The design of technical requirements in public solar auctions: Evidence from India

Solar technology diffuses across the globe as countries transition from fossil to renewable energy. Little solarspecific experience and capacity in newly adopting countries can result in technical failures and lower solar plant performance. This contributes to making the investment in solar plants in newcomer countries risky and may undermine political targets of solar energy deployment. One solution suggested by international organizations is for policymakers in adopting countries to include international quality standards as technical requirements in public auctions. Here, we develop a conceptual framework on how international quality standards could help build a solar sector. As a case study, we analyze the explanatory factors of technical requirements in 100 public auctions of utility-scale solar photovoltaic plants carried out in India between 2013 and 2019. Our findings suggest that more international quality standards are required in auctions in which the government rather than a private actor ultimately carries the commercial risk. On the other hand, local content requirements and attracting foreign investors do not correlate with technical requirements. We argue that using minimal quality standards is unlikely to promote local technological catch-up or attract long-term foreign investments but transfers the techno-commercial risk from the government to the private sector.


Introduction
Solar photovoltaic (PV) technology is currently diffusing across the globe [1], with 106 countries having held renewable energy auctions by the end of 2019 to foster the transition from fossil to renewable energy [2]. However, evidence suggests that solar PV projects may underperform in many low-and middle-income countries, including India. For example, it was reported that 30% of nearly 100 analyzed projects in different countries indicate severe defects [3]. Furthermore, a study of the National Metrology Institute of Germany (PTB) and PI Berlin, a global quality assurance provider, found in all six examined grid-connected solar PV plants in India "sporadic or systemic failures with impact on performance" [4]. Thus, if solar PV projects fail to perform as expected, investors may not recoup their money, undermining future investments and, in turn, political goals for renewable energy deployment [5][6][7].
One solution recommended by international organizations such as the International Renewable Energy Agency (IRENA) and PTB is that policymakers in adopting countries include international quality standards 1 as technical requirements in the design of public auctions [4,7,8]. For instance, it was suggested "that strict technical requirements [ …] should be part of tender requirements [ …]. [ …] Lax tender requirements should be avoided as they give comfort to module suppliers and installation companies as no legal framework is forcing them to provide evidence of long-term durable products and failure free installation works" [4, p.6]. Similarly, it was proposed that "the auctioneer can also define other technological requirements, [ ….], such as specifications on the equipment used. Imposing equipment specifications can help ensure that the sector will be developed using state-of-the-art technology and appropriate quality of components" [8, p.32]. Hence, compliance on the part of solar PV plant developers with specific quality standards is expected to reduce quality defects and technical failures. In turn, investor confidence and trust will increase, as uncertainty about generated electricity decreases [7].
However, the existing literature on renewable energy auctions offers few insights related to the optimal integration of technical requirements into auction design. The effectiveness of renewable energy auctions is widely found to depend on their design, which varies across countries [8][9][10][11][12][13][14][15][16][17]. Yet, very few authors [17,18] have examined how specific auction design elements condition their effectiveness. Publications have rather focused on comparing auctions with feed-in tariffs [6,11,19,20]. As a result, there is a knowledge gap on designing renewable energy auctions optimally, and on how different design elements (e.g. auction mechanism, qualification criteria and bid evaluation, legal/contractual clauses, local content requirements etc.) shape auction outcomes.
To close this gap and help policymakers choose technical requirements for solar PV auctions in a more strategic way, we develop a conceptual framework on how technical requirements could help build up a domestic solar sector. Based on the framework, we derive three hypotheses on how policymakers can choose to use international quality standards as technical requirements in the design of public auctions for solar PV plants. The three hypotheses are empirically verified in a regression analysis of 100 public auctions for industrial, utility-scale solar PV projects with a total capacity of 51 GW carried out in India between 2013 and 2019. We also examine whether other potential risks such as exclusion of bidders have materialized upon including international quality standards.
Interestingly, "most African countries that held renewable energy auctions (dominated by solar) in 2017-2018 did so for the first time" [2, p.11]. Policymakers in these countries need to choose how to introduce technical requirements and their design choice likely affects the development of a national solar sector. India graduated from virtually zero solar energy to the fifth largest photovoltaic market worldwide within less than a decade. The Indian experience with designing technical requirements to mitigate underperforming solar PV plants can help newly adopting countries avoid similar mistakes.
In the following, section 2 illustrates the contextual background and relevant literature. Section 3 outlines the conceptual framework and develops our three hypotheses on the use of international quality standards by Indian policymakers. Section 4 presents and discusses the results of the empirical analysis. Finally, section 5 concludes with policy implications and a research outlook.

