Popular in Social Sciences

Political Science

V-Dem’s Conceptions of Democracy and Their Consequences

University of Notre Dame Michael Coppedge

Central Monetary Authorities

Macroeconomic modelling of CBDC: a critical review

European Central Bank Ulrich Bindseil

Social Identities and Divisions

A new explanation of antisemitism: Jew hatred as a civilisational phenomenon

Boston University Liah Greenfeld

Central Monetary Authorities

Financing Private Credit

Federal Reserve Bank of New York Nina Boyarchenko

Economics

Failing Banks

Massachusetts Institute of Technology Emil Verner

Culture and Society

The City as a Growth Machine: Toward a Political Economy of Place

New York University Harvey Moltoch

American Politics

Evaluating Contradictory Experimental and Nonexperimental Estimates of Neighborhood Effects on Economic Outcomes for Adults

View more insights in Social Sciences - Sociology

View more figures from UC Berkeley

View the page of David J. Harding

Read the Study

Next in Social Sciences

Measurement and Theory of Core Inflation

Argia Sbordone | Federal Reserve Bank of New York

The Liberalism of Refuge

Bryan Garsten | Yale University

David James Harding discusses a within-study comparison of the Moving to Opportunity (MTO) experiment and the Panel Study of Income Dynamics (PSID) to challenge nonexperimental findings on neighborhood effects for adults, revealing that similarities in estimates from both sources suggest selection bias may drive observed neighborhood effects on adult economic outcomes.

Image courtesy of interviewee. October 30, 2024

Log-in or Sign-up to Faculti
Currently viewing this subject insight as a guest. You have insight(s) remaining for this month. Login to view 8000+ figures on the platform.

Generating Workspace...

Export as PDF

This video discusses the impacts of living in high poverty neighborhoods, focusing on children and adults, particularly in relation to social mobility and health. It identifies a consensus on negative effects for children, while findings for adults regarding employment and earnings are mixed and puzzling. The "Moving to Opportunity" study, which provided low-income families with housing vouchers to relocate to lower-poverty areas, yielded null results on employment and earnings, leading to questions about why the expected positive effects didn't materialize.

This video explores various hypotheses, including potential selection bias, differences in study designs, the disruptiveness of moving, and the role of social networks and health. It emphasizes the need for further research on these aspects and the importance of natural experiments in understanding these complex social dynamics. The findings suggest that while vouchers can alter neighborhood conditions, the broader impacts on employment require careful consideration of the nuances involved in such transitions for low-income families, particularly those who are precariously employed.

Ultimately, the text underscores the importance of understanding both the processes leading to neighborhood segregation and the disruptive effects of relocation in policy discussions regarding housing assistance programs.

Language:

Export as PDF

Export as PDF
Here are five key points from the video :

1. **Impact of High Poverty Neighborhoods**: Research in social sciences has consistently shown that living in high poverty neighborhoods adversely impacts children and adults alike, affecting social mobility, opportunity, mental health, and physical health. There is an emerging consensus on these negative outcomes for children, but results concerning adults’ employment and earnings remain contentious and unclear.

2. **Research Methodologies**: The studies on neighborhood effects employ two main methodologies: observational studies that track individuals over time in their real-life convideos and natural experiments that utilize pre-existing variations in housing interventions (e.g., the Moving to Opportunity study) to examine neighborhood impacts without direct random assignment.

3. **Moving to Opportunity Study (MTO)**: This particular study randomized families living in public housing across different cities to either receive a housing voucher to move to lower poverty neighborhoods or remain in their current environment. The study found a significant decrease in neighborhood poverty rates for families receiving vouchers but did not identify clear positive impacts on employment and earnings outcomes.

4. **Selection Bias and Disruptive Moves**: The paper discusses the possibility of selection bias affecting observational study outcomes compared to the quasi-experimental findings from MTO. The researchers also consider that the process of moving can be disruptive for individuals, potentially affecting their social networks and support systems, which may hinder employment opportunities.

5. **Future Research Directions**: The findings point to a need for further exploration of the effects of moving between neighborhoods, especially for socioeconomically disadvantaged groups. Researchers suggest examining the reasons why families end up in either high or low poverty neighborhoods and taking into account the disruptive nature of moving, particularly for low-income families with precarious employment situations, to better understand the overall impact of housing voucher programs.

Export as PDF
Certainly! Here are some key quotes from the video you've , which summarizes findings and discussions around the impacts of living in high poverty neighborhoods and the results of the Moving to Opportunity study:

1. **Impact on Children and Adults**:
- "There’s now what we call an emerging consensus across studies that these impacts on children are real. They're large enough to have an impact on social mobility and opportunity."
- "There's also an emerging consensus that there are mental health and physical health effects of living in high poverty neighborhoods."

2. **Discrepancies in Adult Outcomes**:
- "There's still a remaining puzzle in this literature... different studies show different things... no impact on employment and earnings, whereas other studies show there is a negative impact on things like employment and earnings for adults."

3. **Social Isolation and Job Opportunities**:
- "High poverty neighborhoods tend to be... distant from where job opportunities are... they tend to be more socially isolated, their environments of social distrust."

4. **Research Methods**:
- "This is not a setting in which we can do an experiment... The two types of research designs we can do are either what we call observational studies... and the second type is called a natural experiment."

5. **Moving to Opportunity Study**:
- "The Moving to Opportunity Study... randomizes the being offered a housing voucher that one can use in the private rental market."

6. **Neighborhood Change and Employment**:
- "We do see that there are big impacts of being offered the vouchers... but we’re not finding any effects on employment and earnings."

7. **Selection Bias**:
- "The main weakness of the observational studies is that there’s some remaining thing that we didn’t control for..."

8. **Disruption from Moving**:
- "Every family that is in the Moving to Opportunity study has to make a residential move... those residential moves could potentially be disruptive."

9. **Negative Outcomes for Certain Groups**:
- "Women who were employed at the beginning of the study and especially those who had less than a high school degree... their economic trajectories were actually negatively impacted."

10. **Future Research Directions**:
- "It’s really important to focus future research on the selection process itself... Understanding that process will then help us to interpret our results about the impacts of those neighborhoods."

These quotes encapsulate key themes and findings from the video regarding the complexities associated with high poverty neighborhoods, methodological challenges, and implications for public policy regarding housing and social services.

X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:90000

1
00:00:00.000 --> 00:00:01.965
There's a longstanding literature,

2
00:00:02.065 --> 00:00:05.565
in the social sciences dating back decades,

3
00:00:05.795 --> 00:00:09.805
that is interested in, interested in the kind of the impacts

4
00:00:09.805 --> 00:00:11.445
of living in high poverty neighborhoods.

5
00:00:12.065 --> 00:00:13.565
So both on, children,

6
00:00:13.665 --> 00:00:15.045
if we're thinking about social mobility

7
00:00:15.045 --> 00:00:17.405
and social opportunity, and also on adults,

8
00:00:17.575 --> 00:00:18.765
there opportunity.

9
00:00:19.265 --> 00:00:22.725
Uh, and it's also closely connected to concerns about,

10
00:00:22.725 --> 00:00:24.525
racial inequality and class inequality.

