There are numerous studies regarding water pricing, demand and elasticity for certain regions. However, in Brazil, there are no studies on these matters, even though there is a need to explore the behavior of Brazil's population, especially because the nation is susceptible to extreme water events. São Paulo State, Brazil's most important economic region, has recently experienced a severe water scarcity status. In an attempt to control water demand, the São Paulo Water Agency (SABESP) implemented a ‘bonus and onus’ program. In this context, the aim of this study was to analyze the SABESP programs in terms of their structures and results using a panel model. The econometric results showed that (i) the bonus program was successful and more effective than the onus program, and (ii) water consumption reduction was more significant in regions supplied by water reservoirs where the relative water level was lower, although inhabitants of other regions also reduced their water consumption.

In Brazil, Art. 1 line I of the National Water Resources Policy (law n. 9.433, 1997) (Brasil, 1997) states that water is a public good, which indicates that public authorities are responsible for managing the water supply. Public goods, or environmental ones, have specific characteristics: they are non-rival (meaning that one person's consumption does not diminish the amount of this good available for others to consume) and non-excludable (it is not feasible and practical to selectively allow consumers to use this good) (Kolstad, 2011).

Water managers can choose cost incentives to control water demand in periods of scarcity. Water conservation policies based on economic incentives are more cost-effective than prescriptive policies, even when water prices and allocation across sectors are inefficient (Olmstead et al., 2005). Environmental cost internalization may be implemented through the adoption of command and control mechanisms and market mechanisms. Such mechanisms are complementary and non-exclusive (Motta & Young, 1997).

Traditional command and control regulation comprises the use of rules reinforced by legal sanctions (Baldwin et al., 2000). The authors explain that when a command and control regulation is used, the required behavior is stipulated, standards are fixed, unacceptable actions are defined and outlawed, and penalties for noncompliance are set out. Given environmental issues, command and control regulations (also called prescriptive regulations) establish limits for the use of natural resources, pollution levels, or other acts that may impact the environment (Kolstad, 2011). Their implementation must be overseen by an environmental authority through monitoring, with the possibility of applying penalties when the limits are not followed or demanding environmental damage remediation (Sette, 2014).

Economic incentives, in contrast, provide rewards for consumers/firms/polluters to do what is perceived to be in the public interest (Kolstad, 2011). The study of Gilbertson et al. (2011) on the behaviors of inhabitants in areas with and without water stress shows that economic incentives are effective in public policy. Regarding water issues, pricing may be an effective tool for managing demand (Sebri, 2014); however, such effectiveness depends on economic and social variables of the population (Metaxas & Charalambous, 2005).

Littlefair (1998) showed that variations in the perception of water price are clearly extensive, and it cannot be assumed that people attach the same value or cost for the provision of water at any time or place. Regarding this issue, Magnusson (2004) highlighted that accessibility to water and urban lifestyle are two factors influencing water use behavior more than economic demand management incentives, at a certain consumption level. The author's overall conclusion is that the willingness to respond to demand management is reduced with improved ability-to-pay and access to water. It is important to emphasize as well that, as highlighted by Venkatachalam (2006), consumption determinants, such as willingness to pay (WTP), are site-specific in nature.

Detailed knowledge about the price and income elasticities of residential water demand is available through a substantial number of empirical studies, but these estimated elasticities vary. Dalhuisen et al. (2003) note that such variation may result not only from the differences in spatial and temporal dynamics but also from differences in theoretical microeconomic choice approaches. These authors, as well as others (Gaudin, 2005; Worthington & Hoffman, 2008; Sebri, 2014), analyzed the relations of the sensitivities of price, income, and household size elasticities to various characteristics, including price specification, tariff structure, location of demand, and estimation technique.

There are numerous studies regarding water pricing, demand, and elasticity for certain regions. A region very well explored in such studies, for instance, is California (Renwick & Green, 2000; Nataraj & Hanemann, 2011; Lee & Tanverakul, 2015; Mini et al., 2015). In some developing countries, these issues have also been the object of study (Venkatachalam, 2006; Magnusson, 2004; Saz-Salazar et al., 2014), However, regarding Brazil, there are no studies and, as Sebri (2014) notes, once decision-makers in a given country would not rely on the findings of studies conducted in other countries in formulating their policies, there is a need to explore the behavior of the Brazilian population, especially because the nation is susceptible to extreme water events, which may reduce water availability for the population.

The city of São Paulo, one of the most important regions for the Brazilian economy, has recently experienced a severe water shortage status. From late 2013 until the beginning of 2015, São Paulo State has undergone a severe water crisis. It was the worst drought since 1930, when the water levels of springs started being measured. During this period, rainfall was below the historic average for several months. As a result, the levels of the three most important reservoirs that supplied inhabitants in the city (Cantareira, Alto Tiete, and Guarapiranga) had significant decreases in volume.

This fact influenced the municipality population lifestyle and required unprecedented governmental actions to reduce water consumption and increase water supply. Among these actions, in an attempt to control water demand, the São Paulo State Government added a ‘bonus and onus’ program. In this program, households that experienced a decrease in water consumption during a determined period would receive a discount on their water bill, while those that experienced an increase in water consumption would be penalized financially.

