with an 8% increase in the Covid-19 death rate.
While the above studies provide useful preliminary evidence, Conticini et al. (2020) and Ogen (2020) offer only geographical correlations between Covid-19 cases andpollutionexposure, whereas Travaglio et al. (2020) take a similar approach but control only for differ- ences in population density and do so across only 7 relatively large regions. Establishing a convincing link between exposure to pollutionandCovid-19 cases requires individual-level data with the ability to control for individual characteristics, such as age and the presence of underlying health conditions. Since individual-level data on Covid-19 infections is not available, the next best alternative is to examine a large number of small geographic re- gions with detailed data on the characteristics of those regions. This allows the researcher to assess whether any correlation between Covid-19andpollutionexposure still holds once differences in social deprivation, population density, ethnic composition, and other factors are controlled for. While Wu et al. (2020) come closest to doing this, US counties are still relatively large, raising the question of how well such aggregated data can capture the local
Appendix B: Scientific Background and Potential Mechanisms
Broadly speaking, according to the existing medical literature, airpollution may affect cognition through physiological and psychological pathways.
A few of these physiological pathways have been documented in the literature (Block and Calderón-Garcidueñas 2009). First, multiple pollutants (or toxic compounds bonded to the pollutants) may directly affect brain chemistry. For example, ozone in the air can react with body molecules to create toxins, causing asthma and respiratory problems (Sanders 2012). 20 Particulate matter (PM), especially fine particles, can carry toxins through small passageways and directly enter into the brain. There is evidence that suggests that exposure to high PM concentrations may compromise cognitive performance even for people working indoors (Braniš, Řezáčová, and Domasová 2005). 21
production (meters of defective fabric), allowing us to examine the effect of airpollution on the product of labor along this additional margin.
Our main finding is that higher mass concentrations of PM2.5 in outdoor air, as recorded at the air monitoring site situated 2 km from the plant, have a significant adverse impact on contemporaneously observed worker productivity. While the sign of the estimated impacts is consistent with what has been found in the small extant lit- erature, the range of variation is larger than that, to the best of our knowledge, ever examined. Our main result is best described in Figure 4 below, where the scatter depicts data—mean output per attendant worker for every date by (8-hour) work shift in the sample—and the lines indicate alternative non-linear fits. 6 Starting at the sample mini- mum of 10 µg/m 3 (8-hour means), every additional 10 µg/m 3 of exposure to PM2.5 over a worker’s shift reduces her total output by 4.3 meters of fabric, highly significant both statistically and economically, equivalent to about 0.9% of mean output in the sample (509 m/worker-shift). Estimated marginal effects are similar up to the 75th percentile of the PM2.5 distribution, at 149 µg/m 3 , and halve thereafter (-2.0m of fabric per 10 µg/m 3 increase). Beyond the 90th percentile, at 230 µg/m 3 , the estimated marginal effect is a low and only marginally significant -0.5m per 10 µg/m 3 increase.
information on all residential addresses between the date of conception and that of delivery,
Warren et al. ( 2017 ) show that ignorance of residential mobility during pregnancy does not lead to exposure misclassication. Therefore, mobility should not substantially aect our results. A further indication that residential mobility is likely to be of limited concern in our setup derives from the Italian census data, which points to generally high percentages of owned dwellings, ranging from 61.9% in the region of Campania to 78.8% in the region of Molise, registered in 2001 ( ISTAT , 2001 ). Hence, we expect relatively low mobility among resident families. Taken together, underestimation of the true eects of pollution on health at birth due to residential misclassication does not seem highly relevant in our case. Second, we include in our sample also mothers with region of hospital dierent from region of residence, which might introduce an attenuation bias due to a potential measurement error in the pollution assignment. From the initial total births population of ca. 3,400,000 mothers, only 162,244 of them report region of residence dierent from that of their hospital of delivery (less than 5%). We cannot reduce the mismatch at the provincial or municipal level because mothers might choose to deliver in a hospital located in a dierent province in the same region of residence or might be forced to move to the closest municipality with a hospital if their municipality of residence lacks one. In fact, out of almost 8,100 municipalities in Italy, less than 800 have a hospital with a maternity ward. To check to what extent our estimates are sensitive to the inclusion of mothers declaring region of residence dierent from region of hospital in our sample, we run the estimates by excluding the associated observations. As expected, it turns out that this variation in the sample composition yields slightly larger estimates. 21
Xi Chen thanks the Yale Macmillan Center Faculty Research Fund, the U.S. PEPPER Center Scholar Award (P30AG021342), and two NIH/National Institute on Aging Grants (R03 AG048920 and K01AG053408). Xin Zhang acknowledges financial support from the China Postdoctoral Science Foundation Grants (2017M620653 and 2018T110057), and the Fundamental Research Funds for the Central Universities. We are grateful to Katrin Rehdanz and Heinz Welsch for thoughtful feedback on early drafts, and to seminar participants at Yale Department of Economics and NBER Summer Institute 2018 for helpful comments. We appreciate the Institute of Social Science Survey at Peking University for providing us with the CFPS data.
