Power and Gas Trading and Their Modelling Problems

Electricity cannot be stored at scale, so power prices balance supply and demand in real time rather than through a curve.

Power and natural gas trading differ from most other commodity markets because electricity cannot be stored economically at scale, which forces prices to reflect instantaneous supply and demand balance rather than an inventory-smoothed curve, and because both markets are shaped by physical network constraints, weather and regulation in ways a purely financial model tends to miss. A candidate interviewing for a power or gas desk should expect questions that test physical market understanding as much as pricing theory.

Almost every other commodity discussed in a futures pricing context, crude oil, gold, wheat, can be stored, which is exactly what makes the cost of carry model work: today's price and a future price are linked by the ability to buy now, hold, and deliver later. Electricity has no such link at meaningful scale, since grid-level battery storage remains a small fraction of total generation capacity in most markets, so supply and demand for electricity must balance in real time, second by second, on the physical grid.

This means power prices at different hours of the day, or different days of the week, are not connected by an arbitrage relationship the way futures on a storable commodity are. A trader cannot buy cheap overnight electricity and sell it at the expensive afternoon peak without a battery or another storage mechanism, so the price relationship between time periods reflects genuinely separate supply and demand conditions rather than a carry cost.

Because electricity cannot be stored, an unexpected surge in demand, a heat wave driving air conditioning load, or an unexpected loss of generation capacity from a plant outage, forces the price to whatever level clears the market in that instant, and that level can be extraordinarily high since demand for electricity is highly inelastic in the short run. Prices that spike to ten or twenty times their typical level for a few hours are a normal feature of power markets rather than a rare tail event, and a pattern that would be considered a market malfunction in most other asset classes.

This volatility profile makes standard risk models built for continuous, roughly normally distributed returns a poor fit for power price behavior, and desks trading power typically build models with explicit spike components and regime switching rather than relying on a single volatility parameter.