主题：Incorporating Time Costs into SNAP Allotment Calculation: A Home Food Production Time Use Analysis
主讲人：游雯 Associate Professor, Department of Agriculture and Applied Economics, Virginia Tech
讲座时间：2015年12月1日 10:00~11:30
讲座地点：信息楼530
讲座内容：
讲座时间：2015年12月1日 10:00~11:30
讲座地点：信息楼530
讲座内容：
The Supplemental Nutrition Assistance Program (SNAP), formerly known as the
Food Stamp Program, serves as the largest hunger safety nest in the United States.
The program aims at enabling low-‐income eligible people to purchase foods needed
for a nutritionally adequate diet through providing monthly benefits pre-‐calculated
for achieving the goal. In 2014, the program served about 47 million people with a
total program cost of about $74 billion. The average monthly benefits in 2014 were
around $125 per person and $256 per household (FNS, 2015a). Thorough
evaluations of the SNAP program effectiveness at achieving its goals (i.e., achieving
food security and healthier diets) are wanted given the sizable program costs and
are essential for utilizing the vast coverage of SNAP program to achieve the largest
public health impact.
Several proposals and calls for actions have been brought to the spotlight: such
as proposing different versions of SNAP program modifications through imposing
restrictions on food purchased using SNAP benefits (e.g.,You et al., 2012; Pomeranz
and Chriqui, 2015) and providing incentives for healthy food purchasing (e.g.,
Olshoa et al., 2015; FNS 2014c). All of those theoretical predictions and field
experiments focused on whether or not the arbitrary chosen “boosts” (no matter it
is in the form of rewards or punishment) have significant impact on SNAP
participants’ food security and health outcome improvements. However there is a
remaining unanswered question: how much of the maximum SNAP allotment should
be adjusted to address the majority of the potential inadequacy at reaching SNAP
program goals?
In 2013, the Institute of Medicine/National Research Council committee
published a report per the United States Department of Agriculture (USDA) request
specifically on addressing the SNAP benefit adequacy. The report identifies a set of
adequacy-‐relevant factors and calls for research to directly incorporate those
factors into the SNAP allotment calculations (IOM/NRC, 2013). The three key factors
identified by the committee that are currently absent from the benefit calculation
are: time required to shop and prepare food at home, price variation across regions
and unequal access to healthy food. The ignorance of time factor has been
recognized in the literature as a significant road-‐block for SNAP to achieve its
nutritional goals (Rose 2007; You et al., 2009; Davis and You 2010a, 2010b, 2011,
2013; Rashcke 2012; Davis 2014) and is the focus of this paper.
The time (labor) cost absent from the calculation is an important oversight since
labor is an inseparable part of at-‐home food production. Ignoring labor (time) cost
will lead to an overly optimistic evaluation of SNAP program’s effectiveness of
reaching nutrition goals: the number of people not reaching the SNAP program
nutrition goals will be underestimated or stated alternatively, the number of people
classified as meeting the target will be over estimated. This study calculates the
appropriate scaling factor, k, which can be used to adjust the SNAP benefit amount
so that the benefit amounts are consistent with the Money-‐Time Threshold implied
by the Thrifty Food Plan (which is the base for SNAP benefit calculation). The
Money-‐Time Threshold (MT) formula is: !" = ! !"# + !! !"# − !! !"#$!% , where M TFP
is the actual SNAP benefit according to TFP, p is the opportunity cost of time, T TFP is
the TFP implied home food production time and the T actual is the actual home food
production time by households. This study looks at how much a scaling factor k can
be used in order to increase the SNAP benefit, M TFP , to match the MT threshold
requirement. In other words, it will seek to satisfy: !" = ! !"# ×(1 + !). After
substituting the MT formula, we get the following formula for k: ! = !(! !"# −
! !"#$!% )/! !"# .
This study utilizes American Time Use Survey (ATUS) data to get information on
T actual . For single-‐headed households, the respondent time use data can be treated
as the T actual if the respondent is the main food preparer and main food shopper
which we utilize information provided by Eating and Health Module (EHM)
supplement to ATUS to identify. For dual-‐headed households, the missing spousal
time use information requires imputation. We present an algorithm to be used to
predict spousal food production time use information using information provided in
ATUS respondents’ time use and spousal demographic and employment information
along with the EHM food production role identification information.
The sample was restricted to those households with one or two adults to avoid
predicting other adults’ time use since we do not have any further information on
other adults of the households. We further limit the sample to those households
indicating that their household income is less than or equal to 185% of the poverty
threshold since the focus of scaling the SNAP benefits is to help low-‐income
households. We further drop households with respondent indicating that he/she is
sharing the food production roles with other adults in the household since there is
no information available to assume the percentage split of the responsibility.
Two-‐part models are utilized to estimate and predict time use based on food
production role differences and the scaling factor k is calculated for the following
four categories of households: single-‐headed households with children, single-‐
headed households without children, dual-‐headed households with children, dual-‐
headed household without children. To account for the uncertainty in the
opportunity cost of time, p, and the estimation noises in the actual time use,
parametric bootstrapping is conducted to map out the empirical distribution for the
k factor.
We find that ignoring spousal contribution to household food production
underestimates household home food production time by about 3 hours per week.
Without considering the spousal contribution to food production time, we are likely
to overestimate the labor adjusted full TFP cost deficit by about $50 per week. On
average, the suggested scaling factor, k, is less than 0.5. Furthermore, we compare
the adjusted TFP amount with the Low-‐Cost Food Plan (LCFP) amount and
Moderate-‐Cost Food Plan (MCFP) amount. The adjusted TFP amount has much
higher median than both LCFP and MCFP for single-‐headed households with and
without children. Without considering spousal contribution to the production time,
the dual headed households’ adjusted TFP amount also has slightly larger median
than LCFP and MCFP. However, after considering spousal contribution to the food
production, the dual headed households’ adjusted TFP amount became slightly
smaller than the LCFP and much smaller than the MCFP. This signals that dual-‐
headed households have the flexibility of spousal time substitution therefore their
time constraints are not as tight as the single-‐headed households and require
smaller time-‐cost adjustment to meet the threshold.