IZA World of Labor | August 2017 | wol.iza.org IZA World of Labor | August 2017 | wol.iza.org 3
SEFI ROTH | Airpollution, educational achievements, and human capital formation
effect of airpollution on educational outcomes. For example, airpollution is often correlated with other factors (such as wealth), which are also likely to be correlated with school or student characteristics, generating a potential omitted variable bias problem. Therefore, if wealthier families are sorting themselves into residential locations with lower levels of pollution, a naive ordinary least squares (OLS) analysis may overestimate the true effect of airpollution due to the positive association between wealth and education. To address this and other related statistical concerns, applied econometricians tend to use quasi-experimental designs, which mainly focus on within-school or student variation. This lends further support to the causal interpretation of their analysis, as it relies on comparing students within the same school, or even comparing the same student over time. More generally, this type of analysis controls for all observed and unobserved potential confounders that remain constant over time. Another form of quasi-experimental design that applied researchers often use is a policy change that induces variation in pollutionexposure only for a subsample of the population. For example, China’s Huai River policy, which provided free winter heating via the provision of coal for boilers in cities north of the Huai River but did not provide it to cities south of the river, led to increased levels of airpollution in the north, as combustion of coal in boilers is associated with the release of air pollutants . These exogenous events enable the researcher to construct control and treatment groups that are statistically identical apart from their pollutionexposure. This form of study design can also be seen as a social analogue to a randomized experiment, which would clearly be unethical in this context. Nevertheless, quasi-experimental designs also have their limitations. First, the results do not explicitly explain the underlying mechanism. Second, there could still be time- variant omitted variables that some of these models fail to capture. Third, the challenge of assigning pollutionexposure to individuals may affect the reliability of the estimates. More specifically, the dilemma here is whether to assign pollutionexposure to an individual’s home location, school location, or some combination of the two. While this issue is less profound in younger populations (since they tend to go to schools in close proximity to their homes), it is still likely to yield some degree of measurement error due to the significant spatial variation in pollution, even within finely defined areas.
BP is regulated as follows: (i) sensors estimate the pressure and transfer a signal to an evalua- tor; (ii) evaluators translate the coded signal from the sensors, compare the BP with a set point (a value desirable under current conditions) and trigger, if needed, the compensatory mecha- nisms; and (iii) according to the signal from the evaluators, the effector mechanism can change the heart rate, the cardiac output or the total peripheral vascular resistance to stabilize the BP (Ackermann 2004). BP regulation can produce acute or long-term changes in pressure. The short-term responses are mainly produced by baroreceptors (the non-encapsulated nerve endings located in the arterial wall of the carotid sinus and the aortic arch) and by the stretch sensors in the cardiac atria (Ackermann 2004). The signal is conveyed to the midbrain (neu- rons in the nucleus tractus solitarius) and can trigger (i) the increased production of angioten- sin II, which increases sodium and water retention in the kidneys, thereby increasing the blood volume; (ii) cardiac parasympathetic outflow, which results in lower heart rate, and, correspondingly, lower cardiac output and BP; and (iii) sympathetic nervous outflow, with the release of adrenalin, renin and noradrenaline, activating α- and β-adrenoreceptors in the cell membranes of the target tissues, resulting in increases in heart rate, cardiac contractility and vascular resistance – all of which can lead to a raise in BP (Ackermann 2004).
