The adoption of green energy technologies: The role of policies in Austria, Germany, and Switzerland.

We contribute to the existing research about policy-induced technology adoption in several ways. First, we suggest a new survey design to measure the energy-related policy environment. Second, we simultaneously estimate the policy effects for the adoption propensity and the adoption intensity simultaneously and, third, we compare the policy effects in the three countries, Austria, Germany, and Switzerland. Based on a representative sample of firms for all three countries we find that policies essentially promote the adoption of technologies and they are practically ineffective for the intensity, which poses a great challenge to future policy designs. Voluntary agreements or demand-related factors are among the most important drivers for the adoption propensity of green energy technologies. Given the current institutional framework in the surveyed countries, subsidies are more effective in Austria, taxes are more effective in Germany, and demand-related factors are relatively more effective in Switzerland.

Introduction

Despite the growing literature on the adoption of energy-efficient technologies and the well-elaborated theoretical background on technology diffusion (Battisti and Stoneman 2003, 2005; Karshenas and Stoneman 1995), gaps remain in the empirical literature limiting our understanding of the effects of policies on the adoption of energy-related technologies.

First, current literature either focuses on the propensity of adoption (Arvanitis and Ley 2013; Veugelers 2012) or intensity of adoption (Stucki and Woerter 2016) rather than recognizing that the two activities have different features. Secondly, data availability seriously limits cross-country comparisons depriving the possibility to identify policy effects based on identical definitions, an identical theoretical background, an identical model specification, considering country-specific factors. With the paper at hand, we address both issues. We simultaneously estimate the relationship between energy-related policies including voluntary measures and the propensity to adopt environmentally friendly energy technologies (inter-firm) and the effect of policies on the intensity of investments in such technologies (intra-firm). Based on a representative panel of firms for Switzerland, Germany, and Austria, respectively, we collected data for all three countries using a commonly designed questionnaire. The surveys have been conducted by KOF (Swiss Economics Institute; Switzerland), ZEW (The Centre for European Economic Research, Germany), and WIFO (Austrian Institute of Economic Research, Austria). These are institutions with many years of experience in collecting survey data and the applied survey design has been frequently used to provide survey-based information for official statistics in the respective countries.

Energy technologies that increase the energy efficiency or that use renewable energy sources. Such Technologies may include tangible equipment as well as information and management systems, In the remainder of the paper, we use the term ‚green energy technology’ to denote these technologies.

An important contribution to the existing literature also refers to the measurement of the energy-related policy environment. Ghisetti and Pontoni (2015) conducted a meta-analysis of policy effects on environmental innovation and found that the operationalization of the policy information is crucial for the result. Hence, we made a great effort to measure policies as exactly and unbiased as possible using surveys as a tool for collecting this information. We wanted to avoid the frequently observed issue that policy information is only available for firms that adopted green energy technologies and we also wanted to avoid a potentially “common method bias” by posing a question that contains the type of policy and its effect in one sentence. This led to a new design of the questionnaire where we excluded a direct link between the policy exposure and the adoption behavior of the firm. We asked firms to rate the importance of a series of energy-related factors, including government policies (taxes and levies, regulations, subsidies), voluntary agreements, demand-related factors, and factors related to the energy environment (fluctuation of energy prices, (feared) bottlenecks in energy supply). Clearly, this question asks for a “subjective” response. The usual critic for this type of question (see Bertrand and Mullainathan 2001) does not necessarily apply, however. First, the respondents usually think of an “objective” event in the firm or in its environment, or an important discussion that helps him/her to decide if a policy-related factor is important or not. Second, the list of policy-related factors helps the respondents to assess the relative importance among the factors asked, which is important for the applied econometric analysis.

For instance: “Did the firm introduce and environmentally friendly technology in response to regulation x?”.

Based on a comprehensive data set—which allows for a rich control vector—covering the countries Austria, Germany, and Switzerland, we apply standard econometric procedures to identify the relationship between different policies and the propensity to install green energy technologies. Moreover, we identify their meaning for the investment intensity in such technologies. We can identify country differences for both aspects of adoption (propensity/intensity) and we can compare the results across different countries. We also investigate if firm characteristics—like “energy intensity”—drives the adoption. In a more explorative investigation, we analyze if policy effects differ across technologies and if their effect differs between technology adoption and the adoption of environmental management systems (non-technology measures).

The direct comparisons of policy effectiveness for the propensity of adoption and the intensity of adoption, respectively, reveals some interesting insights. Policies in all three countries essentially promote the adoption of technologies and they are practically ineffective for the intensity, which poses a great challenge to future policy designs; the intensity of adoption is mainly driven by firm specific characteristics, like the energy costs of the focal firm and firm size. Only public subsidies show a significant and positive relationship with the investment intensity in green energy technologies. Voluntary agreements or demand-related factors are among the most important drivers for the adoption of such technologies.

Exploratory estimations on technology specific effects of the investigated factors reveal that demand-related factors and public subsidies are important for the adoption of all identified technologies (excluding transport), whereas the adoption of green energy technologies in the field of production is significantly driven by taxes and in the field of buildings voluntary agreements are very important. The adoption of energy management systems is driven by taxes, voluntary agreements, and public subsidies.

In general, national promotion plans relying on subsidies and demand-related factors appear to be effective for the adoption of a broad spectrum of technologies and green management systems. Taxes and voluntary agreements are also overall effective but they show some technology-specific patterns, which can be due to the lack of initiatives related to such instruments. Given the current institutional framework in the respective countries, subsidies are more effective in Austria, taxes are more effective in Germany, and demand-related factors are relatively more effective in Switzerland.

The paper is structured as follows. Section 2 provides a review of the literature and identifies relevant hypotheses to be tested in the empirical part of the paper. Section 3 presents some important details about the survey design and the data description. Section 4 deals with the econometric framework applied to test the hypotheses. Section 5 presents the results and in section 6 we conclude and point at important policy implications. We briefly introduce the policy environment in the investigated countries in the Appendix.

Literature review and hypotheses

Given the presence of market failures concerning the adoption of environmentally friendly technologies (see e.g. Jaffe and Stavins 1994), theoretical studies argue that long-term incentives to adopt environmentally friendly technologies are stronger under market-based instruments (e.g. subsidies, taxes) than they are under command-and-control instruments (see e.g. Downing and White 1986; Milliman and Prince 1989). There are at least two reasons for this. First, market-based instruments have dynamic advantages. They tend to give permanent incentives to further decrease the environmental burden of the production process, since any further improvement of the environmental performance yields lower costs in the form of permits that can be sold, subsidies that can be received, or taxes that can be saved (see e.g. Cleff and Rennings 1999; Jaffe and Stavins 1995). Secondly, market-based instruments allow for a cost-effective response to environmental issues, the firm can choose the most cost-effective way to improve its environmental performance according to its characteristics. In contrast, environmental standards and regulations allow for a less flexible firm response and have adverse dynamic characteristics, since once a standard or the regulatory requirements have been satisfied, there are hardly any incentives to further adopt newer technologies to abate contamination (Jaffe and Stavins 1995).

See also Stucki and Woerter (2016) for an overview of theoretical arguments in favor of market based policies.

Voluntary agreements are a third type of instruments to pursue the goal of more environmentally friendly production processes (Croci 2005; Veugelers 2012). They can be individual agreements or collective or branch agreements. Voluntary agreements share the dynamic advantages of market-based instruments; however, they are even more flexible, which contributes to efficiency and effectiveness and they can be applied to a variety of circumstances taking into account local and branch peculiarities. Given these advantages from a point of view of a policy maker, it is still questionable “why firms should enter into such agreements”. Croci (2005, 11) suggests a number of reasons. For instance, firms want to avoid stricter regulation (see also Florida and Davison 2001; Rondinelli and Berry 2000), or lower the adaption costs through better information which results from the negotiations to reach an agreement, they can gain reputation and most importantly they obtain their flexibility in choosing the adaptation process. Although these aspects point at the superiority of voluntary agreements, their effectiveness might suffer, since voluntary agreements does not necessarily involve all polluters and enforcement is usually less strictly. Due to their cooperative nature and their flexibility, it is likely that such agreements—at least in markets with a relatively low diffusion rate of green energy technologies—essentially support the diffusion of such technologies.

