Science of the Total Environment 409 (2011) 1774–1785
Contents lists available at ScienceDirect
Science of the Total Environment
j ourna l homepage: www.e lsev ie r.com/ locate /sc i totenv
Assessing the co-benefits of greenhouse gas reduction: Health benefits of particulatematter related inspection and maintenance programs in Bangkok, Thailand
Ying Li a,⁎,1, Douglas J. Crawford-Brown b
a Department of Public Policy, University of North Carolina at Chapel Hill, Chapel Hill, NC, USAb Cambridge Centre for Climate Change Mitigation Research, University of Cambridge, Cambridge, UK
⁎ Corresponding author at: Department of Public PolicyChapel Hill, CB# 3435, Chapel Hill, NC 27599-3435, USA.315 684 6588.
Article history:Received 28 September 2010Received in revised form 25 January 2011Accepted 26 January 2011Available online 21 February 2011
Keywords:Health benefitsAmbient particulate matterInspection and maintenance programsMotor vehiclesBangkok
Since the 1990s, the capital city of Thailand, Bangkok has been suffering from severe ambient particulatematter (PM) pollution mainly attributable to its wide use of diesel-fueled vehicles and motorcycles with pooremission performance. While the Thai government strives to reduce emissions from transportation throughenforcing policy measures, the link between specific control policies and associated health impacts isinadequately studied. This link is especially important in exploring the co-benefits of greenhouse gasemissions reductions, which often brings reduction in other pollutants such as PM. This paper quantifies thehealth benefits potentially achieved by the new PM-related I/M programs targeting all diesel vehicles andmotorcycles in the Bangkok Metropolitan Area (BMA). The benefits are estimated by using a framework thatintegrates policy scenario development, exposure assessment, exposure-response assessment and economicvaluation. The results indicate that the total health damage due to the year 2000 PM emissions from vehiclesin the BMA was equivalent to 2.4% of Thailand's GDP. Under the business-as-usual (BAU) scenario, totalvehicular PM emissions in the BMA will increase considerably over time due to the rapid growth in vehiclepopulation, even if the fleet average emission rates are projected to decrease over time as the result ofparticipation of Thailand in post-Copenhagen climate change strategies. By 2015, the total health damage isestimated to increase by 2.5 times relative to the year 2000. However, control policies targeting PM emissionsfrom automobiles, such as the PM-oriented I/M programs, could yield substantial health benefits relative tothe BAU scenario, and serve as co-benefits of greenhouse gas control strategies. Despite uncertainty associatedwith the key assumptions used to estimate benefits, we find that with a high level confidence, the I/Mprograms will produce health benefits whose economic impacts considerably outweigh the expenditures onpolicy implementation.
, University of North Carolina atTel.: +1 919 360 0636; fax: +1
Urban air pollution associated with transportation is currently amajor public health concern in many large developing metropolitanareas as rapid expansion of vehicle use is a common feature in theseareas, and scientific studies have found suggestive evidence relatingadverse human health effects to exposure to primary traffic-generatedair pollution (HEI, 2010). Although the negative health impacts of urbanair pollution have been widely investigated, the link between specifictraffic-related emission control policies and associated health benefitshas not been well characterized and rarely been integrated into policyanalysis in a developing country context. This issue is especiallygermane under the emerging Copenhagen Accord, because policies
aimed at reducing greenhouse gas emissions find weak support indeveloping nations faced with needed economic growth, but gain insupport as the co-benefits of greenhouse gas reductions are included infinancial analyses. One co-benefit of greenhouse gas reduction in thetransportation sector is reduced emissions of particulatematter, leadingto improved health and economic performance.
This paper studies the capital city of Thailand—Bangkok,a megacity that has been suffering from severe adverse health effectsattributable to ambient particulate matter (PM) for more than adecade. To mitigate the serious urban air pollution and promotesustainable transportation development in Bangkok, a justified goal inits own right but also a co-benefit of climate change policies,enhanced inspection and maintenance (I/M) programs targeting PMemissions from in-use diesel-fueled vehicles and motorcycles havebeen proposed to be adopted. These are relatively new types of I/Mprograms distinct from traditional I/M programs that focus primarilyon other pollutants from gasoline-fueled vehicles. At present, PM-oriented I/M programs have gained a growing interest due to thepublic concern over the health threat posed by fine PM emitted from
1775Y. Li, D.J. Crawford-Brown / Science of the Total Environment 409 (2011) 1774–1785
in-use vehicles. In particular in less developed nations, vehicles areoften poorly maintained—leading to both PM emissions and fuelinefficiency—and there are still many old vehicles running on roads.PM control policy options have been proposed as imperative solutionsto the air pollution resulting from mobile sources in these regions inaddition to setting new vehicle emissions standards and enforcingfuel quality regulations. The goal of this research is to understand thehealth benefits potentially achieved by enforcing these programs sobetter cost–benefit assessments can be developed.
An integrated procedure is used here that combines exposureassessment, exposure–response assessment and economic valuation.Our methods are based on those used by the U.S. EPA in makingregulatory decisions based on cost–benefit comparisons (USEPA, 1999),and others to assess the benefits of reducing ambient air pollution (e.g.Ostro and Chestnut, 1998; Levy and Spengler, 2002; Wang andMauzerall, 2006; Cifuentes et al., 2001). While most previous studieseither focus on the health benefits from achieving a target air qualitystandard, or the health related environmental costs attributable to aparticular sector of energy use, fewer studies have examined specificcontrol policy scenarios targeting emissions from the transportationsector. This studyprovides a systematic evaluationof the health benefitsresulting from new PM-related vehicle I/M programs to evaluate thebenefits of air pollution control policies.
The paper is organized as follows. Section 2 describes the methodsused to quantify the health benefits of PM-related I/M programs in thestudy area, includingpolicy scenario development, exposure assessment,exposure–response assessment, and economic valuation. Section 3presents estimates of health damage costs attributable to vehicle PMemissions in the base year and benefits of the I/M programs in futureyears, and also compares the health benefits with the costs ofimplementing the programs. Section 4 summarizes the major conclu-sions of this study.
2. Methods
2.1. Policy scenario development
This study uses the year 2000 as the base year to establish theemission baseline. This year was chosen because in the late 1990s, inconfronting severe PM pollution in the Bangkok area, the Thaigovernment implemented a series of cost-saving mitigation measures,suchasproviding carengine tune-up service to thepublic for free, pavingstreet shoulders to reduce road dust, etc. Although these measures didsignificantly contribute to preventing Bangkok's air quality from gettingworse, ambient PM concentrations continue to exceed the air qualitystandards. There remains, therefore, significant pressure to bring aboutfurther reductions. An additional advantage of this base year selection isthat the Bangkok Air Quality Management Project sponsored by theWorld Bank, conducted during the late 1990s through early 2000s, hasprovided reliable data for the year 2000 required to estimate the localpublic health benefits of emission control policies. These data are rare inless developed countries.
We consider both a business-as-usual (BAU) scenario, i.e. thescenario with no further control policies relative to the base year2000, and a hypothetical abatement policy scenario enforcingenhanced PM-oriented vehicle inspection and maintenance (I/M)programs targeting all diesel-fueled vehicles and motorcycles in theBangkok area. The programs are assumed to start in 2008 and target atotal reduction of 25% in PM10 emissions (all particulates with anaerodynamic diameter of less than or equal to 10 μm) from all diesel-fueled vehicles and motor vehicles relative to the baseline. We alsotest the sensitivity when the I/M programs do not accomplish theemission cut target, i.e. achieving less than 25% PM10 emission, whichis very likely in reality and will result in smaller benefits from theprograms.
2.2. Simulation of changes in human exposure to ambient particulatematter due to policy implementation
To estimate the changes in ambient PM concentrations under apolicy scenario, we began by defining the modeling domain andobtaining PM monitoring data. The modeling domain of this study isthe Bangkok Metropolitan Area (BMA), which includes Bangkok andits surrounding five provinces, Samut Prakarn, Nonthaburi, PathumThani, Nakhon Pathom and Samut Sakhon, altogether covering an areaof 7761.50 km2 (the area of Bangkok is 1568.20 km2), or about 1.5% ofthe area of Thailand, and with a population of 12 million (as of 2008).The reason that Bangkok's peripheral provinces were also included isthat these areas are closely linked in terms of traffic and economicdevelopment (Oanh and Zhang, 2004). We limited the impacts of theI/M programs on PM air quality within metropolitan Bangkok due tothe reason that, unlike particulates from industrial point sources ornatural sources, which may be transported over very long distanceswith high stacks or climate conditions, the particulate emissions fromvehicle exhaust considered here are expected to have much greaterimpacts at the local scale than at the regional scale. Nevertheless, weare aware that ignoring the impacts in the larger region may slightlyunderestimate the total benefits resulting from policy implementation.
