Coming of the Age of Volatility

May 12, 2011, Denver, Colorado -

Out of the battlefields and playgrounds

The time when the concept of volatility was reserved exclusively for academic battlefields and “big-boy” financial playgrounds is over. Currently, utilities, energy companies and other industry players have realized that to meet the realities of estimating and managing their own market risks associated with resource and portfolio management they need to understand and get comfortable with the concept of volatility. There are of course, other users of volatilities – the most obvious group being option speculators – but they are not a subject of this paper.

Estimating market risk is not an easy task. Energy prices are notorious for their ability to reach stratospheric values (who remembers $20,000/MWh or $50/GJ or dreams of $200/bbl oil). For electricity, the often volatile nature of underlying fuel prices, e.g. natural gas, oil, coal, further complicates the ability to estimate market risk. Electricity cannot be stored, and for many customers, electricity demand is not elastic. This, of course, is just a simplified rational for the occasional price jump which can be so high that it can only be described by extreme value theory. Other factors like hourly, weekly, and annual demand profiles, weather, transmission constraints, and equipment outages also contribute to complex price behavior.

Many concepts, theories, and models have been developed for market risk estimation and management purposes – the majority of which, however, are overly complex and impractical. All risk analysis models have in common two very important but relatively simple parameters: volatility and correlation.

So, what exactly is volatility?

Volatility measures how fast market price changes, therefore, allowing quantification of price changes for different time horizons. The quantification is performed on a probabilistic basis, e.g. there is 95% probability that the electricity price will stay below $80/MWh next May.

So, who needs volatility measures?

Volatility measures are important to those who must commit resources, monetary or otherwise, forward in time; or who, by today’s actions, subject themselves to future risk. Examples of the application of volatility measures can be found in:

One can consider some real life examples based on sample client projects engaged by Zema PowerGroup’s consulting division:

Example one: long-term contract premium. Suppose a generator is long on electricity and proposes to enter into a five-year fixed-price sales contract with an attached price-acceleration clause. How would the generator compare this contract with the alternative opportunity to sell directly into an open market? How would the generator quantify the market risk? Or can the generator estimate the premium that should be embedded into the fixed-price sales contract in comparison to the long-term levelized market forecast? The task is impossible without application of price volatility.

Example two: investment risk evaluation. Suppose the generator is estimating a proposed investment into a hydro generation facility. The proposed generation profile makes the investment questionable; however, under the expected electricity price scenario the investment seems reasonable. To estimate the risk, the generator must determine the probability that the price levels will be insufficient to allow the desired rate of return within the desired time horizon. Data on the volatility of future market prices is needed to determine the probability of meeting the investment objectives.

Example three: risk hedging. A generator owns a natural-gas-fired generation plant. The generator has a customer who wants price stability for the electricity being purchased (fixed price). Naturally, the generator would want to manage/hedge the natural gas exposure when procuring the needed natural gas supply for the term of the agreement. The generator is wondering if it has to hedge the cost of natural gas for all months. One option is for the generator to hedge periods when the market prices are the most volatile (hedge the maximum risk); alternatively, the generator’s option is to minimize the cost of hedging by paying the minimal forward premiums. The choice of either option will be partially based on market volatilities.

Volatility is also a good indicator of market uncertainty at the current price level. A sudden increase in volatility reflects market uneasiness about the current price level and the fundamentals supporting the current price. It is with good reason that the implied volatility index is also known as the “panic index”.

What volatility knowledge is required to deal confidently with market risk?

The examples above do not require minute-to-minute real-time estimations of volatility; nor do they require detailed understanding of complex theories and manipulations with strange-looking mathematics. Still, one really does need to have a good feeling for the current market volatility levels, volatility trends, the mean-reversion and resistance levels and the seasonal pattern of volatility. Volatility needs to be observed over time to develop a good understanding of its behavior; the principal issues are:

What is the difference between historical and implied volatility?

Historical volatility is measured from the movement of past market prices; hence it represents a measure of historical market behavior. Implied volatility is measured/extracted from current options prices and reflects today’s market perception of the future. Although, in perfectly efficient markets, historical and implied volatility theoretically converge, in reality they are often quite far apart. Both types of volatility are used in the energy industry in order to:

Measurement of Volatility There are a multitude of methods that can be used for measurement of volatilities; both historical and implied. These methods often require different types of data. For example, the classical log-return method of measuring historical volatility requires only daily average price, daily settlement price, or daily closing price. On the other hand, the Parkinson method requires both daily high and low prices. The more complex Garman-Klass method needs daily open, high, low, and close prices. The same choices exist for implied volatility measurement. For example, one can use either the Black-Scholes or Heston model, depending on one’s view of the behavior of underlying assets. In classic theory, exemplified by Black-Scholes option pricing, volatility is assumed as a fixed number. In reality, volatility is a complex stochastic variable which exhibits properties of mean-reversion, seasonality, and clustering, as illustrated by the graph of historical volatility variation shown below:

• Volatility smile – for implied volatility vs. strike price
• Volatility term – for implied volatility vs. strike date
• Volatility surface – for combination of smile and term analysis
• Volatility cone – for historical volatility distributions