by David Orrell
17. Cowles 1933.
18. Fama 1965.
19. As opposed to “alternative” areas of research such as green economics, which are not usually aimed at making specific financial predictions.See, for example, Hawken 1994; Daly and Cobb 1989.
20. A number of websites help you analyze charts; see, for example, www.stockcharts.com.
21. For a discussion of the track record of technical trading with references, see Malkiel 1999, p. 160. The statement applies to average performance.
22. Haugen and Lakonishok 1992.
23. Fama 1965.
24. Williams 1938.
25. Sherden 1998, p. 167.
26. Malkiel 1999, p. 170.
27. Fama 1965.
28. On March 6, 1999, The Economist announced on its cover that the world was “drowning in oil.” Prices had been falling in a roughly linear fashion for over a year, and had reached ten dollars a barrel. The article ventured a prediction that prices could soon fall as low as five dollars. At about exactly that moment, oil prices changed direction and reached twenty-five dollars by the end of the year (when the magazine followed up with an article entitled “We Woz Wrong”). Anonymous 1999.
29. The Organisation for Economic Co-operation and Development (OECD) notes that until the mid-1990s, many housing markets tended to follow a cycle with an expansion phase lasting around six years, and a contraction phase lasting about five years (maybe Jevons had a point when he linked the economy to the eleven-year sunspot cycle . . . ).However, this pattern then seemed to break when low interest rates ushered in a long period of expansion (OECD 2005).
30. Woudenberg 1991. Quoted in Sherden 1998, p. 169.
31. Adapted from Koutsogeorgopoulou 2000, figure A2, p. 45. Results are similar for the individual countries.
32. A 1993 comparison of OECD forecasts with those from the International Monetary Fund and the governments of the USA, Japan,Germany, France, Italy, and Canada showed no improvement over naïve forecasts. OECD Economic Outlook, June 1993.
33. The Economist, July 27, 1991, p. 61. Quoted in Sherden 1998, p. 62. “Of 60 recessions in developed and developing economies during the 1990s, two-thirds remained undetected by consensus forecasts as late as April of the year in which the recessions occurred. In one-quarter of cases, the consensus forecast in October of that year still expected positive growth.” Anonymous 2001a.
34. Anonymous 2005b.
35. McCauley 2004, p. 6. For a discussion of models used to predict currency fluctuations, see Cheung et al. 2005.
36. Buchanan 2000, p. 133.
37. McNees and Ries 1992.
38. McCauley 2004, p. 6.
39. Malkiel 1999, p. 242. “Where it counts, the market’s behavior conforms to the rational model.” Bernstein 1998, p. 296.
40. Aristotle et al. 1981, p. 90.
41. Kenneth Arrow described this discrepancy as an “empirical falsification” of orthodox theory. See Ormerod 2000, p. 16.
42. Liu et al. 1999.
43. See McCauley 2004, p. 7; Ormerod, p. 128 for examples with economic variables.
44. Buchanan 2000, p. 157.
45. See Barabási and Albert 1999; Zipf 1949.
46. The number of connections that proteins have in biological networks also follow a power-law distribution. Barabási and Oltvai 2004. See Buchanan 2000 for a general discussion.
47. Dunphy 2004.
48. See Homer-Dixon 2000, pp. 292–96 for a discussion.
49. Simons 1997.
50. Bollerslev 1986; Engle 1982.
51. Mandelbrot and Hudson 2004, p. 104. See also Loomes 1998, Anonymous 1999a.
52. Singer 2005.
53. Ormerod 2000, p. 72.
54. Iamblichus 1918.
55. See, for example, the above quote from Malkiel 1999, as well as Bernstein 1998, p. 296.
56. “Double-think” is Arthur Koestler’s expression. Koestler 1968.
57. Quoted is the Nobel laureate economist Merton Miller, from Malkiel 1999, p. 273.
58. Soros 2000, p. 27. Dangers of an unfettered market: George Soros interview, 20 Sept., 2002. “Wall $treet Week with FORTUNE,” PBS.
59. Bass 1999.
60. See, for example, Lux and Marchesi 1999.
61. See McCauley 2004 for a discussion. Companies using agent-based simulations for specific business problems include NuTech Solutions (http://www.biosgroup.com/) and Volterra Consulting (http://www.volterra.co.uk/).
