Environmental policies and risk finance in the green sector: Cross-country evidence

Energy Policy 83 (2015) 38–56

Contents lists available at ScienceDirect

Energy Policy journal homepage: www.elsevier.com/locate/enpol

Environmental policies and risk finance in the green sector: Crosscountry evidence Chiara Criscuolo a,b,n, Carlo Menon a,c a

OECD, Science, Technology and Innovation Directorate, 2 rue André Pascal, 75775 Paris Cedex 16, France Centre for Economic Performance, the London School of Economics and Political Science, London, UK c Spatial Economics Research Centre, the London School of Economics and Political Science, London, UK b

H I G H L I G H T S

 Risk-finance in the green sector is likely to face more challenges than in other hi-tech sectors.  Supply and deployment policies are associated with more investments relative to fiscal policies.  FITs have a positive effect, but in the solar sector very generous FITs discourage investments.

art ic l e i nf o

a b s t r a c t

Article history: Received 17 October 2014 Received in revised form 19 March 2015 Accepted 21 March 2015

This paper provides a detailed description of venture capital investment in the green sector across 29 countries over the period 2005–2010, and identifies the role that policies might play in explaining observed cross-country differences. The analysis is based on a deal-level database of businesses seeking financing, combined with indicators of renewable policies and government R&D expenditures. The econometric analysis relates the number of deals and their volumes in a country to deployment and supply policies using count data and limited dependent variable (Tobit) models. The results suggest that both supply side policies and environmental deployment policies, designed with a long-term perspective of creating a market for environmental technologies, are associated with higher levels of venture capital relative to more short-term fiscal policies. When focusing on policies related to renewable energy generation, the results confirm the positive association of generous feed-in tariffs (FITs) with venture capital investment. However, in the solar sector excessively generous FITs tend to discourage investment, perhaps reflecting a lack of credibility over the longer term. Thus, both sets of results point to long-term policy stability, sustainability and credibility as important policy features to ensure Venture capital backing of innovative and risky ventures in a country's green sector. & 2015 Elsevier Ltd. All rights reserved.

Keywords: Environmental innovation Environmental policies Environmental technologies Risk finance Venture capital JEL Codes: Q55 Q58 G24

1. Introduction Shifting economies from environment- and resource-intensive trajectories to ‘green growth’ will require structural transformation and technological innovation. For this reason, start-up companies play a crucial role in moving towards green growth, as they often exploit opportunities ignored by incumbent firms. Venture Capital (VC)1 is essential to enable new businesses to

n

Corresponding author at: OECD, Science, Technology and Innovation Directorate; 2 rue André Pascal, 75775 Paris Cedex 16, France. E-mail addresses: [email protected] (C. Criscuolo), [email protected] (C. Menon). 1 In this paper, venture capital includes all forms of financing other than traditional corporate finance tools (e.g., banking loans, corporate bonds, public equity). Therefore it also includes angel finance, public investments and grants, and private equity. http://dx.doi.org/10.1016/j.enpol.2015.03.023 0301-4215/& 2015 Elsevier Ltd. All rights reserved.

grow in emerging sectors such as Information and Communication Technologies (ICT), software and biotech, but also in environmental technology (hereafter referred to as the green sector).2 However, VC backing of green sector companies faces more challenges than other sectors, due to gaps in managerial skills, the long term investment period, risky exit opportunities, and regulatory

2 In this paper, the classification of economic activities into the “green sector” is based on the original selection in the source database maintained by the Cleantech Group called “Cleantech”, and refers to a broad part of the economy. According to the data provider, “Cleantech is new technology and related business models that offer competitive returns for investors and customers”, while “greatly reducing or eliminating negative ecological impact, at the same time as improving the productive and responsible use of natural resources”. We also on occasion employ the term “environmental technologies” to identify the same subset of activities.

C. Criscuolo, C. Menon / Energy Policy 83 (2015) 38–56

uncertainty (Wustenhagen and Teppo, 2006). The latter sources of risk are particularly relevant, as environmental policy has proved, over the last decade, to be remarkably volatile in many countries. This can be a crucial issue for green sector ventures, as their profitability prospects often depend on public regulation. This is probably the most distinctive characteristic of green investments, as compared to VC deals in other sectors (e.g., ICT or medical devices). In fact the most common valuation method used by venture capitalists when deciding whether to invest in a venture is the Discounted Cash Flow analysis (DCF) which is a net present value method (Messica, 2008) based on cash flow projections. Those are highly speculative for non-public and non-traded companies in high risk high technology ventures, which makes the use of DCF extremely uncertain. This is going to be even more so in the case of green deals that have a long time to maturity, high capital intensity and often relate to new goods for which a market might not yet exist (Elton et al., 2009). In this case, valuation becomes more of an art and is based on a variety of metrics (e.g. Baum and Silverman, for the biotech case) but also on gut feelings (Messica, 2008), thus the importance of the perception of political risk when VC invest in the green sector. However, the level of VC in the green sector differs significantly across countries, stages of financing, and sectors. The empirical evidence on what drives these differences is somewhat limited. Most of the discussion on the gaps in VC investment in the green sector is based on case studies, anecdotal and/or survey evidence, but is not supported by econometric analysis. This paper aims to fill this gap by providing new evidence on the national-level determinants of VC investment in the green sector. It investigates the relationship between national level “environmental” policies and VC investment in the green sector using cross-country, cross-industry micro-aggregated data. In particular, the empirical analysis estimates the relation between national level policies, the number of green sector ventures obtaining VC backing, and the amount of funding received. The included policies are both supply-side, such as public Research and Development (R&D), as well as deployment policies, such as regulations and standards. The analysis encompasses different fields and stages of investment, and covers the period 2005–2010. By doing so, the paper is complementary to the recent analysis of Cumming et al. (2013), who explore the role played by oil price, media coverage, and other legal, cultural, and governance variables in explaining the diffusion of Cleantech VC investments around the world. National environmental policies might strongly affect the expected commercial viability and future profitability of nascent ventures in the green energy sector. Although part of the goods and services produced by the green sector are in principle tradable, the domestic policy environment still plays a prominent role, for several reasons. First, barriers to technology diffusion hinder knowledge transfer across borders: for instance, empirical evidence on the wind power sector shows that the marginal effect of domestic policies on innovation is 25 times stronger than that of foreign policies (Dechezleprêtre and Glachant, 2013). Similarly, it is well known that the energy market is heavily regulated in many countries, and there are non-trivial costs in the storage and transfer of electricity; this also limits the international tradability of (electric) energy. As a consequence, domestic policies are of prime importance in the energy generation sector, and are likely to be even more so in some of the other domains considered in the analysis, e.g. wastewater treatment, soil remediation, etc. The analysis exploits comprehensive deal-level information on VC activity and on businesses seeking risk capital in environmental technologies over the period 2005–2010 in 29 OECD and emerging

39

economies,3 combined with indicators of government renewable policies and government R&D expenditures.

2. Methods 2.1. Start-UPS and venture capital investment in the green sector Start-ups have been the engine behind growth and breakthrough innovations in sectors such as software, nanotechnology and biotechnology. More recently the importance of start-ups as a source of radical and architectural innovations4 has also become evident in the green sector.5 While large incumbents are better at, and more likely to, introduce incremental and modular green innovations, start-ups play a crucial role in ensuring the shift to a greener growth paradigm and are complementary to “greening Goliaths” (i.e., incumbent large companies; Hockerts and Wuestenhagen, 2010). In recent years, VC/PE (private equity) also played an increasingly important role especially in the United States, the United Kingdom and more recently in China, although – as shown in Fig. 1 – VC/PE represent only a small percentage of overall funding sources for the green sector.6 These funding sources are relatively more important in countries where VC is already developed, such as the United States, but they are also growing in importance in emerging economies, such as Brazil and China. The largest sources of funding, however, remain asset finance7 and public markets. This is in line with the fact that VC backing is focused on a particular type of project, characterised by high technology risk and low capital intensity. In fact, as exemplified in the typology outlined in Fig. 2, bank loans might be the more appropriate source of funding for projects with low capital intensity and low risk profiles, while project finance is better suited for projects with high capital intensity and low risk profiles (Kerr and Nanda, 2009; Ghosh and Nanda, 2015). On the other hand, venture capitalists are crucial investors for entrepreneurial high growth start-ups operating in young, dynamic and uncertain industries where the selection process of an investment is based on an uncertain valuation, with a lack of a clear performance history and a very high technology risk. In software and biotech sectors, they have been key providers of finance, but have generally financed projects with low capital intensity that can show rapid commercial viability (3–5 years), and can be sold within the life of a fund (about 10 years). This is motivated by the need to diversify their high-risk portfolio and increase the chances of positive “tail” outcomes in their investments' portfolio. Venture capitalists are therefore more likely to finance projects in the bottom right panel of the diagram in Fig. 2. VC is becoming increasingly important for the green industry. Figures for the United States market show that, since 2004, VC in 3 These are Australia, Austria, Belgium, Canada, China, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Ireland, Israel, Italy, Japan, Luxembourg, Netherlands, New Zealand, Norway, Poland, Portugal, Russia, South Africa, Spain, Sweden, Switzerland, the United Kingdom, and the United States. 4 For a general discussion of entrepreneurship and radical innovation see Squicciarini et al. (2013). 5 The definition of the Green sector includes clean energy generation, infrastructure and storage; energy efficiency; land management; natural pesticides; emissions control; recycling and waste, transportation and water conservation and treatment. However, some of the figures presented refer only to clean energy due to a lack of comprehensive data. 6 Note that the figure reports data only for the clean energy sector rather than the whole green sector. 7 Asset finance is defined as “all money invested in renewable energy generation projects, whether from internal company balance sheets, from debt finance, or from equity finance. This excludes re-financings” (Bloomberg New Energy Finance, 2010).

