Choices Among Alternative Risk Management Strategies: Evidence From the Natural Gas Industry

Transcription

1 Choices Among Alternative Risk Management Strategies: Evidence From the Natural Gas Industry Christopher C. Géczy,* Bernadette A. Minton,** and Catherine Schrand * This paper examines risk management strategies for natural gas firms that face multiple risks (e.g., price and volume risk) and have a variety of financial and non-financial tools available to manage those risks. Natural gas firms risk exposures to price and volume risk were changing significantly during the sample period due to a unique series of regulatory changes, including price deregulation. Natural gas firms used a combination of gas storage, cash holdings, line-ofbusiness and geographic diversification to hedge increasing volume risk and changing exposure to price risk. The firms extensively use commodity derivatives when available. Holding cash and storing gas are complementary strategies that are used by similar types of firms. While the use of derivatives and storing gas are related, firms appear to use derivatives to manage price risk and gas storage to manage volume risk. Finally, the various strategies are effective. Natural gas pipeline firms that pursue financial or operational hedging activities have smaller and less variable sensitivities to price changes than firms that do not, especially post-deregulation. JEL classification: G3; L5; Q4 Keywords: Risk management; Hedging; Diversification; Derivative securities; Regulation *University of Pennsylvania. **Ohio State University. Contact author: Catherine Schrand, The Wharton School, University of Pennsylvania, 2427 Steinberg Hall-Dietrich Hall, Philadelphia, PA, 19104, (215) , schrand@wharton.upenn.edu. We thank Ed Glidewell, Josh Goza, Janet Lee, Victoria Mukovozov, Jane Song, and Vivian Sun for excellent research assistance, and Christopher Culp, David Haushalter, Vince Kaminski, Sara Moeller, René Stulz, David Shimko, Sheridan Titman, Amir Yaron, and seminar participants at Cornell University, Darden, the Ohio State University, and the University of Texas at Austin for helpful comments and suggestions. Gene Fama kindly provided access to some data used in the paper. Géczy thanks the Rodney L. White Center for Financial Research for financial support. Minton thanks the Dice Center for Financial Economics, the Dean s 1997 Summer Fellowship Committee, and Ohio State University Seed Grant committee for financial support.

2 Choices Among Alternative Risk Management Strategies: Evidence from the Natural Gas Industry Abstract This paper examines risk management strategies for natural gas firms that face multiple risks (e.g., price and volume risk) and have a variety of financial and non-financial tools available to manage those risks. Natural gas firms risk exposures to price and volume risk were changing significantly during the sample period due to a unique series of regulatory changes, including price deregulation. Natural gas firms used a combination of gas storage, cash holdings, line-ofbusiness and geographic diversification to hedge increasing volume risk and changing exposure to price risk. The firms extensively use commodity derivatives when available. Holding cash and storing gas are complementary strategies that are used by similar types of firms. While the use of derivatives and storing gas are related, firms appear to use derivatives to manage price risk and gas storage to manage volume risk. Finally, the various strategies are effective. Natural gas pipeline firms that pursue financial or operational hedging activities have smaller and less variable sensitivities to price changes than firms that do not, especially post-deregulation.

3 Choices Among Alternative Risk Management Strategies: Evidence from the Natural Gas Industry 1. Introduction Anecdotal evidence and surveys suggest that corporate managers contemplate risk management in a broad context. However, most empirical research on risk management focuses primarily on the use of one risk management tool derivatives for managing a single financial risk such as interest rate risk, foreign exchange rate risk, or a commodity price risk (e.g., Géczy, Minton and Schrand, 1997). Recently, a few studies have examined how firms use derivatives and one other risk management strategy, such as holding cash buffers (Opler, Pinkowitz, Stulz, and Williamson, 1999) or accounting earnings management (Barton, 2000; Rajgopal and Pincus, 2000). However, these studies do not examine which factors affect the strategy or strategies a firm chooses or the relative effectiveness of non-derivative risk management strategies. Examples of papers that consider multiple risk management activities are Tufano (1996) and Petersen and Thiagarajan (1999), both of which examine firms in the gold mining industry. This paper takes a broader view of risk management than the existing literature on two dimensions. First, we examine how firms use multiple risk management strategies to manage risk. The analysis focuses on the natural gas industry, in part because these firms have a variety of measurable risk management tools available. Moreover, significant regulatory changes during the 1980s changed the relative costs and benefits of using the various tools. In fact, much of the sample period prior to price deregulation occurs before the development of a formal natural gas derivatives market. We examine four risk management strategies: using derivatives, holding cash, storing gas, and diversifying, where diversification can be geographic or across lines of business. This paper also takes a broad view of corporate risk management on a second important dimension by acknowledging sources of risk besides price risk. Two significant regulatory changes took place in the natural gas industry during our sample period: the required the -1-

