Demand side management policies for residential water use: who bears the conservation burden?
Increased reliance on demand side management (DSM) policies to manage existing water supplies is stimulating substantial debate among economists and policymakers. Debate focuses on both the effectiveness of alternative policy instruments in increasing efficiency and their equity implications for residential users. While economists generally advocate higher residential water prices as a means of reducing demand, others argue that non-price policies, which do not affect the price of water but place direct controls on water use such as rationing, constitute the only viable means to reduce residential demand. This conclusion relies, in part, on empirical research indicating that residential water demand is price inelastic, making price a relatively ineffective DSM policy (see, for example, Nieswiadomy and Molina 1989: Agthe et al. 1986; Billings and Agthe 1980; and Howe and Linaweaver 1967).(1) Further, some policymakers argue that "the use of price as an allocation mechanism is constrained by the fact that water is generally regarded as a basic necessity, even a right, not an economic good" (Berk et al. 1980). As a consequence, policymakers cannot ignore the equity implications of water allocation decisions. These distributional concerns often favor rationing and targeted use restrictions as well as "block" pricing schemes which tend to provide a basic level of water service at lower cost.
The problem facing water policymakers and water utility managers is a lack of adequate information to determine the performance of price and non-price policy instruments in their communities. Most previous economic analyses of residential water demand have relied on aggregate cross-section (Nieswiadomy 1992; Howe 1982; Foster and Beattie 1979) or times series data (Agthe et al. 1986; Billings and Agthe 1980; Young 1973). These studies have lacked critical information on household demand determinants including domestic water-saving technologies, use of water for landscape irrigation, and type of irrigation technologies. No formal attention has been given to analyzing how household characteristics influence price and non-price policy responsiveness and their associated distributional implications.
This research first develops a theoretical framework to improve our understanding of how DSM policies are expected to influence residential demand for different classes of residential households. The theoretical framework is then tested empirically by utilizing a natural experimental market setting in two communities in Southern California in which households faced a variety of DSM policies. An econometric model of household demand for water is formulated and estimated in a two-stage least squares estimation framework. This econometric model explicitly incorporates endogenous technological change, improving upon earlier specifications. The analysis relies on household-level cross-section monthly time series data in two communities in Southern California for a six-year period.
The relative performance of alternative DSM policies in reducing aggregate demand and their distributional implications, based on select household characteristics, is assessed. The two general hypotheses tested are: (l) the magnitude of the reduction in aggregate demand attributable to a specific policy instrument is a function of the socioeconomic and structural characteristics of households in a given community and (2) the choice of the policy instrument influences the distribution of water savings or "conservation burden" among household classes based on these characteristics.
II. THE ECONOMICS OF DSM FOR RESIDENTIAL WATER: THE COMPOSITION OF AGGREGATE DEMAND MATTERS
In any particular local water market, the aggregate water demand curve equals the horizontal summation of individual household demand curves whose slope and position vary, in part, due to differences in household characteristics. Household characteristics such as income, intensity of water-using capital, and landscaped area are also expected to influence how household demand responds to specific policy instruments. As a result, the composition of aggregate demand in a particular community is an important determinant of aggregate demand reduction and the distribution of conservation burden among household classes.
The Available Set of Policy Instruments
Policymakers have two basic categories of DSM policy instruments to align excess demand with available supplies - price policies and non-price policies, as shown in Figure 1. The initial equilibrium corresponds to point a, with price and quantity equal to [P.sub.0] and [Q.sub.0], respectively. A supply shock, such as due to drought or more stringent environmental regulations, reduces the aggregate supply of water to [Q.sub.1], represented by the short-run inelastic supply curve S[prime]. To reduce excess demand, the water utility can raise the price to [P.sub.1]. If policymakers are unwilling to increase the price of water to equilibrate the market, the only remaining alternatives are non-price policy instruments that shift the aggregate demand curve from D to D[prime] at the prevailing price of [P.sub.0]. One means of reducing excess demand is rationing. However, a strict rationing policy where households are entitled to a fixed amount of water and then cut off from service is no doubt politically unacceptable as residential water use is generally considered a basic necessity. Other policies that shift the demand curve include household specific water allocations with various penalties for noncompliance; water restrictions on selected uses such as irrigation; public information and education campaigns to encourage voluntary conservation; and subsidies for adoption of water efficient technologies.
The Conservation Burden: Differential Impacts of Policy Instruments
The extent to which a particular DSM instrument reduces aggregate demand equals the sum of savings from individual households. Figure 2 shows the reduction in aggregate demand and potential for uneven incidence of conservation burden associated with a uniform price increase from [P.sub.0] to [P.sub.1] for two hypothetical classes of households, each with a different demand curve. The price increase secures a larger reduction in demand from class B households relative to class A households.
The composition of aggregate demand also may influence the reduction in demand and the distribution of conservation burden associated with non-price policies. Figure 3 shows that the incidence of conservation burden associated with a landscape irrigation ban varies among household classes. For class A households, who use water for predominantly domestic uses, the irrigation ban results in only a small shift in their demand curve ([D.sub.a] to [D[prime].sub.a]), leaving the slope unchanged as this is a relatively small and price inelastic component of total demand. For class B households, the irrigation ban changes both the slope and position of their demand curve ([D.sub.b] to [D[prime].sub.b]). The magnitude of the demand curve shift and slope changes reflect the levels of pre-policy irrigation water use and household price responsiveness, relative to total pre-policy demands.