11
00:00:24.745 --> 00:00:28.525
Uh, and so in the US context, the combination of racial

12
00:00:28.785 --> 00:00:32.205
and class segregation in neighborhoods produces,

13
00:00:32.205 --> 00:00:33.685
especially in central cities, a lot

14
00:00:33.685 --> 00:00:34.925
of concentrated poverty neighborhoods

15
00:00:34.945 --> 00:00:36.165
or high poverty neighborhoods.

16
00:00:36.785 --> 00:00:40.805
Uh, and we're, we're very interested in the long term in,

17
00:00:40.985 --> 00:00:43.445
what's the impact of the people who live

18
00:00:43.465 --> 00:00:46.565
or grow up in these neighborhoods on their life prospects.

19
00:00:46.565 --> 00:00:48.525
And across a number of different domains.

20
00:00:48.555 --> 00:00:51.285
There's now what we call an emerging consensus across

21
00:00:51.285 --> 00:00:56.245
studies that there these impacts, on children are real.

22
00:00:56.675 --> 00:00:59.325
They're large enough to have impact on social,

23
00:00:59.785 --> 00:01:02.525
social mobility and, and opportunity,

24
00:01:02.785 --> 00:01:04.445
and also among adults.

25
00:01:04.785 --> 00:01:06.765
Uh, there's also an emerging consensus

26
00:01:06.765 --> 00:01:08.045
that there are mental health

27
00:01:08.105 --> 00:01:10.645
and physical health effects of, of living,

28
00:01:11.065 --> 00:01:12.285
in high poverty neighborhoods.

29
00:01:13.075 --> 00:01:16.045
There's still a remaining puzzle in this literature,

30
00:01:16.055 --> 00:01:18.805
which is that there's kind of an uncertainty

31
00:01:18.945 --> 00:01:22.525
or a discrepancies in research findings about

32
00:01:23.555 --> 00:01:26.335
the impact of living in a high poverty neighborhood on

33
00:01:26.335 --> 00:01:29.935
adults and their employment, and earnings prospects.

34
00:01:30.675 --> 00:01:33.735
So, different studies show different things.

35
00:01:33.845 --> 00:01:35.935
Some studies, particularly the Moving

36
00:01:35.955 --> 00:01:38.575
to Opportunity study show no impact.

37
00:01:39.195 --> 00:01:42.255
Uh, and whereas other studies with other data sets,

38
00:01:42.255 --> 00:01:45.395
with other research designs show, there,

39
00:01:45.395 --> 00:01:47.595
there is a negative impact on things like employment

40
00:01:47.595 --> 00:01:50.435
and earnings, for adults, compared

41
00:01:50.435 --> 00:01:51.995
to similarly situated adults

42
00:01:52.015 --> 00:01:54.835
who live in lower poverty neighborhoods.

43
00:01:55.015 --> 00:01:57.355
So that's kind of the puzzle that this paper,

44
00:01:57.575 --> 00:01:58.755
tries to address.

45
00:01:59.335 --> 00:02:01.955
Uh, and the reason it's an important puzzle also is

46
00:02:01.955 --> 00:02:05.355
that there are, a number of social science theories

47
00:02:05.355 --> 00:02:06.635
that will lead us to expect there

48
00:02:06.635 --> 00:02:07.995
to be these effects, right?

49
00:02:08.055 --> 00:02:10.715
So, high poverty neighborhoods tend to be,

50
00:02:11.055 --> 00:02:14.275
in certain geographic places in the city that are,

51
00:02:14.275 --> 00:02:16.635
distant from where job opportunities are.

52
00:02:17.095 --> 00:02:19.995
Um, they as a, also as a result, they tend

53
00:02:19.995 --> 00:02:22.635
to be more socially isolated, their environments

54
00:02:22.635 --> 00:02:23.635
of social distrust.

55
00:02:23.655 --> 00:02:25.315
And we know that people often find jobs

56
00:02:25.455 --> 00:02:27.675
and find better jobs through social networks.

57
00:02:27.705 --> 00:02:30.875
There's also questions around, you know, norms around work

58
00:02:30.875 --> 00:02:34.155
and employment in, in high poverty neighborhoods.

59
00:02:34.775 --> 00:02:37.275
We know that high poverty neighborhoods actually sort

60
00:02:37.275 --> 00:02:39.995
of face a stigma in the labor markets,

61
00:02:40.225 --> 00:02:41.915
that can affect your chances of employment.

62
00:02:42.255 --> 00:02:44.275
Uh, and then there's also a set of health pathways.

63
00:02:44.575 --> 00:02:46.715
You know, if, if people is mental health is impacted,

64
00:02:46.715 --> 00:02:50.155
that's gonna impact their ability, to, to work

65
00:02:50.295 --> 00:02:51.795
and to, and to hold a job.

66
00:02:51.935 --> 00:02:54.555
So, sort of understand the methodological puzzle here.

67
00:02:54.555 --> 00:02:56.515
It's important to understand a little bit about the

68
00:02:56.515 --> 00:02:58.475
different types of research designs

69
00:02:58.475 --> 00:02:59.845
that are used to study this question.

70
00:03:00.745 --> 00:03:03.565
Uh, so this is not a setting in which we

71
00:03:03.565 --> 00:03:04.605
can do an experiment.

72
00:03:04.625 --> 00:03:06.485
We can't randomly assign someone

73
00:03:06.985 --> 00:03:08.645
to live in a high poverty neighborhood

74
00:03:08.645 --> 00:03:10.405
and someone else to live in a low poverty neighborhood

75
00:03:10.405 --> 00:03:13.405
that would be, unethical and impractical and, and,

76
00:03:13.405 --> 00:03:15.685
and very expensive, as well to,

77
00:03:15.705 --> 00:03:17.485
to do an actual experiment.

78
00:03:17.625 --> 00:03:21.285
So the two types of research designs we can do are either

79
00:03:21.555 --> 00:03:23.245
what we call observational studies,

80
00:03:23.735 --> 00:03:26.685
where we're following people, you know, as they are,

81
00:03:26.865 --> 00:03:28.365
as they go about their lives in the world

82
00:03:28.385 --> 00:03:29.565
and collecting data on them.

83
00:03:30.105 --> 00:03:33.085
Um, and the second type is called a natural experiment,

84
00:03:33.335 --> 00:03:35.405
where it sort of mimics an experiment.

85
00:03:35.565 --> 00:03:39.685
There's some, something out there that is producing

86
00:03:39.685 --> 00:03:42.925
what looks like random variation in who ends up in what type

87
00:03:42.925 --> 00:03:44.525
of neighborhood, in this case.

88
00:03:45.525 --> 00:03:47.465
So, an observational study would be just,

89
00:03:47.465 --> 00:03:48.465
we follow people over time,

90
00:03:48.925 --> 00:03:50.385
and that will naturally,

91
00:03:50.475 --> 00:03:51.625
we'll naturally see people living

92
00:03:51.625 --> 00:03:52.705
in different types of neighborhoods.