4 unbundling of the sale and transmission of gas and intervention in long-term contracting arrangements. These changes not only affected price risk exposure, but also increased volume uncertainty. Regulatory changes related to price deregulation also altered the firms exposure to price risk. The net effects of these regulatory changes on price risk are evident in changes in gas price exposures. Despite the fact that some of the sample firms had the opportunity pass on rate increases to customers through regulatory rate adjustments, 1 the exposures of our sample firms are reliably negative in the early part of the sample period. The exposures are positive or zero in the later part of the sample period when the firms pipeline operations were no longer acting as consumers of natural gas. After documenting the time-series pattern in the firms natural gas exposures, our substantive analysis of the firms risk management practices consists of three parts. First, we assess how firms use various strategies as substitutes and complements by examining the relations among the uses of these strategies over time. Second, we examine cross-sectional variation in the characteristics of firms that use combinations of strategies. Finally, we evaluate the effectiveness of the chosen strategies for reducing gas price sensitivity. Not surprisingly, natural gas firms typically use more than one risk management strategy. The analysis provides strong evidence of a complementary relation between gas storage and cash holdings. In addition, the firms that pursue these hedging activities have similar characteristics and are significantly different from the storage non-hedgers and cash non-hedgers. Specifically, cash hedgers and storage hedgers are more profitable and less financially constrained than are the non-hedgers. There also is a relation between levels of gas storage and derivatives use. However, differences in the types of firms that use these tools suggest that the strategies are distinct. The firms that use derivatives are less profitable and more financially constrained than are non-users. These characteristics are opposite those of storage hedgers, but are consistent with extant 1 As a robustness check, we examine whether firms that can pass on rate adjustments drive the results. -2-

5 empirical research on firms that use financial derivatives and consistent with the theories of risk management (Geczy, Minton and Schrand, 1997). One interpretation of this combination of results is that storage and derivatives are used to manage different types of risks. Anecdotal evidence also suggests that storage is used to hedge volume risk while derivatives are used to hedge price risk (Kaminski, 2001). The complementary relations among physical storage, cash holdings, and derivatives use may not be surprising because these strategies are similar in the sense that a firm can alter the extent of its involvement in each activity dynamically and at relatively low cost. In contrast, diversification is a relatively high-cost risk management strategy. Altering a diversification decision through future acquisitions and divestitures involves significant costs. In addition, there are costs associated with diversification in that diversification can reflect, and might primarily reflect, other strategic motivations besides risk management. Despite this difference in the nature of the costs associated with these strategies, diversification is positively related to storage, cash holdings, and derivatives use. Evidence on the types of firms that diversify, however, suggests that diversification is a distinct risk management activity. The more diversified firms, like derivatives users and unlike firms that use storage and cash as risk management tools, are more financially constrained. This constraint is evidenced by marginally lower interest coverage ratios and worse debt ratings. An analysis of the trends in diversification over time shows that the sample firms became more diversified across lines of business during the period from 1979 through This trend contrasts with the general trend away from diversification by U.S. firms during this time period (Comment and Jarrell, 1995), but it is consistent with a trend toward diversification in the regulated electric utilities industry (Jandik and Makhija, 2000). This difference between our sample, which represents a set of firms that were undergoing significant increases in risk, and the 2 The sample period is one of significant acquisition/divestiture activity in the natural gas industry related to the regulatory changes. Mitchell and Mulherin (1996) note that over 67% of firms in the natural gas industry are the subjects of some form of takeover bid in the 1980 s, more than half of which were hostile. -3-

6 general population of firms suggests an important role for diversification as a risk management tool. Given the compelling evidence that diversification destroys value (Berger and Ofek, 1995), we view studies of the value-enhancing role of diversification to manage risk as important future work. Finally, we provide evidence on the effectiveness of hedging activities. For portfolios of hedgers, defined by both univariate and multivariate classifications of hedgers, gas price sensitivities are lower, less variable, and less frequently significant than are those of nonhedgers. The distinction between hedgers and non-hedgers is strongest between 1985 and 1989, the period that immediately precedes the development of an exchange-traded derivatives market in gas price instruments. During these years, operating strategies, specifically storage, holding cash, and diversification, effectively reduce exposures. These findings for non-derivative strategies add to the growing literature that documents the effectiveness of derivative strategies as a risk management tool. The paper proceeds as follows. Section 2 describes the sample. Section 3 summarizes regulatory changes in the natural gas industry from 1978 to 1995 and uses gas price sensitivities to illustrate the time-series pattern of price risk during this period. In Section 4, we discuss risk management activities and their use by our sample firms. In Section 5, we characterize the effectiveness of the risk management strategies, individually and in combinations, for controlling price risk. Section 6 concludes. 2. Sample The sample represents publicly traded natural gas companies that were (or had subsidiaries that were) major interstate natural gas pipelines, during the period between 1978 and In almost all years, these major interstate pipelines account for over 80% of the total gas receipts, assets, sales, and gas operating revenues in the pipeline industry. Although the sample firms represent the majority of the pipeline industry, the firms are not pure pipeline firms. Only a few pipelines are publicly traded entities; most are subsidiaries of natural gas companies or -4-