Public auctions as an instrument for the global energy transition
In the last two decades governments have increasingly used public auctions rather than other policy tools to guide the expansion of solar and wind energy [2,[11][12][13]21]. Several recent publications emphasized the ongoing need for further research on the recent experience with public auctions for renewable energy [6,9,11,14].
Public auctions for solar PV plants work as follows. The responsible government institution allocates the right to build a renewable energy plant to the private company with the best bid. In a reverse auction, as conducted in India, selection is based on price alone and the bidder who offers the lowest price wins. The most general contractual agreement for large-scale utility solar PV projects is Build-Own-Operate (BOO) associated with a power-purchasing agreement (PPA). The PPA stipulates that the government, mostly represented by a public utility, buys the produced electricity for the following 25 years. Another contractual format is Engineering-Procurement-Construction (EPC), which is less frequent but crucial to our analysis. EPC contracts are associated with a specified Operation and Maintenance (O&M) period, after which ownership and liability for the solar PV plant are transferred to the auctioneer.
The prevalent perception is that policymakers have increasingly adopted public auctions to create a domestic market, reduce the procurement price of electricity, avoid excessive subsidy levels and be able to better steer and plan the expansion and integration of renewables into national electricity grids (e.g. Refs. [2,6,14,15]). Several authors have examined whether auctions represent a more effective policy tool than the previously dominant options of bilateral negotiations with project developers, quota and feed-in tariffs [6,11,19,20]. There seems to be a consensus that auctions, if well designed, are a powerful, cost-effective policy tool to create a domestic market and provide a necessary level of investment certainty. Reverse public auctions have led to fierce price competition and historically low electricity tariffs around the globe [22]. For example, the highest winning bids in Indian solar auctions were on average 36 percent lower than the feed-in tariff [23]. Although this is less emphasized, the use of public auctions also led to a transformation of ownership structures in the electricity sector, bringing about a transition from public ownership to increased private ownership [24]. While the EU and US markets are already dominated by private power producers, with a 75% and 80% market share respectively, this share is much lower in developing countries and emerging markets like India and China -around 30 and 40%, respectively [24].
Yet, this consensus on the effectiveness of renewable energy auctions has emerged in the context of a simultaneous decrease in international prices for solar PV modules, the main cost factor of solar PV plants, and the expansion of public auctions in several countries (proof-by-association). Only few authors were cautious enough to underline that there is no rigorous, counterfactual-based, causal identification of the effect of public auctions on electricity price [15]. The concerns or potential drawbacks of public auctions for the development of renewable energy mentioned in the literature are (1) low(er) project realization rates, as a consequence of adventurous or strategic underbidding, (2) a lack or excess of competition, which may lead to market consolidation over time in both cases, and, as a result, (3) insufficient incentives for dynamic learning-by-doing and innovation [10,15,17,[25][26][27][28]. Lower project realization rates have received most scholarly attention, with most studies focusing on countries in the global north [10,15,16,26,[28][29][30][31], while the dynamic effect of auctions on competition/market consolidation [11,32] and local learning-by-doing and innovation have been understudied (excepting [18]). For instance, prices of initially  -35% below competitive tariffs were found in the first reverse auctions in India and were attributed to underbidding by inexperienced players [5,26]. Several authors have noted that auctions are not a panacea for successful renewable energy deployment given that different countries had varying success rates [6,8,15,25,33]. One conclusion may be that it is crucial to design auctions in line with a country's local institutional and industrial context. Auction design elements, such as technical requirements, may be one powerful mechanism for policymakers to improve the effectiveness of auctions, i.e. reduce or prevent low project realization rates and promote local learning-by-doing (see section 3.2). Several authors noted that countries choose slightly differing auction designs, which makes it difficult to conduct cross-country comparisons, but also suggests that policymakers adjust the auction design to the specific institutional and market context [8][9][10][11]19]. Unfortunately, there are only few dedicated studies, mostly based on one or at most a handful of countries for which several auctions at best are observed over time [9,[13][14][15][16]29]. Hence, there is only a limited understanding of the incentives and rationale driving the auction design choices made by policymakers (exceptions are [17,25]). This shortcoming has first been acknowledged in Refs. [8,10], which both provide theoretical frameworks for auction design. Quantitative evidence of the specific effect of auction design elements on the auction outcome was first given in Refs. [17,18]. By comparing the project realization rates for auctions in different countries and years in a regression framework, it was found that pre-qualification measures and penalties are positively correlated with realization rates [17]. Moreover, local content requirements were shown to increase the bidding price in Indian solar PV auctions [18]. Thus, it is a promising avenue for future research to investigate how specific auction design elements condition both the opportunities and the shortcomings associated with renewable energy auctions. This paper contributes to closing this gap by investigating the factors motivating the use of technical requirements in Indian solar auctions.

