Economic Effects of Petroleum Prices and Production in the Gulf of Mexico OCS on the U.S. Gulf Coast Economy

Transcription

1 OCS Study MMS Coastal Marne Insttute Economc Effects of Petroleum Prces and Producton n the Gulf of Mexco OCS on the U.S. Gulf Coast Economy Trends n Unemployment Rates n the U.S. Coastal Gulf States Unemployment Rates, % ALQUR LAQUR MSQUR TXQUR U.S. Department of the Interor Mnerals Management Servce Gulf of Mexco OCS Regon Cooperatve Agreement Coastal Marne Insttute Lousana State Unversty

2 OCS Study MMS Coastal Marne Insttute Economc Effects of Petroleum Prces and Producton n the Gulf of Mexco OCS on the U.S. Gulf Coast Economy Authors Omowum O. Iledare Wllams O. Olatub October 2006 Prepared under MMS Contract CA by Lousana State Unversty Center for Energy Studes Baton Rouge, Lousana Publshed by U.S. Department of the Interor Cooperatve Agreement Mnerals Management Servce Gulf of Mexco OCS Regon Coastal Marne Insttute Lousana State Unversty

3 DISCLAIMER Ths report was prepared under contract between the Mnerals Management Servce (MMS) and Lousana State Unversty s Center for Energy Studes. Ths draft report has not been techncally revewed by MMS. Approval does not sgnfy that the contents necessarly reflect the vew and polces of the Servce, nor does menton of trade names or commercal products consttute endorsement or recommendaton for use. It s, however, exempt from revew and complance wth MMS edtoral standards. REPORT AVAILABILITY Extra copes of the report may be obtaned from the Publc Informaton Offce (Mal Stop 5034) at the followng address: U.S. Department of the Interor Mnerals Management Servce Gulf of Mexco OCS Regon Publc Informaton Offce (MS 5034) 20 Elmwood Park Boulevard New Orleans, Lousana Telephone Number: GULF or Suggested Ctaton: CITATION Iledare, O.O. and W.O. Olatub Economc Effects of Petroleum Prces and Producton n the Gulf of Mexco OCS on the U.S. Gulf Coast Economy. U.S. Dept. of the Interor, Mnerals Management Servce, Gulf of Mexco OCS Regon, New Orleans, LA. OCS Study MMS pp. ACKNOWLEDGMENTS Ths report s based prmarly on OCS leasng records made avalable to the Center for Energy Studes by the Mnerals Management Servce, New Orleans. Barbara Kavanaugh, Versa Stckle, Rc Pncomb, Yan Zhang, and Amar Dave were very helpful n obtanng and processng these data.

4 ABSTRACT The purpose of ths study s to analyze the dynamc nteracton between changes n crude ol prces, ol and gas ndustry actvty n the OCS (measured n terms of petroleum producton) and selected ndcators of the Gulf Coast economes. The scope of the study s expanded to nclude E&P actvty n the deepwater. A vector auto-regresson (VAR) model framework showng the nteracton between crude petroleum prce, ol and gas producton, the U.S. nterest rates, the U.S. gross domestc product, and selected ndcators of the state of the Gulf Coast economy personal ncome, unemployment rate and revenue was developed and estmated. The model framework enables us to establsh the drecton, symmetry, causaton, duraton, responsveness, and correlaton between ndustry actvty and state economc actvty ndcators and ol prce changes over tme. The emprcal results show that changes n crude ol prces have sgnfcant effects on ol and gas producton n the Gulf of Mexco OCS and on measures of the Gulf Coast economy. The effects of ol prces on the state of the economy n the Gulf Coast are two-pronged. There s an establshed drect effect on the macroeconomc aggregates and there s also an ndrect effect through producton actvty. As expected, the results show that the magntude and duraton of a crude ol prce shock on the state of the economes n the Gulf States, as well as ol and gas producton, dffer sgnfcantly by state. In a broad sense, the study shows that whle the natonal economy may have become less senstve to ol prce shocks n the aggregate, the Gulf Coast economes are stll prone to ol prce shocks, albet wth varatons across the states n the Gulf Coast. Thus, the study reaffrms the need to be cautous about polcy responses that tend to focus only on the natonal response to polcy ssues wth regonal mplcatons. The assumpton that such natonal response s applcable or approprate across regons may be erroneous. Ths demonstrates that understandng the dynamc of ol prces and ther mpacts on macroeconomc aggregates n and wthn the regons/states are as mportant as ever, even as mtgatng natonal polces and response strateges evolve. v

5 TABLE OF CONTENTS Page LIST OF FIGURES... x LIST OF TABLES... x EXECUTIVE SUMMARY.... INTRODUCTION Background Study Objectves Regonal Scope of the Study DATA SOURCES AND DESCRIPTIVE ANALYSIS Sources of Data Key Indcators of Economc Performance VAR MODELING OF THE ECONOMIC EFFECTS OF PETROLEUM PRODUCTION AND PRICES VAR Model Specfcaton Emprcal VAR Model Representaton VAR Model Estmaton and Analyss ESTIMATED VAR MODEL RESULTS: VARIANCE DECOMPOSITION ANALYSIS VAR Results from OCS Aggregate Producton System Equatons OCS Petroleum Producton and the Lousana Economy OCS Petroleum Producton and the Alabama Economy OCS Petroleum Producton and the Msssspp Economy OCS Petroleum Producton and the Texas Economy VAR Results from OCS Deepwater Producton System Equatons OCS Deepwater and the Lousana Economy OCS Deepwater and the Alabama Economy OCS Deepwater and the Msssspp Economy OCS Deepwater and the Texas Economy ESTIMATED VAR MODEL RESULTS: IMPULSE RESPONSE FUNCTION APPROACH IRF Results from OCS Aggregate Producton System Equatons Prce Shock, Gulf OCS Producton, and the Lousana Economy Prce Shock, Gulf OCS Producton, and the Alabama Economy Prce Shock, Gulf OCS Producton, and the Msssspp Economy Prce Shock, Gulf OCS Producton, and the Texas Economy...35 v

6 TABLE OF CONTENTS (Contnued) Page 5.2. IRF Results from OCS Deepwater Producton System Equatons Prce Shock, OCS Deepwater Producton, and the Lousana Economy Prce Shock, OCS Deepwater Producton, and the Alabama Economy Prce Shock, OCS Deepwater Producton, and the Msssspp Economy Prce Shock, OCS Deepwater Producton, and the Texas Economy ECONOMIC INTERPRETATIONS OF THE VAR MODEL RESULTS SUMMARY AND CONCLUSIONS...5 REFERENCES...53 APPENDIX A AN OUTLINE OF THE VAR PROCEDURE...55 v

7 LIST OF FIGURES Fgure Descrpton Page. Trends n Annual Revenue of the U.S. Gulf States Trends n Quarterly Personal Income of the U.S. Gulf States Trends n Unemployment Rates n the U.S. Coastal Gulf States Trends n Crude Petroleum Prce Index, Gulf of Mexco OCS Petroleum Producton by Water Depth Category Lousana Personal Income and OCS Producton Dynamc Paths Lousana Unemployment and OCS Producton Dynamc Paths Dynamc Paths of Lousana Revenue and OCS Producton Responses of Gulf Producton & AL Unemployment Rate to Prce Responses of Gulf Producton & AL Personal Income to Prce Responses of Gulf Producton & AL Revenue to Prce Responses of Gulf Producton & MS Unemployment Rate to Prce Responses of Gulf Producton & MS Personal Income to Prce Responses of Gulf Producton & MS Revenue to Prce Responses of Gulf Producton & TX Unemployment Rate to Prce Responses of Gulf Producton & TX Personal Income to Prce Responses of Gulf Producton & TX Revenue to Prce Responses of Deepwater Producton & LA Unemployment to Prce Responses of Deepwater Producton & LA Personal Income to Prce Responses of Deepwater Producton & AL Unemployment Rate to Prce Responses of Deepwater Producton & AL Personal Income to Prce Responses of Deepwater Producton & MS Unemployment Rate to Prce Responses of Deepwater Producton & MS Personal Income to Prce Responses of Deepwater Producton & TX Unemployment Rate to Prce Responses of Deepwater Producton & TX Personal Income to Prce...44 x

8 LIST OF TABLES Table Descrpton Page. Varable Names, Descrptons, and Transformaton Method Correlaton Matrx of Model Varables...5 3a. Quarterly Summary Statstcs of Model Varables, 976:-999:...5 3b. Annual Summary Statstcs of Model Varables, Decomposton of the Varance of Macroeconomc Varables Due to Changes n Petroleum Prces and OCS Gross Petroleum Producton Decomposton of the Varance of Macroeconomc Varables Due to Changes n Petroleum Prces and OCS Deepwater Petroleum Producton Estmated Range of the Impact of Changes n Prce and OCS Producton on Macroeconomc Varables Usng the Impulse Response Functon Technque (%) Estmated Range of the Impact of Changes n Prce and Deep OCS Producton on Macroeconomc Varables Usng the Impulse Response Functon Technque (%) Prce Elastcty of Macroeconomc Varables and the Quantty Equvalence Condtonal on the Dynamcs of OCS Petroleum Producton and the Gulf Coast Economy Estmated Adjustment Paths to Equlbrum Followng a Prce Shock Impact on Aggregate OCS Petroleum Producton and the Economy...50 x

9 EXECUTIVE SUMMARY There s a general consensus that declnng ol prces stmulate economc growth whle ncreasng ol prces tends to dampen economc performance; the effects are not generally conclusve, however, for sub-natonal economes. Whle the effects of changes n ol prce structure on the U.S. natonal economy are generally understood, the mpacts of such changes on the state or sub-regonal economes are less fully examned. Very few studes have studed the mpact of changes n crude ol prce on state economc performance, and such studes tend to conclude that a rsng ol prce more often than not stmulates economc growth n ol exportng states and hnders growth n ol mportng states. The converse s true for declnng ol prces. For effectve polcy and regulatory gudance wthn the context of the overall natonal energy polcy, agences such as the MMS need relable nformaton at the regonal levels, where most relevant ol and gas actvtes take place. Ths s because each state or regon often possesses unque characterstcs that are at varance wth natonal outlooks. Therefore, such unque stuatons requre a dfferent polcy or regulatory framework. Accordngly, ths study s proposed to fll these gaps by extendng prevous natonal studes to sub-natonal economes, especally to areas where MMS has jursdctonal mandates. Ths study analyzes the nteractons between crude ol prces, ol and gas ndustry actvty n the Outer Contnental Shelf (OCS), and selected economc ndcators of the Gulf Coast States. Total revenue, personal ncome, and the unemployment rate of four states n the U.S. Gulf Coast are used as proxes for measurng the strength of the U.S. Gulf Coast economy. The states were selected on the bass of some unque structural and economc characterstcs as specfed below: Lousana: Represents net ol exporter wth lmted dversfed economy; Msssspp: Represents net ol mporter wth lmted dversfed economy; Texas: Represents net ol exporter wth relatvely dversfed economy; Alabama: Represents net ol exporter wth lmted dversfed economy. Three key ndcators for measurng E&P ndustry actvty and performance that are hghly correlated wth crude ol prce movements both n the short run and long run have been dentfed as drllng rg counts, producton, and captal expendtures. However, due to data lmtatons, ol and gas producton was used as a proxy for measurng the trends n E&P ndustry actvty. The scope of the study was also expanded to cover deepwater operatons. Data and Methods: The data used for ths study are bascally secondary n nature from varous sources. The data source for the ol and gas producton seres s from the MMS ol and gas database. The ol prce data s the crude ol producer prce ndex deflated by the all commodtes prce ndex seres, whch s avalable from the U.S. Bureau of Labor Statstcs. Data for unemployment rates are avalable from the U.S. Bureau of Labor Statstcs. The Bureau of Economc Analyss (BEA) s the source of the followng data: the quarterly personal ncome and the annual revenue seres for the states, U.S. real GDP, GDP mplct deflator and nterest rates. A vector auto-regresson (VAR) model framework showng the nteractons between crude petroleum prce, ol and gas producton, the U.S. nterest rates, the U.S. gross domestc product,

