AUTHOR ACCEPTED MANUSCRIPT FINAL PUBLICATION INFORMATION Energy Efficiency Barriers in Commercial and Industrial Firms in Ukraine An Empirical Analysis The definitive version of the text was subsequently published in Energy Economics, 63, 2017-03 Published by Elsevier and found at http://dx.doi.org/10.1016/j.eneco.2017.01.013 THE FINAL PUBLISHED VERSION OF THIS MANUSCRIPT IS AVAILABLE ON THE PUBLISHER’S PLATFORM This Author Accepted Manuscript is copyrighted by World Bank and published by Elsevier. It is posted here by agreement between them. Changes resulting from the publishing process—such as editing, corrections, structural formatting, and other quality control mechanisms—may not be reflected in this version of the text. You may download, copy, and distribute this Author Accepted Manuscript for noncommercial purposes. Your license is limited by the following restrictions: (1) You may use this Author Accepted Manuscript for noncommercial purposes only under a CC BY-NC-ND 3.0 IGO license http://creativecommons.org/licenses/by-nc-nd/3.0/igo. (2) The integrity of the work and identification of the author, copyright owner, and publisher must be preserved in any copy. (3) You must attribute this Author Accepted Manuscript in the following format: This is an Author Accepted Manuscript by Hochman, Gal; Timilsina, Govinda R. Energy Efficiency Barriers in Commercial and Industrial Firms in Ukraine © World Bank, published in the Energy Economics63 2017-03 CC BY-NC-ND 3.0 IGO http://creativecommons.org/licenses/by-nc-nd/3.0/igo http://dx.doi.org/10.1016/j.eneco.2017.01.013 © 2019 World Bank 1 Energy Efficiency Barriers in Commercial and Industrial Firms in Ukraine: 2 An Empirical Analysis* 3 4 Gal Hochman and Govinda R. Timilsina§ 5 6 7 Abstract 8 Improvement in energy efficiency is one of the main options to reduce energy demand and 9 greenhouse gas emissions. However, large-scale deployment of energy-efficient technologies is 10 constrained by several factors. Employing a survey of 509 industrial and commercial firms 11 throughout Ukraine and a generalized ordered logit model, we quantified the economic, 12 behavioral, and institutional barriers that may impede the deployment of energy-efficient 13 technologies. Our analysis shows that behavioral barriers resulted from lack of information, 14 knowledge, and awareness are major impediments to the adoption of energy-efficient technologies 15 in Ukraine, and that financial barriers may further impede investments in these technologies 16 especially for small firms. This suggests that carefully targeted information provisions and energy 17 audits will enhance Ukrainian firms’ investments in energy-efficient technologies to save energy 18 consumption, improve productivity, and reduce carbon emissions from the productive sectors. 19 20 JEL Classification: O1, O3, Q40 21 22 Keywords: Energy-efficiency, barriers to energy efficiency, discrete choice models, generalized 23 ordered logit model, Ukraine. *The authors would like to thank Mook Bangalore, Robert Borgstrom, Michael Traux, and Mike Toman for very helpful comments and suggestions. We also acknowledge the financial support of the World Bank’s Knowledge for Change Program (KCP). The views and interpretations herein are the authors and should not be attributed to the World Bank Group or the organizations with which it is affiliated. § Hochman (gal.hochman@rutgers.edu) is an Associate Professor, Department of Agricultural, Food, and Resource Economics, Rutgers University, and Timilsina (gtimilsina@worldbank.org) is a Senior Research Economist, Development Research Group, World Bank. 1 1 2 1 1. Introduction 2 The adoption of energy-efficient technologies has been touted as a major policy option to save 3 energy consumption and reduce greenhouse gas (GHG) emissions. The International Energy 4 Agency estimates that energy-efficient technologies offer the highest potential in the total GHG 5 mitigation required to limit global temperature rise by 2050 to 2°C above pre-industrial levels 6 (International Energy Agency, 2012). Many of the studies that develop marginal abatement cost 7 curves for GHG mitigation demonstrate that energy-efficient technologies entail negative costs 8 (i.e., value of energy savings exceeds investment costs even if GHG mitigation benefits are not 9 accounted for) and therefore these options are interpreted as “low-hanging fruit” for climate 10 mitigation (Asian Development Bank, 1998; McKinsey & Company, 2009; Energy Sector 11 Management Assistance Program, 2012). Energy-efficient technologies save firm’s expenditures 12 on energy resulting in their overall cost efficiency. 13 In practice, however, the scale of implementation of such seemingly win-win options are 14 small in relation to their apparent economic potential. The rationale for this disparity is that 15 implementation of these options is constrained by economic and institutional barriers (Jaffe & 16 Stavins, 1994; Howarth & Sanstad, 1995; Sorrell et al., 2004; Mundaca et al., 2013). Moreover, 17 the estimated benefits of deployment of energy efficient technologies may not be realized in 18 practice. This is because the economics of energy-efficient technologies is normally evaluated 19 using engineering benefit-cost analysis (e.g., Goldstein et al., 1990; Blumstein & Stoft, 1995; 20 Brown et al., 1998; McKinsey & Company, 2009; Gillingham & Sweeney, 2012), which usually 21 omits variables such as opportunity costs of investment and also the transaction costs of the 22 deployment (Allcott & Greenstone, 2014). If the transaction costs imposed by the economic and 23 institutional barriers are accounted for, the net benefits that can be achieved from energy- 24 efficient technologies would be smaller as compared to the opportunity costs of the investment 25 thereby causing firms to lose interest in adopting them (Anderson & Newell, 2004). 26 The gap between the level of energy efficiency actually realized and that achieved in 27 theory via the implementation of cost-effective energy-efficient technologies is called the 28 “energy efficiency gap” (Blumstein et al., 1980; DeCanio, 1993; Jaffe & Stavins, 1994; Sanstad 29 & Howarth, 1994; Sorrell et al., 2004; Schleich, 2009). In this paper, we revisit this gap while 3 1 employing survey data collected in 2012 from 509 Ukraine firms. We use the data to empirically 2 examine barriers to the adoption of energy-efficient technologies in Ukraine for both commercial 3 and industrial firms. The empirical analysis employs a generalized ordered logit model (GOL 4 model; Williams, 2006). The selection of the model is based on the fact that GOL models are 5 less restrictive than ordered logit models, whose assumptions are often violated, but are more 6 parsimonious than the multinomial logit models that ignore the ordering of categories. 7 The empirical analysis is carried out for Ukraine, a small developing Eastern European 8 country, located between Russia and Europe. Its geographic position led it to become transit 9 country of natural gas imported from Russia to Europe (Correlje, 2006; Goldthau, 2008; 10 Soderberghn, 2010). Although historically Russia has supplied Ukraine with natural gas and 11 petroleum at prices much lower than market prices (Dimitrova, 2009), recently Russia has begun 12 pushing for higher prices, resulting in several disputes. The culmination of these disputes came 13 when exports to Ukraine of 90 million cubic meters of natural gas per day were halted on 14 January 1, 2009. Increases in energy prices may result in long-run effects that yield significant 15 changes to the economy’s production structure (Schubert & Turnovsky, 2009). 16 The aging infrastructure and the inefficiencies have led Ukraine to call for increased 17 penetration of clean energy and improved efficiency in utilizing energy. The government has 18 enacted several policies to promote the adoption of clean and energy-efficient technologies 19 (Trypolska, 2012). Adoption of alternative energy technologies and the introduction of energy- 20 efficient technologies can address Ukraine’s structural problems, and energy-efficient 21 technologies reduce energy costs and generate economic growth (Gillingham et al., 2014). 2 22 However, are current investments in energy-efficient technologies in Ukraine optimal? 