The Cost of Air Pollution: A Meta-Analysis of Air Quality Valuations Jennifer Chirico, MPH Douglas Noonan, PhD APHA Conference November 5, 2007

Outline Introduction and Background Data Analysis Findings Conclusion

Introduction and Background Air Quality Poor air quality has significant health, environmental, and economic impacts Air quality is a public good (“non-market good”) and difficult to value Significance to public policy How air quality is valued in society is a significant concern to policy makers as they make decisions about the appropriate means that should be allocated to air quality improvement

Introduction and Background Meta-Analysis (“study of studies”) The purpose of a meta-analysis is to measure relationships between reported environmental valuation estimates for goods and non-market goods Types of air quality valuation studies included in this meta-analysis: Revealed preference e.g., Hedonics: How much do people reveal their air quality preferences; may pay more to live in an area with better air quality Stated preference e.g., Contingent Valuation: Surveys that ask individuals/households “How much would you be willing-to-pay for a 50% reduction in air pollution?” Revealed Preference Theory Uses behavior of individuals and communities to analyze their spending and income preferences Looks for correlations between market behavior and non-market goods Includes hedonics – property and rent models break down housing components to see how much of where people live is attributed to air quality. You may pay more to live somewhere with better air quality/visibility Also includes travel costs (demand curve to visit something, costs involved) Averting costs – least common; e.g. spend money on air filters to avoid poor air Stated preference Ask people that preference through surveys Using used in CVM. What would you be WTP for a change in AQ? Choice experiment – which bundle would you pick? Rank order

Introduction and Background Several Air Quality Meta-Analysis Studies have been conducted since 1995 Vassanadrumrongdee et al. (2004). “Meta-analysis of contingent valuation studies on air pollution-related risks.” Focus is on contingent valuation method (CVM) studies and the health effects of air pollution Delucchi et al. (2002). The Health and Visibility Cost of Air Pollution: A Comparison of Estimation Methods. Focus in on health and visibility costs using hedonics and CVM studies Smith, V.K. & Huang, J. (1995). “Can Markets Value Air Quality? A Meta-Analysis of Hedonic Property Value Models.” Focus in on Hedonic Valuation studies There are not many meta-analysis on air quality These are 3 most recent air quality Our study focuses on studies from 1978 to today and looks at both RP and SP and compares studies around the world.

Purpose and Research Question How much are people willing-to-pay for air quality improvements? Purpose/Objectives: To provide a statistical summary of air quality valuations conducted around the world To examine stated and revealed preference studies that measured air quality against health costs and property values To provide insights into the costs to society of air pollution and how much people are willing-to-pay for air quality improvements

Data 104 studies found using Georgia Tech library databases (including UGA and GSU), Environmental Valuation Research Inventory (EVRI), Research Papers in Economics (RePEC),and Google scholar 19 were dropped due to missing information 34 were not in print or had a cost associated 1 was dropped due to language barrier (not available in English) 4 were meta-analyses 46 articles were examined The sample size is defined as the number of willingness-to-pay observations (n=309) 104 studies found through searches Some of the ones that are difficut to find may be lower quality papers that are not readily availabe GT library databases Economic databases Google scholar For the 19 with missing info, we will go back and try to find that info to eventually include in analysis Sample size = wtp observations

Data Variable Definition Air Quality Mean air quality level (e.g., PM10, Ozone, TSP) Change in air quality, % change in air quality measure (e.g., 50% reduction in ozone) Population Population of study area Income Sample income per household Country Study site Year Year published Year Conducted Year the survey was conduction/data gathered Peer-reviewed Was the article peer-reviewed? type of valuation, Hedonics, CVM, 2-stage hedonic, compensating differential, benefit transfer Funder Who funded the study (G=gov’t, N=nonprofit, A=school, W=World Bank, O=other) Good The good the author(s) are valuing (e.g., improving air quality by 50%) Sample Size, Size of the author(s) study sample elasticity Income elasticity Model Type of model used (e.g., R=regression, L=logit, P=probit) demographic groups # of demographic variables in study survey Type of survey (e.g. phone, mail, door-to-door third world Was the study conducted in a third world country?

Data Number of studies (46) Revealed preference studies 30 articles Stated preference studies 14 articles Combined stated and revealed preference studies 2 articles Number of observations (n=309) Revealed preference observations 190 observations Stated preference observations 119 observations

Analysis CAA Amendments 1977 CAA Amendments 1990 CAA 1970 Number of studies per year between 1971 and 2007 As illustrated, there are three big humps, or punctuaated equilibriums in 1981, 1998, 2006. These are when they were published, which is usually several years after the study has been conducted, so we can assume it is more like 1978, 1995, and 2004

Researchers and Funders Distribution of Researchers 28 studies conducted by academic researcher(s) 3 studies conducted by NGO researcher(s) No studies conducted by government researcher(s) 11 combined researcher studies 4 studies did not indicate researcher affiliation Funders of Studies Government funded 20 NGOs funded 6 University 1 Combined 4 Unknown 15 Number of studies by funders Clearly govert agencies are the biggest funders of these studies Followed by unknown. If they didn’t specify who funded it, we left it as unknown, but most likely it was academic work since they were the biggest section of researchers. And they probably would have stated it if it was funded by someone else.

25 countries represented, including Studies by Country This is per country Obviously the US produces the most studies China, Mexico and Thaiwan are about the same. Then Korea. The rest is probably 1-2 observations per country 25 countries represented, including 10 developing countries

Stated and Revealed Preference Studies Varied by Country

Types of Air Measures There are six criteria air pollutants monitored by the epa. Lead No2 So2 Ozone Pm10 Pm2.5 PM2.5 was not measured in any study, probably because it is the most recent to be monitored. Just started monitoring in late 1990s TSP – total suspended particles use to be the measure for particulates and this measure was used for most of the older studies 1970s and 80s

Additional Air Measures There were some additional observations for non-criteria air pollutants Visibility Black fallout days Cooling degree days Heating degree days Days with unhealthy air Odor Noise These were some additional observations that were kind of interesting outside of the basic criteria pollutants. There were several for visibility, I think 2 studies looked at smell Some others were heat, humidity, rain, days of unhealthy air

Change in Air Quality Changes in air quality How much willing to pay for a 50% reduction in one of the air quality measures mentioned. Wtp 100 for a 50% reduction in ozone, smell

Sample Size of Studies Reviewed Per sample size Most fell between 1000-2000 But many in the category of just under 2000 Or above 5000

Local, Regional, or National? Local is highest for both contingent and hedonics, followed by regional and national Benefit transfer was mostly at national level

Survey Modes Surprisingly most were conducted door-to-door May be because they were older studies. More recent one were conducted by mail. Next mail

Final WTP measures were converted to annual estimates per household (2 Final WTP measures were converted to annual estimates per household (2.5) based on a 1% change in air quality Revealed Preference Stated Preference Total WTP Mean $20.99 $48.89 $37.98 Median $4.73 $10.13 $7.67 Max $141.31 $364 $364.31 Min $.06 $.00 SD $30.12 $79.50 $65.32

Average WTP Across Countries

Conclusion Stated preference studies have a higher WTP than revealed preference studies The total mean WTP based on evaluations around the world ranges from $0-$364 This wide range of results demonstrates the difficulty in valuing air quality These varying results make it hard for public policy makers to determine the true value society places on air quality Transferring benefit estimates to other locations can yield confidence intervals based on the location’s unique attributes Total mean 79.86 RP= 100.95 SP= 46.82 Devoped=94.21 Developing 47.18 US 69.86 US falls in between We need to look further into why there is such a difference between RP and SP and US and Developed countries

Questions???