A Top-Down Approach to Evaluating Cross-Border Natural Gas Infrastructure Projects in Europe

There is an ongoing policy debate in Europe about how to select natural gas infrastructure projects for an EU-wide investment support scheme. We contribute to this debate by introducing a model-based project evaluation method that addresses several shortcomings of the current approach and demonstrate its application on a set of shortlisted investment proposals in Central and South Eastern Europe. Importantly, our selection mechanism deals with the complementarity and the substitutability of new pipelines. We ﬁnd that a small number of projects are sufﬁcient to maximize the net gain in regional welfare, but different baseline assumptions favor different project combinations. We also explore the consequences of Russian gas permanently delivered at the EU border from northern and southern routes that bypass Ukraine and ﬁnd modest negative welfare effects.


INTRODUCTION
Cross-border investments in electricity and natural gas transmission networks raise unusual challenges for cost-benefit analysis.Trading on a new interconnector narrows price differences across markets, making consumers better off on one side, but worse off on the other.Producer surplus moves in the opposite direction, while storage units (in case of gas) can either gain or lose profits, depending on seasonal demand-supply patterns.Market participants in directly and indirectly connected third countries are also affected.Moreover, the new infrastructure element might only be needed in case of a serious (but rare) supply disruption.These nuanced complications can forestall neighboring transmission system operators (TSOs) and governments from investing into projects that would otherwise be beneficial for the region as a whole.
Europe, with over 30 interconnected national electricity and gas markets, is ripe for divergent bilateral and regional interests over cross-border transmission investment.Aware of the threat of underinvestment in market interconnection, the European Union (EU) established a system for supporting energy infrastructure projects deemed essential for a fully integrated European market and for security of supply.These projects of common interest (PCIs) may benefit from accelerated 3. The welfare measure is calculated as an unweighted sum of surpluses, whereas national governments might assign unequal weights to different stakeholders.In theory, favoring certain groups (consumers, for example) can decrease the willingness to build a pipeline that harms those groups, but still brings positive net benefits in aggregate.
4. The original role of long-term gas purchase contracts and the underlying fundamentals are described by Neuhoff and  Hirschhausen (2005), Asche et al. (2002), Stern and Rogers (2014) and Henderson and Pirani (2014).
5. Hence our intention is not to model a supply disruption scenario, but rather a permanent shift to alternative delivery routes and destinations.Richter and Holz (2015) carry out simulations for disruptions to transits through Ukraine without assuming a new southern route, and find substantially more severe effects for a similar set of countries as we do.
6. Storage and transmission system operators have exogenously given fees that exceed marginal costs.7.As a result, we have nothing to say about the price development of long-term supply contracts.Models with a strategic upstream sector do address producer price setting, although they tend to abstract away from the take-or-pay clauses.For a more in-depth endogenous contracting model, see Abada et al. (2014).
Somewhat contrary to our expectations, we also find that most of the welfare gains from new investment tend to accrue to the countries building the pipelines, in some cases even making third countries slight net losers.This reflects the limited number of scenarios considered, but it also raises questions about the exact mechanism through which the PCI scheme improves welfare for the EU itself. 3s a secondary application, we also demonstrate the advantages of the EGMM's treatment of long-term supply contracts in conjunction with the proposed PCI evaluation mechanism.We explore a hypothetical baseline scenario in which the contract delivery points of gas arriving from Russia are moved to the member state border where the gas enters the EU, and all transit routes from the east avoid crossing Ukranian territory. 4Although this is a large shift from the current mode of operation, we find that its negative welfare consequences for the CSEE region are modest.However, this result significantly depends on assumptions about a (not-yet-existing) southern route becoming available for Russian transit. 5The new baseline also shuffles the project combination rankings, moving a north-to-south supply route from Poland at the top of the list.Implementation of the proposed PCIs has the potential to neutralize around 30 percent of the welfare decline in the region.We also discuss other potential benefits, such as the improvement of buyer bargaining position, after modeling results are presented.
