Vulnerability to Air Pollution:Is There Any Inequity in Exposure?

Vulnerability to Air Pollution: Is There Any Inequity in Exposure?

E

Economic and Political WeeklyJuly 28, 20073159using 1960 Census data for the Kansas City, Washington DCand St Louis concluded that blacks and low income persons facedthe highest exposure to both sulphur dioxide (SO2) andparticulates.Similarly, Zupan (1973) found a positive correlation betweenSO2level and particulates, and the percentage of low-income house-holds in New York city. Kruvant (1975) analysed the exposureto non-point source emissions of carbon monoxide (CO) andhydrocarbons (HC) in the Washington DC area and discoveredthat low incomes, low rents, low proportions of professional andmanagerial workers, and high proportions of black residentscharacterised the most highly polluted zones. Asch and Seneca(1978) studied 284 cities and 23 states and found a significantpositive relationship between exposure to particulates and moredensely populated cities with lower incomes, lower educationallevels, and higher proportions of non-white residents.Brajer and Hall (1992), in their study of the South Coast AirBasin of California found evidence of higher exposure to ozoneand particulates among low-income, and ethnic groups, mainlyblack and Hispanic residents. The study also found the populationdensity to be positively related to exposure, whereas a high-income group showed a strong negative association with expo-sure. Harrison and Rubinfeld (1978) based on their study for theBoston metropolitan area concluded that the physical benefitsof pollution abatement accrued more to poor than to rich.Most of the studies until the early 1990s used ambient con-centration as an indication of pollution. In a significant departure,Brooks and Sethi (1997) computed an index based on toxicityof 150 most significant air toxics rather than ambient concen-trations for the entire US. The study then looked into whetherexposure varied across various socio-economic characteristicsincluding income, race, educational attainment, housing tenure,and the propensity of communities to engage in collective action.The study found greater exposure in black communities even aftercontrolling for a variety of variables.Of late, some studies have reinforced the above-mentionedresults. Villeneuve et al (2003) evaluated the relationship be-tween daily levels of particulate and gaseous phase pollutantsand mortality for different classes of income. The study foundincreased risk of all-cause and cardiovascular mortality at lowerlevels of income. Similarly, an inventory of pollution releasesin England showed that 90 per cent of polluting factories inLondon were in areas with below-average income [FOE-UK1999],2thereby affecting the population disproportionately.Jarrett et al (2004) in their study of Hamilton city, Canada,found that increased mortality was associated with air pollutionexposure in intra-urban zones with lowersocio-economic char-acteristics. Low educational attainment andhigh manufacturingemployment in the zones were significantly andpositively relatedwith the acute mortality effects of air pollution exposure.Of late, efforts have been put in to see the extent of inequalitiesin exposure to other mediums like noise, waste treatment anddisposal, etc. See for example, Brainard et al (2004) for noiseexposure, Pastor et al (2001) for siting and location of toxicstorage and disposal facilities (TSDFs).Evidence from Developing andTransition CountriesRecent efforts have also been directed to see the impact ofexposure to air pollution in developing and middle-incomecountries. In a study of six regions of Sao Paulo,Brazil, Martinset al(2004) suggested that socio-economic deprivationacted asan effect modifier of the association between airpollution andrespiratory deaths in elderly people for an increase of 10 μg/m3.Thestudy found a negative effect of PM10 on both percentageof people with college education and high family income, andpositive association with the percentage of people livingin slums.Based on the results, the study concluded that poverty representedanimportant risk factor that should be taken into