test scores is larger than that of contemporaneous exposure. As shown in the last row of Panel A and Panel B, an increase in the mean API on the interview date by one standard deviation (SD) lowers verbal test scores by 0.131 point (0.012 SD), while a one SD increase in average API over three years prior to the interview is associated with an up to as 1.139 points (0.109 SD) drop in verbal test scores. Third, airpollutionexposure appears to have a more negative effect on verbal test performance than math test performance. It is evident that the changes in SDs in the parentheses presented at the bottom of Panel A for verbal test scores are more prominent than the corresponding ones in Panel B for math test scores. Given that the cognitive tests we used might be easier and less challenging than the college entrance exams, our identified contemporaneous effects are a little smaller than those obtained in Ebenstein, Lavy, and Roth (2016). For example, Ebenstein, Lavy, and Roth (2016) find that a one SD reduction in airpollution leads to an increase in Bagrut scores by 0.038 SD. However, our estimated cumulative effects are larger than their contemporaneous effects.
was found to be 60% for France from 1 to 20 August, 2003 ); while 655 (6%) excess deaths were estimated for Cali- fornia from 15 July to 1 August, 2006 . However, not only heat waves but also increases in moderate tempera- ture contribute to the observed heat-related mortality. Exposure-response functions between mortality time-ser- ies and continuous temperature measures have shown V-, U- or J-shaped associations, and the range of temperature corresponding with a minimum mortality (“threshold”, “turning point” or “optimum temperature”) was reported to be related with latitude [6,7]. The residents of lower latitudes tended to be more vulnerable only at higher tem- perature values, indicating less susceptibility to heat [8-10]. Excess winter mortality has been well known (which may have also caused that in recent years, particularly in the light of global warming, there were fewer studies particularly focusing on cold spells or temperature decreases). The Eurowinter study  found that annual excess deaths due to cold ranged from 408 to 1,617 for eight European regions on days colder than 18°C. Barnett et al.  compared coronary events occurring in the coldest 25% of periods with those occurring in the rest of periods among the WHO MONICA project population and found an overall increase. In a recent large multi-centre European study (PHEWE, 15 cities), Analitis et al.  found that a 1°C decrease in 16-day- average minimum apparent temperature was associated with 1.25%-3.30% increases of total or cause-specific mortalities.
In recent years, airpollutionexposure has been also linked to cancer development, includ- ing lung cancer [Soberanes et al., 2012, Zhao et al., 2013, Raaschou-Nielsen et al., 2013]. Pathophysiological mechanisms like inflammation and oxidative stress have been found to constitute plausible mediators but despite recent conclusion regarding the strength and the consistency of this scientific evidence, the extent to which these systematic effects are elicited by ambient pollutionand which biological pathways are stimulated is still undetermined and under debate [Peters, 2012]. Therefore, studies that are helping to enlighten and detail the systemic impact of ambient particles need to take into account and substantiate the multi-organ involvement in response to inhalation of particulate matter. Moreover, deepening the knowledge regarding the genome, it has become more evident that genetics alone is not sufficient to explain the risk of common diseases. There are, in fact, non-genetic and extra-genetic factors that play an important role. Focusing on cardiovascular disease, Baccarelli et al. [Baccarelli et al., 2010] produced a clear conceptual model of how the different worlds influence each other. Epigenetics lies in the middle being influenced by genomics but also by the environment and these three elements combine for subclinical diseases that lead to cardiovascular diseases. Therefore it is no surprise to observe how epigenetics has arisen in the recent year as a key research area in both biomedicine and public health. A first and sharp definition of epigenetics was given by Sir Conrad Waddington in 1942, who defined it as "the branch of biology which studies the causal interactions between genes and their products, which bring the phenotype into being" [Waddington, 2012]. Now we can define it as the heritable changes in phenotype and gene expression that are occurring without a change in the genomic sequence. In fact, the prefix "Epi" comes from ancient Greek and means "upon", "above", "on", "on top of", "over" and defines something that is happening on genetics, over genetics, referring to non- and extra-genetics mechanisms. The most understood epigenetic markers are DNA methylation, histone modification and microRNA. In this work we will consider DNA methylation.
