Housing Demand & Affordability for Low-Wage Households: Evidence - - PowerPoint PPT Presentation

Housing Demand & Affordability for Low-Wage Households: Evidence from Minimum Wage Changes

Sam Hughes

Wharton Virtual AREUEA 2020

Agenda

• 1. Research questions & motivation
• 2. Empirical strategy
• 3. Results
• 4. Conclusion

Housing affordability & the “Engel Curve”

Connected to a fundamental theoretical issue

Cobb-Douglas’ Engel curve is flat; income has no causal effect on rent-to-income ratios. Differences are caused by changes in preferences (or maybe prices). (DiPasquale & Murray, 2017)

Using minimum wage to measure causal effect of premanent income increase on housing expenditure share

If Engel curves are sloped, we may need non-homothetic models of housing demand. Want to measure: income elasticity of expenditure share

Connecting policymaking on housing & minimum wage

Connecting policymaking on housing & minimum wage

Research Questions

• 1. What is the effect of minimum wages on housing demand & consumption?
• 3 main outcomes: rent; household income; rent-to-income ratio
• Test homotheticity assumption: Is expenditure share constant?
• In the paper, derive sufficient statistics-style formula for the welfare change
• 2. How do minimum wages affect rental prices?
• Potential policy trade-off: MW helps those whose incomes increase; price increase may

hurt others

• Test of heterogeneity w.r.t. housing supply elasticity

Preview of main results

A 10% increase in minimum wages:

• increases income for affected households by 1.9%
• increases housing consumption by 0.5%
• decreases rent-to-income ratios by 1.4%
• has a near-zero average effect on rental prices nationally
• the rental price effect is heterogeneous w.r.t. local housing supply elasticity, so

rental prices rise more in cities where it is difficult to build housing

Contribution

• 1. Consumption & associated effects of minimum wage

➢ Aaronson, Agarwal & French (2012); Alonso (2016); Dautovic, Hau & Huang (2017) ➢ Employment/poverty: Card & Krueger, 2000; Dube, 2018; Cengiz, Dube, Lindner & Zipperer, 2019 ➢ Other outcomes… crime (Fone, Sabia & Cesur, 2019), health (Wehby, Dave & Kaestner, 2016), credit (Dettling & Hsu, 2017) etc.

• 2. Insights into the fundamentals of housing demand

➢ How inelastic is housing demand (e.g. Mayo, 1981)? How elastic is the expenditure share? ➢ Is Cobb-Douglas good enough?

• YES: Davis & Ortalo-Magné (2011); Moretti (2013); Eeckhout, Pinheiro & Schmidheiny (2014); Diamond (2016); Monte, Redding & Rossi-

Handberg (2018); Nathanson (2019); Molloy, Nathanson & Paciorek (2019); Rappaport (2019); Hsieh & Moretti (2019)

• NO: Albouy, Ehrlich & Liu (2016); Ganong & Shoag (2017); Tsivanidis (2019); Couture & Handbury (2018); Couture, Gaubert, Handbury &

Hurst (2019); Atkin, Faber, Fally & Gonzalez-Navarro (2019)

• 3. Price effects of minimum wage

➢ MW + Prices in general: Wadsworth (2010); Aaronson et al. (2012), Leung (2018), and Renkin et al. (2019). ➢ MW + Migration/Spatial eqm: Monras (2018); Perez Perez (2018); Zhang (2018) ➢ MW + Prices + Spatial eqm: Yamagishi (2019); Tidemann (2018); Agarwal, Ambrose & Diop (2019)

Agenda

• 1. Research questions & motivation
• 2. Empirical strategy
• 3. Results
• 4. Conclusion

Data: ACS 2005-2017; renter households whose household head is working in private sector Variables: household income, rent, imputed wage (usual hours; weeks worked);

• includes state-by-occupation & occupation-by-year FEs, state-specific linear time trend & Census region by year FE

Triple difference:

• 1. pre-/post-MW change;
• 2. across states MW change;
• 3. low-wage vs. higher-wage workers

ln 𝑧𝑗,𝑡,𝑢 = 𝜄𝑡 + 𝜄𝑢 + 𝜒 ln 𝑁𝑋

𝑡,𝑢 + 𝛿 𝟐 wage𝑗 < 𝑏𝑔𝑔 𝑡,𝑢 + 𝛾 ln 𝑁𝑋 𝑡,𝑢 ∗ 𝟐 wage𝑗 < 𝑏𝑔𝑔 𝑡,𝑢 + 𝜁𝑗,𝑡,𝑢 individual state year Common min. wage effect Binary variable: =1 if household head below X percentile in state-year wage distribution; X defined using average percentile where wages are <100% of state min. wage Interaction minimum wage effect Main coefficient of interest: 𝛾

Empirical Approach

Minimum Wage Variation

Using ACS data from 2005-2017 Using ACS data from 2005-2017

Agenda

• 1. Research questions & motivation
• 2. Empirical strategy
• 3. Results
• 4. Conclusion

Results

ln(rent) ln(hhincome) ln(rent/hhincome)

(1) (2) (3) (4) (5) (6) Common Effect: ln(MW)

• 0.0106

(0.0328)

Interaction Effect: ln(MW)*Affected 0.189***

(0.0322)

Affected

• 1.104***

(0.0652) Observations 2,248,921 R-Squared 0.291 Affected Wage Relative to MW 100% Affected Wage Percentile Cutoff 16 Two-Way FE X State-Year Linear Trend X e^(Mean of Outcome Below Affected Pctile) $1,452.44 e^(Mean of Outcome Above Affected Pctile) $3,857.21

Results

ln(rent) ln(hhincome) ln(rent/hhincome)

(1) (2) (3) (4) (5) (6) Common Effect: ln(MW)

• 0.00783
• 0.0106

0.00273

(0.0235) (0.0328) (0.0241)

Interaction Effect: ln(MW)*Affected 0.0543** 0.189***

• 0.135***

(0.0246) (0.0322) (0.0361)

Affected

• 0.213***
• 1.104***

0.892***

(0.0497) (0.0652) (0.0689) Observations 2,248,921 2,248,921 2,248,921 R-Squared 0.327 0.291 0.158 Affected Wage Relative to MW 100% 100% 100% Affected Wage Percentile Cutoff 16 16 16 Two-Way FE X X X State-Year Linear Trend X X X e^(Mean of Outcome Below Affected Pctile) $682.73 $1,452.44 47.0% e^(Mean of Outcome Above Affected Pctile) $854.23 $3,857.21 22.1%

Results

ln(rent) ln(hhincome) ln(rent/hhincome)

(1) (2) (3) (4) (5) (6) Common Effect: ln(MW)

• 0.00783
• 0.00619
• 0.0106
• 0.0158

0.00273 0.00965

(0.0235) (0.0237) (0.0328) (0.0328) (0.0241) (0.0244)

Interaction Effect: ln(MW)*Affected 0.0543** 0.0256 0.189*** 0.130***

• 0.135***
• 0.104***

(0.0246) (0.0234) (0.0322) (0.0277) (0.0361) (0.0254)

Affected

• 0.213***
• 0.155***
• 1.104***
• 0.893***

0.892*** 0.738***

(0.0497) (0.0473) (0.0652) (0.0585) (0.0689) (0.0513) Observations 2,248,921 2,248,921 2,248,921 2,248,921 2,248,921 2,248,921 R-Squared 0.327 0.328 0.291 0.293 0.158 0.157 Affected Wage Relative to MW 100% 125% 100% 125% 100% 125% Affected Wage Percentile Cutoff 16 25 16 25 16 25 Two-Way FE X X X X X X State-Year Linear Trend X X X X X X e^(Mean of Outcome Below Affected Pctile) $682.73 $688.35 $1,452.44 $1,694.43 47.0% 40.6% e^(Mean of Outcome Above Affected Pctile) $854.23 $872.27 $3,857.21 $4,081.01 22.1% 21.4%

Results for retail workers Robustness (individ. controls)

Results: Dynamics

ln(rent) ln(hhincome) ln(rent/hhincome)

Dynamics for retail workers Dynamics Reg Coefficients

Results: Heterogeneity by housing supply elasticity

ln(rent) ln(hhincome) ln(rent/hhincome)

(1) (2) (3) (4) (5) (6) Common Effect: ln(MW)

• 0.0286

0.0809*

(0.0304) (0.0409)

Interaction Effect: ln(MW)*Affected 0.0241 0.0282

(0.0255) (0.0242)

