Determining Water Values with Computable General Equilibrium Models

8 downloads 97 Views 928KB Size Report
We can consider the problem of determining the social cost associated with non-optimal ..... Table 1: Example of a Social Accounting Matrix (millions of $). Taxes.
Determining Water Values with Computable General Equilibrium Models A Report submitted for presentation at “The Importance of Water to the U.S. Economy: Technical Workshop, September 19, National Academy of Public Administration, 900 7th Street NW, Suite 600, Washington D.C.

This report is submitted to Industrial Economics, Inc. 2067 Massachusetts Avenue, Cambridge, MA 02140

Submitted by: Elizabeth Fadali, MS Research Associate Department of Economic University of Nevada, Reno [email protected] Kimberly Rollins, PhD Associate Professor Department of Economic University of Nevada, Reno [email protected] Shawn Stoddard, PhD Senior Resource Economist Truckee Meadows Water Authority Reno, Nevada [email protected]

1

Determining Water Values with Computable General Equilibrium Models This paper assesses the role of computable general equilibrium (CGE) models for determining the value of water to the US economy. Four sub-objectives are (1) to determine the circumstances in which CGE modeling is suited for this role, (2) to examine how a CGE model must be built to specifically handle water for the types of water-related economic changes and policy questions relevant for CGE simulations, (3) to review existing examples of CGE models that determine water values in the US and elsewhere, and finally, (4) to provide recommendations to help guide development of water CGE models for applications in the US. A systematic review of water CGE models is found in the appendix to this paper; however, we use examples from the literature throughout the body of this paper to illustrate concepts and points, so that most of the articles reviewed in the appendix are also referred to throughout this paper. One of the main findings in this paper is that data limitations in the US, relative to data availability in other parts of the world, is a likely explanation for the relatively few examples of water CGE models in use in the US. This paper is organized in the following manner. First, the introductory section describes the complexity of water resources and economic sectors in regional economies. This complexity motivates the need for analytical methods, namely CGE modeling, that can take into account interrelated markets and secondary impacts in evaluating the net effects from changes that affect water resources. The second section describes the basic elements of CGE models, with a focus on water. The third section focuses more specifically on the issues that water-CGE modelers must contend with, and notes that data problems in the US make creation of water accounts for CGE models difficult. The fourth section describes multi-regional CGE modeling approaches, and their uses for water resources. Section five continues to extend CGE approaches by discussing the use of dynamic models for water applications. Finally, the sixth section describes an on-going effort to build a CGE model for the state of Nevada for eventual application to scenarios involving water transfers between regions. The uses of multiregional and dynamic elements are discussed in the context of this application. The concluding section summarizes advantages and limitations of CGE modeling for water, and types of questions that CGE models are particularly good for addressing. Two appendices follow. The first is a literature review of CGE papers published that incorporate water. Appendix B includes details about the data requirements for the Nevada CGE model.

1. The role of water in economic activity 1.1 Why water resources have a complicated value structure Water is a vital resource in any economy and at all stages of economic development. In years past communities were located on or in close proximity to waterways that provided means of transportation, water for irrigation of crops and livestock, and for human consumption. As communities grew the demand for water storage, flood control, irrigation, and power generation resulted in the building of infrastructure in the form of dams and reservoirs which added the dimension of controlling the flow and timing to the value of water resources. As economies become more complex, so do the multiple demands for water resources. In the consideration of the “value of water,” the entire spectrum of consumptive and non-consumptive water uses need 2

to be considered, as these varied uses are often integrated. This spectrum includes various combinations of water attributes: timing of water flows; quantity of water; and physical quality of water. The demand for water for a single given use could be based on specific minimum requirements for all of these attributes. And there are many different uses for water. A single water resource could be used to satisfy many different demands for water within an economy. For example, water within a river system is generally limited to being used within that river’s watershed, but not limited to any one use. Some uses of the water do not consume the water; hydroelectric generation is one such example of a non-consumptive use. Flood irrigation of croplands is an example of partial consumption of water; to move the water across the field a quantity is diverted sufficient to create a flow that will cover the field, but only water that is evaporated or transpired is consumed. The next user downstream has access to the balance of the flow, including returns from irrigation systems (return flows). A municipal water utility could divert water for the production of drinking water for the community. This utility is producing at least two new products from the raw water: drinking (potable) water; and conveyance (delivery) of potable water to the place of use. This water can be used for direct consumption by individuals (residential), used as an input for a business, or used as a source of irrigation for community landscaping. All of these uses of water are limited to the watershed, unless an infrastructure is constructed to allow for the export of water from one basin to another. In the above examples the same water resource is used at different times and in different ways with each use bringing value to the economy, either by producing goods and services or as a final demand product. Each of these uses has a different value associated with the water used. Thus it is not really possible to state any single value for water, but rather a water resource used within a watershed (or economy) results in a total increase of economic value or community welfare. A general equilibrium approach to understanding an entire regional economy through its industrial sectors and market interlinkages presents a reasonable way of valuing water resources. Computable general equilibrium (CGE) models are simulations based on general equilibrium theory. CGE models allow for the multiple uses of a water resource in an economy and return estimates of changes in social welfare for increases or decreases in the water resource, thus providing an estimate of the marginal value of water to the economy. This estimate of marginal value implies that some form of water market system is used to allocate water. The definition of a well-functioning market is one that results in equilibrium prices and quantities traded for commodities. In an ‘ideal’ pure market economy, where water is treated like a normal good, the value of water would equal its marginal value product (MVP) and the quantity of water purchased at that price by firms in each sector will depend only on each sector’s demands for water. There would be a supply and demand for water, an equilibrium price for water, and competition for water. This ideal world does not exist; water is not treated as a normal good. In the case of water and depending on the region, ‘prices’ associated with water use do not necessarily reflect the marginal value of water, but rather may be administratively set and subject to subsidies, regulations, or restrictive water rights institutions. In part, this is because several factors make it difficult to specify exclusive private property rights for water. For example: 

Supplies vary from year to year and season to season. 3

 

Connections between groundwater and surface water are not well defined. Water extractions by one user may affect other users, thereby generating external costs. Water infrastructure is often most efficient at large scales, which leads to natural monopolies and, traditionally, government involvement in either supplying or regulating water capture, storage, delivery and treatment (Wittwer 2012; Young 2005).

For these reasons, the price that water users pay generally cannot be assumed to be a market equilibrium price. Water prices, where they exist in the US, are affected by combinations of government and environmental regulations; water rights and property laws; geographical limits on when and where water can be used; constraints on the quality and quantity of water; and natural variability of water resources from droughts to floods. Often water prices are set administratively in such a way that supply infrastructure might be subsidized and opportunity costs are not recognized. Because of the administrative nature of the water prices, these prices will generally vary by use. A market may be observed but, more often, non-standard techniques of analysis are needed to measure the marginal value product of water to a water user or economy. Evaluation of policies that would alter water infrastructure, regional supplies, interbasin transfer, and water quality would require that water values and changes in welfare be quantified for comparison over alternative scenarios. However, the observed ‘prices’ and quantities of water used by households and firms in various sectors in an economy cannot be assumed to represent optimal societal outcomes. In these cases, economic approaches to measuring the value of water and of possible changes in water use are implemented that do not assume that observed quantities and prices are appropriate for estimating water values. In addition, water resources tend to be regional, and are not traded among regions unless specific infrastructure changes take place (pipelines, aqueducts). Thus, water values measured in one region may not be applicable to other regions. 1.2 Determinants of water value, and variations among regions and over time Just as the rental price of an oceanfront condo unit in Maui would be expected to vary over seasons and to differ from the rental price of the same condo unit located in downtown Indianapolis, the value of water would be expected to vary considerably by region, and by season. The rented condo is an input into the production of a vacation and, as such, its value is a function of the prices of outputs (the value of the vacation in one area versus another, or in one season versus all others) and all other inputs (airfare to the destination, local costs for car rentals, and such). In much the same way, where water is an input to production, water value is a function of the prices of outputs and all other inputs. An illustrative example is Australia’s Watermove market. In Australia, land and water property rights were separated in 1994 in a bundle of changes called the Council of Australian Government’s reforms (COAG). This reform established a fairly well-structured water market which offers some important insights about the value of water. Amongst water users within a single defined trading zone, temporary and permanent water trades take place through a water exchange called Watermove (Schreider, 2009). Trading zones are defined by the infrastructure and topography that make a physical water trade possible. Within zones, and using actual bids from buyers and sellers, Watermove determines a single price for short term water trades on a weekly basis. To illustrate the character of these water prices, Schreider (2009) examines the 4

Temporary Water Right/Diversion License in 1A Greater Goulburn Victoria for a fairly wet season (2005/06) and a drought year (2006/07). In the wet year, prices for a megaliter of water varied from AUD 12 to AUD 80, while in the dry year, prices ranged from AUD 300 to AUD 950. Schreider describes these prices as being characterized by upward and downward jumps, with a level specific to a particular season and a drop towards the end of each season. In regression analysis of another set of Watermove regional water trade prices, Wittwer and Griffin (2012) find that the prices are negatively related to the amount of water available for allocation, and positively related to a drought index and to farm output prices. Over the ten-year period the average price for a megaliter of water varied from AUD 35 to AUD 562. These examples illustrate the effects that variations in water supply and the structure of a regional economy can have on water prices. While the overall supply within a region where water can feasibly be traded is a primary influence on the value of water in the region, Wittwer (2012) notes that the value of water in production can also be influenced by:     

prices and availability of all other inputs of production, such as wages, investments made in multi-period capital, such as perennial crops or livestock substitution possibilities available for water prices of trade goods from other regions opportunity costs of water used in alternative industries and for consumption in households.

Finally, direct observation of prices for trades in long-term water rights, just like land and housing prices, have been shown to be subject to boom and bust cycles. The Watermove markets reflect prices that can reasonably be used to assess the marginal value of water as an input across uses within individual regions. This innovative program and the resulting markets illustrate the fluctuations that occur in water values over time and by use and region. Suppose that potential changes in supply of water or growth of demand in any given region create enough of a differential in prices across regions to justify investment in the means of conveyance between regions. The net gains from trade between regions would be a complex function of impacts on a variety of the many sectors that would be affected in each region, both through primary effects (changes in the price of water in the two regions to a single price, less conveyance costs to the importing region), and secondary effects (associated impacts on related industries and incomes). In this case, where water prices could be assumed to more or less reflect the marginal value of water, a computable general equilibrium modeling approach would be appropriate for simulating the effects of and net gains from opening trade in water between regions. In contrast, in most regions in the US, such markets in water do not exist, and prices are instead set by other means. 1.3 Valuing water as an input to production: partial versus general equilibrium approaches Partial equilibrium analysis may be the best approach for analyzing economic impacts for small changes in water attributes (supply, quality, timing, flow, prices) that are unlikely to affect prices of other goods and services throughout the economy in an appreciable manner. Partial 5

equilibrium approaches hold other prices and markets constant, while focusing on a specific water use. These methods include the hedonic property value method, stated and revealed preference, non-market valuation of the value of water to recreation and ecosystem services, estimation of production functions or demand functions, the residual method (subtracting all other input costs from total revenue), and linear programming input-output models where prices are fixed (Young 2005). However, for non-marginal changes in water supplied or pricing associated with many types of water policies, the direct and secondary influences on other commodity and factor markets may be of consequence throughout an economy. Partial equilibrium approaches cannot account for secondary effects, and therefore estimates of changes in water demand and prices from partial equilibrium approaches could lead to over or underestimates of changes in water values, depending on the extent and type of linkages in the regional economy affected. Potential changes in prices in other commodity and factor markets in turn also affect incomes and can have fiscal impacts on regional governments and water suppliers. CGE models are specifically built to represent these interrelationships among markets and sectors in regional economies, where water pricing and supply can affect multiple markets and sectors in non-marginal ways. CGE models are simulations meant to represent the economies of well-defined regions so that analysis of the impact of changes in any one (or combination of) sectors can be traced through to predict changes that will result throughout the entire regional economy. A CGE modeling approach allows for complete exploration of complex feedbacks throughout the economy, so that the modeler can experiment and isolate the effects of many variables and identify the linkages between them with regard to water shadow prices in industry. CGE approaches break down the net effect of a shock to the system (the ‘value’ of an increase or decrease in the supply of water to a region, for example) into individual changes in prices, outputs and incomes by each affected industrial sector, so that a complete set of gainers and losers are identified, along with the measurements of change. To illustrate the difference between partial and general equilibrium approaches, we consider the application of each approach to the case of water as in input to production. Consider, as an example, the case that water fees are set administratively, and are unlikely to be equal to what would be the market clearing price for water in a region. The value of the marginal product (VMP) of water to producers may be either less or more than the administratively set fee rate for its delivery. If more water is demanded than is available at the given administrative fee, then water is a binding constraint and the value of an additional unit of water is indicated by a shadow price for water, which exceeds the administrative price (Diao and Roe 2000). If, on the other hand, the producer has more water than he can profitably use, the VMP for water is zero for that producer. In the case of an administratively set fee that is not a market clearing price for water, water purchases and uses across industrial sectors may be distorted from what is socially optimal. The analyst may wish to determine the welfare loss associated with constraints to water trading. We can consider the problem of determining the social cost associated with non-optimal water pricing by considering the counterfactual, where well-functioning markets for water purchases and sales exist. Assuming there are no additional distortions in the economy, the equilibrium market clearing price of water would be equalized over all producers and will be the socially optimal price. In either case, that of administratively set fees or market clearing prices 6

for water, we can assume that producers who use water as a variable input to their production processes choose how much to purchase depending on the marginal value of that input to the value of outputs. Let us now examine the difference between partial and general equilibrium approaches to evaluating the social welfare loss. Let y = y(x1,x2, . . .,xn,w) be the production function for a firm. The firm produces an output level of y as a function of n inputs, x1 to xn, and water, w. We assume a representative firm owner maximizes profits subject to their production function. (

)



where py is the price of output, pw is the price of water and pi is the price of the nth input. First order conditions for a maximum dictate that:

The change in profit when one more unit of water is purchased is the value of the marginal product (VMP), that is, how much more of product y can be produced with an additional unit of water, all else equal. This will be the same as the price of the water in a well-functioning marketplace. Where there is no well-functioning marketplace, the VMP of water will still exist from the point of view of the producer, but it will not necessarily be equal to the observed price. A general equilibrium analysis of the value of water in production will be somewhat different than the partial equilibrium analysis given above. The key difference is that all else will not be held constant in a general equilibrium approach. All prices in all markets adjust supply and demand to clear the market. The water input to production is reallocated elsewhere to another consumer or producer. Thus in a general equilibrium approach, the value of water in production is given by: (

)

In a general equilibrium approach, all prices and all quantities are allowed to adjust to a new equilibrium when any other price or quantity changes anywhere in the entire system. Other prices in the market are no longer considered to be parameters. Wherever trading is allowed, water prices will equilibrate, and all other markets that are affected through incomes, prices of competing or substitute inputs and outputs, government revenues, and investments are all potentially affected. Thus the net effect of a given change is modeled to include primary and secondary effects in all related activity in the regional economy. For example, consider the case of a drought which could change the availability of water in a given regional economy. A policy maker wishes to know what the value of water would be to a particular sector (agriculture) under these drought conditions. Widespread drought conditions could increase the market price of agricultural products. The increase in the price of agricultural 7

products could attract more labor to agricultural sectors. As the share of household income spent on food increases, the share spent on other types of goods could fall, causing firms in nonagricultural sectors to lay-off workers. Higher unemployment could change consumer demands. All of these changes in other markets for factors and goods ultimately affect the value of water in agriculture. It is in taking into account this cascade of indirect effects that a general equilibrium model is better than a partial equilibrium model. The literature includes many examples of CGE models that have been used to examine the economic consequences of alternative water projects, allocations, or prices, as well as the effects of increasing scarcity. The existing literature on water-CGE models gives examples of the types of general equilibrium effects that cannot be accounted for in partial equilibrium methods. A good example of how a CGE can identify secondary effects is described by Hassan and Thurlow (2011), who use a multi-regional CGE model of South Africa to compare water trade liberalization policies. They find that creating a water market amongst rural farmers improves the welfare of rural farmers but hurts the urban poor because the prices of cereals increase when the price of irrigation water increases, encouraging farmers to grow higher value vegetable and fruit crops rather than grains. In this example, higher water prices lead to different crop mixes, price changes for agricultural commodities, and different income effects for urban and rural poor. The types of economic problems concerning the value of water resources that lend themselves to CGE approaches tend to include the following elements: (1) the value of water as an input to one or more industrial sectors in a well-defined regional economy is a relatively high proportion of the total value of the output of those sectors, (2) those sectors are integrated into the rest of the regional economy, so that secondary effects in other markets are likely as a result of changes in sectors that rely directly on water resources, (3) the regional borders of the economy to be modeled are well defined in terms of water use, such as a hydrological basin, a watershed, a water utility district, or rivershed, (4) there is sufficient use for a water-CGE model (in developing simulation scenarios that are policy-relevant) to justify the investment in designing, developing and calibrating it. 1.4 Existing water CGE models with applications in the U.S. Relatively few water-CGE models have been built in the U.S. and used in analysis appearing in refereed academic publications. We find seven published studies, based on five CGE models that assess changes in water use and prices for U.S. applications. Berck et al. (1991) created one of the first CGE models to include water use at a time when the necessary regional input-output data was beginning to be more readily available. They modeled the economic effects of reducing the amount of water used for crop irrigation to alleviate salinization of irrigated land in California’s Central Valley, determining a shadow price for water that would be diverted from agricultural production. They evaluated this shadow price in the context of water prices in nearby urban areas, and found that urban water users could easily afford to compensate rural farmers for the marginal value product of the water taken out of irrigation agriculture. Seung et al. (1997, 1998, 2000) used a fixed ratio of water to land to model the economic consequences of water withdrawn from agriculture for various environmental purposes. They found that recreation benefits were not large enough to compensate for lost agricultural activity. 8

Goodman (2000) used a dynamic CGE model to compare economic outcomes of building an additional dam versus allowing short term water trades between agricultural water users and municipalities in southeastern Colorado, concluding that the water trades did not impoverish rural regions and would meet urban demands more cheaply. Similarly, using a model that simulated population growth and increasing water demand, Watson and Davies (2011) find that allowing short-term water trades between agricultural sectors and municipal water providers in northeastern Colorado would mean an increase of about 8% in the price of municipal water and 10% in the price of agricultural water. In contrast, a simulation of population growth without a water market predicted an increase of 25% in the price of municipal water and no increase in the price of agricultural water. Finally, Rose et al. (2011, 2005) used a CGE to model the short-term economic effects of water supply disruptions due to an earthquake in Los Angeles, California and Portland, Oregon, respectively.

2. Incorporating water into static single region CGE models 2.1 Basic concepts for CGE CGE models are simplified representations of entire economies. One approach to constructing a CGE model is through the notion of the circular flow of the economy. Figure 1 presents the core of the conceptualized circular flow in a CGE model, adapted from Ghadimi (2007). First, we start with the producers. A CGE model contains multiple producing sectors such as the agricultural sector, the manufacturing sector, the trade sector, the services sector and the utilities sector. The number of sectors (and model complexity) could vary from only two sectors to hundreds, depending on the level of aggregation of industry activity needed for a particular policy analysis. Water CGE models often include a water utility sector that captures, stores, treats and delivers water for its customers. Each industry sector is represented in the model in aggregate over all firms and therefore with a specific production function. Producers in a CGE model purchase inputs to produce commodities to sell in the product market. For example, the agricultural sector purchases fertilizer, seed, tractors, gasoline and so forth from the product market. These are called inter-industry purchases. Producers also purchase the services of factors of production. The agricultural sector, for example, purchases labor, capital and land from their owners. In the case of water CGEs, water may also be considered a factor of production. An electric utility may own water rights that permit the utility access to a given percentage of water available in a given year and watershed. The electric utility pays a “rent” for the right to use this water in the same way that an agricultural producer might pay rent for the right to use land. In a CGE model, the owners of the factors of production are called households. Households may consist of a single representative household or, if different income levels, locations, ethnicities, or other characteristics are of interest to the modeler, more than one representative household. All factor income accrues to households as the ultimate owners of the factors. To complete the circle, households spend the income that they receive for the use of the factors they own in the product market. In water CGEs one of the commodities purchased in the product market may be water from the water utility sector. 9

Three concepts from neoclassical theory link the firms and households in the circular flow and make up the core of the CGE model (Wing 2011). These are: 1. Zero profit conditions. Because of constant returns to scale and competitive markets, producers do not make a profit. Total revenue is equal to total costs. All firm revenue is used to purchase intermediate inputs or to rent the factors of production from households. 2. Market clearance conditions. The value of a firm’s output will be equal to the value of household and other firms’ intermediate purchases. That is, in each market supply is equal to demand. This will also hold in the factor market. 3. Income balance conditions. In order to maximize utility, all the income that households earn by renting out factors will be spent on purchases of commodities from the product market (we abstract from savings and taxes for government for now). While the above represents a basic description of the core of a CGE model, CGE models typically also contain representations of a government sector, investment and savings, and trade. The government sector is important to help model the lack of a market in the water factor market or the water commodity market. Governments collect taxes, consume commodities, and redistribute some taxes. In the case of some water CGEs, water ‘prices’ are specified as taxes or fees, as opposed to market clearing prices that are determined endogenously through the model, which are redistributed back to households. Investment and savings specifications become important for dynamic CGE models in order to connect savings and investment in the initial time-period with capital formation. This can be especially important for dynamic water-CGEs that consider policy questions about water supply infrastructure over time. Specification of trade flows with other regions are a standard part of CGE models and may also be important for modeling trade liberalization in conjunction with changes in the institutional structure for water rights, as well as in multi-regional models that investigate water trading between regions. The circular flow of the economy is represented in a CGE by a set of financial transactions referred to as a Social Accounting Matrix, or SAM. Table 1 illustrates an example of a SAM. The SAM uses double-entry accounts. By convention, the rows are account receipts and columns are purchases. Total outlays (purchases) are equal to total receipts for each account. In Table 1, for example, reading down the first column, we see that the agricultural sector purchased $1.3 million of agricultural commodity inputs, $400,000 of trade commodities, $700,000 from the manufacturing sector, and so on. The agricultural sector paid out $1.3 million in wages to laborers. Another $3.3 million represented returns to capital and depreciation. Total outlays were $8 million dollars, exactly matching total receipts, which are found by reading across the row. The producers are the “activity” accounts. The producers purchase inputs from other industries and pay wages and rents to the factors of production, as well as indirect business taxes, such as sales taxes, to the government. In the factors columns, labor pays $54 million in income to households as well as $7 million in payroll taxes to governments, and money is allocated from the capital account to household owners of capital, to the government for capital gains taxes and for investment. Reading down the final demands sector columns for households and government we once again close the circle, with households spending income on purchases of commodities. Households also pay direct taxes to government, invest, and transfer money to other households. 10

Figure 1: Circular Flow of Income for a Water CGE

Wages, interest capital gains, rents for water **, land.

Producers – Includes Ag, Mfg, Utility Sectors

Payments for goods and services including water*

Factor Markets

Input Purchases Including Water

Product Market

Payments to labor capital, land and self – supplied water** owners

Households

Consumption of goods and services including water*

* Water “produced” by the water utility sector ** Water owned as a factor of production, i.e. water supplied through ownership of water rights.

11

12

↓↓↓↓↓↓

Receipts

Activities

Final Demand Taxes Factors Commodities

0.7 0.7 1.3 3.3 0.4

Mfg

Services

Labor

Capital

Taxes

5.2

1.8

2.7

0.7

4 12

TOTAL

-

0

Trade 121

5.4

31.8

42.3

29.3

6.1

5.3

0.2

-

-

8

0 25

0.2

3.5

6.6

3.4

7.9

2.1

1

Services Ag

Investment

Mfg

0

30

3.4

5.7

10.8

Trade

Govt

8

0.4

Trade

Households

1.3

Ag

Ag

Services

Mfg

Trade

Ag

Activities

-

-

-

Trade

38

8

0

0 -

-

0 -

30 -

Mfg

Commodities

59

34

0

25

0

149

21

1

5

1

120

0

1

0

-

61

7

54

Taxes

44

-1

29

1

16

-

-

-

Capital Taxes

Factors

↓↓↓↓↓↓↓↓

Services Labor

Purchases

9

9

100

0

2

9

2

54

15

17

1

Households Govt

57

0

6

19

3

19

8

2

0

52

4

6

7

23

1

8

1

2

Investment Trade

Final Demand

70

0

7

0

2

37

11

7

6

8 25

30

836

70

52

57

100

9

44

61

149

59

38

12

121

TOTAL

Table 1: Example of a Social Accounting Matrix (millions of $)

Government purchases goods and services, transfers money back to households, and invests. The final column shows exports, the purchases of commodities made by parties outside of the region. 2.2 Implementation Figure 2 shows an overview of the CGE modeling process (Gillig and McCarl 2002). Computable general equilibrium models are referred to as computable because they are applied to economic data. Data for the SAM is collected and then adjusted and balanced so that total receipts are equal to total outlays for each account. The SAM data described in Table 1 represents the so-called benchmark general equilibrium. This data, along with specific assumptions regarding utility and production functions, represents one equilibrium solution of the economic model. A water CGE model will usually include a set of water accounts that accompany the SAM, which represent water use by industry and final demand sectors at the equilibrium solution.

Figure 2. Diagram of CGE Modeling Process

Since the benchmark is considered to represent an equilibrium solution, once specific functional forms are chosen, the benchmark data is used to calibrate the parameter values for the functional forms. Depending on the functional forms chosen for producers and consumers, some parameter

13

values will not be supplied by the calibration and will have to be supplied exogenously. Values are either taken from the literature or chosen using the modeler’s best judgment. After calibration, the model is checked to see if it correctly replicates the baseline data in the SAM. When it is established that the baseline data can be replicated, the model is “shocked”. For example, an increase in export demand may be imposed exogenously or a tax may be eliminated. The model is solved once again to find the “counterfactual” equilibrium set of prices and quantities for all sectors. These results can then be compared to the base solution or other counterfactual scenarios. To illustrate the type of comparative policy analysis that can be carried out with a water-CGE and what outputs from such a model look like, we elaborate results from Qureshi et al. (2012), who use the static version of the Australian multi-regional model TERM-H2O to examine how water resources and their prices are affected by an increase in the urban population and decreases in water availability. Four scenarios are modeled. The baseline year is shocked with population growth and decreasing water availability under four different policy alternatives: 1. Business as usual: no water trades between regions are allowed and no new water sources are developed. 2. Water trading between rural and urban areas. 3. Water trading is allowed and a “new” water source is built (perhaps a desalinization plant). 4. Scenario 3 is modified by allowing labor mobility between regions. For each scenario, Qureshi et al. report aggregate consumption, real gross regional product, aggregate employment for each region, water use by sector, water price (use charges) and shadow prices of water for each region. They find that without new water sources or water trading, Australian cities will face as much as an eightfold increase in the shadow price of a kilolitre of water. The “business as usual” scenario, (1) above, will result in the lowest level of aggregate consumption. Water trade between urban and rural regions will reduce production in water intensive crops as the shadow price of water increases in rural areas and decreases in urban areas to equilibrate urban and rural water prices. Providing new supplies, even after accounting for infrastructure costs, reduces the economic impact on rural areas. Because the CGE model is a representation of the entire economy, the output from the model gives a complete set of market-clearing prices and quantities in the product and factor markets. Thus almost any economic variable of interest can be compared to the baseline: GDP, employment levels by sector, aggregate consumption, water use by sector, water prices and shadow values, and more. Especially important for the water CGE model is that an explicit measure of welfare, the equivalent variation, can be calculated from the results so that the change in welfare for different simulations can be calculated. This serves as a shadow price for water in some models where simulations change the quantity of water available to the economy.

