| Mathematical Politics |
It is not at all unusual to attempt to predict future events from cur
rent and past ones - and to do so quantitatively with mathematics and experimental measurements. This is the primary professional activity of those who trade stocks and commodities with charts and other quantitative technical indicators.I have carried out many studies of this sort in my scientific work -and some in non-scientific areas. One of the simplest yet most unusual was during President Nixon's first term of office when troops were gradually being withdrawn from Vietnam. After several phases of the withdrawal had been completed, I drew a graph of the number of troops remaining in Vietnam vs. time. This turned out to be a perfect straight line which, when extrapolated into the future, passed through zero in November of the year that Nixon would run for re-election.
It was evident that President Nixon realized that the electorate wanted the troops home but also wanted a victory. He then determined that the troops must be out before the next election but should stay as long as possible. This could have been done in more sophisticated ways, but apparently the bureaucrats in charge used a simple straight edge and constructed a line like mine. A straight line has two adjustable parameters, which determine its slope and location. One parameter was adjusted to make the line fit the current troop deployment number and the other to give a slope such that this number would be zero at the time of the next election.
Some theoretical studies in science involve so-called "
ab initio'' calculations. This means "from first principles.'' These calculations are required to fit theoretical principles but are not constrained to experimental values. They are the most rigorous test of a theory. If the end results of the calculations conform to experiment - without the calculations themselves being adjusted to experimental values - then the theory is remarkably good.Ab initio
calculations usually are restricted to problems in physics and chemistry that are so simple that they can be rigorously described in quantitative terms - such as bonding in a molecule with only two atoms. This was especially so in the era during which scientists had to do calculations by mathematically solving the actual equations. With the advent of computers, problems that could not be solved rigorously could be undertaken much more easily.One of the first such applications was during the development of the atomic bomb. Calculations were necessary (not entirely
ab initio, but partially so) that could not be performed with exactitude. Iterative procedures were necessary wherein a solution is guessed, a complicated calculation evaluates the guess, a new guess is made, and the process repeated over and over until a satisfactory solution is reached. I have heard Richard Feynman, Joseph Mayer, and Edwin York each describe the computer that was used.The scientists had a large room full of desks at which were seated people with the ability to carry out simple arithmetic operations. Each person was a logic element in the "computer.'' IBM punch cards were passed through the room in a predetermined pattern. Each person carried out his calculation, punched the card, and sent it on to the next "computer element.'' This computer was designed and built by Klaus Fuchs, who later confessed to being a spy. The computer worked so well that Richard Feynman, who had been assigned to check it, found himself with little to do and moved on to other problems. Ed York told me that, as late as 1965, he was still using the results of calculations of the propagation of hydrodynamic blast waves in air that were made with this computer.
Extrapolations of current data to predict future values can be carried out with or without a theoretical scientific basis. For example, measurements of the Earth's atmospheric temperature could be fitted by a simple polynomial equation with enough adjustable parameters to assure that the fit was a good one. Then, predicted future temperatures could be calculated from this fitted polynomial equation.
Better results would be expected by fitting the temperature data with equations that reflect natural processes. These are the equations that scientists involved in climate modeling are trying to develop. Since the Earth's climate is a very complex system, these models must have lots of adjustable parameters. The models are partially based on theory, but, since theory here is quite poorly developed and imperfect, they must depend largely upon experimental measurement.
In a case such as this, the calculation is constrained to fit the experimental data. Adjustable parameters are set, by calculation, such that the model is always in agreement with the known data. Then the resulting solution is used to predict additional experimental results. If, when those experiments are performed, the model is right, it is considered to be a good model. Alternatively, the model may be constrained to some of the data and evaluated with respect to other data.
[For example, when we were trading in the copper market, we developed a computer model of that market that called about 50 long and 50 short trades per year. Each trade was of about four hours duration. At the time (1981), we had our computers connected to the Comex by wire and had recorded three years of data including every tick in the market. Our model had about 10 adjustable parameters. We therefore optimized those parameters on the 200 trades in the second and third years back. The results were evaluated on the 100 calculated trades from the most recent year - which had not been used to optimize the model. Then, we used the model to trade - to predict market action four hours ahead twice a week - in real time. This worked quite well. ]
Since temperature measurements are among the most reliable measurements that have been made on the atmosphere, all climate models should be constrained to them. Yet, we see that the "global warming'' models adopted by the United Nations in calling for international rationing of energy do not even agree with the past temperature record of the atmosphere as measured by balloons and satellites. By definition, any good climate model should be forced by calculation to agree at least with past atmospheric temperatures.
Apparently one of two possible "errors'' has been made. Either the U.N. climate models have not been fitted to the known temperatures at all or else they have been fitted to the NASA surface record - which has obviously been selectively compiled to show rising temperatures that are not occurring in the atmosphere - or to some other biased record that rises with time. (See, for example, the section on heat island effects in the NASA-selected measuring stations in California in "Experimental Effects of Increased Atmospheric Carbon Dioxide.'') In either case, this sort of fitting would be a
political activity, not a scientific one. It makes no scientific sense at all to construct a computerized model of the atmosphere and not constrain the adjustable constants in the model to conform to the known past temperature record of the atmosphere. Temperature is such a fundamental variable that it would never be used to test such a model - except by prediction of future temperatures, which is not the case at issue here.If the U. N.-endorsed climate models are fitted to a biased record that shows rises in atmospheric temperature with time (not what the recent record shows), then there really is no need for a fancy "climate model.'' The same result would be obtained by fitting to a simple polynomial that does not contain any climate science at all. The result would be the same - but the
political effect would not be the same! No one could represent the polynomial as "climate science,'' and its link to the flawed temperature record would be impossible to conceal.Instead, the world is being told that "very complicated'' computer models have been designed by "experts'' (who happen to receive large welfare payments from the Clinton Administration). These models are represented as so sophisticated that mere mortals - even other scientists - cannot hope to understand them. In common parlance, this is a "scam.'' The computer models do not fit the past atmospheric data because they have been designed,
for political reasons, to predict that future temperatures will rise. Were that not the case, they would agree with the past experimental data. Since atmospheric temperatures for the past 20 years have not risen, the "modelers'' are unable to get both the politically correct warming prediction and agreement with the immediate past record simultaneously. To please their masters, they have chosen to get the politically desired result rather than to appropriately constrain their models.