7 holding companies that also own other energy or transportation subsidiaries (e.g., railroad and trucking operations). This sample has advantages and disadvantages. Perhaps the most compelling advantage of our sample is that the firms face a common, significant and clearly definable risk related to natural gas prices as we document in Section 3.2. However, price risk is not the only risk these firms face. They also face significant volume risk. A second advantage is that the regulatory changes during the sample period significantly affect the types and magnitudes of risks the sample firms face. This unique setting permits a controlled examination of an array of firms risk management activities. A third advantage is that the firms have a variety of tools available to manage the significant and changing exposures to price and volume risk. In particular, the early part of our sample period is prior to the development of an active spot market and exchange-traded derivatives market for natural gas futures and options. Thus, we can examine the use and effectiveness of operating risk management strategies when firms do not have effective financial hedging tools available. Finally, there are two practical advantages of this sample. Regulatory reporting requirements at the pipeline level provide better data than are typically available in annual reports for measuring the extent to which firms use a wide variety of risk management tools. In addition, the firms are in the same industry, which mitigates potential problems associated with unidentified or omitted industry-specific effects. In addition, one aspect of the sample that is both an advantage and a disadvantage is the heterogeneity of the firms within the natural gas industry. Although we capture most of the pipeline industry, the firms are diversified into other natural gas-related activities, such as production, as well as some businesses that are totally unrelated to the natural gas industry. Thus, all of the firms are exposed to natural gas price risk, albeit for varying reasons and in different ways. This heterogeneity is a strong advantage in that it allows us to examine line-ofbusiness diversification as a risk management tool. -5-

8 On the other hand, the heterogeneity reduces the power of the empirical tests. In the calculations of natural gas betas, we attempt to address the potential problems associated with heterogeneity through sensitivity analyses (see Sections 3.2 and Section 5). We also perform supplemental analyses on subsets of the full sample that more closely represent a pure play on gas prices. All results are qualitatively similar. Another disadvantage of the sample is that we are forced to use a dichotomous specification of derivatives hedgers based on the use of commodity derivatives. Detailed data on the extent of derivatives use are limited, which makes it difficult to create a reliable continuous measure of derivatives hedging and still maintain a reasonable sample size. However, under the dichotomous specification, almost all of the sample firms are hedgers. We mitigate this problem by providing supplemental analyses using only the firms for which more detailed derivatives data are available. A final disadvantage of our sample is that because of the focus on natural gas firms, the results may not generalize to firms outside the natural gas or other extractive industries. We identify major pipelines using the Energy Information Agency Form 2 filings, which are required by the Federal Energy Regulatory Commission (FERC, the industry regulatory agency). 3 We match the subsidiary pipelines to publicly traded entities on the Center for Research in Security Prices (CRSP) database based on reviews of holding company annual reports, selected discussions in the annual Statistics of Interstate Natural Gas Companies, and LEXIS/NEXIS searches. Joint ownership of some pipelines complicates the process of combining the FERC data, which is reported at the pipeline level, with Compustat or CRSP data, which is reported for the entity that owns the pipeline. For jointly owned pipelines, we divide the pipeline data according to the entities ownership percentages. Thus, a sample firm s natural 3 Form 2 defines major interstate natural gas pipeline companies as those firms that have combined gas sales for resale and gas transported (interstate) or stored for a fee that exceeded 50 billion cubic feet during the preceding calendar year or those firms that filed FERC Form 11, Natural Gas Pipeline Company Monthly Statement. -6-

9 gas data are the weighted averages of the data for each FERC pipeline that it owns. Over 95% of the major interstate pipelines identified are matched with one or more publicly traded firms. Table 1 lists our sample of holding companies and the minimum and maximum number of pipelines owned by each firm during its tenure in the sample. After restricting the sample to firms with available data on Compustat, the final sample contains 394 pooled firm-year observations. The annual samples contain a minimum of 21 firms (1978) and a maximum of 25 firms (1985). [INSERT TABLE 1 HERE.] 3. The effects of regulation In this section we summarize the operations and regulation of the natural gas industry that affect risk exposure and illustrate the time-series trends in the sample firms stock return sensitivities to gas returns during the period of deregulation. We base this (necessarily) limited discussion of the effects of regulation on more extensive analyses performed by the American Gas Association (1987), Castaneda and Smith (1996), Fitzgerald and Pokalsky (1995), and the New York Mercantile Exchange (undated). 3.1 Overview of regulation Between the mid-1950s and late1970s, regulated pricing and the bundling of the sale and transportation of gas were two of the central features of the natural gas industry. Pipeline companies purchased gas from producers at regulated wellhead prices. Pipelines transported purchased gas, and possibly their own production, to city gates for resale to local distribution companies and other end users and sold this gas at regulated city-gate prices. 4 The spread 4 Although prices were regulated, they were not fixed. However, some of the sample firms had the ability to attempt to effectively fix their price exposure by passing price changes through to customers. The effectiveness of this strategy is limited because the regulatory appeal process can be lengthy and the outcome is uncertain. Kaminski (2001) notes that the regulated pipelines that attempted to pass through price changes engaged in a costly risk -7-

