Which is responsible for more U.S. deaths - Excessive Heat or Excessive Cold?

The USGCRP Synthesis Report states that data on 19,958 deaths from weather related extreme events from 1970 to 2004 for the US indicates that heat/drought is responsible for the largest share (19.6%), followed by severe weather, defined to include fog, hail, wind and thunderstorm (18.8%) and winter weather (18.1%). This information is sourced to Borden and Cutter (2008), henceforth B&C. [Note that these estimates exclude deaths from excess winter mortality, which is a chronic phenomenon unrelated to extreme weather. Also, note that it's not just global warming, but also the heat island effect that may contribute to excessive deaths in warm weather.]

In contrast to B&C, other researchers have identified deaths from excessive cold as the single largest cause with twice as many dying from excessive cold as excessive heat (e.g., Deschenes and Moretti, 2009; Thacker et al., 2008; Goklany, 2007, 2009; Goklany and Straja 2000).

What accounts for this discrepancy?

As acknowledged by B&C (p. 10 of 13), it depends on data source as well as how the events are grouped. B&C used the Spatial Hazard Events and Losses Database for the United States (SHELDUS; available at www.sheldus.org) which is derived primarily from the National Climatic Data Center (NCDC) monthly Storm Data publications, while the other publications use death certificate data maintained by the Center for Disease Control (CDC) in the Compressed Mortality File. In the US, death certificates must be accompanied by a medical certificate of death signed by the attending physician at time of death. As such, the CDC's Compressed Mortality File is the best source for the cause of death.

B&C justify their choice on the basis that "unlike Storm Data (upon which SHELDUS is based), the Compressed Mortality File is not solely focused on natural hazard events. Although both SHELDUS and the Compressed Mortality File likely suffer from undercounting hazard related deaths [4,39], it is known that the only reason any of the deaths appear in Storm Data (and SHELDUS) is because of some natural event. In the CDC's Compressed Mortality File, deaths are interpreted from classifying the underlying cause listed on death certificates [4], whereas SHELDUS mortality is derived from Storm Data." It also notes that the coding system used by the CDC was revised after 1998.

Neither of these reasons is compelling. First, Storm Data procedures were also changed in the 1990s (B&C, p. 3; Dixon et al., p. 939). More importantly, studies that have attempted to verify numbers from Storm Data or the Annual Summaries based on Storm Data, find that they substantially underestimate deaths (e.g., Ashley and Gilson, 2009; Goklany, 1999, 2007). To quote Dixon et al. (2005): "weather-related catastrophic 'group kills' rather than 'individual kills' are more likely to be included in Storm Data. Therefore, this may tend to give more complete numbers for weather-related categories, such as tornadoes, hurricanes, or heat waves, than for deaths from winter cold, and the multiple categories for excessive cold deaths can also introduce an underreporting of cold deaths." They also observe that, "The CDC NCHS's Compressed Mortality Database is, in general, a more comprehensive database. As such, it would more likely include weather-related 'single kills' than would Storm Data."

This observation is critical because it is possible to have people die from excessive cold even in the midst of a "normal" winter because of exposure to elements as an unintended consequence of intended or unintended actions (e.g., taking a walk in the cold or through loss of heating, for whatever reason).

Dixon et al. also add that, "However, the Compressed Mortality Database is limited by the medical personnel's actual determination of the 'weather relatedness' of death, and the database often runs years behind current events." Regarding the first part of this argument, one should note that determining the cause of death requires medical expertise rather than meteorological expertise such as the NCDC possesses. With respect to the second part, one must respond that in a scientific exercise, speed of reporting cannot take precedence over the accuracy or completeness of data.

Finally, although NCDC is a part of NOAA, the data it provides (based on Storm Data) is sometimes at odds with data from other parts of NCDC. For example, its data on deaths from floods is different from that of the Hydrological Information Center (HIC), the group within NOAA charged with keeping data on flood deaths. And the compilers of the Storm Data-derived Annual Summaries, themselves have in the past suggested using the HIC compilation (Goklany, 1999, footnote 38, pp. 337-338; 2007, footnote 214, p. 457, 2009, pp. 105-106).

For all these reasons, it is more appropriate to use the CDC's Compressed Mortality Database for deaths from excessive heat and cold. And this database indicates that on average twice as many people die from excessive cold than excessive heat.

Additional References
Ashley,W.S., and Gilson, C.W. 2009. A Reassessment of U.S. lightning mortality. Bulletin of the American Meteorological Society doi: 10.1175/2009BAMS2765.1.

Borden, K.A., and Cutter, S.L. 2008. Spatial patterns of natural hazards mortality in the United States. International Journal of Health Geographics doi:10.1186/1476-072X-7-64.

Deschenes, O., Moretti, E. 2009. Extreme Weather Events, Mortality and Migration. Review of Economics and Statistics 91: 659-681.

Dixon, P.G., Brommer, D.M., Hedquist, B.C., Kalkstein, A.J., Goodrich, G.B., Walter, J.C., Dickerson, C.C., Penny, S.J., and Cerveny, R.S.. 2005: Heat mortality versus cold mortality: A study of conflicting databases in the United States. Bulletin of the American Meteorological Society 86: 937-943.

Goklany, I.M. 1999. Richer is More Resilient: Dealing With Climate Change and More Urgent Environmental Problems. In: Bailey, R., ed. Earth Report 2000, Revisiting the True State of the Planet. New York, NY: McGraw-Hill, pp. 155-187.

Goklany, I.M. 2007. The Improving State of the World: Why We're Living Longer, Healthier, More Comfortable Lives on a Cleaner Planet. Washington, DC: Cato Institute: 167.

Goklany, I.M. 2009. Deaths and Death Rates from Extreme Weather Events: 1900-2008. Journal of American Physicians and Surgeons 14: 102-09.

Goklany, I.M., and Straja, S.R. 2000. "U.S. Death Rates due to Extreme Heat and Cold Ascribed to Weather, 1979-1997." Technology 7S: 165-173.

Thacker, M.T.F., Lee, R., Sabogal, R.I., and Henderson, A. 2008. Overview of deaths associated with natural events, United States, 1979-2004. Disasters 32: 303-315.