Flexibilization or biomethane upgrading? Investment preference of German biogas plant operators for the follow‐up of guaranteed feed‐in tariffs

This article reports the results of a discrete choice experiment with 183 German biogas plant operators designed to elicit the respondents' plans for biogas utilization pathways after the end of guaranteed feed‐in tariffs. Participants could choose between ‘flexibilization’ for demand‐based electricity generation and conversion to biomethane upgrading for direct feed‐in into the natural gas grid. A binomial logit model revealed a 37% probability of switching to biomethane upgrading. These plants are characterized by higher capacities, several involved shareholders, secured succession, costly digestate disposal and belonging to the upper performance quartile. Mixed logit estimations conducted separately for the two investment concepts revealed a very high overall willingness to invest: 71% for flexibilization and 82% for biomethane upgrading. The respondents demand a return on investment of 19% for flexibilization and 26% for biomethane upgrading. Within the flexibilization, twofold overbuilding (installed capacity equals 2 times the rated power) is clearly preferred to fivefold overbuilding. For the biomethane upgrading, private ownership of the upgrading plant is preferred to a joint investment in a central upgrading facility. Limiting the use of energy crops reduces the propensity to invest in both models, while a longer utilization period enhances it. The respondents consider lack of planning reliability as the biggest obstacle to invest, followed by long approval procedures and high investment costs due to restrictive legal requirements.

17.62 TWh th of thermal energy (AGEE-Stat, 2022).Since electricity generation from biogas is costly in comparison to alternative renewable energies, the development of biogas production in Germany has mainly been driven by government support in the form of guaranteed feedin tariffs under the Renewable Energies Act (Erneuerbare Energien Gesetz-EEG) (Bahrs & Angenendt, 2019).
Since 2020, 20 years after the launch of the EEG, older biogas plants have been reaching the end of the legally guaranteed feed-in tariffs and a large share of the approximately 9600 plants in operation will be phased out of support in line with the expansion path from 2025 (Güsewell et al., 2019;Lauer & Thrän, 2017).For repowering, that is the ongoing operation of these plants, various follow-up concepts are being considered; these are associated with more or less extensive investment volumes (Grösch et al., 2020).Essentially, the operation of the biogas plants must be adapted to market needs and the requirements of future energy systems.Two basic alternatives for extended operation are emerging: (1) flexibilization, meaning the demand-based electricity generation and (2) biomethane upgrading, meaning to produce biomethane for direct feed-in into the grid as a substitute for natural gas.
This article aims to complement the literature by considering, for the first time, operators' investment preferences regarding this two options for extended operation of their plants.It extends our knowledge of how the parameters of the required investment affect choices, and it provides useful insights for policy makers by revealing necessary support requirements for the trajectory of sustainable biogas pathways.Due to the hypothetical character of the survey on preferred investment concepts a discrete choice experiment (DCE) was implemented.Besides basic characteristics of the investments itself, such as expected profit, capital demand and useful lifetime, the impact of legal requirements and operational characteristics of the biogas plants on investment decisions are investigated.
Discrete choice experiments are widely used in agricultural and environmental economics.Many articles aim to elicit preferences for the design of agri-environmental schemes (Birol et al., 2006;Broch & Vedel, 2012;Kuhfuss et al., 2016;Ruto & Garrod, 2009).Choice experiments have also been used to investigate the determinants of managerial decisions such as conclusion of supply contracts (Petersen & Hess, 2018;Reise, Liebe, & Mußhoff, 2012;Sauthoff et al., 2016), expansions of the production programme (Breustedt et al., 2008) or implementation of innovative production processes (Deressa et al., 2009;Gebrezgabher et al., 2015;Thiermann & Latacz-Lohmann, 2022).Other studies examine consumers' willingness to pay for renewable energies (Bartczak et al., 2017;Knoefel et al., 2018;Lehmann et al., 2021).However, there is a paucity of research on biogas-related investment choices.Zemo and Termansen (2018) studied farmers' preferences for community-based biogas plants.The authors showed that the possibility of cooperation among farmers increases the willingness to invest, which is attributed, among other things, to the distribution of financial and operational risk.If, however, a certain number of partners is exceeded the propensity to invest declines.Reise, Mußhoff, et al. (2012) investigated the investment behavior of German farmers for the expansion of bioenergy by means of hypothetical investment concepts.They found that it is primarily the capital costs and the expected risk that influence the decision.Nonmonetary parameters hardly have any influence, but experience in the field of renewable energies lowers the investment threshold.Burg et al. (2021) used an agentbased modelling approach to investigate Swiss farmers' preferences for investing in biogas plants.The agents' objective functions were calibrated with the results of a discrete choice experiment previously carried out.Decisive for the willingness to innovate are the energy prices and the number of cooperation partners needed to supply the minimum required biomass.As an alternative approach, Venus et al. (2021) applied a Q-experiment to determine stakeholders' views on the future development of biogas in Germany.While the importance of biogas for the energy system is recognised in principle, attitudes towards the general promotion of the plants are divided and the focus on specific aspects, such as flexibility, is favored.
The remainder of the paper is structured as follows.Section 1.2 sheds light on the potential role in the future energy system of flexibilization and biomethane upgrading considered in this paper.Section 2 sets out the methodological approach, describes the experimental design and develops hypotheses.Section 3 presents the results including descriptive statistics, model estimation results and hypotheses testing.Section 4 discusses the results.
1.2 | The potential role of biogas in the energy system As biomass resources are limited and electricity generation from biomass is comparatively costly (Bahrs & Angenendt, 2019), wind energy and photovoltaics are likely to provide the bulk of renewable energy (Hahn et al., 2014).However, these energy sources are characterized by high temporal fluctuation and consequently variable energy supplies, leading to high adaptation needs for the remaining energy sources (Schaber et al., 2012).Biogas plants can compensate for the fluctuations due to the rapid availability and storability of the energy carrier and contribute significantly to the integration of other renewable energies by balancing the residual load (Hahn et al., 2014;Lauer & Thrän, 2017).