Solar energy development in India: National Solar Mission
India's domestic solar power generation before 2009 was negligible and consisted only of a few mini-grid projects [19,34]. In 2009, the Government launched the National Solar Mission. Its objective is to install 100 GW solar PV capacity by 2022 and build up local, globally competitive manufacturing. While about 20 export-oriented solar module manufacturers existed already at the outset of the Solar Mission, there was virtually no large-scale solar PV generation in India [34,35]. No cell manufacturing existed yet.
The Ministry of New and Renewable Energy (MNRE) through its executive agency -the Solar Energy Corporation India (SECI) -has managed to nurture and expand solar energy production. Yet, it remains a challenge to establish globally competitive solar component manufacturing at a large scale in India. Initially, MNRE relied on feed-in tariffs and other policy tools (i.e. purchasing obligations for India's federal governments). Reverse public auctions and their eligibility criteria have become the major policy tool to guide the development of solar electricity production and component manufacturing [18,26,35,36]. Among the eligibility criteria used by SECI are local content requirements, restrictions on foreign bidders, and technical requirements.
While the initial approach to technical requirements can be characterized as "laissez-faire", there have been more recent efforts to foster quality assurance and quality upgrading. In 2017, MNRE published the national "Lab Policy for Testing, Standardisation and Certification for Renewable Energy Sector" [37]. The aim is to upgrade the performance quality of solar PV components and turn testing laboratories into centers of global excellence. The policy also requires all components, in particular PV modules, to be tested for reliability and performance in India again. However, execution was postponed on several occasions due to feedback from manufacturers and testing laboratories about insufficient local testing capacity. In 2020, MNRE created a "Renewable Energy Standardisation Cell", which shall push for further quality upgrading and public policies.

Technical requirements in public auctions of solar PV plants as a policy tool for sustainable, industrial development
Green industrial development at the global frontier focuses on inventing and patenting technologies that are new to the world. In developing and emerging economies, green industrial policy deals mostly with adapting, disseminating and finally catching up with the global technology frontier [38,39]. International quality standards are one mechanism to disseminate state-of-the-art technology and production methods, and foster investment in product testing for trial-and-error innovation in developing and emerging economies [40,41]. Fig. 1 illustrates how governments can require compliance with international quality standards in public auctions for solar PV plants. Firms bidding to develop solar PV plants will need to choose components in compliance with specific technical requirements. Firms have an incentive to invest in upgrading their production methods and product quality accordingly, given that government procurement of electricity creates business opportunities [40,[42][43][44][45].
The benefits to local industrial development are twofold [41,[46][47][48][49]. Firstly, international quality standards promote productivity (see section 3.2.). The productivity increase can come from adoption of international best practices and from product tests in accredited laboratories as defined in international quality standards. Product tests are an opportunity for domestic firms to trial-and-error test their products and compare their quality against international benchmarks. For example, less than a handful of Indian companies conducted in-house research and development in 2012 [18,34]. Secondly, quality standards should have a strong positive signaling effect (see section 3.3). Buyers (investors) have incomplete information about the quality of the sellers' production processes, and thus are either unwilling to buy or only willing to pay less [50]. Incomplete information is particularly problematic when a country enters solar PV energy production, given that there is little local data about potential electricity yields and technical component resilience (i.e. in tropical climate conditions) [51]. Scarce information is also likely to be often paired with investors' mistrust against the general investment climate and institutional context of most developing and emerging countries. Hence, adherence to international quality standards is likely to have a strong, positive signaling effect to investors or banks.
Finally, there are also some risks related to the use of (international) quality standards in public auctions to promote green industrial development. Firstly, quality standards exclude non-compliant competitors. This may be a concern for governments in developing and emerging countries, particularly in smaller countries that may only have a few competitors, especially during the initial transition to solar PV [49,52]. Secondly, international quality standards may notably hurt local firms if there are insufficient, costly, or slow testing services available, while idiosyncratic national standards may scare away international competitors [49,53]. Limited local testing capabilities are a common problem in developing and emerging countries, given that testing equipment has high capital costs and requires specialized personnel [7]. Idiosyncratic national quality standards, by contrast, create additional costs for international competitors, which may prevent them from entering a market, particularly if the market is very small. Thirdly, quality standards may raise the bidding price, which in turn raises the cost of electricity for the government. While this is in principle true, engineering studies suggest that the benefits of stringent quality assurance, at least at the project level, outweigh the additional costs of quality [4,54]. Ultimately, a crucial concern anchored in the industrial policy literature is that policymakers may not possess the capabilities to determine which standards should be selected at what stage of industrial development [55]. In conclusion, too stringent quality standards could force nascent players to exit the market, while too lenient quality standards may entail that there is no local industrial learning along with reluctance from buyers to invest in solar PV projects.
In the following, we formulate three hypotheses related to the use of international quality standards in public auctions for solar PV plants in India.