40

C. Criscuolo, C. Menon / Energy Policy 83 (2015) 38–56

Fig. 1. Investment by financing type in the clean energy sector (billions USD), 2009–2011. Notes: The graphs report the distribution investments by type across countries. Asset finance is defined as “all money invested in renewable energy generation projects, whether from internal company balance sheets, from debt finance, or from equity finance. This excludes re-financings”. Small distributed capacity refers to small-scale projects, i.e. those with less than 1 MW. Public Market is defined as “all money invested in the equity of publicly quoted companies developing renewable energy technology and clean power generation”. Venture Capital/Private Equity includes all money invested by venture capital and private equity funds in the equity of companies developing renewable energy technology. Source: Pew Trust, June 2010 and Bloomberg New Energy Finance (2010).

the green sector has steadily grown and represents almost a quarter of overall investment (Fig. 3). The importance of environmental technologies in the United States VC market is in line with the growth of green VC in North America, Europe, Israel, China and India, both in terms of the number of deals and the amounts invested, as shown in Fig. 4. One particular feature of this increase is that the overall growth in volume of deals and dollar values is not accompanied by an increase in the average value of each deal. This might reflect several trends, but a possible explanation is that, following the financial crisis, venture capitalists have become more selective in their choice of projects. They cherry-pick projects that are less capital intensive and therefore require less financing on average, thus having a lower impact in terms of risk for the fund. In the green sector, VCs are faced with an increased level of managerial and operational risk that could hinder the success of green start-ups. Managerial risk can arise because of the mismatch between the skills needed to manage green sector companies and the skills of CEOs/entrepreneurs/managers that have experience in being backed by VC and, similarly, between skills needed in the initial stage of idea vetting and upscaling/deployment. In addition, the experience of VC investors in mentoring, networking, etc., are not as developed in the green sector as within internet-based start-ups; thus there is the risk of a managerial valley of death, in addition to the commercialisation valley of death discussed below at the stage of deployment and commercialisation. Using the classification of Fig. 2, a financing gap is likely to arise for those projects that have a high technology risk profile and that are capital intensive; this gap is worsened if, in addition, the prospects of the project's commercial viability has a long horizon and VCs' exit opportunities are highly uncertain. This is

particularly true for those projects that are characterised by technology risk at the stage of lab development, but also applies to later stages, such as demonstration and early commercialisation. The main reason for this funding gap is that the longer the time horizon, the higher the financing risk for seed and that the early stage VCs is unable to ensure a successful exit, or raise follow-on funding before the end of the VC fund life (Nanda and RhodesKropf, 2010). These projects are therefore very hard to fund with either project or debt financing or VC, and can fall into a “Valley of the Death” in terms of financing. Projects in certain areas of the green sector fall exactly into this category. Examples include offshore wind farms, advanced biofuel refineries and first commercial plants for unproven solar cell technologies. These are projects that are very capital intensive and have a very high technology risk, not only in the seed stage, but also in the deployment stage. The risk of failure persists even if the project succeeds at the “lab experimentation” level. Other projects in the green sector, however, fall into the category of high-risk, low-capital intensive projects that have been traditionally backed by VC: examples are energy efficiency software, fuel cells etc. Anecdotal evidence (see discussion in Ghosh and Nanda, 2015) suggests that VCs are increasingly shifting their investments towards these latter types of projects. The development and the commercialisation of radical technologies in those green sectors that are both high-risk and capital intensive are jeopardised if they fall short of VC backing. These highly risky and capital intensive projects are not funded through project finance either, even though this source of financing has been steadily growing since 2004 for projects employing proven clean energy equipment (projects that in Fig. 2 would be in the top left quadrant). Recent data reports that even before the

C. Criscuolo, C. Menon / Energy Policy 83 (2015) 38–56

41

Capital intensity

Project Finance/Existing firms

Bank Debt/ Existing firms

Venture Capital

Technology risk Fig. 2. Many Green sector ventures are “hard to fund” Source: Ghosh and Nanda (2015).

Fig. 3. VC investment in the US in Green sector, 2004–2010. (As compared to Biotech, Software and Medical devices and equipment). Source: Cleantech Group, PWC/NVCA Money Tree Report.

financial available nologies, this type 2010).

crisis almost no private project finance capital was for projects whose aim was to deploy unproven techand the financial crisis has made capital availability for of project even scarcer (Bloomberg New Energy Finance,

As mentioned above, exit strategies for VCs, be it under the form of Merger and Acquisitions (M&As) or Initial Public Offerings (IPOs), can be particularly difficult both at the level of manufacturing scale up and at the level of full commercialisation. These difficulties arise also because of the inherent regulatory risk in the

42

C. Criscuolo, C. Menon / Energy Policy 83 (2015) 38–56

Fig. 4. Cleantech venture investments by deal stages. Source: OECD based on the Cleantech Market Insight Database.

Fig. 5. IPOs and M&As in the Green sector, 2005–2010. Note: dollar values are expressed in real terms. Source: Authors' elaboration based on the Cleantech Market Insight Database.

green sector, which might arise because private investors are unsure whether established policies and regulation will remain in place over the longer term. There are still very few IPOs in the green sector, as shown in Fig. 5, with a big drop observed during the crisis. The extent to which the current level of activity in the M&A market of green companies is sufficient to develop a virtuous circle of financing is still open to debate. Recent data however show an upward trend in M&A activity by large corporations, as shown in Fig. 5 (and as discussed in Criscuolo et al., 2013). 2.2. Data The data used in the analysis comes from combining a private commercial database on financing, IPO and M&A activity in the

clean tech sector (from the Cleantech company, www.Cleantech. com) with indicators of relevant policies and institutional settings at the national level. 2.2.1. Data on deals in the green sector In this paper we exploit the information on financing of green companies from the commercial database Cleantech. The dataset includes information on all businesses that seek financing whether they are successful or not. For the businesses that are financed, the database includes information on different forms of financing. More than half of investors are VC firms, while PE funds and corporations represent together more than 20% of all investors. The remaining investors are classified as business angels, public investments, investment banks, non-profit, university, and non-VC funds. Investors that

C. Criscuolo, C. Menon / Energy Policy 83 (2015) 38–56

43

Table 1 Examples of the Cleantech database. Source: Authors' elaboration based on Cleantech Market Insight Database. Some details of the two companies and of investors' names have been replaced with fictitious records to avoid disclosure of proprietary data. Name

Clean Ideas

Greener World

Primary industry Secondary industry

Air & Environment Air & Environment, Cleanup/Safety, Remediation

One sentence description Website

Developer of a variety of bioremediation technologies http://www.cleanideas.com

Address

1345 Green Drive

City Country Deal date Amount Invt. Stage Investors

Ottawa Canada 2Q2006 $1,440,000 Follow-On VentureFun Funds, Covent Group of Funds

Water & Wastewater Oil and Gas, Wastewater Treatment, Water & Wastewater Provider of oilfield wastewater treatment services http://www.greenerworld.cn 3rd Fl., Alley 342, #23 Jeaihi Road Shanghai China 8/19/2009 $500,000 Seed Rich Partners, iSecond Capital, Rockfeller

Table 2 Environmental technology sectors covered by the Cleantech database. Primary industry

Secondary industry

Aquaculturea; land management; natural pesticides Cleanup/safety; emissions control; logistics; monitoring/compliance; trading & offsets; vehicles; water treatment Advanced packaging; biofuels; buildings; chemical; glass; lighting; monitoring & control; monitoring/compliance; smart production; trading & offsetsa; transmission Energy generation Biofuels; geothermal; hydro/marine; solar; wind Energy infrastructure Management; transmission Energy storage Advanced batteries; fuel cells; hybrid systems Manufacturing/industrial Advanced packaging; monitoring & control; smart production Materials Bio; chemical; glassa; nano Recycling & waste Recycling; waste treatment Transportation Advanced batteries; fuels; logistics; structures; vehicles Water & wastewater Bio; cleanup/safety; glassa; wastewater treatment; water conservation; water treatment Agriculture Air & environment Energy efficiency

a The secondary industries “acquaculture” and “glass” are not generally included in the “Green sector”. Similarly, “trading & offsets” may be an unusual area of activity for startups. However, their exclusion does not affect the results of the empirical analysis (results available from the authors upon request).