23 Do barriers to the adoption of the energy efficient technologies lead to underinvestment? A 24 number of studies attempted, through empirical analysis, to understand the energy efficiency 25 barriers in different countries, sectors, and energy end uses (see, e.g., Rohdin & Thollander, 26 2006; Sardianou, 2008; Schleich, 2009). The literature suggests that financial barriers such as 27 high upfront costs, lack of information, and priority setting of upper management impede the 2Although we still do not understand the magnitude of the rebound effect, current literature offers little support to the Jevons hypothesis whereby the introduction of energy efficiency results in a net increase of energy use (Gillingham et al., 2014). 4 1 adoption of energy-efficient technologies. However, to our knowledge, this literature focuses 2 mostly on Organization for Economic Cooperation and Development (OECD) countries, and an 3 in-depth analysis has not been performed to better understand barriers to improvements in energy 4 efficiency among industrial and commercial firms in small developing economies. Our study 5 aims to contribute to filling this research gap. Moreover, the importance of such an analysis has 6 increased tremendously with the current political situation in Ukraine, as energy security has 7 become a key concern. 8 Our analysis shows that financial and economic barriers (e.g., high upfront costs of 9 energy-efficient devices and processes, high costs of financing due mainly to higher risks 10 perceived by financial institutions in new and emerging technologies) may hamper firms’ 11 investments in energy-efficient technologies especially for small firms (measured via revenues). 12 It also shows that, contrary to our intuition, firms do not perceive government regulations and 13 internal corporate bureaucracy as barriers to the adoption of energy-efficient technologies. 14 However, the analysis does suggest that behavioral constraints such as lack of knowledge are 15 major barriers that hamper adoption of energy-efficient technologies and lead to underinvestment 16 in such technologies. Furthermore, we found that the commercial sector, which includes the 17 public sector, is less likely to invest in energy-efficient technologies in the absence of policy 18 interventions, but that mandatory energy audits have a larger effect on the commercial sector. 19 We organized the paper as follows: Section 2 discusses firms’ investment decisions 20 regarding energy-efficient technologies and possible barriers that hinder the adoption of such 21 technology. Section 3 presents the cross-sectional survey data and the empirical analysis, 22 followed by concluding remarks in Section 4 23 2. The firm investment decision model 24 We assume a profit-maximizing firm that contemplates whether to invest in energy-efficient 25 technologies. This firm faces two alternative technologies: an energy-inefficient technology 26 (henceforth, denoted with subscript 0) and an energy-efficient technology (henceforth, denoted 27 with subscript 1). For simplicity, we assume the production function is of fixed proportions and 28 let E denote the fixed-proportion energy-output coefficient; that is, E energy units are used to 5 1 produce one unit of output Q. Also, let denote output price and let denote energy price. In 2 addition, we assume a risk adjustment discount rate of i. 3 When modeling the firm’s investment decisions, we assumed the firm first decides 4 whether to invest in the energy-efficient technology and then, given its choice, the firm produces 5 and generates profits. That is, we assumed two periods: In period 1, the firm decides whether to 6 invest I>0 in the energy-efficient technology, while in period 2, the firm consumes energy and 7 incurs the operating costs of using the energy. The model can be made more complex to include 8 other inputs (e.g., labor), other production functions, and uncertainty, as the literature on the 9 adoption of new technologies suggests (e.g., Sunding & Zilberman, 2001). However, for our 10 purposes, this basic stylized model suffices. 11 The model contains several economic factors affecting the firm’s decision whether to 12 adopt the energy-efficient technology. We assume that an investment of I lowers the firm’s 13 energy intensity from 0 to 1 . In addition, we assume that adoption of the energy-efficient 14 technology encompasses hidden or transaction costs H>0 that are unobserved by the 15 econometrician. Other sources of economic costs include the opportunity cost of capital that 16 affects the risk-adjusted discount rate, which can be affected by imperfect credit markets that 17 prevent firms from accessing the capital needed to upgrade a plant and adopt energy-efficient 18 technologies. We denote these types of barriers as economic barriers. These economic costs 19 suggest that firms may elect not to invest in energy-efficient technologies because the cost of 20 implementing such technologies is greater than the benefits to the firm. 21 Economic barriers may limit firms’ investment in energy-efficient technologies. 22 Although the investment may be optimal from the firm’s point of view, these investments may 23 be socially suboptimal because the calculation leading to firms’ investment decisions do not 24 include market distortions caused by environmental externalities (e.g., pollution generated from 25 energy consumption). Thus, market failure results in underinvestment in energy-efficient 26 technologies. On the other hand, imperfect capital markets suggest that firms face capital 27 constraints that reduce the amount invested in energy-efficient technologies. Then, if imperfect 28 credit markets are of concern, development of informal credit markets (Deb & Suri, 2013) and 29 introduction of state-owned development banks (David, 1984; Pulley, 1989) should be 6 1 contemplated as mechanisms that alleviate the financial constraint. These barriers can be reduced 2 via the introduction of better governance and best practices. 3 We let institutional barriers denote a second group of barriers that may impede the 4 adoption of energy-efficient technologies. The literature suggests that regulation can become a 5 barrier to the adoption of new technologies (Djankov et al., 2002; Graff et al., 2009). 3 Regulation 6 may impose a cost (e.g., licensing, permits), as well as delays to the implementation of the new 7 technology, which may negatively affect firms’ incentives to invest in energy-efficient 8 technologies. Formally, we focus on the cost and assume government regulation is costly and 9 results in an additional upfront cost of > 0 to the firm. 10 The literature, however, identifies another important class of barrier that leads to 11 investment inefficiencies resulting in the energy efficiency gap (Alcott & Greenstone, 2012), 12 namely, behavioral and information barriers. The firm may not be aware of or informed about 13 the monetary benefits of using energy-efficient technologies. The firm may also not be attentive 14 to its own energy costs. These barriers are the outcome of bounded rationality, perceived 15 credibility and trust, and information. We assume these barriers cause the firm to undervalue the 16 benefits of energy-efficient technologies. Formally, we capture underinvestment in energy- 17 efficient technologies by the parameter ∈ (0,1), whereby the larger the , the larger the energy 18 efficiency gap. Lack of information, awareness, and experience results in a larger and in firms 19 undervaluing the benefits of energy-efficient technology. 20 However, even if the firm is informed and aware of the benefits of the technology, it still 21 may elect not to invest because, for example, the firm may be renting its facility and not paying 22 the electricity bill (i.e., split incentives). That is, the firm pays less than for a unit of energy. 23 Clearly, if the company does not pay for energy then it will not invest (recall that the upfront cost 24 is strictly positive). However, even if the firm only pays part of the energy cost, its incentives to 25 invest in energy-efficient technologies are lower than socially optimal. Technically, we assume 26 the firm pays only a share ≤ 1 of the energy costs. These split incentives then result in the 3Regulation may also facilitate the adoption of new technologies. However, this will not contribute to an energy efficiency gap and thus is not included in the stylized model. 