A comprehensive categorization of voluntary agreements includes voluntary public schemes, negotiated agreements, unilateral commitments recognized by the public administration, unilateral commitments, third party initiatives, and private agreements (Croci 2005, 7).

Market demand for environmentally friendly produced products is considered as a fourth type of incentive. Arguing from the perspective of the focal firm, market-based instruments, regulation, and voluntary agreements are strongly related to its cost side, whereas the demand side is largely ignored in studies about technology adoption, although it should be one of the most effective forces for profit-oriented firms, since it provides full technological flexibility and low transaction costs. Hence, market demand (policy driven or not) for environmentally friendly produced products should be an effective factor for adopting environmentally friendly technologies.

Following the literature (see e.g. Cleff and Rennings 1999) it is clear that both, costs and the demand factor are important for the adoption decision of a firm. However, it is difficult to identify, which policy instruments are most effective in encouraging the adoption, since the design of the instruments, the characteristics of the regulator (Requate and Unold 2003), and the characteristics of the affected firms play a role as well (Jaffe and Stavins 1995; Magat 1978; Malueg 1989; McHugh 1985). However, given that only regulation shows serious theoretical adverse characteristics, we want to formulate our first hypothesis:

Regulation is less effective than market-based policies, voluntary agreements, or demand to increase the propensity of technology adoption

In order to observe far reaching positive environmental effects of technology adoption is not sufficient that energy-related technologies are adopted, it is essential that such technologies also diffuse within the firm (intra-firm diffusion) or to express it differently, that they are used intensively. But how policies can also increase the intensity of adoption? Actually we could assume that the expected policy effect resembles the effect for the adoption. However, this is not necessarily the case for at least two reasons.

First, the benefits from the environmentally friendly policy regime are fully exploited with installing a specific technology, for example, an end-of-pipe solution if the regulation requires it, and no further incentives are provided. The same is true for voluntary agreements. Ones the condition of a such an agreement are met, and the environmental friendliness of a new, for example, production technology, is secured and communicated, for example with a label, then there might be no further incentives to invest more in environmental friendly technologies (intensity of adoption). However, this is not necessarily the case with taxes/levies or subsidies. Those instruments potentially provide dynamic advantages, meaning that the benefits increase proportionately with the intensity of adoption. This brings us to the second reason.

Second, a proportional increase of benefits resulting from the policy framework is likely to be insufficient to induce further environmentally friendly investments, since the costs for additional energy savings increase disproportionately, similar to abatement costs. Kuik, Brander, and Tol (2009) in a meta-analysis of marginal abatement costs identified a very skewed distribution meaning that the median abatement cost level is significantly lower than the means of abatement costs. This indicates that the marginal costs increase disproportionally unless technological development significantly lowers them (Baker, Clarke, and Shittu 2008). Hence, firms have to make disproportionately greater financial efforts to reduce energy consumption beyond a level that is suggested by the dominant technology and without incentives resulting from an improved design of the environmental policy (Hemmelskamp 1997; Kerr and Newell 2003) or other economics factors, like the market structure, technological opportunities, demand factors, or the energy intensity, the focal firm is unlikely to improve its environmental performance by increasing its investments. However, the policy effects on investment intensity are not necessarily positive. Zhao (2003) confirmed the results of Milliman and Prince (1989) and Jung, Krutilla, and Boyd (1996) stating that market-based policies, like emission pricing, can even lower the incentives to intensively invest in abatement technologies or—in our case—energy-saving technologies, since tradable permits should become cheaper the more firms adopt environmentally friendly process technologies. However, under investment uncertainty, specific types of policies can reduce investment incentives. Zhao (2003) found that emission charges are likely to reduce investment incentives, even greater than under tradable permits. Our second hypothesis reads as follows:

The policy effects for higher investments in green-energy technologies are lower than for the adoption propensity.

Empirical studies on adoption propensity (inter-firm diffusion)

The empirical evidence vastly confirms H1. Kerr and Newell (2003) investigated the technology-adoption behavior of refineries during the lead phasedown in the USA and they found that the adoption propensity of refineries with low compliance costs was greater under market-based lead regulation than under performance standards. Jaffe and Stavins (1995) investigated the adoption decision for thermal insulation technologies under different policy regimes; these are taxes, subsidies, and technology standards. They found that the technological cost effect is three times larger than the adoption effect of energy savings over the whole lifetime of the investment. Hence, it is comprehensible that subsidies exert a much higher positive effect than taxes on the adoption propensity. Quite contrary to subsidies, the building codes (command-and-control instrument) did not contribute to the diffusion of insulation technologies. Since the observed building codes did not make any significant differences to the existing building practices, the authors cannot exclude that more stringent codes might have had an effect.

Switching to studies based on European data, van Leeuwen and Mohen (2017) used Dutch survey data from four different sources—environmental cost survey, the energy use survey, the community innovation survey, and the production statistic survey—to investigate the policy effects for the adoption of environmentally friendly process innovation. Applying a structural model in the spirit of the CDM (Crepon, Mairessec, and Duguet 1998) model, they found that levies and regulation are positively correlated with environmentally friendly investments in process innovation (0/1 decision). However, due to the dependent variable (eco investments 0/1) it was not possible to insert subsidies in the adoption equation, which might drive the observed effects of the other policy variables (levies, regulation). Cleff and Rennings (1999) used German data and investigated the policy effect on “innovation integrated measures on the process level”. They found evidence that the so-called “soft” factors are very supportive to environmentally friendly process innovations, whereas among other state regulation did not show any effect. The term soft-factors refers to voluntary agreements and informational instruments. Frondel, Horbach, and Rennings (2007) also investigated environmentally friendly innovation decisions based on survey data for Germany. Their results suggest that the adoption of cleaner production technologies tends to be market driven and not so much by regulatory measures. Similarily, Horbach, Rammer, and Rennings (2012) investigated the influence of policies on firms’ decision for adopting environmentally friendly process innovations. They identified technologies according to their environmental impact; for instance, to save energy or to reduce C02 emissions. Concerning energy-related technologies they found that self-commitment (voluntary agreements) and demand for environmentally friendly produced products show a significant and positive effect; regulation and even subsidies were not significantly related to energy-related processes innovations.

Veugelers (2012) used adoption motives as a measure of policy affectedness. Based on Flemish CIS-data she assessed the responsiveness of firms to environmentally friendly policy interventions. Besides the generation of green technologies, she also assessed the effect of the policies on the propensity to adopt environmentally friendly technologies. She found that regulations/taxes show a larger effect than subsidies and voluntary industry codes and agreements are important drivers for introducing environmentally friendly technologies.

Demirel and Kesidou (2011) used data from 289 UK firms stemming from the Government Survey of Environmental Protection Expenditure by Industry’ in years 2005 and 2006. They measured the use of cleaner production technologies by their investment level. Hence they investigated both the effect of policies on the adoption decision (0/1) and on the intensity of adoption (see below). Neither regulation nor the market based instrument (taxes) showed a significant effect in the full model specification.

Empirical studies on the level of investment (intra-firm diffusion)

The empirical evidence about the relative inducement effect of several environmental policies for the intra-firm diffusion of green energy technologies is relatively scarce.

van Leeuwen and Mohen (2017) could use a comprehensive Dutch data set (see above) and they did not only investigate the 0/1 choice for process-integrated, environmentally friendly investments, they also investigated the policy effects on the investment intensity using the Crepon, Duguet, and Mairessec (1998) framework. The 0/1 choice and the intensity effects have been estimated simultaneously taking into account potential selection issues. They found that subsidies and regulation show a significant effect for investment intensity (van Leeuwen and Mohen 2017). Moreover, energy intensity is also significant and positively related with the dependent variable.

However, if they lag the non-policy covariates by two periods the effect of regulation gets insignificant. (see earlier version of the paper: van Leeuwen and Mohnen, 2017.

