Welfare Consequences of Neighborhood

Demographic Changes

Franklin Qian, Stanford University

IO Workshop

April 19, 2018

1
Motivation

Changes in neighborhood demographic makeup affect welfare of

incumbents and in-migrants

• Gentrification debate: newly rich neighborhoods harm welfare of

displaced poor

Two distinct channels for welfare consequences with different implications

for efficiency

• Price effect: rich in-migrants price poor incumbents out

• Preferences for neighborhood demographic mix: externalities imposed
by in-migrants

Design public policies that mitigate displacement and promote equality

• Government affordable housing projects, zoning laws

2
This Paper

1. Descriptive/Reduce-form evidence on how gentrification and other

neighborhood demographic changes affect outcomes of incumbents

2. Estimate housing demand elasticity and preferences for neighborhood

demographic mix by different demographic groups
• Unified neighborhood and city choice model as part of a spatial
equilibrium model, building on (Bayer, Ferreira & McMillan (2007), Diamond
(2016))

• New instrument for changes in local rents and neighborhood demographic

mix

3. Given people’s preferences, how do gentrification and other

neighborhood demographic changes affect welfare of incumbents and
in-migrants

4. Evaluate distributional and welfare consequences of affordable

housing policies (e.g. LIHTC)

3
Plan Today

1. Propose a measure of neighborhood gentrification and implement in SF

• Introduce reduced-form framework for studying impact of gentrification on
incumbents’ outcomes

2. Estimate housing demand elasticity and preferences for neighborhood

demographic mix by different demographic groups
• Unified neighborhood and city choice model as part of a spatial
equilibrium model, building on (Bayer, Ferreira & McMillan (2007), Diamond
(2016))

• New instrument for changes in local rents and neighborhood demographic

mix

4
Outline

Introduction

Data

Descriptive Evidence

Model

Estimation

5
Data

Public Use U.S. Census

• 1990, 2000 & 2010 U.S. Censuses and ACS
• Both at census tract level and micro level

Geocoded PSID
• 1990-2016 annual panel at individual and family level
• Individual migration history at block and tract level
• Detailed demographics: education, income, employment, housing
characteristics, etc

Infutor (In progress)

• Entire address history of U.S. population between 1990-2016
• Exact street address, dates of residence, name of individual, age, and
gender
• Aggregated from credit headers, car registrations, voter registrations,
and marketing data
• Imputing race and citizenship
6
Outline

Introduction

Data

Descriptive Evidence

Model

Estimation

7
What is Gentrification

“replacement of low-income, inner-city working class residents by middle- or

upper-class households, either through the market for existing housing or
demolition to make way for new upscale housing construction.” – Hammel and
Wyly (1996)

“process by which a neighborhood occupied by lower-income households

undergoes revitalization or reinvestment through the arrival of
upper-income households” – U.S. Department of Housing and Urban Development (1979)

8
Characteristics of Gentrifying Neighborhoods

Neighborhoods with potential to gentrify:

Process of gentrification:

9
Characteristics of Gentrifying Neighborhoods

Neighborhoods with potential to gentrify:

1. Central city neighborhoods

Process of gentrification:

9
Characteristics of Gentrifying Neighborhoods

Neighborhoods with potential to gentrify:

1. Central city neighborhoods

2. Populated by low-income households

Process of gentrification:

9
Characteristics of Gentrifying Neighborhoods

Neighborhoods with potential to gentrify:

1. Central city neighborhoods

2. Populated by low-income households
3. Disinvestment

Process of gentrification:

9
Characteristics of Gentrifying Neighborhoods

Neighborhoods with potential to gentrify:

1. Central city neighborhoods

2. Populated by low-income households
3. Disinvestment

Process of gentrification:

4. Influx of relatively affluent households

9
Characteristics of Gentrifying Neighborhoods

Neighborhoods with potential to gentrify:

1. Central city neighborhoods

2. Populated by low-income households
3. Disinvestment

Process of gentrification:

4. Influx of relatively affluent households

5. Increase in investment

9
Operationalize Gentrification Measure

Similar to Freeman (2005) and Governing.com (2015), use 1990 & 2000 Censuses
and 2008–2012 ACS at census tract level

Neighborhoods with potential to gentrify:

Gentrified in next decade if:

10
Operationalize Gentrification Measure

Similar to Freeman (2005) and Governing.com (2015), use 1990 & 2000 Censuses
and 2008–2012 ACS at census tract level