Daily monitored PM10 data were obtained from the PollutionControl Department (PCD), Thailand. Of the 17 permanent air qualitymonitoring stations in the BMA operated by PCD, 8 have daily PM10
data in the base year 2000, with all these stations located in Bangkok.In the case of missing monitoring data, daily PM10 concentrations onthese days were estimated based on data of the previous and nextdays using linear interpolation. If monitoring data were missing forten or more continuous days, the seasonal average PM10 level wasutilized as an approximation of the daily concentrations to reflect thepossible significant difference among seasons (including Winter,Summer and Rainy) throughout a year.
To estimate the ambient PM10 concentrations attributed to mobilesources, it was assumed that the shares of different source categories formonitored PM10 concentrations are proportional to their relativecontribution to the emissions inventory of PM10. For example, if PM10
emitted frommobile sources account for 50% of the total PM10 emissionsfrom all man-made sources, then 50% of the total anthropogenic ambientPM10 concentrations, defined as a monitored concentration less theregional background of PM10, are contributed by these sources. It was notjudged feasible to attempt a more precise allocation of the relationshipbetween emissions and concentration for each source. Therefore, thereduction in ambient PM10 concentrations attributed tomobile sources isalso proportional to the projected emission reduction in these sources. Inaddition, it was assumed that except for mobile sources, all other PM10
emission sources will not be affected by I/M programs considered. Thisapproach allows us to calculate the daily PM10 concentrations atmonitorsin the base year attributable to mobiles sources, and to predict the futurechanges in concentrations observed at monitors that reflect thehypothetical imposition of the I/M programs by using Bangkok's PM10
emission inventory and daily monitoring data in the year 2000.The best available PM10 emission inventory information in the BMA
was examined, and it was found that in this area there still existssignificant uncertainty in the relative share of contribution to the totalPM10 emissions by different source categories, including mobile (motorvehicles), industrial, road re-entrainmentand construction. For example,aWorld Bank (2002) publication stated that emission inventories in theBMA prepared and reported by different organizations can differ by afactor of twenty. In order to reduce the possible biases introduced by asingle study on the emissions inventory in the BMA, this study collectedall available studies and averaged the results. We estimate that in thebase year 2000, there were 15,650 tons of PM10 emitted from motorvehicles in the BMA, accounting for 32% of total anthropogenic PM10
emissions in that year. In uncertainty analysis, we represent theuncertainty of this variable as a Uniform distribution, with a lower
1776 Y. Li, D.J. Crawford-Brown / Science of the Total Environment 409 (2011) 1774–1785
boundbeinghalf of the central value (i.e.16%) andupper boundbeing1.5times the central value (i.e. 47%).
Although particulates from forest burning in Southeast Asia mightbe a significant non-anthropogenic regional source of PM10 in thestudy area, to our knowledge, currently there are very few studies onthe regional background levels of PM10 in Thailand. The study byPuangthongthub (2006) estimated the regional PM10 background indifferent regions of Thailand by using the 5th percentile of monitoringdata collected in each corresponding region during 1998–2003. Theestimated regional background level in different regions in Thailandwas 3.02 μg/m3 in the BMA and the central region, and fell into therange of 7.42–12.33 μg/m3 in the other parts of Thailand, indicatingthat the regional background levels of PM10 in Thailand are generallysmall. The BMA estimate of 3.02 μg/m3 was applied in this study andassumed to be a constant regional background level of PM10 across thestudy area and study period. In uncertainty analysis, we represent theuncertainty of regional background PM10 level as a Triangulardistribution, with a lower bound being zero (theoretical lowest level)and upper bound being theWorld Health Organization suggested PM10
natural background level—10 μg/m3 (Ostro, 2004).In order to estimate the number of people affected by the changes in
PM10 levels at a monitoring site, the total population in Bangkok wasequally assigned to eachof the 8monitoring stations that haddaily PM10
concentration data in 2000. Here it is assumed that the population inBangkok is equally distributedaroundeachstation. Forpopulation in thesurrounding five provinces, daily PM10 concentrations averaged acrossthe eight stations were used as estimates for the daily concentrationsdue to the lack of PM10 monitoring data in these provinces.
2.3. Estimation of the changes in health outcomes attributable to thechanges in particulate matter exposure
2.3.1. Health endpoints consideredA wide spectrum of adverse health effects, from acute respiratory
symptoms to premature deaths, is associated with exposure to ambientPM. Although previous health benefits studies consistently suggestedthat death usually represents the dominant impact in economic terms,for a comprehensive evaluation of the health benefits associated withpollutionmitigation, bothmortality andmorbidity effects are included inthis study. Those health endpoints forwhich the concentration–response(CR) relationships have not beenquantifiedwere excluded. Based onU.S.EPA's list of PMhealth effects and other relevant literature (e.g. Ostro andChestnut, 1998; Thanh and Lefevre, 2001; Cifuentes et al., 2001), adultand infant mortality outcomes and seven morbidity outcomes areincluded in this study, including chronic bronchitis, respiratory hospitaladmissions, cardiovascular hospital admissions, emergency room visits,acute asthma attacks, acute respiratory symptom days and respiratoryrelated restricted activity days.
2.3.1.1. Premature mortality. Earlier epidemiological studies conductedin the U.S. and Europe reported an increase of 0.5–1% in daily all-causemortality per 10 μg/m3 increase in daily-average PM10 (e.g. Stieb et al.,
Table 1Concentration–response coefficients selected to estimate mortality attributable to traffic-re
Health endpoints Reference Study area
Acute adult mortality (AM) Ostro et al., 1999 Bangkok, ThailandVajanapoom et al., 2002Weighted averageAdjusted estimate (*1.5)
Acute infant mortality (AIM) Loomis et al., 1999 Mexico City, MexicAdjusted estimate (*1.5)
The numbers in bold are those used in calculation.
2002; Anderson et al., 2004; Dominici et al., 2005). It is also generallyaccepted that given the disparity in particle characteristics and variousfactors that influence public health status, results from studiesconducted in an area should not be extrapolated to another areawithoutscrutiny of transferability. In particular, the factors affecting the PM-health relationship might be significantly different between developingand developed countries. Given these reasons, to estimate the healthbenefits associatedwith PM control, this study relied on epidemiologicalstudies conducted in Thailand and gave priority to studies in otherdeveloping regions in Asia wherever possible.
To date two epidemiological studies on the associations betweendaily PM10 concentrations and short-term mortality in Bangkok havebeen published by Ostro et al. (1999) and Vajanapoom et al. (2002). Fora 10 μg/m3 increase in daily PM10, the former reported 1.0% increase indaily all-causemortalitywhereas the latter reported0.77%. These resultsindicate a large level of congruence in CR estimates for acute mortalityfrom Thailand and other regions in the world. A weighted average ofthese two studies was calculated. As the second study estimated the CRcoefficient based on data over a longer time period, and used multi-sitemean monitoring data, whereas the first one used only data from onemonitoring site, a larger weight (0.7) was assigned to the second study,resulting in a weighted average of 0.084% (Table 1).
Another issue is that neither study differentiated PM from distinctsource categories, which was often the case in earlier epidemiologicalstudies. In recent years, research is increasingly addressing thecontribution of PM origins to the severity of its health impacts, andstudies have argued that particulates frommotor vehicle sources maypose a greater risk for public health than particulates from othersources such as industrial and wind-blow dust (Laden et al., 2000).Given that this study aims at estimating the health risk reductionsassociated with the decrease of PM emissions from mobile sources,the estimates obtained by studies that did not differentiate PMsources may understate the changes in health outcomes associatedwith the policy scenario considered here.
Currently there is little evidenceon the relativemagnitudeof source-specific PM10 health risk. To investigate this, we conducted a stratifiedmeta-analysis of time-series studies on PM10 and acute mortality, inwhich relevant studies were grouped by the primary sources of PMreported in those studies (see Li, 2008 for details). Commonly used PMsource categories—mobile, industries and crustal sources—wereemployed to stratify the relevant time-series studies. The resultsindicate that the effects of PM10 from mobile sources are about twotimes larger than those of PM10 from industrial and crustal sources, andare approximately 1.5 times larger than the overall health effects ofPM10.This implies that traffic-related PM10 in the air poses the greatestrisk to human health compared to industrial and crustal sources. Basedon thefindings from ourmeta-analysis, the estimates of PM10-mortalityin Bangkok from previous epidemiological studies were multiplied by afactor of 1.5 to obtain the health effects per unit exposure associatedwith mobile sources, given that the earlier epidemiological studies inBangkok were based on the mixed effects of PM10 from all sources. Thesame adjustment factor (1.5) was applied to the infant mortality
lated PM10 in the Bangkok Metropolitan Area.
