Relative Effectiveness of Energy Efficiency Programs versus Market Based Climate Policies in the Chemical Industry.

• Author: Boyd, Gale A.

1. INTRODUCTION

   The Paris Accord has a goal of reducing C[O.sub.2] emissions to limit climate change to less than 2 degrees C. It has been broadly embraced by nearly every nation of the world. Even in the United States, where the Trump administration has stated its intent to pull out of the accord in 2020 and cancel the Clean Power Plan, climate policy is still being pursued by regions, states, cities, and via other existing programs at the Federal level. One might argue that the success of the Paris Accord is that it is not prescriptive regarding the types of policies nations must pursue, much the same way that a variety of policies are being pursued in the U.S., irrespective of the accord.

   The policy options available to reduce C[O.sub.2] emissions are as varied as the sources of C[O.sub.2] themselves; de-carbonization energy use and reducing total energy demand are the overarching goals. The UNFCCC web site identifies 6 classes of policy instruments; (1) we consider three of these policy levers; market approaches, e.g. carbon taxes or tradeable permits; regulatory instruments, e.g. renewable portfolio standards or fuel economy / equipment standards; and voluntary programs, e.g. informational and behavior based interventions. Economists often point out the benefits of market based approaches, while others may point to market failures as a possible justification for the latter two instruments which we group together as policies and programs (Jaffe and Stavins 1994). Market approaches rely on the price responsiveness of the demand sector to generate change in the level or mix of energy use. One important way that policies and programs work, particularly voluntary programs, is to reduce existing inefficiencies that may not be as responsive to price changes, per se. The relative effectiveness of market based approach (price) vs policies and programs (efficiency) will depend on the energy price elasticities of demand vis-a-vis the extent of existing levels of (in) efficiency. Sectors with high price elasticity and low levels of efficiency gap would be best tackled with market approaches; policies and programs would be more effective in sectors with the opposite, low elasticities and high levels of energy efficiency gap.

   The extent and sources of the energy efficiency gap has been the subject of numerous studies and reviewed by a number of economists (Jaffe and Stavins, 1994; Huntington, 1995; Bloom, Genakos et al., 2010; Allcott and Greenstone, 2012; Boyd and Zhang, 2013; Boyd and Curtis, 2014; Gillingham and Palmer, 2014; Boyd, 2016; Gerarden, Newell et al., 2017).

   Jaffe and Stavins (1994) lay out a framework to compare the economic vs technical (engineering) perspectives in the early literature. Huntington (1995) draws a connection between a largely engineering approaches and the emerging data envelopment (DEA) and stochastic frontier analysis (SFA) literature measuring productivity, suggesting these economic tools could bridge the information gap. (Bloom, Genakos et al., 2010; Boyd and Curtis, 2014) consider how the efficiency gap is related to management practices in industry, with mixed results. Bloom et al find a robust positive relationship in UK manufacturing data while Boyd and Curtis find no similar effect for US data. Allcott and Greenstone (2012) are critical of that largely engineering literature on the extent of energy efficiency and conclude that the area is ripe for rigorous study of how policies and programs might impact heterogeneous consumers. Gillingham and Palmer (2014) conclude the true size of the gap is unclear, but also call for more research. (Gerarden, Newell et al., 2017) review the literature from the perspective of private vs. social optimality and market vs non-market failures. They find a wide range of evidence, from strong to weak, of both market and non-market failures, but note that this evidence is largely from the residential sector and less is known about the industrial and commercial sectors.

