1. MOBILITY AND ENVIRONMENTAL EQUITY: Do Housing Choices Determine Exposure to Air Pollution? Brooks Depro North Carolina State University and RTI International and Christopher Timmins Duke University and NBER August 2008

2. Does Mobility Influence Observed Environmental Risk Exposure Patterns?  Fact 1: Minorities and low income households often live in areas with polluted air.  Fact 2: U.S. Census reports 14 percent of the US population move to a new residence in each year.  These facts raise questions about the role of mobility- induced exposure.  However, the relationship between household sorting and exposure is still not well understood.

3. Closely Related Literature: Where Does Our Work Fit? Three broad groups of studies EJ studies: 1.Document correlation with different pollution measures (e.g. proximity to TRI facility, pollution concentrations, traffic studies) (Freeman, 1972; Asch and Seneca, 1978; UCC, 1987;GAO, 1983; GAO, 1995; Brooks and Sethi, 1997; Bullard,2000; Houston et al., 2004) 2.Siting decisions of pollution firms (Hamilton, 1995; Arora and Cason, 1999) 3.Sorting-induced exposure stories (Banzhaf and Walsh, 2008; Sieg, et al, 2004).

4. What Are Our Key Findings?  We find “footprints” in the data that are consistent with the sorting story:  More air pollution = less expensive home.  More housing services = more expensive home.  Tradeoff: a buyer can get more housing services for the same price by moving to a neighborhood with more pollution.  Poor/minority households are more likely to make this tradeoff.  Wealth taken from appreciating housing stocks increases poor/minority ability to avoid this tradeoff.

5. Brief Overview of the Data  The key advantage: we observe individual homebuyer choices on multiple occasions and homebuyer economic circumstances.  This allows us to test the sorting induced exposure story in a more direct way.  Air quality monitor data:  ozone and PM 10 exceedances of pollution standards  House specific exposure (distance to monitor as weights)  Sources: DataQuick (transaction prices), HMDA (race and income), CARB (air pollution data)

6. Visual Patterns of Correlation: Ozone

7. Visual Patterns of Correlation: PM 10

8. Housing Prices and Pollution: Hedonic Price Function  Sample observation for house i, located in zip code j, and selling in year t:  error term ( ) can be decomposed into a  fixed component that is specific to house i ( ) and  a time-varying component ( ).  Year indicators

9. Housing Prices and Pollution: Hedonic Price Function  Houses that sold at least twice during the period between 1990 and 2004  Hedonic Price Function: Fixed house and neighborhood attributes differenced away:  Documents correlation and explaining tradeoff  Validity checks: Compare MWTP ozone and PM10

10. Hedonic Results No House Fixed Effects House Fixed Effects (Differences) Dependent Variable:ln(price)ln(price t+1 ) − ln(price t ) Days exceeded state 1 hour ozone standard (3 year moving average) − 0.05* (0.001) −0.09 * (0.002) Days exceeded state 24 hours PM10 standard (3 year moving average) 0.03* <(0.001) −0.01* <(0.001) Year indicatorsYes Observations271,989148,755 R-Squared0.380.07

11. Examining Housing Services/Pollution Tradeoff  Construct year-specific housing service and neighborhood quality indices  Compute differences in the following variables resulting from move from a first home to a second home:  housing services index,  other neighborhood services index (i.e., zip code fixed effects),  house specific air pollution (measured using the 3-year moving average number of days exceeded for ozone and PM 10 )  Measure correlation changes in these variables for different groups.

12. Low Income Minorities and Asians Make Housing Services/Ozone Tradeoff OzonePM10 Neighborhood Quality Minority White Diff: High Income−0.010.170.08 Minority White Diff: Low Income0.180.080.09 Double Difference:−0.190.09−0.01 Asian White Diff: High Income−0.030.000.08 Asian White Diff: Low Income0.05−0.02−0.01 Double Difference:−0.080.020.09

13. Minorities Experiencing High Gains in Housing Values Avoid Housing Services/Pollution Tradeoff OzonePM10 Neighborhood Quality Minority0.220.040.02 Low Income0.270.030.02 High % Gain0.270.040.00 Low % Gain0.270.030.00 High Income0.100.08−0.08 High % Gain0.12 −0.06 Low % Gain0.080.01−0.10 Diff Income Groups: High % Gain0.15−0.090.05 Diff Income Groups: Low % Gain0.190.020.10 Double Difference:−0.04−0.10−0.05

14. Housing Wealth Effects for Asians Are Mixed OzonePM10 Neighborhood Quality Asian0.11−0.09−0.05 Low Income0.14−0.07−0.08 High % Gain0.19−0.05−0.11 Low % Gain0.10−0.09−0.06 High Income0.08−0.10−0.08 High % Gain0.02−0.05−0.09 Low % Gain0.12−0.14−0.08 Diff Income Groups: High % Gain0.170.00−0.02 Diff Income Groups: Low % Gain−0.020.060.02 Double Difference:0.19−0.06−0.04

15. No Housing Wealth Effect for Whites OzonePM10 Neighborhood Quality White0.09−0.07−0.10 Low Income0.09−0.05−0.07 High % Gain0.08−0.06−0.07 Low % Gain0.10−0.04−0.08 High Income0.11−0.09−0.17 High % Gain0.10−0.11−0.20 Low % Gain0.12−0.08−0.13 Diff Income Groups: High % Gain−0.020.050.13 Diff Income Groups: Low % Gain−0.020.040.05 Double Difference:0.01 0.09

16. Other Implications and Next Steps  Homeowners who move from declining neighborhoods may be more constrained in the housing services/pollution tradeoff.  There may be environmental justice benefits associated with improving access to credit to minority homeowners.  Correlation between race and pollution declines over our sample period.  Use formal estimation methods (systems of equations).  Examine factors the influence the length of housing spells (duration models)