Abstract

Over the past decades rapid population growth in urban areas has promoted the development of high-density housing such as high-rise apartments. In order to properly supply water to this growing sector, it is essential to understand the determinants of its water use. However, this task has largely remained unexplored through the empirical study of water demand mainly due to the scarcity of data in this sector. Using a rich source of GIS-based urban databases in Auckland, New Zealand, this study integrated apartment water consumption, property characteristics, weather, water pricing and census microdata to overcome this issue. This study also compared high-rise apartment water use and its determinants with low-rise apartments. Through applying panel data models, the study revealed that, similar to the low-rise apartments, household size is the most important determinant of high-rise apartment water use in Auckland, where other socioeconomic factors, building features, water pricing and weather variables were not significant determinants. The study also showed that the per capita water consumption in the high-rise apartments in Auckland was higher than in the low-rise apartments, challenging the assumption underlying contemporary urban policy that densifying the central city areas can offer significant savings in water use.

Introduction

Compact development as a contemporary urban growth management strategy has been used in many cities around the world in order to mitigate the social, economic and environmental consequences of uncontrolled low-density urban sprawl (Randolph, 2006; Haarhoff et al., 2012). This strategy has promoted the use of intensive housing development, mainly in the form of multi-unit housing (i.e. apartments), in and around existing urban centres (Haarhoff et al., 2012). Multi-unit housing generally refers to a dwelling type that is a single structure containing one or more housing units, such as flats and apartment buildings. This housing type basically consists of multiple living units within one or more shared structures and common use areas (e.g., laundry, pools, parking, and recreational facilities) (Wentz et al., 2014).

In order to properly supply water and manage demand in this fast growing sector, it is vital to understand the determinants of its water use. While a vast number of studies have investigated the factors affecting residential water use in single-unit housing (e.g. separate houses) or as a total (Wentz & Gober, 2007; Schleich & Hillenbrand, 2009; Chang et al., 2010; House-Peters et al., 2010; Polebitski & Palmer, 2010; Rockaway et al., 2011), few studies have evaluated the determinants of water use in multi-unit housing. This segregation is necessary since there may be substantial differences in the determinants of water use across the single-unit and the multi-unit houses. This distinction can be attributed to the differences in the socioeconomic characteristics of residents and the level of outdoor usage (e.g. gardens and swimming pools) between these two housing types (Russac et al., 1991; Fox et al., 2009).

Water consumption and its determinants also may vary considerably within each of these housing groups based on the property characteristics. For example, the water consumption in the different types of multi-unit residences (i.e. high-rise and low-rise apartments) may be significantly different (Russac et al., 1991; Loh et al., 2003; Troy & Holloway, 2004; Zhang & Brown, 2005; Domene & Saurí, 2006; Fox et al., 2009). In general, smaller multi-unit complexes with fewer housing units are more likely to show similar water habits to single-unit housing (Wentz et al., 2014).

This study focuses on the understanding of water consumption and its determinants in high-rise apartments (i.e. four or more storey buildings) in Auckland, New Zealand. The results are also compared with a study by Ghavidelfar et al. (2016a) evaluating low-rise apartment water use in Auckland. Since in Auckland the majority of high-rise apartments are concentrated in the central city while the low-rise apartments are available across the city with a lower density, this comparison can help to reveal the effects of densifying central city areas on urban water consumption. In general, higher density living is seen as a credible path for improving urban sustainability through reducing the use of urban resources and the need for more infrastructure (Randolph, 2006; Boon, 2010; Haarhoff et al., 2012). However, in terms of water consumption, there is no clear understanding of the benefits of higher density urban development. This study aims to bridge this gap by fully investigating apartment water demand in Auckland.