62. See Ormerod 2000, p. 200, for an example of a conceptual model that incorporates internal prediction errors.
63. Quoted in Bernstein 1998, p. 203.
64. Soros 2000.
7 ⊳ THE BIG PICTURE
HOW WEATHER, HEALTH, AND WEALTH ARE RELATED
1. See Wright 2004.
2. Ehrlich 2000, p. 245; Kirch 1984; Diamond 2005.
3. Maunder 1997.
4. Asimov 1951.
5. Popper 1957, p. 4.
6. Meadows et al. 1972, p. 23.
7. Ibid., p. 58.
8. See The Population Bomb by Paul Ehrlich (1968). He overestimated the rate of U.S. population growth, which slowed after the introduction of the contraceptive pill. The popularity of his book may also have played a contraceptive role.
9. A number of U.S. politicians seem to believe in Armageddon. For example, James Watt, who was President Reagan’s first secretary of the interior (not the inventor of the flyball governor), testified to the U.S. Congress that protecting natural resources was unimportant since “after the last tree is felled, Christ will come back.” Moyers 2004.
10. According to the Serbian mathematician Milutin Milankovitch, whose 1924 theory appears consistent with recent ice age estimates.
11. Plato 1961.
12. Ruddiman 2005.
13. Ehrlich and Ehrlich 2004, p. 201.
14. Ibid., p. 21.
15. Ibid., p. 42.
16. Schnur 2002.
17. Kennedy 2004.
18. Nepstad et al. 1999.
19. It is impossible to accurately measure the rate of species extinction, but scientists generally agree that it has increased dramatically since our arrival. According to Martin Rees, it has increased from about one species in a million per year to one in a thousand. Rees 2003, p. 101.
20. Stocks of smallpox are maintained at the Centers for Disease Control, Atlanta, and the Vector Laboratory in Moscow.
21. 1.012000 is slightly more than 439 million.
22. For example, Margolis 2000, p. 83: “The very contention that a relatively puny species . . . has damaged, and without really trying to, something as vast as an entire planet will strike many as an especially audacious example of the arrogance of the present.” This refers to “the belief of every generation that it alone has been chosen to live in a ‘special’ time. . . . The first cousin of plain egotism.”
23. Crichton 2004, p. 570. See also Allen 2005. Personally, I agree more with the introduction to Crichton’s 2002 thriller Prey, written before he got into climate research: “The fact that the biosphere responds unpredictably to our actions is not an argument for inaction. It is, however, a powerful argument for caution, and for adopting a tentative attitude toward all we believe, and all we do.”
24. This price mechanism is not straightforward, because it only works if the producers are aware of the scarcity. Norgaard 1990.
25. Huber and Mills 2005.
26. Skinner 1973, pp. 10–12.
27. Stoto 1983.
28. See Wright 2004; Diamond 2005.
29. Population Reference Bureau 2000. See www.prb.org.
30. Ehrlich and Ehrlich 2004, p. 185.
31. World Wildlife Fund, www.panda.org.
32. Millennium Ecosystem Assessment, 2005. Statement of the MA Board. See www.maweb.org.
33. Rees 2003.
34. Deacon 1968.
35. Nordhaus 1994a.
36. Suzuki 2005. Suzuki’s website is http://www.davidsuzuki.org.
37. Arrhenius 1896. The warming effect of greenhouse
gases was first noted by the British physicist John Tyndall in 1859.
38. Actually, Arrhenius thought that global warming would bring benefits, at least for Sweden. Extrapolating from the rate of production at that time, he estimated it would take 3,000 years for atmospheric carbon dioxide to double.
39. Kerr 2004b.
40. Ibid. The word “canonical” is used in this way in Kerr 2005.
41. Jasanoff and Wynne 1998, p. 70. Quoted in Rayner 2000, p. 275.
42. Quoted from Stephen Schneider’s website in August 2005 (http://stephenschneider.stanford.edu/).
43. Whitfield 2003.
44. Over millions of years of the planet’s history, trillions of tons of carbon were removed in this way, buried in sediments and slowly transformed into fossil fuels such as coal, oil, and natural gas.