14

3. Challenges posed by CGE models generally and water-CGE models Unlike some types of econometric work, a CGE model is not a method for testing functional forms, their parameters or the structure of the economy. A CGE model assumes very specific functional forms, with parameters that are either supplied by the modeler or calibrated to the assumed base equilibrium solution represented in the SAM (Gillig and McCarl 2002). It does not test a specification, rather it is the specification. The results from a CGE model reflect all the assumptions used to build it. If results confirm the assumptions of a neoclassical economy, it is because the modeler built the CGE to be a neoclassical economy, not a confirmation that the real world works that way. Because of their complexity, CGE models in general have a reputation for being a “black box” (Ian Sue Wing, 2011). The assumptions made within the CGE model structure can be quite powerful in determining results, and not all economists are familiar enough with the models to feel that they understand how results are influenced by them. CGE models take more time, care and skill to specify and interpret than do some partial equilibrium methods. CGE models work because they assume a simplified version of a neoclassical economy that has a point of equilibrium where the right price creates market clearing. Special techniques are required when it is known, as is often the case with water, that a resource’s observed prices do not stem from a well-functioning market. It is a challenge to estimate a starting ‘market’ price for the initial baseline equilibrium which is used to calibrate the model. It can also require a different technique to model the water factor or sector if one would like to assume that there is not a wellfunctioning market that generates market-clearing prices and quantities. We observe a variety of methods for tackling these difficulties: 

  

Estimation of water ‘rent’ in various ways, often using land values. The rent is then subtracted from gross operating surplus and distributed to households. This technique assumes a functioning water factor market (for example, Robinson and Gehlhar, 1995, TERM-H2O models described in Wittwer 2012, Seung et al. 1997, etc.). Assumption that no market for water exists in the baseline, that water is in surplus and its price in equilibrium is zero, becoming positive only as water supplies are withdrawn (Examples are Berritella et al. 2007, 2008, Diao and Roe 2005, Roson et al. 2010). In one case, water ‘rent’ is subtracted from utility fees charged to industries (Hassan and Thurlow 2011). Some models use administratively set utility fees for treated water as if they were determined by a market equilibrium.

3.1 Water Accounts Data Building a standard CGE model is an extremely data intensive enterprise, requiring detailed baseline data for all parts of the economy. Building into a CGE model the ability to address changes in water resources usually requires additional data that links economic sectors with their water use. Wittwer (2012) refers to this data as “water accounts.” Finding sources for water accounts data can be a challenge. Our current version of the Nevada Water CGE model represents a minimum level of water data, i.e. estimates of water use per dollar of output by 15

sector, household water use, and total water use for the Nevada economy as a whole. For many policy questions, however, a greater effort collecting and organizing this data is necessary. The development and availability of water accounts data is major reason why Morocco, South Africa and Australia have an abundance of water CGE models.1 Unfortunately, the United States faces a lack of water accounts data. The existing U.S. water CGE models used a variety of unique datasets on water resources, allocation and usage. Berck et al. and the Seung et al. series of water CGE models consider only agricultural water use (Berck et al. 1991; Seung et al. 1998; Seung et al. 1997; Seung et al. 2000; Seung et al. 1998; Seung 1999; Seung et al. 1997). The agricultural sectors use a fixed proportion of land to water. Water resource and allocation data was derived from the California Dept. of Water Resources in the case of the Berck et al. study and from related federally sponsored environmental impact studies in the case of the Seung et al. series. Goodman and Watson and Davies both consider water use by municipal users as well as agricultural users (Goodman, 2000; Watson and Davies 2011). In those studies, water accounts data is derived from the Colorado Division of Water Resources, a state research institute, the USGS, and a local water board. The amount of water used by each industry sector is typically used to estimate a water intensity factor that gives the amount of water necessary to produce a unit of output. Depending on the purposes of the model, water use by households and government, total water availability in the region, and other more detailed water data may be necessary. A majority of the countries of the world do have water accounts or are planning to collect data for them but, unfortunately, the United States is not one of these countries (Wittwer 2012). Some estimates are available from the United States Geological Survey at a very aggregated level. Blackhurst et al. (2010) used this data to estimate water intensity factors for all BEA industry sectors (Blackhurst et al. 2010a, 2010b). Fadali extended this estimation to the state level (Fadali 2012). County level estimates of water withdrawals and consumption have been attempted for some agricultural sectors (Mubako 2011). The challenge is to match the water use with the appropriate industry, as defined by the North American Industry Classification System. Rollins and Stoddard have data by NAICS code which we plan to use to refine the Blackhurst estimations. Financial flows data is more readily available. Data requirements for building water CGE models in many situations can be met by use of county level data, or aggregations of selected counties. In the literature, county-level data has been shown to have a sufficiently fine level of geography for water-CGE models, as evidenced by existing U.S. water CGE models (Berck et al. 1991; Goodman 2000; Seung et al, 1998; Watson and Davies 2011). In the United States, county or smaller sub-county level SAMs are estimated and readily available for purchase from Minnesota IMPLAN Group (Minnesota IMPLAN Group 2010). The estimates for these financial transaction flows are made using data from various government agencies such as the Bureau of Economic Analysis’ National Income and Product Accounts, U.S. 1

Australian water CGE studies include Dixon, Rimmer and Wittwer (2011), Dixon (1990), Horridge, Dixon and Rimmer (1993), Horridge, Madden and Wittwer (2005), Peterson et al (2005), Qureshi et al (2012), Wittwer and Griffith (2012), Wittwer (2011; 2009; 2006). South African water CGE studies include Hassan and Thurlow (2011), Hassan et al (2008), Letsoalo et al (2007), Mukherjee (1996), van Heerden, Blignaut and Horridge (2008). Moroccan studies include Diao and Roe (2000a; 2000b), Diao, Roe and Doukkali (2005), Hassan and Thurlow (2011), Hassan et al (2008), Tsur et al (2004).

16

Census Bureau’s Survey of Governments and Economic Census, National Agricultural Statistics Services’ Census of Agriculture, and many more (Minnesota IMPLAN Group 2010). CGE modelers may also use such data directly to build a social accounting matrix. 3.2 Sector Aggregation in Water CGE Models The financial flows data in the SAM interact with the water accounts data. One issue that arises is the aggregation level of the financial flows accounts. For example, in the SAM in Table 1 there is only one agricultural sector. This would usually be too grossly aggregated for water CGE models used to find the economic value of water. For example, hay, beef cattle and dairy are the three largest agricultural sectors in Nevada (National Agricultural Statistics Service 2009), each of which have different water intensity factors and different production functions and, where institutions prevent trading, potentially different marginal values for water. The current Nevada water CGE disaggregates these sectors. Because of the complexity and size of the non-linear system involved, the CGE modeler usually must choose to aggregate many industry sectors. This is an important limitation, since water use may vary widely even within a detailed disaggregation of industry sectors. Avoiding aggregation error may mean greater customization of models for a given region, so that the largest water users in a region are adequately described. For example, normally ranches and feedlots are included in the same sector. Watson and Davies (2011) find that they must split this sector into two parts for their model of northeastern Colorado, since “these two operations use land and water in very different ways and interface differently with the regional economy” (Watson and Davies 2011, p. 346). In some cases, especially where the economic impact of a water policy is suspected to have differential effects on different types of households (i.e., rural or urban, low income or high income), two or more representative households will be represented in the SAM data rather than one aggregated representative household. Sector or household disaggregation can help a modeler specify separate water markets with different prices. Watson and Davies (2011) include representative rural and urban households. Rural households own water rights and are compensated for the water rights by tax payments from urban households in the water trade simulations. Rural households receive payments for labor and capital within the agricultural sectors and as such are the ‘losers’ when agricultural activity declines due to lack of water availability. Such a model has some of the advantages of a multi-regional model without requiring a full set of trade flows. 3.3 Defining a ‘region’ for the purposes of a water CGE In defining the regional boundaries for a CGE model with an analytical focus on the market interlinkages of water, a key feature is the area within which water resources can be feasibly traded. To illustrate this, observe the way that trading zones were defined by Watermove, the Australian water market exchange. Amongst water users in a trading zone, both temporary and permanent water trades take place through this water exchange (Schreider 2009). Trading zones are the areas defined by infrastructure and topography that make a physical water trade possible. Within the zones, using actual bids from buyers and sellers, Watermove determines a single price for short term water trades on a weekly basis. If a major focus of a water-CGE model is to find an economic value for water, then in most cases the region of the model must mirror these 17

Australian water-trading zones. Thus, depending on the circumstances, the boundaries may be defined as including all economic activity within the same hydrological basin or that which is connected through pipelines. By definition, water from outside this boundary does not enter the regional economy. Outside of these boundaries, water supplied from different sources is simply not available for the target CGE economy. Therefore, the marginal value of water in separate economic systems, described by separate CGE models, would not be expected to be the same, because by definition there is no trade, and therefore market-clearing conditions don’t hold. In many cases, however, the possibility of bringing water from one watershed into another may be the focus of policy analysis. For example, California governor Jerry Brown has announced plans to build two new pipelines which may increase the amount of water exported from northern California to central and southern California (Green 2012). Decreases in the differentials between marginal values of water between the two regions (due to increased local demands, technological innovations that reduce the cost of conveyance between the regions, or a one-time exogenous subsidy that reduces the fixed cost of constructing the infrastructure to move water) would bring the possibility of increased trade in water between them. As the marginal cost of using water from outside a region declines sufficiently, trade between the regions will induce water to flow (uphill, if necessary) from regions with lower marginal values to regions with higher marginal values until a new equilibrium price of water is achieved. This can be modeled through linked water-CGE models, described in a later section. In order to analyze the economic impacts and net welfare changes from interbasin water transfers, the individual regions must first be carefully defined. As an illustrative example, we consider the case of Southern Nevada, where the greater Las Vegas area that includes Clark County has been dependent on a relatively fixed water supply through the Colorado River agreement. Simplifying details of the situation for our purposes, the demand for water has increased in Clark County to the point that water was seen as a fixed input to production that would limit future growth in the regional economy. All other economic activities, either directly or indirectly, would eventually be constrained by the fixed water supply. This is precisely the type of analysis that CGE models allow an analyst to consider – the total set of interlinkages through which a change in any one sector would impact other sectors. Eventually, should the marginal value of water become high enough, the potential to increase supply through purchases of water inputs from regions outside of Clark County arises. In the case of Nevada such a transfer is possible, and state engineers are evaluating the cost of building a pipeline to convey water to Clark County from rural areas of White Pine and Lincoln Counties. A separate CGE model for the region defined as White Pine and Lincoln Counties combined, the likely suppliers, can be built and linked to the Clark County model to determine how interbasin trade would eliminate the price differential and affect net regional production and income in each regional economy. Thus, for policy-relevant questions related to water values, an approach based on separate regional models that represent the supply of water within each region and the feasibility of moving water between them makes good sense. In contrast, a Nevada-wide water-CGE model that aggregates the entire state economy is not consistent with the physical boundaries and limits to water use. A statewide CGE model could feature water as an input and source of income, but it would give results that average economic values of water for Nevada as a whole, and the important policy-relevant features of water 18

transfers between individual regions would be lost. For this particular example, if we consider the state of Nevada as a whole, there would be a handful of regions that are defined by current separations in water supplies but with potential for inter-basin trade. A ‘state-wide’ set of linked multiregional water CGE models would be more appropriate than a single statewide model for analysis of the value of water as it flows through regional economies, where regional differentials in water values are a target of policy analysis, as much as the total value of water to the state economy. This example also indicates that a statewide single model could create other problems. For at least two reasons, the supply of water through the proposed southern Nevada pipeline could affect counties in neighboring states. The first is that the pipeline can draw water from Utah counties. Second, as water prices decline in the region receiving water, the marginal cost of water from all sources will eventually equalize, altering the amounts of water drawn from existing sources, such as the Colorado River. Levels in Lake Mead could be affected, as well as opportunity costs for the same source of water throughout the states that are signatories to the Colorado River agreement. This is purely speculative and possibly unlikely, but we raise these issues to illustrate that the purpose of a CGE model is to capture linkages, and therefore the definitions of the boundaries of such models must be carefully considered to fit the nature of the economic problems being evaluated. In the case of water, these boundaries are defined by both physical and economic circumstances The literature includes multi-regional water CGE models that allow for separate water markets and trade flows between regions, with the most prominent being the Australian model TERMH2O. In one application,TERM-H2O was used to isolate the effects of a water buyback program, carried out for environmental reasons, from the effects of prolonged drought in Australia’s Murray-Darling Basin (G. Wittwer and M. Griffith, 2011). Wittwer and Griffith find that drought is responsible for most of the unemployment related to lack of water. The multi-regional model allows both for trading of water between regions and trading of farm labor across regions, as well as modeling of downstream effects on food processing sectors in separate regions. The only water-CGE published articles we are aware of that use a region larger than watershed level as a base are the GTAP-W model, one model of South Africa, and our own Nevada Water CGE model (Letsoalo et al. 2007; Berrittella et al. 2007; Berrittella et al. 2008; Calzadilla et al. 2010, 2011; Fadali 2012; Roson and Sartori 2010). 2 All these papers except for Letsoalo et al. investigate the so-called ‘virtual water’ embodied in trade flows across larger regions. They show how welfare, water use, and balance of trade could change under scenarios of increased water prices, increased irrigation efficiency, or situations of water scarcity at a world or multicountry level. Fadali (2012) studied the change in virtual water trade between the state of Nevada and its trading partners. Letsoalo et al. focus on which type of federal water tax will bring about a “triple dividend”: reduced water use, reduced tax distortion by using water tax revenues in place of other tax revenues, and an increase in incomes in low income households. In this application, water is not sold in a market – price is set by government policy and the marginal value product of water is not at issue.

4. Multi-regional CGE Water Models. 2

GTAP-W is a variant of the Global Trade Analysis Project world CGE model of Purdue University especially adapted to include water use at the country level. It includes water accounts for agricultural sectors only.

