Price perception and nonprice controls under conservation rate ...

4 downloads 0 Views 302KB Size Report
the 2010 AWWA and Raftelis Financial Consulting (RFC) Rate. Survey (AWWA .... earlier work may be economically significant, but not practically significant, to ...
E446

Aubuchon & Roberson | http://dx.doi.org/10.5942/jawwa.2012.104.0101 Journal - American Water Works Association Peer-Reviewed

Price perception and nonprice controls under conservation rate structures C r aig P . A u b uch o n 1 AND J. Al a n Roberson 2 1The

Analysis Group, Boston, Mass. Washington, D.C.

2AWWA,

This research evaluates the effect of price and nonprice conservation controls on monthly water system demand and explores differences in rate design, education and outreach programs, population growth, and regional climate variables among a national cross section of utilities. Using the Shin price perception parameter, this study found that under conservation rate structures, aggregate demand was related to something other than marginal or average price. The price–demand response

increases with higher levels of consumption for both the marginal price and the total bill, which may provide preliminary evidence that the price signal of the total bill matters for demand. Nonprice controls were not found to be statistically significant in the study sample. Income elasticities were positive and slightly larger in magnitude than price elasticities, suggesting that over the long term, utility managers may need to increase rates faster than regional income growth for effective demand management.

Keywords: conservation, demand management, utility rate structures As a result of increasing population growth and water demand, many researchers and practitioners advocate a “soft path” approach to water supply development, in which future water needs are met through a combination of various demand-side conservation measures (AWWA Climate Change Committee, 2011; Brandes & Brooks, 2006; Gleick, 2003). The soft path approach recognizes that water is a “normal” good in the sense that water is required for human life and most people in developed countries can turn on a tap, flush a toilet, or take a shower without thinking twice about the source but, more important, that water is normal in an economic sense. As prices rise, consumers tend to use less. Increasingly, water managers, utility councils, and board members are developing a better understanding of this definition of water as a normal good and are using conservation-oriented rate structures as part of the long-term management of water supplies. Conservation rate structures take a variety of forms, but at the root, these structures share a basic tenet that water gets more expensive at higher and more discretionary levels of use. This study considered a conservation rate structure to be either an increasing block rate (IBR) or a seasonal rate structure. In an IBR, the “usage in each succeeding block is charged at a higher unit rate than the previous blocks” (AWWA, 2000). Under a seasonal rate structure, water prices increase during summer months in order to reduce demand that is primarily associated with seasonal irrigation. A few utilities, primarily in the West, have also implemented rate structures based on water budgets. Beecher (2012) provides a comprehensive overview of water budget rate structures and discusses the relative merits of the price and nonprice signals inherent in this rate design. Under a water budget rate, households are typically allocated a quantity of water based on

household size, lot size, and landscaping mixture. Prices generally follow an IBR structure within each household allocation. In contrast, some utilities use a declining block rate structure (DBR), in which the unit price of water decreases with higher levels of consumption; a flat rate, in which all customers pay the same amount per billing period regardless of use; or a uniform rate, in which all users pay the same unit price for any level of consumption. Water budget rates are not considered in the context of this report. This research provides new findings that take advantage of the 2010 AWWA and Raftelis Financial Consulting (RFC) Rate Survey (AWWA & RFC, 2010). This data set allows a comparison of utilities and rate structures across the United States and permits consideration of the variables that most influence the adoption of conservation rate structures. It also allows the effect of prices on water demand to be isolated when the effects of population growth, climate, and other drivers of demand are held constant. By examining aggregate demand at the utility level, the researchers considered the important question of how price and nonprice controls contribute to demand management. Previous research often expressed demand reductions at the household level, but more interesting, and potentially more applicable to utility managers, is how demand is correlated with residential prices at the broader utility scale. Recent research suggests that price elasticity may be a function of rate structure (Olmstead et al, 2007) or influenced by the amount of information on the bill (Gaudin, 2006). The current study updates the approach of Nieswiadomy and Cobb (1993) and considers the relationship between demand and price schedules. Using the 1984 AWWA rate survey, Nieswiadomy and Cobb

2012 © American Water Works Association

E447

Aubuchon & Roberson | http://dx.doi.org/10.5942/jawwa.2012.104.0101 Journal - American Water Works Association Peer-Reviewed

found that under both IBR and DBR structures, consumers responded to average prices. In contrast, Nieswiadomy and Molina (1991) used a random sample of 100 customers in Arizona and found that consumers responded to marginal prices under an IBR. One goal of a conservation rate structure is to maximize the price signal to consumers. Because an estimated 75% of US customers receive water bills that list only the total price of consumption (Gaudin, 2006), it is important to understand how aggregate demand is related to price and which price signal—average price, marginal price, or total bill—sends the most effective message in support of conservation goals. The current study provides price elasticity measures that may be more useful to utility boards and officials who are considering implementation of a conservation rate structure. As a final step, this analysis considered the effects and interaction of nonprice controls, such as demand management programs (e.g., droughtinduced irrigation restrictions, time-of-day watering) and conservation education initiatives, on water demand.

OVERVIEW As more utilities adopt or consider adopting conservation rate structures, it is important to understand why they do so and what part of the rate structure is effective because consumers may respond to the price increase, the stigma associated with moving up to a higher tier of consumption, or more detailed information about indoor versus outdoor water use. As a first step, this study empirically tested the claim that water managers are increasingly adopting conservation rate structures in response to population growth in water-constrained regions or in areas where the gap between system capacity and total demand is narrowing (Gleick, 2010; Cooley & Gleick, 2009). Utilities faced with accelerated population growth may adopt a conservation rate structure as a precursor to developing new supplies. If this is the case, population growth and demand forecasts should be clearly conveyed to water utility boards during new rate proposals. Water utilities may also adopt conservation rate structures in order to reduce both per capita demand and peak load factors associated with residential demand, particularly during summer months, potentially delaying costly expansions of infrastructure or the need for new sources of supply (Beecher, 2010a; Chesnutt & Beecher, 1998). The peak load factor is usually defined as the maximum water demand relative to average water demand, and the time frame can vary (hourly, daily, or monthly are all typical). Because peak day (or peak hour) demands typically drive future infrastructure investments, a lower peak factor usually allows a utility to more reliably forecast future infrastructure investment needs and minimize excess capacity. Water demand and price structures have been studied extensively since the late 1960s, beginning in part with publication of the research of Howe and Linaweaver (1967). This study was one of the first to follow a panel of individual households and to publish empirical estimates of price elasticity. These investigators found that indoor water demand was highly inelastic, whereas outdoor use was more elastic, particularly in the East. Price elasticity refers to the percentage change in water quantity demanded divided by the percentage change in price. Values between 0 and –1 are termed

