who suffers from indoor air pollution? evidence from...

Who suffers from indoor air pollution? Evidence from

Bangladesh

SUSMITA DASGUPTA, MAINUL HUQ, M KHALIQUZZAMAN, KIRAN PANDEY ANDDAVID WHEELERDevelopment Research Group, World Bank, Washington DC, USA

In this paper, we investigate individuals’ exposure to indoor air pollution. Using new survey datafrom Bangladesh, average hours spent by members of households in the cooking area, living areaand outdoors in a typical day are combined with the estimates of pollution concentration in differentlocations in order to estimate exposure. We analyse exposure at two levels: differences withinhouseholds attributable to family roles, and differences across households attributable to income andeducation. Within households, we relate individuals’ exposure to pollution in different locations duringtheir daily round of activities. We find high levels of exposure for children and adolescents of bothsexes, with particularly serious exposure for children under 5 years. Among prime-age adults, wefind that men have half the exposure of women (whose exposure is similar to that of children andadolescents). We also find that elderly men have significantly lower exposure than elderly women.Across households, we draw on results from a previous paper, which relate pollution variation acrosshouseholds to choices of cooking fuel, cooking locations, construction materials and ventilationpractices. We find that these choices are significantly affected by family income and adult educationlevels (particularly for women). Overall, we find that the poorest, least-educated households havetwice the pollution levels of relatively high-income households with highly educated adults.

Our findings further suggest that young children and poorly educated women in poor householdsface pollution exposures that are four times those for men in higher income households organizedby more highly educated women. Since infants and young children suffer the worst mortality andmorbidity from indoor air pollution, in this paper we consider measures for reducing their exposure.Our recommendations for reducing the exposure of infants and young children are based ona few simple, robust findings. Hourly pollution levels in cooking and living areas are quite similarbecause cooking smoke diffuses rapidly and nearly completely into living areas. However, outdoorpollution is far lower. At present, young children are only outside for an average of 3 hours per day.For children in a typical household, pollution exposure can be halved by adopting two simplemeasures: increasing their outdoor time from 3 to 5 or 6 hours per day, and concentrating outdoor timeduring peak cooking periods.

Key words: indoor air pollution, human exposure, household, Bangladesh

Introduction

Indoor air pollution (IAP) from burning wood, animaldung and other biofuels is a major cause of acuterespiratory infections, which constitute the most impor-tant cause of death for young children in developingcountries (Murray and Lopez 1996; Smith et al. 2004).Acute lower respiratory infection (ALRI), the mostserious type of acute respiratory infections, is oftenassociated with pneumonia (Kirkwood et al. 1995).ALRI accounts for 59% of all premature deaths inyoung children, and nearly all of these deaths occur indeveloping countries, with the heaviest losses in Asia andAfrica (Smith et al. 2004). Through its effect onrespiratory infections, IAP is estimated to cause between

1.6 and 2 million deaths per year in developing countries(Smith 2000; Smith et al. 2004). Most of the dead are inpoor households and approximately 1 million are children(Smith 1993; Smith et al. 1993; Smith and Mehta 2000).Table 1 provides estimates of health damage from IAPby region.1

The size of IAP’s estimated impact has prompted theWorld Bank (2001) and other international developmentinstitutions to identify the reduction of IAP as a criticalobjective for the coming decade. The current scientificconsensus is that most respiratory health damage comesfrom inhalation of respirable particles whose diameter isless than 10 microns (PM10), and recent attention hasfocused particularly on fine particles (PM2.5).

� The Author 2006. Published by Oxford University Press in association with The London School of Hygiene and Tropical Medicine. All rights reserved.

doi:10.1093/heapol/czl027 Advance Access publication 9 October 2006

In a previous paper, we analysed variations in averageIAP levels across Bangladeshi households (Dasgupta et al.2004). We found that common variations in fuel use,cooking locations, construction materials and ventilationcharacteristics lead to large differences in IAP. Non-fuelcharacteristics are so influential that some householdsusing ‘dirty’ biomass fuels have PM10 concentrationscomparable with those in households using clean fuelssuch as liquid natural gas. Under adverse conditions,on the other hand, Bangladeshi households using dirtyfuels can experience 24-hour average PM10 concentrationsas high as 800 ug/m3.

Such concentrations are far higher than outdoor PM10

levels considered dangerous for public health in industrialsocieties (Galassi et al. 2000). In those societies, however,use of clean fuels is so pervasive that attention focuses onoutdoor pollution. In biofuel-using Bangladeshi house-holds, particularly in rural areas, the calculus is oftenreversed: IAP may be much worse than outdoor pollution,and health risks may be severe for household memberswho are exposed to IAP for long periods during the day.

In this paper, we use our survey data to estimate IAPexposure for family members by age/sex group, witha particular focus on young children. We investigate thetwo major sources of differential exposure: individuals’time spent in different locations (cooking areas, livingareas and outside), and hourly fluctuations in pollutionfrom cooking. We also assess the effect of parents’ incomeand education on average household pollution levels.

We first present the methods of our analysis, describe thesampling strategy, and how indoor air and exposure toIAP have been monitored for this exercise. The followingfour sections discuss the sources of variation in individ-uals’ exposure to pollution within households. Thesesections analyse individuals’ daily location patterns, theirinteraction with daily cycles in pollution from cooking,the implications for pollution exposure, and some possibleremedies for the most vulnerable family members (parti-cularly young children). We then compare our results withthose from a recent study in India. In the penultimatesection, we assess the effects of income and adulteducation on both determinants of exposure: averagepollution levels across households, and patterns of activity

within households. Finally, we provide a summary andconclusions.

Methods

Selection of study households where indoor air was

monitored

Previous studies have identified several potential determi-nants of exposure to IAP: fuel type, time spent in cooking,cooking location, structural characteristics of housesand household ventilation practices (opening of windowsand doors etc.) (Brauer and Saxena 2002; Freemanand Sanez de Tajeda 2002; Moschandreas et al. 2002;World Bank 2002). All of these factors may be importantin Bangladeshi households, which exhibit significantdiversity in cooking fuels, stove types, cooking locationsand ventilation characteristics of houses. Discussion withlocal experts revealed widespread use of gas, electricity,kerosene, firewood, cow dung, rice husks, straw, jutesticks, bagasse and sawdust as fuel; four cooking locations(separate attached, separate detached, outside/open, singleroom dwelling – no separate kitchen); and thatch, tin,mud and brick as common structural materials of houses.

For this research, we used stratified sampling in urban andperi-urban areas of Narshingdi (Dhaka region) toincorporate representative variations in fuel use, cookingarrangements and structural characteristics that affectventilation.2 We separated the households into groupsdefined by cooking fuel, kitchen type and location, andconstruction material. Then we selected householdsindependently from each group.3 Our sample size was236, given cost constraints.