Considering this scenario, in the present study, we aimed to analyze such a program, its structure, and its impacts. The main goals were as follows: (1) analyze the program concerning its adherence and effectiveness and (2) compare the performances among households considering socio-economic variables.

After this brief introduction, the section below describes the situation of the water shortage in São Paulo during the period of 2014–2015, presenting information on the most important reservoirs used for population supply in the state and explaining the bonus and onus program implemented by the water and sanitation operator. The next section presents the panel method used in the analysis of program effectiveness. This is followed by a section ‘Results and Discussion’ showing the data on water consumption variation for each of the districts and the results from an explanatory analysis and panel analysis when considering each one of the variables determined in the previous section. Also, we present a discussion of the main results to understand consumer behavior regarding water demand during a water crisis. The final section presents a conclusion based on the results and literature research on this topic.

The program created by the São Paulo State Government was established when the levels of the three most important reservoirs that supply the city of São Paulo (Cantareira, Alto Tiete, and Guarapiranga) had significant volume decreases (Figure 1).

Fig. 1.

Status of the relative volumes of the three main water reservoirs in São Paulo State (volume/maximum capacity). Source: SABESP (2015).

Fig. 1.

Status of the relative volumes of the three main water reservoirs in São Paulo State (volume/maximum capacity). Source: SABESP (2015).

Close modal

As shown in Figure 1, in July 2014, Cantareira, which is the most important reservoir for São Paulo's population (supplies over 6.5 million people), reached a negative level, and the government had to start using its technical reserve, which is sited below the floodgates. Pumping systems are needed to retain the water of this technical reserve.

In this water scarcity context, the regulatory agency for water issues in São Paulo State (Agência Reguladora de Saneamento e Energia do Estado de São Paulo, or ARSESP) passed a bill authorizing SABESP to establish a program that provided a financial bonus on water bills for every household that reduced their water consumption below a determined level and, on the other side, applied an economic penalty to households that increased their consumption over a determined level (ARSESP, 2014). This program was used on a population of over 17 million people and 31 municipalities in the state and started in February 2014.

The bonus was based on three reduction levels related to the average consumption for each household between February 2013 and January 2014. Households with a water consumption reduction ranging between 10% and 15% would get a discount of 10% on their water bill. Those that reduced their consumption by 15% to 20% would get a discount of 15%, and every household that reduced the water consumption by more than 20% was awarded with a discount of 30% (Table 1).

Table 1.

Bonus and contingency rate program summary.

Consumption reductionBonus
 10% ≤ x < 15% 10% 
Bonus program 15% ≤ x < 20% 20% 
x ≥ 20% 30% 
Contingency rate programConsumption increaseContingence rate
 y ≤ 20% 40% 
 y > 20% 100% 
Consumption reductionBonus
 10% ≤ x < 15% 10% 
Bonus program 15% ≤ x < 20% 20% 
x ≥ 20% 30% 
Contingency rate programConsumption increaseContingence rate
 y ≤ 20% 40% 
 y > 20% 100% 

After the implementation of this bonus system, the SABESP managers understood that it would be fair if the entire city population could be impacted by economic instruments. Thus, the company also included a tax program that started in January 2015. According to this program, households that increased their water consumption up to 20% would be penalized with an increase of 40% in their water bill, and those households that increased their water consumption by more than 20% would be penalized with an increase of 100% on their water bill1 (Table 1).

Table 1 presents a summary of the bonuses/rates applied for each consumption rate.

To evaluate the effectiveness of the bonus implemented by SABESP, we analyzed information about the average consumption of water per household in the city of São Paulo. Monitoring considered 25 districts in the municipality, and the consumption data refer to the period of 2013 and 2015 for each of these districts (Table 2).

Table 2.

The 25 districts of the city of São Paulo considered in the present study.

Jacana Americanópolis/Cidade Ademar Perus Vila Maria Freguesia do O 
Pirituba Casa Verde/Cachoeirinha São Mateus Arthur Alvim/Itaquera Santana 
São Miguel Capela do Socorro Guaianases Campo Limpo Butanta 
Itaim Paulista Ipiranga Grajau/Parelheiros M'Boi Mirim Penha 
Cidade Tiradentes Mooca Vila Mariana Santo Amaro Se 
Jacana Americanópolis/Cidade Ademar Perus Vila Maria Freguesia do O 
Pirituba Casa Verde/Cachoeirinha São Mateus Arthur Alvim/Itaquera Santana 
São Miguel Capela do Socorro Guaianases Campo Limpo Butanta 
Itaim Paulista Ipiranga Grajau/Parelheiros M'Boi Mirim Penha 
Cidade Tiradentes Mooca Vila Mariana Santo Amaro Se 
For the analysis, we used a panel model, which joins data in a cross-section with the time series. This enables us to identify and incorporate possible heterogeneities among the analyzed data. Wooldridge (2010) shows that the unobserved effects model (UEM) can be written for a randomly drawn cross-section observation i, as follows:
formula
(1)
where represents an observed variable vector; denotes the unobserved effect, representing the individual effects of the cross-section data with a constant time t; and represents the idiosyncratic error. Intuitively, this parameter captures constant aspects that are intrinsic to each location i.
The author also emphasizes that, assuming the validity of this hypothesis, Equation (1) may be estimated by ordinary least squares (OLS) as a compost residue of the specific effect and the idiosyncratic error:
formula
(2)
where represents the residue of the equation to be estimated (i.e., the sum of the unobserved effect and idiosyncratic error).