inelastic, because demand is reduced by less than the price increase; values smaller than –1 are termed elastic. Espey et al [1997] analyzed 124 previous studies and found a mean elasticity of –0.51, whereas Dalhuesin et al [2003] surveyed 314 models and found a mean elasticity estimate of –0.41. A number of other studies have confirmed these general findings and have used new econometric techniques to improve the reliability of and confidence in elasticity estimates (Worthington & Hoffman, 2008; Dalhuisen et al, 2003; and Espey et al, 1997). All three of these studies provide detailed meta-analyses of the literature. In particular, researchers have addressed two questions of bias. Selectivity bias may occur because more conservationoriented municipalities are likely to adopt conservation-oriented rate structures and serve households that are more likely to respond to price incentives. Simultaneous equation bias may occur because prices and demand are simultaneously determined in an IBR structure. When consumers choose a level of consumption, they simultaneously change the marginal price they face. This bias results in inefficient and biased estimates of elasticity. (An inefficient estimate is one that does not converge with the true population parameter as the sample size increases.) This is primarily a concern related to the statistical validity of the model, but because of the sensitive political nature of rates and rate structures, statistical accuracy in this context carries a practical weight beyond the academic literature. The current study updates and extends previous research that used earlier AWWA rate surveys (Nieswiadomy and Cobb [1993] and Nieswiadomy [1992] used the 1984 AWWA rate study, and Gaudin [2006] used the 1996 AWWA rate study) by incorporating a pricing structure selectivity parameter—the probability of adopting a conservation rate study—in an ordinary least squares (OLS) framework. The analysis then uses a price perception parameter to investigate the relationship between demand and price signals. Gaudin (2006) found that 78% of utilities did not include any price information other than the total charge on water bills. This raises a concern that the average and marginal prices studied in earlier work may be economically significant, but not practically significant, to customers and utility decision-makers. Finally, the current work updates much older studies—for example, Nieswiadomy and Cobb found that a utility located in California has a lower probability of adopting a conservation rate structure, which is no longer true—and it addresses a concern that much of the research on water demand has been based on relatively few data studies (Worthington & Hoffman, 2008).

DATA AWWA and RFC have conducted biannual rate surveys together since 2004, although each organization conducted earlier iterations independently. This report draws on the 2010 survey, which contains data on 308 water utilities in 49 states and the District of Columbia. The cross-sectional approach allows researchers to consider the effect of spatially dependent variables such as climate, population growth, and the existence of conservation education programs. Previous studies have found cross-sectional data elasticity estimates to be consistent with estimates from time series data (Chung et al, 2005; Espey et al, 1997).

2012 © American Water Works Association

E448

Aubuchon & Roberson | http://dx.doi.org/10.5942/jawwa.2012.104.0101 Journal - American Water Works Association Peer-Reviewed

Using data from the AWWA–RFC survey, a measure of monthly consumption was calculated as the total gallons sold (on a monthly basis) divided by the number of residential service accounts. This measure is a pooled commercial–residential data point and offers a more comprehensive picture of total capacity constraints for a given utility. This dependent variable is useful to utility managers because it allows a better understanding of how total system demand is correlated with residential prices. According to data from the AWWA–RFC rate study, only 20% of nonresidential water users face an IBR compared with 40% of residential customers. This difference in pricing structure can depict conflicting goals because water managers may implement conservation measures for residential users but encourage economic growth based on the principles of a “water economy” that promotes access to adequate water supplies for commercial and industrial use (Walton, 2010; Beecher, 2010b). This report does not consider the price elasticity of water for industrial consumers. Interested readers should consult De Rooy (1974) or Renzetti (1992). Instead, elasticity estimates in this study are interpreted as changes in aggregate monthly utility demand, and the effects of all other prices were held constant. Because industrial demand is not expected to change with changes in residential prices, the reported price elasticity estimates will be conservative estimates of the price elasticity of residential demand. The AWWA–RFC survey provides consumption information for the total water bill delineated at 0, 500, 1,000, 1,500, and 3,000 cu ft of consumption. Following Nieswiadomy and Cobb, marginal price was calculated as the change in total bill between two consumption tiers divided by the change in quantity for the same two tiers. Average price was calculated at the midpoint between consumption tiers as total bill divided by quantity. Each utility was assigned the marginal price and average price that corresponded to its monthly aggregate demand. For example, a utility with a monthly per account demand of 1,200 cu ft would be assigned a marginal price calculated as “total bill of 1,500 cu ft minus total bill of 1,000 cu ft” divided by “1,500 cu ft minus 1,000 cu ft.” The same utility would be assigned an average price calculated as the average of prices at the 1,000 and 1,500 cu ft reporting levels. Because the survey data provide prices only at these fixed consumption levels, the calculated marginal price may represent an average marginal price for some utilities if they included more than one price block in the AWWA–RFC consumption level. Therefore, care should be taken when the marginal prices of individual utilities are compared with those in this study. The advantage of the AWWA–RFC approach, however, is that it allows a more uniform comparison of price differences across utilities. Demand was also matched with the ratio of residential to industrial customers, the ratio of maximum production to average production, the price structure (either IBR, DBR, uniform, or flat), the number of blocks, a series of dummy variables indicating nonprice controls such as billing frequency or type of bill (water only or water and wastewater combined), and the presence of a demand management or conservation education program. Because climate plays a significant role in discretionary and seasonal water demand, utilities were also matched with their corresponding National Oceanic and Atmospheric Administra-

