The Impact of the Fracking Boom on Arab Oil Producers

Transcription

1 The Impac of he Fracking Boom on Arab Oil Producers Ocober 18, 216 Luz Kilian Universiy of Michigan CEPR Absrac: This paper makes four conribuions. Firs, i invesigaes he exen o which he U.S. fracking boom has caused Arab oil expors o decline since lae 28. Second, he paper quanifies for he firs ime by how much he U.S. fracking boom has lowered he global price of oil. Using a novel economeric mehodology, i is shown ha in mid-214, for example, he Bren price of crude oil was lower by $1 han i would have been in he absence of he fracking boom. Third, he paper provides evidence ha he decline in Saudi ne foreign asses beween mid-214 and Augus 215 would have been reduced by 27% in he absence of he fracking boom. Finally, he paper discusses he policy choices faced by Saudi Arabia and oher Arab oil producers. JEL Code: Key Words: Q43, Q33, F14 Arab oil producers; Saudi Arabia; shale oil; igh oil; oil price; oil impors; oil expors; refined produc expors; oil revenue; foreign exchange reserves; oil supply shock. Acknowledgemens: I hank Chrisiane Baumeiser, Bassam Faouh, Ryan Kellogg, he edior and hree anonymous referees for helpful commens and discussion. This work has benefied from a financial gran from he Economic Research Forum and from commens from wo anonymous referees provided by he Forum. The conens and recommendaions do no necessarily reflec he views of he Economic Research Forum. Correspondence o: Luz Kilian, Deparmen of Economics, 611 Tappan Sree, Ann Arbor, MI , USA. lkilian@umich.edu.

2 1. Inroducion The use of hydraulic fracuring (or fracking) in conjuncion wih horizonal drilling and micro-seismic imaging has made i possible o exrac crude oil from rock formaions characerized by low permeabiliy. Oil exraced by hese echniques is commonly referred o as igh oil or shale oil o differeniae i from crude oil exraced by convenional drilling echniques. To dae, commercial shale oil producion has been largely limied o he Unied Saes. The U.S. oil fracking boom is an example of a echnological change in a single indusry in one counry affecing inernaional rade worldwide. Increased U.S. shale oil producion over ime has displaced crude oil expors from Arab oil producing counries, boh because he Unied Saes no longer relies as heavily on crude oil impors from Arab counries and because U.S. refineries have increasingly expored refined producs such as gasoline or diesel made from domesically produced crude oil, causing oher counries o cu back on heir crude oil impors as well (see Kilian 216). Whereas he gains o he U.S. economy of he fracking boom are well undersood a his poin, lile is known o dae abou he losses his developmen has imposed on foreign oil producers. Undersanding he implicaions of he U.S. fracking boom is imporan no only for policymakers in Arab economies deciding on how bes o respond o he igh oil boom, bu i also provides a prime example of a well-idenified exogenous shock o he erms of rade of primary commodiy exporers. This paper quanifies he impac of he U.S. igh oil boom on global oil producion, on U.S. impors of crude oil and expors of refined producs as well as Arab expors of crude oil. 1 I represen he U.S. igh oil boom as a sequence of exogenous shocks o world oil producion. On he basis of a srucural economeric model of he global marke for crude oil, I consruc an esimae of how he price of oil in global markes would have evolved in he absence of his boom. I show ha he cumulaive effec of he igh oil boom on he Bren price had been building gradually, reaching a peak in mid-214, before declining in lae 214. Whereas in mid-214 he Bren price was lower by $1 han i would have been in he absence of he fracking boom, by mid-215 his price differenial had fallen o $5. My analysis also demonsraes ha a very similar price decline would have occurred beween July 214 and January An excellen nonechnical sudy of he impac of he igh oil boom on Arab oil producers is Faouh (214). 1

3 even in he absence of increased U.S. shale oil producion, suggesing ha increased shale oil producion was no he main cause of his price decline. 2 I use he difference beween he acual and he counerfacual global price of crude oil o quanify he losses in Saudi oil revenue (and hence in Saudi naional income) since lae 28 under he mainained assumpion ha Saudi oil producion remained unchanged by he fracking boom. My analysis shows ha he cumulaive losses in Saudi oil revenue caused by fracking by Augus 215 had reached 12 billion dollars. I relae his esimae o he subsanial decline in Saudi ne foreign asses beween mid-214 and Augus 215. I show ha he shale oil boom accouns for 27% (or 24 billion U.S. dollars) of his decline, wih he remaining 66 billion dollars reflecing increased oil producion by oher oil producers, shifs in oil price expecaions, and he weakening of he global economy. I make he case ha his decline is unsusainable and discuss implicaions of my analysis for Saudi economic policies in he years o come. The remainder of he paper is organized as follows. In secion 2, I review he evoluion of crude oil producion in he Unied Saes and elsewhere, highlighing he conribuion of he U.S. fracking boom o global oil producion growh. Secions 3 and 4 quanify he exen o which he shale oil boom has reduced Arab oil expors since lae 28, disinguishing beween he reducion in U.S. crude oil impors from Arab oil producers and he reducion in Arab oil expors o he res of he world caused by increased U.S. expors of refined producs, as U.S. refiners ook advanage of access o low-cos domesically produced crude oil. In secion 5, I sudy he financial implicaions of he U.S. fracking boom for Arab oil producers, focusing on Saudi Arabia as he leading example. Building on he analysis in secions 2, 3 and 4, I develop and compare wo daa-based counerfacuals aimed a esimaing he implicaions of he shale 2 The consrucion of he counerfacual rules ou an endogenous response of Saudi oil producion o he U.S. shale oil boom, consisen wih he conclusion of Baumeiser and Kilian (216b) and Kleinberg e al. (216) ha here is no documened evidence of such a policy response. In fac, he daa show no apparen reacion by Saudi Arabia unil March 215, when Saudi oil producion increased slighly. If his increase was a Saudi policy shif a all, hen his shif was in response o falling oil prices afer June 214, no in response o he rise in shale oil producion since lae 28. I show ha he shale oil boom was no responsible for his oil price decline. Moreover, given he recursive naure of he counerfacual, even if some of he price decline in 214 had been caused by he effecs of he U.S. shale oil boom, he poenial endogeneiy of he alleged Saudi oil producion response saring in March 215 could effecively be ignored, because only he las six monhs of he 82 monhs, for which I consruc he counerfacual, could possibly be affeced by any Saudi policy response. 2

