Power grid operators need to plan for the future, but how do they know what to plan for? They make forecasts of the future, incorporating expected supply and demand, economic growth, weather trends, and demographic changes.

No one has a crystal ball, but clearly planners could do better.

The Project for a Sustainable FERC recently commissioned a report from the Brattle Group that finds PJM consistently overestimates future demand because they underestimate the effectiveness of energy efficiency programs. By 2022, they estimate PJM will buy an extra 27,245 gigawatt-hours, equal to a 3000 MW power plant running full out for a year.

So PJM ends up with more capacity than they need. Not a problem, right? It’s a good conservative engineering approach, reliability is maintained, and nobody’s lights go out.

But this comes at a significant cost, say $1.3 billion over three years. Not to mention building unnecessary infrastructure and propping up old coal plants that should retire.

Grid planners in Texas have likewise been criticized for predicting too much power demand. In their twice-annual Capacity, Demand and Reserves (CDR) Report, ERCOT forecasts expected reserves for the next five years.

They have consistently overshot in their predictions of peak demand and underestimated the amount of new supply.

The Texas Industrial Energy Consumers group warned against putting too much faith in the forecasts.

“The CDR report systematically over projects reserve margin shortfalls that never materialize,” they wrote in a filing in 2013. “The market does not believe these predictions and [the] Commission should not take drastic action in response to these reports.”

The debate about capacity markets has inevitably raised the question — how much capacity do we need to buy?

Brattle Group, in an earlier report for the Texas PUC, tried to estimate the “economically optimal” margin for ERCOT. While engineers have traditionally used a rule of thumb of “one outage in ten years” — in other words, where demand exceeds supply once in ten years and a blackout occurs — Brattle questioned why that was the rule.

They balanced the cost savings that come from building fewer power plants against the price suppression effects of having a lot of supply, and found the magic number to be 10.2%.

“This risk-neutral, economically optimal reserve margin is substantially below the 14.1% reserve margin needed to meet the traditional 1-in-10 target,” they wrote.

Nevertheless, the annual savings that could be gained by dropping the margin 4% was an “only modest” $100 million, on an annual power bill of about $35 billion.

A few hundred million here, a few hundred million there — pretty soon you’re talking real money. And real enough to nearly double the annual energy efficiency programs of Texas utilities, which would increase reserve margins, clear the air, and save money for consumers.