Neighborhoods with potential to gentrify:

1. Census tract located in central city at beginning of decade with ≥ 500

residents (in central city)

Gentrified in next decade if:

10
Operationalize Gentrification Measure

Similar to Freeman (2005) and Governing.com (2015), use 1990 & 2000 Censuses
and 2008–2012 ACS at census tract level

Neighborhoods with potential to gentrify:

1. Census tract located in central city at beginning of decade with ≥ 500

residents (in central city)
2. Tract’s median household income in bottom 40th percentile in its
metro area at beginning of decade (low-income incumbents)

Gentrified in next decade if:

10
Operationalize Gentrification Measure

Similar to Freeman (2005) and Governing.com (2015), use 1990 & 2000 Censuses
and 2008–2012 ACS at census tract level

Neighborhoods with potential to gentrify:

1. Census tract located in central city at beginning of decade with ≥ 500

residents (in central city)
2. Tract’s median household income in bottom 40th percentile in its
metro area at beginning of decade (low-income incumbents)
3. Tract’s median home value in bottom 40th percentile in its metro area
at beginning of decade (disinvestment)

Gentrified in next decade if:

10
Operationalize Gentrification Measure

Similar to Freeman (2005) and Governing.com (2015), use 1990 & 2000 Censuses
and 2008–2012 ACS at census tract level

Neighborhoods with potential to gentrify:

1. Census tract located in central city at beginning of decade with ≥ 500

residents (in central city)
2. Tract’s median household income in bottom 40th percentile in its
metro area at beginning of decade (low-income incumbents)
3. Tract’s median home value in bottom 40th percentile in its metro area
at beginning of decade (disinvestment)

Gentrified in next decade if:

4. Increase in tract’s college share in top tercile in its metro area (affluent
in-migrants)

10
Operationalize Gentrification Measure

Similar to Freeman (2005) and Governing.com (2015), use 1990 & 2000 Censuses
and 2008–2012 ACS at census tract level

Neighborhoods with potential to gentrify:

1. Census tract located in central city at beginning of decade with ≥ 500

residents (in central city)
2. Tract’s median household income in bottom 40th percentile in its
metro area at beginning of decade (low-income incumbents)
3. Tract’s median home value in bottom 40th percentile in its metro area
at beginning of decade (disinvestment)

Gentrified in next decade if:

4. Increase in tract’s college share in top tercile in its metro area (affluent
in-migrants)
5. Tract’s real median home value increased and percentage increase in
top tercile in its metro area (investment)

10
SF Gentrification Map 1990-2000

Realtor District
11
SF Gentrification Map 2000-2010

12
Extent of Gentrification for Neighborhoods in SF

Others Non-gentrified Gentrified Total

1990-2000 158 34 5 197
(80.2%) (17.3%) (2.5%)
2000-2010 154 33 10 197
(78.2%) (16.8%) (5.1%)

13
Characteristics of Gentrifying Neighborhoods in SF

Others Non-gentrified Gentrified

∆Median Home Value 19.6% 37.7% 133.1%
∆College Share 8.7 6.3 15.0
∆Population 10.5% 8.2% 0.6%
∆Median Household Income 16.4% 17.8% 31.7%
∆Median Rent 13.2% 15.8% 22.6%
∆Homeowner Share 0.7 0.9 5.3
∆Poverty Rate -0.1 -1.1 -4.3
∆White Non-Hispanic Share -2.1 -1.2 4.9

14
Impact of Gentrification on Incumbents’ Outcomes: Literature

Debate about displacement

• No displacement: Vigdor (2002), Freeman and Braconi (2004), Freeman (2005),

McKinnish et al. (2010)

• Displacement: Le Gates & Hartman (1981 & 1986), Atkinson (2003)

Improved amenities such as better schools & lower crime: Florida (2002),
Freeman (2006), McKinnish et al. (2010), Autor et al. (2017), Couture & Handbury
(2017)

Lost of neighborhood capital, loss of housing and business closure: Chernoff

(1980), Zukin (1987)

15
Impact of Gentrification on Incumbents’ Outcomes

Use geocoded PSID (Infutor later) to test following hypothesis:

1. All else equal, incumbent residents in gentrifying neighborhoods more

likely to
• Be displaced
• Move to neighborhoods with worse amenities
• Negatively impacted in their financial outcomes