Our approach is different from most models (except for TIGER) that allow for strategic behavior by upstream firms, and in some cases also in the downstream market, in that we assume all market participants are price takers. 6Working with a perfectly competitive equilibrium has drawbacks, as the presence of market power is an important issue in the upstream market.We remedy this shortcoming with the inclusion of a detailed representation of long-term take-or-pay supply contracts in the model.Our assumption is that most of the upstream market power is exercised through the pricing of these contracts with the market working more like the competitive benchmark in the short run.Accordingly, a single simulation run of our model encompasses one calendar year. 7he price-taking assumption allows us to go into finer detail both geographically and in the temporal dimension.We aggregate demand at the country level, but not across countries.The modeling literature often lumps small neighboring markets (e.g. in South Eastern Europe) together for computational convenience, which rules out the analysis of interconnection between these markets.We also break the modeled time frame into monthly periods, whereas the cited models typically include only 1-3 seasons per year.The monthly output is especially helpful when examining market disturbances, which are typically short-lived, and the extent to which storage units can mitigate them depends on short-run gas withdrawal capacities.In terms of geographical and temporal granularity, our model is close to-but still less detailed than-the TIGER model. 8However, we do allow for price responsive demand functions and can therefore carry out a more informative welfare analysis.The detailed take-or-pay contracting structure in the EGMM also allows us to examine the effect of virtual reverse flows on market integration with limited physical connectivity.
We use a comparative static framework for our project evaluations, contrasting equilibrium outcomes with and without the investments, which could-in theory-also be carried out with other cited models that are sufficiently detailed.Some models even go beyond the static approach and allow for infrastructure changes within the time frame of the simulations.Lise and Hobbs (2008)  extend the GASTALE model to automatically include new pipelines and storage units whenever the forecasted congestion rents exceed a specified threshold value. 9In the Global Gas Model, transmission and storage system operators decide about new investments based on a private costbenefit analysis.
Making the investments of profit-oriented yet geographically limited entities in interconnected markets endogenous is not straightforward.It is not clear, for example, how the substitutability or complementarity of new pipelines can be modeled if the investment decisions are taken project-by-project by non-overlapping (or partially overlapping) sets of TSOs.In reality, policy makers with their own objectives also need to consent to the investments, which complicates a profit-based cost-benefit analysis. 10Finally, even the parties to a single pipeline can have divergent dynamic incentives, which leads to complex bargaining problems and potentially wasteful spending. 11Using the comparative statics approach, we have to make assumptions about what infrastructure will be available in the future, in exchange for a more transparent analysis that avoids the shortcomings of endogenous investment models.Since we focus on a relatively short time frame, we find this trade-off acceptable.
The remainder of the paper is organized as follows.We start with a detailed description of the EGMM, our modeling tool, followed by a section on our project evaluation methodology.We then provide the main modeling results and a ranking of the most beneficial investments for each baseline scenario.In the last part of the paper, we discuss the general lessons from our project evaluation exercise, point out potential pitfalls, and conclude.

