Market-pull policies to promote renewable energy: a quantitative assessment of tendering implementation

Policymakers ideally select the support mechanism that better foments renewable energy production at the lowest cost to comply with international climate agreements. Currently, tendering is the fastest rising scheme. Yet a quantitative assessment of its performance in the literature is missing. We assess the effect of the introduction of auctions in accelerating the addition of renewable capacity through three econometric models: fixed-effects multivariate regression, statistical matching and synthetic control. The dataset includes 20 developed countries, spanning from 2004 to 2014, and both macroeconomic and policy drivers. Results show that tendering has the strongest effects to promote net renewable capacity comparing to other mechanisms like feed-in tariffs. Countries implementing tendering on average have a higher addition of net capacity of renewables in the order of 1000-2000 MW annually. The positive effect of tenders is clearer when analyzing with synthetic controls the case of Italy: while tendering enhances the deployment of renewables, policy instability jeopardizes the sustainability of tendering’s impact.


Introduction
Governments have been supporting the deployment of renewable energy (RE) as a way to tackle climate change and reinforce energy security. In fact, renewables are more expensive than fossil fuels in most cases because the negative externalities associated with the burn of fossil fuels largely remain absent from the prices. Therefore the government need to support RE to become competitive in the market. In 2017, renewable energy accounted for 14.3% of the total primary energy demand (IEA, 2018).
Even though this share is constantly increasing, it is doing so at a modest rate due to the persistent and significant growth in energy demand. A faster deployment rate is needed to limit global warming to 1.5ºC (IEA, 2018).
There is a growing literature on the most efficient policy, or the right "policy mix" (e.g., Del Rio and Bleda, 2012). While the qualitative literature (e.g., Gan et al., 2007) argues for a positive correlation between regulation and RE deployment, the quantitative studies (i.e. panel data analysis on country level) display more ambiguous results. Marques and Fuinhas (2012) argue for the effectiveness of policies in strengthening the use of renewables. Popp et al. (2011) do not find a significant impact of policies on wind generation, and Aguirre and Ibikunle (2014) find a negative relationship between fiscal/financial incentives and deployment of renewable energies.
When comparing the different instruments, Johnstone et al. (2010) find that feed-in tariffs (FiTs) are helpful for the development of less mature technologies, and Zhao et al. (2013) posits that FiTs are the only policy that encourages the development of renewable energy sources. Kilinc-Ata (2016) observes a positive relationship between RE deployment and FiTs or tendering. There is still a debate in the literature about the impact of the policies intended to support the investment in renewable capacity.
Auctions in particular are the object of a growing debate (e.g., Del Rio, 2017). On the one hand, proponents point to their advantages in terms of effectiveness (e.g., European Commission, 2014). On the other hand, opponents argue about their risk to undermine the benefits of existing supporting mechanisms namely in terms of the effects in competition (Butler and Neuhoff, 2008) or in cost reductions (Grau, 2014).
Recent assessments show that tendering may increase the effectiveness and efficiency of support depending on the circumstances (e.g., Winkler et al., 2018). Winkler et al. 3 (2018) has taken a first step in this purpose by performing a preliminary quantitative analysis on a limited number (eight) countries. However, there is a lack of a more systemic assessment of the tendering performance.
This paper aims to fill this gap by performing a cross-country quantitative analysis of the effect of both the introduction and implementation of auctions. It performs a comparative assessment of the policies supporting RE investment over time, with a focus on the introduction of tendering and the conditions of its implementation. For that we use a data set comprising a large and representative sample of 20 countries. The novelty of our approach consists on the study of this new dataset with the combination of three different econometric models: multivariate fixed effects regression, matching estimation, and synthetic controls. In particular, synthetic controls create a counterfactual group that controls for external factors, such as technology cost reductions, and cross-country differences, such as resource endowment, to provide more accurate effects of the policy instruments.
Our interest lies in the supporting policies aimed at directly increasing the share of energy generation from renewable energy sources. We therefore proceed to analyse the different factors affecting renewables deployment in order to identify which are the most effective measures the individual countries can adopt and which are the "external" factors that are not under the countries' control but still impact investors' decisions.
Since tendering has been overcoming feed-in tariffs as primary support mechanism not only in Europe (CEER, 2018) but also worldwide (IRENA, 2015), and given the previous calls for a systematic quantitative assessment of the tendering performance (e.g., Winkler et al., 2018), this paper aims to evaluate the most recent evidence on tendering as an instrument to promote the increase of renewable energy capacity.
The work is organized as follows. Section 2 reviews the literature on support mechanisms-particularly tendering-of renewable energy projects. Section 3 discusses the determinants of the investment in renewable energy in the literature. Section 4 presents the empirical research, namely the data and the econometric models which produce the results presented in Section 5. Section 6 analyzes the robustness of the models and the hypothesis. Finally, Section 7 concludes by discussing the main results and their implications. 4

