The presentations for this panel report research from

1
Sanctions and Welfare Reform The Florida
Project
The presentations for this panel report
research from a project on welfare-to-work in
the state of Florida Principal
Investigators Richard Fording, University of
Kentucky Sanford Schram, Bryn Mawr College Joe
Soss, University of Minnesota
2
Sanctions as a Policy Tool in the Transformed
System of Welfare-to-Work Sanford F.
Schram Graduate School of Social Work and Social
Research Bryn Mawr College Presentation for
the 11th Annual ACF/OPRE Welfare Research and
Evaluation Conference May 28, 2008, Washington, DC
3
Approach Place statistical research on welfare
reform in interpretive context for understanding
welfare reforms relationship to broader changes,
including globalization and neoliberalism. Method
Conduct Problem-Driven Research that combines
different forms of analysis to best understand a
particular social problem. Focus Examine
Get-tough Welfare Policies such as increased use
of sanctions or financial penalties for
recipients who do not follow welfare-to-work
rules.
4
Rethinking the New Forms of Poverty Governance

• Globalization
• Neoliberalism
• Governmentality
• New forms of poverty governance and neoliberal
  governmentality
• Neoliberal Paternalism

5
The Neoliberal Paternalist State

• Laissez Faire at the top and
• Punitive and Disciplinary at the bottom.

6
Significant Features of the Paternalist Bottom

• (1) decreased financial aid to and increased work
  enforcement on the unemployed (Peck 2002)
• (2) decreased rehabilitation and increased
  incarceration for those who commit crimes
  (Wacquant 2001) and
• (3) decreased child welfare services to birth
  families and increased removal of children to
  foster families (Roberts 2002).

7
The Organizational Forms of the New Poverty
Governance

• Devolutionfirst-order and second-order
• Privatization-non-profit and for-profit
• Performance Measurement
• The Business Model

8
New Policy Tools

• Welfare Reform
• Time Limits
• Work Requirements
• Family Cap
• Sanctions-increased use, more punitive

9
Sanctions and Welfare Reform

• Welfare was reformed in 1996 to emphasize a get
  tough approach to moving recipients as quickly
  as possible from welfare to work (with reforms
  including time limits, work requirements,
  compliance monitoring, performance incentives,
  family caps, and sanctions).
• Sanctions are financial penalties for failure to
  comply with welfare-to-work requirements.
• Sanctions have become a key tool for achieving
  the goals of welfare reform.
• From 1997-1999, nearly 500,000 families lost
  benefits due to sanctions approx. one-quarter
  of the caseload reduction for that period
  (Goldberg and Schott 2000).

10
Figure 1. Proportional Change in Rates of
Incarceration and AFDC/TANF Receipt, 1990-2001
11
Figure 2. States with Highest Growth in the
Number of Prisons Number of Prisons in 2000 and
the Percent Change in the Number of Prisons,
1979-2000
Source Sarah Lawrence and Jeremy Travis (2004).
The New Landscape of Imprisonment Mapping
Americas Prison Expansion. Washington, DC
Urban Institute, Justice Policy
Center.http//www.hawaii.edu/hivandaids/The_New_La
ndscape_of_Imprisonment.pdf
12
Figure 3. Black TANF Caseload and State
Corrections Spending by TANF Regime Type
Note TANF regime stringency are based on 2001
state TANF policies as measured by the Urban
Institute and Gainesborough 2003 corrections
spending data are from Sourcebook of Criminal
Justice Statistics 1999 (2000 5-9, Table 1.5).

13
Why Race Still Matters?

• Possible Explanations
• (1) Resisting the Soft Bigotry of Low
  Expectations--welfare policy now more strictly
  holds blacks and whites to the same standards and
  blacks are more likely to not comply
• (2) Classifying by Race--welfare policy still
  is based on racial bias even, or especially, with
  new forms of governance and new policy tools.

14
Figure 4 Scatterplot of the Relationship between
the Racial/Ethnic Composition of the TANF
Caseload and the Sanction Rate, FY 2002
15
Figure 5. Floridas 24 Workforce Regions
16
Table 1. Average cumulative sanction rates
across 24 workforce regions in Florida
Note The sanction rate is the percentage of all
adults sanctioned during their first 12 months on
TANF, of those beginning a new TANF spell between
July 2000 and May 2002.
17
Figure 6. Total Adult TANF Caseload by
Race/Ethnic Identification, January 2000-December
2003

18
Figure 7. Total TANF Sanctions and Non-Sanction
Exits in Florida, January 2000 December 2003
19
Figure 8. Monthly TANF Sanction Rates ( of
Monthly TANF Adult Caseload)
20
Research Question How do client characteristics
and local contexts affect the probability that a
client will be sanctioned in the Florida WT
program?

