Pollution, Mortality Risk, and the Distribution of Income. N.Z. Muller, P. H. Matthews, V. Wiltshire-Gordon Spring, 2017

Transcription

1 Pollution, Mortality Risk, and the Distribution of N.Z. Muller, P. H. Matthews, V. Wiltshire-Gordon Spring, 2017

2 Introduction Considerable attention/interest in distributional issues in economics. Piketty Capital in the 21 st Century, Stiglitz The Price of Inequality, AEA Panels in Sloan Foundation, NBER conference on distributional outcomes of energy/environmental policy, 2016.

3 Introduction. Analyses targeted at ever more rigorous examinations/estimation of income among the top 1%. Imputations for top-coded data Jenkins et al., 2009; Armour, Burkhauser, Larrimore, Analyses showing how the distribution has changed over time. Piketty and Saez, 2003; Piketty, 2014.

4 Introduction This analysis argues for a redefinition of income a la Nordhaus and Tobin (1972). Measures of output such as GDP omit critical aspects of economic activity. Leisure time, home production, natural resources in situ, pollution damage. Each of the above omissions are interesting and likely important for distributional analysis.

5 Introduction This analysis focuses on air pollution damages.

6 Why Local Air Pollution (1)? Source: Muller, 2014

7 Why Local Air Pollution (1)? Source: Muller, [423] 11,548 [11,393] 11, [0.036] [-5.50] 2.42 [2.54] 2.10 Source: Muller, 2014

8 Why Local Air Pollution (2)? Canada GED/VA 1999 Atlantic Ocean Pacific Ocean Mexico Source: Muller, Gulf of Mexico

9 Why Local Air Pollution (2)? Canada GED/VA Atlantic Ocean Pacific Ocean Mexico Source: Muller, Gulf of Mexico

10 Why Local Air Pollution (3)? Excellent data on air pollution: PM 2.5 and O 3 *.

11 Monitoring Network PM2.5

12 Monitoring Network O3

13 Why Local Air Pollution (3)? Excellent data on air pollution: PM 2.5 and O 3 *. Established health impacts from epidemiology**. Non-market valuation of health impacts from economics***. Implication is that estimation of damages from these pollutants quite feasible.

14 Introduction: What the paper does. Adjusted income = H.H. (PM 2.5 and O 3 damage) Compute adjusted income for Estimate Gini coefficient for market and adjusted income. Compute percent shares for both income measures. Decompose results by region, race, and ethnicity.

15 Preview of Results Gini Coefficients for 2011: Market income: (Official Census Gini: 0.475) Market income w/ pollution monitor: Adj. income: More unequal than South Africa and Haiti By Race: African American: Mkt. Gini: 0.484, Adj. Gini: White Mkt. Gini: 0.472, Adj. Gini: 0.671

16 Preview of Results Adj. Gini Mkt. Gini

17 Outline of Talk Introduction. Literature Review and Context. Methods. Market. Damage Estimation. Adjusted. Results. Conclusions/Future Directions.

18 Outline of Talk Introduction. Literature Review and Context. Methods. Market. Damage Estimation. Adjusted. Results. Conclusions/Future Directions.

19 Literature Review Inequality in market income. Kuznets 1955, Piketty, Saez, 2003; Stiglitz, 2012; Piketty, 2014; Autor, Integration of environment/natural resources into national accounts. Leontief, 1970; Nordhaus and Tobin, 1972; Bartelmus, 2009; Muller, Mendelsohn, Nordhaus, 2011 Pollution damage. Mendelsohn, 1980; Burtrawet al., 1998; Muller and Mendelsohn, 2009

