Characteristics of long-term participants in the food stamp program final report

              f. f^^^^SU
United States
Department of
Agriculture
Food and
Nutrition
Service
Office of
Analysis and
Evaluation
Characteristics of
Long-Term
Participants in the
Food Stamp
Program
December 1992
*
v. OMPLETEt
CHARACTERISTICS OF
LONG-TERM PARTICIPANTS
IN THE FOOD STAMP PROGRAM
FINAL REPORT
December 1992
Barbara Fay Murphy
Office of Analysis and Evaluation
Food and Nutrition Service
Marielouise Harrell
Sigma One Corporation
TABLE OF CONTENTS
Executive Summary i
1.0 Introduction 1
2.0 Patterns of Participation in the Food Stamp Program 8
2.1 Cross-Sectional Cohort 15
2.2 Full Panel Population 21
2.3 Censoring of Duration of Participation Spells 23
3.0 Distribution of Food Stamp Benefits By Participation Categy 27
3.1 Benefits Received by the Cross-Sectional Cohort 27
3.2 Benefits Received by the Full Panel Population 29
4.0 Characteristics of Long-Term Participants 34
4.1 Demographic and Household Characteristics of FSP Participants 34
4.2 Income Levels and Sources of Income 42
5.0 FSP Work Registrants 51
5.1 Distribution of Work Registrants by Participation Category 51
5.2 Characteristics of Work Registrants 55
5.3 Characteristics of Long-Term Work Registrants 59
6.0 Conclusions 63
References
Appendix A
LIST OF FIGURES
Figure 1. Cumulative Distribution of the Length of Time an Individual is 10
on the Food Stamp Program, for Single-Spell Participants
Figure 2. Cumulative Distribution of the Length of Time an Individual is on 12
the Food Stamp Program, adjusting for the month of entry in the FSP
Figure 3. Cumulative Distribution of the Total Number of Months Spent on 14
the FSP for Individuals who participated in the FSP for more than
one spell during the 1987 SIPP Panel
Figure 4. Cumulative Distribution of the Length of Time in Months that an 18
Individual is on the FSP for the Single-Spell Participants in the
Cross-Sectional Cohprt
Figure 5. Share of Total Participation Months Throughout the 1987 SIPP Panel ... 20
for the Cross-Sectional Cohort, using Welfare History Data
Figure 6. Composition of Participation Months for Individuals in the Food 24
Stamp Program for the Full Panel Population
Figure 7. Total Food Stamp Program Benefits Received by the Cross-Sectional .... 30
Cohort
Figure 8. Total Food Stamp Program Benefits Received by Participants in 33
the 1987 SIPP Panel
W
LIST OF TABLES
Table 1. Participation Patterns of Cross-sectional Cohort, using welfare 17
history data
Table 2. Participation Patterns of the Full Panel Population , 22
Table 3. Percentage of Participants with Probable Left & Right Censoring 26
of Spell Duration
Table 4. Cost per Household for Individuals Participating in the Food Stamp .... 28
Program in the Cross-Sectional Cohort, using Welfare History Data
Table 5. Cost per Household for the Full Panel Population 31
Table 6. Distribution of Age for the Full Panel Population 35
Table 7. Race or Ethnic Origin of the Full Panel Population 37
Table 8. Last Year of Education Completed by Individuals Age 18 or older 38
in the Full Panel Population
Table 9. Household Composition for the Full Panel Population 40
Table 10. Household with Children under Age 18 for the Full Panel Population ... 41
Table 11. Distribution of Household Income for the Full Panel Population 43
Table 12. The Dynamics of Poverty for the Full Panel Population 45
Table 13. Sources of Income for the Full Panel Population 46
Table 14. Participation in Assistance Programs for the Full Panel Population 48
Table 15. Labor Force Status of Food Stamp Participants Over 15 years of age ... . 50
for the Full Panel Population
Table 16. Distribution of Participants by Work Registration Status for 52
Cross-Sectional Cohort, using welfare history data
Table 17. Distribution of Work Registrants by Participation Category 54
V
LIST OF TABLES
(continued)
Table 18. Education Level of Work Registrants by Participation Category 56
Table 19. Receipt of Training Services by Work Registrants 57
Table 20. Employment History of Work Registrants 58
Table 21. Reasons for Leaving their last job given by Work Registrants 60
Table 22. Characteristics of Long-Term Work Registrants 61
Table 23. Distribution of FSP Population Using Alternate Measures of Participation . 65
VI
EXECUTIVE SUMMARY
Characteristics of Long-Term Participants
in the Food Stamp Program
In the annals of research on participants in Federal assistance programs, three distinct groups
of participants emerge ~ those who rely on assistance to get them over a short period of
difficulty, those who depend on assistance continually for a significant part of their lives, and
those who receive assistance sporadically throughout their lives, but still rely on assistance for
a substantial amount of time. While the existence of these distinct groups is accepted, little is
known about their size or characteristics. It is likely that the characteristics of these groups vary
considerably and that these differences may provide an understanding of why some people
receive assistance for short periods of time while others seem unable to break out of the poverty
trap.
This study uses the 1987 Survey of Income and Program Participation (SIPP) to describe time
patterns of Food Stamp Program (FSP) participation. Participants who had only one
participation spell during the survey period were classified as either short-term (on the program
8 months or less), medium-term (on the program 9-23 months) or long-term participants (on the
program 24 months or longer). Persons who were on and off the FSP during the survey were
classified as multiple-spell participants. Each group is described in terms of the proportion of
the FSP they represent, the share of total benefits they receive, and the characteristics they share
as distinct groups.
The study looked at both a cross-section of participjints who were on the FSP in the first month
of the survey (the cross- sectional cohort) and at all individuals who participated in the FSP for
at least one month of the October 1986 through March 1989 survey period (the full panel
population). From either perspective, a substantial proportion of FSP participants in the SIPP
Panel were long-term participants (59 percent of the cross-sectional cohort and 33 percent of the
full panel population). In fact, when taking multiple-spell participants into account, of the
estimated 18.8 million persons who participated in the FSP at the beginning of the survey
period, 11.6 million persons (62 percent) were also on the program at the end of the panel
period 28 months later.
Not surprisingly, the groups most dependent on assistance (long-term and multiple-spell
participants) consumed the majority of program benefits over time. The cross-sectional cohort
received $17.3 billion in benefits during the 28-month survey period. Although long-term
participants made up only 59 percent of the cross-sectional population, they consumed 72 percent
of the benefits. Multiple-spell participants accounted for an additional 24 percent, so a total of
96 percent of all benefits went to these two groups. Similarly, of the $21.6 billion in benefits
paid to the full panel population, 82 percent went to long-term and multiple-spell participants,
although they made up only 58 percent of all participants. Short-term participants, who made
up about 29 percent of the population, received only six percent of the benefits.
i
I///
There were noticeable differences in household composition between long-term, multiple-spell,
and other participants. The majority of single-elderly households (60 percent) were long-term
participants. Likewise, single-parent households were likely to be dependent on food stamps,
either as long-term or multiple-spell participants. In other words, arguably the two most
vulnerable groups among the poverty population were the most likely to be dependent on the
FSP over time.
Long-term participants were also more likely to be chronically poor. Eighty-nine percent of the
long-term participants lived in households with incomes that were below the poverty line in the
first month of FSP participation and 60 percent had incomes below poverty for all 28 months
of the SIPP Panel. In contrast, only 58 percent of the short-term participants lived in households
with incomes below 100 percent of poverty in their first month on the FSP, and only four
percent had incomes below poverty in every month. Multiple-spell participants, as expected,
were most likely to drift in and out of poverty (76 percent), but were also likely to be poor in
their first month of FSP participation (again, 76 percent).
The income long-term participants did receive was more likely to come from public assistance
than from employment. Sixty-three percent of long-term participants' household income came
from means-tested transfers (compared to 42 percent for the full panel population in general);
only 15 percent was from earnings. On the other hand, half of short-term participants' income
was earned and only 20 percent was from transfers. Short-term participants were nearly four
times as likely as long-term participants to be employed in their first month on the FSP.
Multiple-spell participants were more diverse in terms of their income sources: slightly more
than one-third was from earnings, another one-third was from transfers, and one-quarter was
from other sources, such as unemployment compensation. About half of all adult multiple-spell
participants were either working (presumably at low-wage jobs as their incomes tended to be
below poverty) or unemployed and actively seeking work.
Short-term participants achieved higher educational levels than long-term participants. While
over 60 percent of short-term participants had completed high school, nearly 70 percent of long-term
participants had not. Clearly, the probability of participating for more than two years falls
as education increases. Conversely, the probability of participating for less than eight months
rises as education increases.
In general, those statistics paint a picture of long-term and multiple-spell participants as the most
disadvantage*! among the poor - those with the least education, the fewest ties to the labor
force, and the least available income.
Given the characteristics of long-term and multiple-spell participants, it is no surprise they were
the least likely to be required to register for work under the FSP. Only four percent of long-term
participants and nine percent of multiple-spell participants had to register for work.
However, long-term participants represented a substantial portion of all work registrants - 37
percent of the work registrants in the cross-sectional cohort were long-term participants.
Multiple-spell participants made up another 38 percent. These results indicate that, in a given
,?' l////
month, three-quarters of all work registrants are in the midst of a spell that will last two years
or more or are likely to return to the program after leaving. Since these groups also consume
the greatest amount of program resources over time, targeting them with employment and
training services may have the most impact. On the other hand, since so many work registrants
are long-term or multiple-spell participants, targeting may not be necessary - these individuals
will show up in employment and training programs as a matter of course.
in
IX
1.0 INTRODUCTION
The U.S. Department of Agriculture and other Federal agencies have sponsored a great deal of
research on the dynamics and determinants of participation in the Food Stamp Program (FSP)
and other assistance programs. From this research three groups of participants clearly emerge -
- those who participate in assistance programs for short periods of financial difficulty, those who
depend on assistance for a significant part of their lives, and those who receive assistance
sporadically throughout their lives, but still rely on assistance for a substantial amount of time.
Although the existence of these three distinct groups is accepted, little is known about their size
and characteristics. It is Sikely that the characteristics of these groups vary considerably and that
these differences may provide an understanding of why some people receive assistance for short
periods of time while others seem unable to become self-sufficient.