Barriers specific to renewable energy projects
The investment in renewable energy falls into the broader category of infrastructure investment, and therefore is subject to the general barriers pertaining to this asset class. These are related to the features of infrastructure projects, such as high upfront capital requirements, long asset life, inelastic demand for services and prevalence of fixed costs. Apart from these generic risks of infrastructure investment, there are other specific issues to which renewables projects are subject. Table 1 provides an overview of the typical factors used by Standard & Poor's (2007) to determine the rating of infrastructure projects, along with the risks more specific to RE projects.  Jager et al., 2011). History has proved them right, as demonstrated by the sudden and sometimes retroactive policy changes in a handful of developed markets, including Italy and Spain, to which has followed a contraction in installations, mainly in the wind power sector (REN21, 2017). Policies can be a barrier also in countries where the regulatory system is designed around near-monopoly providers, preventing new players to enter into the market (Sen and Ganguli, 2017). 6 Lack of basic infrastructures and of local competences can delay investments.
These informational barriers are important factors, which need time to be addressed, and can make projects more expensive or even not economically viable. Especially in developing countries like India, the inadequacy of the transmission grid can be a huge problem since potential wind sites are often located in remote locations, far from the main consumption centres (Proparco, 2010). Furthermore, being RE site specific (unlike fossil fuels), the availability of detailed datasets (e.g. irradiation data for solar) is fundamental for the success of the projects (Sen and Ganguli, 2017).
Financial barriers arise from the cost and revenues profile of investments in RE capacity. Firstly, cost of capital can be significantly higher for RE projects, as they are perceived as riskier than traditional power generation projects, due mainly to: (i) higher Socio-cultural barriers may arise from the inadequate attention towards the climate change issue or from the social consequences of some projects (Sen and Ganguli, 2017). For example, social acceptance plays an important role, as recently shown from the ruling to overturn the installation of 14 wind turbines in Scout Moor (which would have created the biggest wind farm in England) due to its potential "significant adverse effect" on the landscape and views (Department for Communities and Local Government, 2017).

