a z}: E 2:...3. it..." .. 1.. sm .2 MM 1 a. s :2. .. \.:x‘x.‘..... i; 1 9 . £51? I... r N i! "I" ":9 6f4$I i Vfi‘xlothl t~eanx Q ‘ w i O r- Atzv - yawn-«~- .“‘...;.‘u , . . . “I‘LL L .h : ... ‘ x . . a”??? .. .;.. ii.“ ,.. {5.5.2.4. .5». :14 unnuflxfr .EE... . . , . ,. M5 q50'1/ This is to certify that the dissertation entitled Two Essays on the Impact of Deregulation on Labor in the Electric Power Industry presented by Chiung-Ying Cheng has been accepted towards fulfillment of the requirements for Ph . D degree in Economics ,/ 7 I 1’ . I ,7 ,- I o g ' f/L _._..- -...-—- » Major professor Date L/ZZ’ 5’2" MS U is an Affirmative Action/Equal Opportunity Institution 0-12771 LIBRARY Michigan State University PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE 6/01 cJCIRC/DateDuepss-p. 1 5 TWO ESSAYS ON THE IMPACT OF DEREGULATION ON LABOR IN THE ELECTRIC POWER INDUSTRY By Chiung-Ying Cheng A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Economics 2002 ABSTRACT TWO ESSAYS ON THE IMPACT OF DEREGULATION ON LABOR IN THE ELECTRIC POWER INDUSTRY BY Chiung-Ying Cheng Chapter One of this dissertation focuses on the impact of deregulation on labor earnings, as well as employment. The findings indicate that union workers in the electric power industry experienced a significant decline in their wage premiums after deregulation (13 percent), while wage premiums of the nonunion workers remained unchanged. Level of employment in the electric power industry exhibited a pattern similar to that of wages, in which the relative employment of union workers was substantially reduced following deregulation (37 percent) while the relative employment of nonunion workers did not show a significant change. The sensitivity analyses find that high electricity prices may have contributed to the wage reductions occurring in those deregulated states, but that the reductions of employment level in the deregulated states were not related to high electricity prices in these states. In other words, we may not observe significant wage reductions in states with low electricity prices one they deregulate. However, deregulation seems to put pressure directly on the power companies to cut employment. The findings of Chapter One are consistent with the labor rent sharing hypothesis, which states that labor in the regulated industry, especially union workers, is likely to share rents with their employers. However, we find that high electricity prices, instead of deregulation itself, might have contributed to wage reductions in the deregulated states. Based on the estimated 13 percent drop in union wage premiums, union workers in the electric power industry shared, at least, modest rents with their employers before deregulation. Furthermore, in light of the dramatic reductions in union employment shown in our results, unions in the electric power industry might have traded off the level of employment against wages. Chapter Two studies unionization and the labor demand using the electric power industry as an illustration. This study finds that the unionization rate has a small but statistically significant effect on the wage elasticity of demand for labor. According to our results, every one percent increase in the unionization rate is accompanied by a .0006 percent decrease in the wage elasticity (less elastic). However, this study does not find significant changes in the union effect on wage elasticity over the course of deregulation after taking into account the effect of technological advances. The finding of a significant union effect on wage elasticity of labor demand supports the argument that unions care about employment as well as wages. The robustness checks suggest some inconsistency between our theoretical assumption about the production function and the estimated results. This could either be a result of the proxy for the cost of capital used in this study being a poor indicator, or be a result of some mis-specifications in the production function. The findings of these two studies suggest that unions have played a very important role in the electric power industry in the pre-deregulation regime. Unionized workers shared rents with their employers. Unions were also able to affect the labor demand by making the wage elasticity of demand for labor less elastic. HI'Z.IH‘THZ .ID a ACKNOWLEDGEMENTS I would first like to express my gratitude to my advisor, Professor David E. Neumark, for his mentorship. My dissertation would not have developed into its finest form without his guidance and support. His thoroughness and efficiency in reading my dissertation were astonishing. His comments always pointed right to where it was weak. It was his insistence on clarity and precision that brought out the best in me. His discipline made me think and write like an economist. My thanks also go to Professor Kenneth D. Boyer and Professor Jeff E. Biddle for their serving on my dissertation committee. Professor Boyer gave me valuable suggestions on the writing of the introduction chapter of my dissertation. Dr. Biddle provided comments that improved the contents of my dissertation. In addition, I would like to thank Professor Robert J. LaLonde, my former advisor, for his support, guidance and inspirations throughout my graduate studies. The discussions with Daiji Kawaguchi and Mau-Sheng Chen have stimulated many interesting and helpful thoughts for the writing of my dissertation. Alaknanda Bagchi and Ralph E. Pyle have tirelessly proofread my dissertation and made themselves available when I needed help. Terry N. Terry arranged a cozy work environment that allowed me to concentrate on my dissertation writing. Lisa A. Knight provided immeasurable emotional support when I was stressed out. Emeritus Professor Subbiah Kannappan’s encouragement and suggestions were always helpful. I am indebted to all of them. iv Last but not least, I want to thank my dearest mother, brothers, and sister for their standing behind me for all these years. My mother receives my deepest gratitude for her love has always been a source of strength to me in those years of struggles, although she was thousands miles away. Finally, I want to thank all the people who helped me along my academic journey and who kept me company in those years of studies and struggles. I want to share my joy and happiness with all of you. I am grateful for your friendship, love, and support. TABLE OF CONTENTS LIST OF TABLES ........................................................................... viii LIST OF FIGURES .............................................................................. ix INTRODUCTION—OVERVIEW OF THE ELECTRIC POWER INDUSTRY ...... 1 I. THE INDUSTRY BEFORE THE 1992 DEREGULATION .................................................. 1 Entry and rate-of-return regulation ....................................................................... 1 Major legislative events ......................................................................................... 4 A decade of changes—the 1980s ............................................................................ 5 II. THE 1992 DEREGULATION AND AFTERWARDS ......................................................... 8 California’s deregulation failure ......................................................................... 10 Obstacles to deregulation ..................................................................................... 12 Labor in the electric power industry ..................................................................... 12 Bibliography ............................................................................................................... 15 CHAPTER 1 THT IMPACT OF DEREGULATION ON LABOR EARNINGS IN THE ELECTRIC POWER INDUSTRY ..... _ - _ -_ -17 I. INTRODUCTION ........................................................................................................ 17 II. REGULATION, DEREGULATION AND UNION RENT SHARING .................................... 25 III. DESCRIBING STATE DEREGULATION ..................................................................... 28 IV. DATA ................................................................................................................... 32 V. ESTIMATING WAGE EFFECTS OF STATE DEREGULATION ......................................... 33 VI. SENSITIVITY ANALYSIS ON THE WAGE EFFECTS OF DEREGULATION ....................... 39 Can the presence of high electricity prices in the deregulated states prior to deregulation have also contributed to the wage reductions? ............................ 39 VII. ESTIMATING EMPLOYMENT EFFECTS OF STATE DEREGULATION ............................. 41 VIII. SENSITIVITY ANALYSIS ON THE EMPLOYMENT EFFECTS OF DEREGULATION .......... 46 Can the presence of high electricity prices in the deregulated states prior to deregulation have also contributed to the reductions ofemployment level? 46 IX. INTERPRETATIONS OF RESULTS .............................................................................. 47 X. CONCLUSIONS ....................................................................................................... 48 BIBLIOGRAPHY ........................................................................................................... 50 APPENDIX l-A ........................................................................................................... 53 CHAPTERZ UNIONIZATION AND THE LABOR DEMAND: AN ANALYSIS OF THE ELECTRIC POWER INDUSTRY ........ - 54 I. INTRODUCTION ........................................................................................................ 54 II. CONSTRUCTING THEORETICAL MODEL ................................................................... 56 III. DATA SOURCES ................................................................................................... 61 IV. DATA DESCRIPTION .............................................................................................. 64 V. ESTIMATING THE EFFECTS OF UNIONIZATION ......................................................... 67 VI. ACCOUNTING FOR THE EFFECTS OF TECHNOLOGICAL CHANGES ............................. 72 VII. ROBUSTNESS CHECKS ......................................................................................... 76 State-level data vs. firm-level data ...................................................................... 76 vi Restriction violation ............................................................................................ 78 VIII. CONCLUSIONS .................................................................................................... 80 BIBLIOGRAPHY ........................................................................................................... 81 APPENDIX 2-A. .......................................................................................................... 83 APPENDIX 2-B ............................................................................................................ 85 CONCLUDING REMARKS ...... - _ -- -- - 87 vii LIST OF TABLES Table 1.3.1. Status of State Electric Industry Restructuring by May 2000 ................ 29 Table 1.5.la. Estimated Wage Premiums for the Electric Power Industry Workers 36 Table 1.5.1b. Estimated Wage Premiums for the Unionized Electric Power Industry Workers .......................................................................................... 36 Table 1.5.lc. Estimated Wage Premiums for the Nonunion Electric Power Industry Workers .......................................................................................... 37 Table 1.6.1.2. Estimated Wage Premium Changes for the Unionized Electric Power Industry Workers for the States with High Electricity Prices But have Not Deregulated ........................................................................................................ 40 Table 1.7.1. Estimated Employment Changes in the Electric Power industry ............ 42 Table 1.7.2. Mean State Relative Employment for the Electric Power Workers: Deregulated vs. Un-deregulated Group ......................................................... 43 Table 1.7.3. Estimated Employment Changes in the Electric Power Industry. ........... 45 Table 1.8.1.2. Estimated Employment Changes for the Unionized Electric Power Industry Workers for the States with High Electricity Prices But Have Not Deregulated ......................................................................................................... 46 Table 2.5.1. Estimating Labor Demand Elasticities Using Model (2.5.l)--Random and Fixed Effects ....................................................................................... 70 Table 2.5.2. Estimating Labor Demand Elasticities Using Model (2.5.1)—GMM-DIF ..71 Table 2.6.1. Estimating Labor Demand Elasticities Using Model (2.6.4)——Random and Fixed Effects ........................................................................................ 74 Table 2.6.2. Estimating Labor Demand Elasticities Using Model (2.6.4)—GMM-DIF . 75 Table 2.7.1. Comparing Labor Demand Elasticities between Using State-level and Firm— level data ............................................................................................. 76 Table 2.7.2. Estimating Labor Demand Elasticities Using Model (2.7.1) .................. 79 viii LIST OF FIGURES Figure 2.4.1. Unionization Rates ............................................................... 64 Figure 2.4.2. Employment Trends ............................................................... 65 Figure 2.4.3. Wage Trends (Hourly) ............................................................ 66 INTRODUCTION—OVERVIEW OF THE ELECTRIC POWER INDUSTRY 1. The Industry Before the 1992 Deregulation The electric power industry can be divided into generation, transmission, and distribution sectors in terms of production functions. Among them, generation is the largest sector of electricity business. For a typical electric utility, generation constitutes about 65 percent of its total cost, while transmission and distribution make up 10 and 25 percent respectively (Crew and Kleindorder, 1986). Throughout the history of the electric power industry, these three sectors have been in large part vertically integrated, since a high degree of coordination and cooperation is required from generation of electricity to delivery of it to the consumers. Entry and rate-of-return regulation The electric power industry has long been characterized as a natural monopoly; therefore, government regulation is needed to protect consumer welfare and promote production efficiency. Entry and rate-of-retum regulation are the two primary forms of regulations. (Electric) utilities are the companies that the entry and rate-of-retum regulations are aimed at. There are four types of utilities—privately owned, federally owned, other publicly owned, and cooperatively owned. Among them, the privately owned utilities have been producing and selling most of the electricity in the United States.‘ Non- utilities are privately owned entities that produce power for their own use and /or for sale to utilities. In general, they are not subject to rate-of-retum regulation as the utilities. ‘ In 1995, the Investor-owned utilities made up around 75 percent of retail electricity sales and around 40 percent of the wholesale electricity sales (EIA, I997). Entry regulation of the electric power industry is implemented through monopoly franchise. The issuance of a monopoly franchise is the legal basis for detailed regulation. The authority of franchising usually rests with the state. Franchise is founded on the premise that the business in question is closely related to the public interest. Nonetheless, the application of franchising to the electric power industry also reflects the government’s principal belief of minimizing control over the private market (Phillips, 1993). In other words, the government could have chosen “publicly owned” as a dominant form for the industry; instead, the government chose a combination of private ownership and public control as a major form for the electric power industry, namely, the privately owned utilities. In fact, the combination of private ownership and public control for the US. electric power industry is unique among most other industrialized countries (Phillips, 1993). The rate-of-return regulation is to assure that the electricity prices (rates) are set in a way that allows the electric utilities to collect enough money to cover all operating expenses (taxes included), and, on the other hand, to have enough operating income left to provide a fair rate of return on the money invested in the business. A fair rate of return, in general, suggests covering the cost of capital. If the profit is too low/high for a fair rate of return, the utilities adjust their electricity rates upward/downward accordingly. How the electricity prices should be differentiated across the customers has always been an issue for both the regulators and the utilities. In fact, the electric utilities have a long history of performing embedded cost-of-service studies, of which the goal is to allocate the total costs among different classes of customers. Total costs are commonly divided into three types of costs, in which the important “factors” that determine the costs of supplying electricity are the load factor, the utilization factor, and the diversity factor. The first type is called demand costs, or capacity costs, which pay the fixed costs of the utility plant. They consist of the operating expenses that do not vary with production of electricity. The second type is called variable costs, which are incurred by providing a certain quantity of service. The third type is called customer costs, which cover the expense of billing, meter reading, accounting, and the capital costs associated with investments in meters and service connections, etc. The electricity demands by industrial and residential customers are very different in terms of load, utilization, and diversity factors. Hence, the costs they incur are also very different--whether it is demand costs, variable costs, or customer costs. However, regulatory agencies and the utilities do not necessarily agree with each other on how the costs should be assigned based on the load responsibilities. For instance, if a plant has to be built to meet the residential demand in a particular area, the regulatory agencies may be reluctant to accept the idea that all capacity costs should be born by the residential customers, since the industrial customers in that area also benefit from building the plant (Hyman, 1992). In addition, some regulators prefer a rate structure that has the low initial cost of service so that the average residential customers can afford the electricity service. Consequently, the loss of revenue from the low initial cost of service will have to be made up by higher rates to big electricity users such as the commercial or industrial customers (Hyman, 1992). Major legislative events Before 1992, the major legislative events that had far-reaching effects were the Public Utility Holding Company Act of 1935 (PUHCA) and the Public Utility Regulatory Policies Act of 1978 (PURPA). The PUHCA of 1935 was in response to the financial abuses and frauds conducted by the gigantic holding companies prevalent prior to and during the 19303. 