tion (NOAA) climate division. (Data accessed through the NOAA National Environmental Satellite Data and Information Service’s National Climatic Data Center: www.ncdc.noaa.gov/climatemonitoring/index.php#drought-icon.) NOAA has designated 344 climate divisions composed of geographically similar monitoring stations. This approach provides the most accurate and unbiased climate data at the level of metropolitan statistical area (MSA). From these data, the researchers calculated the 10-year (2000–09) average precipitation (inches), temperature (degrees Fahrenheit), and Palmer Hydrological Drought Index (PHDI). The PHDI is a measure of the severity of a wet or dry spell and ranges from –6 (extreme drought) to +6 (extreme flooding), with values from –0.5 to 0.5 considered normal. To test the hypothesis that faster-growing cities more readily adopt conservation rate structures, utilities were matched with the estimated rate of population growth between 2000 and 2009 at the MSA level, as defined by the US Census Bureau. Median household income from the five-year American Community Survey (www.census.gov/acs/www/) was also matched with utilities at the MSA level and was included as a control for the income elasticity of water demand. Because both climate and income data were measured at the MSA or larger level and water demand was reported at the utility level, the possibility for measurement error may exist. Any utility that did not report a value for average production, total gallons of water sold, or number of residential accounts was excluded from the analysis. Investor-owned utilities were also removed from the sample because they often face different tax and equity return schedules than publicly owned utilities. This restricted the sample to 243 utilities, with 45 utilities in the Midwest, 18 in the Northeast, 109 in the South, and 71 in the West. Table 1 provides descriptive statistics for the sample utilities, and the map on page E450 shows the distribution of these utilities across the United States. Half of the sample utilities had a conservation rate structure (n = 120) and those that did tended to be larger utilities. Two utilities had fewer than 10,000 accounts (one had a conservation rate structure), 40 utilities served 10,000 to 50,000 accounts (nine had conservation rate structures), 58 utilities served 50,000 to 100,000 accounts (30 had conservation rate structures), and 143 utilities served more than 100,000 accounts (79 had conservation rate structures). On average, 31% (73) of the sample utilities used groundwater as their primary source of supply. The rest of the utilities relied primarily on surface water sources. The fact that the AWWA–RFC survey was not a random survey of utilities may introduce bias into the current study if inclusion in the survey was somehow systematically correlated with aggregate demand. Compared with the full population of community water systems in the US Environmental Protection Agency’s Safe Drinking Water Information System database, the AWWA–RFC survey oversampled utilities in the West and undersampled utilities in the Midwest (Table 2). The practical implication of the survey design is to limit the ability of this study to be generalized to the broader universe of utilities. Thus, caution should be used in extrapolating these findings to any specific utility. This report considers the regional source of the sampling bias and provides

2012 © American Water Works Association

E449

Aubuchon & Roberson | http://dx.doi.org/10.5942/jawwa.2012.104.0101 Journal - American Water Works Association Peer-Reviewed

TABLE 1

Summary statistics for sample utilities

Variable

Number of Sample Utilities

Mean

Standard Deviation

243

12,848

6,253

Monthly demand per residential account—gal Conservation rate adoption vector

241

0.48

0.29

Residential accounts divided by total accounts

243

0.91

0.06

Median household income, 2009—$

238

50,977

9,579

Number of residential accounts

243

79,649

109,137

Net utility income per account—$

243

65.10

194.63

Ratio of maximum production to average daily production

243

1.76

2.41

Utilities relying on groundwater sources—%

243

31.66

41.34

Population growth rate, 2000–09—%

243

11

11

MP—$/1,000 gal

243

3.21

1.62

AP—$/1,000 gal

243

6.95

2.71

Average number of blocks in an IBR

243

1.73

2.22

Average monthly PHDI, 2000–09

241

0.08

2.21

Average monthly precipitation, 2000–09—in.

241

3.48

1.36

Average monthly temperature, 2000–09—°F

241

63.06

8.76

120

3.73

1.79

Utilities With a Conservation Rate Structure MP for utility demand per account— $/1,000 gal AP for utility demand per capita—$/1,000 gal

120

3.63

1.37

MP 1, for consumption of 0–500 cu ft—$/1,000 gal

120

4.58

1.76

MP 2, for consumption of 500–1,000 cu ft—$/1,000 gal

120

2.72

1.29

MP 3, for consumption of 1,000–1,500 cu ft—$/1,000 gal

120

3.39

1.82

MP 4, for consumption of 1,500–3,000 cu ft—$/1,000 gal

120

4.08

2.35

No

Yes

Utilities with a conservation rate structure

127

116

Utilities located in the West

172

71

Utilities billing customers monthly

72

171

Utilities billing for water and wastewater separately

222

21

Utilities with demand management programs

193

50

Binary Variables

AP—average price, IBR—inclining block rate, MP—marginal price, PHDI—Palmer Hydrological Drought Index

results aggregated by rate structure rather than by region. It is unclear whether the current survey under- or oversampled utilities by rate structure, but if western utilities are more likely to have a conservation rate structure, they are also more likely to be represented in the current sample. Therefore, this research may be most generalizable to western utilities.

METHODOLOGY Selectivity parameters. Previous studies have noted that price elasticity may be endogenous with a utility’s price structure choice (Olmstead et al, 2007). To control for this selection bias, a logistic model (model 1) was used to estimate the likelihood that a utility would adopt a conservation-oriented rate structure. The model consisted of utility-level variables for regions (1 = located in the West, 0 = otherwise), the ratio of maximum to average capacity, percentage of groundwater sources, number of accounts, net utility income per account, and ratio of

residential accounts to total accounts, plus MSA-level variables for population growth (2000–09), median household income, 10-year average PHDI, 10-year average precipitation, and 10-year average temperature. Among the independent variables, the ratio of maximum production to average production represents peak demand; utilities with higher peak demand have been found to be more likely to adopt a conservation rate structure (Teodoro, 2010). Regions with faster population growth should be more likely to adopt conservation rate structures in order to avoid costly plant and distribution system expansions and to forecast future demand more accurately (Beecher, 2010a; Chesnutt & Beecher, 1998). Mullin (2008) demonstrated that utilities with higher ratios of residential customers were more likely to adopt conservation rate structures. Because population growth and the ratio of residential customers are likely correlated, population growth is hypothesized to be the more important variable.