Support mechanisms to renewable energy investments
Domestic policies are necessary to overcome the barriers specific to the investments in RE capacity in order to meet the targets set by international agreements.
These policies are typically divided into two main categories: adaptation and mitigation (see Figure 1). Adaptation measures are directed to stimulate "practical steps to protect 7 countries and communities from the likely disruption and damage that will result from effects of climate change" (UNFCCC, 2015). They address the consequences of climate change and not the economic issues to respect the international agreements in order to avoid climate change, and as so, they are not part of our empirical analysis. On the other hand, mitigation measures directly address the economic and investment activity meant to reduce CO2 emissions, namely in electricity generation by increasing the share of RE. Mitigation measures can be further divided in technology-push and market-pull policies (Burer and Wustenhagen, 2009). Technology-push policies, such as government funded R&D, are deployed with the objective of increasing the technology "supply".
Market-pull policies, such as Feed-in Tariffs (FiTs), have the goal of increasing the "demand" of the same technologies (Rickerson et al., 2012). Both are necessary, since innovation is fundamental in providing new technologies and making existing technologies more marketable as they often cannot compete on the market without policy support. 8 Technology push aims to compensate the so-called knowledge market failures (Dechezlepretre and Popp, 2015) in the early stages of research and production of a technology (e.g., new types of bioreactors for producing bio-energy from biological waste treatment (Sepehri and Sarrafzadeh, 2018) ). At this point, the knowledge and experience gains have a public nature that can spill-over to other producers, preventing the innovator from getting the full benefits from them. The government can develop several instruments, such as tax incentives, grants and public funded R&D, to compensate for the insufficient incentives.
Market-Pull policies are important for technologies entering into a wider commercialization. They can themselves be divided into tariff-based instruments and quantity-based instruments (IRENA, 2015). Tariff-based instruments provide economic incentives for electricity production from RE sources, through subsidies or payments for the energy generated (e.g. FiTs, carbon taxes). Feed-in Tariffs (FiTs), for example, are policies designed to guarantee a certain price for a fixed amount of time for electricity generated from RE sources (Couture and Gagnon, 2009 A new category of policies emerged, the so-called hybrid instruments or auctionbased policies (commonly called "tendering"), that combine features of both tariffbased and quantity-based instruments in an attempt to remedy to the downsides of both schemes. Tendering refers to "a procurement mechanism by which renewable energy supply or capacity is competitively solicited from sellers, who offer bids at the lowest price they would be willing to accept" (

Impacts of tendering on renewable energy investments
To understand how tendering effectively promote the investment in RE capacity, by addressing the specific barriers discussed in Section 2.1., several arguments have been risen in the literature about its strengths and weaknesses.
10 Several strengths are attributed to tendering. First and foremost, by setting the quantity in advance, auctions can be more effective than traditional price instruments to address market failures (IRENA, 2015). Theoretically, well-designed auctions are also efficient by selecting the projects with the lowest production costs (Winkler et al., 2018).
Tendering fosters competition among different technologies that can lower RE prices over time, improving dynamic efficiency (Verbruggen & Lauber, 2012). Figure 3 shows that tendering has been at least as efficient as other instruments in bringing down levelized costs of energy (LCOE) for onshore wind, solar photovoltaic, offshore wind and concentrating solar power in the past decade. BNEF (2015) argues that a shift from FiTs or RPS schemes to tendering lowers on average RE project tariffs by 30%. In addition, tenders are a flexible instrument, which can adapt to different jurisdictions with disparate energy sector's structure and maturity. Tenders allow to overcome the information asymmetry between regulators and RE project developers. Often policymakers do not possess the knowledge required to set the support levels at an adequate level, leading to too high or too low tariffs (Del Rio and Linares, 2014).
Tendering instead, if carried out competitively and transparently, can be an effective mean of price discovery (Winkler et al., 2018). This was the case in Germany Some weaknesses have also been pointed to tendering, as well. Transaction costs might be too large compared to the potential profits, thus discouraging potential bidders. This could especially hit smaller size projects, showing that tendering is particularly suitable for large-scale RE projects despite the benefits of smaller-scale and distributed RE generation (Grau, 2014). 1 On the other hand, complexity requires higher structuring costs for the regulator. These costs are nevertheless mostly linked to the learning with the initial design of the policy, and can therefore be amortized over subsequent auctions. Moreover, aggressive bidding may lead to project delays and underperformance (Winkler et al., 2018). The introduction of auctions creates additional risks, such as of penalties for non-realization and delays, that can increase financial costs. This would explain why in some cases FiT was found more efficient and resulting in larger deployment (e.g. Grau, 2014

Determinants of the investment in renewable energy
New energy needs are often supplied to with traditional energy solutions, as they are typically cheaper and more rapidly implementable (Kilinc-Ata, 2016). Hence, a strong change in the incentive schemes is required to increase the use of renewable energy.
Economic and finance literature already provides a guidance on the factors that affect the investment decision in favour to renewable energy. This section systematizes the previous literature by identifying eight factors that could determine the investment in new capacity of renewable energy, discussing their role to foster or hinder the investment in cleaner energy generation. Following the discussion in section 2, we group the factors into three types, encompassing socioeconomic-related aspects, countryspecific features, and regarding institutional/political incentives.