2. Incumbent poor renters in gentrifying neighborhoods are impacted

more

16
Impact of Gentrification on Incumbents’ Outcomes

Yit = αi + βt Ti + γXit + δst + ηzt + εit

• Yit : Outcome of interest for household head, e.g. whether moved or

displaced
• αi : Individual FEs
• βt : Impact of gentrification in year t
• Ti : Treatment is 1 if tract resided in at beginning of decade gentrified; 0
if tract eligible but did not gentrify
• Xit : individual demographic and housing characteristics
• δst : Years at address at beginning of decade×year FEs
• ηzt : Zipcode of initial address×year FEs

17
Measure Incumbents’ Outcomes

Outcome of interest for household head Yit

• Displacement:
• Whether moved in year t: mobility not explained by life-cycle
factors are gentrification induced
• Displacement reported by respondents in PSID: likely overstate
displacement
• Within neighborhood move / Distance moved
• Eviction

• Neighborhood quality: home value, household income, college share,

etc.

• Financial outcome: assets, debts, income, spending broken into

detailed categories

18
Outline

Introduction

Data

Descriptive Evidence

Model

Estimation

19
Model: Overview

• Consider questions jointly through lens of a structural spatial

equilibrium model

• Demand system for neighborhoods and cities: determines worker’s

choice of neighborhood & city

• Model of local labor markets: link between demographic mix and local
wages

• Model of local housing market: link between demographic mix and

local rents

20
Model: Worker Preferences

In year t, worker i of demographic group k chooses which census tract m in

city j to live in by maximizing Cobb-Douglas utility function:
max ζ ln (H) + (1 − ζ) ln (Z) + si (Amjt )
H,Z

s.t. Pt Z + Rmjt H ≤ Wjt

• Worker inelastically supplies one unit of labor and earns wage Wjt
• Consumes local good H with local price Rmjt and freely-traded national
good Z with national price Pt
( )
• Gains utility from local amenities Amjt = xAmjt , amjt , τimjt , xstmj

21
Model: Worker Preferences

In year t, worker i of demographic group k chooses which census tract m in

city j to live in by maximizing Cobb-Douglas utility function:
max ζ ln (H) + (1 − ζ) ln (Z) + si (Amjt )
H,Z

s.t. Pt Z + Rmjt H ≤ Wjt

• Worker inelastically supplies one unit of labor and earns wage Wjt
• Consumes local good H with local price Rmjt and freely-traded national
good Z with national price Pt
( )
• Gains utility from local amenities Amjt = xAmjt , amjt , τimjt , xstmj
1. Exogenous amenities: xAmjt (weather, oceans, etc.)

21
Model: Worker Preferences

In year t, worker i of demographic group k chooses which census tract m in

city j to live in by maximizing Cobb-Douglas utility function:
max ζ ln (H) + (1 − ζ) ln (Z) + si (Amjt )
H,Z

s.t. Pt Z + Rmjt H ≤ Wjt

• Worker inelastically supplies one unit of labor and earns wage Wjt
• Consumes local good H with local price Rmjt and freely-traded national
good Z with national price Pt
( )
• Gains utility from local amenities Amjt = xAmjt , amjt , τimjt , xstmj
1. Exogenous amenities: xAmjt (weather, oceans, etc.)
2. Endogenous amenities:
 ( )   
Nyoung
mjt Nwhite
mjt ( )
amjt = ln , ln   , ln Ymjt 
Nmature
mjt Nminority
mjt

21
Model: Worker Preferences

In year t, worker i of demographic group k chooses which census tract m in

city j to live in by maximizing Cobb-Douglas utility function:
max ζ ln (H) + (1 − ζ) ln (Z) + si (Amjt )
H,Z

s.t. Pt Z + Rmjt H ≤ Wjt

• Worker inelastically supplies one unit of labor and earns wage Wjt
• Consumes local good H with local price Rmjt and freely-traded national
good Z with national price Pt
( )
• Gains utility from local amenities Amjt = xAmjt , amjt , τimjt , xstmj
1. Exogenous amenities: xAmjt (weather, oceans, etc.)
2. Endogenous amenities:
 ( )   
Nyoung
mjt Nwhite
mjt ( )
amjt = ln , ln   , ln Ymjt 
Nmature
mjt Nminority
mjt