| 3 of 22
As described above, the development of biogas production in Germany has mainly been driven by government support in the form of feed-in and premium tariffs guaranteed for 20 years.In this context, on-site electricity generation has barely been linked to market needs, which could be achieved by installing combined heat and power units (CHP) with significantly higher capacity than required for continuous operation, so that extra electricity can be generated at times of increased demand.The intensity of this measure, known as overbuilding, can be described by the power quotient PQ according to Lauer et al. (2017), which is the installed power divided by the rated power. 1  The additional revenues from marketing that were possible in the past are stated to be up to 0.67 ct/kWh (Barchmann et al., 2016;Haensel et al., 2020;Lauer et al., 2017).Depending on the extent of the necessary capacity of two to five times the capacity required for base load operation, Schröer and Latacz-Lohmann (2022) calculated between 0.48 and 0.87 ct/kWh additional revenue generated from demand-orientated marketing, which corresponds to an additional revenue of 2.60%-4.74%compared to the EEG (2021) maximum value of 18.4 ct/kWh. 2 As these additional revenues cannot cover the extra costs involved, flexibilization of biogas plants has been incentivized through support premiums under the EEG since 2012.The flexibilization premium is intended to partially offset the costs of installing the additional electricity generation capacity (Bahrs & Angenendt, 2019;Briest et al., 2022;Daniel-Gromke et al., 2018;Scheftelowitz et al., 2018).Despite the fact that about 40% of biogas plants have registered for the flexibility premium, only 25% have actually implemented flexible CHP units [own calculation according to Marktstammdatenregister (2023)].Nevertheless, the share of flexible plants is continuously rising (Scheftelowitz et al., 2018).While German biogas plants had for many years received the highest feed-in tariff across Europe (Del Pablo-Romero et al., 2017), the 2017 amendment of the EEG saw a drastic decline of feed-in tariffs by requiring operators to submit bids for these tariffs.This system change in combination with the long remaining term of the first support period for the majority of existing plants resulted in low participation rates in the tender rounds (Bundesnetzagentur, 2021;Haensel et al., 2020).This could not be improved by the slight increase in the maximum acceptable bid (bid cap) under EEG (2021).In order to meet the requirements of the future energy system and considering that the EEG is the only direct support option for biogas plants, it is expected that a large proportion of existing biogas plants will eventually submit bids for extended electricity generation (Bahrs & Angenendt, 2019;Brémond et al., 2021).
One such concept is the use of biomethane as a substitute for natural gas, especially since this concept better exploits the technical potential of biogas for GHG reduction (Börjesson & Ahlgren, 2012;Poeschl et al., 2010Poeschl et al., , 2012)).By directly feeding biomethane, after technical treatment, into the natural gas grid, it can be used flexibly in terms of time and space for electricity or heat generation, as well as fuel use, depending on the sectors' needs (Bahrs & Angenendt, 2019;Budzianowski & Brodacka, 2017).However, the specific costs in the typical capacity range of German biogas plants (between 350 and 600 kW rated power) are high (Beil et al., 2019;Skovsgaard & Jacobsen, 2017;Walla & Schneeberger, 2008).In combination with low biomethane prices and a lack of support schemes, biomethane upgrading has been reserved for large plants that are mainly powered by energy crops (Deutsche Energie-Agentur, 2020).Accordingly, biomethane upgrading plays a minor role in Germany with 213 existing plants, while this technology dominates in other European countries such as Denmark and Sweden (Brémond et al., 2021).To make biomethane upgrading financially feasible for small biogas plants, it is necessary to connect several decentralised biogas plants to a central biomethane upgrading facility.This would enable the exploitation of economies of scale (Hengeveld et al., 2016).In addition, the legal framework for biomethane marketing has fundamentally changed in 2020 and 2021.The European Commission's Renewable Energy Directive (RED II), implemented in Germany through the Federal Emissions Control Act (BImSchG, 2021), imposes increasing GHG reduction obligations on mineral oil companies.In the same vein, the Buildings Energy Act (GEG, 2020) grants preferential treatment to biomethane in the calculation of the 'primary energy demand' of buildings.Both legal changes have contributed to a recent massive increase in the price of biomethane (Deutsche Energie-Agentur, 2021).The growing market for biomethane and good price prospects raise the financial attractiveness of this concept for operators who wish to become independent of state support under the EEG.