Stucki and Woerter (2016) based on Swiss survey data could not observe the relationship between different type of policies and non-political, voluntary agreements and the adoption decision (0/1). However, they could observe the effect of different instruments on the number of environmentally friendly, energy-related technologies along the value-added chain of a firm. Considering potential selection issues they found that taxes and regulation are the most effective policy instruments for increasing the intra-firm diffusion of environmentally friendly, energy-related technologies. Taking into account non-political motives, it turned out that “voluntary agreements” do significantly increase the adoption intensity and that they are even more effective than policy measures. Hence, taxes, regulation, and voluntary agreements are the most important motives for the intra-firm adoption of green energy technologies.

Like mentioned above, Demirel and Kesidou (2011) used firm-level information of about 289 UK firms and measured the level of investments into integrated cleaner production technologies. In their main estimations, neither environmental regulation nor environmental taxes significantly increased the investments in cleaner production technologies.

Given these empirical results so far, we get the impression that the observed results strongly depend on the measurement of policies, the number of types of policies controlled for in the model, and the measurement of the intensity variable. Van Leeuwen and Mohen (2015) for the Netherlands and Demirel and Kesidou (2011) for the UK measured the intensity of the adoption of process-integrated technologies by the size of investments and Stucki and Woerter (2016) for Switzerland counted the number of different technologies adopted to measure the intra-firm diffusion of energy-related technologies and all three studies present quite different results. Hence, the empirical findings concerning H2 are mixed and the measurement issues gain in importance.

Survey and data description

We designed a particular questionnaire in order to collect the data necessary to test the above-mentioned hypotheses. This questionnaire laid the basis for a comprehensive survey with the topic “creation and adoption of energy-related technologies”, carried out simultaneously in three countries (Austria, Germany, and Switzerland) in 2015. It was sent to 5789 Swiss firms (KOF-Enterprise Panel), 6374 German firms (ZEW-Enterprise Panel), and 7091 Austrian firms (WIFO Enterprise Panel). The representative firm samples for the respective countries comprise the whole manufacturing, service, and construction sector. It is a stratified random sample, whereas stratification takes place on the two-digit industry level and on three industry-specific firm size classes (with full coverage of large companies).

Publicly available data is not sufficient to identify the relative effects of different policies. It need survey data for at least two reasons (see Stucki et al. 2016). First, to get a complete picture, all relevant policies would need to be identified, which is hardly possible as they can be firm/sector- and technology-specific. Second, also the stringency of single policies matters, which is a difficult to measure and hardly available from public sources (Shadbegian and Wolverton 2010; Xing and Kolstad 2002). The difficulty arises, since different policies typically cover different firms or policies may exist at multiple levels (e.g. federal and local), and monitoring and enforcement are imperfect (Millimet and Roy 2016).

See Arvanitis et al. (2016) for the full sample information.

The response rates are quite different across the countries. The survey yielded valid information for 1815 firms (31.4%), 2321 (36.4%), and 539 firms (7.6%) for the countries Switzerland, Germany, and Austria, respectively. Given the very demanding questionnaire the response rates for Switzerland and Germany are satisfying, whereas the return from Austrian companies was disappointing for various reasons. However, a comprehensive recall action in all three countries ensures that we received a sufficient large number of answers covering all industries and all firm size classes according to the sampling scheme (cells). Only for Austria some cells could not be filled.

In part, the lower return in Austria was expected, since different from the German and Swiss samples, we had no established innovation panel to rely on. Return rates for so called “cold” panels, where all firms must be contacted de novo are generally lower. However, we also experienced unanticipated problems, which relate to the introduction of a new energy efficiency law in Austria (see the previous section).

Due to missing values for some questions the estimations presented in this paper are based on 3465 observations. However, due to specific questions referring only to a subgroup of firms, the number of observation can be lower (see estimations below).

The questionnaire contains questions about some basic firm characteristics (sales, exports, employment, investment, and employees’ education). Moreover, it contains questions about the importance of energy-related factors, like energy-related taxes and levies, energy-related regulations, energy-related voluntary agreements, public subsidies, and demand-related factors. Moreover, we have information about the fluctuation of energy prices and (feared) bottlenecks in energy supply. A unique characteristic of the questionnaire is that all firms needed to answer such policy-related factors, independently if they adopted energy-relevant technologies or not. This has the advantage that we can identify the meaning of the policy environment for all firms and not only for firms that have already adopted environmentally friendly, energy-related technologies, like it is usually found for policy-related questions in the CIS (Community Innovation Survey) type of questionnaire. Hence, we can contrast (for the first time) the effectiveness of policy related factor for the adoption stage (inter-firm diffusion) with their meaning for the intensity stage (intra-firm diffusion).

Table 1 presents the variable definitions and some descriptive information for the variables included in our main estimations. 50% of the responding firms adopted environmentally friendly, energy-related technologies and on average they spent around 39,000 Euros for the adoption of such technologies. On average firms perceive taxes (1.71) as the most relevant energy-related policy followed by regulations (1.49) and public subsidies (1.46). Firms also perceive high/volatile energy prices as a relative important characteristic of their energy-related environment. 54% of our sample firms export, 11% are foreign-owned and 38% have R&D activities. The competition intensity—measured by the number of principal competitors—is moderate on the means level; firms tend to operate in markets with less than 15 competitors worldwide. The energy costs are on average about 115,000 Euros, however, the distribution seems to be rather skewed, since there are few firms with rather high values. Adoption of environmentally friendly, energy relevant technology is no topic or has not been discussed for 39% of the firms.

Table 1.

Descriptive information.

MAIN VARIABLES	Description	N	Mean	SD	Min	Max
Adoption 0/1 (DEPENDENT)	Adoption of green energy technologies, i.e. technologies that contribute to increase energy efficiency and/or enable the use of energy from renewable resources	3535	0.49		0.00	1.00
Energy investments (ln) (DEPENDENT)	Investments for the adoption of green energy technologies	1621	10.56	2.46	0.69	18.39
Taxes	The importance of energy-relevant taxes and levies for the focal firm	3535	1.71	0.74	1.00	3.00
Regulations	The importance of energy-relevant regulations for the focal firm	3535	1.49	0.68	1.00	3.00
Voluntary agreements	The importance of energy-relevant voluntary agreements or standards in the sector of the firm	3535	1.40	0.63	1.00	3.00
Public subsidies	The importance of energy-relevant subsidies for the focal firm	3535	1.45	0.68	1.00	3.00
Demand	The importance of demand for energy efficiently produced products	3535	1.33	0.60	1.00	3.00
CONTROL VARIABLES
High/volatile energy prices	The importance of high/volatile energy prices	3535	1.78	0.73	1.00	3.00
Energy shortage	(feared) bottlenecks in energy supply	3535	1.26	0.55	1.00	3.00
Number of employees (ln)	Number of employees	3535	3.91	1.56	0.00	11.63
Export	Export activities	3535	0.54		0.00	1.00
Foreign owned	Focal firm belongs to a foreign firm	3535	0.11		0.00	1.00
R&D propensity	R&D activities of the focal firm	3535	0.38		0.00	1.00
Share graduated employees	Share of employees with a tertiary degree	3535	17.33	22.68	0.00	100.00
Competition intensity	Number of principal competitors in the main sales market of the focal firm worldwide; 1 (1–5 competitors). 2 (6–10), 3 (11–15), 4 (16–50), 5 (50+)	3535	2.57	1.46	1.00	5.00
Energy costs	Energy costs of the focal firm	3535	11.67	2.21	0.51	20.27
Gross tangible investments (ln)	Gross tangible investments	3535	12.25	2.34	3.94	21.92
Epidemic adoption	Share of firms that adopted energy-related technology in the sector of the focal firm (excluding the focal firm)	3535	0.45	0.13	0.00	0.94

These are unweighted figures; for the representative figures please refer to Arvanitis et al. (2016).

