with previous year’s download.
• Local authority moves constrained to information provided by NHS Central Register
11
Key Internal Migration Quality Issues
Source LA for out-flowsto NI and Scotland
Census and 2001 Patient
Registers
Constraining GP register data to
NHSCR data
Time
Lags
Double counting of School boarders
Not registered
at mid-year
13
Research Methods
• A review of relevant literature.
• Local authority level data analysis
• Review any internal quality assurance.
• Sensitivity analysis
14
Re-registration Time Lag Research
• Comparison of mid-2001 internal migration estimates with 2001 Census migration estimates
• Sex ratios• Propensity to migrate
• Comparison with other data sources• Investigating ‘bumps’ in population age
profiles that sustain over time
15
Birmingham Population Age Profile
16
Provisional Time Lag Findings:Sex Ratios
• Evidence of late-registration of young male migrants
• Geographic variation in sex ratio differences and therefore time lags
Source Migrant Sex Ratio
15-29 years
Census 2001 0.915
Internal Migration 00/01
0.765
17
Provisional Time Lag Findings:Propensity to Migrate
• GP List inflation invalidates analysis to compare Census and internal migration propensities
• Instead, comparing migrant counts for similar populations to identify possible time lags
• Census doesn’t always produce higher LA internal migration estimates
18
Provisional Time Lag Findings:Other Data Sources
• Limited other data sources with which to compare with – No major differences with comparator data sources
• Evidence from survey data of significant late registration (Median 4 months)
19
Provisional Time Lag Findings:Age Profiles
• Some LAs do have age profile bumps that sustain (particularly young adults ages)
• Patterns vary again geographically• Possibly due to:
Imbalance between in and out migrants in LAs with higher education institutions (Males especially)
Increases in International immigrants (young males again)
20
Provisional Time Lag Findings:Summary
• Evidence of Age-Sex Specific Time lags in re-registration.
• Evidence that these vary geographically.• Unclear how much year on year time lags
cancel each other out.• Next Step is to produce an potential error
distribution
21
School Boarder Research
• LAs with largest school boarder populations chosen to identify possible double counting
• Comparing age profile changes in school boarders with LA internal migration estimates
22
Provisional School Boarder Findings
• Similar patterns between school boarder arrivals and internal in-migration
• Therefore, strong evidence of double counting
• Difficult to estimate accurately due to data issues
• Limited impact, for most LAs, on all age internal migration estimates
23
Challenge: Deriving Error Distributions For Each Quality Issue
• Lack of suitable data
• Conflicting evidence
• Somewhat subjective choice of error bounds
- Bias towards larger errors?
- Sensitivity Testing
- Constraining
- Correlation
- User Feedback
24
Challenge: Interpretation of Findings
• In reality, there is uncertainty in these measures of uncertainty, as…
– Only as good as the error assumptions made for each issue
• Therefore exact findings are misleading
• Present approximate indicators
25
Reporting and Future Work
• Short update on progress – August 2009
• Detailed papers on internal migration findings
- November 2009
- 2010
• Potential further work:
- international migration
- quantifying impact of methodological
changes on quality of estimates