Tax shifting in long-term gas sales contracts *

Transcription

1 Tax shftng n long-term gas sales contracts * Frank Asche *, Petter Osmundsen** and Ragnar Tveterås** * Stavanger Unversty College / Norwegan School of Economcs and Busness Admnstraton ** Stavanger Unversty College / Insttute for Research n Economcs and Busness Admnstraton December 2002 Abstract Producers or consumers faced wth an ncrease n taxes are usually able to shft parts of t to other levels n the value chan. We examne who are actually bearng the burden of ncreased taxes on natural gas n the EU-area - consumers or exporters. Strategc trade polcy and cross-border consumer tax shftng are of partcular nterest, as the EU-area ncreasngly s a net mporter of gas. Tradtonal tax ncdence theory presumes spot m arkets. Natural gas n the EU-area, however, s to a large extent regulated by ncomplete long-term contracts. Stll, spot market forces could be ndcatve for tax shftng, by determnng the ex post barganng power n contract renegotatons. By examnng tax shftng n gas sales data we test whether ths s the case. To nvestgate tax ncdence, we estmate natural gas demand elastctes for the household sector n EU countres as well as a reduced form mport equaton. We test whether gas mport prces, whch are predomnantly determned by long-term contracts, have been nfluenced by end-user tax shfts. Keywords: Energy Markets, Incomplete Contracts, Tax Incdence Jel class.no.: K12, L72, Q48, H23, G18, D63, * We are grateful to Dag Morten Dalen, Kjell Hausken, Bengt E. Hernes, Lars Håkonsen, Øysten Håland, Runar Tjersland, Kurt Jacobsen, Tore Wg Jonsbr åten, Espen Moen, Petter Nore, Chrstan Rs, and Krstan Rydqvst for useful comments and suggestons n our research. We are thankful for constructve comments at the 2001 IAEE annual conference, Houston, Aprl 25-27, and at a semnar at Norwegan School of Management, September 19. Fnancal support from the Norwegan Research Councl s gratefully acknowledged. Address of correspondence: Petter Osmundsen, Stavanger Unversty College, Secton of Petroleum Economcs, Po Box 2557 Ullandhaug, N-4091 Stavanger, Norway. Emal: Petter.Osmundsen@tn.hs.no Tel:

2 1. Introducton Energy taxes are mposed for fscal and envronmental reasons, both of whch are much n focus n the EU-area due to Kyoto-requrements for reducng emssons and the EMU fscal requrements of lmtng defcts. 1 Deregulaton of energy markets n the EU-area nduces market shfts that call for a recalculaton of energy taxes. One type of such shfts s reduced proft margns n transmsson and dstrbuton. Energy taxes are also dscussed n relaton to energy securty n the 2000 EU Green Paper. A proposal s put forward for a tax on ol, gas and nuclear energy to fnance a fund for start-up nvestments for renewable energes The EU-countres are net mporters of energy. Accordng to the 2000 Green Paper the EU area s mportng 60 per cent of ts gas consumpton and 90 per cent of ts ol consumpton, and the mport shares are ncreasng. New economc trade theory (strategc trade theory) derves optmal commodty taxes for an mportng country, from a fscal perspectve. 2 An nsght from ths theory s that a net mportng country to some extent may mprove ts terms of trade (reduce the exporters' proft margns) by mposng commodty taxes (energy taxes), and also capture parts of the resource rent. 3 There are also factors that lmt such taxes. Wth an ncreasng moblty of the corporate tax base, natonal energy taxes must be compettve - compared to other countres or regons n order to prevent frms from movng elsewhere. 4 Ths s a matter of tax revenue and employment. As for households, energy taxes are prone to be degressve,.e., to have adverse dstrbutonal effects, snce a low- ncome household typcally spends a large share of ts budget on heatng and transportaton. The "double dvdend" lterature 5 examnes the hypothess that a government can obta n two gans by ncreasng energy taxes: (a) an envronmental gan, and (b) ncreased revenues, that may be used to reduce other dstortve taxes and thereby yeld an effcency gan. An objecton that has been put forward to ths theory s that the result must hnge on 1 If correctly desgned, an energy tax may correct for margnal envronmental damage (Pgouvan tax). However, the same am may be acheved by usng emsson regulatons or quotas. Tradable quotas are partcularly suted as a cost-effectve way to reduce global warmng. If emsson quotas are sold or auctoned, these may also generate smlar revenue as energy taxes. 2 See, e.g., Debashs and Whte (1998). 3 To calculate optmal commodty taxes the mportng country needs knowledge of supply elastctes, demand elastctes and cross prce elastctes for alternatve sources of energy (substtutes). 4 See, e.g., Zodrow and Meszkowsk (1986), Osmundsen, Hagen and Schjelderup (1998), and Olsen and Osmundsen (2000). 5 See, e.g., Pezzy and Park (1998).