Clearly, the composition of aggregate demand matters. It influences both the total reduction in demand attributable to a specific policy instrument and the incidence of conservation burden among household classes.
III. A NATURAL EXPERIMENTAL MARKET SETTING: DSM POLICIES IN SOUTHERN CALIFORNIA
To assess the relative performance of specific DSM policy instruments to reduce aggregate demand and identify the incidence of conservation burden among household classes, this research takes advantage of experience with urban DSM programs implemented during California's statewide drought, which persisted to varying degrees of intensity from 1985-92. While the severity of the drought during this period may have heightened households' sensitivity to water issues and related DSM policies, it also allows examination of a wide range of policy instruments. Household survey data and concomitant utility water use and price data were collected for two communities located along the south coast of Santa Barbara County, California. The communities of Santa Barbara and Goleta were selected for their exclusive reliance on local surface and ground-water supplies, experience with different demand side management policy regimes, and the variety of household types based on socioeconomic and structural characteristics.
These drought years left Santa Barbara County's reservoirs severely depleted. Rainfall from 1987 to 1990 fluctuated from between 94 percent and 30 percent of historical levels. By spring 1990, a 45 percent reduction in deliveries from Lake Cachuma, the principal water source for the region, was required. The water shortage became so acute in July 1990 that the Governor of California declared a state of emergency in the city of Santa Barbara, making it the first entity ever to be proclaimed a state drought emergency site in California.(2) Water managers in both communities were forced to rely almost exclusively on DSM policies to cope with extreme constraints on water availability, instituting increasingly stringent price and non-price policies to reduce demand.
The community of Santa Barbara provides water service to a population of approximately 91,000 (including 16,500 single-family homes). Figure 4 shows Santa Barbara's mean quarterly and annual single-family household water use between 1985 and 1990 and indicates when specific DSM policies were implemented. Between 1985 and 1990, mean household water use in Santa Barbara fell substantially, ostensibly in response to the DSM policies implemented.
Initially in August 1988, Santa Barbara offered free low flow showerheads and rebates for adoption of low flow toilets (REBATE). In June 1989, as water became progressively scarce, Santa Barbara implemented its first price policy and moved from fixed per-unit uniform rates to a moderately increasing block price schedule (PRICE1). As the water shortages became increasingly severe in early 1990, Santa Barbara adopted more aggressive policies. In February 1990, they banned specific water uses, including virtually all landscape irrigation except for hand irrigation and drip systems (RESTRICT). Water police enforced the irrigation restrictions. Two months later, in April 1990, Santa Barbara implemented its second price policy - a steeply increasing block price schedule (PRICE2).(3)
The Goleta Water District serves a population of about 74,000 and includes about 14,000 connections. During the course of the water shortage Goleta also implemented a number of DSM policies. Figure 5 shows quarterly and annual mean single-family household water use for Goleta between 1985 and 1990 and indicates when specific DSM policies were implemented. During this period, mean water use in Goleta fell significantly, ostensibly due to the DSM policies adopted.
In 1986, the District offered rebates for adoption of low flow toilets (REBATE1). In July 1987, the subsidy program was expanded to include free low flow showerheads (REBATE2) and the rebate for low flow toilets was increased. Drought-induced water shortages became increasingly severe and in July 1989 Goleta adopted a mandatory water allocation policy (ALLOCATE) based, in part, upon historical household water use patterns. The policy imposed significant marginal price penalties for households exceeding their allotment. Each household was entitled to a base allocation of 132 hundred cubic feet (hcf) per year plus 55 percent of 1984-88 annual average usage above the 132 hcf.(4) In July 1990, the District moved from a moderately increasing block pricing schedule to relatively high uniform rates (PRICE).(5)
Trends in Water Use by Household Characteristics
Figures 4 and 5 clearly show that demand trended downward during the 1985-90 period. The reductions varied not only over time with policy changes, but also across household types providing support for the argument that policy instruments influence the incidence of conservation burden among household classes. Table 1 shows changes in mean water use over the 1985-90 period by income and housing density, where density is a proxy for landscaped area and concomitant water uses. Low density households possess large landscaped areas and have relatively high irrigation water use, given arid conditions.(6) In contrast, high density households have significantly smaller landscaped areas and thus lower irrigation water use. The data indicate that in Goleta, lower income households tended to reduce their demand most. In Santa Barbara, higher income households reduced their demand most. In both communities, water use fell most among households with larger landscaped areas (low density households).