93
00:03:53.085 --> 00:03:55.785
We can measure many things about them,

94
00:03:56.315 --> 00:03:58.225
especially things we would want to condition on

95
00:03:58.225 --> 00:04:00.065
or control for as we're trying to look at the impacts

96
00:04:00.065 --> 00:04:02.585
of the, of the neighborhoods on their, on their outcomes.

97
00:04:03.165 --> 00:04:04.745
Um, but the downside of, of

98
00:04:04.745 --> 00:04:07.025
that observational study is we can't know

99
00:04:07.025 --> 00:04:08.025
that we've measured

100
00:04:08.085 --> 00:04:09.625
and conditioned on all the things

101
00:04:09.625 --> 00:04:11.505
that might be, might matter, right?

102
00:04:11.505 --> 00:04:13.905
So we can't always distinguish

103
00:04:13.905 --> 00:04:16.305
between is this a real causal effect

104
00:04:16.365 --> 00:04:19.875
or is there just some unmeasured difference between

105
00:04:20.415 --> 00:04:22.275
the families of the adults living in low

106
00:04:22.275 --> 00:04:23.835
and high poverty neighborhoods that are accounting

107
00:04:23.835 --> 00:04:24.915
for the differences in outcomes.

108
00:04:25.695 --> 00:04:28.515
Uh, and so that's why the natural experiments,

109
00:04:28.515 --> 00:04:32.755
which moving to opportunity is one, one example,

110
00:04:33.295 --> 00:04:34.635
are, are useful.

111
00:04:34.935 --> 00:04:37.275
We leverage that variation that sort of appears

112
00:04:37.275 --> 00:04:39.955
to be random relative, to the outcomes.

113
00:04:40.015 --> 00:04:42.435
And so we don't have to worry about measuring all

114
00:04:42.435 --> 00:04:44.915
of those possible, sources of

115
00:04:44.915 --> 00:04:47.235
what we would call confounding, right,

116
00:04:47.235 --> 00:04:49.675
or selection bias as it's called in this paper.

117
00:04:50.055 --> 00:04:52.835
The Moving to Opportunity Study,

118
00:04:53.535 --> 00:04:54.875
is actually an experiment.

119
00:04:55.295 --> 00:04:58.715
Um, but what it randomizes is, the, the

120
00:04:59.525 --> 00:05:01.435
being offered a housing voucher

121
00:05:01.825 --> 00:05:04.435
that one can use in the private, rental market.

122
00:05:04.485 --> 00:05:07.035
These are called Section eight vouchers in the US context,

123
00:05:07.185 --> 00:05:09.595
started with people who lived in public housing developments

124
00:05:09.595 --> 00:05:13.675
in five US cities, and, people signed up for the study.

125
00:05:14.135 --> 00:05:17.635
And then they were randomly assigned to a control group to,

126
00:05:17.855 --> 00:05:19.835
what we call the experimental group,

127
00:05:19.965 --> 00:05:24.155
which they received a housing voucher, assistance with,

128
00:05:24.335 --> 00:05:27.555
relocation, and as well, a requirement, they moved

129
00:05:27.555 --> 00:05:28.955
to a low poverty neighborhood.

130
00:05:29.135 --> 00:05:30.275
And then there's a kind of middle group

131
00:05:30.275 --> 00:05:32.235
that gets a housing voucher, but doesn't get any

132
00:05:32.235 --> 00:05:34.075
of the assistance or the, or the requirements.

133
00:05:34.215 --> 00:05:35.875
What's this randomly assigned is the,

134
00:05:35.895 --> 00:05:37.115
is the housing voucher.

135
00:05:37.495 --> 00:05:39.275
Um, but what

136
00:05:39.275 --> 00:05:42.715
that produces is it also produces variation over time in the

137
00:05:42.715 --> 00:05:43.755
neighborhoods that people live in.

138
00:05:44.815 --> 00:05:46.235
Uh, and so it's that,

139
00:05:47.425 --> 00:05:49.475
randomization in the housing voucher that we're able

140
00:05:49.475 --> 00:05:52.995
to leverage to, to isolate variation in the neighborhoods

141
00:05:53.185 --> 00:05:54.635
that is sort of quasi random.

142
00:05:55.565 --> 00:05:58.845
So this all was done, started in the late 1990s.

143
00:05:59.065 --> 00:06:01.765
The families were recruited, into the study.

144
00:06:01.765 --> 00:06:04.485
They were followed over time in surveys,

145
00:06:04.945 --> 00:06:06.125
and administrative records.

146
00:06:06.545 --> 00:06:11.335
Um, and the PO population of the study was primarily,

147
00:06:11.355 --> 00:06:13.855
low income single mothers whose, who are, you know,

148
00:06:13.855 --> 00:06:17.335
the ones living in these, public housing developments,

149
00:06:17.515 --> 00:06:18.775
in five US cities.

150
00:06:19.275 --> 00:06:20.535
And what we, what we see.

151
00:06:20.535 --> 00:06:22.615
There's a, you know, a very long literature that use

152
00:06:22.615 --> 00:06:25.215
that studies the MTO, many, many papers written on it.

153
00:06:25.555 --> 00:06:27.855
But we do see that there are big impacts

154
00:06:27.915 --> 00:06:29.655
of being offered the vouchers, especially in

155
00:06:29.655 --> 00:06:31.735
that experimental group, on the types of neighborhoods

156
00:06:31.735 --> 00:06:33.055
that people live in later.

157
00:06:33.515 --> 00:06:37.415
So, 18 percentage point difference in the poverty rate,

158
00:06:37.635 --> 00:06:40.735
which is quite a big, a big change.

159
00:06:42.185 --> 00:06:44.565
So it's the MTO study

160
00:06:44.665 --> 00:06:48.205
and looking at the labor market outcomes of the,

161
00:06:48.865 --> 00:06:50.925
the low income mothers in the study,

162
00:06:50.945 --> 00:06:53.725
and not finding any effects on employment and earnings.

163
00:06:53.725 --> 00:06:54.885
It's sort of created this puzzle.

164
00:06:55.345 --> 00:06:57.365
Uh, and we're that we're trying to unpack,

165
00:06:57.625 --> 00:06:59.365
the study uses two data sets.

166
00:06:59.705 --> 00:07:02.525
One is the moving to opportunity dataset,

167
00:07:03.385 --> 00:07:06.725
and then the other, is the panel study

168
00:07:06.725 --> 00:07:11.165
of income dynamics, which is, one of the kind

169
00:07:11.165 --> 00:07:13.565
of workhorse studies of social sciences in the us.

170
00:07:13.865 --> 00:07:16.365
And it's been used to extensively study,

171
00:07:16.365 --> 00:07:18.125
neighborhood effects using this observational

172
00:07:18.465 --> 00:07:19.605
design that we talked about.

173
00:07:20.065 --> 00:07:24.005
And so, the idea that PSID recruited families,

174
00:07:24.025 --> 00:07:27.645
in the late 1960s, and followed them over time,

175
00:07:28.505 --> 00:07:30.565
and measured many of the same things

176
00:07:30.565 --> 00:07:32.605
that the MTO study measures,

177
00:07:32.665 --> 00:07:35.045
and also kept track of the neighborhoods,

178
00:07:35.185 --> 00:07:36.365
and neighborhood conditions that,

179
00:07:36.365 --> 00:07:37.525
that people were exposed to.

