Food Stamp Program Participation and Food Insecurity: An Instrumental Variables Approach

1
Food Stamp Program Participationand Food
InsecurityAn Instrumental Variables Approach

• Steven T. Yen
• The University of Tennessee
• Margaret Andrews
• Economic Research Service, USDA
• Zhuo Chen
• The Centers for Disease Control
• David B. Eastwood
• The University of Tennessee

2
Overview
2

• Food Stamp Program (FSP) is a major USDA food
  program.
• Food insecurity (FI) is an important measure of
  welfare.
• We investigate
• relationship between FSP participation (binary)
  and FI (censored)
• effects of policy/socio-demographic variables on
  FSP participation and FI.

3
Overview

• Motivated by mixed results from previous studies
• insignificant/positive effects of FSP on FI
• insignificant/negative effect of FI on FSP
• inefficient 2-step estimates
• Estimation by maximum-likelihood method (more
  efficient)

4
Overview

• Major findings
• FSP participation decreases FI
• Also playing important roles in FSP FI
• State-level policy variables
• e.g., EBT, short re-certification
• Socio-demographic variables
• e.g., marital status
• Being married reduces participation in FSP
• Being married reduces FI

5
Introduction USDA Food Programs

• USDA implements 16 Food Assistance and Nutrition
  Programs
• to provide low-income families and children with
  access to a healthy diet.
• The 16 programs were funded at 42.9 billion in
  FY 2004.
• 1 in 5 Americans participated in one/more
  programs at some point each year.

6
Introduction Food Insecurity in America

• Most recent USDA food security survey indicates
  that
• 3.5 of U.S. households (3.8 million people) were
  food insecure with hunger
• 7.6 (8.3 million people) were food insecure
  without hunger.
• NFSPS sample (1996-97, low-income) 26.6 were
  food insecure during past 30 days.

7
Empirical Literature

• Earlier studies of food insecurity addressed
  food insufficiency.
• Gunderson and Oliveira (2001)
• investigated food stamp participation and food
  insufficiency, using 19911992 SIPP panels
• used two-step estimator for simultaneous-equation
  probit (Mallar, 1977).
• Findings
• Mutual effects of FSP and FI are insignificant.
• FSP participants had same probability of food
  insufficiency as non-participants.

8
Empirical Literature

• More recent studies used data for FI.
• Huffman and Jensen (2003)
• use the SPD data
• estimated simultaneous-equation probit, also
  using Mallars (1977) procedure
• FS participation (0-1)
• Labor-force participation (0-1)
• food insecurity (0-1).
• Findings
• FI decreases FSP participation
• FSP participation does not affect FI

9
Empirical Literature

• Jensen (2002) estimated the FSP participation
  (0-1) and food insecurity (0,1,2,3), in a
  2-equation SUR system
• Finding FI and FSP are affected in the same
  direction by random shocks or unmeasured
  effects.
• (drawing on significant error correlation)

10
Empirical Literature

• Based on the 1995 and 1999 Food Security
  Supplements to the CPS, Nord (2001) found
  households receiving food stamps registered
  almost no change in the measured prevalence of FI
  or hunger during the period.
• Only one study has shown a clear positive
  association between FI and the use of public
  assistance. Borjas (2004) shows that a 10 cut in
  the fraction of the population that receives
  public assistance increased the percent of FI
  households by about 5.

11
Empirical Literature

• Kabbani and Yazbeck (2004) pooled multiple years
  of data from the CPS and, using a two-stage
  estimation procedure to control for the
  endogeneity of the program participation
  decision, Results suggest that participation in
  the FSP appears to moderate the observed
  differences for households with children aged 5
  to 18, but not significantly.

12
Empirical Literature

• Kabbani and Kmeid (2005)
• Logit analysis
• Sample CPS FSP-eligible households that
  experienced hunger during the year
• Dependent variable (0-1) experienced hunger
  during the last 30 days
• Results
• FSP participation (0-1 exogenous) does not
  affect the odd of hunger
• FSP amount (exogenous) significantly reduces the
  odd of hunger

13
Empirical Literature

• Apparent inconsistency among previous results
  suggests
• that a more careful investigation between FSP
  participation FI is needed.

14
This Study

• Investigates
• relationship between FSP and FI
• effects of socio-demographic variables on FSP and
  FI,
• using an instrumental variables approach.