4 oil boom for he global price of oil and for Saudi Arabia s oil revenues beween December 28 and Augus 215. Secion 6 relaes my preferred esimae of hese effecs o he dramaic decline in Saudi ne foreign asses since June 214 and discusses he policy implicaions for Saudi Arabia in paricular. The concluding remarks are in Secion Recen Changes in Oil Producion in he Unied Saes and Elsewhere Only a few years ago, a common view among pundis was ha world oil producion would no longer be able o keep up wih growing oil consumpion needs. An exreme version of his skepicism was embodied in he peak oil hypohesis which assered ha global oil producion had permanenly peaked by 27 or ha he peak was imminen. 3 Alhough economiss end o be skepical of he peak oil hypohesis for reasons discussed in Holland (28, 213), i was readily apparen a he ime ha many radiional oil fields were in decline and ha he prospecs for discovering more crude oil in regions no ridden wih civil srife were diminishing. Thus, even graning ha high oil prices driven by srong demand for oil end o provide srong incenives for expanding fuure oil producion, as of 27, nohing in pas experience guaraneed ha his supply response o rising oil prices would be adequae or ha i would occur in a imely manner. As he lef panel of Figure 1 shows, wih he benefi of hindsigh, hese concerns were unwarraned. No only did he U.S. fracking boom reverse he long-sanding decline in U.S. crude oil producion, bu Canadian producion of unconvenional crude oil from oil sands soared and Russian oil producion reached unprecedened levels. A he same ime, boh Saudi Arabian and oher Persian-Gulf producion growh, which seemed o have levelled off afer 25, acceleraed again. This surge in oil producion was more han enough o offse he decline in crude oil producion in oher counries. Table 1 shows an increase in global oil producion of 6.17 million barrels per day (mbd) beween November 28 and Sepember 215, corresponding o an average rae of increase of 1.16%. Average growh acceleraed 3 For example, an IMF sudy by Benes e al. (215), using daa up o 29, prediced a near doubling of he price of oil over he coming decade based on he view ha geological consrains would win ou over echnological improvemens in conserving oil use and in oil exracion. 3

5 o 2.63% beween June 214 and Sepember 215. Table 1 dispels he noion ha fracking in he Unied Saes alone has been responsible for his surge. Alhough he Unied Saes was responsible for he bulk of he oil producion increase, accouning for an exra 4.3 mbd over he las seven years, here also have been noable producion increases in Iraq (2.7 mbd), Saudi Arabia (1.23 mbd), Russia (.79 mbd) and Canada (.73 mbd) ha were unrelaed o he fracking boom. The evoluion of U.S. oil producion in paricular is remarkable, given he long-sanding decline in U.S. oil producion ha began in he early 197s and was only briefly reversed by he developmen of he Alaskan oil fields in he lae 197s. The verical line in he lef panel of Figure 1 refers o he dae of November 28, which marks he reversal of his rend. I can be shown ha his reversal is largely due o he U.S. fracking boom. The righ panel of Figure 1 plos he sum of crude oil producion a he Bakken, Eagle Ford, Haynesville, Marcellus, Niobrara, Permian, and Uica shale oil plays. These seven regions accoun for 92% of U.S. domesic shale oil producion growh. Beween November 28 and Sepember 215, U.S. shale oil producion increased by approximaely 3.82 / million barrels/day (mbd), which roughly maches he enry of 4.3 mbd for he change in oal U.S. crude oil producion in he firs column of Table 1. Thus, hroughou his paper, I will rea November 28 as he beginning of he fracking boom. 3. Quanifying he Decline in U.S. Crude Oil Impors from Arab Producers The U.S. oil secor is governed by he following ideniy, which describes he composiion of he oal quaniy of crude oil used by he economy: Q M X C I, where Q is domesic crude oil producion, M is U.S. crude oil impors, X is U.S. crude oil expors, C is he consumpion of crude oil, which refers o crude oil used by U.S. refineries o produce refined producs such as gasoline or diesel for domesic consumpion or for expor, and oil invenories. I denoes he change in crude The lef panel of Figure 2 illusraes how U.S. domesic oil producion, crude oil impors and 4

