Improving Internal Migration Estimates: Update

1
Improving Internal Migration Estimates Update

• Esther R. Miller, Hyo Park, and Barbara van der
  Vate
• Population Division
• Census Bureau
• Presentation to Spring FSCPE Meeting
• Philadelphia, PA March 28, 2005

2
Overview of Presentation

• Enhancements for July 1, 2004 Estimates
• Migration Update System for Estimates (MUSE)
• Planned Enhancements for July 1, 2005 Estimates
• Planned Enhancements for July 1, 2006 Estimates
• Planned Enhancements for post July 1, 2006
  Estimates

3
Enhancements for July 1, 2004 Estimates

• Increased consistency and efficiency by
  consolidating two IRS 1040 sub-processing systems
  
• State and County Totals
• State and County Characteristics

4
Production Enhancement

• Used identical selection criteria to prepare base
  for internal migration files
• - Removed all foreign returns
• - In-migrants will equal Out-migrants at the
  national level
• Removed Y2 only 1040NR (non-resident) returns
• Removed Y1-Y2 matched returns that did not match
  to the Person Characteristics File (PCF)
• - Represents 0.4 percent of all matched returns

5
Production Enhancement Led to a Modification of
Migration Review Procedures

• Pre-production enhancement
• Reviewed county level data and adjusted county
  level migration rates
• Adjustments only affected state and county level
  total migration rates
• Post-production enhancement
• Reviewed county level data and adjusted Geocodes
  at the individual level record
• Adjustments affect migration rates for state and
  county totals AND state and county
  characteristics

6
Migration Update System for Estimates (MUSE)

• Joint project between Population Division (POP)
  and Planning Research and Evaluation Division
  (PRED)
• Move from return-based processing to person-based
  processing approach

7
Main Objectives

• Improve migration rates
• Improve accuracy of demographic characteristics
  of dependents
• Improve the estimates
• Increase coverage
• Reduce bias in the migration rates by
• Increasing the set of administrative records we
  currently work with
• Incorporating models

8
MUSE (cont)

• Long Term Goals
• Complete Project for July 1, 2008 Estimates
• Measure and document incremental impact of adding
  Administrative Records
• Improve Geocoding
• Share results with FSCPE Methodology Research
  sub-committee

9
MUSE (cont.)

• Collaborate with PRED to incorporate full set of
  administrative records
• Electronic Filing File (ELF)
• Medicare
• Other Administrative Record Files

10
MUSE (cont.)

• Collaborate with PRED to
• Model Characteristics for PCF non-matches
• Model for potential bias for non-filers
• Incorporate ACS into the system

11
Planned Enhancements for July 1, 2005 Estimates

• Utilize Geography Division to produce the
  ZIP/CRSS file
• Coding Zip Codes to County
• Research outlier migration data prior to
  production of the migration rates

12
ZIP/CRSS Procedures

• Currently the ZIP/CRSS file is based on the
  postal services Delivery Sequence File code to
  ZIP2-to-county codes
• Geography Division will create the ZIP/CRSS file
• Reflects quarterly updates from the postal
  services Delivery Sequence File
• Cross reference file relates most up-to-date
  ZIP4-to- county codes
• More accurate and consistent with TIGER and MAF
• Consistent with GUSSIE base

13
Research Outlier Migration Data Prior to
Production

• Pre-2005 enhancement
• Receive a 1 sample file representing data
  through the end of June from IRS to build
  ZIP/CRRS coding guide
• Post-2005 Enhancement
• Receive all data through the end of June for
  early research and review

14
Advantages to Planned Enhancement

• Build the Zip/CRSS file earlier
• Ability to conduct thorough review prior to
  production
• Identify possible Geocoding problems in time to
  be resolved prior to producing the IRS Migration
  Rates
• Easier to incorporate Geocoding changes at
  individual record level
• Speed up review process during production

15
Planned Enhancements for July 1, 2006 Estimates

• Incorporate Person Based Approach
• Integrate the Electronic File (ELF) from IRS

16
Assign CharacteristicsReturn Based Methodology

• Age
• Filer and spouse exemptions are assigned filers
  age from PCF
• Child exemptions are assigned to a 19 and under
  age category
• Parent exemptions are assigned to a 65 age
  category
• Sex
• Filer is assigned filers sex from PCF
• Spouse exemption is assigned opposite sex of
  filer
• Child and parent exemptions are assigned sex by
  random number generator

17
Assign CharacteristicsReturn Based Methodology
(cont.)

• Race and Hispanic Origin (HO)
• Filer is assigned filers race and HO from PCF
• Spouse is assigned filers race and HO from PCF
• Child and parent exemptions are assigned filers
  race and HO

18
Assign Characteristics Person Based
Methodology

• Age, sex, race, Hispanic Origin
• Filer is assigned filers characteristics from
  PCF
• Spouse is assigned spouses characteristics from
  PCF
• Child exemptions are assigned childs
  characteristics from PCF
• Parent exemptions are assigned parents
  characteristics from PCF

19
Assign Migration Status Compare FIPS code Y1-Y2

• Return Based
• Filer, spouse, child, and parent exemptions are
  all assigned filers Y1 and Y2 State and County
  FIPS codes
• Person Based
• Filer, spouse, child, and parent exemptions are
  all assigned individual Y1 and Y2 State and
  County FIPS codes

20
Advantage to Person-Based

• Match rates across Y1-Y2 IRS files will increase
• Preliminary evaluation shows that the match rates
  across Y1-Y2 increased
• Improve migration data by picking up additional
  filers
• - Students who move out of the house
• - Divorce or Separated

21
Advantage to Person-Based (cont.)

• Accurate ages for all exemptions
• Migration status is correct
• Accurate sex, race, and HO for all exemptions

22
Preliminary Results Comparing Return-based and
Person-based
23
(Match rates are for person-based returns only)
24
Age-Specific Match Rates for 1999-2000 Migration
Year Using Person-based Records
25
Match Rates by Race and Hispanic origin for
1999-2000 Migration Year using Person-based
Records
26
Migration Rates() for 1999-2000 Migration Year
Return-based vs. Person-based
27
Inter-county Migration Rates () by Age Group for
1999-2000 Migration Year Return-based vs.
Person-based
28
Inter-county Migration Rates () by Race and
Hispanic origin for 1999-2000 Migration Year
Return-based vs. Person-based
29
Scatterplots for the Person-based and
Return-based In-Migration Rates() for 1999-2000
Migration Year
County Level
State Level
30
(No Transcript)
31
Scatterplots for the Person-based and
Return-based Out-Migration Rates() for 1999-2000
Migration Year
County Level
State Level
32
(No Transcript)
33
Scatterplots for the Person-based and
Return-based Net-Migration Rates() for 1999-2000
Migration Year
State Level
County Level
34
(No Transcript)
35
Some Findings from Analysis of Outlier Counties
for Net-Migration

• 60 of the outliers had the population less then
  10,000
• 71 of the outliers less than pop person-based method increased the out-rates.
• 20 of the outliers was associated with major
  colleges and universities.
• 97 of the outliers associated w/ college, the
  person-based method increased the in-rates.

36
Future Research Questions

• Is there a relationship between the outflows from
  small counties and the inflows into the counties
  w/ major colleges?
• Is the relationship stronger for the intra-state
  migration than for the inter-state migration?
• Who are the individuals that move out of the
  counties w/ small population?
• Who are the individuals that move into the
  counties with colleges?

37
Enhancements for post July 1, 2006 Estimates

• Integrate Full Set of Administrative Records
• Integrate Full Model to Reduce Bias due to
  differences in coverage rates