Socioeconomic factors
Macroeconomic variables influence the behaviour of investors, especially in high capital intensive projects such as investments in renewable energies. Some of these socioeconomic factors contribute to foster the investment (e.g., welfare, population, strength of financial markets, cost of fossil energy sources (incumbent) ), others have a 13 negative effect (e.g., interest rates). We analyse next the impact of each one of these factors.

Welfare
Economic activity increases demand for energy and subsequently investment in the energy sector. Wealthier countries can also afford the costs of RE deployment and incentivize it with more ease. However, academic literature is torn on the welfare effect on renewable energy investment. Dong (2012) and Shrimali and Kneifel (2011) maintain that they are uncorrelated, while Aguirre and Ibikunle (2014) welfare variables such as GDP should have a positive or no effect in the RE investment.

Population and human capital
Population can be representative of more than sheer size, as its growth can point to energy needs not always captured in GDP growth. However, new energy needs could be supplied with both fossil fuels and RE. Following general investment theory (Baldacci et al., 2009), we still expect a population growth to positively affect RE deployment.
Human capital is essential for successful RE deployment. Education can also increase attention on the global warming issue. Thus the investment in RE should be positively correlated with the Human Development Index (HDI), yearly computed by the United Nations, which encompasses life expectancy, education and per capita income indicators.

Interest rates
Interest rates are theoretically the result of the equilibrium between demand and supply for funding. They have a negative effect in RE projects, which are usually financed through project finance structures comprising a higher level of debt compared with normal corporate financing (Eyraud et al., 2013). In addition, project finance loans have a tenor in the life of the project which can go up to 30 years. Thus they will be more 14 exposed to long-term interest rates rather than to the fluctuations of short-term interest rates (Eyraud et al., 2013).

Strength of financial markets
Investors need properly functioning capital markets throughout the lifetime of technologies (De Jager et al., 2011). Business angels provide the first equity when demonstrations start; Venture Capital (VC) and Private Equity (PE) help overcoming the "valley of death" in the transition to market; and corporate finance, project finance and public markets assist market expansion and the full deployment of the technology (Grubb, 2004). Furthermore, the credit granted through project finance depends also on  (Awerbuch and Sauter, 2006). Moreover, an increase in fossil fuels prices-the incumbent technology-decreases the relative price of RE generation, hence making it more competitive (Popp, 2001). Thus the increase in the fossil fuel prices should have a positive effect in the RE investment.

Country-specific factors
A second group of variables relates to the circumstances that determine the RE investment. The energy economics literature typically considers two country-specific factors: energy dependency; and renewable potential. This would mean that for the time frame considered we deem the potential to be time invariant. This issue has been debated by Carley (2009), who esteems the potential of renewable sources to be invariant for 9 years. Thus renewable potential should impact little on the investment in RE in the short-run.

Political factors
Finally, the implementation of incentive mechanisms can spur the investment in RE as discussed in the energy economics and policy literature. This concerns in particular the adoption of market-pull policies. Such policies are important to accelerate the growth of emergent technologies (Burer and Wustenhagen, 2009), comprising: tariff-based instruments (e.g., feed-in-tariffs (FiTs), carbon taxes (C_T) ); quantitative-based instruments (e.g., renewable portfolio standards (RPSs); tradable renewable energy certificates (RECs), cap and trade schemes (CaT) ); and auction-based policies or "tendering". A comparison between the different types of support mechanisms is provided in Section 2.2. Recent studies already suggest the superiority of the change to auctions in specific conditions of context (Winkler, 2018) and stability (Tiedemann et al,16 2016). Thus the investment in RE should increase with the existence of incentive mechanisms.