Table 2 presents some country specific information. First, we see different adoption propensity ranging from about 40% in Switzerland to 56% in Austria. Moreover, we see that taxes are perceived as more relevant in Germany than in the other countries, voluntary agreements are of relatively high relevance in Switzerland, and public subsidies is the dominant policy factor in Austria.

Table 2.

Descriptive information per country.

SWITZERLAND	N	Mean	SD	Min	Max
Adoption 0/1	1584	0.40	0.49	0.00	1.00
Energy investments	600	11.28	2.31	3.73	17.29
Taxes	1584	1.63	0.69	1.00	3.00
Regulations	1584	1.56	0.70	1.00	3.00
Voluntary agreements	1584	1.43	0.65	1.00	3.00
Public subsidies	1584	1.40	0.62	1.00	3.00
Demand	1584	1.31	0.57	1.00	3.00
High/volatile energy prices	1584	1.65	0.69	1.00	3.00
Energy shortage	1584	1.27	0.55	1.00	3.00
GERMANY
Adoption 0/1	1633	0.55	0.50	0.00	1.00
Energy investments	845	9.89	2.42	0.69	18.39
Taxes	1633	1.79	0.77	1.00	3.00
Regulations	1633	1.43	0.66	1.00	3.00
Voluntary agreements	1633	1.36	0.62	1.00	3.00
Public subsidies	1633	1.49	0.72	1.00	3.00
Demand	1633	1.34	0.62	1.00	3.00
High/volatile energy prices	1633	1.92	0.74	1.00	3.00
Energy shortage	1633	1.24	0.55	1.00	3.00
AUSTRIA
Adoption 0/1	300	0.56	0.50	0.00	1.00
Energy investments	165	11.34	2.25	3.91	17.27
Taxes	300	1.67	0.75	1.00	3.00
Regulations	300	1.44	0.66	1.00	3.00
Voluntary agreements	300	1.40	0.63	1.00	3.00
Public subsidies	300	1.61	0.75	1.00	3.00
Demand	300	1.37	0.63	1.00	3.00
High/volatile energy prices	300	1.76	0.73	1.00	3.00
Energy shortage	300	1.28	0.58	1.00	3.00

Correlation coefficients for the main model variables are presented in Table 3. Most importantly we see that the dependent variables are significant and positively related with the energy-relevant factors, whereas the correlation coefficients tend to be higher for the adoption variable then for the investment variable (except regulation). Among the policy-relevant factors, we also see positive and significant correlation with relatively high coefficients between demand and subsidies, subsidies and voluntary agreements, voluntary agreements and regulation, and regulation and taxes. Hence, the correlations indicate that it is important to insert all factors in the model to identify the effects precisely.

Table 3.

Correlation between important variables.

	(a)	(b)	(c)	(d)	(e)	(f)	(g)	(h)	(i)	(j)	(k)
(a) Adoption 0/1 (DEPENDENT)	1.000
(b) Energy investments (DEPENDENT)	na	1.000
	na
(c) Energy costs	0.241	0.698	1.000
	0.000	0.000
(d) Number of employees	0.256	0.681	0.787	1.000
	0.000	0.000	0.000
(e) Taxes	0.203	0.158	0.261	0.138	1.000
	0.000	0.000	0.000	0.000
(f) Regulations	0.178	0.252	0.304	0.216	0.506	1.000
	0.000	0.000	0.000	0.000	0.000
(g) Voluntary agreements	0.210	0.205	0.255	0.197	0.380	0.549	1.000
	0.000	0.000	0.000	0.000	0.000	0.000
(h) Public subsidies	0.264	0.141	0.165	0.130	0.318	0.384	0.409	1.000
	0.000	0.000	0.000	0.000	0.000	0.000	0.000
(i) Demand	0.177	0.054	0.075	0.105	0.155	0.274	0.329	0.420	1.000
	0.000	0.030	0.000	0.000	0.000	0.000	0.000	0.000
(j) High/volatile energy prices	0.197	0.178	0.231	0.129	0.597	0.399	0.307	0.291	0.127	1.000
	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
(k) Energy shortage	0.069	0.075	0.110	0.058	0.307	0.282	0.238	0.240	0.171	0.296	1.000
	0.000	0.003	0.000	0.001	0.000	0.000	0.000	0.000	0.000	0.000

Econometric framework

In order to analyze the effects of different types of policies on the inter-firm and intra-firm diffusion of environmentally friendly, energy-related technologies, respectively, we refer to an econometric framework in the spirit of Battisti, Canepa, and Stoneman (2009), which is an extension of Karshenas and Stoneman (1995). Such models have been applied, for example, by Hollenstein and Woerter (2008) for e-commerce adoption, by Arvanitis and Ley (2013) in terms of inter- and intra-firm diffusion of energy-saving technologies, or by Stucki and Woerter (2016) ; concerning the intra-firm diffusion of energy-related technologies.

Stucki and Woerter (2016) provide a very similar econometric framework, however their estimations are limited to intra-firm diffusion and they could not—due to the lack of policy information—compare policy effects for inter- and intra-firm diffusion or apply this framework simultaneously to different countries.

According to Battisti, Canepa, and Stoneman (2009); the first use (inter-firm diffusion) of a new technology and the intensity of its usage (intra-firm diffusion) in time t by firm i in industry j, Di(t), are determined by five categories of variables: First, a vector of characteristics of a firm Ri(t) and its environment Rj(t) reflecting rank effects referring to, e.g., energy intensity, competition, qualification of the employees, or demand. Secondly, the extent of industry usage of new technology SOj(t) to capture inter-firm stock and order effects (i.e., market-intermediated externalities). Thirdly, epidemic effects (i.e., learning and network non-market intermediated externalities) reflecting either the firm’s own experience with the new technology Ei(t), often proxied by the time since the firm’s first adoption, or the experience gained from observing other firms Ej(t) (often measured by the extent of technology diffusion among similar firms in time t). Fourthly, the expected adoption cost of a unit technology Pi(t) that is constituted by two parts: one common to all firms, for example, the price of a new, energy efficient technology; and a second one reflecting firm-specific adjustment and installation costs. Fifthly, in accordance to the particular conditions of the introduction of green energy technologies in Switzerland (as in many other countries), also elements of the literature on induced innovation and technology diffusion (see, e.g., Binswanger 1974) are taken into consideration. The diffusion of green energy technologies can be positively influenced (a) through increases of energy prices and/or taxes (see, e.g., Linn 2008; Jacobs, Kuper, and Van Soest 2009) and (b) through public regulation and/or public incentives (subsidies), and/or “voluntary agreements” to use green energy technologies (see, e.g., Croci 2005; Popp, Newell, and Jaffe 2009). We consider a vector IAi(t) of variables that capture the influence of such factors (inducement effects). We therefore arrive at the following equation that we use for estimating the adoption models:

Rank effects refer to firm heterogeneity that could lead to differing returns to adoption and consequently to different adoption behavior (Davies 1979; Ireland and Stoneman 1986).

Stock and order effects refer to market-intermediated externalities either due to early-mover advantages (order effect) or lower adoption costs due to a higher number of firms utilizing a technology (stock effect) (Battisti, Canepa, and Stoneman 2009).

Epidemic effects refer to learning and network effects either from the firm’s own experience gained from using a technology or the knowledge gained from observing other users of the technology (Battisti, Canepa, and Stoneman 2009; Hollenstein and Woerter 2008).