account whendetermining the health consequences of environmental contami-nation. In another study on air pollution and income, Wheeler(2000) found that low-income families were more likely to livein polluted areas.In India, no study exists looking into environmental justiceaspect. The studies in India like Murty et al (2003), Cropper etal(1997), among others have confined themselves to only quan-tifying the health damages from air pollution. These studies talkabout human population as a whole but are silent on the dis-tribution aspects of the pollution with respect to their socio-economic characteristics. Therefore, the present study fills thegap. In order to establish the relationship, the present studycomputes an exposure index for a household. This index is thenused to find out the nature of the relationship with the socio-economic characteristics of the residents of Delhi.IIIIndex of ExposureIdeally speaking, a household index of exposure should besensitive to (i) the distance of the exposed person from the source,and (ii) the time spent by the exposed person under the exposure.Finding the distance of the exposed person from the source needsa modelling exercise, whereas finding the exposed time requiresthe knowledge of person’s activities during the day. Unfortu-nately, no study exists that tries to compute the exposure indexat the household level. Studies like Brooks and Sethi (1997) havecomputed the index at a region level. The present study assumesthat in and around a region/pollution monitoring station (PMS),each individual is exposed equally, but the average householdexposure varies depending upon the number of household membersand their age distribution.The Central Pollution Control Board (CPCB) measures onlythree major pollutants – SPM, SO2 and NOx – at seven differentlocations in Delhi. The exposure index for the households iscomputed at these seven locations only. The pollution profileof Delhi indicates that among the three major pollutants, it isonly SPM that exceeds the stipulated standards. The data showsthat from 1999 to 2003, SPM varied from 430 μg/m3 to 496μg/m3 in ITO PMS, whereas the value of SO2 and NOx ranged closeto 10 and 72 μg/m3 respectively against the standard of 80 μg/m3 [Kathuria 2005].3 Thus the index is computed for SPM only.Household Air Pollution Exposure IndexThe area specific ambient levels of SPM in μg/m3 are usedto get the household air pollution index for every PMS by usingthe data collected through a primary survey on the age profileof a household in that particular area. For every household, adifferent exposure time is assumed for different categories ofpeople because even if they are exposed for the same duration,the impact would be more on older people and children. It isassumed that an adult male and female are exposed daily for five

(Nam*Tam + Naf *Taf + Nom *Tom + Nof *Tof + Nc*Tc) 24* hh size

Economic and Political WeeklyJuly 28, 20073162Table 6: Robustness Check: Estimates for All PMSs (N=347)S NoVariableLnexpo1Lnexpo2Lnexpo3Lnexpo4Lnexpo5(1)(2)(3)(4)(5)1Education0.006*0.0076*0.00560.0057*0.0059* (1.83)(1.85)(1.59)(1.66)(1.71)2Education2-0.0003-0.0003*-0.0002-0.0002-0.0002(-1.65)*(-1.65)(-1.42)(-1.46)(-1.48)3Income-0.0123*-0.0161*-0.0178*-0.0157*-0.0136*(-2.39)(-2.74)(-3.51)(-3.15)(-2.71)4Caste0.0140.01680.023*0.01920.0156(1.02)(0.98)(1.61)(1.38)(1.13)5Religion (Hindu)-0.019-0.0222-0.0194-0.0175-0.0159(-0.75)(-0.80)(-0.77)(-0.70)(-0.63)6Owned house-0.020-0.0217-0.0179-0.0181-0.0185 (-1.42)(-1.28)(-1.17)(-1.25)(-1.31)7Propchild_old0.287*0.3514*0.2390*0.2637*0.2864*(11.75)(12.34)(8.72)(10.83)(12.37)8Propfem0.022-0.03180.02810.03060.0322(0.72)(-0.88)(0.80)(0.96)(1.07)9Outdoor0.046*0.0637*0.0751*0.0672*0.0591*(1.98)(2.33)(2.98)(2.89)(2.62)10PMS1_dummy-0.401*-0.4021*-0.3994*-0.4026*-0.4052*(-21.82)(-18.21)(-20.20)(-21.55)(-22.20)11PMS2_dummy-0.379*-0.3869*-0.3835*-0.3829*-0.3819*(-20.26)(-17.52)(-19.36)(-20.65)(-21.10)12PMS3_dummy-0.275*-0.2786*-0.2785*-0.2771*-0.2754*(-12.54)(-11.32)(-12.15)(-12.68)(-12.90)13PMS4_dummy-0.308*-0.3058*-0.3027*-0.3048*-0.3064*(-17.32)(-14.33)(-15.81)(-17.06)(-17.78)14PMS5_dummy-0.283*-0.2818*-0.2728*-0.2799*-0.2862*(-15.70)(-13.31)(-14.02)(-15.26)(-16.04)15PMS6_dummy-0.134*-0.1295*-0.1312*-0.1348*-0.1380(-7.68)(-6.13)(-7.20)(-7.76)(-8.14)16Constant4.70* 