Housing Affordability and Material Hardship Experiences during the COVID-19 Pandemic

Forrest Hangen

Sophia Fox-Dichter

Yung Chun

Michal Grinstein-Weiss

Budgeting 101

Income

Expenses

Budgeting 101

Income

Expenses

>

Budgeting 101

Income

Expenses

<

Budgeting 102

Expenses

Income

Budgeting 102

Income

>

Basic Needs

Other

+

Expenses

Budgeting 103

Income

>

Basic Needs

+

Other

Budgeting 103

Income

>

Housing

Food

Medical

Other

Budgeting 103

Income

<

Housing

Food

Material Hardship

Medical

Other

Budgeting 104

Income

<

Cost-burdened

Medical

Food

Housing

Other

Budgeting 104

Low-income

Income

<

Other

 

Medical

Food

Housing

Other

Medical

 

Food

Housing

Income

<

Cost-burdened

Low-income

Other

 

Medical

 

Food

Housing

Income

<

< 80% of AMI

Housing Affordability

Income

>

Other

Medical

Food

Housing

Is it a burden to pay for this housing?
financial

Housing Affordability

Housing

Income

>30%

>50%

<2/3 of Federal Poverty Level

(FPL)

Income

Housing

Ratio Measures

Residual Measures

Housing

Affordability

Ratio Measures

>30%

>50%

Residual Measures

<2/3 of FPL

Low-income

< 80% of AMI

Material

Hardship

Ratio Measures

>30%

>50%

Residual Measures

<2/3 of FPL

Low-income

< 80% of AMI

Material

Hardship

How accurate is this relationship?

Housing

Affordability

COVID-19 Pandemic

Emergency Rental Assistance (ERA)

1. At least one member of your household has:

          - Qualified for unemployment or should qualify

          - Lost income

          - Owed large expenses, OR

          - Had other financial hardships

2.Your household income is below a certain amount, based on where you live

3. At least one member of your household is experiencing housing instability, which means they are at risk of becoming homeless or would have trouble finding a stable place to live

COVID-19 Pandemic

Emergency Rental Assistance (ERA)

 

 

 

 

 

2.Your household income is below a certain amount, based on where you live

3. At least one member of your household is experiencing housing instability, which means they are at risk of becoming homeless or would have trouble finding a stable place to live

1. Financial hardship or income loss

COVID-19 Pandemic

Emergency Rental Assistance (ERA)

 

 

 

 

 

 

3. At least one member of your household is experiencing housing instability, which means they are at risk of becoming homeless or would have trouble finding a stable place to live

1. Financial hardship or income loss

2. Low income

COVID-19 Pandemic

Emergency Rental Assistance (ERA)

1. Financial hardship or income loss

2. Low income

3. Housing instability

COVID-19 Pandemic

Emergency Rental Assistance (ERA)

1. Financial hardship or income loss

2. Low income

3. Housing instability

Material Hardship

COVID-19 Pandemic

Emergency Rental Assistance (ERA)

1. Financial hardship or income loss

2. Low income

3. Housing instability

Material Hardship

Initial Research Questions

Q1: How accurate are simple measures of housing affordability or income level at identifying material hardship? 

Q2: What factors matter for predicting material hardship? 

Q3: Can we use those factors to better predict material hardship? 

Initial Research Questions

Q1: How accurate are simple measures of housing affordability or income level at identifying material hardship? 

Q2: What factors matter for predicting material hardship? 

Q3: Can we use those factors to better predict material hardship? 

 

Initial Research Questions

Q1: How accurate are simple measures of housing affordability or income level at identifying material hardship? 

Q2: What factors matter for predicting material hardship? 

Q3: Can we use those factors to better predict material hardship? 

 

Initial Research Questions

Q1: How accurate are simple measures of housing affordability or income level at identifying material hardship? 

Q2: What factors matter for predicting material hardship? 

Q3: Can we use those factors to better predict material hardship? 

 

Initial Research Questions

Q1: How accurate are simple measures of housing affordability or income level at identifying material hardship? 

Q2: What factors matter for predicting material hardship? 

Q3: Can we use those factors to better predict material hardship? 