Study period Change in daily mortality per 1 μg/m3
1777Y. Li, D.J. Crawford-Brown / Science of the Total Environment 409 (2011) 1774–1785
estimate and all the morbidity estimates (Tables 1 and 2). The selectedepidemiological studies on PM10 mortality as well as the adjustedestimates used are listed in Table 1.
The reductions in PM10 levels and their health effects areestimated on the same day regardless of the possible time-lag afterexposure, since the time-lag of health effects does not affect theaggregated estimate over a year. Acute health effects for each day areaggregated over an entire year to obtain the annual estimates.
2.3.1.2. Morbidity. Comparedwith epidemiological studies onmortality,studies on the association of morbidity and air pollution exposure areless comprehensive (Wang and Mauzerall, 2006). We selected a fewstudies in China for the health endpoints for which Thailand estimatesare not available, and a study in theU.S. for restricted activitydaysdue tothe factor that no estimates are found in developing regions. To reflectthe source-specific PM morbidity risks, in the absence of informationspecific for morbidity, the same adjustment factor found for mortality(1.5) is used, assuming a linear relationship between mortality andmorbidity. Table 2 lists the selected PM-morbidity estimates, theirreferences, and the adjusted estimates.
2.3.2. Estimating adverse health outcomesThe selected CR coefficients for each health endpoint are used in
health impact functions to calculate the changes in health outcomes.This study employs an exponential CR function for PM health effects.2
Therefore, the health impact function is:
Δy = y0 × eβ⋅Δx−1� �
ð1Þ
(Deck et al., 2001; Hubbell et al., 2005; Ostro et al., 2006; Sanhuezaet al., 2003) where Δy is the change in the number of cases of eachhealth endpoint (i.e. mortality or morbidity); y0 is the baselineincidence, equal to the baseline incidence rate (I) times the potentiallyaffected population POP:
y0 = I × POP ð2Þ
β is the concentration–response coefficient estimated from acorresponding damage function (refer to Tables 1 and 2 for values ofβ and uncertain ranges), and Δx is the estimated change in the ambientPM levels as a result of a policy (daily 24-hour averagemeasure for acuteeffects and annual average measure for chronic effects). The baselineincidence rates of somehealthendpoints are not available inThailand. Inthis case, a similar strategy as in obtaining the CR functions is used. Thatis, a baseline incidence rate fromanother developing country in Asia hasthe priority to be selected to approximate the rates in Thailand.Otherwise, a rate in a western developed country is used. Table 3summarizes the baseline incidence rates included in this study and theirreferences. In uncertainty analysis, we represent the uncertainty of theconcentration–response coefficients β as Normal distributions (seeTable 3 for the mean values and 95% CIs).
2.4. Economic valuation
To translate the benefits of reduced health risks associated withreductions in ambient concentrations of air pollution into economicterms, it is well accepted that the most appropriate economicvaluation method is willingness to pay (WTP) (e.g. Freeman, 1993;Hubbell et al., 2005). However, WTP estimates are not available forsome health effects. In the absence of WTP estimates, cost-of-illness(COI), which is more accessible from previous studies, is often used to
2 Some studies use a linear relationship of health response to air pollution at eachhealth endpoint (e.g. Wang and Mauzerall, 2006). A linear health impact function is:Δy=y0×β×Δx, where all parameters are defined the same as in the health impactfunction derived from an exponential model.
evaluate health effects. A COI study estimates two types of costs of anillness: the direct costs (medical and non-medical) associated withthe illness, and the indirect costs associated with lost productivity dueto morbidity or premature mortality (Haddix et al., 1996). Theweakness of COI is that it generally underestimates the true value ofreducing the risk of a health effect, because it does not reflect thevalue of avoided pain and suffering (Hubbell et al., 2005). Therefore,COI should be used along with WTP to provide reliable economicmeasures. In addition, in the case that WTP for a certain healthoutcome is absent,WTP can be estimated based on empirical evidencefor the ratios between WTP and COI.
Another issue of concern is that currently there is no single WTPstudy on health effects of air pollution available in Thailand (Thanhand Lefevre, 2001). Since WTP is significantly affected by factors suchas income, values of WTP may not be readily transferred from acountry to another without adjustment (Thanh and Lefevre, 2001).Scholars have developed several approaches to make such a transfer.The simplest one corrects only for the income difference betweencountries (Thanh and Lefevre, 2001). That is, in transferring the valuesfrom a country to another, only the difference in per capita income isaccounted for and the values are assumed to be proportional toincome. Therefore, based on the disparity in median per capitalincomes of the U.S. ($29,760, Bureau of Economic Analysis, 2003) andThailand ($1345, National Statistical Office Thailand, based on theexchange rate of 1:43.23 in 20003), a ratio of 0.046 is derived.
However, economists have suggested that income elasticity ofWTP is less than one, whichmeans that ‘other things being equal,WTPin low-income countries is lower thanWTP in high-income countries,but proportionately less than the income differentials’ (Thanh andLefevre, 2001). Given these, this study uses a ratio of 0.2 to extrapolatevalues of all health end points from U.S. studies to the currentThailand context, as was used in previous World Bank reports onThailand (Hagler Bailly, 1998). The WTPs or COIs for each healthendpoint used in this study, as well as their probability weights torepresent the uncertainty, are abstracted from existing literature(Table 4).
Premature mortality is often measured based on the years of lifelost (YOLL), due to the fact that when an individual dies prematurelydue to long-term exposure to air pollution, he or she may lose only afew years of his or her life (Rabl, 2003; Wang and Mauzerall, 2006). Itis argued that depending on whether economic valuation is based onthe number of lives lost or YOLL, the perceived health benefits ofpollutionmitigationmay vary sufficiently to alter the results of a cost–benefit analysis (Wang and Mauzerall, 2006). However, given thecomplexity of estimating YOLL and the insufficiency of data in thedeveloping world, this study only considers the number of lives lost ineconomic valuation.
The economic value associated with change in health outcome (Vi)is given by
Vi = Ai × WTPi ð3Þ
where, Ai is the change in the number of a specific health endpoint i,and WTPi is the willingness to pay to avoid the corresponding healthendpoint.
2.5. Estimating the changes in vehicle emissions and population
When the potential health benefits associated with projectedemission reductions in future years are simulated, the possible changesin the key parameters over time need to be included in the model. Thefollowing parameters are expected to change significantly over time:(1) total number of vehicles is likely to increase rapidly in a developing
3 Source of exchange rate: Bank of Thailand, 2007, average exchange rate: historicaldata.
Table 2Concentration–response coefficients selected to estimate morbidity attributable to traffic-related PM10 in the Bangkok Metropolitan Area.
a A study suggested a PM10/TSP ratio of 0.6 in Bangkok (Hagler Bailly, 1998).b Units of acute respiratory symptom days and restricted activity days are per capita CR for 1 μg/m3 change in annual average PM10. These risk factors are abstracted from Hagler
Bailly, 1998. These factors are estimated by using CR coefficients from the original regression analysis and baseline incidence rates, but they are not provided in the documents.Therefore, the risk factors are applied directly here.
c The original study only surveyed adults of ages 18–65, who were working at the time when the study was carried out, for three reasons: first, restrictions in activity are easier todetect for workers since their time is usually more structured than that of non-workers; second, workers daily activity patterns tend to be similar; and third, the time and length ofexposure to a given outdoor air pollutant tend to be similar. In anticipation that PM air pollution will lead to reduced activities, if the effect exists, among the general populationrather other among the working ages exclusively, the CR coefficient is applied to all age groups, assuming that the effect is the same among the entire population.
1778 Y. Li, D.J. Crawford-Brown / Science of the Total Environment 409 (2011) 1774–1785
area such as the BMA; (2) average PM emissions per vehicle-km arelikely to go down due to the tighter emission standards set for newvehicles and the improvement in emission control technologies inmodern vehicles; (3) kilometers traveled per vehicle in the study areamay increase as a result of the increase in demand for travel; and(4) population is likely to grow fast as the result of rapid urbanization.We used the best available empirical data to quantitatively estimate thechanges in these four parameters over time. The detailed methods anddata used are described in Appendix A.