Previous studies that exatnined the issue of long-term participation in the FSP (Burstein and
Visher (1989); Usher et al. (1989)) produced mixed results. Both studies found that a substantial
proportion of the FSP population remained on the program for a long time. Burstein and Visher
found that certain demographic and income characteristics had significant effects on the duration
of FSP spells. In particular, households with more children, those with fewer adults, those with
older heads of household and those with African American heads of household were more likely
to remain on the FSP for long periods. However, Usher et al. found that the impact of
individual and household characteristics on patterns of participation appeared to be weak. Some
of the differences in results may be attributable to differences in data sources used for the study.
Burstein and Visher studied a nationally representative sample of food stamp participants; Usher
1
et al. studied a selected group of work registrants in four counties in Alabama and Washington.
This study uses data from the 1987 Survey of Income and Program Participation (SIPP) to
describe the time patterns of participation in the FSP. Participants were divided into groups
depending on their participation patterns. Each group is described in terms of:
(a) the proportion of the Food Stamp Program they represent;
(b) the share of total benefits they receive; and
(c) the characteristics they share as distinct groups.
There are several alternative ways to measure time patterns of participation in the FSP. One
method is to follow an entry cohort (all persons who initially entered the FSP during a given
period) for a fixed period of time.' A second method is to examine the patterns of participation
for a cohort of individuals who receive food stamps in a given month (i.e., follow all
participants in a particular starting month until they leave the program or the observation period
ends). This group will include many people who have received benefits for some (unknown)
time and some who have just entered the program. A third method examines the patterns of
participation for all persons during a given reference period (e.g., January 1, 1990 through
December 31, 1990).
Each method of analysis presents a very different picture of FSP participation patterns.
'This is the method employed by Usher et al. and Burstein.
2
Examining an entry cohort produces estimates of the proportion of people entering the program
during a given time period who will become long-term participants. Previous research on FSP
participation has used this method of analysis. Looking at all participants at a given point in
time, using a cross-sectional cohort, describes how many participants at a given point in time
are or will become long-term participants. Finally, describing participation patterns over a
reference period (a year, two years, etc.) provides a different view of long-term participants.
It measures the proportion of all participants who participated in the FSPfor a long time during
the reference period.
The estimates of the share of FSP participants who are long-term participants will be smaller
using the third method of analysis as compared to the second method. At any one point in time,
long-term participants will dominate the caseload due to the fact that they remain on the program
for such a long time.2 Because of this, the third method allows closer examination of short-term
and multiple-spell participation patterns, which may be missed altogether when analyzing a
cross-sectional cohort.
Multiple-spell participants are in their own way as dependent on the FSP over time as single-spell
participants who stay on the program for one long spell. However, it is likely that the
characteristics of multiple-spell participants differ from those of long-term, single-spell
participants. Some of these differences may enable these individuals to leave the program for
2David Ellwood (1986) illustrates this point in an analogy of hospital bed usage. While the
majority of persons admitted to the hospital will stay for only brief periods, the majority of beds
are occupied by long-term care patients.
short periods of time, although they inevitably return.
This study analyzes the 1987 SIPP Panel data3 using the second and third methods of measuring
participation patterns. That is, participation patterns and benefit information are presented for
two groups of participants:
(1) a cross-sectional group of participants who were on the FSP in the first month of
the survey4 (the cross-sectional cohort); and
(2) all individuals who participated in the FSP for at least one month of the survey
period (the full panel population).
For the first group, the cross-sectional cohort of participants, retrospective data are used to
adjust the duration of the participation spell which was ongoing in the first month of the
survey.5 For the second group, only FSP participation as reported within the 28 months of the
SIPP Panel is used to analyze participation patterns.
3The 1987 SIPP Panel followed individuals for a 28-month period from October 1986
through March 1989. Data were collected in waves of four month intervals, with each wave of
data constituting the present information plus a retrospective look at the preceding three months.
*The SIPP survey design includes four rotation groups which were interviewed in different
months; thus for any individual, the first month of reported data was any month between
October 1986 and January 1987.
The 1987 SIPP includes a welfare history module that collects retrospective information on
participation in Federally-supported programs prior to the survey period.
Participants who had only one participation spell during the survey period were classified as
either short-term (on the program 8 months or less), medium-term (on the program 9-23 months)
or long-term participants (on the program 24 months or longer). The cutoff values which
distinguished short-term from medium-term participants and medium-term from long-term
participants were determined after examining the distribution of the length of time that
individuals reported receiving FSP benefits during the 28-month survey period.
Persons who were on and off the FSP during the survey were classified as multiple-spell
participants. They were analyzed separately in order to assess if their characteristics were
similar or dissimilar to the single-spell participants.
The number of participants and the proportion of long-term participants is highly dependent on
the method which is used to define participation. For example, we estimate that 18.8 million
persons participated in the FSP in the first month of the panel and that 59 percent were long-term
participants. In contrast, we estimate that 28.7 million persons participated in the FSP for
at least one month between October 1986 and March 1989 and that 33 percent were long-term
participants.
The first estimate describes the participation patterns of a cross-section of participants already
on the program; the second describes the participation patterns of all those receiving benefits at
any time during the 28-month panel and reflects the familiar pattern of turnover in the FSP
(persons who leave the program are replaced by new persons). Aside from the method of
analysis chosen, these estimates vary for other reasons:
(1) welfare history data are used to adjust the duration of the ongoing participation
spell for the cross-sectional cohort of participants;
(2) persons in the full panel who began participating in the FSP late in the survey
may be misclassified as short-term instead of long-term because the true length
of their participation is unobservable; and
(3) analysis of a cross-sectional cohort increases the probability of persons being
long-term participants because at any one point in time, the long-term participants
will dominate merely because they remain on the program for a longer period of
time than other participants.
In addition to quantifying the number of FSP participants and the proportion of the program
benefits they receive, the study objectives include an analysis of the characteristics of each
participant group. By examining the differences between the participant groups, we may gain
further insight into why some people are able to break free from poverty while others not.
The characteristics of FSP participants were examined for both the cross-sectional cohort and
the full panel population. The results did not differ appreciably between these two groups, so
the study presents the analyses for the full panel population only.
A final question to be addressed is what types of participants are subject to work registration
requirements. Work registration requirements are designed to select participants who are more
readily employable. Therefore, it is important to know if some portion of those participants who
are dependent over time is potentially work-ready and could be targeted for employment and
training services. In this study, we simulated the FSP work registration requirements to
determine if there were significant differences in the proportion of work registrants for the
various participation patterns.
2.0 PATTERNS OF PARTICIPATION IN THE FOOD STAMP PROGRAM
As described earlier, this study examines food stamp participation patterns for two groups:
(1) a cross-sectional group of participants who were on the FSP in the first month of
the 1987 SIPP Panel (the cross-sectional cohort); and
(2) All individuals in the FSP at any time during the 28 months of the survey period
(the full panel population).
For both groups FSP participation patterns are presented for individuals, not households. This
is because persons in a given food stamp household may enter or leave the FSP and/or the SIPP
Panel during the panel period for a variety of reasons (births and deaths, marriages and
divorces), and this can alter the configuration of food stamp households.
Participation categories were defined after examining the number and length of spells for the full
panel population. Of the estimated 28.7 million individuals who participated in the FSP for at
least one month during the survey period, three out of four (about 21 million) had only one
participation spell. The remaining 25 percent (nearly 8 million individuals) had anywhere from
two to five participation spells in the 28-month observation period.6 Participation patterns were
6Less than one percent had four or five participation spells during the 28-month survey
period.
8
examined separately for the single- and multiple- spell participants in order to examine any
differences between these two groups.
Single-Spell Participants
For the single-spell participants in the full panel population, spell length was further categorized
as short-term, medium-term, and long-term. The cutoff points for these three categories were
determined by looking at the cumulative distribution of spell lengths for all single-spell
participants. Figure 1 shows that the median duration for a single spell of participation was 19
months;7 the average spell length was 16 months out of a maximum of 28 months.
Twenty-six percent of the single-spell participants were on the program for four months or less.
An additional 12 percent had spells lasting between five and eight months. After the eighth
month there is little change in the probability of exiting from the program from one month to
the next. The sharp increase between months 27 and 28 can be explained by the fact that nearly
40 percent of the full panel population participated in the FSP for the entire panel period - in
other words, these individuals remained on the program for at least 28 months.
The full panel population was divided into three subsamples to further analyze the participation
patterns of this group: (1) those who were on the FSP during the first month of the panei (65
7The sharp jumps in the cumulative distribution plot at four month intervals are a result of
the data collection process. SIPP data are collected in waves at four month intervals with each
wave of data containing the present information and a retrospective look at the preceding three
months.
Figure 1 Cumulative Distribution of the Length of Time an Individual is
on the Food Stamp Program, for Single Spell Participants
10t
Distribution of Sp»l I Length
fr Blnoi. IwlI Krtlelsanta
I I I I I I I I I I I I I I I I I I I I I I I I I I
12345B7i • 1O11iail-M1Siai7iaiaM21Z223»0M27
UM0U* of *>•! I Onotha}
Length Cumulative Length Cumulative
ofspeU percent of spell Frequency
is <= months in months
1 month 6.3 15 months 46.8
2 months 11J 16 months 48.4
3 months 14.9 17 months 49.1
4 months 26.3 18 months 49.3
5 months 29.5 19 months 49.8
6 months 30.6 20 months 53.4
7 months 332 21 months 54.0
8 months 38.4 22 months 54.4
9 months 39.2 23 months 55.1
10 months 40.6 24 months 58.9
11 months 41.3 25 months 59.9
12 months 45.7 26 months 60.5
13 months 46.1 27 months 61.5
14 months 46.6 28 months or more 100.0
Source: 1987 Survey of Income and Program Participation (S1PP).
