Let the Rich Be Flooded: The Unequal Impact of Hurricane Harvey on - - PowerPoint PPT Presentation

Let the Rich Be Flooded: The Unequal Impact of Hurricane Harvey

• n Household Debt

Stephen Billings (U. Colorado Boulder) Emily Gallagher (U. Colorado Boulder & StL Fed) Lowell Ricketts (StL Fed)

The views expressed here are those of the authors only. They do not represent the views of any of the affiliated institutions, data providers, or funders.

Background Background

Hurricane Harvey (Aug-Sep 2017) stalled over Houston.

Source: vox.com

Background Background

Submerged 25–30% of Houston, $125 billion in damage

Background Background

Flooding under Harvey relative to 100 year floodplain

Source: USGS/FEMA

Motivation Motivation

Disconnect: academic literature vs. anecdotal reporting

Based on Katrina, academics view these events as having little effect on household finance Mild, temporary jump in delinquency rates; homeowners used flood insurance payouts to eliminate mortgage debt (Gallagher and Hartley, 2017) Minimal impacts on income and labor markets (McIntosh, 2008; Groen, Kutzbach and Polivka, 2017; Deryugina, Kawano and Levitt, 2018; Deryugina et al., 2018)

Consensus of the literature

“These studies all conclude that the average net financial impact of a large natural disaster is modest and short-lived, even for the most severely impacted victims.” ~ Gallagher, Hartley and Rohlin (2018)

Motivation Motivation

Disconnect: academic literature vs. anecdotal reporting

Meanwhile, reporters tell tales of complete financial devastation from flooding

Anecdotal reporting

“Both families had to start over from nothing. But today, one family is financially stable. The

• ther is facing bankruptcy.”

~ NPR report by Rebecca Hersher and Rober Benincasa (2019)

Motivation Motivation

Rising inequalities in areas affected by disasters

Wealth inequality tends to rise in areas affected by disasters, particularly in areas that receive more FEMA aid (e.g., Howell and Elliott, 2018)

Motivation Motivation

Is Federal disaster assistance regressive?

Flood insurance (from the National Flood Insurance Program) Mandatory for mortgage borrowers located in a 100-year floodplain

• Avg. NFIP payout = $117,000.

For uninsured losses: FEMA grants

◮ Avg. recipient grant = $7,300 (max $33,000) vs. $72,162 in damage for 1 ft of flooding ◮ Eligibility for home repair grants: verified losses > $17k but <50% of the market value of the

home SBA disaster loans

◮ Avg. approved loan = $74,549; interest rate of 1.75%, 30 years for repayment ◮ Eligibility: (1) credit score; (2) debt-to-income (ability to repay) ⋆ Begley, Gurun, Purnanandam and Weagley (2018) show substantial inequalities in

access to individual disaster loans from the SBA IRS disaster refunds

◮ File an amended tax form citing property loss (benefits people with higher incomes and, thus,

greater tax liabilities) ⇒Suggests ↑ access to assistance if you inside the floodplain or are higher income, credit score

Motivation Motivation

This paper

Goal: Apply administrative credit data to the question of whether averages mask important heterogeneity in the financial effects of flooding Setting: Hurricane Harvey Affected a wide variety of income and racial groups, both inside and outside of the designated 100-year floodplain Only 7% if variation in flooding is explained by geo-spacial and socio-economic factors Method: Treatment intensity difference-in-difference design Compare the credit outcomes of Houston residents according to degree of flooding in their Census block Test for differences in treatment effects on those who have more or less resources to absorb a wealth shock

◮ (a) initial financial condition; ◮ (b) location with respect to the floodplain (proxy for flood insurance)

Motivation Motivation

Preview of results

Consistent with earlier research, average effects of flooding are small and temporary

Motivation Motivation

Preview of results

Consistent with earlier research, average effects of flooding are small and temporary But, financially constrained homeowners experience considerably more financial distress (more bankruptcy, more delinquency) after flooding

◮ e.g., top-tercile flooding is associated with a 28% ↑ in bankruptcy rates in high

• wner-occupied, financially constrained areas relative to no-flood areas

◮ Mortgage eliminations, consistent with a lack of resources needed to rebuild

Motivation Motivation

Preview of results

Consistent with earlier research, average effects of flooding are small and temporary But, financially constrained homeowners experience considerably more financial distress (more bankruptcy, more delinquency) after flooding

◮ e.g., top-tercile flooding is associated with a 28% ↑ in bankruptcy rates in high