In general, the empirical studies of water demand targeting multi-unit housing are very limited. In a study in Tucson, Arizona, Agthe & Billings (2002) developed regression models to explain the winter and summer water demand for 308 apartment complexes. They concluded that factors such as the value per bedroom, number of bedrooms, age of apartment, indoor water-saving devices, swimming pools, vacancy rates and water price were the principal determinants of apartment water use. Zhang & Brown (2005) evaluated the effects of household socioeconomics, water amenities and facilities, and attitude toward environmental concerns on apartment water use in Beijing and Tianjin, China. Using these variables they managed to explain around 10% to 55% of variation of water consumption in different types of apartment (i.e. high-rise, multi-storey, low-rise). Mayer et al. (2006) also evaluated apartment water demand across 13 cities in the USA with the main purpose of understanding the benefits of separate billing systems in the multi-unit housing sector. They showed that variables such as average number of bedrooms per unit, existence of cooling tower, fixture efficiency as well as submetering may significantly influence apartment water use. In a recent study, Wentz et al. (2014) used the design features of large apartment complexes to explain the variance in apartment water use in Tempe, Arizona. By examining some of the indoor and the outdoor features of buildings through regression analysis, they concluded that per bedroom water use increased with pool area, dishwashers, and in-unit laundry facilities. However, to the knowledge of the authors, none of these studies fully evaluated the effects of all socioeconomic, property characteristics, weather and water pricing on the high-rise apartment water use and compared it with the low-rise apartments.

One of the main reasons why the study of apartment water demand largely remains unexplored in comparison with single-unit housing is the lack of readily available data in this sector. In order to mitigate this data shortage, this study utilized a geographic information system (GIS) to integrate different sources of data associated with apartments. In this way, the study firstly linked the apartment water consumption data to the property information. Then, the dataset was aggregated at the meshblock scale (i.e. the smallest geographical unit in which the census data are available) in order to include sociodemographic characteristics of households living at the apartments from census microdata. Finally, information on water pricing and weather for different areas was also added into the dataset in order to enable the evaluation of the effects of these variables on apartment water demand as well.

In recent years, data integration in water demand studies has become more plausible due to advances in database technology, data accessibility, computing power, and spatial tools (Polebitski & Palmer, 2010; Dziedzic et al., 2015). In an early attempt at data integration, as a pilot study, Troy & Holloway (2004) linked water demand and property information in six census areas in Adelaide, Australia, in order to examine the water consumption patterns for different types of residential dwellings and areas. Shandas & Parandvash (2010) integrated water consumption, land use and demographic data at parcel level to examine the relationship between land-use planning and water demand. Polebitski & Palmer (2010) integrated utility billing data with census demographic and property appraisal data at the census track level in order to forecast residential use in Seattle, Washington. In a recent study Dziedzic et al. (2015) integrated water billing records, demographic census information, and property information in Ontario, Canada. Through this data integration and subsequent cluster analysis, they identified the pattern of water demand over different areas and groups of customers for the purpose of conservation planning. They emphasized the importance of data integration in order to use the full potential of rich data available from local and national organizations. However, none of these studies has applied data integration in the multi-unit housing sector.

This study, through data integration, developed a database containing information on 147 apartment buildings, with more than 11,000 units, across 126 census meshblock units in Auckland, in order to evaluate the effects of household socioeconomic (e.g. household income and household size), dwelling characteristics (e.g. number of bedrooms, lot size, swimming pool), weather (e.g. temperature), and water pricing on water demand. All of these variables have been frequently reported as the influential factors in empirical water demand studies (House-Peters & Chang, 2011).

In this study, the period of the analysis spans from July 2008 to July 2014. The study utilizes regression methods specific to panel data to evaluate the determinants of apartment water demand. Panel data models are typically preferred to the time-series and cross-sectional models because they include the advantages of both models and can provide more accurate parameter estimates (Arbués et al., 2003; Polebitski & Palmer, 2010). In recent years, with the increase in data availability, these models have been used more frequently (Nauges & Thomas, 2000; Martinez-Espiñeira, 2002; Arbués et al., 2004, 2010; Kenney et al., 2008; Polebitski & Palmer, 2010; Fenrick & Getachew, 2012). However, to the knowledge of the authors, the panel models have never been used for demand analysis in the apartment sector.

This article is organized in the following order. After the introduction, a review of the study area is presented. Then the data integration procedure is discussed, followed by the method of analysis which is briefly discussed. Finally, the results and the conclusions are presented.

Study area

Auckland is the largest city in New Zealand with a population around 1.4 million. This city formerly comprised seven territorial authority areas (i.e. Rodney District, North Shore City, Waitakere City, Auckland City, Manukau City, Papakura District, and Franklin District). However, in 2010 these areas amalgamated to form a unitary authority as the Auckland Council.