45. It is hard to assess what the net effect of this activity is on the biosphere. Some is quickly reabsorbed by plants, to whom more CO2 means more food, so they grow at a faster rate. An acre of healthy, young forest can remove carbon from the atmosphere at a rate of up to a ton per year, and five acres can compensate for the carbon emissions of an average North American. The amount of carbon dioxide that can be absorbed by greenery or by the oceans is, however, limited. Plants can’t grow forever, and when they die and decay, or burn in forest fires, they release their stored carbon. The ability of oceans to absorb CO2 is similarly constrained by ocean chemistry, and will eventually reach saturation.
46. Wohlforth 2004, p. 147.
47. The IPCC writes: “Probably the greatest uncertainty in future projections of climate arises from clouds and their interactions with radiation.” IPCC 2001a.
48. Errico et al. 2002.
49. This question has been raised by skeptic lobbying organizations such as Americans for Balanced Energy Choices, who write on their website: “Predicting weather conditions a day or two in advance is hard enough . . . so just imagine how hard it is to forecast what our climate will be 75 to 100 years in the future.” The point seems naïve, but it’s not. Short-term predictability and long-term predictability are linked because model errors affect each—unless you believe that all error is caused by chaos.For simple examples, see Orrell 2003, and the discussion of the Lorenz system in the Appendices.
50. Quoted in Zöllner and Nathan 2003, p. 76.
51. In 2004, the National Center for Atmospheric Research (NCAR) model indicated that the amount of low-level cloud cover will increase, while the GFDL model said it should decrease. As a result, the GFDL predicts an albedo change that is a factor of three greater. Kerr 2004b.
52. In Charles Wohlforth’s book The Whale and the Supercomputer, he describes a NASA experiment in which satellite measurements of floating Arctic ice were compared over different scales, with observations from an aircraft flying as high as 6,000 metres, an unmanned Aerosonde aircraft at low altitude, a team of scientists standing on the ice, and a Native elder. From outer space, a 200 km2 patch consists of a few pixels in the satellite image. The observer in the airplane can make out large patches of snow, ice, and water, and the Aerosonde can pick out distinct features on the surface (such as small variations in height). To the people on the ground, though, the ice seemed “easier to take in as pure chaos. . . . At every scale there were textures and patterns being broken and superseded.” One of the scientists is depressed at the difference between their models and reality. The Iñupiaq elder thinks that the movement of ice is too complex to model or predict, and what can be known just comes down to common sense: “They use science to prove things we already know.” Wohlforth 2004, p. 90.
53. “Flux adjustments are non-physical in that they cannot be related to any physical process in the climate system and do not a priori conserve heat and water across the atmosphere-ocean interface. . . . The approach inherently disguises sources of systematic error in the models, and may distort their sensitivity to changed radiative forcing.” IPCC 2001a. An alternative is to tune the parameters to give a stable model climate.
54. Stainforth et al. 2005. The paper notes: “The range of sensitivities across different versions of the same model is more than twice that found in the GCMs used in the IPCC Third Assessment Report.” Over 2,000 simulations were performed. The selected runs were compatible with climate observations, so in this sense the perturbed parameter values were not unrealistic. Only six parameters were perturbed, which affected the representation of clouds and precipitation. These were “the threshold of relative humidity for cloud formation, the cloud-to-rain conversion threshold, the cloud-to-rain conversion rate, the ice fall speed, the cloud fraction at saturation and the convection entrainment rate coefficient.” The paper points out that the range of perturbations may be too low because “experts are known to underestimate uncertainty even in straightforward elicitation exercises.” Also, “even the physical interpretation of many of these parameters is ambiguous.” There are, of course, many other parameters in a GCM that could be perturbed. The experiment was carried out not on a supercomputer but by using the idle processing capacity of tens of thousands of computers volunteered by the public. See www.climateprediction.net.
55. Of course, this is very useful information, and it was a good experiment. It has been mistakenly interpreted by some authors to imply that the mean and variance of the results says something about the climate. The mean only reflects the original choice of model around which perturbations were made, and the variance only reflects the model’s sensitivity to parameterization, for those particular (and rather small) perturbations. In other words, the results say a lot about the model, but not much about the climate system itself.
