Valuing instream flows using the hedonic price method
 The Oregon Water Trust (OWT) uses a market-based approach to protect and enhance instream flows in Oregon. We use the hedonic price method to estimate the effect of numerous variables on the annualized price OWT pays for water rights: the amount of water protected by the transaction, transaction type (state approved or contractual agreement), presence of anadromous and/or resident fish, and if a fish is listed under the Endangered Species Act (ESA). We find evidence of a premium for state-approved transactions and for transactions that protect water in streams with listed species. Adjusting the amount of water protected by each transaction to include only rights that will be delivered with a high degree of certainty produces coefficient estimates that are similar, but more accurate, than using unadjusted water rights amounts.
 In 1987 the Oregon Legislature passed the Instream Water Rights Act which marked the first legal protection of instream water rights in the state. The Act was a crucial part of a process for protecting instream flows that began in 1955 with the adoption of minimum perennial flow targets on streams that were not fully appropriated [Neuman, 2004]. The Act strengthened this earlier legislation by ensuring that water rights dedicated instream retained their original priority date, defined instream flows as a beneficial use under the doctrine of prior appropriation, and allowed private organizations to purchase, lease or accept donations of water rights for conversion to instream use.
 The doctrine of prior appropriation is the general governing principle for water law in the Western United States, and is the law in Oregon, first customarily, and then as codified in the 1909 Oregon Water Code [Oregon Water Resources Department, 2006]. Under this doctrine, water that is verifiably diverted and beneficially used is legally protected, with the right dated at the date of first use and limited to the quantity first diverted for that type and place of use and point of diversion [Getches, 1997]. In times of water shortages, water rights are filled in order of their priority until the stream is dewatered or all rights have been delivered. Uses must be “beneficial,” which traditionally included agriculture, industry, and municipal public water services, but was expanded under the Act to include instream flows [Pilz, 2006].
 The Act also provided the legal framework for the creation of the first water trust in the United States: the Oregon Water Trust (OWT). Founded in 1993, OWT is dedicated to “restoring surface water flows for healthier streams in Oregon by using cooperative, free-market solutions” [Paulus, 2007, p. 14]. OWT purchases and leases water rights from willing sellers with rights on small or medium size streams that, with a high degree of certainty, are delivered even during dry summer months. These streams are located in watersheds that have supported historically important anadromous and/or resident fisheries, including several stocks listed under the Endangered Species Act (ESA) [Paulus, 2007]. A pioneer in developing new transaction methods for purchasing and leasing water rights, OWT has permanently protected 60 cubic feet per second (cfs) and, in 2007, had active projects in 10 of the 18 basins in Oregon [Oregon Water Trust, 2007].
 This paper analyzes the determinants of the price OWT pays for purchasing and leasing instream water rights in Oregon. We contribute to the literature in several ways. First, we apply, for the first time, the hedonic price method to analyze the characteristics that influence the price paid to lease or purchase water rights. Second, we develop and describe a methodology to accurately measure the amount of water protected by each transaction. The third contribution is the incorporation of several key variables into our model including dummy variables that categorize transactions by type (state approved or a contractual agreement) and other variables that capture the presence of anadromous and/or resident fish in the stream protected by the transaction and if these stocks are listed under the Endangered Species Act.
 The value of water has been estimated using several techniques including the hedonic price method, agricultural production models, and data from water trusts on the price of leasing and purchasing water rights. Stated preference techniques, such as the contingent valuation method, have been used to estimate the use and nonuse value of fish, including stocks listed under the Endangered Species Act. Faux and Perry  apply the hedonic price method (HPM) to property transactions in Malheur County, Oregon. The estimated value of water used for irrigation ranges from $147 per permanently delivered acre-foot for the lowest-quality soil to $729 per acre-foot for the highest-quality soil [Faux and Perry, 1999]. Butsic and Netusil  use the HPM to estimate a willingness to accept for permanent transfers of $261 per acre-foot and 1 year lease values ranging from $5.33 per acre-foot with a 2% discount rate to $26.10 per acre-foot with a 10% discount rate using property transactions in Douglas County, Oregon.