19

As discussed above, depending on topography and geographic variation, water supplies available to individual regional economies can differ dramatically from watershed to watershed within fairly small geographic areas, and within different regions within a state. For example, supplies in southern Nevada depend almost completely on the Colorado River, where drought and overallocation have made the overall water situation precarious. This is in marked contrast to northern Nevada, where water supplies from snow pack run-off support an agricultural sector that represents a low-priced water supply reserve potentially available for future growth. Regionspecific demands for water resources also vary widely depending on annual precipitation, average temperature, industry mix and technologies, and consumer preferences. Region-specific shocks to water resource supply or demand can also vary immensely by region. Given differences between regions, many water policy questions between economically linked regions can better be answered with a multi-regional model that incorporates regions closely aligned to the watershed level. This is particularly true when a question arises about infrastructure which can link two or more previously separate watershed regions. Multi-regional models allow for separate region-specific supply issues, prices, local government policies, and production functions, as well as imperfect factor mobility between regions (Wittwer 2012). Putting the regions together into a multi-regional model can illustrate inter-regional feedbacks while allowing for greater realism, not only in modeling markets for the exchange of water between the regions, but also in modeling the flows of both factors and commodities that interact with water market trade. The majority of the water-CGE models reviewed in this paper are multi-regional models. The Australian TERM H2O is an example of an inter-regional water CGE model, as is the GTAP-W model, which is a multi-country model of the world. Many of the South African water-CGE models are also multi-regional. Several Moroccan models link a farm level micro model to a national level CGE, approximating a multi-regional model. We are not aware of any U.S. multiregional water CGE models.3 4.1 Additional Modeling and Data for Multi-regional Water-CGEs Multi-regional CGE models add source and destination subscripts to variables and equations (see for example the discussion starting on page 15 in Wittwer 2012). This means, for example, that a production function for beef could be specified differently in each region. If factors are not mobile across regions, different market clearing prices are possible, including water. Final demand sectors, factor supply and demand may all be specific to each separate region. Data describing the economic and water linkages between the regions are required for the multiregional SAM. For example, if we consider two regions, a set of trade matrices consisting of a two by two matrix for each type of commodity in the model is needed. Supposing that Clark and White Pine Counties in Nevada are a closed system, Table 2 illustrates a trade flow matrix for the commodity hay. Reading across the row, the destinations for White Pine hay production are given, with $4600 of the total $71,600 production consumed in White Pine County and $67,000 3

However, the originators of TERM-H2O have created a multi-regional CGE model of the U.S. with all fifty states that does not contain water accounts (Dixon, Rimmer and Wittwer, 2012). See "Usage-R51, a State-Level MultiRegional CGE Model of the US Economy," GTAP, www.gtap.agecon.purdue.edu/resources/download/5933.pdf.

20

of the total production consumed in Clark County. Reading down the columns, the origins of White Pine’s $7400 of hay consumption are given. White Pine produces $4600 of its own hay and consumes another $2800 of hay from Clark County.

Table 2: Additional trade data for 2 region model Hay origin White Pine Clark Total demand

White Pine $4,600 $ 2,800 $ 7,400

Hay destination Clark $67,000 $56,000 $ 123,000

Total supply $ 71,600 $ 58,800 $ 130,400

Data on how much water can flow between the regions also must be incorporated into a multiregional model. In practice almost all CGE models would also account for “rest of the world” imports and exports. Trade flow data amongst regions inside a country is usually sparse, and such is the case in the United States. Considerable effort has been directed at making reasonable estimates of these flows for use with inter-regional models. In the U.S., IMPLAN data now routinely comes with estimated trade flows between regions (Minnesota IMPLAN Group, 2009). Minnesota IMPLAN Group uses commodity trade flow data from the Bureau of Transportation Statistics and the U.S. Census Bureau, travel cost data from Oak Ridge National Labs and a doubly constrained gravity model in conjunction with IMPLAN estimates of commodity supply and demand to estimate these trade flows. The constraints ensure that the gravity model gives back “known” results for regional demand and supply that exist in IMPLAN. The accounting system in IMPLAN ensures that the regional supplies and demands add up to the national supplies and demands and that national supply is equal to national demand (Scott Lindall et al., 2006). TERM H2O also uses a gravity model approach to estimate trade flows between Australian regions (G. Wittwer, 2012). 4.2 Uses for the Multi-regional Water CGE There are many advantages to using a multi-regional water-CGE model. The model can show the distribution of economic impacts across space for a given water policy change. In the case of the TERM model, which is a model of Australia with all its sub-regions, regional impacts of a federal spending project can be modeled at the national level that spending is accounted for. Geographic realism allows better analysis of environmental issues and conforms with the watershed region. There is high demand from regional legislators and local officials to know fine regional level impacts of policy shocks; multi-regional models provide these capabilities.4 The disadvantages of a multi-regional model are primarily the large amount of extra data needed and the heavy computational burden of a model with many regions. The TERM-H2O creators have developed a system for automatically aggregating sectors and regions not of interest in the policy issue at hand. Both GTAP and TERM-H2O use a variety of simplifying assumptions to reduce computational burdens (Wittwer 2012).

4

P. 15 Ibid.

21

5. Dynamic CGE models All CGE models implicitly incorporate time, in that an adjustment process takes place until markets clear (Ghadimi, 2007). Models may be designed to represent short-run changes, in which capital or other factors are not mobile between sectors or regions, or long-run changes that assume full mobility of factors. Many water-CGE models incorporate a more explicit time element in order to observe the temporal effects of a policy adjustment. Many water resource policies are enacted over time and therefore have consequences that unfold over multiple time periods, or questions related to capital stock accumulation. One such important water policy issue is to determine the appropriate amount of investment in supply infrastructure such as dams, pipelines, or desalinization plants. This is an investment and capital stock decision that relates to the economic value of the water that is to be moved and supplied over time. Thus water-CGE models that focus in part or in total on infrastructure supply generally also incorporate time. Examples include: 





Dixon (1990) determines the optimal price for an urban water authority in Melbourne, Australia. To do this, he considers the appropriate level of investment in, as well as timing for, infrastructure related to water capture, storage, treatment, delivery and sewer service as the population increases (Dixon 1990). Goodman (2000) compares the installation of a new dam to meet urban demand, which grows with population over time, to temporary market trades of water between agricultural sectors and municipal water suppliers. Consumers in this model maximize utility over time, while precipitation is allowed to vary from year to year (Goodman, 2000). Wittwer (2009) also conducts an analysis of the appropriate level of supply infrastructure to meet urban water demand in Queensland, Australia. In the TERM-DYN model interregional wage differences cause labor migration over several time periods. Labor supplies and population may therefore change over time depending on the relative economic conditions between regions. This affects the optimal supply of water infrastructure. Precipitation levels may be specified to rise or fall with time, while technical change can increase water use efficiency over time. Capital and labor used to build infrastructure is withdrawn from other activities over several time periods (Wittwer 2009).

Other water policy issues that involve multi-period implications include drought, because of its long-term implications for livestock herds and for perennial crops such as fruit and nut trees (Wittwer and Griffith 2011); climate change scenarios that play out over decades; or policy changes that take time to institute, such as buybacks of water rights for environmental purposes (Seung et al. 2000, Wittwer 2011). Ghadimi (2007) describes two basic types of dynamic CGE models:

22

1. recursive models that solve for a static equilibrium, update time-related variables and solve for the next time period equilibrium in sequence for the required number of time steps, and 2. models that incorporate inter-temporal optimizing behaviors based on expectations. Most water CGEs use the recursive technique. Dixon (1990) and Goodman (2000), however, specify consumers who solve an inter-temporal utility optimization problem. Dynamic models such as TERM-H2O use three different types of adjustments to update variables to a new time period: capital and financial asset/liability accumulation, and lagged adjustments to factor supplies.5 Changes in factor supplies, such as the effect of population growth on the supply of labor, may be imposed exogenously. Changes in water supplies may also be imposed exogenously over a period of time. A baseline for a dynamic model is usually constructed using forecasts of economic growth, population growth, or water availability.6 This baseline model will then be compared with the same model subjected to policy or other exogenous shocks.

6. Example: A CGE Model for Nevada 6.1 Main Features of NV Water CGE Model for Valuation of Water as an Input to Industry The Nevada water CGE model (NVWCGE) was originally built to find the state of Nevada’s water footprint and the response of the footprint to changing policies (Fadali 2012). We are adapting the model to use for estimating the VMP of water as an input to production. The process of adapting the model has allowed us to analyze the current capabilities of the waterCGE, in general, in this capacity, and what is still needed to develop meaningful estimates. Below we discuss the most pertinent features of the more ideal NV Water CGE and how they relate to the task of finding water price or shadow prices, ending with a section on future work. 6.2 Choice of Region As discussed previously, most water-CGE models are built at the level of the watershed, or are a multi-regional model. This is especially so for models used to find water prices, since prices may vary considerably by region. Making a multi-region model would be ideal because of its ability to handle multiple policy scenarios across the state, but it would require considerably more effort. As an intermediate step a new Clark County model has been created that is closer to a watershed level region, but new county level water accounts must still be developed (currently we have only Nevada state level accounts). A process for creating these is suggested in Appendix B. The detailed water use dataset we have for covering urban areas of Nevada would help test the non-agricultural water intensities for accuracy.

5 6

P.38 Ibid. P.45 Ibid.

23

6.3 Customization of SAM data for Water CGE. The state of Nevada SAM used in the NVWCGE was purchased from IMPLAN, as was the Clark County SAM. It is easy to carry out modifications of the SAM in IMPLAN to customize the data. There are many reasons that the water-CGE requires customization. In general, IMPLAN estimates are based on national level production functions and industries and factors related to the BEA data that IMPLAN uses as the base of the SAM. Two prominent and typical customizations for water CGEs are the disaggregation of returns to land and water from a gross operating surplus account (Other Type Property Income) and further disaggregating and customizing of important water-using industries using regionally available data. The water and sewer account has already been customized. Future work includes customization of the hay, cattle and dairy sectors, the most important agricultural sectors for Nevada. Another important question for water-CGE models is the choice of industry sector aggregation. IMPLAN SAMs include 440 sectors. Although disaggregation is preferable, the highly nonlinear nature of CGE models can make solving that big a system quite difficult. Typically CGE modelers must choose to aggregate many sectors together. The aggregation should be chosen so that water using industries that have similar water intensity levels and elasticities are grouped together. In Nevada, the most important activities with regard to self-supplied water use are agriculture, mining, electrical power generation and the water/sewer utility sector itself. For treated water, the hotel-casino sector is the most important commercial water user. Other major water users include construction, real estate, golf courses, parks, hospitals, food services and schools. The NVWCGE currently has an aggregation to 20 sectors chosen on that basis: hay, livestock, dairy, vegetable and melon farming, other agriculture, metal mining, other mining, electricity, transportation and information, water, residential construction, all other construction, finance and real estate, food processing, manufacturing, trade, healthcare, services, recreation (including casino hotels) and food and drinking places. Sector aggregation was chosen for several reasons: 1. Type of water use: treated water from the water utility sector is specified in a different way than is self-supplied water. 2. Nevada’s agricultural sectors, electric utilities, mining and the water and sewer utilities sector remain disaggregated so that the sectors which use large amounts of self-supplied water are adequately described. Other sectors purchase treated water from the water and sewer utility within the input-output intermediate demands block.7 3. Sectors were grouped together with other sectors that had similar water use intensity.

7

It would be preferable to split water and sewer into two sectors or, as in some water-CGE models, account for two different commodities: indoor water, which is more expensive since it requires collection and treatment after use, and outdoor water use, which usually is not treated in any way after use. This is an important topic for future research for any water CGE with a substantial focus on commercial and industrial water demands. One of the constraints, for example, in the Reno/Sparks urban area, is that sewer water is subject to very stringent restrictions on pollutant load. The pollutants are very expensive to remove.

24

4. Suspected variation in responsiveness to price changes in water was considered. For example, businesses with outdoor landscaping might be more responsive to price changes than those that use water indoors primarily for production. This implies a different price elasticity, which in turn implies a different elasticity of substitution for water and other inputs. The current version of the model does not allow for price-responsiveness in water use in any sector since it is specified as a fixed ratio to output. The next model version will have a constant elasticity of substitution (CES) functional form specified for both treated water and self-supplied water to allow substitution possibilities and responsiveness to price. Elasticities of substitution will be chosen from the existing water-CGE literature. Substitution elasticities imply a price elasticity of demand for water inputs. Using the detailed dataset we have for water use in Nevada, we hope to be able to calculate price elasticities of demand by industry. These can in turn be used to calculate a substitution elasticity. Analytic results based on elasticities obtained from the literature can be compared to results using elasticities derived from regional data to determine whether the extra time and effort measuring more elasticities more precisely would be worthwhile. 6.4 Model Base The model is based on the regional CGE model developed at Washington State University (WSU) and described in Stodick (2007). The model was developed for use with IMPLAN data and is available free on the Washington State University School of Economic Science website. The term “regional” means that the model is sub-national, that is, it has trade accounts with the rest of the U.S. as well as with the rest of the world. The program is coded in GAMS software and uses the PATH solver. 6.5 Production/Output Because self-supplied water and treated water are two different types of inputs, they are specified as such. In the current version of NVWCGE, treated water enters the production function as one of the commodities in the intermediate demands aggregate. The assumption is that treated water is used in fixed proportions with other intermediate demands and then combined in another Leontief nest with the labor and capital aggregate. Because we hypothesize that treated water use in the commercial and industrial sector does have substitutes, the next version of the NVWCGE will take treated water out of the aggregate and allow it to combine with the Leontief aggregate of intermediate goods and the value added aggregate within a CES nest (see Figure 3). The current specification of the water utility sector assumes that the water commodity is sold in a market; however, in Clark County and elsewhere in Nevada, most water utilities are operated by government agencies. This may or may not be the preferred technique, depending on the policy issue at hand. In many cases, one could assume that a market will not exist in the water utility sector and that prices will be set exogenously, or by some rule other than profit maximization. A second issue is, if a market assumption is appropriate, whether or not the baseline price, an administratively set price, produces a biased result in counterfactual scenarios, given an

25

assumption of a market going forward. Similar issues remain unresolved for the water factor market and likely are best decided with a specific policy analysis. The original WSU model specifies capital and labor as the primary factor inputs. The Nevada Water-CGE model added self-supplied water as one of the primary factors of production in a Leontief nest with labor and capital. The next model will use the production function outlined in Figure 3 for all sectors. Since in Nevada there is very little non-irrigated agriculture, land is not considered a substitute for water. In essence we assume that land use is not constrained since land without water is of very low value in Nevada.8 This is not precisely true in and around urban areas, so if data on land rents can be summarized by NAICS code, this is another avenue for future work. At the lowest level of the nested production function, capital and water are combined in a CES aggregate. The CES aggregate nest allows the elasticity of substitution between water and capital to be different than the elasticity used for the next higher level of the nest, which combines the capital/water aggregate with labor to make a value-added aggregate. Value-added is combined in fixed proportions with an intermediate input aggregate of all non-primary factors to produce output. These intermediate inputs are combined in fixed proportions. The non-primary factors

Figure 3: Production Function for Nevada Water-CGE Model

8

Evidence of land’s low value may be corroborated by the fact that about 85% of the land in Nevada is owned by the Federal Government.