3. Commuting cost: τimjt

21
Model: Worker Preferences

In year t, worker i of demographic group k chooses which census tract m in

city j to live in by maximizing Cobb-Douglas utility function:
max ζ ln (H) + (1 − ζ) ln (Z) + si (Amjt )
H,Z

s.t. Pt Z + Rmjt H ≤ Wjt

• Worker inelastically supplies one unit of labor and earns wage Wjt
• Consumes local good H with local price Rmjt and freely-traded national
good Z with national price Pt
( )
• Gains utility from local amenities Amjt = xAmjt , amjt , τimjt , xstmj
1. Exogenous amenities: xAmjt (weather, oceans, etc.)
2. Endogenous amenities:
 ( )   
Nyoung
mjt Nwhite
mjt ( )
amjt = ln , ln   , ln Ymjt 
Nmature
mjt Nminority
mjt

3. Commuting cost: τimjt

4. Birthplace amenity: xst
mj
21
Model: Worker Preferences

In year t, worker i of demographic group k chooses which census tract m in

city j to live in by maximizing Cobb-Douglas utility function:
max ζ ln (H) + (1 − ζ) ln (Z) + si (Amjt )
H,Z

s.t. Pt Z + Rmjt H ≤ Wjt

• Worker inelastically supplies one unit of labor and earns wage Wjt

• Consumes local good H with local price Rmjt and freely-traded national
good Z with national price Pt
( )
• Gains utility from local amenities Amjt = xAmjt , amjt
1. Exogenous amenities: xAmjt (weather, oceans, etc.)
2. Endogenous amenities:
 ( )   
Nyoung
mjt Nwhite
mjt ( )
amjt = ln , ln   , ln Ymjt 
Nmature
mjt Nminority
mjt

22
Model: Worker Preferences

Worker i’s Indirect utility function:

( ) ( )
Wjt Rmjt
Vimjt = ln + ζ ln + si (Amjt )
Pt Pt
= wjt − ζrmjt + si (Amjt )

Worker i’s value of amenities Amjt :

si (Amjt ) = xAmjt βix + amjt βia + σi εikbmjt

| {z } | {z } | {z }
Exogenous Endogeneous Idiosyncratic

εikbmjt ∼ Type I Extreme Value

( x a )
• βi , βi , σi : heterogeneous by race × age

23
Outline

Introduction

Data

Descriptive Evidence

Model

Estimation

24
Estimation: Labor Supply to Neighborhoods & Cities

• Change in mean utility for tract m in city j across decades for worker i
of demographic group k:
( )
^ zi
∆ln skmjt = (∆wjt − ζ∆rmjt ) βiw + ∆amjt βia + ∆ξmjt
z
i
∆ξmjt ≡ ∆xAmjt βix

• To identify workers’ preferences for wages, rents and neighborhood

demographic mixes, need variation in neighborhood demographic
z
makeup uncorrelated with ∆ξmjt
• Instrument for ∆rmjt and ∆amjt by imputing exogenous housing demand
shocks
• Instrument for ∆wjt with Bartik demand shocks

25
Instruments for ∆rmjt and ∆amjt : City with Two Neighborhoods

White in-migrants choose

{
2 1
A : 0.8 × 3
+ 0.2 × 3
= 60%
1 2
B : 0.8 × 3
+ 0.2 × 3
= 40%

Minority in-migrants choose

{
2 1
A : 0.2 × 3
+ 0.8 × 3
= 40%
1 2
B : 0.2 × 3
+ 0.8 × 3
= 60%

26
Instruments for ∆rmjt and ∆amjt

• Measure exogenous changes in neighborhood housing demand for

demographic group k:
∑
\ k
ϕbmj,1990 · φkb k
∆N mjt = 1990 · ∆Nj,−m,t
b

• ϕbmj,1990 : share of housing units of type b in tract m city j in year

1990
• type: year built × usecode × tenure

• φkb
1990 : nation-wide share of workers of demographic group k who
choose to live in house type b in year 1990
• ∆Nkj,−m,t : city-wide population change by demographic group k
during each decade, excluding workers in tract m

Yearbuilt Usecode Tenure First Stage

27
Estimation: Threats to Identification

• Neighborhood initial composition of house types might correlate with

local characteristics, which could correlate with local housing supply
shocks

• MSA wide migration might correlate with unobserved local

characteristics that drive local housing supply shocks
• Exclude own neighborhood’s contribution
• Robust to using 50% of tracts randomly drawn in MSA
• Use city-level Bartik shocks as instrument go

• Use migration magnifier instrument (Howard 2017) go

• Use regional or national rural to urban migration

28
Instrument for ∆wjt

• Use Bartik demand shocks (Bartik 1990) to instrument for city-wide wage
growth ∆wjt :
∑ Nl,j,1990
∆Bjt = (wl,−j,t − wl,−j,1990 )
Nj,1990
l