4.6610*4.7004*4.6966*4.693*(140.75)(121.51)(131.36)(143.26)(149.32)17R20.720.670.70.720.7318F66.365163.568.5173.49Notes:Figures in parentheses are t-ratios. * Indicates significance levels atminimum 10 per cent.category, having fewer years of schooling, and belonging to thelower castes are prone to be more exposed to the air pollutionin Delhi.Econometric EstimationIn order to see whether environmental equity exists with respectto exposure or it is disproportionately affecting members ofcertain castes and religions in different parts of Delhi, the modelgiven in Section III is estimated. Table 4 reports the estimatesof the log-linear model. Column 2 reports the results for theaverage SPM for all the PMSs. White’s test shows the presenceof heteroskedasticity, thus the model is re-estimated after correc-ting for heteroskedasticity.The coefficient of the caste variable (row 4) shows a positiverelationship between caste and the level of exposure. Thisimpliesthat lower caste households are more exposed to pollu-tion, but the difference is not statistically significant. Similarly,though minority communities are exposed more to pollution,impact is not statistically significant (row 5). However, withrespect to household income, a rich household seems to be lessexposed to air pollution, thereby indicating their vulnerabilityto pollution.With respect to controlling variables, as hypothesised, ahousehold having a higher proportion of vulnerable people (row7)and higher outdoor active members (row 9) tends to have largerexposure from air pollution, whereas gender distribution (row8)does not have any significant impact on the exposure.Row 1 gives the coefficient of education. Surprisingly, therelation comes out to be contrary to the notion that a literatehousehold due to greater awareness would be less exposed.However, the effect of education is not statistically significant.It is generally argued that education often has a thresholdeffect.Theeffect of education leading to increased awarenessmay be felt only after the household has achieved a certainlevelofliteracy. In order to see this non-linear impact, themodelisa re-run with a square term of education (education2).Column 3 of the table reproduces the results. Based on thesignand significance, it can easily be seen that at lower levelof literacy, the exposure is high but when the literacy level ina household increases, it has a definite negative impact onTable 3: Mean Exposure for Different Categories,1999-2003 (μg/m3)S NoCategoriesMean Exposure1All persons474.932Lower castes (SC/ST)500.973Upper castes469.494No secondary school (≥ 18 years)491.835Secondary school (≥18 years)488.556Graduate466.587Postgraduate448.468Renter occupied481.049Owner occupied473.6610Income (Rs 0.12-1 lakh)484.4111Income (Rs 1.12-3 lakh)470.2512Income (above Rs 3 lakh)440.93Table 4: Estimates of the Log-Linear HeteroskedasticityCorrected Ordinary Least Squares Model (N = 347) (DependentVariable = Household Specific Exposure Index)S NoVariableAll PMSsAll PMSs(1)(2)(3)1Education0.001 (0.77)0.006* (1.83)2Education2-0.0003* (-1.65)3Income-0.014* (-2.91)-0.0123* (-2.39)4Caste0.013 (0.99)0.014 (1.02)5Religion-0.016 (-0.65)-0.019 (-0.75)6Owned house-0.020 (-1.45)-0.020 (-1.42)7Propchild_old0.287* (11.62)0.287* (11.75)8Prop fem0.020 (0.63)0.022 (0.72)9Outdoor0.047* (2.03)0.046* (1.98)10PMS1_dummy-0.400* (-21.60)-0.401*(-21.82)11PMS2_dummy-0.378*(-19.99)-0.379*(-20.26)12PMS3_dummy-0.274*(-12.49)-0.275*(-12.54)13PMS4_dummy-0.308* (-17.07)-0.308*(-17.32)14PMS5_dummy-0.283*(-15.45)-0.283*(-15.70)15PMS6_dummy-0.133*(-7.66)-0.134*(-7.68)16Constant4.72* (140.08)4.70* (140.75)17R20.720.7218F60.0366.36Note:Figures in parentheses are t-ratios.* Indicates significance levels at minimum 10 per cent level.Table 5: Different Variants for Constructing Exposure Index –Hours of Exposure AssumedS NoExposureAdult MaleAdult FemaleOld MaleOld FemaleChildren1Exposure 1559682Exposure 2549683Exposure 3558584Exposure 4558685Exposure 555878Note:Exposure 1 is same as the one mentioned in Section III.

