Daylight time and energy: Evidence from an Australian experiment

https://doi.org/10.1016/j.jeem.2008.02.003Get rights and content

Abstract

Several countries are considering using daylight saving time (DST) as a tool for energy conservation and reduction of greenhouse gas emissions, and the United States extended DST in 2007 with the goal of reducing electricity consumption. This paper assesses DST's impact on electricity demand by examining a quasi-experiment in which parts of Australia extended DST in 2000 to facilitate the Sydney Olympics. Using detailed panel data and a difference-in-difference-in-difference framework, we show that the extension did not reduce overall electricity consumption, but did cause a substantial intraday shift in demand consistent with activity patterns that are tied to the clock rather than sunrise and sunset.

Introduction

One principal socio-economic problem is the optimal allocation of individuals’ activities—sleep, work, and leisure—over 24 h of the day. In today's world of artificial lighting and heating, people set their active hours by the clock rather than by the natural cycle of dawn and dusk. In one of the earliest statistical treatments in economics, “An Economical Project,” Benjamin Franklin [9] observes that this behavior wastes valuable morning daylight as people sleep until long after sunrise, and requires expensive candles to illuminate the night. Franklin calculates that this misallocation causes Paris to consume an additional 64 million pounds of tallow and wax annually.

Governments have since attempted to address this resource allocation problem through the mechanism of daylight saving time (DST).1 Each year, we move our clocks forward by 1 h in the spring, and adjust them back to Standard Time in the fall. The intuition behind this DST adjustment relies on the premise that people's activities shift forward with the clock, so that, during the summer, the sun appears to set 1 h later and the “extra” hour of evening daylight cuts electricity demand.

Today, heightened concerns regarding energy prices and greenhouse gas (GHG) emissions are driving interest in extending DST in several countries, including Australia, Canada, Japan, New Zealand, and the United Kingdom [17], [22], [24], [25], [26]. The United States has already passed legislation to extend DST by 1 month, beginning in 2007, with the specific goal of reducing electricity consumption by 1% during the extension [8]. Therefore, the United States now switches to DST in March rather than in April. The energy legislation also explicitly calls for research into the energy impacts of extending DST, and suggests reverting to the prior DST system if it is demonstrated that the policy goal will not be achieved. Beyond this federal initiative, California is considering even more drastic changes—year-round DST and double DST—that are predicted to save up to 1.3 billion US dollars annually [5].

Our study challenges the energy conservation predictions that have been used to justify these calls for the expansion of DST. Across the studies and reports we surveyed, estimates of an extension's effect on total electricity demand range from savings of 0.6% to 3.5%. The most widely cited savings estimate of 1% is based on an examination conducted over 30 years ago [27]. Arguably, these findings are not applicable today. For example, the widespread adoption of air conditioning has altered intraday patterns of electricity consumption. Further, the 1% savings estimate may be confounded by other energy conservation measures enacted during the oil crisis.

More recent efforts to predict the effect of extending DST on electricity demand employ simulation models, which use data from the status quo DST system to forecast electricity use under an extension. One prominent recent study is being used to argue in favor of year-round DST in California [4]. It predicts three benefits of an extension: (1) a 0.6% reduction in electricity consumption, (2) lower electricity prices, driven by a reduction in peak demand, and (3) a lower likelihood of rolling blackouts. However, this study is not based on firm empirical evidence; it instead uses electricity consumption data under the current DST scheme to simulate demand under extended DST. It may therefore fail to capture the full behavioral response to a change in DST timing.2

Our study obviates the need to rely on simulations by examining actual data from a quasi-experiment that occurred in Australia in 2000. Typically, three of Australia's six states observe DST beginning in October (which is seasonally equivalent to April in the northern hemisphere). However, to facilitate the 2000 Olympics in Sydney, two of these three states began DST 2 months earlier than usual. Because the Olympics can directly affect electricity demand, we focus on the state of Victoria—which extended DST but did not host Olympic events—as the treated state, and use its neighboring state, South Australia, which did not extend DST, as a control. We also drop the 2-week Olympic period from the 2-month treatment period to further remove confounding effects. Using a detailed panel of half-hourly electricity consumption and prices over 7 years, as well as the most detailed weather information available, we examine how the DST extension affected electricity demand in Victoria.

Our treatment effect estimation strategy is based on the difference-in-difference (DD) framework that exploits, in both the treatment state and the control state, the difference in demand between the treatment year and the control years. We augment the standard DD model to take advantage of the fact that DST does not affect electricity demand in the mid-day. This allows us to use changes in mid-day consumption to control for unobserved state-specific shocks via a difference-in-difference-in-difference (DDD) specification. We show that this technique allows us to employ a mild identifying assumption that is more appropriate for the data than that of a standard DD model.

Our results confirm policy-makers’ expectations that the extension of DST causes electricity demand to decrease significantly in the evening. However, we also find an opposing effect in the morning: the Australian extension significantly increased electricity consumption between 07:00 and 08:00. Overall, the evening decrease in demand did not outweigh the morning increase, so that total electricity consumption in Australia was not reduced as a result of the DST extension. These effects are consistent with waking and sleeping behaviors that are tied to the clock, rather than to sunrise and sunset. In particular, the residents of Australia do not appear to have substantially altered the clock time at which they awoke following the extension of DST; they therefore rose before sunrise and needed electric power for lighting.