10 between the city-gate price and the wellhead price represents a pipeline s profit margin to cover the cost of transportation. Throughout this period, there were no centralized spot or futures markets for the purchase of natural gas. In the 1970 s, natural gas firms commonly entered into long-term contracting arrangements. Pipelines purchased gas from producers under take-or-pay contracts and sold gas to end users via contracts containing minimum bill provisions to guard against gas shortages and mitigate price risk (Masten and Crocker, 1985). 5 A take-or-pay contract required that the pipeline pay for a minimum negotiated quantity of gas during the contract period at a preset regulated wellhead price (plus tariffs and other transaction fees). A minimum bill provision similarly required that the end user pay the pipeline for a minimum negotiated quantity of gas during the contract period at the regulated city-gate price. Under both contracts, while the gasreceiving party did not have to take delivery of all purchased gas, it was required to pay for the contracted amount. The significant regulatory changes in the natural gas industry during our sample period were designed to increase the flow of gas into interstate markets and force competition in wellhead prices. The regulatory changes affected pipelines price and volume exposures by deregulating wellhead and city-gate prices, requiring the unbundling of the production and transportation of natural gas, and explicitly intervening in long-term contracting arrangements. The Natural Gas Policy Act (NGPA) of 1978, first discussed in April 1977, was the first regulatory change related to deregulating prices in the industry. The NGPA removed controls over market entry and wellhead gas sales, established categories of gas based on the age of the gas, and initiated the deregulation of wellhead prices for some categories. FERC Order 436 in 1986 and the Decontrol Act of 1989 implemented further price deregulation of various gas management activity maintaining a good relationship with regulators. In Section 4.2.2, we perform a separate analysis of firms with potential access to regulatory rate adjustments to fix prices. 5 The intuition that long-term contracting mitigates price risk is supported by theoretical arguments in Hubbard and Weiner (1986), whose study analyzes markets with both regulation and bargaining possibilities. -8-

11 categories. The changes in regulation during this time period did not explicitly discuss quantity controls or quantity rationing. Figure 1, Panel A, illustrates the time-series pattern in natural gas prices pre- and postderegulation. Prices rose steadily prior to and immediately after the start of deregulation in After a consistent price decline beginning in 1985, price volatility increased substantially in the later part of the sample period. Higher volatility was accompanied by increases in the magnitude of natural gas return exposures for producers, pipelines, and end users, but did not affect the direction of firms exposures. [INSERT FIGURE 1 HERE.] While wellhead price volatility was increasing, regulatory changes also were altering the role of pipelines from that of consumers of gas for transport and resale to that of only transporters of gas between sellers and buyers. These regulatory changes increased volume uncertainty for the pipelines. As the first major step in the unbundling of the sale and transportation of natural gas, FERC Order 436 (October 1985, first discussed in December 1984) required pipelines that chose to transport gas under the order to provide open access transportation services to natural gas producers. FERC Order 636 (April 1992, first discussed in early 1991) mandated that all pipelines unbundle the sale and transportation of natural gas by the end of The required unbundling altered the direction of the pipelines exposure to natural gas prices and also the pipelines volume risk. As a result of unbundling, total gas sales by the sample firms declined steadily during the sample period from 59.99% of total sales and deliveries in 1979 to 7.92% in Gas deliveries, by contrast, increased from 40.01% of total sales and deliveries in 1978 to 92.07% in This regulatory change had a less significant impact on pure natural gas producers and end users, such as utilities. -9-

12 Regulatory changes related to contracting arrangements also affected natural gas firms exposure to price risk and volume uncertainty. In the early 1980s, lower demand for natural gas triggered minimum bill clauses in contracts between end users and pipelines and take-or-pay clauses between pipelines and producers (Henderson, Guldmann, Hemphill, and Lee, 1986). FERC Order 380 (May 1984, first discussed in late 1983) freed end users from their minimum bill provisions in contracts with the pipelines, but did not free the pipelines from their take-orpay contracts with producers. Although some parties continued to honor the contracts believing that prices would turn around (Kaminski, 2001), a common result was huge losses for the pipelines that led to significant industry-wide financial distress (Castaneda and Smith, 1996). This industry-wide distress is partly evidenced by changes in the number of sample firms. After reaching a maximum of 25 in 1985, the number of firms declined to 21 in FERC eventually responded to the impending financial burden in FERC Orders 436 (1985) and 500 (1987), which moved to free pipelines from their take-or-pay contracts with producers. Despite the mitigating effects of FERC orders 436 and 500, regulatory intervention in the contracting process had permanent effects on contracting, and thus price and volume risk, in the industry. The use of privately negotiated long-term contracts declined as end users increasingly purchased gas in the spot market to alleviate concerns about contract risk and credit risk (Fitzgerald and Pokalsky, 1995). Long-term contracting also declined because of acrimony between the contracting parties the pipelines and producers related to numerous lawsuits within the industry. 6 Greater dependence on spot market purchases increased the price risk exposures of pipelines, producers, and end users. However, the asymmetric treatment of pipelines purchase and sale contracts meant that the pipelines were more affected by intervention in the contracting process than were producers or end users. 6 We thank Charles Smithson for noting this dimension of the contracting process. Kaminski (2001) also notes that it became culturally and practically difficult to use these contracts following FERC Order