| Experimental design
Since the survey on preferred investment concepts is hypothetical in nature and goes beyond today's demands, a methodology had to be chosen that allows dealing with this situation.Discrete choice experiments (DCE) meet this requirement and make it possible to determine information on the individual properties of the decision situations (Hensher et al., 2015), i.e. the previously determined attributes and their characteristics.Because of the different parameters of the investment in flexibilization versus biomethane upgrading, an experimental design was formed for each of the two concepts separately.The choice alternatives vary in the five attributes shown in Table 1: profit, capital requirement, investment concept, share of energy crop input and duration.
The experimental design follows from the attributes and their levels described in Table 1 for flexibilization and biomethane upgrading, respectively.In order to verify the selection of relevant attributes and characteristics, indepth interviews were conducted with consultancies and plant operators.Due to the very small total population of 5920 operators, a comprehensive pre-test of the survey among plant operators was not conducted, as it was feared that operators would not participate in the survey a second time.In contrast to the number of choice tasks and attributes, the number of alternatives in a choice set can have a negative effect on response efficiency (Mariel et al., 2021).Due to the increasing complexity with an increasing number of alternatives, the advantage of an alternative that is better suited to the individual preferences can be nullified (Zhang & Adamowicz, 2011).While negative effects on choice behavior were found with four to five simultaneously offered action alternatives, choice sets with fewer alternatives did not show this effect.In order to obtain as many observations as possible, three action alternatives were therefore offered, which should counteract the problem of an overall small population.A D-efficient design was generated with the module Dcreate in Stata 17 (Hole, 2017).The design's D-efficiency according to Kuhfeld (2010) is 98%.The questionnaire was divided into 2 blocks of 6 decision situations each.
In the survey, participants were asked to provide basic characteristics of their biogas plant, such as capacity specifications, substrate input and the number of other biogas plants in the vicinity.The participants were then asked to decide which of the two concepts, flexibilization or biomethane upgrading, they would consider for their plant in the first stage.Since extended operation in the status quo is not possible and an investment has to be made, but no specific investment concepts were presented in this stage, no opt-out variant was considered.Based on this, three investment alternatives of the previously selected concept were presented.In addition, a status quo alternative was offered at this second stage, which effectively implied the decommissioning of the biogas plant.Figure 1 shows the structural procedure of the survey and a choice set for each of the two concepts. 3 The survey was conducted between January and October 2021 via the Unipark platform.Due to few participants and energy market distortions in the aftermath of the conflict in Ukraine a second round was conducted between March and October 2022.The survey was initially distributed by industry interest groups, chambers of agriculture and advisory rings.In addition, the survey was advertised in authors' presentations, and stakeholders were also contacted directly by email or phone.To encourage participation, a two-part video podcast was recorded in cooperation with a biogas consulting company.In the first part, both investment concepts flexibilization and biomethane upgrading were explained, followed by a request to participate in the survey.As an incentive to complete the survey, the second part of the video podcast was offered at the end of the survey, in which profitability analyses were presented for selected concepts.
A total of 192 participants completed the survey in full, of which 9 participants were excluded due to implausible information, 183 operators remain in the data set.The completion rate was 12.3% which is in line with previous experience with online surveys.Taking into account a total of about 8628 operators registered in the German Market Master Data Register (Marktstammdatenregister-MaStR), 2.1% of all German plant operators have completed the survey (Marktstammdatenregister, 2023).

| Hypotheses
H1. Higher profits increase the willingness to invest.
The first attribute is the profit that can be achieved with the investment.For the flexibilization, possible revenues   (Landwirtschaftlicher Buchführungsverband, 2019).The range of typical profits (between €100 and €200 annually per kW of rated power) was considered in the choice sets.
For the biomethane upgrading, revenues were calculated based on current prices reported in the biomethane industry barometer at 7.5 ct/kWh (Deutsche Energie-Agentur, 2020), plus avoided grid fees of 0.7 ct/kWh (GasNZV, 2019).The costs of biomethane upgrading (between 1.1 and 4.7 ct/kWh) were taken from Beil et al. (2019), the costs of biogas production (5.5 ct/ kWh) from Schröer and Latacz-Lohmann (2022).Due to economies of scale, especially in the treatment costs, there is a very wide range of operating results.The possibility of negative profits was excluded, as these did not represent a reasonable investment alternative.The remaining profit range with net profits of €100, 450 or 600 per m 3 of upgrading capacity was considered in the choice sets.

H2.
As capital requirements increase, the willingness to invest decreases.
The capital requirement as the second attribute represents the range of investments.Regarding the flexibilization, the figures are based on investments in new CHPs.In order to determine these, a cost function was created that is based on available cost functions and published calculation examples, taking into account costs for planning and approval, new required components (e.g.new transformers), and typical gas and heat storage facilities (Altrock et al., 2013;Fachagentur Nachwachsende Rohstoffe e.V., 2018;Haensel et al., 2020;KTBL, 2013).Equation ( 1) shows the specific total investment for flexibilization in €/kW of capacity to be installed.In the capacity range between 250 and 2000 kW, which is typical for German biogas plants, investment requirements range between €700 and 1600 per kW.The capital requirement for biomethane upgrading is based on the results of a manufacturer survey by the German Biomass Research Centre (DBFZ), which determined the investment for different upgrading technologies and capacities (Beil et al., 2019).Equation ( 2) represents the specific investment in Euro per m 3 /h upgrading capacity.The capital requirements used in the choice sets correspond to upgrading capacities between 250 and 2000 m 3 /h of biogas H3.PQ 2 leads to a higher willingness to invest, while passive flexibilization and PQ 5 reduces it.
The third attribute relates to the investment concept.The support of the flexibilization of existing biogas plants before is limited to PQ 5 by the flexibility premium.This level has therefore long been recognized in the industry as the upper limit of the power increase, so that this also represents the highest level in the choice sets.Even though the EEG (2021) does not provide for any further upper limit of the PQ, the specification of the operating time represents a limitation of the PQ.So at least 85 percent of the installed capacity must be generated for at least 4000 quarter hours a year.The PQ 5 is expected to have a negative impact on the willingness to invest as it results in significantly higher capital requirements with more extensive planning and approval procedures as well as higher technical requirements for operation (Fachagentur Nachwachsende Rohstoffe e.V., 2018).PQ 2 is the lowest level in the choice sets.This requires less capital with slightly lower profits than PQ 5. Hence, the expectation that PQ 2 will emerge as the preferred investment.Another possibility to achieve flexibility is the so-called passive flexibilization, that is, the reduction of rated power, whereby the existing CHP units are sufficient for flexible electricity generation.Reduced rated output lowers profits, hence the expectation that it will also reduce the likelihood of being chosen.

