tract-level geocoding analysis: identifying communities with low calfresh access
DESCRIPTION
Tract-Level Geocoding Analysis: Identifying Communities With Low CalFresh Access M. Akhtar Khan, PhD . Research Services Branch Chief Aynalem Adugna , Ph.D. Research Program Specialist Research Services Branch California Department of Social Services August 20, 2014. - PowerPoint PPT PresentationTRANSCRIPT
Tract-Level Geocoding Analysis: Identifying
Communities With Low CalFresh Access
M. Akhtar Khan, PhD.Research Services Branch
Chief
Aynalem Adugna, Ph.D.Research Program
Specialist
Research Services BranchCalifornia Department of Social Services
August 20, 2014
Presentation Outline SNAP-CalFresh Access: Big picture
overview.
SNAP-CalFresh Access : As it is measured.
Geo-coding: Looking below county level.
Once below the county level, ability to measure CalFresh access at neighborhood and community levels.
New proposed methodology for identifying “true hot spots” for targeted outreach efforts.
Highlight measurement issues uncovered.
Application of adjusted access measure.
* CalFresh Program Overview Legislative Analyst’s Office, March 2014
California*
Average monthly households 1,890,1
29
Average monthly individuals 4,124,3
73
Average household size 2.2Average monthly benefit per household $333 Average monthly benefit per person $153
Household Characteristics— Income*
Eligible to Receive CalFresh
Income Below Certain Thresholds. • Gross income - 130 percent of federal poverty guidelines
(also known as the federal poverty level, or FPL). Example: $2,008/month for a household of three (2012).
• Net income - income of less than 100 percent of the FPL after certain deductions are applied. Example: $1,545/month for a household of three (2012).
Additional Eligibility Criteria for College Students
Ineligible to Receive CalFresh Citizenship/Immigration Status – the
largest group Drug Felony Convictions SSI/SSP Recipients Ineligible Due to “Cash-
Out.”
Source: CalFresh Program Overview Legislative Analyst’s Office, March 2014
=
Food Distribution Program on Indian Reservations; SSI is Supplemental Security Income
Measuring Program Access
By this measure: California is almost at the bottom in the Union. 3.9 million eligible Californians are not receiving
CalFresh. Questions that come to mind…?• Why is participation low in California?• How counties within the State fare?
• Is it unemployment rate at local level• Is it poverty level at local level?• Rural-Urban divide, or something else
• Any issues with the way program access is measured?
The Program Access Index (PAI): USDA/FNS
6
The search for answers:What sub-county-level geographies should be used to identify areas of low participation?
Regression analyses show the low predictive value of unemployment rates and poverty levels of counties in estimating calFresh access.
6 8 10 12 14 16 18 20 22 24 260
0.20.40.60.8
1
f(x) = 0.0113534671966928 x + 0.361448967437047R² = 0.147718986873891
County % in Poverty vs. % Receiving CalFresh (2011)
% in Poverty, 2011 % R
ece
ivin
g C
alF
resh
, 2
01
1
Distribution by proportions of non-English speakers points to:
o Language as an important factoro The possible role of immigration status
A measure that takes into account ineligibility for CalFresh due to citizenship status is necessary
0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% -
0.20 0.40 0.60 0.80 1.00
f(x) = 1.56602477 x + 0.329909794R² = 0.171480819663686
County % Unemployed vs. % Receiving CalFresh (2011)
Unemployment Rate, 2011% R
eceiv
ing C
alF
resh,
2011
Preliminary - not for reproduction 7
So the Issue!
• Socio-economic indicators at the county level rarely show significant correlations with CalFresh access.
• Explanation to this phenomena may exist at below county level – neighborhoods and communities.
• The question – how do we get down to the neighborhood and community levels.
• Need local level data and tools…
Preliminary - not for reproduction 8
Geo-mapping Analytics Geocoding analytics allow us to
assemble a rich data set using a variety of resources.
Show differences in CalFresh access at below county levels – census tracts, zip codes, neighborhoods, etc. using geocoding analytics.
Highlight population subgroups having CalFresh access lower than expected based on poverty levels using geocoding analytics.
Preliminary - not for reproduction 9
Gecoding is a technique that can help us:
I. map recipient level data, and
II. assess CalFresh reach at community levels.
Based on the poverty estimates for an area, this spatial analytical tool will enable us to identify places:
where potential CalFresh eligibles reside,
where CalFresh reach is low, and more importantly,
where more effective and targeted outreach strategies could be directed.