20 Literature Review Inequality in environmental outcomes. Measures of exposure. Perlin, et al., 1995; Torras, 2005; Lopez, 2009; Clement, and Meunie, 2010; Ash et al., 2013; Boyce, Zwickly, Ash, 2014; Policy Analyses. Gianessi, Peskin, Wolff, 1979; List, 1999; Heiland Wodon, 2000; Ravallion, Heil, Jalan, 2000; Millimetand Slottje, 2002a; Gray, Shadbegian, Wolverton, 2012; Shadbegian and Gray, 2012 Measurement/technical. Millimetand Slottje, 2002b; Oneillet al., 2003; Levy, Wilson, Zwack, 2007

21 Literature Review Epidemiology. Mortality association with pollution exposure. PM 2.5 : Krewskiet al., 2009; Lepeule, et al., O 3 : Bell et al., Valuation of mortality risk. USEPA meta-analysis, 26 revealed and stated preference studies. VSL about $10 million in ($2011). -VSL elasticity 0.4 (Klecknerand Neuman, 1999)

22 Outline of Talk Introduction. Literature Review and Context. Methods. Market. Damage Estimation. Adjusted. Results. Conclusions/Future Directions.

23 Methods: Market Individual and household income data provided by IPUMS. Annual American Community Survey datasets. About 3-7 million observations/year. Household income (>95% positive income). Location (county), age, race, ethnicity. Extract 2011 through Preliminary Results for

24 Methods: Mortality Rates Annual, county-level death counts. CDC Wonder. Age, race, ethnicity. Suppressed and unreliable data.

25 Methods: Pollution data USEPA AIRS monitoring data. Hourly observations of PM 2.5 and O 3 at about 1,000 distinct monitoring stations across the contiguous U.S. Collapsed to annual average: epidemiology. These are historically targeted for enforcement (mostly metropolitan). Matched to households by county.

26 Methods: Pollution damage Mortality risk associated with exposure to both PM 2.5 and O 3. Accounts for ~90% of pollution damage. Method in Section 812 studies and RIAs.,, =,, 1 1 exp, where:,, = baseline all-cause mortality rate, age-cohort (a), location (r), time (t) = empirically estimated concentration-response parameter for pollutant (p)., = ambient concentration of pollutant (p), location (r), time (t).

27 Methods: Pollution damage Monetary damage = mortality risk increment due to air pollution exposure times value of mortality risk.,, =,, VSL adopted from USEPA s guidance: $10.2M. Uniform across all populations. Sensitivity analysis, allowed to vary by income.

28 Methods: Adjusted For individual (i), time (t), location (r), adjusted income is:,, =,,,, In principle, sign indeterminant. Gini Coefficient (Van Kerm, 2010; Yitzhaki, 1998)! = 2#$! %(! ), 1 Ϝ! Generalization permits negative values.

29 Outline of Talk Introduction. Literature Review and Context. Methods. Market. Damage Estimation. Adjusted. Results. Conclusions/Future Directions.

30 Results Summary Statistics Pollution levels Pollution damage levels Region Race Distributional Metrics Preliminary Results for Determinants of county Ginis.

31 Results: Market (nominal) All Households Mean 68,616 A 71,445 75,227 76,742 Median 49,400 51,000 53,300 54,700 Max 3,563,100 2,087,000 2,090,000 1,951,000 Min -18,490-17,300-19,770-14,700 SD 73,368 76,100 82,109 83,350 Count 1,204,830 1,207,725 1,211,264 1,218,240 Gini Est Census Gini B Households in Counties with Pollution Monitors Mean 76,313 77,993 81,867 83,816 Median 54,000 55,055 57,700 59,800 Max 1,760,400 1,711,330 2,052,500 1,802,000 Min -14,800-15,000-14,000-14,000 SD 81,748 83,542 89,301 91,028 Count 455, , , ,894 Gini Est Source: Muller, Matthews, Wiltshire-Gordon, 2016

32 Results: Particulate Matter Levels and Household in 2011 PM 2.5 (ug/m^3) Household ($2011)

33 Results: Baseline Mortality Rate and Household in 2011 Mortality Rate Household ($2011)

34 0 Damages ($/household) Results: Pollution Damages by Age of Exposed Population, Age (Years) 80

