Food deserts - University of Utah

Report
FOOD DESERTS
Lori Kowaleski-Jones
Department of Family and Consumer Studies
University of Utah
Energy Balance Research Group
Food Desert

Broadly conceptualized as limited access to
affordable and nutritious foods, particularly in low
income areas
Food Environment and Health




Overweight increasing in the United States
Recent research has linked food environments to health
2008 Farm Bill called for information about the
identification of food deserts
But how to define…
USDA Food Desert Definition


defines a food desert as a low-income census tract where a
substantial number or share of residents has low access to a
supermarket or large grocery store:
To qualify as a “low-income community,” a census tract must
have either:



1) a poverty rate of at least 20% OR
2) a median family income at or below 80 % of the area's
median family income;
To qualify as a “low-access community,”

at least 500 people and/or at least 33 % of the census tract's
population must reside more than one mile from a supermarket or
large grocery store

(for rural census tracts, the distance is more than 10 miles).
Food Deserts in Salt Lake City Utah (USDA)
Alternate definitions



Does it have to be low income areas?
Are deserts mirages? Estimates suggest that only
15% of people shopped for food within their own
census area.
Should the focus be on supermarket?
 (what
about smaller retailers and farmers markets?)
Research on Food Deserts in SLC




Coauthors are Jessie Fan, Ikuho Yamada, Barbara
Brown, Ken Smith and Cathleen Zick.
Different data sets have been used—do they have
comparable information?
Variety of geographic scales to describe local food
environment and the methods used to describe
distance
Our study was an initial assessment of how sensitive
the analyses are to these issues
Our Study

Assess the comparability of data from
 Reference
USA
 Dun and Bradstreet


•
Roughly, about a 1/3 of the records in each data
set are unique
Research results can only be as good as the data
used
564 census block groups
Research Aims
1.
2.
3.
Where? Explore variation with different data sets to
measure food deserts and their prevalence in Salt Lake
County
 Compare “all grocery stores” (GS) to “large grocery
stores” (LGS)
 Simple in-out measure vs. 1-km radius buffer measure
Who? Use multivariate analysis to relate demographic
characteristics and built environment factors to the likelihood
that a census block group will be a food desert.
Low income only? Incorporate an economic component to
the definition of food deserts
All Grocery Store (GS) and Large Grocery Store (LGS) Counts in Salt
Lake County: by Data Source & Sales Volume
All grocery
DB
RefUSA
243
246
Large
70
Grocery (sales
volume>$1M)
95
Research Aim 1: Where?
Alternative Definitions of Food Deserts



Dichotomous measure, 1= BG (Block Group from census)
contains any grocery store
Dichotomous buffer measure captures whether a BG is covered
by any 1km buffer of grocery stores
1km buffer continuous measure: percentage of the BG land
area that is not covered by the 1km grocery store buffer


These three measures are repeated for large grocery stores (sales
volume > = million)
These measures are repeated for whether the above is in a low income
(median household income < 25th percentile for the county) or higher
income BG
R=1km
Food Outlet A
R=1km
Food Outlet C
Census Block Group
R=1km
Food Outlet B
Definition of Food Desert
BG
Income
No GS in BG All $
Low $
BG > 1km from any GS All $
Low $
% of BG area > 1km from any GS All $
Low $
No Large GS in BG All $
Low $
BG > 1km from any Large GS All $
Low $
% BG > 1km from any Large GS All $
Low $
DB
RefUSA
73%
76%
67%
65%
8%
8%
3%
3%
36.1%
37.3%
22.8%
21.5%
88%
83%
88%
81%
22%
15%
18%
17%
58.6%
49.7%
50.1%
42.6%
dichotomous with all grocery stores
Buffer-based dichotomous with all grocery stores
dichotomous with large grocery stores only
Buffer-based dichotomous with large grocery stores only
Results Aim #1


Using the most basic definition, 73% of Salt Lake
County BG is a food desert; higher when only large
grocery stores are considered
Using the 1 km circular buffer definition, much smaller
estimate of food deserts; variation across data sets;
variation even more distinct when large grocery
stores are considered
Aim 2: Who?
Demographic Characteristics and Built
Environment




Ave. residents per sq. mile (density, BG)
Ave. intersections w/in ¼ mile of residence (design,
BG)
Ave of (1999 – Median Yr. Built) (design, CT)
Within the Block Group, the proportions of
 African
Americans
 Hawaii and Pacific Islanders
 Asians
 Hispanics
Odds ratios for BG not covered by 1 km circular buffer of any grocery store
(GS) and not covered by large grocery store (LGS)
No GS
No LGS
DB
RefUSA
DB
RefUSA
% African American
.98
1.21
.94
.97
% Pacific Islander
.84
.71
.80
.78
% Hispanic
.96
.95
1.01
.78
% Asian
.94
.94
.96
.91
Population density
1.04
1.1
1.01
.99
Age of housing
.95
.74
.94
.87
Intersection density
.85
.95
.99
.95
1.02
1.01
1.00
Median family income 1.03
Results Aim #2




Little variation across data sets
Little variation across size of grocery store
Overall, BG that have older housing stock, greater
intersection and population density have lower odds
of being food deserts
BG with higher median family income are more likely
to be food deserts
Aim #3: Low income only?
Separate Analyses for lower
income and higher income CBGs
Results Aim #3

Overall, similar associations between demographic
characteristics and odds of non coverage across
income level of BG—exceptions
In lower income BGs, higher proportion of Hispanics predicts
lower odds for FD, whereas in higher income BGs, higher
proportion of Hispanics predicts higher odds for FD
 In lower income BGs, higher median income predicts lower
odds for FD, whereas in higher income BGs, higher median
income predicts higher odds for FD

Discussion
•


Aim #1: variation across data sets in identification of food deserts
 Future research and policy activity should be careful to appreciate the
limitations of data used to define and identify food deserts.
 Recommendation: If the focus is to identify FD for policy implementation, more
detailed measurements and multiple data sets give a more comprehensive
picture.
Aim #2: limited variation across data sets in estimating correlates of food deserts
 Surprising, given the lack of agreement across data sets
 Recommendation: If the focus is to estimate correlates of food deserts,
employing a single data source seems to be fine.
Aim #3: some variation in direction of odds ratios across income specification
 Preliminary and likely proxying for more complicated processes
 Recommendation: future research incorporate economic dimensions in food
desert research.
Implications

Controversy that there has been a dearth of
evidence linking food deserts to dietary health.
Consensus that improving access might not change
consumer behavior. Patterns might persist, role of
preferences and prices.
Food Deserts and Public Policy


Recent activity has been the
Healthy Food Financing Initiative
2010 with the aim of eliminating
food deserts
More information on spatial
elements of food environment
 http://www.ers.usda.gov/FoodAtl
as/
 http://www.ers.usda.gov/data/fo
oddesert/

similar documents