Geocoding
Why geocode?
Image source: http://jags-webdesign.com/home-clipart-image-stick-family-
123 Main Street (geocoded)
CalFresh recipient
Poverty level of census tract
Neighborhood characteristics
Languages spoken
Demographic characteristics
Environmental characteristics
Service delivery
Neighborhood organizations
Social/environmental activism
Outreach activities
Geocoding helps us gain a holistic view of the environments surrounding each CalFresh recipient address.
Over two hundred tract-level data elements are linked to each dot.
Example:
Total tract population.
% Below poverty level.
% Non-native. Number of
Hispanics. Number of families
with children under 18.
Number of Female-headed households.
% Speaking languages other than English.
EBT access.
Preliminary - not for reproduction 11
Fresno County CalFresh Distributions (July, 2013)
12
L.A. County: Number of CalFresh Recipients in Tracts with Reported Zero Poverty Levels
13
Monterey County: CalFresh Recipient Addresses and Distance from EBT Locations
(Zip Codes 93905 and 93906)
286 EBT locations7168 / 39167 locations (18 percent) within 0.1 mile18624 / 39167 locations (47 percent) within 0.2 mile
14
Where:FDPRI: Food Distribution Program on Indian ReservationsSSI: Supplemental Security Income Source: http://www.fns.usda.gov/sites/default/files/PAI2011.pdf
PRI=
Program Reach Index (PRI) *
Program Access Index (PAI)
* Geography-based or population-based
Preliminary - not for reproduction 15
Advantages of Using PRI Measure CalFresh access below
county levels.
Measure differences in access among population subgroups.
Use results to devise targeted CalFresh outreach activities.
Help uncover the limitations of PAI as measurement methodology.
Preliminary - not for reproduction 16
Application Example: LA County on the Map
Potential Eligible – Below 130 percent Poverty
vs.
CalFresh Access – PRI
Preliminary - not for reproduction 17
MEDIUM
HIGH
LOW
Preliminary - not for reproduction 18
MEDIUM
HIGH
HIGH
Preliminary - not for reproduction 19
L.A. County Tracts: Percent Speaking Languages other than English by Position Above or Below Median (Median = 58.3)
Preliminary - not for reproduction 20
L.A. County Tracts: Percent Below Poverty by Position Above or Below Median
(Median = 11.2)
21
9th
8th
6th
4th
3rd
2nd
1st 88.2-80.9
80.8-73.8
73.7-66.3
62.2-58.6
99.4-88.3
50.4-40.8
40.7-31.6
23.3-
0.0*
10th
7th
5th* Deciles width/ interval ( percent)
58.5-50.5
31.5-23.4
Population subgroups: LA County language deciles
Preliminary - not for reproduction 22
23
D1 D2 D3 D4 D5 D6 D7 D8 D9 D10
5.1 5.57.4
10.612.8 13.0
16.0
18.8
22.4
27.0
L.A. County: Average Poverty Level Tract by Language (other than English) Deciles
Language deciles
Avera
ge p
overt
y
Application: County Level PRI, 2011High PRI Low PRI
Calaveras 0.90 Santa Barbara 0.38Del Norte 0.83 San Luis Obispo 0.37San Bernardino 0.77 Napa 0.37Fresno 0.75 Marin 0.35Tulare 0.73 Yolo 0.34Sacramento 0.71 Colusa 0.34Yuba 0.71 San Mateo 0.32Solano 0.71 Mono 0.22
1. The Program Reach Index – PRI Raises eligibility from 125 % FPL to the State’s poverty
threshold of 130%FPL
Applying PRI as New Measures of Access
Preliminary - not for reproduction 25
The Program Reach Index (PRI) is as useful an indicator of CalFresh access at the county level as the Program Access Index (PAI).
The ability to apply PRI at below county levels makes it much more valuable in accessing CalFresh reach at community and neighborhood levels.
The PRI can help target outreach activities through mapping techniques that highlight areas needing benefits the most.
HOWEVER, does it help us in identifying “true hot spots” where outreach efforts should be targeted???
Some observations so far!
Preliminary - not for reproduction 26
Nu
mb
er
of
pers
on
s r
eceiv
ing
C
alF
resh
Poverty level of tracts
Federal law
Languages other than
English
Visualphotos.com
????
2. Adjusted Program Reach Index – APRI
Keeps the above adjustment Removes the estimated number of
undocumented persons using data on child-only CalFresh households as proxy
We refer to this method as the Child-only method
Proposed New Measures of Access …contd.