35 Results: Pollution levels and damage in All Regions Household O 3 PM 2.5 Per Capita Damage/ ($) (ppb) (ug/m 3 ) Damage($) Mean 76, , Median 54, , SD 81, , Min. -14, Max. 1,760, , , Obs. 452, , , , ,023

36 Why Local Air Pollution (1)? Source: Muller, [423] 11,548 [11,393] 11, [0.036] [-5.50] 2.42 [2.54] 2.10 Source: Muller, 2014

37 Regional Results: Pollution levels and damage in 2011 Region Household ($) O 3 PM 2.5 (ug/m 3 ) Per Capita Damage($) Damage/ (ppb) East Mean 82, , Median 58, , Obs. 101, , , , ,055 Midwest Mean 66, , Median 49, , Obs. 65,084 65,084 65,084 65,084 64,237 South Mean 73, , Median 50, , Obs. 105, , , , ,946 West Mean 78, , Median 55, , Obs. 180, , , , ,785

38 Results by Race: Pollution Levels and Damage in 2011 Region Household ($) O 3 PM 2.5 (ug/m 3 ) Per Capita Damage($) Damage/ (ppb) White Mean 82, , Median 59, , Obs. 343, , , , ,533 African American Mean 50, , Median 35, , Obs. 60,630 60,630 60,630 60,630 59,267 Asian Mean 96, , Median 73, , Obs. 38, , , , , Hispanic Mean 77, , Median 55, , Obs. 480, , , , ,797.00

39 Results Distributional Metrics Pollution, risk, damage and market income. Gini coefficients: Market and adjusted income. Region Race Percentiles. Sensitivity Analysis. Preliminary Results:

40 Results: Pollution and Mortality Risk Trends Year MR_Index O3_Index PM_Index

41 Results: Pollution and baseline mortality risk Ginis Year Gini_basemort Gini_O3_ppb Gini_PM25_ugm

42 Results: Pollution and baseline mortality risk Ginis-indexed Year MR_Index O3_Index PM_Index

43 Market and Adjusted e e e e ($2011) kdensity hhincome kdensity corhhincopm

44 Gini Coefficients Year Obs. Market , (0.001) A (0.000) , (0.001) (0.001) , (0.001) (0.001) , (0.001) (0.001) All Households Poll. Adj. Damage (0.001) (0.001) (0.001) (0.001) Share PM 2.5 Share O 3 Ratio (Adjusted /Market) Source: Muller, Matthews, Wiltshire-Gordon, 2016

45 Gini Coefficients Year Obs. Market , (0.001) , (0.001) , (0.001) , (0.001) White Households Pollution Adj. Damage (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Share PM 2.5 Share O 3 Ratio (Adjusted /Market) Source: Muller, Matthews, Wiltshire-Gordon, 2016

46 Gini Coefficients Year Obs. Market , (0.001) , (0.001) , (0.001) , (0.001) African American Households Pollution Adj. Damage (0.001) (0.001) (0.001) (0.002) (0.004) (0.003) (0.003) (0.003) Share PM 2.5 Share O 3 Ratio (Adjusted /Market) Source: Muller, Matthews, Wiltshire-Gordon, 2016

47 Quantiles Market Adjusted Market Adjusted < ( ) A ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) Quantiles Market Adjusted Market Adjusted < ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) Source: Muller, Matthews, Wiltshire-Gordon, 2016

48 Quantiles Market Adjusted Market Adjusted < ( ) A ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) Quantiles Market Adjusted Market Adjusted < ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) Source: Muller, Matthews, Wiltshire-Gordon, 2016

49 Quantiles Market Adjusted Market Adjusted < ( ) A ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) Quantiles Market Adjusted Market Adjusted < ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) Source: Muller, Matthews, Wiltshire-Gordon, 2016