For instance, Currie and Neidell (2005) examine the impact of airpollution (CO, O3, and Pm10 4 ) on low birth weight. 5 To address this, they used fixed effects models at the individual level, controlling for zip code-month fixed effects. To associate exposure to airpollution with low birth weight, they impute prenatal pollutionexposure in each trimester using a radius of 10 kilometers (km) (6.2 miles) around the meter device. Results show no significant effect on low birth weight when the mother is exposed to airpollution during pregnancy. Similarly, using fixed effects at the individual level, Currie et al. (2009b) examine the effects of pollution (CO, O3, and Pm10) on birth weight and prematurity. For birth weight, they utilize a panel with a pollution monitor and mother locations fixed effects, in which averages of exposure to pollution are imputed for the three trimesters of pregnancy. Results show that a one-unit increase in CO during the third trimester leads to an average birth weight reduction of 16.65 grams. Currie et al. (2009b) regress levels of pollution during the three trimesters of pregnancy to different birth outcomes (including a model for child mortality). These authors use a rich set of controls as well as fixed effects for the closest airpollution monitor, an interaction between the monitor effect and each quarter of the year (to capture seasonal differences), and mother-specific fixed effects to control for time-invariant characteristics of neighborhoods and mothers. Results show that a one-unit increase in CO during the third trimester reduces birth weight on average by 16.65 grams (results were found at lower levels of CO). Currie and Walker (2011) exploit a policy that reduced traffic congestion in the U.S., in which electronic toll collector technology was implemented to look at the effects of traffic congestion on newborn health. This policy allowed them to implement a difference-in-differences design, in which the treatment group is made up of mothers living within two km of a toll plaza, while the control group is made up of those who live close to a highway, but between two km and 10 km of a toll plaza. Results suggest that implementing the E-ZPass 6 is associated with significant reductions in prematurity, by 8.6%, and in low birth weight, by 9.3%. Finally, Coneus and Spiess (2010) present a study using mother fixed effects and year/zip code effects together with an ample set
In summary, it is recommended to design a study such that the need of precise results, like accurate health risk estimates, and the need for a cost-efficient study design are in a good balance. This means in a concrete way that collecting data of personal exposures might indeed lead to precise risk estimates but are on the other hand often very expen- sive and can thus reduce the sample size because of cost constraints which can then reduce the statistical power of the study. Thus, in some cases exposure measurement from fixed site monitors can be more cost-efficient. As opposed to that, if the surrogate exposure, which is usually the fixed site monitor measurement, is a bad representative of the personal exposure, measurement error will have a severe impact on the study’s results. That is, measurement error reduces the effective sample size which reduces the statistical power of a study to detect truly significant results. Also, measurement error can cause in some cases critical bias in the estimated coefficients that describe the association between exposureand outcome. In the case of severe measurement error due to inaccurate surrogate measurements, the only use of the cheap surrogate mea- surements might be inadequate. In many cases one can achieve a good trade-off with precise results and efficient allocation of resources by conducting a validation study, additionally to the main study. Validation studies have to be well-designed to allocate resources optimally to the main and the validation study. The primary method to ad- just the study design for measurement error is to increase the sample size. Appropriate sample size adjustment approaches have to be applied while using correctly specified measurement error models adapted to individual needs.
