Comparng Changes n Operatonal Productvt and Fnancal Productvt of Electrc Utltes n a Compettve Envronment b Glenn S. Barnette gbarnett@gustavus.edu Gustavus Adolphus College St. Peter, MN 56082 Presented at the MBAA Internatonal Conference Chcago, IL. March 29, 2012 Abstract Ths paper defnes operatonal productvt as a phscal measure of output n the numerator and a phscal measure of nput n the denomnator. Fnancal productvt would have ether the nput or the output expressed n monetar terms. Ths paper measures change n Total Factor Productvt of nvestor owned utltes assocated wth changes n the non-utlt generators market share and ndustr reorganzaton. It measures these effects usng both operatonal productvt and fnancal productvt as the dependent varable. It fnds that the two defntons of productvt are equvalent n evaluatng the effects of competton and restructurng n electrct generaton. However, snce the regressons usng fnancal productvt produce a better ft, the author recommends usng fnancal productvt to evaluate the effects of regulaton. INTRODUCTION Wth the passage of the Energ Polc Act (EPAct) of 1992, the electrc utlt ndustr has moved toward a more compettve wholesale market. Ths act establshes a new class of exempt wholesale generators (EWGs), whch operate ndependentl of state regulaton (Barnette, 1998). To enforce ths act, the Federal Energ Regulator Commsson (FERC) ssued Orders 888 and 889 (FERC 888, 1996 and FERC 889, 1996) n Aprl 1996. Order 888 requres all publc utltes to provde open access to ther transmsson facltes, and Order 889 establshes a code of conduct that requres the utltes to separate ther wholesale merchant and transmsson functons. Whle these rules do not mandate ndustr restructurng, EIA (2000, p. x) reports that, due to state regulaton or busness strateg, Investor Owned Utltes (IOUs) have collectvel ether sold to Non-utlt Generators (NUGs) or transferred to unregulated afflates twent-two percent of ther generatng capact durng the perod 1997 through 2000. As a result NUG net generaton ncreased from nne percent of the market n 1992 to eleven percent n 1998. Whle these developments certanl ntroduced competton n the generaton of electrct, the degree of competton vared among the states. Regulators hope that ncreased competton wll mprove effcenc and lead to lower rates. If these goals are to be acheved, one would expect a postve correlaton between productvt mprovements and changes n the compettve envronment, ceters parbus. Scull (1998) dd examne restructurng n New Zealand and found sgnfcant mprovement n productvt subsequent to the restructurng, but most of the mprovement was the result of changes n labor laws and mergers of adacent dstrbuton companes. Barnette et. al. (2004) dd fnd modest 1
mprovement n total factor productvt n the ear that NUG competton ncreases, but the mprovement dd not contnue n the followng ear. The Federal Energ Regulator Commsson (FERC) mantans a data base of fnancal reports prepared b nvestor-owned utltes (IOUs) snce 1994, but begnnng n 2002 FERC no longer requres frms to report the number of emploees. Consequentl, an tradtonal analss of the effects of deregulaton on productvt s lmted to the perod 1994 through 2001. Snce EPAct was stll beng mplemented durng much of ths perod, such analss wll be lmted to shortterm effects on productvt durng the transton perod. (Barnette et. al. (2004) used these reports.) Blocher, Stout, & Cokns (2010) defnes two measures of productvt operatonal productvt and fnancal productvt. Operatonal productvt s defned as output per phscal unt of nput, and fnancal productvt s output per monetar unt of nput. When measurng Total Factor Productvt (TFP), economsts tpcall use phscal unts to measure labor (emploees or labor hours) and energ. As stated above, ths measure makes t more dffcult to measure productvt changes as a result of changes n regulaton snce 2001. The purpose of ths paper s to use both operatonal productvt and fnancal productvt to measure the effect of regulaton and wholesale competton on the effcenc of Investor Owned Utltes (IOUs). Ths paper frst measures the TFP (operatonal productvt) of nnet-sx electrc utltes durng the perod 1994 to 2001. It then uses two measures of fnancal productvt wth the same frms durng 1994 to 2001. B comparng regressons wth operatonal productvt as the dependent varable wth those usng fnancal productvt as the dependent varable, one