 The cost of improving instream flows in the Pacific Northwest to a level that will support resident and anadromous fish is estimated by Jaeger and Mikesell  using two different approaches: a farm production model and transactions data from Oregon Water Trust and Washington Water Trust. The authors estimate, using Oregon Water Trust data, an annualized price of $9 per acre-foot for permanent transfers and $23 per acre-foot for 1 year leases in Oregon, using a 6% discount rate in 2002 dollars; their farm production model produces similar results [Jaeger and Mikesell, 2002]. Using these values, and those obtained from an analysis of Washington Water Trust data, a broad regional effort to improve instream flows to support healthy fish populations is estimated to cost between $8.2 million to $65.2 million annually or between $1 to $10 per person per year in the Pacific Northwest [Jaeger and Mikesell, 2002]. In addition to benefiting fish stocks, improved instream flow may increase recreation opportunities and hydropower production. An analysis for the Big Hole and Bitterroot rivers in Montana estimates an increase in recreation benefits of $50 per acre-feet and additional hydropower revenue of $25 per acre-feet in 1988 dollars [Duffield et al., 1992].
 These estimated water values are classified as use values, that is, they result from actively using, although not necessarily consuming, the resource. Use values can also be estimated for fish stocks, for example, use values may arise from recreational fishing, commercial fishing, and even from viewing fish at their spawning grounds or as they migrate through fish ladders. Other values include option value, which is the value of keeping open the opportunity to use a resource in the future, and nonuse value, which is derived from existence value (knowing that a resource exists) and bequest value, which is a desire to preserve a resource for future generations. Loomis and White  summarize the literature on rare, threatened, and endangered fish and conclude that existence value is a large part of the overall willingness to pay to protect these species. Annual willingness to pay estimates for salmon and steelhead range, in 1993 dollars, from $63–$181 for San Joaquin River Chinook Salmon to $31 for Pacific Salmon and Steelhead [Loomis and White, 1996]. Layton et al.  estimate willingness to pay for changes in fish populations in Washington State using a stated preference survey of randomly selected Washington State households. Per household annual willingness to pay estimates for a 50% increase in fish populations, compared to maintaining stable fish populations for the next 20 years, range from $9.92 for Eastern Washington and Columbia River Migratory fish to $21.07 for Western Washington and Puget Sound Saltwater fish. An alternative scenario, in which fish populations decline over the next 20 years at the same rate experienced during the past 20 years, elicits a per household annual willingness to pay for a 50% increase in fish populations ranging from $14.55 for Eastern Washington and Columbia River Migratory fish to $31.28 for Western Washington and Puget Sound Saltwater fish. The authors compare their results with other studies and conclude that their estimates are similar. Importantly, the study demonstrates diminishing marginal willingness to pay for increases in fish populations for all types of fish evaluated in the study.
3. Transaction Types and Quantity of Protected Water
 OWT's transactions can be grouped into two categories: state approved and contractual. State approved transactions include standard leases, time-limited transfers, split season leases, conserved water projects, and permanent transfers. These transactions require the approval of the Oregon Department of Water Resources (OWRD) and, once approved, are enforced by water masters who are OWRD officials [OWRD, 2007].
 Standard leases are leases for all or a portion of a water right for a period of up to 5 years; time-limited transfers are similar except they last for more than 5 years. Under split-season leases, water is leased instream for a portion of the irrigation season, but is available for consumptive use for the rest of the season [Paulus, 2007]. Conserved water projects include improvements in irrigation or water transportation techniques, such as lining a canal. The Act allows 25% of the conserved water to be dedicated instream and attaches the original priority date to the conserved water. Permanent transfers are the sale of a water right that is dedicated instream in perpetuity; although paid for by OWT, these rights are held by the OWRD.
 Contractual agreements, which include water use agreements, forbearance agreements, and stored water contracts, do not require approval by the OWRD and are not enforced by water masters, but instead through the court system. These agreements occur when there is certainty that the water left instream will not be diverted by someone else due to the water right holder's location on the stream or because the contract includes every water right holder on the stream.
 A water use agreement is a general term for agreements that specify a change in water use that will improve instream flows. These include agreements where all water right holders on a stream cease diverting if flows fall below a specified level or when a landowner switches the point of diversion from a smaller stream to a larger stream, which provides more water for the tributary where flow is limited but does not impact the landowner's water use. In a forbearance agreement, a water right holder agrees not to divert if specified conditions exist, for example, flow is below a certain level. Stored water contracts are contracts to release water from dams such as those owned by the Bureau of Reclamation.