In Auckland, the housing stock is generally dominated by single-unit houses. At the time of the 2013 Census, single-unit housing (i.e. separate dwelling) made up about three-quarters of occupied private dwellings in Auckland, while the percentage of multi-unit housing (i.e. joined dwelling) was around 25% (Goodyear & Fabian, 2014; Statistics-NZ, 2015).

High-rise apartments (i.e. apartments with four or more storeys) comprise around 14% of multi-unit housing stock in Auckland, where most of them (around 61%) are located within Auckland Central Business District (CBD). Auckland CBD is the economic heart of the Auckland metropolitan area, where more than 75% of residential dwellings are high-rise apartments.

High-rise apartment living is a relatively new lifestyle choice for New Zealanders. By the 1970s a lack of people living in the Auckland CBD was seen as a problem (Boon, 2010). However, since the early 1990s the interest in apartment living has gradually increased. Between 2006 and 2013, the number of people living in the high-rise apartments in Auckland has almost doubled, reaching around 30,000 people in 2013 (Statistics-NZ, 2015). The tendency toward apartment living in Auckland is also boosted by the Auckland council policy of compact city development. Based on the Auckland Unitary Plan, the central areas with good access to high-frequency public transport and other facilities are targeted for higher density living (Goodyear & Fabian, 2014). Although extensive construction of apartments has promoted central city living, the quality of these developments has been questioned. For example, the new developments in Auckland CBD generally were criticized for being too small and having insufficient outdoor space (Boon, 2010; Carroll et al., 2011).

Inner city (CBD) apartments in Auckland are mainly occupied by younger people and students (Statistics-NZ, 2010). There are several tertiary educational institutions within Auckland CBD. According to the 2013 Census, around 30% of Auckland CBD residents are students. The median age of the population in Auckland CBD is around 28 years. This number is substantially different for the out-of-CBD apartments, where the median age of residents is around 42 years. The average household size is about two people per household both within and outside CBD apartments. In Auckland CBD around 80% of apartments are rental, while this number is around 50% for the out-of-CBD apartments (Statistics-NZ, 2015).

Auckland has a subtropical climate with a year-round precipitation. The average annual precipitation is around 1,240 mm. The annual average daily maximum air temperature is around 19 °C. The coldest month is usually July and the warmest month is usually January or February (NIWA, 2015).

Residential water consumption comprises around 56% of total water consumption in Auckland. Watercare Services Limited is the water and wastewater service provider for Auckland. It is a council controlled organization which provides 119 × 109 litres of drinking water annually to 1.35 million customers (Watercare, 2014). Auckland's water supply is obtained from three different sources including dams, rivers and underground aquifers. The exact proportion of water supply from each source varies daily depending on the levels in the storage lakes, forecast rainfall, treatment plant capacity, maintenance requirements and transmission costs. However, in general around 80% of the region's drinking water is supplied by Auckland's dams (Watercare, 2017).

Data integration

This study integrates the data on water consumption, property, weather, water pricing, and census microdata to evaluate the determinants of apartment water use.

In this study the water consumption data was provided by Watercare Services Limited on a monthly basis for all dwellings in Auckland for the period 2008–2014. These data do not include Papakura District meters since the provision of retail water services in that district is franchised to a separate company. Thus, the Papakura district was excluded from this analysis. Up until July 2012, each former district of Auckland had a different water recording span, varying from 6 months to bimonthly periods. From July 2012, the domestic accounts have been read every 2 months by Watercare. To standardize the data across Auckland, Watercare converted these data to a monthly period. In order to estimate the monthly water use for each individual meter, Watercare first estimates the average daily use during the reading period (i.e. the usage on the meter is divided by the number of days between the two readings). Then, this average use is allocated to each month according to the number of days corresponding to that month in that particular reading period. The water consumption database also includes the address of the property and its geographical location (i.e. X and Y coordinates), type of meter (i.e. domestic, commercial, etc.) and its installation date for each individual meter. In contrast to single-unit housing, Watercare only measures the total water use in apartment buildings using master meters and does not meter apartments individually (although the units may be sub-metered individually by the building managers).