13 3.2 Gas price sensitivity This section illustrates changes in the sensitivities of stock returns to natural gas price returns during the period from 1978 to 1995 for a portfolio of the sample natural gas pipeline firms and for a benchmark portfolio of natural gas producers. The measurement of firms gas price sensitivities is complicated by the potential simultaneity between the effects of a firm s risk management decisions and the observed market-based measures of risk. Therefore, although we perform a number of checks designed to test the robustness of the methods and estimated sensitivities presented here and in Section 5, we do not claim that the observed changes in gas price sensitivities were necessarily related to changes in regulation alone. The changes also reflect the firms responses to the regulation and to the same factors that inspired the regulation. Moreover, firms can anticipate and react to the potential regulation prior to its enactment. Thus, the sensitivities presented here illustrate only general trends. We measure natural gas price sensitivities by using an extended Fama-French (1993) model. Eq. (1) regresses returns for an equal-weighted portfolio of the sample natural gas firms (R P ) in excess of the rate of return on 3-month Treasury bills (R TB ) on the equal-weighted market return (R M ) in excess of R TB ; 7 the excess return of average wellhead prices (R NG R TB ); the excess return on the domestic first purchase price of crude oil (R OIL R TB ); the return on a size factor-mimicking portfolio (R SMB ); the return on a value-effect factor-mimicking portfolio (R HML ); the return on a portfolio capturing the slope of the term structure of interest rates (R TERM ); and the return on a portfolio that is long low-grade corporate bonds and short highgrade corporate bonds (R DEF ), capturing the corporate default spread: R P R TB =α+β M (R M R TB ) + β NG (R NG R TB ) +β OIL (R OIL R TB ) + +β HML R HML +β SMB R SMB +β TERM R TERM +β DEF R DEF +ε. (1) 7 Our results are qualitatively robust to the choice of market measure. -11-

14 The coefficient, β NG, represents the portfolio s sensitivity to average wellhead returns. We calculate these returns using monthly Henry Hub wellhead prices obtained from The Monthly Energy Review. We use an equal-weighted portfolio since a small number of very large firms could drive estimates from value-weighted portfolios, especially in the later years in the sample. We include the excess return on crude oil, which we calculate using the EEFPP series from the Standard & Poor DRI database, because natural gas producers generally also engage in oil production. We construct R TERM and R DEF using data from the Standard & Poor DRI macroeconomic database. 8 We obtain market and T-bill returns from CRSP. 9 Consistent with the methodology of Fama and French (1997), we estimate sensitivities using monthly data and rolling 48-month windows from 1978 to 1995 (i.e., we estimate the first observation for the 48-month period ending December 31, 1981). 10 These regressions provide 169 estimates of natural gas price sensitivities using data from the overlapping windows. Motivated by the caveats in Tufano (1998) on potential problems related to using monthly data, we use weekly spot price and market data to verify the robustness of the beta estimates when feasible. Weekly spot prices are available only since the early 1990s. When we use weekly data for this limited period, we find that the natural gas betas are qualitatively similar to the monthly results. [INSERT TABLE 2 HERE.] Table 2 presents summary statistics for β NG and β M for the full sample of natural gas pipelines (Panel A) and for the benchmark sample of producers (Panel B). We report t-statistics for the mean of the rolling beta estimates corrected for the use of overlapping observations 8 Gene Fama kindly provided the HML and SMB time series that we use in the regression. 9 We also use a two-factor market model to estimate gas price sensitivity. The results are qualitatively very similar. We present the results for the Fama-French (1993) model because it includes a control for the potential impact of interest rates, which were extremely volatile during the sample period, on the estimates of natural gas price sensitivities. 10 Results for 36-month windows are similar. -12-

15 (Hansen and Hodrick, 1980). The average wellhead return beta for the portfolio of pipelines over the entire sample period is (significant at the 5% level based on the corrected t- statistic), and the standard deviation is The minimum β NG is and the maximum is At the 10% level, only 18% of the β NG estimates are different from zero, but Figure 1, Panel B, illustrates that the insignificant sensitivities are clustered in the middle of the sample period. The β NG estimates that are reliably different from zero are negative at the beginning of the sample period and positive near the end. This change in the direction of sensitivities corresponds to the required unbundling of pipeline services. The significant negative betas prior to deregulation reflect the role of pipelines as consumers of gas, when the sample firms purchased significant quantities of gas for transportation and resale. Over time as the pipeline operations of the sample firms adopted the new regulations and transitioned into their roles as pure transporters of gas, their sensitivities to gas exposures became less negative. The positive betas at the end of the sample period are consistent with the gas exposures of a distributor rather than a consumer of gas. The positive sensitivities are significant beginning in October 1992, which corresponds to the timing of the final phase of price deregulation and higher price volatility after the Decontrol Act of The timing of the positive significance after 1992 is also consistent with the initiation of New York Mercantile Exchange (NYMEX) natural gas futures and options contract trading in 1990 and 1992, respectively, given that the beta estimate shown for a particular month represents the estimate over the preceding 48 calendar months. The sensitivities are relatively less volatile than in periods before these financial instruments were available. Panel B, Table 2, presents natural gas betas and market betas for an equal-weighted portfolio of oil and gas producers, which we compare to the exposures of the sample natural gas pipeline firms. The producer sample represents all firms on CRSP with an SIC code of This sample selection criterion is consistent with extant research on oil producers (e.g., Haushalter, 2000; Haushalter, Heron, and Lie, 2000). -13-