 Several issues arose when preparing the database for analysis including one previously not considered in the literature: how to treat transactions with multiple priority dates. OWT is interested in deliverable rights. Therefore junior water rights, which are frequently listed as part of a transaction, but are unlikely to be delivered during the critical late summer months, are given no value by OWT staff when calculating the amount they are willing to pay to lease or purchase water rights (K. Schonek, personal communication, 2008).
 The hedonic price model requires all variables to be accurately measured, so we examined the priority dates and water rights for each deal and adjusted the quantity of water protected accordingly. All water rights in contractual agreements were retained because they operate like senior rights because of river and water right conditions present on the specific reach involved in the agreement, but they need not actually have senior priority (D. Pilz, personal communication, 2008). The average unadjusted (CFS_unadjusted) quantity of water protected is 1.91 cfs. Eliminating water rights that are unlikely to be delivered during the late summer months reduces the average amount of water protected to 1.60 cfs: a 17.66% decline.
 Water Strategist, the primary source for transactions information for numerous studies [Loomis et al., 2003; Brewer et al., 2007], and the source for transactions data in the Water Transfer Database (available at http://www.bren.ucsb.edu/news/water_transfers.htm), does not list priority dates, so the deliverability of the water in these transactions is not considered, which potentially biases results. Other concerns with transactions reported in Water Strategist include the reporting of multiyear deals only in the initial year of the transaction and the failure to report contractual agreements, which account for 11.3% of OWT's transactions.
 Because of these limitations we are unable to expand our data set to include transactions undertaken by other organizations in the state that are leasing and purchasing water rights for environmental uses. An analysis of the 69 transactions recorded as “environmental” in the Water Transfer Database as of February 2009 shows that while OWT's transactions account for 53.6% of these deals, the total amount of water it has protected is much smaller than other organizations. In total, OWT transactions accounted for 12.8% of the 874,704 acre-feet of water transfers classified as environmental in the Water Transfer Database. Using our data set, the average price OWT pays per acre-foot of water is $128.86 (2007 dollars). This is comparable to the average price of $169.49 (2007 dollars) paid per acre-foot for transactions classified as a transfer of water to environmental uses in the Water Transfer Database.
 The hedonic price method uses sales data for a marketed good, and detailed information about the good's characteristics, to estimate the implicit price for each characteristic [Freeman, 2003]. Hedonic wage models have been used to value risk, while property value models have been used to value environmental attributes such as open spaces [Lutzenhiser and Netusil, 2001; Qiu et al., 2006], and improvements in air [Kim et al., 2003] and water quality [Poor et al., 2007]. In this paper we apply the hedonic price technique to estimate the marginal implicit prices for water rights transactions that are used to protect instream flows.
 The traditional hedonic price model for property sales combines the sale price of properties with information about their structural, environmental, and locational characteristics [Freeman, 2003]. We apply this model to water rights transactions, using the same conceptual framework, but we associate the annualized price (Pi) for leasing or purchasing water rights with characteristics of the water rights (WRi), such as the quantity of water, the transaction type, locational (Li) characteristics, which include the county where the transaction is located, and environmental (Ei) characteristics such as whether a fish listed under the Endangered Species Act is present in the water protected by the transaction. This is represented by
 Theory does not provide guidance on the functional form for estimation, so it is common for authors to estimate several equations including ones that allow the implicit price to vary with the level of the characteristic [Freeman, 2003]. A Box-Cox transformation of the dependent variable provided evidence that a semi-log functional form was most appropriate. This specification conforms to our intuition of a nonlinear relationship between the annualized price paid for a transaction and its characteristics and is consistent with other applications of the hedonic price method.
5. Data Set
 Data on the price of each transaction, quantity of water protected, priority dates, transaction type (standard lease, permanent transfer, etc.), transaction length (measured in years), days water is dedicated to instream flow, fish species in the stream protected by the transaction, county, and 5th field hydrologic unit code were obtained from the Oregon Water Trust for transactions from 1994 through June 2008. This information was used to generate several additional variables including if a fish species in the stream protected by the transaction is listed under the Endangered Species Act and a dummy variable that equals one if a transaction occurs in an OWT priority watershed, a designation given by OWT to watersheds that have supported historically important fisheries, where local partners exist, and where funding is available to support projects.