7 1 energy efficiency gap and in underinvestment in the technology. Therefore, a profit-maximizing 2 firm invests in the energy-efficient technology if (1 − )(0 − 1 ) ⁄(1 + ) > + + (1) 3 3. The empirical analysis 4 Policy and mechanisms that foster adoption of energy-efficient technologies depend on the 5 specific barriers that hamper the investments in these technologies. Thus, a better understanding 6 of the importance of the various barriers to adoption of the technology is needed. Using the 7 intuition developed above, a survey is developed and used to assess the importance of the 8 aforementioned barriers (i.e., economic and financial, behavioral and information, and 9 institutional barriers). Building on Rohdin and Thollander (2006) and Sorrell et al. (2004) the 10 survey questions were developed (see Table 1 of Appendix A, as well as Fedets et al., 2013). 11 3.1 The cross-sectional survey 12 The survey conducted employed a sample of 509 commercial and industrial firms throughout 13 Ukraine and was implemented through face-to-face and mail interviews with representatives of 14 the chosen firms. To map the sample back to an unbiased representation of the survey 15 population, we weighted the survey data using the prevalence of different firms in the overall 16 economy. 17 The firms were chosen using a two-stage quota sample (Sorrell et al., 2004), accounting 18 for industrial and commercial energy-intensive firms in all regions of Ukraine. The sample 19 included 61.9% industrial firms (315 of 509) and 38.1% commercial firms (194 of 509). With 20 respect to the industrial sector, a focus was on big firms since they are more likely to be energy 21 intensive. Therefore, small firms (those with fewer than 50 employees) amounted to no more 22 than 10% of all industrial firms, while big firms (with more than 500 employees) comprised no 23 less than 20% of all industrial firms. The quota of small commercial firms (up to 50 employees) 24 was set at no more than one third of the sub-sample. Also, in the commercial sector, firms with a 25 venue area of more than 150 m2 were selected for the survey because they are likely more energy 26 intensive than those with smaller venues. 8 1 The industrial sector includes firms that operate in mining, processing, and other 2 industrial sectors, as well as services supplied to those sectors, while the commercial sector 3 includes firms that provide education, health care, and other services and those of construction 4 and trade. 5 An important step in the analysis is evaluating the reliability and validity of the survey 6 data. To this end, various experts on the Ukraine economy and in survey design evaluated and 7 assessed the survey, which was revised accordingly. In addition, statistical tools were used to 8 evaluate internal consistency and unidimensionality (Tavakol and Dennick, 2011). Preliminary 9 analysis suggests inconsistency in variables that measure firms’ perception of the cost 10 component of energy efficient technologies. Therefore, we do not employ these variables in the 11 statistical analysis. The internal consistency of the items (questions) that were employed in the 12 analysis and used to describe a single concept or construct (e.g., barrier), yielded Cronbach’s 13 of 0.78 or higher thus suggesting good internal consistency (Tavakol and Dennick (2011) and 14 references therein). 4 That is, the expected values of half of the sample (after discarding missing 15 observations) resulted in a similar outcome to that of the remaining half. Put differently, the 16 proportion of the observed variance that represents true variance is at least 74%. The use of 17 Principle Component Analysis offers further support for homogeneity or unidimensionality, 18 where each set of items (questions) measure a single latent factor (see section 3.2 and Appendix 19 B). 20 The data collected via the survey are used to quantify firms’ perceived barriers to the 21 adoption of energy-efficient technologies. In the following paragraphs, we first describe the 22 variables used in the analysis and then discuss the model’s specifications. 23 3.2 The dependent variable and its regressors 24 The ordinal dependent variable is the amount a firm has invested in energy-efficient technologies 25 in the past five years. This variable is coded as follows: 26 • 0 if the firm did not invest in energy-efficient technologies (93 firms) 4Regulation may also facilitate the adoption of new technologies. But this will not contribute to an energy efficiency gap and thus is n 9 1 • 1 if the firm made strictly positive investments of less than US$250,000 (232 2 firms) 3 • 2 if the firm invested more than US$250,000 (64 firms) 4 Because the way questions were structured, we could not normalize investment to size of 5 firm via revenues. However, we did introduce revenues explicitly to the regression to control for 6 firm size. 7 The respondents rated the importance of each barrier/question on a scale from 0 to 3 8 based on its perceived influence in the firm on the implementation of energy-efficient 9 technologies. If a barrier’s influence was strong, it was given 3 points, 2 points were given for 10 considerable influence, and 1 point was assigned for little influence. If a barrier had no influence, 11 it was given 0 points, and if it did not apply for a firm, it was marked as “No answer/Not 12 applicable.” The 25 questions pertaining to the different barriers are depicted in Table 1, 13 Appendix A. 14 We reduced the dimensionality of the analysis using factor analysis tools. While using 15 principal components analysis, 5 we reduced the number of variables from 25 to 7 and increased 16 precision. We use the rule of thumb that requires the eigenvalue to be greater than 1 for the 17 factor to be included in the empirical analysis. 6 The results of the principal components analysis 18 are depicted in Appendix B. We obtained one common factor with an eigenvalue greater than 1 19 for each of the barriers analyzed (Table 1). Using the appropriate loading factors, we then 20 calculated the covariates of the various groups of barriers investigated. ot included in the stylized model. loys orthogonal transformation to convert observations of correlated variables (variables that belong to a certain group – e.g., financial barriers) into a set of values of linearly uncorrelated variables that are called principal components. It is used in macroeconomics to aggregate multi-dimension indicators and to clean the noise from observed series in the panel, which is poorly correlated with the rest of the panel (e.g., Avesani et al., 2006; Forni et al., 2000). It has also being applied to complex datasets, which included multiple indicators to construct social capital indexes (Sabatini, 2005). For more on the asymptotic characteristics of factor analysis, see Bai (2003). 6 The eigenvalue is proportional to the portion of the sum of the squared distances of the points from their multidimensional mean. The principal components analysis essentially rotates the set of points around their mean to align with the principal components. This moves as much of the variance as possible (using an orthogonal transformation) into the first few dimensions. The values in the remaining dimensions tend to be small and may be dropped with minimal loss of information. 10 1 Our empirical model also includes the following firm characteristics (Table 1): annual 2 revenues, the share of energy costs to the firm relative to total production costs, a dummy 3 variable that equals 1 if the facility is rented and 0 otherwise, an ownership dummy that equals 1 4 if the firm is domestic and privately owned and 0 otherwise, and an ownership dummy that 5 equals 1 if the firm is foreign owned and 0 otherwise. 6 Table 1. Data summary. Variable Observations Mean Std. Dev. Min Max Dependent variable Invest 389 0.93 0.63 0 2 Firm characteristics Annual revenues 334 2.73 1.64 1 8 Share of energy costs 316 11.43 10.21 0.3 70 Private ownership 491 0.69 0.46 0 1 Foreign ownership 490 0.03 0.17 0 1 Facility rented 509 0.18 0.38 0 1 Barriers to the adoption of energy-efficient measures Financial barriers 356 6.58 2.88 0 14.59 Hidden costs barriers 420 3.76 2.55 0 11.82 Energy costs 443 0.90 1.04 0 3 Split barriers 382 1.03 1.56 0 5.50 Knowledge barriers 433 3.67 3.45 0 12.39 Existing laws and 296 4.83 2.88 0 10.72 regulation Firm’s bureaucracy 401 1.45 1.72 0 7.11 7 8 3.3 The statistical model 9 The parallel lines assumption states that parameters are the same across the different categories. 