2 Prior to the PUHCA, holding companies owned combinations of companies that were not necessarily related to the electricity business. By 1932, 16 holding companies accounted for 76.4 percent of the energy generated (Phillips, 1993). The size and complexity of the holding companies made it very difficult for the states to oversee and regulate their activities. Under the PUHCA, the Securities and Exchange Commission (SEC) was given the authority to break up holding companies if necessary. Also, holding companies were permitted to engage only in business that was appropriate for their utility operation. In other words, they were not allowed to participate in non-utility business. As a result, electric utilities eventually became more simplified integrated systems concentrating on electricity production (Energy Information Administration, 1998). The PURPA of 1978 was a reaction to the sharp rising electricity prices in the US. during the 19705 caused primarily by the oil embargo in 1973. In the early 19705, there was a change in public sentiment towards favoring deregulation, and economists who promoted competition in the electric power industry had contributed to this change 2 A holding company is an enterprise that owns sufficient stock in another company or in a number of companies so that it may influence the management of the company whose stock it holds. There are pure (Trebing, 1987). For instance, Weiss (1975) suggested that the generation sector could support extensive competition if the power plants were under different ownership and had equal access to transmission and distribution. Christensen and Greene (1976) argued that by 1970, most electricity generation was operated by the firms that had exhausted their scale economies. The PURPA was created to encourage competition as well as promote energy conservation. It required electric utilities to purchase power from any non-utility facility meeting the qualifying facility (QF) criteria (ownership, operating, and efficiency) at the utilities’ avoided costs of power production. The avoided cost of power production for a utility is “a rate which does not exceed the incremental cost to the electric utility of alternative electric energy. . .which the utility would generate or purchase from another source.”3 These QFS were not subject to the rate-of -return regulation; usually they were co-generators and small power producers using renewable energy as a primary source. Meanwhile, there were some other non-utility generators such as independent power producers that were not QFS and were unaffiliated with franchised utilities in the market, but they lacked significant market power. A decade of changes—the 19805 For the electric power industry, the 19805 was a decade of changes. First, the electricity prices went down substantially and continuously throughout the 19805, and sustained into the early 19905. Starting from 1983, the financial situation of the electric holding companies that are stockholding firm entirely and operates no properties directly, and operating holding companies that also operate some properties. power industry was Starting to look better than it did during the 19705. An early factor contributing to this was the inclusion in the rate base, by the FERC as well as some states, of a procedure to permit all or part of the construction work in progress (CWIP). This adjustment allowed an electric utility to earn a cash return on investments embedded in the construction work in progress, and thus alleviated the financial hardships brought about by construction programs awaiting completion (EIA, 1994). By the mid-19805, various factors started to contribute to the financial improvement of the electric power industry. They include lower fossil fuel prices, completion or cancellation of much of the ongoing construction activity, less need for new power plant construction due to a decline in the growth rate for electric power, etc. Since the huge construction costs would have reduced the utilities’ profit margins unless they could raise their electricity prices to make up the costs, fewer construction projects certainly would improve the utilities’ financial condition. As a result of the declining fuel prices and improving financial condition, the real price of electricity decreased significantly during the 19805. This is due to the fact that the electric utilities did not have to charge customers high electricity prices to cover their costs in that period. Secondly, due to the PURPA’s encouragement, the percentage of electricity output in the US. produced by non-utilities grew rapidly from 2.9 percent in 1980 to 7.7 percent in 1990 (Hyman, 1992). By the late 19805, the improving gas turbine generators and declining natural gas prices had started to drive down the electricity cost for the newest generating plants. The fact that these improved gas turbine generators were cheap and did not demonstrate substantial economies of scale had alleviated the cost disadvantage the small-scale power generators used to have (Hirsh, 1999). 4 In addition, the regulatory agencies were less willing to give guaranteed returns to utilities or bail those out that were in 3 See the Public Utility Regulatory Policies Act of 1978, Title II, Section 210. ‘ The new gas turbine generators had another desirable characteristic: they had thermal efficiency comparable to or higher than traditional fossil-fueled steam turbines. For example, the combined—cycle units had higher thermal efficiency than the fossil-fueled steam turbines. deep financial trouble. This lack of assurance of risk-free investment also spurred utilities to purchase power from the non-utilities rather than producing on their own. Most of the utilities went into financial trouble due to the failing (nuclear) power projects initiated during the 19705 and the extreme cost overruns. The famous cases of the financially troubled utilities include the Cincinnati G&E’s inability to complete its nuclear station, the default of the Washington Public Power Supply System, and the bankruptcy of the Public Service of New Hampshire. All of their financial trouble was primarily caused by failing to raise enough funds for the construction/completion of the nuclear power plants, partly because the costs for building a nuclear power plant had risen sharply after the Three Mile Island incident in 1979 (White, 1996). These cases showed that the rate-of-return regulation was not an assurance that the investments in the power industry were always risk- free, e.g., the regulatory agencies could put the utilities in a financially dire situation by not allowing the utilities to pass on the increased construction costs to the consumers. This lack of assurance of risk-free investment served as a strong disincentive to utilities’ capital investment. Utilities started to purchase power—often more than was required-- from other power generators rather than building their own generating capacity. During this period, there was less investment in all types of new generating plants by utilities. Also, many utilities abandoned their nuclear power generation projects initiated during the 19705. Thirdly, the fact that utility Stocks had much lower returns in 1990 than in 1980 suggests that investors had revalued their utility stocks (Hyman, 1992). One reason for this revaluation could be the lower interest rates throughout the 19805. Another reason could be that investors were optimistic about the future of this industry, and so were willing to accept lower current returns. Finally, the most disappointing aspect of the 19805 was the failure to improve efficiency in operation by utilities (Hyman, 1992). The cost of generation (other than fuel prices) and distribution for utilities had both gone up and faster than inflation. Hyman (1992) speculated that the electric utilities seemed to have exhausted their economies of scale by 1980. Also, the difficulties of setting up transmission lines in a more environmentally sensitive time might have also pushed up the costs. However, an analysis done by the EIA may shed some more light on this issue (EIA, 1998)’. According to the EIA report, between 1986 and 1990, the average expense (cents per Kilowatthour) on power purchases has trended upward, and the share of power purchase out of the “total electricity generated and received” has increased as well. On the other hand, the average expense (cents per Kilowatthour) on salary and wages has declined during the same period, and the average expense (cents per Kilowatthour) on “administration and general” has also decreased slightly at the same time. Some industry analysts believe that the fact that utilities were required by the PURPA to purchase electricity from the QFS at the utilities’ avoided cost of production was the reason for the escalating Operation expenses (EIA, 1998). II. The 1992 Deregulation and Afterwards To many, the falling electricity prices and the improving cost-effectiveness, primarily due to the independent power producers, during the 19805 seemed to suggest that competition encouraged by the regulators had worked in the generation sector. Hence it was becoming convincing that a large-scale entry and price deregulation in the generation sector would also work well. Stevenson and Penn (1995), White (1996), and Joskow (1997) all mentioned that the potential price differential between the bundled electricity prices charged by utilities and the unbundled prices that consumers could buy directly from the power suppliers was the major force that motivated the interested parties to push for restructuring. These parties included the alternative non-utility suppliers and the large industrials. Other forces such as the federal and state regulatory policies, the improved transmission and monitoring 5 See Chapter 9 Transitional Developments and Strategies: The Industry Prepares for Competition from The Changing Structure of the Electric Power Industry, 1998. technologies that have extended the feasible supplied area, and the reduction in size of generation facilities that have greatly reduced the entry cost, all have worked to attract more rent seekers clamoring for restructuring.6 Moreover, even some utilities sided with the restructuring. Because of the PURPA, many utilities have been saddled with long-term obligations to the overpriced electricity from the QFs. The avoided-cost prices were usually calculated from forecasted fossil fuel prices, which tended to be overestimated by the regulatory agencies (White, 1996). Plus, the regulatory agencies often imposed long-term obligations and limited flexibility to adjust prices in the standard contracts (White, 1996). Motivated by the commonly overestimated avoided cost prices, and increasing cost-attractive alternative generation technologies, some utilities have advocated restructuring for a better profit prospect.7 The Energy Policy Act (EPACT) of 1992 marked the beginning of deregulation in the generation sector at the wholesale level. The EPACT created a new category of generators called the Exempted Wholesale Generators (EWGs), which were not subject to the rate-of-return regulation and the QF standards. The EPACT also gave the Federal Energy Regulatory Commission (FERC) authority to order utilities to open their transmission lines to other generators. This authorization was crucial in creating competition since nearly all of the transmission lines were owned by the utilities. Although the EPACT left the issue of retail level competition to be decided by the states 6 One example of the federal regulatory policies leaning towards competition was that the Federal Energy Regulatory Commission (FERC) started to replace the cost of service standard with the market-based rates since the 19805. The example of the state regulatory policy was that a number of state public utility commissions had implemented bidding programs for power suppliers since the late 19805. 7 Although the QFs were often paid above-market rates, Hirsh (1999) however contended that there was increasing competition among QFs. First, in the mid-19805, some states including California, Colorado, Connecticut, Maine, Massachusetts, New York, and Wisconsin states began to experiment with competitive bidding programs. Second, the PURPA’s applying “pay for performance”, meaning QFs only earned money when they churned out power, also added to the competitive pressure among QFs. themselves, more than half of the states had started to engage in restructuring their retail- level generation market since 1992. States with high electricity rates such as the Northeast region and California were particularly aggressive in pursuing deregulation because of the sizable amount of potential rent for the independent power producers in and out of state (White, 1996). Since utilities could own the EWGs, they also shared some common interest with those independent power producers. Among them, California was the one that had a large-scale overhaul of the pre-existing regulatory system. The new set of rules included mandatory open access, vertical disintegration, centralized power exchange, utility retail price caps, and the Independent System Operator (ISO). California’s deregulation seemed to work reasonably well from 1998 to the spring of 2000. California ’s deregulation failure However, the subsequent price hikes, blackouts and brownouts demonstrated that the California style of deregulation could not handle economic shocks. Although is was widely recognized that California’s power crises were caused primarily by surging natural gas prices, unusually low hydroelectric generation reserves, and extremely high demands for electricity due to unusual temperatures during the year 2000 (Taylor, 2001; Kane, 2001), there was governance failure involved as well (Krapels, 2001). The criticism centered on the fact that the state required the three major electric utilities in the state to divest all of their power plants. As a result, the utilities started to purchase all of their 10 electricity from the Centralized Power Exchange in the day-before spot market after their divestiture. However, even though the market had become competitive since 1998, the retail prices were still capped by the state, and the utilities ended up losing money when the wholesale prices rose above the retail prices. Capping retail prices after deregulation is one of the major items in California’s restructuring bill passed in 1996. The purpose of controlling retail electricity prices was to prevent consumer exploitation but at the same time permit utilities to cover their costs and still have a fair rate of return (Kane, 2001). However, the capped electricity prices ended up failing to reflect the costs that the utilities had born from purchases in the wholesale market. In addition, some also argued that market manipulation was one of the reasons behind California’s power crises. This is owing to the fact that some of the price anomaly during the price hike period cannot be readily explained by the gas price hikes (Krapels, 2001). Plus, some of the power outages were suspected to be the results of generators purposely withholding electricity to drive up the electricity price.8 Electricity prices, however, have decreased in some of the Middle Atlantic States after they deregulated. In Pennsylvania alone, as of March 2001, consumers had saved $3 billion on their electricity bills since the utilities were deregulated in 1999 (Kahn, 2001). The argument was that utilities in this region could still retain their power plants, as well as Sign up long-term contracts with generating companies following deregulation, even though this region did not come across any suddenly heightened demand as California did. 11 Obstacles to deregulation There are some issues need to be dealt with for the electricity generation market to retain its competitive environment after deregulation. Stevenson and Penn (1995) noted that economics of scale might still be significant in the generation market after deregulation. According to annual load factors, 60 percent of the United States’ overall capacity should be baseload. Although additional baseload capacity will not be needed until 2005, given the average 7-9 year lead time to put the new baseload capacity to operation, most of the new baseload capacity should have been in the process of construction. Since constructing baseload units requires large capital investment, the barrier to entry seems to remain to a certain degree even after deregulation. Stevenson and Penn (1995) also noted that there had been increasing corporate consolidation and utilities dominating the independent power sector, which raised the concern of emerging horizontal integration. Joskow (1997) also discussed this issue in his study. Labor in the electric power industry The electric power industry is a capital- and fuel-intensive industry. Labor spending, on average, constitutes around 17 percent of the industry’s total expenses.9 However, with an average annual payroll of roughly 16 billion dollars, and with the " The FERC report also questions the pattern of California’s power outages (Krapels, 2001 ). 