10
percent of all single-spell participants); (2) those who entered the FSP in months 2 through 5 of
the panel (10 percent of single-spell individuals); and (3) those who began participation on or
after the 6th month of the panel (25 percent of single-spell participants).8
Figure 2 presents the cumulative distribution of the length of time spent on the FSP for single-spell
participants for these three subsamples. The sharp increase between months 27 and 28 for
the month one cohort is again a result of the large percentage of individuals who participated in
the FSP for the entire survey period (60 percent of this group). In fact, when taking multiple-spell
participants into account, ofthe estimated 18.8 million persons who participated in the FSP
at the beginning of the survey period, 11.6 million (62 percent) were also on the program 28
months later when the survey ended.
As seen in Figure 2, the shapes of the cumulative distributions of spell duration were similar for
each subsample. Using Figures 1 and 2, the following categories of participation for single-spell
participants were defined:
(1) Short-term participants (those persons who participated in the FSP for eight
months or less);
(2) Medium-term participants (those persons who participated in the FSP for more
than eight months, but less than 24 months); and
(3) Long-term participants (those persons who participated in the FSP for at least 24
consecutive months).
"Entry months other than month one were aggregated to ensure adequate sample sizes.
11
Figure 2 Cumulative Distribution of die Length of Time an Individual
is on the Food Stamp Program, adjusting for the month of entry in the
FSP
Distribution of Spell Length
for SIno la Spall Partlolpanta
I
O nonth 1 cotartCnO
Unath of Bpal I
+ amrtha 3-3 cmo
•♦Case
Lang*
of Spell of tpelt ra <» month*
P-tfcipwlng in FSP
in Month l
Started raP in
Month 2-5
Started PSP in
Month 6+
2i
3i
4 months
Smooth.
6 month.
7l
Si
•l
IOI
Hi
121
13 1
141
131
161
171
Hi
19 1
20 1
21
22 1
23 1
24 1
23 1
26 1
27i
2«i
2.3
3.7
3.1
12.7
13.6
14.7
16.7
23.0
23.2
23.3
23.9
28.1
28.3
2S.3
28.4
29.9
30.6
30.6
31.2
34.7
33.1
33.2
33.6
394
39.6
39.1
40.4
I0O0
3J
IZ4
19.4
37.2
43.3
43.3
343
33J
33J
37.6
39.3
60.3
61.1
643
64.2
65.4
634
63.4
634
63.8
63.8
63.8
67.3
81.8
88.8
94.4
100.0
17.4
300
36.3
36.6
63.7
63.3
67.0
713
73J
774
7».7
84.6
83.3
86.1
86.8
89.0
89.1
90.3
91.1
96.0
97.3
98.3
100.0
30>MC9t lavVr IWPr^y Of (SOP).
12
Multiple-Spell Participants
Of the 28.7 million FSP participants in the full panel population, one fourth experienced more
than one spell of participation during the 28-month survey period. Figure 3 presents the
frequency distribution of the total number of months that the multiple-spell individuals
participated in the FSP. The median duration of the total number of months on the FSP was 17
months (out of a maximum of 28 months) for the multiple-spell participants, compared to 19
months for single-spell participants.
Using the cutoff values obtained from the analysis of the single-spell participants to distinguish
long-term participants from other participants, approximately 30 percent of the multiple-spell
participants would be classified as long-term participants if the lengths of their individual spells
were added together. Twenty-two percent would be short-term participants. In the case of the
multiple-spell participants who could be considered long-term, there is obviously little time
between each spell of participation, given that the survey period was 28 months.
Multiple-spell participants who had a short first spell were likely to have a second spell of short
duration. Specifically, we estimated that out of the multiple-spell participants whose first spell
lasted 8 months or less (i.e. short-term category), about two-thirds had a second spell of short
duration (8 months or less).
The results of the analysis of the participation patterns for single- and multiple-spell participants
were used to define four categories of participation for FSP participants:
13
Figure 3 Cumulative Distribution of the Total Number of Months Spent on the FSP for
Individuals who participated in the FSP for more than one spell during the 1987 SIPP
panel
100
CumuIo11VG Frequency for AI I SpeI Is
for Multipla SMI I Participant*
90 -
80 -
70 -
SO
40
30
80
•H) -
I I I I I I
2 3 4 S ■ 7 ■ S 10 11 12 13 14 15 11 17 10 19 3D 31 22 23 24 25 28 27
Laojth of AM SpaI la (amtNQ
Source: 1987 Survey of Income and Program Participation (SIPP).
14
(1) Short-term, single-spell participants (those persons who participated in the FSP
for eight months or less);
(2) Medium-term, single spell participants (those persons who participated in the FSP
for more than eight months, but less than 24 months);
(3) Long-term, single-spell participants (those persons who participated in the FSP
for at least 24 consecutive months); and
(4) Multiple-spell participants.
Using these four categories, participation patterns were analyzed for both the cross-sectional
cohort and the full panel population.
2.1 Cross-sectional Cohort
The length of each participation spell was computed for all individuals who participated in the
FSP in the first month of the 1987 SIPP Panel. The length of the participation spell which was
ongoing in the first month of the 1987 SIPP Panel was then adjusted using information available
in the Welfare History Module.9 Using this information reduces the problem of underestimating
'The 1987 SIPP Panel Welfare History Module contains explicit questions regarding the
initiation dare of the on-going participation spell. These questions were revised and improved
from earlier SIPP panels.
15
the duration of spells which is typically found in such panel surveys.I0 Only the duration of
the initial spell was adjusted; any information regarding prior participation in the program which
was not linked to the ongoing spell was ignored. Eleven percent of the cross-sectional cohort
did not report a history of FSP participation prior to month one of the 1987 Panel. For these
persons, it was assumed that the current spell of participation began on the first month of the
panel."
Table 1 presents estimates of the proportion of the cross-sectional cohort that falls into each
category of participation. The cross-sectional cohort was primarily composed of long-term (58.9
percent) and multiple-spell (26.6 percent) participants. Short-term participants represented only
6.7 percent of the population and medium-term participants made up the remaining 7.8 percent.
In other words, the vast majority ofparticipants on the FSP in thefirst month ofthe SIPP Panel
either were in the midst ofa spell that would last two years or would experience more than one
spell ofparticipation during the survey period.
Figure 4 presents the distribution of the estimated length of time that single-spell participants in
"This problem is typically referred to as "left censoring" because information prior to the
beginning of ths panel study is usually not available. The 1987 SIPP Panel Welfare History
Module contains explicit questions regarding the initiation date of the on-going spell. These
questions were revised and improved from earlier SIPP panels.
"Welfare history data was only available for the primary informant, and not for all
individuals in food stamp households. The welfare history of the primary informant was
imputed to all other household members except in the case of young children for whom the first
spell was computed to begin from their birth or from the reported beginning date, whichever was
later.
16
Table 1. Participation Patterns of Cross-Sectional Cohort
Number
('000)
Percentage
of Participants
Single Spell Participants
Short Term (1-8 months)
Medium Term (9-23 months)
Long Term (_> 24 months)
1,266.2
1,447.9
11,065.6
6.7
7.8
58.9
Multiple Spell Participants 5,004.8 26.6
All Participants 18,784.5 100.0
Average Monthly Participation* 19,175.0
Source: 1987 Survey of Income and Program Participation (SIPP) and Welfare History Module (Wave 2).
* administrative caseload data (first quarter of FY87)
n = 1,733 persons
n
Figure 4 Cumulative Distribution of the Length of Time in Months that an Individual is
on the FSP for the Single-Spell Participants in the Cross-Sectional Cohort
100
90 -
80
70 -
BO -
0 - ^■^ *W w ^^^
0 12 24384880729496 108120132144198188180192204218228240252284278
tonech of «juotad SMII cmtia)
•This estimate adjusts ate length of the ongoing spell at the beginning of the SIPP panel by using recipiency
histoiy data.
Source: 1987 Survey of Income and Program Participation (SIPP) and Welfare History Module (Wave 2).
18
the cross-sectional cohort received FSP benefits. About half participated in the FSP for almost
five years (median duration is 58 months), and 20 percent participated in the FSP for more than
10 years. As single-spell participants represent three-quarters of all participants in the cross-sectional
cohort, this figure illustrates why, at any point in time, a large portion of the food
stamp caseload has depended or will depend on food stamp benefits for a long time.
If the total duration of all spells were used to classify the multiple-spell individuals, 72.5 percent
of them would be classified as long-term, 25.3 percent as medium-term, and only 2.2 percent
as short-term participants. In fact, when combining single- and multiple-spell participants, 78
percent ofall individuals participating in the FSP in the first month of the SIPP panel were or
would become long-term participants.
Obviously, long-term participants dominate the cross-sectional cohort. Another way to examine
the extent of long-term participation is to examine participant-months. '2 The distribution of
total participant-months during the 1987 SIPP Panel for the cross-sectional cohort is presented
in Figure 5. Single-spell participant-months account for about three-fourths of the total
participant-months in the survey period, with long-term participants comprising the bulk of
these. Multiple-spell individuals account for 25 percent of the participant-months in the panel.
l2Looking at participant-months illustrates how long-term participants dominate over time.
Each month an individual participates in the FSP is equal to one participant-month. A short-term
participant, therefore, would be the equivalent of 1-8 participant-months. A long-term
participant would be the equivalent of 24 or more participant-months.
19
Figure 5. Share of Total Participation Months Throughout the 1987 SIPP Panel for
the Cross-Sectional Cohort, Using Welfare History Data.
Short Term (1.5%)
Long Teim (70.7%;
Multiple Spells (24.7%)
Medium Term (3.1%)
Source: 1987 Survey of Income and Program Participation (SIPP) and Welfare History Module (Wave 2).
20
2.2 Full Panel Population
We also examined participation patterns for all individuals who participated in the program for
at least one month between October 1986 and March 1989. Looking at the full panel population
allows closer examination of short-term and multiple-spell participation patterns, which may be
missed altogether when analyzing a cross-sectional cohort.
We computed spell lengths using data from each month of the 1987 SIPP Panel and did not use
the welfare history data for this group. This tends to underestimate the duration of spells which
were underway during the first month of the panel study as well as those that are on-going at
the end of the survey period.
We further classified participants with a single spell into short-term, medium-term, and long-term.
Table 2 presents the number and percentage of participants in the full panel population
by participation category. As in the cross-sectional cohort, long-term participants are the largest
category although the proportion (33.4 percent) is significantly smaller than that found in the
cross-sectional cohort. The short-term category represented 28.6 percent of the full panel
population compared to 6.7 percent of the cross-sectional cohort. This reflects the influence of
the large number of persons who move onto and off the FSP relatively quickly.