The dependent variable analyzed was the average water consumption reduction per district. This variable was obtained from the difference between water consumption in a month in relation to the previous month (month(t) − month(t − 1)), and the information used was provided by SABESP. Negative values indicate water consumption reduction, whereas positive values indicate an increase in consumption.

The set of explanatory variables includes average household income (deflated by the inflation rate from the period, using the Broad Consumer Price Index – IPCA) and percentage of slums located in the district. The sources of the dependent variables are shown in Table 3.

Table 3.

Sources of information for the used socio-economic variables.

VariableSource
Average household income Non-governmental organization (NGO) ‘Rede Nossa São Paulo’ 
% of slums São Paulo City Government 
VariableSource
Average household income Non-governmental organization (NGO) ‘Rede Nossa São Paulo’ 
% of slums São Paulo City Government 

To determine the use of either the random effect (RE) or the fixed effect (FE), the Hausman test was used. Such a test has a H0 in which both the RE and FE are consistent estimators, and since RE is more efficient, it is preferred to the FE. We cannot reject the null hypothesis using the Hausman test in this analysis; therefore, we chose to use the random effect.

Concerning the variable of average income, in addition to considering the average value for each district (in Brazilian reals), income groups were used. Such income groups were determined based on the minimum wage, as determined by the Brazilian Institute of Geography and Statistics (IBGE). It is important to highlight that in no district did the average income match the income for Class A or Class E; therefore, in the econometric model presented in this study, Class D was considered the comparison group (i.e., it was omitted). The classes considered are those shown in Table 4. Class A was randomly chosen as the control group, to avoid multicollinearity.

Table 4.

Social classes considered.

Social classMonthly household income
Class A Over 15 minimum wages 
Class B From 5 to 15 minimum wages 
Class C From 3 to 5 minimum wages 
Class D From 1 to 3 minimum wages 
Class E Below 1 minimum wage 
Social classMonthly household income
Class A Over 15 minimum wages 
Class B From 5 to 15 minimum wages 
Class C From 3 to 5 minimum wages 
Class D From 1 to 3 minimum wages 
Class E Below 1 minimum wage 

Information about average household income for the years 2013, 2014, and 2015 was not available for the districts in the city of São Paulo. Thus, the data used were the available ones, which refer to the years 2008, 2010, and 2012, meaning that there is a discrepancy in time. The results do not have any significant impact due to this time discrepancy since there were no significant shifts in the economic profile of districts in recent years (Rede Nossa São Paulo, 2016).

Time variables (month and year) were used as controls. Using months as a control prevented distortions due to seasonality throughout the year because water consumption depends, for instance, on weather. Conversely, years were also used as control variables since it is possible that multiple factors that occur in one specific year, in addition to the analyzed variables, may affect water consumption by citizens. The control group was the year 2014, since in that year, the bonus program was already implemented. December was randomly chosen as the control group, to avoid multicollinearity.

To detect whether there was any influence of the water spring used on consumption reduction for households, we also included the variable water reservoirs (Cantareira, Alto Tiete or Guarapiranga). Cantareira was randomly chosen as the control group, to avoid multicollinearity.

The SABESP bonus was considered a binary variable, in which the value 0 was assigned for months in which the bonus program had not begun, while the value 1 was assigned for every month in which the bonus program was being used. The bonus program started in February 2014 for the inhabitants supplied by the Cantareira water reservoir and, in April 2014, the program was extended to the rest of the city population.

Likewise, the variable ‘contingency tariff’ was treated as a binary variable: the value 0 was used for data in which the tariff was not present, and the value 1 was used for information in which the tariff was already implemented. As already discussed, the contingency tariff started for all households in the city as of January 2015.

Table 5 gives a summary of the variables used, their form of measurement and source.

Table 5.

Summary of used variables.

VariableForm of measurementSource
Bonus/contingency tariff Dummy (1 for present; 0 for absence) SABESP 
Slums % of total residences are considered slums São Paulo City Government 
Household average income R$ Rede Nossa São Paulo 
Income class Class A/Class B/Class C/Class D/Class E Rede Nossa São Paulo and IBGE 
Year Year of the consumption measurement – 
Month Month of the consumption measurement – 
Reservoir Which reservoir the consumption measurement refers to SABESP 
VariableForm of measurementSource
Bonus/contingency tariff Dummy (1 for present; 0 for absence) SABESP 
Slums % of total residences are considered slums São Paulo City Government 
Household average income R$ Rede Nossa São Paulo 
Income class Class A/Class B/Class C/Class D/Class E Rede Nossa São Paulo and IBGE 
Year Year of the consumption measurement – 
Month Month of the consumption measurement – 
Reservoir Which reservoir the consumption measurement refers to SABESP 

