Responding to Natural Hazards: The Effects of Disaster on Residential Location Decisions and Health Outcomes

Size: px

Start display at page:

Download "Responding to Natural Hazards: The Effects of Disaster on Residential Location Decisions and Health Outcomes"

Vernon Parsons
5 years ago
Views:

1 Responding to Natural Hazards: The Effects of Disaster on Residential Location Decisions and Health Outcomes James Price Department of Economics University of New Mexico April 6 th,

2 Introduction Analyses Overview 1. Residential Sorting and the Value of Hazard Risk Reduction 2. County Migration Patterns and the Risk of Natural Hazards 3. Determinants of Mental Health & Displacement Following Hurricanes Katrina and Rita 2

3 Residential Sorting Residential Sorting Current Efforts: Improve Disaster Risk Management (DRM) investments Attempts to quantify the benefits and costs of DRM interventions Objective: Estimate WTP for reductions in hazard risk within the United States 3

4 Residential Sorting Previous Literature Theoretical Framework Ehrlich and Becker (1972) Berger et al. (1987) Hedonic Housing Analyses Residential Sorting Timmins (2007) Bayer et al. (2009) 4

5 Residential Sorting Theoretical Model Two-Stage Optimization Problem 1. Determine the optimal allocation of income between consumption goods 2. Select the location that maximizes utility, taking into account location-specific attributes Assumptions Knowledge of markets and amenities at each location Labor and housing market equilibrium Costly migration 5

6 Residential Sorting Theoretical Model (cont.) Expected Utility Function and Budget Constraint E(U ij ) = (1" #)U ND (C i,h i ;X j, M ij ) + #U D (C i,h i ;X j,m ij ) I ij = C + $ j H C = Composite Numeraire Good H = Housing Services X = Location - Specific Attributes M = Migration Cost " = Probability of Hazard Occurance I = Household Income # = The Price of Housing Services 6

7 Residential Sorting Theoretical Model (cont.) Indirect Expected Utility Function E(V ij ) = (1" #)V ND (I ij ND,$ j ND ;X j ND, M ij ) + #V D (I ij D,$ j D ;X j D,M ij ) Select Optimal Location E(V ij ) > E(V ik ) " j # k, k =1,2,..., j 7

8 Residential Sorting Empirical Model Utility Function and Budget Constraint " U ij = C C " i H H i e X " X j +M ij +# j +$ ij I ij = C + % j H Demand Functions C i = # " C % $ " C + " H & ( I ij H i = ' # " H % $ " C + " H & ( I ij ' ) j 8

9 Residential Sorting Empirical Model (cont.) Indirect Utility Function ln( V ij ) = " + # I ln( I ij ) $ # H ln (% j ) + # X X j + M ij + & j + ' ij ( where " = # c ln # + ( c * - + # H ln # + H * - ), ), and # I = # C + # H # I Marginal Willingness-to-Pay # I MWTP i = MU X MU I = " X " I I ij 9

10 Residential Sorting Empirical Model (cont.) Indirect Utility Function ln( V ij ) = " + # I ln( I ij ) $ # H ln (% j ) + # X X j + M ij + & j + ' ij ( where " = # c ln # + ( c * - + # H ln # + H * - ), ), and # I = # C + # H # I # I ln( I ij ) = ln( Î ij ) +" ij # j = # j * M ij = $ MS M ij S + $ MD M ij D + $ MR M ij R 10

11 Residential Sorting Empirical Model (cont.) Indirect Utility Function * ln( V ij ) = " + # I ln( Î ij ) $ # H ln % j ( ) + # X X j + # MS M ij S + # MD M ij D + # MR M ij R + & j + # I ' ij + ( ij Indirect Utility Function ln( V ij ) = " I ln( Î ij ) + " MS M S ij + " MD M D ij + " MR M R ij + # j + $ ij where # j = % & " H ln ' j and $ ij = " I ) ij + * ij ( * ) + " X X j + ( j 11