2.6. Estimating the costs of I/M programs
In order to compare the benefits of the I/M programs (the primarybenefits are health-relatedbenefits)with the social costs of implementingthem, we also estimated program costs. The primary cost components ofan I/M program consist of fixed costs including program start-up costs,and variable costs including operating andmaintenance costs and vehicleemission repair costs. Parsons (2001) estimated the annual costs of the
Table 3Estimated baseline rates of selected health endpoints in the Bangkok Metropolitan Area in
a Units of rates are cases per year per person in the exposed population.b Natural deaths are all deaths other than accidents, suicides and homicides. In Bangkokc No baseline rates are needed to calculate acute respiratory and restricted activity days,
proposed I/M programs targeting diesel-fueled vehicles and motorcyclesin the BMA. However, their estimates of costs only included start-up costs(capital and land use costs) and inspection costs (operating andmaintenance costs), but excluded problem vehicle repair costs, whichcan be a substantial portion of the total social costs of I/M programs(National Research Council, 2001). In order to derive the total social costsof the I/M programs considered here, the repair costs are estimated andaggregated with the other costs estimated by Parsons (2001). The detailsof cost estimation are described in Appendix B.
3. Results and discussion
3.1. Health damage in the BMA attributable to PM10 from motor vehiclesources in 2000
Using the integrated approach described above, this study estimatedthe total health damage costs due to exposure to traffic-related fineparticulatematter to be 2678 million 2000U.S. dollars in the BMA in the
al. (2002); Wang and Mauzerall (2006) Chinaailly (1998) Bangkok, Thailandailly (1998) Bangkok, Thailandet al. (2004) Atlanta, USAal. (2002); Wang and Mauzerall (2006) Chinaal. (2002); Wang and Mauzerall (2006) China
, about 93% of deaths are natural (Hagler Bailly, 1998).given that the units of these two health endpoints are cases per capita (See Table 2).
Table 4Valuation of selected health endpoints.
Health endpoints US values (2000$)a
(Selected probability weights in parenthesis)Thailand values (2000$)b
(Selected probability weights in parenthesis)Type of estimate
a The source of the values and their selected probability weights is Ostro and Chestnut (1998), with the exception of the value of acute asthma attacks, which is abstracted fromUSEPA (1997). Inflation is calculated using the Inflation Calculator by U.S. Department of Labor, Bureau of Labor Statistics: http://data.bls.gov/cgi-bin/cpicalc.pl.
b Thailand value=US value X 0.2.c Adjusted COI is COI multiplied by 2 to approximate WTP, calculated by the authors of the reference.
1779Y. Li, D.J. Crawford-Brown / Science of the Total Environment 409 (2011) 1774–1785
base year 2000. Premature deaths account for 1369.2 million dollarsor 51.1% of the total health damage costs. The estimated healthimpacts (number of cases) and health damage costs in 2000 aresummarized in Table 5. The results indicate that the economic loss dueto exposure to PM10 emissions from transportation is substantial in theBMA, which accounts for about 2.35% of Thailand's Gross DomesticProduct (GDP) in 2000 (113.73 billion dollars, National Statistical OfficeThailand).
3.2. Impact on emissions and concentrations of I/M programs
Using 1999 data in Table 7, it was estimated the total PM10emissions from vehicles in BMA were 15650 tons per year in 2000,and this number continues to increase over time. Fig. 1 illustrates theprojected total PM10 emissions frommotor vehicles in the BMAduring2008–2015. The graph indicates that with no further control policy(the BAU scenario), despite the decrease in emission rates per vehicleas a result of improved vehicle performance, total vehicle PMemissions in the BMA increase significantly, mainly attributable tothe rapid growth in vehicle population. In 2008, it was estimated to bemore than 20,000 tons per year, and by the year 2015, BMA's totalannual PM10 emissions from motor vehicles will be more than3000 tons—double the emissions in the base year 2000. On the otherhand, as Fig. 1 shows, the I/M programs are expected to considerablyreduce the total PM emissions compared to the BAU scenario: cuttingthe emissions in 2008 close to the level in 2000 and continuing toreduce emissions every year. As a result of emission reductions,ambient PM10 concentrations are expected to decline with the I/Mprograms compared to the BAU scenario. Fig. 2 shows the projectedaverage PM10 concentrations in the BMA under the two scenarios.
Table 5Health damage due to exposure to traffic-related PM10 in the Bangkok MetropolitanArea in 2000.
Health endpoints Total number of cases Health damage costs(Million 2000 U.S.$)
Death 1683 1369.2Chronic bronchitis 17,304 860.3Respiratory hospitaladmissions
Restricted activity days 15,575,427 218.2Total costs 2677.7
3.3. Health benefits of the I/M programs
As a result of the reduction of PM concentrations, the healthdamages attributed to exposure to ambient PM are expected todecrease with I/M compared to the BAU scenario, as shown in Fig. 3.The upper line in Fig. 3 illustrates the annual economic costs of PMemissions from motor vehicles in the BMA under the BAU scenario. Itindicates that health damages due to PM10 from transportationsources will considerably increase relative to the baseline year 2000 ifno control policy is implemented. For instance, the total damage costin 2015 (6436.0 Million $) represents an increase by about a factor of2.5 times relative to the base year of 2000 (2677.7 Million $, seeTable 5). However, the lower line in Fig. 3 indicates that implementingsome emission control policies such as the PM-oriented I/M programs,if successful, is anticipated to considerably reduce PM10 emissionsfrom the existing vehicle fleets and consequently produce significanthealth benefits. Table 6 summarizes the health benefits associatedwith the I/M programs if a 25% vehicle PM10 emission reductionrelative to the baseline is achieved by the programs. Table 6 indicatethat implementing I/M programs targeting all diesel-fueled vehiclesandmotorcycles in the BMAwill produce significant benefits to publichealth resulting from air pollution mitigation, if the programs canachieve the target emission reductions from motor vehicles. Ingeneral, as in the BAU case, avoided premature deaths account forslightly more than 50% of the total benefits annually.
In comparing the costs of I/M to the health benefits, this studyestimated the total annual costs of the I/M programs in the BMA to beapproximately 147 million 2000 US dollars (Table 8). Therefore, ourbest estimates indicate that the total annual benefits of the programsare expected to significantly outweigh the total costs.
0
5000
10000
15000
20000
25000
30000
35000
2008 2009 2010 2011 2012 2013 2014 2015
PM
10 E
mis
sion
(T
ons/
Yea
r)
Year
BAU Scenario I/M Scenario
Fig. 1. Projected total PM10 emissions frommotor vehicles in the Bangkok MetropolitanArea: BAU vs. I/M scenarios (25% emission reductions), 2008–2015.
Fig. 2. Projected average PM10 concentrations in the Bangkok Metropolitan Area: BAUvs. I/M scenarios (25% emission reductions), 2008–2015.
1780 Y. Li, D.J. Crawford-Brown / Science of the Total Environment 409 (2011) 1774–1785
3.4. Sensitivity and uncertainty analysis
In this study, the estimates of total health benefits under the policyscenarios considered are based on several key assumptions that areuncertain. Given this, sensitivity and uncertainty analyses areperformed to determine how sensitive the best estimates of healthbenefits associated with control policies are to the uncertainty in eachpremise employed. The uncertainty analysis was conducted usingMonte Carlo simulation and was performed in Crystal Ball AnalysisSoftware. The sample size was 5000.
3.4.1. Sensitivity analysisFour premises are examined for their roles in changing the estimates
of health benefits of the I/M programs achieved.
3.4.1.1. Premise 1: Source-specific vs. non-source-specific PM10 healthrisk. In the earlier analysis, in order to reflect the fact that the effects ofPM10 from mobile sources are larger than the overall health effects ofPM10, an adjustment factor of 1.5 was used to estimate the mobilesource-specific health risk of PM10, based on the findings from thestratified meta-analysis conducted in this study. In the sensitivityanalysis, the effect of the source-specific PM10 health risk assumptionis examined by estimating the total health benefits with and withoutusing the 1.5 adjustment factor. Removal of the adjustment factorresulted in a 35% decrease in the total health benefits estimated.However, we argue that the health benefits of the I/M programs arevery likely to be underestimated if the source-specific PM10 healthrisk is not taken into account, since scientific studies consistently findthat particles frommobile sources pose a greater risk to human health
0
1000
2000
3000
4000
5000
6000
7000
2008 2009 2010 2011 2012 2013 2014 2015
Year
Tot
al A
nnua
l Hea
lth
Dam
ages
(200
0 M
illio
n U
S$)
BAU Scenario I/M Scenario
Fig. 3. Total annual health damages attributable to PM10 emissions frommotor vehicles:BAU vs. I/M (25% emission reductions) scenarios, 2008–2015.
per unit exposure than particles from other sources. Our adjusted CRcoefficient for short-term all-cause mortality is 0.13% per 1 μg/m3
increase in daily PM10. A study in the U.S. reported a 0.34% increase indaily mortality per 1 μg/m3 increase in PM2.5 from mobile sources.Therefore, our estimate of the CR coefficient for traffic-related PMpollution might still be conservative.