3 cases where the tax desgn was not optmal n the frst place, and that a re-desgn of taxes therefore s not partcular to energy taxes. Our focus n ths paper s on tax ncdence for natural gas,.e., on whom s actually bearng the burden of natural gas consumer taxes. 6 In the energy market natural gas s of partcular nterest as ths s an energy source wth an ncreasng market share. Furthermore, because of a specal supply structure regulated by long term contracts, one cannot apply classcal ncdence theory whch s based on spot markets. Ths s n contrast to competng energy sources lke ol that adheres to classcal ncdence theory snce they to a large extent are traded on spot terms. Producers of natural gas, however, are to a large extent locked n (n the short run) as market access s determned by nfrastructure and ppelne capacty. Moreover, beng traded on long-term contracts, tax ncdence of natural gas s partly determned by the explct terms of these contracts and partly settled by renegotatons. Snce a large part of the natural gas consumed n the EU s mported, strategc trade polcy and cross-border tax shftng are also nterestng ssues n ths market. We address tax shftng n the context of long-term gas supply contracts. Tax ncdence s now determned by renegotatons of ncomplete supply contracts, an ssue not covered by exstng ncdence theory. 7 The theoretcal analyss allows us to derve several testable hypotheses for emprcal analyses. Accordng to partal tax ncdence analyses t s vtal to estmate supply and demand elastctes to obtan nformaton on ncdence. We estmate demand elastctes for natural gas on a country panel of EU household sector data. To obtan country specfc elastctes wth the lmted number of observatons avalable for each country, the recently developed shrnkage estmator of Maddala et al. (1997) s employed. The advantage of ths approach s that t avods the homogeneous slope coeffcent straghtjacket of conventonal econometrc panel data models, and at the same tme provdes much more reasonable parameter estmates compared to separate econometrc regressons on each country tme seres. A dynamc econometrc demand model s specfed snce customers are lkely to have larger substtuton opportuntes n the long run. In addton to the partal ncdence effects, general equlbrum effects n energy markets must be consdered. The relevant second -order effects are n our case prmarly determned by the cross-prce elastctes. 6 Polcy mplcatons of ths ssue are prevously surveyed, e.g., by Austvk (1997). We extend hs dscusson by addressng theoretcal aspects of energy tax ncdence, and by undertakng emprcal analyses to shed lght on actual ncdence n the OECD-area. 7 See, e.g., Hamlton (1999), Itaya (1995), and Kotlkoff and Summer (1987). 2

4 Presently, the domnant share of natural gas n the EU-area s not traded on spot terms, but rather on long-term supply contracts. The UK, however, has had an actve spot market for gas for some years. Recently, a spot market has developed n Zeebrugge, after the completon of the Interconnector gas ppelne between the UK and the Contnent, but the prces tend to follow the prces set n the long-term contracts. Snce there s stll avalable capacty (call optons) n the long-term contracts, they represent the margnal source of supply and can thereby dctate prces. The tradtonal (spot market) tax ncdence models are not drectly vald n a settng of long-term contracts. Tax shftng s now nstead determned by the contractual terms and the system for renegotatons. We let contract ncdence denote tax shftng regulated by contracts as opposed to determned by spot markets. Stll, spot market forces may affect underlyng barganng power n contract renegotatons, and thus be ndcatve for contract tax ncdence. To ascertan actual tax ncdence for natural gas mport prces, we specfy and estmate a reduced form equaton for mports of gas to a group of EU countres. To test for the spot market nfluence on gas contract renegotatons, the actual contract ncdence s compared to the spot ncdence predcted by classcal ncdence theory. 2. Tax ncdence Basc nsghts from tax ncdence theory may be derved n a smple partal and statc model. An ene rgy tax, τ, ntroduces a wedge between prces to be pad by the consumers, P c, and prces receved by the producers, P p : (2.1) P P + τ, c = p By dfferentatng the after-tax equlbrum gven by D P ) = S ( P ), we get the classcal tax ncdence formulae ( c p (2.2) dp p dτ ed =, e e S D where e D and e S are the demand and supply elastctes. Ths smple partal equlbrum approach does not capture all relevant aspects of tax ncdence. Stll, t provdes an ntutve 3

5 approach to tax ncdence and accordng to Kotlkoff and Summers (1987) two mportant prncples that emerge from the analyss reman vald also n a more fully specfed model. Frst, tax ncdence does not depend on whch part of a market the tax s assessed on,.e., the person who effectvely pays a tax s not necessarly the person upon whom the tax s leved. In partcular, t s of no materal relevance for actual tax bearng whether an energy tax s leved on extracton companes, transmsson or dstrbuton companes, or consumers. However, the fscal mplcatons may dffer, we should add, f the dfferent partes are located n dfferent countres. Accordng to the nternatonal source prncple of taxaton, companes are taxed at source,.e., where the economc actvty s located. Thus, government revenues of petroleum exportng versus mportng countres are affected by whch level n the value chan the taxes are leved upon. It s reasonable to consder ths as a revenue game between exportng and mportng natons. The second prncple emergng from the basc tax ncdence model s that taxes wll be shfted by those agents and factors that are more elastc n supply and demand,.e. those who can escape the tax. Energy taxes can be shfted forward (downstream) to transmsson companes, dstrbuton companes or consumers, or backwards (upstream) to producers. Generally, the taxes wll be borne by those who cannot easly adjust. Thus, taxes are borne by nelastc buyers or sellers, as s evdent from Eq. (2.2). Ths can also be shown n a dagram. In Fgure 1 taxes are llustrated as a negatve shft n the demand curve, equal to the tax wedge. 4

6 Prce per unt S(P) P c a b τ P d c P p f e D(P) D(P c ) Quantty Fgure 1. Tax ncdence wth a unt tax t In pre-tax equlbrum, pont c, the producer prce equals the consumer prce. We see that the ntroducton of a tax reduces the quantty traded, leadng to a tradtonal deadweght loss (tax neffcency) gven by the trangle bce. 8 (If the tax correctly measures margnal envronmental costs not pad for by the tradng partes, however, ths s not to be conceved as a deadweght loss but rather as a correcton of a negatve externalty.) Also, both the producer and the consumer prce are reduced,.e., the tax burden s shared among the partes, wth the area abcd representng the reducton n consumer surplus and cdfe the reducton n producer surplus. 9 Tax ncdence s most easly llustrated by examnng some extreme cases. If supply s perfectly elastc (horzontal supply curve; e = ), all of the energy tax s accordng to Eq. (5) s borne by the customers ( dp p / dτ = 0 ). Ths would also be evdent from Fgure 1; f the supply curve were a lnear curve parallel wth the q-axs, the producer prce would be gven and all the tax would be shfted to the consumers. Ths could be the case of a unlateral tax ncrease 8 We should p ont out that to get an exact measure of the deadweght loss, compensated demand functons must be appled, see the fgure n Hnes (1999). In our emprcal work n Secton 6 we develop compensated demand elastctes. 9 We have assumed a pece tax. The economc effects of an ad valorem (percentage based) tax are analogous. 5

