Project 3: (a) The adoption of energy-saving technologies in energy-extensive firms (UvT) and (b) The adoption of energy-saving technologies in energy-extensive firms (ECN)

In the literature, many possible barriers are cited including lack of information on energy-saving options, scarcity of managerial time or lack of skilled personnel, capital rationing, energy efficiency being just one of many criteria affecting the choice of equipment, type of decision criterion used, and principal-agent problems (the investor not being the same agent as the one who reaps the benefits of the investment). Taking this list of criteria as a starting point, we conducted several surveys (Van Dril and Roos 2004; Vollebergh et al. 2005a; Aalbers et al. 2004a) to identify the major barriers to adoption (see also Van Dril 2005). Among the factors found were uncertainties regarding energy prices, the rate of technological change, lack of awareness of the merits of energy savings, lack of knowledge of existing energy-saving options, and the type of decision criterion used for energy investments (especially whether a clear criterion is used, such as the net present value, the payback period, or rules of thumb). Firm size was found to matter here; with smaller firms applying more often rules of thumb, investment decisions being less driven by relative energy price changes and more by changes in output, facing larger information problems, and proportionally more negatively affected by uncertainty.
These insights have resulted in a variety of theoretical papers regarding especially the role of information and learning and, not unrelated, uncertainty (see Hunter 2007; Van Soest 2005 and Dijkstra and van Soest 2006). Hunter (2007) introduces a model of technology adoption into a competitive industry that allows for firms to learn from each other’s adoption decisions. This learning process allows information about the costs of adopting the new technology to take on aspects of a public good. In light of the presence of the public good, he allow firms to consider forming coalitions to reduce the free-rider problem and ensure that there is adequate provision, and shows that firms form stable coalitions to share information about the cost of the new technology, even as the output market remains competitive. Regarding uncertainty, Van Soest (2005) analyzes what policy instrument is most conducive to investment in energy saving technologies, when firms are confronted with the possibility that new generations of technologies may become available in the (near future).
Interestingly, the results are found to crucially depend on the required stringency of environmental policy: if the government aims to reduce energy use substantially, taxes provide a stronger incentive to adopt the new technology than quotas, whereas the results are reversed in case of more lenient policies (see also Van Soest 2004). Building on Van Soest (2005), Dijkstra and van Soest (2006) explore how stringent policies need to be. More specifically, they analyze the combined impact of uncertainty (regarding environmental damage) and irreversibility, and find that firm heterogeneity (regarding, for example, energy intensity) matters here. If the industry is sufficiently heterogeneous, a larger expected damage translates into more stringent environmental policies, but this is not necessarily the case if the industry is more heterogeneous. The reason is the possibility of overinvestment; if firms are homogenous, a specific energy tax will induce either no firm or many firms to adopt the energy-saving technology. The government wishes some firms to adopt the new technology, but also wants to prevent overinvestment (generally, full adoption is not really socially optimal because the environmental benefits of adoption of the marginal firm adopting falls short of the investment costs). Therefore, the optimal environmental policy tends to be much laxer than in case of more heterogeneous industries.

Another type of uncertainty has been explored by Van Soest and de Groot (2007), and this pertains to uncertainty about energy prices. One of the most important developments in the energy markets over the past decade has been the liberalization of especially the electricity markets (especially relevant for energy-extensive firms as electricity is their main source of energy). Van Soest and de Groot explore the theoretic implications of liberalization, taking into account that (i) liberalization tends to decrease energy prices, but (ii) also may result in larger uncertainty (more price volatility). If indeed electricity market liberalization results in these two changes, investments in energy-saving technologies will be discouraged. Lower electricity prices reduce the profitability of energy0saving technologies, and this is exacerbated by higher volatility; as more volatile energy prices tend to increase the likelihood of regretting having made the investment (if the price of electricity turns out to be even lower than expected), firms decide to postpone their investments. Van Soest and de Groot explore the relevance of these two effects using a small simulation model.

Related to this theoretical work are the empirical analyses as conducted by Kuper and van Soest (2003), De Nooij et al. (2003) , Van Soest (2006), and Kuper and van Soest (2006). These studies use industry-level data to analyze the relative importance of the various determinants of energy use, including energy price changes, changes in (the demand for) output, and uncertainty. De Nooij et al. (2003) provide a benchmark along which the performance of countries can be evaluated in this respect, and Kuper and van Soest (2003) provide a first step in the analysis trying to determine how the substitutability of labor, energy and capital depends on the business cycle. The latter analysis is taken a step further by Kuper and van Soest (2006) who develop a new measure of uncertainty that contains more information than just the usual measures (in particular, the unconditional variance in the variable under consideration). They construct a conditional variance measure that takes into account the general trends in the variable of interest, and hence come up with a cleaner measure of uncertainty than commonly used in the literature. Van Soest (2006) applies this measure to analyze the impact of both the level and the uncertainty regarding output and energy prices on adoption behavior by energy-extensive firms (as proxied by the amount of energy used). This study clearly indicates that output (and the conditional variance thereof) is a much more important determinant of investments by energy-extensive firms than relative price changes. It also indicates that government policy is likely to be much more effective in times of economic booms than in times of recessions because only in these periods energy taxes (or similar policy instruments) can substantially effect the extent to which new technologies are energy-saving or energy-consuming.