Monitoring of indoor air

At each household, PM10 concentrations in the cookingand living areas were monitored for a 24-hour periodduring December 2003 – February 2004. We used twodevices for monitoring indoor air: (1) a real-time monitor-ing instrument, the Thermo Electric Personal DataRAM(pDR-1000) (Thermo Electron 2004), and (2) a 24-hourinstrument, the Airmetrics MiniVol Portable Air Sampler(Airmetrics 2004). The pDR-1000 uses a light scatteringphotometer (nephelometer) to measure airborne particleconcentrations. The operative principle is real-time mea-surement of light scattered by aerosols, integrated overas wide a range of angles as possible. This instrumentoperated continuously for 24-hour periods, recording PM10

concentrations at 2-minute intervals. The AirmetricsMiniVol Portable Air Sampler, on the other hand, is amore conventional device that samples ambient air for24 hours. The MiniVols were generally programmed todraw air at 5 litres per minute through PM10 particle sizeseparators. The particles were caught on the filters, and thefilters were weighed pre- and post-exposure with a preciselycalibrated microbalance at Airmetrics, Inc.4 The readingsof pD-RAM and MiniVol air samplers provide a detailedrecord of IAP concentration in each household.

Table 1. Annual disease burden from indoor air pollution, 2000

Region Deaths (000s) DALYs (000s)

AFR 392 12 318AMR 27 780EMR 118 3 572EUR 21 544SEAR 559 15 227WPR 503 6 097Total 1 619 38 539

AFR¼African Region; AMR¼Americas Region; EMR¼EasternMediterranean Region; EUR¼European Region; SEAR¼ SoutheastAsian Region; WPR¼Western Pacific Region.

Indoor air pollution in Bangladesh 445

Household questionnaire administration

A short questionnaire was administered to each householdon the same day as the air monitoring to obtaininformation on socio-economic characteristics of thehousehold members, fuel type, fuel quantity, stovelocation, cooking time, number of people cooked for,duration of fire after cooking, the use of iron, mud, thatchand concrete for construction of the house and kitchen,the placement and size of windows, doors and ventilationspaces between walls and roofs, ventilation practices suchas opening doors and windows after cooking, smokingpractices, and the use of lanterns and mosquito coils.In addition, all members of the households werequestioned regarding their time activity pattern: averagehours spent in the cooking area, living areas and outdoorsin a typical day to get an indication of exposure.

Regression analysis to explore the determinants of IAP

A regression analysis for these 236 households wasconducted to explore the determinants of IAP.Households’ PM10 concentrations were regressed onfuels used during the monitored day, cooking time,duration of fire after cooking, number of people cookedfor, stove location, the use of iron, mud, thatch andconcrete for construction of the house and kitchen, theplacement and size of windows, doors and ventilationspaces between walls and roofs, ventilation practicessuch as opening doors and windows after cooking,smoking practices, and the use of lanterns and mosquitocoils. Among these variables, a small set were foundto significantly affect household PM10 concentrations:fuel type, stove locations, building materials, and openingdoors and windows after cooking.

Extrapolation and exposure reconstruction

Since we have analysed the determinants of IAP usinga stratified sample of urban and peri-urban householdsin the Dhaka region, the sample was not intended torepresent all Bangladeshi households. In order to assessthe broader implications, a representative householdsurvey was conducted in six districts in six geographicalregions: Rangpur in the Northwest, Sylhet in theNortheast, Rajshahi and Jessore in the West, Faridpurin the Centre, and Cox’s Bazar in the Southeast. Table 2documents total population, urban population and totalnumber of villages in each district.

In each district, we have attempted to randomly survey25 peri-urban households, 25 urban households5 and50 rural households.6 The same household questionnairewas administered to each household to obtain informationon socio-economic characteristics of the householdmembers, fuel, cooking time, structural characteristics,ventilation practices and time activity pattern of eachhousehold member. The regression results for Dhakaregion were then extrapolated to estimate cooking andliving area PM10 levels for this random sample of 600households. Outdoor 24-hour PM10 concentrations

were monitored for a number of peri-urban and ruralmonitoring points. Respondent estimates of exposureduration (in cooking areas, living areas and outside) fromthe household surveys were then combined with therespective estimates of pollutant concentration to estimateexposure to PM10. We estimated IAP exposure for familymembers by age/sex group, with a particular focus onyoung children.

Analysis of differential exposure

Finally, we investigated the two major sources ofdifferential exposure: individuals’ time spent in differentlocations (cooking areas, living areas and outside), andhourly fluctuations in pollution from cooking. Exposurewas analysed at two levels: differences within householdsand differences across households. Within households,we examined individuals’ exposure to pollution indifferent locations during their daily round of activity.Across households, we assessed the effects of income andeducation on average household pollution levels withregression analysis.

Daily location patterns in Bangladeshihouseholds

Within households, individuals’ pollution exposuremay vary significantly because they spend very differentamounts of time in cooking areas, living areas and outsidethe house. Table 3 reports average daily hours in the threelocations for a representative sample of 4612 individualsdrawn from surveyed households in rural, peri-urbanand urban areas of seven Bangladeshi regions (Figure 1):Rangpur (491 individuals) in the Northwest, Sylhet (578)in the Northeast, Rajshahi (491) and Jessore (490) in theWest, Faridpur (497) and Dhaka (1493) in the Centre, andCox’s Bazar (572) in the Southeast.

Table 3 presents statistics by sex, because gender rolesare quite different in Bangladeshi households. Amongage groups, we distinguish infants (age 0–1) and youngchildren (age 1–5) because they are most vulnerable toair pollution, and most tied to their mothers’ patterns ofactivity. We divide school-age youths into two groupswith different patterns of school attendance that may

Table 2. Population and number of villages: six districts ofBangladesh

District Total population Population ofthe municipality

No. of villages

Rangpur 2 527 060 251 699 1521Sylhet 2 547 320 682 000 3320Rajshahi 2 274 340 383 000 1860Jessore 2 469 680 178 273 1434Faridpur 1 742 720 99 634 1945Cox’s Bazar 1 759 560 60 234 983

Source: Bangladesh Bureau of Statistics (2002).

446 Susmita Dasgupta et al.

have implications for exposure to indoor pollution:students aged 6–8 years, who attend school in themorning (9 a.m.–12 a.m.), and students aged 9–19 years,who attend from midday until late afternoon (11.30 a.m.–4.30 p.m.). We divide adults into prime-age (20–60) andolder (60þ) categories.

Table 3 shows that time-location patterns are very similarfor infants of both sexes. They spend relatively shortperiods in cooking areas (1 hour per day), very longperiods in living areas (20 hours/day), and the residualtime (3 hours) outside the house. Infants spend moretime indoors than any other age group. Children from 1–5exhibit a very similar pattern, with little differencebetween the sexes. After age 5, however, gender differ-ences emerge strongly. In cooking areas, the gap betweenfemale and male hours rises steadily through maturity(to 3.6 extra hours for women aged 20–60), and then fallssubstantially for older women. Living areas exhibit asimilar pattern, with the gap between women and menincreasing steadily from early adolescence through oldage. Men spend much more time outside of the house thanwomen: the gap is 3.4 hours for adolescents, 6.5 hoursfor adults aged 20–60, and 4.5 hours for older adults.