Participants who had more than one spell of participation during the survey period represented
approximately 25 percent of the full panel population (about the same proportion as in the cross-sectional
cohort). Approximately three out of four multiple-spell participants had two
21
Table 2. Participation Patterns of the Full Panel Population
Number
(•000)
Percentage
of Participants
Single Spell Participants
ro
Short Term (1-8 months)
Medium Term (9-23 months)
Long Term (>. 24 months)
8,200.6
3,557.0
9,572.5
28.6
12.4
33.4
Multiple Spell Participants 7,356.1 25.6
All Participants 28,686.2 100.0
Average Monthly Participation* 16,951.0
Average Monthly Participation** 18,845.3
Source: 1987 Survey of Income and Program Participation (SIPP).
* average monthly participation in 1987 SIPP panel
** administrative caseload data (October 1986 - March 1989)
n = 2,671 persons
&
participation spells during the 28-month survey period. Another 20 percent had three
participation spells. For all multiple-spell participants, the average length of the first spell was
seven months. The average length of the second spell was also seven months with a break of,
on average, 4 months between these two spells.13
Figure 6 shows that long-term participants comprised over half (56 percent) of the participant-months
in the full 28-month period of the SIPP Panel. Multiple-spell participants accounted for
26 percent, medium-term participants comprised 12 percent, and short-term participants
comprised 6 percent.
2.3 Censoring of Duration of Participation Spells
When analyzing data collected over a fixed time period, such as the 1987 SIPP, two types of
bias (or censoring) arise. The first is commonly referred to as "left censoring." This occurs
when participation spells that were initiated prior to the beginning of the survey period are
truncated due to the data collection process. Analyzing an entry cohort minimizes left censoring,
because only households with newly-initiated participation spells are included in the study
sample.
However, this study did not utilize an entry cohort, so some left-censoring did occur. For the
cross-sectional cohort, this censoring was minimized by using welfare history data to adjust the
,3It is possible that the four-month average elapsed time between spells is an anomaly of the
SIPP data collection methodology.
23
Figure 6. Share of Total Participation Months for Individuals in the Food Stamp
Program for the Full Panel Population.
Short Term (7.0»)
Multiple Spelb (25.9% Mediem Term (11.6ft)
Long Term (35.6ft)
Source: 1987 Survey of Income and Prognm Participation (SIPP).
24
length of spells initiated prior to the survey period. No adjustments were made for the full panel
population.
A second type of bias is "right censoring." This occurs when participation spells initiated near
the end of the survey period are truncated when data collection ends. Both the cross-sectional
cohort and the full panel population have some right-censored participation spells.
The result is that some participation spells are misclassified as short- or medium-term because
the full duration of the spell is not known. Table 3 shows the extent of right censoring for the
cross-sectional cohort and both right and left censoring for the full panel population. While
censoring was not a problem in the cross-sectional cohort, in the full panel population over half
of all short-term spells and over 80 percent of the medium-term spells were truncated due to the
data collection process14.
Even though some censoring did occur, especially among the full panel population, the clear
result is that a substantial portion of the food stamp population participates for more than 24
months. It is, therefore, important to determine who these long-term participants are and what
proportion of benefits they receive. The following sections attempt to do just that.
,4For more discussion on this topic, see Appendix A.
25
Table 3. Percentage of Participants with Probable Left & Right Censoring of Spell Duration
3
Cross-Sectional Cohort* Full Panel Population1M
% % %
Number
('000)
Right
Censored Number
('000)
Right
Censored
Left
Censored
Single Spell Participants
Short Term (1-8 months)
Medium Term (9-23 months)
Long Term C> 24 months)
1,266.2
1,447.9
11,065.6 74.2
8,200.6
3,557.0
9,572.4
17.2
34.3
93.0
38.6
48.9
92.7
Multiple Spell Participants 5,004.8 67.5 7,356.1 63.2 68.0
All Participants 18,784.5 61.7 28,686.2 56.4 65.5
Source: 1987 Survey of Income and Program Participation (SIPP) and Welfare History Module (Wave 2).
* n = 1,733 persons
** n = 2,671 persons
W
3.0 DISTRIBUTION OF FOOD STAMP BENEFITS BY PARTICIPATION
CATEGORY
The SIPP data were used to estimate the average monthly benefit and total benefits received by
the different participation categories for both the cross-sectional cohort and the full panel
population. Also shown is the share of total benefits received by each group of participants.
Food Stamp Program allotments are distributed at the household level; therefore, this portion
of the analysis looks primarily at household benefits.
3.1 Benefits Received by the Cross-sectional Cohort
The 18.8 million participants in the cross-sectional cohort represented approximately 6.6 million
food stamp households.13 Table 4 presents estimates of the average and median household food
stamp allotment for each group of participants. Short-term participants lived in food stamp
households that received the smallest monthly allotments, on average. They also had the
smallest average household size. Multiple-spell participants had the largest allotments and the
largest households, on average. On a per capita basis, the benefit levels for the short-term,
long-term and multiple-spell categories were comparable (around $42/month) whereas the per
capita benefit for the medium-term participants was slightly higher (around $47/month).
"For this analysis, the term "household" refers to a food stamp unit. In general, individuals
living in the same dwelling unit were considered to be in the same food stamp household.
However, for approximately IS percent of the SIPP dwelling units, the data indicated more than
one primary recipient for food stamps. In these cases, the dwelling units were split into multiple
food stamp households unless visual inspection of the data identified inconsequential multiple
food stamp units (i.e., if husband and wife alternated as the primary recipient).
27
Table 4. Cost per Household for the Cross-Sectional Cohort
oo
Number
C000)
Percentage
of Participants
Average
Household
Size
Average
Allotment
per Household*
Per
Capita
Allotment
Median
Allotment
per Household*
Single Spell Participants
Short Term (1-8 months)
Medium Term (9-23 months)
Long Term (>. 24 months)
1,266.2
1,447.9
11,065.6
6.7
7.8
58.9
2.5
2.9
2.8
$104.90
$138.33
$116.22
$41.96
$47.70
$41.51
$81.00
$121.00
$100.00
Multiple Spell Participants 5,004.8 26.6 3.3 $138.27 $41.90 $124.00
All Participants 18,784.5 100.0 2.9 $121.80 $42.00 $102.00
Source: 1987 Survey of Income and Program Participation (SIPP) and Welfare History Module (Wave 2).
* The benefit level is for the first month of the panel,
n = 1,733 persons
2X
Figure 7 presents the total benefits received by the cross-sectional cohort. This group received
$17.3 billion in benefits during the 28 months of the S1PP Panel, an average of about $619
million per month. Long-term participants received 72 percent of the benefits, multiple-spell
participants received 24 percent, medium-term participants received 3 percent, and short-term
participants received only 1 percent of total program benefits.
As expected, this is comparable to the distribution of participation months presented in Figure
5. However, both the distribution of benefits and the distribution of participant months vary
significantly from the distribution of total participants seen in Table 1. Although long-term
participants made up only 59percent ofthe cross-sectional population, they consumed 72 percent
oftotal benefits; when multiple-spell participants are added in, these two groups accountfor 96
percent of all benefits. In contrast, although short-term participants made up about seven
percent ofthe population, they only received one percent ofthe benefits.
3.2 Benefits Received by the Full Panel Population
Table 5 presents estimates of the average and median food stamp allotment received by full panel
population. Again, short-term participants lived in households that received the smallest monthly
allotments, on average and had the smallest average household size. Multiple-spell participants
had the largest allotments and the largest households, on average. On a per capita basis, there
is little difference between the various categories of participants (each received approximately
$42/month).
29
Figure 7. Total Food Stamp Program Benefits Received by the Cross-Sectional
Cohort (Millions of dollars).
Short torn <1.2ft) 1 »»•"■» I
MaMpbT«n» (24.1ft
Mediam term (3.1 ft)-ftS3Xt Mn
Lou, term (71.7ft)
Total Food Stamp Program Benefits Received: $17333.6 Miffion.
Source: 1987 Survey of Income and Program Participation (SIPP) and Wdfiue History Moduk (Wave 2).
30
Table 5. Cost per Household for the Full Panel Population
CO
/•
Number
(•000)
Percentage
of Participants
Average
Household
Size
Average
Allotment
per Household*
Per
Capita
Allotment
Median
Allotment
per Household*
Single Spell Participants
Short Term (1-8 months)
Medium Term (9-23 months)
Long Term (>. 24 months)
8,200.6
3,557.0
9,572.5
28.6
12.4
33.4
2.9
2.6
2.8
$119.41
$111.37
$120.05
$41.18
$42.83
$42.87
$104.00
$98.00
$107.31
Multiple Spell Participants 7,356.1 25.6 3.3 $139.69 $42.33 $133.82
All Participants 28,686.2 100.0 2.9 $123.26 $42.50 $110.53
Source: 1987 Survey of Income and Program Participation (SIPP).
* Average benefit level for the months on the Food Stamp Program in the SIPP panel.
n = 2,671 persons
3/
Figure 8 presents the distribution of total benefits received by the full panel population over the
28 months of the survey. Long-term participants received 57 percent of total program benefits,
multiple-spell participants received 25 percent, medium-term participants received 12 percent,
and short-term participants received 6 percent. Total benefits during the survey period were
$21.6 billion or about $771 million per month on average. Long-term participants received
$12.3 billion during the survey period, while the other participants combined received only $9.3
billion.
As expected, the distribution of benefits is comparable to the distribution of participation months
presented in Figure 6. However, both the distribution of benefits and the distribution of
participant months vary significantly from the distribution of total participants seen in Table 2.
Although long-term participants made up only 33 percent of the full panel population, they
consumed 57 percent oftotal benefits; when multiple spell participants are added in, these two
groups accountfor 82 percent ofall benefits. In contrast, although short-term participants made
up about 29 percent of the population, they only received six percent of the benefits.
32
Figure 8. Total Food Stamp Program Benefits Received by Participants in the 1987
SIPP Panel (Millions of dollars).