could not be categorised in any of the above groups were included in a residual category “other”. The information is provided at the deal level and includes details on companies receiving funding, (number and identity of investors – including names, addresses and web pages), the development stage of their investment activity, deal dates and amounts of the investment. However, when a syndication of investors is involved in the deal, information on the exact contribution of each individual investor is not reported. In a few cases – less than 30% of the sample – the year in which the business seeking investment was founded and the names of the management team of the company are also included. Table 1 reports the example of two deal records (entries have been replaced with fictitious records to avoid the identification of the two companies and of the investors). The deal data contain detailed information on the stage of the investment, which are consistently aggregated into the five following categories: seed; early stage; later stage; buy out; and other. The data cover Europe, North America, and a few emerging economies (for a detailed list of countries covered see the descriptive analysis in Section 2) over the period of 2000–2011. Given that only the first semester of 2011 was included and that coverage

up to 2004 was likely to be less comprehensive, the regression analysis is limited to 2005–2010. As described in Table 2 the “green” sectors covered in the data span from energy generation, to waste water treatment, to materials. Most of the deals in the sample are in the energy generation sector, but there is wide variation across countries (for further detail on sectors included see Table A1 in the Appendix A). In addition to companies receiving investments, the data include information on companies that are actively seeking funding but are not successful in securing it. Although the point in time when the company starts looking for funding cannot be identified, the data provide an approximate date for when the company makes this information public. This is used to build a control group including those companies actively but unsuccessfully looking for financing. In the analysis, this information is used to construct a control variable that includes all firms seeking financing in the different green sectors across economies to proxy for the entrepreneurial activity and the level of demand for VC and thus can help reach a cleaner identification of the factors leading to successful deals. The data have been extensively checked for inconsistencies and cleaned using information from different sources, including the web pages of the companies in the sample. When possible (for

44

C. Criscuolo, C. Menon / Energy Policy 83 (2015) 38–56

Table 3 Renewables policy measures. Source: REN, 2005–2011. Country and year of adoptiona

Category

Policy Measures

Regulation price

Feed-in tariffs set a fixed price at which power producers can sell renewable power into the electric power network. Countries can offer a fixed tariff or provide fixed premiums added to market – or cost related tariffs. Some provide both.

Germany, Sweden, Denmark, China, France, Austria, Spain, Israel, Ireland, Canada, Switzerland (2004 or before); Greece, Indonesia, Italy, Netherlands (2005); Czech Republic, Hungary, Brazil (2006); South Africa, Portugal (2007); Poland, Luxembourg (2008); United Kingdom, Finland, Japan (2009); India (2010). Australia, Belgium, Italy, Japan, Sweden, United Kingdom, United States Regulation quantity Renewable portfolio standards or quotas (RPS) require that a minimum percentage of generation sold or capacity installed be provided by (2004 or before); Poland (2005); China (2006); India (2006); India (2007); Portugal (2010). renewable energy. Obligated utilities are required to ensure that the target is met, either through their generation, power purchase from other producers, or direct sales from third parties to the utility's customers. Tradable renewable certificates represent the certified generation of Australia, Austria, Belgium, Denmark, Finland, France, Ireland, Italy, units of renewable energy. They allow trading of renewable energy Japan, Norway, Sweden, United Kingdom (2004 or before); Netherlands obligations among consumers and/or producers, and in some markets (2005); Czech Republic, Hungary, Russia (2006); India, Poland, Spain, like the United States allow anyone to purchase separately the green United States (2010). power “attributes” of renewable energy. Australia, Canada, China, France, Germany, India, Japan, Spain (2004 or Public competitive bidding means that public authorities organise before); Austria, Denmark, Poland (2005); Brazil, Hungary, New Zealtenders for a given quota of renewable supplies or capacity, and reand (2006); Greece, Italy, Norway, Portugal, South Africa, Sweden, munerate winning bids at prices that are typically above standard United Kingdom (2008), United States (2010). market levels. Sales tax reductions Sales tax, energy tax, excise tax or VAT reduction are fiscal policy tools Canada, China, Finland, France, Germany, India, Sweden, United Kingdom (2004 or before); Belgium, Denmark, Poland (2005); Czech Reproviding consumption tax exemptions or reductions on the sale of public, Hungary, (2006); Portugal (2007); Indonesia, Israel, Italy, Luxrenewable energy and equipments. embourg, Netherlands, Norway, South Africa, Spain, Switzerland (2008); Brazil, United States (2010). Australia Canada China France, Germany India, Japan, Spain (2004 or Fiscal incentives Public investments, loans or financing are public policies aimed at directly acquiring renewable power generators, or at providing ad-hoc before); Austria Denmark, Poland (2005); Brazil, Hungary, New Zealand (2006); Greece, Italy, Norway, Portugal, South Africa, Sweden, United subsidised financing for private investors. Kingdom (2008); United States (2010). Capital subsidies, consumer grants or rebates are one-time payments Australia, Austria, Belgium, Canada, China, Finland, France, Germany, by the government or utility to cover a percentage of the capital cost of India, Ireland, Italy, Japan, Norway, Spain, Sweden, United Kingdom, United States (2004 or before); Greece, Netherlands, Poland, South an investment, such as a solar hot water system or rooftop solar PV Africa (2005); Czech Republic, New Zealand, Russia (2006); Portugal system. (2007); Denmark, Luxembourg, Switzerland (2008); Hungary (2009). Investment or other tax credits allow full or partial deduction from tax Austria, Belgium, Canada, China, Denmark, France, Germany, India, obligations or income for investments in renewable energy. Ireland, Italy, Norway, Spain, Sweden, United States (2004 or before); Greece, Netherlands, (2005); Czech Republic (2006); Portugal (2007); Brazil, Luxembourg (2008); Hungary, Indonesia, Japan (2009). Finland, Sweden, United States (2004 or before), India, Netherlands Energy production payments or tax credits provide investors or owners of qualifying property with an annual tax credit or a payment (2005); China, (2008); United Kingdom (2010). based on the amount of electricity generated by that facility. a

For a few country-year pairs for which the information was not available it has been imputed based on information on the previous or following year.

about a third of the sample), the data have been complemented with punctual information from another VC database (Thomson One).8 This allowed us to improve the coverage and quality of the information on investees’ and investors’ type, location and founding year. 2.2.2. Data on policies The main source of information on environmental policies is the Renewable Energy Policy Network for the 21st Century (REN21) with reports starting in 2005 and updated yearly since then. The information has been complemented and cross-checked with information from the International Energy Agency (IEA); Pew Trust reports for information relating to 2009 and 2010 and from the OECD Environment Directorate (Johnstone, Haščič and Popp, 2010) for information relating to Feed in Tariffs (FITs) and tradable energy certificates in 2005, which is the latest year for which information from this source is available. Policies to promote renewable energy that are covered in the analysis can be classified as (i) price-driven regulation policies, such as FITs; (ii) quantity-driven regulation policies, e.g. tradable 8 This information was very kindly provided by NESTA. We are particularly grateful to Yannis Pierrakis, Liam Collins and Albert Bravo-Biosca for their help.

renewable certificates; (iii) sales tax or Value Added Tax (VAT) reduction and related mechanisms; and (iv) fiscal incentives, such as direct capital investment subsidies, grants or rebates, tax incentives, and tax credits for renewable energy. Given that generally, the only information available for the deployment policies considered is their presence, we construct dummy variables on whether they exist in a particular country and year. In addition when we consider the different group we construct a group-wide policy measure that is based on the number of policies in each group enacted in a given country and year. Table 3 describes in detail the classification used in the analysis and the policies included in each group, as well as the countries and the dates when these policies were enacted. As shown in Table 3, government support for renewable energy differs both across countries and over time in the sample included in the analysis. One of the most commonly used renewable policies is FITs, which were first introduced in 1978 in the United States (although the programme was discontinued in most US states in the 1990s). In the early 1990s feed-in policies were introduced in Denmark, Germany, Greece, India, Italy, Spain, and Switzerland and in 2005 in Ireland. Among the emerging economies, India was the first to adopt FITs, followed by Brazil, Indonesia, and – in the first half of 2005 – by China, as part of a comprehensive renewable energy promotion law enacted in

C. Criscuolo, C. Menon / Energy Policy 83 (2015) 38–56

February 2005. Table A3 in the Appendix A reports the share of countries in which any given policy is activated in a given year: for most policies the share is growing over time. Renewable portfolio standards (RPS) and public competitive bidding for specified quantities of power generation are also increasingly being used. Most of the RPS policies require renewable power shares in the range of 5–20%. Most RPS targets translate into large expected future investments. Other forms of regulatory policies for renewables include tradable renewable energy certificates, and a number of other policies with more limited impact, such as appliance/equipment efficiency standards, building energy codes, energy efficiency resource standards, energy standards for public buildings, equipment certification requirements, generation disclosure, green power purchasing policies, interconnection standards, mandatory utility green power option, and net metering. Sales tax reductions are also demand-side policies which have been increasingly implemented in a number of countries in recent years. Other forms of support for renewables are also being used: fiscal incentives, such as direct capital investment subsidies or rebates, tax incentives and credits, direct production payments or tax credits (i.e., per kWh). Some type of direct capital investment subsidy, grant, or rebate is offered in as many as 30 countries globally. Governments are also increasingly providing direct public investment or financing. Some countries and regional authorities have established renewable energy funds that are used to directly finance investments, provide low-interest loans, or loan guarantees. Additional support includes supply-push policies, of which research and development funding is a significant part. R&D and Demonstration (RD&D) budgets in the energy sector, i.e. the total amount and allocation of funding by category of energy technologies for research, development and demonstration, would have been the best proxy for supply-push policies in energy. However, due to concerns over comparability and availability of data for non-IEA member countries, data on government R&D budgets are used as a proxy for government energy RD&D. The policy measures used in the analysis have the advantage of being precise on the nature of the policy tool in place, encompassing nine different definitions, which are then grouped into four categories. This allows the disentangling of the different effects of each policy domain, which would not be possible using general indicators of environmental policy stringency.9 Furthermore, they are timely (the latest available year is 2010) and are available for both the OECD and non-OECD economies included in the sample. On the other hand, the measures present some drawbacks. Firstly, they only contain information on whether a policy is in place, but they do not provide detailed information on the characteristics of the policy (generosity of support; stringency; and other features of the policy design). For this reason, we tried to gather wherever possible alternative policy measures that could provide a more precise picture of the generosity of the support or the nature of the target considered. This information was only available for the solar and wind sectors from the newly available Renewable Energy Policy Database (OECD-EPAU, 2013) providing harmonised country-level data on FIT levels and REC quotas over the period 1978–2011. Secondly, the policies considered are arguably more relevant for the promotion of renewable energy in the electricity generation sector; for this reason, the empirical analysis includes a robustness check limiting the sample to the energy sector. Thirdly, the policies considered are all at the national level; policies offered at the regional or local level are not taken into account. This can be particularly limiting in the case of 9

For a review of environmental policy indicators and their properties, see Botta and Kozluk (2015).