H4. The possibility of central upgrading facility by means of a gas collection pipeline increases the willingness to invest.
For the biomethane upgrading, owning the upgrading plant is the obvious option (Burg et al., 2021).An alternative to counter the high specific investment costs in the capacity range of typical biogas plants is the connection of neighboring biogas plants to a central upgrading facility by means of a gas collection pipeline (Hengeveld et al., 2014(Hengeveld et al., , 2016)).Zemo and Termansen (2018) show that the option to invest in a biogas plant in partnership significantly increases willingness to invest, which can be attributed to economic factors such as the distribution of financial risk.

H5. A longer utilization period or contract duration increases the willingness to invest.
The useful life or contract duration is the fourth attribute.The EEG (2021) provides for a follow-up guaranteed feed-in tariff for a period of 10 years.However, since further use is also possible via power-purchase-agreements (PPA) or direct-marketing at the energy exchange, longer utilization periods of 15 or 20 further years are also taken into account in the choice experiment.In the biomethane sector, short-(1 year), medium-(5 years), and long-term (10 years) contracts are considered, which correspond to practice.In the literature, the tendency towards longer contract durations is presented, especially in the context of larger investments (Zemo & Termansen, 2018).H6.Alternatives with a higher share of energy crop input lead to a higher willingness to invest.
Under EEG (2021), apart from limiting maize silage and cereal grain to 40% of substrate input, no further requirements are placed on the use of substrates.Small slurry-based biogas plants represent an exception in this respect.These plants are specifically promoted under the EEG (2021) under the condition that at least 80% animal manure is used in the substrate mix.In general, the use of energy crops for biogas production is increasingly criticized due to monocultures and negative environmental effects, and substitution with agricultural residues is favored (Bacenetti et al., 2016;Meyer et al., 2018).The limitation of the share of energy crops primarily affects maize silage as the main feedstock for biogas production, which, however, has many advantages from the operator's point of view.The dry matter content is high, the ensiling, storage and fermentation properties are good, methane yields are high and supply costs low (Auburger et al., 2017).Venus et al. (2021) show a consensus among plant operators and other stakeholders that a complete ban on energy crops is not expedient.H7.Farms with high availability of sustainable substrates (i.e.those with high livestock densities or availability of other farm residues) are more likely to invest.
Since legislation is expected to change towards more sustainable substrates (Brémond et al., 2021), it follows that these farms can adapt comparatively easily to the new requirements.Furthermore, the literature indicates that farm size has a positive influence on the implementation of new technologies or sustainable production (Gebrezgabher et al., 2015;Rahman & Bulbul, 2015;Ren et al., 2019).
H8. Farms facing difficulties in manure management (e.g.farms located in nitrate-sensitive areas, farms without sufficient slurry storage capacity, or farms that have to pay for digestate disposal) are less likely to invest.
Contrary to the previous hypothesis, difficulties in manure management may be less favorable for an investment.For example, the German Fertilizer Ordinance (Düngeverordnung-DüV) stipulates a reduction of nitrogen fertilization to 20% below crop needs, as well as smaller application windows, when located in a nitrogen-sensitive area (DüV, 2021).In these regions, the disposal costs for manure and digestate are usually high, reducing the profitability of biogas plants which, in turn, may lower the propensity to invest.
H9. Operators of higher-capacity biogas plants are more likely to opt for the biomethane upgrading.
As biomethane upgrading benefits from strong economies of scale, plants with higher capacity are more likely to switch to gas upgrading.In general, 250 m 3 of biogas per hour, corresponding to 550 kW el installed capacity, is considered the lower limit of financial feasibility (Beil et al., 2019).

H10. The majority of participants choose the flexibilization.
Due to the promotion of flexibility since EEG (2021), the procedure, the possible additional revenues and technical requirements, as well as the risks have been known to plant operators for a long time.Due to the long experience with the technology, we expect that the majority of operators will opt for it.It is known that experience influences investment decisions (Reise, Mußhoff, et al., 2012) and a high share of flexible on-site electricity generation is expected (Bahrs & Angenendt, 2019;Brémond et al., 2021).