26

may be produced inside the region or outside. Imported materials are combined with domestically produced materials in a CES aggregate. This specification allows imports and domestically produced goods and services to be imperfect substitutes for each other. This is referred to as the Armington assumption, and allows for the so-called cross-hauling observed in trade data (i.e., where the same good is both imported and exported). The imports are an aggregate in turn of different origins. Imports may either originate in the rest of the United States (RUS) or the rest of the world (ROW). The two different origins make the WSU model a regional, rather than national, model. 6.6 Consumption There are three representative households in the model: low, average and high income. Consumer income is allocated to households as the owners of the primary factors of labor, capital and water; consumer income is also allocated to households from government transfers. A CES utility function is specified. The Armington aggregate is again used to model the way households combine imports and regional goods within the same sector, again with a two-level nest that allows for choice between domestic and foreign imports before combining the aggregate of both types of imports with locally produced goods. Consumers are assumed to use only treated water purchased from the water utility sector.9 6.7 Government and Investment Two government sectors are part of final demand in the model: federal government and state/local government combined. The treated water supply commodity is produced in part by local government. Local government collects fees for these services and spends them in a general government expenditure vector. The model is static. Investment is specified as a single fixed proportion vector of purchases. Savings is investment-driven for this experiment. This fixes the savings adjustment variable and CPI and allows investment value and quantity to adjust. 6.8 Factor Markets Factor market closures are chosen to reflect different time periods, depending on the type of policy being analyzed. The WSU model closures are easily switched so that multiple experiments can be carried out with different assumptions about time periods. For a long-term policy analysis, mobility of capital between both sectors and other regions is more likely. Over the short-term, labor is usually still mobile between most sectors, while capital is typically less mobile. Both capital and labor have fixed total supplies. Over the long-term, migration may allow for growth or reduction of both of these factors where returns do not keep pace with other regions. Untreated water is also included as a factor of production. Water is fixed in supply and not mobile across regions. Base water endowment is assumed to be equivalent to USGS estimates of 9

A relatively small proportion of the domestic household water supply is from self-supplied wells or ditches. Selfsupplied water is allocated proportionally to the three representative households using purchases of treated water.

27

water withdrawals in Nevada. We will allow for two different closures concerning water mobility between sectors. In one model, water is allowed to be mobile across all self-supplying sectors. This would normally yield one equilibrium price for water across all sectors. However, in order to allow for the difference in urban and rural water, a distortion factor is specified that allows for different prices in different sectors. In a second model we will assume water is not mobile across sectors, which yields a different shadow price for each self-supplying sector. 6.9 Parameterization and Trade Closures Most of the parameters follow from the assumption that the base SAM represents an equilibrium solution. The remaining parameters are: 1. The elasticity of substitution between labor and capital is set at 0.9 for all industry sectors. It implies a relatively gentle curve, close to a straight line, for relative ease of factor substitution. Self-supplied water is currently specified as a factor that combines in a Leontief production function with the capital labor aggregate and the intermediate input aggregate. In future it is planned to specify as in Figure 3. 2. Income elasticity is set to one for all goods and households. 3. A parameter that controls the subsistence level of consumption for all commodities is currently set to -1, which implies that there is no subsistence level of commodity required. 4. CES and CET functions for imports and exports were given a moderately elastic value of two for elasticity of substitution (transformation). This implies that imports (exports) may substitute (transform) relatively easily for locally produced goods and vice versa. 5. In future work, a return flows parameter will be added for each sector. Trade closures are chosen to reflect a small country assumption. Small countries face a fixed price for their exports and imports. Thus, export and import prices are exogenous. The exchange rate is also exogenous. Balances for both foreign and domestic trade are allowed to adjust. 6.10 Future Work A first step towards a multi-regional state of Nevada model would be to develop county level water intensity factors that can be used to create sub-state regional models closely aligned with watershed regions. Water intensity factors for industrial and commercial sectors served by water utilities can eventually be tested and compared to more detailed Nevada specific water intensity factors found using the Rollins-Stoddard dataset from Nevada water utilities. Later, with the additional trade flow data, these could be the basis of a multi-regional model. For analysis of interbasin transfers such as the construction of a water pipeline from northern Nevada to southern Nevada, additional data regarding the capacity, costs and timing of the pipeline would be collected. New production functions will be specified for self-supplied water and treated water. Treated water will have a CES aggregation with value added and other inputs to allow for a substitution response to price. Similarly, untreated water will be combined with capital in a CES production function nest to allow for substitution possibilities. As a part of this change in specification, a 28

baseline value for water rent is needed. To find this, NAICS specific rents for land, or land and water combined, must be estimated from property sales data. The CES production functions and current assumptions about the elasticity of substitution imply values for the price elasticity of demand for water as an input. These can eventually be compared with elasticities found using partial equilibrium econometric methods on the Rollins-Stoddard Nevada water utility dataset. Several other data issues remain to be addressed with future work. Currently, IMPLAN and underlying BEA data assume a combined sewer and the water treatment industry and/or commodity. This is not ideal for a water-CGE, since sewer costs may be very different across industries. Addressing this data concern would also allow for better modeling of water quality issues. It is not known what kind of data might be available to split these sectors. Also, irrigation water fees (raw water delivered on-site) currently are not separated from treated water fees in the water utility sector. This may cause distortions in our assumptions for baseline prices and quantities of water use.

7. Discussion and Conclusions In Nevada, a pipeline connecting rural northeastern Nevada with the Las Vegas region is under consideration as a way to alleviate constraints on water supplies to southern Nevada, which currently relies on the multistate Colorado River Agreement and management of Lake Mead. In this situation, a dynamic, multi-regional model containing the urban and rural regions could be used to model what would happen to water values in the two areas. A full CGE model that would allow exploration of potential outcomes of various “with and without pipeline” scenarios for regional economies would ideally include expectations of changes in land value in the rural area, costs of the capital infrastructure as it is built and paid for over time, prices as they equilibrate under different water withdrawal scenarios, different water supply situations on the Colorado river, and different economic situations in Las Vegas (recession or continuation of population growth). The model could then be used repeatedly to examine many different scenarios and explore under which circumstances building the pipeline would make sense. In addition, such a model would be capable of measuring changes in welfare, as well as impacts on different industrial sectors, government, and different income groups. The example we use demonstrates that in Nevada, like in many other situations, the economic value of water to a regional economy is a moving target. There isn’t just one value. The value is constantly in flux due to shifts in water supply, demand, changes in input and factor prices, output prices, incomes, government policies and institutions. CGE models help us to understand these movements and the reasons for them. Water-CGE models are particularly good for exploring the effects that changes in water use attributable to out-of-sample scenarios, such as global warming, would have as they work through the economy. This is because CGE modeling builds on structural assumptions about how an economy works generally. CGE modeling may be the only way to understand important indirect effects, which can affect the value of water in production and are important when non-marginal changes are under consideration. Factor prices, input prices, output prices, household incomes, and government taxes in a CGE model are all allowed to interact. A CGE model can help determine winners and losers as a result of specific 29

changes, and provide measures of welfare changes induced by policy changes. For agriculture, which makes up 75% of total water consumption in the U.S., water-CGE models are relatively well-developed for use in finding economic values for water as an input to production. Municipal water values have also been modeled with CGE techniques, usually finding one value for all uses in an urban region, but specific applications to industries other than agriculture are rare and not well developed. The advantages of using CGE modeling approaches should be weighed against the disadvantages and challenges of building a CGE model. Water accounts data in the U.S. has not been systematically developed. Much more work on data development must occur if water-CGE models are to become routine tools in the United States. However, this should not be seen as a show-stopper. Leontief proceeded with input-output techniques long before good data had been developed for his models. Ways to estimate water data can be developed and tested with more detailed data from individual projects. Water CGE models usually require benchmark values for water prices, water rents and elasticities. These can come from partial equilibrium studies. Water CGE models are not suitable for estimating these initial values, but rather for exploring how values change. An important limitation to the use of water CGE models is that they are a data- and time intensive technique that requires a high level of skill. In addition, not all economists are familiar enough with the models to be good consumers of the results. Strong assumptions can drive the results of CGE models and, without awareness of how the assumptions work, a CGE model has a tendency to become a black box. The CGE modeler should be aware of the structure of the model and the underlying assumptions about how the modeled economy works. Results do not prove this structure but stem from it.

30

Appendix A Water-CGE Literature Review A review of all types of water CGE models that appear in the literature, published and unpublished, was conducted by searching three databases: Econlit, Web of Knowledge and Google Scholar. Publications in this area have multiplied in recent years. From 1990 to 1997 we found only eight papers, compared to 38 papers that appeared from 2008 to 2012 (see Figure 4). The adoption of CGE models to model greenhouse gas emissions may be increasing the technical skills and data availability for other types of environmental CGE models, including those that specialize in water resources.

Figure 4. Rough count of all types of water-CGE models, published and unpublished by time period. Water CGE papers - all types

Number of papers

40 35 30 25 20 15 10 5 0

38

23

8

1990-1997

11

1998-2002

2003-2007

2008-2012

Time Period

We distinguish water-CGEs from other CGE models either by their inclusion of a water satellite account or, in a few cases, by their focus on a water-related issue, such as drought, with a clever specification that avoids the need for a water satellite account. At minimum, a water satellite account specifies how much water, in physical units such as acre-feet, is used by producers and consumers in the baseline scenario. The accounts will also specify how much total water is available for production in the region under consideration. Sometimes much more hydrological detail is included. The following review concentrates mainly on those models published in academic literature in the English language. A large number of water-CGE models exist, but articles on many of these models have not been published. As a result, not all models known to us have been included. 31

Water CGE models have been used to address a variety of water-related policies. Three of the common policies addressed, which can be inter-related are: 1. Economic and water use impacts of re-allocation of water amongst agricultural water users, often by means of liberalizing institutional barriers to water markets (e.g., Roe et al. 2005) 2. Economic and water use impacts as well as distributional effects of re-allocating water use between agricultural and municipal water users (e.g., Watson and Davies 2011) 3. Finding the welfare maximizing amount of water supply infrastructure such as dams, pipelines and desalinization plants (e.g., Wittwer 2009) Other applications are measuring the economic effects of drought (Horridge et al. 2005, Wittwer and Griffith 2011), global warming (Roson et al. 2010, Qureshi et al. 2012), or population increase (Watson and Davies 2011, Wittwer 2009, Qureshi et al. 2012); interaction of water trade liberalization with trade tariffs on agricultural goods (Roe et al. 2005, Robinson and Gehlhar 1995, Tsur et al. 2004); winners and losers in world trade should water supplies be restricted (Berritella et al. 2007), priced differently (Berritella et al. 2008), or used more productively (Calzadilla et al. 2011); analyzing trade-offs between agricultural water use and environmental water use (Seung et al. 1997, 1998, 2000, Dixon et al. 2011, Wittwer 2011); economic effects of sudden supply disruptions (Rose et al. 2005, 2011); water tariff and tax policy analysis (Letsoalo et al. 2007, van Heerden et al. 2008); and the estimation of the value of water quality (Brouwer et al. 2008). Many, but not all of these models are used to find a price or shadow price for water in agriculture or for municipal water. The great majority of water-CGE models address agricultural water use. This is true for a variety of reasons:     



Irrigation agriculture is by far the largest consumer of water withdrawn for human use. It is the economic activity that spends the largest proportion of its outlays on water, which makes it easier to analyze than other industries, which in comparison spend only a very small fraction of outlays on water (Young 2005). Institutional barriers as well as subsidies often isolate agricultural water use from other water uses, so that well-functioning water markets do not exist and large price differentials can occur, making a subject for policy study and change. Data on crop water use is usually available and has sometimes been studied intensively. Data on water use for commercial and many industrial enterprises is not available separate from municipal water use and/or is not of much interest, since commercial and industrial water demand usually is overshadowed by the much larger proportion of municipal demands for household water use. Lack of data and previous work on water valuation for industrial demands make the subject more difficult (Young 2005).

Very few water CGEs attempt to find MVP for water in industries outside of agriculture, though a few models do include some separate treatment of nonagricultural industries (Gomez et al. 2004; Mukherjee 1996; Rose et al. 2005, 2007; Letsoalo et al. 2007; van Heerden 2008; Hassan

32

and Thurlow 2008; Goodman 2000).10 Often industry and commercial water use is considered collectively under the rubric of municipal water, for which there is often assumed to be only one market and one price. What is true of finding the economic value of water in production with CGE models is also true for other methodologies, for identical reasons. There is generally little research on the economic value of water to industry or commercial enterprises (Young 2005). A large amount of the work that has been done is from Canada, where a survey done every five years continues to give researchers access to high quality data on water use by industry (Renzetti 2005). The framework for the water-CGE model is widely varied, depending on the specific problem under investigation. Many problems arise that are unique to water-CGE models:     

collecting the appropriate data on sectoral intensity of water use and other hydrological data on water availability specifying the appropriate elasticity of substitution for water and other factors choosing the appropriate specification of water as a factor of production or as an intermediate input finding baseline values for water or a way to avoid needing a baseline value when water does not have a market price aggregation issues for region and industry sectors

Many of these problems have a connection to the specification of the production function, which is a key element in finding the value of water as an input. Below we discuss the production function in relation to these aspects of a water-CGE, as well as some of the other important equation blocks of the CGE model.

A1 Structure of Water-CGE Models A1.1 The Production Function The backbone of the water-CGE specification is usually the production function. A particular feature of water CGE models is that land or water or both are usually included as a factor of production in the specification. It is in choosing this specification that the modeler must decide how water is used to produce goods. Water will enter the production function in at least one of three ways: 1. As an intermediate input (for example, like fertilizer or marketing services) 2. As a factor of production (for example, like capital or labor or land) 3. Implicitly (for example, in a land factor which has differential productivities due to different levels of precipitation or irrigation water applied) Some models of municipal water demands do not include full water accounts, instead assuming “one price” for municipal water and using total municipal water use and the financial data for 10

Four of the studies noted here are South African studies. Water data in South Africa apparently routinely includes data about water use in industries such as mining.