16 The results in Table 2 suggest that the market exposure of the producers is similar to that of the pipelines. However, the producers gas sensitivities are never reliably different from zero. 11 Because the pipelines exhibit significant changes in gas price sensitivity over the period of deregulation, we focus on the pipelines and exclude producers from our analysis Alternative Risk Management Strategies As is evident from the preceding section, deregulation of the natural gas industry created a number of uncertainties for pipeline firms, the most significant being price and volume uncertainty. In this section, after describing the various risk management strategies available to natural gas pipelines, we use two different analyses to provide evidence about how firms combined these strategies to manage risk. First, we examine whether tools are substitutes or complements. Second, we examine the characteristics of firms using different risk management strategies. Each analysis individually suffers from some econometric limitations, resulting primarily from the relatively small sample size. However, together the two analyses provide consistent and corroborative results. We consider four risk management strategies: derivatives use, holding cash, storing gas, and diversification. 13 The four strategies differ on the important dimension of their relative costs. We consider a strategy to be low cost if a firm can alter the extent of its involvement in the 11 For our sample period from , the mean oil price beta for the benchmark sample of oil producers is The oil return sensitivity estimates are significant in 69% of the time-series regressions (at the 10% level). For the pipelines, the mean oil price beta is 0.078, significant in 38% of the regressions. The sensitivities for the producer sample are comparable to those reported in other oil and gas industry studies. Haushalter, Heron, and Lie (2000) report a mean (median) sensitivity to daily percentage changes in oil prices of (0.208) for 68 oil and gas producers over the period Using a similar methodology, Rajgopal and Venkatachalam (2000) report a mean (median) sensitivity of 0.09 (0.10) for a sample of 25 oil refiners over the period , with 75% of the estimates significant at the 5% level. 12 For an analysis of producers risk management activities, see Haushalter (2000). 13 We also considered examining contracts as risk management tools but do not for various reasons. Take-or-pay contracts, as described previously, were used during the early part of the sample period. Firms continue to use purchase and sale contracts, but without take-or-pay clauses. The current contracts are generally between producers, marketers and end users; the pipelines act merely as distributors with no commitment (Kaminski, 2001). Data are not available to measure firms involvement in either take-or-pay contracts or purchase and sale contracts. Insurance contracts also are not commonly used (Kaminski, 2001). We consider appealing to regulators as a risk management tool separately later in this section. -14-

17 activity from period to period at relatively low cost. Activities that require significant costs of initiation or termination in the short run are defined as relatively high cost strategies. Thus, using derivatives and holding cash are low cost strategies. Diversification is a high cost strategy and storage is somewhere in between. The benefits of each risk management tool differ across firms because the types of risks that firms manage (i.e., price versus volume) differ and because the incentives for risk management vary across firms. For example, some theories of risk management predict that firms will optimize firm value by reducing cash flow volatility. The underinvestment story of Froot, Scharfstein and Stein (1993) is an example. Other theories imply that firms will optimally manage earnings variability. Examples are the bonus compensation story of Smith and Stulz (1985) or the signaling story of DeMarzo and Duffie (1994). Yet others imply that firms increase firm value by managing the market values of assets and liabilities (e.g., the equity compensation stories of Smith and Stulz or the debt overhang story of Myers, 1977) or by managing taxable income (Smith and Stulz, 1985). Finally, theories of holding inventories (e.g. Working, 1962 and Williams, 1986) imply that firms have incentives to manage volume exposures. Since the incentives for hedging and thus the goals of hedging differ across firms, the benefits of a particular risk management tool also can differ. We discuss these differences as we discuss each risk management tool. We take as given that firms want to reduce risk and examine the tools they use to do so. However, the theories of risk management imply predictable cross-sectional differences in the firms that would choose one tool over another given its potential benefits and costs. Following this analysis of the substitutability and complementarity of the various tools for managing risk in Table 4, we analyze characteristics of firms that choose one tool versus another or that choose various combinations of tools. -15-