9 The number is calculated using data in the annual “Financial Statistics of Major U.S. Investor-Owned Electric Utilities,” published by the Energy Information Administration. 12 quality of service being much emphasized in the electric power industry, the importance of labor in the electricity production certainly can not be overlooked.10 In particular, since labor in the electric power industry has been highly unionized, the impact of deregulation on labor, especially unionized one, could be substantial and severe, as deregulation of an industry usually weakens its unionization after the market becomes competitive. In fact, there is ample evidence that workers in the electric power industry have been under tremendous pressure since deregulation started in 1992. On the firm side, there were state-mandated divestiture of generation assets of utilities, voluntary sell-off of generation plants by the utilities, significant amount of consolidating activities (Hamilton, 1996), and substantial increases of contract labor (Cavanaugh, 1994). On the union side, unions have been aggressive in demanding that their welfare be protected. In several states, unions have fought hard for worker protections through legislation (Schuler, 1997). Some even threatened to strike (New York Times, 2000). There have been very few studies looking at the labor market in the electric power industry. As deregulation unfolds, it becomes necessary to understand how workers in the electric power industry have been and will be affected by deregulation. Chapter One of this dissertation focuses on the impact of deregulation on labor earnings, as well as employment. Chapter Two studies unionization and the labor demand using the electric power industry as an illustration. The purpose of these studies is to not only understand the labor market in the electric power industry in particular, but also provide some '” The numbers are estimated based on the data from Employment, Hours. and Eamings U.S. 1909-1994, published by the Bureau of Labor Statistics, and by the Statistical Abstract of the United States, 1997. 13 empirical evidence regarding labor-rent sharing hypothesis and regarding how unionization affects labor demand. 14 Bibliography Cavanaugh, Herbert A. “15 Contract Labor a Temporary Trend?” Electrical World, September 1994. Crew, MA. and PR. Kleindorder. The Economics of Public Utility Regulation. 1986. First MIT Press edition. Christensen, R Laurits and William H.Greene, “Economies of Scale in U.S. Electric Power Generation,” Journal of Political Economy, 1976, Vol 84, no. 4, pt. 1 edited by Carl F. Christ, Stanford, Calif: Stanford Univ. Press, 1963. Energy Information Administration. “Transitional Developments and Strategies: The Industry Prepares for Competition.” (Chapter 9) in The Changing Structure of the Electric Power Industry. 1998. Energy Information Administration. “Federal Legislative Impact,” (Chapter 4) in The Changing Structure of the Electric Power Industry, 1998. Energy Information Administration. “Financial Impacts Of Nonutility Power Purchase on Investor-Owned Electric Utilities.” June 1994, U.S. Department of Energy, Washington, DC 20585. Freeman, Richard B. and Lawrence F. Katz. “Industrial Wage and Employment Determination in an Open Economy” in Immigration, trade and the labor market, edited by Abowd, John M. and Richard B. Freeman. 1991. The University of Chicago Press. Joskow, Paul “Restructuring, Competition and Regulatory Reform in the U.S. Electricity Sector,” Journal of Economic Perspectives, Vol 11, No.3, Summer 1997, pp.1 19-138. Hirsh, Richard. Power loss. 1999, MIT Press, Cambrigdge, Massachusetts, London, England. Hyman, Leonard S. America ’s Electric Utilities: Past, Present and Future, Public Utilities Reports, Inc., Arlington, Virginia, 1992. Kahn, Jeremy. “Where Deregulation Isn’t a Disaster.” Fortune. March 19, 2001, Vol. 143, 155. 6, pp. 40-44. Kane, Tim D. “Deregulation California Style,” USA Today. July 2001, Vol. 130, Issue 2674, pp. 16—19. Krapels, Edward N. “Was gas to blame? Exploring the Cause of California’s High Prices.” Public Utilities Fortnightly. Feb. 15, 2001, Vol. 139, 155. 4, pp. 28-36. Hamilton J, Michael “Measuring the Merger: Fact, Fiction.” Public Utilities Fortnightly, October 1, 1996. 15 New York Times. Strike is Authorized against 4 Utilities. Metro News Briefs: New York. May 18, 2000. Phillips , Charles F. Jr. The Regulation of Public Utilities. 1993. Public Utilities Reports, INC. Schuler Jr., Joseph F. “The Union Label: Electric Restructure. ” Public Utilities Fortnightly, September 15, 1997 pp. 20-27. Stevenson, Rodney E. and David W. Penn. “Discretionary evolution: restructuring the electric utility industry.” Land Economics. August 1995, 71(3): 354-367. Taylor, Jerry. “Did deregulation kill California?” Ideas on Liberty. June 2001, Vol. 51, Iss. 6, pp. 45-50. Trebing, Harry M. “Regulation of Industry: An Institutionalist Approach ” Journal of Economic Issues. Vol. 21, N04, Dec 1987 White, Matthew W. Power Struggles: Explaining Deregulatory Reforms in Electricity Markets. Brookings Papers on Economic Activity. Microeconomics. 1996, pp. 201-50. Weiss, Leonard W. “Antitrust in the Electric Power Industry.” in Promoting Competition in Regulated Markets. Edited by Phillips, Almarin. 1975. The Brookings Institution/ Washington, DC. 16 CHAPTER 1 THT IMPACT OF DEREGULATION ON LABOR EARNINGS IN THE ELECTRIC POWER INDUSTRY 1. Introduction Economic theory and evidence suggest that union workers are likely to have relatively higher wages than nonunion workers (Lawrence and Lawrence, 1985). Further, the higher the percentage of unionized workers in a given industry, the higher its union workers’ wages tend to be (Freeman and Medoff, 1981). The rationale is that unions reduce the opportunity for substituting nonunion for union products, and thus lower the elasticity of demand for unionized workers and the potential loss of employment for a given wage increase. A regulated environment is particularly favorable to unions. The entry barrier reduces the number of firms on the market and thus lowers the cost of organizing employees. The rate regulation allows industries to pass on costs to consumers and thus serves as a disincentive for management to resist union wage demands. However, once regulation is removed from the industry, it is also true that union workers are the most likely to suffer as a result of wages and/or employment 1055 (Rose, 1987; Peoples, 1998). Like workers in other previously regulated industries, workers in the electric power industry were highly unionized prior to deregulation.ll During the decade before deregulation took place in 1992, on average, close to 40 percent of the workers in the electric power industry were union members, while only about 17 percent of the workers in the private sector were union members (See Appendix l-A). The International ” Before deregulation, the union densities of the trucking, railroad, airlines, and telecommunications industries were all more than double of the union density of all other industries. For instance, the union 17 Brotherhood of Electrical Workers (IBEW) and the Utility Workers Union of America (UWUA) were the two major unions among others. Labor-management relations in the industry then could be described as relatively stable, since most contract negotiations had been resolved without strikes (Johnson, 1996). In 1991, the electric power industry employed 447,700 workers. These workers made up about 0.4 percent of the total non-farm workforce for the same year, and had an annual payroll of roughly 16 billion dollars, which amounted to 0.3 percent of the 1991 GDP. ‘2 The employment of electric power workers was distributed fairly evenly across the states. In most states, the electric power industry accounted for between 0.25 and 0.5 percent of the state total employment (McDermott, 1999). As the electric power industry started deregulating, its union density and employment fell sharply within just a few years. The union density fell from 38 percent in 1991 to 31 percent in 1996, while union density of other industries fell only 1 percent during the same period (See 1- A). Employment in the electric power industry dropped 24 percent between 1991 and 1997, whereas employment in the total non-farm sector rose 18 percent in the meantime. The substantial declines in the unionization rate and employment could be related to the restructuring activities engaged by the electric power companies since deregulation. They included divestiture of generation assets of utilities mandated by the states, voluntary sell-off of generation plants by the utilities, a significant amount of consolidation, and substantial increases of contract labor. In density of the airline industry in 1973 was 46 percent while that of all the other industries was 23 percent (Peoples, 1998). 18 reaction, unions have been aggressive as well in demanding that their welfare be protected. In several states, unions have fought hard for worker protections through legislation, especially with regard to the successor clauses (Schuler, 1997).13 Some even threatened to strike (New York Times, 2000). When there is a shock to the product market, normally employment responds in the same direction as the product demand does. Meanwhile, wages adjust to buffer the magnitude of job losses or gains (Freeman and Katz, 1991). The extent of wage adjustment is likely to depend on whether wages are set in a competitive labor market by supply and demand or set through collective bargaining. In a market where wages are set by collective bargaining, the above-market wages negotiated by the unions could create a practically perfect elasticity of labor supply (Freeman and Katz, 1991). Wages under this circumstance could respond to the increasing product market competition in different directions. On the one hand, wages can be unresponsive to the increasingly competitive product market. The usual explanation for the sticky wages is that the decisions of unions operating under a seniority system and a majority voting rule are influenced by the senior workers whose probability of being laid off is small. In other words, the median voters in unions may tend to be senior workers who prefer keeping wages high at the expense of employment. On the other hand, Freeman and Katz (1991) argue that union workers’ wages are likely to fall more than non—union workers’ as competition increases, since union workers’ wages often exceed the outside alternatives. Wages of nonunion workers in a partly unionized sector could change as a result of increasing market competition as well. First of all, there could be several factors affecting non-union wages as a result of union rents. Non-union wages could rise as a '2 The numbers are estimated based on the data from Employment, Hours, and Earnings U.S. 1909-1994, published by the Bureau of Labor Statistics, and by the Statistical Abstract of the United States, 1997. 19 consequence of union threat or of shifted demands towards non-union labor due to its lower relative cost. Non-union wages could also fall if there is less employment in the union sector, thus triggering an increase in the supply of non-union labor. However, Freeman and Medoff (1981) found little or no association between unionization and the wages of nonunion workers in the U.S. manufacturing industries. Likewise, Rose (1987) found insignificant union effects on the wages of nonunion workers in the once-regulated trucking industry. An industry in deregulation provides a natural setting to study how the increasing competition in a product market affects the response of wages that are mostly set through collective bargaining. It also provides an opportunity to examine the extent of rents shared by labor in the pre-deregulation regime. The rationale is that the increasing competition in the product market tends to weaken the unions and thus lead to losses of labor rents through the declines in wages or level of employment. Earlier studies analyzing the impact of deregulation on labor earnings show various degrees of wage response in the face of increasing product market competition. Rose (1987) estimated a more than 20 percent decrease in the wages of union workers over those of nonunion workers, and concluded that union workers in the trucking industry shared considerable rents with their employers. Card (1996) found an average 10 percent earnings drop among the airline industry workers a decade after the industry deregulated in the late 19705. Hendricks (1994) concluded that no significant wage decreases were discovered in '3 Successor clauses refer to the contract provisions that a new owner would have to honor current labor agreements. 20 the telephone, bus transportation, airline, and railroads industries. Peoples (1998), too, found only the trucking industry experienced a substantial wage decline, whereas the airline and telecommunications fell somewhat, and the railroads barely fell. Using earnings data from the Current Population Survey (CPS) from the period of 1992-1999, this study examines how wages in the electric power industry respond to increasing product market competition following state deregulation in the generation sector at the retail level. In particular, this study focuses on the difference of wage response between union and non-union workers. This study hypothesizes that we may witness, in the electric power industry, reductions of rents shared by union/nonunion workers with their employers-through wage losses-- as a result of deregulation. The unique aspect of this study is the focus on the wage effects of “state-level” deregulation. Because the state-level deregulation in the electric power industry happens at different points in time, we can control for the power industry-specific shocks by comparing the changes of wages in the electric power industry in the deregulated states to changes of wages in the electric power industry in the states that are not deregulating at that time. In other words, if there is a shock to the electric power industry nationwide while some states are deregulating, for example, a sudden increase in fuel prices across the country at the time some states are deregulating, we will be able to control for the effect of this shock on power workers’ wages by doing this comparison. In fact, since the wholesale-level deregulation in the power generation market (a nationwide deregulation) started in 1992, the electric power industry as a whole and its workers might be affected as a result. For example, there might have been a downward trend in wages of the electric power workers since 1992. Controlling for the industry-specific Shocks therefore makes perfect sense under this circumstance. Black and Strahan’s (1999) study provides the empirical evidence of the necessity for controlling for the industry-specific shocks. At first, they found that wages in the banking industry had been trending upward since the industry’s deregulation started, 21 controlling for worker age and education. However, after controlling for the banking industry-specific Shocks, workers’ wages were found to decrease by four to six percent after deregulation. Until recently, most studies on the wage effects of deregulation were only able to use “deregulation” that occurred at one point in time as a control variable (Hirsch and Mcpherson, 2000; Peoples, 1998; Talley and Schwarz-Miller, 1998; Card, 1996; Cremieux, 1996; Hendricks, 1994; Hirsch, 1988; Rose, 1987). Since this approach does not allow one to control for the industry-specific shocks and thus might lead one to over- or under- estimate the impact of deregulation (Black and Strahan, 1999). For instance, Another distinct aspect of this study is the estimation of the employment effects of deregulation, which is omitted in most previous studies looking at the impact of deregulation on wages. Freeman and Katz (1991) found a significant wage-employment trade-off in the labor market within the U.S. manufacturing sector. Peoples (1998), in his study, also pointed out a roughly inverse pattern between wage and employment changes in the trucking, railroad, airline, and telecommunications industries, but presented no more than a verbal description. This study conducts a statistical analysis of the employment effects of deregulation. By doing so, we have a more exact estimation of the impact of deregulation on employment. At the same time, this estimation may indirectly Show how unions in the electric power industry trade off wages against employment when facing deregulation. One thing noteworthy is that the CPS data are not sufficient to distinguish labor in the electricity generation sectors from that in the transmission and distribution sectors. Presumably it would be adequate to control for sectors, as the transmission and distribution sectors of the electric power industry still remain regulated. However, since transmission and distribution are usually owned by the utilities that are vertically integrated, it seems reasonable to assume that utilities as a whole are subject to the impact of deregulation, and the effect of omitting control for sectors might be negligible. 