Because of this, I'm very interested in determining what influences the movement of the stock price and whether the downward move that the company has experienced recently was caused by outside factors (i.e. changes in commodities prices, the weakening economy) or internal factors such as market distaste with the company or other problem areas that have yet to crop up in tangible news.

Blurbs that I've read, such as Bob Pisani's Stock Talk from March 31st, 2010, point to natural gas as being a big mover for utility companies like Mirant. Commenting on the movement in independent power producers, Pisani notes:

"Natural gas happened. Independent power producers usually work in regulated markets, where the price they can charge is often tied to natural gas prices. Nat gas went from $6 to $4…a disastrous impact on profits, since fixed costs did not change."

So I thought I'd take Pisani's hypothesis to the data and see what the data had to say. I performed a linear regression in STATA to try and see what connection daily returns in natural gas had on energy producers such as Mirant (MIR), AES Corporation (AES), Dynegy (DYN), RRI Energy (RRI) and NRG Energy (NRG).

The short answer: except for Mirant, natural gas has no statistically significant explanatory power for the companies' stock prices for the time period considered. Even the explanatory power of natural gas for Mirant becomes statistically insignificant when a proxy for market performance (i.e. the S&P 500) is added in.

What I Did

To begin, I used Yahoo! Finance to collect historical daily stock price data for all of the companies mentioned above and for the S&P 500 index. The West Texas Natural Gas Wellhead price was used for natural gas and this data was obtained from GFD (Global Financial Data). For further testing, I also used the WTI Crude price, GSCI Energy Index, Moody's Commodity Index and Dow Jones U.S. Electricity index, all obtained from GFD.

All data was converted to percent return, and to try and control for changing correlation and dependency in the data I only looked at the time elapsed since the beginning of 2009 through the 9th of April 2010. The cut off at Apirl 9th was done to correct for the fact that Mirant and RRI just entered in to a merger agreement.

For the first step, I simply performed a linear regression of the daily percent change in natural gas prices on the daily percent change in the company stock price. Below please find the adjusted R-squared values, coefficients on the natural gas variable and corresponding t statistics from my regressions.

To put the coefficient in to perspective, for Mirant, if there is a 1% positive change in natural gas prices we would expect the stock price to go up by 0.04%.

As we can see above, in AES, Dynegy and RRI adding in natural gas rendered a negative adjusted R-squared value, meaning that it'd be better to guess randomly than look to natural gas as an explanatory variable. It is very easy, from the results shown above, to see that natural gas clearly does not have an impact on utility stock prices for the time period considered.

From a Longer Time Horizon

When I used data spanning back to 2006, the answer started to change. Natural gas became statistically significant for all companies except for AES, and this was even when market performance (i.e. the S&P 500) was added in. The negative coefficient of natural gas in the Dynegy instance is somewhat strange, but the coefficient is so low one can essentially assume it is zero. In fact, in all of the companies the impact of natural gas on stock price was far dwarfed by the S&P 500 variable, in all cases by at least a degree of magnitude.

The conclusion from this study would be that while over the long term utility stock returns are more generally tied to natural gas movements this effect absolutely does not exist on a shorter time frame (i.e. 1-1.5 years going back). This could be because most utilities today hedge for inputs, so they probably are not sweating the day to day movement. These hedges typically go out a year in advance, which could explain why data periods over longer time periods, which can take in to account longer trends, show more significance.

The extremely low values on the coefficients for natural gas, in both the longer and shorter term analysis, suggest that Pisani is mistaken is his connection between natural gas prices and stock price moves of power producers.

Whoops.

Blurbs that I've read, such as Bob Pisani's Stock Talk from March 31st, 2010, point to natural gas as being a big mover for utility companies like Mirant. Commenting on the movement in independent power producers, Pisani notes:

"Natural gas happened. Independent power producers usually work in regulated markets, where the price they can charge is often tied to natural gas prices. Nat gas went from $6 to $4…a disastrous impact on profits, since fixed costs did not change."

So I thought I'd take Pisani's hypothesis to the data and see what the data had to say. I performed a linear regression in STATA to try and see what connection daily returns in natural gas had on energy producers such as Mirant (MIR), AES Corporation (AES), Dynegy (DYN), RRI Energy (RRI) and NRG Energy (NRG).

The short answer: except for Mirant, natural gas has no statistically significant explanatory power for the companies' stock prices for the time period considered. Even the explanatory power of natural gas for Mirant becomes statistically insignificant when a proxy for market performance (i.e. the S&P 500) is added in.

What I Did

To begin, I used Yahoo! Finance to collect historical daily stock price data for all of the companies mentioned above and for the S&P 500 index. The West Texas Natural Gas Wellhead price was used for natural gas and this data was obtained from GFD (Global Financial Data). For further testing, I also used the WTI Crude price, GSCI Energy Index, Moody's Commodity Index and Dow Jones U.S. Electricity index, all obtained from GFD.

All data was converted to percent return, and to try and control for changing correlation and dependency in the data I only looked at the time elapsed since the beginning of 2009 through the 9th of April 2010. The cut off at Apirl 9th was done to correct for the fact that Mirant and RRI just entered in to a merger agreement.

For the first step, I simply performed a linear regression of the daily percent change in natural gas prices on the daily percent change in the company stock price. Below please find the adjusted R-squared values, coefficients on the natural gas variable and corresponding t statistics from my regressions.

To put the coefficient in to perspective, for Mirant, if there is a 1% positive change in natural gas prices we would expect the stock price to go up by 0.04%.

As we can see above, in AES, Dynegy and RRI adding in natural gas rendered a negative adjusted R-squared value, meaning that it'd be better to guess randomly than look to natural gas as an explanatory variable. It is very easy, from the results shown above, to see that natural gas clearly does not have an impact on utility stock prices for the time period considered.

From a Longer Time Horizon

When I used data spanning back to 2006, the answer started to change. Natural gas became statistically significant for all companies except for AES, and this was even when market performance (i.e. the S&P 500) was added in. The negative coefficient of natural gas in the Dynegy instance is somewhat strange, but the coefficient is so low one can essentially assume it is zero. In fact, in all of the companies the impact of natural gas on stock price was far dwarfed by the S&P 500 variable, in all cases by at least a degree of magnitude.

The conclusion from this study would be that while over the long term utility stock returns are more generally tied to natural gas movements this effect absolutely does not exist on a shorter time frame (i.e. 1-1.5 years going back). This could be because most utilities today hedge for inputs, so they probably are not sweating the day to day movement. These hedges typically go out a year in advance, which could explain why data periods over longer time periods, which can take in to account longer trends, show more significance.

The extremely low values on the coefficients for natural gas, in both the longer and shorter term analysis, suggest that Pisani is mistaken is his connection between natural gas prices and stock price moves of power producers.

Whoops.

## No comments:

## Post a Comment