2012 © American Water Works Association

Aubuchon & Roberson | http://dx.doi.org/10.5942/jawwa.2012.104.0101 Journal - American Water Works Association Peer-Reviewed

E450

Mullin found that specialized governance districts are more likely to adopt a conservation rate structure. Demand specifications. Price elasticity was estimated within an OLS framework, a statistical method that minimizes the sum of the squared error, measured as the distance between each data point and the regression line (see Rockaway et al [2011] for an overview of the OLS specification). If certain statistical assumptions are met, OLS produces estimates that are both unbiased and best—that is, they converge to the true value of the population and contain minimum variance among all other estimators. A major assumption in OLS demand estimation is that the independent price variable is not related to the error term. Under an IBR, this assumption may not be satisfied because when consumers Source: AWWA & Raftelis Financial Consulting, 2010 Water and Wastewater Rate Survey choose their level of consumption, The distribution of US water utilities selected for this study is shown in this map. The red dots indicate they simultaneously become subject utilities with a conservation rate structure; the gray dots indicate some other type of rate structure. to a different marginal price. This endogeneity represents a potential source of bias. At the utility scale, however, prices and price schedules are fixed before demand, and therefore consumer-level endogeneity is not a problem in this utilityTeodoro (2010) demonstrated that number of accounts, net level study. Shin (1985) also posited that the use of aggregate data utility income per account, and percentage of groundwater minimizes any endogeneity between price and quantity in consources were all positively related to a utility’s likelihood of sumer choices. Instead, the choice of rate schedule may be endogadopting an IBR structure. Larger utilities, measured by the enous with demand, particularly under conservation rate strucnumber of accounts or the net utility income per account, may tures designed to manage total demand. Inclusion of a vector for be able to afford more in-depth rate analyses and presentations conservation rate adoption, measured as the probability of adoptto local boards, as well as the higher administrative costs associing a conservation rate structure from the logistic regression ated with more complicated rate structures. Consistent with (model 1), is an important control for this potential endogeneity. previous research (Teodoro, 2010; Mullin, 2008), the number of To consider the effect of price and nonprice controls (such as accounts and median household income were included in the log conservation education programs and billing frequency) on form because household size and income effects are theorized to demand while controlling for the composition of a utility’s cusbe nonlinear in nature. Teodoro also hypothesized that utilities tomer class and regional climate variables, the researchers develrelying on groundwater sources are more vulnerable to concerns oped the following equation: about scarcity than utilities relying on surface water sources and that those relying on groundwater are more likely to implement conservation measures for long-term source water protection. Log (Aggregate Demand per Account) = (1) Utilities in regions with more precipitation, less risk of drought,                                    b1 + biXi + bjXj + bkXk + ei and lower temperatures should be less likely to implement conservation rate structures because peak outdoor watering demands in which Xi represents the utility-specific variables including the are lower in these areas. This study does not consider the relaratio of residential accounts to total accounts, the probability of tionship between local government structure and adoption of a selecting a conservation rate structure that includes information conservation rate structure, which may be a potential source of on regional climate conditions and population growth derived bias. Teodoro found that among local government frameworks, from the logistic regression, and a regional dummy variable for cities with a council–manager charter or an at-large election are utilities in the West; Xj represents the log of the price variables more likely to adopt a conservation rate structure. Similarly, for each utility; Xk represents indicator variables for common 2012 © American Water Works Association

E451

Aubuchon & Roberson | http://dx.doi.org/10.5942/jawwa.2012.104.0101 Journal - American Water Works Association Peer-Reviewed

TABLE 2

Region 

US water utilities by region SDWIS Expected Number of Utilities Proportion if Sample Included All US % Community Water Systems

Sample Utilities

North

13

  30.11

 18

South

44

 102.85

109

Midwest

29

  67.85

 45

West

14

  32.87

 71

2 of sample—57.15 2 (p value = 0.01, 3 DF)—16.27 DF—degrees of freedom, p value—indicates the statistical significance of the finding, SDWIS—Safe Drinking Water Information Statistics (2009 database), 2—chi-square test statistic

nonprice controls on demand, such as monthly billing frequency, separate water and wastewater bills, and demand management programs; b1, bi , bj, and bk represent the parameter estimates of interest, and ei represents the error term. To investigate the correlation of total demand with various price signals, this study used the Shin price perception parameter. Under standard microeconomic theory, consumers should re­­ spond to the marginal price of consumption, but because assimilating information beyond total price requires time and energy on the part of consumers, Shin (and others) hypothesized that consumers may react more readily to the average price, which can easily be calculated as total price divided by total demand. Shin (1985) defined the price perception parameter as P*:



Average Price   P* = Marginal Price ×  Marginal Price



k

(2)

When k = 0, P* is equal to marginal price; when k = 1, P* is equal to average price. If k is between 0 and 1, demand is correlated with some combination of marginal and average price, and if k > 1, demand is correlated with some price signal other than marginal or average price. It is unclear what this other price signal might be, but Olmstead et al (2007) found that nearly 40% of demand observations lie within 5% of a change in marginal price. This suggests that consumers may be responding to the perceived price of higher consumption tiers. To estimate k, P* is entered into the demand equation in log (represented as ln) form.





Average Price i ln (P*) = i ln (Marginal Price) + i k × ln  (3) Marginal Price

Within an OLS model, the parameters indicate how a one-unit change in the X variable affects the variable for monthly household water demand, when the effects of all other variables are held constant. When both demand and price are expressed in the log– log model, all parameter estimates indicate a percent change in the quantity demanded for a 1% change in the price variable. The other variables indicate the semi-elasticity of demand: a one-unit

change in the X independent variable causes demand to change by 100 × bi percent. Because the dependent variable includes both residential and nonresidential demand divided by the total number of system accounts, the elasticity estimates provided in this study may help decision-makers better understand the effect of residential prices and rate structures on total system demand. This finding can be used to balance tradeoffs associated with revenue stability and conservation for the utility as a whole. Regions with higher amounts of precipitation and less drought risk have been shown to experience lower overall water consumption (Arbués et al, 2003). The current study controlled for average precipitation from 2000 to 2009, average temperature from 2000 to 2009, and aggregate PHDI from April to September 2010. Although the PHDI measurement captured the effect of local climate conditions on water demand in 2010, long-term precipitation and temperature averages are important in shaping customer expectations of local irrigation needs resulting from seasonal weather patterns. Demand management consists of both price and nonprice controls, including drought restrictions or water rationing, education campaigns, and rebates or subsidies for households that install water-efficient fixtures and appliances (Olmstead & Stavins, 2009). Because implementing new conservation-oriented rate structures can be politically difficult, it is important for utilities to understand the relative effectiveness of both price and nonprice conservation controls at reducing total system demand. As for nonprice controls, several studies have found that monthly billing raises awareness about water consumption and thereby lowers overall consumption (Howe, 2005; Arbués et al, 2003), whereas previous studies have found mixed results with regard to demand management or conservation-focused education programs. Although conservation education programs have been found to be insignificant in a statistical sense (Nieswiadomy, 1992), they are commonly considered an important component in establishing long-term change (Brandes et al, 2010). Within the context of the year during which this study was conducted, it is unclear what impact conservation programs will have on demand.