 The dependent variable for our analysis, the real annualized price paid for a transaction, was adjusted to 2007 dollars using the Consumer Price Index for all Urban Consumers. Price is assumed to represent an equilibrium condition where the willingness-to-pay of the buyer (OWT) equals the seller's willingness-to-accept. Because transactions are for different lengths of time, we annualized the real price paid for a transaction using discount rates that varied by transaction length [Office of Management and Budget, 2008].
 The primary explanatory variables used in our analysis, and their expected signs, are described in Table 1. County-level dummy variables were created to capture differences in economic conditions, agricultural crops and growing conditions, and variation in the length of the irrigation season since each county was decreed separately (K. Schonek, personal communication, 2008). The year the transaction was initiated was also included to capture learning effects and the maturation of the water rights market in Oregon.
Table 1. Explanatory Variables
|CFS_adjusted||Amount of water adjusted to include only deliverable rights||Positive|
|CFS_unadjusted||Amount of water listed in the transaction||Positive|
|Fish species variables:||Dummy variables for transactions with only anadromous fish present in the stream, resident and anadromous fish, or only resident fish.||Transactions with both resident and anadromous fish protect more species, so they may have a higher relative price|
| Resident & Anadromous|
| Resident (excluded category)|
|ESA||Dummy variable for transactions protecting water with fish stocks listed under the ESA||Positive|
|Transaction types:||Dummy variables for each of the nine transaction types||Positive for Standard Lease, Split Season Lease, and Conserved Water Projects compared to the excluded category (Permanent) because these transactions are for short time periods during which landowners need to pay fixed costs. Negative for other transaction types compared to the excluded category (Permanent) because these contractual agreements provide less protection for deliverability.|
| Permanent (excluded)|
| Standard Lease|
| Split Season|
| Time Limited|
| Conserved Water|
| Stored Water|
| Water Use|
| Minimum Flow|
|State approved||Dummy variable that equals 1 if the transaction type is state approved: Permanent, Standard Lease, Split Season Lease Conserved Water Project||Positive because state approved transactions provide a greater degree of protection about deliverability|
|Transaction length:||Dummy variables for transactions lasting 1 year, between 2 and 9 years, and 10 years or more||Transactions lasting for 1 year are expected to have a higher relative price since water rights holders need to pay fixed costs, e.g., payments to maintain irrigation pipes, tractors, etc.|
| One Year|
| Two–Nine Years|
| Ten Years or More|
|DaysInstream||Number of days the water right is dedicated to instream flow||Positive|
|Priority Watershed||Dummy variable that equals one if the transaction is in an OWT priority watershed||Uncertain|
|Year||Initial year for the transaction, 1994 = 0||Negative, because of learning effects|
|County||Sixteen dummy variables for the counties in which the transaction takes place. Jackson is the excluded category||Uncertain|
| Hood River|
 OWT has executed a total of 177 transactions protecting, on average, 1.91 cfs (CFS_unadjusted). Adjusting water rights to eliminate rights that are unlikely to be delivered during critical summer months reduces the amount of instream flow protected to 1.60 cfs (CFS_adjusted) on average. The vast majority of transactions, 93.2%, protect water in streams with a stock listed under the Endangered Species Act, are standard leases (62.1%), last for 1 year (52.5%), and are located in an OWT priority watershed (57.6%). Summary statistics for the variables used in the regression analysis are presented in Table 2.
Table 2. Summary Statistics
|Annualized price per transaction (2007 dollars)||$8936.48||$26,963.47||$40.66||$191,098.60|
|Resident & Anadromous||37||20.90|
|Resident (excluded category)||6||3.39|
|Ten Years or More (excluded category)||48||27.12|
 Three regressions were developed to explore the relationship between the explanatory variables and the dependent variable, the natural log of the annualized price paid for a transaction. Model 1 uses the amount of water protected by each transaction after adjustments were made to eliminate rights that would not be delivered during the critical late summer months. The estimated coefficient on CFS_adjusted is positive, as expected, and statistically significant at the 1% level. In Model 1, an additional cfs is estimated to increase the annualized price per transaction by 50.08% or, when evaluated at the mean annualized transaction price of $8,936, by approximately $4,475.