Proceedings of The National Conference and Training Workshop “Wildlife and Aerobiology” held on February 6-7, 2015 Lahore, Pakistan RESULTS AND DISCUSSION
Mass Concentration of PM: The calibration factor for
the Grimm analyzerswas determined as 0.80 and data was adjusted accordingly.The temperature and humidity during various days of measurements ranged from 30 – 35C° and 50 – 70%, respectively. Table 1 presents the summary of results for different size fractions of PM over the period of monitoring. During the weekdays (Monday - Friday) the mean hourly average concentration of PM 10 , PM 2.5 , PM 1 and PM 10 – PM 2. 5at the road sites was 305μg/m 3 , 84μg/m 3 , 61μg/m 3 and 222μg/m 3 , respectively (Table 1). The coarse size fraction was particularly large but also exhibited a wide variation. Comparison of these concentrations with levels reported by the road sides in an earlier study carried out in Lahore by Colbeck et al, (2011) revealed that the average levels of PM2.5 and PM1 between two measurement campaigns were comparable. However, PM 10 was lower during this campaign. Similarly the mean concentrations of PM 2.5 and PM1during weekdays in this study were in agreement with Alam et al, (2011). They reported average concentration of 91μg/m 3 and 68μg/m 3 for PM 2.5 and PM 1
has an average PM 2.5 level of 4.9 µg/m 3 ; the second is 5.9, third is 6.9, and the highest wind direction is
associated with mean PM 2.5 level of 7.9 µg/m 3 . These categories illustrate that any given county has wide
variation in PM 2.5 levels associated with wind direction, ranging an average of 3 µg/m 3 from lowest to
highest wind direction-pollution combination. Therefore, any given county would be expected to see at least 5% more confirmed cases on the worst wind-pollution combination days in comparison to the best ones, or at least a 1.7% increase in confirmed cases from even a slight step up in PM 2.5 associated with a marginal change in wind direction. For comparison, Persico and Johnson (2020) found that a one unit increase in fine particulate matter led to an approximate doubling of confirmed cases and deaths. Recall that the effects in Persico and Johnson (2020) are induced by regulatory rollback of pollution control at Toxic Release Inventory sites. Because these sites release many harmful pollutants other than fine particulate matter, it seems reasonable that the magnitude of their findings would be higher than those we observe. Further, populations especially exposed to pollution from Toxic Release Inventory sites may be more vulnerable for a variety of reasons, including past exposure to the many harmful pollutants released at those sites. Persico and Johnson (2020) analyze heterogeneous effects along standard socioeconomic dimensions and find worse pollutionexposure for counties with a higher fraction of Black individuals, with higher unemployment, and with lower incomes. Our approach allows isolating the effect of PM 2.5 while inferring from a much broader set of pollution sources, thus generalizing previous results. A wide set of policy implications follow from our empirical findings, which we discuss in Section 6.
Studies reported a decrease in FMD in association with an increase in airpollution as well as with changes in air temperature (17, 18, 23, 24). Schneider and colleagues (18) reported an immediate decrease in FMD in association with increases in PM 2.5 in diabetic individuals. Moreover, increases in PM 2.5 were associated with a one day delayed decrease in NTGMD in the same study (18). This finding indicated that PM 2.5 immediately affected the bioavailability of NO, but had a delayed effect on the function of the smooth muscle cells. A further U.S. study conducted in Boston reported a decrease in brachial artery diameter in association with an increment in the 5-day PM 2.5 -average in diabetics. Moreover, brachial artery diameter increased with a temperature increase, but no associations were found when investigating FMD and NTGMD (25). Inverse associations between UFP and endothelial function were also reported and discussed in a review of the physiological effects of UFP (26). For example, inhalation of high UFP levels during exercise was associated with decreases in FMD in healthy men in the U.S. (27). A further U.S. study on controlled laboratory exposure to ultrafine carbon particles reported reduced bioavailability of NO in healthy adults (28). However, no associations were found between UFP and endothelial function in diabetics in a study conducted in Boston (29).
We add to these closely related papers in two important ways. First, whether or not results from previous pandemics and/or other regional settings are generalizable to other ex- ternal settings, both in case of disease and locality, is yet to be determined. To the best of our knowledge, we are the first to provide evidence on the linkage between airpollutionand the spread as well as the severity of COVID-19 for Germany. Second, we use fine-grained data on daily pollution levels which allow us to identify critical time windows relative to the onset of the illness when pollution levels most crucially affect the mortality risk. This allows us to discriminate between different mechanisms at play. As effects materialize only after the onset of illness, our results support mechanisms causing inflammatory reactions, reducing the im- mune response and aggravating symptoms. Our results do not support proposed mechanisms of higher ability of the virus for airborne infections due to higher airpollution. By focusing on short-term changes in the exposure to airpollution, our results do not speak to potential effects through longer-run exposure to airpollution on respiratory preconditions.