Synthesis and hypothesis
The economics, finance and management literature suggest a number of determinants for the investment in renewable energies that can be grouped in socioeconomic, country-specific and institutional factors. Given that this paper addresses the benefits with the change to an auction-based instrument, we focus on the factors that drive net capacity additions including the differences between countries in terms of resource endowment and environment stability, without considering specific investor and project characteristics. These drivers of the investment in RE will serve to test empirically the significance of the following explanations: H1: The investment in RE depends more on socioeconomic and country-specific factors.
H2: The investment in RE depends more on the policy instruments.
H3: The investment in RE increase with the change to auction-based mechanisms.
The last hypothesis (H3) is the central one of this study. The other hypothesis H1 and H2 are mutual exclusive and serve to clearly distinguish the most important set of drivers that increase the investment in RE. Of course, drivers from different groups of factors can determine the investment. But this strategy helps to more clearly isolate (and compare) the positive effects that can be attributed to the change of the policy (to tendering) from the impacts of contextual variables.
Data sources, the selection of variables and the empirical approach are explained in the next section.

Data
To assess the effect of the introduction of tendering mechanism on renewables' deployment, we collect data for a sample of developed countries (all OECD members).  Table 2 summarizes the different variables used for our analysis, providing detailed information for each one of them. Tables 3-4 show the descriptive statistics for all the variables and the correlation matrix for the independent variables, respectively.
2 These are the OECD countries represented in the top 70 of the List of countries by electricity production from renewable sources (https://en.wikipedia.org/wiki/List_of_countries_by_electricity_production_from_renewable_sources last accessed in 4/7/2019). We removed the following countries from the sample because of their specificities: Norway (hydropower legacy, which represents 99% of electricity production), Chile (late arrival to OECD in 2010 in the middle of the surveyed period); Poland (small share of renewables in the mix 13,7% in 2016); Greece (financial crisis impact).

Models
To assess the effect of the introduction of auctions and of the conditions of its implementation, we take a novel approach based on the combination of econometrical analysis. The first step is a multivariate regression with the objective of testing the relationship between tendering and the addition of renewable energy capacity, taking into account other policy instruments and external factors. In order to detect a possible causality between tendering and renewables' deployment, we further implement two 21 different estimation techniques: statistical matching and synthetic control (more details next). These two models also test for country-specific conditions while controlling for external factors (e.g., technology cost decrease over time).

Multivariate Regression
We perform a multivariate regression with fixed effects in order to control for unobserved country heterogeneity (e.g., resource endowment). Our general specification is the following: where is the amount of net capacity of renewable energy added in year t in country i, is the intercept,, Xi is a matrix of the explanatory variables described in Table 2, β is the coefficient matrix to be estimated, and , is the error term. We are therefore testing the following model: Where: and f is a linear function.