10 The assumption suggests that the correlation between independent variables and the dependent 11 variable is not affected by the dependent variable’s categories, (i.e., the dependent variable’s 12 categories are parallel to one another). However, because this assumption is usually violated 13 when analyzing complex surveys, and because the ordinal response of the questionnaire 14 implicitly captures ever-increasing levels of investment in energy-efficient technologies, we 11 1 employ the GOL model (Williamson, 2006). 7 Because we aggregate the investment response 2 variable to three categories (0, 1, and 2), the GOL model yields two response functions (i.e., for 3 categories 0 and 1). 4 Let 0 and 1 denote the vector of parameters of the response function estimated for 5 investment categories 0 and 1, respectively. Let x denote the vector of regressors, and assume 6 F(.) the cumulative distribution function. Then, the probability that investment will take the 7 value 0, 1, or 2 is equal to 8 ( = ) = (− ) 9 ( = ) = (− ) − (− ) 10 ( = ) = − (− ) 11 Assuming a logistic cumulative distribution function allows the interpretation of the GOL model 12 in terms of logits.8 13 Each set of coefficients can be interpreted as those of a binary logit regression. The first 14 set of response function parameters can be interpreted as the outcome of binary regression where 15 the dependent variable’s categories are either no investment or positive investments (i.e., 0 16 versus 1 or 2). On the other hand, the second response function parameters can be interpreted as 17 the parameters of a binary logit regression where the dependent variable’s categories are either 18 null to some investment or large investments (beyond $250,000 US in the last 5 years), that is, 0 19 and 1 versus 2. 20 When presenting the results, we used an alternative but equivalent presentation of the 21 parameters of the GOL model; instead of assuming 2 sets of , we assumed only one set of 22 coefficients, namely, Beta. We also assumed a second set of coefficients, Gamma. If the null 23 hypothesis that the specific Gamma parameter equals 0 cannot be rejected, then we assumed the 7 A key problem with the cumulative logit model is the assumption of parallel lines. This key assumption is often violated; that is, one or more of the coefficients of the regressors differs across the various categories. The GOL model relaxes the parallel lines assumption and generalizes the cumulative logit models. 8 Because probabilities are constrained to be in the range of [0,1], the generalized ordered logit model imposes explicit restrictions on the range of the x variables: ≥ 12 1 parallel line assumption held and set the relevant Gamma parameter to zero. However, if we 2 reject the hypothesis that the value of the Gamma parameter is 0 at a 5% significance level, then 3 we report the parameter and the parallel line assumption is violated for this specific parameter. 9 4 This alternative presentation suggests that the coefficient of the specific 0 parameter equals the 5 relevant Beta parameter, whereas the coefficient of the specific 1 parameter equals the sum of 6 the relevant Beta and Gamma parameters. 7 3.4 The model’s specifications 8 In developing the empirical model, we follow Hosmer and Lemeshow’s (2000) incremental 9 process. Initially, we performed a bivariate analysis of the relationship of the dependent variable 10 investment and each of the individual regressor candidates, where each specification is defined 11 by its set of regressor candidates. We then selected variables associated with investment whose 12 coefficient differs from 0 at a significance level of less than 0.25 (the output of the bivariate 13 regression is presented in Appendix C). 14 The next step in developing the model’s specification includes the assessment of the 15 validity of the parallel lines assumption. Because the data are survey data, the likelihood ratio is 16 not appropriate (Williams, 2005). We, therefore, use an iterative process to evaluate the validity 17 of the parallel lines assumption at a 5% significance level (Williams, 2006). The analysis 18 suggests that annual revenues, being foreign owned, and knowledge barriers do not meet the 19 parallel lines assumption but other variables do. The Wald chi-squared statistic of the model (i.e., 20 model I in Table 2) is 46.69. 21 In addition to the aforementioned model (Model I in Table 2), we estimated other 22 specifications including one where hidden costs were explicitly included in the regressor 23 candidates set (Model II in Table 2). We used the adjusted Wald test, as well as the Bonferroni 24 adjustment (Korn & Graubard, 1990), to evaluate the null hypothesis that some of Model II’s 25 parameters are zero (private ownership, rented facility, and hidden costs). 10 The Wald test 26 resulted in a test statistic of F=3.14 and P-value of 0.018, whereas the Bonferroni adjustment 9 While focusing on the baseline model, assuming a 10% significance level (or a 1% significance level), instead of a 5% significance level, would not change our conclusion that the parallel line assumption is violated. 10 When using survey data, the assumption that observations are independent of each other may be violated. Therefore, we cannot use the LR test, BIC, and the AIC. However, we can compute the F statistic and the t statistic and perform the Wald test instead. 13 1 yielded F=3.04 and P-value of 0.021. Thus, we could not reject the null hypothesis that yields the 2 more parsimonious specifications of Model I at a 10% significance level. Although we chose 3 model I, we used the other models (Model II and others not reported because of space 4 constraints) to better understand the data. 5 3.5 The empirical analysis 6 As noted above, when presenting the results, we used an alternative but equivalent presentation 7 of the parameters of the GOL model. The alternative presentation suggests that the coefficient of 8 the specific parameter equals the sum of the relevant Beta and Gamma parameters. For 9 example, (see Table 2, model I) 10 0 ( ) = 12.69 but 1 ( ) = 14.16 − 12.53 = 1.62; and 11 0 ( ) = 1 ( ) = −0.14. 12 Finally, the Alpha variables are simply the cutpoints, i.e., the negative of the constant of the 13 relevant response categories: while Constant (category 0) is the cutpoint when comparing no 14 investment (category 0) with positive investment (categories 1 or 2); Constant (category 1) is the 15 cutpoint when comparing no to medium levels of investment (categories 0 or 1) with large 16 investments (category 2). 17 Table 2. The estimated parameters of the (generalized) ordered logit model. Variables Ordered Logit Model The GOL The GOL Baseline Model: Alternative Model: Model I Model II Beta Revenues 0.38** 0.10 -0.31 (0.18) (0.18) (0.19) Energy expenditure 0.03 0.04 0.05* (0.02) (0.03) (0.02) Private ownership -0.14 -0.37 (0.49) (0.60) Foreign owned 2.57** 14.16*** 14.37*** (0.89) (0.72) (0.95) Rented facility -0.50 -0.42 (0.58) (0.53) Financial barriers -0.24** -0.14 -0.22** (0.10) (0.10) (0.10) Split barriers 0.18 0.10 0.04 (0.26) (0.28) (0.34) Knowledge barriers -0.18* -0.33** -0.37* (0.10) (0.16) (0.20) 14 Hidden costs 0.53*** -0.08 (0.16) (0.24) Hidden costs and mandatory 0.09*** 0.10 energy audit (0.04) (0.06) Low priority for energy costs -0.20 -0.07 0.51 (0.43) (0.47) (0.68) Firm’s bureaucracy -0.09 -0.11 -0.16 (0.19) (0.16) (0.14) Gamma Revenues 0.64*** (0.16) Foreign owned -12.53*** -11.30*** (1.15) (1.39) Knowledge barriers -0.48*** -0.59** (0.17) (0.25) Hidden costs 0.71*** (0.23) Low priority for energy costs -1.39** (0.61) Alpha Constant (category 0) -0.95 1.52 1.82* (0.81) (0.94) (1.14) Constant (category 1) -3.01*** -2.61*** -2.12** (0.92) (0.93) (1.06) Statistics N 97 97 97 F 3.32 46.65 37.82 1 Legend: * p<.1; ** p<.05; *** p<.001. Numbers in parenthesis denote standard errors. 2 3.5.1 Economic barriers 3 Financial Barriers: Financial barriers quantify the importance of upfront capital costs, risk, and 4 the opportunity cost of the adoption of energy-efficient technologies. The Model II coefficient of 5 the financial barriers equals −0.22 and is statistically different from 0 at a 5% significance level 6 (0 ( ) = −0.22). Part of the effect of upfront costs might also be observed in 7 firms’ revenues parameter. The effect of revenues on high investment levels is large. The 8 parameter is significant at the 1% level, and an increase of 1 unit in the revenue category results 9 in the odds of a high investment compared to a low investment changing by a factor of 2.1. Large 10 firms are more likely to make large investments in energy-efficient technologies than firms that 11 are more capital constrained. When analyzing the raw data, the industrial sector ranks various 12 financial barriers higher than the commercial sector, although the differences are not large. 