Energy Efficiency: What Has It Delivered in the Last 40 Years?

I'm one of nineteen authors of a new review of energy efficiency economics. It was commissioned for the Annual Review of Environment and Resources, where it is still in (second-round) review. The team was put together and led by Harry Saunders and Joyashree Roy.

Over the past four decades different disciplinary approaches independently adopted different definitions of energy efficiency to answer specific problems. Even within economics there are at least three different ideas of energy efficiency. Technical efficiency in economics compares the quantity of inputs used to produce given outputs (or vice versa) to the best practice or frontier level. This is a relative measure of energy efficiency. But economists often talk about energy efficiency in absolute terms too,  measured as either simply an increase in energy services per unit input or using the concept of energy augmenting technological change where the amounts of other inputs and the technology associated with them are held constant. Energy augmenting technological change is usually used when modeling economy-wide rebound, whereas the energy services per unit input might be used when investigating the energy efficiency gap.

The energy intensity of economies (a metric measuring energy consumption per unit of GDP), which is often interpreted as a proxy for energy efficiency, has trended downwards (increasing efficiency) globally and in many major economies over the last century. But as panel (a) below shows, in many regions of the world, especially poorer or hotter regions, energy intensity instead increased. Today, energy intensity is more similar around the world than in the past.

Innovation in energy-saving technologies is an important driver in improving aggregate energy efficiency deployment by lowering costs and inducing adoption. The productivity of numerous energy-using products has improved dramatically. (e.g., lighting had a 10,000-fold improvement in lumens/Watt since the start of the industrial revolution). 

Energy efficiency improvements, in themselves, generally increase economic welfare. But when we consider negative externalities, such as pollution emissions, welfare effects are more ambiguous. Interventions such as improperly calibrated subsidies to improve energy efficiency or mandates to use costly technologies can lead to a reduction in household welfare.

There is still uncertainty and difficulty in measuring economy-wide rebound effects. Rebound may limit the ability to reduce or constrain overall energy use. In general, it makes more sense to address the environmental impacts of energy use with specific environmental policies rather than trying to reduce energy use with energy efficiency policies.

The contribution of different factors to the persistent “energy efficiency gap”, i.e., the difference between the energy consumption observed and the potential energy consumption levels that would result from the adoption of cost-efficient energy efficient technologies and strategies, is still not well understood. Market and regulatory failures, departure of consumer behavior from rational choice theory, lack of information, the principal-agent problem, among other issues may all contribute to the energy efficiency gap.

Policy interventions aimed at overcoming or reducing barriers to energy efficiency deployment target behavioral anomalies and perceived market failures. They include provision of feedback to energy users, the use of social norms, commitment devices, rewards and regulatory mechanisms such as taxes, subsidies, building codes, etc. The literature and evidence are mixed on the effectiveness of each of these, but all seem to show promise to some degree. 

Methodological advances for examining energy efficiency effects on energy use have been substantial. Primary advances include randomized control trials coupled with appropriate econometric methods, developments in econometric methods and lab/field experiments, agent-based modeling formulations, general equilibrium methods, and behavioral science. 

The following diagram summarizes the state of knowledge across different scales and the needed scope of future research:

Future research should bring together researchers from different fields to shed new light onto energy efficiency questions. Examples of such endeavors include: (i) at the micro-level, a better understanding of consumer choice and behavior by combining insights from engineering and the advanced metering and sensing infrastructure, with those from micro-economic theory as well as the theory of choice and with behavioral economists’ models; (ii) at the program evaluation level, there is a need to continue to develop methods to understand causal inferences using econometrics as well as machine learning to better understand program outcomes; (iii) at the macro-level, developing flexible and credible general equilibrium models that also capture environmental and climate externalities outcomes, and that have good input data to enable us to understand the dynamics of energy efficiency improvements across the economy, the environment, and society, are needed.

How Large is the Economy-Wide Rebound Effect?

Last year, I published a blogpost about our research on the economy-wide rebound effect. The post covers the basics of what the rebound effect is and presents our results. We found that energy efficiency improvements do not save energy. In other words, the rebound effect is 100%. This doesn't mean that improving energy efficiency is a bad thing. It's a good thing, because consumers get more energy services as a result. But it probably doesn't help the environment very much.