The program implemented by SABESP uses two different types of economic incentives combined. The bonus, which is applied to the population that reduced their water consumption, is a type of subsidy: there is a governmental reward (i.e., a discount on the water bill) for those who act in a sustainable way (i.e., water consumption reduction). In Motta & Young (1997), the use of a subsidy is more adequate in the case of specific markets, where there is a significant economic impact, or when there is an emergency need for adjustments. Conversely, the contingency tariff that was used for consumers that increased their water demand is a kind of tax/rate/charge. In this way, the government punishes those who raised their water consumption during a scarcity situation. Once again, we mention Motta & Young (1997), who argue that this is a way to guide economic players to value environmental goods and services according to their social opportunity costs.

We compared consumption reduction information for the months of January and July of each year. The purpose of comparing the same months in different years is to obtain a more realistic analysis because, for many different reasons, water consumption throughout the year may vary (due to weather, dry/wet seasons, and other factors). Information showed that between January 2013 and January 2014, there was an increase in water consumption in households. This increase was noted in almost all districts analyzed, with only one exception: the district of Ipiranga, where the average water demand of households remained almost the same. Conversely, between January 2014 and January 2015 (a period when the SABESP bonus program was ongoing), the water consumption in all districts decreased. The average reduction in this period was 25.0% (average of 25.98% and standard deviation of 2.5; Table 6).

Table 6.

Water consumption variation in each district of São Paulo City.

RegionVariation Jan 2013–Jan 2014Variation Jul 2013–Jul 2014Variation Jan 2014–Jan 2015Variation Jul 2014-Jul 2015Reservoir
Jacana +10.55% −11.61% −31.4% −17.7% Cantareira 
Vila Maria +9.15% −12.31% −29.3% −16.7% Cantareira 
Freguesia do O +8.30% −10.76% −28.9% −17.1% Cantareira 
Pirituba +7.79% −11.51% −28.5% −15.0% Cantareira 
Santana +5.44% −14.54% −28.0% −14.2% Cantareira 
Butanta +8.52% −10.00% −27.5% −18.9% Cantareira 
Casa Verde/Cachoeirinha +8.74% −11.52% −27.3% −15.7% Cantareira 
Arthur Alvim/Itaquera +10.16% −11.23% −26.5% −13.6% Alto Tiete 
Ipiranga −0.53% −12.60% −26.3% −15.4% Cantareira 
Vila Mariana +6.13% −10.74% −26.2% −16.0% Guarapiranga 
Perus +10.95% −10.99% −26.1% −15.3% Cantareira 
Jardins/Pinheiros +3.60% −6.31% −25.9% −17.7% Cantareira/Guarapiranga 
Penha +9.98% −12.19% −25.2% −12.8% Alto Tiete 
Mooca +5.79% −10.47% −24.7% −15.3% Cantareira 
São Miguel +9.86% −10.42% −24.2% −14.7% Alto Tiete 
São Mateus +7.54% −9.40% −24.1% −14.1% Alto Tiete 
Santo Amaro +7.33% −8.15% −23.6% −15.5% Guarapiranga 
Se +3.30% −11.41% −23.0% −14.5% Cantareira 
Guaianases +8.95% −5.72% −23.0% −15.4% Alto Tiete 
M'Boi Mirim +7.78% −9.26% −22.6% −16.0% Guarapiranga 
Itaim Paulista +10.69% −10.52% −22.3% −13.2% Alto Tiete 
Cidade Tiradentes +6.16% −8.46% −22.1% −13.6% Alto Tiete 
Capela do Socorro +6.79% −10.67% −21.9% −13.6% Guarapiranga 
Grajau/Parelheiros +5.48% −9.48% −21.3% −12.6% Guarapiranga 
Campo Limpo +9.71% −6.23% −20.2% −15.6% Guarapiranga 
Americanópolis/Cida de Ademar +4.63% −9.17% −20.2% −13.7% Guarapiranga 
RegionVariation Jan 2013–Jan 2014Variation Jul 2013–Jul 2014Variation Jan 2014–Jan 2015Variation Jul 2014-Jul 2015Reservoir
Jacana +10.55% −11.61% −31.4% −17.7% Cantareira 
Vila Maria +9.15% −12.31% −29.3% −16.7% Cantareira 
Freguesia do O +8.30% −10.76% −28.9% −17.1% Cantareira 
Pirituba +7.79% −11.51% −28.5% −15.0% Cantareira 
Santana +5.44% −14.54% −28.0% −14.2% Cantareira 
Butanta +8.52% −10.00% −27.5% −18.9% Cantareira 
Casa Verde/Cachoeirinha +8.74% −11.52% −27.3% −15.7% Cantareira 
Arthur Alvim/Itaquera +10.16% −11.23% −26.5% −13.6% Alto Tiete 
Ipiranga −0.53% −12.60% −26.3% −15.4% Cantareira 
Vila Mariana +6.13% −10.74% −26.2% −16.0% Guarapiranga 
Perus +10.95% −10.99% −26.1% −15.3% Cantareira 
Jardins/Pinheiros +3.60% −6.31% −25.9% −17.7% Cantareira/Guarapiranga 
Penha +9.98% −12.19% −25.2% −12.8% Alto Tiete 
Mooca +5.79% −10.47% −24.7% −15.3% Cantareira 
São Miguel +9.86% −10.42% −24.2% −14.7% Alto Tiete 
São Mateus +7.54% −9.40% −24.1% −14.1% Alto Tiete 
Santo Amaro +7.33% −8.15% −23.6% −15.5% Guarapiranga 
Se +3.30% −11.41% −23.0% −14.5% Cantareira 
Guaianases +8.95% −5.72% −23.0% −15.4% Alto Tiete 
M'Boi Mirim +7.78% −9.26% −22.6% −16.0% Guarapiranga 
Itaim Paulista +10.69% −10.52% −22.3% −13.2% Alto Tiete 
Cidade Tiradentes +6.16% −8.46% −22.1% −13.6% Alto Tiete 
Capela do Socorro +6.79% −10.67% −21.9% −13.6% Guarapiranga 
Grajau/Parelheiros +5.48% −9.48% −21.3% −12.6% Guarapiranga 
Campo Limpo +9.71% −6.23% −20.2% −15.6% Guarapiranga 
Americanópolis/Cida de Ademar +4.63% −9.17% −20.2% −13.7% Guarapiranga 

