By the Numbers: Part Two

• Welcome

• By The Numbers – Part II

• Questions and Answers

• On The Horizon

Overview

2

3

• Alicia VanOrman, PhD

– Population Reference Bureau

• Today we will explore technical questions related to data disaggregation in support of evidence-informed advocacy work

• For more info: [email protected]

Our Technical Assistance Partner

POPULATION REFERENCE BUREAU | www.prb.org

By the Numbers Part II

Disaggregating Data by

Race and Ethnicity

Alicia VanOrman, Research Associate

October 6, 2016

© 2016 Population Reference Bureau. All rights reserved. www.prb.org

Today’s presentation

Methodological challenges

Classification schemes

Dealing with small numbers

Accessing data


Percent minority* by age, 2015-2050

3840

53

4850

61

3841

55

2224

40

0

10

20

30

40

50

60

70

2015 2020 2050

Total

Ages 0-17

Ages 18-39

Ages 65+

*Includes those who are not non-Hispanic white (alone).

Source: Population Reference Bureau analysis of 2014 U.S. Census Bureau Population Projections.



3840

53

4850

61

3841

55

2224

40

0

10

20

30

40

50

60

70

2015 2020 2050

Total

Ages 0-17

Ages 18-39

Ages 65+





3840

53

4850

61

3841

55

2224

40

0

10

20

30

40

50

60

70

2015 2020 2050

Total

Ages 0-17

Ages 18-39

Ages 65+





3840

53

4850

61

3841

55

2224

40

0

10

20

30

40

50

60

70

2015 2020 2050

Total

Ages 0-17

Ages 18-39

Ages 65+




What are you measuring?

Racial categories are defined by

social, economic, political

institutions

Categories change over time,

depends on social context

Ethnicity is separate from race

Refers to ancestry or heritage

Federal statistical system only

collects data on Hispanic/Non-

Hispanic ethnicity


Guidelines for collecting data

OMB guidelines adopted in 1997

Two question format

Option to select multiple racial designations

Department of education fully implemented in 2010

Minimum

American Indian or Alaska Native

Asian

Black or African American

Native Hawaiian or Other Pacific Islander

White

Hispanic or Latino (asked in a separate question)


Guidelines for collecting data

Variation across data sources

Mutually exclusive groups (e.g., non-Hispanic Black or African

American)

Combining categories (e.g., National Survey of Children’s Health)

Coverage

Did all local entities report data in the same way (e.g., police

jurisdictions)?


Working with small numbers

Maintaining confidentiality

Smaller populations = easier to identify an individual

Providing reliable estimates

Smaller populations = more sampling variability

Multiple options for assessing and dealing with these

challenges


Are you working with confidential data?

Data provider may have prevented disclosure already

E.g., ACS tables, birth and death data through CDC Wonder

Check for guidelines from the data source

E.g., numerator less than 3

Assess the numerator and denominator

Compared to data source guidelines

General guidelines

Numerator/event data: 0 or >2

Denominator/population: >300, 100-300 use caution, <100 use extreme caution

Avoiding disclosure


Assessing reliability

Reliability measure Reliability estimate

Standard error• Estimated deviation from actual population value .6

Margins of error• Maximum amount of difference between sample

estimate and actual population value

±1 percentage point

(90% confidence level)

Confidence interval• Range of values that describe the uncertainty around an

estimate

19.9%-21.8%

(90% confidence interval)

Coefficient of variation• Relative amount of sampling error

• 𝐶𝑉 =𝑆𝐸

𝑋∗ 100

3%

20.9% of children in Nevada are living in poverty.

Source: Population Reference Bureau analysis of 2015 1-year ACS data, Table B17001


No hard and fast rules; depends on the application

Confidence intervals for percents

Less than 10 percentage points

Coefficients of variation

Smaller CVs (<15%) indicate greater reliability; larger CVs

(>30%) indicate unreliable data

Don’t use when proportion is close to zero

What is an acceptable amount of error?


Options for presenting data

Reliable estimates

Present the estimate with the reliability measure

Borderline reliable estimates

Present the estimate with note to use caution

Aggregate to increase sample size

Unreliable estimates and/or disclosure risk

Aggregate to increase sample size

Suppress estimates


What the federal government does

National Center for Health Statistics Less than 10 events: suppress count and rate

Less than 20 events: suppress rate

National Health Interview Survey, other health surveys Relative Standard Error >30%

U.S. Census Bureau Population and geography thresholds (e.g., 65,000 for 1-year ACS estimates)

Significance tests (e.g., >half of estimates are not significantly different from 0)

Relative standard errors (median CV>61%)


What KIDS COUNT does

Guidelines provided by data source

Aggregate data across years

ACS: 3 and 4 year olds not enrolled in school

Data suppression

ACS confidence intervals greater than 10 percentage points

Less than 3 cases


Aggregating data

Combine multiple years of data

5-year ACS data

Custom multiyear estimations: ACS Data User Group Webinar

Expand the geographic area

Combine smaller levels of geography into a larger group (e.g.,

combine counties to create county groups)

https://acsdatacommunity.prb.org/p/custom-multiyear-estimates-webinar


Poverty rates for Latino children,

Selected years and geographies

0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0

United States 2014

United States 2010-2014

Maryland 2014

Maryland 2010-2014

Baltimore 2014

Baltimore 2010-2014

Source: U.S. Census Bureau, American Community Survey.


