Detection of multiple bubbles in South African electricity prices

ABSTRACT

This paper investigates the existence and dating of electricity price bubbles in South Africa from 1965 to 2013. In the literature, it is agreed that such a task is difficult due to the explosive nature of price bubbles and labeling their presence’s occurrence. To overcome the predicament, the authors followed the methodological approach suggested by Phillips et al. (2013): a recursive right-tailed Generalized Sup Augmented Dickey-Fuller (GSADF). Two significant bubbles were detected in the output-adjusted nominal prices: the first one was a long one from 1971 to 1998 and can be attributed to the monopolistic unregulated nature of the electricity market at the time, while the second one lasted for a shorter period of time (2008–2009) coinciding with the severe supply crisis of 2008, and the massive price hikes that followed it.

KEYWORDS:

• Bubbles
• electricity prices
• GSADF
• South Africa

Notes

¹ If we used a 10% critical value, instead of the 5%, results were similar, except that now a third bubble was detected for the year 2011, under the case of both the output-adjusted nominal electricity prices and the nominal electricity prices. We also tested the robustness of our results based on an alternative size of the initial window, also suggested in Phillips, Shi and Yu (2013), based on the following formula: (0.01 + 1.8/ (T)^1/2)*T, with T being the sample size. Now 13 observations were lost for the initial estimation. For the case of the output-adjusted nominal electricity prices as well as nominal electricity prices, only one major bubble is detected. For the output-adjusted nominal electricity prices, the bubble originated in 1977 and collapsed in 1998 or 1999 depending on whether we used the 5% or 10% critical value. For nominal electricity prices, irrespective of whether we use a 5% or 10% critical value, the bubble started in 1977 and collapsed in 2000. However, as in Phillips, Shi and Yu (2013), our preferred initial window of estimation is 10% of the observations, because the results obtained under this case, especially in terms of the second bubble, are in line with the history and events related to the South African electricity market. All the results discussed in this footnote are available upon request from the authors.

² Note that we work with natural logarithmic values of the nominal electricity prices.