12 Residential Sorting Empirical Model (cont.) Conditional Logit Model P[ln(V ij ) " ln(v ik )] = j % k=1 e# I ln ( º ij )+# MS M S ij +# MD M D ij +# MR M R ij +$ j e # I ln ( º ij )+# MS M S ij +# MD M D ij +# MR M R ij +$ j 12

13 Residential Sorting Empirical Model (cont.) Quality-of-Life Decomposition " j = # $ % H ln & j ( * ) + % X X j + ' j # " H i = H % $ " C + " H & ( I ij ) * " H = " j H i I ' ) j I ij " j + # H ln $ j ( * ) = % + # X X j + & j 13

Residential Sorting Data 2005-2009 American Community Survey

14 Residential Sorting Data American Community Survey Housing services regression Wage regressions Conditional logit model 14

15 Data (cont.) Residential Sorting 15

16 Data (cont.) Residential Sorting Expected Number of Disaster Events 16

17 Data (cont.) Residential Sorting Expected Number of Disaster Events by MSA 17

18 Results: Conditional Logit Residential Sorting * p<0.1 ** p<0.05 *** p<0.01 N=50,000 18

19 Results: Quality-of-Life Index Residential Sorting 19

20 Residential Sorting Results: Quality-of-Life Decomposition * p<0.1 ** p<0.05 *** p<0.01 N=296 20

21 Results: WTP Estimates Residential Sorting TEMP ( F) Variable MWTP (Median Income) $759 PRECIP (Inches) $383 EMISSIONS (Lb/Per) ($154) NPLSITES (Sites) ($213) HRISK (Events/1000 Years) ($275) 21

22 Residential Sorting Conclusions Residential location decisions are partially determined by high-consequence lowprobability events Households are WTP $275 annually for a marginal reduction in the number of expected hazard events per 1000 years. 22

23 Net Migration Net Migration The spatial equilibrium model suggests household select their residential location so as to maximize utility--taking into account economic conditions and amenities Objectives: Quantify the relationship between countylevel migration rates and natural hazard risk Identify possible spatial heterogeneity in the migration-risk relationship 23

24 Net Migration U.S. Migration (cont.) 24

25 Net Migration Empirical Model: SAC Spatial Simultaneous Autoregressive M = "WM + E# E + D# D + A# A + u u = $Wu + e M = Net In - Migration Rate E = Economic Characteristics D = Demographic Characteristics A = Environmental Amenities W = Spatial Weight Matrix 25

26 Net Migration Empirical Model: SAC (cont.) Net Inmigration Rate ) + Net Inmigration Rate i = * 2009 " t= 2001 Spatial Weight Matrix, Net Domestic Migration it. # 1 % &.* (* " Population $ 9' it. - t= 2001 W ij = n i=1 d ij " d ij # 1 if counties i and j are neighbors where d ij = $ % 0 otherwise. 26

27 Results: SAC Net Migration * p<0.1 ** p<0.05 *** p<0.01 N=

28 Net Migration Empirical Model: GWR Geographically Weighted Regression M i = " ie E i + " id D i + " ia A i + # i # ~ i.i.d. N(0,$ 2 ) Estimate Parameter Values ) " i = ( # X i W i X i ) $1 # X i W i Y i Weight Matrix # W ik = exp "d & ik % ( $ ' b 2 28

29 Net Migration Results: GWR (Environmental Variables) 29

30 Results: GWR (Hazard Risk) Net Migration 30

31 Net Migration Conclusions County migration patterns are negatively correlated with hazard risk Hurricane and flood risk have a substantially greater affect on migration than earthquake risk There is significant spatial heterogeneity in the relationship between migration and hazard risk This migration-risk relationship is greatest along the Gulf Coast 31

32 Mental Health and Displacement Determinants of Mental Health and Displacement Hurricanes Katrina and Rita devastated parts of the Gulf Coast in 2005 Mass displacement $191 billion in property damage Extreme physical and psychological stress Objectives: Evaluate the effects of post-disaster stress on long term mental health status Identify determinants of displacement and displacement duration 32