• Individual Characteristics
• Gender Number of Children Race Martial Status
• Work History Age of Youngest Child Age
  Education

Contextual Factors County Unemployment
Rate County poverty rate Country Wage Rate
County racial composition County Caseload Size
County population Local Political Ideology
21
Construction of Index of County Political
Ideology To construct our index of local
ideology we collected data on 18 ideologically
relevant constitutional amendments that appeared
on a statewide ballot for ratification from 1996
through 2004. We computed the percentage of
yes votes for each amendment, for each county,
and conducted a factor analysis using all 18
amendments (thus 18 variables, N67 counties).
The subjects of these amendments are listed in
the table below.

22
(No Transcript)
23
Statistical Approach An Event History Model of
TANF Sanctions

• We analyze all new adults entering the TANF
  program from January 2001 to December 2002
• 24 separate cohorts based on shared month of
  entry
• We track these adults only through their first
  spell of receiving cash assistance and for a
  period that is no more than 12 months.
• After applying these criteria, our analysis is
  based on approximately 60,000 TANF adults

24
Table 2. Cox Proportional Hazard Models of
Effect of Individual and Community-Level
Characteristics on Sanction Initiation
25
Figure 9. Cumulative Survival Function for a
Typical TANF Adult, by Local Political Ideology
Note Survival rates are estimated for a 31
year-old white woman with 12 years of education
and average level of wage income. These
estimates are based on the results presented in
column II, Table 3.
26
Sanctioning and Organizational Incentives

• Its all about the numbers now, unfortunately.
  We have to meet our numbers because it affects
  our funding. I wish we werent so obsessed with
  the numbers. Sometimes I feel it performance
  pressure makes me call people just to say I
  need that number so we can get credit for it
  rather than talking about helping them. But I
  guess it is the nature of the beast. We need to
  be concerned about the numbers so we can get our
  funding.
• - WT Case Manager, 2005

27
Sanctioning and Organizational Incentives

• The performance feedback hypothesis is based on
  the observation that in addition to chronic
  performance pressures, regional officials and
  case managers experience increased performance
  pressure in the wake of negative performance
  feedback communicated through periodic state
  reports. Therefore, we expect that in response to
  declining regional performance, regional sanction
  rates will increase.
• The client group hypothesis suggests that the
  effect of performance feedback on regional
  sanction rates will be strongest among
  hard-to-serve clients who are least prepared to
  be successful in the workforce, and therefore
  most likely to be perceived as problem clients.
• The regional variation hypothesis suggests that
  the effect of negative performance feedback on
  regional sanctioning rates will be strongest
  among local welfare offices that embrace the get
  tough approach to welfare reform that recognizes
  an important role for sanctions and other
  punitive practices to motivate clients.

28
Sanctioning and Organizational Incentives

• Data Panel data for Fla counties (10/01
  03/04)
• Dependent variable is the monthly sanction rate
• ( sanctioned / caseload 100)
• Performance Ranking is defined as the change in
  the average monthly ranking (1-24) for the
  entered wage rate, the welfare return rate, and
  the entered employment rate. This variable is
  measured at a lag of three months and is measured
  cumulatively within each fiscal year.
• Estimation - OLS, with panel corrected standard
  errors, full set of fixed effects for workforce
  regions and month of analysis, controls for
  caseload characteristics, economic conditions.

29
Table 3. Effect of Regional Performance Measures
on Regional Sanction Rates

p
30
Table 4. Effect of Regional Performance Measures
on Regional Sanction Rates, by Local Political
Culture

p
31
Table 5. Effect of Regional Performance Measures
on Regional Sanction Rates, by Local Political
Culture and Race of Client
p
32
Figure 10. Summary of Effects of Performance
Feedback on Regional Sanction Rates
33
The New Poverty Governance

• New organizational forms, and new policy tools
  are being used to roll out a more punitive
  approach to poverty management.
• Decentralized, privatized system allows politics
  to affect the extent to which communities enforce
  the punitive approach.
• The new system re-encodes racial hierarchy,
  identifying nonwhites as less deserving of aid
  and more deserving of punishment.