should determne whether fnancal productvt s a good prox for operatonal productvt when phscal unts are unavalable. If these two measures provde comparable polc conclusons, then one could use the FERC-1 Reports to measure fnancal productvt for ears subsequent to 2001 and use the results to evaluate polc changes snce 2001. TOTAL FACTOR PRODUCTIVITY Norsworth and Jang (1992) defne the level of total factor productvt as w TFP = [1] v x where s the phscal quantt of output, x s the phscal quantt of nput, w s the share of output n total revenue, and v s the share of nput n the total cost, where and q w = [2] q p x v = [3] p x so that q s the prce for output and p s the unt cost of nput. 2
The data for calculatng TFP s taken from the Federal Energ Regulator Commsson (FERC) Form No. 1, Annual Report of Maor Electrc Utltes (FERC-1 Report). Ths report ncludes fnancal statements and detaled schedules of assets, revenues, and expenses. The followng dscusson gves the FERC-1 schedules from whch we derve the output and nput measures. The page numbers for these schedules are gven n parentheses. Ths stud calculates TFP wth three dfferent sets of nputs. The three dfferent nput sets are Operatonal Productvt (phscal unts), Fnancal Productvt (current dollars) and a modfed Fnancal Productvt (1994 dollars). All TFP calculatons use mwh of electrct sold (both for resale and to fnal user) as the sngle output for a utlt. Electrc Energ Account (400) gves both the unts and total dollars of sales. Snce ths paper recognzes onl one output, [1] reduces to TFP = v [1a] x Ths paper recognzes sx nputs: (1) owned captal, (2) leased captal, (3) electrct purchased from others, (4) fuel used n generaton, (5) labor, and (6) purchased goods and servces. We obtan the phscal measures and cost of these nputs from the followng FERC-1 schedules. (1) Owned Captal. For calculatng Operatonal Productvt and 1994 Dollar Fnancal Productvt, ths paper estmates the value of Propert, Plant, and Equpment (PPE) at 1994 prces as a prox for captal nput. Ths allows the captal measure to accommodate dfferences n vntage, technolog, and tpes of equpment. I start wth the begnnng balance of plant and equpment n1994 Electrc Plant n Servce (204-207). To calculate the value for subsequent ears at 1994 prces, ths model uses Addtons, Retrements, Adustments, and Transfers n ths schedule and Deprecaton Expense and Adustments n the schedule Accumulated Provson for Deprecaton of Electrc Utlt Plant (219). To calculate change n captal stock, I add Addtons (adusted to 1994 prces) and subtract deprecaton expense, retrements (net of deprecaton adustments), and transfers at current prces. For calculatng Fnancal Productvt, ths paper uses the end of ear balance of fxed assets n Electrc Plant n Servce (204-207). For the captal nput weght, ths paper follows Jorgenson and Yun (1991) and defnes the rental cost of captal as propert tax, plus deprecaton expense, plus the opportunt cost of captal. 1 Propert tax s gven n the schedule of Taxes Accrued, Prepad, and Charged Durng the Year (262-263). I use Accumulated Provson for Deprecaton of Electrc Utlt Plant and Equpment (219) for deprecaton expense. Ths paper uses fnancal accountng prncples to estmate Weghted Average Cost of Captal (WACC). From the Statement of Income (114-117a), we obtan nterest expense and net ncome before ncome taxes for the utlt. From the Balance Sheet (110-113), I add current notes paable to total long-term debt to obtan total nterest bearng debt. The nterest cost of captal s Interest Expense Rate = (Begn Debt + End Debt)/2 and the equt cost of captal s 1 Jorgenson and Yun (1991) also adust for nvestment tax credts and tax benefts from deprecaton n the US ncome tax laws. Ths paper gnores nvestment tax credts, because the dd not appl to the perods of our stud. Determnng the tax benefts from deprecaton would requre more data than provded n the FERC- 1 reports. 3