15 1 Furthermore, firms rank high upfront costs as the most important factor impeding adoption of 2 energy-efficient technologies (i.e., I in Eq. (1)). The statistical analysis finds evidence that 3 supports this observation. However, the statistical analysis suggests that it is likelier to hinder 4 small firms’ investment in energy-efficient technologies. 5 While Worrell (2009) found that financial barriers limit investment in energy-efficient 6 technologies, Carpenter and Chester (1984) did not. Carpenter and Chester (1984) analyzed the 7 conservation tax credits of the early 1980s in the U.S. and found that although 86% of those 8 surveyed knew about the credit, only 35% used it, and of those firms that used it, 94% of 9 investments made in energy efficiency would have been made regardless of the financial 10 incentives (e.g., in the absence of policy). 11 Hidden costs: Hidden costs prevent the scaling up of energy-efficient technologies. Similar to 12 upfront costs and financial barriers, hidden costs may yield higher costs and thus less adoption; 13 that is, a higher H in Eq. (1) results in fewer firms benefiting from the adoption of energy- 14 efficient technologies. The raw data suggest that installation of energy-efficient technologies 15 needs substantial reconfiguration of production processes and that industrial firms do not trust 16 new devices. However, we cannot reject the null hypothesis that hidden costs do not hamper 17 adoption of energy-efficient technologies at a 10% significance level (the Wald test, as well as 18 the Bonferroni adjusted Wald test, rejects Model II in favor of Model I – the difference between 19 the two models being the initial set of regressor candidates). 20 Because preliminary analysis suggests that missing observations of hidden costs are 21 likelier with firms that have undergone a mandatory energy audit, we introduced an interaction 22 term between hidden costs and a dummy variable that equals 1 if the firm has undergone a 23 mandatory energy audit and 0 otherwise. The analysis rejects the hypothesis that the interaction 24 term equals 0 at a 1% significance level (Model I in Table 2). 25 Energy costs and their priority for the firm: Energy costs play an important role, as the 26 literature predicts (Gillingham et al., 2009). Hughes (1991) suggested that the energy sector is 27 important to Eastern European countries for two reasons: Eastern European countries have 28 higher energy prices than countries with equivalent levels of income and Eastern European 16 1 countries are also some of the most energy-intensive economies in the world. 11 On average, the 2 firms surveyed reveal that they would decrease their energy costs by one third if there were no 3 barriers to adopting energy-efficient technologies. The share of energy expenditure of total cost 4 is significant at the 10% significance level in the baseline model (Model I in Table 2). That is, 5 the higher is 0 in Eq. (1), the more likely a firm is to adopt energy-efficient technology. 6 3.5.2 Behavioral and information barriers 7 Split incentive barriers: Split barriers combine barriers that show the influence of splitting the 8 responsibility of using energy resources with another party (see Appendix A). However, the 9 analysis does not find support for split incentive barriers and does not reject the null hypothesis 10 that the split barrier parameter equals 0 at a 10% significance level (however, we should be 11 careful in interpreting this result because in the sample the number of firms whose bills are paid 12 by a building or the owner facility is only 3.4% for commercial firms and 0.6% for industrial 13 firms). The analysis also suggests that rent is not a factor that hinders the adoption of energy- 14 efficient technologies (Table 2). The analysis does not find support for < 1 in Eq. (1). 15 Knowledge barriers: However, we found strong support for knowledge barriers impeding the 16 adoption of energy-efficient technologies. The iterative estimation process does not support the 17 parallel lines assumption and suggests that the effect of knowledge barriers at low investments is 18 different (smaller) than at high investments. The analysis finds support for > 0 (see Eq. (1)) 19 and that increases with investment. When comparing categories 0 versus 1 and 2, the point 20 estimate of the Beta knowledge barrier parameter is -0.33 and is significant at the 5% level (i.e., 21 0 = −0.33), but when comparing categories 0 and 1 versus 2 the Gamma knowledge parameter 22 is -0.48 and 1 = −0.8 and is significant at the 1% significance level. This indicates that Ukraine 23 loses opportunities because it lacks awareness of and experience in energy-efficient technologies. 24 Furthermore, the negative effect on investment is more pronounced at higher levels. 25 How will the amount invested in energy-efficient technologies change as knowledge and 26 awareness of the technology change? To answer this question, while holding all other variables 27 of the sample at their means, we plot the predicted (cumulative) probabilities of the various 11 Recall that Ukraine is very energy intensive and inefficient in its use of energy (Apergis & Payne, 2013; Ogaranko & Hubacek, 2013). 17 1 categories (i.e., categories 0, 1, and 2) in Figure 1. The figure suggests that larger knowledge 2 barriers result in less investment in energy-efficient technologies than otherwise. 3 4 Figure 1. Knowledge barriers. 5 The analysis also suggests that mandatory energy audits can help firms better understand 6 the benefits of energy-efficient technologies, yielding more adoption of these technologies; 7 likewise, policy that incentivizes adoption of these technologies leads to their greater adoption. 8 Several studies have tried to disentangle the effects of information provision from other factors. 9 Newell and Siikamaki (2014) suggested that lack of relevant information can lead to significant 10 undervaluing of energy-efficient technology and that information on economic value of the new 11 technology is an important component affecting cost-effective decisions. Similarly, firms may 12 fail to invest in energy-efficient technologies because they are unaware of them (Anderson & 13 Newell, 2004; Bloom et al., 2013). Although there isn’t much work on firms’ adoption of 14 energy-efficient technologies, the empirical work on residential energy-efficient programs 15 suggests that existing programs may affect consumer behavior and may even result in some 16 consumers relying too much on appliances with energy-efficient certifications, such as the 17 Energy Star certification in the U.S. (Houde, 2014), thus resulting in unintended consequences. 18 1 A well thought out mechanism that assigns information provisions, however, might address such 2 concerns (Schultz et al., 2007; Gerarden et al., 2015). 3 3.5.3 Institutional barriers 4 Existing rules and regulations: A lack of effective government policies to facilitate energy- 5 efficient programs ranks highest among the rules and regulation factors, with an average 6 assessment of 2.2 points. However, statistically, existing rules and regulations do not yield large 7 barriers to the adoption of energy-efficient technologies (see Appendix C). When estimating the 8 bivariate effect of regulation barriers on investment levels, the P-value of the parameter is 0.45. 9 Following Hosmer and Lemeshow’s (2000) incremental process resulted in existing rules and 10 regulations not being included in the final model. 11 Firm’s bureaucracy: Although the firm’s bureaucracy barriers are included in the final model 12 and the sign of the parameter suggests that an increase in this institutional barrier results in less 13 adoption, we cannot reject the hypothesis that the parameter is not different than 0 at a 10% 14 significance level. We, thus, conclude that regulatory and internal bureaucracy does not affect 15 firms’ decision whether to invest in energy-efficient technologies. That is, the statistical analysis 16 cannot reject the null hypothesis that = 0 in Eq. (1). 17 The analysis does suggest that once controlling for the other factors, foreign ownership 18 (when the firm is 100% foreign owned) has a significant effect on the likelihood of investing in 19 energy-efficient technologies, but this effect declines at high investment levels (Table 2). 