45

the United States, as there is a significant degree of variation across State authorities in the number and nature of policies in place. One of the reasons why only national level policies are considered is that evidence from other policy domains, e.g. regional and innovation policies (Wilson, 2009; Greenstone et al., 2010) suggest that local and regional measures might yield a relocation of investments across areas within a country and thus result in a zero sum game at the national level. This would imply that we could not detect their impact using our cross-country data. Since the country for which this might be more of an issue is the United States, robustness tests excluding the United States are reported in the empirical analysis. 2.3. Empirical analysis The aim of the paper is to look at the policy determinants of VC financing in the green sector across countries. This is achieved by investigating the relationship between “green” VC investment and environmental policies aggregating the available deal-level information (number and amount) into year-country-industry cells. The following two equations are estimated. In the first equation the dependent variable is total number of VC deals in sector s, at time t, in country c:

No_dealssct = f (α, regulation_pricect , regulation _ quantityct , sales _tax _reductionct , public _incentivesct , Government _R &Dc, t − 2 , Zct , Dyear . Dsector )

(1)

where Z is a matrix of country-year specific control variables, and Dyear and Dsector are a set of year and sector dummy variables, respectively. The number of VC deals in sector s, at time t, in country c takes only non-negative integer values with a significant share of zeros, which rules out the possibility of a logarithmic transformation. Therefore, we estimate Eq. (1) via a Negative Binomial model which is preferable to log linearising a transformation of the dependent variable – generally taking logs of one plus the dependent variable – when the dependent variable is a non-negative count in order to avoid out-of-sample predictions.10 Since in the Negative Binomial model the estimated coefficients correspond to semi-elasticities, coefficient estimates can be directly converted into marginal effects. For a continuous regressor x, the marginal effect is (∂E [y|x])/∂xj = exp(xβ) βj . The table reports marginal effects calculated at the mean. In the second equation, the dependent variable is the total amount of VC investment in sector s, at time t, in country c:

Deal _Amountsct = f (α, regulation _pricect , regulation _ quantityct , sales _tax _reductionct , public _incentivesct , Government _R &Dc, t − 2 , Zct , Dyear . Dsector )

(2)

The total amount of VC investment received in sector s, at time t, in country c is a continuous variable, but it equals zero for 18% of the industry–country–year cells sample: the sample is therefore censored. Given that the underlying data generating process results in a zero corner solution, i.e. in some sector–year–country cell the dollar amount of VC investment is zero, the regression equation is estimated using the Tobit model, which yields – under 10 Another more commonly used alternative for count data is the Poisson model. However, different tests indicate that the Negative Binomial better fits the data under scrutiny in this paper.

46

C. Criscuolo, C. Menon / Energy Policy 83 (2015) 38–56

certain assumptions – consistent estimates in the case of censored samples. The table reports the marginal effects of regressors on the dependent variable conditional on the cell having positive investment amounts, as suggested in Wooldridge (2002). The explanatory variables of interest are deployment and supply side policies. Deployment policies are proxied by the number of policies enacted in a given country and year, grouped in the four following groups: regulation price, regulation quantity, sale tax reduction, and fiscal incentives, as described in Table 3. Supply-side policies are proxied by the government R&D expenditure in country c at time t 2.11 Regression equations also control for other country-level time varying variables – included in the matrix Z – that might affect the overall level of VC activity in a country. These include GDP per capita, to control for aggregate economic activity (see for example Kraussl and Krause, 2012 and references therein); the number of days needed to open a business, to proxy for entry barriers and general framework conditions for entrepreneurship (Tyebjee and Bruno, 1984); the stock market capitalisation, as a proxy for the development of the stock market in a country (Black and Gilson, 1998), which in turn correlates positively with the expected exit opportunities for Venture capitalists.12 The model also includes the number of green sector companies seeking funding, to proxy for the entrepreneurial dynamism in green sector s and country c; the variable is one year lagged to minimise endogeneity. Finally, the model also includes time dummies to capture the effect of global macroeconomic shocks – such as variation in oil prices and media coverage (see Cummings et al., 2013) – and industry dummies to control for time invariant unobserved industry level characteristics. We have conducted a series of robustness checks relative to the baseline specification. We start by testing the robustness of our results to the sample of countries and sectors used. Firstly, to investigate possible heterogeneous effects of policies at different investment stages, the models are also estimated separately for investments at the seed/early, follow-on, and other investment stages, respectively. Secondly, to address concerns regarding the role played by the United States and measurement errors in the policy variable for this country, the baseline equations are estimated excluding the United States. Thirdly, given concerns on the aptness of the indicators for environmental policies used to capture the policy environment in sectors other than energy generation, we limit the sample to the energy generation sector. We then test the robustness of our estimates to the presence of outliers and to the skewness of the dependent variables. Finally, we also test the robustness of our results to the choice of policy indicators and their classification and we use a new harmonised dataset (OECD-EPAU, 2013) reporting the level (and not just the presence) of FITs adopted by each country in the solar and wind sector.13 We exploit this information by limiting our analysis to these two sectors, which in our dataset are identified by the secondary industry definition within the energy generation industry. The fact that the policy variables are now continuous and sector-specific allows for the inclusion of a country dummy which controls for the effect of other policies (on average over the period) and time-invariant national factors. The FITs variables are interacted with a wind and solar sector dummy, respectively, i.e., the variable is set to zero in the sector which it does not apply to. 11 The choice of the lag is driven mainly by data availability, but the government R&D expenditure is serially correlated and therefore the results are unlikely to be affected. 12 GDP data come from the OECD; the other control variables are sourced from the World Bank. 13 The dataset also contains information on FITs in the marine and geothermal sectors, but the number of related deals in our dataset is very small.

Formally, the estimated models are specified as follows:

Financingsct = f (α, Ds ∑ FITsct , Zct , Dc , Dt , εsct ) s

(3)

where the dependent variable financing is measured as the (i) total amount of VC investment received in sector s, at time t, in country c and (ii) as the total number of deals in sector s, at time t, in country c. The model also includes the other environmental policies and economy-wide controls Z included in Eq. (1), although most of their effect is absorbed by the country dummies.

3. Results 3.1. Descriptive analysis The dataset contains 7268 observations, of which 4898 are VC deals, and 2370 are reports of companies seeking VC funding. It covers 72 different countries over the time period 1999 to July 2011. We dropped from the sample deals and seekers from countries for which it was not possible to find reliable data on environmental policies. For the same reason and to ensure an adequate coverage of deals in the green sector, the analysis is restricted to the period 2005–2010 for 29 OECD and emerging economies, i.e. to a sample covering 4792 observations, of which 3007 are deals, and 1785 are seekers. Fig. 6 reports the number of deals by year and industry. As shown in the figure, energy generation represents the largest recipient of VC deals (30% of deals in 2005 and up to 44% of deals in 2007), followed by energy efficiency, which is also the sector that grew most in relative terms, from 8% of all green sector deals in 2005 to 21% in 2010.14 This pattern is consistent with anecdotal evidence that venture capitalists have shifted their activity within the green sector towards ventures that are less capital intensive and have features similar to ICT start-ups, such as those in the energy efficiency sector. In addition, in some industries – such as energy generation – the size of the market for the goods or services offered by the firm seeking VC backing could depend heavily on the regulatory framework in place in a country, given the inherent public nature of the good in question. Others, such as energy efficiency, might be more resilient to policy changes. A similar pattern is shown in Fig. 7, which aggregates the number of deals according to the stage of the investment: seed and early stage; later stage; and buy-out. Almost half of all the deals are later stage investments, while seed and early stage investments represent about one third of the sample. Over the years, seed investments have become less of a target for investors, relative to later stage and buy-out investment. Again, this suggests a shift by venture capitalists towards less risky investments in the green sector. In terms of geographical distribution of deals and of companies seeking funding in the sample, around half of the deals involve companies based in the United States, and about 40% of companies seeking funding are in the US. This suggests that the larger number of deals partly reflect the higher overall entrepreneurship rate. The United Kingdom and Canada are ranked second and third respectively in the number of deals. When comparing the ratio of companies receiving funding to those seeking funding, the United 14 The distribution of deals and seekers across industries is also fairly concentrated, with the energy generation sector accounting for more than a third of observations of deals and seeking firms, and energy efficiency accounting for 15% of observations. Both sectors are therefore very dynamic, and not necessarily the ones in which firms have the highest likelihood of receiving funding. Other sectors, such as energy storage and recycling and waste, show a much higher funded/seekers ratio.