33

purchases from the water utility sector to model water use (Wittwer, 2009). Water in this case is treated only as an intermediate input or final demand good. In the case of two other CGE models that focus on municipal water demands as intermediate inputs or final consumption goods, many types of water such as “indoor water” and “outdoor water” by season and for peak loads are included in the specification (Dixon 1990; Horridge et al. 1993). Where trade-offs between urban water demands and water for crop irrigation are of central concern, treated water for urban use and untreated water for irrigation enter the production function both as an intermediate input and as a factor of production. The third case of implicit inclusion of water is most often seen in water CGE models oriented towards agricultural production only. In the case of Horridge et al. (2005), water is only implicitly a factor of production. A separate model is used to calculate how varying degrees of drought change the productivity of land across regions. The productivity of the land is then incorporated into the production functions of the CGE. A related type of model has been termed land-water models by Robinson and Gehlhar (1995). In these models, however, water accounts are introduced. In Seung et al. (1997), Seung et al. (1998) and Seung et al. (2000), water is assumed to be used in a fixed ratio with land and land is a factor of production only in agricultural sectors; thus, water irrigation use is equivalent to a given amount of land use. This type of specification may mirror legal and natural resource constraints in arid Nevada, where agricultural sectors such as hay and livestock predominate and water rights are strongly tied to a specified acreage. In Robinson and Gehlhar (1995) and Berck et al. (1991), land and water are also used in fixed ratios. In these models, however, different agricultural sectors have different ratios of land to water use and land is mobile or partially mobile across these different sectors. No rain-fed agricultural sectors are specified, limiting the extent to which water and land can be substituted for each other in production. This is in contrast to Goodman (2000), where land and water can be substitutes for each other. The production block in water-CGE models generally follows the typical specification observed in many other types of CGE models. One of the important features of CGE production functions is that functional forms are generally chosen that allow for substitution between various factors of production such as labor, capital, land and water. Often a Leontief specification that assumes a proportional relationship between quantity of output and demand for factors of production is used for water, usually in a nested structure, allowing non-water factors to substitute for each other but not for water. Cobb-Douglas or constant elasticity of substitution (CES) specifications for labor and capital are most popular. To allow for different degrees of substitutability between different factors, multi-level nests are typically used. Most of the water-CGE models use some variation of well-known specifications (see Figure 5): 1.

Nesting on the lowest level usually makes use of the Armington assumption, allowing imports and regionally produced goods to be imperfect substitutes with a CES function. The producer chooses between imported and regionally produced goods or services to furnish intermediate inputs.

2.

At the next nesting level, the aggregate of imported and regionally produced goods is combined in fixed proportions to make up a composite of intermediate material inputs.

34

3.

Next the composite intermediate inputs are combined, usually in fixed proportions, with a composite of factor inputs such as labor, capital and, for most water-CGE models, land and water. The composite of factor inputs often has been formed with several levels of nesting as well. The factor composite typically is specified with a Cobb-Douglas or CES multi-level function.

Many of the water-CGE models focus on agriculture to such an extent that water use for urban purposes is not incorporated into the model. For models concerned more with urban water use, or urban and agricultural water use trades, a water utility sector is typically included. The water utility sector uses water as a factor of production to produce treated water. Non-agricultural producers then purchase treated water (and perhaps sewer services) from the water utility as an intermediate input to production. In some models, households may also purchase the treated water and sewer services as a consumer commodity.

Figure 5: Sample of Production Technology

A1.2 Private and Government Demands The water-CGE papers reviewed tend to focus more on production than on private demand, given the predominance of agricultural water uses. A linear expenditure system derived from the generalized Cobb-Douglas (Stone-Geary) utility function was most often used to represent the utility of consumers. The Armington aggregate is again typically used when modeling how consumers will combine imports and regional goods within the same sector.

35

A few models include and track in a satellite account the water used by consumers. Goodman (2000) specifies the utility function so that it directly includes water use, as does Dixon (1990). Government and investment demands typically are not a major focus of interest in the waterCGE models included in this survey. Some models (Robinson and Gehlhar 1995) fix government and investment demands. In many cases (Horridge et al. 2005), government and investment spending shares are also exogenous to the model. In Gomez et al. (2004) the government sector merely receives taxes and pays them out again in transfers, with no demand for goods and services. Some modelers simplify by not including government (Goodman 2000). Naturally, government revenues are important where government taxes, subsidies or tariffs are modeled. Roe et al. (2005) and Tsur et al. (2004) model macroeconomic trade reforms in Morocco, and Robinson and Gehlhar (1995) model the effects of removing sectoral subsidies, taxes and tariffs in Egypt.11 Government can be an important piece of the technical specification of the water sector or factor. Often, water ‘taxes’ are collected and redistributed to help ensure the non-market nature of the sector is accounted for. For example, Berritella et al. (2007) and Letsoalo (2007) both model the water sector with administratively set prices as a non-market sector. Taxes are recycled back to households in various ways. Complex investment modeling can be of significant interest for the water-CGE modeler concerned about the timing of investments in large infrastructure projects, such as dams. For example, Dixon (1990) focuses on modeling demand for investment in water projects such as dams and water treatment plants in a dynamic model, as do Qureshi et al. (2012). A1.3 Factor Markets, including Land and Water In CGE models, wages, rents and returns to factor supplies are normally determined endogenously. The interplay of the producing sectors’ demands for factors such as water and land and the owner’s (household’s) supplies of these factors will determine price so that the market clears. Factor mobility decisions, often referred to as ‘closures’, are very influential in CGE simulation results. The obility of factors is critical to determining the value of water in production. Water-CGE models, like most CGE models, almost always include labor and capital as factors of production. The degree of both inter-regional and inter-sectoral mobility of labor and capital, as well as of water as a factor of production, is of some interest to water-CGE modelers. In particular, when the issue involves water trading between urban and rural groups, assumptions regarding regional labor mobility may be quite important. A high degree of mobility may mean workers leave an agricultural area if agricultural water use decreases, while a lack of mobility may mitigate negative rural economic impacts from reduced agricultural activity when the worker remains and is employed in a non-agricultural sector (Seung et al. 1998). Inclusion of non-irrigated agricultural activities may also allow alternative employment of labor, land and

11

Roe et al. (2005) and Tsur et al. (2004) are two of a suite of papers and book chapters related to the same Moroccan water-CGE issues. Other publications on the issue include Goldin and Roland-Holst (1995), Diao and Roe (2000, 2003, 2005).

36

capital as water prices rise, and factors are often modeled as perfectly or partially mobile between irrigated and non-irrigated agricultural sectors. The mobility of land and water across activities is also typically an important issue. In agricultural models, water and land are typically mobile or partially mobile across agricultural activities. Land and water are typically not mobile across regions. In one exception, Peterson et al. (2005) allow water mobility across regions, although the scenario is used to demonstrate the limiting case of pure mobility, with the realization that interregional water transfers would likely entail high costs not included in the model. In fact, the physical realities about the region within which water trades can take place is an important facet of modeling water, especially where the focus is finding shadow values for water. Within an area that can physically and institutionally trade water between sectors the value of the marginal product of water should equilibrate. In modeling water trade between rural and urban activities, Gomez et al. (2004) initially assume that water is not mobile between the agricultural sectors and the drinking water supply sector. This barrier is removed and trade is allowed with the results compared to the baseline model. Goodman (2000), also concerned with water trade between rural and urban uses, does not allow water mobility from agriculture to municipal use initially. In Goodman (2000) water is a factor of production in all sectors of the economy, including non-agricultural sectors, and raw water directly enters the consumption function as a consumer good. Complete mobility is permitted in one counterfactual scenario. Here again, we can see that institutional barriers to trade may be as important to the determination of water values as are physical barriers. Sectors which do not have mobility of water between them will be separate markets with separate equilibrium prices. A1.4 Product Markets for Water CGE models determine prices endogenously by equating supply and demand in a Walrasian general equilibrium framework. This process assumes a perfectly competitive market, but it is often the case that observed water prices reflect government policies rather than the workings of a marketplace. Typically, water is available to agricultural sectors at a lower than market price. In a competitive market without any distortions, the shadow price will be equal to the market price, but this is not the typical case for water. Several water-CGE models assume that water is initially unconstrained so that its initial price is zero. A constraint is then added, and the constraint makes it possible to find the shadow price of water in various activities. For example, the water CGEs using GTAP-W (Berritella et al. 2006, 2007, 2008; Roson 2010) assume that baseline water supply is greater than demand, as does Gomez et al. (2004). This implies a price of zero in the baseline. Simulations decrease water supply and a new market for water is created so that water has a MVP. In the GTAP-W models an industry-specific water price elasticity changes water use intensity factors as water price increases. This is presumably because of increased efficiency of water use but is not explicitly due to substitution of other factors. Other water CGEs specify a market for water as a factor of production whether there is an existing market or not. To do this, a baseline value of rents for water by sector must be found and subtracted from the gross operating surpluses in the SAM. Water rights sales prices can be 37

annualized for this purpose. Sometimes the difference in productivity between dryland and irrigated land is used to proxy water rents. In other cases a short-term market does exist and average lease price can be used (Goodman 2000, Watson and Davies 2011). Yet other water CGEs are not concerned with a market value or shadow price for water but rather model changes in administratively set prices. Letsoalo et al. (2007) are primarily interested in how changes in water tariffs can bring about a triple dividend of reduced water use, reduced overall tax distortion and increase in income for poor households; thus, the price of water is a tariff imposed by the government and the water factor is a non-market commodity. A1.5 Incorporation of Water Use Water-CGE models require the same data as traditional CGE models, but need water data in addition to the standard social accounting matrix. Data on water use as a factor or input of production in all water-using sectors usually must be collected or estimated. The most water-use intensive sectors are usually the agricultural sectors. As a result, many models only consider agricultural water use. Typically, the modeler collects data on current water use or tries to estimate the usage with crop water intensity factors. Industrial water use is often not readily available or at best difficult to obtain. While the modeler could estimate agricultural water use from various sources of crop production data, municipal industrial, commercial, and residential water use must generally be obtained from local water utilities. This presents a data challenge because many utilities do not track use by economic sector. A1.6 Parameterization and Solution Most CGE models, including water-CGE models, are calibrated by using the baseline data to find parameters and by searching the literature for existing parameters that may fit the current model. The assumption that the baseline SAM represents the economy at an equilibrium point allows many of the parameters to be calculated from baseline SAM data, depending on the functional forms chosen for production functions and the utility function. In addition to the water intensity factors discussed in the previous section, an important consideration for the water-CGE model is the elasticity of substitution of water with other factors of production. For agricultural sectors, these are typically assumed to be low because water is a vital input for which there are limited substitutes.

A2 Applications of Water-CGE Models Computable general equilibrium models have frequently been used to examine the economic impacts of various environmental or resource policies. One such resource is water. Given the fact that 70% of human water use is for agriculture, water-CGE models arise naturally from examination of irrigated agriculture. Various applications have examined reductions of irrigation water to agriculture for environmental reasons, because of drought or the consequences of groundwater mining and because of increased competition for the water from urban areas. Five CGE models in the literature focus solely on urban water use.

38

A2.1 Efficient Water Use in Agricultural Sectors. Many applications (Robinson and Gehlhar 1995; Tsur et al. 2004; Berrittella et al. 2005; Horridge et al. 2005; Peterson et al. 2005; Roe et al. 2005, Roson et al. 2010, Wittwer 2011, Wittwer and Griffith 2012) focus almost solely on water use and policy in the agricultural sector. Most use the CGE model to experiment with different government policies thought to bring about a more rational and efficient use of water. In their studies of Morocco, for example, Tsur et al. (2004) and Roe et al. (2005) investigate the effects of trade reform at the macro level on agricultural water use, as well as instigating water trading or other water policy reforms at the farm level. They find that while both reforms can improve water allocation decisions, the order in which the reforms are carried out is quite important. If water policy reforms such as a water trading scheme are carried out before protectionist policies are removed, water use for the protected crops may increase. This could mean the adjustment process when trade policies are liberalized will be even more wrenching. Similarly, Robinson and Gehlhar (1995) find that the introduction of a water market while current tax, subsidy and tariff policies are in place would give an actual price of water near zero when the shadow price is much higher. In other words, introduction of a water market would not bring about more efficient water use under the tax and subsidy policies current at the time of the paper. Conversely, the elimination of distortionary policies would cause the market value of water to rise and create incentives to cheat if the current system of zero cost water for farmers is continued under the new policies. Peterson et al. (2005) explore the ramifications of introducing water trading amongst agricultural sectors in Australia under conditions of decreased water availability. Simulations which decrease water availability are found to reduce gross regional product. Allowing either inter-sectoral or inter-regional trades of irrigation water adds flexibility to the system, which reduces the negative impacts of decreased water availability. Horridge et al. (2005) also use a multi-regional CGE model of the Australian economy and quantify the effects of drought on national GDP. Their findings indicate that the severe 2002-03 drought cost Australia 1.6 percent of GDP. Berrittella et al. (2007), and Roson et al. (2010) recognize that water trades take place whenever agricultural goods are traded due to the virtual water embedded in the products. They use an adaptation of the Purdue University GTAP multi-regional model of the world economy to model virtual water trade under conditions of water scarcity. In Berritella et al, for example, water scarcity is introduced into the model by assuming that use of groundwater is decreased to sustainable levels worldwide, introducing a water constraint and a market for water. They find a fall in global welfare, but locate regional winners and losers. Agricultural prices are found to rise in relation to industrial prices. Roson et al. hypothesize a change in average water availability in the countries of the Mediterranean by using global warming forecasts of precipitation. They find that trade in agricultural goods helps alleviate the negative effects but that effects are quite variable across countries, with some actually experiencing positive effects. Wittwer (2009) and Wittwer and Griffith (2011) model the effects of the recent severe droughts in the Murray Darling basin of Australia. They find that the drought has had significant and long-lasting effects on output and employment.