Daily pollution cycles

To assess the implications of Table 3 for pollutionexposure, we need information on the levels and dailyvariations of PM10 pollution in cooking areas, living areasand outside the house. Results from Dasgupta et al.(2004), cited in Appendix 3, documented extensive indoorair-quality monitoring in the peri-urban area ofNarshingdi (Dhaka region).7 In 236 Narshingdi house-holds, the average 24-hour PM10 concentration was260 ug/m3 (s.d.¼ 156) for cooking areas and 210 ug/m3

(s.d.¼ 115) for living areas.8 To estimate outdoorconcentrations in peri-urban and rural areas, we took24-hour ambient readings in Narshingdi (8 monitoringpoints) and in rural areas of Jessore (16 points) andRangpur (11 points).9 Average outdoor concentrations inthe three locations were 36, 48 and 62 ug/m3, respectively(s.d.¼ 11, 18 and 11, respectively). The overall averagefor 35 monitoring points was 50 ug/m3 (s.d.¼ 17),

which we have adopted as our estimate of outdoorpollution for this exercise.

In Dasgupta et al. (2004), we have shown that pollutiongenerated in cooking areas diffuses almost immediatelyinto living areas (for illustrations of the close relationship,see Appendix 2). As a result, pollution in both areasexhibits strong pollution cycles in response to fuelcombustion for cooking. Drawing on information fromcontinuous, 24-hour monitoring of PM10 in 27 householdsin Narshingdi with the pDR-1000,10 Figure 2 displaysa typical daily pollution cycle as a 24-hour plot of theratio of the hourly mean PM10 concentration to the dailymean concentration.11 Its distinguishing features includetwo peaks during morning and evening cooking times,when pollution rises to over 3 times the daily average, andextensive periods in the afternoon and evening whenpollution is substantially lower than the daily average.Daily indoor pollution cycles are also reflected in outdoorambient cycles, as many houses emit cooking smoke.Figure 3 illustrates a typical ambient cycle, drawnfrom 24-hour monitoring at 7 points in rural villages ofJessore and Rangpur.12 We combine mean indoor andoutdoor pollution concentrations with the hourly ratiosin Figures 2 and 3 to produce Table 4, which provideshourly estimates of PM10 in cooking areas, living areasand outside the house.

To interpret these results, it is useful to note that India’s24-hour standard13 for rural exposure to PM10 is100 ug/m3. This 24-hour standard is exceeded in theindoor cooking area by 2.5 times, and in living areasby 2 times, in Bangladesh. It is also noteworthy thatduring peak cooking periods, 1-hour PM10 concentrationsrise to 845 ug/m3 in the cooking area and 683 ug/m3 in theliving areas.

Daily exposure for different householdmembers

We combine the information in Table 3 and Table 4to produce estimates of daily average PM10

exposure concentration by age/sex group. To incorporateour hourly estimates for PM10 concentrations in cooking,

Table 3. Meana daily hours by location: seven regions in Bangladesh (standard deviation in brackets)

Age (years) Female Male Female – Male

Cooking area Living areas Outside Cooking area Living areas Outside Cooking area Living areas Outside

0–1 1.2 (1.1) 20.1 (3.3) 2.9 (3.1) 1.0 (0.9) 19.3 (3.0) 3.7 (2.8) 0.2 0.8 �0.82–5 1.1 (1.1) 18.3 (2.8) 4.7 (3.0) 1.0 (0.9) 18.1 (3.3) 5.0 (3.4) 0.1 0.2 �0.36–8 1.0 (1.1) 16.4 (2.1) 6.6 (2.1) 0.5 (0.7) 16.4 (2.3) 7.2 (2.4) 0.5** 0 �0.6*9–12 1.3 (1.2) 15.6 (2.4) 7.2 (2.6) 0.3 (0.7) 15.6 (2.2) 8.1 (2.3) 1** 0 �0.9**13–19 2.4 (1.6) 15.7 (2.6) 6.0 (2.9) 0.3 (0.7) 14.3 (2.5) 9.4 (2.5) 2.1** 1.4** �3.4**20–60 3.8 (1.6) 16.1 (2.9) 4.3 (3.2) 0.2 (0.6) 13.1 (2.5) 10.8 (2.6) 3.6** 3** �6.5**60þ 1.5 (1.6) 19.8 (2.90) 2.9 (2.5) 0.2 (0.4) 16.6 (3.6) 7.4 (3.6) 1.3** 3.2** �4.5**

aSimple average.*Significant at 5% level of significance; **Significant at 1% level of significance.


living and outside areas, we adopt a set of conventionsfor assigning family members to locations during the24-hour cycle. We provide a complete accountingof assignments for all age/sex groups in Appendix 1.We assume that infants and children aged 0–5 spend their

single hour in the cooking area during the morningpeak cooking time. We assume that women’s cooking-area time is during peak cooking periods, and thatyoung children have their outside time during themid-afternoon.

Figure 1. Survey areas in Bangladesh


For older children, part of outside time reflects schoolingschedules; we assume that the balance is devoted to play inthe mid to late afternoon. Men’s outside work times liein the interval 7 a.m.–6 p.m., with total hours reflecting thetotals in Table 3. After accounting for cooking-area andoutside times, all family members are assigned to insideliving areas for periods that match the totals in Table 3.We compute daily exposure for members of each age/sexgroup by adding the 24 hourly PM10 concentrationsfor their assigned locations and calculating the meanconcentration. Table 5 presents the results.

To check the general validity of our estimates for a typicalhousehold,14 we replicate our approach 236 times,using the mean PM10 concentration for each householdmonitored by our study. These concentrations varywidely, for reasons explored in Dasgupta et al. (2004).We average the results and present them alongside thetypical household results in Table 5.

Differences in the two sets of results stem from variationsin average pollution levels, household age/sex composi-tions and time allocations by household members in thefull set of monitored households. We present the results

for the typical household in order to display exposurevariations when inside/outside pollution concentrationsand individuals’ time allocations are held constant.In any case, the two sets of results are quite similar.The most striking finding is the high exposure – around200 – for infants and children, regardless of gender.Exposures for student-age individuals (6–19) are some-what lower (although still quite high), and also similarfor both sexes. The real gender-based divergence occursamong adults, with women’s exposures nearly twice thosefor men in the age group 20–60, and about 40% higher forolder women (over 60).

Table 5 indicates that only adult males aged 20–60 havedaily PM10 exposures low enough to approach the Indianstandard (100 ug/m3). All other household membershave significantly higher exposure levels, and the youngestchildren of both sexes have exposures that are amongthe most dangerous. Mortality from respiratory diseaseamong children in this age range attests to the potentimpact of such pollution levels.