22 Section II of this study discusses regulation, deregulation and union rent sharing in the electric power industry. Unions have been subject to various sources of pressure since deregulation. With the electric power industry being more financially sound since the mid- 19805, unions could have shared possibly moderate rents with their employers prior to deregulation. Section III describes state deregulation. Although states have deregulated at fairly different paces, the states with higher electricity prices, especially in the Northeast and California, have been particularly active in restructuring their generation market. By May 2000, California, Massachusetts, Pennsylvania, and Rhode Island have started retail-level full-scale competition in the generation market. Section IV describes the data source. The annual Outgoing Rotation Group (ORG) files of a period 1992-1999 are used in this study. Section V estimates the wage effects of state deregulation using a differences-in- differences-in differences (DDD) approach. The wage premiums of union workers in the electric power industry over the wages of union workers in the private sector were found to drop 13 percent following deregulation, which is also statistically significant (See Table 1.5.1b). The wage premiums of nonunion workers in the electric power industry, however, remained unchanged irrespective of the comparison groups used. Section VI conducts sensitivity analyses on the wage effects of deregulation. This section checks whether the presence of high electricity prices in the deregulated states prior to deregulation has also contributed to the wage reductions. The finding suggests that high electricity prices may have contributed to the wage reductions occurring in those deregulated states. Section VII estimates the employment effects of state deregulation using an approach similar to the conditional logit model adopted in Neumark and Wascher’s (1995) analysis. Union employment in the electric power industry is found to decrease significantly by 52 percent after deregulation, while the nonunion employment in the electric power industry shows a 26 percent reduction (See Table 1.7.3), which does not 23 reach the significance level. Section VIII conducts sensitivity analyses on the employment effects of deregulation. Our finding suggests that reductions of employment level in the deregulated states were not related to high electricity prices in these states. Section IX provides an interpretation of the results obtained in the previous sections. Section X concludes with some policy implications based on the findings we have. 24 11. Regulation, Deregulation and Union Rent Sharing Starting from 1983, the financial situation of the electric power industry was starting to look better than it did during the 19705. The average return on utility stockholders’ equity in 1983 and 1984 exceeded the average for Fortune 500 companies. In 1984, the market-to-book ratios for the electric power industry increased from 0.6 during the 19705 to 1.0 (Sawhill and Silverrnan, 1985). The financial situation of the electric power industry has continued to improve throughout the 19805 (Energy Information Administration, 1994; Hyman, 1992).” An early factor contributing to this was the inclusion in the rate base, by the FERC as well as some states, of a procedure to permit all or part of the construction work in progress (CWIP). This adjustment allowed an electric utility to earn a cash return on investments embedded in the construction work in progress, and thus alleviated the financial hardships brought about by construction programs awaiting completion (EIA, 1994). By the mid-19805, various factors started to contribute to the financial improvement of the electric power industry. They include lower fossil fuel prices, completion or cancellation of much of the ongoing construction activity, less need for new power plant construction due to a decline in the growth rate for electric power, etc. Since the huge construction costs would have reduced the utilities’ profit margins unless they could raise their electricity prices to make up the costs, fewer construction projects certainly would improve the utilities’ financial condition. The financially robust electric power industry would very likely be conducive to unions’ rent sharing. For instance, there was evidence that the collective bargaining '4 The Energy Information Administration will be abbreviated as EIA. 25 agreements negotiated were usually above the all-industries median during the 19805 (Scott, Simpson and Oswald, 1993). Nevertheless, during the 19805, the utility regulatory agencies were becoming more reluctant to bail out utilities that were in financial trouble. In other words, utilities that failed to contain their costs would run the risk of bankruptcy. Subject to the pressure of cost containing, utility management might have more incentive to resist unions’ wage demands as well. Since deregulation took place in 1992, there has been a relatively significant increase in non-utility ownership of electricity supply. The number of nonutilities grew at 7 percent between 1992 and 1998 (from about 1,800 in 1992 to more than 1,900 in 1998), while the number of utilities decreased by eight percent (from 262 in 1992 to around 242 in 1998). The nonutility nameplate capacity grew by 73 percent during the same period, while the nameplate capacity of the utilities decreased by 5 percent. The nonutility additions to capacity increased at an average annual rate of 7 percent since 1992, while the utility additions to capacity on average decreased by about one-half percent. Between 1991 and 1999, electricity generation by nonutilities increased 84 percent, while generation by utilities increased merely 12 percent (EIA, 2000). As workers in the non-utility sector are less likely to be unionized, the increasing competition between the utility and non-utility sectors can have an adverse effect on unions’ rent sharing. One other source that would likely affect unions’ rent sharing was mergers and acquisitions (M&A). Since most previous labor agreements were drafted for a single bargaining unit within a vertically integrated utility, there was a strong incentive for the potential buyers who are not vertically integrated firms to challenge the adequacy of the labor agreement (Miller, 1998). Even more, M&A could threaten the status quo of unions as well. In reacting to the substantial increase in the number of M&A following deregulation (Diamonds and Edwards, 1997), some states have passed worker protection 26 legislation, which gives workers 24 to 30 months of guaranteed employment through the transition. However, the worker protection legislation alone cannot protect labor’s bargaining units. According to the National Labor Relations Act, unless 51 percent of existing employees are kept through a company-to-company transaction, the bargaining unit expires (Schuler, 1997). Another trend that could affect unions’ rent sharing following deregulation is the increasing adoption of contract labor. From meter reading to the business of power plant construction, electric power companies were gradually turning away from internal resources to the hiring of a floating workforce—temporary employees and outside contractors (Cavanaugh, 1994). Outsourcing can put a union’s bargaining power at risk if jobs are transferred from union to nonunion employers. Nevertheless, even if outsourcing simply transfers jobs from one union to another union employer, it still may have significant long-run deleterious effects on the union’s bargaining power, since it takes labor out of a sheltered market and puts it into a competitive one—union vs. union (Perry, 1997). In sum, unions have been subject to various sources of pressure since deregulation. With the electric power industry being more financially sound since the mid-19805, unions could have shared possibly from modest to moderate rents with their employers prior to deregulation. 27 III. Describing State Deregulation To model the impact of state deregulation on labor earnings, we first need to understand which states have deregulated and when they deregulated and the phase of deregulation they are in. Column 3 of Table 1.3.1 presents the states that started full- scale or phase-in competition at the retail level and records the years when this occurred. F ull-scale competition indicates that all or most consumers can purchase electricity from their preferred generation supplier; it does not necessarily suggest that the electricity prices charged to consumers are deregulated. For instance, in California, utility rates are still regulated by its Public Utility Commission after deregulation (Hirsh, 1999), while in Massachusetts and Pennsylvania, electricity prices are solely competitive after deregulation (Massachusetts Department of Telecommunications and Energy, 2001; Pennsylvania Public Utility Commission, 2001). Likewise, phase-in competition denotes that only a portion of the consumers can choose their electricity suppliers one at a time. California, Massachusetts, Pennsylvania, and Rhode Island started the retail-level full- scale competition during 1998 and 1999; New York, Montana, Michigan, and Arizona started the phase-in competition at the retail level during 1998 and 1999. Column 4 of Table 1.3.1 indicates the year that a state enacted restructuring bills, or the year that a state issued regulatory orders if legislative activities were absent in that state. This is to give a picture of where each state stands on the way towards deregulation. The primary reason for using the year that a state enacted restructuring bills as an indicator is that not all of the state utility commissions were empowered to manage deregulating the power market, or even if they were, the state legislature tended to step in and establish guidelines as to how the State should engage in the deregulatory activities (EIA, 1998; 7). 28 Table 1.3. 1. Status of State Electric Industry Restructuring by May 2000 N o . State( 2) FULL-SC ALE IPH ASE-IN Year when Average (1) COMPETITION (3) Restructuring bills electricity enacted/regulatory retail prices orders issued(4) rank" (5) 1 Rhode Island Full-scale competition at the 1996 1 retail level started in 1998 2 California Full-scale competition at the 1996 1 retail level started in 1998 3 Pennsylvania Full-scale competition at the 1996 1 retail level started in 1999 4 New York Phase-in competition at the 1996 1 retail level started in 1998 5 New Hampshire 1996 l 6 Vermont 1996 l 7 Massachusetts Full-scale competition at the 1997 1 retail level started in 1998 8 Montana Phase-in competition at the 1997 4 retail level started in 1998 9 Illinois 1997 1 10 Nevada 1997 3 11 Maine 1997 1 12 Mississippi 1997 2 13 Michigan Phase-in competition at the 1998 1 retail level started in 1999 14 Arizona Phase-in competition at the 1998 1 retail level started in 1999 15 DC. 1998 2 16 Oklahoma 1998 3 17 Connecticut 1998 l 18 Lirginia 1998 2 19 Ohio 1999 2 20 New Mexico 1999 1 21 Oregpn 1999 4 22 Maryland 1999 2 23 New Jersey 1999 l 24 Arkansas 1999 2 25 Texas 1999 2 26 Delaware 1999 2 27 West Viginia 2000 3 28 South Carolina 2000 3 29 Missouri 2000 2 29 Table 1.3.1. Continued NO- State(2) FULL-SCALE/PHASE-IN ye” “he“. . AVERAGE (1) COMPETITION (3) Restrwmmg “"5 ELECTRICI enacted/regulatory Ty RETAIL orders issued(4) PRICES RANK" (5) 30 Indiana N0 3 31 Kansas N0 2 32 Iowa N0 3 33 Wisconsin N0 3 34 Kentucky N0 4 35 Alabama N0 3 36 Tennessee N0 3 37 Idaho N0 4 38 Wyoming N0 4 39 Utah No 3 40 Colorado No 2 41 Washington No 4 42 Alaska No l 43 Hawaii No 1 44 Georgia N0 2 45 North Carolina N0 2 46 Florida NO I 47 Nebraska N0 3 48 North Dakota N0 3 49 Minnesota N0 3 50 South Dakota N0 2 51 Louisiana N0 2 Notes: * 1. Over 7.00 cents. 2. 6.00-6.99 cents. 3. 5.00-5.99 cents. 4. 0-4.99 cents. The electricity price for each state is derived from averaging 5 years (1990- 1995) of state electricity prices. Sources: Energy lnforrnation Administration: EIA-861, "Annual Electric Utility Report," and EIA-826, "Monthly Electric Utility Sales and Revenue Report with State Distributions." As of May 2000, there were 28 states (including D.C.) that had bills regarding restructuring enacted or regulatory orders regarding restructuring issued. Every year since 1996, there have been as many as seven states which began the deregulatory campaign. Column 5 of Table 1.3.1 shows each state’s rank with regard to electricity prices prior to 1996, a year when the state restructuring activities first began. The electricity price for each state is derived from averaging 5 years (1990-1995) of state electricity prices. The annual state electricity prices here represent the annual electric utility average revenue per kilowatthour for the states. Rank 1 has the price range of over 7.00 cents; 30 rank 2 ranges from 6.00 to 6.99 cents; rank 3 ranges from 5.00 to 5.99 cents; rank 4 ranges from 0 to 4.99 cents. As noted earlier, the potential price differential between the bundled electricity prices charged by utilities and the unbundled prices that consumers could buy directly from the power suppliers was the major factor that had contributed to deregulation across the states. Hence, not surprisingly, almost all of the states that already started full-scale/phase-in competition, except Montana, have their electricity price in the first rank. 31 IV. Data The primary data source for this study is the Current Population Survey (CPS) Outgoing Rotation Groups (ORG) files. Each monthly ORG file comprises a quarter sample of the monthly CPS and contains the information on earnings and union status. It also contains data relating to individual characteristics of subjects such as gender, age, education, marital status, occupation, and location (Specific Metropolitan Statistical Areas). This study uses the ORG files compiled by the National Bureau of Economic Research (N BER). The NBER’s ORG files are annual data compiled from the monthly samples. The data from 1992 through 1999 are included in this study. 32 V. Estimating Wage Effects of State Deregulation The first model (1.5.1) is a pooled log—earnings linear regression equation for estimating the effect of full-scale state deregulation on workers’ wages in the electric power industry. Note that it is indicated in section IH that as of May 2000, there were 4 states that have full-scale deregulation, which 3 of them were deregulated in 1998 and one in 1999. Since states that had phase-in deregulation began either from 1998 or 1999, they seem less comparable to the states that started full-scale deregulation in 1998 or 1999 (in terms of the extent of competitive pressure on the markets). They are thus not included as deregulated states. Data used in this model include full-time, private-firm workers in the electric power industry and the private sector from 1992 to 1999.15 LN WAGEijt = a0 + Xkit g1 + YEARI g2+ STATE). 531 + POWERi or, + DEREGULATIONjt or5 + DEREGULATION), * POWERi a6 + em (1.5.1) where: i indexes individual subjects; j indexes states; t indexes years; LNWAGEijt is the log of relative wages; Xkil is a vector of variables representing individual characteristics and geographical location: X“, is the highest grade of school attended; X2“ is the years of work experience, imputed from age-EDUC-6; X3“ is the square of X2"; X4it=l if residing in the Specific Metropolitan Statistical Areas (SMSA), 0 otherwise; X5“ is a vector of marital status dummy variables (MARITAL), MARITAL,=1 if married, 0 otherwise, MARITAL,=I if ‘5 A random sample of 30% of the data on the private sector was used in the estimation. The estimated parameters differ only to the second decimal place between using the 30% of the sample and the full sample. 33 widowed/divorced/separated, 0 otherwise; X6, is a vector of race dummy variables (RACE), RACE,=1 if African American, 0 otherwise; RACE2=1 if other minority, 0 otherwise; YEAR, is a set of year dummies, t=1992 to 1999; STATEj is a set of state dummies; POWER is a dummy equal 1 if electric power industry, 0 all other private sector; DEREGULATION], is a dummy equal 1 if an individual lives in a state when the full-scale generation competition at the retail level takes effect and all the years afterwards, 0 otherwise; DEREGULATION], * POWER, is an interaction identifies the effect of deregulation. e,,, is the residuals. Parameters underlined, such as _O_tl £2. and g1 indicate a vector of parameters. LN WAGEU, = [3,, + x,,,§,+ YEAR, §2+ STATEj 15, + POWER, [3, + STATEj * POWER, 1;, + YEAR, * POWER, 1;, + DEREGULATION, B. + DEREGULATION, * POWER, [3,, + e,,, (1.5.1)' Model (1.5.1)’ differs from (1.5.1) in that it includes STATEj * POWERi and YEAR, * POWER,. These two interactions allow state and year effects to vary between the electric power industry and the private sector. This is to control for state and year variations of wage premiums of the electric power workers. The drawback of this model is that the average wage premium prior to deregulation cannot be estimated directly, but needs to be derived by averaging across the estimated parameters for STATE], POWER,, and STATEj * POWERi in model (1.5.1)’, which are also denoted as coefficients gr. [2,,“ and [31in (1.5.1)’. 34 Both these models constitute a DDD estimation. The fixed effects YEAR, controls for the time-series changes in wages ([32 ); STATE, controls for the time-invariant characteristics of each state (Dr ); POWER, controls for the time-invariant characteristics of the electric power workers nationwide as a group ([34 ). The second-level interactions include STATE, * POWER,, YEAR, * POWER,, and DEREGULATION,,. STATE, * POWERi controls for the time-invariant characteristics of the electric power workers in each state ([351 ). YEAR, * POWER, controls for changes over time for the electric power workers nationwide (fl, ). DEREGULATION, controls for changes over time in the states that have started full-scale deregulation ([37 ). The third-level interaction POWER, *DEREGULATION,, catches all variation in wages specific to the electric power workers in the deregulated states in the years after deregulation started ([38 ). For the DDD estimator to be unbiased, it requires that there should be no contemporaneous shock that affects the relative earnings of the electric power workers in the deregulated states in the same year as deregulation (Gruber, 1994). Table 1.5.1a presents the results from estimating equations (1.5.1) and (1.5.1)’. Prior to deregulation, workers in the electric power industry have a premium of 35 percent above the wages of workers in the private sector. Deregulation leads to a 7 percent decline in the wage premium, which is not statistically significant. After controlling for state and year variations in the wage premium (using (l.5.1)’), the decline reduces to 5 percent, and is much less significant. Robust standard errors are only slightly larger than the unadjusted standard errors. Robust standard errors are used to adjust the effects of aggregate variables, YEAR, STATE,, and POWER,, on the micro units. According to Moulton (1990), they are usually larger than the unadjusted ones. This study creates a new variable, YEAR,*STATE,*POWER,, in the STATA program as a cluster variable for adjusting the effects of aggregate variables. 