I now have a new CAMA working paper, which surveys the literature on this question. Contributions to the literature are broadly theoretical or quantitative. Theory provides some guidance on the factors affecting rebound but does not impose much constraint on the range of possible responses. There aren't very many econometric studies. Most quantitative studies are either calculations using previously estimated parameters and variables or simulations.

Theory shows that the more substitutable other inputs are for energy in production the greater the rebound effect. This means that demand for energy services by producers is more elastic and so reducing the unit costs of energy services increases the amount used by more.

The most comprehensive theoretical examination of the question is Derek Lemoine's new paper in the European Economic Review: "General Equilibrium Rebound from Energy Efficiency Innovation." Lemoine provides the first mathematically consistent analysis of general equilibrium rebound, where all prices across the economy can adjust to a change in energy efficiency in a specific production sector. He shows that the elasticity of substitution in consumption plays the same role as the elasticity of substitution in production: the greater the elasticity, the greater the rebound, ceteris paribus.

Beyond that, the predictions of the model depend on parameter values. The most likely case, assuming a weak response labor to changes in the wage rate, is that the general equilibrium effects increase energy use relative to the partial equilibrium direct rebound effect for energy intensive sectors and reduce it for labor intensive sectors.

Lemoine uses his framework and previously estimated elasticities and other parameters to compute the rebound to an economy-wide energy efficiency improvement in the US. The result is 38%. There are two main reasons why the real rebound might be higher than this. First, most of the elasticities of substitution in production that he uses are quite low because of how they were estimated. Second, an energy efficiency improvement in any sector apart from the energy supply sector does not trigger a fall in the price of energy. A fall in the price of energy would boost rebound. This is because there are no fixed inputs and there are constant returns to scale in energy production.

There are similar issues with simulations from computable general equilibrium models (CGE). The assumptions that modellers make and the parameter values they choose make a huge difference to the results. Depending on these choices, any result from super-conservation, where more energy is saved than the energy efficiency improvement alone would save, to backfire, where energy use increases, is possible.

Rausch and Schwerin estimate the rebound using a small general equilibrium model calibrated to US data. This is somewhere between the typical CGE model and econometric models. They use the putty-clay approach to measuring and modeling energy efficiency. Increases in the price of energy relative to capital are 100% translated into improvements in the energy efficiency of new capital equipment. Once capital is installed, energy and capital must be used in fixed proportions. Rebound in this model depends on why the relative price changes. If the price of energy rises, energy use falls. However, if the price of capital falls energy use increases. These are very strong assumptions, which determine how the data are interpreted. Are they realistic? Rausch and Schwerin find that historically rebound has been around 100% in the US.

Historical evidence also hints that the economy-wide rebound effect could be near 100%. Energy intensity in developing countries today isn't lower than it was in the developed countries when they were at the same level of income. This is despite huge gains in energy efficiency in all kinds of technologies from lighting to car engines. This makes sense if consumers have shifted to more energy intensive consumption goods and services over time. Commuters and tourists on trains in the 19th and early 20th centuries have been replaced by commuters and tourists in cars and on planes in the late 20th and early 21st centuries.

I only found three fully empirical econometric analyses. One of them is our own paper. The others are by Adetutu et al. (2016) and Orea et al. (2015). Both use stochastic production frontiers to estimate energy efficiency. This is a potentially promising approach. Adetutu et al. then model the effect of this energy efficiency one energy use, using an autoregressive model. This includes the lagged value of energy use as an explanatory variable, which means that the long-run effect of all variables is greater in absolute value than the short-run effect. As in the short run, energy efficiency reduces energy use, in the long run it reduces it even more. The result is super-conservation even though short-run rebound is 90%. In Orea et al.'s model, the purely stochastic inefficiency term is multiplied by [1-R(γ'z)] where z is a vector of variables including GDP per capita, the price of energy, and average household size. R(γ'z) is then supposed to be an estimate of the rebound effect. But really this is just a reformulation of the inefficiency term – nothing specifically identifies R(γ'z) as the rebound effect.