Matching Estimation
In order to determine a possible causality between tendering and renewables' deployment, i.e., a causal treatment effect, we apply matching estimation. This method permits to control for external developments such as technology cost reductions. The ATE represents the difference in expected outcomes after participation and nonparticipation. However, the ATE can sometimes be misleading as it includes the effect on individuals for whom the treatment was not designed. The solution to this is the ATT, which represents the difference in expected outcome with and without treatment for those who actually were subject to the treatment. We will proceed to estimate both parameters in order to assess the effect of tendering on net renewable capacity addition, and we will do it using two of the most popular matching techniques: the propensity score matching (PSM) and the nearest neighbour matching (NNM).
PSM relies on the idea of solving the curse of dimensionality by using the relevant characteristics X to compute the probability that an individual will enrol in the treatment (Rubin and Rosembaum, 1983); this value p(X), such that: is the so-called propensity score. By attributing a propensity score to each individual, we obtain the distribution of p(X) which helps us to individuate the common support area, i.e. the overlap between the PS distribution in the two groups. The sample will then be restricted to the individuals that fall in the common support area.
Using the propensity score we can then compute our average treatment effect ( We can see that the propensity score acts weighting the mean difference in outcomes over the common support, giving a higher weight to individuals with a higher p(X).
NNM works by imputing the missing potential outcome Yi(1) or Yi(0) by using average outcomes for individuals with "similar" values for the covariates that have received the other treatment level (Abadie and Imbens, 2002): hence, since both treated and control units are matched, matching is carried out with replacement, so that every unit can be used as match more than once. "Similarity" is accounted for by using a weighted function of the covariates for each individual. Formally, the distance between xiϵX and xjϵX, two vectors of covariates for individuals i and j, is parametrized by the vector norm: where S is a given symmetric, positive definite matrix. By default, Stata set S to be the Mahalanobis scaling matrix, in which weights are based on the inverse of the covariates' variance-covariance matrix. We can then rank each individual in terms of "distance" based on the norm, and call jm(i) the index of the unit that is the m th closest to individual i. We then set the desired number of matches M (in our analysis we will use M=1) and denote with JM(i) the set of indices for the first M matches for unit i: At this point, we can obtain an estimate for all the unobserved potential outcomes, by using the following estimates: Using these estimates, we can compute our ATE and ATT as: average of the untreated individuals from the control group. We suppose (0) is given by a factor model, such that: where is an unknown common factor with constant factor loadings across units, is a (r x 1) vector of observed covariates (not affected by the treatment), is a (1 x r) vector of unknown parameters, is a (1 x F) vector of unobserved common factors, is an (F x 1) vector of unknown factor loadings, and the error terms are unobserved transitory shocks at the region level with zero mean.
We also consider a (J x 1) vector of weights = ( , … , )′, with J being the number of untreated individuals, such that ≥ 0 for = 1, … , and + ⋯ + = 1. Each particular value of the vector W represents a potential synthetic control, which means a weighted average of untreated individuals. The value of the outcome variable for each synthetic control indexed by W is: In a similar fashion to the method employed with the NNM, we look for the W* that minimizes the distance: where S is a (k x k) symmetric and positive definite matrix, Xi is a (k x 1) vector of pretreatment characteristics for the treated individual and XJ is a (k x J) matrix containing the same variables for the untreated individuals; the characteristics taken into account can include covariates as well as pre-treatment values for the outcome of interest Y. The need for a sufficiently long pre-treatment and post-treatment period compels us to focus on countries which implemented tendering at a mid or late stage of our time frame: therefore, we apply SC to Italy (which implemented tendering in 2010).
The pool of donors for synthetic Italy is composed by the ten countries in our sample which did not implement tendering in the time frame considered, which are: Austria, Finland, Germany, Korea, Netherlands, New Zealand, Spain, Sweden, Switzerland and United Kingdom.

27
The findings of the multivariate regression are presented in Table 5. We report both the global regression including all the variables at the same time, and two partial regressions that isolate Political Factors from Socioeconomic Factors and Country-Specific Factors.

29
The global regression explains a high share (42%) of the variation in net RE capacity during the period surveyed. Among the socioeconomic factors, only GDP is positively associated with addition of renewable energy capacity in both regressions (while population is only moderately positively associated with dependent variable in the global regression), confirming that absolute size (in economic or demographic sense) is an important factor in renewables deployment. This is consistent with previous works (e.g. Marques et al, 2010) which used geographical size as proxy for resource potential.
Contrary to the initial expectations, the coefficients of HDI, FDI, and NBP are not significant and negative. Oil price appears to have a positive effect on RE capacity addition (although the statistical significance is moderate), as expected: higher oil prices turn renewable electricity more competitive relative to other modes of power generation (oil, natural gas, etc.), attracting investments in renewable energy capacity.
Energy dependency appears to be significant in both model specifications, suggesting that the quest for energy security leads to renewables deployment. Finally, looking at the last group of variables on policies, which is also the most relevant for our study, we can see that both carbon pricing schemes are significant (even though cap and trade systems only in the partial regression), while FiTs, RPS and RECs do not show significance in our model. However, we observe that tendering has strong statistical significance (1% level). At this stage, we can support the view that tendering mechanisms are on average associated with the increase in renewables' investments but we are unable to claimon the sole basis of this result -for a causal relationship. The coefficient of the Tendering dummy implies that having a national tendering scheme in place is associated with about 1200-2200 MW of renewables installed capacity per country annually.
The findings from the matching estimation are reported in Table 6. We show the results of matching on each category of covariates with both PSM and NNM.
30 Using the matching estimation, the positive impact of tendering on addition in net capacity of renewables is generally confirmed. Especially encouraging is the result by matching with the policy covariates, that is strongly confirmed and is sign that given a similar set of policies, tendering implementation has a considerable positive effect on the addition of net capacity of renewables.
Finally, regarding the implementation of the synthetic control method, and since large sample inferential techniques are not well suited, placebo tests are used instead by applying the SC method to every potential control in the sample (Abadie et al. 2010;2015). Therefore, we consider whether the prediction error -when considering the actual treated country -is "unusually" large relative to the distribution of prediction errors for the countries in the donor pool. This analysis rejects the null hypothesis when the post-intervention mean squared prediction error (MSPE) for the SC estimate is greater than the post-intervention MSPE for the placebo estimates (Abadie et al., 2015).  The introduction of tendering appears to have had a positive bringing an additional 6000 MW circa installed capacity in Italy (Table 7). However, the implementation of tendering lead to contrasting results in the two cases. Figure 4 shows the added capacity for synthetic Italy overcomes the value for actual Italy in 2014.

