This paper, "Selling Wind," provides a game-theoretic look at how the geographic distribution of wind farms affects market competition and social welfare. By modeling wind producers as strategic actors with private information about their own "state" (how much wind is currently blowing at their site), the authors uncover a compelling link between meteorology and market power.
TL;DR
Increasing the "dispersion" (heterogeneity) of wind resources across different locations doesn't just stabilize the grid through diversification; it also weakens the ability of strategic firms to withhold energy. While this always benefits social welfare, it can sometimes hurt producer profits, suggesting that market-driven investment may lead to sub-optimal geographic clustering of renewable assets.
What Problem Does This Paper Address?
Traditional research on market power in electricity sectors usually focuses on fossil-fuel generators, where fuel supply is certain and the primary constraints are technical or regulatory. Renewable energy changes the game because its "fuel"—the wind or sun—is stochastic (random) and often highly correlated across nearby locations.
The authors address a specific gap: How does the stochastic dependence (correlation) of wind availability across different firms impact their strategic behavior? If two wind farms are close together, a "high wind" day for one is likely a "high wind" day for the other. In a market where producers have the power to influence prices, this correlation provides them with information they can use to strategically withhold (curtail) production to keep prices high.
Methods
The authors develop a parsimonious Cournot duopoly model (and later an oligopoly extension) with incomplete information.
- The "Types": Each producer has a private "state"—either High ($H$) or Low ($L$) wind availability.
- The Dispersion Parameter ($d$): This is the core of the paper. It represents the level of heterogeneity. When $d=0$, wind availability is perfectly correlated; when $d=1$, it is independent.
- Strategic Interaction: Firms compete on quantity. Because they don't know their competitor's exact state, they must form a "Bayesian Nash Equilibrium" based on the signal provided by their own realized wind availability.
- Extensions: The model is tested against linear and concave demand functions and extended to include traditional fossil-fuel generators and potential collusion scenarios.
Datasets
While primarily a theoretical paper, the authors use empirical data to motivate their model:
- MISO Wind Data: Hourly energy production data from wind farms in the MISO (Midcontinent Independent System Operator) region from 2014 to 2016.
- Purpose: This data demonstrates that real-world wind farms exhibit varying degrees of correlation based on their location, justifying the use of the dispersion parameter $d$.
Key Insights and Ideas
- The Information Signal: In a highly correlated market, a firm’s own high wind realization is a strong signal that its competitor is also flush with energy. This "information effect" encourages firms to withhold production to prevent a price collapse (the "merit order effect").
- Strategic Curtailment (SC) vs. Wind Diversification (WD): Increasing dispersion ($d$) helps society in two ways. First, through WD, it reduces the variance of total supply (the "portfolio effect"). Second, through SC, it reduces the quality of the information signal, making firms less certain about their competitors and thus less likely to withhold production.
- The Profit Paradox: While social welfare always increases with dispersion, producer profits do not. If the "low state" of wind is relatively high, firms actually prefer to be close together (low $d$) to maintain their market power. This implies that private investment might naturally favor "clustering" wind farms in a way that is socially sub-optimal.
- Forecasting Contradiction: The paper shows that public sharing of high-quality weather forecasts is always socially beneficial. However, it can reduce producer profits by eliminating the information advantage firms hold, meaning strategic producers might actively oppose transparency in forecasting.
Key Results
- Welfare Gains: Social welfare is a monotonically increasing function of dispersion ($d$). More "spread out" wind is always better for the public.
- Price Impacts: In a linear demand model, increasing dispersion consistently decreases the expected market price.
- Threshold Effects: The authors identify specific thresholds for "low state" production ($L$). If $L$ is small, firms want to be far apart to maximize profit; if $L$ is large, they prefer to be close together to exercise market power.
- Collusion: The model suggests that wind producers will always choose to collude if there are no penalties, as sharing information about their stochastic states allows them to achieve monopoly-level profits more effectively.
Limitations and Future Work
The authors acknowledge that their model is "parsimonious," meaning it simplifies reality to highlight specific mechanisms.
- Two-State Model: Real-world wind availability is continuous, not just "High" or "Low."
- Market Complexity: The study focuses on real-time spot markets. Future work could investigate how these strategic incentives play out in day-ahead markets or capacity auctions.
- Transmission Constraints: The model assumes a single market node. In reality, transmission congestion often dictates where wind can be "sold," which could either amplify or mitigate the market power effects described here.
- Policy Interaction: A promising direction for future research is how specific subsidies (like the Production Tax Credit) influence the geographic investment choices of firms in light of these strategic incentives.
Summary updated