 The estimated coefficients on Anadromous and Resident&Anadromous are positive and statistically significant, but an F test fails to reject the hypothesis that the estimated coefficients are equal to each other (F(1, 145) = 0.69; p value = 0.407). All observations with only anadromous fish in the stream protected by the transaction have stocks listed as threatened under the ESA, 31 of the 37 transactions where both resident and anadromous fish are present have listed stocks, but none of the transactions with only resident fish have a listing. The significant t tests and insignificant F test leads us to believe that ESA status is driving the results, so Models 2 and 3 use a dummy variable (ESA) to reflect transactions that protect water in a stream with a listed stock.
 The difference in sign between state approved transactions (all positive) and contractual agreements (all negative), compared to the excluded category of permanent transactions, is not surprising since state approved transactions provide a greater degree of certainty about the delivery of water rights. We reject the hypothesis that the estimated coefficients for transaction types are equal to each other (F(7, 145) = 2.80; p value = 0.009), but are unable to reject the hypothesis that the coefficients on the included state approved transactions (StandardLease, SplitSeason, TimeLimited, ConservedWater) are different from each other (F(3, 145) = 1.69; p value = 0.172) and that the coefficients on contractual agreements (Forbearance, StoredWater, WaterUse, MinimumFlow) are equal to each other (F(3, 145) = 1.51; p value = 0.214). These results suggest that it is appropriate to collapse transaction types into two categories: one for state approved transactions and another for contractual agreements.
 The second model incorporates dummy variables for state approved transactions (StateApproved) and for transactions that protect streams with ESA listed stocks (ESA). The estimated coefficient on CFS_adjusted drops, but still has a positive and statistically significant effect of 39.24% ($3,506 annualized, in 2007 dollars) for an additional cfs of water. ESA listing provides an 82% ($7,328 annualized, in 2007 dollars) increase in the price of a transaction, while state approved transactions are estimated to command a 128% premium over contractual agreements. The signs and magnitude of the other control variables are consistent with Model 1.
 To explore the potential bias introduced by not adjusting the amount of water protected to account for junior rights, a third model was developed using CFS_unadjusted and the StateApproved and ESA dummy variables. The coefficient on CFS_unadjusted is lower, as expected, when compared to the coefficient on CFS_adjusted in Models 1 and 2. In Model 3, an additional cfs is estimated to increase the annualized price of a transaction by 31.78% ($2,840 annualized, in 2007 dollars). The magnitude of the coefficient on ESA is similar to Model 2, but it is not significant at the 10% level. The most noticeable change in Model 3 is the statistically insignificant negative sign on StateApproved. Failing to adjust the amount of water protected has completely eliminated the estimated 128% premium for state approved transactions.
 The coefficients on transaction length (OneYear, Two-NineYears) indicate a statistically significant premium compared to the excluded category (TenYearsorMore) after controlling for transaction type. An F test fails to reject the hypothesis that the estimated coefficients on these variables are different from each other for all three models. This result is expected because water rights holders must still pay expenses for shorter transaction periods, such as maintaining irrigation pipes and farm equipment, which may not be incurred for longer periods.
 While we do not test our models for spatial relationships, we believe that if one exists it is spatial error, which is a result of omitting spatial factors. A spatial lag model is appropriate if nearby transactions influence the price paid by OWT for a water right, but OWT uses many different methods to arrive at a price. These include the staff's valuation of the environmental benefit from acquiring the water right, agricultural production models, comparable sales drawn from the same basin, or from adjacent basins, and a comparison of the sale price of properties with water rights to those without water rights.
 The consequence of spatial error is that standard errors are inflated. However, as stated by Mueller and Loomis [2008, p. 212], “inefficiency of OLS in the presence of spatially correlated errors may not always be economically significant, suggesting nonspatial hedonic property models may provide results useful for policy analysis…”
 As stated in Table 1, we were uncertain about the effect of priority watershed on price. The estimated coefficient for this variable in all three models is positive, but is never statistically significant. This finding can be explained by recalling that watersheds receive this designation if they have supported historically important fisheries, are located where local partners exist, and have funding available to support projects. While more staff time is devoted to developing deals in these areas, these results indicate that OWT is not paying a premium for transactions in these watersheds.