Interaction Effect: ln(MW)*Saiz Elasticity

• 0.0590***

(0.0166)

Affected

• 0.149***
• 0.153***

(0.0507) (0.0486) Observations 1,586,930 1,586,930 R-Squared 0.316 0.328 Affected Wage Relative to MW 100% 100% Affected Wage Percentile Cutoff 16 16 Two-Way FE X X State-Year Linear Trend X X

Results: Heterogeneity by housing supply elasticity

ln(rent) ln(hhincome) ln(rent/hhincome)

(1) (2) (3) (4) (5) (6) Common Effect: ln(MW)

• 0.0286

0.0809*

• 0.00495

0.0601

• 0.0237

0.0208

(0.0304) (0.0409) (0.0376) (0.0451) (0.0261) (0.0234)

Interaction Effect: ln(MW)*Affected 0.0241 0.0282 0.204*** 0.206***

• 0.180***
• 0.178***

(0.0255) (0.0242) (0.0398) (0.0386) (0.0486) (0.0490)

Interaction Effect: ln(MW)*Saiz Elasticity

• 0.0590***
• 0.0350***
• 0.0239***

(0.0166) (0.0107) (0.00773)

Affected

• 0.149***
• 0.153***
• 1.152***
• 1.154***

1.003*** 1.001***

(0.0507) (0.0486) (0.0805) (0.0781) (0.0922) (0.0932) Observations 1,586,930 1,586,930 1,586,930 1,586,930 1,586,930 1,586,930 R-Squared 0.316 0.328 0.296 0.297 0.162 0.162 Affected Wage Relative to MW 100% 100% 100% 100% 100% 100% Affected Wage Percentile Cutoff 16 16 16 16 16 16 Two-Way FE X X X X X X State-Year Linear Trend X X X X X X

Concluding thoughts

• Evidence that minimum wage: raises income, may raise housing consumption,

decreases rent-to-income ratios

• Suggests: income-inelastic demand; non-homothetic preferences
• Little average effect on rental prices, but significant heterogeneity in effects by

housing supply elasticity

• Suggests: imperfections in labor & housing markets interact
• In the paper, theory-motivated formula for thinking about welfare change

using the income elasticity of housing expenditure share

Goodbye!

If you have idle or constructive thoughts or ideas, my email is: skhughes@wharton.upenn.edu

There is also an Appendix!

Back

Panel A. Outcome: ln(rent) Common Effect: ln(MW)

• 0.00783

0.0149

• 0.00278
• 0.000511

0.000115 0.0188 0.00318

• 0.0111

(0.0235) (0.0363) (0.0264) (0.0239) (0.0238) (0.0386) (0.0259) (0.0220)

Interaction Effect: ln(MW)*Affected 0.0543**

• 0.0270
• 0.0277

0.0579***

(0.0246) (0.0428) (0.0424) (0.0199)

Affected

• 0.213***
• 0.104***
• 0.159*
• 0.158*
• 0.210***

(0.0497) (0.00478) (0.0850) (0.0841) (0.0399) Observations 2,248,921 2,250,515 2,250,515 2,248,921 2,248,921 2,250,515 2,250,515 2,248,920 R-Squared 0.327 0.212 0.214 0.324 0.327 0.227 0.228 0.386

Panel B. Outcome: ln(hhincome) Common Effect: ln(MW)

• 0.0106

0.0393

• 0.00351

0.0127 0.0171 0.0214

• 0.0118
• 0.0180

(0.0328) (0.0422) (0.0359) (0.0323) (0.0321) (0.0400) (0.0358) (0.0316)

Interaction Effect: ln(MW)*Affected 0.189*** 0.118*** 0.118*** 0.182***

(0.0322) (0.0403) (0.0396) (0.0320)

Affected

• 1.104***
• 0.728***
• 1.185***
• 1.184***
• 1.030***

(0.0652) (0.0135) (0.0830) (0.0816) (0.0624) Observations 2,248,921 2,250,515 2,250,515 2,248,921 2,248,921 2,250,515 2,250,515 2,248,920 R-Squared 0.291 0.047 0.048 0.226 0.291 0.169 0.169 0.400