Beelen, R., Hoek, G., Vienneau, D., Eeftens, M., Dimakopoulou, K., Pedeli, X., Tsai, M.-Y., Künzli, N., Schikowski, T., Marcon, A., Eriksen, K.T., Raaschou-Nielsen, O., Stephanou, E., Patelarou, E., Lanki, T., Yli-Tuomi, T., Declercq, C., Falq, G., Stempfelet, M., Birk, M., Cyrys, J., von Klot, S., Nádor, G., Varró, M.J., Dėdelė, A., Gražulevičienė, R., Mölter, A., Lindley, S., Madsen, C., Cesaroni, G., Ranzi, A., Badaloni, C., Hoffmann, B., Nonnemacher, M., Krämer, U., Kuhlbusch, T., Cirach, M., de Nazelle, A., Nieuwenhuijsen, M., Bellander, T., Korek, M., Olsson, D., Strömgren, M., Dons, E., Jerrett, M., Fischer, P., Wang, M., Brunekreef, B., de Hoogh, K. (2013) Development of NO2 and NOx land use regression models for estimating airpollutionexposure in 36 study areas in Europe – The ESCAPE project. Atmospheric Environment 72, 10-23.
The second challenge is the endogeneity of pollution. A large share of pollution is caused by business activity. As firms increase their production (where reliance on workers is a major means of achieving this), they emit more pollutants into the environment. This leads to “reverse causality,” where higher levels of productivity lead to higher levels of pollution. Another concern is that individuals may choose where to live (and work) in part based on the level of air quality in that location. When this is the case, it leads to a non- random assignment of pollution, which may give rise to a spurious relationship between pollutionand productivity. For example, higher-paid individuals may choose to live in cleaner areas. These higher-paid individuals may have higher levels of human capital or other unobserved factors that affect their productivity, and these are now correlated with their pollutionexposure because of their decision regarding where to live.
As mentioned, the measure of excess deaths that we have employed thus far (i.e., the difference between the log of the number of years in two subsequent years) has the ad- vantage of allowing an interpretation in terms of deaths’ annual rate of growth (since it is corresponds to the log of the ratio of deaths in 2020 over deaths in 2019). Several al- ternative definitions of excess deaths have been proposed. For instance, one could simply take the difference between the raw number of deaths in 2020 and 2019. However, this approach would not allow a proper comparison between population groups of different size, which is precisely our goal. A percentage measure, that divides the difference be- tween deaths in 2020 and 2019 by deaths in 2019, would also be inappropriate, since the structure of our dataset involves cells associated with zero deaths. Any of the above three definitions can be computed with reference not just to deaths in 2019, but to an average of deaths in multiple previous years. Averaging over multiple years should produce an approximation of a normal year death count, and indeed it should smooth out yearly trends. However, the larger the number of years involved, the larger the weight of long- term trends in population growth or mortality that would invalidate the approximation. With this caveat in mind, in Figure B7 we show that our results are robust to a more conventional definition of excess deaths involving an average of three years prior to 2020, defined as deaths in 2020 minus average deaths in 2017-2019 (multiplied by 1,000).
The urgency of the racial issue and the need for data collection to account for ethnic and racial factors has been widely recognized also within the medical and epidemiological literature. Preliminary evaluations suggest that the high risk of COVID-19 death for mi- nority ethnic groups can be explained by pre-existing health conditions, such as diabetes, obesity, hypertension, and asthma, that are more common among these groups, possibly because of genetic and biological factors. However, the emerging consensus is that the race differential in the prevalence of COVID-19 is also associated with socioeconomic cor- relates. As argued by Yancy (2020), a large share of the black population in the US lives in poor areas characterized by high unemployment, low housing quality, and unhealthy living conditions, making low socioeconomic status a critical risk factor. Furthermore, the higher prevalence of comorbidities in blacks is also itself associated with highly per- sistent socioeconomic factors. Other relevant health-related and behavioral risk factors, such as smoking, drinking, and drug abuse, are deeply ingrained in cultural norms that are also driven by social inequalities. African Americans suffer for further disadvantage in their ability to adhere to social distancing norms, as working from home, avoiding public transportation, and finding refuge in second homes away from crowded cities, are indeed privileges that are denied to the majority of them. Anecdotal evidence also sug- gests the possibility of a different response by health practitioners to black individuals showing COVID-19 symptoms. All the above considerations, as summarized by Yancy (2020) resonate with the long-standing debate on the racial disparities that are deeply entrenched in US history and point to a need to account for race in COVID-19 research and to investigate the role not only of biological factors but also of socioeconomic ones, while acknowledging that the latter may be rooted on inequalities that have been long neglected.