50 Quantiles Market Adjusted Market Adjusted < ( ) A ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) Quantiles Market Adjusted Market Adjusted < ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) Source: Muller, Matthews, Wiltshire-Gordon, 2016

51 Figure 1: Gini Coefficients by Census Division. Source: Muller, Matthews, Wiltshire-Gordon, 2016

52

53 Figure 2: Ratios of Median by Race White/Af-Am Median Market White/Af-Am Median Adjusted White/Hisp Median Market White/Hisp Median Adjusted Source: Muller, Matthews, Wiltshire-Gordon, 2016

54 Sensitivity Analysis 2011 Modeling Scenario Obs. Market Externality Market - Externality Default 455, (0.001) A (0.000) (0.001) Low VSL (0.001) (0.001) (0.001) Alternative PM Dose-response (0.001) (0.001) (0.003) VSL (0.001) (0.001) (0.001) Quantile Default < ( ) ( ) ( ) ( ) ( ) Low VSL Alternative PM Dose- Response ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) VSL ( ) ( ) ( ) ( ) ( ) Source: Muller, Matthews, Wiltshire-Gordon, 2016 A = bootstrap standard errors.

55 Sensitivity Analysis 2011 Modeling Scenario Obs. Market Externality Market - Externality Default 455, (0.001) A (0.000) (0.001) Low VSL (0.001) (0.001) (0.001) Alternative PM Dose-response (0.001) (0.001) (0.003) VSL (0.001) (0.001) (0.001) Quantile Default < ( ) ( ) ( ) ( ) ( ) Low VSL Alternative PM Dose- Response ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) VSL ( ) ( ) ( ) ( ) ( ) Source: Muller, Matthews, Wiltshire-Gordon, 2016 A = bootstrap standard errors.

56 Gini Coefficients: Adj. Gini Mkt. Gini

57 Gini Coefficients: Adjusted, By Census Division Div. 1 Div. 2 Div. 3 Div. 4 Div. 5 Div. 6 Div. 7 Div. 8 Div

58

59 Gini Coefficients: Adjusted, By Race White Afr. American Asian Hispanic

60 Gini Coefficients: Market, By NAAQS Attainment Current NA All Never NA

61 Gini Coefficients: Adjusted, By NAAQS Attainment Current NA Never NA All

62 How does environmental policy affect inequality? The NAAQS form the backbone of the CAA. Periodically, the NAAQS are adjusted PM 2.5 NAAQS review led to adjustment. 71 FR 61144: Standard didn t change but attainment designations did. Compute county Ginis. Use DID framework to test whether attainment status affects Gini.

63 NAAQS Attainment Status and Ginis ), = * +, ) + -, ). +/ + ) +0 ), Treatment is: Counties in NA with only 2006 NAAQS Counties in NA with only 1997 NAAQS Counties in NA with both 1997, 2006 NAAQS After is post 2009: when 2006 NAAQS attainment status first announced. County Ginis:

64 County Level Gini Estimates Density Household Gini

65 County Level Gini Estimates Density Pollution Damage Gini

66 County Level Gini Estimates Density Adjusted Gini

67 Determinants of County Gini Coefficients (1) (2) (3) Household Adj. Pollution Gini Gini Damage Gini NA: 1997 & *** ( ) (0.0104) ( ) adj. R-sq N NA: * ( ) (0.0150) ( ) adj. R-sq N NA: ( ) (0.0181) ( ) adj. R-sq N Standard errors in parentheses Note: * p<0.10, ** p<0.05, *** p<0.01

68 Parallel Trends Adj. Gini Year Solid = treatment; dash = control

69 Outline of Talk Introduction. Literature Review and Context. Methods. Market. Damage Estimation. Adjusted. Results. Conclusions/Future Directions.

70 Conclusions Gini for adjusted income on par with South Africa & Haiti. Pollution damages behave like a symmetric and sharply regressive tax. Ramifications concentrated in African American households and white households. Ramifications concentrated in Central U.S. Counties never in non-attainment with CAA NAAQS, higher Gini for adjusted income.