International quality standards as technical requirements: a performance benchmark to complement local content requirements?
The Indian government aspires to expand domestic solar PV electricity production and create globally competitive domestic manufacturing. Therefore, local content requirements (LCR) were introduced in India's public auctions for solar PV plants. LCRs obligate independent power producers and project developers to source solar PV modules and/or cells in India [18,35].
LCRs often increase procurement costs, as they are typically introduced when local prices are higher than world market prices [18]. The rationale for LCR is that local producers are temporarily protected from foreign competition, can gain hands-on experience with a certain technology ("learning-by-doing"), and catch up with international players [49,56,57]. Firms learn and improve their efficiency/productivity when beginning to manufacture, in this case solar PV modules and cells, but learning rates decrease over time [56,58]. Once learning-by-doing enables local producers to lower their costs to match world market prices, LCRs are no longer necessary and can be withdrawn. A common criticism towards LCR is that they often fail to provide sufficient incentives for local beneficiaries to improve performance and reduce prices, e.g. because of rent-seeking and government failure, and thus fail to transform nascent industries into competitive ones [59][60][61].
We are interested, however, in the interplay between LCR and quality standards. Innovation through rule-setting has proven effective, for example in environmental regulation [62] and technological upgrading due to foreign restrictions [40]. In a similar vein, technical requirements can be combined with LCR to provide performance incentives. Governments can thus require bidders to source components locally and comply with national or international quality standards. Hence, performance standards could be used to drive domestic producers to engage in learning-by-doing rather than import intermediation [59]. For instance, Denmark's aggressive promotion of standards and quality certification helped the country become and remain a world leader in wind turbines [49]. Similarly, it was argued that the incentives and protection provided by industrial policies ("carrots") need to be complemented through a "stick" to guarantee that beneficiary firms also engage in learning, quality upgrading and productivity increases [55].

Hypothesis 1.
Policymakers use technical requirements such as international quality standards to push local firms to catch up with the global solar PV production frontier.

International quality standards as technical requirements: a signaling tool to de-risk investors?
Globally, the majority of investments in renewable energy since 2016 have been taking place in developing and emerging economies [8,14,63]. In India, the government orchestrates the expansion of solar PV electricity. Yet, it is the large Indian industrial conglomerates, Indian banks, and non-banking institutions such as international venture capital and hedge funds that develop and finance utility-scale PV plants [22,[64][65][66].
We apply insights from economic theory on quality uncertainty and adapt it to the context of solar PV auctions [46,47,50,[67][68][69]. As mentioned in section 3.1, compliance with international quality standards has been shown to overcome the buyer's insufficient information or lack of trust, for example in international trade [47,67,69] and foreign direct investment [48,68]. International quality standards create trust, as they signal use of state-of-the-art production processes. Signaling applies in principle to both domestic and international investors, but information asymmetry and lack of trust are especially relevant for foreign investors [46,47,68,69]. In the following, we outline how international quality standards introduced as technical requirements in public auctions could help remedy the lack of information and trust in solar PV performance.
Investments in solar PV projects follow the risk-return profile of the respective project [53]. Technical risk is significant in developing countries with low solar-specific industrial capacity, experience and no historical performance data for credible yield estimates [30,51,[70][71][72]. Even in the EU and the US, investors rate technology performance standards as third-best among 12 market-pull policies [73]. A solar PV plant is an attractive asset if it generates the expected energy yield and revenue over a period of up to 25 years. Payback and break-even times increase when components malfunction and underperform. PV modules are the key component of solar PV plants and make up 50% of the costs [7]. In fact, independent assessments of modules have shown they often underperform relative to manufacturer indications at up to 3% already before operation [7] and degrade over time at 0.4% per year in the US and Europe and at about 0.6% up to 5% in PV plants in India [74][75][76]. Modules can also underperform from day one if they are transported or installed incorrectly, as has been often observed in India [76]. Other components can have similar impacts on plant performance (techno-commercial overviews are available in Refs. [51,54]). Uncertainty is particularly high at the time of market creation and in environments with low information about input parameters, such as irradiation, temperature, component degradation, component peak performance and resistance to environmental conditions, construction and installation capacities [51,54,72,77]. A mixture of low awareness and lack of experience complements scarce publicly available information in countries that kick-start solar PV [51,78].
Based on the above considerations, one can view more stringent technical requirements in public auctions as a quality signal and derisking instrument to attract foreign investment [66,79]. More stringent international quality standards try to tackle the underlying risk of technical failure and related uncertainty about electricity yields. In addition, international quality standards function as a proxy, which signals high solar PV plant performance in a situation where no prior information exists, information is expensive for international investors, and trust in the institutional framework is low.

Hypothesis 2.
Policymakers introduce international quality standards as technical requirements in public auctions to attract foreign investment.