MODEL
The EGMM is a competitive, dynamic, multi-market equilibrium model for natural gas production, trade, storage, and consumption in Europe.It explicitly includes a supply-demand 12. 26 member states of the European Union (all except for Cyprus and Malta) and the following non-members: Albania, Bosnia and Herzegovina, Macedonia, Moldova, Serbia, Switzerland, Ukraine.
13. Any number of links between two markets are possible in each direction.14.Because of take-or-pay obligations, gas sometimes flows from high-priced to low-priced markets.Even if there is no physical connection in the reverse direction, the price difference can be profitably exploited by backhaul shipments.representation of 33 European countries, 12 as well as their gas storages and transportation links to each other and to the outside world.The time frame of the model is 12 consecutive months, starting in April.Market participants have perfect foresight over this period.

Market Participants
There are four kinds of players within the model: consumers, local producers, importers, and traders.Consumers in each market within the region are represented by a linear monthly gas demand function that only depends on the contemporaneous local wholesale price of gas.For outside markets, we use perfectly elastic demand functions at exogenously given prices.
Local producers have piecewise linear short-run cost functions, with upper and lower limits on monthly production and a separate upper constraint on yearly output.
Importers own long-term take-or-pay (TOP) contracts that are sourced from gas exporters in outside markets, most importantly from Russia, Norway, Algeria, and a number of LNG exporting countries.Each contract specifies a price, a delivery route, and a minimum and maximum delivered quantity per month and per year.The monthly minimum delivery constraint alone is flexible: it can be violated, but most of the undelivered gas must be paid for according to the TOP rules.
The function of traders is to move gas between markets using cross-border pipelines and LNG shipments, and between time periods using storages.

Transportation
Transportation links between two markets consist of cross-border pipelines and LNG routes.From a modeling perspective, they operate similarly.Each link is unidirectional 13 and has a maximum transportation capacity.Pipelines have entry and exit tariffs, and LNG connections have shipping costs, regasification prices, and network injection fees.
Transportation infrastructure can be used for delivering gas into the target market through long-term contracts or spot trade, and into the source market via virtual reverse flow ("backhaul").Virtual reverse flow can only be carried out if there is already a pre-contracted gas flow in the default direction. 14Even then, its availability may be limited by the legal environment, which is captured as an additional constraint in the model.
Formally, physical flows on a piece of transportation infrastructure are given by: where indexes the infrastructure, the time period (month), and the long-term contracts.is s f s c t f the amount of spot trade and is the amount of backhaul shipment on infrastructure . is the total delivered quantity on contract , and specifies what portion (typically 0 or 1) of contract s c X fc flows through infrastructure in month .c f s We also include in the model upper constraints on linear combinations of physical flows to separately account for the regasification capacity of LNG terminals and the total shipment capacity of LNG exporters. 15

Storage
Storage units reside within 23 out of the 33 markets we explicitly model.Each has a monthly injection and withdrawal capacity and associated fees.In addition, storages have a working gas capacity and exogenously given starting and year-end inventory levels.Traders can decide in each month how much gas they want to inject into, or withdraw from, a storage unit.Inventory levels can never fall below zero or rise above the working gas capacity of a facility.Storages must eventually be reloaded to the specified year-end inventory level (usually equal to the starting inventory).

Decision Variables
Local producers determine production levels ( ), importers determine the quantity of s e p delivered gas up to ( ) and above ( ) the monthly TOP quantities, and traders determine spot as participants in the model, their consumption is derived from the other decision variables as: where indexes markets, and and are zero-one parameters indicating whether producer m P C mp mg and storage are in market .is also an indicator: it equals if market is the target of p g mU 1 m mf transportation infrastructure if it is the source, and otherwise.f, -1 0

Equilibrium
A crucial assumption in the EGMM is that producers, importers, and traders are all pricetakers.In equilibrium, all arbitrage opportunities across time and space are therefore exhausted to the extent that storage facilities, transportation infrastructure, and contractual conditions permit.As a result, the competitive equilibrium yields an efficient outcome and can equivalently be computed as the solution to a constrained welfare maximization problem with the following multi-period objective function: 17. Uniqueness might not hold, for example, if there is more than one unconstrained marginal producer in the same location.Since marginal production costs are assumed to be constant, any feasible distribution of the joint output across the equal-cost producers is part of an equilibrium outcome, even though all such equilibria lead to the same market price.This multiplicity is not an issue in practice, however.
18.The full set of complementarity conditions are available from the authors upon request.
In the welfare expression, is the month-to-month discount factor, is the (linear) inverse  limit of long-term contracts. 16he objective function is continuous and weakly concave with a non-empty, closed and convex domain, ensuring the existence of an equilibrium.With minor additional assumptions, the equilibrium also becomes a unique one. 17We find it by solving the first-order linear complementarity conditions of the constrained optimization problem using standard pivoting techniques (Gabriel et  al., 2013). 18