C. Criscuolo, C. Menon / Energy Policy 83 (2015) 38–56

47

Fig. 6. Number of deals by primary industry and year. Source: Authors' elaboration based on the Cleantech Market Insight Database.

seed

early stage

later stage

buy-out

other

500

600

2005 2006 2007 2008 2009 2010 0

100

200

300

400

700

800

Number of deals Fig. 7. Number of deals by investment type. Source: Authors' elaboration based on the Cleantech Market Insight Database.

Millions 2005 USD

Mil. $ (total)

Total

Mil. $ (av.)

Average

40,000

140

35,000

120

30,000

100

25,000 80 20,000 60 15,000 40

10,000

20

5,000 0

0 seed

early stage

later stage

Fig. 8. Total (scale on the left) and average (scale on the right) amount by investment type. Source: Authors' elaboration based on the Cleantech Market Insight Database.

buy-out

other

48

C. Criscuolo, C. Menon / Energy Policy 83 (2015) 38–56

Millions 2005 USD Millions $

Total

Average

Millions $ 30

40,000

25 30,000 20 15

20,000

10 10,000 5

0

0

Fig. 9. Total (scale on the left) and average (scale on the right) amount invested by primary industry. Source: Authors' elaboration based on the Cleantech Market Insight Database.

Kingdom shows a much higher ratio between funded and seeking companies (504 to 143, as compared to 2456 to 1519 for the United States, and 308 to 176 for Canada), suggesting that, once the higher number of companies seeking funding is accounted for, the United States is not necessarily the most attractive country for green sector VC investors. Up to now the focus has been on the number of deals; however deals might differ significantly in their value. Thus, we also present figures on average and total amounts invested, aggregated over stages of investment, industry, and year. Looking across stages of investment, the data shows that the average investment amount is smaller than USD 20 million for all but buy-out investments, that reach an average amount of around USD 120 million; i.e. 80 times the average seed investment (worth USD 1.5 million) and 8 times the average follow-on investment (which is around USD 14 million – Fig. 8). Given the larger number of follow-on deals, follow-on investment accounts for 37% of total amount invested, and buyout, in which there are fewer but much more sizeable deals, accounts for 38% of total amount invested. Seed and early stage investment together account for only 5% of aggregate total investment amount.

The average investment amount by primary industry is largest in the energy generation and manufacturing/industrial sectors ( 25 million); it is equal to around 10 million in energy efficiency, while in all other industries is lower than 5 million (Fig. 9). The average and total amount of investment have been growing over time, but with some volatility (Fig. 10). The year with the highest average and total volume of investment is 2008, while figures for 2009 and 2010 are somewhat lower, following the international financial crisis (Lerner, 2011). The empirical analysis assesses whether the presence and size of “green” VC investment in a country is correlated with the type and number of environmental policies in force. In the analysis, policies are grouped into regulatory policies, fiscal incentives and public finance. Fig. 11 and Table 3 (and A3 in the Appendix A) show that there is indeed some variation in the extent to which these different policies are used, both across countries and time. A number of countries increase the number of policies in force during the 2004–2010 period, although at a relatively slow rate. As shown in Table 3, the Netherlands, Sweden, and the United States are the countries with the most intense deployment of environmental fiscal incentives, while India and Italy have the highest

Fig. 10. Total amount (scale on the left) and number of deals (scale on the right) by year quarter. Note: dollar values are expressed in real terms. Source: Authors' elaboration based on the Cleantech Market Insight Database.

C. Criscuolo, C. Menon / Energy Policy 83 (2015) 38–56

49

Fig. 11. Changes in number of deployment policies over the 2004–2010 period. Source: Authors' elaboration based on the Cleantech Market Insight Database.

Table 4 Industry–country level regressions. Estimation model:

Number of deals (1)

(2)

(3)

(4)

(5)

(6)

(7)

Energy generation

No US

Energy generation no US

9.37

2.71 0.266 (0.360) 1.077c (0.381) 1.634c (0.387) 1.151a (0.636) 8.434c (2.179) YES YES YES 551

4.96 0.584 (0.763) 1.846b (0.881) 4.395c (1.028) 0.0642 (1.243) 15.27c (5.919) YES YES YES 114

Negative binomial Dep. var.: Sample

Number of deals All Early stage

Follow-on

Other stages

Mean dep. var. Regulation price

4.99 0.394 (1.082) 3.744a (2.100) 2.267 (1.975) 4.179 (2.947) 35.51b (17.02) YES YES YES 621

2.37 0.0336 (0.396) 1.020b (0.418) 0.453 (0.369) 0.211 (0.752) 11.06c (2.164) YES YES YES 621

6.26 1.298 (2.792) 7.144 (11.74) 3.084 (6.035) 8.607 (14.33) 65.73 (106.3) YES YES YES 621

Regulation quantity Sale tax reductions Fiscal incentives Gov. R&D as % GDP Other controls Sector F.E. Year F.E. Observations

1.07 0.0716 (0.198) 0.525a (0.286) 0.390 (0.247) 1.715c (0.517) 5.337c (1.364) YES YES YES 621

3.494 (2.660) 6.527a (3.590) 14.60a (7.877) 4.094 (4.812) 41.35b (20.07) NO NO NO 120

Note: Estimates reported in the table are average marginal effects. The unit of observation is the year–industry–country combination (year 2005–2010, countries with at least 5 deals and cells with at least one deal or one company seeking investments). All regressions also include controls for: total stock traded as % GDP; days needed to start a business; GDP p.c. (t 1) in PPP. Robust standard errors clustered at country–year level in parentheses. a b c

p o 0.1. p o0.05. po 0.01.

number of regulatory policies. Most countries increased the number of deployment policies over time. Others – France, Germany and Spain – did not experience any change in the number of policies implemented. Finally, Japan, Ireland, Sweden, Belgium and Australia decreased the number of policies in place. 3.2. Results from the econometric analysis Table 4 reports the results of the Negative Binomial estimates of Eq. (1), while Table 5 reports the results of the Tobit estimation of Eq. (2). All estimates also include unreported controls: GDP per capita, number of days necessary to open a business in the country, the value of the stock market and year and sectoral dummies, as reported in the tables' notes; dummy variables are meant to control

for other country and industry level determinants of VC investment and for macroeconomic shocks. Regression coefficients are expressed as marginal effects; the Tobit marginal effects are conditional on the amount of investment being positive. Results in the first column of Table 4 suggest that the number of deals is robustly associated with regulation policies affecting the quantity of renewable energy, like renewable energy certificates (REC) or renewable energy quotas. Government R&D intensity is also positively associated with the number of deals. Columns 2–7 show that these findings hold across different investment stages, as well as when the United States is excluded or when the sample is limited to the energy generation sector. Sales tax reductions, on the other hand, are only positive and significantly correlated with the number of deals in the energy generation sector (columns 5 and 7) and if the United States is excluded from the sample

50

C. Criscuolo, C. Menon / Energy Policy 83 (2015) 38–56

Table 5 Industry-country level regressions. Amount of fundinga (1)

(2)

(3)

(4)

(5)

Estimation model:

Amount of funding

Sample

Regulation price Regulation quantity Sale tax reductions Fiscal incentives Gov. R&D as % GDP Other controls Sector F.E. Year F.E. Observations

(7)

Tobit

dep. var.:

Mean dep. var.

(6)

All 83.03 b

Early stage

Follow-on

Other stages

Energy generation

no US

Energy generation no US

5.50

36.22

97.86

202.47

38.45

88.65

c

c

c

c

21.456 (8.368) 22.038c (7.639) 19.538b (9.310) -12.197 (13.050) 161.296c (51.913)

1.399 (0.813) 1.839c (1.017) 0.313 (0.786) -4.169b (1.613) 16.448c (6.064)

4.099 (4.755) 17.991c (6.798) 5.107 (4.698) -8.292 (9.392) 96.592c (33.178)

34.526 (10.514) 40.665c (13.995) 2.952 (11.945) -37.451c (22.246) 366.525c (86.417)

81.110 (22.712) 44.345c (24.708) 5.595 (24.911) 13.528 (31.873) 338.383b (166.695)

12.780 (4.431) 12.773c (4.782) 9.675c (5.089) -2.053 (8.198) 99.570c (35.584)

46.283c (14.254) 13.934 (13.152) 16.954 (16.628) 16.065 (16.397) 150.653 (108.703)