Considering water consumption in the month of July, the water consumption reduction was less substantial. Between July 2013 and July 2014, the average reduction was 10.2%, and between July 2014 and July 2015, the average reduction was 15.2% (Table 6). Climatic reasons may be used to explain this variation: in January, when it is summer in Brazil, people usually consume more water; thus, there is a wider reduction margin than in July, when it is winter in the southern hemisphere.

Another important result is that from the list of top ten districts with water reduction, between January 2013 and July 2015, eight of these districts used the Cantareira water reservoir, which was the most affected water spring in the city. However, the district with the most significant reduction was Ipiranga (showed a decrease in water consumption of 29.1% during the period), which is supplied by the Guarapiranga reservoir.

The evolution of average water consumption in each district is shown separately according to the supplying reservoir, as shown in Figure 2. We can infer that from the moment the bonus program started for users of the Cantareira reservoir, in February 2014, all districts showed consumption reduction. In other words, even though the bonus program had been established only for those that used water from Cantareira, the population supplied by other reservoirs also started decreasing their water demand. We called this movement the ‘awareness effect’, which influenced the severity of the water crisis in the city as a whole.

Fig. 2.

Evolution of the average household water consumption between January 2013 and July 2015 for districts supplied by each one of the three main reservoirs in São Paulo. *1,000 L. **Each line represents a district. ***The darker vertical line represents the moment when the bonus program was implemented to the citizens in each of the reservoirs, and the lighter vertical line represents the introduction of the contingency tariff.

Fig. 2.

Evolution of the average household water consumption between January 2013 and July 2015 for districts supplied by each one of the three main reservoirs in São Paulo. *1,000 L. **Each line represents a district. ***The darker vertical line represents the moment when the bonus program was implemented to the citizens in each of the reservoirs, and the lighter vertical line represents the introduction of the contingency tariff.

Close modal

In their study on behavior and attitudes towards water conservation, Gilbertson et al. (2011) concluded that substantially more people from water scarcity locations were supportive of most water conservation behaviors. The results from the present study, meanwhile, indicate that even though the region supplied by Cantareira had a worse situation, water scarcity encouraged inhabitants from other regions to also change their behaviors.

Considering the econometric analysis, we evaluated several models using panel data by a random effect estimator. The results are shown in Table 7, where information in the rows is related to a different variable, while information in the columns (A to E) is about the estimations using different combinations of variables to test the robustness of the model.

Table 7.

Results from the panel data.