20 3.6 Industrial versus commercial firms 21 Is the effect of the barriers the same across sectors? While using the GOL model, we increased 22 the means of the various factors by one standard deviation and calculated the predicted 23 probability that a firm either does not invest (investment category 0, henceforth denoted 0), 24 makes a moderate investment (investment category 1, henceforth denoted 1), or makes a large 25 investment (investment category 2, henceforth denoted 2). In Figure 2, we illustrate the predicted 26 probabilities of a firm choosing a level of investment, while separating between industrial and 27 commercial firms. For each factor that appears on the y-axis, we increase its value by one 28 standard deviation from its mean but fix the value of other factors at their mean. We then use the 19 1 estimated parameters to calculate the predicted probability that the firm will locate in a category 2 (i.e., 0 or 1 or 2). We depict in Figure 2 the change in the predicted probability that a firm locates 3 in a certain category. The change in the predicted probability is measured on the x-axis. No 4 change in the predicted probability of a category is depicted with the category’s notation locating 5 on the zero-line (i.e., the vertical line starting at 0 on the x-axis); an increase in predicted 6 probability is denoted via the category’s notation locating to the right of the zero-line while a 7 decrease is denoted with a notation of a category locating to the left of the zero-line. 8 Figure 2 suggests that revenues have a larger effect on industrial than commercial firms, 9 although the sign of the impact is similar between the two sectors. When we increase the value of 10 revenues by one standard deviation, the predicted probability of a firm not investing in energy- 11 efficient technologies does not change (the 0 locates at the zero-line). However, the increase in 12 revenues by one standard deviation results in a decline in the predicted probability of a firm 13 making moderate investments (1 is to the left of the zero-line) but an increase for a firm making 14 large investments (2 is to the right of the zero-line). Furthermore, the net effect on 1 and 2 is 15 larger for the industrial sector (1 and 2 are further away from the zero-line). 16 Figure 2 also suggests that foreign ownership matters to industrial firms, but since there 17 are only a few foreign firms in the commercial sector it has no statistical significance for these 18 firms. While for the commercial sector all notations (i.e., 0, 1, and 2) are bundled around the 19 zero-line, for the industrial sector a one standard deviation increase in the foreign factor results in 20 the predicted probability of making no investment declining (0 locates to the left of the zero- 21 line), but the predicted probability of making an investment increases, with a larger increase 22 calculated for a large investment (1 and 2 are to the right of the zero-line, where 2 is further to 23 the right than 1). 24 Information and its dissemination have different effects on the two sectors (Figure 2). 25 While an increase in knowledge barriers causes commercial firms to stop investing in energy- 26 efficient technologies (1 and 2 locate to the left of the zero-line), a similar increase leads 27 industrial firms that would have made large investments in the technology to make smaller 28 investments (2 locates to the left of the zero-line, but 0 and 1 locate to the right of the zero-line). 20 1 2 Figure 2. Marginal effect of various factors: Industrial versus commercial firms. 3 We found more support for differences across sectors when looking at the interaction 4 term of hidden costs and mandatory energy audits (denoted Audit in Figure 2). Although the 5 interaction term does not have much effect on firms that do not invest in energy-efficient 6 technologies (although the 0 locates to the left of the zero-line, the difference is relatively small), 7 it has a negative effect on small to moderate levels of investment (1 locates to the left of the 8 zero-line) but a positive effect on large levels of investment (2 locates to the right of the zero- 9 line). On average, the data suggest that commercial firms are slightly more energy intensive: The 10 average share of energy costs in total production costs is slightly larger for commercial firms 11 (19.9% versus 17.7%, respectively), and the empirical analysis suggests that the effect of an 12 increase of one standard deviation of the interaction term is larger for commercial firms than 13 industrial firms (for commercial firms, both 1 and 2 are further away from the zero-line; Figure 14 2). 21 1 4. Concluding remarks 2 This study examines energy efficiency barriers to the industrial and commercial (including 3 public) sectors in Ukraine by conducting a survey of 509 firms throughout the country. The 4 results from the survey are then used in empirical models to understand the importance of 5 various barriers to the adoption of energy-efficient technologies. 6 The study finds that lack of knowledge and awareness are major barriers to the adoption 7 of energy-efficient technologies, more so for large firms. Knowledge barriers likely reduce the 8 size of investments made by industrial firms in energy-efficient technologies and result in zero 9 investment by commercial firms. This is because the lack of awareness and information results in 10 firms placing a low priority on energy efficiency improvements. The results suggest that 11 information provision and energy audits help reduce this barrier. This finding is consisting with 12 the existing literature (e.g. Thollander et al., 2007). The study also suggests that financial and 13 economic barriers (e.g., high upfront costs of energy-efficient devices and processes, high costs 14 of financing due mainly to higher risks perceived by financial institutions in new and emerging 15 technologies) may further impede firms’ investments in energy-efficient technologies in Ukraine, 16 especially for small firms. We did not find firms perceiving government regulations and internal 17 corporate bureaucracy as barriers to the adoption of energy efficient technologies. The study also 18 suggests that an increase in energy prices would lead to greater adoption of energy-efficient 19 technologies. 20 Our findings imply that improvement of energy efficiency require a major policy shift in 21 Ukraine to reduce the barriers to energy efficient technologies. The new policies should aim to 22 increase awareness of the benefits from energy efficient technologies among the industrial and 23 commercial sectors. Moreover, the government can lower the financial barriers through various 24 policy incentives, such as creation of dedicated energy efficiency improvement funds, providing 25 guarantee (or insurance) to financial institutions that offer energy efficiency loans. 26 The study also finds sectorial heterogeneity to the perceived barriers to the adoption of 27 energy efficient technologies. Such heterogeneity should be accounted for when designing 28 policies and incentives to improve energy efficiency. In future work, we plan to seek further 22 1 understanding of the importance of heterogeneity among sectors and investigate how sectorial 2 heterogeneity might affect the design of energy efficiency policies and incentives. 3 23 1 References 2 1. Allcott, H, and M. Greenstone. 2012. Is there an energy efficiency gap? MIT working paper 12-03. 3 2. Anderson, S. T. and R. G. Newell. 2004. “Information programs for technology adoption: The case of energy-efficiency audits.” Resource 4 and Energy Economics. 26 (1): 27–50. 5 3. Apergis, Nicholas and James E. Payne. “Energy consumption and economic growth: Evidence from the Commonwealth of Independent 6 States.” Energy Economics, 2013; 31: 641-647. 7 4. Asian Development Bank (1998), ALGAS Summary Report, ADB, Manila. 8 5. Avesani, R. G., A. G. Pascual, and J Li. 2006. A New Risk Indicator and Stress Testing Tool: A Multifactor Nth-to-Default CDS Basket. 9 IMF Working Paper WP/06/105 10 6. Bai, J. 2003. “Inferential theory for factor models of large dimensions.” Econometrica. 71 (1): 135–171. 11 7. Berkhout, PHG, JC Muskens, and JW Velthuijsen. 2000. “Defining the rebound effect,” Energy Policy. 28(6): 425–432 12 8. Bloom, Nicholas, Benn Eifert, Aprajit Mahajan, David McKenzie, and John Roberts. 2013. “Does management matter? Evidence from 13 India.” Quarterly Journal of Economics 128 (1): 1–51. http://dx.doi.org/10.1093/qje/qjs044. 14 9. Blumstein, C., B. Kreig, L. Schipper, and C. York. 1980. “Overcoming social and institutional barriers to energy efficiency.” Energy. 