YES YES YES 621

YES YES YES 621

YES YES YES 621

YES YES YES 621

NO NO NO 120

YES YES YES 551

YES YES YES 114

Note: the unit of observation is the year-industry-country combination (year 2005-2010, countries with at least 5 deals and cells with at least on deal or one company seeking investments). All regressions also include controls for: total stock traded as % GDP; days needed to start a business; GDP p.c. (t-1) in PPP. Reported Tobit marginal effects are for the expected value of the dependent variable conditional on being uncensored, E(y | y4 0). Robust standard errors clustered at country-year level in parentheses. a b c

p o 0.1 po 0.05 po 0.01

(column 6). Finally, fiscal incentives are never significant except in columns 2 and 6, where the dependent variable is limited to early stage investments or the United States is excluded, respectively: in these cases, the coefficient takes a negative sign. One tentative explanation of the result may refer to the perceived instability of these kinds of policies, which might be particularly relevant for investments still far from the commercialisation phase. While these estimates are suggestive of the importance of both supply-side and deployment policies, they abstract from the

magnitude of aggregate investment. Therefore, Table 5 reports the marginal effects of these same policies on the total amount of funding across industry–country– year cells, conditional on the value being positive. The previous results on the positive effect of “quantity” regulation and public R&D are confirmed; in addition, the estimates point to a positive effect of price-targeting policies, such as FITs. The positive correlation between price regulation and amount of funding is robust to excluding the United States from the sample, restricting the

Table 6 Industry–country level regressions: the role of individual regulation quantity polices. Sample

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

6.021 (9.457) YES YES YES YES 621

25.077c (9.323) 43.149 (17.819) -20.057 (14.895) YES YES YES YES 621

All Estimation method

Negative binomial

Dependent variable

Number of deals

Public competitive Bidding

Total amount a

0.615 (0.808) 4.188c (0.963)

Renewable portfolio Standard or quota Tradable Certificates Other policy variables Other controls Sector F.E. Year F.E. Observations

Tobit

YES YES YES YES 621

YES YES YES YES 621

3.202 (0.967) YES YES YES YES 621

1.757 (0.911) 4.747c (1.352) 0.156 (1.581) YES YES YES YES 621

20.887b (8.639) 19.582a (10.217)

YES YES YES YES 621

YES YES YES YES 621

Note: the unit of observation is the year–industry–country combination (year 2005–2010, countries with at least 5 deals and cells with at least on deal or one company seeking investments). All regressions also include controls for: regulation price polices; sales tax reduction policies; fiscal incentives; total stock traded as % GDP; days needed to start a business; GDP p.c. (t 1) in PPP. Reported Tobit marginal effects are for the expected value of the dependent variable conditional on being uncensored, E(y| y40). Robust standard errors clustered at country–year level in parentheses. a b c

p o 0.1. p o0.05. po 0.01.

C. Criscuolo, C. Menon / Energy Policy 83 (2015) 38–56

analysis to the energy generation sector and to seed and early stage financing, but becomes insignificant when restricting it to follow-on investments. Fiscal incentives, however, seem to play no role in attracting VC; when the sample is restricted to seed and early stage deals, the coefficient is significant and negative, as was the case when the dependent variable was number of deals. Unreported year dummies are always highly significant, both in the Negative Binomial and in the Tobit model, suggesting the importance of global time varying factors e.g. oil prices and media coverage, as recently found by Cummings et al. (2013). Table 6 further explores the positive effect of quantity regulation by disentangling the variable into its three components, i.e., renewable portfolio standards or quotas (RPS); tradable renewable certificates; and public competitive bidding. The three dummy variables, indicating whether the related policy is enacted in a given country and year, are included in the regression model oneby-one in three different regressions (columns 1–3 for the Negative Binomial, and columns 5–7 for the Tobit), as well as all together (columns 4 and 8 for the Negative Binomial and the Tobit estimations respectively). In all other respects, the regression models are identical to those reported in column 1 of Tables 4 and 5. The results show that Renewable portfolio standards or quotas variable is the most robust and impactful one, as it keeps it significance both in the Negative Binomial and in the Tobit regressions, irrespectively whether is included alone or with the other two dummies. Public competitive bidding is also significant in both the Negative Binomial and Tobit regression, while the tradable certificates dummy is significant only in the Negative Binomial when the two other dummies are excluded. In Table A3 in the Appendix A we report a further robustness test that tackles the issue of skewness of the two main dependent variables and the presence of some outliers – i.e., cells with an extremely high number of deals and, especially, amount of investment. In order to ascertain that these outliers are not driving our results, we run the main regressions using two different transformations of the dependent variables: a winsorized form (the top 1% of the distribution is set equal to the 99th percentile) and the square root. As it is possible to see in Table A3, the signs of the coefficients are unaffected, both in the Negative Binomial and in the Tobit estimations; while their significance level shows only slight differences with respect to the baseline estimates, namely for the coefficients on sale tax reduction polices and fiscal incentives. To summarise, the results show that the number of deals and the total amount of funding are robustly associated with the national regulatory policies in force in a country, as well as with the share of government R&D expenditures in GDP. Among regulatory policies, those targeting the quantity of clean energy produced seem to affect both the extensive (number of deals) and intensive margin (amount of funding) of investment. On the other hand, policies targeting price are not correlated with the extensive margin, nor with the amount of investment in follow-on stages; however, they are positively correlated with seed and early stage investments (both in the whole green sector and in the energy generation sector). Importantly, sales taxes have a much weaker and less robust positive effect, while fiscal incentives are never significant.

51

Table 7 Secondary industry-country level regressions. Energy generation only; number of deals and amount of funding; regression coefficientsa 1 Estimation method Dep. var.

2

3

4

Negative Binomial

Tobit

Number of deals

Amount of funding

Mean dep. var. FIT solar FIT wind

2.177c (0.429) 2.575b (1.201)

5.516b (2.274) 1.477 (4.867) -5.381 (3.425) 5.781 (14.21)

-42.179 (159.363) 189.992 (427.274)

2,420.917c (681.319) 553.089 (861.319) -3,845.922c (1,130.185) 887.920 (2,486.396)

YES YES YES 125

YES YES YES 125

YES YES YES 125

YES YES YES 125

FIT solar quadratic FIT wind quadratic other controls country FE year FE Observations

Note: the unit of observation is the year-secondary industry-country combination (year 2005-2010, countries with at least 5 deals and cells with at least on deal or one company seeking investments). All FIT variables take positive values only for observation in the relevant sectors. All regressions also include controls for: number of active regulation quantity, sale tax reductions, and fiscal incentives policies; tot. stock traded as % GDP; days needed to start a business; GDP p.c. (t-1) in PPP. Robust standard errors clustered at country-year level in parentheses. a b c

p o0.1 p o 0.05 p o 0.01

solar and in the wind sector. The second column includes also quadratic terms, which allows one to explore whether at high levels of FITs, the effect might become marginally smaller, or even negative. This is motivated by existing evidence showing that excessively generous FITs tend to have a negative effect due to expectations that the policy might be soon revoked due to its perceived fiscal unsustainability (Criscuolo et al., 2013). Results in column 4 tend to support this hypothesis for the solar sector: the coefficient on the quadratic term is indeed negative, implying that the estimated association between FITs and the total amount of VC funding follow an “inverse U” curve, sloping upwards for lower values of the policy, and downwards for high values. This is not the case for wind power. Given the much higher average values for FITs for solar than for wind power in most countries, the estimates seem to be consistent with such an interpretation. Moreover, this might also explain the lack of significance of the price regulation coefficient in column 3 of Table 5 when the analysis is restricted to follow-on investment that is on average much larger than seed and early stage. Rather than reporting marginal effects of the FIT variable – which enters as a quadratic polynomial – Table 7 reports coefficients’ estimates while Fig. 12 plots marginal effects of FITs. The graph clearly shows the “inverse U” relationship between the generosity of FITs in the solar sector and the amount of funding.

3.3. Analysis of the effect of FITs in the solar and wind sector

3.4. Caveats on the econometric analysis

The results from the estimation of Eq. (2), limited to the solar and wind sectors and including sector-specific policy variables, are reported in Table 7. The first column shows Negative Binomial estimates with the number of deals as the dependent variable. As shown, FITs are positively associated with the number of funded deals both in the

As with all econometric estimates, the results reported in Tables 4–7 need to be interpreted with caution. Firstly, the policy measures included in the analysis are very broad and do not take into account the details of environmental policy designs, and are only based on countries' renewable energy policies. However, they provide a first indication of countries' predisposition toward

52

C. Criscuolo, C. Menon / Energy Policy 83 (2015) 38–56

Fig. 12. Plot of marginal effects of FITs in the solar sector. Panel A: Tobit regression – amount of funding. Panel B: Poisson regression – number of deals. Note: the marginal effects are calculated based on the coefficients reported in column 2 (Panel A) and column 4 (Panel B) of Table 6. The x-axis reports the level of FITs and the y-axis the estimated marginal effects at the mean value of all other variables.

environmental policies. The choice for the indicators used was mainly driven by the lack of timely comparable environmental policies indicators for both OECD and non-OECD economies included in the sample. Secondly, given the limited time variation in the deployment policies considered and the strong serial correlation of government R&D intensity, we could not estimate the role of deployment and supply side controlling for country unobserved fixed effects (except in the regressions restricted to energy generation). We hope that the availability of more detailed information on environmental policies, at the industry level, could help overcome this limitation of the study.