Short Term (6.2%)
Long Term (56.9%)
Medium Term (11.8%)
Multiple Spells (25.1%)
Total Food Stamp Program Benefits Received: $21,574.8 Million.
Source: 1987 Survey of Income and Program Participation (SIPP).
33
4.0 CHARACTERISTICS OF LONG-TERM PARTICIPANTS
In addition to describing each participant group in terms of their size and the proportion of
benefits they receive, this study set out to describe the characteristics shared by participants in
these distinct subgroups to determine if those who are dependent on food stamp over time differ
systematically from those who leave the program quickly and never return. The following
sections describe individual and household characteristics of long-term and multiple-spell
participants and compare these characteristics to those found among other FSP participants.16
Some socioeconomic characteristics (for example, race and gender) are relatively stable across
time. However, other variables, such as age or household income, may change from month to
month or year to year. Unless otherwise noted, characteristics presented here were obtained for
the first month of FSP participation.
4.1 Demographic and Household Characteristics of FSP Participants
Table 6 presents the age distribution for the full panel population. Long-term participants were
predominantly children, as was the case for the other participation groups. However, although
all participation groups contained a significant number of children, children were most likely to
be long-term or multiple-spell participants (36 and 27 percent, respectively). None of the groups
l6We analyzed the characteristics of long-term participants in both the cross-sectional cohort
and the full panel population. As the results did not differ greatly between the two populations,
we present only the results for the full panel population here.
34
Table 6. Distribution of Age for the Full Panel Population
Total
Number of
Participants
Percentage of Participants in each age group* and participation category
Under 18
years
% of % of
parti- age
cipants group
18-34
years
%of
parti-cipants
%of
age
group
35-59
years
%of
parti-cipants
%of
age
group
over 59
years
% of % of
parti- age
cipants group
CO en Single Spell Participants
Short Term (1-8 months)
Medium Term (9-23 months)
Long Term t> 24 months)
Multiple Spell Participants
All Participants
8,200.6
3,557.0
9,572.5
41.5
45.2
52.1
24.8
11.7
36.2
32.9
25.8
18.3
35.6
12.1
23.1
19.7
19.7
17.2
32.0
13.9
32.6
5.9
9.2
12.4
21.0
14.2
51.7
7,356.1 51.1 27.3 30.1 29.2 14.7 21.5 4.1 13.1
28,686.2 47.9 100.0 26.4 100.0 17.6 100.0 8.0 100.0
Source: 1987 Survey of Income and Program Participation (SIPP).
* Age is as given in the first month in which the person participated in the FSP.
n = 2,671 persons
5r
had a significant number of elderly individuals, but the elderly were most likely to be long-term
participants (over half participated for two years or longer).
White non-Hispanics made up 45 percent of the full panel population; African Americans made
up 33 percent and Hispanics accounted for 16 percent (Table 7). Long-term participants were
about equally split between whites and African Americans; however, African Americans were
more likely to be long-term or multiple-spell participants (65 percent of all African Americans
fell into one of these two groups). Hispanics also were likely to be long-term participants,
whereas white participants were more likely to participate for eight months or less.
Table 8 presents educational attainment information for the full panel population. In general,
short-term participants had higher educational levels than long-term participants. While over
60 percent of short-term participants had completed high school, nearly 70 percent of long-term
participants had not. In addition, short-term participants were three times as likely to have some
post-secondary education. Nearly half of all participants with post-secondary education were
short-term participants. The general conclusion that can be made is that the probability of
participating for more than two years falls as education increases. Conversely, the probability
ofparticipating for eight months or less rises as education increases.
Tables 9 and 10 present FSP household composition by participation categories.17 As seen in
,7When describing household characteristics, it is conceivable that individual household
members may have different participation patterns. In all cases, household classifications of
(continued...)
36
Table 7. Race or Ethnic Origin of the Full Panel Population
Total
Number of
Participants
African-
White American
Non-Hispanics Non-Hispanics
Hispanics Other"
•^1
% of % of
parti- White
cipants Non-
Hispanics
%of *of *of %of
parti- African parti- Hispanics
cipants American cipants
Non-Hispanics
% of % of
parti- Other
cipants
Single Spell Participants
Short Term (1-8 months)
Medium Term (9-23 months)
Long Term C> 24 months)
Multiple Spell Participants
All Participants
8,200.6 55.2 34.8 26.7 22.8 13.0 23.4 5.1 27.5
3,557.0 48.4 13.2 28.9 10.7 17.3 13.5 5.4 12.8
9,572.5 37.1 27.3 38.1 38.0 19.1 40.1 5.7 35.6
7,356.1 43.7 24.7 37.1 28.5 14.2 23.0 5.0 24.1
28,686.2 45.4 100.0 33.4 100.0 15.9 100.0 5.3 100.0
Source: 1987 Survey of Income and Program Participation (SIPP).
* Other category includes American Indians, Eskimos, Aleuts, Asians or Pacific Islanders.
n = 2,671 persons
37
Table 8. Last Year of Education Completed by Individuals 18 years or over in the Full Panel Population
Percentage of Participants Ages 18 years or More
Participants
18 years or more
(•000)
Number (%)
Attended
Grades
1-8
%of %of
Ptrtici- Educa-pation
tion
Category Level
Attended
Grades
9-11
%of %of
Partici- Educa-pation
tion
Category Level
Completed
High
School
%oF %"of
Partici- Educa-patioo
tion
Category Level
Post-
Secondary
Education
%"of iTof
Partici- Educa-pation
tion
Category Level
CO
00 Single Spell Participants
Short Term (1-8 months)
Medium Term (9-23 months)
Long Term [>_ 24 months)
4,638.5
1,939.9
4,589.2
31.5
13.2
31.2
14.3
27.3
36.0
17.9
14.3
44.6
24.9
26.0
31.3
27.8
12.1
34.5
40.5
29.8
25.8
39.2
12.2
24.7
20.3
16.9
6.9
45.4
15.8
15.3
Multiple Spell Participants 3,549.4 24.1 24.1 23.2 30.0 25.6 32.3 23.9 13.8 23.6
All Participants 14,717.0 100.0 25.2 100.0 28.2 100.0 32.5 100.0 14.1 100.C
Source: 1987 Survey of Income and Program Participation (SIPP).
n = 2,671 persons
3<f
Table 9, nearly one-third of all FSP households consisted of a single adult participant.18 About
half of these single adults were elderly, although among long-term participants nearly three-quarters
were elderly. Another 37 percent of FSP households were composed of single parents
and their children, and the remainder were two-adult households, with and without children (25
and 9 percent, respectively).
The majority (nearly 60 percent) of single-elderly households were on the program for 24
months or longer. Likewise, single-parent households were likely to be highly dependent on
food stamps, either as long-term or multiple-spell participants (over two-thirds of all single-parent
households fell into one of these categories). In other words, the most vulnerable among
the poverty population tended to participate for two years or longer or to go on and off the
program repeatedly.
Table 10 shows the percentage of households with children by the children's age. Nearly two-thirds
of all households contained children, with an average of about two children per household.
About one-third "f all households with children had only preschool-age children; another third
had only school-age children. Long-term and multiple-spell households tended to have more
children than short- or medium-term households.
I7(...continued)
multiple-spell or short-, medium-, or long-term are based on the participation patterns of the FSP
household head.
'"Single adult participants did not necessarily live alone; rather, they were the only members
of their dwelling units who received food stamps.
39
Table 9. Household Composition for the Full Panel Population
PERCENTAGE OF HOUSEHOLDS WITH:
Total
Number of
Households
('000)
Single Adult
Alone
Two Adults
with No
Children
Single Adult
with children
under 18 years
of age
Two Adults
with children
under 18 years
of age
< 59 years 60+ ye«ri
1 of % of % of % of
participant persons participant persons
category < 59 category 60+
%of
Parti-cipant
Category
%of
House-hold
type
*of
Parti-cipant
Category
«of
House-hold
type
%of
Parti-cipant
Category
*of
House-hold
type
O
Household head is:
Single Spell Participant
Short Term (1-8 months)
Medium Term (9-23 months)
Long Term t> 24 months)
2,952.7
1,311.6
3,601.4
18.6
21.1
7.5
40.1
20.2
19.8
9.4
14.4
24.4
18.7
12.7
59.3
9.8
10.6
9.3
32.2
15.5
37.3
28.6
24.1
41.8
22.4
8.4
39.8
32.2
29.8
16.9
38.0
15.6
24.3
Multiple Spell Participant 2,235.8 12.2 19.9 6.1 9.3 6.5 16.2 40.2 23.8 35.0 31.2
All Households 10,101.5 14.4 100.0 14.2 100.0 8.9 100.0 37.4 100.0 24.8 100.0
Source: 1987 Survey of Income and Program Participation (SIPP).
* Age of persons in the household is as given in the first month of the survey.
n = 1,433 households (2,671 persons)
qo
Table 10. Households with Children Under Age 18 years for the Full Panel Population
Total Percent of
Number of Percent of Percent of households with
Households Average Households with Households with both preschoolers
with children Number preschoolers only children ages and schoolage
(•000) of Children (< 6 years old) 6-18 only children
*of *of %of
Houie- Houae- Houte-
Sof holdi % of holdi % of holdi with
partici- with pre- partici- with partici- both pre-pation
schoolers pation children palion Khoolera
Category only Category 6-18 Category & achoolage
Household head is:
Single Spell Participant
Short Term (1-8 months)
Medium Term (9-23 months)
Long Term Q> 24 months)
1,662.1
637.1
2,073.9
1.9
1.8
2.3
34.6
55.1
30.2
27.5
16.8
29.9
40.1
26.1
38.3
30.3
7.5
36.1
25.3
18.8
31.5
25.7
7.4
40.1
Multiple Spell Participant 1,553.9 2.2 34.6 25.8 37.1 26.1 28.3 26.8
Households with Children 5,927.0 2.1 35.2 100.0 37.3 100.0 27.5 100.0
Source: 1987 Survey of Income and Program Particir/ation (SIPP).
* Age of persons in the household is as given at the first month of the survey.
n = 588 households with children
til
4.2 Household Income
Tables 11 and 12 present household income for the full panel population by participation
categories. For each FSP household in the SIPP Panel the monthly value of household income
was compared to the poverty line for a given household size in order to calculate where the
household fell relative to the poverty line in each month.