International quality standards as technical requirements: a tool to reduce risk exposure of the public sector?
The bathtub curve in Fig. 2 illustrates that technical failures in solar PV plants occur mainly at the beginning (or even before operation starts, e.g. malfunctional modules due to microcracks from incorrect transportation) and at the end of their lifetime due to wear and tear [7].
As in a principal-agent model, the Indian government represented by SECI (principal) delegates the task to build and operate a solar power plant to the private company (agent) that offers the lowest bid price in a public auction. The agent enters into an EPC plus O&M or BOO plus PPA contract with the principal. There is a key difference between these two types of contracts. Under EPC plus O&M contracts, the project developer (agent) constructs, operates and maintains the solar PV plants only for a limited number of years and then transfers ownership to SECI (principal). Under BOO contracts, the project developer (agent) is responsible for everything from construction to operation and assumes ultimate ownership. The allocation of technical risk over time thus differs significantly in these two contract arrangements. Fierce price competition in reverse auctions forces project developers to bid close to their expected electricity costs per kilowatt hour. Low, winning bids can turn into a curse if project developers are too optimistic, e.g. about the O&M costs, particularly if they choose low-quality components to be more price-competitive, or if interest rates change etc.
Since SECI ultimately assumes the risk of underperformance of the solar plant over the lifecycle, EPCs have an incentive to ensure the shortterm performance and quality of solar plants. By contrast, EPCs have no (economic) incentive to focus on quality beyond their O&M period. In a BOO contract, remuneration is fully performance-based and the owner has a strong incentive to maximize the output over the lifetime of the plant, as it may take several years to amortize costs and generate profits. In line with the principal-agent theory, SECI uses upfront payments, sanctions, and performance targets to monitor and guarantee that the power producers do not simply neglect their contractual obligations.
In section 2.1, underbidding and subsequent lower project realization rates were highlighted as a problem for policymakers in reverse auctions for renewable energy projects, including in India [5,23,26,30,   . Different types of standards and their economic effects in the context of public procurement were discussed in Ref. [80]. Here, the central role of minimum quality standards is to reduce uncertainty and risk for the public procurer. Furthermore, policymakers were recommended to include minimum viability criteria to prevent underbidding in renewable energy auctions [31]. In a similar vein, underbidding could be countered through stringent qualification criteria as a risk management strategy [26].

Hypothesis 3.
Policymakers require more stringent international quality standards in EPC auctions to reduce the government's risk exposure as owner and investor (while only minimum quality standards are required in BOO auctions, where private producers assume the ultimate risk as owners and investors).

Data
The primary data we use are tender documents -so-called "Requests for Proposals" -issued by SECI to invite solar companies to submit their bids. These documents contain all the crucial information bidders need to structure their offers, including financial and technical eligibility criteria as well as general and procedural information. Firstly, we downloaded all publicly available tender documents from the SECI website, 2 which provided us with a sample of 59 solar PV auctions. SECI kindly provided another 41 tender documents, which gave us a sample of 100 public auctions conducted since SECI's creation in 2013 until 2019. In addition, we conducted informal interviews with relevant stakeholders in public institutions, such as SECI, MNRE, the National Institute of Solar Energy, the Indian Renewable Energy Development Agency and other public banks.

Method
We proceeded in three steps. Initially, we screened the tender documents manually to find out how SECI integrates technical specifications and quality standards into the tender documents.
Secondly, we conducted an automated text analysis. For this purpose, we collaborated with PI Berlin, a well-established testing and auditing laboratory, which has audited several solar PV plants in India since 2013. PI Berlin provided a list of 58 international quality standards considered as the global benchmark in terms of stringency. This includes quality standards that are specifically relevant to the Indian context, such as IEC 60068 environmental testing part 2-68 "dust and sand". Based upon this list, we wrote an algorithm to "mine" our sample of 100 tender documents and count the occurrence of each quality standard in the respective auction.
Note that we took several measures to guarantee that the chosen quality standards truly reflect a global optimal and are relevant to the Indian context. First of all, we refined the list of quality standards during several iterations with PI Berlin. We also took into account the feedback from SECI, e.g. regarding equivalent Indian standards. Furthermore, we reviewed the list with solar PV experts from the National Metrology Institute of Germany to check on the relevance of the chosen quality standards. Moreover, we also compared and complemented the automated results with our manual search.
Thirdly, we manually searched the tender documents for further characteristics of each auction to develop bivariate graphs and conduct a multivariate regression analysis.

Descriptive statistics
The 100 investigated tenders intended to auction a total of 51,455 MW (~51.5 GW). 3 Table 1 provides an overview of the main characteristics of the analyzed auctions. The dependent variable -the required number of quality standards -varies between zero and 22 quality standards. One has to note that one quality standard can encompass several hundred pages that lay out various tests to control product conformity with design, security and performance characteristics. Hence, a standard deviation of about 5 quality standards can entail substantial increase in costs or exclusion of suppliers that cannot conduct the required tests. The three major explanatory variables -global versus domestic auction, auction with or without LCR, and EPC versus BOO -provide substantial variation to explore. About two thirds of the auctions were limited to domestic players (including foreign firms with subsidiaries in India), while one third of the auctions were open to global competition (foreign firms need to set up an Indian company in case they win the auction). Finally, about 60% of the auctions were BOO with an associated PPA while 40% were EPC plus O&M arrangements for at least two and up to 35 years. Fig. 7 in the appendix illustrates that the average (median) BOO project size (296 MW (50 MW)) is about 10 times higher than the average (median) EPC project size (23 MW (6 MW)).