6 expors of crude oil have evolved since he 197s. Three facs sand ou. Firs, as already noed, following a secular decline in U.S. oil producion, U.S. oil producion sharply acceleraed in lae 28, reflecing he fracking boom. Second, hisorically crude oil impors have been rising during economic expansions and falling during economic downurns. This paern changed in 25, even before he fracking boom, when he U.S. economy learned o make do wih less crude oil during an economic expansion amids high and rising oil prices. The decline in crude oil impors acceleraed during he fracking boom, when increased domesic shale oil producion lowered he U.S. price of oil compared wih he global price of oil, reducing U.S. demand for high-cos crude oil impors. The reason for he comparaively low price of shale oil was a glu of ligh swee crude oil in he cenral Unied Saes ha refleced ransporaion bolenecks wihin he Unied Saes as well as a ban on U.S. crude oil expors (see Kilian 216). There have been some exempions from his ban, noably for oil expors o Canada, bu overall his resricion has been binding. As a resul, U.S. crude oil expors have remained negligible hroughou mos of his sample. Only in 214 and 215 U.S. oil expors increased somewha, as he Obama adminisraion allowed for more flexibiliy in he inerpreaion of he law governing U.S. oil expors. As of December 215, Congress passed a law lifing he oil expor ban, bu ha policy shif is no ye refleced in he daa in Figure 2. The righ panel of Figure 2 quanifies he oal use of crude oil by he U.S. economy, compued as Q M X. In quanifying he imporance of oil impors for he U.S. economy, i is useful o express U.S. crude oil impors as a share of he amoun of crude oil used by he U.S. economy. 4 The lef panel of Figure 3 shows he evoluion of he share of U.S. crude oil impors from he res of he world as well as he share of U.S. crude oil impors from OPEC, from Arab OPEC (defined as Algeria, Iraq, Kuwai, and Saudi Arabia for our purposes), and from Wes Africa (Nigeria, Angola). 5 The overall impor share 4 This approach conrols for flucuaions in he size of he U.S. economy and for slow-moving changes in he energy efficiency of he U.S. economy. During a recession, for example, all else equal, boh he use of crude oil by he economy and is oil impors would be expeced o fall wihou affecing he oil impor share. In conras, if he fracking boom displaces crude oil impors from Arab oil producers, his should be refleced in a reducion in he share of U.S. impors from Arab counries in he oil used by he U.S. economy. 5 I approximae crude oil impors from hese counries by peroleum impors under he assumpion ha U.S. produc impors from Arab counries are zero. This assumpion can be verified for seleced daes using daa in he EIA s daabase for U.S. crude oil impors a he company level. Figure 3 excludes Libya from he Arab oil producers 5

7 declined from a peak of almos 7% in he mid-2s o abou 45% in 215. This paern is also found for OPEC oil impors and is componens. By 215 crude oil impors from Wes Africa fell o near zero. Wes African crudes are in direc compeiion wih domesically produced ligh swee crude oil in he Unied Saes. As U.S. refiners along he Eas Coas found ways of subsiuing shale oil (or lower priced convenional ligh swee crude oil from domesic sources) for hese impors, Wes African crudes were almos enirely displaced. Crude oil impors from Arab OPEC as well fell from abou 2% of U.S. oil use as recenly as 28 o under 1% in 215. The reason presumably was ha hese crudes no longer could compee wih he low price of domesically produced U.S. crude oil and U.S. impors of heavy Canadian crudes (see Kilian 216). The righ panel of Figure 3 races he impor shares of four Arab oil producers before and afer November 28, which marks he beginning of he fracking boom (see Figure 1). We are ineresed in wheher here are imporan changes in he dependency of he U.S. economy on oil impors from hese counries. Figure 3 shows a sriking drop in he share of U.S. oil impors from Saudi Arabia afer November 28. Afer some flucuaions, he Saudi share reaches near 5% in 215, compared wih abou 1% iniially. Figure 3 shows a seady decline in he share of U.S. oil impors from Algeria (anoher producer of ligh swee crude oil no unlike he Wes African oil producers) as well. The share of U.S. oil impors from Iraq also declines over ime. The higher volailiy of he laer share is likely explained by poliical consrains on Iraqi oil producion more han he fracking boom. Only he share of oil impors from Kuwai remains largely sable. 4. Quanifying he Displacemen of Arab Oil Expors o he Res of he World How quaniaively imporan is he reducion in U.S. expor marke share for Arab oil producers? A naural counerfacual is how Arab oil expors o he Unied Saes (measured in millions of barrels per day) would have evolved afer November 28, if domesically produced U.S. crude oil had no displaced because Libyan oil expors in recen years are driven primarily by poliical evens in Libya. The UAE and Qaar are excluded, because he EIA does no provide monhly daa for U.S. impors from hese counries, given heir small size. 6

8 hese crude oil impors. Thus, one componen of he reducion in Arab oil expors is he amoun of crude oil ha he Unied Saes would have impored from Saudi Arabia, Kuwai, Algeria, and Iraq, respecively, if he share of U.S. impors from hose counries would have remained consan a is average value beween Sepember 24 and November There also is a second componen o be considered, however, arising from he displacemen of foreign crude oil impors. Given he U.S. oil expor ban, very lile of he shale oil producion generaed by he fracking boom was expored. I may seem ha herefore his producion would no maer for Arab oil expors beyond he reducion in U.S. crude oil impors. This conclusion would be misaken because he glu of ligh swee crude oil in he cenral Unied Saes ha emerged following he fracking boom depressed he price of domesically produced crude oil. U.S. refiners wih access o low-priced domesic crude oil herefore increased expors of refined producs such as gasoline or diesel, noably o Europe and o Lain America. These expors of refined producs effecively circumvened he ban on crude oil expors. To he exen ha counries in he res of he world ha bough hese refined producs no longer needed o impor as much crude oil for heir refineries as before, his rade in producs reduced he demand for Arab oil expors, adding o he coss of he fracking boom o Arab oil producers (see Kilian 216). The lef panel of Figure 4 shows a surge in U.S. expors of refined producs in recen years. Assuming ha he share of U.S. produc expors in he oil used by he U.S. economy had sayed consan a is November 28 value, one can consruc a counerfacual for how expors of refined producs would 6 This benchmark is reasonable in ha he Saudi share, for example, remained remarkably sable over his period (see Figure 3). Disregarding earlier daa in compuing his share helps avoid variaion in he share caused by exogenous evens such as he 23 Iraq War. In he absence of alernaive explanaions of he variaion in Arab oil expors, he difference beween his counerfacual pah of Arab oil expors and he observed pah of Arab oil expors o he Unied Saes can be inerpreed as he loss in Arab oil expors caused by he fracking boom. Clearly, he assumpion ha all of he variaion in he share of Arab impors can be aribued o he fracking boom may be quesioned. There could be alernaive explanaions of he disproporionae drop in he U.S. impor share from Saudi Arabia beween 29 and 211, for example. Moreover, here are reasons o believe ha U.S. impors from Saudi Arabia are less sensiive o he shale oil boom han oher oil impors, because some U.S. refineries have been designed specifically for Saudi medium-densiy crude and because Saudi crude oil faces lile direc compeiion form ligh swee shale oil (also see Kilian 216). Finally, one could make he case ha oil producers responding o a decline in he global price of oil riggered by reduced U.S. oil use would reduce heir oil producion proporionaely o heir share in global oil producion, in which case he impor share would no remain consan over ime, bu may acually fall in response o a decline in he use of oil by he U.S. economy. For now we pu hese concerns aside. The realism of he working assumpion of impor shares will be assessed in secion 5 by confroning counerfacual oil price esimaes based on his assumpion wih exraneous evidence. 7