In conclusion, the economy-wide rebound effect might be near 100%. But I wouldn't describe the evidence as conclusive. Both our research and the historical investigations might be missing some important factor that has moved energy use in a way that makes us think it is due to changes in energy efficiency, and Rausch and Schwerin make very strong assumptions about analysing the data.

Energy Efficiency Improvements Do Not Save Energy

I have a new working paper out, coauthored with Stephan Bruns and Alessio Moneta, titled: "Macroeconomic Time-Series Evidence That Energy Efficiency Improvements Do Not Save Energy". It's another paper from our ARC funded project: "Energy Efficiency Innovation: Diffusion, Policy and the Rebound Effect". We estimate the economy-wide effect on energy use of energy efficiency improvements in the U.S. We find that the rebound is around 100%, implying that in the long run energy efficiency improvements do not save energy or reduce greenhouse gas emissions.

At the micro level, we might naïvely expect a 1% improvement in energy efficiency to reduce energy use by 1%. But people adjust their behavior. Efficiency improvements reduce the cost of energy services like heating, transport, or lighting. Because these are now cheaper to produce, people consume more of them, and so the percentage reduction in energy use is less than the improvement in efficiency. This is known as the direct rebound effect.

People might also redirect their spending to consume more of complementary goods, like larger houses in the case of residential heating improvements, and reduce their consumption of substitute goods and services, like bus rides or cycling, in the case of car fuel economy improvements. These changes have implications for the energy used to produce these goods and services. Additionally, the reduction in demand for energy should lower the price of energy further boosting the rebound in energy use. Finally, the improvement in energy efficiency is an increase in productivity, which should result in economic growth. Higher incomes mean higher demand for energy. Adding these indirect rebound effects to the direct rebound effect we get the economy-wide rebound effect.

The size of the economy-wide rebound effect is crucial for estimating the contribution that energy efficiency improvements can make to reducing energy use and greenhouse gas emissions. Our study provides the first empirical general equilibrium estimate of the economy-wide rebound effect. Previous studies use simulation models, known as computable general equilibrium models, or partial equilibrium econometric models that don't allow the price of energy to adjust. Some of the latter studies also measure rebound incorrectly, for example assuming that energy intensity – energy used per dollar of GDP – measures energy efficiency. In fact, the majority of the rebound effect happens when energy intensity rebounds as people shift to more energy intensive consumption after an energy efficiency improvement. Economic growth induced by the efficiency improvement is expected to contribute less to total rebound.

We use a structural vector autoregressive model, or SVAR, that is estimated using search methods developed in machine learning. We apply the SVAR to U.S. monthly and quarterly data. An SVAR explains changes in the vector of variables, x, in terms of its past values and a vector of serially and mutually uncorrelated shocks, ε:

In our basic model, the vector, x, contains three variables: primary energy use, GDP, and the price of energy. The first of the shocks is a shock to energy use, holding constant shocks to GDP and the price of energy and the past values of all three variables. We think this is a reasonable definition of an energy efficiency shock. The other two shocks are income and price shocks.

The matrix, B, which transmits the shocks to the dependent variables cannot be estimated without imposing some restrictions or conditions on the model. Usually economists use economic theory to impose restrictions on the coefficients in B (short-run restrictions) and the Π_i (long-run restrictions). Alternatively, they sample a range of models, rejecting only those that don't meet qualitative "sign restrictions" on the matrix B. Instead, we use independent component analysis, an approach that is relatively new to econometrics. This imposes conditions on the nature of the shocks instead and estimates B without direct restrictions. Unlike the short- and long-run restrictions approach, it doesn't impose a priori restrictions on the data, and unlike the sign restrictions approach, it estimates a unique model.

Using the estimated SVAR model we compute the impulse response functions of the dependent variables to the shocks:

The top left graph shows the effect of an energy efficiency shock on energy use. The grey shading is a 90% confidence interval, the x-axis is in months, and the y-axis in log units.

Initially, an energy efficiency shock strongly reduces energy use, but this effect wears off over the following years as consumers and the economy adjusts. Eventually, there is no change in energy use so that rebound is 100%.

The other graphs in the first column show the effect of the energy efficiency shock on GDP and the price of energy. The second column shows the effect of a shock to GDP, and the final column an energy price shock.