Model
Given that the number of quality standards per auction is strictly positive and constitutes a count, we use a Poisson count regression model. We opt for a Poisson rather than a negative binomial regression model as the dependent variable is not strongly overdispersed (the unconditional variance is not more than two times the mean of the dependent variable). 5 We detect no presence of multicollinearity given a variance inflation factor below 10.
In our primary specification, we estimate the number of international quality standards IQS required in a public auction a in the following form: where P stands for the Poisson transformation, α is a constant, LCR is a dummy for local content requirements, Global indicates whether an auction was open to foreign firms and investors, and EPC states whether the contractual arrangement for the auctioned solar PV plant was EPC plus O&M or BOO with PPA. We also consider several control variables, including plant location, contract length, solar cell technology to be used, maximum auctioned project size, whether a subsidy was offered, whether the plant will be located in a solar park, and a control for the auction year to cover time trends.
In a second step, we also estimate two further regressions that relate to the concern that too stringent technical requirements may repel bidders, leading in the worst case to the cancellation of an auction due to an insufficient number of bidders.
The second model is a logistic model of the following specification: where canceled is a dummy variable (1 = auction canceled; 0 = auction conducted), G is the cumulative distribution function of the standard logistic distribution, IQS is the number of international quality standards, and all other variables are the same as in equation (1). The third model is a negative binomial model 6 of the following form: bidders a = exp (∝ + ß 1 IQS a + ß 2 LCR a + ß 3 Global a + ß 4 EPC a + γ Controls a ) where bidders is the number of bidders (a strictly positive count), exp stands for exponential as the conditional mean of the outcome variable is modeled with an exponential function, 7 and all other variables are the same as in equation (2). Table 2 summarizes the results obtained from regressing several explanatory factors on the dependent variable "number of international quality standards" across all 100 public auctions investigated.

Results
The first coefficient of Table 2 in column (3) indicates no statistically significant correlation between the use of LCR and the number of international quality standards required in public auctions of solar PV plants in India between 2013 and 2019. We read this as evidence that  5 Accordingly, the dispersion parameter "Alpha" in Table 8 column (3) is not significant when estimating a negative binomial regression model in Stata 15, which speaks for the Poisson model. The results also hold if estimated with an Ordinary Least Square model. 6 Here, we use the negative binomial model because the dispersion parameter is significant and hence this model is more appropriate than a Poisson model. For completeness, the interested reader can find the estimated coefficients for Poisson, negative binomial, and Ordinary Least Square regression in Table 5. 7 See Ref. [81] p. 74 onwards for the formal specification of the model, p. 241 for an empirical application, and the Stata 15 base reference manual p. 1707 for further documentation of the negative binomial regression implemented here.
Indian policymakers have not used technical requirements as a performance benchmark to guarantee that local products perform according to international norms (Hypothesis 1). The second coefficient in column (3) suggests that the scope of the auction -whether it was open not only to domestic but also international competitors -did not relate to the number of required quality standards either. This suggests that Indian policymakers did not use technical requirements as a tool to reduce foreign investors' (perceived) uncertainty about solar energy yields by guaranteeing the performance of PV plants per international quality standards (Hypothesis 2). The third coefficient in column (3) illustrates that the dummy variable defining the type of contractual arrangement -EPC vs. BOO -is statistically significant at the 1% level. EPC auctions increase the number of international quality standards by an estimated 54.7% relative to BOO auctions. We read this as evidence for the hypothesis that SECI has used technical requirements as a means to reduce its own technological and commercial risk exposure (Hypothesis 3). Fig. 4 highlights the higher number of international quality standards required by Indian policymakers in EPC relative to BOO auctions. We argue that the difference arises because ultimate ownership and thus risk lies with SECI in EPC auctions, while it lies with the private electricity provider in BOO auctions. This line of reasoning is backed up by an additional analysis of the role of contract length and the composition of additional quality standards employed by SECI in EPC tenders.