Welfare
We use storage and transportation fees in our objective function (3) as if they were the marginal costs of these activities, even though the fees typically exceed the marginal costs to ensure a sufficient return on capital invested into the infrastructure.Similarly, using a low marginal price for the gas bought below the monthly delivery limit in a long-term contract does not reflect the full cost of the gas, because the take-or-pay obligation creates a large fixed cost element as well.Despite these qualifications, the welfare formulation in expression (3) is the one that allows us to replicate the competitive equilibrium.
In our ex post welfare calculations, we adjust the maximized value of our objective function to properly account for actual welfare in the market.We add the operating profit of transmission and storage system operators using estimates for their marginal costs, and increase the expenditure on import contracts by the take-or-pay fixed cost element.The augmented welfare used for evaluating the benefit of new infrastructure is therefore given by: long-term contract.All numerical welfare values in the paper are the result of evaluating expression (4) at the competitive equilibrium with given input parameters.
We assign all welfare components to regional and outside markets based on location.For consumer and local producer surplus, long-term contract profit, 19 storage operating income and congestion rent, the assignment is straightforward.Pipeline operating income is shared in the ratio of entry and exit fees and pipeline congestion rent is shared equally by the neighboring markets.LNG-related welfare components are assigned to the market hosting the terminal.

METHODOLOGY
We evaluate the social benefit of new cross-border pipelines by running market simulations with and without the new infrastructure and comparing aggregate welfare in the two equilibria.All simulations are for the year 2020 using forecasted values for demand, indigenous production, and long-term contracts in place. 20In our baseline ("before investment") scenarios, we include all existing infrastructure, as well as all current projects that already have final investment decisions and are planned to be commissioned before 2020. 21Most important among these for our analysis is the Trans Adriatic and the Trans Anatolian Pipeline (TAP/TANAP) connecting Turkey with Italy through Greece and Albania. 22e use three baseline scenarios for our evaluations.The first one is a business-as-usual case (Reference) with no additional assumptions.The second baseline scenario explores the market consequences of Russian gas being delivered at the border of the EU on routes that avoid Ukraine (No Ukraine transit).Finally, we examine the effects of a new LNG terminal in Croatia on the usefulness of cross-border pipeline investments in the region (Croatia LNG).
In each case, we proceed by adding to the baseline setup all possible combinations of cross-border pipelines, chosen from a shortlist of proposed projects of common interest in the North-South Gas Corridor in the CSEE region, 23 and comparing the changes in our welfare measure relative to its baseline value.Evaluating the benefit of all project combinations, as opposed to single projects in a selected order, allows us to capture both the complementarity and the substitutability between pipelines.The shortlisted projects are shown in Table 1.
Our main result is a list of the most beneficial project combinations based on welfare improvement net of annualized investment costs.We estimate investment costs using publicly available information on the projects (pipeline length and diameter) and benchmark unit investment costs published recently by the European Agency for the Cooperation of Energy Regulators 2.8 † A pipeline currently operates from Hungary to Croatia, which could be physically reversed with a single compressor station on the Croatian side.A new 308 km internal pipeline may, however, be necessary to transport larger quantities of gas from the coast to Hungary if the LNG terminal in Croatia is built.We therefore calculate assuming this higher investment cost in the Croatia LNG scenario, but not otherwise.24.Our cost figures are only imprecise estimates, since there is considerable (typically ) variation around the ‫%03ע‬ mean benchmark values in the cited dataset.In an actual policy application, the numbers can be substituted by the cost estimates of the project promoter, for example.
25. Belarus, Kosovo, and Montenegro are not included in the model.
(ACER).For example, the average unit investment cost is 1.06 million €/km for a pipeline with diameter 28Љ-35Љ (  , 2015).Calculating with a length of 185 km and a single 18 MW compressor station (a typical size in the ACER database), the Greece-Bulgaria pipeline has a total investment cost of €234 million.This is equivalent to €15 million annually, assuming a 25-year lifetime and a 4% yearly discount rate (recommended parameters by ACER, 2014).Table 1 lists our estimates of the annualized investment costs for all the projects we include in the analysis. 24