10 and selected ndcators of the Gulf Coast States economes personal ncome, unemployment rate and revenue was developed and estmated. The VAR approach has been used generally for forecastng systems of nterrelated tme seres and for analyzng the mpact of a random dsturbance on a system of varables. In ths formulaton, every endogenous varable s modeled to depend on ts own lag(s), lags of other endogenous varables, and any exogenous varables that may also be ncluded. Varance decomposton and mpulse response functons represent two complementary ways to characterze the dynamc effects of an unexpected shock to a gven economc system that s represented by a VAR model. The varance decomposton procedure provdes a way to decompose the effects of a shock on the system to ther component parts. The percentage share of the effect of each partcular shock provdes an ndcaton of ts relatve potency n explanng the observed varatons n each varable experencng the shock. The mpulse response functons, on the other hand, provde a way to examne the paths of the effects of an exogenous shock of one varable on other varables and to further characterze the stablty and ts duraton for the varables. The persstence of such shocks reveals the pace and pattern of the adjustment process of the system to ts long-run equlbrum. The faster t takes a shock to dampen, the faster the adjustment process back to equlbrum (Brown and Yucel, 995). Economc Effects of Ol Prce and E&P Actvty n the Gulf OCS: On Lousana Economy: The dynamc VAR analyss of the nteractons between changes n crude ol prces, ol and gas producton n the Gulf of Mexco (GOM) OCS and Lousana unemployment rate shows that prce s sgnfcant and t explans, on average, approxmately percent of the observed varaton n unemployment over tme. Crude ol prce nteractng wth ol and gas producton n the Gulf of Mexco OCS also explans at most 4 percent of the expected varaton n personal ncome and between to 6 percent of the varaton n revenue. To our surprse, however, autonomous ol and gas producton has no drect sgnfcant effect on unemployment accordng to our VAR results. The mpulse response results present the adjustment paths assocated wth prce shocks. The results show that t can take more than 0 years for unemployment, about 3 years for personal ncome, and up to 20 years for revenue to be restored to ntal equlbrum. On Alabama Economy: The model results descrbng the nteractons between ol prce, ol and gas producton n the Gulf of Mexco OCS and Alabama unemployment rate shows that prce explans up to 30 percent of the expected varaton n Alabama unemployment. The results also show that a prce shock condtonal on OCS ol and gas producton profle explans up to percent of the observed varaton n personal ncome n Alabama. Further, a prce shock exposed to ol and gas producton path n the Gulf also has a potental mpact of at most 29 percent n the long-term on Alabama revenue. The ensung mpulse response functons reveal the adjustment paths for the nteractons between a prce shock and gross ol and gas producton n the Gulf. The response paths show that t takes approxmately 6 years for unemployment, 2 years for personal ncome, and 2 years for revenue to be restored to ther ntal equlbrums subsequent to the shock. The autonomous drect mpact of ol and gas producton n the Gulf OCS on Alabama unemployment s also not sgnfcant, accordng to the VAR model results. 2

11 On Msssspp Economy: The model results, whch descrbe the nteractons between ol prce, ol and gas producton n the Gulf OCS, and Msssspp economc varables show that the percentage of the varaton n the state s unemployment accounted for by prce s less than 0 percent on the average, but sgnfcant. Smlarly, the emprcal results ndcate that the effect of prce on personal ncome subject to OCS producton path can be up to 5.5 percent. The prce mpact on revenue accordng to the VAR model results s as hgh as 6.7 percent. Further analyss of the mpulse response results and subsequent adjustment paths to a prce shock ndcate that unemployment rate takes more than 8 years, personal ncome takes about 2 years, and revenue takes 5 years to adjust to ther ntal equlbrum levels. Just as s the case wth Lousana and Alabama, ol and gas producton n the Gulf has no drect sgnfcant mpact on the state unemployment rate. On Texas Economy: The estmated model results of the effect of ol prce nteractons wth Gulf ol and gas producton and state economc varables wth respect to the Texas economy show that the mpact of a prce shock on the Texas unemployment rate s relatvely small, although sgnfcant. As much as 9 and 8 percent of the varatons n personal ncome and revenue n the state are explaned by prce shocks, respectvely. Wth regard to the adjustment paths over tme, unemployment rate takes less than 0 years, personal ncome takes approxmately 4 years, and revenue takes about 7 years for ntal equlbrum to be restored. The effect of OCS producton on Texas unemployment rate, unlke n the other Gulf States, s sgnfcant, but small. Economc Effects of Ol Prces and Deepwater E&P Actvty n the Gulf OCS: On Lousana Economy: The estmated model results for the nteractons between ol prces and deepwater ol and gas producton show that the effect of a prce shock on Lousana unemployment s relatvely small (2 percent). However, deepwater producton shows no sgnfcant and drect mpact on the state unemployment. The model results further show that a prce shock, condtoned on OCS deepwater producton path, explans as hgh as 6.5 percent of the varaton n Lousana personal ncome. The paths of adjustment to prce changes f deepwater producton s restrcted show a lag of 8 quarters for unemployment and 6 quarters for personal ncome. On Alabama Economy: The model results explanng the dynamc nteractons between prce and deepwater ol and gas producton show that prce explans up to 33 percent of the varablty of Alabama unemployment. Also, wthn the same context, but unlke the estmated effect on Lousana, OCS deepwater producton has a hghly sgnfcant effect on Alabama unemployment. OCS deepwater producton explans up to 22 percent of the observed varaton n Alabama unemployment. The effect of prce shocks on personal ncome n Alabama s also found to be sgnfcant wth respect to deepwater producton. Under that scenaro, a prce shock explans up to 5.9 percent of the varaton n personal ncome as well. On Msssspp Economy: The VAR results descrbng the effect of a prce shock and ol and gas producton from OCS deepwater on Msssspp economc varables show that the changes n prce explan a relatvely small proporton of the observed varaton n unemployment (roughly 4 percent). A shock to deepwater producton has a sgnfcant effect on unemployment. Prce shocks explan up to 5.3 percent of the observed varaton n personal ncome. The paths of 3

12 adjustment to prce changes subject to deepwater producton profle show a lag of 3 quarters for unemployment and 6 quarters for personal ncome. On Texas Economy: Accordng to the VAR model results, the response of Texas unemployment to changes n ol prce subject to the nteractons between ol and gas producton from OCS deepwater and prce s not statstcally sgnfcant. However, deepwater producton has a drect and sgnfcant mpact on unemployment. The results further suggest that prce shocks explan up to 6.3 percent of the observed personal ncome varaton. The paths of adjustment to prce changes nteractng wth deepwater producton from the Gulf OCS show a lag that s more than 24 quarters for unemployment and 0 quarters for personal ncome. In an overall sense, ths study suggests that petroleum producton n the Gulf of Mexco OCS responds postvely to a postve prce shock n the economy. Further, the study shows that: Unemployment rates n coastal Gulf States n the U.S. tend to declne n response to ncreases n petroleum prces. It s nterestng to note, however, that the responsveness of unemployment rates to changes n prces dffer sgnfcantly across the Gulf States. Texas has the least unemployment responsveness to a prce shock and Alabama has the hghest among the four Gulf States. Accordng to the VAR model results, personal ncome tends to ncrease followng a postve shock to petroleum prces n the presence of rsng petroleum producton. The degree of ncome responsveness to prce shocks vares across the U.S. Gulf States. In general, personal ncome responsveness n Texas s greater than that of Msssspp, Lousana, and Alabama, n that order. The emprcal results also suggest that the Texas economy, because of ts sze, tends to experence a more lngerng path to adjustment for personal ncome than Lousana, Msssspp and Alabama. Smlarly, personal ncome n Lousana tends to experence more lngerng effects than Msssspp and Alabama followng a petroleum prce shock. Postve changes n petroleum prces lead to ncreases n annual revenue n Lousana, Texas, and Alabama. The responsveness of revenue to prce changes, however, vares across Gulf States just as changes n unemployment and personal ncome vary across the Gulf States. Surprsngly, unemployment rates n the Gulf States appear to be relatvely less senstve to producton actvtes n the Gulf States than expected. In many nstances n the U.S. Coastal Gulf States, the drect mpacts of changes n producton on unemployment rates are nsgnfcant. 4

13 Fnally, there s statstcal evdence suggestng sgnfcant dfferences n the duraton of the lngerng effects of a prce shock on the economc performance of the Coastal Gulf States we nvestgated n ths study. 5

14 . INTRODUCTION.. Background The Mnerals Management Servce, a federal agency n the U.S. Department of the Interor, manages more than one bllon offshore acres and has collected about 4-5 bllon dollars n mneral revenues annually over the past fve years (USDOI, MMS, 2003). The Gulf of Mexco OCS regon accounts for about 25 percent of the ol and gas produced n the U.S. (USDOE, EIA, 2002). Thus, the ol and gas ndustry n the Gulf Coast s mportant to the naton s economy, especally to the states n the U.S. Gulf Regon. Hence, whatever happens n the ol market portends a certan trend for the natonal or regonal economes, ether n the short or long run. Perhaps the most mportant varable n the ol market s crude ol prces. Thus, a few economc mpact studes supported by the MMS have focused on the effect of ol prces on the economes of Gulf of Mexco (GOM) communtes. Ths s because ol prces, n addton to affectng the revenue base of adjacent states and communtes, also have profound effects on the profts of ol companes operatng n the regon, and consequently, the levels of ndustry actvtes n the GOM. Over the past three decades, polcy makers have become overtly concerned wth the effects of ol prces on the economc performance of natons or regons. The very hgh ol prces n the 970s and the very low prces n the md-980s and the early 990s amplfy these concerns. Most studes of natonal economes have concluded that changes n ol prce sgnfcantly affect varatons n macroeconomc aggregates and hence, the growth of economes. It s generally agreed that a declnng ol prce stmulates economc growth whle an ncreasng ol prce tends to dampen economc performance. These effects are often exacerbated dependng on whether the naton s net ol mportng or net ol exportng. The semnal work by Hamlton (983) lad the foundaton for the observed lnkage between crude ol prce movements and the level of economc actvty n the U.S. economy. Other studes have snce been revealed to debunk or support Hamlton s clam that a sharp ncrease n ol prces produced the recessons between the end of World War II and early 980. For example, Consdne (988) shows that gans n output and employment n the U.S. economy were relatvely small followng the 986 drop n ol prce. Tatom (988) also shows that changes n ol prces affect real GNP, productvty, and terms of trade and that these effects are asymmetrc. A revew of several econometrc models by Hckman (984) supports Tatom s conclusons. On the other hand, Prescott (986) mantans that ol prce shocks have lttle or no effect on natonal producton possbltes. These studes suggest that a sustaned declne or ncrease n ol prce and ts effects on natonal economes can be predcted to some degree. Although the effects of ol prce on the natonal economy are generally understood, the mpacts of such movements on state or sub-regonal economes are less fully examned. Few studes (Brown and Hll, 988; Brown and Yucel, 995; Yucel and Guo, 994) have studed the effects of crude ol prce on state economc performance. Most of these studes, unlke natonal studes, tend to show that a rsng ol prce stmulates economc growth n ol exportng states and hnder growth n energy mportng states. The reverse s the case for declnng ol prces. These studes 7