Water consumption reduction
ABCDE
Explanatory variablesCoef.Coef.Coef.Coef.Coef.
(p-value)(p-value)(p-value)(p-value)(p-value)
Bonus −1.896 −1.636 −1.629 − 1.589  
(0.000) (0.000) (0.000) (0.000)   
L-slum 0.028 0.029 0.024 0.015 0.015 
(0.525) (0.390) (0.596) (0.760) (0.750) 
L-income −0.705 −0.730     
(0.000) (0.000)       
Income class D (income ≥1 and <3 minimum wages)     0.000 0.000 0.000 
Income class C (income ≥3   −0.442 − 0.411 −0.410 
and <5 minimum wage)     (0.000) (0.000) (0.000) 
Income class B (income ≥5   −0.682 − 0.600 −0.600 
and <15 minimum wages)     (0.000) (0.004) (0.004) 
Year 2014 0.000 0.000 0.000 0.000 0.000 
Year 2015 −0.783 −1.186 −1.329 − 1.345 −1.953 
(0.000) (0.000) (0.000) (0.000) (0.000) 
January  2.007 2.012 2.040 3.138 
  (0.000) (0.000) (0.000) (0.000) 
February  1.866 1.870 1.889 2.638 
  (0.000) (0.000) (0.000) (0.000) 
March  1.698 1.701 1.720 2.469 
  (0.000) (0.000) (0.000) (0.000) 
April  1.901 1.904 1.923 2.672 
  (0.000) (0.000) (0.000) (0.000) 
May  2.351 2.352 2.360 2.664 
  (0.000) (0.000) (0.000) (0.000) 
June  2.499 2.500 2.508 2.812 
  (0.000) (0.000) (0.000) (0.000) 
July  2.528 2.529 2.537 2.841 
  (0.000) (0.000) (0.000) (0.000) 
August  1.745 1.745 1.745 1.745 
  (0.000) (0.000) (0.000) (0.000) 
September  1.645 1.645 1.645 1.645 
  (0.000) (0.000) (0.000) (0.000) 
October  1.475 1.475 1.475 1.475 
  (0.000) (0.000) (0.000) (0.000) 
November  0.593 0.593 0.593 0.593 
  (0.002) (0.002) (0.002) (0.006) 
December  0.000 0.000 0.000 0.000 
Cantareira reservoir      0.000 0.000 
Alto Tiete reservoir    0.099 0.350 
      (0.350) (0.001) 
Guarapiranga reservoir    0.137 0.387 
      (0.185) (0.000) 
Constant 4.458 2.262 −2.993 − 3.122 −5.460 
(0.000) (0.007) (0.000) (0.000) (0.000) 
Number of observations 475 475 475 475 475 
R-sq. within 0.491 0.725 0.729 0.729 0.633 
Hausman test 5.13 11.19 2.53 0.61 0.32 
(0.274) (0.739) (0.999) (1.000) (1.000) 
Water consumption reduction
ABCDE
Explanatory variablesCoef.Coef.Coef.Coef.Coef.
(p-value)(p-value)(p-value)(p-value)(p-value)
Bonus −1.896 −1.636 −1.629 − 1.589  
(0.000) (0.000) (0.000) (0.000)   
L-slum 0.028 0.029 0.024 0.015 0.015 
(0.525) (0.390) (0.596) (0.760) (0.750) 
L-income −0.705 −0.730     
(0.000) (0.000)       
Income class D (income ≥1 and <3 minimum wages)     0.000 0.000 0.000 
Income class C (income ≥3   −0.442 − 0.411 −0.410 
and <5 minimum wage)     (0.000) (0.000) (0.000) 
Income class B (income ≥5   −0.682 − 0.600 −0.600 
and <15 minimum wages)     (0.000) (0.004) (0.004) 
Year 2014 0.000 0.000 0.000 0.000 0.000 
Year 2015 −0.783 −1.186 −1.329 − 1.345 −1.953 
(0.000) (0.000) (0.000) (0.000) (0.000) 
January  2.007 2.012 2.040 3.138 
  (0.000) (0.000) (0.000) (0.000) 
February  1.866 1.870 1.889 2.638 
  (0.000) (0.000) (0.000) (0.000) 
March  1.698 1.701 1.720 2.469 
  (0.000) (0.000) (0.000) (0.000) 
April  1.901 1.904 1.923 2.672 
  (0.000) (0.000) (0.000) (0.000) 
May  2.351 2.352 2.360 2.664 
  (0.000) (0.000) (0.000) (0.000) 
June  2.499 2.500 2.508 2.812 
  (0.000) (0.000) (0.000) (0.000) 
July  2.528 2.529 2.537 2.841 
  (0.000) (0.000) (0.000) (0.000) 
August  1.745 1.745 1.745 1.745 
  (0.000) (0.000) (0.000) (0.000) 
September  1.645 1.645 1.645 1.645 
  (0.000) (0.000) (0.000) (0.000) 
October  1.475 1.475 1.475 1.475 
  (0.000) (0.000) (0.000) (0.000) 
November  0.593 0.593 0.593 0.593 
  (0.002) (0.002) (0.002) (0.006) 
December  0.000 0.000 0.000 0.000 
Cantareira reservoir      0.000 0.000 
Alto Tiete reservoir    0.099 0.350 
      (0.350) (0.001) 
Guarapiranga reservoir    0.137 0.387 
      (0.185) (0.000) 
Constant 4.458 2.262 −2.993 − 3.122 −5.460 
(0.000) (0.007) (0.000) (0.000) (0.000) 
Number of observations 475 475 475 475 475 
R-sq. within 0.491 0.725 0.729 0.729 0.633 
Hausman test 5.13 11.19 2.53 0.61 0.32 
(0.274) (0.739) (0.999) (1.000) (1.000) 

In column A, we used the variable bonus and the socio-economic variables household income (as a continuous variable) and percentage of slums, in addition to the variable of year, which is used as the control. In column B, we added the variables for each month to capture monthly seasonality. From the results in these two columns, we may infer that the bonus program was effective in reducing water consumption of households and, on average, the higher the income was, the higher the reduction.