The child-only method is an indirect method of estimating the number of persons ineligible to receive CalFresh due to citizenship status.
Child-only CalFresh households are households in which all participants are minors and all adults:
• receive SSI• are convicted felons
• undocumented immigrants
The Child-Only Method
Proposed New Measures of Access
.
Three important questions:
Q. 1 What proportions of child-only households are headed by parents/adults who are ineligible to receive CalFresh due to citizenship status?
Q. 2 How many undocumented adults live in a child-only household?
Q. 3 How many undocumented adults live in households where there are no children?
The Child-only method…contd.
30
County*
Proportion of child-only
households (June, 2013)
Amador 0.0360Bute 0.0365
Calaveras 0.0369Del Norte 0.0379Humbolt 0.0415Inyo 0.0507Lake 0.0521Lassen 0.0564Mariposa 0.0585Mendocino 0.0618Modoc 0.0689Nevada 0.0746Plumas 0.0837Shasta 0.0961Siskiyou 0.1005Tehama 0.1075Tuolumne 0.1108
Average 0.0653Median 0.0585
Nor
th a
nd Mou
ntain
Bay A
rea
Farm
Bel
tL.
A.
South
ern (
w/o
L.A
.)
0.7269
0.5452
0.7075
0.59 0.6237
PAI by Region*, 2010
Answer 1: 6 percent of child-only cases in every county are due to non-immigration causes such as or parents on SSI and that the remaining 94 percent are due to immigration status
*Alpine, Mono, Sierra and Trinity are excluded
Regions are per. Thomas MaCurdy, Mancuso and Margaret O’Brien-Strain, The Rise and Fall of C a l i f o r n i a ’s We l f a r e Caseload: Types and R e g i o n s , 1 9 80– 1 9 9 9, Public Policy Institute of California, PPIC, 2000.
31
Number of adults in child-only CalFresh households
• Answer 2: There are 1.77 undocumented adults in each child-only household.
• Answer 3: Number of adults in non-participating non-child-only households
For every 177undocumented adults residing in child-only households there are 124 undocumented persons residing in households without children.
Undocumented Households with Children
Number of households
(000s)
Adultsper
householdTotal
adults
2-parent 507 2 1,014
1-parent 155 1 155
Other 9 2 18
Total 671 1,187
Average 1.77
Undocumented Households without Children
Number of households (000s)
Personsper
householdTotal
persons
Married couple 118 2 236
Other families 29 3 87
Solo adult men 435 1 435
Solo adult women 137 1 137
Total 719 895
Average 1.24
32
Calculation Objective Result15,136 x 0.94 Estimate the number of households that are
child-only due to the parents’ citizenship status
14,228
14,228 x 1.77 Estimate the number of undocumented adults residing in child-only households
25,184
25,184 + (25,184 x (124/177)) Estimate the total number of undocumented persons (those residing in households with
children + those residing in households without children)
42,827
Validation
23.4 ÷ 15.07 Fresno’s poverty rate relative to the statewide average 1.55
1.55 x 32.5* Estimate the number of undocumented persons in Fresno that are CalFresh-poor (130% FPL or below) for
every 100 undocumented persons in the county
50.5
42,812 x (100/50.5) Estimate the total number of undocumented persons in Fresno County
84,858
*This is a minimum percentage for all counties (based on simulation)
Validating the Child-Only Method Fresno County, 2011 (15,136 Child-Only Households)
33
Validation: Ten largest child-only counties, 2011 Total Number of
Undocumented Persons
Child-Only Households
Undocumented (130% FPL)
Child-Only, 2011
PPIC (2011)*
Los Angeles 119,837
338,954
964,501 916,000
Orange 29,734
84,101
258,774 289,000
San Bernardino 21,870
61,858
179,320 150,000
San Diego 17,967
50,819
156,366 198,000
Riverside 17,706
50,081
154,095 146,000
Fresno 15,136
42,812
84,858 49,000
Santa Clara 13,468
38,094
117,211 180,000
Kern 11,845
33,503
72,614 46,000
Sacramento 11,260
31,848
97,995 65,000
Alameda 10,808
30,570
94,062 124,000
Statewide 356,627 1,008,705
2,864,504 2,874,500
34
The Child-Only Method vs. PPIC’s method
The Child-Only Method (poverty-based) appears to give a better estimate of undocumented persons than PPIC’s:
o for counties with higher poverty level than the statewide average
o in general, for counties where agriculture is the predominant economy
The PPIC method (tax-return-based) appears to give a better estimate of undocumented persons than the Child-only method:
o for counties with lower poverty level than the statewide average
o for counties with predominantly non-agricultural economies
35
SUMMARY: Statewide PAI Gain Using Different Estimates
Statewide Program AccessUnder Six Scenarios
Numerator 2011
Denominator 2011
%Receivin
g
%age Point GAIN
FNS FNS (2011) 3,760,866 7,684,310 49
--
Adjusted Program Reach Index - APRI
Child-Only
0.94 x 356,627 3,760,866 7,349.091 51
2
Pew
0.94 x 356,627 x 1.78 3,760,866 7,087,60
2 53
4
0.94 x 356,627 x 1.78 x 1.94
3,760,866 6,526,696 58
9
Urban Institute
0.94 x 356,627 x 1.77 3,760,866 7,090,954 53
4
0.94 x 356,627 x 1.77 x 1.70
3,760,866 6,675,605 56
7
36
L.A. County (June, 2013) True + False “hot-spots”
LA County: Program Reach Index - PRI, June 2013
37
True “hot-spots”
LA County: Adjusted Program Reach Index -
APRI, June 2013
38
Child-o
nly HHs
Receiving
CalFr
esh
0-49
% F
PL
50% -
99% F
PL
100%
-124
% F
PL
SSIS
SP
Poten
tial C
F (o
n Med
i-Cal)
English
spea
kers
Non
-Eng
lish
17616178082
92731
911516977559
L.A. Northwest (58 Zip codes)
Child-o
nly HHs
Receiving
CalFr
esh
0-49
% F
PL
50% -
99% F
PL
100%
-124
% F
PL
SSIS
SP
Poten
tial C
F (o
n Med
i-Cal)
English
spea
kers
Non-E
nglis
h
541797912
37530
360254342963
L.A. Northeast (30 Zip codes)
Child
-onl
y HHs
Recei
ving
Cal
Fres
h
0-49
% F
PL
50%
- 99
% F
PL
100%
-124
% F
PL
SSIS
SP
Pote
ntia
l CF
(on
Med
i-Cal
)
Engl
ish
spea
kers
Non
-Eng
lish
125129528 15362
763154
350498
L.A. Southwest (47 Zip codes)
Child
-onl
y HHs
Recei
ving
Cal
Fres
h
0-49
% F
PL
50%
- 99
% F
PL
100%
-124
% F
PL
SSIS
SP
Pote
ntia
l CF
(on
Med
i-Cal
)
Engl
ish
spea
kers
Non
-Eng
lish
86656
842578
363803
2003667
3566000
L.A. Southeast (152 Zip codes)
PRI = 0.64APRI = 0.72
PRI = 0.47APRI = 0.54
PRI = 0.21APRI = 0.22
PRI = 0.53APRI = 0.63
L.A. County Regions
39
Kern
40
41
LIMITATIONS OF THE CHILD-ONLY METHODNumerator and Denominator Data Issues
The reliability of APRI is affected by percentage of unmatched addresses due to P. O. Box addresses.
The denominator becomes negative for tracts with 0 eligibles and where the eligibles based on 130% poverty (formula below) is lower than the SSI component and/or the child-only households component.
Due to small sample sizes and large margins of error, the ACS shows many tracts with fewer eligible persons than the number of persons receiving CalFresh; this leads to APRI greater than1.
A single address is used to provide CalFresh benefits to hundreds or thousands of beneficiaries (over 6,100 recipients in one L.A. County address and 5,100 recipients in one Fresno County address) making it difficult to interpret APRI maps for surrounding areas.
APRI =
42
The reliability of APRI increaseswith increasing geographic scale
Tract Zip code CalFresh region
County
Denominator margin of error
Numerator includes recipients from outside of
geographic unit
LIMITATIONS….Summary
43
Conclusions Geocoding enables us to analyze CalFresh data in the
context of the environments in which recipients live.
Estimates of undocumented persons from the Child-Only Method are broadly consistent with county-level estimates from PPIC and state-level estimates from DHS and PPIC.
o Any discrepancies most likely reflect differences in methodological focus - persons receiving public assistance (Child-Only) vs. persons receiving taxable income (PPIC).
The Child-Only Method can be used with confidence at county levels and for regions within a county.
o In some instances, Zip code or tract-level analysis may be feasible.
It appears that in places where non-English speakers are a minority (example: LA Southwest) the participation rate is significantly lower (APRI = 0.22) than in places where they are a majority.
THANK YOU