3.4.1.2. Premise 2: Acute vs. chronic mortality effects of particulate matterpollution. Earlier analysis considers acute mortality effect by usingshort-term (time-series) studies in computing the total health benefitsdue to the fact that only these studies are readily available in Thailand.Relying on short-term studies may underestimate the health benefitsbecause it is likely that air pollution also causes chronic health effects,including mortality (Künzli et al., 2001). In this analysis the sensitivityof the total health benefits to the short-term vs. long-term mortalityeffects are examined by replacing time-series studies with cohortstudies. By surveying a recent comprehensive review of literature onhealth effects of traffic-related air pollution (HEI, 2010), we selected thecohort study by Jerrett et al. (2005), which reported a 1.1% in all-causemortality per 1 μg/m3 increase in PM2.5.
4 The study by Woodruff et al.(1997) is selected for postneonatal infant (28 days–1 year) mortalityestimation. Moreover, the long-term effects of ambient particulatematter are currently considered to be solely attributable to theexposure to the fine portion of PM10, i.e. PM2.5. Due to the lack ofPM2.5monitoring data in the baseline year 2000 in Thailand, the ratio ofPM2.5/PM10 was abstracted from the literature and used to estimateambient PM2.5 concentrations based on the available PM10 monitoringdata. Note that this study focuses on diesel exhaust particulates. Anemission inventory developed by California Air Resource Boardindicates that approximately 98% of the particles emitted from dieselengines are less than 10 μm in diameter, 94% less than 2.5 μm indiameter, and 92% less than 1.0 μm in diameter (National ToxicologyProgram, 2005). Based on this information, it is estimated that the ratioof PM2.5/PM10 is about 0.959 (94%÷98%≈0.959).This sensitivityanalysis results in an estimate of the total annual health benefits ofthe I/M programs that is approximately twice that of the originalestimate. Therefore, it is likely that the I/M will provide moresubstantial benefits than the estimates listed in Table 6.
3.4.1.3. Premise 3: Registered vs. real population. In the earlier analysis,the total health benefits are estimated using the registered populationto calculate the total number of people affected by air pollution.However, as is typical in large urban areas in developing countries, theactual population in the study area is still largely unknown, as thereare many people who commute to work in Bangkok or live in the citywithout registration (UNEP, 2004). A study of the un-registeredpopulation in Bangkok indicated that the ratio of true population tothe registered population is approximately 1.57 (UNEP, 2004),indicating that the total benefits attributable to air pollution controlmay be underestimated by relying on registered population. Thesensitivity analysis is conducted by assuming that the actualpopulation is 1.57 times the registered population, and this ratio isapplied to Bangkok and its five surrounding provinces. This results in a45% increase in the estimated health benefits of the I/M programs.
3.4.1.4. Premise 4: Exposure to the PM10 levels at the permanent vs.roadside monitoring stations. The last sensitivity analysis is conductedto examine the changes in the expected health benefits if somefraction of the population in the BMA is exposed to the roadside PM10
levels. Our analysis so far uses the ambient PM10 concentrations fromeight air quality monitoring stations in the BMA as the surrogate ofexposure, presuming that the general population is exposed to the
4 The HEI report abstracted two estimates from Jerrett et al (2005): 1.7% and 1.1%,depending on different covariates used. We selected the lower estimate to provide aconservative estimate of the health benefits.
Table 6Potential health benefits of I/M programs targeting 25% particulate matter emission reductions from motor vehicles in the Bangkok Metropolitan Area (Year: 2008–2015).
Economic valuation (million 2000 U.S. dollars)Health benefits of avoided deaths 457.9 485.7 517.1 552.5 592.2 636.8 686.8 743.0Health benefits of avoided illness 450.0 478.3 510.2 546.1 586.5 632.0 683.1 740.5Total health benefits 907.9 964.0 1027.3 1098.6 1178.7 1268.8 1369.9 1483.5
1781Y. Li, D.J. Crawford-Brown / Science of the Total Environment 409 (2011) 1774–1785
pollution levels at the monitoring sites. All of the eight are permanentstations located in residential, business or school areas, and they areconsidered to be representative of most residents’ exposure levels.There are also temporary roadside stations in the BMA for short-term(e.g. from 15 days to 1 month) air quality monitoring, since the PCDare concerned that the air quality in areas near road traffic can bemuch worse than the residential areas. For example, in the year 2000,20 roadside temporary stations monitored daily 24-hour averagePM10 for approximately 300 days (each station was only used forabout half to onemonthwith only one station used on any single day).It was found that daily PM10 concentrations at permanent stations androadside stations correlated well (see an example in Appendix C). Bycomparing the available daily PM10 concentrations at the roadsidestationswith the daily concentrations at the 8 permanent stations, it isfound that daily PM10 levels at the roadside stations (CR) weregenerally 2–3 times higher than those at the permanent stations (CP).The average ratio of CR/CP was 2.5.
It is largely uncertain what fraction of exposure takes place at theroadside levels in the study area. In general, people who live or workclose to road traffic (e.g. traffic policemen, street vendors, etc.) areexposed to more traffic-related air pollution. In testing the sensitivityof total health benefits to different levels of exposure (PM10 levels atgeneral residential and business areas vs. levels at road traffic areas), itis assumed that 80% of total exposure is at the levels found in thepermanent stations and the remaining 20% is at the levels found in theroadside stations. It is also assumed that the roadside PM10 levels are
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
-200 0 200 400 600 800 1000 1200
CD
F
Annual Net Bene
4% 7%
Fig. 4. Cumulativedistribution functions (CDF) of the total net Benefits atdifferent levels of over
2.5 times the levels at the permanent stations, based on the averageratio. Therefore, the average PM10 exposure level in this sensitivityanalysis can be expressed as:
Exposure = 0:8 × CP + 0:2 × 2:5 × CPð Þ ð4Þ
This contrasts with the exposure in our earlier analysis, whereExposure=CP.
Changing this assumption resulted in a 48% increase in the estimatedtotal health benefits of I/M. In general, most people are not expected tobe exposed to the roadside level of air pollution. However, in the BMA,there may be a higher fraction of people that are exposed to roadsidepollution levels due to less prevalence of air-conditioning, and morepeople who live or work in the roadside areas. The total health benefitsof the I/M programs will significantly increase if a large fraction ofexposure is at the roadside levels. Therefore, it is likely our earliercalculation understates the benefits of I/M. Nevertheless, furtherresearch is warranted to verify this assumption.
3.4.2. Uncertainty analysisMonte Carlo simulation indicates that when the uncertainty in the
four key variables noted above (regional background PM10 level, therelative contribution to total PM10 emissions by mobile sources,concentration–response coefficients, and monetary values of eachhealth endpoints) is taken into account, the estimated total annual
1400 1600 1800 2000 2200 2400 2600 2800
fits (Million 2000$)
9% 11% 25%
all PM10 emission reduction frommotor vehicles in theBangkokMetropolitanArea in2008.
1782 Y. Li, D.J. Crawford-Brown / Science of the Total Environment 409 (2011) 1774–1785
health benefits of the I/M programs in 2008 fall into the range of [340,2000] (95% CI), with a mean of 926 million 2000 U.S. dollars.
As noted above, we also applied different levels of overall PM10
emission reductions achieved by the I/M programs (less than the levelof 25% used in the deterministic estimates) to quantify the overalluncertainty in the health benefits of the programs. Fig. 4 shows thecumulative distribution functions (CDF) of the net health benefits ofthe I/M (total health benefits minus the total costs, estimated to be147 million 2000 US dollars per year) in 2008 when four differentlevels (4%, 7%, 9% and 11%) of overall PM10 emission reductions areassumed, as well as the original CDF when a 25% reduction level isassumed. It is found that when the level of PM10 emission reductionsdecreases, the total health benefits of the programs, and consequentlythe net benefits, will decrease significantly. As shown, when a 4%overall emission reduction level is achieved by the I/M programs,there is only 45–50% confidence that the benefits of the programsexceed the costs (the net benefits are positive). When the emissioncut level reaches 25%, which is assumed in the earlier analysis in thispaper, the confidence level will be greater than 99%. A 9% overallemission reductions is required to improve the confidence level to95% (at the 5th percentile in Fig. 4). Therefore, it is considered that 9%is the lowest acceptable level of PM10 emission reduction achieved bythe I/M programs.
4. Conclusions
Using the health benefits analysis framework that integrates airquality modeling, exposure assessment, exposure–response assess-ment, economic valuation and policy analysis, it was estimated that inthe year 2000, there were 1682 deaths and a large number of illnessesin the BMA attributable to PM10 emissions from motor vehicles,resulting in a total economic loss of 2.68 billion 2000 U.S. dollars,which approximately equals 2.4% of Thailand's GDP in that year.Mortality and morbidity are responsible for approximately 51% and49% of the total health damage costs, respectively. If no further PMemission control policy targeting motor vehicles is introduced, totalPM10 emissions from vehicles in the BMA will increase considerablyover time due to the rapid growth in vehicle population, even ifthe fleet average emission rates are projected to decrease over timeas the result of improved emission control technologies in modernvehicles.