| Principles of discrete choice analysis
According to McFadden (1974), the utility of one of the available alternatives of action is derived from its characteristics, in the case of this paper, the attributes and characteristics of the hypothetical investment alternatives.
It is assumed that a respondent n chooses the alternative i which provides him or her with the greatest utility U ni compared to another alternative U nj : U ni > U nj (Train, 2009).Since utility is a latent variable that is deterministic for the decision-maker but only incompletely observable by the analyst.Utility must therefore be regarded as a random variable.The reasons for the randomness of utility are unobservable characteristics, variations in preferences, measurement errors and misspecification (Auspurg & Liebe, 2011).The random utility function U nj is thus divided into a deterministic component V nj and a stochastic component ε nj .An additive function is usually assumed for the deterministic component.The vector x njk describes the characteristics K of the alternatives, β is the coefficient of these characteristics.In addition, the inclusion of J-1 alternative-specific constants (ASC) β j is possible, which describe systematic influences on utility that stem from factors not included in the model (Train, 2009) Due to the randomness of the stochastic component, statements can only be made about the probability that a particular alternative is chosen (Hoyos, 2010).The probability of choice P ni of an alternative i by agent n is calculated as (Train, 2009): Taking into account the density of the stochastic component f = (ε n ), the cumulative choice probability can be expressed as follows (Train, 2009) DCE are estimated using standard methods of discrete choice analysis.Depending on the distribution (3) (5) of ε n , logit, or probit models can be used to estimate Equation ( 5) (Auspurg & Liebe, 2011).Logit models are most frequently used, because the assumption of the iid (independently and identically) extreme value distribution of the stochastic component for all participants allows for a simple solution (Train, 2009).If the choice situation contains two alternatives, the model is called binary probit or logit model (Auspurg & Liebe, 2011).
For the conditional logit model, the following choice probability results: In order to test whether there is a difference between the participants who choose flexibilization and those who choose biomethane upgrading at the first stage of the survey, a binomial logit model was estimated.In case of this particular question, the personal and operational characteristics of the participants are used as explanatory variables in the model estimation to explain the choice between the two options.
For the estimation of the stated-preference data collected at the second stage of this studies survey, a mixed logit model is used because of the three known advantages over the standard logit model: (1) the IIA assumption ('Independence of Irrelevant Alternatives') resulting from the iid distribution of ε nj does not have to be fulfilled; (2) panel data can be used; and (3) unobserved heterogeneity between decision-makers is taken into account (Train, 2009).This is implemented by considering a vector β n instead of uniform coefficients β as in the standard logit model.The mixed logit probability is then the integral of the standard logit probability over the distribution of β n , where θ describes the moments of the distribution (Train, 2009) Estimation of the mixed logit model occurs through a simulation by drawing different values β r from the distribution function.After the draw, the logit probability is calculated.This procedure is repeated many times and the results are averaged.The simulated choice probability Pni results from R draws (Train, 2009), The choice data of the two subsamples were analyzed separately using a mixed logit model, drawing on the mixlogit module in Stata 17 (Hole, 2007).The estimation results provide insights into the factors affecting operators' willingness to invest in the future of their biogas plants compared to the decommissioning of the plants.Such analyses were carried out separately for the two alternative investment concepts: flexibilization and biomethane upgrading.
The mixed logit model accounts for preference heterogeneity by forming interaction terms of participants' socio-demographic characteristics with the alternative specific constant (ASC) or attributes (Auspurg & Liebe, 2011).In order to implement this interaction terms of personal and operational characteristics of the biogas plants with the ASC were formed from which it can be interpreted whether these have an influence on the choice of an alternative course of action.In the course of data preparation, missing values for information on heat prices were replaced with the mean value.Missing values on animals kept, special crops and arable or grass land were replaced with zero.The selection of personal and operational characteristics is based on the likelihood ratio test.1000 Halton draws were performed for each model.The mixlogit models Pseudo-R 2 for the flexibilization is 0.3396 (Chi 2 18.7; Prob > Chi 0.4104) and for the biomethane model 0.2288 (Chi 2 8.61; Prob > Chi 0.8968).
As one of the main outcomes of a DCE, willingness-to-accept (WTA) estimates are calculated as the ratio of a coefficient β k of an attribute to the coefficient β c of the monetary variable (Mariel et al., 2021): WTA = −β k /β c .The WTA describes by how much the profit needs to be increased if a negatively valued attribute is offered such that the probability of participation remains unchanged.In the case of this paper, it can be estimated how high additional profits must be in order for higher plant operation requirements to be accepted.