• wner-occupied, financially constrained areas relative to no-flood areas

◮ Mortgage eliminations, consistent with a lack of resources needed to rebuild

Being in a floodplain and, hence, having a higher likelihood of flood insurance, largely mitigates negative effects

Motivation Motivation

Preview of results

Consistent with earlier research, average effects of flooding are small and temporary But, financially constrained homeowners experience considerably more financial distress (more bankruptcy, more delinquency) after flooding

◮ e.g., top-tercile flooding is associated with a 28% ↑ in bankruptcy rates in high

• wner-occupied, financially constrained areas relative to no-flood areas

◮ Mortgage eliminations, consistent with a lack of resources needed to rebuild

Being in a floodplain and, hence, having a higher likelihood of flood insurance, largely mitigates negative effects Mechanism? Due, in part, to a regressive allocation of federal disaster assistance

◮ Not just SBA, FEMA grants also appear to be regressive

Identification Identification

Data & Method

Data NYFED/Equifax Consumer Credit Panel, Census Data, FEMA flood maps Method Treatment intensity difference-in-difference yibt =

8

• τ=−8

βτ

• Tk

b × Dτ

• + αi + Dτ + κf(Ait) + η (Cb × Dτ) + φCb + εibt

yibt is quarterly outcome (bankruptcy, delinquency, mortgage debt) Tb is the treatment intensity, WAvg. Flood Depth; assigned according to block where that individual lived in as of Q2 2017; in terciles of flooding β: quarterly change in blocks of a given flood intensity relative to the blocks that did not flood

◮ Unconstrained vs. Constrained ◮ Inside vs. Outside 100-year floodplain (proxy for insurance)

Results Results

Block-level bankruptcy flag rate

Data source: Federal Reserve Bank of NY/Equifax Consumer Credit Panel

Results Results

Block-level bankruptcy flag rate

Sample All All Own<p50 Own≥p50 FlP=0% FlP ≥ 50% & Own≥p50 & Own≥p50 T1

b × Pt

0.09 0.13

• 0.01

0.17 0.02 0.21 (1.03) (1.25) (-0.05) (1.42) (0.17) (0.46) T2

b × Pt

• 0.05
• 0.01
• 0.13

0.04

• 0.03

0.38 (-0.57) (-0.13) (-0.66) (0.35) (-0.18) (0.86) T3

b × Pt

• 0.02

0.01 0.13

• 0.01
• 0.37

0.07 (-0.18) (0.14) (0.61) (-0.06) (-1.46) (0.17) T1

b × Pt × Constrainedb

• 0.1
• 0.14

0.13 0.62**

• 0.88

(-0.57) (-0.52) (0.50) (2.17) (-1.00) T2

b × Pt × Constrainedb

• 0.09
• 0.15

0.11 0.15 0.01 (-0.49) (-0.57) (0.36) (0.38) (0.02) T3

b × Pt × Constrainedb

• 0.07
• 0.65**

0.66** 1.48*** 0.08 (-0.37) (-2.20) (2.28) (2.97) (0.10) N 518,428 518,428 228,545 289,883 215,824 25,283

Heavy flooding & constraint is associated with a 0.66%pt rel. ↑ in bankruptcy rate, which represents 28% of the avg. pre-Harvey bankruptcy rate in Constrained, owner-occupied blocks (2.4%).

Data source: Federal Reserve Bank of NY/Equifax Consumer Credit Panel

Results Results

Individual’s Share of debt that is severely delinquent

β estimates of the effect of terciles of flooding relative to no flooding

Sample: All LoEquity HiEquity LoEquity & FlP=0% LoEquity & FlP≥50% T1

b × Pt

• 0.09

0.11

• 0.01

0.05

• 0.28

(-0.81) (1.30) (-0.04) (0.42) (-1.38) T2

b × Pt

0.14 0.09

• 0.07

0.17

• 0.50

(1.11) (0.63) (-0.36) (0.95) (-1.33) T3

b × Pt

• 0.01

0.11 0.04

• 0.15
• 0.30

(-0.09) (1.10) (0.19) (-0.64) (-1.44) T1

b × Pt × Constrainedi

0.04 0.56 0.68 0.59 2.81 (0.11) (1.07) (1.59) (1.06) (1.41) T2

b × Pt × Constrainedi

0.68** 0.96* 0.61 1.12* 1.54 (2.23) (1.98) (1.03) (1.90) (1.68) T3

b × Pt × Constrainedi

0.79** 1.38***

• 0.36

1.75** 0.84 (2.28) (2.91) (-0.64) (2.16) (0.71) N 1,605,162 297,002 288,532 191,942 20,828 Effect is modest in the full sample: 0.8%pts represents 4.4% of the Constrained sample’s pre-Harvey delinquency share (17.7%) Constrained, mortgage-holders with little equity see a relative 1.4%pt (35%) ↑ in their pre-Harvey delinquency share Being outside the floodplain also has an amplifying effect on delinquency