The implications for policy are that encouraging energy efficiency innovation is unlikely to make a contribution to reducing greenhouse gas emissions. This is one reason why I am skeptical of projections that predict that energy intensity will fall much faster in the future than in the past because of energy efficiency policies.

On the other hand, if these policies raise rather than reduce the costs of producing energy services then the direct rebound (and presumably the economy-wide rebound) will be negative rather than positive. As, apart from their environmental effects, these would reduce economic welfare, it seems that there would be better options to reduce emissions by switching to low carbon energy.

Flying More Efficiently

I have another new working paper out, coauthored with Zsuzsanna Csereklyei on airline fleet fuel economy. Zsuzsanna worked as research fellow here at the Crawford School on my Australian Research Council funded DP16 project on energy efficiency and the rebound effect. This paper reports on some of our research in the project. We also looked at energy efficiency in electric power generation in the US.

The nice thing about this paper is that we have plane level data on the aircraft in service in 1267 airlines in 174 countries. This data is from the World Airliner Census from Flight Global. We then estimated the fuel economy of 143 aircraft types using a variety of data sources. We assumed that the plane would fly its stated range with the maximum number of passengers and use all its fuel capacity. This gives us litres of fuel per passenger kilometre. Of course, many flights are shorter or are not full, and so actual fuel consumption per passenger kilometre will vary a lot, but this gives us a technical metric which we can use to compare models.

The graph shows that the fuel economy of new aircraft has steadily improved over time. One of the reasons for the scatter around the trendline is that large aircraft with longer ranges tend to have better fuel economy than small aircraft:

This is also one of the reasons why fuel economy has improved over time. Still, adjusted for size, aircraft introduced in earlier decades had (statistically) significantly worse fuel economy than more recent models. We used these regressions to compute age and size adjusted measures of fuel economy, which we used in our main econometric analysis.

The main analysis assumes that airlines choose the level of fuel economy that minimizes costs given input prices and the type of flying that they do. There is a trade off here between doing an analysis with very wide scope and doing an analysis with only the most certain data. We decided to use as much of the technical aircraft data as we could, even though this meant using less certain and extrapolated data for some of the explanatory variables.

We have data on wages in airlines and on the real interest rates in each country. The wage data is very patchy and noisy and we extrapolated a lot of values from the observations we had in the same way that, for example, the Penn World Table extrapolates from surveys. There are no taxes on aircraft fuel for international travel and the price of fuel reported by Platts does not vary a lot around the world. But countries can tax fuel for domestic aviation. We could only find data on these specific taxes for a small number of countries in a single year. So, we used proxies, such as the price of road gasoline and oil rents, for this variable. We proxy the type of flying airlines do using the characteristics of their home countries.

The most robust results from the analysis – that hold whether we use crude fuel economy or fuel economy adjusted for size and age – are that – all things constant – larger airlines select planes with higher fuel economy, higher interest rates are associated with poorer fuel economy, higher fuel prices are associated with higher fuel economy (but the elasticity is small), and fuel economy is worse in Europe and Central Asia than other regions.

It seems that for a given model age and size, more fuel efficient planes cost more. This would explain why, even holding age and size factors constant, higher interest rates are correlated with worse fuel economy. Also, if larger airlines have more access to finance or a lower cost of capital they will be able to afford the more fuel efficient planes.

What effect could carbon prices have on fleet fuel economy? The most relevant elasticity is the response of unadjusted fuel economy to the price of fuel. This allows airlines to adjust the size and model age of planes in response to an increase in the price of fuel. We estimate that this elasticity is -0.09 to -0.13, which suggests the effect won't be very big. Because we use proxies for the price of fuel, we expect that the true value of this elasticity is actually higher. The elasticity also assumes that there is a given set of available aircraft models. Induced innovation might result in more efficient models being developed. There might also be changes in the types of airlines and flights. So the effect could be quite a bit larger in the long run.

Energy Intensity, Growth, and Technical Change

I have a new working paper out, coauthored with Akshay Shanker. Akshay recently completed his PhD at the Crawford School and is currently working on the Energy Change Institute's Grand Challenge Project among other things. This paper was one of the chapters in Akshay's thesis. Akshay originally came to see me a few years ago about doing some research assistance work. I said: "The best thing you could do is to write a paper with me – I want to explain why energy intensity has declined using endogenous growth theory." This paper is the result. Along the way, we got additional funding from the College of Asia and the Pacific, the Handelsbanken Foundation, and the Australian Research Council.