These results contradict the claims made by prior studies that extending DST will conserve energy, and indicate that proposals in Australia to extend DST permanently are unlikely to reduce energy use and GHG emissions. Furthermore, the morning peak demand caused by Australia's 2000 extension is associated with significantly higher wholesale electricity prices, indicating that the steep morning ramp-up in demand likely caused an increase in generation costs. This outcome undercuts claims that extending DST leads to generation efficiencies by smoothing the hourly demand profile.

While we cannot directly apply our results to other countries without adjustments for behavioral and climatic differences, this study raises concern that the recent DST extension in the United States is unlikely to result in energy conservation. To investigate the degree to which our results extend to the US, we reconstruct the simulation model that was used to forecast energy savings for California [4], and apply it to the Australian data. Noting that Victoria's latitude and climate are similar to those of central California, we find that the simulation systematically overstates energy savings in both the morning and evening, casting further suspicion on claims that extending DST in California and the rest of the United States will reduce electricity consumption.

Section snippets

Background on daylight saving time in Australia

The geographical area of interest is the southeastern part of Australia. Three states in the southeast of the mainland observe DST: South Australia (SA), New South Wales (NSW), and Victoria (VIC). DST typically starts on the last Sunday in October and ends on the last Sunday in March. Queensland, the Northern Territory, and Western Australia do not observe DST. Table 1 provides summary statistics and geographical information for the capitals of these states, where the populations and

Data

Our study uses detailed electricity consumption and wholesale price panel data, obtained from Australia's National Electricity Market Management Company Limited (NEMMCO).5 These consist of half-hourly electricity demand and wholesale prices by state from 13 December 1998 to 31 December 2005. Wholesale prices are market prices paid by utilities to generators, while end-use customers instead pay a regulated price for

Identification

While we have noted that the DST extension was implemented solely to facilitate the Olympic Games, and that we are not aware of any energy-based justifications for it, identification of the extension's effect on energy use is made difficult by the presence of potentially confounding factors. In particular, there are reasons to suspect that the Olympics may have changed electricity consumption in Australia significantly, even absent a DST extension. The 2000 Games were the most heavily visited

Reference case results

We use two separate methods to obtain the overall effect of the DST extension. First, we estimate a version of (1) in which the treatment variables are pooled overall half-hours of the day. These results indicate that the extension of DST did not significantly affect overall electricity consumption in VIC in 2000. Our point estimate indicates a 0.02% increase in consumption due to the extension, with a clustered standard error of 0.43.

Second, we estimate a separate treatment effect for each

Evaluation of the simulation technique

It is natural to ask whether the simulation technique used in [4] to predict energy savings in California would have accurately predicted the outcome of the Australian DST extension. A successful validation would lend credence to the model's results in California, and suggest that California may experience reduced energy use due to an extension, even if Australia did not.

The simulation approach uses data on hourly electricity consumption under the status quo DST policy to investigate the impact

Conclusions

Given the economic and environmental imperatives driving efforts to reduce energy consumption, policy-makers in several countries are considering extending daylight saving time (DST), as doing so is widely believed to reduce electricity use. Our research challenges this belief, as well as the studies underlying it. We offer a new test of whether extending DST decreases energy consumption by evaluating an extension that occurred in the state of Victoria, Australia, in 2000. Using half-hourly

Acknowledgments

We are indebted to Michael Anderson, Maximilian Auffhammer, Severin Borenstein, Jennifer Brown, Kenneth Chay, Michael Hanemann, Ann Harrison, Guido Imbens, Enrico Moretti, Jeffrey Perloff, Susan Stratton, Muzhe Yang, David Zilberman, and two anonymous referees for valuable discussions and suggestions. We further thank Alison McDonald from NEMMCO and Lesley Rowland from the Australian Bureau of Meteorology for helping us understand the electricity and weather data, and Adrienne Kandel from the

References (30)

  • B.A. Rock

    Impact of daylight saving time on residential energy consumption and cost

    Energy and Buildings

    (1997)
  • M. Arellano

    Computing robust standard errors for within-groups estimators

    Oxford Bull. Econ. Stat.

    (1987)
  • M. Bertrand et al.

    How much should we trust differences-in-differences estimates?

    Quart. J. Econ.

    (2004)
  • J. Bushnell et al.

    Consumption under noisy price signals: a study of electricity retail rate deregulation in San Diego

    J. Ind. Econ.

    (2005)
  • California Energy Commission (CEC), Effects of daylight saving time on California electricity use, Report P400-01-013,...
  • California Joint Senate Resolution, Bill No. SJRX2-1, 25 June...
  • M. Downing

    Spring Forward: The Annual Madness of Daylight Saving Time

    (2005)
  • Emergency Daylight Savings Time Energy Conservation Act, US Public Law 93-182, December...
  • Energy Policy Act of 2005, US Public Law 109-58, July...
  • B. Franklin

    An economical project, essay on daylight saving

    J. Paris

    (April 1784)
  • J. Gruber

    The incidence of mandated maternity benefits

    Am. Econ. Rev.

    (1994)
  • D.S. Hamermesh, C.K. Myers, M.L. Pocock, Time zones as cues for coordination: latitude, longitude and letterman, NBER...
  • S.P. Holland et al.

    Is real-time pricing green?: The environmental impacts of electricity demand variance

    Rev. Econ. Stat.

    (2008)
  • M.J. Kamstra et al.

    Losing sleep at the market: the daylight saving anomaly

    Am. Econ. Rev.

    (2000)
  • B.D. Meyer

    Natural and quasi-experiments in economics

    J. Bus. Econ. Stat.

    (1995)
  • Cited by (0)

    View full text