Reference Scenario
The stylized map of Europe in Figure 1 provides an overview of market prices in the Reference baseline scenario.It shows the modeled markets with a shaded background and the exogenously-priced outside markets (Norway, Russia, and Turkey) in white. 25Consumptionweighted yearly average wholesale gas prices (in €/MWh) are displayed for a selected set of countries.Darker shaded areas indicate higher prices.
Markets in South Eastern Europe, including even Croatia and Hungary, have the most problematic supply-demand situations.The region lacks sufficient interconnection to Western Europe for cheaper gas to enter and push down prices.Greece is an exception, due to access to LNG and the Trans Adriatic Pipeline, but markets to its north see little of these benefits.Surveying the potential projects in Table 1, we would therefore expect the GR-BG, BG-RS, and HR-RS pipelines to allow for the largest increase in welfare as a result of increased trading and price equalization.
Starting from the baseline scenario depicted in Figure 1, we add all 127 possible pipeline combinations from Table 1 and find the competitive equilibrium in each case.We then aggregate social welfare (as measured by equation ( 4)) across all Central and South Eastern European markets,

If building Pipelines
A and B has a positive ANSB, but building Pipeline A by itself has an even higher ANSB, then we drop the two-element combination, because it is more costly and less valuable than the single pipeline investment.In the reverse case (when B adds value to A), we keep both pipeline combinations in the list, because financial constraints might, for example, prevent the A + B investment, but allow for the less costly A investment to proceed.and calculate the annual net social benefit (ANSB) of an investment by subtracting the estimated annualized investment cost from the change in our welfare measure between the baseline and the after-investment scenario.We rank all pipeline combinations in decreasing ANSB order, and drop the ones that are dominated by another investment with a subset of the pipelines. 26able 2 shows the main result of our simulations for the Reference baseline.Each column lists the pipelines included in the investment scenario, followed by the social cost-benefit balance.There are only two pipeline combinations with a positive net social benefit.The route from Greece to Serbia through Bulgaria is the most valuable investment, followed by a single pipeline from Greece to Bulgaria.The Romania-Hungary pipeline also yields some additional welfare for the region, but not enough to justify its cost.The pipelines in the 4th and 5th place do not increase welfare at all, because they connect markets with a price difference smaller than the cost of transportation.
In Table 3, we take a closer look at price and surplus changes resulting from the Greece-Bulgaria-Serbia pipeline investment.The first two columns show the before-after market prices, followed by changes in consumer surplus, producer surplus, net profit from long-term contracts, and infrastructure income.The latter category includes operating income and congestion rent on pipelines, LNG terminals, and storage units.We list the markets that are substantially affected in some dimension in separate rows, and add an Other CSEE line for the total welfare change in the region to match with the value in Table 2.
The largest price decrease is in Bulgaria, closest to the low-priced Greek market.Consumers in Hungary, Romania, and Serbia also benefit to some extent, while the rest of the region remains largely unaffected.Overall, consumers gain about €270 million, which is partially mirrored by losses on the producer and long-term contract holder side.The net effect on infrastructure income 27.In the past, natural gas disputes between Russia and Ukraine have led to supply disruptions to consumers in Central and South Eastern Europe.The aborted South Stream and Turkish Stream projects have both sought to displace Ukraine as a transit country.In this scenario, we assume that a similar project in the south might eventually be completed, although recent political tensions between Russia and Turkey make this unlikely to happen until 2020.
28.In the welfare expression, trading profit is actually turned into congestion rent collected by TSOs and SSOs.29.Detailed welfare results are available in the Online Appendix which can be accessed at http://bit.ly/2eNMpLw.
is neutral, although the area gets more of its gas from the south, which redistributes transportation profits from Hungary and Slovakia to Bulgaria and Greece.In the end, the countries benefitting most from the investment are the ones with jurisdiction over the territory where the pipelines run.