4. Discussion The results suggest that national environmental deployment policies designed with a long-term perspective of creating a market for environmental technologies, such as price and quantity renewable policies, e.g. FITs and tradable certificates, are associated with higher investment levels relative to more short-term fiscal policies, such as tax incentives and rebates. This result suggests that long-term policy stability, sustainability and credibility are important to ensure financing of innovative and risky ventures in the green sector. This result is consistent with econometric evidence from

Johnstone et al. (2010), who find a positive correlation between perception of environmental policy stability and patenting applications in environmental technologies. This result is also in line with existing evidence from survey results (Barradale, 2010) that suggests that patterns of repeated expiration and short-term renewal of the US federal production tax credit might have been a cause of uncertainty for returns on investments in wind power and thus might in part be responsible for increased volatility in this sector. The same survey, on the other hand, suggests that regionaland national-level portfolio standards were considered by respondents as stable enough to influence long-term investment planning. The analysis does not find any significant correlation between public investment loans or financing and the amount of private financing of innovative ventures in the green sector. However, this does not mean that government financing, e.g. through public VC funds invested in partnership with private and corporate VCs, could not represent a possible solution to the financing gaps in the green sector. This gap arises because the level of investment required by the green sector is on average much larger than in other sectors and the time span of green projects from the seed to the scaling-up phase is much longer than the average life of a VC fund; both of these issues might be less problematic for government investments. Results also suggest that supply push policies have a positive impact on financing: we find that government R&D is an important predictor for the level of investment in clean sectors. On the contrary, sales taxes have a much weaker and less robust positive effect. This might suggest that both supply and deployment policies might be useful for countries that would like to diversify the portfolio of environmental technologies and resources to advance, for example, energy security goals (“letting a thousand flowers bloom”). At the same time, the results suggest that fiscal incentives and public finance policy measures do not appear to matter for this purpose. Finally, analysis restricted to the energy generation sector focusing specifically on FITs suggests that these policies tend to show marginally decreasing returns, to the point that they might even discourage investments in the case of very generous provisions. The finding is consistent with other contributions in the literature, as well as with anecdotal evidence, suggesting that a very high level of FITs could raise credibility concerns about the fiscal sustainability of the policy in the medium and long term, as FITs are often at least partly funded by the public budget.

5. Conclusions and policy implications The study has looked at the linkages between both supply-side R&D support and market demand policies (such as FITs and RECs) and high-growth financing deals in the green sectors. Policy makers can play an important role at different points in the market development process to foster innovation activity and entrepreneurship in the “green energy” sector. In the first instance, governments play an important role in inducing invention through grants, loans and prizes and advanced R&D support, or through funding very early stage investments via special programmes, such as the Advanced Research Projects Agency-Energy (ARPA-E) programme in the United States. The US Department of Energy has its National Renewable Energy Laboratory (NREL) which has contributed to a number of significant renewable energy advances. China has a number of government related green R&D programmes, such as the National High-Tech R&D programme (USD 2.9 billion) and the National Basic Research Programme with funding of USD 585 million; environmental technology research institutes and laboratories can be found in several Chinese

C. Criscuolo, C. Menon / Energy Policy 83 (2015) 38–56

universities (Cleantech, 2010). Further downstream, many national governments in the last decades have adopted regulation and deployment policies, sales tax or VAT reduction mechanisms, in order to stimulate both demand and supply in the green energy sector. Direct support for capital investment has also been provided by a number of countries through loan guarantees, grants, and accelerated tax depreciation schedules. And finally, policies which affect returns on the sale of electricity generation from renewable energy sources – such as feed-in-tariffs and renewable energy certificates – have been adopted by most OECD governments in one way or another. The challenge for policy makers at all stages of technological development is not to pick “national champions” or “technological winners” and allow as much experimentation as possible, especially in the early stage of research and technological development. However, even the use of downstream policies such as FIT policies which are differentiated according to energy source can have a significant influence on the direction of early-stage market development in the renewable energy sector as VC investment is responsive to policy inducements at all points in market development. However, despite the existence of upstream R&D support policies and downstream market demand policies entrepreneurs/ inventors still face a significant financing challenge when evolving from the laboratory phase to the development of the technology (the so-called “technological valley of death”). In recent years,

53

VC/PE has played an increasingly important role in helping firms’ transition from the laboratory to the market. This is particularly true in the United States, the United Kingdom and more recently in China. However, VC and PE still only represents a small percentage of overall funding sources for the green energy sector.

Acknowledgements We thank Fabio Bertoni, Massimo Colombo, Stefano Elia, Ivan Haščič, Nick Johnstone, Pietro Moncada-Paternò-Castello, Sandro Montresor, Dirk Pilat, Victoria Shestalova, Mariagrazia Squicciarini, Peter Voigt, Andrew Wyckoff, and Karen Wilson for useful discussions and comments. We are also grateful to the members of the OECD Committee on Industry, Innovation and Entrepreneurship (CIIE), to participants in the seminar at the Dipartimento di Ingegneria Gestionale at Politecnico di Milano and in the ZEWCODE Conference 2012 for valuable comments, to Hélène Dernis for help with the patent data, and to Teresa Taló and Linda HaieFayle for excellent editorial assistance.

Appendix A See Appendix Tables A1–A4 and Fig. A1.

Table A1 Primary, secondary and tertiary industries in the Environmental technology sector covered by the Cleantech database. Primary industry

Secondary industry

Tertiary industry

Agriculture

Aquaculture Land management Natural pesticides

Farms, health & yield Crop yield, erosion control, precision agriculture, smart irrigation, soil products/composting, sustainable forestry Antimicrobial, beneficial insects, biological control, organic fungicides, sustainable fertilisers

Air & environment

Cleanup/safety Emissions control Logistics Monitoring/compliance Trading & offsets Vehicles Water treatment

EHS & ERM, fire suppression, hazardous waste/toxins control, leak detection, remediation, sensors & controls Bio-filtration, cartridge/electronic, catalytic converters, clean coal, indoor air quality, sorbents Transportation efficiencies Measurement & testing, sensors, software/systems Carbon/emissions, water Water transport Filtration

Energy efficiency

Advanced packaging Biofuels Buildings Chemical Glass Lighting Monitoring & control Monitoring/compliance Other

Containers biodiesel Building automation, building envelope & insulation, energy saving windows, home automation, HVACR/R Coatings Chemical, electronic Ballasts & controls, smart lighting systems, solid state lighting Software Software/systems Appliances, consulting/facility management, consumer education, efficient motors, energy saving, monitoring, metering & control, sensors & controls Construction/fabrication, process efficiency Carbon/emissions Smart grid

Smart production Trading & offsets Transmission

Energy generation

Biofuels Geothermal Hydro/marine other

Wind

1st generation, algae biodiesel, biodiesel, biogas, biomass, cellulosic ethanol, grain ethanol Development, generation, hardware, power plants, wells current/tidal, ocean floor, wave Combined heat/power, electro textiles, hydrogen production, natural gas, on-site systems, renewable energy providers, turbines, waste heat, waste to energy Cells & modules, combined heat/power, concentrated PV, concentrated solar thermal, photovoltaics, systems, thin films Farms, gearboxes & components, renewable energy providers, turbines

Management Other

Power conservation, power monitoring & metering, power protection, power quality & testing, smart grid Waste to energy

Solar

Energy infrastructure

54

C. Criscuolo, C. Menon / Energy Policy 83 (2015) 38–56

Table A1 (continued ) Primary industry

Energy storage

Secondary industry

Tertiary industry

Transmission

Sensors & controls, smart grid

Advanced batteries

Advanced Pb-acid, charging & management, flow batteries, lithium-ion, nickle zinc, other technologies, thin film, ultra-capacitors Methanol, other technologies, PEM, solid oxide, systems integrators, zinc air flywheels, heat storage, hydrogen storage

Fuel cells Hybrid systems

Manufacturing/industrial Advanced packaging Monitoring & control Other Smart production

Containers, packing Automation, sensors, software, systems Energy saving companies Construction/fabrication, process efficiency, resource utilisation, toxin/waste minimisation

Materials

Bio Chemical Glass Nano Other

Cdvanced processes, biodegradable products, catalysts, sustainably harvested lumber Coatings, composites, polymers Chemical Catalysts & additives, detectors/sensors, gels & coatings, lubricants/films, membranes, powders adhesives, ceramics, efficient motors

Recycling & waste

Other Recycling Waste treatment

Combined heat/power, consulting/facility management, waste heat, waste to energy Chemicals, food, metals, mixed wastes, oils/lubricants, paper, plastic/rubber, services, sorting technologies, wood Biological breakdown, environmental disposal, gasification, hazmat destruction, plasma destruction

Transportation

Advanced batteries Fuels Logistics Other Structures Vehicles

Lithium-ion CNG/engine conversion, fuelling infrastructure, improved petroleum based Fleet tracking, mass transit, ride sharing, traffic monitoring software, transportation efficiencies Efficient motors, hydrogen production Light-weight Bicycles & scooters, electric & hybrids, rail transport, vehicle components/engines, water transport

Water & wastewater

Bio Cleanup/safety Glass Wastewater treatment Water conservation Water treatment

Catalysts Leak detection Chemical Biological, mechanical Recycling & management, sensors & controls, smart metering & control, water saving appliances Contaminate detection, desalination, filtration, filtration & purification, purification

Table A2 Summary statistics: distribution of the variables used in the regression analysis. Variable

Nr

Mean

s.d.

p25

p50

p75

Skewness

Number of deals Total amount of funding Regulation: price Regulation: quantity Public finance Fiscal incentive Gov. R&D exp. (% GDP) Value of total stock traded (% GDP) Nr of days to start a business GDP PPP p.c. Nr of seekers

