Laws on pollution in Toronto – failing? Pollution levels haven’t changed despite efforts as indicated by this University of Toronto study.
Looking over the last decade, there has been no overall reduction in smog in the GTA, despite best efforts to control some of the contributing factors,
However, claims of health impacts due to pollution in Toronto and other Canadian cities are up according to some other studies. Ross McKitrick says in a new peer reviewed study that the models and claims don’t add up.
Study Questions Link Between Air Pollution, Serious Health Effects
University of Guelph News Release
Challenging accepted wisdom, a University of Guelph professor says claims about the health effects of air pollution are not supported by data from Canadian cities.
Guelph economist Ross McKitrick, along with Gary Koop of the University of Strathclyde in Glasgow and Lise Tole of the University of Edinburgh, analyzed a new database from 11 Canadian cities over a 20-year period. Unlike most earlier studies, this one included controls for effects of smoking and income.
They found no evidence that air pollution affected either hospital admission rates or time spent in hospitals. However, they did determine that both smoking and income levels directly affect respiratory health. Their findings appear this week in the journal Environmental Modelling and Software.
The researchers compared monthly hospital admission rates between 1974 and 1994 for all lung ailments to ambient levels of five common air contaminants. “We were looking for predictable, common physical effects from standardized exposure levels,” McKitrick said, adding the researchers examined data over a longer time span than most previous studies, and used advanced econometric methods called Bayesian Model Averaging to ensure they considered all possible combinations of effects.
“Our examination of data back to the early 1970s was motivated in part by the fact that air pollution was much higher compared to today,” he said. “If today’s air pollution levels are causing thousands of hospitalizations, the effects should have been even stronger in the 1970s when air quality was much worse.”
“But the data showed no evidence of changing health effects at the pollution levels observed in Canada over recent decades.”
The findings contradict hundreds of studies that have connected urban air pollution levels and respiratory health problems. Such studies have resulted in calls for tighter air pollution regulations and more stringent emission standards.
McKitrick said the discrepancies between this study and earlier research stem from the common practice of not examining long enough data sets and not controlling for model uncertainty, smoking rates and socioeconomic variables. He added that their study drew data samples from the 1970s, when many Canadian cities had high pollution levels, through the 1980s, when steady reductions began, and into the 1990s, when pollution levels were historically low.
“It’s important to get accurate measures of the potential benefits of air pollution regulations, namely improved quality of life and reduced health-care costs, in order to guide regulatory decision-making,” McKitrick said.
“We did find consistent evidence that lower smoking rates lead to fewer hospital admissions and shorter stays,” he said. The researchers also found evidence that, all else being equal, regions with larger economies tend to have higher hospital admission rates. This may indicate more hospitals and longer patient treatment regimens, McKitrick said.
See the paper and supporting data here in Dr. McKitricks web page.
Models that predict thousands of smog-related hospitalizations in Toronto don’t hold up
By Ross McKitrick
For many years we have heard that air pollution in Canada is responsible for thousands of annual deaths and hospitalizations. In 2004 Toronto Public Health claimed that 1,700 premature deaths and 6,000 hospitalizations occur each year in Toronto alone, due to air pollution. The Ontario Medical Association, provincial and federal governments, lung associations and other groups regularly cite these kinds of figures in support of calls for new regulatory initiatives. These death and hospitalization rates are astonishing. It is like suffering a 9/11-sized terrorist attack every 10 months.
But is it really true? The estimates are derived by taking correlations in the epidemiological literature between observed pollution levels and health indicators, like hospital admission rates, and then extrapolating across populations to estimate how many deaths and illness diagnoses can, in theory, be attributed to pollution. In other words, the numbers come from statistical models, not from direct observations. That means we need to pay close attention to how the statistical modeling is done.
Together with my coauthors Gary Koop of Strathclyde University and Lise Tole of the University of Edinburgh, I have just published a peer-reviewed study in the journal Environmental Modelling and Software that does just that. What we found gives us reason to believe that the kind of statistical modeling behind common claims about air pollution may need a careful second look.
There are hundreds of studies in the epidemiological literature that have reported correlations between air pollution and health measures. But there are some common weaknesses to this literature. First, the results are not consistent across studies. Some studies find particulate matter (PM) affects health, but not sulphur dioxide (SO2) or carbon monoxide (CO). Others reported SO2 has an effect, but not PM. Another reports CO has an effect but not ozone (O3), while another finds O3 matters in some cities but not others. One large U.S. study found PM increased mortality risk a little bit across the U.S., except in 20 out of 88 cities in which it actually reduced mortality risk. These kinds of inconsistencies should not occur if the health effect is based on a real physiological response. This is a second puzzling aspect of the literature: Despite decades of testing, clinical investigations have not found experimental support for the idea that current ambient air pollution levels cause lung disease or mortality.
We found, not surprisingly, that smoking is bad for lung health. We found that regions with higher Gross Domestic Product (GDP) tend to have higher hospital admission rates, depending on the model specification, which may indicate that those regions have more hospital services. And we found evidence that hot days with high air pressure tend to produce more hospital admissions.
What we did not find was any evidence that increases in air pollution levels are associated with increased rates of hospital admissions. We looked at the data every which way imaginable. If we were to cherry pick, by looking only at a sub-sample of the time or by picking just the right form of the model, we could find evidence that CO or nitrogen dioxide (NO2) have positive effects on lung disease, but those results do not get strong support in the data. The models that get consistent support either show no pollution effects or — paradoxically — negative effects. In other words, in some cases as air pollution rises, hospital admissions go down. As odd as that sounds, we are by no means the first to report negative coefficients in the literature. Nobody is trying to argue that air pollution is good for you: this is either just noise in the data, or it might be an effect from “averting” behaviour, where people who are susceptible to lung problems stay indoors on days with bad air quality.
Based on our analysis, we could estimate what the effect on hospital admissions would be if all the pollution currently observed in Toronto air were to disappear. Toronto Public Health claims about 6,000 fewer hospitalizations would occur. But this claim gets no support in the data. We found that there would be no reduction in lung-related hospitalizations. If anything there might be somewhere between 20 and 200 more admissions, if we apply the statistical results in a mechanical fashion.
Very few studies over the past decade have controlled for socioeconomic covariates (including smoking), fewer still have looked at long data panels back to the 1970s and fewer still have dealt with model uncertainty. Those that have addressed one or more of these issues typically find the effect of air pollution shrinks or disappears outright. Thus our results are actually quite consistent with the relevant group of previous studies. The popular idea that current ambient air pollution has a powerful effect on lung health might look like it is based on a large empirical foundation, but on closer inspection the pile contains a lot of weak results.
So the bottom line is that, for the purpose of assessing the link between air pollution levels and hospital admissions, one needs to look closely at the kinds of studies being done and how they did the statistical modeling. More studies need to be done using long time series that go back to the 1970s or earlier, more studies need to control for socioeconomic covariates and more studies need to take account of model uncertainty. Based on evidence to date, as these things begin to happen, we should not be surprised if current estimates of the health effects of air pollution turn out to be in need of major revision.
Ross McKitrick is a professor of economics at the University of Guelph.