Results Results

Individual’s Share of debt that is severely delinquent

β estimates of the effect of terciles of flooding relative to no flooding

Sample: All LoEquity HiEquity LoEquity & FlP=0% LoEquity & FlP≥50% T1

b × Pt

• 0.09

0.11

• 0.01

0.05

• 0.28

(-0.81) (1.30) (-0.04) (0.42) (-1.38) T2

b × Pt

0.14 0.09

• 0.07

0.17

• 0.50

(1.11) (0.63) (-0.36) (0.95) (-1.33) T3

b × Pt

• 0.01

0.11 0.04

• 0.15
• 0.30

(-0.09) (1.10) (0.19) (-0.64) (-1.44) T1

b × Pt × Constrainedi

0.04 0.56 0.68 0.59 2.81 (0.11) (1.07) (1.59) (1.06) (1.41) T2

b × Pt × Constrainedi

0.68** 0.96* 0.61 1.12* 1.54 (2.23) (1.98) (1.03) (1.90) (1.68) T3

b × Pt × Constrainedi

0.79** 1.38***

• 0.36

1.75** 0.84 (2.28) (2.91) (-0.64) (2.16) (0.71) N 1,605,162 297,002 288,532 191,942 20,828 Effect is modest in the full sample: 0.8%pts represents 4.4% of the Constrained sample’s pre-Harvey delinquency share (17.7%) Constrained, mortgage-holders with little equity see a relative 1.4%pt (35%) ↑ in their pre-Harvey delinquency share Being outside the floodplain also has an amplifying effect on delinquency

Results Results

Individual’s Share of debt that is severely delinquent

β estimates of the effect of terciles of flooding relative to no flooding

Sample: All LoEquity HiEquity LoEquity & FlP=0% LoEquity & FlP≥50% T1

b × Pt

• 0.09

0.11

• 0.01

0.05

• 0.28

(-0.81) (1.30) (-0.04) (0.42) (-1.38) T2

b × Pt

0.14 0.09

• 0.07

0.17

• 0.50

(1.11) (0.63) (-0.36) (0.95) (-1.33) T3

b × Pt

• 0.01

0.11 0.04

• 0.15
• 0.30

(-0.09) (1.10) (0.19) (-0.64) (-1.44) T1

b × Pt × Constrainedi

0.04 0.56 0.68 0.59 2.81 (0.11) (1.07) (1.59) (1.06) (1.41) T2

b × Pt × Constrainedi

0.68** 0.96* 0.61 1.12* 1.54 (2.23) (1.98) (1.03) (1.90) (1.68) T3

b × Pt × Constrainedi

0.79** 1.38***

• 0.36

1.75** 0.84 (2.28) (2.91) (-0.64) (2.16) (0.71) N 1,605,162 297,002 288,532 191,942 20,828 Effect is modest in the full sample: 0.8%pts represents 4.4% of the Constrained sample’s pre-Harvey delinquency share (17.7%) Constrained, mortgage-holders with little equity see a relative 1.4%pt (35%) ↑ in their pre-Harvey delinquency share Being outside the floodplain also has an amplifying effect on delinquency

Results Results

Mortgage debt after flooding (mortgage holders)

β estimates of the effect of top-tercile flooding relative to no flooding

Discussion Discussion

Mechanism? Inequality in access to disaster assistance

FEMA registration data, Hurricane Harvey Lower-income and minority areas get less in disaster assistance, controlling for insurance, flood depth, and FEMA-assessd damage

Dependent var: Share of blocks’ Individual FEMA registrants Eligible that are granted Individual FEMA for assistance assistance ($) SBA loan Median Income (10k) 0.174*** 171.94*** 1.176*** (5.90) (9.34) (-21.01) Minority Share

• 0.085***
• 0.69}
• 0.032***

(-23.70) (-0.37) (-7.37) Sample All Homeowners; assistance>0 Damage>0 Y-mean 14.19 9,353 10.21 N 32,659 62,485 150,339

Controls: Flood insurance, property damage, share of housing units with damage, share of housing units registered with FEMA,

• wgt. avg. flood depth, density, owner-occ. share.