Household income as a percentage of poverty is presented for the full panel population in Table
11. Long-term participants lived in households with the smallest monthly incomes ($617, on
average), while short-term participants lived in households with higher incomes ($965, on
average). Ninety percent of long-term participants lived in households whose incomes were
below the poverty line in their first month on the FSP. In contrast, only 58 percent of short-term
participants lived in such households. One-fifth of short-term participants lived in
households whose incomes in the first month of FSP participation were above 130 percent of the
poverty level, compared to 12 percent for all FSP participants". Forty percent of all
participants with incomes below poverty in the first month were long-term participants. Nearly
half of all participants with incomes above poverty in the first month were short-term
participants. Clearly, the probability ofbeing a long-term participant increases as incomefalls;
conversely, the probability ofparticipating for a short time increases as income rises.
"Previous research (Martini, 1992) has shown that some households that report participating
in the program appear to be ineligible according to the income and assets information they
provide during the SIPP interview. This could result from underreporting income when applying
for benefits, misreporting FSP participation in the SIPP survey, or other reasons.
42
Table 11. Distribution of Household Income* for the Full Panel Population
Total
Number of
Participants
Average
Monthly
Income
< 100%
Poverty Level
100-130%
Poverty Level
Over 130 % of
Poverty Level
% of % of
participant income
category group
% of % of
participant income
category group
% of % of
participant income
category group
CO
Single Spell Participants
Short Term (1-8 months)
Medium Term (9-23 months)
Long Term (>. 24 months)
Multiple Spell Participants
All Participants
7,509.0
3,425.2
9,538.0
$964.87
$701.39
$617.47
57.8
78.0
89.5
20.7
12.7
40.7
20.8
9.8
4.8
46.5
10.0
13.7
21.4
12.2
5.6
48.7
12.7
16.4
7,164.2 $748.54 75.8 25.9 14.0 29.8 10.2 22.2
27,636.4 $760.80 75.9 100.0 12.2 100.0 11.9 100.0
Source: 1987 Survey of Income and Program Participation (SIPP).
* Household income is as reported for the first month in which the household participated in the FSP.
n = 2,671 persons
ctf
Table 12 displays income dynamics for the full panel population throughout the 28 months of
the survey. The majority of all participants (64 percent) lived in household with income that
was below the poverty level for some, but not all, months of the survey. However, there were
some striking differences between long- and ion-term participants and the rest of the full panel
population.
While only 28 percent of the full panel population was poor in every month of the survey, the
majority (71 percent) of these individuals were long-term participants. Nearly 60 percent of
long-term participants lived in households whose incomes were below the poverty level in all
28 months of the survey. In contrast, only four percent of short-term participants lived in poor
households in every month. Short-term participants made up nearly 60 percent of the
participants who were never poor. Multiple-spell participants, as expected, were most likely to
drift in and out of poverty (76 percent).
Clearly, long-term participants differ from other groups with respect to the amount of income
they have available. About 70 percent ofthe FSP households with incomes always at or below
poverty were long-term participants; nearly 60percent ofhouseholds with incomes above poverty
in all months were short-term participants. In other words, long-term participants could be
described as chronically poor.
Table 13 presents income sources for participant households to explore the question of whether
their sources of income differ as well. The table also illustrates what portion, on average, of
44
Table 12. The Dynamics of Poverty for Food Stamp Participants in the Full Panel Population
Total
Participants
C000)
Always At or
Below Poverty Level
Sometimes Below1
Poverty Line
Always above
Poverty Level2
%of
Participation
Category
% of
Income
Level
%of
Participation
Category
% of
Income
Level
ft of
Participation
Category
ft Of
Income
Level
Single Spell Participants
Short Term (1-8 months)
5 Medium Term (9-23 months)
Long Term t> 24 months)
8,153.7
3,557.0
9,572.5
4.1
18.2
59.3
4.2
8.0
70.8
78.9
72.9
37.4
35.3
14.2
19.7
17.0
8.9
3.4
57.5
13.1
13.5
Multiple Spell Participants 7,356.1 18.6 17.0 76.2 30.8 5.2 15.9
All Participants 28,639.3 28.0 100.0 63.6 100.0 8.4 100.0
Source: 1987 Survey of Income and Program Participation (SIPP).
n = 2,671 persons
1 Below 100% of poverty at least one month, but not all months of survey.
2 Income is above 100 % of the poverty level for all months of the panel.
(4$
Table 13. Sources of Income for the Full Panel Population
Total
Number of
Participants
Percent of Household Income* Accounted for by:
Earned
Income
%of
parti-cipants
%of
earned
income
Transfer
Income
%of
parti-cipants
%of
transfer
income
Property
Income**
%of
parti-cipants
%of
property
income
Other
Income***
% of % of
parti- other
cipants income
Single Spell Participants
Short Term (1-8 months)
Medium Term (9-23 months)
Long Term (J> 24 months)
7,509.0
3,425.2
9,538.0
49.2
27.3
14.6
42.0
10.6
15.8
20.7
41.8
62.7
13.6
12.5
52.1
1.6
0.2
0.1
69.7
4.0
5.5
28.5
30.7
22.5
29.7
14.6
29.8
Multiple Spell Participants 7,164.2 38.6 31.5 34.9 21.8 0.5 20.8 26.0 25.9
All Participants 27,636.4 31.8 100.0 41.5 100.0 0.6 100.0 26.1 100.0
Source: 1987 Survey of Income and Program Participation (SIPP).
* Household income is as given in the first month in which the household participated in the FSP.
** Transfer Income includes income from means-tested government programs such as Food Stamps, AFDC, General Assistance and other
forms of public assistance.
*** Other Income includes pensions, Social Security, Railroad Retirement, unemployment compensation, disability payments, alimony, child
support, and income from savings.
n = 2,671 persons
Uifi
total household income came from these sources.20 For long-term participants, the majority
of household income came from means-tested transfers (63 percent of all household income,
compared to 42 percent for the full panel population in general); only IS percent was from
earnings. On the other hand, half of short-term participants' income was earned and only 20
percent was from transfers. Multiple-spell participants were more diverse in terms of their
income sources: slightly more than one-third was from earnings, another one-third was from
transfers, and one-quarter was from other sources, such as unemployment compensation.
Table 14 provides information on participation in other assistance programs. About one-third
of all FSP participants also participated in the AFDC program and about 6 percent received
Supplemental Security Income (SSI). However, participation in these programs varied greatly
by FSP participation category. While 53 percent of long-term participants received AFDC, only
17 percent of short-term participants did so. Similarly, about 12 percent of long-term
participants received SSI, compared to four percent among short-term participants and three
percent among multiple-spell participants. This is not surprising given that long-term
participants were most likely to be single-parents and their children or elderly individuals.
Only 24 percent of all FSP participants over age 15 were employed in their first month on the
FSP, and 59 percent were not in the labor force, i.e., they were not working, laid off, or
"Although the share of income from different sources varied somewhat from month-to-month,
the differences across participation categories remained relatively stable. The average
share of income from various sources throughout the survey period also showed similar
differences across participation categories. Table 14 describes income sources in the first month
of FSP participation.
47
@MfflP«
Table 14. Participation in Other Assistance Programs* for the Full Panel Population
£
Total
Participants
('000)
Received
AFDC
Received
SSI
Received
General Assistance
Received
WIC
%of
Participant
Category
%
AFDC
Participants
%of %
Participant SSI
Category Recipients
%of
Participant
Category
%
General
Assistance
%of
Participant
Category
%
WIC
Participants
Single Spell Participants
Short Term (1-8 months)
Medium Term (9-23 months)
Long Term (>. 24 months)
8,200.6
3,557.0
9,572.5
16.8
33.8
52.7
13.9
12.1
50.7
4.0
5.2
11.7
17.7
10.1
60.4
4.9
10.5
8.8
21.3
19.8
44.7
3.4
6.2
4.0
20.4
16.0
27.8
Multiple Spell Participants 7,356.1 31.5 23.3 2.9 11.8 3.6 14.2 6.7 35.8
All Participants 28,686.2 34.7 100.0 6.4 100.0 6.6 100.0 4.8 100.0
Source: 1987 Survey of Income and Program Participation (SIPP).
* participation in the month in which the person first participated in the FSP; participation in assistance programs is not mutually exclusive,
i.e. a person could receive benefits from both AFDC and WIC.
n = 2,671 persons
W
searching for work (Table 15). However, short-term participants were nearly four times as
likely as long-term participants to be employed. Multiple-spell participants were also more likely
to be working.
In general, these characteristics paint a picture of long-term and multiple-spell participants as
the most disadvantaged among the poverty population. They have the least education, the least
available income, and the fewest ties to the labor force.
49
Table 15. Labor Force Status* of Food Stamp Participants Over 15 years of Age for the Full Panel
Total
Participants Percent
over 15 years old Percent Percent Not in
('000) Employed Unemployed** Labor Force***
Single Spell Participants
ion
Short Term (1-8 months)
Medium Term (9-23 months)
Long Term (>. 24 months)
5,297.5
2,207.1
5,260.3
38.1
14.9
10.4
18.3
16.2
12.9
43.6
68.8
76.7
Multiple Spell Participants 4,244.1 26.8 22.3 50.9
All Participants 17,009.0 23.7 17.4 58.9
Source: 1987 Survey of Income and Program Participation (SIPP).
* Labor Force Participation is as given in the first month in which the person first participated in the Food Stamp Program.
** Unemployed include persons on layoff or looking for a job.
*** Not in the Labor Force includes persons without a job, not on layoff, and not looking for work.
n = 2,671 persons
sd
5.0 FSP WORK REGISTRANTS
A final question to be addressed is whether there is a distinct subgroup of long-term participants
who may be subject to work registration requirements. Work registrants are more likely to be
employable than other participants. Therefore, it is important to know if some portion of the
long-term participant population is work-ready and could be targeted for employment and
training services. To determine if this is the case, we simulated the FSP work registration
requirements to determine if there were significant differences in the proportion of work
registrants for the various participation patterns.