Two features of our results warrant particular scrutiny.First, most studies have traditionally focused on

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

12:00 AM 4:00 AM 8:00 AM 12:00 PM 4:00 PM 8:00 PM 12:00 AM

Time

PM

10 r

elat

ive

to m

ean

dai

ly P

M10

Figure 2. Daily PM10 pattern for cooking and living areas

0.0

0.5

1.0

1.5

2.0

2.5

12:00 AM 4:00 AM 8:00 AM 12:00 PM 4:00 PM 8:00 PM 12:00 AM

Time

PM

10 r

elat

ive

to m

ain

dai

ly P

M10

Figure 3. Daily ambient (outdoor) pollution pattern in rural villages


pollution in cooking areas. However, our monitoring ofBangladeshi households suggest that simultaneous pollu-tion in living areas is the root cause. This is particularlytrue for young children, who spend only 1 hour per day incooking areas, on average. Living-area pollution is onlymoderately below cooking-area pollution and followsthe same cycle, so most daily inhalation of particulatesoccurs in the living areas. Adult males have lowerexposures simply because they are out of the house formany more hours per day.

Secondly, we should qualify our results with a cautionarynote about the impact of very intense pollution on women

and children during peak cooking periods. It is possiblethat peak pollution during a few hours per day causesdisproportionate health damage. Scientific evidencecurrently available suggests that health damage isassociated with daily average exposure levels, not peakhourly exposures. However, the evidence is far fromconclusive, and it is mostly derived from research onoutdoor pollution effects in industrial economies.15 Giventhe large peaks of PM10 concentration during cookingin many Bangladeshi households, further research on thisissue seems justified.

Reducing exposure for young children

We use the case of infants (age 0–1) in our typicalhousehold to illustrate some implications of our results.We focus on infants because their documented vulner-ability to indoor pollution seems to be the greatest.Our example applies to both male and female infants,since Table 5 shows insignificant gender difference fordaily exposure concentrations (216 ug/m3 for females vs.214 for males).

Figure 4 presents a simple experiment with data for thetypical household. Starting with the status quo, withinfants spending 3 hours outside in mid-afternoon, weoptimize the 3-hour outside period by switching to othertimes (e.g. 8 a.m.) that provide the greatest relief fromindoor pollution. Then we add 3 more hours outside,sequentially choosing times that yield the greatest incre-mental reductions in daily PM10 exposure. We plot theresults in Figure 4.

Figure 4 indicates that keeping infants outside duringan optimally chosen 6-hour period during peak cookingtimes (7–10 a.m.; 5–7 p.m.) would reduce daily PM10

exposure to the Indian standard level (100 ug/m3). Sinceapproximately half of the sample households have PM10

concentrations above the mean level for our surveypopulation (260 ug/m3 in cooking areas, 210 in livingareas), the potential reduction in exposure for infants inmany homes could be much greater.16

Table 4. Daily PM10 (ug/m3) concentrationsa, by location in atypical householdb

Time Kitchen Living area Ambient

12 a.m. 78 63 401 a.m. 78 63 352 a.m. 78 63 373 a.m. 78 63 394 a.m. 78 63 415 a.m. 142 114 426 a.m. 257 207 477 a.m. 466 376 578 a.m. 845 683 449 a.m. 568 459 3310 a.m. 382 308 2811 a.m. 257 207 3312 p.m. 173 139 391 p.m. 116 94 372 p.m. 78 63 343 p.m. 78 63 404 p.m. 78 63 375 p.m. 257 207 366 p.m. 845 683 667 p.m. 525 424 1018 p.m. 326 263 889 p.m. 202 163 7210 p.m. 126 101 5011 p.m. 78 63 46

aConcentrations are averages over the hour, for example, 11.30–12.30 average has been reported for 12.00.bDrawing on mean values from continuous, 24-hour monitoring of27 households in Narshingdi.

Table 5. Daily average PM10 (ug/m3) exposure by age and gendera

Age Typical householdb 236 monitored householdsc

Female Male Female Male

0–1 216 214 209 1951–5 212 212 199 1926–8 173 172 156 1639–19 207 174 196 19420–60 227 116 221 11860þ 220 161 264 188

aOutdoor PM10¼ 50.bPM10 concentrations: cooking area 260; living area 210.cAverages for 236 separate calculations using monitored PM10 levelsof respective households.

50

75

100

125

150

175

200

225

0 1 2 3 4 5 6

Hours outside

Dai

ly e

xpo

sure

(u

g/m

3)

Figure 4. Optimum outside hours and daily PM10 exposure forinfants


Our results suggest that for households whose youngchildren are kept inside during peak cooking periods,simply moving the children outside when weather permitscould yield significant reduction in exposure. Householdmembers assigned to outside supervision would alsobenefit from reduced pollution. In cases where familyhelp is scarce, it might be possible for several householdsto pool supervision during peak periods. While this mightcreate some inconvenience, families might well considerthis option if they recognized the potential health benefitsfor their children. By the same logic, of course, otherfamily members would benefit from extending theiroutside time during peak cooking periods.

Comparison with results for India

A recent study of IAP exposure in India (World Bank2002; Balakrishnan et al. 2004a) provides a useful point ofcomparison for our results. The India study also considersdifferent age/sex groups, cooking vs. living areas, andcooking vs. non-cooking times. Its conclusions are similarto our findings in some respects:

Whereas women, in their traditional capacity as cooks,suffer from much greater average daily exposuresthan other family members, adult men experience theleast exposure. Among non-cooks, those who are mostvulnerable to the health risks of IAP — young childrenand elderly people — tend to experience higher levelsof exposure because they spend more time indoors.

(World Bank 2002, p. 3)

As Table 5 shows, we also find large differences betweenwomen and men in exposure to air pollution.17 However,we find essentially no difference in exposure for womenand young children of both sexes. We also find relativelysmall differences between women’s exposures and expo-sures for adolescents of either sex. The essential differencebetween our results and the India results lies in ourfindings for pollution in living areas. We find averageliving-area pollution concentrations to be much closer tocooking-area concentrations, and our 24-hour monitoringdata indicate that daily pollution cycles are close toidentical for the two areas.18 As a result, time spent in

living areas does not provide much relief from pollutionexposure.19 Time spent outside the house thereforeemerges as the key variable in our analysis, and adultmales have much smaller pollution exposure simplybecause of their outside orientation.

Income, education and exposure

Households’ PM10 concentrations and individuals’ loca-tion patterns combine to produce pollution exposures forfamily members. As we have shown in Dasgupta et al.(2004), household pollution levels are highly sensitive tofuel choices, cooking locations and construction materialsthat affect ventilation. Each of these factors may, in turn,be affected by household income and adult educationlevels. If such effects are significant, they may be related inpart to pollution-related problems, and in part to otherfactors (e.g. status issues associated with choice ofbuilding materials and cooking locations). The positiveincome elasticity of clean fuel choice should also havea significant effect, at least in urban areas where cleanfuel is a feasible option. Education may affect adults’awareness of the relationship between pollution exposuresand health risks for infants and young children. If this isthe case, then we might expect children to spend moretime outside in more highly educated families. We areagnostic about whether income could also have an effectin this context.