Change of Russian Long-term Contract Delivery Points
In this section, we explore hypothetical baseline and investment scenarios in which the contract delivery points of gas arriving from Russia are moved to the country where the gas enters the European Union.We will further assume that all transit routes from the east avoid crossing the territory of Ukraine.Instead, three other routes will be used: the Nord Stream pipeline with its current capacity under the Baltic Sea to Germany, the Yamal pipeline through Belarus to Poland, and a not-yet-existing pipeline through Turkey to Bulgaria.For simplicity, we will refer to this set of scenarios as No Ukraine transit. 27e assume that all re-routed Russian long-term contracts are delivered to Germany, Poland, or Bulgaria.Specifically, the gas currently destined to Austria, France, Hungary, and Italy will first be sold in Germany, the Czech, Dutch, and Slovakian contract will be sold in Poland, and the gas transported through the Trans-Balkan pipeline will instead arrive through Turkey in Bulgaria.We calculate the revenues of contract holders as if they sold all the deliveries in these three countries, and allow traders in the spot market to profit from distributing the gas across the continent. 28n the spirit of a fair contract renegotiation, we assume that the (implied) wellhead prices of contracts are changed to reflect the difference in transportation fees between the new and the old delivery routes.As a result, if the gas found its way to its former target market, its border price (wellhead price + transportation cost) would be the same as before.
Figure 2 shows the equilibrium outcome of the No Ukraine transit baseline scenario.It uses a map of Europe similar to Figure 1, but the displayed numbers now represent differences in yearly average prices relative to the Reference baseline.Lighter shaded areas experience price decreases, darker shaded ones see their prices rise.
Changing contract delivery points does change market prices across the continent, although only mildly.Prices in Italy and most of Central and South Eastern Europe rise, while consumers in Western Europe become slightly better off.Prices in the smaller designated delivery countries, Poland and Bulgaria, are especially depressed.Overall, Figure 2 suggests that a lack of cross-border capacity prevents some of the gas from reaching its nominal destination.
Taking the No Ukraine transit baseline as a starting point, Table 4 shows the pipeline combinations from Table 1 that are most beneficial to the CSEE region.Connecting Bulgaria to Serbia and Poland to the Czech Republic and Slovakia is the most valuable investment with an ANSB of €195 million.All other projects in the top 5 list are variations on the same idea: letting surplus gas "trapped" in Poland and Bulgaria find its way towards the markets that have previously been supplied through Ukraine.
Welfare calculations show that changing the long-term contract delivery point results in a welfare loss of about €1.5 billion/year relative to the Reference baseline for all modeled countries together. 29Two-thirds of the loss is borne by Ukraine because of the re-routed transits.Contract One might argue that a change in contract delivery points affects the entire European continent to some extent, which will likely spark new investment projects other than the ones listed in Table 1.In this case, the rearrangement of transit routes will likely result in even smaller welfare losses for gas importing countries (although not for Ukraine).
holders also have to forgo about €1 bn/year by selling gas in areas with depressed prices.Consumers, on the other hand, end up being better off across the continent on average, even if the gains are unevenly distributed.Specifically, Italy, Ukraine, Romania, Hungary, and Austria are among the countries whose consumers lose most from the change, although this is more than offset by gains in Poland, Bulgaria, Germany, the Netherlands, and France.The overall losses in the CSEE region amount to €680 million/year.Around 30% of this amount can be avoided by carrying out the most beneficial pipeline investment indicated in Table 4. 30