180
00:07:38.345 --> 00:07:42.385
And so the basic study design here is to say, you know,

181
00:07:42.525 --> 00:07:45.065
we have an example of the observational, data.

182
00:07:45.085 --> 00:07:48.145
We have the, the quasi or natural experimental data,

183
00:07:48.805 --> 00:07:52.465
and we're gonna try to work with those two data sets, to

184
00:07:53.345 --> 00:07:55.795
examine all the possible hypotheses about

185
00:07:55.795 --> 00:07:58.155
where this discrepancy in the results are coming from,

186
00:07:58.415 --> 00:08:00.595
and try to go through and kind of eliminate the ones that,

187
00:08:00.595 --> 00:08:03.435
that are no longer po plausible.

188
00:08:03.935 --> 00:08:05.995
One other note about the way we're using the moving

189
00:08:06.015 --> 00:08:08.795
to opportunity data is we're using what's called,

190
00:08:08.795 --> 00:08:10.555
within Study Comparison design.

191
00:08:11.295 --> 00:08:15.955
And so what we do is we take the control group

192
00:08:16.495 --> 00:08:17.955
in the Moving to Opportunity Study,

193
00:08:18.575 --> 00:08:21.795
and we treat it as if it were just an observational study.

194
00:08:21.815 --> 00:08:23.955
We ignore the fact that we, we randomized

195
00:08:23.955 --> 00:08:25.155
and we just look in that group.

196
00:08:25.895 --> 00:08:27.875
Um, and that allows us to test a bunch

197
00:08:27.875 --> 00:08:30.195
of hypotheses about possible differences,

198
00:08:30.195 --> 00:08:33.355
because now we're comparing the data from the MTO study

199
00:08:33.665 --> 00:08:37.715
collected in the way that the MTO data were, with AS,

200
00:08:37.855 --> 00:08:40.155
and analyzing it as an observational study.

201
00:08:40.295 --> 00:08:42.955
And we're comparing that with the, analysis of the

202
00:08:43.555 --> 00:08:45.155
PSID observational study.

203
00:08:45.185 --> 00:08:47.555
There's a, a, a long list of possible reasons

204
00:08:47.575 --> 00:08:50.235
of what's going on here, so I'll try to move

205
00:08:50.235 --> 00:08:52.075
through those relatively quickly.

206
00:08:52.655 --> 00:08:55.365
Um, the, the big one is this idea of selection bias

207
00:08:55.365 --> 00:08:58.285
that really, this is just a mythological artifact of the,

208
00:08:58.305 --> 00:09:00.165
the main weakness of the observational studies,

209
00:09:00.435 --> 00:09:02.565
that there's some remaining thing that we didn't control

210
00:09:02.585 --> 00:09:04.325
for in the moving to opportunity.

211
00:09:04.825 --> 00:09:07.245
Uh, on the other hand, because we have this quasi random

212
00:09:07.245 --> 00:09:08.685
variation that we can leverage

213
00:09:08.685 --> 00:09:10.765
and sort of treat like an experiment,

214
00:09:11.865 --> 00:09:13.085
it doesn't have that problem.

215
00:09:13.425 --> 00:09:16.205
And so that ends up being kind of a residual explanation,

216
00:09:16.425 --> 00:09:18.685
you know, because we can't directly

217
00:09:19.465 --> 00:09:22.165
impose a natural experiment on the observational study.

218
00:09:22.625 --> 00:09:25.285
Um, so that's, that one's sort of,

219
00:09:25.665 --> 00:09:27.725
the residual explanation that as we knock off the others,

220
00:09:27.725 --> 00:09:29.885
that one becomes, much more likely.

221
00:09:30.665 --> 00:09:32.885
So there's a couple of hypotheses that are,

222
00:09:33.535 --> 00:09:36.245
about just differences in the way the key

223
00:09:36.485 --> 00:09:37.645
variables are measured across studies.

224
00:09:37.785 --> 00:09:40.325
No, you know, no two surveys are gonna be exactly the same.

225
00:09:40.825 --> 00:09:43.525
Um, and it basically comes down to different ways

226
00:09:43.525 --> 00:09:45.805
of measuring the independent and independent variables.

227
00:09:45.805 --> 00:09:46.845
And maybe if those are different,

228
00:09:46.845 --> 00:09:49.565
that's producing different results, there's a set of,

229
00:09:50.065 --> 00:09:52.105
explanations.

230
00:09:52.105 --> 00:09:54.265
They're really about differences between

231
00:09:55.085 --> 00:09:58.625
the study composition, in the moving to opportunity data

232
00:09:58.965 --> 00:10:00.225
and the PSID data.

233
00:10:00.325 --> 00:10:03.385
So the moving to opportunity fam families are particularly,

234
00:10:04.085 --> 00:10:05.145
disadvantaged.

235
00:10:05.175 --> 00:10:07.705
They're living in public housing developments.

236
00:10:08.085 --> 00:10:10.065
Um, so that's, that wasn't surprising to see

237
00:10:10.065 --> 00:10:11.705
that was expected in the study design.

238
00:10:12.925 --> 00:10:16.745
Uh, and also the range of neighborhood poverty in which,

239
00:10:16.935 --> 00:10:19.865
there's variation in MTO study is kind of at the higher end.

240
00:10:20.325 --> 00:10:24.785
Uh, whereas the PSID study, the range of, of, of,

241
00:10:24.785 --> 00:10:27.425
of variation in neighborhood poverty rates is much broader.

242
00:10:27.545 --> 00:10:29.185
'cause it's a, a representative sample

243
00:10:29.185 --> 00:10:30.745
of the entire, us.

244
00:10:31.685 --> 00:10:35.385
So, that can sort of produce possible differences.

245
00:10:35.445 --> 00:10:38.025
So, for example, if certain types of,

246
00:10:38.385 --> 00:10:39.905
families are more susceptible to

247
00:10:40.585 --> 00:10:43.825
negative neighborhood effects and they're, you know,

248
00:10:43.895 --> 00:10:48.105
more, more common in, the PSID data, then that might,

249
00:10:48.485 --> 00:10:51.065
you know, explain, the result.

250
00:10:51.085 --> 00:10:54.745
It may be that the MTO families, many of them actually,

251
00:10:55.045 --> 00:10:57.305
you know, moved to, lower poverty neighborhoods,

252
00:10:57.325 --> 00:10:58.945
but then moved back,

253
00:10:59.045 --> 00:11:00.665
or moved to other, you know, back

254
00:11:00.665 --> 00:11:01.745
to higher poverty neighborhoods.

255
00:11:01.765 --> 00:11:04.785
So maybe the duration of their exposure to the,

256
00:11:05.085 --> 00:11:07.585
the negative, neighborhood environment wasn't,

257
00:11:07.585 --> 00:11:10.305
wasn't really long enough to, have a positive effect,

258
00:11:10.485 --> 00:11:11.545
on employment.