TABLE 1 CHANGE IN MEAN WATER USE BY HOUSEHOLD CHARACTERISTICS: 1985-1990 (PERCENT REDUCTION) Santa Barbara Goleta Income Class Less than $20,000 -38 -60 $20,000-$59,999 -60 -53 $60,000-$99,999 -57 -43 Greater than $99,999 -74 -47 Density Class Low -69 -61 High -37 -46
IV. A MODEL OF HOUSEHOLD DEMAND FOR WATER INCORPORATING ENDOGENOUS TECHNOLOGICAL CHANGE
An econometric model of household water demand, which explicitly incorporates technological change, is specified and estimated to identify the reduction in aggregate demand attributable to specific policies and their distributional implications. The household water demand model is composed of two basic components: technology adoption equations (four equations) and a water demand equation. The technology adoption equations capture endogenous household technology adoption decisions. Explicit measures of domestic and landscape irrigation technologies include: low flow toilets, low flow showerheads, and water efficient irrigation (drip and hand held) and traditional irrigation technologies (permanent and moveable sprinklers and hoses). Predicted values of these four technology variables are used in the second stage of the analysis to help explain changes in household water demand and to assess the relative contributions of price policy, non-price policy, and technological change in the reduction of household water demand.
The model of residential water demand takes the following form:
Domestic Technology Adoption
[Mathematical Expression Omitted] 
[Mathematical Expression Omitted] 
Landscape Irrigation Technology Adoption
[Mathematical Expression Omitted] 
[Mathematical Expression Omitted] 
[Mathematical Expression Omitted] 
[Mathematical Expression Omitted]
[Mathematical Expression Omitted].
Variable definitions are presented in Table 2.
Equations  and  capture the adoption of domestic water efficient technologies represented by low flow toilets ([LFT.sub.it]) and showerheads ([LFS.sub.it]). Equations  and  capture the adoption of landscape water efficient ([WSAV.sub.it]) and traditional technologies ([WUSE.sub.it]). Both traditional and water efficient irrigation technologies are included in the model since households frequently use both types of technologies simultaneously. Following Nieswiadomy (1992), Nieswiadomy and Molina (1989), and Moncur (1987), the water demand equation (equation ) is specified as a linear functional form. A two-stage least squares (2SLS) estimation procedure is employed given the simultaneity inherent in the model.
DSM policies are expected to influence demand directly and indirectly by inducing technological change. Higher water prices are expected to directly reduce demand in the short run and stimulate the demand for water efficient technologies by increasing the relative benefits associated with adoption in the medium to long run. Thus, the marginal price variable (MP) is included in both the water demand equation and the technology adoption equations. In the water demand equation (equation ), a difference variable (D) is incorporated to account for the effect of intra-marginal rate changes on water demand under increasing block pricing schedules, in accordance with the "Taylor-Nordin" specification (Taylor 1975; Nordin 1976).(7) Both the marginal price (MP) and difference (D) variables are lagged one period to avoid contemporaneous correlation with water use under block pricing. Under increasing block pricing schedules, the difference variable (D) acts as a lump sum income transfer and is expected to correlate positively with water use. Non-price policies are also expected to influence water demand both directly, by restricting demand, and indirectly, by encouraging adoption of water efficient technologies. Dummy variables representing Goleta's administrative water allocation policy (ALLOCATE) and Santa Barbara's landscape irrigation use restrictions policy (RESTRICT) are included in the water demand equation and adoption equations. These policy dummies are expected to correlate negatively with water demand and positively with adoption of water efficient technologies (equations -). In addition, a dummy variable representing Santa Barbara's and Goleta's low flow toilet and showerhead rebate programs (REBATE) is used to account for the reduced costs of adoption for domestic water efficient technologies, and is expected to have a positive impact on the adoption of domestic water efficient technology.
TABLE 2 NOTATION AND VARIABLE DESCRIPTION Variable Indicator Variable Description [W.sub.it] Water consumption for the ith household in the tth period [MP.sub.it] Marginal price of water faced by the ith household in the tth period [D.sub.it] "Taylor-Nordin" difference variable for the ith household in the tth period [RESTRICT.sub.t] Policy dummy for Santa Barbara's landscape irrigation use restriction policy [ALLOCATE.sub.t] Policy dummy for the quantity-related component of Goleta's water allocation policy [REBATE.sub.t] Policy dummy for the low flow toilet and showerhead subsidy programs [WUSE.sub.it] Adoption of traditional irrigation technologies by the ith household in the tth period [WSAVE.sub.it] Adoption of water efficient irrigation technologies by the ith household in the tth period [LFS.sub.it] Adoption of low flow showerheads by the ith household in the tth period [LFT.sub.it] Adoption of low flow toilets by the ith household in the tth period [INC.sub.it] Gross monthly household income for the ith household in the tth period [HH.sub.it] Number of household members for the ith household in the tth period [LOW.sub.it] Dummy variable for low density households (lot size greater than .55 of an acre) [MED.sub.it] Dummy variable for medium density households (lot size between .08 and .55 of an acre) [FAU.sub.it] Number of faucets for the ith household in the tth period [SB.sub.i] City of Santa Barbara dummy variable [CPI.sub.t] Consumer price index for Santa Barbara and Ventura Counties for the tth period [RAIN.sub.t] Cumulative monthly rainfall for the tth period MONTH[(j).sub.t] Month dummy variables (j = 2, . . ., 12)
To separately identify domestic from landscape irrigation demands, dummy variables representing landscaped irrigation uses (as proxied by density size - LOW and MED) are included in the water demand equation (equation ). Households with larger landscaped areas are expected to demand more water, all other factors held constant. The amount of landscaped area is also expected to influence irrigation technology adoption and thus density dummy variables are included in the irrigation technology adoption equations (equations  and ).