World and U.S. energy intensities have declined over the past century, falling at an average rate of approximately 1.2–1.5 percent a year. As Csereklyei et al. (2016) showed, the relationship has been very stable. The decline has persisted through periods of stagnating or even falling energy prices, suggesting the decline is driven in large part by autonomous factors, independent of price changes.

In this paper, we use directed technical change theory to understand the autonomous decline in energy intensity and investigate whether the decline will continue. The results depend on whether the growing stock of knowledge makes R&D easier over time – known as state-dependent innovation – or whether R&D becomes harder over time.

Along a growth path where real energy prices are constant, energy use increases, energy-augmenting technologies – technologies that improve the productivity of energy ceteris paribus – advance, and the price of energy services falls. The fall in the price of energy services reduces profitability and incentives for energy-augmenting research. However, since the use of energy increases, the "market size" of energy services expands, improving the incentives to perform research that advances energy-augmenting technologies. In the scenario with no state dependence, the growing incentives from the expanding market size are enough to sustain energy-augmenting research. Energy intensity continues to decline, albeit at a slower rate than output growth, due to energy-augmenting innovation. There is asymptotic convergence to a growth path where energy intensity falls at a constant rate due to investment in energy-augmenting technologies. Consistent with the data, energy intensity declines more slowly than output grows, implying that energy use continues to increase.

This graph shows two growth paths – for countries that are initially more or less energy intensive – that converge to the balanced growth path G(Y) as their economies grow:

This is very consistent with the empirical evidence presented by Csereklyei et al. (2016).

However, the rate of labor-augmenting research is more rapid along the balanced growth path and there will be a shift from energy-augmenting research to labor-augmenting research for a country that starts out relatively energy intensive. This explains Stern and Kander's (2012) finding that the rate of labor-augmenting technical change increased over time in Sweden as the rate of energy-augmenting technical change declined.

The following graph shows the ratio of the energy-augmenting technology to the labor-augmenting technology over time in the US, assuming that the elasticity of substitution between energy and labor services is 0.5:

Up till about 1960, energy-augmenting technical change was more rapid than labor-augmenting technical change and the ratio rose. After this point labor-augmenting technical change was more rapid, but the rise in energy prices in the 1970s induced another period of more rapid energy-augmenting technical change.

In an economy with extreme state-dependence, energy intensity will eventually stop declining because labor-augmenting innovation crowds out energy-augmenting innovation. Our empirical analysis of energy intensity in 100 countries between 1970 and 2010 suggests a scenario without extreme state dependence where energy intensity continues to decline.

How Accurate are Projections of Energy Intensity?

A new short working paper about how accurate projections of future energy intensity are. It's an extension of comments I made at Energy Update 2016 here at the ANU.

Energy intensity is one of the four factors in the Kaya Identity, which is often used to understand changes in greenhouse gas emissions. It is one of the two most important factors together with the rate of economic growth. The 2014 IPCC Assessment Report shows that less than 5% of models included in the assessment project that energy intensity will decline slower than the historic rate under business as usual:*

Is this likely? In the paper, I evaluate the past performance of the projections implied by the World Energy Outlook (WEO) published annually (except in 1997) by the International Energy Agency (IEA). The following graph shows the average annual difference between the projected and actual rate of change in energy intensity in subsequent years** for each WEO since 1994:

Positive errors mean that energy intensity declined slower than projected in the following years while negative errors mean it declined faster. So, for example, the error of -0.4% for 2000 means that over the years 2001-2015, on average energy intensity declined by 0.4% a year faster than was projected in the 2000 WEO.