Economic and Political WeeklyJuly 28, 20073163exposure. This negative relationship could be due to the factthatahousehold with higher aggregate education will be moreaware about the negative impacts of air pollution. Hence, mayhave taken preventive measures accordingly (for subsequentanalysis, education in square term has been included in theanalysis). With respect to all other variables, the model behavesidentically.Testing for RobustnessAs mentioned earlier, not a single study has attempted toconstruct an outdoor exposure index. Therefore, in order tocheckfor the robustness of the results, the exposure index is re-computed assuming different exposure hours for different cat-egory of household members. Table 5 indicates how each typeof members have been assumed to be exposed.Table 6 gives the results for each variant of the Exposure Index.From Table 6 it can easily be inferred that the results are fairlyrobust to the alternate mode of construction of exposure index.Income (row 3) has a negative relation with the exposure. Simi-larly, low caste households (row 4) are significantly more exposedto pollution in variant 3. Religion (row 5) has no impact on theexposure.With respect to controlling variables, a household having alarger proportion of active members (row 9) and more vulnerablepeople (row 7) tends to have larger exposure. However, the impactof education becomes insignificant in some variants of the model,though the variable retains the same sign. On the other hand,owning a house (row 6) makes people less vulnerable to pollutionexposure but the effect is not statistically significant.Based on the results, it can be concluded that the evidenceof environmental inequity with respect to caste and religion isfairly weak in the Indian context. However, poor people facediscrimination in terms of exposure.Environmental Equity across PMSs TypesAs mentioned, of the seven PMSs, two are on industrial areas– the environmental inequity may be less prevalent in industrialareas, given the fact that an industrial area is more polluted. Thus,separate analysis is carried out for residential and industrial areasto verify environmental inequity. Table 8 reports the resultsaccordingly. Columns 2 and 3 report results for residential andindustrial PMSs respectively. It needs to be mentioned at theoutset that since the pollution level is higher at industrialareas,accordingly, the exposure would be more. As a result, theimpactofdifferent variables like literacy, income, etc, would berelatively less.From Table 8, it can easily be seen that the exposure isdifferently impacted by socio-economic characteristics, depend-ing upon where the households are staying. However, caste(row4), and religion (row 5) play no role in either locations, butincome (row 3) certainly plays a key role in reducing exposureto pollution in industrial areas. A higher value of the coefficientin industrial areas (-0.0157) than the residential ones (-0.008)means that the location of such households is an important