39

A2.2 Trade-offs between Urban and Agricultural Water Use As urban water demands for household and industry use increase, these demands compete with agriculture for scarce water resources. Often water shadow prices for urban use can be many times the shadow price of water used in agriculture. The price paid by farmers for irrigation water often does not reflect the value of the water to society in terms of the next best use. Berck et al. (1991) give an example: in the San Joaquin Valley in California at the time of writing, the price charged for irrigation water from Federal projects was $20 to $30 per acre-foot, while the value in agricultural production was estimated to be $60 to $70, and the value of the “next unit sold to the highest bidder” was somewhere between $1,000 and $2,000 for nearby urban areas. The large price differential suggests that water was not going to its highest and best use and was not being used efficiently. One way to allow water to flow to its most valued use would be to create a water market between agricultural sectors and municipal water utility sectors. Most of these models compare a scenario that allows water trades with a scenario that does not allow water trades. Unsurprisingly, since a CGE model is built to embody neo-classical theory, market outcomes are generally found to be superior to non-market outcomes. A mixture of results is found, however, regarding the winners and losers should such a water market be imposed. The basic agricultural water-CGE model is extended to include urban water use in the model in Gomez et al. (2004), Goodman (2000), Mukherjee (1996), Hassan and Thurlow (2011), Watson and Davies (2011), and Qureshi et al. (2012). Typically, raw water is an input to a water utility sector which uses water as a factor of production along with the agricultural sectors. The nonagricultural sectors then buy water as an intermediate input from the water sector, usually in fixed proportions. However, each of the models treats water inputs to non-agricultural production in a somewhat different way. Gomez et al. (2004) allow substitution between drinking water and capital. In Goodman (2000), water is a factor of production in each of four production sectors in the economy and also enters directly into consumer utility functions. Watson and Davies (2011) allow urban consumers to purchase water rights and the proceeds of the water rights sales accrue to rural households. Gomez et al. (2004) find that temporary water trades between urban and agricultural sectors during times of drought are mutually beneficial, reducing income losses in important tourism related sectors on the Balearic Islands and in some cases increasing overall income to the agricultural sectors. The modelers conclude that water trading is more efficient than building additional desalinization plants. Similarly, Goodman (2000) finds that reservoir storage and temporary water trades between urban water users and agricultural water users give similar outcomes. In Colorado, where some groups adamantly oppose new dams, Goodman concludes that temporary water trading may be a viable option. Over a 40-year period, he finds that when permanent land transfers to non-agricultural sectors are allowed, these cause a much larger reduction in agricultural activity than does temporary water trading. Mukherjee finds that some non-agricultural sectors are surprisingly water-intensive, so that movement away from commercial farming under water supply reductions does not take place to the extent that was 40

hypothesized at the beginning of the modeling experiments. Qureshi et al. models increases in population and simultaneous decreases in water supply. They find that allowing for water trade reduces negative impacts to the economy as a whole but increases negative impacts to rural regions. Building new urban infrastructure helps to reduce negative impacts on the rural economy due to water trading. A2.3 Water Demand Curves Berck et al. (1991), Gomez et al. (2004), Mukherjee (1996), Robinson and Gehlhar (1995) and Tsur et al. (2004) all use the CGE model to map out a demand curve for water. The technique used is to constrain water supply by reducing the baseline amount in steps (e.g., 95 percent of base, 90 percent of base, 85 percent of base and so on). Because of trade policy distortions or constraints on labor or water factor mobility between sectors, shadow prices may not always be equivalent to market prices for water. Robinson and Gehlhar (1995) and Gomez et al. (2004) trace out various shadow prices and market prices, while Tsur et al. (2004) and Berck et al. (1991) find the demand curve of shadow prices and Mukherjee (1996) finds a market price demand curve. A2.4 Trade-offs between Environmental and Agriculture Water Use Berck et al. (1991), Seung et al. (2000), Seung et al. (1998), Seung et al. (1997) and Wittwer (2011) examine trade-offs between water use for environmental or recreational goods and agricultural water use. Berck et al. (1991) model large reductions in irrigation as a possible solution to salinity build-up in the San Joaquin Valley. They do not directly attempt to measure benefits due to slower importation of salts, but rather present the economic losses to agriculture as a point of departure for comparison with other salt reduction projects. The three Seung et al. (2000, 1998, 1997) papers have in common measurement of economic trade-offs between increased recreation due to increased water at wetlands or lakes versus losses in agricultural sectors due to decreased irrigation. Recreation benefits are incorporated using exogenous models which give impacts on the service and trade sector due to increases in recreation when lakes are fuller or wetlands are larger in area. In each case, the authors find that increased recreation benefits in sparsely populated rural Nevada do not outweigh the decreases in agricultural production. There is no attempt to incorporate non-market benefits. Wittwer (2011) is able to use the TERM-H2O model to isolate the impacts of water rights buybacks for environmental purposes from the effects of a drought in Australia’s Murray Darling Basin. The drought’s effects are determined to far outweigh the consequences of the buyback; however, once again, no attempt is made to quantify the environmental benefits or incorporate them into the CGE model. A 2.5 Urban Water Use Of the five CGE models in the literature that focus solely on urban water use, two find the ideal price for an urban utility to charge (Dixon 1990, Horridge 1993), two examine short run impacts from earthquake-disrupted water supplies (Rose and Liao 2005, Rose et al. 2011), and another explores the circumstances under which the infrastructure investments for municipal water demand would have been optimal (Wittwer, 2009). Dixon (1990) uses a water-CGE model to find the appropriate price for water utilities to charge their customers, a task usually 41

accomplished with a partial equilibrium model. Horridge et al. (1993) explore the same question with a more sophisticated approach that includes stochastic water supplies. The shadow price of water is used to guide decisions about how much water supply infrastructure to build and when. Wittwer (2009) uses the multi-regional dynamic form of TERM, the Australian regional CGE model, to evaluate the need for a dam or other supply infrastructure projects in SE Queensland. He finds that if precipitation holds to long-term averages, the infrastructure investment will likely not increase welfare.

A.3 Conclusions Several problems remain for future researchers. In general, a lack of data hinders many possible directions of new research. Hydrological detail is generally omitted from water-CGE models because of data limitations and complexity. Data about groundwater use is often limited or unavailable and omitted from models, despite its importance in the economy and the looming or already present difficulties of non-sustainable use. Seasonality of water use may impose a different set of constraints than is given when using annualized data on water use. Water may be of varying qualities, and therefore of varying value, both because of naturally occurring quality differences and differences occurring because of human-caused pollutants, but this remains difficult to incorporate into a water-CGE model. Given the importance of water quality both as an input to production and consumption and as an output, usually negative, of production, this may be an area in need of further effort. One issue related to water quality may be industrial and commercial water use outside of agriculture, which has largely not been addressed in the water-CGE literature thus far. Detailed data on water use in non-agricultural sectors is limited, so little effort has been concentrated on water use trade-offs amongst them. Urban return flows as well as issues relating to recycling of treated water have yet to be incorporated into a water-CGE model. Modeled water markets are usually beneficial in reducing or eliminating water inefficiencies. These models, however, generally depend on assumptions of zero transmission losses when moving water from one user to another. In reality, transmission losses may be quite high in some situations. Additionally, positive or negative third party effects may occur as a result of water transfers, such as groundwater recharge, maintenance of open green space or trees, return flows to rivers, or loss of instream flows for fisheries and recreation. Regional leakage of water-rights compensation from water markets is another unexplored area. Enhanced tourism, amenity values, and environmental or recreational non-market values have yet to be fully incorporated into a water-CGE model. Given the importance of these issues, this may be a fruitful field for future research. Research on urban water and land use is limited. Increasing urban demands for water and land are a likely topic for future water-CGE models. Given a world in which populations and economies are growing, demands for both agricultural water and urban water will continue to grow, and may often be in conflict. Additional pressure may be added if ethanol made from thirsty crops becomes a significant fuel source. Reliability of water supplies is always a question in many parts of the world, due to variable climate 42

conditions. When pressing environmental needs for water are added to the mix, it isn’t hard to imagine how carefully water resources will need to be managed in the future. Given the large role that water plays in the economy, the CGE model, with its ability to handle system feedbacks, may be an important tool for examining these issues. As water becomes more valuable, data on water uses and related values is likely to improve. Thus it is likely that more water-CGE models will be developed and that the thirty-five models in this survey are only the beginning forays.

43

Appendix B Adaptation of Blackhurst et al. method for obtaining industry sector water intensity factors Blackhurst et al. (2010a; 2010b) develop a method to calculate water withdrawal intensity factors by industry sector at a national level. We have adapted their method for use at a state level, for both withdrawals and consumption, and for use with state-level IMPLAN data. We have used the method to calculate water intensity factors for all 440 industry sectors in IMPLAN for the state of Nevada in a manner that can be duplicated for any other state. The process can also be used at the county level if adaptations are made to the process that is used to derive the crop irrigation water intensity factors. Blackhurst et al. use USGS national level water use data in combination with the irrigation survey from NASS and several other sources to find national water use coefficients. USGS water use data by major sector by state and by county is produced every five years (Kenny et al 2009). More detail on this method is provided below. The most current water use data available from the USGS is for 2005, while the data for the IMPLAN social accounting matrix that we use in the CGE model is for 2010. After examining total water use data from the major utilities in Nevada, it was determined that urban water demand did not increase over the period. It was assumed that most self-supplied industry water use, which is tied to limited water rights, has not changed substantially over the period from 2005 to 2010.12 Thus no upward adjustment in water use was made to the 2005 USGS data. B.1 Withdrawal versus Consumption Water consumption is defined as water that leaves the region of interest through evaporation, transpiration, or because it is contained within trade goods. Consumption does not include water returned to the regional water supply through, for example, sewer systems and water treatment facilities, run-off, or seepage into groundwater. Water withdrawals are actual water volumes used at a site for treated water and at a source for supplied water. Smith et al. (2011) adjust water withdrawals to water consumption using state-specific values.. Using the USGS estimates given in Smith et al. for state and overarching category sectors (domestic supply, industrial and commercial, irrigation, etc.), consumption rates are first applied to the USGS categories of water use; the Blackhurst method is then used with these totals. Table 3 shows how water consumption totals are estimated from water withdrawal data for Nevada. Overall, the USGS estimated that 2.7 million acre-feet (AF) of water was withdrawn for human use in 2005. Of this, about 50%, or 1.4 million AF, was evaporated, transpired, or otherwise taken out of the region. Thermoelectric plants are estimated to have the highest rate of consumption out of all Nevada sectors, with 75% of the withdrawn water consumed, whereas industrial and commercial uses are estimated to consume 15% of the water withdrawn for their 12

Mining activity increased substantially over this time period. The ramifications of this for water use have not yet been investigated.

44

use. Water consumption was the measure used in the Nevada model being developed for statewide water footprints. For finding shadow prices, however, water withdrawal is a better estimate. The USGS estimates for proportion consumed may be used to allow for return flows in the future version of the water-CGE model. Ideally, particularly for urban water, sewer water returns should be tracked separately, since waste water is an input into costly water treatment and legal requirements for water quality can be a constraint on total water use.

Table 3. 2005 Nevada Water Withdrawals Adjusted to Water Consumption Category

Water withdrawals (AF)*

Percent consumed or evaporated**

Estimate of water consumption (AF)

Public Supply Domestic Industrial and commercial Self-supplied

472,085

25.6%

120,789

285,446

15.0%

42,817

Domestic Self Supply

41,871

25.6%

10,713

Industrial self-supplied

6,609

15.0%

991

Irrigation Livestock/Aquaculture use Mining

1,678,211

66.7%

1,118,807

26,715

37.0%

9,890

110,961

27.3%

30,262

Total thermoelectric 41,255 75.7% Total withdrawals, in 2,663,153 51.3% AF/Yr Sources: *(Kenny et al 2009), **(Smith et al 2011), author’s calculations

31,220 1,365,490

B.2 Public Supply USGS public supply for the state of Nevada is allocated across industry sectors and final demand using purchases of the water, sewage and other systems utility industry in the IMPLAN social accounting matrix. This requires assumptions such as a single price for water across sectors, negligible or equal percentages of the purchases spent on sewer by each industry sector and final demand institution, and national level estimates that fit local production functions. This method needs further testing. For Nevada, it was found that top water users, such as the casino industry, were correctly estimated to be top water users with this methodology. Intuitive results for different income levels of households were also obtained. One puzzle was the very low water use by government institutions and sectors. For example, public schools appear to be a major water user with large areas of irrigated landscape, but water use for the local and state education final demand sector is near zero in the IMPLAN data. 45

B.3. Self-Supplied water use in power generation, agriculture, manufacturing and mining USGS water use estimates are given for the power generation, agricultural and mining sectors. The Blackhurst methodology allocates these estimates to subsectors. If the above method showed water purchases from the treated water utility for these sectors, the two amounts are added together for intensity factors. However, the self-supplied water use total is maintained. B.4. Power generation Self-supplied withdrawals for power generation can be directly applied to the power generation sector (in our case, sectors 31, 428 (federal electric utilities) and 131 (state and local electric utilities) in IMPLAN). B.5. Irrigation withdrawal allocations for agriculture Irrigation water consumption represents the vast majority of total water consumption and water withdrawals in Nevada. Irrigation for agriculture is the second largest withdrawal after thermoelectric cooling and is the largest proportion of water consumption nationwide. The Blackhurst et al. method uses the NASS 2008 Farm and Ranch Irrigation Survey (United States Department of Agriculture 2009) combined with Census of Agriculture data on crop acreage (National Agricultural Statistics Service 2009). The survey contains data on irrigated acreage by crop and average acre-feet applied by crop for each state. The irrigation survey does not contain county-level estimates of crop irrigation withdrawals. However, the USGS does give a county-level estimate of total crop irrigation water applied, and the Census of Agriculture gives crop acreage data by county. Using state-level water intensity levels with crop acreages, followed by raking total irrigation water use back to USGS as a control, represents one method of approximating new water intensity factors for agriculture at the county level. Other, more sophisticated methodologies have been developed for estimation of water consumption by crop, using weather station data and reference crop evapotranspiration rates as described in Mubako (2011). Some models include rainfall as part of the agricultural sector specification. Rainfall variability is an important driver of irrigation water demand. The Nevada model does not address rainfall issues, since it is a small proportion of total water needed and used. For states with a larger amount of rainfall, this would be an important model consideration. B.6. Manufacturing Statistics Canada has developed values for water withdrawal and water consumption per employee for manufacturing by NAICS code. Blackhurst et al. use these values to help allocate industry self-supplied water across manufacturing sectors. This method has been adapted for use with IMPLAN sectors. The industries most likely to use self-supplied water are the largest water using manufacturing sectors: food processing, petroleum refineries, coal, metal refining, paper and wood processing and computers and electronics manufacturing. None of these types of manufacturing are prominent in Nevada, which makes these estimates of negligible importance 46

for our model. Other states with more activity in these manufacturing sectors would need to examine the Canadian values more closely. B.7 Mining Blackhurst et al. use a variety of source material to determine per employee water use for the mining sectors, which they then apply, raking back to the USGS national mining water withdrawal total. The water intensity factors that Blackhurst et al. give in terms of water use per dollar of output are applied and raked to the USGS estimates for Nevada or Clark County.