15 5:355–72. 16 10. Blumstein, C., and S. E. Stoft. 1995. “Technical efficiency, production functions and conservation supply curves.” Energy Policy. 23 (9): 17 765–68. 18 11. Brown, Marilyn A., Mark D. Levine, Joseph P. Romm, Arthur H. Rosenfeld, and Jonathan G. Koomey. 1998. “Engineering–economic 19 studies of energy technologies to reduce greenhouse gas emissions: Opportunities and challenges.” Annual Review of Energy and the 20 Environment 23, 287–385. 21 12. Carpenter E.H., Chester T.S. 1984. “Are federal energy tax credits effective? A Western United States survey.” Energy J. 5:139–49. 22 13. Correlje, A., and C. van der Linde. “Energy supply security and geopolitics: A European perspective.'' Energy Policy, 2006; 34(5): 532-543. 23 14. David, I. P., 1984. Estimates of the total external debt of the developing member countries of ADB: 1981–1983. Statistical Report Series, 1. 24 Asian Development Bank Economics Office, Manila. 25 15. Deb, R., and T. Suri. 2013. “Endogenous emergence of credit markets: Contracting in response to a new technology in Ghana.” Journal of 26 Development Economics. 101: 268–283. 27 16. Djankov, S., R. La Porta, F., Lopez-de-Silanes, and A. Shleifer (2002). “The regulation of entry.” Quarterly Journal of Economics, Vol. 28 CXVII (1): 1-37 29 17. DeCanio S.J. 1993. “Barriers within firms to energy-efficient investments.” Energy Policy 21:906– 14. 30 18. Dimitrova, Antoaneta. “Constraining external governance: Interdependence with Russia and the CIS as limits to the EU's rule transfer in the 31 Ukraine.” Journal of European Public Policy, 2009; 16(2): 2009. 32 19. Energy Sector Management Assistance Program. 2012. Planning for a Low Carbon Future: Lessons Learned from Seven Country Studies, 33 World Bank, Washington, DC. 34 20. Fedets, I., Kuziakiv, O., and Naumenko, D.. 2013. “Quantifying barriers to energy efficiency measures to reduce GHG emissions.” Institute 35 for Economic Research and Policy Consulting: Kyiv, Ukraine. January 2013. 36 21. Forni, M., Hallin, M. Lippi, M., and Reichlin, L. 2000. Coincident and leading indicators for the EURO area. Review of Economics and 37 Statistics. 82 (82): 540–554. 38 22. Gerarden, T., Newell, R. G., and Stavins, R. N. 2015. “Assessing the energy-efficiency gap.” Cambridge, Mass.: Harvard Environmental 39 Economics Program, January 2015. 24 1 23. Gillingham, K., Newell, R. G., Palmer, K. 2009. Energy Efficiency Economics and Policy. NBER Working Paper 15031. 2 http://www.nber.org/papers/w15031 3 24. Gillingham, K., and Sweeney, J. (2012). Barriers to implementing low-carbon technologies. Climate Change Economics. 3(4): 1-21. DOI: 4 1250019 (2012) [21 pages] DOI: 10.1142/S2010007812500194 5 25. Gillingham, K., Rapson, D., and Wagner, G. “The rebound effect and energy efficiency policy.” Fondazione Eni Enrico Mattei #107.2014. 6 2014. 7 26. Goldstein, D., Mowris, R., Davis, B., and Dolan, K. 1990. Initiating Least-Cost Planning in California: Preliminary Methodology and 8 Analysis, Natural Resources Defense Council and the Sierra Club, prepared for the California Energy Commission Docket No. 88-ER-8, 9 Revised May 10. 10 27. Goldthau, A.. ''Rhetoric versus reality: Russian threats to European energy supply.'' Energy Policy, 2008; 36(2): 686-692. 11 28. Graff, G. D., Zilberman, D., and Bennett, A.B. 2009. “The contraction of agbiotech product quality innovation,” Nature Biotechnology, 12 27(8): 702-704. 13 29. Greening, L.A., D.L. Greene, A.B., and Difiglio, C. 2000. “Energy efficiency and consumption—the rebound effect—a survey.” Energy 14 Policy. 28(6): 389–401 15 30. Hosmer, D. W. and Lemeshow, S.. 2000. “Introduction to the logistic regression model.” In Applied Logistic Regression, Second Edition. 16 Wiley Online Library. 1—30. 17 31. Houde, S. 2014. How Consumers Respond to Environmental Certification and the Value of Energy Information. Working Paper 20019. 18 National Bureau of Economic Research. http://www.nber.org/papers/w20019. 19 32. Howarth R.B., and Sanstad A.H. 1995. Discount rates and energy efficiency. Contemp. Econ. Policy 13:101–9. 20 33. Hughes, G. 1991. The energy sector and problems of energy policy in Eastern Europe. Oxford Review of Economic Policy. 7(2): 77-99. 21 34. International Energy Agency. 2012. Energy Technology Perspectives 2012: Pathways to a Clean Energy System. IEA, Paris. 22 35. Jaffe, A., and Stavins, R. 1994. The energy paradox and the diffusion of conservation technology. Resource and Energy Economics. 16(2): 23 91–122. 24 36. Korn, E. L., and Graubard, B. I. 1990. “Simultaneous testing of regression coefficients with complex survey data: Use of Bonferroni t 25 statistics.” American Statistician, 44: 270–276. 26 37. Mundaca, L., Mansoz, M., Neij, L., and Timilsina, G.R. 2013. Transaction costs analysis of low-carbon technologies, Climate Policy, 27 Vol.13, No. 4, pp. 490-513. 28 38. McKinsey & Company (2009), Pathways to a Low Carbon Economy: Version 2 of the Global Greenhouse Gas Abatement Cost Curve. 29 McKinsey & Company. http://www.mckinsey.com/clientservice/ccsi/pathways_low_carbon_economy.asp. 30 39. Newell, R. G., and Siikamä ki, J. 2014. “Nudging energy efficiency behavior: The role of information labels.” Journal of the Association of 31 Environmental and Resource Economists 1 (4): 555–98. http://dx.doi.org/10.1086/679281. 32 40. Ogaranko, L., and Hubacek, K. 2013. Eliminating indirect energy subsidies in Ukraine: Estimation of environmental and socioeconomic 33 effects using input–output modeling. Journal of Economic Structure. 34 41. Pulley, R. 1989. Making the Poor Creditworthy: A Case Study of the Integrated Rural Development Program in India. World Bank 35 Discussion Paper #58. 36 42. Rohdin, P.m and Thollander, P. 2006. Barriers to and driving forces for energy efficiency in the non-energy intensive manufacturing 37 industry in Sweden. Energy. 31 (12): 1836–1844. http://dx.doi.org/10.1016/j.energy.2005.10.010. 38 43. Sabatini, F. 2005. The Role of Social Capital in Ecnomiic Development: Investigating the Casual Nexus through Structural Equations 39 Models. 40 44. Sanstad A.H., and Howarth R.B. 1994. Consumer rationality and energy efficiency. Presented at Proc. Summer Study Energy Effic. Build., 41 Berkeley, CA 25 1 45. Sardianou, E. 2008. “Barriers to industrial energy efficiency investments in Greece,” Journal of Cleaner Production, Vol. 16 (13): 1416– 2 1423. 3 46. Schleich, J. (2009). “Barriers to energy efficiency: A comparison across the German commercial and services sector,” Ecological 4 Economics, Vol. 68 (7): 2150–2159 5 47. Schubert, S. F., and S. J. Turnovsky. “The impact of oil prices on an oil-importing developing economy.” Journal of Development 6 Economics, 2011; 94:18-29 7 48. Schultz, P. W., Nolan, J. N., Cialdini, R. B., Goldstein, N. J., and Griskevicius, V. 2007. “The constructive, destructive, and reconstructive 8 power of social norms.” Psychological Science 18 (5): 429–34. http://dx.doi.org/10.1111/j.1467- 9280.2007.01917.x. 9 49. Soderberghn, B., Jakobsson, K., and Aleklett, K. ''European energy security: An analysis of future Russian natural gas production and 10 exports.'' Energy Policy, 2010; 38(12): 7827-7843. 11 50. Sorrell, S., O'Malley, E., Schleich, J., and Scott, S. 2004. The Economics of Energy Efficiency - Barriers to Cost-Effective Investment, 12 Edward Elgar, Cheltenham. 13 51. Sunding, D., and D. Zilberman, (2001) “The agricultural innovation process: Research and technology adoption in a changing agricultural 14 sector” In Handbook of Agricultural Economics. 15 52. Tavakol, M, and R Dennick, (2011). “Making sense of Cronbach’s alpha.” International Journal of Medical Education, 2:53-55. DOI: 16 10.5116/ijme.4dfb.8dfd 17 53. Thollander, P., Danestig, M., and Rohdin, P. (2007). “Energy policies for increased industrial energy efficiency: Evaluation of a local 18 energy programme for manufacturing SMEs.” Energy Policy, 35(11): 5774-5783. 19 54. Trypolska, G. 2012. Feed-in tariff in Ukraine: The only driver of renewables’ industry growth? Energy Policy. 45; 645-653. 20 55. Williams, R.. 2005. “Gologit2: A Program for Generalized Logistic Regression/ Partial Proportional Odds Models for Ordinal Variables.” 21 Retrieved July 07, 2014 (http://www.nd.edu/~rwilliam/stata/gologit2.pdf). 22 56. Williams, R. 2006. "Generalized ordered logit/partial proportional odds models for ordinal dependent variables." Stata Journal 6(1):58-82. 