RESULTS Table 3 shows results of the logistic regressions. After controlling for population growth, and in contrast to the work of Teodoro and Mullin, the number of accounts is not related in a statistically significant way to a utility’s likelihood of adopting a conservation-oriented rate structure. However, the number of accounts does include the correct sign (the direction of the parameter estimate—whether it is positive or negative), and larger utilities are more likely to have conservation-oriented rates. Median household income also had no statistical effect on a utility’s likelihood of adopting a conservation rate structure. The negative relationship between household income and conservation rates may provide some support to Mullin’s hypothesis that opposition to these rates comes from high-income residents. Model 2 includes the statistically significant variables from model 1. The effect of population growth is the emphasis of the current study and was the most influential variable in both

2012 © American Water Works Association

E452

Aubuchon & Roberson | http://dx.doi.org/10.5942/jawwa.2012.104.0101 Journal - American Water Works Association Peer-Reviewed

TABLE 3

Estimates of a utility’s likelihood of selecting a conservation rate structure Odds Ratio Estimates Model 2

Model 1 (Standard Error)

Model 2 (Standard Error)

Located in the West

  1.08   (0.62)

  1.27*  (0.58)

  3.57

Residential accounts divided by total accounts

–1.10  (2.96)

Log (accounts)

 0.22  (0.18)

  5.33‡  (1.66)

205.31

  0.77*  (0.37)

  2.19

Variable

    –0.000526      (0.000768)

Net utility income per account Population growth rate —%

 5.26†  (1.88)

Log (median household income)

 –0.33  (1.02)

Ratio of maximum production to average daily production

 0.92*  (0.40)  –0.0014   (0.0042)

Ratio of groundwater sources to surface water sources Average monthly PHDI

 0.37  (0.22)

 0.40  (0.21)

  1.493

Average monthly precipitation—in.

–0.48‡   (0.18)

 –0.42*  (0.17)

  0.66

Average monthly temperature—°F

 0.16†  (0.03)

  0.16†   (0.028)

  1.17

Constant

–8.46 (11.40)

–10.98†  (2.12)

242

247

Likelihood ratio (chi-squared statistic)

98.5†

96.1†

Cases correctly predicted—%

85.4

84.7

Number of utilities

p value—indicates the statistical significance of the finding, PHDI—Palmer Hydrological Drought Index *p value = < 0.05 †p value = < 0.001 ‡p value = < 0.01

model 1 and 2. Dropping statistically insignificant variables does not drastically reduce model fit, as measured by the chisquared likelihood ratio, and allows for a more intuitive interpretation of the vector for adoption of conservation rates included in later models. On the basis of the median values of these independent variables, 37% of utilities would be expected to have a conservation rate structure. Table 3, column 2, shows the maximum likelihood estimates, which indicate a unit change in the odds ratio; column 3 shows the odds ratio estimates. The odds ratio estimate indicates the multiplicative effect on the odds ratio of a one-unit change in the independent variable. If the population growth rate increases by 10 percentage points (0.1 unit), the likelihood ratio increases by a factor of 20.5. Therefore, the faster a city grows, regardless of region or climate, the greater the likelihood that a given utility in that city has a conservation rate structure. For example, a utility with a 10-year population growth rate of 20.3%—10 percentage points higher than that of the median utility—is 13% more likely to have a conservation rate structure.

Population growth often results in increasing costs for utilities, but an IBR or seasonal rate structure can be a cost-effective way to control peak demand and delay additional infrastructure investments. Accurate forecasts of population growth should therefore be an important component of any conservation rate proposal. If the median utility (with a population growth rate of 10.3%) is located in the West, it would be 68% likely to have an IBR. This suggests that regional attributes play a significant role either in the decision to adopt a conservation rate structure or in the ability to implement it, when the effects of population growth and climate are held constant. For utilities in cities that are growing faster than the median and are located in the West, the probability of their adopting a conservation rate structure increases substantially to 78%. Table 4 shows the modeled probability of a utility adopting a conservation rate structure under other scenarios. Table 5 gives the results of Eq 1 estimated with marginal price, average price, and Shin’s price perception parameter for the subsamples of utilities with conservation rate structures and those

2012 © American Water Works Association

E453

Aubuchon & Roberson | http://dx.doi.org/10.5942/jawwa.2012.104.0101 Journal - American Water Works Association Peer-Reviewed

TABLE 4

Modeled probability of a utility having a conservation rate structure (model 2) Variables Evaluated

Modeled Probability %

Median value

37

With 20% population growth

50

Located in the West

68

With 20% population growth in the West

78

With PHDI = –1

27

With PHDI = –1, 20% population growth, and located in the West

69

PHDI—Palmer Hydrogical Drought Index

without (models 3–8). Separating the sample by rate structure leads to a better model fit and allows for a comparison of the price perception parameter by rate structure. After initial regressions, six observations were dropped as outliers, determined by a Cook’s difference value greater than 4/250, a conservative estimate of sample size. (Cook’s difference is a measure of how much effect the residual has on the overall model.) The final models correctly explain 36–50% of the variation in monthly water demand, a value consistent with previous studies (Gaudin, 2006). The vector for likelihood of adopting a conservation rate structure is included from model 2. This vector addresses some of the previously noted concerns about statistical validity and is highly significant in all models. This suggests that if a utility is more likely to have a conservation rate structure, it is also more likely to have higher monthly household water demand. The conservation rate adoption vector also accounts for the correlation of climate with demand but helps to remove the effects of multicollinearity among climate, region, and median household income in the final OLS models. Multicollinearity occurs when independent variables are highly correlated. It does not violate any of the necessary assumptions for OLS but instead increases the standard errors on correlated parameter estimates. Larger standard errors can lead to lower t-statistics, greater acceptance of the null hypothesis of no statistical significance, and an increase in type 2 errors (Gujarati & Porter, 2009). The significance of the conservation rate adoption vector was expected, even after the study controlled for climate and household income influences, because conservation rate structures are often implemented to decrease discretionary water use. The significance of this vector also suggests that previous studies that have ignored the endogeneity of price elasticity and rate structure selection may have produced biased estimates. None of the models showed any evidence of heteroscedasticity in the data, based on White’s test, which is shown as the chi-square statistic in Tables 5 and 6. This final check ensured that the model met the OLS assumptions. The price elasticity estimates in models 3–8 show several important conclusions. First, the price elasticity of demand is higher for both marginal price and average price under nonconservation rate structures than under conservation rate struc-