15 also mply sustaned declnes or ncreases and ther effects can be ascertaned and polcy responses desgned approprately. For effectve polcy and regulatory gudance, keepng n perspectve the overall natonal energy polcy objectves, agences such as the MMS desre relable nformaton at the regonal levels, where ol and gas actvtes take place. Ths s because each state or regon possesses unque operatng envronments that are at varance wth the natonal outlook. Hence, a dfferent polcy or regulatory framework s requred. The purpose of ths study s to fll these nformaton gaps by extendng prevous natonal studes to sub-natonal economes, especally to areas where MMS has jursdctonal mandates n the Gulf of Mexco OCS regon. Analyses of mcroeconomc data at the level of ndvdual ndustres, frms, or workers have establshed that there s sgnfcant correlaton between ol prce shocks and output, employment, or real wages (Keane and Prasard, 996; Davs et. al., 996; Lee and N, 999). For example, an ncrease n the prce of ol leads to an upward shft n frm s cost curve, reducng proft levels, and hence, lowerng employment and subsequently output. Such an ncrease may lead to substtuton away from ol to other nputs wth potental for further pressure on the cost curves as the prces of those nputs rse. A prce declne n the ol market may have the opposte effect, although not lkely of a smlar magntude. At the general economc level, because of lnkages n the economy among sectors, an ncrease n ol prce may nduce nflaton, hence, the noton that ncreases n ol prce have preceded most recessons n the U.S. (Carruth et al., 998). Changes n crude ol prces do affect revenue and the personal ncomes of communtes n many natons where the ol and gas ndustry looms large n the economy. For most ol producng regons, tax revenue from ol s a major source of general fscal revenue, hence, a declne or ncrease n the levels of frm s profts can nfluence ths tax base sgnfcantly. Besdes, an ncrease n ol prce wll generally nduce cost-cuttng measures by frms. The frst casualty n ths stuaton s often labor nputs. To get to equlbrum followng an energy prce ncrease as frms cut output and employment, wages are often cut, thus ncome of households wll become negatvely affected. In ol mportng natons, ol prce ncreases may be nflatonary and lead to a dramatc fall n real wages and ncome. In a general sense, the above theoretcal descrpton may be true for natonal or cross-natonal economes, but the realty may be dfferent and more complex n some states or regons. For example, an exporter of ol may receve some benefts from an ol prce ncrease, but the non-ol frms located n that regon may face ncreases n nput costs. The converse may be true n an ol mportng state. On the other hand, a decrease n ol prce may also produce a depressed demand n some sectors of the state economy, and unemployed labor s not mmedately shfted elsewhere. Ths effect may be qute pronounced because states wthn natons may possess economes wth rgdtes that are substantally dfferent from ther natonal economes as a whole. The overall net effect n each case s therefore subject to emprcal verfcaton and therefore the relevance of our focus on state-level analyses. Potental structural rgdtes and the degree of sectoral dependences n a partcular regon s economes wll largely nfluence ths stuaton. A regon wth a hgh concentraton of ol dependent sectors wll be especally complex to analyze. 8

16 .2. Study Objectves Ths study develops economc and econometrc models that examne the effects of changes n crude ol prces on both the E&P ol ndustres and the relevant regonal economes n the Gulf of Mexco. The research uses recent econometrc tools to provde quanttatve estmates of the responsveness and correlaton between past and current actvtes of the ol ndustres and Gulf States economc growth and ol prce changes and volatlty. Specfcally, the followng objectves are addressed: examne the changes n some specfc economc ndcators of E&P actvtes of the OCS ol ndustres as a result of ol prce changes and prce volatlty over tme; examne the type of relatonshps that exst between ol prce changes and the level of economc actvtes of the Gulf Regon; forecast potental mpacts of future changes n ol prces on ndustry actvtes and state aggregate economc varables; and dentfy possble polcy responses to these changes by the ndustry and the relevant government n the Gulf. In order to meet the above challenges, recent developments n tme seres econometrc modelng tools are employed. These tools enable us to establsh the drecton, causaton, duraton, responsveness, and correlaton between ndustry and states economc actvty ndcators and ol prce changes over tme..3. Regonal Scope of Study Ths study covers selected representatve states n the GOM Regon. Specfcally, we selected the followng states based on ther unque structural and economc characterstcs specfed n each case. Lousana: Represents net ol exporter wth lmted dversfed economy; Msssspp: Represents net ol mporter wth lmted dversfed economy; Texas: Represents net ol exporter wth relatvely dversfed economy; Alabama: Represents net ol exporter wth lmted dversfed economy. In terms of ndustry-level, the project focuses on two levels of actvtes. Frst, at the aggregate level the study examnes ol prce mpact on ndustry and state-level macro-aggregates usng ndustry actvtes for the entre OCS n the GOM. It s hoped that the results of such analyss wll provde a broad pcture of the potental mpact of ol prce drven polcy varables of OCS actvtes on the ndvdual state. Second, because MMS polcy s often appled by plannng area or water depth, the study repeats the same exercse at a more dsaggregated level of ndustry actvty n the deepwater. 9

17 2. Sources of Data 2. DATA SOURCES AND DESCRIPTIVE ANALYSIS Most of the prevous research on the economc effects of ol prce shocks on macroeconomc varables have reled on natonal data, whch are easly avalable from a varety of sources. One of the reasons for paucty n regonal/state-level analyses s because relable sources of statelevel nformaton n the preferred format are lmted. The data collecton efforts n ths study were very focused on fndng accurate sources of data that are both comprehensve and tenable. In order to establsh the robustness of our model, both from statstcal and economc theory perspectves, we also used other U.S. macroeconomc aggregate data n the estmaton procedures. The natonal-level aggregate economc varables used n the model nclude quarterly and annual data on real gross domestc product, crude ol producer prce ndex, all commodtes prce ndex, nterest rates (the 3-month treasury bll rates), and mplct gross domestc product deflator seres. These natonal-level aggregate data are mportant nputs nto the decson makng process of the ol and gas ndustry for makng exploraton and producton nvestment decsons. For example, gven an ol prce, the choce of the level of nvestment and hence, potental ndustry output, may be drven by the prevalng nterest rates. Wth regards to the states, t s also expected that states economc varables at the state-level wll to a large extent correlate wth mportant natonal aggregates such as the overall GDP, whch measures natonal economc output n the U.S. The data on ol and gas producton came from the MMS ol and gas database. The ol prce s the crude ol producer prce ndex deflated by the all commodtes prce ndex. Both seres are avalable from the U.S. Bureau of Labor Statstcs. The data on unemployment rates for the states also came from the U.S. Bureau of Labor Statstcs. The Bureau of Economc Analyss (BEA) s the source for the followng economc varables: quarterly personal ncome and annual revenue data for the states; U.S. real GDP, GDP mplct deflator and nterest rates. Table descrbes the nature of the data used n ths study. Data for the followng state macroeconomc aggregates: personal ncome, unemployment rate, and OCS ol and gas producton are reported quarterly. However, for any aspects of the analyses whch requre the use of state revenue data, we have used annual seres of the relevant varables. Ths was because quarterly data on revenue at the state level was not avalable. Table shows the data that were avalable for dfferent tme spans for each of the varables. The modelng framework appled s thus restrcted by the most tme-lmtng seres. Crude ol equvalence as a measure of ol and gas actvtes n the OCS was used.

18 2.2. Key Indcators of Economc Performance The followng macro-aggregates 2 or ndcators are used as proxes for gaugng the economc strength at the state-level: Revenue: Many GOMR States derved a large percentage of ther budgetary revenue from the ol and gas ndustry located n ther areas and some have ndustry sectors that are hghly energy-dependent; Unemployment: A lot of people n most of the states n the GOMR are employed drectly or ndrectly n the ol and gas sector, hence, any unusual developments n the sector wll reflect on states welfare; unemployment level s one such closely watched varable; Personal Income: Apart from the substantal number of jobs produced by the ol and related sectors, wages n the ol sectors are often hgher than other sectors, thus overall personal ncome levels n the states may be affected by downturns or booms n the ol sector. There are several potental ndcators of ndustry actvtes n the ol and gas ndustry. Three key ndcators of the level of economc actvtes n the ol and gas ndustry that may ndrectly affect state economc performance and drectly mrror the potental performance of the ndustry tself are drllng rg counts, exploraton and development drllng, and producton. These ndcators are hghly ted to the prce of ol n the short-run as well as n the long run dependng on current and expected proft margns n the ndustry. However, due to data lmtatons, especally at the more dsaggregated levels of water depths, ol and gas producton s used as a proxy for ndustry actvty n our model formulaton and estmaton. The modelng approach developed and estmated n ths secton s premsed on our desres to answer the followng questons: a. What have been the general trends n ol prces, ndustry ndcators, and macroeconomc varables over tme? b. Are ol prce movements correlated wth dentfed macroeconomc and ndustry ndcators and to what degree? c. Is there strong causalty between ol prce movements and dentfed performance varables n the state? 3 d. How long does the effect of an ol prce change persst before equlbrum s restored n these relevant aggregates? e. What s the degree of responsveness (.e. elastcty), f any, of these economc ndcators to changes n ol prce? 2 Output was orgnally proposed as one of the ndcators but could not be used because of the length of the seres at the state level. GSP s only avalable from 977 and only on an annual bass. 3 Causalty s defned n the Granger-sense here, not n the commonly understood sense. See appendx for detals. 2

19 Table Varable Names, Descrptons, and Transformaton Method Varable Descrpton Perod* Length Seasonally Transformaton Deflated by Adjusted ALQPI AL Quarterly Personal Income 969:-2000:2 26 No Log Dfference GDPI LAQPI LA Quarterly Personal Income 969:-2000:2 26 No Log Dfference GDPI MSQPI MS Quarterly Personal Income 969:-2000:2 26 No Log Dfference GDPI TXQPI TX Quarterly Personal Income 969:-2000:2 26 No Log Dfference GDPI ALQUR AL Quarterly Unemployment Rates 976:-2000:4 00 Yes Non Dfferenced LAQUR LA Quarterly Unemployment Rates 976:-2000:4 00 Yes Non Dfferenced MSQUR MS Quarterly Unemployment Rates 976:-2000:4 00 Yes Non Dfferenced TXQUR TX Quarterly Unemployment Rates 976:-2000:4 00 Yes Non Dfferenced QCPPI Quarterly Crude ol PPI 947:-2000:4 26 No Log Level QAPPI CPPIV Quarterly Crude ol PPI Volatlty 947:-2000:4 26 No Non Dfferenced QAPPI Quarterly All Commodtes PPI 947:-2000:4 26 No Non Dfferenced RGDP Real GDP n 996 Dollars 947:-2000:4 26 Yes Log Dfference GDPI Implct GDP Deflator 947:-2000:4 26 Yes Log Dfference TRBR Three Month Treasury Bll Rate 947:-2000:4 26 No Non Dfferenced GOSHA Gulf: Ol & Gas Producton Shallow. Waters 948:-2000:4 22 No Log Dfference GODEP Gulf: Ol & Gas Producton Deep Waters 979:3-2000:4 86 No Log Dfference GOTOT Gulf: Ol & Gas Producton Total 948:-2000:4 22 No Log Dfference ALARV AL Annual Revenue No Log Dfference GDPI LAARV LA Annual Revenue No Log Dfference GDPI MSARV MS Annual Revenue No Log Dfference GDPI TXARV TX Annual Revenue No Log Dfference GDPI * Year:Quarter-Year:Quarter. Table 2 presents the basc correlaton coeffcents among macroeconomc aggregates and selected exogenous varables. In general, the crude petroleum prce ndex s shown to be negatvely correlated wth personal ncome, but postvely correlated wth unemployment ratesexcept n Lousana. The correlaton coeffcents between prce and unemployment rates are, however, relatvely small n value. Personal ncome s hghly and postvely correlated wth the overall OCS ol producton. The correlaton coeffcents between unemployment rates and OCS producton, n general, are smlar n magntude to those between producton and personal ncome, but the sgns of the correlaton coeffcents are reversed. State revenue shows a postve correlaton wth both prce and crude petroleum producton n the OCS. It should be noted that these results are only ndcatve of the potental relatonshps among the varables; correlaton s not causaton. Therefore, a more robust tool of analyss such as a VAR s often requred for an n-depth examnaton of such relatonshps among varables. Descrptve statstcs of all the varables dscussed n the estmaton process are shown n Tables 3a and 3b. Average personal ncome s hghest n Texas, followed by Lousana, Alabama and Msssspp, respectvely. However, the range n average personal ncome between the states s relatvely large. Over the perod, unemployment rates n these states are qute hgh, rangng from a mean value of 6.2 percent n Texas to 8.08 percent n Lousana. The Gulf OCS gross ol and 3