We test the effects of income and education on pollutionexposure factors, using Ordinary Least Square regressionsadjusting for heteroskedasticity, for our full sample ofhouseholds in seven regions of Bangladesh. We use theeconometric results presented in Dasgupta et al. (2004) toestimate PM10 in each household, incorporating thecombined effects of fuel choices and structural character-istics (cooking locations and construction materials).We regress estimated PM10 on household income(in US$ per day) and the average education levels20 ofmen and women in the household. For infants and youngchildren, we regress time spent outside on the samevariables.

We present the regression results for both exposurecomponents in Table 6.

Table 6. Education, income and pollution exposure for children

(1) (2) (3)Cooking area PM10 (ug/m

3) Living area PM10 (ug/m3) Hours outside house (Children 0–5)

Adult education level (range: 0–4)Female �20.305 (7.77)** �14.404 (7.11)** 0.130 (0.69)Male �7.556 (3.36)** �6.546 (3.76)** �0.612 (3.85)**

Income per capita (US$ per day) �9.753 (4.82)** �4.752 (3.03)** 0.033 (0.18)Constant 298.616 (87.64)** 252.849 (95.78)** 5.145 (21.73)**Observations 4174 4174 483R-squared 0.07 0.06 0.04

Absolute value of t statistics in parentheses.*Significant at 5%; **Significant at 1%.


Education and household per capita income have theexpected effects on determinants of indoor concentrationof PM10. All estimated effects are large, highly significantand have the expected negative sign. Women’s educationhas a particularly large effect. Our results indicate thatwhen men’s and women’s education levels jointly increasefrom 0 (no primary schooling) to 4 (post-secondaryeducation), predicted PM10 in the cooking area decreasesby about 110 ug/m3. This is a very large effect, since theaverage PM10 concentration for our 236 monitoredhouseholds is 260 ug/m3. Each increase of US$1.00/dayis associated with a decline of 10 ug/m3, so the predictedreduction over the sample income range (less than $0.50/day to $15.00/day) is approximately 150 ug/m3.

Although education and income strongly reduce averagepollution, they do not seem to change children’s dailyactivity patterns in ways that reduce pollution exposure.Children’s hours spent outdoors are not significantlyaffected by women’s education or income per capita, andthe measured effect of adult male education is actuallyperverse (children of more educated men spend less timeoutdoors).

Hence, we tentatively conclude that parents’ educationand income do affect children’s pollution exposure, butonly through the determinants of pollution. We extend theanalysis of income and education further by estimatingseparate Ordinary Least Square regressions for pollution

determined by fuel choice and structural determinantsof pollution (cooking locations, construction materials).We combine information on fuel choices and structuralcharacteristics into two separate pollution indices,21 withweights determined by the regression coefficients forcooking-area pollution determinants in Table 7 ofDasgupta et al. (2004). Table 7 presents the results,which suggest that female education, male education andfamily income all have large, highly significant effectson pollution via fuel choice. Female education hasan equivalent effect on the structural determinants, butmale education and family income do not appear to besignificant. As in the composite result (Table 6), femaleeducation appears to be the strongest and most pervasivedeterminant of arrangements that reduce IAP.

We illustrate our overall results in Table 8, which presentspredicted average PM10 levels in cooking and living areasby income and women’s education.

Table 8 is generally consistent with our econometricresults, while suggesting that some non-linear effects arenot captured by the linear regression model in Table 6.For example, education seems to have minimal effects onpollution for the lowest income group, but pronouncedeffects when income is higher. Similarly, increased incomedoes not reduce pollution consistently for the lowesteducation group, but it has a strong effect at higherlevels of education. The combined effects of income andwomen’s education are sufficient to approximatelyhalve PM10 pollution, from near 300 ug/m3 in the poorest,least-educated groups to around 150 in the highest-income, best-educated groups. As our sample compositionstatistics clearly show in Table 9, most of the individuals(and households) in our survey are in the four table cellsassociated with the lowest income and education levels.

Summary and conclusions

In this paper, we have investigated individuals’ exposureto IAP in Bangladesh. We have analysed exposure attwo levels: differences within households attributable tofamily roles, and differences across households attrib-utable to income and education. Within households, wehave related individuals’ exposure to pollution in differentlocations during their daily round of activity. We find

Table 7. Income, education and determinants of indoor airpollution

Fuel choice Construction materialsof cooking location

Adult education level(range: 0–4)Female �8.687 (12.00)** �6.926 (3.85)**Male �7.555 (12.14)** �2.151 (1.39)

Income per capita(US$ per day)

�12.760 (22.79)** 1.547 (1.11)

Constant 15.511 (16.44)** 22.031 (9.40)**Observations 4174 4174R-squared 0.36 0.01

Absolute value of t statistics in parentheses.**Significant at 1%.

Table 8. Average predicted PM10 concentrations (ug/m3) by income and women’s education level

Income per capita (US$/day) Women’s education level (range: 0–4)

Cooking area Living areas

0–1.00 1.01–2.00 2.01–3.00 3.01þ Total 0–1.00 1.01–2.00 2.01–3.00 3.01þ Total

0–0.50 267 305 205 261 272 229 257 183 234 2330.51–1.00 314 222 220 206 260 269 193 196 185 2251.01–2.00 267 205 168 162 211 230 187 163 158 1912.01–5.00 266 212 151 148 188 236 198 152 148 1795.01þ 373 135 150 135 163 308 147 157 147 166Total 278 246 192 181 254 239 213 178 173 221


high levels of exposure for children and adolescents ofboth sexes, with particularly serious exposure for childrenunder 5. Among prime-age adults, we find that menhave half the exposure of women (whose exposure issimilar to that of children and adolescents). We alsofind that elderly men have significantly lower exposurethan elderly women.

Across households, we draw on results from a previouspaper (Dasgupta et al. 2004), which relates pollutionvariation across households to choices of cooking fuel,cooking locations, construction materials and ventilationpractices. We find that these choices are significantlyaffected by family income and adult education levels(particularly for women). Overall, we find that thepoorest, least-educated households have twice the pollu-tion levels of relatively high-income households withhighly educated adults.

To summarize, we find that young children and poorlyeducated women in poor households face pollutionexposures that are four times those of men in higher-income households organized by more highly educatedwomen. In Dasgupta et al. (2004), we recommendedfeasible changes in cooking locations, constructionmaterials and ventilation practices that could greatlyreduce average household pollution levels. In this paper,we consider measures for narrowing the exposure gapwithin households. We focus particularly on changes forinfants and young children, since they suffer the worstmortality and morbidity from IAP, but our findings alsoapply to women and adolescents. Our recommendationsfor reducing their exposure are based on a few simple,robust findings: hourly pollution levels in cooking andliving areas are quite similar because cooking smokediffuses rapidly and nearly completely into living areas.At the same time, outdoor pollution is far lower.At present, young children are only outside for an averageof 3 hours per day. For children in a typical household,pollution exposure can be halved by adopting two simplemeasures: increasing their outdoor time from 3 to 5 or6 hours per day, and concentrating outdoor time duringpeak cooking periods. We recognize that weather andother factors may intervene occasionally, and that childsupervision outdoors may be difficult for some house-holds. However, the potential benefits are so great that

neighbours might well agree to pool outdoor supervisiononce they are aware of the implications for their children’shealth.