259
00:11:12.045 --> 00:11:16.505
Um, it could also be that, there's other aspects

260
00:11:16.505 --> 00:11:18.585
of the neighborhood besides the poverty rate that matter.

261
00:11:18.685 --> 00:11:21.345
So, one thing we see is that the moving

262
00:11:21.365 --> 00:11:22.665
to opportunity families that move

263
00:11:22.665 --> 00:11:26.745
to low poverty neighborhoods, they tended to move to,

264
00:11:27.275 --> 00:11:29.625
lower poverty neighborhoods that were, that were still sort

265
00:11:29.625 --> 00:11:32.425
of, sort of racially segregated, right?

266
00:11:32.445 --> 00:11:36.345
And opportunities also strongly patterned, by race,

267
00:11:36.365 --> 00:11:37.545
in the us the type

268
00:11:37.605 --> 00:11:39.425
and the magnitude of the, of the change in,

269
00:11:39.445 --> 00:11:41.185
in the neighborhoods that we can observe in the two data

270
00:11:41.185 --> 00:11:42.185
sets might be different.

271
00:11:43.045 --> 00:11:45.985
Uh, and then finally, you know, by the very nature

272
00:11:46.045 --> 00:11:49.065
of the moving to opportunity study design,

273
00:11:49.365 --> 00:11:50.385
the first word is moving.

274
00:11:50.485 --> 00:11:53.065
And so every family that is in the Moving

275
00:11:53.125 --> 00:11:56.925
to opportunity study has to make a residential move,

276
00:11:56.985 --> 00:11:58.965
you know, to get to a lower poverty neighborhood.

277
00:11:59.225 --> 00:12:01.685
And those residential moves could potentially be

278
00:12:01.895 --> 00:12:03.245
disruptive, right?

279
00:12:03.265 --> 00:12:05.965
So people's social networks are afraid, their source sources

280
00:12:06.025 --> 00:12:07.885
of support, the institutions that they rely on,

281
00:12:07.885 --> 00:12:09.045
they're gonna have to find new ones.

282
00:12:09.435 --> 00:12:12.805
It's not clear what their reception is gonna be in these new

283
00:12:12.805 --> 00:12:14.205
low, lower poverty neighborhoods are.

284
00:12:14.205 --> 00:12:15.405
They're gonna be able to make new social

285
00:12:15.405 --> 00:12:16.885
connections and so on.

286
00:12:16.885 --> 00:12:18.605
So maybe there's something about, you know,

287
00:12:18.605 --> 00:12:21.405
the residential moves that are happening that is also sort

288
00:12:21.405 --> 00:12:23.965
of dampening the, the potentially positive effects.

289
00:12:24.215 --> 00:12:25.365
Lemme say a little bit about

290
00:12:25.585 --> 00:12:27.645
how we actually analyze the data, right?

291
00:12:27.645 --> 00:12:30.285
So how do you deal with the fact that these two samples are

292
00:12:30.285 --> 00:12:32.525
so, are, are potentially different?

293
00:12:32.565 --> 00:12:35.685
A lot of our explanations have to do, with, with that.

294
00:12:36.575 --> 00:12:38.675
Um, so right away we were, well, lemme say right away

295
00:12:38.675 --> 00:12:40.115
that we were able to dismiss kind

296
00:12:40.115 --> 00:12:41.595
of the measurement type explanations

297
00:12:41.865 --> 00:12:45.275
because as soon as we looked in the, the,

298
00:12:45.555 --> 00:12:47.715
MTO control group and just analyzed it separately,

299
00:12:48.135 --> 00:12:51.555
we saw the same, effects as we see in the,

300
00:12:51.555 --> 00:12:52.875
in the PSID, right?

301
00:12:52.895 --> 00:12:54.235
So it can't really be the,

302
00:12:54.335 --> 00:12:55.675
the measurement that's driving it.

303
00:12:55.875 --> 00:12:59.195
'cause, we're seeing kind of the same things when we,

304
00:12:59.195 --> 00:13:01.435
when we treat the MTO study as,

305
00:13:01.535 --> 00:13:03.115
as an observational, study.

306
00:13:03.335 --> 00:13:06.915
So the other ones require us to, to do some statistical work

307
00:13:07.255 --> 00:13:10.275
to kind of line up the, the samples

308
00:13:10.455 --> 00:13:12.915
and the experiences of the people in the two data sets.

309
00:13:13.495 --> 00:13:17.115
We use, some statistical methods, related to sort

310
00:13:17.115 --> 00:13:20.875
of matching cases and re-weighting, the data,

311
00:13:20.875 --> 00:13:24.035
especially in the MTO, sample, which is much more,

312
00:13:24.065 --> 00:13:27.195
much more disadvantaged samples so that the characteristics

313
00:13:27.195 --> 00:13:30.675
of the two samples look more similar, to one another

314
00:13:30.675 --> 00:13:32.155
or look, look similar to one another.

315
00:13:32.495 --> 00:13:34.475
Uh, and, you know, the number of moves

316
00:13:34.495 --> 00:13:35.875
and the types of neighborhoods they move

317
00:13:35.875 --> 00:13:37.315
to also start to line up.

318
00:13:37.975 --> 00:13:40.675
And so we were able, to do that

319
00:13:40.815 --> 00:13:44.115
and, basically that process allowed us

320
00:13:44.375 --> 00:13:46.075
to dismiss all.

321
00:13:46.135 --> 00:13:49.355
But two of the, that long list of explanations

322
00:13:49.355 --> 00:13:51.225
that I gave you,

323
00:13:51.285 --> 00:13:56.025
and those two are the, the residual, explanation

324
00:13:56.045 --> 00:13:57.185
of selection bias.

325
00:13:57.485 --> 00:14:01.705
Um, and then also, something about,

326
00:14:02.015 --> 00:14:05.145
potentially something about, the disruptiveness

327
00:14:05.285 --> 00:14:07.625
of the moves that the moving to opportunity families,

328
00:14:07.725 --> 00:14:08.785
are experiencing.

329
00:14:09.645 --> 00:14:11.305
So one thing that we noticed

330
00:14:11.405 --> 00:14:15.665
as we were doing this analysis is that when we lined up the,

331
00:14:16.125 --> 00:14:19.385
ex the ex, the MTO,

332
00:14:20.425 --> 00:14:24.825
experimental sample with the PSID sample, we started to see

333
00:14:24.825 --> 00:14:27.145
that actually the impact of moving

334
00:14:27.205 --> 00:14:30.705
to a low poverty neighborhood in the MTO data,

335
00:14:30.845 --> 00:14:32.465
became negative, right?

336
00:14:32.525 --> 00:14:34.265
So, which is not just zero,

337
00:14:34.445 --> 00:14:36.345
but like the opposite of the direction

338
00:14:36.345 --> 00:14:38.065
that we were expecting it to see.

339
00:14:38.065 --> 00:14:40.825
And so this was kind of a puzzle no one expected really to,

340
00:14:41.205 --> 00:14:43.785
to, to see this negative effect.