9 have evolved in he absence of he fracking boom saring in lae 28. The difference beween he pah of acual and counerfacual U.S. produc expors provides a rough measure of he magniude of he implied reducion in Arab oil expors o he res of he world caused by he shale oil boom. The righ panel of Figure 4 expresses his loss in erms of is crude oil equivalen. The crude oil equivalen is compued by muliplying he shorfall of produc expors compued in he lef panel by This conversion facor is based on he fac ha U.S. refineries in 214 on average produced 12 gallons of diesel and 19 gallons of gasoline (for a oal of 31 gallons of refined produc) from 42 gallons (one barrel) of crude oil, according o he U.S. Energy Informaion Adminisraion. Assuming an average mix of diesel and gasoline, his implies a conversion facor of 42/31= This baseline esimae of he Arab crude oil expors los o U.S. produc expors, however, ignores he fac ha rising Canadian expors of crude oil allowed he Unied Saes o expor more refined producs han oherwise would have been possible. Impors of heavy crudes from Canada in paricular appear o have displaced U.S. impors of heavier crudes from Mexico and Venezuela wih he laer counries impor shares afer lae 25 seadily falling from 9% each o abou 4% and Canada s share seadily rising from 15% o 21% over he same period. 7 Thus, he increase in U.S. oil impors from Canada exceeds he combined reducion in U.S. oil impors from Mexico and in U.S. oil impors from Venezuela. I would be misleading o aribue he resuling produc expors o he fracking boom. Subracing he ne increase in crude oil impors from Canada, Mexico and Venezuela from he baseline expor loss series in he righ panel of Figure 5 ends o reduce he esimae of he oil expor losses of he res of he world subsanially. I ake he laer series as he benchmark when assessing he role of he fracking boom in driving up U.S. refined produc expors. In quanifying he Arab oil expor losses caused by he U.S. expors of refined producs, for simpliciy, I assume ha he oal loss of expors o he res of he world (measured as he crude oil equivalen of increased U.S. produc expors) can be apporioned based on he average share of each counry s oil producion in oil producion in he world 7 These esimaes were obained by muliplying U.S. peroleum impors from hese counries by he counry-specific share of crude oil in hese impors obained from Ocober 214 daa in he Energy Informaion Adminisraion s company level repors. 8

10 excluding he Unied Saes. Figure 5 combines for each Arab oil producer he reducion in oil expors esimaed in secion 3 and ha esimaed in secion 4. For Saudi Arabia, Algeria and Iraq, here is a endency for he expor losses expressed in million barrels of crude oil per day o increase over ime. In he case of Saudi Arabia, he combined losses reach one million barrels per day in 214; Algeria reaches.5 million barrels per day in 213 and Iraq in 215. Only Kuwai has been largely unaffeced wih no pronounced endency for increased expor losses over ime. Aggregaing hese esimaes across he four counries under consideraion, Arab oil expors losses over ime reached a maximum of 5.3 mbd. This indirec esimae is somewha higher han he direc esimae of he increase in U.S. shale oil producion in secion How Much Oil Revenue Did Arab Oil Producers Lose? So far we have quanified he physical oil expors foregone by Arab oil producers as a resul of he surge in U.S. igh oil producion (measured in barrels per day). An equally imporan quesion is how large he financial losses have been in U.S. dollars. For concreeness, my discussion in his secion focuses on Saudi Arabia. Answering his quesion involves consrucing a counerfacual for he price of crude oil based on a srucural model of he global oil marke. A naural quesion o explore is how differen he price of crude oil would have been, if all oil producers oher han he Unied Saes had mainained heir observed oil producion levels, bu he U.S. fracking boom had never happened. This counerfacual level of global oil producion may be consruced eiher by lowering acual global oil producion by he amoun of Arab oil expor losses shown derived in secions 3 and 4 or alernaively by subracing he direc esimae of shale oil producion in secion 2. Given a suiable srucural model of he oil marke, one may hen infer he sequence of flow supply shocks required o produce his pah of oil producion afer November 28, holding consan he remaining srucural shocks in he model. If his counerfacual shock sequence does no differ sysemaically from hisorical shock sequences, we may feed i back ino he srucural model o simulae how he price of oil would have evolved under he counerfacual. Based on his counerfacual price sequence, he implied increase in Saudi revenues may be compued as 9