35 Table 1.5.la. Estimated Wage Premiums for the Electric Power Industry Workers (1 )Variables (2.1 Coefficient Std. E (p value) Robust Std E (p value) POWER (1.5.1) .348 .006(.000) .007(.000) POWER*DEREGULATION -.073 .047(. 1 18) .062(.026) (1.5.1) = POWER*DEREGULATION -.049 .053(.354) .048(.3l 1) (1.5.1 )’ Note: N: 154250. Column 2 of Table 1.5.1b shows the results from estimating the wage premiums of union workers in the electric power industry over the wages of workers in the private sector. Before deregulation, union workers have a wage premium of 40 percent above the wages of workers in the private sector. It drops 14 percent after deregulation, which is statistically significant at or: 10 (using robust standard errors). However, after controlling for state and year variations in the wage premium, the drop reduces to 9 percent and becomes insignificant. Table 1.5.1b. Estimated Wage Premiums for the Unionized Electric Power Industry Workers (1)Variables (2) union/total (3) union/union Co-efficient Std. E (p Robust Std Co- Std. E (p Robust value) E (p value) efficient value) Std E (p value) POWER (l .5.1) .403 .009(.000) 009(.000 .204 .009(.000) 0094.000 POWER“ -.l38 .062(.026) .073(.O60) .060(.l32) .064(.159) DEREGULATION 1.5.1 POWER“ .068(.065) .065(.053) DEREGULATION (1.5.1) Notes: 1. In column (2), N=152226; in column (3), N=25796. Column 3 of Table 1.5. lb shows the results from estimating the wage premiums of union workers in the electric power industry over the wages of union workers in the private sector. The reason to use wages of union workers in the private sector as a comparison is to 36 control for the general union trend. In other words, the change of wage premiums of union workers in the electric power industry might reflect more or less the general union trend, and thus the general union trend needs to be controlled for. After controlling for state and year variations in the wage premiums, the wage premium of the electric power workers shows a significant drop of 13 percent following deregulation (p=.065). The result remains significant using the robust standard errors (p=.053). Column 2 of Table 1.5.1c shows the results from estimating the wage premiums of nonunion workers in the electric power industry over the wages of workers in the private sector. N onunion workers in the electric power industry have a wage premium of 29 percent above the wages of workers in the private sector before deregulation. The change of wage premium after deregulation, however, is minute and not statistically significant. The result remains insignificant after controlling for state and year variations in the wage premium. Using the wages of nonunion workers in the private sector as a comparison shows similar results (column 3). In other words, our models show that wages of nonunion workers in the electric power industry have not been seriously affected by deregulation. Table 1.5.lc. Estimated Wage Premiums for the Nonunion Electric Power Industry Workers (1)Variables (2) nonunion/total H (3) nonunion/nonunion Coefficient Std. E (p Robust Std Co- Std. E (p Robust Std value) E (p value) efficient value) E (p value) POWER(1.5.1) .294 .009(.000) .010(.000) .338 .009(.OOO) .010(.000) POWER* -.006 .071(.933) .034(.859) -.009 .070(.888) .036(.783) DEREGULATION 1.5.1 _____ POWER" .015 .080(.855) .070(.834) .016 .078(.834) .073(.822) DEREGULATION (1.5.1)’ Note: In column (2), N=15223l; in column (3), N=l28454. As mentioned in section one “Introduction,” most previous research with regard to the impact of deregulation on labor earnings was only able to use “deregulation” that occurred at one point in time as a control variable. It is also mentioned that because of this 37 limitation (not able to compare the earnings difference in the states that have deregulated to that of those states that have not deregulated at that time), most previous research failed to control for the industry-specific shocks. Also because of these limitations, these studies used differences-in-differences (DD) estimations rather than DDD estimation. For instance, in Rose’s (1987) models, the second-level interactions are either “union (union vs. nonunion workers) *deregulation (before vs. after deregulation)” or “truck drivers (truck drivers vs. non-truck drivers) * deregulation”. The former captures variation in wages of the union workers (relative to nonunion) in the years after deregulation (relative to before deregulation), while the latter captures variation in wages of the truck drivers (relative to non-truck drivers) in the year after deregulation. In Card’s study, the second-level interaction is “airline (airline workers vs. non-airline) * deregulation (before vs. after deregulation). Likewise, in People’s analysis, the second-level interaction is “workers in the previously regulated industry (trucking, or airline, or railroad, or telecommunications vs. workers in other industries) * deregulation”. An early study by Hendricks (1977) on wage premiums for workers in the electric power industry even lacks the variable of “deregulation” like those used in the analysis of Rose’s, Card’s, etc., to control for the time-series changes in wages. Hendricks’ analysis compares some occupational workers’ (such as electricians, foremen etc.) earnings between the electric power industry and the manufacturing industries using a set of 1970 data. In other words, his model is a “differences” model. Although he found that labor in the electric power industry received insignificant regulatory rents, the finding, however, seems to be questionable on account of the failure to control for time-series changes in wages and the industry-specific shocks on wages. 38 VI. Sensitivity Analysis on the Wage Effects of Deregulation Can the presence of high electricity prices in the deregulated states prior to deregulation have contributed to the wage reductions? In section 111, this study reports that the states with high electricity prices tend to have a more rapid pace on deregulation. In fact, the four states that have deregulated in full scale all have high electricity prices. (See Table 1.3.1.) However, it is not clear how much the high electricity prices have contributed to the wage falls in those deregulated states. Since the states with high electricity prices have been pressured to engage in deregulation to reduce costs of producing electricity, the pressure to lower costs might have contributed to the decreases of workers’ wages in these states following deregulation. To explore this possibility, this study evaluates Whether those states with high electricity prices but have not deregulated (HP states in short) also experienced wage reductions as the states that had started full-scale deregulation did, using (5.1)’. We use year 1998, the year that most of the deregulated states started full-scale deregulation, as the cut-off point to test the wage effects of high electricity prices for these HP states. The implicit assumption is that these HP states may experience similar shocks as most of the deregulated states started full-scale deregulation. Table 1.6.1.2 shows the results of estimating wage changes for the union workers in these HP states using all the states with lower electricity prices (ranks 2-4 in Table 1.3.1) as the comparison group. The estimated wage premiums for the unionized electric power workers decrease by 8 percent, and the reduction is very close to significance at Ot=.10 (p=. 12). In other words, these HP states, although not being deregulated yet, seem to have experienced wage reductions similar to those in the states that started full-scale deregulation. This result suggests that more than deregulation itself, high electricity prices may have contributed to the wage reductions occurring in those deregulated states as well. 39 Table 1.6.1.2. Estimated Wage Premium Changes for the Unionized Electric Power Industry Workers for the States with High Electricity Prices But have Not Deregulated (1)Variab|es (2) union/union Coefficient Robust Std E (p value .051(.1 15) POWER" DEREGULATION ( l .5. 1)’ Notes: 1. In column (2), N=21324. .045(.078) 40 VII. Estimating employment effects of state deregulation In this section, this study estimates the employment effects of state deregulation. Note that although Section I already mentioned that total employment in the electric power industry dropped sharply between 1991 and 1997, yet we do not know specifically the employment changes in the states that have started full-scale deregulation. Equations (1.7. la) to (1.7.1c) evaluate the impact of full-scale deregulation on employment in the electric power industry. These equations were inspired by Neumark and Wascher ‘s (1995) conditional logit model using grouped state-year observations to estimate the minimum wage effects on employment and school enrollment. In a similar fashion, equations (1.7.1a) to (1.7.1c) are estimated with grouped state-year observations. They represent ratios of employment in the electric power industry to employment in the other private industries. Data are grouped using the years from 1992 to 1999, of full-time and part time, private-firm workers in the electric power industry and the other private industries. LN(ELECEMP,,/ PRIVEMP,,) = 7,0 + YEAR, yu+ STATE, er+ DEREGULATION, 7,, + e,,, (1.7.1a), LN (UELECEMP,,/ PRIVEMP,,) zyzo +YEAR, er+ STATE, y12+ DEREGULATION, 723 + e,,, (1.7. lb), LN(NUELECEMP,,/ PRIVEMP,,)='y30 +YEAR, Yrr +STATE, 112+ DEREGULATION, 733 + e,,[ (1.7.lc), ELECEMP, in (1.7. la) denotes the number of labor employed in the electric power industry. UELECEMP,, in (1.7. lb) denotes the number of union labor employed in the 41 electric power industry. NUELECEMP,, in (1.7.1c) represents the number of nonunion labor employed in the electric power industry. PRIVEMP. the denominators in all three .1" equations, represents the number of labor employed in all other private industries. YEAR,, STATE,, and DEREGULATION, are as defined in (1.5.1). Table 1.7.1 shows the results from estimating (1.7.1a) to (1.7. 1c). The relative employment in the electric power industry falls significantly by 37 percent after deregulation starts. The relative union employment in the electric power industry decreases 43 percent, while the relative nonunion employment decreases 27 percent. The former is significant at OI=.10 and the latter is not statistically significant. Table 1.7.1. Estimated Employment Changes in the Electric Power Industry. (l)Variables (2) 1.7.la (3) 1.7.1b (4) 1.7.1c union/total nonunion/total Coefficient P value Coefficient P value Coefficient (Std. P value (Std. E) (Std. E) E) DEREGULATION -.373(.154) .016 -.434(.251) .085 -.269(.186) l .149 Notes: In column 2, N=408; in column 3, N=379; in column 4, N=408. Since the 37 percent drop is quite large, we compare it to the change in the mean relative employment for the power workers in the states that have had full-scale deregulation in the years after their deregulation (called deregulated group) relative to the mean relative employment for the power workers in the states that have had full-scale deregulation in the years prior to their deregulation and in the states that have not deregulated at all (call un-deregulated group). Table 1.7.2 reports the relative employment in the deregulated and un-deregulated groups and the percentage difference between them. The mean (state) relative employment for the deregulated group is 52 percent less than the mean relative employment for the un-deregulated group. In other words, without 42 controlling for the state and year variations, the effect of deregulation on employment is 15 percent larger than that controlling for the state and year variation (52 percent vs. 37 percent). Table 1.7.2 Mean State Relative Employment for the Electric Power Workers: Deregulated vs. Un- deregulated Group. Un-deregulated Deregulated group Percentage F test on group (1) (2) difference significance between (1) and of between (2) group differences Mean state relative .0073 (.0049) .003 5(.0021) 52 PH, 406] employment for the =6.663; electric power workers P=.01 (to workers in the private sector) Note: N=408. Further, to explain what an average 37 percent drop annually of the relative employment in a state that has had full-scale deregulation after they deregulate compared to the relative employment in the states before they deregulated and in the states that have not deregulated at that time, we use the State of California as an example. In 1997, the year before California’s deregulation, the relative employment for the electric power workers is .0023. A 37 percent drop means that the relative employment for the electric power workers in 1998 and in 1999 would both be 0.0015. In 1998, total private sector employment in California is 11,430,000. Then the relative employment Of0.0015 would amount to the employment of 17,145 electric power workers in 1998. Compared to the employment of 21,978 electric power workers in 1997, calculated from .0023 (the relative employment in 1997)* 10,989,000 (the total privatesector employment in California in 43 1997), it is equivalent to a loss of 4,833 electric power workers in California the year right after deregulation. A couple of reasons might explain the large estimated employment loss after deregulation. As discussed in section II, outsourcing has been the increasing trend in the electric power industry. If a considerable portion of the contracting business went to workers in other industries, e.g., construction workers from the manufacturing sector, it would mean that employment in the electric power industry did not shrink as severe as it seemed. In addition, the increasing M&A could have caused employees previously classified under the electric power industry to be reclassified under other industries as the purchasers were not in the electric power industry. Moreover, the increasing share of the electricity market of nonutility power producers also suggests that more electric power workers could have been classified under other industries, since a considerable portion of the nonutilities is large industrial facilities that are not classified under the electric power industry, e.g., General Motors, K Mart (McDermott, 1999). Next, to control for the general union trend in employment, equations (1.7.3a) and (1.7.3b) use UPRIVEMPJ-t, union labor employed in all other private industries, and NUPRIVEMPJ-t, nonunion labor employed in all other private industries, as the denominators. LN(UELECEMP,,/ UPRIVEMP,,)= 1,, +YEAR, 1,,+ STATE, 1,, +DEREGULATION,, 1,, + e. Ijt (1.7.3a), 44 LN(NUELECEMP,,/ NUPRIVEMP,,)= I3O+YEAR, 111+ STATE, 3,, + DEREGULATION,, 133 + C,,, (1.7.3b), Column 2 of Table 1.7.3 includes the results from estimating (1.7.3a); column 3 shows the results from estimating (1.7.3b). After controlling for the general union trend, the relative union employment decreases significantly by 52 percent after deregulation, while the drop in the relative nonunion employment remains at 26 percent and insignificant. Table 1.7.3. Estimated Employment Changes in the Electric Power Industry. (1)Variables (2) 1.7.33 (3) 1.7.3b union/union nonunion/nonunion Coefficient (Std. P value Coefficient P value Errors) (Std. Errors) DEREGULATION -.524(.253) .039 -.262(.186) I .160 Notes: In column 2, N=379; in column 3, N=408. 45 VIII. Sensitivity analysis on the employment effects of deregulation Can the presence of high electricity prices in the deregulated states prior to deregulation have contributed to the reductions of employment level ? Here this study asks the question similar to the one asked in section VI--how much could the high electricity prices have contributed to the drop in employment level in those deregulated states. To explore this possibility, this study evaluates whether those states with high electricity prices but have not deregulated (HP states in short) also experienced employment reductions as the states that had started full-scale deregulation did, using (1.7.3a). Again, we use year 1998, the year that most of the deregulated states started full-scale deregulation, as the cut-off point to test the employment effects of high electricity prices for these HP states. Table 1.8.1.2 shows that there is no significant change in the estimated employment in these HP states after 1998. In other words, these HP states did not experience significant employment reductions at the time other states started full-scale deregulation. The results above suggest that the reductions of employment level in the deregulated states were not related to high electricity prices in these states. This finding is interesting because in section VI we have found that high electricity prices may have contributed to the wage reductions occurring in those deregulated states. Table 1.8. 1.2. Estimated Employment Changes for the Unionized Electric Power Industry Workers for the States with High Electricity Prices But Have Not Deregulated (1)Variables (3) 1.7.3a union/union Coefficient (Std. P value Errors) DEREGULATION .036(. 147) .808 Note: N=348. 46 IX. Interpretations of results This study finds that wage premiums for union workers in the electric power industry over the wages of union workers in the private sector decreased significantly by 13 percent following deregulation. Wage premiums of nonunion workers in the electric power industry, however, remained unchanged irrespective of the comparison groups used. Based on these results, unions apparently shared at least modest rents with their employers prior to deregulation. In addition, relative union employment in the electric power industry dropped significantly by 52 percent. The estimated decline in relative nonunion employment is 26 percent, which is shy of statistical significance. The sensitivity analyses in this study finds that high electricity prices may have contributed to the wage reductions occurring in those deregulated states, but that the reductions of employment level in the deregulated states were not related to high electricity prices in these states. In other words, deregulation at the state level has caused reductions on employment level much more than reductions in wages. These results suggest that we may not observe significant wage reductions in states with low electricity prices once they deregulate, since these states were not under pressure to cut wages as much as the states with high electricity prices were. However, deregulation seems to put pressure directly on the power companies to cut employment, regardless of the electricity prices. 