20 gas producton averaged about MMB annually. The average dstrbuton of annual revenue n the states also shows a smlar pattern to the dstrbuton of quarterly personal ncome. Texas s consderably ahead of the others n state revenue on both an absolute and per capta bass. The trends n unemployment rates, personal ncome and annual revenue, macroeconomc ndcators of the strength of the U.S. Gulf Coast economy, are dscussed brefly below and depcted n Fgures through 3. The trends n annual state revenue as depcted n Fgure also show smlar patterns to the trends n personal ncome of the four states (see Fgure 2). Lousana has had the lowest growth rate n revenue, especally snce the early 990s. Pror to the late 980s, revenue derved from the ol and gas sector accounted for more than one thrd of the state government aggregate revenue. Presently, however, the ol and gas sector of the economy accounts for less than 2.5 percent of government revenue (Iledare and Olatub, 2004). Fgure 2 shows the trends n quarterly per capta personal ncome n the four Gulf States over tme. It shows that the growth rate n Texas personal ncome s much hgher than the growth n the other three Gulf States. Personal ncome n Alabama and Msssspp has grown n tandem over ths perod and the growth s better than the growth n Lousana. Fgure 3 presents the trends n another mportant macroeconomc varable--unemployment rates n the Gulf States. Employment levels provde an mportant ndcaton of the level of economc actvty n a state. Unlke personal ncome and revenue trends dscussed earler, the trends n unemployment rates follow smlar patterns n all of the states. Generally, there were low unemployment rates untl the early 980s, when t ncreased dramatcally. It s nterestng to note that the net-petroleum mportng states Alabama and Msssspp experenced the hghest reported unemployment rates n the early 980s. Many people n the Gulf States are employed drectly or ndrectly n the ol and gas sector, so any unusual developments n the petroleum sector wll reflect on the state s welfare. The trend n quarterly crude petroleum producer prce ndex (QCPPI), a measure of composte ol prces, s presented n Fgure 4. In general, ol prce was stable untl the md-970s. From the md-970s, the crude ol prce ndex rose sharply to ts hstorcal hgh n the early 980s. Although the prce fell n the md to late 980s relatve to the prevous decade, t was relatvely more volatle n the 990s. In fact, the 990s wtnessed at least two spkes n ol prces. 4

21 Table 2 Correlaton Matrx of Model Varables ANNUAL SERIES: Prce Index Producton Real GDP Treasury Bll Revenue, AL Revenue, LA Revenue, MS Revenue, TX Producton QUARTERLY SERIES: 976:-999:4 Prce Index Producton Real GDP Treasury Bll Income, AL Income, LA Income, MS Income, TX Unemp., AL Unemp., LA Unemp., MS Unemp., TX Producton Table 3a Quarterly Summary Statstcs of Model Varables, 976:-999: Mean Medan Max Mn Std. Dev. Obs. ALQPI* 56,564 53,748 02,073 9,22 24, LAQPI 59,655 54,557 0,460 2,07 22, MSQPI 3,506 29,0 58,53,4 3, TXQPI 273, ,886 55,782 78,828 29,79 96 ALQUR** LAQUR MSQUR TXQUR Prce Index Real GDP 6,26 6,255 9,084 4,266,3 96 Treasury Bll OCS Total Prod * XQPI represents quarterly personal ncome measured n mllons of real dollars for state X. ** XQUR s unemployment rates n percent for state X. Producton s measured n Mllons of barrels of ol equvalent and the real GDP s n trllon dollars. 5

22 Table 3b Annual Summary Statstcs of Model Varables, Mean* Medan Max Mn Std. Dev. Obs. Revenue n AL 4, , , , Revenue n LA 4, ,26.5 4, , Revenue n MS 2,644.05, , , Revenue n TX 4, , , , Prce Index Real GDP 4, ,5.80 8, ,099.50, Treasury Bll OCS Producton , * Annual revenue s reported n mllon dollars ALARV LAARV MSARV TXARV Revenue Index (950=00) Fgure : Trends n Annual Revenue of the U.S. Gulf States

23 ALQPI LAQPI MSQPI TXQPI 970Q= Fgure 2: Trends n Quarterly Personal Income of the U.S. Gulf States Unemployment Rates, % ALQUR LAQUR MSQUR TXQUR Fgure 3: Trends n Unemployment Rates n the U.S. Coastal Gulf States. 7

24 Fgure 5 shows that ol and gas producton from the OCS has ncreased sgnfcantly snce the late 950s. There was a rapd growth n ol and gas producton from 959 to the late 970s. However, from the md-970s to the late 990s, the rate of growth n producton moderately declned. Snce the late 990s, there appears to be a sharper declne n producton rate than any other tme n hstory. In terms of water depth, most of the producton actvtes n the GOM have hstorcally occurred n the shallow waters. However, snce the early 990s, producton has declned n the shallow waters whle the producton n the deep waters has been rsng Index(982=00) Fgure 4: Trends n Crude Petroleum Prce Index,

25 400 Ol-Equvalent(MMB) GODeep GOShall Note: For ths report, a lease s consdered to be located n deepwater f the average water depth s at least 200 m. Fgure 5: Gulf of Mexco OCS Petroleum Producton by Water Depth Category. 9

26 2 3. VAR MODELING OF THE ECONOMIC EFFECTS OF PETROLEUM PRODUCTION AND PRICES 3.. VAR Model Specfcaton Recent developments n tme seres analyss, especally n the theory of co-ntegraton, errorcorrecton and Granger-causalty, have extended the opportuntes avalable to analyze, n-depth, economc equlbrum relatonshps. In ths study, as n most studes of macroeconomc mpact of ol prce change, a VAR modelng approach s adopted. VAR modelng s a mult-stage process nvolvng unt roots tests, co-ntegraton examnaton, and Granger-causalty exploraton 4. The VAR approach s commonly used for forecastng systems of nterrelated tme seres and for analyzng the dynamc mpact of random dsturbance on the system of varables. In ths formulaton, every endogenous varable s modeled as beng dependent on ts own lag(s) and the lags of other endogenous varables. Exogenous varables may also be ncluded n the specfcaton of the systems. The general mathematcal formulaton usually takes the form: t t p t p t t Bx y A y A y ε =... () where y t s a vector of k dependent varables, x t s a vector of m ndependent varables, A,., A p and B are matrces of coeffcents to be estmated. The term ε t s dsturbance term that may be correlated wth each other but may not be correlated wth ther mmedate past values (ε t- ) or other varables on the rght-hand-sde Emprcal VAR Model Representaton The VAR model, whch descrbes the nteractons among the Gulf Coast economc ndcators, ol and gas producton n the Gulf OCS, and changes n crude petroleum prce ndex s represented by the followng system of equatons (2): t t t t t t t t t t t t t t t t t t t t D p X X p y p y p y p y D p X X p y p y p p y y D p X X p p y y p y µ δ ϕ ω λ γ β α µ δ ϕ ω λ γ β α µ δ ϕ ω γ β α + + = + = + = + = + = + = + + = + = + = + = + = + = + + = + = + = + = + = (2) where: y t (=,2,3): = natural log of crude petroleum prce ndex; 2= natural log of ol and gas producton; and 3= quarterly unemployment rates or natural log of annual state revenue or natural log of quarterly personal ncome; 4 A bref outlne of a typcal VAR procedure s gven n Appendx A.

27 X t = the U.S. three-month treasury bll rate n levels (a proxy for nterest rates); X 2t = natural log of real U.S. gross domestc product; D = a determnstc dummy whch equals for the perod 979 to 986 and 0 otherwse; p = the number of past values (lags) of the dependent varables n the system equatons ncluded as ndependent varables. The dummy varable D s ncluded n each equaton of the system to capture the perod when ol prces declned and crashed. In addton, the proxy for economc ndcator, y 3t, does not appear n the prce equaton because the ncluded measures of the economy n the Gulf States are not expected to have a drect nfluence on the crude petroleum prce ndex because most economc actvtes n the Gulf States are prce takers n the overall global petroleum economy. The number of past values of the dependent varables (length of lags) n each system of equatons s determned statstcally usng a combnaton of Schwartz Bayesan Crtera (SBC) and Akake Informaton Crtera (Iledare and Olatub, 2004). Further, the general formulatons represented n the above system of equatons (2) are ndeed a standard format of VAR model representaton. In the prmtve forms, the current levels of the other varables are ncluded n the rght-hand-sde of the equaton defnng the evoluton of that varable. From a statstcal perspectve, the prmtve system of these equatons suffers an dentfcaton problem. In addton, not all of the parameters of the prmtve forms can be recovered from estmatng the standard form. To dentfy the prmtve system, restrctons have been mposed on some of the parameters. Such restrctons are based on economc theory or the ntuton of the researcher. A common type of restrcton s to order the varables (and hence, the error terms) accordng to the effects that are beleved to be a pror. For example, n ths study, we order the varables as follows: [ol prce OCS actvty economc ndcators]. Ths orderng mples that the shocks on economc varables flow from the shock to ol prce and OCS actvty n that order. By mplcaton, ol prce s not drectly affected by ether OCS actvty or economc varables. A dfferent orderng may produce a dfferent response path, hence, we chose carefully the approprate orderng based on economc theory or alternatve plausble results from dfferent orderngs VAR Model Estmaton and Analyss Generally, a VAR model such as the type we specfed n equaton () can be estmated usng ordnary least squares (OLS), f each equaton n the system contans the same number of varables and has smlar lags on the rght-hand-sde. OLS n ths case provdes estmates that are both consstent and asymptotcally effcent. The system formulaton n equaton (2) does not fully meet ths crtera; hence, the specfcaton n ths paper can be descrbed as near-var models. The near-var model n each of the cases formulated s estmated usng seemngly unrelated regresson (SUR) technques. A dynamc formulaton of the VAR-type has been found to perform better n macroeconomc forecastng than theoretcally based large structural models of the past. Hence, VAR has become a popular means of studyng the structural path of dynamc seres. Its usefulness for economc 22