11 Q ( P P ), where Saudi, acual counerfacual acual Q Saudi, acual denoes he observed level of Saudi oil producion in monh and P counerfacual P acual denoes he difference beween he acual dollar price of oil and he dollar price of oil obained under his counerfacual The Srucural Model of he Global Oil Marke The counerfacual price of oil is consruced wih he help of a srucural vecor auoregressive (VAR) model of he global marke for crude oil (see, e.g., Kilian and Murphy (214); Kilian and Lee (214); Baumeiser and Kilian (214, 216a)). This srucural model includes four variables: (1) he growh in global crude oil producion, (2) an updaed measure of cyclical flucuaions in global real economic aciviy originally proposed by Kilian (29), (3) he log of he real price of crude oil (obained by deflaing he U.S. refiners acquisiion cos for crude oil impors by he U.S. CPI), and (4) he change in above-ground global crude oil invenories measured by a suiable proxy. We follow he lieraure in esimaing he model a monhly frequency using seasonal dummies and 24 auoregressive lags. Absracing from he pre-sample observaions, he esimaion period sars in The model is reesimaed on daa exending o Augus 215, as described in Kilian and Lee (214). The srucural shocks are idenified based on a combinaion of sign resricions and bounds on he shor-run price elasiciies of oil demand and oil supply. The key idenifying assumpions are resricions on he signs of he impac responses of he four observables o each srucural shock. There are four srucural shocks. Firs, condiional on pas daa, an unanicipaed disrupion in he flow supply of oil causes oil producion o fall, he real price of oil o increase, and global real aciviy o fall on impac. Second, an unanicipaed increase in he flow demand for oil (defined as an increase in oil demand for curren consumpion) causes global oil producion, global real aciviy and he real price of oil o increase on impac. Third, a posiive speculaive demand shock, defined as an increase in invenory demand driven 8 There are several poenial caveas o keep in mind. Firs, his hough experimen is based on he premise ha increased U.S. shale oil producion was enirely due o exogenous oil supply shocks raher han being caused by oil demand shocks. This assumpion is consisen wih he view ha in he absence of echnological innovaion, he shale oil boom would no have aken place, bu i makes no allowance for effecs working hrough oil price expecaions. Second, he counerfacual effecively posulaes ha Saudi Arabia would have been unable or unwilling o increase is oil producion in he absence of he fracking boom. 1

12 by expecaions shifs no already capured by flow demand or flow supply shocks, in equilibrium causes an accumulaion of oil invenories and raises he real price of oil. The accumulaion of invenories requires oil producion o increase and oil consumpion o fall (associaed wih a fall in global real aciviy). Finally, he model also includes a residual demand shock designed o capure idiosyncraic oil demand shocks driven by a myriad of reasons ha canno be classified as one of he firs hree srucural shocks. 9 As all sign-idenified srucural VAR models, his model is se-idenified and does no yield a unique soluion. I follow Kilian and Lee (214) in selecing he unique model ha comes closes o a onemonh price elasiciy of oil demand of -.26, among all admissible soluions of he model (see Kilian and Murphy 214; Coglianese e al. 216). 1 Figure 6 shows how much shocks o he flow demand for oil, o he flow supply of oil and o speculaive (or sorage) demand shocks driven by oil price expecaions have conribued cumulaively o he evoluion of he real price of oil, according o he model esimaes. The esimaes confirm ha all hree shocks conribued o he decline in he real price of oil since June 214 raher han he oil supply shock alone, which is broadly consisen wih he conclusions of Baumeiser and Kilian (216b) using an alernaive mehodology The Counerfacual Level of World Oil Producion Our main ineres in his paper is in he sequence of flow supply shocks in he srucural model. Given a pre-specified counerfacual pah of global oil producion, we may infer he sequence of flow supply 9 In addiion o hese saic sign resricions, he model imposes he dynamic sign resricion ha srucural shocks ha raise he price of oil on impac do no lower he real price of oil for he firs 12 monhs following he shock. The raionale for his resricion is ha an unexpeced flow supply disrupion would no be expeced o lower he real price of oil wihin he same year nor would a posiive flow demand or speculaive demand shock. Finally, he model imposes he resricions ha he impac price elasiciy of oil supply is close o zero and ha he impac price elasiciy of oil demand canno exceed he long-run price elasiciy of oil demand, which convenional esimaes pu a These are he only sudies in he lieraure ha esimae he price elasiciy of demand explicily accouning for sorage behavior and for he endogeneiy of price and quaniy. This makes hem more credible han earlier sudies ha do no, resuling in elasiciy esimaes ha are biased oward zero. Oher recen sudies have obained even larger esimaes of he shor-run price elasiciy of oil demand ranging from -.35 o -.41 using alernaive srucural models, bu he laer sudies do no accoun for sorage and are based on lower frequency daa (see, e.g., Serleis e al. 21; Bodensein and Guerrieri 212; Baumeiser and Peersman 213). 11 Esimaes of he responses of each model variable o each oil demand and supply shock are shown in Appendix A. 11

13 shocks required o produce his pah of oil producion afer November 28, holding consan he remaining srucural shocks in he model. Similar echniques are discussed in Waggoner and Zha (1999) and Baumeiser and Kilian (214). Knowledge of his sequence of flow supply shocks hen allows us o deermine how much higher he price of oil would have been under his counerfacual compared wih he acual price of oil. A more deailed discussion of his procedure can be found in Appendix B. Figure 7 shows wo measures of he counerfacual pah of world oil producion corresponding o compeing views of how large world oil producion would have been in he absence of he U.S. fracking boom. The indirec measure in he lef panel is derived from he effecs of fracking on inernaional rade in crude oil. This measure consiss of he reducion in U.S. crude oil impors from he res of he world and he crude oil equivalen of he increase in U.S. expors of refined producs, adjused for ne increases in U.S. crude oil impors from Canada, Mexico, and Venezuela, as discussed in secions 3 and 4. The direc measure in he righ panel is based on he change in shale oil producion since November 28, as discussed in secion 2. As noed earlier, he indirec measure is slighly larger han he direc measure, bu he overall paern is similar. One poenial difference is ha increased shale oil producion need no ranslae o increased producion of refined producs wihin he same monh, given bolenecks in he shipping of his oil o U.S. refineries. Insead i may be accumulaed in invenories and poenially refined a a laer poin (see Kilian 216). The indirec measure accouns for his problem and hence may be more reliable. I may involve oher approximaion errors, however, ha do no arise when consrucing he direc measure The Counerfacual Oil Supply Shock Sequences Given he change in he quaniy of oil produced in equilibrium under he counerfacual, i is sraighforward o infer how much he supply curve mus shif each monh, condiional on pas daa, in order o produce he counerfacual equilibrium quaniy. Figure 8 shows he sequence of flow supply shocks ha would have o be imposed in he global oil marke model of Kilian and Lee (214) o make global oil producion follow he counerfacual pah of world oil producion in Figure 7. An imporan 12