Aggregating by collapsing categories

Asian, Pacific Islander, American Indian/Alaska Native,

and two or more races

17.5% ± 1.8

Asian and Pacific Islander and

American Indian/Alaska Native

12.0% ± 2.2

Asian and Pacific Islander

12.0% ± 2.4

Asian

11.8% ± 2.4

Native Hawaiian or other Pacific

Islander

18.8% ± 9.6

American Indian/Alaska

Native

12.0% ± 6.1

Two or more races

20.5% ± 2.5


35.6% ± 1.3

Non-Hispanic White

14.5% ± .7

Hispanic

28.2% ± 1.1

Source: Population Reference Bureau analysis of 2015 1-year ACS data, Table C17001 series

Percent of children in poverty by race and ethnicity, Florida, 2015


Aggregating data

Technical considerations

Microdata is more flexible

Use aggregate numerators and aggregate denominators for rates

𝑃𝑦1 + 𝑃𝑦2 + 𝑃𝑦3

3≠

𝑁𝑦1 + 𝑁𝑦2 + 𝑁𝑦3𝐷𝑦1 + 𝐷𝑦2 + 𝐷𝑦3

Standard errors/margins of error ACS Accuracy of Data Documentation

ACS Data User Group Webinar

https://www2.census.gov/programs-surveys/acs/tech_docs/accuracy/ACS_Accuracy_of_Data_2015.pdf

https://acsdatacommunity.prb.org/p/custom-multiyear-estimates-webinar


Aggregating data

Conceptual considerations

Tradeoffs between data that is less current, has less geographic

detail or less sub-group specificity

Best option will vary based on goals and data availability

Some data is better than no data


Accessing data

ACS collects detailed race

and ethnicity data

Alone: a single race

Alone or in combination: a

single race and those who

designated multiple races

White alone: does not

consider the question

about Hispanic origin

White, non-Hispanic or

Latino: White only and not

Hispanic


Accessing data by race and ethnicity: American Community Survey

Basic counts for total population by detailed racial and

ethnic groups (alone)

(there are also tables by alone and in combination)

1-year ACS Table 5-year ACS Table

Asian B02015 B02006

Native Hawaiian and other Pacific

Islander

B02016 B02007

American Indian or Alaska Native B02014 B02005

Hispanic or Latino B03001 B03001


Social and economic characteristics by detailed race

and ethnicity

Very little data available

Selected Population Profiles: 1-year data only

Advanced Search -> Race and Ethnic Groups -> Detailed groups

-> All available races



American FactFinder

detailed and subject tables

by race and ethnicity

Identified with a letter after

the table number B17001B: Poverty Status, Black or

African American Alone

Categories available: White


American Indian and Alaska Native

Asian

Native Hawaiian and other Pacific Islander

Other races

Two or more races

White, not Hispanic or Latino

Hispanic or Latino



Table identifier on

American FactFinder

Category on KIDS COUNT Data Center

C American Indian

D + E Asian and Pacific Islander

B Black or African American

I Hispanic or Latino

H Non-Hispanic White

G Two or more races

Add tables together

using aggregation

methods previously

discussed



List of all tables disaggregated by race and ethnicity Advanced Search -> Race and Ethnic Groups -> Basic Groups

-> Select group of interest




Accessing Data by Race and Ethnicity:

American Community Survey


Counts for selected tribal groupings B02014 (AIAN alone): 1-year data, US and very few states

B02017 (AIAN alone or in combination): 1-year data, few states

B02005 (AIAN alone): 5-year data, all states

Social and economic characteristics by tribe Selected Population Profiles, 1-year data only, US and few states

By state: 2006-2010 Data Profiles 2, 3, and 4

American Indian Area/Alaska Native Area/Native

Hawaiian Home lands 5-year data: 693 areas available

1-year data: 12 areas available



Public Use Micro Sample Data (PUMS)

Flexible, custom tabulations

Geographic data is limited

Need statistical software (e.g., SAS)

Calculating margins of error can be technical See PUMS Accuracy of Data document

Download data files on AFF or Census website


http://www2.census.gov/programs-surveys/acs/tech_docs/pums/accuracy/2014AccuracyPUMS.pdf



PUMS via IPUMS

Custom tabulations, 1-year and 5-

year ACS data Recode and create variables

Some effort to make race and ethnicity

categories match those in KIDS COUNT

Produces standard errors

Download microdata Need statistical software (e.g., SAS)

Multiple years of data at once, data are

harmonized

https://usa.ipums.org/usa/


CDC Wonder

Births

Deaths

Data by state and by

county Publically available birth data

limited to counties larger than

100,000

Suppressions rules have

already been applied

Accessing data by race and ethnicity: Vital Statistics

https://wonder.cdc.gov/


Civil Rights Data Collection

Data pre-tabulated by state

Data available at school district

and school levels Use the data analysis tools to compare

multiple schools or districts

Can request a data file to use with

statistical software to analysis all

districts or schools at once

Accessing data by race and ethnicity: Education-related data

http://ocrdata.ed.gov/DataAnalysisTools

http://ocrdata.ed.gov/DataAnalysisTools


National Center for Education

Statistics and Ed Data

Express

State-level data

Accessing data by race and ethnicity: Education-related data

http://nces.ed.gov/

http://eddataexpress.ed.gov/


Looking ahead

Combining two-question format into one question

Middle Eastern or North African (MENA) category

2020 Census & 2019 ACS 2015 National Content Test

Office of Management and Budget Seeking review and comments on possible changes

Federalregister.gov for more information


For more information

Alicia VanOrman

[email protected]

202-939-5474

mailto:[email protected]

41

Questions?

42

• Availability of recorded content for future reference

• The discussion continues at the KIDS COUNT Institute and through future webinars co-hosted with our state partners.

On The Horizon

Developing solutions to build a brighter future for children, families and communities

www.aecf.org

Data & Analytics

By the Numbers: Part Two