The model is estimated using a 2SLS estimation procedure to account for endogeneity of technology. In the first stage, the technology adoption equations are estimated. For the domestic technology adoption equations (equations  and ), a classical additive disturbance term with zero expectation and finite variance is added to each equation to reflect random errors in adoption behavior, and they are estimated using OLS. The home ownership (HOME) and rebate policy (REBATE) variables are excluded from the demand equation to ensure identification of the model. For the landscape irrigation technology equations (equations  and ), LIMDEP's maximum likelihood estimation (Green 1990) was used to maximize the log likelihood function for the probit model to pro, vide consistent estimates of [Mathematical Expression Omitted] and [Mathematical Expression Omitted].
In the second stage, the water demand equation (equation ) is estimated using predicted values from the technology adoption equations ([Mathematical Expression Omitted], [Mathematical Expression Omitted], [Mathematical Expression Omitted], [Mathematical Expression Omitted]). A classical additive disturbance term with zero expectation and finite variance is added to the equation to reflect random errors in consumption. The water demand equation is estimated using OLS with White's formulation (1980) to correct the standard errors for possible hetereoscedasticity introduced by the data.(8)
Data were collected on a representative stratified random sample of 119 single-family households in the communities of Goleta and Santa Barbara over the six-year (198590) period. Stratification was based on actual housing densities for the communities from Santa Barbara city and county zoning maps and supplemental planning and zoning data. Actual water use and costs were obtained from each household's monthly utility bill for the period. Sample households took part in a lengthy telephone survey on household water use including such factors as income, number of people per household, water-related technologies in use for domestic and landscape irrigation uses, explicit measures of domestic and landscape irrigation technological change over the six-year period, home ownership status, and landscaped area as an approximate measure of the allocation of total household use between domestic and landscape uses. Secondary sources were used for climatic data likely to affect demand. In addition, extensive corroborating interviews were conducted in the field with households, utility staff, and individuals working in the area of water resources.
VI. RESULTS AND DISCUSSION
The estimation results of the water demand equation for the region, shown in column 1 of Table 3, indicate good model performance.(9) All coefficients generally exhibit expected signs and statistical significance. The results appear robust to changes in specification, time period, and sub-sample analysis.(10) The adjusted R-square for the two-stage least squares (2SLS) model for the region, which includes predicted values for the technology variables, is .31. This adjusted R-square is comparable to earlier studies (Nieswiadomy 1992; Nieswiadomy and Molina 1989; Jones and Morris 1984).(11)
The signs and statistical significance of the socioeconomic, structural, and other control variables demonstrate the model's robust performance. The coefficient on the number of household members (HH) is, as expected, positive and significant. The number of interior household water connections (FAU), which serves as a proxy for intensity of water-using capital, has the predicted positive effect on water demand. Households with more water-using capital, all other factors held constant, demand more water.
The estimated coefficients on the income variable (INC) and the increasing block price implicit income subsidy variable (D) - known as the difference variable - exhibit [TABULAR DATA FOR TABLE 3 OMITTED] the anticipated positive effect on household water use. The magnitude of the income variable implies that a 10 percent increase in income will increase monthly household water demand by 3.6 percent.(12) This income elasticity of demand is comparable with other residential water demand studies (Howe and Linaweaver 1967; Jones and Morris 1984; Nieswiadomy 1992).(13)
Effects of Price Policy on Household Water Demand
The coefficient on the marginal price of water (MP) is, as expected, negative and statistically significant (column 1, Table 3). The estimated short-run own-price elasticity of demand equals -.33, implying a 10 percent increase in water price will reduce aggregate demand by 3.3 percent.(14) Given the actual price changes in Santa Barbara and Goleta over the period, the average household demand fell 9.3 percent and 26.2 percent, respectively.(15) The estimated long-run own-price elasticity of demand equals -.39, implying that a 10 percent increase in prices will result in a 3.9 percent reduction in demand over the longer run due to the direct and indirect effects of price policy. The indirect effect captures the adoption of water efficient technologies and abandonment of traditional irrigation technologies as a result of higher water prices.(16)
The own-price elasticity estimates for aggregate water demand are within the order of magnitude of earlier studies that utilized panel data. These estimates ranged from -.27 to -.52 (Nieswiadomy and Molina 1989; Billings 1987; Moncur 1987; Agthe et al. 1986). More importantly, the own-price elasticity estimates are similar to those previously estimated for three communities in the South Coast of the Santa Barbara County region (Berk et al. 1980). Using utility-level data from the 1970s, Berk et al. estimated the mean own-price elasticity for the South Coast region as -.29. For the individual communities of Carpenteria, Summerland, and Montecito the own-price elasticities of demand equaled -.28, -.37, and -.22, respectively.
While aggregating water demand across households provides policymakers with a measure of the price responsiveness of aggregate demand, it does not tell them about the distributional consequences of higher prices on specific household groups. Results indicate that price responsiveness varied significantly according to income and other household characteristics. Households with lower incomes responded more to higher water prices than wealthier household groups, as hypothesized (Table 4). The magnitude and statistical significance of the estimated own-price elasticities of demand by income group imply that a 10 percent increase in water prices would result in a 5.3 percent reduction in demand for low income households, a 2.2 percent reduction for moderate to high income households, and a 1.1 percent reduction for wealthy households.(17) As a consequence, price policy employed in the region distributed the conservation burden to lower income households. These results corroborate the argument that price policy results in a non-uniform incidence of conservation burden and highlights the importance of examining the effects of policy instruments based on household characteristics.