14 concern is wheher he flow supply shocks required o implemen he pre-specified pah of global oil producion afer November 28 are oo large by hisorical sandards or oo predicable o mainain he assumpion of a ime-invarian srucural VAR model. Figure 8 illusraes ha neiher of hese shock sequences involves shocks ha are unusually large by hisorical sandards, addressing he firs concern. As o he second concern ha he counerfacual shocks mus no be predicable, i is useful o focus on evidence of a large number of consecuive flow supply shocks having he same sign (referred o as a run ). In he upper panel he longes run observed afer December 28 is seven monhs. Runs of his lengh are no unprecedened. The longes run prior o December 28, in fac, lass nine monhs. Thus runs in he flow supply shock daa are neiher paricularly unusual nor do hey necessarily imply ha marke paricipans would have found i easy o anicipae hese shocks. Raher heir exisence is consisen wih he view ha he rapidly increasing growh rae of shale oil producion coninued o surprise marke paricipans more ofen han no. The corresponding shock sequence in he lower panel of Figure 8, in conras, exhibis one run lasing for foureen monhs, which is considerably longer han he longes run in he hisorical daa, bu similar runs are no unprecedened in oher srucural shock sequences in his model and in he lieraure more generally How Much Higher he Bren Price Would Have Been Under he Counerfacuals Figure 9 plos he Bren price of crude oil in U.S. dollars ( P of he Bren price ( P counerfacual acual ) along wih he counerfacual evoluion ) based on hese shock sequences. These esimaes are obained based on counerfacual simulaions of he log of he real price of oil in he global oil marke model discussed earlier, esimaed on daa unil Augus 215. The simulaed oil price daa have been adjused for differences beween he Bren price and he U.S. refiners acquisiion cos for crude oil impors. The Bren price of crude oil may be inerpreed as a proxy for he global price of crude oil. The lef panel of Figure 9 shows ha under he indirec measure of he counerfacual evoluion of world oil producion he 12 In a relaed conex, Kilian and Hicks (213) documened predicion errors in professional global real GDP growh forecass of he same sign for nearly five years during As long as such paerns average ou in he long run, hey do no violae he premise of raional expecaions. 13

15 Bren price of oil would have been higher by as much as $5 in 29, bu mos of he cumulaive effecs of he fracking boom would have been observed beween 211 and mid-214, wih he counerfacual price exceeding he acual Bren price by as much as $9 a imes. Thereafer, he price differenial P counerfacual P becomes negligible again. This decline can be raced o he fac ha he growh rae acual of shale oil producion slowed unexpecedly in 214 and 215. In conras, under he direc measure of he counerfacual evoluion of world oil producion he Bren price of oil would have been largely he same as he acual price unil lae 21, as shown in he righ panel of Figure 9. Beween 211 and mid- 214, he Bren price of oil would have been higher by as much as $1 in he absence of he U.S. fracking boom, reaching $12 repeaedly and a some poin even $13. Unlike in he firs counerfacual, even in 215, he Bren price would have been higher han acually observed price by as much as $5. 13 The esimaes in Figure 9 also shed ligh on he exen o which he U.S. fracking boom caused he decline in he global price of crude oil in lae 214 and early 215. Boh counerfacuals agree ha in he absence of he fracking boom his oil price decline would have occurred a he same ime, as i did, bu ha i would have sared a a higher level. According o Figure 9, he cumulaive decline in he Bren price would have been $7 under boh specificaions of he counerfacual (and herefore would have been $6 higher han in he acual daa). Thus, here is no suppor for he view ha U.S. fracking was he primary cause of his specific oil price decline. This resul complemens he analysis in Baumeiser and Kilian (216b) who were unable o pin down he quaniaive role of fracking specifically, bu who provided evidence ha a slowdown in he global demand for oil was a major conribuor o his specific oil price decline in addiion o a mix of shocks o acual and/or expeced global oil supplies prior o July 214 and a shif in oil price expecaions in July 214. Figure 9 helps refine ha inerpreaion. I also complemens he analysis of he cumulaive impac of all global oil supply shocks combined in Figure 6 13 The order of magniude of he esimaed cumulaive effecs seems reasonable in ligh of hisorical evidence on he effec of oil supply shocks. When he Iran-Iraq War broke ou in lae 198, abou 4% of world oil producion was removed from he marke, followed by approximaely a 1% increase in he price of crude oil. U.S. shale oil producion amouns o an increase of abou 4% in global crude oil producion, as shown in Kilian (216), and hence would be expeced o be associaed wih a fall in he price of oil of abou 1%, which amouns o abou $1 per barrel. 14