 DaysInstream is consistently negative, a result counter to our expectations, but significant only in Model 1. County dummy variables are included in all models, but are not reported in Table 3 (full regression results are available from the authors). F tests for all models show the joint significance of these variables. Year, a continuous variable that represents the year a transaction was initiated, is consistently negative and of similar magnitude in all models. This variable, which was included to capture learning effects and the maturation of the market for water rights in Oregon, shows a per year decline in the annualized price of a transaction ranging from 5.64% (Model 1) to 6.95% (Model 3).
Table 3. Regression Resultsa
|CFS_adjusted||0.406*** (0.075)||0.331*** (0.037)|| |
|CFS_unadjusted|| || ||0.276*** (0.026)|
|Anadromous||0.776*** (0.231)|| || |
|Resident & Anadromous||1.140** (0.506)|| || |
|ESA|| ||0.599* (0.361)||0.539 (0.374)|
|Standard Lease||0.368 (0.504)|| || |
|Split Season||0.713 (0.533)|| || |
|Time Limited||1.314 (0.908)|| || |
|Conserved Water||0.282 (0.507)|| || |
|Forbearance||−0.738 (0.835)|| || |
|Stored Water||−0.217 (0.775)|| || |
|Minimum Flow||−2.634** (1.273)|| || |
|Water Use||−0.331 (0.640)|| || |
|State Approved|| ||0.826* (0.427)||−0.231 (0.290)|
|One Year||0.968** (0.441)||1.192*** (0.202)||1.106*** (0.199)|
|Two–Nine Years||0.660* (0.357)||0.902*** (0.271)||0.806*** (0.241)|
|Days Instream||−0.003* (0.002)||−0.002 (0.002)||−0.002 (0.001)|
|Priority Watershed||0.004 (0.229)||0.118 (0.212)||0.040 (0.218)|
|Year||−0.058** (0.023)||−0.064*** (0.025)||−0.072*** (0.024)|
|Constant||6.269*** (0.498)||5.585*** (0.765)||6.969*** (0.658)|
 As stated by Loomis [1998, p. 1008], “The demands for water driven by the ESA is a reflection of broad public concern, not just that of recreationists.” OWT was founded in 1993 by a diverse group of stakeholders including farmers, ranchers, environmentalists, and irrigators who created an organization to take advantage of changes in water law resulting from the 1987 Instream Water Rights Act [Neuman and Chapman, 1999; Neuman, 2004]. Since its founding OWT has focused on protecting instream flows for historically important but vulnerable fisheries. OWT's actions, and those of other water trusts and government agencies that purchase instream water rights, have benefited users, such as commercial and recreational fishermen and water recreationists, as well as individuals with “nonuse” values for resident and anadromous fish, especially those listed under the ESA.
 Our preferred models, which adjust the quantity of water protected by a transaction to remove junior rights, show an economically and statistically significant premium for streams with stocks listed under the ESA and for state approved contracts. While many valuation studies provide dollar/fish amounts [Loomis and White, 1996], our study provides a general estimate of value: the annualized price paid for a transaction in streams with listed species is 82% higher, or $7,328 annualized in 2007 dollars, than those without. A biological function that links increased instream flow with changes in fish populations could be used to generate a dollar/fish estimate, but determining this biological relationship is outside the scope of our analysis.
 Adjusting water rights to more accurately measure the quantity of water protected by each transaction proved crucial since the large and statistically significant premium on state approved contracts disappears when water amounts are not adjusted. While the coefficient on the ESA dummy has a similar magnitude across models, it lost its statistical significance in the model with CFS_unadjusted. Failing to recognize the limitations of the major data source (Water Strategist) for water rights transactions may lead to biased results and potentially inappropriate policy suggestions because priority dates are not included with transaction information and contractual agreements, which make up 11.3% of OWT's transactions, are not reported.
 We gratefully acknowledge the assistance of David Pilz, Kim Schonek, Scott Peerman, Steve Parrett, Molly O'Sullivan, Kacy Markowitz, Fritz Paulus, and Lois Hobbs. Jan Crouter, Jon Kadish, participants in the 2008 Washington Department of Ecology Instream Values Symposium, the 2008 Reed College Social Sciences Summer Workshop, and a seminar at the University of Nevada–Reno Department of Resource Economics provided valuable suggestions on our empirical work and earlier drafts. Funding provided by a 2008 Bernard Goldhammer Summer Collaborative Grant.