A simultaneous 1σ ↓ in block income and 1σ ↑ in minority share would:

◮ ↓ the prob. of a FEMA grant by 3%pts (21%) ◮ ↓ a homeowner’s FEMA grant amount by $653 (7%) ◮ Cut in half a FEMA registrant’s chances of an SBA loan

Discussion Discussion

Mechanism? Inequality in access to disaster assistance

FEMA registration data, Hurricane Harvey Lower-income and minority areas get less in disaster assistance, controlling for insurance, flood depth, and FEMA-assessd damage

Dependent var: Share of blocks’ Individual FEMA registrants Eligible that are granted Individual FEMA for assistance assistance ($) SBA loan Median Income (10k) 0.174*** 171.94*** 1.176*** (5.90) (9.34) (-21.01) Minority Share

• 0.085***
• 0.69}
• 0.032***

(-23.70) (-0.37) (-7.37) Sample All Homeowners; assistance>0 Damage>0 Y-mean 14.19 9,353 10.21 N 32,659 62,485 150,339

Controls: Flood insurance, property damage, share of housing units with damage, share of housing units registered with FEMA,

• wgt. avg. flood depth, density, owner-occ. share.

A simultaneous 1σ ↓ in block income and 1σ ↑ in minority share would:

◮ ↓ the prob. of a FEMA grant by 3%pts (21%) ◮ ↓ a homeowner’s FEMA grant amount by $653 (7%) ◮ Cut in half a FEMA registrant’s chances of an SBA loan

Discussion Discussion

Mechanism? Inequality in access to disaster assistance

FEMA registration data, Hurricane Harvey Lower-income and minority areas get less in disaster assistance, controlling for insurance, flood depth, and FEMA-assessd damage

Dependent var: Share of blocks’ Individual FEMA registrants Eligible that are granted Individual FEMA for assistance assistance ($) SBA loan Median Income (10k) 0.174*** 171.94*** 1.176*** (5.90) (9.34) (-21.01) Minority Share

• 0.085***
• 0.69}
• 0.032***

(-23.70) (-0.37) (-7.37) Sample All Homeowners; assistance>0 Damage>0 Y-mean 14.19 9,353 10.21 N 32,659 62,485 150,339

Controls: Flood insurance, property damage, share of housing units with damage, share of housing units registered with FEMA,

• wgt. avg. flood depth, density, owner-occ. share.

A simultaneous 1σ ↓ in block income and 1σ ↑ in minority share would:

◮ ↓ the prob. of a FEMA grant by 3%pts (21%) ◮ ↓ a homeowner’s FEMA grant amount by $653 (7%) ◮ Cut in half a FEMA registrant’s chances of an SBA loan

Discussion Discussion

Conclusions

Allocating disaster assistance in a way that appears to be (unintentionally) regressive Therefore, important to differentiate treatment effects by initial financial condition and floodplain status

◮ Flooding has much more deleterious effects on the bankruptcy rates and

delinquencies and of homeowners that are Constrained

◮ A lack of resources to rebuild is potentially leading to forced home sales ◮ Constrained vs. Unconstrained outcomes are much more equal inside the

floodplain, suggesting that flood insurance helps Unless disaster assistance becomes less regressive, expect rising inequality in areas affected by disasters

Discussion Discussion

Discussion Discussion Begley, T.A., Gurun, U., Purnanandam, A.K., Weagley, D., 2018. Disaster lending: "fair" prices, but "unfair" access. Working paper. Deryugina, T., Kawano, L., Levitt, S., 2018. The economic impact of hurricane katrina on its victims: Evidence from individual tax returns. American Economic Journal: Applied Economics 10, 202–233. Gallagher, J., Hartley, D., 2017. Household finance after a natural disaster: The case of hurricane katrina. American Economic Journal: Economic Policy 9, 199–228. Gallagher, J., Hartley, D., Rohlin, S., 2018. Weathering an unexpected financial shock: The role of cash grants on household finance and business growth following a natural disaster. Working paper. Groen, J.A., Kutzbach, M.J., Polivka, A.E., 2017. Storms and jobs: The effect of hurricanes on individuals’ employment and earnings over the long term. U.S. Bureau of Labor Statistics, Working paper. Howell, J., Elliott, J.R., 2018. Damages Done: The Longitudinal Impacts of Natural Hazards on Wealth Inequality in the United States. Social Problems . McIntosh, M.F., 2008. Measuring the labor market impacts of hurricane katrina migration: Evidence from houston, texas. American Economic Review 98, 54–57.