28 analyss also les n the flexblty offered to test varous hypotheses of causaton (n the Granger sense) among the varables. In addton, the structure of the VAR can be exploted through what s generally referred to as nnovaton accountng. Two processes n nnovaton accountng mpulse response functon (IRF) and varance decompostons are adopted to study effects of shocks (.e. unexpected polcy changes) on the system represented n equaton (). To estmate the system of equatons whch nvolve personal ncome or unemployment rates, quarterly data for all model varables for Alabama, Lousana, Msssspp, and Texas were collected, processed and organzed nto a regresson format. However, all estmaton procedures nvolvng state revenue data utlzed annual data for the estmaton procedure because of a lack of quarterly data for state revenue. As mentoned earler, Lousana (LA) represents a petroleum producng and net petroleum exportng state wth a lmted dversfed economy. Msssspp (MS), on the other hand, represents a net petroleum mportng state wth a lmted dversfed economy and Texas (TX) s a relatvely more dversfed economy than Lousana. The economc base of Texas s also consderably larger than Lousana. Texas s also a net exporter of natural gas but a net ol mporter. Lastly, Alabama (AL) s a borderlne net petroleum mporter (hgh net ol mporter and low net gas exporter) and ts economy s less dependent on petroleum than Texas or Lousana. The long-run mpact of a polcy change affectng one of the varables n the system can be nvestgated usng the mpulse response functon and the proporton of these changes that are attrbutable to each varable n the system can be evaluated usng varance decomposton analyss. Accordngly, the central focus of VAR analyss s the fndng and understandng of the nterrelatonshp among varables over tme and not necessarly on the assessment of pont estmates. Thus, the VAR results are dscussed generally n terms of the varance decomposton and mpulse response functons generated from estmatng the VAR model represented by the system of equatons n (2). The emprcal results reported n ths report have been derved from estmatng the system of equatons n (2) ndvdually for employment, real personal ncome, and state revenue n combnaton wth OCS petroleum producton by plannng area and water depth one at a tme. 5 Varance decomposton and mpulse response functon analyses for each of the Gulf States have been appled to the VAR model results. The varance decomposton procedure provdes a way to decompose the mpact of a shock on the economc system nto ts component parts. The relatve proporton of the decompostons ndcates the relatve potency of the effect of a partcular shock n explanng the observed varatons n each varable experencng the shock. On the other hand, the mpulse response functon technque characterzes the dynamc effects of an unexpected shock n a gven economc system. It shows the dynamc paths of the effects of an ndependent shock of one varable on another varable and t s also useful for characterzng the 5 Ths mples estmatng several dfferent models/systems for each state: () prce, OCS producton, and employment, (2) prce, OCS producton, and personal ncome, (3) prce, OCS producton, and revenue, (4) prce, OCS deepwater producton, and employment, (5) prce, OCS deepwater producton, and personal ncome, and (6) prce, OCS deepwater producton and revenue. Interest rate, tme dummes, and GDP appear n each model/system as exogenous varables. 23

29 stablty and duraton of such effects. The persstence of such a shock reveals how fast the system wll return to ts orgnal equlbrum. The faster t takes a shock to dampen, the shorter the adjustment perod (Brown and Yucel, 995). 24

30 4. ESTIMATED VAR MODEL RESULTS: VARIANCE DECOMPOSITION ANALYSIS The emprcal results reported n ths report have been derved from estmatng the system of equatons n (2) ndvdually for employment, real personal ncome, and state revenue n combnaton wth OCS petroleum producton n the OCS and deepwater one at a tme. 6 Varance decomposton and mpulse response functon analyses for each of the Gulf States have been appled to the VAR model results. The varance decomposton procedure provdes a way to decompose the mpact of a shock on the economc system nto ts component parts. The relatve proporton of the decompostons ndcates the relatve potency of the effect of a standard devaton prce or producton shock n explanng the observed varatons n each varable experencng the shock. 4.. VAR Results from OCS Aggregate Producton System Equatons 4... OCS Petroleum Producton and the Lousana Economy: Accordng to the results reported n Table 4, the dynamc VAR analyss of the nteractons between changes n crude petroleum prces and ol and gas producton n the Gulf of Mexco OCS, and Lousana unemployment rates shows a sgnfcant prce effect on unemployment rates. Prce explans about percent of the observed varaton n unemployment over tme. Crude ol prce nteractng wth ol and gas producton n the Gulf of Mexco OCS also explans about percent of the expected varaton n personal ncome and between.45 to 6.8 percent of the varaton n revenue. The autonomous ol and gas producton shows no sgnfcant drect effects on unemployment accordng to the VAR results. Nonetheless, a relatvely sgnfcant varaton n personal ncome and state annual revenue s explaned by changes n autonomous producton. In an overall sense, both ol prces and Gulf ol producton have more mpact on revenue than they have on Lousana unemployment rates and personal ncome OCS Petroleum Producton and the Alabama Economy: The model results descrbng the nteractons among ol prces and ol and gas producton n the Gulf of Mexco OCS and Alabama unemployment rates ndcate that petroleum prce varaton explans up to 30 percent of the expected varaton n Alabama unemployment. The results also show that a prce shock condtonal on the OCS ol and gas producton profle explans up to percent of the observed varaton n personal ncome n Alabama. Further, a prce shock nteractng wth ol and gas producton also has a potental mpact of at most 29 percent n the long-term on Alabama revenue. The autonomous drect mpact of ol and gas producton n the Gulf OCS on Alabama unemployment s also not sgnfcant, accordng to the VAR model results. 6 Ths mples estmatng several dfferent models/systems for each state: () prce, OCS producton, and employment, (2) prce, OCS producton, and personal ncome, (3) prce, OCS producton, and revenue, (4) prce, OCS deepwater producton, and employment, (5) prce, OCS deepwater producton, and personal ncome, and (6) prce, OCS deepwater producton and revenue. Interest rate, tme dummes, and GDP appear n each model/system as exogenous varables. 25

31 Table 4 Decomposton of the Varance of Macroeconomc Varables Due to Changes n Petroleum Prces and OCS Gross Petroleum Producton A B C D States/Varables Perod LA Unemployment OCS Producton Prce Index LA Personal Income OCS Producton Prce Index LA Revenue OCS Producton Prce Index AL Unemployment OCS Producton Prce Index AL Personal Income OCS Producton Prce Index AL Revenue OCS Producton Prce Index MS Unemployment OCS Producton Prce Index MS Personal Income OCS Producton Prce Index MS Revenue OCS Producton Prce Index TX Unemployment OCS Producton Prce Index TX Personal Income OCS Producton Prce Index TX Revenue OCS Producton Prce Index

32 4..3. OCS Petroleum Producton and the Msssspp Economy: The model results, whch descrbe the nteractons between ol prce and ol and gas producton n the Gulf OCS and Msssspp economc varables demonstrate that the varaton n the state s unemployment accounted for by petroleum prces s less than 0 percent on average, but sgnfcant. Smlarly, the emprcal results ndcate that the effects of petroleum prces on personal ncome nteractng wth OCS producton may be about 5.5 percent. The prce mpact on revenue, accordng to the VAR model results, reaches as hgh as 6.7 percent. The mpact of a change n ol and gas producton n the Gulf, as s the case wth Lousana and Alabama, has no drect sgnfcant mpact on the state unemployment rate. However, the mpact of producton on revenue and personal ncome s statstcally sgnfcant as evdent n Table OCS Petroleum Producton and the Texas Economy: The estmated model results reported n Table 4 show that the mpact of a prce shock on Texas unemployment rates s relatvely small, although sgnfcant. The varatons n personal ncome and revenue n Texas explaned by prce shocks are 9 and 8 percent, respectvely. The effects of OCS producton on Texas unemployment rates, unlke n the other Gulf States, s sgnfcant, but small. Producton effect on Texas revenue ranges from 0.04 percent n the short-run to 2.4 percent n the long-run. Ths s a sgnfcant departure from the trends observed for Lousana, Alabama and Msssspp VAR Results from OCS Deepwater Producton System Equatons The emprcal results reported n Table 5 have been derved from estmatng the system of equatons n (2) for employment, real personal ncome, and state revenue n combnaton wth OCS deepwater petroleum producton and by usng the varance decomposton procedure for each of the Gulf States. The relatve mportance of changes n petroleum prces and producton n explanng volatlty n economc actvty n these states s dscussed brefly as follows OCS Deepwater and the Lousana Economy: The deepwater model results ndcate that varaton n prce and deepwater producton has lttle or no nfluence on the observed varaton on Lousana unemployment rates over tme. Ths s contrary to expectaton n comparson to the other Gulf States. On average, however, prce and deepwater producton explans about 6 and 2.6 percent of the observed varaton n Lousana personal ncome, respectvely. We dd not estmate the deepwater system of equatons for revenue because of data lmtatons. 27

33 Table 5 Decomposton of the Varance of Macroeconomc Varables Due to Changes n Petroleum Prces and OCS Deepwater Petroleum Producton A B C D States/Varables Perod LA Unemployment OCS Producton Prce Index LA Personal Income OCS Producton Prce Index LA Revenue OCS Producton Prce Index AL Unemployment OCS Producton Prce Index AL Personal Income OCS Producton Prce Index AL Revenue OCS Producton Prce Index MS Unemployment OCS Producton Prce Index MS Personal Income OCS Producton Prce Index MS Revenue OCS Producton Prce Index TX Unemployment OCS Producton Prce Index TX Personal Income OCS Producton Prce Index TX Revenue OCS Producton Prce Index

34 OCS Deepwater and the Alabama Economy: The model results descrbng the nteractons among ol prces and deepwater producton n the Gulf of Mexco OCS and Alabama unemployment rates ndcate that petroleum prce varaton explans up to 33 percent of the expected varaton n Alabama unemployment. The autonomous drect mpact of deepwater producton n the Gulf OCS on Alabama unemployment s sgnfcant, accordng to the VAR model results, explanng between 3-25 percent of the observed varaton n Alabama unemployment. The results also show that a prce shock condtonal on the OCS deepwater producton profle explans percent of the observed varaton n personal ncome n Alabama. The varaton n Alabama personal ncome explaned by changes n deepwater producton ranges from 0.80 to 7.00 percent OCS Deepwater and the Msssspp Economy: The VAR results descrbng the effect of a prce shock and ol and gas producton from OCS deepwater on Msssspp economc varables show that the changes n prce explan a relatvely small proporton of the observed varaton n unemployment (roughly 4 percent on average). A shock to deepwater producton also has a sgnfcant effect on unemployment. The results show that approxmately 8 percent of the observed varaton n unemployment s explaned as a result of producton shocks. Prce shocks also explaned up to percent of the observed varaton n personal ncome over the perod. Producton mpact, on the other hand, explaned less than 2.5 percent of the varaton n personal ncome over the perod OCS Deepwater and the Texas Economy: Accordng to the VAR model results, the mpact of changes n ol prces on Texas unemployment subject to varaton n OCS deepwater ol and gas producton s not statstcally sgnfcant. However, deepwater producton has a drect and sgnfcant mpact on Texas unemployment. The results further suggest that prce shocks explaned up to 6.3 percent of the observed personal ncome varaton, and deepwater producton explaned a lttle less than 6 percent of the observed varaton n Texas personal ncome. 29

35 5. ESTIMATED VAR MODEL RESULTS: IMPULSE RESPONSE FUNCTION APPROACH To further quantfy the responsveness of the economc performance ndcators to prce shocks and OCS producton n the Gulf States, the mpulse response functon technque for characterzng the dynamc effects of an unexpected shock n a gven economc system s appled separately to data from Alabama, Lousana, Msssspp, and Texas. Generally, the mpulse response functon (IRF) shows the dynamc paths of the effects of an ndependent shock of one varable on another varable and t s also useful for characterzng the stablty and duraton of such effects. 5.. IRF Results from OCS Aggregate Producton System Equatons 5... Prce Shock, Gulf OCS Producton, and the Lousana Economy: The mpulse response of Gulf ol producton and Lousana unemployment rate to a one-tme postve shock to crude ol prce s presented n Fgure 7. Unemployment rate falls and ol producton ncreases n response to the shock. Unemployment rate reaches ts hghest level wthn 0 quarters after the shock. Ths corresponds to about 0.6 percent above ts ntal equlbrum. The mnmum level of unemployment rate (0.26 percent n below equlbrum) was attaned wthn three quarters subsequent to the shock. Unemployment rate gradually moves towards equlbrum after reachng ts maxmum. Gulf aggregate producton, on the other hand, rses wthn fve quarters to a maxmum of 0.35 percent above the ntal equlbrum and falls to a mnmum of 0.26 percent below ts ntal level wthn three quarters. Ol producton fluctuates around ts equlbrum level over the tme horzon. It s also noted that both ol and gas producton and the unemployment rate return to ther orgnal equlbrum levels, although the dynamc paths to equlbrum are dfferent; ol producton fluctuates much more than unemployment rate. The dynamc response of Lousana personal ncome and Gulf OCS producton to prce s depcted n Fgure 6. A postve shock to prce ntally leads to a postve response from both ol producton and personal ncome. The affected varables return to the ntal equlbrum levels quckly. Fgure 8 shows that the dynamc paths of producton and revenue rose followng a prce shock. Revenue rose to a maxmum 0.35 percent of ts ntal level before the shock. However, all varables fluctuated wdely, albet towards equlbrum restoraton, and movements n producton and revenue were much more n tandem durng the perod Prce Shock, Gulf OCS Producton, and the Alabama Economy: The mpulse responses of aggregate OCS petroleum ol producton and Alabama unemployment rate, personal ncome and gross revenue to a one-tme postve shock to crude ol prce are presented n Fgures 9 through. Fgure 9 presents the response of Gulf ol producton and Alabama unemployment rate to a one-tme postve -standard devaton shock to crude ol prce. The mmedate effect s a decrease n unemployment rate and an ncrease n ol producton. The hghest level of unemployment reached s about 0.85 percent (n 6 quarters) above ts ntal equlbrum level whle the mnmum reached s 0.5 percent (n 6 quarters) below equlbrum. 3