7 on ol n a sngle country. If the producton s shpped by ol tankers, ol companes have great flexblty wth respect to destnaton. To attract ol, an mportng country would have to offer the same producer prce as other countres. Note that ths does not necessarly apply to gas supples, as large rreversble nvestment s n ppelnes and take-or-pay supply contracts lock the suppler nto a long-term relatonshp wth ts customers. The reverse result, where all the tax s borne by producers ( dp p / dτ = 1), would be the outcome f demand were perfectly elastc ( e = ),.e., f the demand curve n Fgure 1 were horzontal and parallel wth the q- d axs. Ths could be a case of perfect substtuton, e.g., wth ndustral customers havng dual burners. The most lkely scenaro, however, s that both supply and demand have fnte elastctes, and that the tax burden s shared among sellers and buyers. As for natural gas supply elastctes, adjustments have to be made to the theoretcal tax ncdence model. The model s cast n a framework of spot tradng. Most of the natural gas delvered to the European Contnent, however, s not traded on spot terms. The gas delveres are typcally regulated by long-term contracts, so-called take-or-pay contracts. Tradtonal spot market ncdence s thus replaced by ncdence regulated by contract regulatons and procedures for re-negotaton. Tax shftng n a settng of long-term contracts s not addressed n the present lterature. The most mportant elements of the gas sales contracts, and the mplcatons for tax ncdence, are the topcs of the next Secton. We have dscussed tax ncdence emprcally n relaton to a basc tax ncdence theory and contract theory, wth an emphass on the effect of dfferng demand elastctes among countres. The effects of elastctes are general,.e., they also apply to more advanced ncdence models. In advanced ncdence theory, however, addtonal factors are derved that may affect ncdence. Ideally, we would calculate the general equlbrum before and after the tax change. The changes would then provde a descrpton of the ncdence of the tax. Ths s obvously too complcated to pursue. Even at a theoretcal level ths s complex, snce the economy s usually not dsturbed n one dmenson, but rather a package of polces. 10 Stll, the necessary condton for gnorng general equlbrum effects - that the product n queston has a market that s small relatve to the entre economy - s typcally not satsfed for dfferent types of energy nputs. It s therefore necessary also to explore general equlbrum effects. We do ths by dervng cross prce elastctes wth substtutes such as lght fuel ol (LFO) and electrcty. 10 See Atknson and Stgltz (1980). 6

8 The effect of energy taxes (nput factor taxes) depends on whether the customer s a frm or a household. As for households, an ncrease n the prce of one partcular type of energy nduces substtuton effects. We mght expect substtuton towards alternatve means of energy. In the short term the household may be locked nto a partcular technology (demandng a certan type of energy), and the substtuton effect s therefore lkely to be hgher n the long run than n the short run. As for frms, there wll be substtuton effects and output effects. Some ndustres, lke power generators, have nstalled dual burners, whch are lkely to produce a strong substtuton effect, even n the short run. Thus, we mght expect a strong nterfuel competton. Frms are facng dfferent types of compettve pressure n dfferent market segments, callng for a dfferentated tax shftng. Ths s analogous to prce dfferentaton. Prce dfferentaton could generally be possble for natural gas, due to lmted capacty n ppelnes, and for other energy nputs to households, snce they may lack access to the spot market. 3. Gas Sales Contracts The majorty of gas sales n the EU-area s regulated by long-term gas sales contracts. To examne tax ncdence of natural gas we must understand the prce structure of these contracts. Snce there s no regonal, let alone global, lqud traded market prce for natural gas as there s for ol, the market value for natural gas n each sector s typcally determned relatve to the prce of the prncpal competng fuel. In the presence of long-term contracts the tax ncdence theory based on spot markets does therefore not apply. Instead, the tax-shftng pattern s determned by the prcng formulae n the sales contracts. In regulatng contract volumes, the exportng and the mportng companes have conflctng nterests. Snce gas storage s expensve and n lmted supply, the mporter would lke to have flexblty wth respect to volumes, thus beng able to adjust to changes n downstream demand. Demand fluctuates, especally over the seasons, wth a hgher demand n wnter than n summer. The exporters, on the other hand, have to snk large rreversble nvestments n extracton, processng, and transportaton facltes. Before dong so, they would lke to have assurances that they wll be able to sell the gas over a consderable perod of tme, thus securng a return on ther nvestments. Also, to explot the extracton, processng and transportaton capacty, the seller would prefer to delver a stable gas stream at maxmum capacty utlsaton. The exporter would before makng large rreversble nvestments lke 7