47

References Berck, Peter; Sherman Robinson and George Goldman. 1991. "The Use of Computable General Equilibrium Models to Assess Water Policies," In The Economics and Management of Water and Drainage in Agriculture., ed. P. Berck, S. Robinson and G. Goldman, 489-509. Norwell, MA: Kluwer Academic Publishing. Berrittella, M.; A. Y. Hoekstra; K. Rehdanz; R. Roson and R. S. J. Tol. 2007. "The Economic Impact of Restricted Water Supply: A Computable General Equilibrium Analysis." Water Research [Water Res.], 41(8). Berrittella, M.; K. Rehdanz; R. Roson and R. S. J. Tol. 2008. "The Economic Impact of Water Taxes: A Computable General Equilibrium Analysis with an International Data Set." Water Policy [Water Policy], 10(3). Berrittella, M.; K. Rehdanz and R. S. J. Tol. 2006. "The Economic Impact of the South-North Water Transfer Project in China: A Computable General Equilibrium Analysis," In Working Paper FNU-117. Palermo, Italy: Fondazione Eni Enrico Mattei. Blackhurst, Michael; Chris Hendrickson and Jordi Sels i Vidal. 2010a. "Direct and Indirect Water Withdrawals for U.S. Industrial Sectors." Environmental Science and Technology, 44(6), pp. 2136-30. ____. 2010b. "Direct and Indirect Water Withdrawals for U.S. Industrial Sectors, Supplemental Material." Environmental Science and Technology, 44(6), pp. 2136-30. Brouwer, Roy; Marjan Hofkes and Vincent Linderhof. 2008. "General Equilibrium Modelling of the Direct and Indirect Economic Impacts of Water Quality Improvements in the Netherlands at National and River Basin Scale." Ecological Economics, 66(1), pp. 127-40. Calzadilla, A.; K. Rehdanz and R. S. J. Tol. 2010. "The Economic Impact of More Sustainable Water Use in Agriculture: A Computable General Equilibrium Analysis." Journal of Hydrology, 384(3-4), pp. 292-305. Calzadilla, A.; K. Rehdanz and R. S. J. Tol.. 2011. "Water Scarcity and the Impact of Improved Irrigation Management: A Computable General Equilibrium Analysis." Agricultural Economics, 42(3), pp. 305-23. Diao, Xinshen and Terry Roe. 2003. "Can a Water Market Avert the "Double-Whammy" of Trade Reform and Lead to a "Win-Win" Outcome?" Journal of Environmental Economics and Management, 45, pp. 708-23. Diao, Xinshen and Terry Roe. 2000a. "Economy-Wide Gains from Decentralizing Water Allocation in a Spatially Heterogeneous Agricultural Economy.," In The Political Economy of Water Pricing Reforms., ed. A. Dinar. New York: Oxford University Press. Diao, Xinshen and Terry Roe. 2000b. "The Win-Win Effect of Joint Water Market and Trade Reform on Interest Groups in Irrigated Agriculture in Morocco," In The Political Economy of Water Pricing Reforms, ed. A. Dinar. Oxford: Oxford University Press. Diao, Xinshen; Terry Roe and Rachid Doukkali. 2005a. "Economy-Wide Gains from Decentralized Water Allocation in a Spatially Heterogenous Agricultural Economy." Environment and Development Economics, 10(3), pp. 249-69. Dixon, P. B.; M. T. Rimmer and G. Wittwer. 2011. "Saving the Southern Murray-Darling Basin: The Economic Effects of a Buyback of Irrigation Water." Economic Record, 87(276), pp. 153-68. Dixon, Peter B. 1990. " A General Equilibrium Approach to Public Utility Pricing: Determining Prices for a Water Authority." Journal of Policy Modeling, 12(4), pp. 745-67. Fadali, Elizabeth. 2012. "Implan Water Footprints," In. Minneapolis, MN: Mid-continent Regional Science Association 43rd Annual Conference. Fadali, Elizabeth. 2012. "The Economics of the Regional Water Footprint," In Western Regional Science Association. Poipu, Hawaii.

48

Ghadimi, Hodjat 2007. "Cge Modeling and Applications: A Short Course," In. Morgantown, West Virginia. http://rri.wvu.edu/CGECourse/index.htm Gillig, Dhazn and Bruce A. McCarl. 2002. "Introduction to Computable General Equilibrium Model (Cge)," In. College Station, TX. Goodman, D Jay. 2000. "More Reservoirs or Transfers? A Computable General Equilibrium Analysis of Projected Water Shortages in the Arkansas River Basin." Journal of Agricultural and Resource Economics, 25(2), pp. 698-713. Gomez, C.; D. Tirado and J. Rey-Maquieira. 2004. "Water Exchanges Versus Water Works: Insights from a Computable General Equilibrium Model for the Balearic Islands." Water Resources Research, 40(10). Green, Emily. 2012. "Tunneling under California's Bay: Delta Water Wars," In High Country News. Paonia, CA. Hassan, R. and J. Thurlow. 2011. "Macro-Micro Feedback Links of Water Management in South Africa: Cge Analyses of Selected Policy Regimes." Agricultural Economics, 42(2), pp. 235-47. Hassan, R.; J. Thurlow; T. Roe; X. Diao; S. Chumi and Y. Tsur. 2008. "Macro-Micro Feedback Links of Water Management in South Africa: Cge Analyses of Selected Policy Regimes," In. Horridge, Mark; Peter B Dixon and Maureen T. Rimmer. 1993. "Water Pricing and Investment in Melbourne: General Equilibrium Analysis with Uncertain Streamflow," In, 1-25. Clayton, Australia: Centre of Policy Studies, Monash University. Horridge, Mark; John Madden and Glyn Wittwer. 2005. "The Impact of the 2002-2003 Drought on Australia." Journal of Policy Modeling, 27(3), pp. 285-308. Kenny, J. F., Barber, N. L., Hutson, S. S., Linsey, K. S., Lovelace, J. K., & Maupin, M. A. (2009). Estimated Use of Water in the United States in 2005. nvco2005.xls, from United States Geological Survey, http://water.usgs.gov/watuse/data/. Letsoalo, A.; J. Blignaut; T. de Wet; M. de Wit; S. Hess; R. S. J. Tol and J. van Heerden. 2007. "Triple Dividends of Water Consumption Charges in South Africa." Water Resources Research, 43(5). Lindall, Scott; Doug Olson and Greg Alward. 2006. "Deriving Multi-Regional Models Using the Implan National Trade Flows Model." Regional Analysis & Policy, 36(1), pp. 76-83. Minnesota IMPLAN Group. 2010. "Version 3.0 User's Guide," In. Hudson, WI: Minnesota IMPLAN Group, Inc. Mubako, Stanley. 2011. "Frameworks for Estimating Virtual Water Flows among U.S. States," In Department of Environmental Resources and Policy, 251. Carbondale, IL: Southern Illinois University. Mukherjee, Natasha. 1996. "Water and Land in South Africa: Economy-Wide Impacts of Reform a Case Study for the Olifants River," In TMD Discussion Papers. Washington, D.C.: International Food Policy Research Institute Trade and Macroeconomics Division. National Agricultural Statistics Service. 2009. "2007 Census of Agriculture Nevada State and County Data," In. Washington, D.C.: United States Department of Agriculture. National Agricultural Statistics Service. 2008. "2008 Farm and Ranch Irrigation Survey,"United States Department of Agriculture,. Peterson, Deborah; Gavan Dwyer; David Appels and Jane Fry. 2005. "Water Trade in the Southern Murray-Darling Basin." Economic Record, 81(Special Issue Aug. 2005), pp. s115-27. Qureshi, Muhammad Ejaz; Wendy Proctor; Mike D. Young and Glyn Wittwer. 2012. "The Economic Impact of Increased Water Demand in Australia: A Computable General Equilibrium Analysis." Economic Papers, 31(1), pp. 87-102. Renzetti, S. 2005. "Determining the Economic Value of Water: Concepts and Methods. Resources for the Future." Environmental & Resource Economics, 32(3), pp. 439-41. Robinson, Sherman and Clemen Gehlhar. 1995. "Land, Water, and Agriculture in Egypt: The Economywide Impact of Policy Reform," In TMD Discussion Paper No. 1, 1-50. Washington, D.C.: International Food Policy Research Institute. 49

Roe, T.; A. Dinar; Y. Tsur and X. S. Diao. 2005. "Feedback Links between Economy-Wide and Farm-Level Policies: With Application to Irrigation Water Management in Morocco." Journal of Policy Modeling, 27(8), pp. 905-28. Rose, A. and S. Y. Liao. 2005. "Modeling Regional Economic Resilience to Disasters: A Computable General Equilibrium Analysis of Water Service Disruptions." Journal of Regional Science, 45(1), pp. 75112. Rose, A.; S. Y. Liao and A. Bonneau. 2011. "Regional Economic Impacts of a Verdugo Scenario Earthquake Disruption of Los Angeles Water Supplies: A Computable General Equilibrium Analysis." Earthquake Spectra, 27(3), pp. 881-906. Roson, Roberto and Martina Sartori. 2010. "Water Scarcity and Virtual Water Trade in the Mediterranean," In Working Paper Department of Economics Ca' Foscari University of Venice, 1-13. Venice, Italy: Ca' Foscari University of Venice. Schreider, S. . 2009. "Water Price Dynamics, Water Derivatives and General Equilibrium Modelling," In 18th World IMACS/ MODSIM Congress. Cairns, Australia. Seung, C.; T. Harris and R. Narayanan. 1998. "A Computable General Equilibrium Approach to Surface Water Reallocation Policy in Rural Nevada." American Journal of Agricultural Economics, 80(5), pp. 119797. Seung, C. K.; J. Englin and T. Harris. 1997. "Application of Computable General Equilibrium Model to Derive Impacts of Surface Water Reallocation Policy." Journal of Agricultural and Resource Economics, 22(2), pp. 395-95. Seung, C. K.; T. R. Harris; J. E. Englin and N. R. Netusil. 2000. "Impacts of Water Reallocation: A Combined Computable General Equilibrium and Recreation Demand Model Approach." Annals of Regional Science, 34(4), pp. 473-87. Seung, Chang; Thomas R. Harris; Thomas R. MacDiarmid and W. Douglass Shaw. 1998. "Economic Impacts of Water Reallocation: A CGE Analysis for the Walker River Basin of Nevada and California." Journal of Regional Analysis and Policy, 28(2), pp. 13-34. Seung, Chang K. 1999. "Application of a Computable General Equilibrium Model to Evaluate Surface Water Reallocation Policies." Review of Regional Studies, 29(2), pp. 139-55. Seung, Chang K.; Thomas R. Harris and Thomas R. MacDiarmid. 1997. "Economic Impacts of Surface Water Reallocation Policies: A Comparison of Supply-Determined Sam and CGE Models." Journal of Regional Analysis and Policy, 27(2), pp. 55-76. Smith, C. A., Simon, A. J., & Belles, R. D. (2011). Estimated Water Flows in 2005. (LLNL-TR-475772). Livermore, CA: Retrieved from https://flowcharts.llnl.gov/. Tsur, Yacov; Terry Roe; Rachid Doukkali and Ariel Dinar. 2004. "Chapter 5. Interaction between Economywide Policies and Irrigated Agriculture in Morocco," In Pricing Irrigation Water: Principles and Cases from Developing Countries. Washington, D.C.: Resources for the Future. van Heerden, J. H.; J. Blignaut and M. Horridge. 2008. "Integrated Water and Economic Modelling of the Impacts of Water Market Instruments on the South African Economy." Ecological Economics, 66(1), pp. 105-16. Watson, Philip S. and Stephen Davies. 2011. "Modeling the Effects of Population Growth on Water Resources: A Cge Analysis of the South Platte River Basin in Colorado." Annals of Regional Science, 46(2), pp. 331-48. Wing, Ian Sue. 2011. "Computable General Equilibrium Models for the Analysis of EconomyEnvironment Interactions," In Research Tools in Natural Resource and Environmental Economics, ed. A. Batabyal and P. Nijkamp. Boston: World Scientific Publishing Company. Wittwer, G. ed. Economic Modeling of Water: The Australian CGE Experience. Dordrecht: Springer, 2012.

50

Wittwer, G. and M. Griffith. 2012. "The Economic Consequences of a Prolonged Drought in the Southern Murray-Darling Basin," In Economic Modeling of Water: The Australian Experience, ed. G. Wittwer, 119-41. Dordrecht: Springer. Wittwer, G. and M. Griffith. 2011. "Modelling Drought and Recovery in the Southern Murray-Darling Basin." Australian Journal of Agricultural and Resource Economics, 55(3), pp. 342-59. Wittwer, Glyn. 2011. "Confusing Policy and Catastrophe: Buybacks and Drought in the Murray-Darling Basin." Economic Papers, 30(3), pp. 289-95. Wittwer, Glyn. 2009. "The Economic Impacts of a New Dam in South-East Queensland." Australian Economic Review, 42(1), pp. 12-23. Wittwer, Glyn. 2006. "Modelling Future Urban and Rural Water Requirements in a CGE Framework," In, 1-19. Centre of Policy Studies, Monash University. Young, Robert A. 2005. Determining the Economic Value of Water: Concepts and Methods. Washington, D.C.: Resources for the Future.

51