LNG terminal in Croatia
The third baseline scenario simulates the completion of a long-planned LNG regasification terminal with a capacity of 6.5 bcm/year on the Adriatic coast of Croatia.By itself, the LNG terminal has a modest effect on the CSEE region.Average prices decrease by €4.8/MWh in Croatia relative to the Reference baseline (welfare increase of €80 million/year), but almost none of it flows over to neighboring markets due to the lack of interconnectivity.As a secondary effect, competition for LNG sources results in a loss of €20 million/year for Greece.Overall, the LNG terminal in Croatia brings a surplus of €48 million/year to the CSEE region without additional investments.
The right-hand side of Table 4 shows the investments with the highest social return in the Croatia LNG baseline.By far the most valuable project combination is the pair of pipelines from Croatia to Hungary and Serbia, which would increase regional welfare by €124 million/year, net of investment costs.Table 5 shows a breakdown of surplus changes in the countries most affected by the new infrastructure.
Croatia benefits most from the two new export pipelines.Its income from the LNG terminal and the cross-border infrastructure rises by €173 million/year, largely consisting of profits from LNG imports.Consumers in Serbia and Hungary also derive massive benefits from indirect access to the LNG market.At the same time, producers and (especially) contract holders suffer some losses due to decreased prices.Pipelines carrying gas from the north towards Hungary and Serbia are also less heavily utilized.Similarly to the Reference baseline, the countries that derive most of the benefits from the projects are the ones involved in the investment decision.

DISCUSSION
In this section, we discuss the lessons from our modeling exercise and point out potential pitfalls.

Welfare-based Regional Project Evaluation
Our starting point was a Reference baseline scenario for 2020, depicted in Figure 1.In our model runs, the central areas of South Eastern Europe have the highest prices.Thus pipelines bringing gas into this region from lower-priced markets have the highest potential for welfare improvement.
The projects on the shortlist in Table 1 offer three distinct directions in which the region, roughly around the core of Serbia, could be supplied.The first one is from the north using expanded capacity from Poland to the Czech Republic and Slovakia; connections from Austria and Slovakia to Hungary already exist.Our simulations suggest that this direction would not yield welfare improvements, because the three countries to be connected already have similar prices.
The second direction is from the south, exploiting the backhaul capabilities of the newly built Trans Adriatic Pipeline and low-priced LNG imports to Greece.In the Reference scenario, this is the most promising alternative, suggesting that the Greece-Bulgaria and Bulgaria-Serbia interconnectors could bring reasonable net welfare improvements across the region, especially in these three markets.
The third direction is from the west, using cross-border pipelines from Croatia to Hungary and Serbia.These projects could substantially increase welfare in the region, but only in connection with a (long-planned) LNG regasification terminal in Croatia that brings in new gas sources.Otherwise, prices in the three markets set to be connected are too similar for the new infrastructure to be worth building.
The welfare analysis in Tables 3 and 5 also illustrates the advantages of a model-based project evaluation.By comparing equilibrium outcomes with and without the new infrastructure, we can provide a nuanced picture of which market participants are likely to gain and lose from the investments across the greater interconnected system.It is possible, for example, that third countries receive indirect positive effects from an investment that would not bring enough direct benefits to its sponsors otherwise.In this case, regional subsidies can be welfare-improving.The opposite outcome, where direct effects are large enough but indirect effects are negative, can also occur.In fact, many of our results fall into the second category, questioning the exact rationale for subsidizing the investments in Table 1.
We have also analyzed the net social benefits of PCIs inclusive of all modeled European countries in our welfare measurement, rather than the CSEE region alone.It turns out that some of the welfare gains in CSEE are actually a redistribution from other EU countries, tilting the costbenefit analysis against the new investments.In the Reference baseline, all investments yield net welfare reductions (the Greece-Bulgaria pipeline is only marginally unprofitable).The most beneficial project combinations in the other two baselines are also worthwhile investments from an EU point of view.
To check for robustness, we have performed the analysis assuming lower demand for natural gas in 2020.(To ´h et al., 2014) A low-demand baseline leads to more uniform prices across the continent, decreasing the potential benefit of additional pipelines.Accordingly, we find that no project combinations result in a positive ANSB in the Reference baseline, although both the Croatia-Hungary and the Romania-Hungary pipelines are on the verge of profitability for the region.In the Croatia LNG scenario, the Croatia-Serbia interconnector proves to be a robust project.
Our investment analysis is not without a number of caveats.In an actual policy application, one would extend the analysis for several years (possibly decades) after 2020, and examine alternative assumptions about our exogenous parameters (e.g., LNG prices, long-distance transportation infrastructure, contract characteristics).Ideally, there would be some feedback over time from shortrun model outcomes into the assumptions driving the results in subsequent years.Such an analysis is beyond the purposes of this paper, however.
The competitive nature of our modeling approach does not allow us to analyze the effects of new investment on the market power of upstream, downstream, and infrastructure firms either.Such effects might be relevant, especially at the time of long-term contract negotiation.We also omit the analysis of short-run disruption scenarios, even though some of the cross-border pipelines could bring considerable benefits in extreme supply shortages.Our results should therefore be taken as an illustration of a regional welfare-based investment (or subsidy) decision method, rather than a definitive list of recommended projects.