Endnotes

1 The ‘Regions’ are defined as in the World HealthOrganization’s Global and Regional Burden of Disease Report 2004,online at [http://www.who.int/publications/cra/en], and do notcorrespond to continent/region boundaries.

2 Although we use the term ‘peri-urban’ to describe areasproximate to Dhaka, our sample includes many rural farmhouseholds.

3 Our stratification had been designed for cell values largeenough to test fuel and ventilation effects on IAP, and was notintended to represent all Bangladeshi households.

4 While the MiniVol is not a reference method sampler, itproduces results that closely approximate data from US FederalReference Method samplers.

5 In each district, uniformly rich/poor neighbourhoods inurban and peri-urban areas were left out after discussion with thelocal experts. One neighbourhood was selected randomly from theheterogeneous neighbourhoods. Interviewers then approached eachconsecutive house, starting from one end, until 25 surveys werecompleted.

6 In each district, on two consecutive days, the survey teamtravelled exactly the same number of kilometres away from the maintown boundary in an East, West/North or South direction (directionwas decided randomly). From the first village then encountered,starting from one end, each consecutive house was approached until25 surveys were completed.

7 Although this region is near Dhaka, many of the sampledhouseholds are in rural settings.

8 PM10 concentration was monitored in 236 cooking areasand 244 living areas continuously, without intervention, for 24-hours, with MiniVol.

9 Ambient readings were undertaken, with MiniVol, for 4days per location.

10 The pDR-1000 monitors need calibration, being secondaryinstruments. Calibration is usually done by colocated experiments.In our case, we conducted colocated experiments with MiniVols.The calibration factors used were as follows:

(1) PDRAM Unit – 05217: average (165 ug/m3) andMinivol reading (259 ug/m3), Calibration factor¼ 1.57

(2) PDRAM Unit – 5190: average (170 ug/m3) and Minivolreading (259 ug/m3), Calibration factor¼ 1.52

Zeroing of the instruments was done by putting the pDR-1000s in the zero air pouch and adjusting the reading to zero.

11 Of course, households differ significantly in the timing ofdaily peaks; some have three rather than two, and there are alsosignificant differences in peak levels and change gradients. Figure 2represents the central tendency in the observed patterns for 27households.

12 This cycle is clearly different from the cycle in Figure 2, fortwo apparent reasons. First, the outdoor cycle reflects the combinedeffect of fuel burning in many households, whose daily cycles reachpeaks at different times, and at different PM10 emissions intensities.However, this does not explain the large difference in the relative sizeof the two outdoor peaks, as compared with rough parity for theindoor peaks. Since the outdoor pattern is reproduced across severalwidely separated monitoring points, we hypothesize that differentatmospheric conditions lead to more pronounced accumulation andduration of suspended particulates in the evening. Alternatively,preparation of evening meals may cluster more tightly in time,

Table 9. Number of adult females by income and education level:sample composition

Income per capita(US$/day)

Adult female education level

0–1.00 1.01–2.00 2.01–3.00 3.01þ Total

0–0.50 1 703 455 81 52 2 2910.51–1.00 585 553 172 78 1 3881.01–2.00 184 251 121 76 6322.01–5.00 32 59 41 61 1935.01þ 11 18 25 54 108Total 2 515 1 336 440 321 4 612


http://www.who.int/publications/cra/en

leading to more concentrated loading of PM10 in the evening. Futureresearch may shed more light on this phenomenon.

13 Source: Central Pollution Control Board, Delhi, undated.National Ambient Air Quality Standards, online at [http://www.cpcb.nic.in/as.htm].

14 With mean values of cooking and living area concentra-tions from continuous 24-hour monitoring of 27 households inNarshingdi.

15 Studies of short exposures to outdoor particulateconcentrations suggest some impact on heart rate variability andthe rate of heart attacks. However, a recent study in Palm Springs,California suggests that the short-period effect disappears when 24-hour average exposure is controlled for. Similarly, average exposuresseem to dominate day-to-day variations in daily time-series studies.Our thanks to Dr Bart Ostro, Chief, California Office ofEnvironmental Health Hazard Assessment, for his insights.

16 We recognize that our illustration achieves such largereductions by assuming that infants are outside during the leastdesirable time in the status quo situation (i.e. the mid-afternoonperiod when indoor pollution is also relatively low). However, webelieve that our essential point is generally valid – optimally chosenoutside time for infants has the potential for considerable reductionof health damage.

17 As we note in Dasgupta et al. (2004), average PM10

concentrations are smaller in our study.18 The correlation coefficient was 0.93 for median PM10

concentrations in cooking and living areas for households whereindoor air was monitored with pDR-1000.

19 High pollution indoors has been documented in ruralhouseholds in Tamil Nadu (Balakrishnan et al. 2002; Parikh et al.2003) and Andhra Pradesh, India (Balakrishnan et al. 2004b).

20 Education, in the questionnaire, was a categorical variable:no schooling, less than or equal to 5 years of schooling, 6–10 years ofschooling, 11–12 years of schooling, more than 12 years ofschooling.

21 The indices are constructed from thefollowing equations: Fuel Choice¼ 258.563 – 45.233 * Jute – 106.729 *Kerosene� 112.523 *Lpg/Lng� 155.285 * PipedNatural Gas; Construction Materials/Cooking Location¼258.563þ 253.896 *Mud Walls – 124.058 *Mud Walls, DetachedKitchen – 9.887 *Open-Air Kitchen – 37.599 *Detached Kitchen.The fuel choice index (mean 237.3, s.d. 48.2) and constructionmaterials/cooking location index (mean 266.3, s.d. 87.4) arepredicted indoor particulate concentrations, measured in micro-grams of PM10 per cubic metre. The constant term in each equationis the average indoor concentration in sample households for allfuels and structural characteristics that are not explicitly controlledfor in the regression analysis. The fuel choice equation incorporatesthe pollution-reducing effects of clean fuels (jute, kerosene, lpg/lng,piped natural gas). The construction materials/cooking locationequation incorporates the pollution-increasing effect of mud-wallconstruction, with a downward adjustment for the mud-wall casewhen the kitchen is detached, along with pollution-reducingadjustments for open-air and detached kitchens. All explanatoryvariables in this regression are coded as 0–1 dummy variables.

References

Airmetrics. 2004. MiniVol Portable Air Sampler, description onlineat [http://www.airmetrics.com/products/minivol/index.html].

