**2014 IAEE Best Student Paper Award Winner**

** **

**Electricity Market Price Volatility:**

**The Importance of Ramping Costs**

Within the past fifteen years, most electricity markets across the United States have restructured to allow competition in the generation of electricity. Since inception, high price volatility has plagued wholesale electricity prices, creating major implications for risk-averse market participants and system operators tasked with maintaining grid reliability. The expected rise in intermittent renewable generation has augmented these volatility concerns, since various generator types may affect volatility differently.

The mainstream view is that high price volatility within electricity markets is due to the lack of hourly retail pricing in combination with the lack of cost-effective electricity storage mechanisms. Although daily demand follows a relatively predictable pattern, hourly prices vary greatly throughout the day because electricity cannot be directly stored to stabilize prices. However, the inputs of electricity can be stored in conventional generators (e.g. coal, natural gas, etc), which suggests that dispatchable generators can stabilize prices if generators have low “ramping costs” (costs of adjusting output) and adequate capacity exists.

The subsequent analysis seeks to understand the importance of generator type in price volatility, providing the first rigorous empirical analysis of related questions. More specifically, what is the impact of additional natural gas capacity on electricity price stability and how does this compare to inflexible capacity such as nuclear? Finally, what is the value of such volatility reductions to power purchasers and how does this change with the rise of intermittent renewable generators?

First I developed a formal theoretical model of competitive generation with ramping costs that suggests there is both a “supply shift effect” and “ramping cost effect” from adding more flexible natural gas capacity. The outward shift in the supply curve should yield a decrease in intra-day price variance because demand intersects on a flatter supply curve convexity. The second effect is the decrease in ramping costs which squeezes together the intra-day dynamic supply curve shifts, which also yields a decrease in intra-day price variance.

It is possible to separate out these two effects using capacity changes that only affect volatility through outward supply curve shifts. For example, nuclear power faces relatively low marginal costs in addition to binding ramping constraints on the technology. For this reason, it is often used as a base-load power source and occupies the left most region of the supply curve in addition to renewable generators that have zero marginal cost. With this information it seems reasonable to assume that nuclear power outages will only shift the supply curve inward, without changing the intra-day dynamics involved from ramping costs. Meanwhile, natural gas generators will have both a “supply shift effect” and “ramping cost effect”, allowing for greater price stabilization when compared to inflexible generation.

To empirically explore these issues, I took advantage of high-frequency wholesale electricity price data at the hourly level in New England, which were collapsed into daily observations with mean and variance. The baseline econometric model used a pooled event study approach, which assumed exogenous capacity additions within the event window. However, natural gas capacity is arguably endogenous with electricity price and intra-day price variance because it is the marginal generating unit. To correct for any endogeneity-induced bias, I also performed a two-stage least squares (2SLS) regression to instrument for natural gas capacity using a 31-day rolling average of the “spark spread”, lagged by 24 months. The spark spread is the gross margin between electricity price and the cost of generation using natural gas, which provides a measure of natural gas generator profitability and drives capacity investment decisions. Finally, a third specification used a generalized autoregressive conditional heteroskedasticity (GARCH) model, which has been used in the literature on electricity price behavior.

The regression results showed that increasing natural gas capacity significantly reduced price volatility more than nuclear capacity. For example, the average gas generator in my sample decreased intra-day price volatility by 5.6%, compared to only 0.7% for the equivalent amount of nuclear capacity. Thus, empirically it appears that the reduction of volatility from the “supply shift effect” is actually quite small compared the “ramping cost effect”, although it is still statistically significant. The bulk of the volatility reduction came through supply flexibility via decreased ramping costs, as this represents a more effective indirect storage of electricity.

Lastly, the coefficient estimates for volatility reductions were used as inputs for the Black-Scholes asset pricing model to quantify the value of such volatility reductions due to increased natural gas capacity. The average natural gas generator addition is estimated to decrease the price of the option by approximately 4%. Adding the assumption that power purchasers fully hedge away from spot price risk suggests annual savings of $1.13 million each year for the total New England market, or 1.6% of recent natural gas generator construction costs. Finally, a simulation explored how these values change over time in the presence increasing wind generation. The results suggest that gas generators provide a natural counterbalance to the expected volatility increases from wind generation.

My analysis provides several contributions to the existing literature on electricity markets, as it describes and quantifies the additional benefits from adding flexible and dispatchable generation capacity. In the absence of cost-effective storage, ramping costs are a major contributor to price volatility in the electricity market. The theoretical model showed that adding generation capacity with lower ramping costs and lower marginal costs will unambiguously decrease intra-day price volatility. Further, the implications of the model easily generalize to all non-storable, or perishable, commodities where there are marginal costs of adjusting output. In brief, flexible production can serve a similar role to storage in ensuring price stability when the inputs can be stored.

Taken together, the results of this analysis point electricity market regulators towards specific policies. First, market design and policies should acknowledge that there are additional benefits around adding capacity that has both low ramping costs and low marginal costs, such as natural gas generators. The pecuniary externality from different generation types has real value that impact consumer welfare. This is increasingly important when considering the future growth of non-dispatchable generators such as wind. To the extent that price stability benefits are not priced under the current design of electricity markets, incentives should be created to ensure such benefits are internalized into long-run capital investment decisions. However, any policy incentives for specific generator types must also consider current research on environmental externalities that need be weighed against the benefits described in this analysis.