9 to have some sort of guarantee of recoupng hs nvestments. One way of dong ths would be to mpose a specfc prce, a mnmum prce, or other types of prce guarantees for the entre perod of delvery. However, ths may elmnate the upsde potental for the seller, and t would also be a bad barganng strategy for the seller to reveal hs reservaton prce. The buyers, on the other hand, would lke the gas prce to be responsv e to the prce of substtutes (such as ol products), so that they reman compettve. It s n the nterest of both partes to have some flexblty n the gas prce so that t can adjust to market changes and keep gas a compettve energy source. The soluton s to let the prce be flexble but to guarantee the seller that he wll be able to sell certan mnmum amounts (volume commtments). The challengng task for gas contract desgn s to trade off conflctng nterests wth respect to volume and prce. The exact contents of these contracts are secret, but the general contract structure s common knowledge n the gas ndustry. The major part of gas export to the EU-regon has untl now been sold on long-term take-or-pay contracts, see Brautaset et al. (1998). In these contracts, the buyer agrees to receve a certan volume of gas per year or, alternatvely, to pay for the part of ths gas volume that t does not want to receve. At the same tme, the buyer has an opton to take out more gas than these mnmum annual amounts, thus conveyng flexblty. Substantal volume flexblty s also avalable on a daly bass. The contracts specfy two types of reference volumes, Daly Contract Quantty (DCQ) and Annual Contract Quantty (ACQ). The annual flexblty s regulated by an nterval around the ACQ, e.g., the buyer s commtted to take or pay per cent of ACQ, and may have specfc optons on annual volumes exceedng ACQ. As for the daly flexblty and commtments, the buyer may be commtted to take or pay per cent of DCQ, and the seller may be commtted to delver up to 110 per cent of DCQ. Addtonal flexblty for the buyer s provded by the rght to receve at a later tme gas that has been pad but not taken (Make Up Gas), and the rght to reduce future delvery f gas take exceeds the commtments n some years (Carry Forward Gas). The current prce on gas delvered accordng to the long-term take-or-pay contracts s determned by a prce formula. The formula lnks the current gas prce to the prce of relevant energy substtutes, thus contnuously securng the buyer compettve terms. 11 The prce formula conssts of two parts, a constant bass prce (fxed term) and an escalaton supplement lnkng the gas prce to alternatve forms of energy (varable term). 12 Examples of alternatve 11 Adjustments n the gas prce are not automatcally mposed, though, but by perodcal (monthly or quarterly) recalculatons of the contract prce by usng the prce formula and updated prces on substtutes. 12 Ths s the basc structure of most gas contracts n Europe. 8

10 energy commodtes used n prcng formulas for natural gas are lght fuel ol, heavy fuel ol, coal, and electrcty. Usually a combnaton of alternatves s used for escalaton purposes (weghted average of energy prces) to reflect the markets for substtutes. 13 Dfferent technques are used, e.g., usng dfferent types of prce lags n the prce formulas. The bass prce reflects the partes evaluaton of the value of the gas at the tme of enterng nto the contract. Each of the alternatve energy commodtes s assgned a certan weght n the escalaton element, reflectng the compettve stuaton between natural gas and the substtute. The prce change of each energy commodty s multpled by an energy converson factor, to make the substtute and natural gas commensurable. Thereafter, the ndvdual escalaton terms are multpled by pass through factors,.e., the change n the prce of the substtute s not fully reflected n the gas prce. A typcal prce formula s gven by (3.1) Pp P0 + α j( AE j AE j 0) EK AEjλ j =, j where P p s the gas prce pad to the extracton company (producer prce), P O s the bass prce, α s the weght n the escalaton element for substtute j (often wth α = ), j ( AEj AEj o ) s the prce change for substtute j (actual mnus reference prce), EK AEj s an energy converson factor, and j j 1 λ j s the pass through factor for prce changes n substtute j. The pass through factors are typcally hgh, e.g., 0.85 or Thus, natural gas prces n these contracts are hghly responsve to prce changes n subs ttutes, and exhbt a hgh volatlty. Ths mples that the exporters are carryng a large fracton of the prce rsk. Prce adjustments for substtutes are based on the dfference between current and hstorc prces. Current prces are calculated as ave rage prces for a reference perod, rangng from three to nne months. Ths gves relable prce data and mples a certan lag n the prce adjustments, both upwards and downwards. 4. Tax Incdence for Natural gas The take-or-pay contracts are complex, contanng a number of detaled regulatons of contngences related to quanttes and prces. For example, the contracts specfy the changes n gas prces that wll take place n response to a change n the ol prce. Stll, there are a 13 Some contracts also contan adjustments for nflaton. 9

11 number of feasble contngences that are not explctly covered by the contracts, e.g., the contractual response to deregulaton. The contracts must therefore be consdered as ncomplete, and revsons and renegotatons take place. Accordng to Hart (1995), an ncomplete contract s best seen as provdng a sutable backdrop or startng pont for such renegotatons rather than specfyng the fnal outcome. The contract should be desgned to ensure that, whatever happens, each party has some protecton aganst bad luck and opportunstc behavour by the other party. Tax ncdence of ol taxes adheres to classcal ncdence theory, snce ol to a large extent s traded on spot terms. Beng traded on long-term contracts, tax ncdence of natural gas s partly determned by the terms of the take-or-pay contracts and partly settled by renegotatons. Under certan condtons and at certan tme ntervals the partes of a gas sales contract may demand prce revsons. The bass for such renegotatons s that (outsde the control of the contractng partes) the value of gas has changed substantally - relatve to the avalable substtutes - n the buyer s home country. The overall objectve s to mantan the compettveness of gas supples. As for changes n energy taxes, however, ol tax changes should not call for renegotatons, to the extent that they are covered by the prcng formula (3.1). Let us take the Kyoto case where natural gas s tax favoured due to less envronmental damage. In the presence of an ncrease n ol taxes, and no adjustments n taxes on gas, the producer gas prce would ncrease by the full extent of the tax accordng to the prcng formula, to the extent that the ol prce that s part of the prcng formula s tax nclusve. 14 An solated ncrease n ol taxes s therefore fully shfted to the gas customers. Thus, there s a cross tax ncdence effect from the ol market to long-term gas sales, specfed by the gas sales contracts. Ths s also reasonable, as the compettveness of natural gas has ncr eased due to the change n relatve tax rates. But what happens f there s an ncrease n gas taxes n the customer's country, e.g. due to EMU revenue requrements? The mplcaton of such a tax change s not explctly regulated by the prcng formula or other terms of the take-or-pay contracts. If the gas taxes are changed relatve to alternatve sources of energy, however, the tax change may nstgate the buyer to demand renegotatons of the producer prce. Tax changes are reportedly an essental part of contract renegotatons. It s therefore nterestng to characterze ths stuaton n terms of standard contract theory The export contracts dffer at ths pont. 15 See, e.g., Hart (1995), Laff ont and Trole (1993), and Mlgrom and Roberts (1992). 10