15
This Study

• NFSPS 1996-97 data offer a unique opportunity for
  such a pursuit
• include the 18 items of the food security module
  (used for annual monitoring of FI)
• include follow-up questions that allow
  examination of FI status in the past 30 days
• variations in state-level variables
• EBT adoption
• re-certification periods
• provide better instruments than those used in
  previous studies (except Kabbani and Wilde, 2003
  Kabbani and Yazbeck, 2004, 2005)

16
This Study

• Methodology accommodates endogeneity of FSP in
  the FI equation
• Results
• FSP improves () FI.

17
Econometric Models
18
Econometric Models
19
Econometric Model
20
Data

• National Food Stamp Program Survey (NFSPS),
  199697.
• Low-income sample (below 150 of poverty
  threshold).
• Endogenous variables
• FSP participation (0-1)
• FI constructed from responses to 18 questions
  in Food Security Module
• Censored (considered in this study)
• Binary (considered elsewhere)
• Ordinal (considered elsewhere)

21
Data
Food Security Module Sample Questions 3. The
food that (I/we) bought just didnt last, and
(I/we) didnt have money to get more. Was that
often, sometimes, or never true for (you/your
household) in the last 12 months?
Often true Sometimes true
Never true DK or
Refused 4. (I/we) couldnt afford to eat
balanced meals. Was that often, sometimes, or
never true for (you/your household) in the last
12 months? Often true
Sometimes true Never true
DK or Refused
22
(No Transcript)
23
(No Transcript)
24
(No Transcript)
25
Results
26
(No Transcript)
27
(No Transcript)
28
(No Transcript)
29
Results
Effect of FSP participation on FI
Sample Mean 7.36 SD1.84 (FIgt0, FSP0) Mean
6.88 SD1.68 (FIgt0, FSP1)
30
Marginal Effects on Probabilities
31
Marginal Effects on Probabilities
32
Marginal Effects on Probabilities
33
Marginal Effects on Probabilities
34
Marginal Effects on Probabilities
35
Marginal Effects on FI
36
Marginal Effects on FI
37
Marginal Effects on FI
38
Marginal Effects on FI
39
Concluding Remarks

• Nord, Andrews, and Carlson (2004) cautioned that
  the relationship between food assistance programs
  and FI is complex due to the two-way causality.
• This study answers that call to address the
  causality issue by developing a instrumental
  variables model of FSP participation and FI.

40
Concluding Remarks

• Results suggest
• lack of association between FSP participation
  and FI in previous studies is likely due to the
  failure to accommodate the simultaneity, sampling
  weight, and the error covariance structure.

41
Concluding Remarks

• Our results
• FSP reduces FI.
• Obvious policy implications

42
Concluding Remarks

• Results in sharp contrast to previous findings on
  effect of FSP on FI
• Opposite (Huffman and Jensen, 2003)
• Predominantly insignificant/non-existent
  relationship
• (Gundersen and Oliveira, 2001 Gibson-Davis and
  Foster, 2005 Jensen, 2002 Kabbani and Yazbeck,
  2004, 2005 Oberholser and Tuttle, 2004).

43
Concluding Remarks

• Use of state-level FSP policy variables
• used improve the quality of instruments (without
  which the model would have been difficult to
  estimate).
• Allow drawing more policy-relevant conclusions.
• Econometric issues state-level variables which
  vary only across states but not observations
  use of hierarchical models (Raudenbush and Bryk,
  2002)??

44
Concluding Remarks

• Short re-certification ( on FSP on FI prob
  level)
• Improve to promote FSP participation and
  eliminate FI
• EBT implementation (n.s. on FSP/FI)
• may be related to the combined effect of
• reduced stigma
• difficulty to use among the poor who are less
  likely to own a debit card.

45
Concluding Remarks

• Other determinants of FSP participation and FI
  have important policy implications as well. For
  instance,
• married and widowed individuals have a lower FSP
  participation rate than individuals who are never
  married
• this segment of the population can be targeted
  for promotion of FSP participation
• which, in reference to the negative effect of FSP
  participation on FI, can help reduce FI among
  these individuals (true for married not for
  widowed)

46
Concluding Remarks

• Dissatisfied-Shopping ( on FI probability and
  level)
• Establishment of more easily accessible and
  friendly shopping facilities may also improve
  (reduce) on consumer dissatisfaction with
  shopping in the neighborhoods,
• which will help reduce FI.