TABLE 4 OWN-PRICE ELASTICITY OF DEMAND FOR WATER BY INCOME CLASS Income Class Own-Price Elasticity(a) Total Sample -.33 [less than]$20,000 -.53 $20,000-$59,999 -.21 $60,000-$99,999 -.22 [greater than]$99,999 -.11 a Calculated based on estimated marginal price/income class interaction coefficients and 1990 average marginal and consumption data. All estimates significant at the .05 level or greater.
Effects of Non-Price Policy on Household Water Demand
The impact of Goleta's water allocation policy (ALLOCATE) and Santa Barbara's irrigation use restriction policy (RESTRICT) on water consumption and the associated distributional consequences are investigated separately. Whereas the marginal price variable (MP) captures the marginal price penalty surcharge component of Goleta's allocation policy, the ALLOCATE variable captures the quantity-related component of this policy.
Goleta's water allocation policy. The analysis indicates that Goleta's water allocation policy significantly reduced water demand. The coefficient on the allocation policy dummy variable (ALLOCATE) is negative and statistically significant, implying that average household water demand in Goleta was reduced by 4.58 HCF (or 28.2 percent) as a result of this policy, as shown in column 3 of Table 3.(18)
To better understand the distributional implications of Goleta's allocation policy, the model was estimated for density subsamples, as shown in columns 4 and 5 of Table 3. Estimating by density subsamples indicates differential rates of policy responsiveness based on lot size. In terms of magnitude, the negative and statistically significant coefficients on this policy variable (ALLOCATE) imply the quantity-related component of the allocation policy reduced demand more among low density respondents with larger landscaped areas (7.24 HCF) than high density respondents (3.40 HCF) with smaller amounts of landscaping.
Santa Barbara's landscape irrigation use restrictions policy. Results clearly affirm a priori expectations that Santa Barbara's landscape irrigation use restrictions policy reduced demand, as shown in column 1, Table 3. The magnitude of the negative and statistically significant coefficient on the irrigation restrictions variable (RESTRICT) for the Santa Barbara subsample implies that implementation of this policy reduced average household water demand in the city by 4.37 HCF (or by 16 percent).(19)
To identify the incidence of conservation burden associated with irrigation restrictions for household classes, the model was estimated based on landscape area (or density) subsamples. As predicted, irrigation use restrictions created differential negative effects on household water demand depending on landscaped area. The coefficients on the irrigation restrictions variable (RESTRICT) for low and high density subsamples are negative and statistically significant, as shown in columns 4 and 5 of Table 3. The magnitudes of the coefficients imply this policy reduced average household demand on an absolute basis among low and high density households by 6.69 HCF and 4.52 HCF, respectively.
Effects of Technological Change on Household Water Demand
Technological change also reduced household water demand. The estimated coefficient on the predicted low flow toilet variable [Mathematical Expression Omitted] has the anticipated negative impact on water demand, as shown in column 1 of Table 3. It implies that increasing the number of low flow toilets by one decreases household water demand by 10 percent.(20) This result closely parallels estimated savings of approximately 11 percent associated with adoption of low flow toilets found in a recent study using data from southern California (Chesnutt, Moynaham, and Bamesai 1992).
Low flow showerheads also reduce household water demand. The estimated coefficient on the predicted low flow showerhead adoption variable [Mathematical Expression Omitted] implies that increasing the number of low flow showerheads by one decreases household water demand by 8 percent.(21) Earlier studies estimate household water savings associated with adoption of a low flow showerhead to be: 9.7 percent (Whitcomb 1991), 6.4 percent (Whitcomb 1990), and 2 percent (Chesnutt and McSpadden 1991).(22, 23) Adoption of low flow toilets and showerheads reduced water demand in the region by improving the technical efficiency of these fixtures. The main advantage of retrofitting domestic plumbing fixtures is that they secure significant long-term water savings and require no behavioral changes on the part of households.
With respect to irrigation technological change, the estimation results demonstrate that both water-efficient and traditional irrigation technologies influence water demand. The estimated coefficients for these predicted irrigation technology variables ([Mathematical Expression Omitted] and [Mathematical Expression Omitted]) are shown in column 1 of Table 3. Results indicate that adoption of water efficient irrigation technologies [Mathematical Expression Omitted] reduce average household water demand by 11 percent in the region.(24) As expected, reductions in water use associated with adoption of more efficient irrigation technologies varied based on density, a proxy for landscaped area (see columns 4 and 5 of Table 3). The estimated coefficients on the predicted water-efficient irrigation technology adoption variables for density subsamples can be interpreted to mean that adoption of water-efficient irrigation technologies among low and high density households reduced average total water usage by 31 percent and 10 percent, respectively.(25)
Households who irrigate their landscaping with traditional technologies, all other factors held constant, use more water as expected. The coefficient on the predicted traditional irrigation technology variable [Mathematical Expression Omitted] for the region is positive and statistically significant implying that households using traditional irrigation technologies, such as moveable and permanent sprinklers, use on average 9 percent more water than households who do not (see column 1, Table 3).(26) Chesnutt and McSpadden (1991) found that single-family households in Los Angeles with automatic sprinkler systems consumed on average 11.2 percent more than those households using manually controlled sprinkler systems or watered by hand. The estimated increase in average household demand associated with traditional irrigation technologies in this study is slightly lower than Chesnutt and McSpadden (1991), perhaps as a consequence of the fact that many households retrofitted hoses with flow control devices.