Balakrishnan K, Parikh J, Sambandam S et al. 2002. Daily averageexposure to respirable particulate matter from combustionof biomass fuels in rural households of Southern India.Environmental Health Perspectives 110: 1069–75.

Balakrishnan K, Mehta S, Kumar P et al. 2004a. Indoor air pollutionassociated with household fuel use in India: an exposureassessment and modeling exercise in rural districts of AndhraPradesh, India. Washington, DC: World Bank, Energy SectorManagement Assistance Program (ESMAP).

Balakrishnan K, Sambandam S, Ramaswamy P et al. 2004b.Exposure assessment for respirable particulates associatedwith household fuel use in rural districts of Andhra Pradesh,India. Journal of Exposure Analysis and EnvironmentalEpidemiology 14: S14–S25.

Bangladesh Bureau of Statistics. 2002. Statistical Yearbook ofBangladesh 2002. Dhaka: Bangladesh Bureau of Statistics.

Brauer M, Saxena S. 2002. Accessible tools for classification ofexposure to particles. Chemosphere 49: 1151–62.

Bruce N. 1999. Lowering exposure of children to indoor airpollution to prevent ARI: the need for information andaction. Capsule Report 3. Arlington, VA: EnvironmentalHealth Project.

Dasgupta S, Huq M, Khaliquzzaman M et al. 2004. Indoor airquality for poor families: new evidence from Bangladesh.Development Research Group Working Paper No. 3393.Washington, DC: The World Bank.

Freeman NCG, Saenz de Tejada S. 2002. Methods for collectingtime/activity pattern information related to exposure tocombustion products. Chemosphere 49: 979–92.

Galassi C, Ostro B, Forastiere F et al. 2000. Exposure to PM10 inthe eight major Italian cities and quantification of the healtheffects. Epidemiology 11: S126.

Kirkwood B, Gove S, Rogers S et al. 1995. Potential interventionsfor the prevention of childhood pneumonia in developingcountries: a systematic review. Bulletin of the World HealthOrganization 73: 793–8.

Moschandreas DJ, Watson J, D’Aberton P et al. 2002. Methodologyof exposure modeling. Chemosphere 49: 923–46.

Murray C, Lopez A (eds). 1996. The global burden of disease.Cambridge, MA: Harvard University Press.

Parikh J, Balakrishnan K, Laxmi V et al. 2003. Exposure to airpollutants from combustion of cooking fuels: a case study ofrural Tamil Nadu, India. Accessed online at: [http://www.repp.org/discussiongroups/resources/stoves/Countries/exposure.doc].

Smith K. 1993. Fuel combustion, air pollution exposure, and health:the situation in developing countries. Annual Review ofEnvironment and Energy 18: 526–66.

Smith K. 2000. National burden of disease in India from indoor airpollution. Proceedings of the National Academy of Sciences USA97: 13286–93.

Smith K, Metha S. 2000. The global burden of disease from indoorair pollution in developing countries: comparison of estimates.Prepared for the WHO/USAID Global Technical Consultationon Health Impacts of Indoor Air Pollution in DevelopingCountries.

Smith K, Liu Y, Rivera J et al. 1993. Indoor air qualityand child exposures in highland Guatemala. Proceedings ofthe 6th International Conference on Indoor Air Qualityand Climate, University of Technology, Helsinki, volume 1:441–6.

Smith K, Mehta S, Maeusezahl-Feuz M. 2004. Indoor air pollutionfrom household use of solid fuels. In: Ezzati M, Rodgers AD,Lopez AD et al. (eds). Comparative quantification ofhealth risks: the global and regional burden of disease due toselected major risk factors volume 2. Geneva: World HealthOrganization, pp. 1435–93.

Thermo Electron. 2004. Personal DataRAM, pDR-1000, descrip-tion online at [http://www.thermo.com/eThermo/CMA/PDFs/Product/productPDF_18492.pdf].


http://www.cpcb.nic.in/as.htm

http://www.airmetrics.com/products/minivol/index.html

http://www.repp

http://www.thermo.com/eThermo/CMA/PDFs/

World Bank. 2001.Making sustainable commitments: an environmentstrategy for the World Bank. Washington, DC: World Bank.

World Bank. 2002. India: household energy, indoor air pollution, andhealth. Washington, DC: World Bank, Energy SectorManagement Assistance Program (ESMAP), South AsiaEnvironment and Social Development Unit.

Acknowledgements

Authors’ names in alphabetical order. We would like to express ourappreciation to the field survey team of the Development PolicyGroup, for their excellent air-quality monitoring work underdifficult conditions. We are also grateful to Kseniya Lvovsky,Maureen Cropper, Douglas Barnes, Bart Ostro and Paul Martinfor useful comments and suggestions. Financial support for thisstudy has been provided by Trust Funds through the Knowledgefor Change Program of the World Bank’s DevelopmentEconomics Vice-Presidency, and by the Development ResearchGroup. The findings, interpretations and conclusions are entirelythose of the authors. They do not necessarily represent the view ofthe World Bank, its Executive Directors or the countries theyrepresent.

Biographies

Susmita Dasgupta is Senior Economist in the Infrastructure/Environment Unit of the World Bank’s Development ResearchGroup. Dasgupta is a specialist in empirical research in environ-mental economics and has done extensive research on settingpriorities in pollution control, estimation of pollution abatementcost, cost-effective regulations, and poverty/environment nexus. Shereceived her undergraduate degree from Presidency College, India;her Masters degree from University of Calcutta, India; and her PhDfrom the State University of New York at Albany.

Mainul Huq has extensive research experience in developmenteconomics and environmental issues. He is currently working as anEconomist for the Development Policy Group (DPG) ofBangladesh.

M Khaliquzzaman has more than 20 years of experience in studiesrelated to environmental sciences including ambient and indoor airquality. He has been involved in the measurement and sourceapportionment of size-fractionated airborne particulate matter(PM2.5 and PM10) in Bangladesh using PIXE as an analyticaltool. Khaliquzzaman obtained his PhD from London University in1971 and he is currently a consultant with the World Bank Office,Dhaka.

Kiran D Pandey, PhD, is Senior Environmental Economistat the Global Environment Facility. His interests include:(a) indicators and system development for environmental healthand environmental policy, and (b) global environment financingand reform. He is also a member of the World Health Organization(WHO) working group on Urban Air Pollution for WHO’sGlobal Burden of Disease Comparative Quantification of HealthRisks Study.

David Wheeler is Lead Economist in the Infrastructure/Environment Unit of the World Bank’s Development ResearchGroup. He received his undergraduate degree from PrincetonUniversity (1968) and his PhD in economics from theMassachusetts Institute of Technology (1974). He has also taughton the economics faculties of Boston University, MassachusettsInstitute of Technology and the National University of Zaire(Congo).