• wl,−j,t : average log wage of workers in industry l in year t,

excluding workers in city j
• Nl,j,1990 : number of workers employed in industry l in city j in year
1990

• Initial industry shares might correlate with local characteristics that

drive supply shocks
• Residualize ∆Bjt on initial local characteristics

First Stage Relation with Local Characteristics

29
Parameter Estimates: Worker Labor Supply

Table 1: GMM Model Estimates: Year FE

Worker Preferences for Neighborhoods

White Minority
Young Mature Young Mature
Wage 11.561*** 5.018*** 4.107*** 2.370***
(1.132) (0.528) (0.344) (0.219)

Rent -4.538*** -0.729** -3.733*** -1.030***

(0.593) (0.297) (0.243) (0.190)

Expenditure Share 0.393*** 0.145*** 0.909*** 0.435***

(0.028) (0.047) (0.087) (0.083)

Young/Old 0.382*** -0.928*** 1.050*** -0.484***

(0.096) (0.054) (0.068) (0.051)

White/Minority -1.110*** -0.731*** -1.547*** -1.160***

(0.091) (0.042) (0.031) (0.022)

Household Income -5.744*** -2.612*** 2.117*** 1.301***

(0.860) (0.420) (0.229) (0.130)
Observations 73,190 73,170 70,781 70,780
Hansen J (p-value) 0.116 0.229 0 0

30
Variable Definition Fixed Exp Share
Conclusions

• Proposed a measure of extent of neighborhood gentrification

• A spatial equilibrium model that links local wages, rents and

demographic composition to worker’s neighborhood choice
• Estimated migration elasticity wrt. wages, rents and neighborhood
demographic mix ratios

Next steps:

• Extent gentrification to top 50 cities in U.S.

• Reduced form evidence of impact of gentrification on incumbents
• Incorporate Infutor panel to estimate model
• Enrich labor demand model and add commuting model
• Counterfactual exercise on sorting behavior under government
affordable housing projects

31
Thank You!

32
SF Neighborhood Map

Source: Paragon real estate group

0
SF Gentrification Map
 Bu Share
ilt
19 0 .1 .2 .3
39
Ea
rli
er

Bu
ilt
19
40
s

Bartik Housing Demand

Bu
ilt
19
50
s

Young
Bu
ilt
19
60
s
Bu
ilt

(a) By Age Group

19
70

Mature
Bu s
ilt
19
80
La
te
r

Bu Share
ilt 0 .05 .1 .15 .2 .25
19
39
Ea
rli
er

Bu
ilt
19
1990 Nation-Wide Preferences for Year Built of Houses

40
s
Bu
White, Alone

ilt
19
50
s
Bu
ilt
19
60
Black

s
(b) By Race

Bu
ilt
19
70
Bu s
ilt
19
80
La
te
r
White, Hispanic

1
1990 Nation-Wide Preferences for Usecode of Houses

.8

.8
.6

.6
Share

Share
.4

.4
.2

.2
0

0
r y ily ily r y ily ily
he mil he mil
Ot Fa am am Ot Fa am am
gle lti F lti F gle ltiF
ult
iF
Sin ll- mu -m
u
Sin mu -m
ma
e all- e
S L arg Sm Larg

Young Mature White, Alone Black White, Hispanic

(a) By Age Group (b) By Race

Bartik Housing Demand

2
1990 Nation-Wide Preferences for Tenure of Houses

.8

.8
.6

.6
Share

Share
.4

.4
.2

.2
0

0
d d d d
pie pie pie pie
ccu ccu Oc
cu ccu
rO rO er rO
ne nte n nte
Ow Re Ow Re

Young Mature White, Alone Black White, Hispanic

(a) By Age Group (b) By Race

Bartik Housing Demand

3
Systematic Preferences for House Types Across Cities

β = 0.912 (0.011) β = 0.845 (0.018)