| Descriptive statistic and factors affecting the choice of investment concept
Table 2 shows the descriptive statistics.Compared to the German average of 407 kW rated power, 558 kW installed capacity and commissioning in 2009 (Marktstammdatenregister, 2023) it is evident that older plants with comparatively high capacity are present in the survey.In addition, the share of energy crops is about 10% higher and that of animal manure 10% lower than the national average (Fachagentur Nachwachsende Rohstoffe e.V., 2022).To identify possible systematic differences between both samples for flexibilization and biomethane ( 6) (7) T   upgrading, the means of variables were compared using Welch test, or Wilcoxon test, depending on the variable types.Information on the test statistics is given Tables A1 and A2 in the Appendix.The binomial logit model, presented in Table 3, was estimated to determine whether these variables affect the choice between the two investment concepts.A large proportion of the participants (63%) opted for the flexibilization, thus the binomial logit model estimates a choice probability for biomethane upgrading of 37%, lending support to H10.Participants who opted for biomethane upgrading have a higher rated capacity, which equates to a larger capacity of the biogas plant (p = 0.0032).The 95% confidence interval for flexibilization is between 445 and 586 kW el , while the interval for the biomethane choice lies between 612 and 895 kW el .The binomial logit model thus reveals a higher choice probability for participants with a higher rated power.This lends support to H9.Furthermore, participants who see themselves in the lower quartile of farm performance are more likely to choose the biomethane upgrading.This is not clearly indicated by the comparison of both samples (p = 0.0598) but is significant in the binomial logit model.Also, the mean values for animal intensive (farm located in a livestock-intensive region, p = 0.0002) and digestate disposal (farm has to dispose of digestate at a charge due to the fertiliser ordinance, p = 0.0003) are higher for participants in the biomethane sample.The binomial logit reveals a higher choice probability for operators who have to dispose of digestate at a charge and those whose plants are located in regions with high livestock densities, thus lending support to H7.The share of collectively operated plants in the biomethane sample is significantly higher (p = 0.0151) and has a positive coefficient in the binomial logit model.This finding is in support of H4, whereby the possibility of a jointly operated biomethane upgrading plant increases the willingness to invest.
For the other variables, no significant differences in the means between the two samples could be detected.However, the binomial logit model shows a positive influence of successor on the choice of biomethane upgrading.By contrast, operators who stated the need for financial support are less likely to choose biomethane upgrading.Interestingly, operators who took the survey after the outbreak of the Ukraine war are less likely to choose biomethane upgrading.confidence intervals for the flexibilization.Table 5 displays this numbers for the biomethane model.The estimated overall probabilities of choosing an investment over and above the decommissioning of the plants are 71% for the flexibilization and 82% for the biomethane upgrading.Interestingly, 51% of the participants always chose an investment alternative, 38% chose both investment and decommissioning selectively, and only 11% chose decommissioning in every choice situation.interpreting the WTA, it should be noted that it is the result of hypothetical decision situations.Further research may be needed before adjusting funding policies, as participants in choice experiments tend to have excessive payment demands (Fifer et al., 2014;Hensher, 2010;Lusk & Schroeder, 2004;Moser et al., 2014).

| Factors affecting operators' willingness to invest in the follow-up use of their biogas plants
The results of the mixed logit models lend support to most of the hypotheses.Beginning with economic characteristics, higher expected profit increases the investment probability (H1✓), while higher capital requirements reduce it (H2✓).The WTA of €0.19 per kW el rated power for an additional capital requirement of 1 Euro per kW installed capacity means that the profit would have to increase by €0.19 per kW el rated power in order to keep the choice probability unchanged.For biomethane upgrading the WTA is €0.26.
When considering the flexibilization, the attribute level PQ2 shows a significant positive influence on investment behavior, whereas the PQ5 level has a negative impact (H3✓).For biomethane upgrading the mixed logit model (Table 5) reveals highly significant coefficients for both on-site_upgrading and collective_upgrading, which means that both investment concepts (refer to Table 1) are preferred to the option not to invest and decommission the plants.Looking back at the results of the binomial logit model (Table 3), it can be said that the operation of a collective plant with several shareholders increases the willingness to switch towards biomethane upgrading, but joint operation with other stakeholders (negative coefficient of cooperative_plant in Table 5) tends to limit the subsequent willingness to invest (H4⨯).The presence of other plants in the immediate vicinity does not seem to be the reason for this, as the variable neighbours could be excluded from the biomethane model on the basis of a likelihood ratio test.
Higher shares of energy crops in the substrate mix shows the expected effect in both estimations, i.e. it increases the propensity to invest, thus lending support to H6.The WTA for one percent more energy crop allowed in the substrate mix is estimated at €-9.12 per kW rated power for the flexibilization alternatives.With a simplified assumption that 1 m 3 /h of upgrading capacity corresponds to 2 kW el rated power (see explanations in Table 2.), the WTa of €-26.69 per m 3 upgrading capacity for one percent more energy crop allowed in the biomethane alternatives is equivalent to €-13.35 per kW rated power and thus higher than in the case of flexibilization.
A positive effect of a longer utilization period could be determined for both concepts, lending support to H5.
Considering personal and operational characteristics, some of the variables which determine the choice of investment concept in the binomial logit model also show up in the mixed logit models.Operators who rate themselves in the lower 25% bracket show a reduced investment probability in both the biomethane and flexibilization model.Although many variables do not influence the choice of investment concept and were excluded from the binomial logit model, the coefficients of these variables have different signs in the mixed logit models for flexibilization and biomethane upgrading.It seems that operational linkages, while not influencing the choice of the investment concept, are significant for the assessment of a final investment alternative.Hypotheses H7 (Farms with high availability of sustainable substrates are more likely to invest) and H8 (Farms facing difficulties in manure management are less likely to invest) can therefore not be answered conclusively.
The comments and remarks made by the respondents at the end of the survey highlight the importance of political aspects.A quarter of the participants complained about long approval procedures, costly expert opinions and high replacement investment (for example agitators) due to unfavorable legal framework conditions.The lack of planning reliability with regard to uncertain development of future EEG subsidies, restrictive conditions attached to these subsidies and complex approval procedures of the tender system as well as increasing technical requirements for plant operation were criticized by 29% of the respondents.The mean values of the Likert scales (see Table 2), which were to elicit barriers to a possible investment, underline this picture.Lack of planning reliability and uncertainty over government subsidies are seen as high risks, while maintenance requirements and safety standards seem less relevant.
The mixed logit models could determine significant effects for only some of the Likert scales.For flexibilization, operators are less likely to invest if they believe that the costs of a general overhaul of the biogas plant are high.In contrast, they are more likely to invest if they have already dealt with possible concepts for the follow-up use beforehand.It can therefore be concluded that lack of knowledge has a significant influence on the decision to invest.
While the mixed logit model for biomethane upgrading identifies a negative development in the willingness to invest since the beginning of the Ukraine conflict, the opposite can be observed in the model for flexibilization.
T A B L E 5 Mixed logit model, marginal effects, and willingness to accept (WTA) for operators preferring biomethane upgrading.German biogas plants are reaching the end of the legally guaranteed feed-in tariffs, and plant operators are facing two possible alternatives for repowering-flexibilization or biomethane upgrading to produce a natural gas substitute.A DCE with 183 plant operators was conducted to elicit operators' investment preferences regarding this two options for extended operation of their plants.We found evidence in support of most of the hypotheses (Table 7).