341
00:14:44.605 --> 00:14:46.705
And what that made us think of is

342
00:14:46.705 --> 00:14:48.905
that perhaps when we're looking,

343
00:14:49.085 --> 00:14:53.765
at the overall MTO results, that those kind of null results,

344
00:14:53.765 --> 00:14:57.725
those zero effects were a, a, a hidden average

345
00:14:57.865 --> 00:14:59.285
of positive and negative effects.

346
00:14:59.545 --> 00:15:00.845
And that there were some people who were

347
00:15:00.845 --> 00:15:01.925
having these negative effects.

348
00:15:02.625 --> 00:15:05.125
And so we tried to drill down into the moving

349
00:15:05.125 --> 00:15:07.965
to opportunity data to understand, okay, which subgroups

350
00:15:08.025 --> 00:15:10.725
of low income mothers in this data seem

351
00:15:10.725 --> 00:15:14.955
to be experiencing these negative effects,

352
00:15:14.975 --> 00:15:17.595
of moving, to a lower poverty neighborhood on their,

353
00:15:17.735 --> 00:15:19.355
on their employment and earnings.

354
00:15:19.895 --> 00:15:24.115
And what we saw is there was this one key group that seemed

355
00:15:24.115 --> 00:15:26.635
to be driving a lot of this, this, negative effect,

356
00:15:26.635 --> 00:15:29.515
which was women who were employed at the beginning

357
00:15:29.515 --> 00:15:32.155
of the study and had,

358
00:15:32.255 --> 00:15:34.155
and especially those who had,

359
00:15:34.155 --> 00:15:35.755
less than a high school degree, right?

360
00:15:35.755 --> 00:15:39.305
So these are gonna be, people who,

361
00:15:39.365 --> 00:15:41.825
in our contemporary US labor market are gonna be most

362
00:15:41.945 --> 00:15:44.145
precariously employed, right?

363
00:15:44.405 --> 00:15:46.425
Uh, given their, given their credentials.

364
00:15:46.815 --> 00:15:50.025
There's something that was potentially,

365
00:15:50.155 --> 00:15:53.585
disruptive about these moves to this set

366
00:15:53.585 --> 00:15:55.705
of low income mothers who were employed

367
00:15:55.765 --> 00:15:57.425
and were very low education.

368
00:15:57.995 --> 00:16:00.745
Their economic trajectories were actually,

369
00:16:01.345 --> 00:16:02.825
negatively impacted at this point.

370
00:16:02.965 --> 00:16:05.905
We are really stretching the data thin, whereas zero,

371
00:16:06.005 --> 00:16:07.625
you know, cutting the data into smaller

372
00:16:07.685 --> 00:16:09.025
and smaller slices of data.

373
00:16:09.445 --> 00:16:12.105
And so we don't wanna say too strongly that

374
00:16:12.105 --> 00:16:14.425
that is entirely, what's going on.

375
00:16:14.605 --> 00:16:17.585
So I, you know, there are ways to sort of post hoc think of

376
00:16:18.185 --> 00:16:19.385
possible explanations for this.

377
00:16:19.965 --> 00:16:24.545
Um, so there been some, some other groups of researchers

378
00:16:24.565 --> 00:16:26.665
who have done qualitative like interview

379
00:16:27.085 --> 00:16:29.785
and participant observation studies of the families in MTO.

380
00:16:29.885 --> 00:16:32.105
And one of the things that they document is

381
00:16:32.105 --> 00:16:34.865
that it's actually very hard to find an apartment

382
00:16:34.865 --> 00:16:36.025
with one of these vouchers.

383
00:16:36.025 --> 00:16:37.065
This is a, a,

384
00:16:37.265 --> 00:16:41.025
a widespread problem in the us housing system.

385
00:16:41.645 --> 00:16:44.625
Um, and it, it, these families experience it as well.

386
00:16:44.805 --> 00:16:49.415
And so if it's very hard to, to get a landlord to,

387
00:16:49.755 --> 00:16:51.335
take one of these vouchers

388
00:16:51.335 --> 00:16:52.935
and to find an apartment in a,

389
00:16:52.935 --> 00:16:55.455
in a low poverty neighborhood, then, you know,

390
00:16:55.455 --> 00:16:57.095
it's gonna be very disruptive to look for this.

391
00:16:57.315 --> 00:17:00.095
Uh, and also to, to make,

392
00:17:00.655 --> 00:17:02.975
whatever new place you are living, sort of work

393
00:17:03.045 --> 00:17:04.455
with all your social supports

394
00:17:04.455 --> 00:17:05.975
and maybe childcare arrangements that you have

395
00:17:05.975 --> 00:17:07.295
that's allowing for this employment.

396
00:17:07.755 --> 00:17:10.015
Um, so that's, you know,

397
00:17:10.075 --> 00:17:11.815
one possible story one would construct,

398
00:17:11.815 --> 00:17:13.735
but that's really just a set of,

399
00:17:14.015 --> 00:17:15.175
hypotheses for future research.

400
00:17:15.235 --> 00:17:17.775
We think it's really important to be, to continue

401
00:17:17.775 --> 00:17:20.935
to look very carefully in future studies at these potential

402
00:17:20.935 --> 00:17:23.375
disruptive effects of, of residential moves,

403
00:17:23.375 --> 00:17:26.535
especially in this, a population of people that's very,

404
00:17:26.635 --> 00:17:27.935
precariously employed.

405
00:17:28.075 --> 00:17:30.175
We also think it's important to,

406
00:17:31.105 --> 00:17:35.335
focus future research on the selection process itself.

407
00:17:36.235 --> 00:17:39.215
Uh, right. So in other words, how is it,

408
00:17:39.405 --> 00:17:41.855
that some families end up in, in high poverty neighborhoods

409
00:17:41.855 --> 00:17:43.935
and some families end up in low poverty neighborhoods?

410
00:17:43.935 --> 00:17:45.935
Understanding that process will then help us

411
00:17:46.035 --> 00:17:48.775
to interpret our results about the impacts of

412
00:17:48.775 --> 00:17:49.775
Those, those neighborhoods. 'cause we'll have

413
00:17:49.775 --> 00:17:50.765
a better sense of how similar

414
00:17:50.825 --> 00:17:53.205
or different they are and, and, and why.

415
00:17:54.145 --> 00:17:56.565
Um, and then from more of a policy perspective,

416
00:17:56.665 --> 00:17:59.165
if we're thinking about the housing voucher programs

417
00:17:59.355 --> 00:18:00.965
that are in the us which are one

418
00:18:00.965 --> 00:18:02.685
of the main ways in which we have

419
00:18:02.705 --> 00:18:05.685
to both stabilize people's housing, who are,

420
00:18:05.785 --> 00:18:09.725
who are low income, and also move them to,

421
00:18:09.725 --> 00:18:12.125
neighborhoods with, with greater economic opportunity

422
00:18:12.315 --> 00:18:13.845
that those programs need

423
00:18:13.845 --> 00:18:16.645
to take very seriously the disruptive impacts of,

424
00:18:16.705 --> 00:18:18.365
of those, of those moves.