Panel C. Outcome: ln(rent/hhincome) Common Effect: ln(MW) 0.00273

• 0.0244

0.000726

• 0.0132
• 0.0170
• 0.00264

0.0150 0.00689

(0.0241) (0.0407) (0.0250) (0.0236) (0.0239) (0.0392) (0.0269) (0.0245)

Interaction Effect: ln(MW)*Affected

• 0.135***
• 0.145***
• 0.145***
• 0.124***

(0.0361) (0.0405) (0.0405) (0.0359)

Affected 0.892*** 0.623*** 1.026*** 1.026*** 0.820***

(0.0689) (0.0128) (0.0765) (0.0765) (0.0678) Observations 2,248,921 2,250,515 2,250,515 2,248,921 2,248,921 2,250,515 2,250,515 2,248,920 R-Squared 0.158 0.014 0.015 0.101 0.158 0.101 0.102 0.211 Affected Wage Relative to MW 100% 100% 100% 100% 100% 100% 100% 100% Affected Wage Percentile Cutoff 16 16 16 16 16 16 16 16 State FE X X X X X X X X Year FE X X X X X X X X State-Time Trend X X X X X X State-Occupation FE X X X X Occupation-Year FE X X X X Household/Individual Controls X

Results: Dynamics

ln(rent) ln(hhincome) ln(rent/income)

ln(MW) Average of Leads

• 0.024
• 0.008
• 0.016

(0.034) (0.034) (0.034)

ln(MW) Average of Lags 0.008 0.013

• 0.005

(0.021) (0.032) (0.024)

ln(MW) Difference 0.032 0.021 0.012

(0.026) (0.033) (0.023)

ln(MW)*Affected Average of Leads

• 0.026
• 0.042

0.015

(0.024) (0.029) (0.040)

ln(MW)*Affected Average of Lags 0.021 0.055

• 0.034

(0.019) (0.023) (0.022)

ln(MW)*Affected Difference 0.047 0.096

• 0.049

(0.021) (0.034) (0.042)

Observations 2,063,501 2,063,501 2,063,501 R-Squared 0.324 0.290 0.157 Affected Wage Relative to MW 125% 125% 125% Affected Wage Percentile Cutoff 25 25 25 Two-Way FE X X X State-Year Linear Trend X X X e^(Mean of Outcome Below Affected) $688.35 $1,694.43 40.6% e^(Mean of Outcome Above Affected) $872.27 $4,081.01 21.4%

Back

Results: Households headed by cashiers, clerks & salespersons

ln(rent) ln(hhincome) ln(rent/income)

(1) (2) (3) (4) (5) (6) Common Effect: ln(MW)

• 0.0365
• 0.0344
• 0.0659
• 0.103

0.0294 0.0683

(0.0509) (0.0518) (0.116) (0.119) (0.112) (0.113)

Interaction Effect: ln(MW)*Affected 0.122* 0.0620 0.260*** 0.201***

• 0.139**
• 0.139**

(0.0664) (0.0456) (0.0826) (0.0744) (0.0656) (0.0559)

Affected

• 0.327**
• 0.209**
• 1.152***
• 0.933***

0.825*** 0.724***

(0.133) (0.0916) (0.170) (0.152) (0.129) (0.113) Observations 150,880 150,880 150,880 150,880 150,880 150,880 R-Squared 0.196 0.196 0.148 0.139 0.076 0.066 Affected Wage Relative to MW 100% 125% 100% 125% 100% 125% Affected Wage Percentile Cutoff 31 47 31 47 31 47 Two-Way FE X X X X X X State-Year Linear Trend X X X X X X e^(Mean of Outcome Below Affected Pctile) $634.48 $641.69 $1,245.51 $1,460.89 50.9% 43.9% e^(Mean of Outcome Above Affected Pctile) $703.31 $714.58 $2,486.67 $2,615.47 28.3% 27.3%

Back

Results: Dynamics Retail Occupations

ln(rent) ln(hhincome) ln(rent/hhincome)