Control policies targeting PM10 emissions from motor vehicles, ifsuccessful, are expected to yield substantial public health benefitsresulting from the avoided premature death and illnesses. Forexample in 2015, if the PM-related I/M programs targeting alldiesel-fueled vehicles and motorcycles in the BMA achieve 25%overall PM10 emission reductions from motor vehicles, it is estimatedthat these programs will save 913 premature deaths, about 56,000cases of illnesses as well as about 55 million acute respiratorysymptom days or restricted activity days, resulting in a total economicbenefits of 1484 million 2000 U.S. dollars. Avoided premature deathsaccount for slightly more than 50% of the total benefits annually. Oursensitivity analyses indicate that the health benefits achieved mightbe even greater, given several conservative assumptions made toestimate the benefits.
The total costs of an I/M program generally include the costs ofequipment, land use, testing as well as vehicle repairs. Using empiricaldata from earlier studies, it is estimated that the total costs of the I/Mprograms targeting all diesel-fueled vehicles and motorcycles in theBMA are 147 million 2000 U.S. dollars annually. Therefore, it isexpected that the total benefits of the I/M programs considerablyoutweigh the total costs of them, provided that a 25% emission cuttarget is achieved. Our analysis also suggests that 9% is the lowestacceptable level of PM10 emission reduction in order for the I/Mprograms to achieve positive net benefits with a high level ofconfidence.
Finally, we return to the issue of I/M programs being viewed notsolely through the lens of cost-effectiveness in PM emissions reduction,but as a co-benefit of greenhouse gas emissions reduction. Like alldeveloping economies, Thailand is not included as an Annex I nationunder the United Nations Framework Convention on Climate Change.Under both the UNFCCC and the emerging Copenhagen/CancunAccords, Thailand is not required at present to establish or achievetargets for reduction of greenhouse gas (GHG) emissions, in part as aconcession to the needs of suchnations for economic growth to alleviatepoverty. As a result, reduction of GHG emissions in developing nationsoften must be supported by other policy aims such as directimprovement in air quality and health, the topic of the current study.
The I/M program assessed here will bring about reduction in GHGemissions from the transport sector, which represents approximately40% of such emissions in Thailand (World Bank, 2009). Sinceemissions in Thailand are currently on the order of 3.5 tCO2 perperson per year (IEA, 2010), the transport-related emissions areapproximately 1.4 tCO2 per person per year. Under the scenariosexamined here, improvements in the fuel efficiency of vehiclesthrough the I/M program as well as turnover of the fleet to newervehicles results in a 30% decrease in emissions below that projectedfrom future growth of vehicle travel (see Fig. 1). The vehicle programconsidered in the present analysis, therefore, may produce a GHGemissions reduction of approximately 0.4 tCO2 per person per year. Asdescribed, the cost of the I/M program is 147 million USD annually, orapproximately 12 USD per year per person. This a cost effectivenessfor GHG emissions reduction of 30 USD per tCO2, well within therange of costs for emissions reduction by other measures. Equallysignificant, this reduction is actually achieved at a net cost savings(Implementation of the policy minus avoided cost of illness) if the PMemissions reduction is at least the 9% figure mentioned previously.
Acknowledgements
We thank the Thailand Pollution Control Department for providingmonitored air quality data as well as other relevant data anddocuments for this study. We also thank Drs. Richard Andrews,Richard Kamens, Brian Morton (all at University of North Carolina atChapel Hill) and Dr. Mort Webster (at Massachusetts Institute ofTechnology) for providing useful comments for our study.
Appendix A. Estimating the changes in vehicle emissionsand population
A.1. Projecting changes in vehicle population and total vehicular emissionsunder the business-as-usual scenario
Using vehicle emission factors, vehicle registration data andannual kilometers traveled data, Parsons (2001) estimated the 1999mobile source components of PM emissions (Table 7). According tothe Road Transport Statistics of Thailand, the vehicle fleet in the BMAgrew at a rate of 6.2% per year on average over 1983 to 1999 (Parsons,2001). Therefore, the annual growth rate in vehicle registrations inthe BMA is much higher than other more saturated markets. Whenthe types of vehicles are examined for growth, the average annualgrowth rates vary significantly among different subcategories ofvehicles. For example, light-duty trucks are the fastest growingcategory, growing at about 16% per year over the last ten years;motorcycles and passenger cars have increased by approximately 6%per year; whereas the total number of registered buses has remainedvirtually constant (Parsons, 2001). The annual growth rates for eachautomotive category used to estimate the vehicle population in futureyears are summarized in Table 7. It is assumed that the growth rates inthe past will continue into the near future.
The 2000 vehicle PM emission rates in Table 7 are the baselineemission rates in this study. In the business-as-usual (BAU) scenario,
Table 7Vehicle population, annual growth rates and emissions data in the Bangkok Metropolitan Area in 1999 (Adapted from Parsons, 2001).
Vehicle type Number ofvehicles (1999)
Estimated annualgrowth rate of thenumber of vehicles (%)
6 Vehicle registration data source: Beijing Transportation Center, 2005 AnnualTransportation Report for Beijing.
7 There may also be changes in the passenger car fleet. For example, some travelersare likely to switch to passenger cars if other vehicles are regulated more rigorously.However, these effects are not likely to have significant impacts on the total vehiclePM emissions since the emission projection in the BAU scenario shows that passenger
1783Y. Li, D.J. Crawford-Brown / Science of the Total Environment 409 (2011) 1774–1785
although no further policy to control emissions from in-servicevehicles is implemented, the fleet-average PM emissions per vehicle-kilometer should decline over time due to: (1) newer vehicle modelsentering the fleet as older ones are scrapped; (2) tighter emissionstandards for new vehicles are enforced, and emission performance inmodern vehicles is greatly improved. For example, the emissionstandards for new light-duty vehicles in Thailand from 1995 to 2014are: 1995–2000 (Euro 1), 2001–2003 (Euro 2), 2004–2011 (Euro 3)and 2012–2014 (Euro 4).
However, Table 7 is the only available information on fleet-averageemission rates for different types of vehicles in the BMA. It is ratherdifficult to estimate the change in average emission rates in futureyears resulting from the improvement in emission performanceassociated with new vehicles without conducting the same type ofstudies for the years after 1999. Moreover, fleet-average emissionrates for in-use vehicles as well as their changes over time areinsufficiently studied to date. Some researchwas conducted in Beijing,China, for vehicle pollutants including NOx, CO and HC, but not PM,due to the lack of PM emission data in the past (He et al., 2002). Giventhese facts, this study assumes a 5% annual PM emission factordecrease rate for all types of vehicles in the BMA.5
Another issue is that the annual vehicle kilometers travelled (VKT)per vehicle may increase, as people's travel needs increase as theresult of rapid economic growth and urbanization, particularly in adeveloping country, resulting in a potential increase in total emissionsfrom motor vehicles if VKT growth exceeds the decline in emissionsper vehicle-kilometer. However, the increase in VKT per vehicle ismore likely to happen among private vehicles such as private trucks,cars and motorcycles, since a large portion of commercial vehiclessuch as city buses and trucks usually run on fixed routes. Currentlyaccurate information regarding the change in VKT per vehicle overtime is not available in the BMA. When comparing the 1999 VKT inTable 7 with the VKT per vehicle in Bangkok for several vehicle typesin 2005 reported by Perera (2006), based on available data on citybuses, passenger cars and motorcycles, there was no significantincrease in VKT per vehicle found during this six-year period. Giventhis, we assume that the average annual VKT per vehicle for all vehicletypes remain unchanged over time.
The projected total annual PM10 emissions from motor vehicles infuture years under the BAU scenario are estimated using the projectedaverage emission rates, VKT, and projected number of vehicles.
A.2. The impacts of emission control policies on the growth of vehiclesand emissions.
If new inspection and maintenance programs targeting all diesel-fueled vehicles and motorcycles are implemented, vehicle growthrates may diminish due to the regulation. However, data on new
5 The same annual PM emission decrease rate was assumed in a World Bank studyon the reductions in emissions from motorcycles in Bangkok (World Bank, 2003).
vehicle registrations in Bangkok in the past decade indicate that thegrowth of vehicle numbers is more significantly affected by the overalleconomic situation than by environmental regulation. Nevertheless,government policy interventions may still alter the trend in vehiclegrowth. Parsons (2001) reports that during 1990–1991, there was adecline in new vehicle registration in Bangkok, whereas new vehicleregistrations in Thailand increased during the same period. Theyargue that the changes in Bangkok may be due to some policiesimplemented exclusively in the city, such as a change in the licensingfee and local value added tax.