47 X. Conclusions This study estimated the wage effects of deregulation using the data on state deregulation and a DDD approach to avoid biased estimation of the wage effects of deregulation owing to failure to control for industry-specific shocks. Further, this study estimated the employment effects of deregulation. The findings indicate that union workers in the electric power industry experienced a significant decline in their wage premiums after deregulation (13 percent), while wage premiums of the nonunion workers remained unchanged. Level of employment in the electric power industry exhibited a pattern similar to that of wages, in which the relative employment of union workers was substantially reduced following deregulation (37 percent) while the relative employment of nonunion workers did not Show a significant change. The sensitivity analyses find that high electricity prices may have contributed to the wage reductions occurring in those deregulated states, but that the reductions of employment level in the deregulated states were not related to high electricity prices in these states. In other words, we may not observe significant wage reductions in states with low electricity prices once they deregulate. However, deregulation seems to put pressure directly on the power companies to cut employment. The findings of this study are consistent with the labor rent sharing hypothesis, which states that labor in the regulated industry, especially union workers, is likely to share rents with their employers. However, we find that high electricity prices, instead of deregulation itself, might have contributed to wage reductions in the deregulated states. Based on the estimated 13 percent drop in union wage premiums, union workers in the electric power industry Shared, at least, modest rents with their employers before deregulation. Furthermore, in light of the dramatic reductions in union employment shown 48 in our results, unions in the electric power industry might have traded off the level of employment against wages. In other words, it is possible that the employment level might have dropped less severely if unions could have made more wage concessions, since we do not observe similar degrees of employment reductions in the nonunionized sector. Based on the evidence presented in this study, workers, the union workers in particular, in the electric power industry have endured substantial welfare losses. Usually an industry is deregulated because the policymakers believe that there will be considerable efficiency gains from deregulation, and that the consumers will be beneficiaries of these gains. In other words, it is accepted that the welfare of the regulated producers will be transferred to consumers, but the gains will be greater than the losses. The inflation-adjusted average retail price of electricity in the states that have started full-scale deregulation dropped more than one dollar following deregulation.l 6 This suggests that welfare transfer has indeed occurred as intended by the policymakers. Since this study has shown that unions shared regulatory rents with their employers prior to deregulation, it seems justifiable, from a policy perspective, that these union rents dissipated as the market became competitive. On the other hand, since the employment losses appear to be more severe than the wage losses for union workers in the electric power industry, the issue of fairness of the welfare redistribution resulting from deregulation should not be overlooked either. 1" The inflation-adjusted average retail price of electricity prior to deregulation for these states was $6.41, while it is $5.26 after their deregulation. These average prices are calculated based on the information in the I 990-2000 Annual Electric Utility Average Revenue per K ilowatthour for All Sectors by State, from EIA. 49 Bibliography American Gas, “Unions say restructuring Uncertainty Will Spur Utility Organizing Efforts.” February 1999. Volume 81, Issue 1, pp.7-8. Black, Sandra E. and Philip E. Strahan, “ The Division of Spoils: Rent-Sharing and Discrimination in a Regulated Industry,” European Labor Economic Conference, September 1999. Card, David. “Deregulation and Labor Earnings in the Airline Industry,” NBER Working Paper 5687, July 1996. Cavanaugh, Herbert A. “Is Contract Labor a Temporary Trend?” Electrical World, September 1994. Cremieux, Pierre-yves. “The Effect of Deregulation on Employee Earnings: Pilots, Flight Attendants, and Mechanics, 1959-1992,” Industrial and Labor Relations Review, Vol.49, No. 2 (January 1996). pp. 223-242. Diamonds, Joseph and John D. Edwards. “Mergers, Acquisitions, and Market Power in the Electric Power Industry.” April 1997. Energy Information Administration. Energy Information Administration. Net Generation, 1991 through 2000. 2000. Energy Information Administration, “Evolving Regulatory Reform: The Federal and State Role in Promoting Competition.” (Chapter 7) in The Changing Structure of the Electric Power Industry. 1998. Energy Information Administration. “Transitional Developments and Strategies: The Industry Prepares for Competition.” (Chapter 9) in The Changing Structure of the Electric Power Industry. 1998. Energy Information Administration. “Financial Impacts of Nonutility Power Purchase on Investor-Owned Electric Utilities.” June 1994. U.S. Department of Energy, Washington, DC 20585 . Freeman, B. Richard and Lawrence F. Katz. Industrial Wage and Employment Determination in an Open Economy, in Immigration, Trade, and the labor Market, edited by John M. Abowd and Richard B. Freeman. The University of Chicago Press, 1991. Freeman, Richard B.and James L. Medoff. “The Impact of the Percentage Organized on Union and Nonunion Wages.” Review of Economics and Statistics 63 (November 1981): 561-72. Grossman, Gene M. “International Competition and the Unionized Sector. ” Canadian Journal of Economics 17 (August 1984): 541-56. Gruber, Jonathan. “The Incidence of Mandated Maternity Benefits.” American Economic Review, June 1994: 622-41. 50 Hendricks, Wallace. “Deregulation and Labor Earnings,” Journal of Labor Research, Vol. XV, No 3, Summer 1994. Hendricks, Wallace. “Regulation and Labor Earnings”. Bell Journal of Economics 8 (Autumn 1977): 583-596. Hirsch, Barry T. and David A. Macpherson. “Earnings, Rents, and Competition in the Airline Labor Market,” Journal of Labor Economics, 2000, vol. 18, no. 1, pp. 125-155. Hirsh, Barry T. “Trucking Regulation, Unionization, and Labor Earnings: 1973-85.” The Journal of Human Resources, XXIII, Vol. 3, 1988, pp. 296-319. Hirsh, Richard. Power Loss. 1999, MIT Press, Cambrigdge, Massachusetts, London, England. Hyman, Leonard S. America ’s Electric Utilities: Past, Present and Future, Public Utilities Reports, Inc., Arlington, Virginia, 1992. Johnson, Ronald M. “Unions and Deregulation: Some Lessons for Utilities. " Public Utilities Fortnightly, September 1, 1996. McDermott, David. “Employment and other trends in the Electric Services Industry, ” Monthly Labor Review September 1999, pp.3-8. Miller, Christopher S. “Power Plant acquisitions: Workforce Management from the Buyer’s Perspective.” Public Utility Fortnightly, December 1998: 26-32. Moulton, Brent R. “An Illustration of a Pitfall in Estimating the Effects of Aggregate Variables on Micro Units.” The Review of Economics and Statistics, Volume 72, Issue 2 (May 1990): 334-38. Neumark, David and William Wascher. “Minimum Wage Effects on Employment and School Enrollment.” Journal of Business and Economic Statistics. April 1995, Volume 13, Issue 2, pp 199-206. New York Times. “Strike is Authorized against Four Utilities.” May 18, 2000. Peoples, James. “Deregulation and the Labor Market,” Journal of Economic Perspectives, Summer 1998, Vol. 12, No. 3, 111-130. Perry, Charles R. “ Outsourcing and Union Power.” Journal of Labor Research. Volume 18, Number 4, Fall 1997. Rose, Nancy. “Labor Rent-Sharing and Regulation: Evidence from the Trucking Industry,” Journal of Political Economy, Dec. 1987, 95, pp. 1146-8. Sawhill, John C. and Lester P. Silverman. “Transformed Utilities: More Power to You.” Harvard Business Review 63 (July/August 1985): 88-96. 51 Schuler Jr., Joseph F. “The Union Label: Electric Restructure, ” Public Utilities Fortnightly, September 15, 1997 pp. 20-27. Scott, Clyde, Jim Simpson, and Sharon Oswald. “An Empirical Analysis of Union Election Outcomes in the Electric Utility Industry,” Journal of Labor Research, Vol. XIV, number 3, surmner 1993. Talley, Wayne K. and Ann V. Schwarz-Miller. “Railroad Deregulation and Union Labor Earnings,” in Regulatory Reform and Labor Markets edited by James Peoples. Kluwer Academic Publishers, 1998. 52 Appendix l-A Table l.A.1. Union Densities: the Electric Power Industry vs. All other Industries Year Electric All other power industries ( %) industry ( %) 1983 44 19 1988 38 16 1991 38 15 1996 31 14 Sources: Union data books, 1983-1996. 53 CHAPTER 2 UNIONIZATION AND THE LABOR DEMAND: AN ANALYSIS OF THE ELECTRIC POWER INDUSTRY 1. Introduction A firm’s labor demand curve explains how elastic its labor demand is with respect to changes in input prices. The magnitude of the price elasticity of demand for any input depends primarily on the elasticity of substitution between the input in question and other inputs, the share of total cost of that input, and the price elasticity for the product or service being produced (Nicholson, 1992). It is believed that the presence of unions, however, reduces a frrm’s wage elasticity of labor demand, i.e. unions can raise wages without suffering a proportionate reduction in employment (Freemand and Medoff, 1981; Thorton, 1979). It is not uncommon for unions to affect employment levels indirectly as a result of bargaining over work rules, such as the minimum number of workers that are assigned to an operation (labor-capital ratio) or the work intensity. These practices have been prevalent in some previously regulated industries where the unions had strong bargaining power. For example, unions in the airline industry used to negotiate crew size on the airplane; or unions in the railroad industry bargained over having a fireman on diesel locomotives in freight and yard service. Nevertheless, although it is recognized that the wage elasticity of demand for labor tends to be negatively associated with unionization, the relationship has neither been explicitly modeled nor statistically tested. Maki and Meredith’s (1987) analysis, which finds a negative association between unionization and labor elasticity of substitution, is a relevant study in the sense that labor elasticity of substitution constitutes part of the wage elasticity of labor demand. Hsing and Mixon’s (1995) research, which finds that the wage elasticity of labor demand for the railroad industry has risen in absolute value following deregulation in the railroad industry, only indirectly supports the negative relationship between wage 54 elasticity and unionization, since they attribute the rise in wage elasticity after deregulation to the decline in union bargaining power in the railroad industry. This study models the relationship between unionization and labor demand, and tests the hypothesis that unionization is negatively associated with the wage elasticity of demand for labor, i.e., the higher the unionization rate is, the stronger the bargaining power is, and the lower the wage elasticity of labor demand should be. This study uses panel data from the electric power industry of a time span covering the years before and after the 1992 deregulation. The advantage of using data covering the deregulation period is that deregulation usually leads to substantial drops in unionization rates and thus allows us to evaluate how changes in unionization affect the wage elasticity of demand for labor. Section II constructs a theoretical model that characterizes the relationship between unionization and the labor demand. Section III describes the data sources of this study. The state-level data of the electric power industry for the period 1987-1996 is used to test our hypothesis. Section IV describes the trends of the key variables in this study, such as unionization rate, labor employment, and wages, from the early 19805 to 1996. In addition, these trends are compared to the trends in the private sector. Section V estimates the effect of unionization on the wage elasticity of demand for labor. After correcting the endogeneity of wages, the results suggest a statistically significant effect of unionization on the wage elasticity of labor demand. Section VI explores the possibility of a varying effect of unionization on the wage elasticity of labor demand over time due to some technological advances brought about by deregulation. However, the results do not support this possibility. Section VII checks the robustness of the findings in this study. Section VIII summarizes the conclusions of this study. 55 II. Constructing Theoretical Model To derive a conditional labor demand curve, this study employs a Cobb-Douglas function to characterize the production technology of the electric power industry. Two issues are noteworthy with regard to modeling the production technology of the electric power industry. First, it has been suggested that the substitution possibility in electricity generation is scant at the plant level (Komiya, 1962). However, significant substitutions have been found at the firm level (Christensen and Greene, 1976). Second, some studies found capital-energy complementarity in the manufacturing sector, but studies looking at the electric power generation sector have found significant substitutions between capital and fuel (Christensen and Greene, 1976; Atkinson and Halvorsen, 1984). The reason for this finding can be that, for instance, an increase in fuel prices may spur firms to invest in fuel-efficient machines. In light of the evidence indicating significant substitutions among inputs for electricity production, a Cobb-Douglas function should be adequate in characterizing the electric power industry’s production technology. Assuming that the representative utility firm has a Cobb-Douglas production function: Q=ALaKbF°, where L denotes labor, K denotes capital, and F denotes fuel, the following describes the assumptions about this production function. Assumption 1. A, a, b, c are all positive constants. b>a, and c>a, since the electric power industry is a capital- and fuel-intensive industry. 56 Assumption 2. a=¢, +0,u, b=t1>2 +02u, c=<1r3 +03u, where (1),, (1),, (I), are positive numbers, “u” denotes unionization rate, and 0,+ 02+03=O. Assuming that a, b, and c are linear functions of “u” indicates that unionization affects a firm’s production technology in terms of factor share and elasticity of factor substitution. This is not an implausible assumption since many view labor unions as a device for increasing labor’s share in the distribution of income of a firm (Rees, 1989). There is also research that finds a negative association between unionization and the elasticity of factor substitution (Maki and Meredith, 1987). This assumption also implies that unionization increases the marginal productivity of labor if 0, is positive. This is not implausible since there are studies finding that unions have a positive impact on the frrm’s productivity (Addison and Hirsch, 1989).17 The assumption that 9,-1- 02+03=0 indicates that a firm’s unionization rate affects its factor allocation, but not its scale of production, i.e., a + b + C: ¢,+¢, +63. The objective function of the representative utility firm is: The firm minimizes WL + rK+ pf F s.t. Q=ALaKbFC = AL¢I+GluK¢2+92uF¢3+93u, (2.2.1) where w is wage, r is rental price of capital, and p,- is price of fuel. 2 8Q 8L8“ = A61(L¢]+9]u-IK¢2+92uF¢3+63u)+ A(¢l + 61 u){L¢l+91u—1K¢2+62u (in + 93u)F¢3+93““ + [no + 62u>L¢I+91““K¢2+92“-' + ($1 + an —1)t¢t+9w-1 "K¢2+92“ IF¢3+93“ }. It will be positive if 6, is positive. In equation (2.2.2) below, we show that 61 determines the sign of the effect of unionization rate on the wage elasticity of demand. 57 It can be derived that the conditional labor demand is : —1 —(¢2+¢3+92u+93u) ¢2+92u —(¢2+¢3+62u+93u) ¢2+62u ¢3+63u ] L = A a+b+c.(_:) a+b+c Lg) a+b+c w a+b+c ,. a+b+c PWC We “1 -(¢2+¢3+92u+93u) ¢2+92u -(¢2+¢3+92U+93u) ¢2+92u =Am(¢;+93u) 4t+¢z+¢3 (:3+93u) + +3”, th+¢2+¢3 W: +311! 2+ 211 l ¢3+93u PWQ“ :5 :5 (2.2.2) (See Appendix 2 - A) 58 Take logs on both sides, and it becomes: LnL=————_1 LnA+—————_(¢2+¢3) Ln (”“63“ + $2 Ln ——¢3+93“ ¢I+¢2+¢3 ¢I+¢2+¢3 ¢I+91u ¢I+¢2+¢3 ¢2+92“ + 91 Ln ¢3+93u)u+ 92 [1,,(il’3r‘93uw+ -(¢2+¢3) an ¢I+¢2+¢3 ¢I+9lu ¢I+¢2+¢3 ¢2+92u ¢1+¢2+¢3 ) + 6‘ uan—El—Lnr+ —-l——uLnr+ __£’L_anf ¢I+¢2+¢3 ¢I+¢2+¢3 ¢I+¢2+¢3 ¢I+¢2+¢3 +——03—uLnPf + —l—LnQ (2.2.3) ¢I+¢2+¢3 ¢I+¢2+¢3 =ot0 +01, lnw+azulnw+a3lnr+a4ulnr+a51nPf +ot6ulnPf +017 lny, (2.2.4) where do = + ——(¢2 + ¢3) Ln(——¢3 + 63“) + ————¢2 Ln ——-¢3 + 03“) ¢I+¢2+¢3 ¢I+¢2+¢3 ¢I+91u ¢I+¢2+¢3 412+qu + 91 __an)u+—9—2——Ln——¢3+63u)u. ¢1+¢2+¢3 ¢I+91u ¢I+¢2+¢3 ¢2+92“ According to (2.2.3) and (2.2.4), the wage elasticity of labor demand, 81.nL =al+a2u= ’(¢2+¢3) + 9‘ u. dan ¢1+¢2+¢3 ¢1+¢2+¢3 a, <0. The sign of a2, the effect of unionization rate on the wage elasticity of demand, depends on the sign of 61. If 01 > 0, unionization rate has a positive effect on the wage elasticity of labor demand. In other words, the higher the unionization rate is, the less elastic the wage elasticity of labor demand will be. The labor demand elasticity with respect to rental price of capital, 01“” = a3 + a4u = 4’2 + 62 u. 3L!" ¢I+¢2+¢3 ¢I+¢2+¢3 0:3 > 0. 