33 Mental Health and Displacement Data Panel Survey of Income Dynamics 2005 and 2007 Supplemental questionnaire for residents of hurricane-affected areas Federal Emergency Management Agency Geospatial data regarding hurricane damage 33

34 Mental Health and Displacement Empirical Model: Mental Health Simultaneous Equations Model MH i = f (E i,b i,ss i,pdvi i ) PDVI i = f (E i,b i,ss i,ds i ) MH = Mental Health Indicator E = Socioeconomic Characteristics B = Behavioral and Health Characteristics SS = Social Support Index PDVI = Post Disaster Vulnerability Index DS = Disaster Severity 34

35 Mental Health and Displacement Empirical Model: Mental Health Post-Disaster Vulnerability Index Displacement Duration Property Damage Food Shortages Water Shortages Unsanitary Conditions Loss of Electricity 35

36 Data (cont.) Mental Health and Displacement 36

37 Mental Health and Displacement Results and Conclusions: Mental Health Several socioeconomic variables are correlated with adverse mental health outcomes The SSI is negatively correlated with adverse mental health outcomes The PDVI is positively correlated with adverse mental health outcomes 37

38 Mental Health and Displacement Empirical Model: Displacement Probit-Weibull Hurdel Model x 1 j = f (H j,ds j, E h j,b h j,ss j ) x 2 j = f (H j,ds j, E h j,b h j,ss j ) H = Housing Damage DS = Disaster Severity E = Socioeconomic Characteristics B = Behavioral and Health Characteristics SS = Social Support Index 38

39 Mental Health and Displacement Displacement Duration Plot of Kaplan-Meier Estimator 39

40 Mental Health and Displacement Results and Conclusions: Displacement Housing damage is positively correlated with both displacement and displacement duration Most socioeconomic variables are not significantly correlated with displacement The SSI is positively correlated with displacement and negatively correlated with displacement duration Remittances from family are negatively correlated with displacement duration 40

41 Mental Health and Displacement Questions? 41

42 Introduction Exposure to Natural Hazards Major hazard events since (Globally) 246 (United States) Exposure to natural hazards is rapidly increasing Population growth within hazard-prone areas Climate Change 42

43 Residential Sorting Theoretical Model (cont.) 43

44 Residential Sorting Empirical Model (cont.) Conditional Logit Model P[ln(V ij ) " ln(v ik ) # j $ k] = ev ij (I ij,% j ;X j,m ij ) j & e V ik (I ij,% j ;X j,m ij ) k=1 44

45 Residential Sorting Empirical Model (cont.) Price of Housing Services Estimates ln(uc ij ) = " 0 + ln(# j ) + " D D + $ ij UC = User Cost # = MSAfixed - effects D = Dweling Characteristics Income Estimates ln(w ij ) = " 0 + " S S ij + " P P(R B,R D SC) + " P P(R B,R D SC) 2 + # ij W = Hourly Wage Rate S = Socioeconomic Charateristics P( ) = Probability that Person Born in R B resides in R D 45

46 Residential Sorting Data (cont.) Data Sources for Location-Specific Attributes City and County Database Core of Common Data County Business Patterns National Climate Data Center Environmental Protection Agency Global Risk Data Platform 46

47 Results: Housing Regression Residential Sorting N=1,599,627 R 2 = * p<0.1 ** p<0.05 *** p<

48 Results: Wage Regression Residential Sorting N: Mean=5405 Max.=99324 Min.=547 R 2 : Mean=0.373 Max.=0.517 Min.= % of coefficients are significant at p<0.1 48

49 Results: WTP Estimates Residential Sorting 49

50 Net Migration U.S. Migration 50

51 Results: GWR (Emissions) Net Migration 51

Commuting, public transport investments and gentrification

Commuting, public transport investments and gentrification Evidence from Copenhagen Ismir Mulalic Technical University of Denmark and Kraks Fond June 12, 2018 Ismir Mulalic (DTU and Kraks Fond) Commuting