36 Producton Income Fgure 6: Lousana Personal Income and OCS Producton Dynamc Paths Producton Unemployment Fgure 7: Lousana Unemployment and OCS Producton Dynamc Paths. 32

37 Producton Revenue Fgure 8: Dynamc Paths of Lousana Revenue and OCS Producton..00 Producton Unemployment Fgure 9: Responses of Gulf Producton & AL Unemployment Rate to Prce. 33

38 Producton Income Fgure 0: Responses of Gulf Producton & AL Personal Income to Prce Producton Revenue Fgure : Responses of Gulf Producton & AL Revenue to Prce. 34

39 Petroleum producton also rses to a maxmum of 0.27 percent (n quarter) and falls to a mnmum of 0.24 percent (n 3 quarters) below ts ntal level. Alabama unemployment and OCS aggregate ol producton returns to ntal equlbrum level at about the 24 th perod. It s noted that whle all varables return to ther orgnal equlbrum, the dynamc paths are not the same. Ol producton fluctuates much more than unemployment rate. The response of Alabama personal ncome to prce n the context of all Gulf ol producton s shown n Fgure 0. A postve shock to prce leads to postve response from both ol producton and personal ncome. In ths case the affected varables fluctuate around ther base-levels, although ths pattern s more pronounced for ol producton. The dynamc paths for producton and revenue are depcted n Fgure. Alabama s state revenue responds postvely to postve prce shock. The response s wthn 0.38 percent of ts ntal levels before a shock. However, the movements n revenue and ol producton are not n tandem Prce Shock, Gulf OCS Producton, and the Msssspp Economy: The response of Gulf ol producton and Msssspp unemployment rate, MSQUR, to a one-tme postve shock to crude ol prce s shown n Fgure 2. The mmedate effect s a decrease n unemployment rate and an ncrease n ol producton. Unemployment rate reaches ts hghest level of about 0.29 percent above ts ntal equlbrum n 6 quarters and ts mnmum level of 0.3 percent n the frst quarter. Aggregate OCS producton fluctuates around ts equlbrum throughout, rses to a maxmum of 0.25 percent n the frst quarter and falls to a mnmum of 0.24 percent below ts ntal level n 3 quarters. Producton returns farly quckly to ts ntal level after the prce shock whle unemployment takes longer to return to full equlbrum. Fgure 3 shows the tme path of Msssspp personal ncome to prce n assocaton wth aggregate petroleum producton n the Gulf OCS. A postve response to prce shock by producton and personal ncome s evdent n Fgure 3. Both varables fluctuate, although the pattern s more persstent and pronounced for ol producton. The dynamc paths of producton and revenue n response to a postve prce shock are depcted n Fgure 4. The Fgure shows that gross revenue n Msssspp responds negatvely to a postve prce shock. Revenue ntally falls to 0.35 percent of ts ntal levels before the shock before rsng towards ts ntal equlbrum level. But, as t s the case wth Alabama, all varables fluctuate wdely Prce Shock, Gulf OCS Producton, and the Texas Economy: Accordng to Fgure 5, the mmedate effect of a prce shock to the nteracton among Texas economy and OCS aggregate petroleum producton s a decrease n unemployment rate and an ncrease n ol producton. The overall effect s qute small for both varables. Although producton and unemployment return to ther orgnal equlbrum, the dynamc paths are not the same; unemployment rate path s slghtly dfferent longer and less cyclcal. Ol producton fluctuates n a much more cyclcal trend but around ts equlbrum level. 35

40 .00 Producton Unemployment Fgure 2: Responses of Gulf Producton & MS Unemployment Rate to Prce Producton Income Fgure 3: Responses of Gulf Producton & MS Personal Income to Prce. 36

41 Producton Revenue Fgure 4: Responses of Gulf Producton & MS Revenue to Prce..0 Producton Unemployment Fgure 5: Responses of Gulf Producton & TX Unemployment Rate to Prce. 37

42 Fgure 6 depcts the response of Texas personal ncome to a prce shock. A postve prce shock leads to postve response from both ol producton and personal ncome. Both varables fluctuate, but the pattern s more pronounced for ol producton than for ncome. The former s more cyclcal. Texas revenue ncreases ntally n response to a postve prce shock n the context of Gulf ol and gas producton. However, all varables quckly trend toward equlbrum although the path to equlbrum s faster for revenue than producton (see Fgure 7) IRF Results from OCS Deepwater Producton System Equatons Prce Shock, OCS Deepwater Producton, and the Lousana Economy: The mpulse response of OCS deepwater producton and Lousana unemployment rate to a one-tme postve shock to crude ol prce s presented n Fgure 8. Lousana unemployment and deepwater producton decrease followng a postve prce shock. The negatve producton response s contrary to our expectaton. However, ths response s small and probably transtory, reflectng a lagged responsveness. The response path for unemployment s also relatvely short. Further, the response paths for deep OCS petroleum producton and Lousana quarterly personal ncome to a postve shock to crude petroleum prces are depcted n Fgure 9. The fgure shows that the mpact of prce on personal ncome s postve and t reaches a maxmum of 0.33 percent wthn a year (3 quarters). The restoraton to ts orgnal equlbrum s also n less than 2 quarters Prce Shock, OCS Deepwater Producton, and the Alabama Economy: In Fgure 20 response of unemployment to a prce shock s negatve ntally and deepwater producton response s also unexpectedly negatve. The response path for unemployment s strkng. The unemployment response at ts maxmum s much larger (.24 percent) than the ntal shock to prce. The subsequent hke n unemployment rate follows the ensung sharp declne n prce after the ntal shock. The restoraton to ntal equlbrum takes much longer for unemployment and producton n comparson to other scenaros. The response to postve prce shock by Alabama personal ncome and OCS deepwater producton are presented n Fgure 2. The Fgure shows that personal ncome fluctuates wthn 0.3 percent of ts orgnal equlbrum, whereas the varaton n deepwater producton nearly doubled the fluctuatons n personal ncome followng a prce shock Prce Shock, OCS Deepwater Producton, and the Msssspp Economy: As evdent n Fgure 22, the response of deepwater producton or Msssspp unemployment rate to prce shock s negatve. The negatve responsveness s unexpected wth respect to deepwater producton. The response, however, shows mnmal effects over tme as devatons from equlbrum levels appear nsgnfcant. The response to postve prce shock by deepwater producton and Msssspp personal ncome s presented n Fgure 23. The fgure shows that changes n personal ncome are never above or below 0.2 percent and smlarly producton devatons are less than 0.25 percent. Hence, mpacts of ol prce shock on Msssspp economy are very small n magntude. 38

43 Producton Income Fgure 6: Responses of Gulf Producton & TX Personal Income to Prce Producton Revenue Fgure 7: Responses of Gulf Producton & TX Revenue to Prce. 39

44 Producton Unemployment Fgure 8: Responses of Deepwater Producton & LA Unemployment to Prce Producton Income Fgure 9: Responses of Deepwater Producton & LA Personal Income to Prce. 40

45 Producton Unemployment Fgure 20: Responses of Deepwater Producton & AL Unemployment Rate to Prce..00 Producton Income Fgure 2: Responses of Deepwater Producton & AL Personal Income to Prce. 4

46 .00 Producton Unemployment Fgure 22: Responses of Deepwater Producton & MS Unemployment Rate to Prce Producton Income Fgure 23: Responses of Deepwater Producton & MS Personal Income to Prce. 42

47 Prce Shock, OCS Deepwater Producton, and the Texas Economy: The mpact of a postve prce shock on Texas unemployment rate and deepwater producton s presented n Fgure 24. The fgure shows that unemployment and producton fall ntally n response to a prce shock. Unemployment rate rses to a maxmum of about 0.26 percent and producton declnes at a smlar magntude n the opposte drecton. The restoraton to equlbrum takes at least 24 quarters for unemployment rate n Texas. The dynamc paths for OCS deepwater producton nteractng wth Texas quarterly personal ncomes are depcted n Fgure 25. The Fgure shows that personal ncome rses to about 0.3 percent of ts ntal state and a postve devaton from deepwater producton equlbrum s at a slghtly smaller level..00 Producton Unemployment Fgure 24: Responses of Deepwater Producton & TX Unemployment Rate to Prce. 43

48 .00 Producton Income Fgure 25: Responses of Deepwater Producton & TX Personal Income to Prce. 44

49 6. ECONOMIC INTERPRETATIONS OF THE VAR MODEL RESULTS The mpulse response functon results and the correspondng graphcal representatons have been used n quantfyng the prce responsveness of state macroeconomc varables. The results are shown n Tables 6 and 7. Each elastcty reported n Table 7 s estmated by normalzng the varable response to ol prce shock at the correspondng maxmum. Hence, the mplct assumpton of a constant-elastcty has been nvoked (Brown and Yucel, 995). Wth regard to state macroeconomc varables unemployment, revenue, and personal ncome the dfferences among states do not appear to be large. The respectve means of each varable s ncluded to gve a sense of what each elastcty may mean n quantfable terms. In general, all three macroeconomc varables are more elastc to prce changes than ol and gas producton. The hghest ol prce elastcty of unemployment s n Alabama (2.575) whle Texas shows the least response (.97). These represent a change of 0.2 and 0.9 n unemployment rates for Alabama and Texas, respectvely. In the case of personal ncome response, each elastcty s smlar n magntude. In quanttatve terms, and unlke n the case of unemployment rates, Texas s more responsve whle Alabama s the least responsve. The responsveness of revenue to prce n all Gulf States s elastc, except for Msssspp (0.738). Two major nferences can be made from the emprcal analyss as reported n Tables 6 and 7. Frst, the effects of producton on macroeconomc varables n the Gulf Coast States are generally less than those of changes n prces. In other words, unemployment rates, personal ncome, and annual revenue are more drectly affected followng changes n petroleum prces than they are affected consequental to the changes n petroleum producton, whch ensue from changes n prces. Ths pattern ndcates that the drect effects of prce shocks on the economy n general should be of greater nterest to polcy makers n the Gulf Coast States than the drect effects of prces on ol and gas producton. The only excepton to ths pattern s the prce effects on the annual state revenue n Msssspp. The second nference s the lack of symmetry n the effects of a prce shock and changes n petroleum producton on economc performance ndcators across the Gulf States. The emprcal results show sgnfcant dfferences n the responsveness of economc performance ndcators to changes n prces across the states n the Gulf Coast. These dfferences are even more notceable when the effects are translated nto quantfable terms (quantty equvalence of unemployment, state revenue and personal ncome wth respect to the mean of the varable) than what the elastcty measures tend to portray (see Table 8). 45