Currently there is little empirical evidence on the impact ofinspection and maintenance programs on vehicle growth in Asiandeveloping countries. Beijing launched I/M programs targeting CO, HCand NOx emissions from in-use vehicles in 1999 (Hao et al., 2006).Vehicle registration data in Beijing show that the average annualvehicle growth rate during 1999–2005 after the implementation ofthe programs was 12.1%, whereas the average rate during 1993–1998before the programs was 18.3%, which means that there was a 34%decrease in the average annual growth rate after the programs.6
However, there were other key vehicle emission control measuresbeing launched in Beijing in 1999, such as the introduction of a morestringent (Euro 1) emission standard for new vehicles, completelyphasing out lead in gasoline, etc (Hao et al., 2006). Therefore, it isdifficult to separate the impact of I/M policies from that of others.
Based on this limited evidence, this study assumes that a 10%decrease, or approximately one-third of the overall 34% decrease, inannual growth rate is attributable to the I/M programs. The assumptionis used to project the change in annual growth rate for all diesel-fueledvehicles and motorcycles after the new PM-related I/M programstargeting these vehicles are introduced. The growth rate for passengercars is assumed to be unchanged since they are excluded from the PM-related I/M programs.7
Parsons (2001) proposed PM10 emission reduction targets foreach type of vehicle to be achieved by the new I/M programs targetingall diesel-fueled vehicles and motorcycles as follows: Bus (20%emissions reduction per VKT), city truck (25%), light duty truck(25%), long haul truck/bus (25%), motorcycle (30%). These emissionreduction targets are adopted in the health benefits estimation here.These emission reduction targets would result in an approximate 25%reduction in the total annual vehicle particulate emissions relative tothe BAU scenario.
cars only contribute to less that 1% of total vehicle PM10 emissions over time. Giventhis, assuming the same the growth rate of passenger cars under both the BAU and theI/M scenarios will not have significant impacts on the projection.
9 The value is for traditional I/M programs targeting emissions of CO, HC and NOx. In
Table 8Social costs of inspection and maintenance programs in the Bangkok Metropolitan Region.
a Parsons, 2001.b CAF, 2002.c Ando et al., 1999.d World Bank, 2003.
1784 Y. Li, D.J. Crawford-Brown / Science of the Total Environment 409 (2011) 1774–1785
A.3. Projected increase in population in the Bangkok Metropolitan Area
Population data are used to estimate the size of population exposedto air pollution. In projecting into the future years, the trends inpopulation change in thepast are used to forecast future population. Theaverage annual growth rates over the 1999–2005 period for each of thesix provinces in the BMA8 are used to predict the population in futureyears based on the population data in the base year 2000.
Moreover, in order to forecast the infant mortality rates due to PMpollution exposure in future years, the number of births needs to bepredicted. Available data indicate that in Bangkok the average birthrate over the period of 2000–2005 increased 0.43% per year (BangkokStatistics, 2005). Due to the lack of data for the surrounding fiveprovinces, the percentage birth rate increase in Bangkok is applied toall the six provinces in the BMA to predict the increase in the numberof total births in future years.
Appendix B. Estimating the costs of I/M programs
Currently information about vehicle repair costs as well aseffectiveness of repair associated with I/M programs is ratherincomplete in the U.S. and elsewhere (National Research Council,2001), in particular for PM-related I/M programs, which are still intheir infancy. A recent study on reducing emissions from motorcyclesin Bangkok (World Bank, 2003) reports 60% of motorcycles that failthe emission test will pass the test after a minor tune-up, and theremaining 40% that fail the initial test will need major repairs,remanufacturing to meet current emissions standards, or scrapping.The study also reports the costs of repairs: the average minor repaircost is 30 baht per vehicle ($0.75, based on the exchange rate of 39.79in 2000); major repair costs vary from 100 baht to 10,000 baht,depending on the motorcycle conditions (on average 2204 baht, or$58.3); and engine remanufacturing costs from 5000 to 7000 baht (onaverage 6000 baht, or $150.8). Since their study was conducted in2000, it is expected that much less engine remanufacturing is neededto satisfy the emission standards in later years. Here only minor andmajor repairs are considered, as it is assumed that engine remanu-facturing has already been accomplished for the large majority ofcandidate vehicles.
Given the lack of repair costs for all other vehicle types in Thailand,relevant information in the U.S. was examined. A study at Colorado
8 Source: Bangkok Metropolitan Administration, Strategy and Evaluation Depart-ment. 2006. Statistical Profile of Bangkok Metropolitan Administration 2005(simplified as Bangkok Statistics, 2005).
State reports that estimated average repair cost is $465 per failingvehicle for heavy-duty diesel vehicles including heavy trucks andbuses (CAF, 2002). Furthermore, Ando et al. (1999) report that themean repair costs for light trucks are about $110 per vehicle.9 In orderto apply the U.S. vehicle repair costs to Thailand, the values need to beadjusted to reflect the fiscal difference between the two countries. Ingeneral, vehicle repair costs include parts and labor costs and it isassumed that each accounts for 50% of the total repair costs. Regardingvehicle parts, the World Bank study reports that a motorcycle sparkplug costs 36 baht or approximately $1 in Thailand, whereas itscounterpart in the U.S. costs about $2 per sparkplug. Based on this,this study assumes that the costs of vehicle parts in Thailand arehalf of the costs in the U.S. Second, in comparing the labor costs, it isfound that the monthly salary of skilled laborers and technicians inThailand falls into the range of $215–$286 (Thailand Board of Invest-ment, 2006), whereas the monthly salary of similar working groupsin the U.S. falls into the range of $2500–$3000 (U.S. Department ofLabor, 2006). Based on this information, this study assumes thelabor costs in Thailand are one tenth of the costs in the U.S. Therefore,a Thailand/U.S. ratio of 0.3 (0.5×0.5+0.1×0.5=0.3) is used toextrapolate the U.S. costs to Thailand.
Empirical evidence in the U.S. indicates that the percentage offailing vehicles ranges from 10% to 25% in traditional I/M programstargeting light-duty vehicles such as passenger cars and light trucks(Parsons, 2001). Given this, an assumption is made that 17.5% (themidpoint of the range) of the light duty trucks are problem vehiclesand thus need repairs. For motorcycles, this value is 25% based on thefindings by World Bank (2003).10 Moreover, heavy diesel I/Mprograms are relatively new. Past experience shows that a typicalfailure rate was 12–15% at the beginning, and declined to 5–6% aftersome years of implementation (Duleep, 2004). This implies that thepercentage of failing heavy diesel vehicles may be lower than that offailing light diesel vehicles. Based on this, this study makes anassumption that 10% of the buses and heavy trucks in the BMA areproblematic and will fail inspection.
Using the data discussed above, this study estimated the totalannual costs of I/M in the BMA to be 147 million 2000 US dollars, assummarized in Table 8.
lack of information about costs to reduce PM emissions from light-duty trucks, thisvalue is adopted in this study.10 The study found that 65% motorcycles in Bangkok were 5 or less years old, and 20%of them failed the emission test, whereas among the remaining 35% motorcycles thatwere more than 5 years old, 35% failed the test. Based on this, it is assumed that 25%motorcycles fail the emission test (0.65×0.2+0.35×0.35=0.25).
y = 1.921x + 86.689
R2 = 0.7306
0
50
100
150
200
250
300
350
0 20 40 60 80 100 120
Daily PM10 Concentrations (ug/m3) at a Permanent Station(Ramkhamhaeng University)
Dai
ly P
M10
Con
cent
rati
ons
(ug/
m3 )
at a
Roa
dsid
e St
atio
n(S
ukhu
mvi
t R
oad)
Fig. 5. Correlation between daily 24-hour average PM10 concentrations at a permanentstation and a roadside station, January 6–21, 2000.
1785Y. Li, D.J. Crawford-Brown / Science of the Total Environment 409 (2011) 1774–1785
Appendix C. Correlation between daily PM10 concentrations atpermanent stations and roadside stations
Fig. 5 shows the correlation between a roadside station and apermanent station on 16 days in January 2000.
Fig. 5 indicates that daily PM10 concentrations at the two stationscorrelated well (R2=0.73).
References
Anderson HR, Atkinson RW, Peacock JL, Marston L, Konstantinou K. Meta-analysis oftime-series studies and panel studies of particulate matter (PM) and ozone (O3):report of a WHO task group. Copenhagen: World Health Organization; 2004.
Ando A, McConnell V, Harrington W. Costs, emissions reductions, and vehicle repair:evidence from Arizona. Discussion Paper 99-23-REV, Washington, DC: Resource forthe Future; 1999.