Back

Results: Dynamics Retail Occupations

ln(rent) ln(hhincome) ln(rent/income)

ln(MW) Average of Leads

• 0.084
• 0.013
• 0.071

(0.088) (0.146) (0.135) ln(MW) Average of Lags

• 0.059

0.019

• 0.078

(0.043) (0.094) (0.091)

ln(MW) Difference 0.025 0.032

• 0.007

(0.073) (0.126) (0.115)

ln(MW)*Affected Average of Leads

• 0.011
• 0.103

0.092 (0.055) (0.083) (0.089) ln(MW)*Affected Average of Lags 0.154 0.094 0.061 (0.050) (0.101) (0.074)

ln(MW)*Affected Difference 0.165 0.197

• 0.031

(0.060) (0.116) (0.103) Observations 140,156 140,156 140,156 R-Squared 0.194 0.138 0.067 Affected Wage Relative to MW 125% 125% 125% Affected Wage Percentile Cutoff 47 47 47 Two-Way FE X X X State-Year Linear Trend X X X e^(Mean of Outcome Below) $641.69 $1,460.89 43.9% e^(Mean of Outcome Above) $714.58 $2,615.47 27.3%

Back

Housing Demand: Set-Up

Straightforward consumer demand, Stone-Geary in housing services:

max

𝑌𝑗,𝐼𝑗 𝑉(𝑌𝑗, 𝐼𝑗) = 𝑌𝑗 𝜀 𝐼𝑗 − 𝜄 1−𝜀 𝑡. 𝑢.

𝑛 𝑗 = 𝑞𝑌𝑗 + 𝑟𝐼𝑗

Motivating Stone-Geary by graphing rent-to-income ratios by monthly income for different preference parameters, 1 − 𝜀:

Composite good Housing services Subsistence level

• f housing

Income (index by i) Price per unit of housing services Observed rent in the real world

Housing Demand: Set-Up

Straightforward consumer demand, Stone-Geary in housing services:

max

𝑌𝑗,𝐼𝑗 𝑉(𝑌𝑗, 𝐼𝑗) = 𝑌𝑗 𝜀 𝐼𝑗 − 𝜄 1−𝜀 𝑡. 𝑢.

𝑛 𝑗 = 𝑞𝑌𝑗 + 𝑟𝐼𝑗

Housing Demand: 𝐼𝑗 = 1 − 𝜀

𝑛 𝑗 𝑟

+ 𝜀𝜄 Housing Expenditure Share: 𝑡𝐼𝑗 =

𝑟𝐼𝑗 𝑛 𝑗 = 1 − 𝜀 + 𝜀 𝑟𝜄 𝑛 𝑗

Housing Demand: Mapping an Income Shock to Welfare

In Cobb-Douglas, can map ∆ income or ∆ consumption → ∆ utility. (holding prices fixed) In Stone-Geary, you cannot. Income expansion path is non-linear. Utility in terms of housing demand → Consider: shock to income/consumption

𝑊 𝑞, 𝑟, 𝐼𝑗, 𝜄 = 𝜀 1 − 𝜀 𝑟 𝑞

𝜀

𝐼𝑗 − 𝜄 → %∆𝑊 = 𝑊𝑞𝑝𝑡𝑢 − 𝑊𝑞𝑠𝑓 𝑊𝑞𝑠𝑓 = 𝐼𝑗

𝑞𝑝𝑡𝑢 − 𝜄 − 𝐼𝑗 𝑞𝑠𝑓 − 𝜄

𝐼𝑗

𝑞𝑠𝑓 − 𝜄

= 𝐼𝑗

𝑞𝑝𝑡𝑢 − 𝐼𝑗 𝑞𝑠𝑓

𝐼𝑗

𝑞𝑠𝑓 − 𝜄

To map to estimable parameters, rearrange terms:

1 %∆𝑊 = 𝐼𝑗

𝑞𝑠𝑓

𝐼𝑗

𝑞𝑝𝑡𝑢 − 𝐼𝑗 𝑞𝑠𝑓 −

𝜄 𝐼𝑗

𝑞𝑝𝑡𝑢 − 𝐼𝑗 𝑞𝑠𝑓 =

1 %∆𝐼 − 𝑟𝜄 1 − 𝜀 1 ∆𝑛

Housing Demand: Mapping an Income Shock to Welfare

To map to estimable parameters, rearrange terms:

1 %∆𝑊 = 𝐼𝑗

𝑞𝑠𝑓

𝐼𝑗

𝑞𝑝𝑡𝑢 − 𝐼𝑗 𝑞𝑠𝑓 −

𝜄 𝐼𝑗

𝑞𝑝𝑡𝑢 − 𝐼𝑗 𝑞𝑠𝑓 =

1 %∆𝐼 − 𝒓𝜾 1 − 𝜀 1 ∆𝑛

Subsistence rent is not observed. Back it out using change in expenditure share.

