chapter  17
20 Pages

Modeling behavioral response to changes in reservoir operations in the Tennessee Valley region: Paul M. Jakus, John C. Bergstrom, Marty Phillips, and Kelly Maloney

ByPAUL M . JAKUS , JOHN C . BERGSTROM , MARTY PHILLIPS

Introduction In 1933, the United States Congress created the Tennessee Valley Authority (TVA) to help bring economic development and quality-of-life improvements to people living in the Tennessee River watershed. The TVA was one of the hallmarks of President Franklin D. Roosevelt’s “New Deal” program responding to the US “Great Depression.” From its inception, the primary goals of TVA have been providing flood control within the valley, assuring a navigable channel on the Tennessee River, and the provision of electric power (TVA 2010). A fourth goal, regional economic development, was added in later years. Over the past 75 years, electricity from TVA power plants has played a key role in the economic development of the TVA service region composed of parts of seven states: Alabama, Georgia, Kentucky, Mississippi, North Carolina, Tennessee and Virginia. TVA generates electricity through hydroelectric dams, coal-fired power plants, nuclear plants and some “green power” (e.g. wind, solar) sites. Hydroelectric plants currently account for about 10 percent of total TVA electricity generation. TVA operates an interconnected system of 49 dams and reservoirs, of which 35 projects are important for the generation of hydropower. During the previous two decades, especially the 1990s, recreational use of TVA reservoirs increased dramatically. This recreational use comes not only from those who live adjacent to the reservoirs but also from visitors who live much farther away. Increased recreational use of TVA reservoirs can sometimes compete with other reservoir benefits. For example, many recreation users desire high reservoir water levels to support recreational activities such as boating, water skiing and swimming. However, if the TVA were to reduce reservoir water releases during the summer to support higher lake levels, there may be a loss of public power generation and navigation benefits. Thus, an important management policy issue facing the TVA is balancing the multiple objectives and benefits of TVA reservoirs including growing recreational use and continued power generation, navigation and flood control needs. To facilitate reservoir management, the TVA initiated a comprehensive reservoir operations study (ROS) in 2001. The ROS examined the environmental, social and economic impacts of reservoir management alternatives. The results

of the ROS were published in 2004 in the form of a final ROS programmatic environmental impact statement (TVA 2004). One of the objectives of the ROS was to assess the effects of fluctuating reservoir water levels associated with different management alternatives on recreation visitation and economic impacts. The effects of fluctuating reservoir water levels on recreation visitation were examined using a combined travel cost and trip response model. Standard travel cost models are classified as revealed preference (RP) nonmarket valuation approaches. Revealed preference approaches are so called because individuals reveal their preferences or demand for a nonmarket good or service through actual behavior (for a discussion of the travel cost model as revealed preference method, see Parsons 2003). For example, individuals reveal their demand for recreation trips to reservoirs at different water levels through the actual number of trips they make to reservoirs at fluctuating water levels. A trip response model is a type of intended or contingent behavior model (Bergstrom et al. 2004; Betz et al. 2003; Loomis 1993; Teasley et al. 1994). Intended or contingent behavior models are classified as stated preference (SP) nonmarket valuation approaches. Stated preference approaches are so called because individuals state their preferences or demand for nonmarket goods and services directly through survey questions. For example, in a survey, individuals may state how many recreational trips they would take to a reservoir at different water levels. The combined RP-SP recreation demand model estimated in this study presents a behavioral response model that combines actual behavior under current reservoir operating guidelines and intended behavior contingent on future reservoir conditions under proposed management alternatives. Water levels at different times during the recreation season are treated as quality shifters, which are then used to predict changes in recreation behavior for any of the potential management alternatives. We supplement our RP data with SP data because a revealed preference model based on a single year of observed recreation could not include a meaningful measure of policy-relevant water levels: at any given site the actual water level is what it is as determined by current management and environmental factors (e.g. rainfall patterns). Management alternatives proposed by the TVA result in different water elevations at a given site, so a model must be able to include responses to higher or lower water levels at a site. One modeling option is to treat observed changes in actual trip behavior with changes in actual water levels as “natural experiment” data and estimate a RP recreation demand model showing the relationship between changes in actual trips and water levels. However, even if one could introduce site level variation by using multiple years of RP data, the water levels of many of the proposed management alternatives would be outside the levels experienced in the past, thus introducing the problem of “out-of-sample” prediction. Another complicating factor in the design of our modeling strategy was that TVA had developed 65 different management scenarios at the time the survey

was to be implemented, and the model had to accommodate any and all of these alternatives. Combining RP and SP data allowed us to incorporate more variation in water levels within and across reservoirs in our study than would have been available using the “natural experiment” revealed preference data alone. The end result was a flexible, empirical recreation demand model capable of handling many “what if ” scenarios related to fluctuating water levels. This flexible model helped to meet one of the primary policy needs and purposes of the ROS is study which was to assess changes in recreation visitation to TVA projects with changes in water level management.