12 Frst of all, the ssues of contract desgn. Should the contract, accordng to contract desgn prncples, contan more specfc regulatons of whom s to bear the burden or receve the gans of changes n natural gas taxes? A typcal reason for havng ncomplete contractual terms s enforcement problems, e.g., when crtcal parameters are not perfectly observable to one or both of the contractng partes, or when such parameters are not verfable to a court. Nether of these enforcement problems seems to apply to energy taxes that are easly measured. Second, the ssue of rsk sharng. Increases n energy taxes can be perceved as a poltcal rsk. Standard contract theory prescrbes that each of the contractng partes should be held accountable for the rsk wthn ther own control sphere. The resdual rsk should be borne by the party wth the lowest rsk averson. As for the la tter, both the sellers and buyers of natural gas n the EU-area are large dversfed companes, callng for a sharng of the resdual rsk. Whether corporatons can nfluence energy taxes s an open queston. Large transmsson companes are lkely - on behalf of ther customers - to have some nfluence on energy taxes n ther home countres, at least much more nfluence than the seller f he s located n a dfferent country. Overall, therefore, theory prescrbes that the buyer should carry more of the energy tax rsk, thus provdng them wth ncentves to keep taxes low, to the beneft of both contractng partes. The take-or-pay contracts can be perceved as a sequence of gas futures and supply optons. Each year the buyer s commtted to pay for a mn mum quantty. In addton, the buyer has an opton to take addtonal volumes for a gven prce (relatve to substtutes). Ths, of course, devates from the standard spot tax ncdence model. In lack of markets, market forces are replaced by contracts, and the ncdence of taxes on natural gas s determned by renegotatons. Stll, underlyng market forces could have bearng on renegotaton processes. It s therefore nterestng to outlne an analogy to standard ncdence theory. An mportant reference pont s the mnmum quanttes that both the seller and the buyer are commtted to, for a gven relatve prce. Ths means that we have perfectly nelastc demand and supply, whch accordng to Eq. (2.2) mples that the tax ncdence s undetermned. We get the same result from a graphcal analyss, as both curves would be vertcal n Fgure 1. Accordngly, the shftng of natural gas taxes s nstead determned by renegotatons, n whch the market prncple,.e., the compettveness of natural gas n nterfuel competton, s decsve. For the addtonal volume the buyer has an opton to buy, the stuaton may be a bt dfferent. For addtonal volumes the buyer may arbtrage between dfferent sources of gas 11

13 supply,.e., the suppler s lkely to face a decreasng demand curve for addtonal volumes. Wth a decreasng demand curve and a vertcal supply curve, one would expect that all the gas tax s shfted to the producer. Ths presumes, however, that the prce s flexble for extra sales. Ths s not the case for the take-or-pay contracts, as the relatve prce s fxed by the contract. Hence, f the contractual gas prce s not perceved compettve - or f the buyer does not need addtonal gas - the opton s not exercsed. In ths case, however, there s excess capacty n the ppelnes. The seller may thus be tempted to reduce prces for margnal volumes. But ths may turn out to nduce prce pressure on the man supples and on new gas contracts. Snce t s the producer that undertakes the hghest specfc nvestments n nfrastructure, t s ths contractng party that most needs contractual protecton aganst opportunsm from the buyer. In the context of taxaton, the seller also needs protecton aganst opportunstc tax settng by the buyer's government whch may want to capture parts of the resource rents from the exportng country. It s unclear whether the exstng contracts gve the sellers due protecton n ths respect. The tax shftng between sellers and buyers of natural gas s determned by renegotatons of ncomplete gas sales contracts. Lettng x denote utlty, ξ the reference pont (threat pont) of contract party, and α the relatve barganng power; the standard Nash barganng soluton (Roth, 1979) s where the weghted Nash product ( ) α 1 x ξ ( ξ ) α 1 1 x2 2 s maxmsed. Ths barganng abstracton generates three alternatve hypotheses: (a) mport taxes are shared among the buyers and sellers, (b) mport taxes are borne by the buyers, and (c) mport taxes are borne by the sellers. Hypothess (a) corresponds to a case of symmetrc barganng power n contract renegotatons (α close to 0.5), or a case where spot markets are dctatng renegotatons of long-term contracts. Hypothess (b), on the other hand, corresponds to a case where the seller has the most barganng power (α close to 1), or where the status quo of the barganng game s man taned. An argument n favour of the latter s that asymmetrc barganng power s necessary for the seller to recoup rreversble nvestments n nfrastructure. Correspondngly, there are usually no explct regulatons n the take-or-pay contracts as to the response to hgher taxes n the mport country. Ths mght be to the seller's advantage n the case of tax ncreases, as the status quo s no tax shftng. Hypothess (c) corresponds to the case where the buyer has most barganng power (α close to 0). Ths 12