425
00:18:18.865 --> 00:18:22.365
Uh, and also, you know, as others have said, to deal

426
00:18:22.365 --> 00:18:25.565
with this sort of challenge of, of quote, leasing up with,

427
00:18:25.825 --> 00:18:28.245
one of these, one of these vouchers.

428
00:18:28.665 --> 00:18:30.005
And then finally, I would say, you know, it,

429
00:18:30.005 --> 00:18:31.965
it does illustrate the value of these types

430
00:18:31.965 --> 00:18:35.205
of natural experiments, in, in,

431
00:18:35.265 --> 00:18:38.005
in helping us understand, these, the, these sort

432
00:18:38.005 --> 00:18:38.485
of phenomenon.

The Impacts of High Poverty Neighborhoods on Social Mobility and Well-being

00:00:00 - There's a longstanding literature, in the social sciences dating back decades, that is interested in, interested in the kind of the impacts of living in high poverty neighborhoods. So both on, children, if we're thinking about social mobility and social opportunity, and also on adults, there opportunity. and it's also closely connected to concerns about, racial inequality and class inequality. and so in the US context, the combination of racial and class segregation in neighborhoods produces,

00:00:33 - of concentrated poverty neighborhoods or high poverty neighborhoods. and we're, we're very interested in the long term in, what's the impact of the people who live or grow up in these neighborhoods on their life prospects. And across a number of different domains. There's now what we call an emerging consensus across studies that there these impacts, on children are real. They're large enough to have impact on social, social mobility and, and opportunity,

00:01:04 - there's also an emerging consensus that there are mental health and physical health effects of, of living, in high poverty neighborhoods. There's still a remaining puzzle in this literature, which is that there's kind of an uncertainty or a discrepancies in research findings about the impact of living in a high poverty neighborhood on adults and their employment, and earnings prospects. So, different studies show different things.

Exploring Employment Impacts of High Poverty Neighborhoods

00:01:35 - to Opportunity study show no impact. and whereas other studies with other data sets, with other research designs show, there, there is a negative impact on things like employment and earnings, for adults, compared to similarly situated adults who live in lower poverty neighborhoods. So that's kind of the puzzle that this paper, tries to address. and the reason it's an important puzzle also is that there are, a number of social science theories

00:02:06 - to be these effects, right? So, high poverty neighborhoods tend to be, in certain geographic places in the city that are, distant from where job opportunities are. they as a, also as a result, they tend to be more socially isolated, their environments of social distrust. And we know that people often find jobs and find better jobs through social networks. There's also questions around, you know, norms around work and employment in, in high poverty neighborhoods. We know that high poverty neighborhoods actually sort

00:02:40 - that can affect your chances of employment. and then there's also a set of health pathways. You know, if, if people is mental health is impacted, that's gonna impact their ability, to, to work and to, and to hold a job. So, sort of understand the methodological puzzle here. It's important to understand a little bit about the different types of research designs that are used to study this question. so this is not a setting in which we can do an experiment. We can't randomly assign someone to live in a high poverty neighborhood

Research Designs in Neighborhood Impact Studies

00:03:10 - that would be, unethical and impractical and, and, and very expensive, as well to, to do an actual experiment. So the two types of research designs we can do are either what we call observational studies, where we're following people, you know, as they are, as they go about their lives in the world and collecting data on them. and the second type is called a natural experiment, where it sort of mimics an experiment. There's some, something out there that is producing

00:03:42 - of neighborhood, in this case. So, an observational study would be just, we follow people over time, and that will naturally, we'll naturally see people living in different types of neighborhoods. We can measure many things about them, especially things we would want to condition on or control for as we're trying to look at the impacts of the, of the neighborhoods on their, on their outcomes. but the downside of, of that observational study is we can't know that we've measured and conditioned on all the things that might be, might matter, right?

00:04:13 - between is this a real causal effect or is there just some unmeasured difference between the families of the adults living in low and high poverty neighborhoods that are accounting for the differences in outcomes. and so that's why the natural experiments, which moving to opportunity is one, one example, are, are useful. We leverage that variation that sort of appears to be random relative, to the outcomes. And so we don't have to worry about measuring all

Exploring the Moving to Opportunity Study: Housing Vouchers and Neighborhood Impact

00:04:44 - what we would call confounding, right, or selection bias as it's called in this paper. The Moving to Opportunity Study, is actually an experiment. but what it randomizes is, the, the being offered a housing voucher that one can use in the private, rental market. These are called Section eight vouchers in the US context, started with people who lived in public housing developments in five US cities, and, people signed up for the study.

00:05:17 - what we call the experimental group, which they received a housing voucher, assistance with, relocation, and as well, a requirement, they moved to a low poverty neighborhood. And then there's a kind of middle group that gets a housing voucher, but doesn't get any of the assistance or the, or the requirements. What's this randomly assigned is the, is the housing voucher. but what that produces is it also produces variation over time in the neighborhoods that people live in.

00:05:47 - randomization in the housing voucher that we're able to leverage to, to isolate variation in the neighborhoods that is sort of quasi random. So this all was done, started in the late 1990s. The families were recruited, into the study. They were followed over time in surveys, and administrative records. and the PO population of the study was primarily, low income single mothers whose, who are, you know, the ones living in these, public housing developments,

Impacts of Housing Voucher Programs on Neighborhoods and Employment Outcomes

00:06:19 - And what we, what we see. There's a, you know, a very long literature that use that studies the MTO, many, many papers written on it. But we do see that there are big impacts of being offered the vouchers, especially in that experimental group, on the types of neighborhoods that people live in later. So, 18 percentage point difference in the poverty rate, which is quite a big, a big change. So it's the MTO study and looking at the labor market outcomes of the,

00:06:50 - and not finding any effects on employment and earnings. It's sort of created this puzzle. and we're that we're trying to unpack, the study uses two data sets. One is the moving to opportunity dataset, and then the other, is the panel study of income dynamics, which is, one of the kind of workhorse studies of social sciences in the us. And it's been used to extensively study, neighborhood effects using this observational design that we talked about.