621 621 621 621 621 621 621 621 621 621 621

82.75 4.94 0.59 0.65 0.66 0.91 0.23 118.86 16.02 33,313.02 2.18

320.70 11.51 0.49 0.48 0.48 0.28 0.08 83.86 13.30 11,106.56 7.76

1.15 1.00 0.00 0.00 0.00 1.00 0.18 54.01 6.00 31,139.27 0.00

10.00 2.00 1.00 1.00 1.00 1.00 0.22 96.71 13.00 35,033.42 0.00

41.00 4.00 1.00 1.00 1.00 1.00 0.30 169.67 22.00 38,623.22 1.00

9.39 6.72 0.36 0.61 0.66 2.97 0.23 1.08 1.51 0.75 8.94

Kurtosis 108.27 58.60 1.13 1.37 1.44 9.81 2.49 3.68 7.16 5.27 107.59

Table A3 Summary statistics: cross-country average of policy variables by year. Year/ Policy

Renewable portfolio standards or quotas (RPS)

Capital subsidies, consumer grants or rebates

Investment or other tax credits

Sales tax, energy tax, excise tax or VAT reduction

Energy production payments or tax credits

Public investments, loans or financing

Public competitive bidding

Feedin tariffs

Tradable re- Regulation quantity newable certificates

Sales tax

2005 2006 2007 2008 2009 2010 Total

0.33 0.32 0.32 0.32 0.31 0.30 0.32

0.83 0.86 0.84 0.84 0.96 0.96 0.88

0.72 0.59 0.52 0.56 0.62 0.70 0.61

0.39 0.45 0.48 0.44 0.77 0.87 0.58

0.17 0.18 0.16 0.16 0.19 0.22 0.18

0.33 0.41 0.44 0.40 0.62 0.65 0.48

0.22 0.27 0.24 0.20 0.35 0.43 0.29

0.50 0.55 0.52 0.60 0.73 0.83 0.63

0.28 0.27 0.24 0.24 0.23 0.30 0.26

0.39 0.45 0.48 0.44 0.77 0.87 0.58

0.56 0.55 0.48 0.48 0.58 0.70 0.55

C. Criscuolo, C. Menon / Energy Policy 83 (2015) 38–56

55

Table A4 Robustness tests: winsorization and square root of the dependent variables. Sample

(1)

(2)

(3)

(4)

(5)

(6)

All Estimation method

Negative binomial

Dependent variable Mean dep. var.

All deals 4.99

All deals winsorized 3.24

All deals Square root 1.67

Total amount 83.03

Total amount winsorized 35.68

Total amount Square root 5.34

Regulation price

0.394 (1.082) 3.744c (2.100) 2.267 (1.975) 4.179 (2.947) 35.51a (17.02) YES YES YES 621

0.154 (0.412) 1.483b (0.514) 0.787c (0.438) 1.519c (0.861) 13.54b (2.380) YES YES YES 621

0.0825 (0.124) 0.609b (0.170) 0.267c (0.142) 0.396 (0.253) 4.636b (0.776) YES YES YES 621

21.456a (8.368) 22.038b (7.639) 19.538a (9.310) 12.197 (13.050) 161.296b (51.913) YES YES YES 621

8.466b (2.866) 8.775b (3.041) 2.830 (3.504) 1.274 (4.875) 90.199b (17.256) YES YES YES 621

1.119 (0.319) 1.204b (0.335) 0.260 (0.354) 0.086 (0.622) 10.450b (1.907) YES YES YES 621

Regulation quantity Sale tax reductions Fiscal incentives Gov. R&D as % GDP Other controls Sector F.E. Year F.E. Observations

Tobit

Note: the winsorization threshold is set at 2% of the distribution (1% on each tail). The unit of observation is the year–industry–country combination (year 2005 2010, countries with at least 5 deals and cells with at least on deal or one company seeking investments). All regressions also include controls for: total stock traded as % GDP; days needed to start a business; GDP p.c. (t 1) in PPP. Estimates reported in the Poisson regressions are average marginal effects. Estimates reported in the Tobit regressions are marginal effects for the expected value of the dependent variable conditional on being uncensored, E(y|y40). Robust standard errors clustered at country–year level in parentheses. a b c

p o 0.05. p o0.01. po 0.1.

Number of deals

seed

first round

follow-on

private equity

800 700 600 500 400 300 200 100 0

Fig. A1. Number of deals by primary industry and investment type. Source: Authors' elaboration based on Cleantech Market Insight Database.

References Barradale, M.J., 2010. Impact of public policy uncertainty on renewable energy investment: Wind power and the production tax credit. Energy Policy 38 (12), 7698–7709. Baum, J.A., Silverman, B.S., 2004. Picking winners or building them? Alliance, intellectual, and human capital as selection criteria in venture financing and performance of biotechnology startups. J. Bus. Ventur. 19 (3), 411–436. Black, B., Gilson, R., 1998. Venture capital and the structure of capital markets: banks versus stock markets. J. Financ. Econ. 47. Bloomberg New Energy Finance, 2010. Crossing the Valley of Death: Solutions to the Next Generation Clean Energy Project Financing Gap. Working paper. Botta, E., Kozluk, T., 2015. Measuring Environmental Policy Stringency in OECD Countries – A Composite Index Approach. OECD Economics Department, Working Paper (Forthcoming).

Cleantech, 2010. A Future Of Innovation And Growth. Cleantech Group Reports, San Francisco. C., Criscuolo, Menon, C., Shestalova,V., 2013. Renewable Energy Policies and Crossborder Investment: Evidence from M&A in Solar and Wind Energy, OECD. Working paper. Cumming, D.J., Henriques, I., Sadorsky, P., 2013. Cleantech’Venture Capital Around the World. Available at SSRN 2323589. Wilson, Daniel, J., 2009. Beggar Thy neighbor? The in-state, out-of-state, and aggregate effects of R&D tax credits. Rev. Econ. Stat. 91 (2), 431–436. Dechezleprêtre, A., Glachant, M., 2012. Does foreign environmental policy influence domestic innovation? Evidence from the wind industry. Evidence from the Wind Industry. Environ. Resour. Econ., 1–23. Elton, E.J., Gruber, M.J., Brown, S.J., Goetzmann, W.N., 2009. Modern Portfolio Theory and Investment Analysis. John Wiley & Sons, New York. Ghosh, Shikhar, Ramana, Nanda, 2015. Venture Capital Investment in the Clean Energy Sector. Harvard Business School Working Paper, 11-020, also published

56

C. Criscuolo, C. Menon / Energy Policy 83 (2015) 38–56

as MIT Industrial Performance Center Working Paper 10-004. To appear in Richard Lester (edited), Essays on America's Energy Innovation Problem (working title), MIT Press (forthcoming). Greenstone, M., Hornbeck, R., Moretti, E., 2010. Identifying agglomeration spillovers: evidence from winners and losers of large plant openings. J. Polit. Econ. 118 (3), 536–598. Hockerts, K., Wüstenhagen, R., 2010. Greening Goliaths versus emerging Davids – theorizing about the role of incumbents and new entrants in sustainable entrepreneurship. J. Bus. Ventur. 25 (5), 481–492. Johnstone, N., Haščič, I., Popp, D., 2010. Renewable energy policies and technological innovation: evidence based on patent counts. Environ. Resour. Econ. 45 (1), 133–155. Kerr, W., Nanda, R., 2009. Financing Constraints and Entrepreneurship (No. w15498). National Bureau of Economic Research, Cambridge, MA. Kräussl, R., Krause, S., 2012. A comparative analysis of venture capital investment in the United States and Europe In: Cummings, D. (Ed.), The Oxford Handbook of Venture Capital. Oxford University Press, Oxford. Lerner, J., 2011. Venture capital and innovation in energy In: Henderson, R., Newell, R.G. (Eds.), Accelerating Innovation in Energy: Insights from Multiple Sectors, pp. 225–260, Available at: http://www.nber.org/books/hend09-1.

Messica, A., 2008. The valuation of cash-flowless high-risk ventures. J. Private Equity 11 (2), 43–44. Nanda, R., Rhodes-Kropf, M., 2010. Financing Risk and Bubbles of Innovation. Harvard Business School, Harvard, MA. OECD-EPAU, 2013. Renewable Energy Policy Dataset, version February 2013. Compiled by the OECD Environment Directorate's Empirical Policy Analysis Unit (Johnstone, N., Haščič, I., Cárdenas Rodríguez M., Duclert, T.) in collaboration with an ad hoc research consortium (Arnaud de la Tour, Gireesh Shrimali, Morgan Hervé-Mignucci, Thilo Grau, Emerson Reiter, Wenjuan Dong, Inês Azevedo, Nathaniel Horner, Joëlle Noailly, Roger Smeets, Kiran Sahdev, Sven Witthöft, Yunyeong Yang, Timon Dubbeling). Tyebjee, Tyzoon T., Bruno, Albert V., 1984. A model of venture capitalist investment activity. Manag. Sci. 30 (9), 1051–1066. Wooldridge, J., 2002. Econometric Analysis of Cross Section and Panel Data. MIT Press, Cambridge, MA, p. 2002. Wustenhagen, R., Teppo, T., 2006. Do venture capitalists really invest in good industries? Risk-return perceptions and path dependence in the emerging European energy VC market. Int. J.f Technol. Manag. 34 (1), 63–87.