56. Allen et al. 2002. See also Smith 2000.
57. Wohlforth 2004, p. 169.
58. Kerr 2005.
59. Rayner 2000. Koestler described Ptolemy’s final version of the Greek Circle Model as “a monumental and depressing tapestry, the product of tired philosophy and decadent science” (Koestler 1968, p. 69).Have we come full circle? The former Secretary-General of the UN World Meteorological Organization, Aksel Wiin-Nielsen, wrote in 1996, “The most important explanation as to why so much extensive theoretical work in the development of climate models has been done during the last ten years is that the development of models sustains funding and secures jobs at research institutions.” Quoted in Lomborg 2001, p. 37.
60. Another example is the yeast galactose network discussed in Chapter 5. The model accurately predicted that removing regulatory feedback loops would make the yeast behave in a more erratic fashion, but it was impossible to be sure of this until the experiment was performed.It would have been very hard to defend the model against a skeptic who didn’t believe the prediction, or to assign probabilities to different outcomes, because we knew that the model was only a coarse approximation of the real thing. We had no idea if there were other feedback loops that could compensate for those that had been removed, or if the mutant yeast would even function as an organism. Pretending to make a probabilistic forecast for a single yeast cell would have been inappropriate, to say the least.
61. For evidence based on measurements of the ocean’s heat content, see Hansen et al. 2005. For information about the state of sea ice, see the National Snow and Ice Data Center (http://nsidc.org). See also Kolbert 2005 for a general discussion.
62. Parmesan and Galbraith 2004.
63. The climate has always been variable, and life on this planet has somehow survived numerous ice ages, along with smaller effects (such as volcanoes or comets) that cool the climate by throwing dust into the atmosphere. In the past, species could often migrate with the climate, moving to cooler locations as it warmed and vice versa. Some went extinct, as species do, but a robust ecosystem is flexible and can adapt well to external perturbations. Such migration is now far more difficult, because of obstacles such as roads, farms, urban conurbations, and so on. Winged species have an advantage, but laggards arriving by foot will take longer, and some trees and plants, which can move only by dispersing seeds, wi
ll find it hard to survive a rapid climate perturbation. Ocean-dwelling species will have their own challenges. Schneider and Root 2001; Tol and Dowlatabadi 2002; Byers 2005; Barnett and Adger 2003.
64. What matters in climate is not so much the average warming over the globe but the local climate in particular regions. If, for example, some areas experienced extreme heating while others were cooled, or some became very wet while others suffered from drought, then the small average change worldwide would mask significant local effects. It is even harder for climate models to pick up these subtle local variations than it is to detect the broader warming trend. IPCC 2001. For example, while the NCAR and GFDL models agree quite well on average warming, the first predicts a wetter United States, the second a drier.
65. See http://www.seaaroundus.org/ for information on the state of fisheries.
66. Baker et al. 2005, Watson 2005. The average historical return is about 6.5 percent. The equations used are simple and make all sorts of assumptions about productivity growth, population growth, and the world at large. Results therefore depend on the judgment and biases of the forecaster.
67. IPCC 2000.
68. IPCC 2001.
69. The latter adjusts wealth according to domestic purchasing power.Anonymous 2005d.
70. Peter Schwartz and Doug Randall, “An Abrupt Climate Change Scenario and Its Implications for United States National Security,”October 2003. Quoted in Byers 2005.
71. Nordhaus 1994. See Rotmans and Dowlatabadi 1998 for a discussion of integrated assessment models. Information about RICE and DICE is available at the homepage of William Nordhaus (http://www.econ.yale.edu/~nordhaus/homepage/homepage.htm).
72. This from The Skeptical Environmentalist by Bjorn Lomborg, who writes that the model “gives the same qualitative conclusions as all other integrated assessment models.” Lomborg 2001, pp. 306, 310. Costs are adjusted for the year 2000.
73. Quoted in Bernstein 1998, p. 203.
74. Ziegler 1965.
75. Osterholm 2005.
76. Matthews and Fraser 2000.
77. The name H5N1 refers to proteins on the surface: hemagglutinin type 5 and neuraminidase type 1. The 1918 virus has been artificially reconstructed. See Tumpey et al. 2005.