factor,thus, supporting our earlier results (Table 3) that the poorhouseholds are more exposed as a whole.Education has a direct impact on exposure in residential areas(rows 1 and 2), however in industrial areas, where the level ofpollution is relatively high, education plays no role in defensiveactivities. The coefficient of the house-ownership variable (row6) behaves differently in the areas, though it is not significantat either location. Row 7 examining the vulnerability of olderpersons and children indicates that irrespective of location –residential or industrial – this group is exposed more to pollution.The impact on outdoor active members has come out to be correctin residential areas. However, when a dwelling has a house inindustrial locations, it does not matter whether the householdTable 7: Estimating the Impact on Residential and IndustrialAreas Separately (Dependent Variable: Log Exposure)S NoVariableResidential AreasIndustrial Areas(1)(2)1Education0.0108* (2.66)-0.0035 (-0.55)2Education2-0.0005* (-2.26)0.0001 (0.59)3Income-0.0083 (-1.24)-0.0157* (-1.86)4Caste0.0263 (1.43)-0.0154 (-0.69)5Hindu-0.0157 (-0.47)-0.3375 (-1.39)6Owned house-0.0219 (-1.37)0.0053 (0.16)7Propchildold0.2926* (9.99)0.2880* (6.92)8Propfem0.0350 (0.81)0.0030 (0.08)9Outdoor0.0603* (2.10)0.0047 (0.12)10PMS1_dummy-0.4005* (-20.20)11PMS2_dummy-0.3808* (-19.73)12PMS3_dummy-0.2746* (-12.58)13PMS4_dummy-0.1757* (-9.70)14PMS5_dummy-0.2791* (-14.41)15Constant4.6677* (119.03)4.673* (83.80)16R20.740.6617F63.0423.1118N245102Notes:Figures in parentheses are t-ratios. * indicates significance levels atminimum 10 per cent.Table 8: Robustness Check: For Residential Areas (N = 245)S NoVariableLnexpo1Lnexpo2Lnexpo3Lnexpo4Lnexpo5(1)(2)(3)(4)(5)1Education0.0108*0.0127*0.0099*0.0100*0.0103*(2.66)(2.71)(2.39)(2.50)(2.55)2Education2-0.0005*-0.0006*-0.0005*-0.0005*-0.0005*(-2.26)(-2.28)(-2.07)(-2.13)(-2.14)3Income-0.0083-0.0115-0.0139*-0.0119*-0.0100(-1.24)(-1.51)(-2.08)(-1.81)(-1.49)4Caste0.02630.02890.0357*0.0310*0.0267(1.43)(1.31)(1.90)(1.69)(1.46)5Hindu-0.0157-0.0190-0.0217-0.0169-0.0125(-0.47)(-0.53)(-0.65)(-0.51)(-0.38)6Owned house-0.0219-0.0262-0.0200-0.0198-0.0197 (-1.37)(-1.39)(-1.15)(-1.20)(-1.24)7Propchild_old0.2926*0.3601*0.2374*0.2662*0.2924*(9.99)(10.83)(7.14)(9.19)(10.68)8Propfem0.0350-0.02220.04120.04540.0481(0.81)(-0.47)(0.86)(1.06)(1.19)9Outdoor0.0603**0.0821*0.0893*0.0812*0.0728*(2.10)(2.54)(2.84)(2.87)(2.67)10PMS1_dummy-0.4005*-0.4027*-0.3971*-0.4010*-0.4043*(-20.20)(-17.14)(-18.46)(-19.85)(-20.58)11PMS2_dummy-0.3808*-0.3906*-0.3838*-0.3839*-0.3835*(-19.73)(-17.33)(-18.33)(-19.84)(-20.57)12PMS3_dummy-0.2746*-0.2799*-0.2780*-0.2765*-0.2748*(-12.58)(-11.35)(-11.98)(-12.61)(-12.91)13PMS5_dummy-0.2791*-0.2791*-0.2675*-0.27549*-0.2825*(-14.41)(-12.33)(-12.80)(-13.96)(-14.71)14Constant4.6677* 4.616*4.6675*4.6589*4.6520*(119.03)(106.30)(105.25)(120.84)(129.58)15R20.740.700.720.740.7516F63.0450.4760.1364.9471.21Notes: Figures in parentheses are t-ratios.* Indicates significance levels at minimum 10 per cent.

1li