It turns out that these errors are strongly negatively correlated (r = -0.8) with the error in projecting the rate of economic growth, which IEA outsources. Csereklyei et al. (2016), similarly, find that reductions in energy intensity tend to only occur in countries with growing economies. If we divide and multiply the growth rate of energy intensity g(E/Y) by the growth rate of GDP g(Y) we get the following identity:

The first term on the right hand side can be seen as the elasticity of energy intensity with respect to GDP.*** The following graph plots the elasticity as projected and as subsequently realized for each WEO:

The two seem to have tracked each other quite well. But there is a complication. The 1994 to 96 WEOs only projected future energy use up to 2010. 2010 is the only recent year when global energy intensity actually increased. This end point reduces (in absolute value) the actual elasticities for these three WEOs. From 1998 on, the difference between the projected and actual rate of change in energy intensity is calculated up to 2015. But through the 2011 WEO, 2010 is one of the years in the projection period. From 2012, 2010 is no longer include in the projection period and there is a sharp step down in the actual elasticity over the projection period. I think that the elasticities for 2012-16 probably under-estimate the true long-run elasticities and that the relatively stable values from 1998-2011 are more representative of what the future elasticities will be over the full projection horizon to 2030 or 2040.

If that is the case, then the projected elasticity of -0.6 in the 2016 WEO probably over-estimates the the elasticity that will be realized in the long run. Why would this be the case?

Early WEOs largely modeled energy intensity trends based on historical trends. This is not the case for recent WEOs. Over time, the IEA has endogenized more variables in their model of the world energy system and included more and more explicit energy policies. It is likely that the model under-estimates the economy-wide rebound effect. It's also possible that energy efficiency policies are not implemented as effectively as expected.

As part of our ARC funded DP16 project, we hope to contribute to improving future projections of energy intensity by empirically estimating the economy-wide rebound effect.

* The light grey area indicates the projections between the 95th and 100th percentile of the range for the default scenario.

** The base year for each WEO is 2-3 years before the publication date. Therefore, we can already assess the 2015 and 2016 WEO's.

*** We can use the identity to decompose the projection errors:

Over time the contribution of errors in the projected growth rate has increased relative to the contribution from errors in the elasticity. But I think that if we revisit this experiment in 2030 we will find a larger contribution from errors in the elasticity for what are currently recent issues of the WEO.

P.S. 23 June 2017

The paper is now published in Climatic Change.

Discovery Projects 2016

We - myself together with Stephan Bruns and Alessio Moneta - got an ARC Discovery Projects grant. Thanks also to Zsuzsanna Csereklyei who contributed to the development of the proposal and who we may hire as a post-doc using the grant - depending if she wants to come to Australia for the time we will be able to afford to fund with the money we received ($A273k - 65% of what we requested). This is my second ARC grant following the DP12 grant that we got a few years ago. The title of our project is: "Energy Efficiency Innovation, Diffusion and the Rebound Effect". We will be looking at the diffusion of energy efficiency innovations and trying to measure the economy-wide rebound effect empirically. This was our second attempt at applying for a grant on this topic. Last time around we were rated in the top 10% of unfunded proposals and so I thought it was worthwhile revising and resubmitting!

Other good news today is that my colleague, Paul Burke got a DECRA grant. I think this is our fourth DECRA at Crawford. Congratulations to Paul! Also Peter MacDonald and Robert Sparrow got a DP16 grant. Congratulations to Robert and Peter!

Energy Leapfrogging (or Not)

Arthur van Benthem has a recent paper in the Journal of the Association of Environmental and Resource Economists titled "Energy Leapfrogging". The main thesis of the paper is that despite presumed improvements in the energy efficiency of individual technologies such as cars and refrigerators, energy intensity in developing countries today is similar to what it was in today's developed countries when they were at a similar income level. There is no "energy leapfrogging". This is also an implication of our paper "Energy and Economic Growth: the Stylized Facts". If there has been an almost constant log-linear relationship between energy use and GDP per capita then there is no energy leapfrogging.

van Benthem suggests that a major contributor to this is that the consumption bundle in developing countries today is much richer in energy services like personal transport than was the consumption bundle at a similar level of development in today's developed countries. Consumers have substituted towards these now cheaper energy services (what are they consuming less of though?).

On the face of it, this suggests that there would be a very large rebound effect due to substitution towards energy services. This is on top of any indirect rebound effect due to increased energy productivity boosting income and thus energy demand as originally proposed by Harry Saunders.

On the other hand, there must be some shift away from energy services as income increases so that energy intensity is lower in richer countries. Anyway, this is pretty speculative but worth thinking about, I think.