tures. More important, under nonconservation rate structures, model 8 indicates that the price perception parameter is not statistically different from model 1, and aggregate demand is correlated with average price when the effects of climate, region, household income, and conservation education programs are held constant. This finding is consistent with previous research, and according to estimates from model 7, a 10% increase in average price is correlated with a utility-level reduction in aggregate demand of roughly 3.3%. More surprising is the estimate of k from model 5. In the current sample, an F-test rejects the null hypothesis that the price perception is equal to either 0 (marginal price) or 1 (average price). Instead, demand is correlated with something other than marginal or average price under conservation rate structures. The number of blocks in the rate structure has no impact on overall demand, which lends support to the idea that it is ultimately the price signal and not the number of blocks in the rate structure that is correlated with demand. This does not suggest, however, that rate design and the number of blocks are insignificant in satisfying other utility goals such as revenue stability or affordability. To the extent that higher demand influences rate design, there may be some endogeneity (and therefore bias) between the number of blocks and demand. One hypothesis is that aggregate demand may be correlated with either the total bill or the price of the highest marginal consumption tier instead of average or marginal price. To investigate this claim, Eq 1 was estimated independently for utilities with conservation rate structures at each marginal price and total bill reported in the AWWA–RFC survey. Table 6 shows results from models 9–16. For both measures of price (total bill and marginal price), the relationship of price and demand is strongest at the 1,500 cu ft consumption level. With consumption at 1,500 cu ft (model 11), a 10% increase in the total bill is associated with a 2.7% decrease in demand. These models provide preliminary evidence that utility demand is related to the price signal embedded in the total bill, the marginal price for higher levels of consumption, or both. This could be consistent with Shin’s hypothesis that because assimilating additional information “is costly” in terms of time and energy, consumers likely fall back on the easiest or most available price signal. In the past, this may have been average price, but as conservation rate structures have become more prevalent, the most meaningful or memorable price signal may be associated with the higher prices of a rate schedule. Future research should investigate this relationship more carefully using household-level data. Within the current sample of utilities, and when the study controlled for prices and household income, there is no statistical evidence that nonprice controls affected residential water demand, but the researchers note that the variables all have the expected sign. Monthly billing and demand management programs are correlated with reductions in water consumption, whereas separate water and wastewater bills are correlated with greater water consumption. A monthly bill provides the price signal more frequently, but a water-only bill will appear to have a smaller price than a combined water and wastewater bill. Both

2012 © American Water Works Association

E454

Aubuchon & Roberson | http://dx.doi.org/10.5942/jawwa.2012.104.0101 Journal - American Water Works Association Peer-Reviewed

TABLE 5

Shin price perception parameter and determinants of water demand Utilities With Conservation Rate Structure (Standard Error) Demand Per Account

Utilities Without Conservation Rate Structure (Standard Error)

Model 3

Model 4

Model 5

Model 6

Model 7

Model 8

Residential accounts divided by total accounts

–1.67* (0.51)

–1.55* (0.50)

–1.42* (0.49)

–1.06† (0.50)

–1.03† (0.48)

–1.02† (0.49)

Log (median household income)

0.24 (0.21)

0.29 (0.21)

0.25 (0.20)

 0.077 (0.15)

0.10 (0.15)

0.10 (0.15)

Conservation rate adoption vector

0.50* (0.16)

0.53‡ (0.15)

 0.50‡ (0.15)

 0.38† (0.15)

0.37† (0.15)

0.37† (0.15)

Located in the West

0.33‡ (0.07)

 0.32‡  (0.066)

 0.34‡  (0.065)

 0.36‡  (0.091)

 0.38‡  (0.088)

0.37‡ (0.089)

Log (marginal price)

–0.088  (0.061)

–0.28‡  (0.081)

–0.23‡  (0.057)

–0.21*  (0.078)

Log (average price)

–0.34‡ (0.069) –0.34‡  (0.069) –0.31† (0.12)

–0.60* (0.18)

Log (average price divided by marginal price) –0.010  (0.018)

 –0.0098  (0.017)

–0.016  (0.017)

–0.077 (0.10)

–0.087  (0.099)

–0.090  (0.097)

–0.021  (0.062)

–0.030  (0.060)

–0.029 (0.060)

Billed separately for water and wastewater

  0.0055 (0.13)

 0.015 (0.12)

 0.031 (0.12)

 0.31* (0.10)

0.30*  (0.098)

0.30* (0.099)

Demand management programs

–0.051  (0.067)

–0.051  (0.065)

–0.068  (0.064)

–0.043  (0.097)

 0.004  (0.096)

0.00067 (0.097)

7.50* (2.45)

 6.35* (2.41)

 6.55*  (2.35)

 7.82‡   (1.70)

 7.19‡ (1.68)

7.184‡ (1.68)

112

112

112

124

124

Increasing consumption blocks Billed monthly

Constant Number of utilities r2

124

0.36

0.39

0.42

0.46

 0.489

0.49

F statistic

  7.99*

 8.99*

 9.02*

14.04*

15.69*

13.85*

2

40.08

39.94

70.04

41.95

43.53

61.00

Adjusted

Shin price perception parameter Ho: k = 0

Ho: k = 1

2.17

0.91

F statistic

18.51‡

10.25*

p value

 0.000

 0.002

F statistic

 5.39†

0.10

p value

0.02

0.75

Ho: k = 0—statistical representation of “the null hypothesis is k = 0,” Ho: k = 1—statistical representation of “the null hypothesis is k = 1,” p value—indicates the statistical significance of the finding, r 2 —correlation coefficient, 2—chi-squared statistic for heteroscedasticity *p value = < 0.01 †p value = < 0.05 ‡p value = < 0.001

of these changes in the total bill are correlated with expected changes in demand, providing further evidence that the total bill price signal matters. The magnitude of these semi-elasticities is also greater than that of price elasticity. When the effects of price and household income are held constant, monthly billing is correlated with a 7–9% reduction in demand (although this is not statistically significant), whereas a separate bill is correlated with a greater demand of 1–4%. The estimate for the parameter of demand management programs is not statistically significant but suggests that the presence of such a program is correlated with a 3–6% reduction in

demand. This suggests that any reduction in demand from nonprice controls may be short-lived, not adequately measured in the data set, or not statistically different from random fluctuations. From a benefit–cost perspective, utility managers may be unwise to commit significant financial resources to nonprice demand management controls without also including a more permanent price signal. Finally, income elasticity is positive but not statistically significant in any model. The income elasticity parameter under conservation rate structures is between 0.22 and 0.38 and, similar to price elasticity, increased at higher tiers of consumption. In most cases, the magnitude of the effect of income