EBIT ROE = (Begn Equt + End Equt)/2 where ROE s Return on Equt and EBIT s earnngs before ncome taxes. The WACC s the weghted average of the Interest Rate and ROE (weghted b average debt and average equt). 2 The opportunt cost of captal s the WACC tmes the Net Utlt Plant on the Summar of Utlt Plant and Accumulated Provsons for Deprecaton (200-201). The user cost of captal s the sum of opportunt cost of captal, deprecaton expense, and propert taxes. (2) Leased Captal appears as an expense on the Electrc Operaton and Mantenance Expense (320-323) schedule. For both Operatonal Productvt and 1994 Dollar Fnancal Productvt ths paper converts the expense to 1994 dollars as a prox for phscal unts (operatonal productvt). It uses current dollars to assgn the TFP weght (p x n Equaton 3). The current dollar Fnancal Productvt calculaton does not convert to 1994 dollars. (3) Purchased Power appears n mwh n the Electrc Energ Account (401a) schedule. The cost appears on the Electrc Operaton and Mantenance Expense (320-323) schedule. Operatonal Productvt calculatons use mwh as the nput. 1994 Dollar Fnancal Productvt converts the cost of power purchased to 1994 dollars. Current dollar Fnancal Productvt does not convert to a base prce. The current dollar cost of power purchased s used for assgnng a weght to ths nput. (4) Fuel Used. Ideall, one should use a common phscal measure such as BTU to calculate Operatonal Productvt, but FERC-1 does not consstentl gve ths nformaton n useable form. Therefore, ths paper uses the dollar cost on Electrc Operaton and Mantenance Expense (320-323). It converts to 1994 dollars as the nput for both Operatonal Productvt and 1994 Fnancal Productvt. It uses current dollar prces for calculatng the TFP weght. (5) Labor. For labor productvt, ths paper defnes one full tme emploee equvalent (FTE) used n operaton and mantenance (O&M) as a unt of labor for calculatng Operatonal Productvt. FERC-1 (p. 323) lsts the number of full tme and part tme of electrc department emploees. To calculate FTEs, we assume that a part tme emploee s the equvalent of one half of a full tme emploee. Total paroll cost s the sum of operaton and mantenance paroll (Dstrbuton of Salares and Wages, pp. 354-355), emploee benefts cost (Electrc Operaton and Mantenance Expense pp. 320-323), and paroll taxes (Taxes Accrued, Prepad, and Charged Durng the Year, pp. 262-263). For calculatng nput usng 1994 Dollar Fnancal Productvt, ths paper adusts the total paroll cost to 1994 wages. (6) Purchased Goods and Servces. To calculate the cost of purchased goods and servces, I take total Operaton and Mantenance Expense (Electrc Operaton and Mantenance Expense pp. 320-323) mnus the sum of the cost of rental captal, purchased power, fuel, and labor calculated above. For both Operatonal Productvt and 1994 Dollar Fnancal Productvt, ths paper converts ths cost to 1994 dollars. 2 Usng accountng statements to calculate WACC dffers from Jorgenson and Yun (1991). Ths model calculates nterest rates based on the nterest rates on outstandng debt, and t uses ROE as a prox for the equt cost of captal. Jorgenson and Yun (1991) use market rates to determne debt and equt costs of captal. 4
The above dscusson makes frequent references to 1994 dollars. To convert to 1994 dollars, ths paper uses three prce ndces mantaned b the U. S. Bureau of Labor Statstcs. To convert captal purchases and captal rentals, I use the Turbne and Power Equpment prce ndex. For fuel cost and purchased power, I use the Fuels and Related Products and Power prce ndex. For labor costs, I use the Electrc Utlt Unt Labor Cost ndex. For other costs, we use the Electrc Power Dstrbuton prce ndex. All of these ndces are unts of the Producer Prce Index. In the regresson model developed n the next secton, ths paper s nterested n measures of absolute TFP and changes n TFP. To calculate changes n TFP, we follow equaton [1] and defne the change n TFP durng ear t as ΔTFP = v, t ln ln [4] x, t x, t 1 where v,t s the nput weght assgned usng costs n ear t, and the productvt of nput n ear t. The term factor productvt of nput n ear t. ln ln x, t x, t 1 REGRESSION MODELS ln s the natural logarthm of x t, s the growth rate of the sngle Barnette et. al. (2004) develops a model to show the relatonshp between NUG competton and utlt restructurng to Total Factor Productvt. It defnes the relatonshp as ΔTFP = a + b t + b ΔNUG + b ΔNUG 1 + b ΔTFP 5 2 t 1 6 3 + b Dvest 1 t 1 7 + b TFP 4 + b Dvest where a s the ntercept, the b s are coeffcents, and ε s the error. t s a number for the ear, where 1996 s ear 1 and 2001 s ear 6. NUG s the dfference n the logarthms of the nonutlt market shares from the prevous ear to the current ear. TFP t-2 s the absolute productvt for two ears pror to the current ear. TFP t-1 s the change n total Total Factor Productvt n the prevous ear. Dvest 1 and Dvest 2 are two dumm varables to ndcate whether