Correspondence: Susmita Dasgupta, Development Research Group,World Bank, 1818H Street, NW, Washington DC 20433, USA.E-mail: [email protected]


Appendix 1. Hourly household member locations, by age and gender (X ¼ presence in alocation)

Sex Age Place Time (0¼ 12.00 a.m.; 23¼ 11.00 p.m.)

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Total

F 0-1 K X 1F 0-1 L X X X X X X X X X X X X X X X X X X X X 22F 0-1 O X X X 1F 1-5 K X 1F 1-5 L X X X X X X X X X X X X X X X X X X X 19F 1-5 O X X X X 4F 6-8 K X 1F 6-8 L X X X X X X X X X X X X X X X X X 17F 6-8 O X X X X X X 6F 9-18 K X X 2F 9-18 L X X X X X X X X X X X X X X X X 16F 9-18 O X X X X X X 6F 19-60 K X X X X 4F 19-60 L X X X X X X X X X X X X X X X X X X 18F 19-60 O X X 2F 60þ K X X 2F 60þ L X X X X X X X X X X X X X X X X X X X 19F 60þ O X X X 3M 0-1 K X 1M 0-1 L X X X X X X X X X X X X X X X X X X X X 20M 0-1 O X X X 3M 1-5 K X 1M 1-5 L X X X X X X X X X X X X X X X X X X X 19M 1-5 O X X X X 4M 6-8 K X 1M 6-8 L X X X X X X X X X X X X X X X X 16M 6-8 O X X X X X X X 7M 9-18 K 0M 9-18 L X X X X X X X X X X X X X X X X 16M 9-18 O X X X X X X X X 8M 19-60 K 0M 19-60 L X X X X X X X X X X X X X 13M 19-60 O X X X X X X X X X X X 11M 60þ K 0M 60þ L X X X X X X X X X X X X X X X X 16M 60þ O X X X X X X X X 8


Appendix 3. Determinants of indoor airpollution in Bangladesh

In an earlier paper, Dasgupta et al. (2004), we investigatedthe determinants of indoor air pollution in Bangladesh,using monitoring data for a stratified sample of 236households in urban and peri-urban areas of Dhaka. Thehouseholds were separated into groups defined by cookingfuel (gas, electricity, kerosene, firewood, cow dung, ricehusks, straw, jute sticks, bagasse, sawdust), kitchen typeand location (separate attached, separate detached,outside/open, single room dwelling - no separate kitchen),and construction material (thatch, tin, mud, brick).Samples were then selected independently from eachgroup. In each household, we monitored respirableairborne particulate (PM10) concentrations in the kitchenand living room during the period December 2003 –February 2004. Two devices – (1) a real-time monitoringinstrument, the Thermo Electric Personal DataRAM(pDR-1000) (Thermo Electron 2004), and (2) a 24-hourinstrument, the Airmetrics MiniVol Portable Air Sampler(Airmetrics 2004) – were used for monitoring indoor air.We monitored most houses for 1 day, and a few for 2 days.

The results were then extrapolated to estimate indoorair pollution levels for a random sample of 600 rural, peri-urban and urban households in six regions of Bangladesh:Rangpur, Sylhet, Rajshahi, Faridpur, Jessore and Cox’sBazar. Overall, our results suggest:

� IAP is alarmingly high for many poor families inBangladesh. Twenty-four-hour average concentrationsof PM10: 300 ug/m

3 or greater are common, implyingwidespread exposure to a serious health hazard.

� Pollution from cooking diffuses into living spacesrapidly and fairly completely in many cases, so thatexposure is similar for all household members who areindoors during the same periods.

� Differences in ventilation factors produce great varia-tions in potential exposure for all household members,even for poor families that use biomass fuels forcooking.

� After allowing for the effect of household ventilationcharacteristics, we find very significant differencesbetween biomass and ‘clean’ fuels. For example,mean PM10 in clean-fuel households is 133 ug/m3

Appendix 2. Comparative PM10 concentrations in four Bangladeshi houses: kitchens andliving rooms

34

56

7

0 5 10 15 20 2524-Hour Time

Log Kitchen Concentration Log Living Room Concentration

45

67

89

4.5

55.

56

6.5

7Lo

g Ki

tche

n C

once

ntra

tion/

Log

Livi

ng R

oom

Con

cent

rat

0 5 10 15 20 2524-Hour Time


45

67

8Lo

g Ki

tche

n C

once

ntra

tion/

Log

Livi

ng R

oom

Con

cent

ra

0 5 10 15 20 2524-Hour Tim e


7

6

5

4

3

Log

kitc

hen

conc

entra

tion/

log

livin

g ro

om c

once

ntra

tion

0 5 10 15 20 25

24-Hour Time

Log kitchen concentration Log living room concentration

(1)

Log

kitc

hen

conc

entra

tion/

log

livin

g ro

om c

once

ntra

tion(3)

9

8

7

6

5

4

Log

kitc

hen

conc

entra

tion/

log

livin

g ro

om c

once

ntra

tion(2)


0 5 10 15 20 25

24-Hour Time

7

Log

kitc

hen

conc

entra

tion/

log

livin

g ro

om c

once

ntra

tion(4)

8

7

6

5

4

6.5

6

5.5

5

4.5

0 5 10 15 20 25

24-Hour Time


0 5 10 15 20 25

24-Hour Time



(s.d.¼ 20, n¼ 46), little higher than ambient pollutionin the urban area where most clean fuels are used. Forbiomass-using households, on the other hand, theaverage concentration is 242 ug/m3 (s.d.¼ 143, n¼ 404).

� For biomass fuels, a cautionary note is introduced bylarge differences in median and mean exposuresrecorded by our 24-hour monitors. Very large peaksof concentration for short periods (e.g. 10 000 ug/m3 orhigher) under some conditions can have large effects onestimated mean concentrations. Median concentrationsare often much lower.

� Although fuel choice certainly affects indoor airpollution, our results suggest that its role is secondaryto the role of ventilation factors for Bangladeshihouseholds. Moving from indoors to an open kitchenlowers the PM10 concentration in cooking and livingareas by almost the same magnitude as switching from

firewood to kerosene. Switching from mud walls toother materials lowers PM10 far more than any otherfactor, and use of thatch roofs also has a large effect.

� Our extrapolation results further suggest greatgeographic variation, even for households in the sameper capita income group. This variation reflectslocal differences in fuel use and, more significantly,construction practices that affect ventilation. Forthe poorest households, rural PM10 concentrationsvary from 410 ug/m3 (s.d.¼ 128, n¼ 48) in Cox’sBazar to 202 ug/m3 (s.d.¼ 19, n¼ 46) in Faridpur.Even in urban areas, concentrations differ by almost100 ug/m3 between the highest areas, Jessoreand Rajshahi, and the lowest, Sylhet. The pooresthouseholds in Rangpur face the same mean indoorconcentration, 198 ug/m3 (s.d.¼ 26, n¼ 34), as thehighest-income households in Cox’s Bazar.


who suffers from indoor air pollution? evidence from...

Documents