2010 Census Coverage Measurement Discussion by Kirk Wolter May 22, - - PowerPoint PPT Presentation

Discussion by Kirk Wolter

May 22, 2012

2010 Census Coverage Measurement

• 1. Coverage measurement results
• 2. Methods used to produce the CCM results
• 3. Census costs
• 4. Implications for 2020

2

Outline of My Remarks

Net Undercount Rate: National Level, 2010

• CCM = -0.01% (or overcount of 36,000 people)
• DA = -0.09% (or overcount of 270,538 people)
• Seems amazingly good

3

• 1. Coverage Measurement Results

Census Year Net Undercount (in %)* 1990 1.61 2000 ‐0.49 2010 ‐0.01

*Results based on the PES method

4

How Does 2010 Compare to Recent Censuses?

• Race
• Ethnicity
• Sex
• Age
• Type of enumeration area
• Housing tenure
• Mover status
• Urban/rural
• State

5

Differential Undercount

How Does 2010 Compare to Recent Censuses?

6

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 1940 1950 1960 1970 1980 1990 2000 2010 Difference in Percent Net Undercount Census Year

Differential Undercount v. Census Year: National Level

DA Difference by Sex (Male ‐ Female) DA Difference by Race (Black‐Nonblack) PES Difference by Race (NonHispanic Black ‐ NonHispanic White)

Differential Undercount by Age: Males in 1980, Demographic Analysis Estimates

7

‐1 1 2 3 4 5 0 to 4 5 to 9 10 to 14 15 to 19 20 to 24 25 to 29 30 to 34 35 to 39 40 to 44 45 to 49 50 to 54 55 to 59 60 to 64 65 to 69 70 to 74 >= 75 Net Undercount (in %) Age Groups

DA Estimated Percent Net Undercount by Age: Males

Differential Undercount by Age: Males in 1990, 2000, and 2010, PES Estimates

8

‐2 ‐1 1 2 3 4 5 0 to 4 5 to 9 10 to 17 18 to 29 30 to 49 50+ Net Undercount (in %) Age Groups

PES Estimated Percent Net Undercount by Age: Males

2010 2000 1990

• Top three states
• DC = 2.23%
• VT = 1.29%
• TX = 0.97%
• Bottom three states
• AK = -0.85%
• OK = -1.08%
• WV = -1.43%

Differential Undercount by State: 2010, CCM Estimates

9

CCM Estimate Census Estimate Correct Enumerations 284,668 Correct Enumerations 284,668 Omissions 15,999 EE+II 16,035 Total Population 300,667 Total Population 300,703

2010 was Accurate Overall because of Well Targeted Additions

10

Additions at the State Level

11

1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 EE+II (in %) Omissions (in %)

EE+II v. Omissions at the State Level: 2010

Additions at the County Level

12

0.0 5.0 10.0 15.0 20.0 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 EE+II (in %) Omissions (in %)

EE+II v. Omissions at the County Level: 2010

Additions at the County Level

13

0.0 5.0 10.0 15.0 20.0 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 EE+II (in %) Omissions (in %)

EE+II v. Omissions at the County Level: 2010

}Net Undercount

Duplicates are a Major Part of the Census Additions

14

8521 1520 5993

Duplicates Other EE II

But Do They Distort Relationships?

15

Key Variables CE Duplicates Other EE II Net Undercount Omissions Persons White 95.2 2.7 0.4 1.7 ‐0.5 4.3 Black 92.6 3.6 0.7 3.1 2.1 9.3 Owner 95.7 2.4 0.4 1.5 ‐0.6 3.7 Renter 92.5 3.7 0.7 3.0 1.1 8.5 Mailout/Mailback 94.8 2.7 0.5 2.0 0.0 5.2 Update/Leave 92.7 4.7 0.5 2.2 ‐1.4 6.1 Update/Enumerate 91.1 3.0 0.5 5.3 7.9 16.0 OK 92.6 5.0 0.7 1.4 ‐1.1 6.4 TX 94.0 2.7 0.6 2.6 1.0 6.9 Housing Units Owner 98.6 0.6 0.8 0.2 1.6

Renter 97.2 1.3 1.5 ‐0.3 2.5

Type of Enumeration Areas

16

Quality Declines with Time

17

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 2/25‐3/17 3/18‐3/24 3/25‐3/31 4/1 ‐ 4/7 4/8 ‐ 4/15 4/15 ‐ 4/30 5/1 ‐ 9/30 No Valid Return Not in Mail Return Universe

Components of Census by Mail Return Date

CE Duplicates Other EE II

Quality Declines with Time - continued

18

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% April May June July and August Unknown Month In VDC and in NRFU In VDC but not NRFU Not in NRFU or VDC but in NRFU Reinterview or Residual Not in any NRFU Universe

Components of Census by NRFU Domain and Month of Enumeration

CE Duplicates Other EE II

• Analysis by mover status
• Ethnographic studies
• Analysis of omissions
• Within household
• Whole household

Wish List

19

Analysis of Omissions

20

1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Omisions of Persons (in %) Omissions of HUs (in %)

Omissions of Persons v. Omissions of HUs: State Level, 2010

• Basic Method
• Problems?
• Not enough information for matching
• Innovations
• Assessment of undercount by stage of operations
• Persons and HUs
• Logistic regression
• Correction for correlation bias
• 2. Methods Used to Produce the CCM Results

21

• Great work!
• 2. Methods - continued

22

• 3. Census Costs

23

$0 $20 $40 $60 $80 $100 $120 $140 1950 1960 1970 1980 1990 2000 2010 2020 Dollars Census Years

Costs in 2009 Real Dollars per HU v. Census Year

Approximate Distribution of Total Costs in 2010

24

$2,266 $436 $2,129 $148 $364 $2,986 $4,147 Communications and technology Printing and postage Office space and staff Fingerprinting National processing center ops Headquarters staff Field ops

Approximate Distribution of Field Ops Costs in 2010

25

$386 $2,744 $341 $108 $118 $450

Address canvassing NRFU VDC Update/enumerate Update/leave Other field ops

• Simulate cost and quality of alternative census scenarios
• Address canvassing
• NRFU
• VDC
• CFU
• Update/leave
• Update/enumerate
• Ascertain key reasons why people and HUs are being

missed

• 4. Implications for 2020

26

• Duplicates, other EE, and II
• Ascertain key reasons for EE+II
• Ascertain how well they balance omissions within blocks, tracts,

and counties

• Are they a cost-effective, worthwhile, and acceptable

compensation for the omissions

• Internet census
• Any new problems of EE, II, or omissions?
• Any new problems of coverage measurement?
• Administrative records census
• Any new problems of EE, II, or omissions?
• Any new problems of coverage measurement?

2020 - continued

27

Thank You!