   Most of these economic reviews recognize the important role of market based approaches, but few studies look at simultaneous price and efficiency effects and fewer still focus on the industrial sector. This paper addresses the basic empirical question with an industry case study of energy intensive chemical manufacturing in the U.S. Using stochastic frontier analysis (SFA) on the most detailed, plant-level data available, this paper econometrically estimates the two core elements needed for our comparison. The first is persistent and time varying energy efficiency gap, accounting for both industry-sector specific and plant level heterogeneity. The second is energy price elasticities, possibly accounting for energy price endogeneity. A two stage SFA is applied to estimate energy demand frontiers for electricity and fuel separately for 4 segments of the industry. This provides an estimate of the potential energy demand reduction in response to policies and programs designed to close the efficiency gap; the price elasticities can be used to assess a similar response to market based approaches. While the cost-effectiveness of such programs are not considered, their potential impact is assumed to be limited by the estimated, pre-existing efficiency levels. We then compute the carbon price that would be needed to reduce demand by the same amount that is implied by the estimates of the energy efficiency gap. This price provides a metric to compare the two sources of carbon reductions.

   The need for better information on industry energy efficiency was raised by Gerarden Newell et al. (2017) and the importance of the industrial sector in terms of energy demand can't be overstated. In the U.S. 2017 Annual Energy outlook (U.S. Energy Information Administration 2017), industrial is the only sector that energy use is forecast to grow; residential, commercial, and transportation energy consumption are flat. This does not mean that industry doesn't respond to prices or experience technical change. While these types of improvements in efficiency occur in all sectors, improvements in the industrial sector are not forecast to keep pace with economic growth. It may even be the case that policies and programs in the non-industry sectors are easier to implement, e.g. CAFE standards in transport or appliance & lighting standards in residential and commercial sectors, and are responsible for the improvements in energy efficiency that have led to declining or stable demand. Regardless, it is likely that potential impact of energy policies and programs targeted at the industrial sector will be limited to reducing levels of current inefficiency, at least in the near term. The relative size of efficiency vs price response will be key to determining what policies are likely to be more impactful.

   This paper provides estimates of energy efficiency and energy price response in the energy intensive chemical manufacturing sector and shares important features with the methodology presented in and applied to analyze metal based durables (Boyd and Lee, 2016) in that it measures the distribution of energy efficiency of demand relative to local (plant level) energy prices. This paper proposes a slightly different approach from the MLE and LIML reviewed by (Amsler, Prokhorov et al., 2016), because we are concerned with both price endogeneity and systematic plant level heterogeneity in energy use which are not accounted for by Amsler et al. This paper uses a two stage variant of SFA that allows us to account for both plant level energy price endogeneity and plant specific heterogeneity in energy use. The two stage method developed by (Kumbhakar, Lien et al., 2014) handles heterogeneity and can be readily modified to account for plant level price endogeneity controls. This modification is a contribution to the literature. The two stage approach allows the decomposition of efficiency into a plant specific (persistent) and time-varying components, which can be compared across new and continuing plants. This is another contribution to the literature to explore the dynamic aspects of efficiency.

   Jointly estimating the plant level price response and efficiency gap is important. For example, (Gerarden, Newell et al., 2017) point out that divergence of plant level price from average prices may overstate engineering estimates of the efficiency gap. Whether the energy efficiency measure from SFA or its non-parametric counterpart, Data Envelopment Analysis (DEA), is a purely technical efficiency measure or includes allocative efficiency depends on the inclusion of the relevant prices. (Filippini and Hunt, 2015) discuss the difference in the treatment of technical and allocative efficiency in more detail. The energy sub-vector (directional) distance function as defined by (Boyd, 2008) presents this approach as a measure of energy efficiency. Including prices makes the resulting measure of efficiency depend on prices but not explicitly measuring allocative efficiency by altering the direction in which energy efficiency is being measured. To illustrate this, Figure 1 shows the production isoquant for energy and all other quasi-fixed inputs and fixed output; the interior point A to the isoquant is inefficient. It shows three different directions that one could consider measuring efficiency, one is the purely technical efficiency measure in the absence of prices used by (Boyd, 2008). The other two are based on different energy prices. The difference between the energy sub-vector (directional) distance function, AC, where the pure technical efficiency gap is measured as a reduction in energy use, holding other quasi-fixed inputs and production constant and one that includes both technical and allocative efficiency is embodied in the direction that efficiency is measured depending on prices. In this example, when one accounts for energy prices the efficiency gap is lower than when considering pure technical efficiency. In figure...