I appreciate the opportunity to respond to Laurence Kalkstein's critique of my paper on Health and Amenity Effects. As you can guess I disagree with most of his points, but they are important. He has obviously spent a good bit of time evaluating my paper and I wish to thank him for his insights. From: Laurence Kalkstein: > Mr. Moore begins by noting that retirees "...flee to southern and >warmer locales," implying that they enjoy warmer weather better. >Certainly they enjoy warmer weather better in the winter, but certainly >not in summer. Do people move to Phoenix to enjoy its summer climate? >How about southern Florida? In fact, Mr. Moore should recognize that >many of these people are snowbirds, and they "flee" back north during >summer if they can afford to. Even with possible air conditioning >mitigation, I am quite certain that most people would rather not stay in >these summer climates. A good many of the elderly who retire to Florida or Arizona take up year around residence in those states. Clearly they did not move there for the summer temperatures, but with airconditioning they prefer living in those areas year around rather than in the north year around. > 1. Mr. Moore only uses three years of data. Most studies >which evaluate weather/mortality use many more years. Thus, his sample >size for deaths and weather data is only about 36, really too small to >draw proper conclusions. I assume that Kalkstein is referring to my regressions dealing with Washington DC. The size of the sample is perfectly adequate to find statistically significant results which I do turn up. If the relationship had been weaker I might have needed a larger sample but the relationship is quite strong. > 2. Importantly, Mr. Moore uses monthly means and totals as >his variables. There are real problems doing such a project in this >manner. For example, he has removed the daily variability which is so >important in evaluating both weather and mortality data. If he had >looked at daily data, it would have been clear (actually unequivocal) >that daily death totals rise enormously during days when it is very hot. >All of this is lost with a monthly means approach. So, if there were six >very hot days imbedded within a rather innocuous month, the variability >associated with these six days would be lost. We acknowledge in our >research that mortality rises on only about the 10 percent hottest days in >the summer; Mr. Moore cannot evaluate weather "thresholds" which lead to >these increases by using monthly data. I disagree. It may be that very hot weather increase the death rate by causing people who would have died a few days later to die a few days prematurely. Thus a month is better. It averages out daily fluctuations which may be due to heat or maybe due to other things. If hot weather -- which usually comes in streaks of a few days or a week -- has a significant effect on mortality it should show in the monthly data. My data included the summer of 1988 for Washington, which "enjoyed" one of its hottest Julys. Nevertheless the data show lower deaths in the summer than in the winter. Other things can foul up daily figures. For example, Chicago had an extraordinarily hot July in 1995. The coroner reported a marked increase in deaths. What was very curious was that on Friday, Saturday and Sunday, July 14, 15, and 16, the reported deaths were way below the normal of 78 per day, (only 14 people were reported to have died on Saturday) but in the two next days -- Monday and Tuesday -- were well above normal. The previous record low deaths for any day in the last 30 years was 46! These numbers are quite remarkable and quite suspicious given that on July 14 a record temperature of 106 deg F was set at Midway Airport. Could it have been that most people in the coroner's office took the weekend off and counted bodies on Monday and Tuesday? > 3. Mr. Moore lumps winter and summer data together. This is a >serious flaw, as different mechanisms affect mortality across the >seasons. In summer, he is correct in noting that general (baseline) >mortality is comparatively low, but he again misses the point involving >the very high mortality days imbedded in the record. During winter, the >baseline is higher, but the day-to-day variability is much less. By not >separating the seasons, the results shown in Table 2 are largely driven >by the higher winter totals. Thus, even the hottest summer months >probably have lower mortality totals than the typical winter months, and >regressing against temperature across season yields results that show >actual decreases with increasing temperature. Exactly! > Let's put it another way. >Assume you have average summer and winter temperatures. Summer mortality >will be lower than winter for reasons that have nothing to do with the >direct impact of temperature (we'll get to this in a second). Since the >summer temperatures will always be higher, and the monthly summer mortality >will always be lower, is it a surprise that he found an inverse >relationship between temperature and mortality? At the very least, Mr. >Moore should have disaggregated by season and used temperatures and >mortality totals which are variations from the monthly mean. I would bet >him a monthly mortgage payment that his results would be quite >different! By the way, I think that even if Mr. Moore used his >procedures on winter months only, he would find that the coldest winter >months are not those with the highest mortality. The IPCC (1995) study >found that regardless of winter temperatures, winter mortality is always >higher. In addition, it is often not the coldest winter months which >possess the highest mortality from infectious and respiratory diseases, >which Mr. Moore isolates as the most important causes of death in winter. > > Now let's get to the question as to why winter mortality is >generally higher than summer. In winter, more people are confined >indoors, rendering them more susceptible to infectious and respiratory >diseases, as well as possible complications from these diseases which >may kill the elderly and other susceptibles. In fact, Table 1 in Mr. >Moore's paper indicates this quite >clearly. So temperature is indirectly involved, as it confines people >indoors in northern and midwestern cities. However, the difference in >month to month winter temperature changes things very little. High rates >of influenza and other infectious diseases (as well as respiratory and >circulatory diseases) are not necessarily associated with the coldest >months. Mr. Moore should check this out for a variety of cities to see >this for himself. > > To summarize, it is really only summer mortality which is >directly related to temperature. Although some European scientists have >found a weak winter relationship, most agree that weather-related >mortality variability is much higher in summer. Mr. Moore's results are >misleading, as he did not disaggregate by season. I agree that the increased in mortality due to respiratory diseases probably results from