18 4.1 Summary of strategies for managing risk Financial derivative instruments The sample firms can use financial derivative instruments to reduce gas price risk, but exchange-traded natural gas price derivatives are a relatively recent innovation. The first natural gas futures contract and options on futures began trading on the New York Mercantile Exchange (NYMEX) in April 1990 and October 1992, respectively. In addition to exchange-traded derivatives, firms can utilize over-the-counter options, forwards, swaps and other derivative instruments. For most derivatives, although the contract is based on the price of natural gas, the firm does not take (or make) delivery of the gas purchased (or sold). Thus, derivatives are generally used to hedge price rather than volume risk. Derivative instruments can be used to hedge variability in cash flows, earnings, or market values of assets and liabilities. According to the Wharton/CIBC Wood Gundy survey of derivatives use in 1995, 49% of firms reported that the most important objective of their hedging strategy was to manage cash flows, 42% to manage accounting earnings, 8% to manage the market value of the firm, and 1% to manage balance sheet accounts. Derivative instruments are available at relatively low cost regardless of a firm s hedging goals. Basis risk is a potentially significant cost of using derivatives. Prior to the development of natural gas futures contracts 1990s, firms wanting to use exchange-traded instruments had to cross-hedge with exchange-traded products based on another commodity such as oil prices. However, this strategy involves basis risk. The coefficient of variation of monthly natural gas prices from January 1991 to January 1995 was and the coefficient of monthly oil prices was (Haushalter, 2000). After the introduction of contracts on natural gas prices, basis risk remains because the most actively traded futures contract is based on the Henry Hub (in Louisiana) price of natural gas. However, natural gas spot prices, like most commodity prices, can vary geographically (Chambers, 1999). During our sample period, the correlations between wellhead prices at various geographic locations were relatively stable. An unprecedented basis -16-

19 blowout occurred only after our sample period in 1996 (Lagrasta, Kaminski, and Prevatt, 1999). 14 We obtain data on the use of derivatives from 10-K filings and annual reports. The data are available consistently only from 1993 through 1995, since firms have been required to report the use of derivative instruments only for fiscal years ending after December 15, 1994 (SFAS No. 119). We use the 1994 disclosures, which report outstanding positions at the previous yearend, to ascertain derivatives use beginning in We use an indicator variable (DERIV) to measure a firm s use of financial derivative instruments. Continuous measures of the magnitude of off-balance sheet activities are frequently inconsistent or missing. For each sample year, DERIV is equal to one if the firm reports outstanding commodity derivatives at fiscal year-end or reports using commodity derivatives during the year, and equal to zero otherwise. We designate all commodity derivatives users as hedgers even though the use of derivatives can imply speculation rather than risk reduction (Géczy, Minton, and Schrand, 1997). Several firms (i.e., El Paso Natural Gas, Questar, and Enron) report that they use derivatives for both risk management and trading. The use of natural gas derivatives for trading supports the claim that trading in derivatives can be a positive NPV activity for a firm that believes it has insider knowledge of prices or that can effectively act as a market maker. 15 Although the use of derivatives can reflect speculation, the data do not allow a better specification of hedgers. Table 3 reports that 83.3% of the sample firms use some form of commodity derivatives as of The high percentage of derivatives users is similar to that in other industries with 14 A unique feature of the natural gas contract environment is that there is a very vibrant stand-alone basis market associated with the commodity. The basis market has become increasingly important after the basis blowout in 1996 (Lagrasta, et al., 1999). 15 Campbell and Kracaw (1999) provide a theory of optimal speculation based on asymmetric information. Stulz (1996) discusses the trading of interest-rate based derivatives by Banc One, as an example. 16 Although the statistics show that almost all the sample firms use derivatives, the statistics do not reveal which natural gas segments are using them. Kaminski (2001) provides some anecdotal evidence about the use of derivatives across segments. Some pipeline operators stayed out of the derivatives market because of management conservatism. Other pipelines historically had been successful at passing price changes on to customers and thus were less likely to use derivatives. Some production segments did not use derivatives because the market is too small to hedge all production-related price risk, so they did not hedge any. Still other firms saw derivatives losses -17-

20 price risk exposure concentrated in a single commodity. For example, Haushalter (2000) reports that in 1994, 58% of his sample of oil and gas producers uses commodity derivatives. Tufano (1996) reports that between 1990 and 1993, over 85% of his sample of gold mining firms manage gold price risk using financial instruments Holding cash A second relatively low-cost risk management strategy is to hold an internal buffer of liquid assets. The cost of holding large cash balances is the firm s opportunity cost of capital. Holding cash benefits firms attempting to reduce costs associated with cash flow volatility, such as underinvestment (e.g., Froot, et al., 1993). However, holding cash does not mitigate costs associated with earnings or value fluctuations (e.g., taxes, bonus compensation, etc.). Thus, we expect firms whose costs of underinvestment are high to hold more cash, which implies that we expect firms that are financially constrained and have high growth opportunities to hold more cash (Geczy, Minton, and Schrand, 1997). Internal cash (CASH) for the gas subsidiary is the ratio of cash to total assets minus cash, following Opler, Pinkowitz, Stulz, and Williamson (1999), using FERC Form 2 data. 17 The results in Table 3 do not show a clear time-series pattern in the trend of cash holdings. However, using the cash balance only at year end likely measures the construct of cash holdings with error. The cross-sectional relations between cash holdings and other activities presented in Section 4.2 are more revealing. and considered hedging to be a failure. In summary, the high rates of derivative use by the parent company do not provide a clear picture of pipeline company derivatives use. 17 We also compute cash for the pipeline holding company using Compustat data. Extant research on the funding of investment by internal capital markets does not provide a clear indication of whether cash holdings in this context should be measured at the subsidiary level or the parent level. In particular, Shin and Stulz (1998) document that investment is more sensitive to segment-level cash flows than to firm-level cash flows, but there is still a statistically significant relation between firm-level cash flows and investment. Lamont (1997) concludes that oil segment cash flow affected investments of non-oil segments in 1985, but only subsidiary-level cash flow affected investment in He concludes that parent companies stopped subsidizing their non-oil segments after a 1985 price decline and left the segments to depend only on their own cash flow. Regardless of the metric that we use to measure cash holdings, however, the results are similar. -18-