34
Deciding to Discipline A Multi-Method Study of
Race, Choice, and Punishment at the Frontlines
of Welfare Reform

• Linda Houser
• Bryn Mawr College
• Presentation for the 11th Annual ACF/OPRE Welfare
  Research and Evaluation Conference
• May 28, 2008, Washington, DC

35
The paper on which this talk is based is
available as a NPC working paper
(07-33)http//www.npc.umich.edu/publications/u
/working_paper07-33.pdf
36
Race and Sanctions

• Historically, race has been associated with
  disparities in treatment in welfare provision
  (Lieberman 1998 Quadagno 1994 Ward 2005).
• Race-coded appeals and racialized public
  responses played a key role in the national
  debates that led up to reform (Hancock 2004
  Reese 2005).
• More recently, race has been associated with
  state decisions to adopt more punitive welfare
  policies (Soss et al. 2001).
• Most recently, race has been identified as a
  primary predictor of the likelihood of being
  sanctioned (Meyers et al. 2006).

37
Race and Caseworker Profiling

• Part of our project has focused on the role of
  race in welfare reform from the national
  government down to the state and local levels.
• Caseworkers have long held discretion in dealing
  with their clients (Lipsky 1980), but they have
  been given a variety of new powers and
  responsibilities under welfare reform
  (Watkins-Hayes 2008).
• The specific research presented here examines
  whether race influences what goes on at the
  bottom of the chain of decisionmakingi.e.,
  frontline caseworkers decisions to impose
  sanctions.

38

• Racial Classification Model (RCM)
• 1. Policy actors rely on salient social
  classifications and group reputations in
  designing social policies and applying policy
  tools to particular target groups.
• 2. When racial minorities are salient in a policy
  context, race will be more likely to provide a
  basis for social classification of targets and,
  hence, to signify target group differences
  perceived as relevant to the accomplishment of
  policy goals.
• 3. The likelihood of racially patterned policy
  outcomes will be positively associated with the
  degree of policy-relevant contrast in policy
  actors perceptions of racial groups. The degree
  of contrast, in turn, will be a function of
• (a) the prevailing cultural stereotypes of
  racial groups
• (b) the extent to which policy actors hold
  relevant group stereotypes and
• (c) the presence or absence of
  stereotype-consistent cues.

39
Race and Expectancy Confirmation

• Stereotype-consistent cues have been shown to
  activate pre-existing racial bias in a number of
  areas
• retail sales (Ayres and Siegelman 1995)
• mortgage loans (Munnell et al. 1996) insurance
  (Wissoker et al. 1998)
• healthcare (Schulman et al. 1999) housing
  (Yinger 1995)
• labor markets (Bertrand and Mullainathan 2003
  Pager 2007)
• The RCM might help us understand how
  stereotype-consistent markers can function as a
  source of what Darley and Gross (1983) call
  expectancy confirmation.
• The cue strengthens the effects of racial status
  on decision making by activating a pre-existing
  expectation about the racially marked persons
  behavior.
• Importantly, these processes can emerge from
  cognitive biases in decision making even in the
  absence of conscious racial animus, out-group
  threat, or in-group favoritism (cf. Key 1949
  Blalock 1967).

40
Racial Cues and Profiling ClientsCaseworkers
may be vulnerable to relying on racial cues for
profiling clients just as citizens rely on racial
cues in making voting decisions.
Valentino, Hutchings, and White (2002
86) When the black racial cues are
stereotype-inconsistent, the relationship between
racial attitudes and the vote disappears.
Likewise the presence of black images alone
does not prime negative racial attitudes. The
effect emerges only when the pairing of the
visuals with the narrative subtly reinforces
negative stereotypes in the mind of the viewer.
41
Sanctioning on the Frontlines

• Data Sources
• Web-based case manager survey
• Embedded 2x2 randomly assigned experimental
  vignettes
• Administrative data on TANF adults
• Longitudinal data on client characteristics,
  sanctioning history and earnings (2001-2004)

42
Why Study Florida?

• Florida relies on immediate and full-family
  sanctions, the strictest (i.e., most
  disciplinary) combination of sanctioning choices
  (Pavetti et al. 2003).
• Our interviews suggest that caseworkers
• are for the most part committed to the overall
  goals of TANF and to the tools provided to them
  for goal achievement
• are at times ambivalent about the effectiveness
  of sanctions
• perceive a lack of alternative tools for bridging
  the gap between performance expectations and
  client situations

43

• Within the high-pressure, frontline work of
  Welfare Transition case managers, does race
  intersect with stereotype-consistent cues to
  place clients in increased jeopardy of receiving
  a welfare sanction?