Quality standards, contract length and risk allocation
While there is generally a standard contract length of 25 years for a PPA associated with a BOO tender, the contract length varies between 2 and 25 years for O&M periods associated with EPC tenders (very rarely up to 35 years). The longer the O&M period, the longer an EPC company has to bear the risk of technical failure and incur associated costs of reparation, replacement, and lost electricity sales. Fig. 5 shows that SECI requires on average as many technical requirements in EPC tenders with O&M periods above 10 years as in BOO tenders. Yet, SECI requires on average 55% more international quality standards (6 more, a total of 18 on average) in EPC tenders with an O&M period of 10 years or less. In other words, the technical requirements are more stringent if SECI (or another government institution) assumes ownership and full responsibility for the solar PV plant after less than 10 years of operation. This confirms our finding that Indian policymakers employ international  Note: Column (1) provides the average number of international quality standards included as technical requirements in solar PV auctions in India between 2013 and 2019 ("~14 ′′ ) with the standard deviation in square brackets. Columns (2) and (3) provide average marginal effects, which can be multiplied by 100 to be interpreted as percentage changes in the number of required international quality standards. Standard errors are given in round parentheses.
quality standards as a tool to reduce the government's own technocommercial risk. At the same time, they pursue a minimum quality standards laissez-faire strategy if private companies assume the risk for more than 10 years. Table 3 presents some of the additional quality standards required in EPC compared to BOO auctions. While some relate to additional activities that SECI has to assume in EPC auctions (e.g. IEC 62446 commissioning tests and inspection of the solar PV plant), the remaining ones relate to other components of the solar PV plant. Examples include IEC standard 62271 for high-voltage switchgear, IEC 60076 for power transformers, and IEC 62852 for cable connectors, all related to the choice of components. On the other hand, IEC 60364, 61724 and 62446 relate to the quality of labor efforts, including electrical installations and measurement and monitoring of plant performance at the commissioning stage. The purpose of these additional international quality standards is to reduce the handling options for EPC companies seeking to decrease their costs and increase their profits, e.g. by selecting lowquality components or minimizing labor efforts.

Which additional standards are required in EPC tenders, and why?
In section 3.1, we outlined the concerns related to including more international quality standards as technical requirements in public auctions of solar PV plants. In the following, we investigate whether some of these concerns are manifest in the Indian solar PV auctions.

Did international quality standards lead to the cancellation of public auctions due to an insufficient number of bidders?
A major concern raised in our discussions with Indian policymakers is that technical requirements may exclude bidders and, in the worst case, lead to cancellation of public auctions if no (domestic) bidder can comply. In fact, in our sample of 100 public auctions, about 27 have been canceled, according to SECI. The main reason for the cancellation was an insufficient number of bidders. In the following, we investigate whether cancellation correlates with the number of required international quality standards. Table 4 shows the results for regressing the cancellation dummy "canceled" (1 = auction canceled; 0 = auction conducted") on several auction-level explanatory variables. At first sight, column (3) suggests that international quality standards are positively associated with cancelling public auctions of solar PV plants. However, this positive correlation suffers from an omitted variable bias, as illustrated in column (4). Here we drop all the eight auctions that were somewhat "experimental" or "unconventional", namely three auctions of "floating" power plants, four open renewable auctions (wind or solar) and one open solar auction (ground mounted, rooftop or floating). Once these auctions are removed, the positive association of international quality standards with the cancellation variable becomes insignificant. This suggests that SECI has used more international quality standards in these eight PV auctions, likely as an insurance mechanism given the increased uncertainty related to the experimental nature of these auctions.
Instead, two other factors are primary reasons for auction cancellations. We had to drop plant type and plant location from the regression, as they perfectly predict an auction cancellation. In terms of plant type, SECI canceled 5 out of the 8 experimental auctions mentioned above. In  and Madhya Pradesh (two, both canceled). The 24 auctions with no specific location requirement were the most successful, with only one auction canceled. Table 5 illustrates the results for regressing the number of bidders on all auction-level explanatory variables. Firstly, international quality standards are shown not to affect the number of bidders for each auction. We also checked whether this finding depends on the type of contractual arrangement, given our result that EPC auctions drive the increase in technical requirements. However, we do not find that international quality standards in EPC auctions reduce the number of bidders.
Moreover, the results provide interesting information about other factors that have affected the number of bidders. LCRs reduce and subsidies seem to increase the number of bidders. This suggests that another possible way of introducing technical requirements would be to make access to subsidies conditional on compliance with international quality standards. At the same time, it is not clear whether linking technical requirements to LCR would reduce or increase the number of bidders. For example, international quality standards may attenuate the negative effect of LCR on the number of bidders if this were due to the reluctance of international competitors to bid. On the other hand, international quality standards could further reinforce the negative effect of LCR on the number of bidders if bidders in LCR auctions are mainly young domestic companies for whom compliance with technical requirements is more costly and complicated.