2012 © American Water Works Association

E455

Aubuchon & Roberson | http://dx.doi.org/10.5942/jawwa.2012.104.0101 Journal - American Water Works Association Peer-Reviewed

TABLE 6

Total bill and marginal price by consumption block for utilities with a conservation rate structure Model 9 (Standard Error)

Model 10 (Standard Error)

Model 11 (Standard Error)

Model 12 (Standard Error)

Model 13 (Standard Error)

Model 14 (Standard Error)

Model 15 (Standard Error)

Model 16 (Standard Error)

Residential accounts divided by   total accounts

–1.58* (0.51)

–1.48* (0.51)

–1.42* (0.50)

–1.56* (0.49)

–1.58* (0.51)

–1.41* (0.50)

–1.42* (0.48)

–1.66* (0.49)

Log (median household income)

0.22 (0.21)

0.26 (0.21)

0.32 (0.21)

0.32 (0.21)

0.22 (0.21)

0.32 (0.20)

0.38 (0.20)

0.29 (0.21)

Conservation rate adoption vector

0.52* (0.16)

0.51* (0.15)

0.53† (0.15)

0.53† (0.15)

0.52* (0.16)

0.50* (0.15)

0.52† (0.15)

0.52† (0.15)

Located in the West

0.36† (0.067)

0.34† (0.065)

0.32†‡ (0.065)

0.30† (0.065)

0.36† (0.067)

0.31† (0.066)

0.27† (0.066)

0.30† (0.066)

Increasing consumption blocks

–0.015 (0.018)

–0.015 (0.017)

–0.011 (0.017)

–0.0061 (0.017)

–0.015 (0.018)

–0.013 (0.017)

–0.0043 (0.017)

–0.0057 (0.017)

Billed monthly

–0.071 (0.10)

–0.083 (0.10)

–0.089 (0.098)

–0.078 (0.10)

–0.071 (0.10)

–0.085 (0.098)

–0.085 (0.096)

–0.074 (0.099)

Billed separately for water and  wastewater

0.026 (0.13)

0.0069 (0.12)

0.018 (0.12)

0.0080 (0.12)

0.026 (0.13)

–0.080 (0.12)

0.037 (0.12)

0.0038 (0.12)

Demand management measures

–0.064 (0.066)

–0.062 (0.066)

–0.059 (0.064)

–0.040 (0.064)

–0.064 (0.066)

–0.062 (0.064)

–0.054 (0.063)

–0.034 (0.065)

Log (total bill = 500 cu ft)

–0.14 (0.082)

Demand per Account

Log (MP 1) Log (MP 2)

–0.21‡ (0.088)

Log (total bill = 1,000 cu ft)

–0.23* (0.069)

Log (total bill = 3,000 cu ft) Constant Number of utilities Adjusted

r2

F statistic 2

–0.22† (0.064)

Log (MP 3)

–0.27* (0.083)

Log (total bill = 1,500 cu ft)

–0.14 (0.082)

–0.24† (0.06)

Log (MP 4)

–0.16* (0.055)

8.54†‡ (2.29)

8.38† (2.27)

7.92† (2.23)

8.03† (2.22)

7.37* (2.42)

5.75‡ (2.37)

4.92‡ (2.36)

6.52* (2.35)

112

112

112

112

112

111

112

112

0.37

0.38

0.41

0.41

0.37

0.42

0.44

0.40

8.18

8.66

9.55

9.63

8.18

9.80

10.59

9.24

38.77

39.64

43.07

40.29

38.77

60.42

38.63

37.51

MP—marginal price, p value—indicates the statistical significance of the finding, r2—correlation coefficient, 2—chi-squared statistic for heteroscedasticity *p value = < 0.01 †p value = < 0.001 ‡p value = < 0.05

elasticity is greater than that of price elasticity. Therefore, to effectively reduce demand over the long run, utility managers should attempt to increase rates at or above relative growth rates in regional household income.

CONCLUSION AWWA’s 2010 State of the Industry Report ranked business factors, infrastructure spending, and new source development as the top three long-term challenges for the drinking water community (Mann & Runge, 2010). Adoption of a conservation rate structure helps to address all three of these concerns. The most significant factor that might prompt a utility to consider adopting a conservation rate structure, even after the variables of size, region, and climate are held constant, is population growth. Utilities experiencing rapid population growth should consider a conservation rate structure as one tool to manage peak demand and to potentially delay costly infrastructure investments or the need for new source development.

In addition to addressing these stated industry concerns, this study provides new evidence that the price signal contained within rate structures matters for conservation goals and that under conservation rate structures, demand is correlated with something other than average or marginal price. Preliminary evidence suggests that the price signal of the total bill or the marginal price at higher tiers of consumption may be more relevant for communicating conservation goals. Future work should consider the impact of these different price signals on demand. Finally, within the current study sample and with the effects of price and household income held constant, demand management programs did not appear to be statistically correlated with demand, but the magnitude of the parameter estimates suggests that these programs may reduce demand by 3–6%. Utility managers interested in reducing aggregate demand should carefully consider the benefits and costs associated with conservation education programs and attempt to embed a more permanent price signal in such efforts.

2012 © American Water Works Association

E456

Aubuchon & Roberson | http://dx.doi.org/10.5942/jawwa.2012.104.0101 Journal - American Water Works Association Peer-Reviewed

ACKNOWLEDGMENT The authors thank Rocky Craley of Raftelis Financial Consultants and several members of AWWA’s Rates and Charges Committee for helpful comments during previous drafts of this report. Craig Aubuchon’s internship with AWWA was funded by the Water Industry Technical Action Fund, administered by AWWA, and supported by the dues of AWWA organizational members. The fund supports information collection and analysis and other activities in support of sound and effective legislation, regulation, and drinking water policies and programs.