.2

.2
2 2
R = 0.988 R = 0.921
Obs = 96 Obs = 192
.15

.15
βkb 1st half

βkb 1st half

.1

.1
.05

.05
0

0
0 .05 .1 .15 .2 0 .05 .1 .15 .2
βkb 2nd half βkb 2nd half

95% CI Linear Fitted Value 95% CI Linear Fitted Value

(a) By Age Group (b) By Race

Bartik Housing Demand

4
 ( )
First Stage for ln Young/Mature

Bartik Housing Demand

5
 ( )
First Stage for ln White/Minority

Bartik Housing Demand

6
First Stage for ln (Hhld Income)

Bartik Housing Demand

7
Bartik Labor Demand Shocks as Instrument for City-Level
Migration

• Following Notowidigdo (2013), predicted city-wide employment change

∆\
ln (Njt ):
∑ Nl,j,1990 ( νl,−j,t − νl,−j,1990 )
∆\
ln (Njt ) =
Nj,1990 νl,−j,1990
l

• νl,−j,t : national employment share of industry l in year t, excluding

workers in city j
• Nl,j,1990 : number of workers employed in industry l in city j in year
1990

Identification Threats

8
Shift-Share Labor Supply Shocks as Instrument for City-Level
Migration

• Following Altonji and Card (1991) & Howard (2017), predicted city-level
djt :
inmigration ∆N
∑ mc→j,1990
djt =
∆N · mc→−j,t
mc→−j,1990
c∈−j

• −j is the set of all counties that are sufficiently far from j

• mc→j,t : migration from county c to city j in year t

Identification Threats

9
First Stage for ln (Wage)

Bartik Labor Demand

10
Relationship between Bartik Demand Shocks and Local Characteristics

(1) (2)
Bartik 2000 Bartik 2010
Male -0.015 0.043
(0.012) (0.029)
White -0.018∗∗∗ -0.057∗∗∗
(0.004) (0.010)
College 0.025∗∗∗ 0.083∗∗∗
(0.006) (0.014)
Immigrant -0.002 -0.027∗∗
(0.005) (0.013)
Mature -0.035∗∗∗ -0.145∗∗∗
(0.011) (0.026)
Constant 0.361∗∗∗ 0.630∗∗∗
(0.010) (0.024)
N 238 238
R2 0.307 0.420

Bartik Labor Demand

11
Definition of Variables

Variable Sample Notes

Metropolitan All MSAs identified across all 3 censuses. Rural areas of each
Statistical Area state are excluded.
Census Tracts Census tracts in all non-rural MSAs identifiied across all 3 cen- Boundaries are standardized to 2010 bound-
suses aries.
Wages by De- All individuals with no business or farm income ages 25-64 Local wages in each MSA are averages of
mographic working at least 35 hours per week and 48 weeks per year and workers of each demographic group living in
Group earn no business or farm income. each city. Demographic groups are combina-
tions of race and young/mature age group of
workers
Local Housing Renter-occuplied housing units paying cash rent in each census Local rents in each census tract are medium
Rents tract gross rent.
Local All individuals ages 25-64. A young worker is a worker who ages 25-40, a
Young/Mature mature worker is a worker who ages 41-64.
Ratio
Local All individuals ages 25-64. A worker is classified as white if he is white,
White/Minority non-hispanic. A minority worker is defined as
Ratio either black or white, hispanic.
Local Medium All households
Household
Income
Local House All housing units House types are combinations of the year
Type Shares built decade, tenure and usecode of housing
units
Nation Wide All individuals living in non-rural MSAs with no business or farm
Preferences for income ages 25-64 working at least 35 hours per week and 48
House Types by weeks per year and earn no business or farm income.
Demographic
Group
City Wide Net All individuals ages 25-64. When measuring Bartik local housing de-
Migration by mand shock for a given census tract, the pop-
Demographic ulation of that census tract is removed.
Group

12
Parameter Estimates
Parameter Estimates: Worker Labor Supply

Table 2: GMM Model Estimates: Year FE, Fixed Local Expenditure Share

Worker Preferences for Neighborhoods

White Minority
Young Mature Young Mature
Wage 5.248*** 2.468*** 4.957*** 2.162***
(0.250) (0.173) (0.277) (0.195)

Rent -3.254*** -1.530*** -3.074*** -1.340***

(0.155) (0.107) (0.172) (0.121)

Expenditure Share 0.62 0.62 0.62 0.62

- - - -

Young/Old 0.448*** -0.771*** 1.046*** -0.443***

(0.049) (0.036) (0.067) (0.047)

White/Minority -0.729*** -0.576*** -1.609*** -1.130***

(0.034) (0.024) (0.026) (0.017)

Household Income -0.729*** -0.395*** 1.320*** 1.525***

(0.125) (0.091) (0.121) (0.069)
Observations 70,775 70,757 70,781 70,780
Hansen J (p-value) 0 0 0 0
13
Parameter Estimates