Italy Weights
Germany ( (the so-called "conto energia") (Tiedemann et al., 2016). These policy changes have undermined investors' confidence in the stability of the regulatory framework (Mahalingam and Reiner, 2016

Conclusion and policy implications
Tendering is the fastest rising mechanism to support the increase of electricity generation from renewable sources in the past two decades. So far the literature gives an unclear picture about the advantages of tendering in comparison with other instruments. The current debate is controversial with some authors assessing the effects of switching to auctions in terms of improved effectiveness, while others considering it would rise financial risks and costs. Recent assessments show that tendering may increase the effectiveness and efficiency of support depending on the circumstances, however no systematic cross-country analysis has yet been done. This paper performs a comparative analysis of the policies supporting RE investment over time, providing in particular a quantitative assessment for the effect of the introduction of tendering and of the conditions of its implementation.
An innovative econometric approach combined three different models (multivariate fixed effects regression, matching estimation, and synthetic controls) to evaluate the data from a representative sample of countries. Synthetic controls in 36 particular create a counterfactual group that controls for external factors (e.g., technology cost reductions) and cross-country differences (e.g., resource endowment).
Our empirical analysis confirms the advantages of tendering over other support mechanisms in promoting the investment in cleaner, renewable capacity. The model with fixed effects shows evidence that tendering has a positive effect in net additions of RE capacity. This effect is significant and higher than of the other instruments like feedin-tariffs. The matching estimator confirmed the superiority of tendering over other mechanisms on the countries surveyed, during the period under analysis. The results also indicate good performance of auctions to deal with financial and socio-cultural barriers to RE investments. Finally, synthetic controls revealed benefits of switching to tendering but these are contingent on other factors such as policy stability. The three approaches lead to estimates of additional renewable capacity installed in the order of 1000-2000 MW per country per year.
The analyses still have some limitations. Data was only available for the period up to 2014. However, this already includes the beginning of recovery from the financial crisis (e.g., Italy). In the future it would be possible to perform analysis including more countries, over a longer period. On the other hand, we were only able to assess the average effects of the mechanisms in the net addition of RE capacity. In practice, tendering performance might be different by technology. These limitations are however unlikely to affect the meaningfulness and the direction of the results.
Two main policy implications derive from the results. Firstly, as shown by Italy's example, tendering is not a panacea for the transition to a low-carbon economy, as it can be effective only if investors trust the regulatory system in place. Here lies a lesson for policymakers, who have to design a portfolio of support schemes that is attractive for investors as well as sustainable in the long run: infrastructure investors have a time horizon that is lengthier than any legislature and lose confidence quickly in the political framework of a country. Secondly, tendering alone cannot substitute all the other RE support policies: they are specifically devised for the creation of large RE generation projects. This might be a challenge for new technologies in the early years of formation for which the size of the projects is typically smaller. In these cases, policies should gradually move from price-based mechanisms (e.g., FiT) to auctions as the new