For the empirical implementation of the model, we follow as far as possible Arvanitis and Ley (2013) and Stucki and Woerter (2016). Firm-specific rank effects are measured by (a) the firm’s number of employees, (b) gross physical investments, (c) the qualification level of the employees, (d) R&D activities, (e) export activities, (f) foreign ownership, and (g) the relevance of energy-related technologies for the focal firm. The R&D activities and the qualification level of the employees also control for the firms’ absorption capacity, which is important as not each company has the same conditions to deal with the adoption of new energy-related technologies. Moreover, we inserted the control variable for the relevance of energy-related technologies for the focal firm controlling for firms without a relevant technological base or a non-technological orientation and, hence, such technology issues do not apply and consequently have not been discussed internally. Rank effects as to the firms’ market environments are measured by (a) the intensity of competition, (b) the energy environment characterized by high/volatile energy prices and energy shortage, (c) industry affiliation, and (d) country dummies. Based on cross-sectional data it is hardly possible to separate epidemic effects from stock and order effects. Hence we measure a net effect of the two by including the mean of adopted technologies within the firm’s two-digit industry (without the focal firm). Based on the survey results the Adoption costs could not be measured effectively, since the perception of obstacles frequently increases with adoption activities and consequently is clearly endogenous. Information about market prices of technologies and internal adjustment costs are not available for the sample of firms we could use. Finally, in order to capture inducement effects we control for (a) the firm’s energy costs and most importantly for this paper (b), the firm’s political environment, which is measured by variables referring to regulation, subsidies and energy taxes. Furthermore, we can distinguish these standard policy factors from alternative drivers of energy-related technology adoption, we control for (d) “voluntary agreements”, and (e) the demand for environmentally friendly produced products. See Table 1 to Table 3 for the variable definitions and descriptive statistics, respectively.

In order to separate the epidemic effects from other industry-specific effects, we add industry fixed effects to the model.

An important feature of existing empirical investigations about policy induced adoption based on survey data is that the meaning of the policy environment of non-adopter is not observed. For instance, in the CIS the policy effects are usually identified as a motivational question referring only to firms that have already adopted such technologies, for example “how important were the following factors in driving your decisions to introduce innovations with environmental benefits?” Consequently, information about the policy affectedness of non-adaptors is missing meaning that such data sets lack policy information for firms that did not adopt an environmentally friendly technology although they are potentially exposed to the same policy pressure than adopting firms. For the study at hand, we designed a specific set of questions that allows for policy information for non-adaptors as well. Hence we can compare policies that drive the adoption (inter-firm diffusion) with policies that drive the intensity of adoption (intra-firm diffusion).

Although we have policy information for adoptors and non-adoptors, we can only observe investments in energy-related, environmentally friendly technologies (our dependent variable), if a firm adopted such a technology, and we do not observe the investment behavior if such technologies have been not adopted. Moreover, we have an interest in the parameters of the selection equation and of the intensity equation. Hence, we apply a sample selection estimator which is formulated according to Heckman (1979) and Amemiya (1985).

A firm might have invested in clarifying a potential adoption and finally did not do it, since the adoption project failed.

In econometric terms we face the following problem:

Let y 1 be our investments in adoption of environmentally friendly, energy-related technologies and y 2 is the binary variable indicating the adoption of such a technology; x and y 2 are always observed and y 1 is only observed if y 2 equals 1. u 1 and v 2 are independent of x and have zero mean, whereas v ~ Normal(0,1); x 1 is a strict subset of x; and E(u 1|v 2) = γ 1 v 2.

Basically we could estimate (2) consistently and efficiently if the (here) unobserved γ 1 = 0, indicating that we do not have a selection problem. However, since we cannot be sure that this is the case, we follow the Heckman (1979) procedure, which allows us to estimate

where λ(×) is the inverse mills ratio. Here, γ 1 can be observed by inserting the inverse mills ratio in the model and if it is zero, equation (2) does not suffer from a selection bias and the coefficients are consistently and efficiently estimated (see Wooldridge 2010, 803). δ 2 is available from the first stage probit estimation, which is only consistently estimated if x contains one variable that affects y 2—the selection equation—but does not have a partial effect on y 1 (see Wooldridge 2002, p. 562). In the paper at hand we can use the epidemic variable in the selection equation as the necessary exclusion restriction; it is significantly positively related to the adoption propensity (y2) and has no effect on y1.

In order to estimate the heckman procedure we use the heckman command as implemented in STATA 13 with heteroscedasticity robust standard errors; we use the full maximum likelihood version of the estimator.

A potential problem is the possible endogeneity of some of the right-hand variables that would imply inconsistent estimates. Endogeneity might arrive not so much from the policies per se, since the design of the policies is clearly beyond the influence of a single firm and they likely perceive them as an exogenous shock, however, the endogeneity issue might result from the “subjective” assessment of the importance of the respective policies. We asked the firms to assess how important (on a 3 point Likert-scale) are taxes, subsidies, voluntary agreements, regulation, and demand for them. These assessments might be correlated with some unobserved firm factors that can bias our coefficients. However, since a broad set of observables that generally affect the firms’ adoption activities is included in the estimation equations besides the policy variables, our main results should at least not be affected by an omitted variable bias. We do not see why the used policy measures should systematically share a common unmeasured cause with the firms’ adoption intensity ; we do not see a “common response” bias, since the questions about the policy environment are not related to the questions about technology adoption. We thus expect that the policy variables affect the firms’ adoption intensity directly and endogeneity is not a main concern. However, we do not claim causal relationships.

We control for unobserved heterogeneity on the sector level (2-digit level). Moreover, we have a very comprehensive control vector correcting for most of the factors that might influence the assessment; such factors are the energy intensity of the focal firm, its R&D activities, the education level of the employees, foreign ownership, size, basic relevance of environmental issues, intensity of competition, export activities, and other investment activities.

Even more so as we are primarily interested in the effects of the different policy types relative to each other.

We are not only interested in the policy effects for the adoption propensity (inter-firm diffusion) and the level of investments of adoption (intra-firm diffusion), we are also interested, if there are differences among the participating countries. In order to identify differences, we interacted all covariates with the country dummies, respectively. This has the advantage that we directly observe, if the detected differences are significant.

Results

Table 4 presents the main results of our estimations. The results of both stages of the heckman estimation are presented in columns 2 and 3.

Table 4.

Main estimations.