14 hypothess s of partcular nterest n the gas market, snce t has been argued that the EU explots market power to obtan a share of the resource rent. Before testng these hypotheses we gve a bref outlne of the general trend n EU natural gas taxes. 5. Development of Natural Gas Taxes Most EU countres have leved taxes on natural gas to the household sector. For the ndustry sector and the electrcty sector fewer countres have mposed taxes, and the relatve sze of the tax s much smaller for these sectors than for households, reflectng the noton that governments are concerned about the nternatonal compettveness of ther domestc ndustres. Hence, wth our focus on tax ncdence, we focus partcularly on the household sector n the followng. A more comprehensve overvew of the development n household natural gas taxaton n EU countres s provded n Table 1. The table presents tax rates for the frst and last year that data are avalable for each country. We see that there were large dfferences n the tax rate at the end of the 1990s. In 1997, the last year wth observatons for all countres, UK had the lowest tax rates (7%) and Italy had the hghest tax rate (76.5%). 16 However, for all countres there s a postve trend n the tax rate from the late 1970s to the end of the 1990s. To get a clearer pcture of the overall development over tme, we aggregate the natonal tax rates of the eleven countres to a smple non-weghted average n Fgure 2. From ths fgure we see that the average tax rate has ncreased from around 10% n 1978 to around 30% n One possble explanaton to especally hgh gas tax n Italy s that Italy s perceved - relatve to the other EU - countres - to have hgh tax collecton costs. Gas taxes are easy to enforce, due to small montorng problems. 17 When Italy s excluded, we fnd that the average tax rate has ncreased from 7% to 21%. 13

15 Table 1. Taxes and Natural Gas Prces Pad by the Resdental Sector* (Data source: IEA) Country Year Excse tax Value added tax Total Tax Prce ncl. tax USD Percent USD Percent USD Percent USD Austra % % % % % % Belgum % % % % % % Denmark % % % % % % Fnland % % % % % % France % % % % % % Germany % % % % % % Ireland % % % % % % Italy % % % % % % The Netherlands % % % % % % Span % % % % % % UK % % % % % % * Taxes and prces have been deflated to 1995=100 and multpled by the USD exchange rate n the last observaton year for each country. Taxes are n percent of ex tax prce Tax n % of ex tax prce Fgure 2. Average Tax n 11 EU Countres on Natural Gas to Household Sector

16 6. Econometrc Analyss of Household Gas Demand In the prevous Secton we found that taxes on natural gas to households n the EU have ncreased. As shown n Secton 2, elastcty estmates are requred for emprcal analyss of tax ncdence. In ths Secton we present an econometrc model of household sector energy demand. We have a panel data set wth annual observatons from several OECD countres at our dsposton. In panel data analyss t s customary to pool the observatons from each unt (here: country) together, and estmate models wth unt-specfc ntercepts whch are specfed as random or fxed. However, the coeffcents assocated wth the explanatory varables, the slope coeffcents, are assumed to be equal across unts. In our context ths mples, for example, that the own prce elastcty of natural gas s dentcal across countres. In lght of results from prevous econometrc studes and other emprcal evdence, homogenety restrctons on slope coeffcents are not very appealng (Maddala et al., 1997; Garca- Cerrutt, 2000). Hence, our models should be specfed to allow for cross-country heterogenety n demand elastctes Shrnkage Estmaton Estmaton of separate demand models for each country gves the greatest degree of flexblty wth respect to elastcty estmates. However, earler studes have demonstrated that such regresson models often provde mplausble elastcty estmates, for example, postve ownprce elastctes (Atknson & Mannng, 1995). In ths paper we employ a Bayesan shrnkage estmator on our demand models (Maddala et al., 1997). Ths estmator allows for slope coeffcent heterogenety, but mposes some addtonal structure on the generaton of the true coeffcent values compared to separate regresson models. Ths addtonal structure s the assumpton of a common probablty dstrbuto n from whch the true parameter values of the demand models are drawn for each country. The coeffcents estmated by the shrnkage method wll be a weghted average of the overall pooled estmate and separate estmates from each country. 18 In ts most general form the demand model s specfed as 18 The shrnkage estmator shrnks estmates from separate regresson models towards a populaton average. 15

17 (6.1) y = X β + u, = 1, 2,..., N, where y s a T 1 vector, X s a T k matrx of observatons on the k explanatory varables, β s a k 1 vector of parameters, and u s a T 1 vector of random errors whch s dstrbuted as u ~ N(0, σ 2 I). We assume that (6.2) β ~ IN(µ,Σ), or equvalently that (6.3) β = µ + v, where v ~ N(0,Σ). Equaton (6.2) specfes the pror dstrbuton of β n the Bayesan framework. The posteror dstrbuton of β depends on µ and Σ. If µ and Σ are not known, prors must be specfed. When µ, σ 2 and Σ are known, the posteror dstrbuton of β s normal wth mean and varance gven by * 1 ' 1 1 ' (6.4) β = + Σ βˆ 1 X + Σ µ 2 X X 2 X, and σ σ (6.5) V ( ) 1 1 * 1 ' β = X + Σ 2 X, σ respectvely. βˆ s the OLS estmate of β. * If the matrx X ncludes lagged values of y,, the normalty of the posteror dstrbuton of β holds only asymptotcally and under the usual regularty condtons assumed n dynamc regresson models. In the emprcal Bayesan approach that we employ, we use the followng samplebased estmates of µ, σ 2 and Σ n equaton (6.4): 16