Bangkok Metropolitan Administration, Strategy and Evaluation Department. 2006.Statistical Profile of Bangkok Metropolitan Administration 2005.
CAF, Clean Air Fleets. Colorado diesel inspection and maintenance programs. Denver,CO; 2002.
Chen C, Chen D, Green C, Wu C. Benefits of expanded use of natural gas for pollutantreduction and health improvement in Shanghai. Sinosphere J 2002;5(2):58–64.
Cifuentes L, Borja-Aburto VH, Gouveia N, Thurston G, Davis DL. Assessing the healthbenefits of urban air pollution reductions associated with climate changemitigation (2000–2020): Santiago, Sao Paulo, Mexico City, and New York City.Environ Health Perspect 2001;109(Suppl 3):419–25.
Deck LB, Post ES, Smith E, Wiener M, Cunningham K, Richmond H. Estimates of thehealth risk reductions associated with attainment of alternative particulate matterstandards in two U.S. cities. Risk Anal 2001;21:821–36.
Dominici F, McDermott A, Daniels M, Zeger SL, Samet JM. Revised analyses of thenational morbidity, mortality, and air pollution study: mortality among residents of90 cities. J Toxicol Environ Health 2005;68(13–14):1071–92 (Part A).
Duleep KG. Heavy-duty diesel I/M: Lessons fromNorth America. Better Air Quality 2004Workshops, Agra, India; 2004.
Freeman III AM. The measurement of environmental and resource values: theory andmethods. Washington, DC: Resources for the Future; 1993.
Haddix AC, Teutsch SM, Shaffer PA, Dunet DO, editors. Prevention effectiveness: a guide todecisionanalysis and economic evaluation.NewYork,NY:OxfordUniversity Press; 1996.
Hagler Bailly, Inc.. Final Report: health effects of particulate matter air pollution inBangkok. A report prepared for Air Quality and Noise Management, PollutionControl Department, Bangkok, Thailand; 1998.
Hao J, Hu J, Fu L. Controlling vehicular emissions in Beijing during the last decade.Transp Res A Policy Pract 2006;40(8):639–51.
He K, Huo H, Zhang Q. Energy and environmental issues for transportation sector ofNorth Asiamega-cities. Proceedings of IGES/APNMega-City Project. Japan: Institutefor Global Environmental Strategies; 2002.
HEI, Health Effects Institute. Traffic-related air pollution: a critical review of theliterature on emissions, exposure, and health effects. Spec Rep 2010;17.
Hubbell B, Hallberg A, McCubbin DR, Post E. Health-related benefits of attaining the 8-hr ozone standard. Environ Health Perspect 2005;113(1):73–82.
IEA, International Energy Agency. CO2 emissions from fuel combustion 2010—highlights; 2010.
Jerrett M, Burnett RT, Ma R, Pope III CA, Krewski D, Newbold KB, Thurston G, Shi Y,Finkelstein N, Calle EE, ThunMJ. Spatial analysis of air pollution andmortality in LosAngeles. Epidemiology 2005;16:727–36.
Künzli N, Medina S, Kaiser R, Quénel P, Horak Jr F, Studnicka M. Assessment of deathsattributable to air pollution: should we use risk estimates based on time series oron cohort studies? Am J Epidemiol 2001;153(11):1050–5.
Laden F, Neas LM, Dockery DW, Schwartz J. Association of fine particulate matter fromdifferent sources with daily mortality in six U.S. cities. Environ Health Perspect2000;108(10):941–7.
Levy JI, Spengler JD. Modeling the benefits of power plant emission controls inMassachusetts. J Air Waste Manage Assoc 2002;52:5-18.
Li Y. Health benefits of traffic-related particulate matter control policies: the case ofBangkok, Thailand. Ph.D. Dissertation. University of North Carolina at Chapel Hill,Chapel Hill, NC. 219 pages; 2008.
Loomis D, Castillejos M, Gold DR, McDonnell W, Borja-Aburto VH. Air pollution andinfant mortality in Mexico City. Epidemiology 1999;10(2):118–23.
Metzger KB, Tolbert PE, Klein M, Peel JL, Flanders WD, Todd K, Mulholland JA, Ryan PB,Frumkin H. Ambient air pollution and cardiovascular emergency department visits.Epidemiology 2004;15(1):46–56.
National Research Council. Evaluating vehicle emissions inspection and maintenanceprograms. Washington, DC: National Academy Press; 2001.
National Toxicology Program. Diesel exhaust particulates. Report on carcinogens 11thEdition. 2005.
National Statistical Office Thailand. 2000. The 2000 Population and Housing Census.Oanh NTK, Zhang B. Photochemical smog modeling for assessment of potential impacts
of different management strategies on air quality of the Bangkok MetropolitanRegion, Thailand. J Air Waste Manage Assoc 2004;54(10):1321–38.
Ostro BD. Air pollution and morbidity revisited: a specification test. J Environ EconManage 1987;14:87–98.
Ostro BD. Outdoor air pollution: assessing the environmental burden of disease atnational and local levels. World Health Organization Environmental Burden ofDisease Series, No.5; 2004.
Ostro BD, Chestnut L. Assessing the health benefits of reducing particulate matter airpollution in the United States. Environ Res 1998;76:94-106 (Section A).
Ostro BD, Chestnut L, Vichit-Badakan N, Laixuthai A. The impact of particulate matter ondaily mortality in Bangkok, Thailand. J Air Waste Manage Assoc 1999;49(9):100–7.
Ostro BD, Tran T, Levy JI. The health benefits of reduced tropospheric ozone inCalifornia. J Air Waste Manage 2006;56(7):1007–21.
Parsons International Ltd.. Final report for the Bangkok Air Quality ManagementProject. A report prepared for Bangkok Metropolitan Administration, Thailand;2001.
Perera R. Promoting travel demand reduction in transport sector in cities of Asiandeveloping countries: case of Bangkok. Workshop paper, Japan: Institute for GlobalEnvironmental Strategies; 2006.
Puangthongthub S. Bayesian Maximum Entropy space/time analysis of ambientparticulate matter and mortality in Thailand. Ph.D. Dissertation, University ofNorth Carolina at Chapel Hill, Chapel Hill, NC, 164 pages; 2006.
Rabl A. Interpretation of air pollution mortality: number of deaths or years of life lost?J Air Waste Manage Assoc 2003;53:41–50.
Sanhueza P, Reed GD, Davis WT, Miller TL. An environmental decision-making tool forevaluating ground-level ozone-related health effects. J Air Waste Manage Assoc2003;53:1448–59.
Stieb DM, Judek S, Burnett RT. Meta-analysis of time-series studies of air pollution andmortality: effects of gases and particles and the influence of cause of death, age, andseason. J Air Waste Manage Assoc 2002;52:470–84.
Thailand Board of Investment. Labor costs: median monthly salaries for selectedpositions, survey date: 2006.
Thanh BD, Lefevre T. Assessing health benefits of controlling air pollution from powergeneration: the case of a lignite-fired power plant in Thailand. Environ Manage2001;27(2):303–17.
UNEP, United Nations Environment Program. Bangkok State of Environment, 2003.Bangkok, Thailand; 2004.
United State Department of Labor. 2006 national occupational employment and wageestimates.
USEPA, United States Environmental Protection Agency. The benefits and costs of theclean air act, 1970 to 1990: main report. EPA-410-R-97-002, United StatesEnvironmental Protection Agency Office of Air and Radiation Policy; 1997.
USEPA, United States Environmental Protection Agency. The benefits and costs of theClean Air Act, 1990–2010. EPA-410-R-99-001, United States EnvironmentalProtection Agency Office of Air and Radiation Policy; 1999.
VajanapoomN, Shy CM, Neas LM, Loomis D. Associations of particulate matter and dailymortality in Bangkok, Thailand. Southeast Asian J Trop Med Public Health 2002;33(2):389–99.
Wang X, Mauzerall DL. Evaluating impacts of air pollution in China on public health:implications for future air pollution and energy policies. Atmos Environ 2006;40:1706–21.
Woodruff TJ, Grillo J, Schoendorf KC. The relationship between selected causes ofpostneonatal infant mortality and particulate air pollution in the United States.Environ Health Perspect 1997;105:608–12.
World Bank. Thailand environment monitor; 2002.World Bank. Thailand: reducing emissions from motorcycles in Bangkok. Report
275/03, World Bank Energy Sector Management Assistance Programme; 2003.World Bank. Thailand: making transport more energy efficient, World Bank and
National Economic and Social Development Board report; 2009.Xu X, Li B, Huang H. Air pollution and unscheduled hospital outpatient and emergency
room visits. Environ Health Perspect 1995;103(3):286–9.