59 The Sign of (14, the effect of unionization rate on the capital price elasticity of labor demand, depends on the signs of 62. The labor demand elasticity with respect to fuel prices, 0]”an = a5 + aéu = 4’3 + 93 u. 81.nP ¢I+¢2+¢3 ¢1+¢2+¢3 as > 0. The sign of a6, the effect of unionization rate on the fuel price elasticity of labor demand, depends on the sign of 63. 8LnL 1 = a8 : > O. 3LnQ in + $2 + $3 The output elasticity of labor demand, 60 III. Data Sources This study uses the state-level data of the electric power industry for the period 1987-1996 to test our hypothesis. Ideally, we could study unionization and the labor demand using the firm-level data, but since union data at the firm level are usually not available for the public, and since unionization rate is the key variable in this study, we had to use the best available source. This study uses the union membership information provided in the Current Population Survey (CPS) Outgoing Rotation Group (ORG) files, aggregates them at the state level, and then computes the unionization rate for each state in each year. Consequently, data on other variables used in this study are therefore also grouped at the state level so that we have an internally consistent data set. Data on employment, wages, other factor prices, and output for the electric power industry are aggregated from the firm-level data. These firm-level data consist of a panel of 153 major privately owned utilities documented in the annual “Financial Statistics of Selected Electric Utilities” and the annual “Cost and Quality of Fuels for Electric Utility Plants” published by the Energy Information Administration (EIA) over the period 1987- 1996. '8 The EIA has been required to collect and publish the utilities’ financial data as a result of the Federal Power Act of 1935, which requires the Federal Energy Regulatory Commission (later the EIA) to collect financial information on the major privately owned electric utilities. '8 Major privately owned utilities are defined as those private utilities that have had, in the past three consecutive calendar years, sales or transmission services that exceeded one of the following: 1 million 61 During the data period, there were on average 260 privately owned utility firms in the United States, and usually about 180 of them were selected and their operational data were published in the “Financial Statistics of Selected Electric Utilities” by the EIA. Even though the utilities documented in the EIA’s publications were only a selected set, they accounted for more than 99 percent of the revenues from sales to ultimate consumers, and for around 96 percent of the revenues from sales for resale of all investor- owned electric utilities. In other words, the effect of leaving part of the utilities out of our data should be negligible. Since not every firm was included in every year, and since not every firm was included in both the annual “Financial Statistics of Selected Electric Utilities” and the annual “Cost and Quality of Fuels for Electric Utility Plants,” this study uses data from only 153 utility firms. How well might the state unionization rates computed according to the CPS ORG data match the utility data aggregated at the state level? In other words, how do we reconcile the fact that the state unionization rates computed based on the ORG files reflect the unionization rates in the whole electric power industry while the data on employment, wages, etc., reflect only what happened in the utility sector? First, the state unionization rate information based on the ORG files may underestimate the true unionization rates in the utility sector, since the utility sector has been highly unionized. However, since this study focuses primarily on the changes in the unionization rates, the problem of underestimation may be ignored. Second, do the megawatt-hours of annual sales for resale, 500 megawatt-hours of annual gross interchange out, or 500 megawatt-hours of wheeling for others (deliveries plus losses). 62 changes in the ORG unionization rates accurately reflect the changes in the unionization rates in the utility sector? Since deregulation affects the privately owned utilities the most among all types of power suppliers, changes in the ORG unionization rates may well reflect those in the utility sector. 63 IV. Data Description This section describes the trends of the key variables in this study, such as the unionization rate, labor employments and wages, from the early 19805 to 1996. In addition, these trends are compared to the trends in the private sector. Figure 2.4.1 shows the changes in the unionization rate during the period of 1983- 1996 based on three sources of data. The blue line denotes the unionization rates for the electric power industry, and is drawn based on the union densities calculated using the CPS ORG data. The red line denotes the unionization rates for the private sector (excluding the electric power industry), and is drawn also according to the CPS ORG data. Figure 2.4.1. Unionization Rates 50 40 percentage + eleCtriC power + private sector 10 0 '5 1'0 o rt. «3 ‘b ‘b ‘b 9 0.: ,9 .31 ,3.) ,3: ,3 year As we can see, the unionization rate for the electric power industry dropped substantially in the early 19805, stayed roughly flat afterwards, and gradually declined after 1992. The unionization rate for the private sector, however, declined more steadily during this period. Figure 2.4.2 shows the employment trends in the electric power industry, including the trend in electric utilities and the trend in the electric power industry as a whole, compared with that in the private sector, using the year of 1987 as the base year. Although the employment trends in the electric power industry had been different from that in the private sector prior to the deregulation in 1992, the trends started to diverge after 1992, with the electric power industry employment dropping while the private sector employment rising continuously. One thing noteworthy is that employment in the electric utility sector has fallen more rapidly than that in the electric power industry.'9 Figure 2.4.2 Employment Trends 20 15 :3. cumulative 1g lelectric utilities- percent _g ; this study change :‘1lg l electric power 3g [:1 private sector '9 The source of the macro data used for Figures 2.4.2 and 2.4.3 is the Employment, Hours . and Earnings: 1988-1996 published BLS. 65 Figure 2.4.3 shows the wage trends in the electric power industry, compared to those in the private sector. The hourly wages for the electric utility sector were derived using the yearly price of labor (in real dollars) divided by 52 weeks and then by the non- supervisory-worker average weekly hours.20 The yearly price of labor was computed using the yearly total salaries and wages paid and employee pensions and benefits in a utility, divided by the yearly total employment in that utility, and adjusted for inflation. It should be noted that the hourly wages for the electric power industry (from macro data) and for the private sector do not include employee pensions and benefits. Wages in the electric utility sector show more fluctuations and a more obvious upward movement than those in the electric power industry. This could have resulted partly from including employee pensions and benefits in our wage calculation for the utility sector. Figure 2.4.3. Wages Trends (Hourly) —o— electric utilities- this study —-— private sector- macro data electric power- macro data dollars in real D '1' b 6 % D W b: 6 ‘b ‘b 95 $ % 9 9 9 9 r9 '9 N9 '3 r9 r9 r9 r9 r9 year 2“ The source of data on the non-supervisory worker average weekly hours is Employment, Hours. and 66 V. Estimating the Effects of Unionization Equation (2.5.1) is a two-way error component model, where i denotes states, and t denotes years. It is the econometric model of the labor demand function based on model (2.2.4). LnL”: (10+ a,LnW,,+ or,uLnW,, + (13Lnr,.,+ a4uLan, + (XSLnPf,,+ aéuLnPf,, + 0t7LnQ,t +U, + A, + V,, (2.5.1) LnL ,, is the logarithm of the average employment (full-time plus one-half of the part-time employment) in electric utilities in state i and in year t. LnW,‘, is the logarithm of the average yearly price of labor in real value in state i and in year t. The yearly real price of labor for an electric utility is calculated using the sum of yearly total salaries and wages paid and employee pensions and benefits in a utility, divided by the yearly total employment in that utility, and adjusted for inflation. Lnr,., denotes the average cost of capital of firms in state i and in year t. This study uses “annual rate of return on common equity” as a proxy for the annual price of capital. Calculating the cost of capital requires knowledge of the detailed capital structure of each company, i.e., how the capital structure is proportioned among debt, equity and preferred stock, and the returns on equity and stock and the interests on debt. Tax issues also need to be factored into the calculation. For simplicity, this study uses only one element of the cost of capital as our indicator. This might lead to an under- or over-estimation of the cost of capital. Hence, caution is required in interpreting the results pertaining to this variable. LnPf ,, denotes the average price of fuel of firms in state i and in year t. The annual price of fuel of a firm is calculated based on the average of annual gas, petroleum, and coal prices that a firm pays, and each is weighted by its BTU percentage of the firm’s total fuel consumption. The following equation, (2.5.2), is the formula of it. Eamings: [988-1996 published BLS. 67 (Cents per million BTU of coal * BTU percentage of coal) + (cents per million BTU of gas * BTU percentage of gas) + (cents per million BTU of petroleum * BTU percentage of petroleum): average cents per million BTU: average price of fuel (2.5.2). LnQ ,, indicates the average output of firms in state i and in year t. A firm’s annual output is the total net energy it generates in megawatt-hours in a year. Ui captures state heterogeneity. It can be a state’s power production ecology—the distribution of mode of power production, such as steam, nuclear, or hydraulic. For instance, a state that relies heavily on nuclear power generation may tend to have higher factor prices, since construction of a nuclear power plant has become very expensive since the early 19805, It may also lead to a higher rate of labor employment (for monitoring) due to a high security risk (Electrical World, 1992). V ,, is a disturbance ~ III) (0, 01,). There are two issues noteworthy about model (5.1). First, LnW ,, can be correlated with V ,.,. The endogeneity of LnW ,, can come from two sources. In general, since wages are determined by the bargaining between unions and firms, and since unions care about employment as well as wages, LnW ,, (Addison and Hirsch, 1989), W ,, can thus be endogenous. In particular, wages (W ,, ) in this study is computed from the sum of total salaries and wages paid and employee pensions and benefits for a year, divided by the yearly total employment, L,',. The presence of L,, in the denominator can also lead to . LnP'. I.t’ I.t’ correlation of LnW ,_, with the residuals, V ,J. Second, Lnr are assumed and LnQ ,.,, strictly exogenous, and uncorrelated with V ,,. Normally, costs of capital are determined in the capital market, and fuel prices are determined in the fuel market. Output is basically determined by the demands for electricity. The fixed effects models and random effects models are used to estimate the equation (2.5.1). In addition, the general method of moments (GMM) first-differences estimator developed by Arellano and Bond (1991) is used to correct the possible 68 endogeneity of LnW”. The advantage of using this estimator is the efficient use of the number of instruments for the endogenous explanatory variables (Konings and Roodhooft, 1997). Arellano and Bond (1991) used the GMM technique for estimating the dynamic panel data models with endogenous variables. However, this technique can be applied to the static panel data models with endogenous variables as well (Konings and Roodhooft, 1997). (See Appendix 2-B for details of Arellano and Bond’s (1991) GMM estimator.) 69 Table 2.5.1 presents the results from estimating equation (2.5.1). The Hausman specification test suggests that the GLS estimator for the random effects model is biased and inconsistent. In other words, the fixed effects model fits the data better than the random effects model. In column 2, the effect of unionization rate on the wage elasticity of labor demand (coefficient of U,,* LnW“) is .08 and not statistically significant. The only coefficient that reaches the significance level is the coefficient for the wage elasticity of labor demand (LnW,t ), which is -.87. Table 2.5.1. Estimating Labor Demand Elasticities Using Model (2.5.1)-Random and Fixed Effects Variable (1)Random effects (2)Fixed effects Coefficient (std P value Coefficient (std P value E) E) LnW,, -.827(.087) .000 -.87l(.084) .000 U,,* LnW,t .087(.055) .117 .076(.053) .153 Lnr“ .014(.O66) .829 .Ol3(.063) .840 U,_, *Lnr,, -.321(.230) .163 -.286(.219) .193 LnP’ ,, .149(. 125) .234 .016(. 125) .900 U,, *LnP‘ ,, -.283(.212) .181 -.173(.204) .396 LnQ ,, .187(.031) .000 .038(.O38) .309 specification test Notes: 1. N=412. 2. All of the models in table 2.5.1 include separate year dummies. Table 2.5.2 reports the results of regressing (2.5. 1) using the GMM first differences technique. Both the AR(l) and AR(2) tests do not reject the hypothesis that the 70 GMM estimates are consistent.21 After treating LnW,, and U,,* LnW,, as endogenous, the effect of the unionization rate on the wage elasticity of labor demand becomes significant at 01:. 10. An increase in the unionization rate by 1 unit is accompanied by a decrease in the wage elasticity by 0.06 percent in absolute value (less elastic). That is, every one percent increase in the unionization rate is accompanied by a .0006 percent decrease in the wage elasticity. The wage elasticity of demand for labor equals -.75+.06* U,,. For example, the average union density of the electric power industry in 1991 was 38 percent(U,,=.38), then the wage elasticity equaled -.727. In 1996, the average union density of the electric power industry was 31 percent (U,,=.31), and the wage elasticity becomes -.731. Table 2.5.2. Estimating Labor Demand Elasticities Using Model (2.5.1)—GMM-DIF Variable (l)GMM-DIF Coefficient (std P value E) LnW,,. -.754(.l65) .000 U.,,* LnW,, .059(.036) .097 Lnr“ .033( .030) .279 U,,,*Lnr,,t -.226(.l44) .118 LnP’ ,, .20 I (.084) .018 U,,*LnP' ,_, -. I48(. 140) .291 LnQ o, .017(. 02_7)_ .537 _ fl- N(0, l)=-. 893 Notes: 1.N=368. 2. The GMM-DIF estimation includes separate year dummies. 3. The instruments for the first difference LnW, tand Ui ,* LnW, ,are the second and all further lagged variables of LnW, and Ui ,* LnW,,, i. e., LnW,,_ _, aiid U, ,s* LnW,,s, s=2,.. ..9 Note that the variable In the first- difference equation for estimation is ALnW,,= LnW,,- -LnW,, ,, hence the valid instruments starting from the second lag rather than the first lag of LnW”. The same principle applies to A(U,n,* LnW,,). 2‘ The insignificance of AR( 1) and AR (2) tests suggests that the errors in this model in levels are not serially correlated. These tests that failed to reject the hypothesis that the errors in levels or in differences are serially correlated assure the consistency of the GMM estimator. 71 VI. Accounting for the Effects of Technological Changes This section explores the possibility of a varying effect of unionization on the wage elasticity of labor demand over time due to some technological advances brought about by deregulation. Deregulation of an industry may be accompanied by significant technological changes owing to the increasingly intensified competition (Card, 1996). As technological changes often affect the existing production function of a firm, the union bargaining power of the employee is also likely to be affected, given the same level of unionization. A firm may be less constrained by its labor and therefore has more bargaining power as a result of a relevant technological improvement. For example, following deregulation in the electric power industry, some electric power companies started adopting the newly developed automated meter reading system rather than using the traditional labor- intensive meter reading operations (Pollom, 1999). This new technology could have a strong negative influence on union bargaining power, since meter readers had become more substitutable by computers. Thus, the effect of unionization on wage elasticity of labor demand can change as a result of technological advances. In a technical sense, 01 in equation (2.2.1) that reflects the effect of unionization on a utility’s production function may vary over time based on some technology reason. To adjust our model for the possible effects of technological changes on the unionization effects, the following specification allows 0, in equation (2.2.1) to change over time. For the purpose of this study, we treat 02 and 03 as constant, since they are not the parameters of interest. 72 The production function in this case is defined as : Q = ALaK ch = AL¢' +91't'luK‘I2 +We?3 +93", wheretij =1ifi = j,0ifi a j;i,j = 1, ..... ,T. (2.6.1) The log conditional labor demand curve can be derived as the following : LnL: ‘1 LnA+ -(¢2 W3) Ln( ¢3+93u 3+ 4’2 Ln(¢___)3+63u + ¢I+¢2+¢3 ¢I+¢2 +¢3 ¢I+9nliju ¢I+¢2+¢3 ¢2+92“ 9“”7 Ln (p3 +93u )u+ 92 Ln ¢3___)+ 93” u+ _(¢2+¢3an ¢I+¢2+¢3 ¢I+91,-’iju ¢I+¢2 +413 ¢2+92u ¢1+¢2 +¢3 9 't" 1' y “an+ ¢2 Lnr+ 92 uLnr + $3 JJIPf ¢I+¢2+¢3 ¢I+¢2+¢3 ¢I+¢2 +¢3 ¢1+¢2+¢3 + 93 uLnPf + 1 LnQ (2.6.2) ¢I+¢2+¢3 ¢I+¢2+¢3 = a0 +ot, lnw+a2itijulnw+a3 lnr +a4ulnr+a51nPf + a6uln Pf + (17 In y (2.6.3) Econometrically, equation (2.6.3) can be written as a variation of the two-way error component model (2.5. 