18 (6.6a) µ N * 1 = N = 1 β * 2 1 * ' * (6.6b) σˆ = ( y X β ) ( y X β ) T k N 1 * * ' (6.6c) Σ ˆ = ( µ )( β µ ) N 1 = 1 β. We see that the pror mean µ * s an average of the β *, the estmate of the pror varance Σ* s * obtaned from devatons of β from ther average µ * 2, and the estmate of σ s obtaned from the resdual sum of squares usng β *, not the OLS estmator βˆ. The equatons (6.6) are estmated teratvely. In the ntal teraton the OLS estmator βˆ s used to compute µ * 2, σ and Σ*. To mprove convergence (6.6c) s modfed as N ' (6.6c ) Σˆ 1 * * = R + ( µ )( β µ ) N = 1 1 β, where R s a dagonal k k matrx wth small values along the dagonal (e.g ). Accordng to a Monte-Carlo study by Hu and Maddala (1994), the teratve procedure gves better estmates n the mean squared sense for both the overall mean µ and the heterogenety matrx Σ than two-step procedures Emprcal Model and Results The demand for natural gas (y N ) n the household sector of country n year t s specfed as (6.7) ln y = β + β ln y + β ln p + β ln p t N,, t 0 y N,, t 1 f f,, t ft f = E, F, N f = E, F, N + β ln I + β ln I t + u I, t It, t, t f,, t, where p f s the prce pad by the resdental sector for fuel f, f = E(lectrcty), F(uel ol), N(atural gas), I s ncome measured by prvate consumpton, and t s a tme trend varable. The terms nvolvng t allow for changes n demand elastctes over tme. The own-prce elastcty of natural gas demand s e NS = (β N + β N t) and e NL = (β N +β N t)/(1 - β Y ) n the short and 17

19 long run, respectvely. Analogous measures apply to the other prces and ncome. We expect lght fuel ol (LFO) to be a substtute for natural gas. Electrcty s also a substtute, but n many countres electrcty s prmarly used for electrc applances and to a smaller extent for heatng. Hence, we expect smaller cross-prce elastctes between natural gas demand and the prce of electrcty. Table 2 shows the shrnkage estmates of the natural gas demand elastctes n the household sector. All elastctes have the expected sgns, and are less elastc n the short run then n the long run. Demand seems to be bascally untary elastc wth respect to ncome n the short run and elastc n the long run. Ths ndcates that the budget share of gas s constant relatve to ncome n the short run, but ncreasng wth hgher ncome. Gven further economc growth, ths ndcates that demand for natural gas s lkely to contnue to ncrease, and that natural gas wll contnue to gan hgher market shares. All own-prce elastctes are negatve, but demand s relatvely nelastc, and lght fuel ol seems to be a stronger substtute than electrcty. Accordng to the Kyoto agreement fuel ol taxes should be hgh relatve to tax on natural gas due to hgher CO2 emssons per unt. One mplcaton of the emprcal results n Table 2 s that an ncrease n lght fuel ol taxes relatve to natural gas taxes, reflectng the CO2 emssons, would typcally lead to an ncrease n natural gas demand. Table 2. Shrnkage Estmates of Natural Gas Demand Elastctes n the Household Sector* Gas short run Gas long run LFO short run LFO long run Electrcty short run Electrcty long run Income short run Income long run Austra Belgum Denmark Fnland France Germany Ireland Italy Netherlands Span UK * Evaluated n the sample mean year for each country. It s worthwhle to note the substantal varaton n the magntude of the elastctes. In partcular, French natural gas demand s twce as elastc as Spansh demand n own prce. Hence, the degree to whch consumers n dfferent EU countres have to bear taxes can dffer substantally. Snce demand elastctes are ndcatve of ex post barganng power, the 18

20 elastctes n Table 2 provde vtal nformaton to gas sellers as to n whch countres specal protectve contract provsons are called for. However, as demand s farly nelastc n both the short and the long run, n all the countres, t s clear that consumers have to pay a substantal part of the taxes f the supply s not extremely nelastc. Gven the estmated short-run and long-run demand elastctes we can calculate tax ncdence predcted by spot ncdence theory by means of formula (2.2), f one assumes a certan level for the supply elastcty. One ntermedate result s that as expected a larger fracton of the tax ncrease wll be borne by the supplers n the long run than n the short run, due to more elastc long-run demand and hgher substtuton possbltes. The technologcal structure of the natural gas ndustry means that the elastcty of supply s well below 1 and decreases rapdly wth hgh capacty utlzaton rates. 19 For example, f the elastcty of supply s 0.7, we calculate the ncdence on producer prce predcted by the spot ncdence model for the two European countres representng the extremes n Table 2 wth respect to own-prce elastctes,.e. Span and France. The short-run elastctes are 0.154/( ) = and -0,308/(0,7+0,308) = , respectvely. Fgure 3 plots the tax ncdence for these two countres for dfferent supply elastctes. 0 Change n producer prce / tax cha -0,1-0,2-0,3-0,4-0,5-0,6-0,7-0,8-0,9 0 0,2 0,4 0,6 0,8 1 1,2 1,4 1,6 Span France 1,8 2-1 Supply elastcty Fgure 3. Rato of Change n Producer Prce to Tax Change for Span and France 19 For emprcal evdence on the structure of natural gas supply n Europe, see e.g. Mathesen, Roland & Thonstad (1987) and Golombek, Gjelsvk & Rosendahl (1998). 19