16 based on he srucural model of Kilian and Lee (214) Choosing Among he Two Counerfacuals The firs wo panels of Figure 1 show he price differenials P counerfacual P obained from Figure 9 acual under each of he wo alernaive counerfacuals. I highlighs ha only he indirec measure implies a posiive price differenial during 29 and early 21, whereas only he direc measure implies a posiive price differenial afer lae 214, raising he quesion of which of hese price differenials is economically more plausible. This quesion may be assessed based on he informaion in he boom panel of Figure 1, which plos he dollar spread beween he Bren price and he Wes Texas Inermediae (WTI) price of crude oil. 14 This spread conveys imporan informaion abou when he fracking boom began o affec he U.S. marke for crude oil. The fac ha he Bren-WTI spread only widened in lae 21 suggess ha he posiive price differenial during 29 and early 21 in he firs panel canno be aribued o he fracking boom, casing doub on he indirec counerfacual. As discussed in Kilian (216), radiionally hese crude oil prices have been closely racking one anoher. This paern only changed when he U.S. igh oil boom creaed a glu of crude oil in he cenral Unied Saes, causing WTI oil o rade a a discoun relaive o he global price for oil, as measured by he Bren price. We know ha he iniial drop in he WTI price relaive o he Bren price occurred because here was no enough ransporaion capaciy o ship he crude oil arriving from he Bakken and oher shale oil plays in large enough quaniies from Cushing o refiners elsewhere in he Unied Saes. As a resul, mos U.S. refiners, including he refiners along he Gulf Coas and he Eas Coas, coninued o rely on oil impors a his sage (see Kilian 216). Nor were U.S. refiners able o ake advanage of he glu of oil in Cushing o increase expors of refined producs. These ransporaion bolenecks loosened only gradually over ime, as rail, barge and pipeline capaciy increased. There are of course shale oil plays such as Eagle Ford in Texas wih easier access o refineries in he Houson area and beyond, bu producion in Eagle Ford was negligible prior o 211. Given ha 14 The Bren price refers o he spo price of ligh swee crude oil raded in London, UK, whereas he WTI price is he spo price of ligh swee crude oil for delivery in Cushing, OK. 15

17 neiher of he wo mechanisms by which shale oil producion affecs he global price of oil were in operaion a his sage of he shale oil boom, i would be unreasonable o expec shale oil producion o have had a large effec on he global price of crude oil before In shor, he evoluion of he Bren- WTI spread suppors he conclusion ha he price differenial in he second panel based on he direc measure of he counerfacual is a more accurae approximaion. The remainder of he analysis herefore focuses exclusively on he laer counerfacual Implicaions for Saudi Oil Revenue The sequence of price differenials counerfacual P acual P may be used o compue, monh by monh from December 28 unil Augus 215, he addiional revenue Saudi Arabia would have received afer November 28 in he absence of he U.S. fracking boom, given he hisorically observed levels of Saudi oil producion. The upper panel of Figure 11 shows he evoluion of he losses in Saudi oil revenue measured in billions of U.S. dollars under he preferred counerfacual based on he direc measure of igh oil producion. There is evidence ha he Saudi oil revenue losses were negligible iniially, bu gradually increased over ime, reaching 2 billion dollar per monh in lae 212 and peaking a 2.9 billion dollars per monh in mid-214, before levelling off. Of course, his evidence does no speak o he overall loss of Saudi oil revenue following he decline in he price of oil afer June 214, bu only o he componen of his loss caused by he U.S. fracking boom. I should also be noed ha he dip in Saudi oil 15 I is imporan o sress ha he Bren-WTI price differenial does no by iself represen a proper counerfacual for he global price of oil because an increase in igh oil producion all else equal will have a larger effec on he U.S. price of oil han on he global price. All we can say is ha if here is no effec on he U.S. price of oil unil lae 21, hen fracking is unlikely o have affeced he global price before his poin in ime. The evoluion of he Bren- WTI spread beween 211 and 214, in conras, ells us lile abou he counerfacual for he global price because i reflecs in imporan par he evoluion of ransporaion and refining bolenecks in he Unied Saes (see Kilian 216). 16 One may be concerned ha he model-based resuls in Figure1 are driven primarily by he choice of he price elasiciy of oil demand. This is no he case. I can be shown ha he overall paern of he cumulaive effec of he shale oil revoluion on he Bren price of oil is he same for all reasonable elasiciy values. The cumulaive effec on he price of oil rises saring in 211, peaks by mid-214 and drops hereafer. In no case is shale oil an imporan facor in driving he oil price decline in he second half of 214. Alhough here is no monoonic relaionship beween his elasiciy and he impac of he U.S. shale oil boom on he price of oil, i is difficul o generae much larger cumulaive effecs for reasonable elasiciy values. For example, he cumulaive effec on he Bren price is bounded by $6.54/barrel for an elasiciy of -.21 and by $3.86/barrel for an elasiciy of -.31, compared wih $9.74/barrel in he baseline model. For an elasiciy of -.11, ouside of he range of reasonable values, he effec rises o $14.59/barrel, bu he paern remains similar. 16