a frm has been forced b regulators to sell a sgnfcant porton of ts generaton assets. Durng the frst ear of dvestture, Dvest 1 assumes a value of 1, and durng subsequent ears Dvest 2 assumes a value of 1. (Dvest 1 reverts to 0.) These dumm varables reflect the fact that durng dvestture, the utlt s n effect substtutng purchased power for generaton captal, fuel, and labor. Whle nput substtuton for busness reasons ma be common, forced dvestture ma not necessarl have the same effect on productvt. I obtan NUG market shares from the Energ Informaton Admnstraton (EIA) web ste. To obtan values for the dvestture dummes, I rel on the notes to the fnancal statements ncluded n the FERC-1 reports. Ths paper repeats the regresson sx tmes, twce for each measure of productvt: Operatonal Productvt, 1994 Dollar Fnancal Productvt, and Current Dollar Fnancal Productvt. For 2 t 2 + ε [5] 5
the frst regresson for each productvt measure, I defne NUG as the percentage of total capact n a state that s owned b a non-utlt generator. For the second seres of regressons, I defne NUG as the percentage of total generaton provded b non-utlt generators. REGRESSION RESULTS Table 1 (below) shows the regresson results for the three measures of productvt where the NUG varable s defned as the non-utlt share of total generaton capact n a state. For all regressons the Tme coeffcent s not sgnfcant at the -level. For the two measures of fnancal productvt, the coeffcent for LN NUG (t-1) s sgnfcant at the -level; all other coeffcents are sgnfcant at the 95%-level of confdence. All coeffcents, other than tme, have the same sgn n each regresson. In other words the sgn for LN NUG s the same across all three regressons. The same s true for other coeffcents. All coeffcents except tme are sgnfcant at the 95%-level when operatonal productvt s the ndependent varable. Table 1: REGRESSION RESULTS WITH NUG CAPACITY Operatonal Productvt 1994 Dollar Fnancal Productvt Current Dollar Fnancal Productvt Independent Varable Coeffcent t-statstc Coeffcent t-statstc Coeffcent t-statstc Intercept - 0.05201-1.40205-0.059938-1.49969-0.08203* - 1.98090 Tme - 0.00018-0.02842 0.005043 0.87783 0.00775 1.33404 LN NUG - 0.04287* - 2.67275-0.029813* - 2.06097-0.03172* - 2.19849 LN NUG (t-1) 0.04079* 2.88215 0.023931** 1.86600 0.02366** 1.84911 TFP (t-1) 0.00024* 4.83800 10.176529* 14.89663 10.81020* 14.05411 TFP (t-2) - 0.00015* - 2.60266-9.665016* - 12.71375-10.10121* - 12.32041 Dvest 1 0.25692* 4.50054 0.244347* 4.71243 0.24095* 4.66572 Dvest 2 0.49709* 8.31272 0.494491* 9.05940 0.50087* 9.21644 R-Square 0.19448 0.39989 0.37685 Ad. R-Square 0.18456 0.39250 0.36917 *Sgnfcant at the 95% level, **Sgnfcant at the level The purpose of ths paper s to determne the extent to whch operatonal productvt s equvalent to fnancal productvt when analzng the effect of polc or exogenous varables on Total Factor Productvt. Table 2 (below) shows the confdence ntervals for the four exogenous varables under each defnton of the ndependent varable. The coeffcents for these measures for each defnton of productvt (Table 1) are wthn the confdence ntervals for the other two measures. When ths observaton s combned wth the fact that all coeffcents have the same sgn when compared to the other productvt defnton, one can conclude that for the purpose of evaluatng polc alternatves all three methods of measurng productvt eld equvalent results. 6
Table 2: CONFIDENCE INTERVALS FOR EXOGENOUS VARIABLES WITH NUG CAPACITY Operatonal Productvt 1994 Dollar Fnancal Productvt Current Dollar Fnancal Productvt Independent Varable Lower Upper Lower Upper Lower Upper LN NUG - 0.06929-0.01644-0.053646-0.00598-0.05550-0.00795 LN NUG (t-1) 0.01747 0.06410 0.002802 0.04506 0.00258 0.04474 Dvest 1 0.16287 0.35098 0.158920 0.32978 0.15587 0.32603 Dvest 2 0.39857 0.59561 0.404563 0.58442 0.41133 0.59040 Tables 3 and 4 (below) show the coeffcents, t-statstcs, and confdence ntervals for the regressons when the NUG varable s defned as the change n the non-utlt generators share of output n the state n whch the utlt s operatng. The results are smlar to those obtaned when the NUG varable s based on capact, but for both regressons of fnancal productvt the coeffcents related to changes n the NUG share of the market are not sgnfcant at the level of confdence. The coeffcents for the exogenous varables all ft wthn the confdence ntervals of the other regressons. Table 