	Heckman		Country Differences: Adoption				Country Differences: Intensity
			Reference: Switzerland		Reference: Germany		Reference: Switzerland		Reference: Germany
	Intensity	Adoption	Germany	Austria	Switzerland	Austria	Germany	Austria	Switzerland	Austria
MAIN VARIABLES
Taxes	−0.043	0.084*	0.206**	−0.155	−0.206**	−0.361**	0.101	0.211	−0.101	0.110
	(0.065)	(0.043)	(0.091)	(0.173)	(0.091)	(0.167)	(0.148)	(0.256)	(0.148)	(0.237)
Regulations	0.070	−0.010	−0.119	−0.217	0.119	−0.099	0.037	0.127	−0.037	0.090
	(0.065)	(0.046)	(0.095)	(0.189)	(0.095)	(0.188)	(0.143)	(0.241)	(0.143)	(0.230)
Voluntary agreements	−0.008	0.101**	−0.088	−0.478***	0.088	−0.389**	−0.049	−0.215	0.049	−0.167
	(0.065)	(0.047)	(0.095)	(0.185)	(0.095)	(0.186)	(0.141)	(0.223)	(0.141)	(0.218)
Public subsidies	0.182***	0.300***	−0.070	0.378**	0.070	0.448***	0.194	0.400**	−0.194	0.207
	(0.056)	(0.041)	(0.085)	(0.159)	(0.085)	(0.156)	(0.121)	(0.187)	(0.121)	(0.178)
Demand	−0.018	0.176***	−0.159*	−0.049	0.159*	0.110	0.056	−0.305	−0.056	−0.361
	(0.069)	(0.046)	(0.094)	(0.158)	(0.094)	(0.157)	(0.148)	(0.235)	(0.148)	(0.227)
CONTROL ENERGY ENVIRONMENT
High/volatile energy prices	0.222***	0.098**	−0.010	0.042	0.010	0.052	−0.063	−0.158	0.063	−0.095
	(0.063)	(0.042)	(0.085)	(0.173)	(0.085)	(0.170)	(0.138)	(0.252)	(0.138)	(0.242)
Energy shortage	−0.075	−0.098**	−0.018	−0.101	0.018	−0.083	0.013	0.044	−0.013	0.031
	(0.067)	(0.048)	(0.098)	(0.178)	(0.098)	(0.176)	(0.151)	(0.222)	(0.151)	(0.210)
CONTROL FIRM CHARACTERISTICS
Number of employees	0.261***	0.078**	−0.065	−0.271**	0.065	−0.206*	0.105	0.237	−0.105	0.133
	(0.053)	(0.031)	(0.063)	(0.120)	(0.063)	(0.118)	(0.118)	(0.165)	(0.118)	(0.151)
Export	0.130	0.103*	−0.094	0.070	0.094	0.165	0.125	0.058	−0.125	−0.067
	(0.101)	(0.059)	(0.115)	(0.239)	(0.115)	(0.240)	(0.203)	(0.467)	(0.203)	(0.462)
Foreign owned	0.210*	−0.150*	0.019	−0.111	−0.019	−0.130	0.110	0.157	−0.110	0.047
	(0.121)	(0.077)	(0.182)	(0.224)	(0.182)	(0.257)	(0.274)	(0.330)	(0.274)	(0.344)
R&D propensity	−0.088	0.289***	−0.052	−0.097	0.052	−0.045	−0.136	0.267	0.136	0.403
	(0.097)	(0.059)	(0.114)	0.000	(0.114)	(0.203)	(0.193)	0.000	(0.193)	(0.316)
Share graduated employees	−0.001	0.000	−0.002	−0.001	0.002	0.002	0.005	0.005	−0.005	0.000
	(0.003)	(0.001)	(0.003)	(0.005)	(0.003)	(0.005)	(0.006)	(0.011)	(0.006)	(0.010)
Competition intensity	−0.008	0.008	−0.050	0.032	0.050	0.082	−0.061	0.148	0.061	0.209**
	(0.030)	(0.017)	(0.036)	(0.070)	(0.036)	(0.069)	(0.064)	(0.104)	(0.064)	(0.100)
Energy costs	0.186***	0.044**	0.012	0.057	−0.012	0.045	−0.053	0.048	0.053	0.101
	(0.035)	(0.021)	(0.044)	(0.044)	(0.044)	(0.085)	(0.083)	(0.083)	(0.083)	(0.090)
Gross tangible investments	0.449***	0.082***	0.020	0.176***	−0.020	0.155**	−0.092	−0.347**	0.092	−0.255**
	(0.040)	(0.018)	(0.038)	(0.064)	(0.038)	(0.063)	(0.091)	(0.141)	(0.091)	(0.125)
COUNTRY CONTROLS
Swiss firm	−0.124	0.027			−0.836				−0.615
	(0.139)	(0.105)			(0.516)				(0.916)
German firm	−0.529***	0.165*	0.836				0.615
	(0.140)	(0.092)	(0.516)				(0.916)
Austrian firm				−1.410		−2.246**		2.230*		1.615
				(0.925)		(0.911)		(1.242)		(1.104)
EPIDEMIC ADOPTION		2.517***	−0.607	0.530	0.607	1.138
		(0.353)	(0.723)	(0.783)	(0.723)	(0.813)
Constant	0.435	−3.978***
	(0.671)	(0.506)
No. of observations			3439				1621
Censored/uncensored	1818/1621
R2							0.63
Wald test of indep. eqns. (rho = 0): chi2	1.05
	Wald chi2(48) = 2525.57***		Wald chi2(48) = 719.56***				F(78, 1542) = 39.84***

Note: heteroscedasticity robust standard errors; all estimations include 2-digit sector controls; significance levels: (*0.1,** 0.05, ***0.01).

Adoption propensity (inter-firm diffusion)

Most importantly, we see that taxes, voluntary agreements, public subsidies, and demand significantly induce the adoption of environmentally friendly, energy-related process innovations (green energy technologies); firms that identified such factors as important are significantly more likely to adopt such technologies. The results for the control variables in the model indicate that the adoption propensity is higher for larger firms, firms with R&D activities, and firms with greater gross tangible investments. The epidemic variable is also significantly and positively related with the adoption propensity of the focal firm. The results also unveil that (feared) bottlenecks in energy supply significantly decrease the adoption propensity and forces firms to stick to traditional technologies and that high/volatile energy prices urge firms to adopt green energy technologies. Referring to the econometric framework presented above we see that firm-specific rank effects and epidemic effects are important, however, the inducement effects resulting from the policy environment appear to be very important for the adoption propensity; here, only regulations (including CO2 certificates, which are toothless in all three countries) do not show a significant effect indicating that the regulatory requirements have been already fulfilled by the firms before the period under investigation, hence, lacking any further incentives for technological improvements. This clearly points at the “dynamic” disadvantages of regulations, presented in the literature review; regulations usually do not require to adopt the best available technological solution; ones the requirements are fulfilled no further environmentally friendly activities can be expected. Hence, we cannot reject H1 meaning that regulations are indeed less effective than market-based policies, voluntary agreements, or demand to increase the propensity of green energy technology adoption.

Level of investments (intra-firm diffusion)

The heckman estimation presented in columns 1 (intensity) and 2 (selection) in Table 4 shows that both equations are independent and that the intensity equation is not affected by the selection issue; the Wald test on independence of equations (rho = 0) is insignificant.

The most striking result is that all but one proxy for energy-related (policy) factors are insignificant, indicating that the current political framework and policy related factors are less relevant for the firms’ decision to increase its investment in green energy technologies (intra-firm diffusion); this is in stark contrast to the results for the adoption decision presented above. Only the importance of subsidies is significant and positively related with the amount of investments. Referring to H2, our results clearly show that the policies are less relevant for the amount of investments in green-energy technologies rather than for the adoption propensity.

Please note that the perception of subsidies as a stringent policy measure does not “mechanically“ increase the amount of investments.

Moreover, we see that the energy costs are also significant in the intensity equation signalling that firm specific energy-related factors are more important than policy inducement for the amount of investments. It is also remarkable that investments tend to increase in an energy environment that is characterized by high/volatile energy prices, whereas feared energy shortage does not affect the amount of investments.

The estimations for the single countries are presented in the Appendix Table A1. Basically the general results presented above hold for all three countries with few exceptions. Concerning the propensity: voluntary agreements are significant and negative for Austria and taxes are only significant and positive for Germany. Concerning the intensity, subsidies are not significant in Switzerland.

Table A1.

Determinants for propensity and intensity per country.

	Propensity			Intensity
	CH	D	AT	CH	D	AT
Taxes	−0.032	0.178***	−0.135	−0.110	−0.012	0.059
	(0.072)	(0.057)	(0.172)	(0.127)	(0.080)	(0.270)
Regulations	0.070	−0.041	−0.165	0.044	0.091	0.156
	(0.069)	(0.066)	(0.190)	(0.115)	(0.088)	(0.248)
Voluntary agreements	0.194***	0.100	−0.360*	0.041	−0.004	−0.180
	(0.067)	(0.069)	(0.190)	(0.109)	(0.094)	(0.251)
Public subsidies	0.308***	0.234***	0.746***	−0.021	0.207***	0.532***
	(0.065)	(0.056)	(0.148)	(0.097)	(0.076)	(0.188)
Demand	0.251***	0.102	0.332**	0.009	0.009	−0.213
	(0.068)	(0.067)	(0.161)	(0.120)	(0.095)	(0.253)
High/volatile energy prices	0.084	0.082	0.168	0.226**	0.176**	−0.098
	(0.065)	(0.056)	(0.167)	(0.113)	(0.086)	(0.275)
Energy shortage	−0.069	−0.091	−0.379**	−0.086	−0.061	−0.046
	(0.071)	(0.068)	(0.177)	(0.118)	(0.096)	(0.225)
Number of employees	0.131***	0.073*	−0.174	0.207*	0.249***	0.419***
	(0.047)	(0.043)	(0.135)	(0.106)	(0.074)	(0.148)
Export	0.141	0.067	0.351	0.102	0.142	0.258
	(0.087)	(0.087)	(0.272)	(0.161)	(0.149)	(0.519)
Foreign owned	−0.146	−0.116	−0.387*	0.250	0.352*	0.319
	(0.097)	(0.158)	(0.231)	(0.181)	(0.204)	(0.366)
R&D propensity	0.319***	0.315***	0.443**	−0.079	−0.179	0.079
	(0.090)	(0.083)	(0.225)	(0.165)	(0.136)	(0.402)
Share graduated empl.	0.003	−0.001	−0.005	−0.005	0.001	−0.014
	(0.003)	(0.002)	(0.006)	(0.005)	(0.003)	(0.012)
Competition intensity	0.036	−0.021	0.034	−0.002	−0.049	0.108
	(0.028)	(0.023)	(0.077)	(0.050)	(0.042)	(0.106)
Energy costs	0.032	0.037	0.111	0.180***	0.165***	0.264***
	(0.033)	(0.030)	(0.100)	(0.069)	(0.052)	(0.090)
Gross tangible investments	0.053*	0.070***	0.280***	0.545***	0.445***	0.164
	(0.028)	(0.026)	(0.066)	(0.082)	(0.047)	(0.127)
Epidemic adoption	3.037	−2.740	1.580
	(1.860)	(3.783)	(2.576)
Constant	−4.222***	−1.288	−5.464***	0.473	0.277	1.199
	(0.612)	(1.435)	(1.036)	(0.922)	(0.625)	(1.465)
N	1584	1633	300	600	845	176
Wald chi2/R2	309.55***	268.50***	101.59***	0.58	0.63	0.65
Log Likelihood	−894.73	−969.87	−138.46	−1091.56	−1529.57	−296.71

