An investigation of thresholds in air pollution-mortality effects

Gary Koop, Lise Tole

Research output: Contribution to journalArticle

17 Citations (Scopus)

Abstract

In this paper we introduce and implement new techniques to investigate threshold effects in air pollution-mortality relationships. Our key interest is in measuring the dose-response relationship above and below a given threshold level where we allow for a large number of potential explanatory variables to trigger the threshold effect. This is in contrast to existing approaches that usually focus on a single threshold trigger. We allow for a myriad of threshold effects within a Bayesian statistical framework that accounts for model uncertainty (i.e. uncertainty about which threshold trigger and explanatory variables are appropriate). We apply these techniques in an empirical exercise using daily data from Toronto for 1992-1997. We investigate the existence and nature of threshold effects in the relationship between mortality and ozone (O3), total particulate matter (PM) and an index of other conventionally occurring air pollutants. In general, we find the effects of the pollutants we consider on mortality to be statistically indistinguishable from zero with no evidence of thresholds. The one exception is ozone, for which results present an ambiguous picture. Ozone has no significant effect on mortality when we exclude threshold effects from the analysis. Allowing for thresholds we find a positive and significant effect for this pollutant when the threshold trigger is the average change in ozone two days ago. However, this significant effect is not observed after controlling for PM.
LanguageEnglish
Pages1662-1673
Number of pages11
JournalEnvironmental Modelling and Software
Volume21
Issue number12
DOIs
Publication statusPublished - May 2004

Fingerprint

Air pollution
Ozone
atmospheric pollution
mortality
ozone
particulate matter
effect
Air
pollutant
dose-response relationship
total ozone
Uncertainty

Keywords

  • threshold air pollution–mortality effects
  • Bayesian model averaging
  • air pollution
  • econometrics
  • environmental economics

Cite this

@article{def5e296c1df4a14a21bc3f5eb4b220d,
title = "An investigation of thresholds in air pollution-mortality effects",
abstract = "In this paper we introduce and implement new techniques to investigate threshold effects in air pollution-mortality relationships. Our key interest is in measuring the dose-response relationship above and below a given threshold level where we allow for a large number of potential explanatory variables to trigger the threshold effect. This is in contrast to existing approaches that usually focus on a single threshold trigger. We allow for a myriad of threshold effects within a Bayesian statistical framework that accounts for model uncertainty (i.e. uncertainty about which threshold trigger and explanatory variables are appropriate). We apply these techniques in an empirical exercise using daily data from Toronto for 1992-1997. We investigate the existence and nature of threshold effects in the relationship between mortality and ozone (O3), total particulate matter (PM) and an index of other conventionally occurring air pollutants. In general, we find the effects of the pollutants we consider on mortality to be statistically indistinguishable from zero with no evidence of thresholds. The one exception is ozone, for which results present an ambiguous picture. Ozone has no significant effect on mortality when we exclude threshold effects from the analysis. Allowing for thresholds we find a positive and significant effect for this pollutant when the threshold trigger is the average change in ozone two days ago. However, this significant effect is not observed after controlling for PM.",
keywords = "threshold air pollution–mortality effects, Bayesian model averaging, air pollution, econometrics, environmental economics",
author = "Gary Koop and Lise Tole",
note = "Working paper version",
year = "2004",
month = "5",
doi = "10.1016/j.envsoft.2005.07.012",
language = "English",
volume = "21",
pages = "1662--1673",
journal = "Environmental Modelling and Software",
issn = "1364-8152",
number = "12",

}

An investigation of thresholds in air pollution-mortality effects. / Koop, Gary; Tole, Lise.

In: Environmental Modelling and Software, Vol. 21, No. 12, 05.2004, p. 1662-1673.

Research output: Contribution to journalArticle

TY - JOUR

T1 - An investigation of thresholds in air pollution-mortality effects

AU - Koop, Gary

AU - Tole, Lise

N1 - Working paper version

PY - 2004/5

Y1 - 2004/5

N2 - In this paper we introduce and implement new techniques to investigate threshold effects in air pollution-mortality relationships. Our key interest is in measuring the dose-response relationship above and below a given threshold level where we allow for a large number of potential explanatory variables to trigger the threshold effect. This is in contrast to existing approaches that usually focus on a single threshold trigger. We allow for a myriad of threshold effects within a Bayesian statistical framework that accounts for model uncertainty (i.e. uncertainty about which threshold trigger and explanatory variables are appropriate). We apply these techniques in an empirical exercise using daily data from Toronto for 1992-1997. We investigate the existence and nature of threshold effects in the relationship between mortality and ozone (O3), total particulate matter (PM) and an index of other conventionally occurring air pollutants. In general, we find the effects of the pollutants we consider on mortality to be statistically indistinguishable from zero with no evidence of thresholds. The one exception is ozone, for which results present an ambiguous picture. Ozone has no significant effect on mortality when we exclude threshold effects from the analysis. Allowing for thresholds we find a positive and significant effect for this pollutant when the threshold trigger is the average change in ozone two days ago. However, this significant effect is not observed after controlling for PM.

AB - In this paper we introduce and implement new techniques to investigate threshold effects in air pollution-mortality relationships. Our key interest is in measuring the dose-response relationship above and below a given threshold level where we allow for a large number of potential explanatory variables to trigger the threshold effect. This is in contrast to existing approaches that usually focus on a single threshold trigger. We allow for a myriad of threshold effects within a Bayesian statistical framework that accounts for model uncertainty (i.e. uncertainty about which threshold trigger and explanatory variables are appropriate). We apply these techniques in an empirical exercise using daily data from Toronto for 1992-1997. We investigate the existence and nature of threshold effects in the relationship between mortality and ozone (O3), total particulate matter (PM) and an index of other conventionally occurring air pollutants. In general, we find the effects of the pollutants we consider on mortality to be statistically indistinguishable from zero with no evidence of thresholds. The one exception is ozone, for which results present an ambiguous picture. Ozone has no significant effect on mortality when we exclude threshold effects from the analysis. Allowing for thresholds we find a positive and significant effect for this pollutant when the threshold trigger is the average change in ozone two days ago. However, this significant effect is not observed after controlling for PM.

KW - threshold air pollution–mortality effects

KW - Bayesian model averaging

KW - air pollution

KW - econometrics

KW - environmental economics

U2 - 10.1016/j.envsoft.2005.07.012

DO - 10.1016/j.envsoft.2005.07.012

M3 - Article

VL - 21

SP - 1662

EP - 1673

JO - Environmental Modelling and Software

T2 - Environmental Modelling and Software

JF - Environmental Modelling and Software

SN - 1364-8152

IS - 12

ER -