18 revenue losses in January 215 mirrors he emporary ighening of he Bren-WTI spread, adding credibiliy o his esimae. A rough measure of he cumulaive impac of he fracking boom on Saudi Arabia is obained by simply cumulaing hese losses. The cumulaive decline in Saudi oil revenue beween December 28 and Augus 215 is 12.1 billion dollars. This amouns o an average loss per monh of 1.3 billion U.S. dollars. The incremenal Saudi oil revenue losses beween July 214, when he price of oil began o fall precipiously, and Augus 215 are 24.3 billion dollars. To pu hese esimaes in perspecive, he lower panel of Figure 11 expresses he reducion in Saudi oil revenues caused by he fracking boom as a percenage of he Saudi oil revenue in November 28, when he Bren price of oil sood a $ I is imporan o be clear abou he inerpreaion of his esimae. Figure 11 shows ha in mid-214, for example, Saudi Arabia in he absence of he U.S. fracking boom would have earned 12% of he oil revenue i generaed in November 28, if i had produced exacly as much oil as i acually did in mid In oher words, he plo does no measure he acual change in he oil revenue generaed by he Saudi economy, bu raher he los revenue relaive o wha Saudi Arabia would have earned if he Unied Saes sill relied on convenional crude oil producion only A Case Sudy of Saudi Arabia Saudi Arabia s sraegy for dealing wih he fracking boom so far has been o preserve is marke share and o avoid idle capaciy (see, e.g., Faouh and Sen 215; Faouh, Poudineh and Sen 215). There is no sign of Saud Arabia having curailed is oil producion nor is i likely ha a unilaeral sraegy of 17 I should be kep in mind ha his esimae is only an approximaion, firs, because Saudi oil expors ha were divered from he Unied Saes o Asian economies, for example, were no necessarily priced based on he Bren benchmark, and, second, because my mehodology absracs no only from changes in Saudi oil producion, bu also from any Saudi effors o marke is oil more aggressively. Wih regard o he firs poin, my analysis assumes ha he price of Saudi crude oil declined as much as he price of Bren crude, allowing for differences in he level of hese prices. My analysis also absracs from differences beween Saudi expors of heavy and ligh crude oil. The concern is no so much ha Saudi Arabia could have subsiued expors of heavy crude o he Unied Saes for is expors of ligh crude. This possibiliy seems unlikely because, as I showed, impors of lower-priced Canadian heavy crudes increasingly subsiued for U.S. impors of heavy crudes from oher sources, a he same ime as shale oil producion displaced U.S. impors of ligh swee crude oil. Raher he concern is ha by applying he counerfacual price differenial indiscriminaely o Saudi producion of ligh and heavy crude oil, one overesimaes he loss in oil revenue ha can be aribued o shale oil because shale oil compees wih Saudi heavy crude oil only indirecly hrough U.S. expors of refined producs. 17

19 resricing Saudi oil producion would have helped preserve Saudi oil revenue (see Baumeiser and Kilian 216b). Insead, he Saudi governmen has deal wih he decline in is oil revenue mainly by apping is financial reserves, as prediced by sandard models of precauionary savings (see, e.g., Bems and de Carvalho Filho 211). IMF daa show a reducion in Saudi ne foreign asse holdings of near $9 billion beween mid-214 and Augus 215. This esimae is considerably larger han he loss in oil revenue over he same period ha is aribuable o he fracking boom. The laer we esimaed o be abou $24 billion. In oher words, he analysis in his paper enaively suggess ha in he absence of he fracking boom he reducion in Saudi foreign exchange reserves since mid-214 would have been abou wo hirds of he acual decline wih he remainder reflecing higher crude oil producion elsewhere in he world, shifs in oil price expecaions affecing he demand for oil socks, and he effecs of a slowing global economy. An imporan quesion is how long Saudi Arabia can susain such losses in is foreign exchange reserves. If he decline in Saudi ne foreign asses were o coninue a he rae experienced beween January 215 and January 216, one would expec Saudi ne foreign asses o be exhaused by early 22. Of course, his process would accelerae if he price of oil were o fall furher, or i would slow if he price were o recover somewha. There are wo poenial miigaing facors, however. Firs, unlike some of is neighbors, Saudi Arabia has very lile exernal deb because i resised he empaion o leverage is increased oil revenues during he 23-8 oil price boom by running up exernal deb (see Tornell and Lane 1998). As a resul, for now, he Saudi governmen has been able o borrow in global financial markes, given he expecaion ha he price of oil will ulimaely recover. Second, he Saudi governmen hus far has been relucan o impose fiscal rerenchmen, because of he poliical coss of such measures, bu in December 215 i announced a budge plan ha envisions cus in spending from $975 billion in 215 o $84 billion in 216 (represening a 14% drop in governmen spending). The finance minisry also announced ha i would raise axes and adjus subsidies for waer, elecriciy and peroleum producs over he nex five years. The cenral quesion for Saudi Arabia is o wha exen i should rely on exernal borrowing and o wha exen i should cu fiscal spending. Economic heory ells us ha i makes sense for Saudi Arabia 18

20 o borrow in response o a shor-lived fall in oil prices. If he decline in he price of oil were expeced o persis for a long ime, in conras, exernal borrowing would no be he appropriae response and he adjusmen insead would have o come from fiscal rerenchmen. Thus, he answer o he quesion of how much fiscal adjusmen Saudi Arabia requires is ulimaely deermined by how long one expecs low oil prices o persis. How soon he price of oil is expeced o recover depends in imporan par on why he price of oil declined and how imporan hese deerminans will remain in he fuure (see Baumeiser and Kilian 216b). For example, o he exen ha his price decline was caused by he U.S. fracking boom, he quesion is how long his boom can persis a curren prices wih many commercial igh oil producers already experiencing heavy operaing losses, and how quickly hese firms could resume igh oil producion a higher oil prices, if hey were forced o close down. The same concern applies o unconvenional oil producion in Canada. Even if he curren low price of oil were o pu shale oil producers ou of business, however, an obvious concern would be ha shale oil producion is likely o resume as soon as world oil prices recover sufficienly, as long as here remains easily accessible shale oil in he ground. A $6 per barrel, for example, many U.S. shale oil producers are likely o generae profis, even if hey do no a $4 per barrel. If we accep his reasoning, we should rea much of he shale oil componen of he Saudi foreign exchange losses as persisen and respond by cuing Saudi fiscal spending raher han borrowing. Of course, shale oil is only one reason for he ample supply of crude oil in world markes. To he exen ha oher oil producers including Russia and Iraq have expanded oil producion in recen years, or, as in he case of Iran, are abou o do so, he quesion becomes one of how long sae-conrolled oil producers wih few oher opions o generae foreign exchange and no accounabiliy o shareholders or financial markes will pu pressure on he price of oil. Counries such as Russia, for example, rely on oil revenue o susain heir economy. There is no reason o expec hese counries o reduce heir expors, excep o he exen ha inadequae invesmen in he oil secor limis heir abiliy o susain high levels of oil producion in he long run. 19