Note: significance levels: (*0.1,** 0.05, ***0.01).

Country comparisons

Given the different institutional framework in the investigated countries (see above), it is plausible to assume that the investigated types of policies show different effects. And in fact we find that taxes are significantly more effective for the adoption decision of a firm in Germany than in Austria and Switzerland (Table 4, columns 6 and 7). We also see differences in terms of voluntary agreements, which are significantly less effective in Austria than in Germany and Switzerland. To the contrary, public subsidies are more effective (important) in Austria and demand-related factors for environmentally friendly produced products tend to be more effective in Switzerland compared to Germany. Concerning regulation, we do not see significant differences across countries.

The results are quite different for the amount of investments (intra-firm diffusion). Public subsidies are significantly more effective in Germany and Austria than in Switzerland. We do not detect any differences for all other policy related factors.

Extensions

Marginal effects

Although we cannot claim to present causal effects, we want to give an impression about the size of the effects of the observed policies and policy-related factors; Table 5 presents the marginal effects. The probability of a randomly chosen firm to adopt any green energy technology—in our sample—is 49%. This figure fluctuates between 40% in Switzerland and 56% in Austria. Moreover, we see that the adoption probability significantly increases with the relevance of taxes (47–53%), voluntary agreements (48–58%), regulation (50–53%), and demand (47–58%), whereas public subsidies are related with the relatively strongest increase in the adoption probability (44–68%). There are country differences. The relatively largest increase in adoption probability results from public subsidies in all three countries. Voluntary agreements and demand are increasing the adoption propensity in Switzerland by 21%-points and 23%-points, respectively. Taxes increase the adoption propensity in Germany by 13%-points and demand shows a relatively strong effect in Austria (+27%-points).

Table 5.

Margins.

	Margins
	Overall	Switzerland	Germany	Austria
Probabilities (constant)	0.49	0.40	0.55	0.56
Taxes
not relevant (1)	0.47	0.40	0.52	0.63
relevant (2)	0.50	0.40	0.56	0.62
very relevant (3)	0.53	0.34	0.67	0.53
Regulation
not relevant (1)	0.50	0.39	0.58	0.64
relevant (2)	0.46	0.37	0.54	0.52
very relevant (3)	0.53	0.49	0.56	0.59
Voluntary agreements
not relevant (1)	0.48	0.37	0.55	0.66
relevant (2)	0.51	0.40	0.61	0.54
very relevant (3)	0.58	0.58	0.61	0.37
Public subsidies
not relevant (1)	0.44	0.35	0.52	0.42
relevant (2)	0.55	0.44	0.60	0.75
very relevant (3)	0.68	0.66	0.72	0.89
Demand
not relevant (1)	0.47	0.37	0.55	0.56
relevant (2)	0.56	0.44	0.64	0.67
very relevant (3)	0.58	0.60	0.57	0.83
Fluctuating/high energy prices
not relevant (1)	0.47	0.37	0.54	0.55
relevant (2)	0.49	0.41	0.56	0.64
very relevant (3)	0.54	0.44	0.61	0.66
(Feared) bottlenecks
not relevant (1)	0.50	0.40	0.58	0.66
relevant (2)	0.46	0.39	0.53	0.38
very relevant (3)	0.44	0.33	0.52	0.49

Note: Probabilities (constant) describe the probability to adopt an environmentally friendly, energy relevant technology. This means that the overall probability to adopt such technologies without taking into account policies or firm characteristics etc. is 50%.

Technology-specific effects and certified management systems

Table 6 presents the results for specific types of green energy technologies and energy-saving certified management systems (e.g. environmental certification ISO 140001, Energy-management ISO 50001, environmental-/energy-target declaration, periodical environmental-/energy-reports). We included the certified management systems to the simultaneous estimation procedure, since previous literature shows that they are immediately related to technology adoption. For instance, Frondel, Horbach, and Rennings (2007) found that organizational measures are very important for the adoption of environmentally friendly process innovations. Also Khanna, Deltas, and Harrington (2009) present evidence for a strong and significant relationship between totally quality management and the adoption of pollution prevention technologies.

Table 6.

Adoption by type of technology and adoption of energy management systems.

	Production	ICT	Transport	Buildings	Renewables	Manage-ment
Taxes	0.092*	0.001	0.010	0.047	0.047	0.194***
	(0.047)	(0.045)	(0.049)	(0.044)	(0.058)	(0.047)
Regulations	0.062	0.030	−0.080	−0.032	−0.078	0.045
	(0.047)	(0.046)	(0.051)	(0.045)	(0.059)	(0.047)
Voluntary agreements	0.073	0.032	0.079	0.139***	0.019	0.277***
	(0.047)	(0.045)	(0.052)	(0.045)	(0.058)	(0.048)
Public subsidies	0.189***	0.152***	0.040	0.257***	0.423***	0.073*
	(0.041)	(0.041)	(0.044)	(0.040)	(0.050)	(0.043)
Demand	0.153***	0.151***	0.266***	0.144***	0.144***	0.033
	(0.048)	(0.045)	(0.048)	(0.045)	(0.055)	(0.047)
High/volatile energy prices	0.145***	0.068	0.129***	0.075*	−0.019	0.030
	(0.045)	(0.042)	(0.048)	(0.042)	(0.054)	(0.046)
Energy shortage	0.001	0.057	−0.005	−0.066	−0.063	−0.020
	(0.051)	(0.047)	(0.052)	(0.049)	(0.060)	(0.048)
Number of employees	−0.036	0.075**	0.167***	0.062**	0.019	0.203***
	(0.035)	(0.032)	(0.035)	(0.031)	(0.039)	(0.034)
Export	0.187***	0.152**	−0.145**	0.056	−0.009	0.178***
	(0.069)	(0.062)	(0.073)	(0.061)	(0.078)	(0.066)
Foreign owned	−0.042	−0.023	−0.117	−0.200**	−0.504***	0.300***
	(0.080)	(0.080)	(0.091)	(0.079)	(0.119)	(0.081)
R&D propensity	0.334***	0.238***	0.185***	0.253***	0.168**	0.202***
	(0.063)	(0.061)	(0.069)	(0.059)	(0.072)	(0.062)
Share graduated employees	−0.001	0.005***	−0.001	−0.001	0.001	0.002
	(0.002)	(0.001)	(0.002)	(0.001)	(0.002)	(0.002)
Competition intensity	−0.021	0.007	0.033	−0.020	−0.035	−0.022
	(0.021)	(0.018)	(0.020)	(0.018)	(0.024)	(0.020)
Energy costs	0.116***	−0.008	−0.003	0.033	−0.010	0.070***
	(0.026)	(0.022)	(0.023)	(0.021)	(0.026)	(0.023)
Gross tangible investments	0.098***	0.059***	0.051**	0.103***	0.111***	0.040**
	(0.022)	(0.019)	(0.023)	(0.018)	(0.025)	(0.020)
Epidemic effect*	3.318***	3.433***	4.627***	2.443***	4.428***	2.659***
	(0.412)	(0.479)	(0.664)	(0.350)	(0.627)	(0.464)
Swiss firm	−0.138*	−0.056	−0.127*	−0.206***	−0.067	−0.091
	(0.073)	(0.073)	(0.072)	(0.071)	(0.081)	(0.077)
Austrian firm	−0.166	−0.128	−0.183	−0.155	−0.091	−0.033
	(0.102)	(0.099)	(0.120)	(0.094)	(0.127)	(0.108)
Constant	−5.034***	−2.959***	−4.177***	−4.064***	−4.336***	−5.110***
	(0.503)	(0.404)	(0.459)	(0.369)	(0.581)	(0.403)