23 57. Worrell, E., L. Bernstein, J. R., Price, L., and Harnisch, J. 2009. Industrial energy efficiency and climate change mitigation. Energy 24 Efficiency. 2(2): 109-123. 25 26 1 Appendix A: 2 Table 1A: Specific questions asked to analyze energy efficiency barriers 1) Economic barriers 1.1) Financial barriers High upfront costs: Are upfront capital costs of energy efficient appliances and devices high? Lack of capital: Do financial institutions (Banks and other financial institutions) perceive energy efficiency investment as risky and therefore charge high premium? Low opportunity costs: Are there other priorities for capital investment, which can produce high returns? Low opportunity costs of appliances to be replaced: Is there any resale value of the replaced appliances, which still has a long operational life? Long payback period: Is payback period of efficient appliances/devices too long to discourage their implementation? 1.2) Hidden costs Skilled personnel: Lack of skilled personnel to handle the efficient devices and processes, especially in their maintenance Supplies: Lack of local supplies for equipment parts and very expensive purchasing from abroad, as well as long lead time to get equipment parts Reconfiguration: Installation of energy efficiency measures needs substantial reconfiguration of production process Malfunction and poor performance: Higher probability of malfunction or poor performance thereby disrupting production process 1.3) Low energy costs Low priority: Low priority of the firm to reduce energy consumption; energy cost is not a big component of production costs 2) Behavioral and informational barriers 2.1) Split incentives Bills paid by landlord: No incentives for the firm to reduce energy consumption as energy bills are paid by building/facility owners Split bills: No incentives for the firm to reduce energy consumption as energy bills are split among the building/facility tenants 2.2) Knowledge, information and experience Metering: Lack of gas, electric and heat metering Awareness: Lack of awareness of the availability and/or benefits of deploying energy efficient processes and devices Information: Difficulties with obtaining necessary information Confidence: Lack of confidence on energy efficient devices and processes (they do not deliver the services at the level their promoters advocate) 27 Experience: Lack of experience in energy efficiency measures 3) Institutional barriers 3.1) Existing Rules and Regulation Government permits: Need to obtain government permits to deploy energy efficient devices and processes Property rights: Lack of legal protection of property rights Policy instruments: Administrative price setting, subsidies and cross subsidies Government policy: Lack of effective government policies to facilitate energy efficiency programs Unofficial payments: Unofficial payments demanded to receive government permits 3.2) Firm’s bureaucracy barriers Decision chain: Long decision chain on the firm The future: Uncertainty about the firm’s future Conflict of interest: Conflict of interests inside the firm 1 2 Appendix B: 3 In computing the common factor attributed to each barrier, unrotated Principle-Component 4 methods were employed. The output of the various runs is depicted in Table 1B.a. 5 Table 1B-a. Principle component analysis Financial barriers: Number of observations 356 Factors retained 1 Number of parameters 5 Factor Eigenvalue Difference Proportion Cumulative Factor1 2.68267 1.97536 0.5365 0.5365 Factor2 0.70732 0.03498 0.1415 0.678 Factor3 0.67234 0.13459 0.1345 0.8125 Factor4 0.53776 0.13785 0.1076 0.92 Factor5 0.39991 . 0.08 1 Chi2(10)= 457.19 Split barriers: 28 Number of observations 382 Factors retained 1 Number of parameters 1 Factor Eigenvalue Difference Proportion Cumulative Factor1 1.68226 1.36452 0.8411 0.8411 Factor2 0.31774 . 0.1589 1 chi2(10)= 238.34 Information and knowledge barriers: Number of observations 433 Factors retained 1 Number of parameters 5 Factor Eigenvalue Difference Proportion Cumulative Factor1 3.41871 2.84219 0.6837 0.6837 Factor2 0.57652 0.13668 0.1153 0.799 Factor3 0.43985 0.08195 0.088 0.887 Factor4 0.35789 0.15086 0.0716 0.9586 Factor5 0.20703 . 0.0414 1 chi2(10)= 1181.82 Hidden costs: Number of observations 420 Factors retained 1 Number of parameters 4 Factor Eigenvalue Difference Proportion Cumulative Factor1 2.31472 1.56538 0.5787 0.5787 Factor2 0.74934 0.2142 0.1873 0.766 Factor3 0.53514 0.13434 0.1338 0.8998 Factor4 0.4008 . 0.1002 1 chi2(10)= 413.15 Existing rules and regulations: Number of observations 296 Factors retained 1 Number of parameters 5 Factor Eigenvalue Difference Proportion Cumulative 29 Factor1 2.57632 1.66628 0.5153 0.5153 Factor2 0.91004 0.31833 0.182 0.6973 Factor3 0.59171 0.07042 0.1183 0.8156 Factor4 0.52128 0.12064 0.1043 0.9199 Factor5 0.40065 . 0.0801 1 chi2(10)= 363.58 Firm’s administrative barriers: Number of observations 401 Factors retained 1 Number of parameters 3 Factor Eigenvalue Difference Proportion Cumulative Factor1 1.88518 1.19138 0.6284 0.6284 Factor2 0.69381 0.27279 0.2313 0.8597 Factor3 0.42101 . 0.1403 1 chi2(10)= 238.16 1 2 We also obtained the extracted sum of the squared loading (Table 1B-b), and use the 3 loading coefficients to calculate the factor that we employ in the empirical analysis. 4 Table 1B-b. Factor loading (pattern matrix) and unique variance Financial barriers Loading Variable Uniqueness coefficient High Up front costs 0.7184 0.484 Lack of Capital 0.7952 0.3677 Low opportunity cost 0.7205 0.4809 Zero or very small monetary value 0.723 0.4773 Long pay back period 0.7018 0.5074 Split barriers Loading Variable Uniqueness coefficient Energy bills paid by building/facility 0.9171 0.1589 owner 30 Energy bills shared among 0.9171 0.1589 building/facility owner and firm Information and knowledge barriers Loading Variable Uniqueness coefficient No metering 0.7696 0.4077 Lack of awareness 0.8895 0.2088 Difficulty obtaining information 0.8623 0.2564 Lack of confidence in these measures 0.8003 0.3596 Lack of experience 0.8069 0.3489 Hidden costs Loading Variable Uniqueness coefficient Skilled labor 0.7026 0.5064 Expensive imports and lack of 0.7938 0.3699 domestic supply Requires substantial changes to the 0.8116 0.3413 production process High probability of malfunction 0.7296 0.4676 Existing rules and regulations Loading Variable Uniqueness coefficient Government permits required 0.7873 0.3802 Lack of property rights protection 0.7467 0.4425 Administrative price setting 0.7174 0.4853 Government policy not effective 0.5878 0.6544 Unofficial payments demanded 0.7341 0.4612 Firm’s administrative barriers 31 Loading Variable Uniqueness coefficient Long decision chains 0.7044 0.5039 Uncertainty about firm's future 0.8139 0.3375 Conflict of interest inside the firm 0.8524 0.2734 1 2 Appendix C: 3 Table 1C depicts the bivariate regressions of the various variables. 4 Table 1C. The bivariate regressors Coefficient Std. Err. t P>t Investment-revenues Revenues 0.52 0.08 6.88 0.00 Cutpoint 1 0.05 0.16 0.34 0.73 Cutpoint 2 2.98 0.23 12.98 0.00 Investment - energy cost share Energy cost share -0.02 0.01 -2.71 0.01 Cutpoint 1 -1.55 0.18 -8.82 0.00 Cutpoint 2 -0.82 0.16 5.01 0.57 Investment - private ownership Private Ownership -0.38 0.12 -3.24 0.00 Cutpoint 1 -1.43 0.10 -14.84 0.00 Cutpoint 2 1.06 0.10 10.69 0.00 Investment - financial barriers Financial Barriers -0.17 0.04 -4.90 0.00 Cutpoint 1 -2.25 0.29 -7.82 0.00 Cutpoint 2 0.17 0.26 0.64 0.52 Investment - split barriers Split barriers -0.24 0.06 -4.27 0.00 Cutpoint 1 -1.47 0.13 -11.08 0.00 Cutpoint 2 1.30 0.14 9.23 0.00 Investment - knowledge barriers Knowledge barriers -0.18 0.02 -7.61 0.00 Cutpoint 1 -2.02 0.15 -13.32 0.00 32 Cutpoint 2 0.63 0.12 5.08 0.00 Investment – hidden costs barriers Technical barriers 0.05 0.03 1.76 0.08 Cutpoint 1 -1.14 0.14 -8.09 0.00 Cutpoint 2 1.35 0.15 8.98 0.00 Investment - existing rules and regulations Existing rules and regulations -0.02 0.03 -0.76 0.45 Cutpoint 1 -1.22 0.20 -6.02 0.00 Cutpoint 2 0.97 0.21 4.64 0.00 Investment - low priority of energy prices Low priority of energy prices -0.24 0.06 -3.99 0.00 Cutpoint 1 -1.44 0.11 -13.75 0.00 Cutpoint 2 1.01 0.11 9.48 0.00 Investment - firm's bureaucracy Firm's bureaucracy -0.06 0.05 -1.24 0.22 Cutpoint 1 -1.50 0.13 -11.18 0.00 Cutpoint 2 1.15 0.13 8.67 0.00 Investment - foreign owned Foreign owned 1.35 0.32 4.21 0.00 Cutpoint 1 -1.13 0.07 -16.60 0.00 Cutpoint 2 1.36 0.08 16.82 0.00 Investment - rent Rent -0.72 0.12 -5.96 0.00 Cutpoint 1 -1.30 0.08 -16.98 0.00 Cutpoint 2 1.20 0.08 14.85 0.00 1 2 33