VII. CONCLUSIONS AND POLICY IMPLICATIONS
The main objective of this research was to assess the performance of alternative demand side management policy instruments in terms of their effectiveness in reducing aggregate demand and their distributional implications. Results indicate that DSM policies were effective in reducing demand. Importantly, the magnitude of the reduction in demand associated with different policy instruments varied significantly with the characteristics of the households. These results highlight the importance of conducting analyses of residential water demand at the household level.
Overall, household demand was responsive to price changes. However, price responsiveness was found to vary by income group, as anticipated. Lower income households were more price responsive. In particular, low income households were found to be more than five times as price responsive as relatively wealthy households reflecting the fact that their water bill typically constitutes a larger share of the household budget. These results suggest that price policy will achieve a larger reduction in residential demand in a lower income community than in a higher income community, all other factors held constant. Results also suggest that if price policy is the primary DSM instrument in a particular locale, lower income households will bear a larger share of the conservation burden.
Results indicate that socioeconomic and structural characteristics of households influence the outcome of non-price policies. Landscape irrigation technological change reduce demand more among households with larger landscaped areas. These households also tend to be higher income. Obviously, these types of policies will reduce demand more in suburban communities where households maintain larger amounts of landscaping than in higher density urban areas. The "joint" nature of the policy outcome suggests that the same policy instrument applied to two different communities may result in different outcomes in terms of the total reduction in aggregate demand and the distribution of conservation burden among household classes.
What are the implications of this research for urban water utility managers and policymakers designing demand side management plans? A key policy implication of these results for regional urban water policy planners is that to achieve required reductions in aggregate demand as efficiently as possible, given equity considerations, regional water demand needs to be disaggregated based on selected characteristics of the communities.
One of the greatest operational difficulties encountered by water policymakers in designing demand side management programs is a lack of adequate information to determine how specific policy instruments may influence demand in their respective localities. This issue is especially problematic since water policymakers often make policy decisions in vastly different environments. Policymakers need three important interrelated pieces of information to effectively evaluate alternative demand management instruments to ensure they achieve the required reduction in demand while meeting their equity objectives. First, they need to have some sense of the characteristics of households in their service area to determine the feasible set of policy instruments. For example, what are the income, landscaping, and other characteristics of the households in their service area? Second, they need to know the extent to which specific policy instruments are expected to reduce aggregate demand, given these particular characteristics. Third, they need to understand how different types of households are expected to reduce their demand in response to specific policy instruments to assess the distributional implications of the policy.
Once the characteristics of aggregate demand are understood, policymakers can use the results of household-level residential water demand analyses to identify the feasible set of policy instruments, to estimate the expected reduction in aggregate demand associated with each policy instrument, and to assess the potential distribution of conservation burden among household classes. When faced with a water shortage, policymakers can use this menu of policy options to determine which policies will achieve the required reduction in demand while meeting their equity objectives.
Further research on household-level water demand is needed to address two important issues. First, more information is needed on the extent to which other non-price policies (such as public education programs and off-peak irrigation policies) reduce aggregate demand and how household characteristics influence policy responsiveness. Off-peak irrigation policies generally significantly limit or ban irrigation during peak water demand periods (such as during the high evapotranspiration hours of 10:00 a.m. to 4:00 p.m. during the summer months). Second, since water policymakers frequently implement more than one policy during the course of a water shortage, more research is needed to understand how the interaction of policy incentives influence the reduction in demand and the distribution of conservation burden. In particular, under what circumstances, if any, do policy incentives interact synergistically to achieve a larger reduction in demand than if each policy were implemented in isolation. What are the lags in responsiveness of demand to policy changes, and how does the amount of time that lapses between when each policy is implemented influence the reduction in demand?
1 While economic theory suggests that residential water demand should be relatively price inelastic, the argument that residential consumers do not respond to higher prices is seriously flawed for several reasons. Economic theory suggests that residential water demand should be price inelastic for three reasons: (1) there exists no close substitutes for water in most of its uses, (2) the amount of money spent on water is generally a relatively small share of the typical household budget, and (3) water is frequently demanded jointly with some other complementary good. Yet, the argument that residential consumers do not respond to higher prices because demand is price inelastic is seriously flawed for at least two reasons. Since a market demand curve may be inelastic in some price ranges and elastic in others, reference to a demand curve as inelastic or elastic must be made to a specific range of prices. As prices rise demand is expected to become less inelastic. particularly over the long run. Second, many advocates of non-price policy have erroneously equated price inelasticity with no price responsiveness. The description of residential demand as price inelastic is a technical definition; it simply means that a one-percent increase in price results in a less than one-percent decrease in consumption. In other words, consumers respond to higher prices, but at less than proportionate to the price increase.