Implications
The gradual increase in the stringency of technical requirements in EPC auctions illustrates SECI's accrued technical expertise and underlines that the minimal requirements in BOO auctions represent a strategic political choice. This strategy to transfer risks to project developers benefits the public budget. However, the strategy may have a few unintended, negative consequences. Firstly, given that fierce price competition from cheap, imported components erodes profits, it is unlikely that project developers will invest in research and development and quality upgrading. High-end companies likely either exit the market Note: Column (1) provides the average number of bidders ("7") with the standard deviation in square brackets. Column (2) provides average effects for an Ordinary Least Square model, which can be read as unit changes in the number of bidders. Columns (3) and (4) are based on a Poisson and negative binomial model, respectively. Coefficients in columns (3) and (4) can be multiplied by 100 and interpreted as percentage changes in the number of bidders. Standard errors are given in round parentheses. "Lnalpha" suggests that the Nbreg model is the most appropriate to use, as the dependent variable is a count and overdispersed. Note: Column (1) provides the mean cancellation rate ("27%") with the standard deviation in square brackets. Columns (2)-(4) provide average marginal effects, which can be read as percentage changes in the cancellation rate. Standard errors are given in round parentheses.
or lower the quality/costs of their components. As a result, the actual project costs may be higher and the generated electricity lower than expected. Secondly, it seems plausible that more stringent technical requirements could have helped to attract investors with a longer time horizon. Technological risk is one factor that explains why some banks and international investors remain reluctant to finance solar power projects in India [24]. In the early days, it was easier to get equity finance from international venture capital with higher risk appetite than debt from national and international banks. Given the short time horizon of these players and the variable loan rates in India, non-performing power plants could lead to a sudden withdrawal or reduction in equity/debt finance. Suppose many project developers have engaged in speculative or Ponzi financing [24,82]. In that case, this could lead to bankruptcies due to project developers' inability to pay the interest with the returns from low electricity generation. Project developers with more stringent technical requirements will also resell their solar PV plants on the secondary market, which has started in recent years, more easily and for higher prices. In contrast, others may sell at a loss [70,83].

Conclusion and outlook
After analyzing 100 public auctions for solar PV plants in India, we find that policymakers seem to use technical requirements as a tool to reduce the government's own techno-commercial risk exposure rather than a signaling device to promote foreign investment or to encourage domestic industrial upgrading. Within three years since its creation, SECI started to require more stringent technical requirements in EPC auctions where the government assumes the final risk as the ultimate owner of the project. However, there has been no increase in the quality level required in BOO auctions, representing about 60% of the auctions and the lion's share of auctioned electricity. We believe this is a missed opportunity: had international quality standards been connected to LCR or subsidies, Indian and foreign bidders would have had a stronger incentive to catch up with the global frontier, and more international and long-term investment may have been gained.
Based on these results, we recommend policymakers in India and other countries adopt a strategic and proactive approach to the use of technical requirements. A gradual increase in the number and stringency of international quality standards, ideally combined with forward guidance (early announcement, transition period), incentives to adopt international quality standards (e.g. making access to subsidies conditional on compliance with international quality standards), and support measures to set up domestic laboratories or vouchers for product testing, should promote productivity increases through learning from international standards. This gradual increase could prevent negative consequences from a too ambitious use of international quality standards, such as excluding nascent local companies. To facilitate learning-bydoing, the communication of technical requirements should be improved, as this is particularly important for new players. For example, guidelines and videos could be added to the required international quality standards, potentially in collaboration with and financed by international institutions (i.e. IEC, IRENA, World Bank). Governments should also actively promote the market entry of private laboratories by guaranteeing that specific tests will be required in public auctions in the future and offering low-interest credits for investments in new laboratories or equipment for new tests. Finally, governments should seek to enable real-time, central monitoring of the amount and efficiency of electricity generation from PV plants to reduce information asymmetry between project developers, investors, and state utilities.
In the future, it would be interesting to investigate whether there are any potential unintended consequences of the Indian government's minimum quality standard approach. For example, we suggest exploring whether certain firms exited the market or adapted their business strategies to favor price competition over quality. It would be equally relevant to analyze whether minimum quality standards have led to low performance of solar PV plants, e.g. due to the use of low-quality components, inadequate transportation or installation of components. Finally, it would be important to examine whether international quality standards could attract more investment including investors with a longer time horizon as well as incentivize firms that benefit from LCR.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.    Full version of the regression analysis summarized in Table 2 Variables ( Note: Columns (1) to (3) are the same as in Table 2 and were described in detail there. Here, Andhra Pradesh is the reference category for the location controls and floating for the plant type controls.

Table 8
Sensitivity of the estimates presented in Table 2 to the regression model used. Note: ***p < 0.01, **p < 0.05, *p < 0.1. All standard errors are robust. Column (1) provides the mean number of international quality standards required in the 100 public auctions. Column (2) provides estimates based on the Poisson model. Column (3) provides estimates for a negative binomial model, which correspond to a Poisson model given that the dispersion parameter alpha is insignificant and thus a Poisson model is estimated. Column (4) presents estimates for an ordinary least squares (OLS) model. Location and plant-type control variables are the same as in Table 7 of the appendix.

Table 9
Sensitivity of the estimates presented in Table 2 IQS  IQS  IQS  IQS  IQS  IQS  IQS  IQS  IQS  IQS