In the models, with the results shown in columns C, D, and E, the income variables used were explanatory variables. The results for the models used in columns A and B are also observed in columns C, D, and E. However, in columns D and E, the variable of water spring was also considered, and, in column E, the variable of bonus was excluded for comparison. Of all the five models created, we considered the model in column D the most suitable for presenting a complete analysis when considering the variables included.

The analysis demonstrated that water consumption reduction had a strong correlation with bonus implementation. In other words, bonus implementation is significantly and negatively correlated with consumption reduction (coef = −1.589). Therefore, the results demonstrated that the bonus program was relevant in decreasing the demand of water in households in São Paulo.

The results also showed that household average income (columns A and B) had an influence on consumption reduction (coef = −0.705 and coef = −0.730, respectively): the greater the average household income was, the greater the reduction. However, it is important to highlight that, as shown in Figure 2, regions with higher income (such as Santo Amaro and Vila Mariana) are those that have higher water consumption, which leads us to believe that there is greater margin in reducing consumption since inhabitants may use water for less essential activities. This result differs to what Magnusson (2004) found, in his study on management incentives in Namibia. The author reports that the effectiveness of economic incentives is overestimated when applied to middle- and high-income areas.

Regarding social class (columns C, D, and E), the explanation in the former paragraph is still pertinent. Class B presented the highest water consumption reduction, followed by class C and class D. However, the variable ‘slum’ was not significant in explaining the reduction in water consumption. Two explanations for this are that (i) the effect of this variable may have been captured by the effect of income, as districts with lower average household income usually have more people living in slums, or (ii) usually, in slums, the water consumed comes from illegal connections and, therefore, the bonus/onus program does not apply.

When we included the variable water spring in columns D and E, the objective was to detect whether there was any influence of the water spring used on consumption reduction in households. Our hypothesis was that households supplied by a reservoir with a more critical situation would be more willing to reduce their water consumption, regardless of the implementation of the bonus. However, we found that the bonus program was more effective for water reduction than the water spring used for water supply. This is because the so-called water spring effect is significant only in the absence of the bonus variable (column E). When the bonus program was omitted, the districts supplied by the Cantareira reservoir showed lower average water consumption per household, followed by the districts supplied by Alto Tiete and, finally, the districts supplied by Guarapiranga.

Considering the contingency program, since it was implemented only at the beginning of 2015, it was not possible to isolate its effects from other events that may have contributed to water consumption reduction in households. Even so, it is possible to assert that there were more events in 2015 that contributed to water demand reduction than in 2014, which may be assigned the contingency fee.

However, the coefficient estimated for the bonus program is greater than that estimated for the contingency tariff. This indicates that the bonus program was more effective than the tariff. Indeed, in all districts analyzed, the average household water consumption was higher in 2014 (when only the bonus program existed) than that in 2015, when both the bonus and contingency tariff existed. There are different possible explanations for these results. For instance, (i) since household water consumption had already decreased widely after the implementation of the bonus program, it was harder to significantly reduce the water demand after 2015, and (ii) for behavioral reasons, the population reacted better to a bonus system than to a contingency tariff.

This study has demonstrated that the use of economic incentives was effective in reducing water consumption in the city of São Paulo, Brazil. Thus, the goal of preserving water resources in a scarcity situation was achieved.

The strategy used by SABESP combined two different kinds of economic incentives: a bonus for customers who saved water and a contingency tariff for customers who increased their water consumption during the crisis period. It is important to emphasize, however, that this kind of incentive is typically used to ensure the quality of environmental goods; in the case of the present study, the main objective was related to quantity (i.e., saved water), not quality.

Global literature on economic incentives diverges regarding the results of water demand and lacks studies of population behaviors in developing countries, such as Brazil. This paper tried to overcome this lack and showed that the SABESP program may be considered a valid solution for lowering water demand.

The results of the econometric analysis support the results of the explanatory analysis. From these results, we note that (i) the implementation of the bonus was effective in encouraging water consumption reduction and more efficient than the contingency tariff, and (ii) consumption reduction was more meaningful in districts that used water originating from springs under more critical conditions but was adopted by citizens from all analyzed districts.

In addition, the econometric analysis also demonstrated that income was relevant for water demand reduction both when we considered a continuous variable and when we considered a binary variable (in ranges). Districts in higher social classes were more willing to reduce consumption. The analysis also demonstrated that the quantity of slums was not relevant to water consumption variation.

A relevant fact is that water consumption reduction in districts that are not supplied by the Cantareira reservoir started before the implementation of the bonus/onus system: in districts supplied by the Alto Tiete reservoir, the reduction started in February 2014, and in districts supplied by the Guarapiranga reservoir, the reduction started in March 2014. These facts reveal that despite the importance of an economic incentive, a considerable part of the population in São Paulo did not wait for the implementation of the bonus/onus to start saving water in their households.

Monitoring for a longer period of time would be of interest to confirm the influence of the socio-economic variables considered in the present study. It would also be important to use updated variables to prevent time gaps in the analysis. Moreover, we understand that it would be appropriate to include all municipalities in São Paulo State that took part in the SABESP program. Therefore, once again, the lack of information was an obstacle to a more detailed assessment.