About the authors Craig Aubuchon (to whom correspondence should be addressed) is an associate with Analysis Group, 10th Floor, 111 Huntington Ave., Boston, MA 02199; craig.aubuchon@ gmail.com. He wrote this article during a summer internship in AWWA’s Washington office. He holds a master’s degree in public administration and environmental science with an emphasis on water resources from the School of Public and Environmental Affairs, Indiana University, Bloomington, Ind., and a bachelor’s degree in economics from Washington University, St. Louis. J. Alan Roberson is AWWA’s director of federal relations in Washington.

Peer review Date of submission: 08/12/2011 Date of acceptance: 05/30/2012

Chung, C.; Dong, D.; Schmit, T.M.; Kaiser, H.M.; & Gould, B.W., 2005. Estimation of Price Elasticities From Cross-sectional Data. Agribusiness, 21:4:565. Cooley, H. & Gleick, P.H., 2009. Urban Water Use Efficiencies: Lessons From United States Cities. The World’s Water: 2008–2009. The Biennial Report on Freshwater Resources. Island Press, Washington. Dalhuisen, J.M.; Florax, R.; de Groot, H.; & Nijkamp, P., 2003. Price and Income Elasticities of Residential Water Demand: A Meta-Analysis. Land Economics, 79:2:292. De Rooy, J., 1974. Price Responsiveness of the Industrial Demand for Water. Water Resour. Res., 10:3:403. Espey, M.; Espey, J.; & Shaw, W.D., 1997. Price Elasticity of Residential Demand for Water: A Meta-Analysis. Water Resour. Res., 33:6:1369. Gaudin, S., 2006. Effect of Price Information on Residential Water Demand. Applied Economics, 38:4:383. Gleick, P.H., 2010. Roadmap for Sustainable Water Resources in Southwestern North America. Proc. Natl. Acad. Sci., 107:50:21300. Gleick, P.H., 2003. Global Freshwater Resources: Soft-Path Solutions for the 21st Century. Science, 302:5650:1524. Gujarati, D.N. & Porter, D.C., 2009 (5th ed.). Basic Econometrics. McGraw-Hill, New York. Howe, C.W., 2005. The Functions, Impacts, and Effectiveness of Water Pricing: Evidence From the United States and Canada. Water Resource Devel., 21:1:43. Howe, C.W. & Linaweaver, F.P., 1967. The Impact of Price on Residential Water Demand and Its Relation to System Design and Price Structure. Water Resource Res., 3:1:13. Mann, J. & Runge, J., 2010. State of the Industry Report 2010: How Water Professionals Are Meeting Ongoing Challenges and Economic Uncertainty. Jour. AWWA, 102:10:44. Mullin, M., 2008. The Conditional Effect of Specialized Governance on Public Policy. American Jour. Political Sci., 52:1:125. Nieswiadomy, M. & Cobb, S.L., 1993. Impact of Pricing Structure Selectivity on Urban Water Demand. Contemporary Economic Policy, 11:3:101.

REFERENCES Arbués, F.; Garcia-Valinas, M.A.; & Martinez-Espineira, R., 2003. Estimation of Residential Water Demand: A State of the Art Review. Jour. Socio-Economics, 32:1:81. AWWA Climate Change Committee, 2011. Committee Report: Sustainability of Water Resources Depends on Implementing Our Knowledge on Climate Variability. Jour. AWWA, 103:6:42. AWWA & RFC, 2010. 2010 Water and Wastewater Rate Survey. AWWA, Denver, & Raftelis Financial Consulting, Charlotte, N.C. AWWA, 2000. M1: Principles of Water Rates, Fees and Charges: Manual of Water Supply Practices. AWWA, Denver. Beecher, J.A., 2012. The Ironic Economics and Equity of Water Budget Rates. Jour. AWWA, 104:2:73. Beecher, J.A., 2010a. Water Pricing Primer for the Great Lakes Region. Alliance for Water Efficiency, Chicago. http://www.allianceforwaterefficiency.org/ (accessed June 2011).

Nieswiadomy, M., 1992. Estimating Urban Residential Water Demand: Effects of Price Structure, Conservation, and Education. Water Resource Res., 28:3:609. Nieswiadomy, M. & Molina, D.J., 1991. A Note on Price Perception in Water Demand Models. Land Economics, 67:3:352. Olmstead, S. & Stavins, R.N., 2009. Comparing Price and Nonprice Approaches to Urban Water Conservation, Water Resource Res., 45, W04301, doi:10.1029/2008WR007227. Olmstead, S.M.; Hanemann, W.M.; & Stavins, R.N., 2007. Water Demand Under Alternative Price Structures. Jour. Envir. Economics & Mngmnt., 54:2:181. Renzetti, S., 1992. Estimating the Structure of Industrial Water Demands: The Case of Canadian Manufacturing. Land Economics, 68:4:396. Rockaway, T.D.; Coomes, P.A.; Rivard, J.; & Kornstein, B., 2011. Residential Water Use Trends in North America. Jour. AWWA, 103:2:76.

Beecher, J.A., 2010b. The Conservation Conundrum: How Declining Demand Affects Water Utilities. Jour. AWWA, 102:2:78.

Shin, J-S., 1985. Perception of Price When Price Information Is Costly: Evidence From Residential Electricity Demand. Review Economics & Statistics, 67:4:591.

Brandes, O.M.; Renzetti, S.; & Stinchcombe, K., 2010. Worth Every Penny: A Primer on Conservation-Oriented Water Pricing. Polis Project on Ecological Governance. University of Victoria, B.C. www.poliswaterproject.org (accessed June 2011).

Teodoro, M.P., 2010. The Institutional Politics of Water Conservation. Jour. AWWA, 102:2:98.

Brandes, O.M. & Brooks D.B., 2006. The Soft Path for Water: Backgrounder. Horizons, 9:1:71.

Walton, B., 2010. The Price of Water: A Comparison of Water Rates, Usage in 30 U.S. Cities. Circle of Blue. www.circleofblue.org/waternews/2010/world/ the-price-of-water-a-comparison-of-water-rates-usage-in-30-u-s-cities/ (accessed April 2010).

Chesnutt, T.W. & Beecher J.A., 1998. Conservation Rates in the Real World. Jour. AWWA, 90:2:60.

Worthington, A.C. & Hoffman, M., 2008. An Empirical Survey of Residential Water Demand Modeling. Jour. Economic Surveys, 22:5:842.

2012 © American Water Works Association