44
Table 6. Distribution of Responses by Region
45
Table 7. Characteristics of Caseworker Respondents
46
Experimental Vignettes
Vignette 1 Emily OBrien/Sonya Perez is a 28
year-old single mother with one child aged 7 /
four children who is currently in her fourth
month of pregnancy. She entered the Welfare
Transition program six months ago, after leaving
her job as a cashier at a neighborhood grocery
store where she had worked for nine months. Emily
was recently reported for being absent for a week
from her assignment for community service work
experience. Immediately after hearing that Emily
had not shown up for a week of work, Emily's
caseworker mailed a Notice of Failure to
Participate (Form 2290) and phoned her to ask why
she had missed her assignment. Emily was not home
when the caseworker called. However, when she
responded to the 2290 three days later, she said
she no longer trusted the person who was looking
after her child, and she did not want to go back
to work until she found a new childcare provider.
Emily returned to work the next day. Vignette
2 Emily OBrien/Lakisha Williams is a 26
year-old single mother with two children. She has
been in the Welfare Transition program for five
months. Lakisha was recently reported for failing
to show up for a job interview that had been
scheduled for her with a local house-cleaning
service. Immediately after hearing about the
missed interview, Lakishas caseworker mailed a
Notice of Failure to Participate (2290) and
phoned her to ask why she had not shown up.
Lakisha said she had skipped the interview
because she had heard that a better position
might open up next month with a home health
agency She had been sanctioned two months
earlier for failure to complete her hours for
digital divide.
47
Variation within Vignettes

• RACE
• Vignette 1 Hispanic-sounding name vs.
  White-sounding name
• Vignette 2 Black-sounding name vs.
  White-sounding name
• REPUTATIONAL DIFFERENCE
• Vignette 1 Young mother with multiple children
  who is also pregnant
• Vignette 2 Repeat welfare recipient who is not
  only returning to TANF but also was previously
  sanctioned

48
Figure 11. Florida Sanction Flow Chart

49
Figure 12. Sanction Rate by Client Name Condition
50
Table 8. Analysis of Vignette Experiments
51
Figure 13. Predicted Probabilities by Race and
Condition
52
Strengths, Limitations Triangulation

• Confidence in results based on
• Random assignment of case narratives
• Consistency of results from two very different
  vignettes
• Limitations include
• Scenarios are, in the final reckoning,
  hypothetical
• Sanctioning can occur without a face, story, or
  detailed file
• Sanctioning can occur (or not) with no concern
  about performance numbers
• Therefore, we triangulate our findings with
  administrative data provided by the Florida
  Department of Children and Families (DCF)

53
Table 9. Cox Proportional Hazard Models of
Effects of Minority Status and Number of Children
on Sanction Initiation
54
Tab le 10. Weibull Selection Model of Effects of
Minority Status and Sanction History on Sanction
Initiation during Second TANF Spell
55
Key Findings Experiment

• White clients suffer no statistically discernible
  negative effects when linked to characteristics
  that hold negative meanings in the
  welfare-to-work context.
• Minority clients, by contrast, are vulnerable to
  the attachment of discrediting,
  stereotype-consistent markers, such as having
  multiple children and having received a prior
  sanction.
• More experienced case managers (i.e., those with
  more than two years experience) are significantly
  less likely to impose sanctions in either case.
• White case managers were no more likely than
  nonwhite case managers to sanction clients
  overall nor were they more likely than nonwhite
  case managers to sanction nonwhite clients.

56
Key Findings Administrative Data

• Hispanic clients do not emerge as more likely
  than white clients to be sanctioned, and this
  null finding holds regardless of number of
  children.
• Among second-spell participants, black clients
  with a prior sanction are more likely than their
  white counterparts to be sanctioned again.