3: REGRESSION RESULTS WITH NUG GENERATION Operatonal Productvt 1994 Dollar Fnancal Productvt Current Dollar Fnancal Productvt Independent Varable Coeffcent t-statstc Coeffcent t-statstc Coeffcent t-statstc Intercept - 0.05304-1.42706-0.058764-1.46361-0.08093** - 1.94518 Tme - 0.00076-0.11783 0.005131 0.88784 0.00776 1.32678 LN NUG - 0.02928* - 2.11573-0.019155-1.53645-0.02038-1.63754 LN NUG (t-1) 0.03345* 2.47863 0.014889 1.21512 0.01469 1.20164 TFP (t-1) 0.00024* 4.87398 10.206336* 14.87194 10.83085* 14.01139 TFP (t-2) - 0.00015* - 2.62455-9.727108* - 12.75808-10.15201* - 12.33529 Dvest 1 0.24983* 4.38058 0.238329* 4.60152 0.23463* 4.54680 Dvest 2 0.49913* 8.33707 0.496975* 9.08968 0.50329 9.24362 R-Square 0.18984 0.39623 0.36507 Ad. R-Square 0.17986 0.38878 0.21928 *Sgnfcant at the 95% level, **Sgnfcant at the level 7
Table 4: CONFIDENCE INTERVALS FOR EXOGENOUS VARIABLES WITH NUG GENERATION Operatonal Productvt Lower Upper 90.0% 90.0% 1994 Dollar Fnancal Productvt Lower Upper 90.0% 90.0% Current Dollar Fnancal Productvt Lower Upper 90.0% 90.0% Independent Varable LN NUG - 0.05209-0.00648-0.039695 0.00139-0.04088 0.00012 LN NUG (t-1) 0.01121 0.05568-0.005299 0.03508-0.00545 0.03484 Dvest 1 0.15587 0.34379 0.152997 0.32366 0.14961 0.31965 Dvest 2 0.40049 0.59776 0.406896 0.58705 0.41359 0.59300 CONCLUSION Ths paper provdes some support to the hpothess that when one s examnng the mpact of a polc on changes n Total Factor Productvt t makes lttle dfference whether one defnes nput n terms of phscal unts or monetar costs. Whether one defnes the NUG share of the market n terms of capact or generaton, the coeffcents on the exogenous varables have the same sgn regardless of how we measure productvt. Lkewse, the values of the coeffcents n the regresson for one measure of productvt fall wthn the confdence nterval of the same coeffcent wth another measure. However, the fact that the values of the coeffcents of the NUG varables are not sgnfcant when the NUG share s defned n terms of output rases the queston whether we can unversall accept operatonal productvt as equvalent to fnancal productvt. There s a maor dfference between the Adusted R-Square for operatonal productvt and fnancal productvt. Ths could be due to the dffcult n developng phscal measures of fuel nput and captal nput. For future studes, t ma be useful to calculate the productvt at the dvsonal level generaton, transmsson, and dstrbuton. Dong so wll make t possble to come up wth a common measure of captal for each dvson. Also, determnng an opportunt cost for nternall generated electrct would make t possble to calculate the benefts of ncreasng relance on purchased power. Whle ths stud does not show clear evdence that the change n operatonal productvt s equvalent to the change n fnancal productvt, the author suggests that future studes could safel use one of the two measures of fnancal productvt. For one, the Adusted R-Squares for the regressons for fnancal productvt are much stronger. Ths suggests a better ft. More mportantl s the purpose of encouragng competton n generaton s to reduce the costs of producng electrct. Ths purpose can be acheve both b the utlt becomng more effcent n usng ts resources and b beng more dlgent n obtanng a better prce for ts nputs. Fnancal productvt wll capture both effects, but operatonal productvt wll not. 8
REFERENCES Barnette, G. S. 1998. Investment Incentves n Electrc Utltes: The Analss of Impared Assets. Unpublshed PhD dssertaton, Rensselaer Poltechnc Insttute. Barnette, G. S.; E. Gurgen; and J. R. Norsworth. 2004. Electrc Utlt Competton and Total Factor Productvt. Research Journal of Busness Dscplnes, Vol. 10, Blocher, E. J.; Stout, D. E.; and Cokns, G. Cost Management: A Strategc Emphass, 5th Edton, McGraw-Hll Irwn, 2010. Energ Informaton Admnstraton (EIA). October 2000. The Changng Structure of the Electrc Power Industr 2000: An Update. DOE/EIA-0562(00). Jorgenson, D. W. and K.-Y. Yun. 1991. Tax Reform and the Cost of Captal. Oxford, UK: Oxford Unverst Press. Norsworth, J. R. and S. L. Jang. 1992. Emprcal Measurement and Analss of Productvt and Technologcal Change: Applcatons n Hgh Technolog and Servce Industres. Vol. 211 n Contrbutons to Economc Analss, D. W. Jorgenson and J.-J. Laffont, eds. North-Holland. Scull, G. W. 1998. Reform and Effcenc Gans n the New Zealand Electrcal Suppl Industr. Journal of Productvt Analss 11: 133-147. 9