2 Santa Barbara County had repeatedly rejected delivery of State Water Project (SWP) water, in part because of a belief that increased water supplies would induce growth. In 1990, the severity of water shortages dominated growth control concerns, at least in the short run, and resulted in voter approval of SWP water and an emergency desalinization plant. At present the desalinization plant is in production and it is anticipated that SWP water will come on-line within three to four years.
3 Santa Barbara chose block water price schedules and landscape irrigation use restrictions as their primary DSM instruments during the most severe period of drought-induced water shortages because they thought it was the "fairest way to get the conservation needed." The City adopted block pricing "to convey a strong price signal." City water officials believed that because of increased water rates most households would reduce their usage, but they also anticipated many wealthier households would continue to use large amounts of water and pay higher water bills. They thought that the irrigation use restrictions, in conjunction with higher water rates, "distributed the pain better" then higher prices alone. (Source: Steve Mack, City of Santa Barbara.)
4 One hcf equals 756 gallons.
5 Goleta chose allocations because they needed to reduce demand immediately given the severity of the water shortage. In their decision-making process, District Board members sought an equitable plan and ultimately determined that "allocations were the fairest way to achieve the desired level of demand reduction." They chose to base allocations in part on historical usage as a compromise between their equity considerations and administrative feasibility since historical usage data was readily available. The adoption of uniform rates was motivated, in part, by a need to stabilize district revenues. With DSM, revenues became increasingly unreliable under block pricing. The district also faced financial hardship as a result of its successful low flow toilet rebate program. In addition, uniform rates simplified district billing and made water bills easier for consumers to understand, particularly under the allocation policy where marginal price penalties were set at four or ten times the marginal rate, depending on the frequency with which a household exceeded its allotment. (Source: Darcy Aston, Santa Barbara County Water Agency, formerly with the Goleta Water District.)
6 Low density households have lot sizes in excess of .55 of an acre, medium density have lots of between .08 and .55 of an acre, and high density households maintain less than .08 of an acre.
7 The difference variable (D) is defined as the difference between what a consumer would have paid if all units were purchased at the marginal price and the amount paid under the block pricing schedule. D = [P.sub.k]Q - ([summation of] + [P.sub.j][Q.sub.j] where j = 1 to k - 1 + [P.sub.k] (Q - [summation of] [Q.sub.j] where j = 1 to k - 1)), where k identifies the block in the price schedule in which consumption occurs, [P.sub.j] is the marginal price in the jth block, [Q.sub.j] is the threshold quantity in the jth block, and Q is the total quantity demand and is contained in the kth block.
8 Possible sources of hetereoskedasticity include greater variation in water use among low density households than their higher density counterparts. Low density households, with substantially larger lots, may exhibit a wider range of variability in landscape irrigation water use as some households chose to landscape intensively with ornamental species while others chose native or natural landscaping.
9 Estimation results for the Stage I technology adoption equations are reported in Appendix Table 1.
10 The estimation results for the city and density subsamples are shown in columns 2 through 5 of Table 3.
11 The adjusted R-squares for the Jones and Morris (1984) marginal price water demand equations ranged from .25 to .28, for the Nieswiadomy and Molina (1989) models they ranged from .11 to .46, and for the Nieswiadomy (1992) marginal price models they ranged from .07 to .44.
12 Estimated using the income coefficient in column 1 of Table 3 and 1990 mean income and consumption data, as shown in Appendix Table 2.
13 Howe and Linaweaver (1967) estimated the income elasticity of demand as .32 for domestic water uses and .66 for irrigation demand. The estimated income elasticities of demand calculated by Jones and Morris (1984) for their three marginal price models ranged from .40 to .55. In the three marginal price models estimated by Nieswiadomy (1992) he found the income elasticity of demand to range from .28 to .44.
14 Estimated using the marginal price coefficient in column 1 of Table 3 and 1990 mean price and consumption data, as shown in Appendix Table 2. The estimated own-price elasticity assumes an equal percentage (or proportionate) changes in all blocks of the rate structure.
15 Percentage reduction based on 1985-88 average monthly consumption for Santa Barbara and Goleta of 27.46 hcf and 16.2 hcf, respectively.
16 Estimated using the coefficients in column I of Table 3 and the marginal price coefficients from the estimated technology adoption equations in Appendix Table 1.
17 The own-price elasticity is defined with respect to equal percentage (or proportionate) changes in all blocks of the rate structure.
18 Percentage reduction based on 1985-88 average monthly consumption in Goleta of 16.26 hcf.
19 Percent reduction based on 1985-88 average consumption in Santa Barbara of 27.46 hcf.
20 Calculated using the estimated coefficient from column 1 of Table 3 and 1990 mean low flow toilets and water use, as shown in Appendix Table 2.
21 Calculated using the estimated coefficient in column 1 of Table 3 and 1990 mean consumption and low flow showerhead adoption data, as shown in Appendix Table 2.
22 The estimate from Chesnutt and McSpadden (1991) likely underestimates the potential savings associated with installation of low flow showerheads because this study measured the savings associated with preexisting low flow showerheads, which were adopted prior to the DSM program, of likely inferior technical efficiency.
23 Whitcomb (1991) used data from Contra Costa County, California, and his 1990 study utilized data from Seattle, Washington. Chesnutt and McSpadden (1991) used data from southern California.