Coefficient
A binomial logit model reveals a probability of 37% for choosing biomethane upgrading, supporting the statements of industry experts who see a large proportion of plants continuing to be subsidized under the EEG for on-site electricity generation in the future.The fact that Participants who opted for biomethane upgrading have a higher rated capacity and are more likely to choose biomethane upgrading can be explained because these plants benefit from stronger economies of scale compared to smaller ones.Since participants who see themselves in the lower quartile of farm performance are more likely to choose the biomethane upgrading, it is possible that these participants see a potential to improve by switching from on-site electricity generation to biomethane upgrading.Interestingly, the mean values for animal intensive and digestate disposal are higher for participants in the biomethane sample, indicating better availability of farm manure.This may be due to the marketing of biomethane as a fuel, higher GHG reductions and thus higher prices achieved when generated from farm manure.In regions with high slurry volumes, the switch to biomethane upgrading is therefore financially more interesting.The significantly higher share of collectively operated plants in the biomethane sample and the positive coefficient in the binomial logit model indicate a higher willingness to take the risk of changing the utilization pathway when other operators participate in the investment.The positive influence of successor on the choice of biomethane upgrading in the binomial logit model suggests that operators with secured succession are more willing to go down that route.By contrast, operators who stated the need for financial support are less likely to choose biomethane upgrading.In contrast to the free market of the biomethane sector, the EEG as the only available support program offers a safety net for electricity generation in the form of a guaranteed feed-in tariff determined at auction.On the other hand, operators who choose the conversion of the plant for biomethane upgrading may not believe that state subsidies are necessary or possibly wish to avoid dependence on a state subsidy program.Operators who took the survey after the outbreak of the Ukraine conflict are less likely to choose biomethane upgrading, which may be due to the dynamic development of the energy markets since February 2022.
The mixed logit models show overall high investment probabilities of 71% for flexibilization and 82% for biomethane upgrading and reflect a strong willingness to further extend the mainstay of biogas production and its importance for the farm income.The expected return on investment is high and economic considerations are decisive.The WTA of €0.19 implies an imputed return on investment (ROI) of 19%.For the biomethane investment the WTA of €0.26, represents an even higher required ROI of 26%.This is likely to reflect the higher risk involved in changing the utilization path (adoption of a less wellknown technology) and operating in the volatile biomethane market in the absence of the safety net of the EEG.Operators therefore include a higher risk premium in their investment return.This finding corroborates the results of other studies which have demonstrated that investors in the renewable energy sector demand a high return on their investment (Gamel et al.,  T A B L E 6 Comparison of possible annual profits and WTA of flexibilization levels (€/kW rated power).et al., 2012;Salm et al., 2016).However, the participants' selection behavior suggests that major changes in operational strategy are also accepted, provided they are financially viable.Within the flexibilization, PQ 2 is clearly preferred to PQ 5.The high additional costs of PQ 5 provided by the literature offer an explanation.Haensel et al.'s (2020) calculation reveals additional costs between PQ 2 and PQ 4 of €205.8 to €249.5 per kW rated power.According to own calculations, the difference between PQ 2 and PQ 5 is €171 per kW rated power.If the estimated WTA is deducted from the possible profits, it can be seen that the 'perceived profit' of PQ 5 is well below the corresponding figure for PQ 2 (Table 6).This may explain the low number of biogas plants that have hitherto invested in PQ 5 flexibilization.
For the biomethane upgrading, private ownership of the upgrading plant is preferred to a joint investment in a central upgrading facility.Highly significant coefficients indicate a strong desire for the follow-up use of the plants.
A closer look at the marginal effects reveals a preference for investing in one's own biomethane upgrading plant than in a collective upgrading facility with other operators.The preference for individual ownership of upgrading plants could be explained by transactions costs arising from complex coordination and management procedures and possible quarrels over the distribution of economic benefits among the shareholders of a jointly operated upgrading facility.These factors can increase the potential for conflict and thus act as deterrent to cooperation (Zemo & Termansen, 2018).The results are consistent with the literature, which identifies the flexibility to exit an investment as easily as possible as the most valuable attribute (Gamel et al., 2016;Zemo & Termansen, 2018).While the divestment of shares of a plant owned by several participants is relatively easy to handle, the separation of one's own plant from the joint biomethane upgrading via a biogas grid cannot be done without further ado as, for example, the cost structure of all other plants involved is affected.
Limiting the use of energy crops reduces the propensity to invest, while a longer utilization period enhances it.While in the flexibilization case the end of electricity generation is legally fixed by the EEG, the effective lifetime of a biogas plant can be extended through power purchase agreements.For biomethane upgrading, generation beyond the contract period is conceivable and probable, so that longer payback periods reduce the risk of financial failure.
While the binomial logit indicates a higher choice probability for biomethane upgrading for operators who rate themselves in the lower 25% bracket, this characteristic reduces the investment probability in both the biomethane and flexibilization model.It is possible that these operators extended to expect problems in plant operation, which makes them reluctant to invest.
Following the Ukraine conflict the newly emerged high volatility of the natural gas market may deter some operators from choosing the biomethane upgrading.By contrast, extended electricity generation, like the gas market, offers high revenue opportunities but at reduced risk since the EEG remuneration offers a price floor.The wholesale prices for electricity have risen from €74 per MWh in June 2021 through €128 per MWh in February 2022 and €252 Source: Own illustration.