5.1 Distribution of Work Registrants by Participation Category
Potential work registrants were identified in the SIPP data by simulating the FSP exemption
criteria. Any FSP participant who was under age 18, over age 59, disabled, employed for more
than 30 hours per week, a full-time student, a caretaker of a child under six, or a participant in
AFDC was classified as exempt under work registration requirements.2I These exemption
categories are not mutually exclusive; i.e., individuals could be exempt for multiple reasons.
Table 16 presents the exemption categories and the percentage of work registrants for the cross-sectional
cohort and compares these data to administrative data from the same time period. Only
6.3 percent of the participant population would have been required to register for work in the
2IAFDC-FSP participants who participated in the Work Incentive Program (WIN) were
exempt from FSP work registration requirements. As SIPP does not contain information on
WIN participation, receipt of AFDC was used as a proxy when determining the work
registration status of FSP participants.
51
Table 16. Distribution of Participants by Work Registration Status for Cross-Sectional Cohort
Short Medium Long Multiple Total Administrative
Term Term Term Spells (%) Data*
(%) (ft) (%) (%)
Required to register for Work 10.0 10.6 4.0 9.1 6.3 7.0
Less than 18 years 43.6 47.3 52.0 55.2 51.8 49.1
Disabled and/or Elderly 17.8 16.2 25.0 14.4 21.0 13.7
AFDC adult recipient (proxy for WIN) 3.3 5.7 11.7 7.7 9.6 9.1
Caretakers of Children under 6 10.1 9.4 3.9 7.4 5.6 12.0
Employed more than 30 hrs/wk 13.8 9.4 2.7 6.0 5.0 4.3
Students 1.4 1.4 0.7 0.3 0.7 2.5
Other or unknown 2.3
Total Exempt Participants ** 90.0 89.4 96.0 90.9 93.7 91.2
All Participants 100.0 100.0 100.0 100.0 100.0 100.0
Total Number of Participants ('000) 1,266.2 1,447.9 11,065.6 5,004.8 18,784.5 -
Source: 1987 Survey of Income and Program Participation (SIPP) and Welfare History Module (Wave 2).
* Source for Administrative data is the Summer 1988 Food Stamp Quality Control Sample.
** The exempt categories include persons who would be exempt for more than one reason, e.g. disabled and elderly.
n = 1,733 persons
SI-
first month of the panel. Given the characteristics of long-term and multiple-spell participants,
it is no surprise they were the least likely to be work registrants; only four percent of long-term
participants and nine percent of multiple-spell participants would have been work registrants.
About half of all participants were exempt from registration because they were too young. Long-term
and multiple-spell participants were most likely to be exempt for this reason. One-fifth
were exempt because they were elderly or disabled; again, this was especially true of long-term
participants. Short-term participants were the most likely to be exempt because they were
working or caring for young children. The differences in reasons for exemption are consistent
with the characteristics data in Section 4. Table 9 showed that long-term and multiple-spell
participants tended to live in households with children or, in the case of long-term participants,
to be elderly. Table 13 indicated that short-term participants were most likely to have some
earned income.
As expected, the distribution of work registrants by participation pattern (Table 17) differs from
the distribution of the total population as seen in Tables 1 and 2. These differences occur
because short-term participants are more likely, and long-term participants are less likely, to be
work registrants. However, although a small percentage of long-term participants were work
registrants, they still represented a substantial portion of the work registrant population. In the
cross-sectional cohort, slightly more than a third of the work registrants were long-term
participants (Table 18). In the full panel population, nearly 20 percent participated for at least
24 consecutive months. When multiple-spell participants are added in, these percentages
53
Table 17. Distribution of Work Registants by Participation Category
Cross-Sectional Cohort* Full Panel Population**
Number
CO00)
%
of Work
Registrants
Number
(•000)
%
of Work
Registrants
Single Spell Participants
Short Term (1-8 months)
Medium Term (9-23 month)
Long Term (_>_ 24 months)
142.2
153.5
440.7
11.9
12.9
37.0
745.5
208.2
370.7
37.7
10.5
18.7
Multiple Spell Participants 456.0 38.2 652.4 33.0
Short and Medium Term
Long Term
135.1
320.9
11.3
26.9
137.5
514.9
7.0
26.0
All Participants Required
to Register for Work 1,192.4 (6.3%) 100.0 1,976.8 (6.8%) 100.0
Exempt from Work
Registration
17,592.1 — 26,709.4 —
Source: 1987 Survey of Income and Program Participation (SIPP) and Welfare History Module (Wave 2).
* n = 1,733 persons
** n = 2,671 persons
M
increase to 75 and 51, respectively. In other words, in any given month, three quarters ofall
work registrants are in the midst of a spell that will last two years or longer or are likely to
return to the program after leaving.
5.2 Characteristics of Work Registrants
Tables 18 through 20 describe the education and employment history of work registrants in the
full panel population. As shown in Table 18, less than half of the participants required to
register for work had a high school diploma or the equivalent GED. Short-term participants
were most likely to have a high school education, as was true for the FSP population in general
(Table 9).
Many work registrants had already received some type of employability training. Table 19
shows that about one-fifth of the work registrants had received training through CETA, JTPA,
the Veterans Administration (VA), or some other source. About half of these individuals
received training through the VA. The majority of registrants received skills training and basic
education services.
Data from the employment history module was used to analyze patterns of labor force
participation for individuals required to register for work. Table 20 shows that two-thirds of the
work registrants had not worked in the past year and over half had not worked in three years.
Approximately 21 percent had never been in the labor force. Of these participants, the principal
reason given for never entering the labor force was care of home and family (48 percent of those
55
in
Table 18. Percentage of Work Registrants with High School Diploma or GED, by Participation Category
Percent
Number with High School
('000) Diploma or GED
Work Registrants
Short Term (1-8 months) 883.0 54.5
Medium Term and Long Term
(> 9 months)* 1,093.8 33.8
All Work Registrants 1,976.8 43.0
All Food Stamp Participants
18 Years of Age or Over 14,717.0 52.5
Source: 1987 Survey of Income and Program Participation (SIPP) and Education and Training Module (Wave 2).
* Combined due to small sample size
n = 239 persons
Jh
Table 19. Receipt of Training Services by Work Registrants
Number Percent Percent Percent Percent Percent
of Who Receiving Receiving Receiving Receiving
Work Received JTPA CETA Veterans Other
Registrants Training Training Training Training Training
Work Registrants:
Short Term (1-8 months) 883.0 19.6 4.4 2.3 10.6 2.3
Medium and Long Term
(9-28 months) 1,093.8 23.8 7.3 4.3 11.4 0.9
All Work Registrants 1,976.8 21.8 5.9 3.4 11.0 1.6
Source: 1987 Survey of Income and Program Participation (SIPP) and Education and Training Module (Wave 2).
n = 239 persons
J7
B.l
s
Table 20. Employment History of Work Registrants
Number
('000)
Percent
Employed
in
Last Year
Percent
Employed
in Last
Three Years
Percent
Never
Worked
Work Registrants
Short Term (1-8 months) 883.0 31.2 37.7 14.0
Medium and Long Term
t> 9 months)* 1,093.8 34.8 42.5 25.9
All Work Registrants 1,976.8 33.2 40.4 20.6
Source: 1987 Survey of Income and Program Participation (SIPP) and Employment History Module (Wave 2).
* Combined due to small sample size
n = 239 persons
St
who never worked). Sixteen percent were students and the remaining 36 percent stated that they
either could not find work, didn't want to work or had another reason.
Table 21 shows the reason given for leaving their last job, as reported by work registrants who
had left a recent job. Approximately one-third gave layoff or temporary job as the reason for
leaving their last job. Twenty percent stated family reasons such as pregnancy, health reasons
or other family reasons.
5.3 Characteristics of Long-term Work Registrants
The characteristics of long-term participants who were not exempt from work registration are
presented in Table 22.22 As seen in the table, the majority of long-term work registrants lived
in households with children (57 percent). The vast majority of these households (87 percent)
were not single-parent households. Households with children on average contained 3.1 children.
One-fifth of the long-term work registrants were single adults living alone, and an additional 22
percent lived in households of two or more adults without children. The average household size
across all long-term work registrants was 4.2 persons, and the average monthly allotment was
$130. This average benefit is about $10 higher than the average household benefit for all long-
22The sample size for v ork registrants in the SIPP Panel was small (239 individuals). In
order to examine the characteristics of long-term work registrants, multiple spell participants
who were on the FSP for a total of 24 months or longer were combined with the single-spell
long-term participants.
59
Table 21. Reasons for Leaving their last job given by Work Registrants
Reason:
Number %
C000)
309.4 23.7
126.1 9.6
164.7 12.6
95.9 7.3
19.4 1.5
161.2 12.3
54.6 4.2
35.4 2.7
56.7 4.3
68.7 5.3
142.7 10.9
73.6 5.6
Layoff/plant closed
Discharged
Temporary Job only
Found a better job
Retired
Did not like work or location
Dissatisfied with earnings
School
Pregnant/had Child
Health Reasons
Other family reason
Other reason
Total 1,308.4 100.0
Siurce: 1987 Survey of Income and Program Participation (SIPP) and Employment
History Module (Wave 2).
n = 239 persons
60
Table 22. Characteristics of Long-Term Work Registrants
Individual Characteristics:
Average Age
Average Years of Schooling
Average FSP Spell Length
% Receiving General Assistance
% Female
% African American
Household Characteristics;
Households with Children
Pre-schoolers only
School-age children only
Both preschoolers and school-age children
Single Parent Households
Multiple Adult Households with kids
Single Adult only
Multiple Adults, no kids
Average Household Size
Average # of Children
Average Monthly Food Stamp Benefit
Administrative
SIPP Data Data*
34.0 34.4
8.9 —
27.2 months —
19.2 16.5
58.4 51.1
53.7 32.2
56.9% 55.8 %
16.6 % ** 17.3 % **
56.7 % ** 59.6 % **
26.7 % ** 22.8 % **
7.5 % 13.6%
49.5 % 41.8 %
20.7 % 28.6 %
22.4% 15.6 %
4.2 persons 3.0 persons
3.1 ** 2.2**
$130.00 $151.00
Source: 1987 Survey of Income and Program Participation (SIPP).
* Source: Summer 1987 Food Stamp Quality Control Sample. Administrative data is for all work registrants,
not just long-term registrants.