50 Table 6 Estmated Range of the Impact of Changes n Prce and OCS Producton on Macroeconomc Varables Usng the Impulse Response Functon Technque (%) Prce Effect Producton Effect Varables/VAR system Hgh Low Hgh Low A B C D Lousana (LA) Unemployment Personal Income Revenue Alabama (AL) Unemployment Personal Income Revenue Msssspp (MS) Unemployment Personal Income Revenue Texas (TX) Unemployment Personal Income Revenue

51 Table 7 Estmated Range of the Impact of Changes n Prce and Deep OCS Producton on Macroeconomc Varables Usng the Impulse Response Functon Technque (%) Prce Effect Producton Effect Varables/VAR system Hgh Low Hgh Low A B C D Lousana (LA) Unemployment Personal Income Revenue Alabama (AL) Unemployment Personal Income Revenue\ Msssspp (MS) Unemployment Personal Income Revenue Texas (TX) Unemployment Personal Income Revenue 47

52 Table 8 Prce Elastcty of Macroeconomc Varables and the Quantty Equvalence Condtonal on the Dynamcs of OCS Petroleum Producton and the Gulf Coast Economy Quarterly Unemployment Quarterly Personal Income Annual Revenue Lousana Mean 8.09% $59,650 Mllon $4,963 Mllon Elastcty Quantty Equvalent* 0.20% $675 Mllon $ Mllon Alabama Mean 7.78% $56,560 Mllon $4,32 Mllon Elastcty Quantty Equvalent* 0.20% $603 Mllon $ Mllon Msssspp Mean 7.89% $3,50 Mllon $2,644 Mllon Elastcty Quantty Equvalent* 0.9% $369 Mllon $9.50 Mllon Texas Mean 6.22% $273,960 Mllon $4,273 Mllon Elastcty Quantty Equvalent* 0.2% $3,38 Mllon $6.99 Mllon * The correspondng average change n macroeconomc varables due to a percent change n prce. 48

53 Although the unemployment rates across the states tend to declne followng an ncrease n petroleum prces, the hghest ol prce elastcty of unemployment rates occurs n Alabama (2.575), whle Texas shows the least responsveness of unemployment rates to prce shocks (.97). These represent a quantty equvalence of and 0.9 percent change wth respect to the mean value of unemployment rates n Alabama and Texas, respectvely. Table 8 presents prce elastcty of macroeconomc varables and the correspondng quantty equvalence. The elastcty estmates are condtonal upon the nteractons among OCS petroleum producton, changes n petroleum prces, and the economy. Further analyss of the mpulse response functons also reveals dfferent adjustment paths to equlbrum for the Gulf States followng a prce shock (see Table 9). The emprcal results ndcate that t may take unemployment rates, personal ncome and government revenue more than ten years, about 3 years, and up to 20 years, respectvely to be restored to ntal equlbrum n Lousana. For the Alabama economy, the response paths show that t may take approxmately 6, 2, and 2 years, respectvely, to restore unemployment, personal ncome, and revenue to ther ntal equlbrum subsequent to any prce shock. The adjustment paths to a prce shock to the Msssspp economy ndcate that unemployment rates take more than 8 years, personal ncome takes about 2 years, and revenue takes 5 years to adjust to ther ntal equlbrum levels. The adjustment paths over tme for unemployment rate take less than 0 years, personal ncome takes more than 4 years, and revenue takes about 7 years for ntal equlbrum to be restored n response to a prce shock to the Texas economy. The fact that t takes longer for the employment levels n Texas and Lousana than Alabama and Msssspp to return to ntal equlbrum after a prce shock s most lkely due to the fact that ol and gas producton and ol and gas related busnesses are more prevalent n Texas and Lousana than Alabama and Msssspp. However, because Texas has a larger and more dversfed economc base than Lousana, t s more able to dampen the lkely destablzng effects of a prce shock on employment levels than Lousana. On the other hand, the economc sze of Texas seems to cause the effects of changes n crude petroleum prces on personal ncome to lnger longer than n Lousana, Alabama and Msssspp, n that order. The gross annual revenue n Lousana seems to be the most susceptble to an unexpected prce shock and Msssspp annual revenue s more reslent than Lousana, Alabama and Texas n ths regard. The declne n petroleum revenue n Lousana as a result of declnng ol prces has tended to push Lousana to the brnk of a budget defct n the more recent tme than Texas (Brown and Yucel, 995). The results for Alabama and Msssspp are also consstent wth the declnng relatve exposure to the petroleum ndustry vagares over tme (Scott, 2002). 49

54 Table 9 Estmated Adjustment Paths to Equlbrum Followng a Prce Shock Impact on Aggregate OCS Petroleum Producton and the Economy Indcators Alabama Lousana Msssspp Texas Unemployment (Quarters) Personal Income (Quarters) State Revenue (Years)

55 7. SUMMARY AND CONCLUSIONS Ths study examnes the nteractons between ol prce changes, ol and gas producton and selected macroeconomc varables of each economy of the Gulf States. Rather than focus on pont estmates from regresson analyses, we employed a VAR approach to understand both the composton of potental effects of a prce change and the adjustment paths of the economc varables and ol and gas producton over tme. By decomposng and examnng the mpulse responses of forecast errors, we are able to predct the relatve magntude and the dynamc adjustments of the selected varables to ol prce shock. Specfcally, the study shows that: Ol and gas producton n the Gulf as a whole responds postvely to a postve shock n crude ol prce. Ths s an expected result gven that frms operatng n the Gulf OCS desre to maxmze return on nvestment, hence, an ncrease n the prce of output s a sgnal from the market of a hgher demand for ol and gas products. Lkewse, a decrease n prce wll have the opposte effect. Unemployment rates across all the states tend to declne followng an ncrease n prce of crude ol. Ths result s consstent wth the fact that an ncrease n prce of ol and gas ndustry output wll spur the ndustry to expand output, and ceters parbus, more workers are needed to meet the new desred levels of output. Thus, employment levels n the states wll rse (means unemployment rates declne) to meet ndustry needs. It s noted that ths s a net effect, because an ncrease n prce of crude ol should also ncrease producton cost n ndustres where ol and gas are the prmary producton nput, e.g. chemcal and alled products. Ths fnal result may therefore mply that prce-elastcty of employment n the ol and gas producng ndustry s greater than prce-elastcty of ncome for ol and gas consumng ndustry n these states. Unemployment rates n the Gulf States appear to be relatvely less senstve to ol and gas producton actvtes. That s, Gulf producton role n states unemployment varaton s relatvely dmnutve over tme. In many nstances, the mpact of producton shocks on unemployment s nsgnfcant. Personal ncome ncreases followng a crude ol prce postve shock. Ths result s also consstent wth the fact that all the states consdered are ol producers although n varyng magntude. In general, the ol and gas ndustry pays relatvely hgher wages than most other ndustres n these states. It follows, therefore, that personal ncome n these states rses followng an ol prce ncrease. Of course, the degree of ths ncrease n ncome s asymmetrcal across states. In general, Lousana and Texas have hgher responsveness to prce change than Msssspp and Alabama. Revenue ncreases n each of the Gulf States followng an unexpected ncrease n the prce of crude ol and gas, except for Msssspp. However, as wth employment, ths must be regarded as a net-result. Ths s because other ol and gas usng ndustres may decrease ther producton capacty and output leadng to a declne n ncome tax base, 5

56 another sgnfcant source of state revenue. It s also noted that because the mpact of a prce ncrease s not necessarly unform across all states wth postve response, the revenue effects are not unform ether. In fact, as noted, Msssspp s revenue declne may be an ndcaton that the postve effect of producton n the Gulf s not enough to overrde the negatve effect of ncome tax base eroson. Further, the response lags n the mpulse response analyss gve an ndcaton of the length of the effects of a prce shock on a state economy. At one extreme, a persstent effect wll ndcate a long-lastng mpact that may n fact change the structure of the state economy. Thus our study suggests that: The mpact of a prce change takes about 45 quarters to return to equlbrum wth respect to employment n Lousana. Smlar measures for Alabama, Msssspp, and Texas are 25, 35, and 38 quarters, respectvely. In other words, such a prce change may have more destablzng effects on Lousana, Texas, Msssspp, and Alabama, n that order. Ths change may be destablzng because the change s a shock.e. unexpected, whch means frms and polcy makers may fnd t dffcult to respond adequately. Ths pattern among states may be explaned by the fact that ol and gas producton and ol and gas dependent ndustres are more prevalent n Texas and Lousana than n the other states. However, Texas has a larger and more dversfed economy than Lousana and s more able to dampen the destablzng effect of ts exposure to the prce shocks than Lousana. Personal ncome can take about 8 quarters n Texas, 2 quarters n Lousana and 8 quarters n Msssspp and Alabama to restore ntal equlbrum after a prce shock. In ths case, Texas economc sze seems to prolong the ncome effect of the change n prce more than any other state n the Gulf States. However, the net ol exportng states are stll far more exposed to such a shock than ther net ol mportng counterparts. In an overall sense, two major observatons are evdent n ths study. Contrary to our ntal hypothess, the effects of ol and gas prce shock on coastal Gulf States are more drect than ndrect (through ol and gas producton). In other words, employment, personal ncome, and revenue are mpacted more drectly followng a prce change rather than through changes n ol and gas producton followng a prce shock. Further, accordng to our emprcal results, there s a strong statstcal evdence to suggest an asymmetrc response of each of the three macroeconomc varables to prce n the four coastal Gulf States. 52

57 REFERENCES Brown, S.P.A. and J.K. Hll Lower ol prces and state employment. Contemporary Polcy Issues 6: Brown, S.P.A. and M.K. Yucel Energy prces and state economc performance. Economc Revew, Federal Reserve Bank of Dallas (Second Quarter). Pp Carruth, A.A., A.M. Hooker, and A.J. Oswald Unemployment equlbrums and nput prces: Theory and evdence from the Unted States. Revew of Economcs and Statstcs 80: Consdne, T.J Ol prce volatlty and U.S. macroeconomc performance. Contemporary Polcy Issues 6: Davs, S.J., P. Loungan, and R. Mahdhara Regonal labor fluctuatons: Ol shocks, mltary spendng, and other drvng forces. Workng paper, Unversty of Chcago, IL. Hamlton, J.D Ol and the macro economy snce World War II. Journal of Poltcal Economy 9(Aprl): Hckman, B.G Macroeconomc effects of energy shocks and polcy response: A structural comparson of fourteen models. Energy Forum Workng Paper 7.4, Stanford Unversty, n Hckman, B.G., H.G. Huntngton, and J.L. Sweeney, eds. 987, The Macroeconomc Impacts of Energy Shocks, Amsterdam: Elsever Scence Publshers. Iledare, O. O. and W.O. Olatub Impact of changes n crude ol prces and offshore ol producton on economc performance of U.S. coastal Gulf States. The Energy Journal 25(2): Keane, M.P. and E. Prasad Employment and wage effects of ol prce changes: A sectoral analyss. Revew of Economcs and Statstcs 78: Lee, K. and S. N On the dynamc effects of ol prce shocks: A study usng ndustry level data. Workng paper, Unversty of Mssour, Colomba. Prescott, E.C Theory ahead of busness cycle measurement. Quarterly Revew, Federal Reserve Bank of Mnneapols. (Fall):9-22. Scott, L The energy sector: Stll a gant economc engne for the Lousana economy. Lousana Md-Contnent Ol and Gas Assocaton, Baton Rouge, LA. Tatom, J.A Macroeconomc effects of the 986 ol prce declne. Contemporary Polcy Issues 6(July):

58 U.S. Dept. of Energy. Energy Informaton Admnstraton U.S. Crude Ol, Natural Gas, and Natural Gas Lquds Reserves Annual Report - Hstorcal. Internet webste: erves/reserves_hstorcal.html. U.S. Dept. of the Interor. Mnerals Management Servce MRM Statstcs Home. Internet webste: Yucel, M.K. and S. Guo Fuel taxes and co-ntegraton of energy prces. Contemporary Economc Polcy 2(July):