 

Preliminary Qs & Results

Data & Methods

Social Economic Impacts of COVID-19 Survey (SEICS)

conducted by the Social Policy Institute at Washington University in St. Louis (2021)

Nationally Representative

5 waves

April 2020 and June 2021

~5000 / wave

19593 total

Data & Methods

Housing Affordability

Ratio >30%

Ratio >50%

Residual < 1/2 of FPL

Residual <2/3 of FPL

 

Income

<80% of AMI

 

Material Hardship

 

Trouble paying rent/mortgage

Eviction or foreclosure

Trouble paying utility bill

 

Food insecurity

Financial Trouble

Data & Methods

AUC

 

Efficiency

 

Coverage

 

Harmonic Mean of Coverage and Efficiency

 

 

Data & Methods

AUC

(Area under the ROC Curve)

How well does measure distinguish between positive and negative instances

 

.5 = no better than chance

Efficiency

 

Coverage

 

Harmonic Mean of Coverage and Efficiency

 

 

 

Data & Methods

AUC

(Area under the ROC Curve)

How well does measure distinguish between positive and negative instances

 

.5 = no better than chance

Efficiency

n individuals who experienced a hardship and are

    burdened    

n burdened indivuduals


How well does measure identify those who experience a hardship?

Coverage

 

Harmonic Mean of Coverage and Efficiency

 

 

Data & Methods

AUC

(Area under the ROC Curve)

How well does measure distinguish between positive and negative instances

 

.5 = no better than chance

Efficiency

n individuals who experienced a hardship and are

    burdened    

n burdened indivuduals


How well does measure identify those who experience a hardship?

Coverage

n individuals who experienced a hardship and are

    burdened    

n hardship individuals

 

How well does measure identify everyone who experienced a hardship?

Harmonic Mean of Coverage and Efficiency

 

 

Data & Methods

AUC

(Area under the ROC Curve)

How well does measure distinguish between positive and negative instances

 

.5 = no better than chance

Efficiency

n individuals who experienced a hardship and are

    burdened    

n burdened indivuduals


How well does measure identify those who experience a hardship?

Coverage

n individuals who experienced a hardship and are

    burdened    

n hardship individuals

 

How well does measure identify everyone who experienced a hardship?

Harmonic Mean of Coverage and Efficiency

reciprocal of the arithmetic mean of the reciprocals

 

How well does measure balance coverage and efficiency?

 

Data & Methods

AUC

(Area under the ROC Curve)

How well does measure distinguish between positive and negative instances

 

.5 = no better than chance

Efficiency

n individuals who experienced a hardship and are

    burdened    

n burdened indivuduals


How well does measure identify those who experience a hardship?

Coverage

n individuals who experienced a hardship and are

    burdened    

n hardship individuals

 

How well does measure identify everyone who experienced a hardship?

Harmonic Mean of Coverage and Efficiency

reciprocal of the arithmetic mean of the reciprocals

 

How well does measure balance coverage and efficiency?

 

Predictive Validity

30% 2/3 of FPL 80% AMI
% identified 18.77% 13.20% 43.17%
AUC 0.604 0.613 0.653
Efficiency 43.14% 53.74% 35.95%
Coverage 34.78% 30.48% 66.67%
HM 38.51% 38.90% 46.72%
30% 2/3 of FPL 80% AMI
% identified 18.77% 13.20% 43.17%
AUC 0.604 0.613 0.653
Efficiency 43.14% 53.74% 35.95%
Coverage 34.78% 30.48% 66.67%
HM 38.51% 38.90% 46.72%

Predictive Validity

30% 2/3 of FPL 80% AMI
% identified 18.77% 13.20% 43.17%
AUC 0.604 0.613 0.653
Efficiency 43.14% 53.74% 35.95%
Coverage 34.78% 30.48% 66.67%
HM 38.51% 38.90% 46.72%

Predictive Validity

30% 2/3 of FPL 80% AMI
% identified 18.77% 13.20% 43.17%
AUC 0.604 0.613 0.653
Efficiency 43.14% 53.74% 35.95%
Coverage 34.78% 30.48% 66.67%
HM 38.51% 38.90% 46.72%

Predictive Validity

30% 2/3 of FPL 80% AMI
% identified 18.77% 13.20% 43.17%
AUC 0.604 0.613 0.653
Efficiency 43.14% 53.74% 35.95%
Coverage 34.78% 30.48% 66.67%
HM 38.51% 38.90% 46.72%

Predictive Validity

30% 2/3 of FPL 80% AMI
% identified 18.77% 13.20% 43.17%
AUC 0.604 0.613 0.653
Efficiency 43.14% 53.74% 35.95%
Coverage 34.78% 30.48% 66.67%
HM 38.51% 38.90% 46.72%

Predictive Validity

30% 2/3 of FPL 80% AMI
% identified 18.77% 13.20% 43.17%
AUC 0.604 0.613 0.653
Efficiency 43.14% 53.74% 35.95%
Coverage 34.78% 30.48% 66.67%
HM 38.51% 38.90% 46.72%

Good news!

