Why a Wind Tax Credit Is Only Part of the Negative Electricity Price Story

When electricity is plentiful and demand is low, wholesale energy prices can fall below zero. In those hours, generators effectively pay to supply energy rather than receiving the usual payment. Why would they do that, and how much responsibility lies with subsidies for renewable power?

Economists Erik P. Johnson and Daniel H. Karney investigate those questions using data from five wholesale electricity markets in the eastern United States. Their preprint finds evidence that the federal production tax credit for wind contributes to negative energy prices. But it also points to other mechanisms, including the difficulty of quickly turning some power plants down and back up.

The study examines the energy component of wholesale electricity prices, not the full price at every point on the grid. The full local price also includes effects such as congestion and transmission losses, so a negative energy component is not the same thing as a negative local price or a negative household electricity bill.

Why bid below zero?

The production tax credit gives qualifying generators a credit for each unit of electricity produced during their first ten years of eligibility. A wind generator receiving the credit may be willing to accept a negative market price if the credit makes generating worthwhile overall. If enough generators submit negative bids relative to demand, the market energy price can become negative.

That is not the only possible route. Some thermal power plants cannot rapidly change output. Their operators may accept a loss in a low-demand hour to stay running and be ready to generate when demand and prices rise later. Other production incentives can also affect bids.

The researchers assembled hourly price and demand data covering portions of 2000–2023 in New England, the Midcontinent, New York, PJM and the Southwest Power Pool. Coverage varies by market. About 18% of the observed negative energy price hours had prices more negative than the value of the tax credit. Under the authors’ bidding model, the credit alone could not account for prices that low, although it could still have contributed alongside other factors.

Separating wind from the subsidy

Wind generation and negative prices can occur together without the tax credit being the sole cause. To distinguish the credit’s role, the authors used the ten-year eligibility limit: older wind capacity can lose eligibility while newer capacity remains eligible.

They combined generator operating dates and monthly production records with hourly market-wide wind generation. Because they did not observe each wind generator’s output every hour, they estimated it by allocating monthly output according to the market’s hourly wind pattern. Their main statistical model estimated the probability that the energy price was at least as high as the negative value of the credit but below zero. It compared estimated eligible and ineligible wind generation while accounting for current and future demand and differences across markets and time. The authors analyzed prices below the negative value of the credit in a separate regression.

The authors interpret their results as evidence of a causal effect of the credit on negative prices. That interpretation depends on assumptions, including that wind does not systematically favor eligible over ineligible generators within a market and that the timing of eligibility is not related to unobserved influences on current prices. The hourly allocation also assumes plants within a market experience the same wind pattern, an approximation that may not hold everywhere.

The estimated effect varied substantially by market and was largest in the Southwest Power Pool. The paper reports that a 1% increase in eligible wind generation raises the probability of an energy price in the main model’s range—from the negative value of the credit up to, but not including, zero—by 3.4 percentage points on average. Its discussion of the Southwest result, however, describes a change inconsistent with the corresponding reported percentage-point estimate. That makes the precise magnitude and interpretation of these estimates uncertain from the paper as written.

A separate calculation assigns eligible wind generation about 10% of the model’s explanatory power for prices between the negative value of the credit and zero. This is a measure of the variable’s contribution to a statistical model, not a direct count showing that the credit caused exactly 10% of negative-price hours.

Costs and environmental benefits

For 2014–2022, the authors estimate that negative-price payments attributable to the credit could be as much as $369 million using a pooled model, or $570 million using separate market estimates. They compare these figures with estimated benefits of $655 million to $1.09 billion from avoided emissions in the selected negative-price hours.

That comparison is conditional on several assumptions, including how many hours to attribute to the credit and how much emissions-intensive generation each extra unit of wind displaces. Negative-price payments are also not a complete measure of the credit’s social cost, and the calculation omits effects in hours with positive prices. It therefore does not establish the policy’s overall net benefit. The preprint offers a way to investigate the competing drivers of negative prices, but its causal and benefit estimates remain dependent on its methods and assumptions.