people spending more time indoors -- although in Arizona and Florida, people probably spend more time indoors in airconditioning in the summer. However, all other major causes of death, also peak in the winter. Cancer (only slightly higher in the winter) and circulatory disease (22 percent higher) deaths cannot be assumed to come from living indoors. Moreover, if the climate were warmer, people would spend less time indoors being exposed to disease. My section on death rates by metropolitan areas gets around any seasonal effects. It turns out that it is healthier to live in a warm climate than in a cold. Kalkstein fails to mention these results, which are quite powerful. I explain between 93 and 95 percent of the variance in death rates among these urban areas by age, race, income and various measures of temperature. These results suggest very strongly, that warmer is healthier. > > Mr. Moore's discussion about vector-borne infectious disease is >similarly inaccurate. He cites only developed country examples in his >discussion; what about potential spread of vectors in countries where >public health facilities and sanitation are poor? Also, the argument >that he uses about winter and night temperatures rising more than summer >and day temperatures works against him here. The spread of tropical >vectors will be enhanced if winter and night temperatures increase, as >the freezing line (and other cold-temperature thresholds which retard >tropical insect development) will move poleward. The fact that summer >temperatures don't increase as much will be a benefit to these vectors, >many of which find dessicating summer conditions unsuitable. I will >restrict my remarks on the vector-borne disease question, as there are >others more qualified than I to talk about this. I agree that a warmer climate can increase the spread of tropical vectors, but as I said in the paper, apparently Kalkstein missed it, poor people have few resources to protect themselves. Rich countries are largely immune. I cited Singapore and Hong Kong as well as Hawaii as examples of rich but tropical areas with Western type life expectancies. Richer is healthier. As has been shown in a recent study in Africa, simple mosquito netting with a cheap insecticide can cut malaria significantly. Rich countries have screens -- a low tech but effective barrier to insect spread diseases. The southern part of the United States use to suffer from malaria and before there is any climate warming has many of the conditions necessary for the spread of tropical vectors. We are rich enough that we need not fear these diseases becoming a major problem. > The impact of a potential global warming on mortality is more >difficult to assess. First, the impact of air conditioning and >acclimatization is not easy to quantify, and it must be acknowledged that >strong uncertainty exists. However, most research shows that summer >mortality will increase across the United States, especially in >the Northeast and Midwest. Why should it? Dallas, Miami, Phoenix, and Houston are very hot in the summer, and all of them are in my sample of 89 counties. As can be seen from my Table 3B, the higher the SUMMER temperature, the lower the death rate (regression 8). Also the more cooling degree days (i.e. the hotter the temperature) the lower the death rate (regressions 6 and 7). Is it going to get hotter in Chicago than it is now in Phoenix? If people in Phoenix don't die of the heat, why should those in Chicago? This also counters some of Mr. Moore's >findings. The frequency of intense heat waves (with temperatures >exceeding the threshold) will increase. It is true, as Mr. Moore states, >that winter minimum temperatures are expected to increase more >dramatically than summer maximum temperatures. But it is not certain >what will happen to climate variability; one plausible scenario is that >the frequency of extreme heat waves will increase. Although humans may >acclimatize somewhat to this increased frequency, the urban structures >are not likely to change. Thus, the tenement housing common to >vulnerable cities (red brick houses, black tar roofs, windows on two >sides only) will most likely remain, and most victims of heat-related >mortality are pulled out of these type of buildings (Mr. Moore should >talk to the medical examiners of Chicago, Philadelphia, and other large >cities). Thus, a summer increase in mortality is likely, as >acclimatization will be incomplete. Mr. Kalkstein asserts that tenement housing conditions will mostly likely remain. I doubt it. Most tenement housing is old and poor and will not be around in 100 years when these conditions are forecast to occur. Cities are constantly being rebuilt and these are already old structures with a very limited life expectancy. Most probably will not be around in 20 years. > There is much disagreement about the direction of winter >mortality if global warming takes place. One argument does agree >somewhat with Mr. Moore's assessment - winter mortality might decline to >such a degree that it will offset increases in summer mortality (Mr. >Moore is incorrect when he states that our studies have, "...ignored the >reduction in mortality that warmer winter months would bring." This is >always a major discussion among those of us who work in this area, and it >has been addressed by the IPCC. We are constantly battling with this >question). Studies that discuss a commensurate winter decline under >global warming come from Europe, and notably Britain, where summer >weather is not marked by the severe heat waves found in the U.S. Most >U.S. studies (including additional ones from China and Egypt) clearly >suggest that declines in winter mortality will not offset much >larger increases in summer mortality. The prevailing thought is that a >few degree increase in winter temperatures will still keep people >confined for most winter days, and not reduce infectious and respiratory >illnesses which lead to many winter deaths. Remember again that the >direct relationship between winter temperatures and mortality is much >weaker than for summer; this is especially true for infectious and >respiratory illnesses. Even if winter temperatures increase more rapidly >than summer temperatures, people in New York, Chicago, Philadelphia, and >other mid-latitude cities vulnerable to such deaths will still remain >largely indoors. > > Again, I want to emphasize the fact that the summer-winter >compensation issue is not resolved, and the IPCC chapter on human health >and climate change clearly states this. However, the prevailing thought >points to the suggestion that the reduction in winter deaths will not >compensate fully for the increase in summer deaths. Again I point to the data from the 89 counties which show that higher average annual temperatures, higher summer temperatures, and higher winter temperatures all are related to lower death rates. Thomas Gale Moore Senior Fellow Hoover Institution (415) 493-7358 http://hoover.stanford.edu/~moore/moore.html -------