Nevertheless, the results of this study may be used as a subsidy for government decision-making, in situations of water scarcity. Actions that prevent water scarcity are a priority, but the findings we present here may be useful, especially when considering that prevention actions are sometimes difficult to implement, especially in underdeveloped countries.

1

Hospitals, police stations, prisons, and every household with a monthly consumption below 10,000 L were exempt from the program.

ARSESP
(
2014
).
Deliberação Arsesp n° 469 (Arsesp’ s Deliberation n. 469)
.
Agência Reguladora de Saneamento e Energia do Estado de São Paulo
,
São Paulo
,
Brazil
.
Baldwin
R.
Hutter
B.
Rothstein
H.
(
2000
).
Risk Regulation, Management and Compliance – A Report to the BRI Inquiry
.
London School of Economics and Political Science
,
London
,
UK
.
Brasil
(
1997
).
Lei n° 9.433, de 8 de janeiro de 1997 (Act n. 9.433, of January 8th, 1997)
.
Presidência da República
,
Brasília
,
Brazil
.
Dalhuisen
J.
Florax
R.
de Groot
H.
Nijkamp
P.
(
2003
).
Price and income elasticities of residential water demand: a meta-analysis
.
Land Economics
79
(
2
),
292
308
.
Gaudin
S.
(
2005
).
Using water bills to reinforce price signals: Evidence from the US
. In:
IWA International Conference on Water Economics, Statistics, and Finance
,
8–10 July
,
Rethymno, Greece
.
Gilbertson
M.
Hurlimann
A.
Dolnicar
S.
(
2011
).
Does water context influence behaviour and attitudes to water conservation?
Australasian Journal of Environmental Management
18
(
1
),
47
60
.
Kolstad
C. D.
(
2011
).
Environmental Economics
, 2nd edn.
Oxford University Press
,
Oxford
,
UK
, p.
496
.
Lee
J.
Tanverakul
S. A.
(
2015
).
Price elasticity of residential water demand in California
.
Aqua
64
(
2
),
211
.
Littlefair
K.
(
1998
).
Willingness to pay for Water at Household Level: Individual Financial Responsibility for Water Consumption (Occasional Paper n. 26)
.
School of Oriental and African Studies (SOAS), University of London
,
London
,
UK
, p.
46
.
Magnusson
T. S.
(
2004
).
Household responsiveness to water demand management incentives in Windhoek
.
Namibia. Water Policy
6
,
453
471
.
Metaxas
S.
Charalambous
E.
(
2005
).
Residential price elasticity of demand for water
.
Water Supply
5
(
6
),
183
188
.
Mini
C.
Hogue
T. S.
Pincetl
S.
(
2015
).
The effectiveness of water conservation measures on summer residential water use in Los Angeles, California
.
Resources, Conservation and Recycling
94
,
136
145
.
Motta
R. S.
Young
C. E. F.
(
1997
).
Projetos de Instrumentos Econômicos Para A Gestão Ambiental (Economic Instruments Projects for Environmental Management)
.
IPEA
,
Rio de Janeiro
,
Brazil
.
http://www.ie.ufrj.br/gema/pdfs/projeto.pdf. (accessed 28 December 2016)
.
Nataraj
S.
Hanemann
W.
(
2011
).
Does marginal price matter? A regression discontinuity approach to estimating water demand
.
Journal of Environmental Economics and Management
61
(
2
),
198
212
.
Olmstead
S. M.
Hanemann
M. W.
Stavins
R. N.
(
2005
).
Do Consumers React to the Shape of Supply? Water Demand Under Heterogeneous Price Structure – Discussion Paper, 05-29
.
Resources for the Future
,
Washington, DC
,
USA
.
Rede Nossa São Paulo
(
2016
).
Mapa da Desigualdade 2016 (Inequality Map 2016)
, p.
59
.
Renwick
M. E.
Green
R. D.
(
2000
).
Do residential water demand side management policies measure up? an analysis of eight California water agencies
.
Journal of Environmental Economics and Management
40
(
1
),
37
55
.
SABESP (Companha de Saneamento do Estado de São Paulo)
(
2015
).
Situação dos Mananciais (Springs Daily Status)
. .
Saz-Salazar
S.
González-Gómez
F.
Guardiola
J.
(
2014
).
Willingness to pay to improve urban water supply: the case of Sucre, Bolivia
.
Water Policy
17
(
1
),
112
125
.
Sebri
M.
(
2014
).
A meta-analysis of residential water demand studies
.
Environment, Development and Sustainability
16
,
499
520
.
Sette
M. T.
(
2014
).
Manual de Direito Ambiental (Environmental Law Guide)
, 3rd edn.
Juruá Editora
,
Curitiba
,
Brazil
.
Wooldridge
J. M.
(
2010
).
Econometric Analysis of Cross Section and Panel Data
, 2nd edn.
MIT Press
,
Cambridge, MA
,
USA
, p.
1065
.
Worthington
A. C.
Hoffman
M.
(
2008
).
An empirical survey of residential water demand modelling
.
Journal of Economic Surveys
22
(
5
),
842
871
.