Modified Delivery Routes for Long-term Contracts
The second contribution of this paper is the analysis of a large-scale change in the delivery routes and destinations of long-term contracts from Russia.In our hypothetical (but not entirely unrealistic) situation, no gas arrives via Ukraine, resulting in a large loss of transit income for the country.The effect on welfare in the European gas sector is less drastic.The redistribution of welfare from Eastern to Western Europe is more noticeable than the overall loss resulting from the increased geographic mismatch of supply and demand.A substantial part of the negative effect can even be mitigated by north-to-south pipeline investments within the European Union.
Again, our conclusions are derived from a single baseline scenario, and a more nuanced picture would emerge from the use of a multitude of alternative assumptions.The necessary renegotiation of contract terms could, for example, lead to somewhat different outcomes in terms of contract prices and/or delivery locations.
There are two additional aspects worth discussing.Contrary to our assumptions about the border (destination) prices of contracts remaining the same as before, one could imagine that the current system of price discrimination by destination country would disappear if all contracts were to be delivered at an EU entry point.Although the CSEE region loses welfare when transits through Ukraine stop, it would benefit from having its long-term supply contracts priced according to the market realities in Germany, for example.
With the change in delivery points to the EU border, the long-term capacity bookings on intra-EU pipelines (such as the Poland-Germany, Slovakia-Austria, or Austria-Italy pipeline) could also be freed up and made available for spot trade.These pipelines could even carry gas in the reverse direction with moderate additional investment, further contributing to the emergence of an integrated European gas market.

CONCLUSION
In this paper, we have introduced a competitive dynamic natural gas market model for Europe and applied it to evaluate infrastructure projects in Central and South Eastern Europe.Project rankings are based on the resulting net increase in regional welfare, taking into account all possible combinations of pipelines to be built from the shortlist of proposed PCIs.In a separate baseline case, we also used the EGMM to explore the consequences of changing the delivery points of longterm gas supply contracts from Russia to bypass the territory of Ukraine.
The goal of the paper is to outline an alternative project selection methodology, and not to provide an exhaustive analysis of costs and benefits.For reasons of conciseness, we have for example omitted the simulation of actual supply disruption scenarios.
One important element missing from our paper is the feedback from infrastructure investment into the pricing of long-term contracts.The destination-based price discrimination currently seen in take-or-pay contracts is largely consistent with the lack of alternative supply sources in countries that receive Russian gas at higher prices.As new pipelines bring gas from lower-priced regions, the bargaining position of these countries will also improve.Accounting for these effects in contract pricing is an important area for additional work on our modeling approach.
gas storages ( indexes the producers and the storages).Although consumers are listed prices of gas below and above the minimum monthly delivery

Figure 1 :
Figure 1: Weighted Average Market Prices for 2020 in the Reference Baseline Scenario

Figure 2 :
Figure 2: Price Changes from the Reference to the No Ukraine transit Scenario function in market and period , and the (quadratic, concave) integral term measures

Table 3 : Major Price and Surplus Changes from Best Investment in the Reference Baseline (GR-BG, BG-RS)
Prices are in €/MWh, all other values in million €/year.

Table 4 : Ranking of Investments in the No Ukraine transit and the Croatia LNG Baseline
Welfare changes and net social benefits are valid for the CSEE region.