21 7. Emprcal Analyss of Natural Gas Import Prce Incdence In Secton 6 we showed that demand for gas n the EU s hghly nelastc, and that consumers accordng to spot ncdence theory are lkely to bear a substantal part of any tax except for n stuatons wth extremely nelastc supply. We now turn to actual tax bearng n long-term gas sales contracts. Gven the complexty of the value chan t s not possble to specfy a full structural model. However, to assess how the tax burden s dvded t s suffcent to specfy a reduced form model that gves the mpact on the prce of taxes and other exogenous varables n the system. We assume that the exogenous factors determnng the mport prc e for natural gas n the EU, are the ol prce, the tax level, and prvate fnal consumpton. The Brent spot prce s ncluded to account for the fact that natural gas prce contract formulas are heavly nfluenced by the prce of ol. Hence, there should be a strong postve relatonshp between the mport prce and the ol prce. Prvate consumpton s ncluded to account for shfts n the demand curve. In addton, we nclude the lagged mport prce for natural gas to account for adjustment costs. The eq uaton to be estmated s: (7.1) lnimp,t = α + α IMP1 lnimp,t-1 + α OIL lnoil,t +α TAX lntax,t + α CONCAP lnconcap,t +u,t, where IMP,t s the mport prce to country n year t, OIL s the spot prce of Brent blend ol, TAX s the tax on natural gas n the household sector, and CONCAP s prvate fnal consumpton per capta. Country-specfc effects α are ncluded to capture cross-country dfferences n technology, nfrastructure, competton and regulaton whch nfluence the mport prce. The log-log specfcaton mples that the coeffcents can be nterpreted as short-run elastctes. The parameter on the TAX varable s of partcular nterest, snce t provdes nformaton on the tax ncdence. In partcular, αtax=0 corresponds to the case whe re the buyers bear the full mpact of the tax. If 0 < α TAX < 1, the tax s shared. The closer α TAX s to 0, the smaller part of the burden s shfted to the sellers, and vce versa for α TAX = 1. If α TAX = 1, the seller bears the full burden of the tax. Even wth these smplfyng assumptons, t s dffcult to obtan the requred data for all EU-countres. Our data set therefore becomes smaller than n Secton 6, as suffcent data are avalable only for Belgum. Fnland, France, Germany, Italy, the Netherlands, Span and 20

22 the UK. However, as all the man gas consumng countres n the EU are ncluded n ths sample, t should stll gve a good ndcaton of the total effect. Table 3. Emprcal Results from Import Prce Regresson Model* Elastcty Short-run elastctes Long-run elastctes Est. St.Err. t-rato Est. St.Err. t-rato E OIL E TAX E CONCAP * N = 49 observatons. R 2 adjusted = The short run elastctes correspond to the parameters α OIL, α TAX, and α CONCAP. The estmate of α IMP1 s wth standard error Table 3 presents the short- and long-run elastctes from the estmated mport prce regresson model. Use of a country-specfc ntercept s supported by an F-test wth a test statstc of 3.11, whch leads to rejecton of homogenety at the 1% level ((7, 37) df). Accordng to Table 3 the mport prce ncreases sgnfcantly wth a hgher ol prce both n the short and long run. Furthermore, a hgher per capta ncome leads to a sgnfcantly hgher gas mport prce. Both these relatonshps are as expected, gven the structure of the market, the long-run contracts, and the ncreased demand for natural gas. However, there s no sgnfcant negatve relatonshp between the mport prce of gas and the gas taxes. In other words, these results ndcate that the mport countres have not been able to shft parts of the tax ncreases backwards to the producers,.e., the data support hypothess (b) n Secton 4. Thus, t seems that spot market forces have not been decsve n the renegotatons of the takeor-pay contracts. The spot ncdence model does not gve good predctons for contract ncdence n gas markets. One reason for the lack of spot nfluence on contract renegotaton s that there has not been an effcent spot market at the European Contnent n the perod we analyse. The gas transmsson companes have had several sources of gas to select from, manly Russa, The Netherlands, Norway and Algera. The long-term contracts wth these countres, frst of all The Netherlands and Norway, also contan swng elements (call optons) that allow for addtonal volumes. But even f the take-or-pay contracts have perodcal renegotatons, they probably do not mmc spot contracts. Frst of all, contract volumes are to a large degree fxed, and not subject to renegotaton. Second, a few of the contracts have specal provsons dealng wth tax ncreases. The devatons from spot tradng were necessary - at the tme - to nduce large rreversble nvestments n nfrastructure, takng nto account the poltcal rsk of ncreased energy taxes n the mportng countres once gas penetraton s hgh. 21

23 8. Concluson Gas exports are - for techncal and economc reasons - most often transmtted by means of ppelnes. Presently, ths mples lmted export flexblty, and often the entre gas supply from a ppelne goes to a sngle customer, typcally a large transmsson company. The suppler s thus locked nto a long-term relatonshp wth one commercal party. Due to the hgh level of rreversble and specfc nvestments, ths nvolves a hgh hold -up rsk, whch s why natural gas exports often are regulated by long-term contracts that protect each of the contractng partes from opportunstc behavour. We focus on a smlar case where the seller s exposed to rsk from the government n the customer s country. By ntroducng energy taxes, mportng countres may capture parts of the resource rent. 20 The shftng of the tax s determned by a process of renegotaton of ncomplete gas sales contracts, and hence devates from the normal spot based theory of tax ncdence. The fact that both sellers and buyers may be locked n s also nterestng n a strategc trade polcy perspectve. In ths paper we study the market for natural gas n the EU. Our demand analyss ndcates that demand s relatvely heterogeneous between the dfferent countres, and longrun responses are substantally stronger then short-run responses. Stll, demand s always farly nelastc, and hence consumers are lkely to bear a substantal part of any tax, provded that the supples are not extremely nelastc. Both lght fuel ol and electrcty seem to be substtutes to natural gas, wth fuel ol as the strongest. Hence, the Kyoto requrement that ol s taxed more heavly s lkely to lead to a postve envronmental effect as energy demand wll shft towards natural gas. When nvestgatng whethe r the ncreases n taxes n the EU have had an mpact on the mport prce for natural gas, we can strongly reject ths hypothess both n the short and long run. Hence, the long-run contracts or the current market structure have so far been able to protect the exporters of natural gas from rent capture by the mportng countres, and the ncreased taxes on natural gas have then manly transferred revenue from consumers to governments wthn the EU. Wth ths structure, t s also clear that one cannot say that taxng natural gas has been used as a tool n strategc trade polcy. 20 The gas exporter s locked nto a long run relatonshp wth both the mporter and the tax authortes n the mportng country. Thus, barganng over the resource rent could be perceved as a mult-prncpal mult-agent game. 22