57
Our findings suggest that, while TANF is
ostensibly a race-neutral public policy, it is
carried out today in a way that allows
pre-existing racial stereotypes and race-based
disadvantages to produce large cumulative
disadvantages (Schram 2005, 2006).
58
Evaluating Sanctioning Outcomes Subjective
Caseworker Assessments and Objective Impacts
Richard Fording Department of Political
Science University of Kentucky Presentation for
the 11th Annual ACF/OPRE Welfare Research and
Evaluation Conference May 28, 2008, Washington,
DC
59
What Difference Does It Make?The Consequences of
Sanctioning

• In addition to studying why sanctions are
  imposed in some cases more than others, we are
  also studying how sanctions affect several key
  outcomes in Floridas WT program
• Return Rates and Future TANF Usage
• Odds of Being Sanctioned in the Future
• Client Earnings after exit

60
What Difference Does It Make?The Consequences of
Sanctioning

• Data sources
• Case manager survey
• CM opinions about sanctions and their effects
• Administrative data on TANF adults
• Longitudinal data on client characteristics,
  sanctioning history and earnings (2000-2004)

61
Do Case Managers Support Sanctions?
62
Figure 14. In general, do you favor or oppose
the sanction policy of the WT program?
N 125
63
Figure 15. For each, which option for an initial
sanction do you favor most?
N 128
64
Figure 16. For each, which option for an initial
sanction do you favor most?
N 127
65
Figure 17. For each, which option for an initial
sanction do you favor most?
N 127
66
Figure 18. On the whole, do you feel that the
sanctioning policies used in your region are too
lenient, too strict, or just about right?
N 123
67
Do Case Managers Believe that Sanctions are
Effective?
68
Figure 19. Sanctions are effective in helping
clients become self-sufficient.
N 126
69
Figure 20. Sanctions make clients more likely to
follow program rules.
N 128
70
Conclusions from Case Manager Survey

• Case managers
• display strong support for the sanction policy in
  Florida
• generally believe that sanctions are an effective
  tool for helping clients learn to be
  self-sufficient

71
Administrative Data Analysis

• How does sanctioning affect
• TANF Recidivism
• Odds of being sanctioned in the future
• Client earnings
• Nonexperimental data our analyses are limited
  in their ability to test for the causal effects
  of sanctioning

72
Administrative Data Analysis

• Sample
• All Florida TANF recipients (adults)
• New TANF recipients (off TANF for preceding 12
  months)
• First cohort enters January 2001
• Last observation month (Sept. 2003 April 2004)
• Definition of spell consecutive months of
  receipt

73
Sanctions and Return Rates
Table 11. Compared to clients who exit for other
reasons, sanctioned clients are more likely to
return to the TANF program
Note This table is based on 15 cohorts of new
TANF clients, entering TANF from January 2001
through March 2002.
74
Sanctions and Time Between Spells
Table 12. Compared to clients who exit for other
reasons, sanctioned clients return to TANF more
quickly
Note This table is based on 15 cohorts of new
TANF clients, entering TANF from January 2001
through March 2002.
75
Table 13. Cox Proportional Hazard Model of Return
to TANF
Note The sample includes all clients who entered
TANF for a first spell beginning January 2001 or
later, and exited by December 2002. All clients
who have not returned by August 2003 are treated
as censored. The dependent variable is the hazard
of return for second spell (for each month after
exit). The model was estimated in Stata 10.0
using the stcox procedure. Cell entries are
hazard ratios, with p-values based on robust
standard errors (adjusted for error clustering at
the county level). p 76
Effects of First-Spell Sanction on TANF Return,
by Race/Ethnicity and Education

• Hazard ratios for specific client groups
• White clients 1.20
• Black clients 1.22
• Hispanic clients 1.48
• Less than H.S. education 1.19
• H.S. Education or more 1.35
• Hazard ratios estimated based on Cox model
  described in previous slide p

77
Do sanctions reduce the odds that a client will
be sanctioned in the future?

• After sanctioned clients return to TANF, are
  they more or less likely to be sanctioned than
  otherwise similar clients?
• Sanctions may teach clients to comply with
  program rules (negative effect)
• Sanctions may be due to client characteristics
  (barriers to employment) that cannot easily be
  changed (positive effect)
• Case managers may be less likely to give
  previously sanctioned clients the benefit of the
  doubt (positive effect)

78
Table 15. Compared to clients who were not
previously sanctioned, clients sanctioned in the
first spell are just as likely or more likely to
be sanctioned in the second spell
Note The sample for this analysis consists of
all clients identified in the previous two tables
who returned for a second spell prior to April
2003.
79
Methodological Issues