Predictive Validity

Factors Predicting Material Hardship

Logistic Regression Models

 

HA or Income Measure

 

Household Size

Children

Renter

Age

Job Loss

Race/Ethnicity

Geographic Region

 

Factors Predicting Material Hardship

Logistic Regression Models

 

HA or Income Measure

 

Household Size

Children*

Renter

Age*

Job Loss

Race/Ethnicity*

Geographic Region

 

Factors Predicting Material Hardship

Logistic Regression Models

 

HA or Income Measure

 

Household Size

Children

Renter

Age

Job Loss

Race/Ethnicity

Geographic Region

 

Original Measures

30% 2/3 of FPL 80% AMI
% identified 18.77% 13.20% 43.17%
AUC 0.604 0.613 0.653
Efficiency 43.14% 53.74% 35.95%
Coverage 34.78% 30.48% 66.67%
HM 38.51% 38.90% 46.72%

Original Measures + Job loss

30% + Job 2/3 + Job 80% + Job
% identified 33.97% 29.48% 52.83%
AUC 0.676 0.685 0.653
Efficiency 41.83% 45.74% 35.59%
Coverage 61.04% 57.91% 80.77%
HM 49.64% 51.11% 49.41%

Original Measures + Job loss

30% + Job 2/3 + Job 80% + Job
% identified 33.97% 29.48% 52.83%
AUC 0.676 0.685 0.653
Efficiency 41.83% 45.74% 35.59%
Coverage 61.04% 57.91% 80.77%
HM 49.64% 51.11% 49.41%

Original Measures + Job loss

30% + Job 2/3 + Job 80% + Job
% identified 33.97% 29.48% 52.83%
AUC 0.676 0.685 0.653
Efficiency 41.83% 45.74% 35.59%
Coverage 61.04% 57.91% 80.77%
HM 49.64% 51.11% 49.41%

Factors Predicting Material Hardship

Decision Tree Classifier

Threshold that best separates positives from negatives

Factors Predicting Material Hardship

Decision Tree Classifier

Job loss?
Yes = 
at Risk
No
Income < 73% of AMI
No
Yes = 
at Risk

DT Thresholds

31.75% + Job $34,834 + Job 73.25% AMI + Job
% identified 36.11% 46.53% 51.26%
AUC 0.692 0.710 0.722
Efficiency 43.72% 40.77% 38.75%
Coverage 63.79% 76.67% 80.27%
HM 51.88% 53.24% 52.27%

DT Thresholds

31.75% + Job $34,834 + Job 73.25% AMI + Job
% identified 36.11% 46.53% 51.26%
AUC 0.692 0.710 0.722
Efficiency 43.72% 40.77% 38.75%
Coverage 63.79% 76.67% 80.27%
HM 51.88% 53.24% 52.27%

DT Thresholds

31.75% + Job $34,834 + Job 73.25% AMI + Job
% identified 36.11% 46.53% 51.26%
AUC 0.692 0.710 0.722
Efficiency 43.72% 40.77% 38.75%
Coverage 63.79% 76.67% 80.27%
HM 51.88% 53.24% 52.27%

DT Thresholds

31.75% + Job $34,834 + Job 73.25% AMI + Job
% identified 36.11% 46.53% 51.26%
AUC 0.692 0.710 0.722
Efficiency 43.72% 40.77% 38.75%
Coverage 63.79% 76.67% 80.27%
HM 51.88% 53.24% 52.27%

DT Thresholds

31.75% + Job $34,834 + Job 73.25% AMI + Job
% identified 36.11% 46.53% 51.26%
AUC 0.692 0.710 0.722
Efficiency 43.72% 40.77% 38.75%
Coverage 63.79% 76.67% 80.27%
HM 51.88% 53.24% 52.27%

Takeaways

ERA Eligibility Requirements provided High Coverage, but Lower Efficiency

30% Ratio Measure Generally Underperformed

Residual & Income Measures Show Promise

Future Directions

Explore more general thresholds (ACS to SEICS)

Use panel data to understand shifts in validity through pandemic

Your Suggestions!

Examine geographic and racial inequities

Thank You

Forrest Hangen

 

 

hangen.f@northeastern.edu

forresthangen.com