The chain of causality in a weather-driven market such as natural gas allegedly works as follows: It gets cold, heating fuel use rises, and inventories are reduced thereby. Traders jump in front of the anticipated demand for rebuilding those inventories, and firms that supply natural gas, or whatever the commodity in question is, see their shares rise accordingly. It all looks like a testable set of assumptions, so let's take a look.

A Degree a Day

The measure of energy demand for either heating or cooling is the degree-day, which is the maximum of (0, (65 degrees - current temperature)) for a heating degree-day or the maximum of (0, (current temperature - 65 degrees)) for a cooling degree-day. Considerations such as the wind-chill factor, the temperature-humidity index or whether anyone actually turns on the air conditioning at 67 degrees do not factor into the equation. Can we predict the price of natural gas using temperature as the sole predictive variable?

No, not at all. Neither the price of natural gas, nor overall economic activity, prices and availability of alternative energy sources, nor the health of natural gas-consuming industries such as fertilizer or petrochemicals enters into the equation. Each of these factors is quite important; I suspect, for example, that those who use gas heat have turned down their thermostats and have taken other conservation measures, now that $6 per million BTU natural gas is common. During the 1990s, when $3 natural gas was high, buyers could afford to be more profligate in their habits.

Any model for either trading natural gas directly or for pricing weather derivatives -- bets that pay off on the state of weather variables such as temperature or precipitation -- that seeks to link natural gas demand solely to heating-degree days is likely to be too simplistic.

We can compare the Department of Energy's Eastern Consuming Region inventories for natural gas against heating-degree-days in the East North Central region expressed as a percentage of normal over the November-March heating season. The American Gas Association weather data were used prior to November 28, 1997, and National Weather Service data were used thereafter. The winters of 1995-96, 2000-01 and 2002-03 stand out for their combination of cold weather and inventory drawdowns.

The winter of 2000-01 in particular began the heating season from a low starting point in inventories, as it followed a hot summer that forced utilities to use natural-gas-fired generators to meet their demand surges. This also was the summer when water reservoir levels in the Pacific Northwest were low, a condition that reduced hydroelectric power available to the western states and led to the various machinations of Enron (ENRNQ - commentary - Cramer's Take) and others in the California power market.

Even with all of the special factors taken into consideration, the intuitively pleasing conclusion that colder-than-normal weather will lead to more rapid inventory drawdowns of natural gas appears acceptable.

Natural Gas Inventories and Cold Weather

Source: Bloomberg