21 4.1.3 Storage Working (1962) was among one of the first economists to develop a theory explaining why firms store inventories to manage risk. Williams (1986) relies on Working s theory to develop a theory of demand accessibility and outlines four main reasons about why firms hold inventories: speculative demand, transactions demand, precautionary demand, and pure storage. 18 Firms have a speculative demand for storage if they anticipate a price increase and project that storage costs will be less than the expected savings from purchasing the commodity in the current period. Firms have a transactions demand when storage costs are less than the costs of purchasing inventory on an as-needed basis (e.g., administrative or transportation costs). Firms have a precautionary demand in the face of demand uncertainty when storage costs are less than stock-out costs. Firms have a pure demand when they can earn abnormal rents from the storage activity because of a comparative cost advantage. Speculative demand in Williams taxonomy corresponds to price hedging while transactions and precautionary demand correspond volume hedging. Williams pure demand theory of storage does not necessarily apply to pipelines, because storage is not their core business. 19 For all four theories of demand accessibility, firms engage in costly storage if the marginal benefit exceeds the marginal cost. The cost of storage is a function of the price of natural gas. Holding the benefits of storage constant, increasing gas prices through the mid- 1980s suggest a decrease in storage, and the declining gas prices beginning in 1985 suggest an increase in storage. However, predictions on changes in the absolute level of storage are ambiguous because the benefits of storage depend on which theory describes a firm s motivation for storing gas. If 18 Williams (1986) provides an extensive discussion about the differences between his theory of demand accessibility and Working s theory of the supply of storage. See also Culp (2001) for a discussion of Working s theory in the context of risk management. 19 An analysis of footnotes to financial statements and management discussions in SEC filings of sample firms indicates that some firms do store gas for others. However, inconsistent disclosures about levels of storage for others or profits from storage activities prohibit us from developing a proxy for gas stored for others. The lack of data on this topic suggests that it is not a significant activity for many firms and it is never a primary activity. -19-

22 storage is motivated by speculative demand, the benefits depend on price expectations and are uncertain. However, Williams (1986) claims that commodity storage generally results from transactions or precautionary demand. Negative correlations between wellhead prices and storage and between wellhead price changes and changes in storage are consistent with this claim. 20 Anecdotal evidence also supports Williams (1986) claim. In its 1992 annual report, Questar argues that precautionary demand motivates greater storage following deregulation: As the industry moves into the post-order No. 636 era, storage appears to be increasing in value by providing shippers needed flexibility in responding to changing market conditions. Producers may find storage a useful way to aggregate production and balance fluctuating demand. Thus, based on the Williams analysis, storage is expected to hedge volume risk and should therefore increase through the sample period as deregulation increased volume uncertainty. Kaminski (2001) also notes that storage is viewed primarily as a volume risk management tool in the industry. However, Culp and Miller (1995) contend that physical storage should never be more expensive than derivatives. If a firm s marginal storage costs exceed the marginal storage costs priced into the derivatives basis, the firm will use derivatives. If not, the firm will store the commodity. In either case, the firms that use storage more extensively should be those firms whose motivation is totally precautionary (i.e., the firm faces greater volume risk) or for which storage is a low-cost solution. Storage (STORAGE) represents gas in underground storage in billions of cubic feet (bcf) at fiscal year-end scaled by the total quantity of gas sales and deliveries (in bcfs) for the year. Gas in underground storage includes the amounts in both the current and non-current sections of the FERC 2 balance sheet. The storage data in Table 3 indicate no obvious trend in underground storage activities during the sample period. The table shows an increase in the mean and median levels of storage 20 The average within-firm correlation between wellhead prices and STORAGE for the 24 sample firms with five or more time-series observations is -0.17; the range is to 0.46 (results not tabulated). All but six of the 24 firms exhibit the predicted negative correlation. The average firm-specific correlation between STORAGE and the percentage wellhead price change for the year is -0.21, with a range of to 0.63 (17 of 24 are negative). -20-