Note: multivariate probit estimator with heteroscedasticity robust standard errors. (*) the epidemic effect refers to the respective technology or the management systems. Number of observation 3480, Wald chi2 (294) = 4444.34***, Likelihood ratio test of rho21 = rho31 = rho41 = rho51 = rho61 = rho32 = rho42 = rho52 = rho62 = rho43 = rho53 = rho63 = rho54 = rho64 = rho65 = 0: chi2(15) = 1493.89, Prob > chi2 = 0.0000

Demand related factors are important for the adoption of all identified technologies, whereas the adoption of energy saving technologies in the field of production are also significantly promoted by taxes and public subsidies. Concerning building technologies, the adoption stimulating effects of voluntary agreements, subsidies, and demand are significant. Public subsidies and demand essentially drive the adoption of technology to produce energy from renewable sources.

We also can compare the effects of policy related factors for green technology adoption with firm behaviour related to the introduction of energy saving certified management systems. Management systems are essentially driven by taxes, voluntary agreements, and subsidies. Demand-related factors do not play an important role for management systems, which is in stark contrast to green technology adoption.

National promotion plans appear to be relatively ineffective in the transportation sector. Public subsidies and demand related factors appear to be effective independent of the type of technology.

Policy discussion

From the results of this investigation, we can draw a number of policy conclusions. Most importantly, all types of policy instruments that we looked at—taxes, regulations, voluntary agreements, subsidies, stimulating demand—effectively stimulate the adoption of green energy technologies and energy management systems. The strongest effect is observed for subsidies. However, most instruments have no effect on the intensity of adoption, that is the amount of investment into green technology. Public subsidies are the only exception. But the size of investment matters if the environmental burden is to be decreased significantly. For pushing investment into green technologies upwards, subsidies are the most powerful, but also a costly policy instruments. We find the strongest impacts of subsidies for Austria which also applies the most generous subsidy schemes. Given the limited budgets of governments to subsidies technology adoption, policy makers should focus subsidies for green technology on those areas where the largest marginal positive environmental impact can be produced. Dynamic policies like taxes should be focused on those areas of green energy technologies where adoption costs increase proportionate per unit of reduced environmental pollution since taxes are less effective, if costs increase disproportionally.

Apart from direct policy intervention, the amount of energy costs and high/volatile energy prices are important drivers for higher investment into green energy technologies. Both factors are mainly determined by the developments in energy markets. However, policy can indirectly affect a firm’s energy costs through energy taxes. This implies that taxes work in two ways: they increase the propensity of firms to engage in the adoption of green energy technologies. The more taxes raise firms’ energy costs, they indirectly also increase the amount of investment into these technologies.

Demand for products and services that are produced with a high energy efficiency is another important factor for the propensity of green energy technology adoption. Policy instruments that raise awareness toward energy-efficient behaviour among consumers as well as policies that inform consumers about the energy efficiency of production processes are an effective stimulus for the adoption of green energy technologies in the firm sector. This effect is observed in all three countries and seems to be rather independent of the respective institutional framework, though demand factors tend to be more effective for Swiss firms. Voluntary agreements and standards are significantly related to the adoption of green energy technologies and the adoption of energy management systems. They are partly motivated by avoiding public regulations so that one may suspect that the threat of a potential public intervention to be the underlying cause for their effectiveness.

Policy makers relying on a portfolio of different types of policies including subsidies and demand related factors are on the safe side in terms of their general effectiveness. Taxes and voluntary agreements are also overall effective but they show some technology-specific patterns, which can be due to the lack of initiatives related to such instruments; taxes are more effective for green energy-related production technologies and similar to voluntary agreement they are especially effective for the adoption of energy management systems and voluntary agreements also are significantly effective for building technologies (see Table 6).

Given the current institutional framework in the three countries we studied (see Appendix), subsidies are more effective in Austria, taxes are more effective in Germany, and demand-related factors are relatively more effective in Switzerland. This result can be interpreted in two ways. On the one hand, subsidies in Austria and taxes in Germany are significantly higher than in Switzerland because they may be more stringent and effective. This would imply that these policies are broadly accepted by firms. On the other hand, subsidies in Austria and energy taxes in Germany may be better designed than the same instruments in the other countries, for example by more consistently considering the particularities of the firms in the respective country. Both reasons cannot be tested in this framework and must be left for more detailed policy analyses. However, the main lesson is that such policies are not effective independent of designs and institutional acceptance.

Moreover, our study does not investigate the costs of the respective policy instruments applied in these countries. They depend upon the type of policy and the design of the policies in the respective country. For instance, De Groote and Verboven (2016) found that an upfront subsidy program can promote the adoption of photovoltaic systems at much lower costs than future production support. See for an overview about policy design and social costs (McCann 2013).

Conclusions

We have seen a significant increase in the empirical literature about the adoption of environmentally friendly technologies over the last years. Nevertheless, there are some important blind spots that have not been addressed adequately, not at least due to the lack of firm-level data. First, we have still an insufficient understanding of the effect of environmental policies (taxes, subsidies, and regulations), voluntary agreements, and demand related factors, on the adoption of green energy technologies. Moreover, we know too less about their effects on the intensity of technology use and how they impact the adoption of non-technical measures (certified management systems). Second, policy makers tend to apply best-practice from other countries. However, it is not sure that, for example, a CO2 tax or a feed-in tariff has the same effects in Germany and in Switzerland; the institutional framework matters, and should be seriously considered.

In the study at hand we try to address those issues by simultaneously estimating the policy effects on the propensity and intensity of adoption of green energy technologies. Moreover, we compare the policy effects for different technologies and non-technological measures (management systems) in a multivariate estimation setting and we simultaneously investigate country-specific differences concerning the effects of different types of policies for the adoption behaviour.

The results are based on a comprehensive data set stemming from a firm-level survey in three countries (Austria, Germany, and Switzerland) conducted at the same point in time and based on an identical questionnaire (see data section). The results reveal that the current policy framework in all three observed countries is partly effective for the adoption of such technologies and non-technical measures and it is ineffective in terms of the intensity of adoption (intra-firm). Firm-specific characteristics drive intra-firm adoption whereas primarily voluntary agreements and demand-related factors as well as subsidies and taxes drive the propensity of adoption (inter-firm adoption). Country-specific characteristics play a significant role in terms of policy effectiveness. Taxes are more effective in Germany and demand related factors are more effective in Switzerland. Public subsidies are more effective in Austria. The results are different for the investment intensity (intra-firm diffusion). Here we see that public subsidies tend to be more effective in Austria than in Switzerland. However, in order to identify the efficiency of policy instruments, we would have to include the economic costs of policies which is beyond the focus of this study and must be left for future investigations.

Obviously, our results suggest that the effectiveness of policies is country specific and can hardly be generalized. The institutional framework, the policy attitude, and the economic environment significantly differs across countries and such factors seem to be important for the effectiveness of energy policies. Hence, our analytical approach can be used to investigate the effectiveness of policies in other countries. Our results suggest that national evidence-based policy making requires country-specific studies.