** Average computed for households with children
n = 42 persons
61
term participants in the full panel population.
The average age of the long-term work registrants was 34. The majority (80 percent) did not
finish high school. Approximately one-fifth received general assistance at some time during the
28-month panel. Their average duration on the food stamp program was 27.2 months (out of
a possible 28 months). More than half (58 percent) were female and a similar proportion were
African American (54 percent). These characteristics are not significantly different from the
characteristics of work registrants in general, as seen in administrative data from that time
period.
62
6.0 CONCLUSIONS
The U.S. Department of Agriculture and other Federal agencies have sponsored a great deal of
research on the dynamics and determinants of participation in the FSP and other assistance
programs. From this research, three distinct groups of participants emerge - those who rely
on assistance to get them over short periods of financial difficulty, those who depend on
assistance continually for a significant part of their lives, and those who receive assistance
sporadically throughout their lives, but still rely on assistance for a substantial amount of time.
While the existence of these distinct groups is accepted, little is known about their size or
characteristics. It is likely that the characteristics of these groups vary considerably and that
these differences may provide an understanding of why some people receive assistance for short
periods of time while others seem unable to become self-sufficient.
This study uses the 1987 Survey of Income and Program Participation (SIPP) to describe time
patterns of Food Stamp Program (FSP) participation. Participants who had only one
participation spell during the survey period were classified as either short-term (on the program
8 months or less), medium-term (on the program 9-23 months) or long-term participants (on the
program 24 months or longer). Persons who were on and off the FSP during the survey were
classified as multiple-spell participants. Each group is described in terms of the proportion of
the FSP they represent, the share of total benefits they receive, and the characteristics they share
as distinct groups.
63
There are several alternative ways to measure time patters on FSP participation. These methods
include following an entry cohort for a fixed period of time, examining participation patterns of
all individuals who participate in a given month, and examining the participation patterns of all
individuals during a given reference period. Each method of analysis presents a very different
picture of FSP participation patterns.
This study looked at both a cross-section of participants who were on the FSP in the first month
of the survey (the cross-sectional cohort) and at all individuals who participated in the FSP for
at least one month of the October 1986 through March 1989 survey period (the full panel
population). From either perspective, a substantial proportion of FSP participants in the SIPP
Panel were long-term participants (59 percent of the cross-sectional cohort and 33 percent of the
full panel population). In fact, when taking multiple-spell participants into account, of the
estimated 18.8 million persons who participated in the FSP at the beginning of the survey
period, 11.6 million persons (62 percent) were also on the program at the end of the panel
period 28 months later.
Table 23 provides a comparison of the distribution of the full panel and cross-sectional cohort
populations when using alternative measures of participation (percent of participants, percent of
participant-months, and percent of benefits). From this table it is apparent that the groups most
dependent on assistance over time (long-term and multiple-spell participants) consumed a
disproportionate share of program benefits. Although long-term participants made up only 59
percent of the cross-sectional population, they consumed 72 percent of the benefits. Multiple-
64
Table 23. Distribution of FSP Population Using Alternate Measures of Participation
Full Panel Population Cross-Sectional Cohort
%of
participants
%of
months
%of
benefits
%of
participants
%of
months
%of
benefits
Single Spell Participants
Short Term (1-8 months)
Medium Term (9-23 months)
Long Tt.m C^. 24 months)
28.6
12.4
33.4
7.0
11.6
55.6
6.2
11.8
56.9
6.7
7.8
58.9
1.5
3.1
70.7
1.2
3.1
71.7
Multiple Spell Participants 25.6 25.9 25.1 26.6 24.7 24.1
Totals 18.84 million — $21.6 billion 18.78 million .... $17.3 billion
n - 2,671 n = 1,733
bS
spell participants accounted for an additional 24 percent, so a total of 96 percent of all benefits
went to these two groups. Similarly, of the $21.6 billion in benefits paid to the full panel
population, 82 percent went to long-term and multiple-spell participants, although they made up
only 58 percent of all participants. Short-term participants, who made up about 29 percent of
the population, received only six percent of the benefits.
In addition to describing participation patterns and benefit consumption, we also looked at the
characteristics of each participant group. By examining the differences between the participants
groups, we may gain further insight into why some people remain on the program for a long
time while others leave relatively quickly.
There were noticeable differences in household composition between long-term, multiple-spell,
and other participants. The majority of single-elderly households (60 percent) were long-term
participants. Likewise, single-parent households were likely to be dependent on food stamps,
either as long-term or multiple-spell participants. In other words, arguably the two most
vulnerable groups among the poverty population were the most likely to be dependent on the
t
FSP over time.
Long-term participants were also more likely to be chronically poor. Eighty-nine percent of the
long-term participants lived in households with incomes that were below the poverty line in the
first month of FSP participation and 60 percent had incomes below poverty for all 28 months
of the SIPP Panel. In contrast, only 58 percent of the short-term participants lived in households
66
with incomes below 100 percent of poverty in their first month on the FSP, and only four
percent had incomes below poverty in every month. Multiple-spell participants, as expected,
were most likely to drift in and out of poverty (76 percent), but were also likely to be poor in
their first month of FSP participation (again, 76 percent). Clearly, the probability of
participating for a long time increases as income falls. Conversely, the probability of leaving
the program within eight months falls as income rises.
The income long-term participants did receive was more likely to come from public assistance
than from employment. Sixty-three percent of long-term participants' household income came
from means-tested transfers (compared to 42 percent for the full panel population in general);
only 15 percent was from earnings. On the other hand, half of short-term participants' income
was earned and only 20 percent was from transfers. Short-term participants were nearly four
times as likely as long-term participants to be employed in their first month on the FSP.
Multiple-spell participants were more diverse in terms of their income sources: slightly more
than one-third was from earnings, another one-third was from transfers, and one-quarter was
from other sources, such as unemployment compensation. About half of all adult multiple-spell
participants were either working (presumably at low-wage jobs as their incomes tended to be
below poverty) or unemployed and actively seeking work.
Short-term participants achieved higher educational levels than long-term participants. While
over 60 percent of short-term participants had completed high school, nearly 70 percent of long-
67
term participants had not. Clearly, the probability of participating for more than two years falls
as education increases. Conversely, the probability of participating for less than eight months
rises as education incieases.
In general, these statistics paint a picture of long-term and multiple-spell participants as the most
disadvantaged among the poor ~ those with the least education, the fewest ties to the labor
force, and the least available income.
Given the characteristics of long-term and multiple-spell participants, it is no surprise they were
the least likely to be required to register for work under the FSP. Only four percent of long-term
participants and nine percent of multiple-spell participants had to register for work.
However, long-term participants represented a substantial portion of all work registrants — 37
percent of the work registrants in the cross-sectional cohort were long-term participants.
Multiple-spell participants made up another 38 percent.
These results indicate that, in a given month, three-quarters of all work registrants are in the
midst of a spell that will last two years or more or are likely to return to the program after
leaving. Since these groups also consume the greatest amount of program resources over time,
targeting them with employment and training services may have the most impact. On the other
hand, since so many work registrants are long-term or multiple-spell participants, targeting may
not be necessary — these individuals will show up in employment and training programs as a
matter of course.
68
REFERENCES
Burstein, Nancy R. and Mary G. Visher. "The Dynamics of Food Stamp Program
Participation." Report to the Food and Nutrition Service, U.S. Department of Agriculture.
Washington, D.C.: Abt Associates, Inc., 1989.
Ellwood, David T. "Targeting 'Wouid-Be* Long-Term Recipients of AFDC." Report to the
U.S. Department of Health and Human Services. Princeton, NJ: Mathematica Policy Research,
Inc., 1986.
Martini, Alberto. "Participation in the Food Stamp Program: A Multivariate Analysis." In
Current Perspectives on Food Stamp Program Participation. Alexandria, VA: Food and
Nutrition Service, U.S. Department of Agriculture, 1992.
Usher, Charles L., Harlene C. Gogan and Helen P. Koo. "Long Term Participation in the Food
Stamp Program by Work Registrants." Report to the Food and Nutrition Service, U.S.
Department of Agriculture. Research Triangle Park, NC: Research Triangle Institute, 1989.
#
APPENDIX A
EFFECTS OF CENSORING IN SPELL DURATION
As discussed in the text, when analyzing data collected over a fixed time period, such as the
1987 SIPP, two types of bias (or censoring) arise. The first is commonly referred to as "left
censoring." This occurs when participation spells that were initiated prior to the beginning of
the survey period are truncated due to the data collection process. Analyzing an entry cohort
minimizes left censoring, because only households with newly-initiated participation spells are
included in the study sample.
Table 3 in the text shows the extent of right censoring for the cross-sectional cohort and both
right and left censoring for the full panel population. While censoring was not a problem in the
cross-sectional cohort, in the full panel population over half of all short-term spells and over 80
percent of the medium-term spells were truncated due to the data collection process. The
question then can be asked, how would this have changed if welfare history data were used to
adjust spell length for the full panel population, and thereby reduce left-censoring?
Table A-l presents a comparison of classifications of the full panel population with and without
the use of the recipiency history data. Use of retrospective information would reclassify 11
percent of the short-term participants as long-term and 12.2 percent as medium-term. Likewise,
36.5 percent of the medium-term recipients would be reclassified as long-term.
10
Table A-l. Comparison of the Classification of FSP Participants Based on Full Panel Population and
Incorporation of Historical Data
Definition Incorporating Recipiency History
Definition
Based on
SIPP Panel
Only
Short term
(1-8 months on FSP)
Number
Medium Term
(9-23 months on FSP)
Number
Long Term
(24 months or more)
Number
SINGLE SPELL PARTICIPANTS:
Short Term 6,300,381
(76.8%)
998,449
(12.2%)
901,500
(11.0%)
Medium Term 0
(0%)
2,247,505
(63.6%)
1,288,206
(36.5%)
Long Term 0
(0%)
0
(0%)
9,572,109
(100.0%)
Total Single
Spell Participants
6,300,381
(29.6%)
3245,954
(15.2%)
11,761,814
(55.2%)
Source: 1987 Survey of Income and Program Participation (SIPP) and Welfare History Module
(Wave 2).
n = 2,015 persons
II