71 Conclusions While Gini for market income has risen over time, that for adjusted income has fallen sharply. Fallen most rapidly for African American households and white households. Fallen most rapidly in Central U.S. Fallen most rapidly in counties never in nonattainment with the CAA NAAQS.

72 Future Directions Continue with work on Test whether there are effects on adjusted Gini due to policy interventions. Is environmental policy a means to reduce inequality? Are changes in attainment status associated with the Ginis? Other pollutants.

73 Thank you.

74 Determinants of the Ginis? Test whether NAAQS attainment affects the adjusted income Gini. 4 1, = * +. 1, + 1,, -, 3 1,, + 1 +/ +0 1, Several ways to code NAAQS attainment. Default is change in attainment status from prior period. Requires estimation of state Ginis.

75 Determinants of the Ginis? Test whether NAAQS attainment affects the pollution damage Gini. 1, = 4 * +. 1, + -, 3 1,, + 1 +/ +0 1, 1,, Several ways to code NAAQS attainment. Default is change in attainment status from prior period. Requires estimation of state Ginis.

76 Attainment Status: All Determinants of State Gini Coefficients (1) (2) (3) Market Adjusted Pollution Damage Current NA *** * ( ) ( ) ( ) NA Lag *** ( ) ( ) ( ) NA Lag *** ( ) ( ) ( ) NA Lag ( ) ( ) ( ) NA Lag *** ( ) ( ) ( ) NA Lag *** ** ( ) ( ) ( ) Frac. White *** ( ) (0.0294) (0.0157) Frac. Afr.~ ** (0.0365) (0.166) (0.0542) Frac. Hisp~c ** (0.0217) (0.0868) (0.0389) _cons 0.438*** 0.689*** 0.673*** ( ) (0.0122) ( ) adj. R-sq N Standard errors in parentheses Note: * p<0.10, ** p<0.05, *** p<0.01

77 Attainment Status: PM Determinants of State Gini Coefficients (1) (2) (3) Market Adjusted Pollution Damage Current NA ** *** ( ) ( ) ( ) NA Lag ** *** ( ) ( ) ( ) NA Lag ( ) ( ) ( ) NA Lag *** ( ) ( ) ( ) NA Lag ** ( ) ( ) ( ) NA Lag *** ** ( ) ( ) ( ) Frac. White *** ( ) (0.0306) (0.0152) Frac. Afr.~ ** (0.0348) (0.156) (0.0530) Frac. Hisp~c (0.0222) (0.0786) (0.0369) _cons 0.435*** 0.677*** 0.673*** ( ) (0.0103) ( ) adj. R-sq N Standard errors in parentheses Note: * p<0.10, ** p<0.05, *** p<0.01

78 Determinants of the Ginis? Test whether NAAQS attainment affects the Ginisin DID framework. 1, = * +, 1, + -, 1,. +/ +0 1, Requires estimation of state Ginis. In DID: implementation of PM 2.5 NAAQS. States in compliance with 97 NAAQS. Multi-periods before and after intervention.

79 State Ginis: Frequency Gini Household

80 State Ginis: Frequency Gini Adjusted

81 State Ginis: Frequency Gini Pollution Damage

82 Determinants of State Gini Coefficients NA 2006/ (1) (2) (3) A 1997 Market Adjusted Pollution Damage Treatment post ** ** ( ) (0.0221) ( ) adj. R-sq N Treatment post ** *** ( ) (0.0221) ( ) adj. R-sq N Treatment post *** *** ( ) (0.0231) ( ) adj. R-sq N Treatment post ** ** ( ) (0.0261) ( ) adj. R-sq N Treatment post ( ) (0.0346) ( ) adj. R-sq N Standard errors in parentheses Note: * p<0.10, ** p<0.05, *** p<0.01