𝜁𝑡𝐼,𝑛 = ∆𝑡𝐼𝑗 ∆𝑛 𝑗 𝑛 𝑗 𝑡𝐼𝑗 = −𝜀 𝑟𝜄 𝑛 𝑗 2 𝑛 𝑗 𝑡𝐼𝑗 = −𝜀 𝑟𝜄 𝑟𝐼𝑗

𝑞𝑠𝑓

⟺ 𝒓𝜾 = −𝜁𝑡𝐼,𝑛 𝑟𝐼𝑗

𝑞𝑠𝑓

𝜀

Housing Demand: Mapping an Income Shock to Welfare

To map to estimable parameters, rearrange terms:

1 %∆𝑊 = 𝐼𝑗

𝑞𝑠𝑓

𝐼𝑗

𝑞𝑝𝑡𝑢 − 𝐼𝑗 𝑞𝑠𝑓 −

𝜄 𝐼𝑗

𝑞𝑝𝑡𝑢 − 𝐼𝑗 𝑞𝑠𝑓 =

1 %∆𝐼 − 𝒓𝜾 1 − 𝜀 1 ∆𝑛

Subsistence rent is not observed. Back it out using change in expenditure share.

𝜁𝑡𝐼,𝑛 = ∆𝑡𝐼𝑗 ∆𝑛 𝑗 𝑛 𝑗 𝑡𝐼𝑗 = −𝜀 𝑟𝜄 𝑛 𝑗 2 𝑛 𝑗 𝑡𝐼𝑗 = −𝜀 𝑟𝜄 𝑟𝐼𝑗

𝑞𝑠𝑓

⟺ 𝒓𝜾 = −𝜁𝑡𝐼,𝑛 𝑟𝐼𝑗

𝑞𝑠𝑓

𝜀

Now, estimate:

1 %∆𝑊 = 1 %∆𝐼 + 𝜁𝑡𝐼,𝑛(𝑟𝐼𝑗

𝑞𝑠𝑓)

𝜀 1 − 𝜀 1 ∆𝑛

Observed rent Estimated ∆ income Income elasticity of the housing expenditure share Estimated %∆ housing services

Results: Calculating Welfare Change

Graphical

Negative where behavior can only be rationalized by subsistence level θ<0

Reminder, calculated using:

All Private Sector Cashiers, Clerks & Salespersons <25th Percentile Wage (125% MW) <25th Percentile Wage (125% MW) <47th Percentile Wage (125% MW) <47th Percentile Wage (125% MW)

Table Reference Table 1 Table 3 Table 2 Table 4

Inputs to Welfare Calculation Income elasticity of expenditure share

• 0.80
• 0.51
• 0.69
• 0.16

Average rent ($2016) $688 $688 $642 $642 Average change in monthly income ($2016) $22 $16 $29 $29 Change in housing consumption 0.0026 0.0047 0.0062 0.0165 Welfare Change Calculations Homothetic model 0.26% 0.47% 0.62% 1.65% Non-homothetic model Preference Parameter, 1-δ = 0.5 0.34% 0.79% 0.99% 2.15% Percent Difference from Homothetic 34.45% 68.31% 59.94% 30.14% Preference Parameter, 1-δ = 0.4 0.35% 0.81% 1.02% 2.17% Percent Difference from Homothetic 36.41% 73.24% 64.04% 31.79% Preference Parameter, 1-δ = 0.3 0.37% 0.91% 1.12% 2.28% Percent Difference from Homothetic 43.89% 93.49% 80.56% 38.06% Preference Parameter, 1-δ = 0.2 0.43% 1.28% 1.50% 2.59% Percent Difference from Homothetic 66.77% 173.34% 141.31% 56.70% Preference Parameter, 1-δ = 0.1 0.89%

• 3.69%
• 15.10%

4.63% Percent Difference from Homothetic 246.92%

• 885.00%
• 2534.91%

180.31%

Results: Graphical Representation of Welfare Change from Change in Income

Asymptote where welfare change increases to infinity (i.e. marginal utility gain where H=θ)