59 Step : Model Formulaton APPENDIX A AN OUTLINE OF THE VAR PROCEDURE A VAR analyss begns wth the selecton of a sutable model nformed by economc theory. Usually, each varable n the system s treated symmetrcally. Consder a two-varable case consstng of y and y 2, each affectng the tme-path of the other such that: y ( t) + v0 + v2 y2( t) + a y( t ) + a2 y2( t 2) + e( t) (A) y 2( t) + v20 + v2 y( t) + a2 y( t ) + a22 y2( t 2) + e2( t) (A2) In a general matrx form wth m varables and p lags, y t = v + A 0 y t + A y t- + A 2 y t-2 + A 3 y t A p y t-p + e t (A3) Where y t, v and e t are m x column vectors and A 0, A, A 2, A 3,..... A p are m x m matrces of coeffcents. The m-element vector e t are whte nose resduals that are d satsfyng E{e t e t`}= D, where D s a dagonal matrx. Note also that e (t) and e 2(t) are uncorrelated and are pure nnovatons (or shocks) n y (t) and y 2(t), respectvely. Equatons (A) and (A2) are referred to as prmtve or structural form of a VAR. Often ths prmtve form s ether over-dentfed or under-dentfed and the presence of the current levels of the other varable n ts own equaton mples correlaton of the regressed wth the error terms. Hence, consstent estmaton of these forms cannot be obtaned. To estmate each of these equatons by OLS, one must obtaned reduced forms. The system of equatons s solved smultaneously to extract the reduced or standard VAR form: (I A 0 ) yt = v + A 0 y t + A y t- + A 2 y t-2 + A 3 y t A p y t-p + e t (A4) Whch reduces to y t = (I A 0 ) - v + (I A 0 ) - A y t- + (I A 0 ) - A 2 y t-2 + (I A 0 ) - A 3 y t (I A 0 ) - A p y t-p + (I A 0 ) - e t. (A5) In general matrx form, equaton A5 becomes: y t = b + B y t- + B 2 y t-2 + B 3 y t B p y t-p + u t (A6) Where b = (I A 0 ) - v, B = (I A 0 ) - A, B 2 = (I A 0 ) - A 2 B 3 = (I A 0 ) - A etc., and u t = (I A 0 ) - e t. The varance-covarance matrx of resduals of the vector u t equals 55

60 [((I A 0 ) - ] D [(I A 0 ) - ]. Each of the descrbng equatons of A6 can be estmated by OLS. However, OLS can only be used f the system contans the same number of varables and lags n the rght-hand sdes. In ths study, as may be observed n equaton A4, the rght-hand varables n each equaton are not the same thus SUR s utlzed. Step 2: Unt Root Tests Havng formulated an approprate theoretcal model, the next step s to test for unt roots (or statonary) n all the varables. It has been shown that an OLS or SUR regresson of the long-run relatons mpled by each descrbng equaton of A6 s vald (non-spurous). Non-spurousness of long-run relatons means that the varables are co-ntegrated. To be co-ntegrated there must be unt roots n at least two or more of the varables. A common method to test for a unt root n a varable s by the Augmented Dckey Fuller (ADF) Test. Equaton (A7) s estmated to perform the ADF test: y t = µ + γy t- + δ y t- + δ 2 y t δ p y t-p + ε t (A7) Where y t = (y t y t- ), γ = ρ-, whle the null and alternatve hypotheses are Unt root: H 0 : γ = 0 No Unt Root: H : γ < 0 There s no consensus as to what should be done to the varable(s) subsequent to VAR estmaton f a unt root s confrmed. Some suggest that the varable be dfferenced to remove the unt root(s). Others argue otherwse. Those who argue for non-dfferencng beleve that snce the goal of a VAR analyss s not to determne parameter estmates, but uncover dynamc nterrelatonshps among varables, dfferencng throws away valuable nformaton. However, the majorty vew s for dfferencng because a VAR should mmc the true data generatng process. In ths study, we adopt the majorty vew. Step 3: Exogenety and Excluson Tests Although n theory we have formulated A6 such that every endogenous varable s present n each equaton and the lag length s also equal across equatons, n realty, t may be that a varable or some of ts lags does not really add to the forecastng performance of another varable and may therefore be excluded from the determnaton of that varable. The procedure to determne f a varable s a causal factor n predctng another s often the Granger causalty and excluson tests. If y does not mprove the forecastng performance of y 2, then y does not Granger-cause y 2 and therefore nothng s ganed by ncludng t n the equaton determnng y 2. The common F-test can be used to evaluate Granger-causalty for a sngle equaton. A test for exogenety s techncally dfferent and more restrctve than Granger-causalty, however. A necessary condton for the exogenety of y s that the current and past values of y 2 does not affect y. A multvarate approach to carryng out the exogenety and excluson test s to use the so-called block causalty test. 56

61 To perform the test, run the system of equatons wth all the lags and varables (unrestrcted form, U), and obtan the varance-covarance matrx, Σ u. Then regress the system agan excludng all the lags of the varable from the equatons where t s theorzed to be exogenous, and obtan the restrcted Σ r. The results are evaluated usng the lkelhood-rato test (T-c)(log Σ r - log Σ u ), whch s dstrbuted as a ch-square wth the degrees equal to the number of restrctons. T s the number of observatons and c s the number of parameters estmated n the unrestrcted form. Ths logc may be extended to the queston of the ncluson of dummy varables as well. Step 4: Lag-Length Selecton The selecton of the approprate lag-length n the system of equatons s an mportant consderaton. As n the selecton of the approprate varable(s) n the rght-hand sdes, the lkelhood-rato test s often used to select the approprate lag length. The goal here s to ensure a parsmonous system wth errors that are whte nose as the theoretcal model presumed. Ths test may also use the Akake Informaton Crtera (AIC) and or the Schwartz-Bayesan Crtera (SBC). In the case of the AIC and SBC, we look for the model wth the lowest value of the AIC or SBC estmates. Step 5: Estmaton Wth steps 4 and 5 completed, the system of equatons may stll be symmetrc. In ths case, OLS s stll the approprate choce estmator applcable to each of the equatons. However, t s possble that the resultng system after the prevous two-steps produces a non-symmetrc system such that ether the rght-hand varables are not the same across equatons, or the lag-lengths dffer across equatons. In the non-symmetrc stuaton, OLS s no longer an approprate estmator as ponted out prevously, we have to use another estmator such as a SUR. Step 6: Innovaton Accountng Because of the restrctons mpled n the reduced system n A6, not all of the parameters of the prmtve forms can be recovered wthout even further restrctons. In addton, further restrcton may be necessary to obtan consstent estmates of A6. Thus the man focus of a VAR s not on parameter estmates, rather t s to understand the tme-path and dynamc nterrelatonshps among ncluded (endogenous) varables. One approach to obtan useful nformaton from a VAR s to focus on the error terms n A6 snce by desgn these are contemporaneously related across equatons. In essence, we want to see what happens to a varable and to the other varables to whch t s related f there s an nnovaton (or shock) to t. One method to accomplsh ths s to use a movng average representaton of the system. For example, the system gven by A6 s transformed such that: y t = C 0 u t + C u t- + C 2 u t-2 + C 3 u t C s u t-s + y 0 (A8) Where y 0 equals ntal value of yt. 57

62 Equaton A8 does not gve a proper ndcaton of how the system responds to shocks to the ndvdual structural equatons. Ths s because the shocks to the equatons contaned n the vector u t are correlated wth each other. It s therefore not possble to determne the effects on the m varables of a shock to an ndvdual structural equaton would be as the observed u t represents the combned shocks to a number of equatons. It s noted that u t = (I A 0 ) - e t.. To obtan unencumbered ndvdual shocks n the structural system, t s necessary to solve the system for A 0 and thus obtan (I A 0 ) -, whch wll enable us to transform the u t-j s n nto e t-j s. The transformaton s done by selectng an approprate matrx to orthogonalzed the errors so that A 0 s dentfed. Then y t = Z 0 e t + Z e t- + Z 2 e t-2 + Z 3 e t Z s e t-s + y 0 (A9) Where Z j = CjG ; e t-j = G - u t-j and G = (I A 0 ) -. The standard approach to dentfy the elements of A 0 and hence decompose the matrx of reduced form resdual n a VAR analyss s by the so-called Cholesk Decomposton: Where D = I. u t u`t = Ω =Ge t. (Ge t. )` = Ge t. e`t. G` = GDG` The Cholesk Decomposton of the matrx Ω s obtaned such that ~ ~ ) 0 ( I A = G ~ Whch mples A 0 = I G and A0 s a representaton of A 0 after scalng of the varables n order to obtan D = I. Wth ths G matrx the matrces Z j n equaton A9 wth the errors, e t, of unt varance. The Z j matrces are called mpulse-response functons. In ths partcular method of decomposton, a partcular orderng of the varable s mposed on Ω. A dfferent form of orderng wll produce a dfferent mpulse response. Hence, the analyst must choose a plausble orderng guded by economc theory. (In ths study we use the orderng: ol prce, ol producton, and state economc varable. Ths orderng mples that ol prce s not affected by the other varables and the flow of causal relaton s from prce to producton and then state economc varable). A plausble way to determne the mportance of dfferent exogenous shocks n explanng the dependent varables s by calculatng the fractons of the forecast error varance of these varables attrbutable to such shocks. That s, the fractons of these forecast errors that are due to ndvdual shocks can be obtaned from equaton A9. In the two-varable case consdered here, the varance decomposton may be estmated as descrbed below. 58

63 Let 0 z j be the j-th element of Z 0, we can express the current-perod forecast error thus: y y t t = z = z e t e t + z + z 0 2 e t e 2t Then, Var{ y Var{ y t t } = ( z } = ( z ) ) ( z + ( z ) ) 2 2 For e and e 2 are ndependent shocks wth unt varance. The standard devatons of these estmates are ther respectve square roots and the fracton of the error varance attrbutable to the shock to the frst and second equatons are ( z 0 ( z ) ) + ( z 0 2 ) ( z2 ) and ( z ) + ( z ) 2 59

64 The Department of the Interor Msson As the Naton's prncpal conservaton agency, the Department of the Interor has responsblty for most of our natonally owned publc lands and natural resources. Ths ncludes fosterng sound use of our land and water resources; protectng our fsh, wldlfe, and bologcal dversty; preservng the envronmental and cultural values of our natonal parks and hstorcal places; and provdng for the enjoyment of lfe through outdoor recreaton. The Department assesses our energy and mneral resources and works to ensure that ther development s n the best nterests of all our people by encouragng stewardshp and ctzen partcpaton n ther care. The Department also has a major responsblty for Amercan Indan reservaton communtes and for people who lve n sland terrtores under U.S. admnstraton. The Mnerals Management Servce Msson As a bureau of the Department of the Interor, the Mnerals Management Servce's (MMS) prmary responsbltes are to manage the mneral resources located on the Naton's Outer Contnental Shelf (OCS), collect revenue from the Federal OCS and onshore Federal and Indan lands, and dstrbute those revenues. Moreover, n workng to meet ts responsbltes, the Offshore Mnerals Management Program admnsters the OCS compettve leasng program and oversees the safe and envronmentally sound exploraton and producton of our Naton's offshore natural gas, ol and other mneral resources. The MMS Mnerals Revenue Management meets ts responsbltes by ensurng the effcent, tmely and accurate collecton and dsbursement of revenue from mneral leasng and producton due to Indan trbes and allottees, States and the U.S. Treasury. The MMS strves to fulfll ts responsbltes through the general gudng prncples of: () beng responsve to the publc's concerns and nterests by mantanng a dalogue wth all potentally affected partes and (2) carryng out ts programs wth an emphass on workng to enhance the qualty of lfe for all Amercans by lendng MMS assstance and expertse to economc development and envronmental protecton.