24 Calculated using the estimated coefficient in column 1 of Table 3 and 1990 mean water use and adoption of water efficient technology data, as shown in Appendix Table 2.
25 Calculated using the estimated coefficients in columns 4 and 5 of Table 3 and 1990 mean water use and adoption of water-efficient technology data, as shown in Appendix Table 2.
26 Calculated using the estimation coefficient in column 1 of Table 3 and 1990 mean water use and adoption of traditional technology data, as shown in Appendix Table 2.
Agthe, Donald E., Bruce R. Billings, John L. Dobra, and Kambiz Raffiee. 1986. "A Simultaneous Equation Demand Model for Block Rates." Water Resources Research 22 (1):14.
Berk, Richard A., et al. 1980. "Reducing Consumption in Periods of Acute Scarcity: The Case of Water." Social Science Research 9 (2):99-120.
Billings, R. Bruce. 1987. "Alternative Demand Model Estimators for Block Rate Pricing." Water Resources Bulletin 23 (2):341-45.
Billings, R. Bruce, and Donald E. Agthe. 1980. "Price Elasticity for Water: A Case of Increase Block Rates." Land Economics 56 (Feb.):73-84.
Chesnutt, Thomas, and Casey McSpadden. 1991. "A Model-based Evaluation of Westchester Water Conservation Program." A&N Technical Services, San Diego.
Chesnutt, Thomas, Michael Moynahan, and Anil Bamezai. 1992. "Ultra-Low-Flush Toilet Rebate Programs in Southern California: Lessons for Water Managers and Planners." A&N Technical Services, San Diego.
Foster, Henry S., Jr., and Bruce R. Beattie. 1979. "Urban Residential Demand for Water in the United States." Land Economics 55 (Feb.): 43-58.
Greene, William. 1990. Econometric Analysis. New York: Macmillan Publishing Co.
Howe, Charles W. 1982. "The Impact of Price on Residential Water Demand: Some New Insights." Water Resources Research 18 (4): 713-16.
Howe, Charles W., and F. P. Linaweaver Jr. 1967. "The Impact of Price on Residential Water Demand and Its Relation to System Design and Price Structure." Water Resources Research 3 (1):13-31.
Jones, C. Vaughan, and John Morris. 1984. "Instrumental Price Estimates of Residential Water Demand." Water Resources Research 20 (2): 197-202.
Moncur, James. 1987. "Urban Water Pricing and Drought Management." Water Resources Research 23 (3):393-98.
Nieswiadomy, Michael L. 1992. "Estimating Urban Residential Water Demand: Effects of Price Structure, Conservation, and Public Education." Water Resources Research 28 (3): 609-15.
Nieswiadomy, Michael L., and David J. Molina. 1989. "Comparing Residential Water Demand Estimates Under Decreasing and Increasing Block Rates Using Household Demand Data." Land Economics 65 (Aug.):280-89.
Nordin, J. A. 1976. "A Proposed Modification of Taylor's Demand Analysis: Comment." The Bell Journal of Economics 7 (2):719-21.
Renwick, Mary E. 1996. "An Econometric Model of Household Water Demand With Endogenous Technological Change Under Demand Side Management Policies." Ph.D. diss., Stanford University.
Taylor, Lester D. 1975. "The Demand for Electricity: A Survey." The Bell Journal of Economics 6 (1):74-110.
Whitcomb, John. 1990. "Water Use Reductions from Retrofitting Indoor Water Fixtures." Water Resources Bulletin 26 (6):921-26.
-----. 1991. "Water Reductions for Residential Water Audits." Water Resources Bulletin 27 (5):761-67.
White, H. 1980. "A Heteroscedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroscedasticity." Econometrica 48:817-38.
Young, Robert. 1973. "Price Elasticity of Demand for Municipal Water: A Case Study of Tucson, Arizona." Water Resources Research 9 (4): 1068-72.
Diane Hite was a visiting assistant professor at the Ohio State University, Department of Economics and Center for Human Resource Research when this paper was written; she is currently assistant professor, Department of Agricultural Economics, Mississippi State University. The author would like to thank Wen Chern, Fredrick Hitzhusen, Hajime Miyazaki, Alan Randall, and David Schmeidler for contributing to the conceptual framework of this research. Lucia Dunn's assistance in creating the survey instrument was invaluable. Funding and support from the Ohio Coal Development Office, Electric Power Research Institute, American Electric Power Co., Ohio Edison Co., and the Ohio State University Department of Agricultural Economics is gratefully acknowledged. Finally, the author wishes to thank two anonymous referees whose critical comments and constructive suggestions improved the manuscript immeasureably; any errors are attributable to the author, naturally.
Mary E. Renwick and Sandra O. Archibald are, respectively, research associate and professor at the Humphrey Institute of Public Affairs, University of Minnesota.
|Printer friendly Cite/link Email Feedback|
|Author:||Renwick, Mary E.; Archibald, Sandra O.|
|Date:||Aug 1, 1998|
|Previous Article:||Cross-tenure differences in home maintenance and appreciation.|
|Next Article:||The economic effects of land registration on smallholder farms in Kenya: evidence from Nyeri and Kakamega districts.|