00:07:24 - in the late 1960s, and followed them over time, and measured many of the same things that the MTO study measures, and also kept track of the neighborhoods, and neighborhood conditions that, that people were exposed to. And so the basic study design here is to say, you know, we have an example of the observational, data. We have the, the quasi or natural experimental data, and we're gonna try to work with those two data sets, to

Analyzing Discrepancies: Methodological Approaches in Neighborhood Impact Studies

00:07:55 - where this discrepancy in the results are coming from, and try to go through and kind of eliminate the ones that, that are no longer po plausible. One other note about the way we're using the moving to opportunity data is we're using what's called, within Study Comparison design. And so what we do is we take the control group in the Moving to Opportunity Study, and we treat it as if it were just an observational study. We ignore the fact that we, we randomized

00:08:25 - and that allows us to test a bunch of hypotheses about possible differences, because now we're comparing the data from the MTO study collected in the way that the MTO data were, with AS, and analyzing it as an observational study. And we're comparing that with the, analysis of the PSID observational study. There's a, a, a long list of possible reasons of what's going on here, so I'll try to move through those relatively quickly. the, the big one is this idea of selection bias

00:08:58 - the main weakness of the observational studies, that there's some remaining thing that we didn't control for in the moving to opportunity. on the other hand, because we have this quasi random variation that we can leverage and sort of treat like an experiment, it doesn't have that problem. And so that ends up being kind of a residual explanation, you know, because we can't directly impose a natural experiment on the observational study. so that's, that one's sort of,

Measuring Neighborhood Effects: Variability in Study Designs

00:09:27 - that one becomes, much more likely. So there's a couple of hypotheses that are, about just differences in the way the key variables are measured across studies. No, you know, no two surveys are gonna be exactly the same. and it basically comes down to different ways of measuring the independent and independent variables. And maybe if those are different, that's producing different results, there's a set of, explanations. They're really about differences between the study composition, in the moving to opportunity data

00:10:00 - So the moving to opportunity fam families are particularly, disadvantaged. They're living in public housing developments. so that's, that wasn't surprising to see that was expected in the study design. and also the range of neighborhood poverty in which, there's variation in MTO study is kind of at the higher end. whereas the PSID study, the range of, of, of, of variation in neighborhood poverty rates is much broader. 'cause it's a, a representative sample

00:10:31 - So, that can sort of produce possible differences. So, for example, if certain types of, families are more susceptible to negative neighborhood effects and they're, you know, more, more common in, the PSID data, then that might, you know, explain, the result. It may be that the MTO families, many of them actually, you know, moved to, lower poverty neighborhoods, but then moved back, or moved to other, you know, back

Impacts of Neighborhood Moves on Employment Outcomes

00:11:01 - So maybe the duration of their exposure to the, the negative, neighborhood environment wasn't, wasn't really long enough to, have a positive effect, on employment. it could also be that, there's other aspects of the neighborhood besides the poverty rate that matter. So, one thing we see is that the moving to opportunity families that move to low poverty neighborhoods, they tended to move to, lower poverty neighborhoods that were, that were still sort of, sort of racially segregated, right?

00:11:36 - in the us the type and the magnitude of the, of the change in, in the neighborhoods that we can observe in the two data sets might be different. and then finally, you know, by the very nature of the moving to opportunity study design, the first word is moving. And so every family that is in the Moving to opportunity study has to make a residential move, you know, to get to a lower poverty neighborhood. And those residential moves could potentially be disruptive, right?

00:12:06 - of support, the institutions that they rely on, they're gonna have to find new ones. It's not clear what their reception is gonna be in these new low, lower poverty neighborhoods are. They're gonna be able to make new social connections and so on. So maybe there's something about, you know, the residential moves that are happening that is also sort of dampening the, the potentially positive effects. Lemme say a little bit about how we actually analyze the data, right? So how do you deal with the fact that these two samples are so, are, are potentially different? A lot of our explanations have to do, with, with that.

Statistical Analysis of MTO Control Group Consistency

00:12:38 - that we were able to dismiss kind of the measurement type explanations because as soon as we looked in the, the, MTO control group and just analyzed it separately, we saw the same, effects as we see in the, in the PSID, right? So it can't really be the, the measurement that's driving it. 'cause, we're seeing kind of the same things when we, when we treat the MTO study as, as an observational, study. So the other ones require us to, to do some statistical work

00:13:10 - and the experiences of the people in the two data sets. We use, some statistical methods, related to sort of matching cases and re-weighting, the data, especially in the MTO, sample, which is much more, much more disadvantaged samples so that the characteristics of the two samples look more similar, to one another or look, look similar to one another. and, you know, the number of moves and the types of neighborhoods they move to also start to line up.

00:13:40 - and, basically that process allowed us to dismiss all. But two of the, that long list of explanations that I gave you, and those two are the, the residual, explanation of selection bias. and then also, something about, potentially something about, the disruptiveness of the moves that the moving to opportunity families, are experiencing.

Unexpected Negative Impacts of Neighborhood Change

00:14:11 - as we were doing this analysis is that when we lined up the, ex the ex, the MTO, experimental sample with the PSID sample, we started to see that actually the impact of moving to a low poverty neighborhood in the MTO data, became negative, right? So, which is not just zero, but like the opposite of the direction that we were expecting it to see. And so this was kind of a puzzle no one expected really to,

00:14:44 - And what that made us think of is that perhaps when we're looking, at the overall MTO results, that those kind of null results, those zero effects were a, a, a hidden average of positive and negative effects. And that there were some people who were having these negative effects. And so we tried to drill down into the moving to opportunity data to understand, okay, which subgroups of low income mothers in this data seem to be experiencing these negative effects,

00:15:17 - on their employment and earnings. And what we saw is there was this one key group that seemed to be driving a lot of this, this, negative effect, which was women who were employed at the beginning of the study and had, and especially those who had, less than a high school degree, right? So these are gonna be, people who, in our contemporary US labor market are gonna be most precariously employed, right? given their, given their credentials.

Challenges of Relocation for Low-Income Employed Mothers

00:15:50 - disruptive about these moves to this set of low income mothers who were employed and were very low education. Their economic trajectories were actually, negatively impacted at this point. We are really stretching the data thin, whereas zero, you know, cutting the data into smaller and smaller slices of data. And so we don't wanna say too strongly that that is entirely, what's going on. So I, you know, there are ways to sort of post hoc think of

00:16:19 - so there been some, some other groups of researchers who have done qualitative like interview and participant observation studies of the families in MTO. And one of the things that they document is that it's actually very hard to find an apartment with one of these vouchers. This is a, a, a widespread problem in the us housing system. and it, it, these families experience it as well. And so if it's very hard to, to get a landlord to,

00:16:51 - and to find an apartment in a, in a low poverty neighborhood, then, you know, it's gonna be very disruptive to look for this. and also to, to make, whatever new place you are living, sort of work with all your social supports and maybe childcare arrangements that you have that's allowing for this employment. so that's, you know, one possible story one would construct, but that's really just a set of, hypotheses for future research. We think it's really important to be, to continue to look very carefully in future studies at these potential

Researching Neighborhood Impacts on Employment

00:17:23 - especially in this, a population of people that's very, precariously employed. We also think it's important to, focus future research on the selection process itself. right. So in other words, how is it, that some families end up in, in high poverty neighborhoods and some families end up in low poverty neighborhoods? Understanding that process will then help us to interpret our results about the impacts of Those, those neighborhoods. 'cause we'll have a better sense of how similar or different they are and, and, and why.

00:17:56 - if we're thinking about the housing voucher programs that are in the us which are one of the main ways in which we have to both stabilize people's housing, who are, who are low income, and also move them to, neighborhoods with, with greater economic opportunity that those programs need to take very seriously the disruptive impacts of, of those, of those moves. and also, you know, as others have said, to deal with this sort of challenge of, of quote, leasing up with,

00:18:38 - And then finally, I would say, you know, it, it does illustrate the value of these types of natural experiments, in, in, in helping us understand, these, the, these sort of phenomenon.

Faculti AI can make mistakes

Research

Policy

Analysis