• Need to control for other client characteristics
• Selection bias (nonrandom selection into second
  spell)
• The effect of a first-spell sanction on
  second-spell behavior may vary depending upon the
  clients belief that she will be sanctioned

80
Figure 21. Average Monthly Sanction Rate across
Florida Counties, 2000-2004
Note The monthly sanction rate is calculated as
( of adults sanctioned) / (number of adults
receiving TANF)100
81
Table 16. Weibull Hazard Models of Second-Spell
Sanction
(continued)
Note The sample includes all adult TANF
recipients who started a first TANF spell between
January 2001 and December 2002, and returned to
TANF by September, 2003. The model was estimated
in Stata 10.0 using the streg procedure for the
single equation model, and the dursel procedure
for the selection model (Boehmke 2005). Cell
entries are hazard ratios, with p-values based on
robust standard errors (adjusted for error
clustering at the county level). p 82
Table 17. Weibull Hazard Models of Second-Spell
Sanction
(continued)
Note The sample includes all adult TANF
recipients who started a first TANF spell between
January 2001 and December 2002, and returned to
TANF by September, 2003. The model was estimated
in Stata 10.0 using the streg procedure for the
single equation model, and the dursel procedure
for the selection model (Boehmke 2005). Cell
entries are hazard ratios, with p-values based on
robust standard errors (adjusted for error
clustering at the county level). p 83
Figure 22. Predicted Effect of First-Spell
Sanction on Second-Spell Sanction (by Local
Sanction Rate)
84
How Do Sanctions Affect Client Earnings?

• Pre-existing Differences as noted earlier,
  clients with weaker earning histories are more
  likely to wind up being sanctioned. Thus, we
  should expect sanctioned clients to continue to
  have lower earnings after program exit.
• Key Question Do sanctions reduce or widen this
  gap in earnings? That is, do sanctions put this
  low-earning group of clients at a greater
  disadvantage, or do they push them toward the
  higher earnings of non-sanctioned clients?
• Method To answer this question we compare
  quarterly earnings of sanctioned clients to
  non-sanctioned clients, before entering and after
  exiting TANF

85
Analysis of Client Earnings

• Sample
• Panel data for 15 quarters, starting January 2000
  and ending September 2003
• Include all clients who started a first TANF
  spell anywhere between Jan. 2001 - Dec. 2002, and
  who exited prior to June 2003

86
Analysis of Client Earnings

• Dependent Variable
• Quarterly earned income (Florida UI records)
• Estimation
• Separate regression models for sanctioned clients
  and non-sanctioned clients
• Individual fixed effects
• Include dummies for
• 4 quarters prior to entry
• Entry quarter
• Exit quarter
• 4 quarters after exit
• Include dummies for TANF receipt, observation
  quarter

87
Figure 23. Predicted Client Earnings through the
First TANF Spell (Never-Sanctioned Clients)
88
Figure 24. Predicted Client Earnings through the
First TANF Spell, by Sanction Experience
89
Figure 25. Predicted Client Earnings as a Percent
of Prior Earnings, by Sanction Experience
90
Figure 26. Predicted Client Earnings as Percent
of Prior Earnings, by Sanction Experience, for
Whites
91
Figure 27. Predicted Client Earnings as a Percent
of Prior Earnings, by Sanction Experience, for
Blacks
92
Figure 28. Predicted Client Earnings as a Percent
of Prior Earnings, by Sanction Experience, for
Hispanics
93
Figure 29. Predicted Client Earnings as a Percent
of Prior Earnings, by Sanction Experience, for
94
Figure 30. Predicted Client Earnings as a Percent
of Prior Earnings, by Sanction Experience, H.S.
Education or More
95
Summary of Findings

• How does sanctioning affect
• TANF Recidivism
• Sanctioned clients more likely to return to TANF
• Odds of being sanctioned in the future
• Sanctioned clients more likely to be sanctioned
  in subsequent TANF spells (although this effect
  declines in high-sanctioning counties)
• Client earnings
• Never-sanctioned clients experience a small
  increase in earnings (compared to pre-entry
  levels)
• Sanctioned clients often do not return to
  earnings levels seen prior to entry
• The earnings gap between sanctioned and
  non-sanctioned clients expands after exit
• Nonexperimental data our analyses are limited
  in their ability to test for the causal effects
  of sanctioning