1): LnLi,l = 0‘0"' allT88 + 0‘1sz + (1,3T90+ al4T9l + O‘15T92 'i' 0‘16T93 + 0‘I7T94‘i' 01,,,T95 +OI,,T96 + OtanW,l + a3u,,LnW,, + or,,u,,Ln W,,T8,, + auuan W,,T,9 + a43u,,Ln W,,T90 + (x44u,,Ln W,.,T91 + a45u,,Ln W,.,T92 + 01,6u,,Ln W,.,T93 + anuan W,', T94 + 014,,u,,Ln W,t T95 + 0149u,,Ln W,, T9,, + OtsLnr,, + a6u,,Lnr,, + a7LnPf ,_, + 018% LnPf,,+ 0.9LIIQ ,, +U, + V ,, (2.6.4), where ngwm T9,, are year dummies. 014, to or” measure the effect of unionization rate on wage elasticity of labor demand by years. By allowing the effect to vary by years, we allow it to be influenced by technology changes. Table 2.6.1 shows the results from estimating model (2.6.1), using fixed effects and random effects models. The Hausman specification test suggests that the fixed effects model is more adequate than the random effects model. In column 2, none of the 73 yearly effects of the unionization rate on the wage elasticity of labor demand is significantly different from zero. However, the coefficients for 1994 and 1996 are positive and close to the significance level at OI=.10. They suggest that the effect of the unionization rate on the wage elasticity of labor demand could have increased after deregulation. In other words, changing the unionization rate could have a larger effect on the wage elasticity of labor demand after deregulation. Table 2.6.1. Estirnatin Labor Demand Elasticities Usinj Model (2.6.4)——Random and Fixed Effects Variable (l)Random effects (2)Fixed effects Coefficient (std P value Coefficient (std E) P value E) LnW,, -.827(.088) .000 -.87l(.084) .000 U,,* LnW” .038(.065) .557 .044(.063) .482 U,,* LnW,,* T88 .020(.022) .376 .017(.021) .433 U,,,* LnW,,* T89 .011(.023) .632 .009(.022) .702 U,,* LnW,,* T9,, .027(.023) .241 022(022) .320 U,_,* LnW,,* T9, .024(.023) .295 .020(.022) .364 U,,,* LnW,,* T92 .027(.024) .260 022(023) .359 U,,* LnW,,* T9, 035(024) .152 .028(.023) .231 U,,* LnW,,* T9, .046(.025) .065 .039(.024) .109 U,,* LnW,,* T,5 .004(.025) .857 -.005(.024) .831 U,,* LnW,,* T96 .047(.024) .048 .037(.023) .111 Lnr,, .002(.069) .980 .008(.066) .904 U,t *Lnr” -.242(.244) .321 -.246(.234) .296 LnP',, .059(.137) .667 -.036(.137) .794 U,, *LnP‘,, -.094(.256) .713 -.O65(.248) .794 LnQ,, .182(.032) .000 .045(.O38) .245 Hausman x2(25)=246.91 .000 specification test Notes: 1. N=412. 2. All of the models in table 6-1 include separate year dummies. 74 Nevertheless, as Table 2.6.2 shows, after treating LnW,J and Um“ LnWL, as endogenous, none of the yearly effects of the unionization rate on the wage elasticity of labor demand is statistically significant. Table 2.6.2. Estimating Labor Demand Elasticities Using Model (2.6.4)-—GMM-DIF Variable ( l )GMM-DIF Coefficient (std E) P value an,, -.779(.185) .000 U,,,* LnW,_, .078(.063) .216 Un* LnW...* -.014(.009) .132 T89 Utt“ LnWi,t* -.003(.009) .760 T90 U1.t* ann’“ -.003(.010) .757 T91 Un* Lan* -.006(.013) .614 T92 Un“ LnW.,.* 0041.011) .758 T93 U...* Lan* 0141.013) .252 T94 ULS" LnW,.,* -.032(.044) .469 T95 Uni‘ an,,* .008(.014) .575 T96 Lug, .052(.051) .314 U,, *Lnr,l -.279(.211) .187 LnP'... .196(.097) .045 U,t *LnP’ ,, -23 l (.200) .249 LnQ 1.1 .014(.023) .537 AR(2) test N(0,l)=-.724 .469 Notes: 1. N=368. 2. Separate year dummies are included in the model. 3. The year dummy starts with year 1989 because of the first differences and the lagged LnW” used as instruments for Law”. 75 VII. Robustness Checks State-level data vs. firm-level data Koonings and Roodhooft (1997) found a long-run wage elasticity of well above 1 in absolute value and noted that the estimated wage elasticities from microeconomic studies were higher than those from macroeconomic studies. They suggested a number of reasons for these findings. The first reason is that their data had a large cross-section dimension relative to the time dimension and thus better represented the long-run wage elasticities. Another reason is that the firm-level data capture the individual firm’s responses to shocks while the macro-level data do not. Since this study uses the state- level data aggregated from the firm-level data, the estimated results might be downwardly biased. Table 2.7.1 compares the estimates from using the state-level data with those from using the firm-level data.(The former data set is an aggregation of the latter.) The model estimated here is similar to model (2.5.1), but with the interactions of unionization rate and factor elasticities omitted. Table 2.7.1. Comparing Labor Demand Elasticities between Using State-level and Firm-level data Fixed effects models Variable (1)5tate-level data (2) firm-level data Coefficient (std P value Coefficient (std P value E) E) LnW,t -.861(.067) .000 -.560(.029) .000 Lnl',‘l -.06 l (.034) .077 -.Ol9(.014) .182 LnP' ,, -.076(.098) .437 -.199(.059) .001 LnQ ,, .031(.033) .41 l .005(.010) .649 Notes:l.In column 1, N=412; in column2, N=1056. 2. Separate year dummies are included in both models. 76 The results in Table 2.7.1 show that the wage elasticity of labor demand estimated for the firm level is actually smaller in absolute value than that estimated for the state level. However, the fuel price elasticity of labor demand estimated for the firm level is much larger than that estimated for the state level. These results suggest that the interpretations in Koonings and Roodhooft ‘s(1997) study do not apply to ours. It could be that the within-state variation in employment responses to wages, on average, is smaller than the between-state variation in employment responses to wages. Hence after aggregation, the employment responses to a wage change actually become larger. 77 Restriction violation Next we check whether our results violate the primary assumption 01+ 02+03=0 indicated in Section II. In section II equation (2.2.3), we derive the following: LnL: ‘1 InA+ ’(¢2+¢3)rn¢3+03“o+ “2 , ¢3+93u ¢1+¢2 +A ‘ ¢1+¢2 +A ‘ A +91u’ A +152 +A ”(A +92ui + 91 I. $3 +03u),,+ 02 Ln ¢3+63u u+ —(¢2 +¢3) an ¢I+¢2+¢3 ¢I+91u ¢I+¢2 +¢3 412+qu ¢I+¢2+¢3 + 61 uan 02 Lnr + 92 uI.nr + 4’3 LnP f ¢I+¢2+¢3 ¢I+¢2 +453 ¢I+¢2+¢3 ¢I+¢2 +¢3 + 93 urnPf + 1 LnQ ¢I+¢2+¢3 ¢I+¢2+¢3 Using the restriction that 91 +02 +63 = 0, Substitute 92 = _(91 +63) ¢1+¢2 +¢3 4’1 +¢2+¢3 into (2.2.3). Collect terms, :>LnL= '1 LnA+ _(¢2+¢’3) Ln(¢3+03u)+ “2 Ln(¢3+63u ¢1+¢2+¢3 ¢I+¢2 +953 ¢1+HIu h+¢2+¢3 2+ 2” + 91 Ln(¢3 +93“ u, 92 in ¢3+93u u+ -(¢2 +¢3) an ¢I+¢2+¢3 1+1” ¢I+¢2+¢3 2+ 2“ ¢I+¢2+P3 + 91 (uan—uLnr) + 422 Lnr+ ¢3 LnPf ¢I+¢2+¢3 ¢1+¢2 +53 ¢1+¢2 +¢3 + 63 (uLnPf —uLnr) 91 +902 +¢3 + 1 LnQ 91 +452 +453 = 010 +01, lnw+a2(uln w—uLnr)+a3 lnr +065 In Pf +a6(uln Pf —uLnr)+ a7 lny, (2.7.1) 78 Table 2.7.2 reports the results from estimating equation (2.7.1) using a fixed effects model. Compared with the estimates shown in Table 2.5.1, the sizes of the estimated 012 and 016 here are quite different from those of the estimated 012 and 016 in Table 2.5.1, although neither of them are statistically significant. The cause of these differences could be either that the proxy for rm used in this study is a poor indicator of the cost of capital, or that there might be some mis-specifications in our production function. Table 2.7.2. Estimating Labor Demand Elasticities Using Model (2.7.1) Variable (l)Para- (2)Fixed effects (3)Fixed effects using model meters using model (2.7.1) (2.5.1) Coefficients(std E) Coefficient (std E) P value LnW,t or, -.843(.083) .000 U,',* (LnWm- Lnr,,) or2 .076(.053) .025(.042) .560 Lnr” OI3 -.038(.055) .494 LnP' ,, o, -.087(. 108) .419 U,, *(LnPr ,, -Lnr,,) ot6 -.173(.204) .064(.141) .650 LnQ ,, .032(.038) .391 Notes: 1. N=412. 2. Separate year dummies are included. 79 VIII. Conclusions This study estimated the union effects on the wage elasticity of demand for labor using the state-level data aggregated from the firm-level data of the electric power industry. One of the unique aspects of this study is the theoretical characterization of the relationship between unionization and the wage elasticity of labor demand. Another is the consideration of the impact of technological changes on the union effect on wage elasticity, as technological advances brought about by deregulation usually relax the employers’ production constraints and make labor a more substitutable factor. This study finds that the unionization rate has a small but statistically significant effect on the wage elasticity of demand for labor. According to our results, every one percent increase in the unionization rate is accompanied by a .0006 percent decrease in the wage elasticity (less elastic). However, this study does not find significant changes in the union effect on wage elasticity over the course of deregulation after taking into account the effect of technological advances. The finding of a significant union effect on wage elasticity of labor demand supports the argument that unions care about employment as well as wages. The robustness checks suggest some inconsistency between our theoretical assumption about the production function and the estimated results. This could either be a result of the proxy for the cost of capital used in this study being a poor indicator, or be a result of some mis- specifications in production function. The direction of the future studies should be to look for a better indicator for the cost of capital for the electric power industry, and to adopt different production functions in the theoretical models. 80 Bibliography Addison, John T. and Barry T. Hirsch. “Union Effects on Productivity, Profits, and Growth: Has the Long Run Arrived?” Journal of Labor Economics, Volume 7, No. 1, January 1989. Arellano, Manuel and Stephen Bond, “Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations, ” Review of Economic Studies (1991) 58, 277-97. Atkinson, Scott and Robert Halvorsen. 1984. “Parametric Efficiency Tests, Economies of Scale, and Input Demand in U.S. Electric Power Generation. ” International Economic Review, 25: 647-62. Card, David. “Deregulation and Labor Earnings in the Airline Industry,” NBER Working Paper 5687, July 1996. Christensen, R Laurits and William H.Greene, “Economies of Scale in U.S. Electric Power Generation,” Journal of Political Economy, 1976, Vol. 84, no. 4, pt. 1 edited by Carl F. Christ, Stanford, Calif: Stanford Univ. Press, 1963. Electrical World. “How Many People Does it Take to Run a Nuclear Power Plant?” July 1992, pp. 9-13. Freeman, Richard B. and James L. Medoff. “The Impact of the Percentage Organized on Union and Nonunion Wages.” Review of Economics and Statistics 63 (November 1981): 561-72. Hsing, Yu and Franklin G. Mixon, JR. “The Impact of Deregulation on Labor Demand in Class-I Railroads.” Journal of Labor Research, Volume XVI, Number 1, Winter 1995. Komiya, R. “Technical Progress and the Production Function in the United States Steam Power Industry,” Review of Economics and Statistics. 44, no. 2 (May 1962): 156-67. Konings, Jozef and Filip Roodhooft, “How Elastic is the Demand for Labor in Belgian Enterprises? Results from Firm Level Accounts Data, 1987-1994.” De Economist 145, No. 2, 1997 Maki, Dennis R. and Lindsay N. Meredith. “A Note on Unionization and the Elasticity of Substitution.” Canadian Journal of Economics, Volume 20, Issue 4 (November, 1987), 792-801. Nicholson, Walter. Microeconomic Theory. 1992, fifth edition. The Dryan Press. Pollom, Brian. “AMR: The Key to Enhanced Operations.” Transmission & Distribution World, September 1999, Vol. 51,155. 11 pp. 26-32. Thornton, Robert J. “The Elasticity of Demand for Public School Teachers. ” Industrial Relations 18 (Winter 1979): 86-91. 81 Rees, Albert. The Economics of Trade Unions. Third Edition. 1989. The University Of Chicago Press. 82 Appendix 2-A. The firm minimizes wL + rK + pf F (1) st. Q = ALaKbFC (2) I F = (QA’1L_“K—b)c (3), plug into (1) 1 => Min wL +rK +Pf(QA"L'“K"’)c (4) Take derivative of (4) with respect to L, l —a—c :> w-3 Pf(QA"K"’)c L c :0 (5) C -a—c l :1 :>L c =w3P‘_ (QA“K"’)c (6) a C C I __1 —a—c =>L= w—Pf— (QA'1K_b) C (7) a Take derivative of (4) with respect to K, 1 —b—c => r-B Pf(QA"L_a)C K 6 =0 (8) C C C , :1 —b—c 2.» K: rB-Pf— (QA'1L'“)C (9). plug into (7) C -1 c c =>L={WPf-lQA"((r-C,;Pf—](QA"L’“)C)‘b‘cfb 3‘“ (10) Collect terms, -(b+C) I) “I -(b+(‘) b c I 3 L = (E) a+b+c (2) a+b+c A a+b+c w a+b+c ra+b+cpfa+b+c Qa+b+c (1 1) a b fora=¢1 +6111, b =¢2+62u, c =¢3+63u (12), plug into (11) 83 ’(¢2+92"+¢3+93ul ¢2+92u *1 -(¢2 +92u+¢3+03u) 2, L: (13—+931) A+A+A (M)A+A+A A¢1+¢2+¢3w A+A+A ¢1 + 91 14 $2 + 02a —____¢2+92u ¢3+93u l r¢1+¢2+¢3 pf¢, +¢2+¢3 Q¢l +412 +¢3 Assuming that 91 + 92 + 93 = 0, take logs on both sides, and collect terms, 23LnL= -1 LnA+ —(¢2+¢3) Ln(¢3+93u + ¢2 Ln ¢3+63u ¢1+¢2+¢3 ¢1+¢2+¢3 ¢1+91u ¢1+¢2+¢3 ¢2+92u + 91 Ln(¢3+63u)u+ 92 Ln(¢3+63u)u+ -(¢2+¢3) an ¢I+¢2+¢3 ¢l+91u ¢1+¢2+¢3 ¢2+92u ¢1+¢2+¢3 + 91 uan $1 ‘1' 4’2 '1' 4’3 ¢2 Lnr+ 62 uLnr+ LDIP” + 63 uLnPf ¢1+¢2+¢3 ¢1+¢2+¢3 ¢1+¢2+¢3 ¢l+¢2+¢3 l + Ln (l3) ¢1+¢2+¢3 Q ) 84 Appendix 2-B. The GM estimator developed by Arellano and Bond (1991) was used to obtain consistent estimates in a dynamic panel data model where a right-hand-Side variable (lagged dependent variable in this case) is correlated with the disturbance. This study applies the GM estimator to a panel data model where a right-hand-side variable is potentially correlated with the disturbance. The principle is that additional instruments can be obtained if one utilizes the orthogonality conditions that exist between lagged values of the endogenous variable and the disturbance V ,,. For example, for equation: L,,= 5 W,, + V,,‘ (1), the instruments are [Wt ,W.,r-2 ], (2) Then, the matrix of instruments is m=[W’,,. ...., W’N] and the moment equations described above are given by E((fi’,AV 10:0 Pre-multiplying the first-difference equation (1) by 65’, one gets m’AL = (11’ (AW,,) 6 + (D’AV (3) , since E((U’,AV i,)=0, we can perform GLS on (3), and then we get the GMM estimator. In addition, for the GMM estimator to be valid, we need three assumptions as follows: (1) E(U ,)=0, E(V ,, )=0, E(v,, U,)=0, for i=1, ...... ,N and 1:2,. . ...,T. 85 In other words, errors, including unobservable individual effects and the remainder disturbance, are independently distributed and are not correlated to each other. (2) E(V,, V i's)=0 for i=1,. . ...,N and V(for all) t at s. This indicates that the disturbance is un-correlated across time. (3) E(W,I V i,)=0 for i=1, ...... ,N and t=2,......T. With the above 3 assumptions, E(w m AV ,.,)=0 for t=3,. . ..,T and 522 is valid, and is the set of moment conditions used in the first-differenced GMM models. This study uses Ox DPD (dynamic panel data) programs developed by Doomik, Arellano, and Bond (1999) for the GMM estimation. 86 CONCLUDING REMARKS Chapter One of this dissertation focuses on the impact of deregulation on labor earnings, as well as employment. The findings indicate that union workers in the electric power industry experienced a significant decline in their wage premiums after deregulation (13 percent), while wage premiums of the nonunion workers remained unchanged. Level of employment in the electric power industry exhibited a pattern similar to that of wages, in which the relative employment of union workers was substantially reduced following deregulation (37 percent) while the relative employment of nonunion workers did not show a significant change. The sensitivity analyses find that high electricity prices may have contributed to the wage reductions occurring in those deregulated states, but that the reductions of employment level in the deregulated states were not related to high electricity prices in these states. In other words, we may not observe significant wage reductions in states with low electricity prices one they deregulate. However, deregulation seems to put pressure directly on the power companies to cut employment. The findings of Chapter One are consistent With the labor rent sharing hypothesis, which states that labor in the regulated industry, especially union workers, is likely to share rents with their employers. However, we find that high electricity prices, instead of deregulation itself, might have contributed to wage reductions in the deregulated states. Based on the estimated 13 percent drop in union wage premiums, union workers in the electric power industry shared, at least, modest rents with their employers before deregulation. Furthermore, in light of the dramatic reductions in union employment shown in our results, unions in the electric power industry might have traded off the level of employment against wages. In other words, it is possible that the employment level might 87 have dropped less severely if unions could have made more wage concessions, since we do not observe similar degrees of employment reductions in the non-unionized sector. Chapter Two studies unionization and the labor demand using the electric power industry as an illustration. This study finds that the unionization rate has a small but statistically significant effect on the wage elasticity of demand for labor. According to our results, every one percent increase in the unionization rate is accompanied by a .0006 percent decrease in the wage elasticity (less elastic). However, this study does not find significant changes in the union effect on wage elasticity over the course of deregulation after taking into account the effect of technological advances. The finding of a significant union effect on wage elasticity of labor demand supports the argument that unions care about employment as well as wages. The robustness checks suggest some inconsistency between our theoretical assumption about the production function and the estimated results. This could either be a result of the proxy for the cost of capital used in this study being a poor indicator, or be a result of some mis- specifications in the production function. The findings of these two studies suggest that unions have played a very important role in the electric power industry in the pre-deregulation regime. Unionized workers shared rents with their employers. Unions were also able to affect the labor demand by making the wage elasticity of demand for labor less elastic. As deregulation unfolds, it is as important to know how unions will continue to adjust themselves to a competitive environment in terms of their future wage and employment policies. 88 111111111111111111111111111111171 3 93 02334 0213