| 17 of 22
MWh in May 2022 to €465 per MWh in August 2022.This dynamic development in combination with the safety net offered by the EEG may tip the decision for plant managers to invest in flexibilization rather than biomethane upgrading.
Whereas the feeding-in of biomethane into the natural gas grid offers the most versatile options for its use, flexible on-site generation electricity is likely to gain importance as the conversion to a carbon-neutral energy sector progresses.The biggest advantage of biogas plants is the temporal controllability of the electricity supply.The rapid availability and storability of the energy carrier makes it possible to compensate for the temporal fluctuations of the other renewable energy sources, thus increasing the share of biogas in the energy system.A maximum of flexibility is therefore particularly beneficial to the system, but does not receive the necessary support.As the results of the study suggest, the perceived profit of PQ 5 is well below the PQ 2, and plant operators clearly prefer the less ambitious flexibilization.A significant increase in the subsidy rates for particularly high flexibility is therefore necessary to make up for the difference and to achieve acceptance of the highest possible power quotients.This calls for the flexibilization subsidy to be differentiated by flexibilization ambition, with higher subsidy rates being offered for greater power quotients.
Another policy conclusion that can be drawn from the analysis is that follow-up support for highly flexibilised biogas plants should be extended beyond the current tenyear period.This follows directly from the observation that a longer useful life increases operators' willingness to invest.With Germany's commitment to become carbon neutral by 2045, strong investment in high flexibilization seems important to enable the biogas sector to gradually take over the role of CO 2 -emitting gas power plants in supplying positive control energy.
As respondents consider lack of planning reliability as a key obstacle to invest, followed by long approval procedures and high investment costs due to restrictive legal requirements, simplification of the legal framework could contribute significantly to the further development of biogas use.So far, the majority of plants have had to go through the complex process of obtaining new permits.Therefore, an extension of the technical grandfathering to biogas plants in case of investments in a subsequent use, which prevents the upgrade to the plant technology of new BGP, could be a very effective policy measure.
on the maximum bid value of the EEG tender as of 2021 (18.4 ct/kWh), the annual flexibility surcharge of 65 €/kW (equivalent to 1.55-3.88ct/kWh), as well as possible additional revenues from demand-side marketing (0.48-0.87 ct/kWh)(Schröer & Latacz- Lohmann, 2022).The costs of on-site electricity generation were derived from data of the Agricultural Accountancy Association for the 2014/15 to 2018/19 marketing years

F
I G U R E 1 Structural procedure of the survey.Source: Own illustration.
specific Invest flexibilization = 44.875* installed capacity −0.566 , (2) specific Invest upgrading = 90.705* upgrading capacity −0.547 .| 7 of 22 of the dummy variable, for example top 25 = 0.47 means that 47% of the respondents would place themselves in the top 25% of the company comparison.b Only the animal numbers of farms keeping this type of animal are included in the calculation of the mean values.Source: Own calculation.

T A B L E 3
Binomial logit model to explain the choice of biomethane upgrading (reference = flexibilization).T A B E 4 Mixed logit model, marginal effects, and willingness to accept (WTA) for operators preferring flexibilization.
profits increase the willingness to invest.✓ ✓ H2-As capital requirements increase, the willingness to invest decreases.✓ ✓ H3-PQ 2 leads to a higher willingness to invest, while the passive flexibilization, or PQ 5 reduces it.✓n/a H4-The possibility of central upgrading facility by means of a gas collection pipeline increases the willingness to invest.n/ax H5-A longer utilization period or contract duration increases the willingness to invest.✓ ✓ H6-Alternatives with a higher share of energy crop input lead to a higher willingness to invest.✓ ✓ H7-Farms with high availability of sustainable substrates (i.e.those with high livestock densities or availability of other farm residues) are more likely to invest.o o H8-Farms facing difficulties in manure management (e.g.farms located in nitrate-sensitive areas, farms without sufficient slurry storage capacity, or farms that have to pay for digestate disposal) are less likely to invest.o o H9-Operators of higher-capacity biogas plants are more likely to opt for the biomethane upgrading.✓ ✓ H10-The majority of participants choose the flexibilization.✓ ✓ Note: Hypothesis could not be rejected (✓), partly rejected (o), rejected (x).

T A B L E 1
Characteristics of the hypothetical choice alternatives.
Note: Bold: Expression for non-participation, that is no investment in either of the two options (opt-out).
2016; Reise, Mußhoff, 18.4 ct/kWh maximum possible bid of the EEG auctions for the year 2021. c