The general conclusion of the theoretical and empirical literature addressed above is that the instrument that economists like best, energy taxes, is not likely to be very effective in inducing adoption of energy-saving technologies by energy-extensive firms. The stakes are not sufficiently high, they do not solve the information problem regarding which technology to adopt, etc. Therefore, we turned to analyzing the potential effectiveness of adoption subsidies in inducing investment. We did so in several ways, by means of a survey (Aalbers et al. 2004a,b) and by means of a series of economic experiments (Vollebergh et al. 2005a,b); for an overview, see van Soest and Vollebergh (2007). Aalbers et al. (2004a,b; 2007) use a survey to identify factors that influence adoption of energy-saving technologies, and analyze whether and how relatively small enterprises use investment criteria when making investment decisions. It turns out that both the size of the firm and the sector in which the firm operates matters. Interestingly, it does not seem to matter whether the firm operates in a market environment or is a non-profit firm. Using a survey the paper also estimates whether a subsidy scheme on specific investments, like the Energie Investerings Aftrek (EIA) in the Netherlands, affects behavior of those who are subsidized. It turns out that ‘windfall gains’ of this EIA-subsidy depend on the profitability of the technology and whether the firm uses an investment decision criterion, or not.

These insights are useful, but it is hard on the basis of survey data to infer the actual motivations underlying the decisions made by firms. Therefore, we resorted to an alternative approach – using economic experiments to infer the effectiveness of subsidies in various circumstances. Economic experiments can be used to systematically vary the context in which investment decisions are being made (including search costs, uncertainty of search success, the existence of adoption externalities, the type of subsidy provided — lump sum, or variable). Students are used as subjects, but also managers of small and medium-sized enterprises. The results are revealing in the sense that, not very surprisingly, subsidies are effective in inducing adoption – even if the technology purchased is not profitable according to a standard net present value criterion – especially because of the attention value of subsidies. Offering subsidized technologies mitigates the information problem (and hence the search costs) of the entrepreneur, and even if subsidies are insufficient to make the investment opportunity a profitable one, entrepreneurs may decide to adopt it. This effect is strong and robust for differences in subsidy design or decision environments. The policy implication is that a list with technologies that is ex ante available is very important for the effectiveness of subsidy schemes, but also that the list need not be large. Looking at both studies together tax relief and (flexible) depreciation allowances seem to be most effective. A technology list also directs firms in their decisions. Overall effectiveness is largest with a small subsidy on many technologies. Technologies should also be screened on their profitability. Finally, it seems useful to link the absolute level of subsidy to the additional savings.

So, subsidies are found to be effective in inducing adoption, but they are not always efficient in the sense that only firms that actually need financial support in order to adopt energy-saving technologies claim subsidies. Indeed, as firms are heterogeneous with respect to the benefits they reap from adoption, some may adopt the technology even if they are not subsidized. This problem was empirically assessed by Aalbers et al. (2007) , but the issue whether firms can be separated according to whether or not they need a subsidy, was taken on by Arguedas and van Soest (2005, 2006). Their starting point is that governments can induce adoption by means of lump-sum subsidies, but also by means of taxes -in fact, very often both instruments co-exist, as is the case in the Netherlands. For the same tax level, firms always prefer receiving larger subsidies to smaller subsidies; to induce separation, a menu of tax/subsidy combinations is called for, with -in the case of two firm types-, one combination consisting of a small subsidy but also a low tax rate, and one with large subsidy and a high tax rate. Because firms differ with respect to the variable costs associated with energy use, for the same increase in subsidies low-cost firms are willing to accept a higher tax rate than high-cost firms. Thus, the two types can always be separated, but the conditions under which such a separation scheme is preferable to a uniform system, are fairly specific. That means that Arguedas and van Soest come to a very strong conclusion: the subsidy free-rider problem can be solved, but at substantial costs only.

Having analyzed the barriers to adoption (theoretically and empirically) as well as the policy
responses to that, we have been able to infer implications for the way in which energy investments should be treated in the various energy models available in the Netherlands. Various papers address this issue, including Van Dril (2004), Vollebergh and de Zeeuw (2004), Daniels (2007). Models in use in the Netherlands have been intensively analyzed and various suggestions have been made for improvement. For the SAVE model on industry and agriculture, adjustments are planned at ECN for 2007.
Van Groenendaal (2004) shows the effects three different theoretical starting points for energy models have on policy evaluation and advice. Three energy models, NEMO of Netherlands Bureau for Economic Policy Advice (CPB), SAVE of the Energy Research Centre of the Netherlands (ECN), and MEI-Energy of the National Institute of Public Health and Environment (RIVM), have been used to analyze the same policies for four energy extensive sectors. The results obtained are surprisingly different. This is remarkable since each of these sector models uses the same data on technology and economic growth -which is exogenous in all three the models-, and has a vintage structure.