The property information was obtained from the publicly available databases at Auckland Council (Auckland Council, 2015) and Land Information New Zealand (LINZ, 2015). The developed property dataset contains information such as housing type (i.e. single-unit, flats or apartments, etc.), assessed value of the property, structure size of building (i.e. building footprint), impervious area, the issue dates of section (as a proxy of age of property), and the address of the property.

The weather data, including air temperature and rainfall, was provided by New Zealand's National Climate Database (CliFlo, 2015) for the period 2008–2014. These data came from 15 weather stations across Auckland and was interpolated in GIS to estimate average air temperature and rainfall over different areas.

The water and wastewater charges for six districts of Auckland, from 2008 to 2014, were also provided by Watercare. The water tariff in Auckland consists of an annual fixed charge and the volumetric charges for water and wastewater. Watercare calculate the volume of wastewater based on the water volume measured by the water meter. The water, wastewater and fixed charges have undergone substantial changes over the last few years in Auckland. Before 2010, the water and wastewater charges were determined by the local councils thus every district had its own tariff. However, after amalgamation of the Auckland councils in 2010, Watercare took over the water sector in Auckland and gradually changed all the water and wastewater tariffs over the local councils to a unified tariff for all of Auckland after July 2012. Watercare usually adjusts water and wastewater charges annually in July each year.

The socioeconomic information for households was collected from Statistics New Zealand Data Lab (Statistics-NZ, 2015) for census 2006 and 2013. The Data Lab provided access to the microdata (i.e. data about specific people, households, or businesses). From census microdata it is possible to estimate household and housing information (e.g. household income, household size, education level, number of bedrooms, etc.) for different types of housing. For this study, the census information for households living in the high-rise apartments was collected at the meshblock level. Meshblock is the small geographical unit for which the census information is available.

In this study, data integration was carried out using a geographic information system (GIS). In this way, first, water consumption and property data were arranged in GIS and linked using the addresses and geographical coordinates. By this integration the information on water consumption and property for around 350,000 housing units including single-unit and multi-unit (i.e. low-rise and high-rise apartments) housing became available for the demand analysis.

The present article only focuses on the evaluation of water demand in the high-rise apartments. High-rise apartments made up around 4% of housing stock in Auckland. Thus, after filtering the database based on the property type, around 15,000 apartments from 190 residential apartment buildings remained for the rest of the analysis. From this dataset the apartment buildings with missing water use records or shared meters with the commercial sector (i.e. non-residential customers such as restaurants and cafés) were excluded from the database, leaving 147 apartment buildings with 11,832 units for the final demand analysis. Figure 1 shows the location of the studied apartments across Auckland.

Fig. 1.

Locations of studied apartments in Auckland.

Fig. 1.

Locations of studied apartments in Auckland.

Through the first step of data integration, information such as average water consumption per apartments, number of units in each apartment building, age of buildings, lot size of buildings and presence of swimming pools and gardens in the buildings became available at the apartment building level. Since the apartment water consumption data was only available at the apartment building level, in order to estimate the average water use per apartment, the total water consumption of each apartment building was divided by the total number of apartment units in the building. The presence of a garden and swimming pool in each apartment building was also investigated using high-resolution aerial images through visual inspection.

In the second step of data integration, the water consumption and property information was aggregated at the meshblock scale in order to add the socioeconomic characteristics of households living at the apartments into the dataset. By this integration, information such as household size, household income, age of residents, and property ownership became available for the demand analysis.

Finally, the water pricing and weather information was also assigned to each meshblock based on its geographical location in order to enable understanding of the effects of these variables on apartment water demand as well.

Water demand models

This study applies regression methods specific to panel data to understand the determinants of apartment water demand. A panel of data is the repeated observations for specific subjects over multiple time periods (Hill et al., 2010). In this study, the subjects are meshblock units and the repeated observations are changes of water consumption, water pricing, socioeconomic factors, and weather across meshblock units over 6 years. This study examined three common panel data methods (i.e. pooled, fixed, and random effects models) for demand analysis. In the pooled method, the regression model has a single intercept; however in fixed effects and random effects models the intercept is allowed to vary between subjects (Hill et al., 2010). Therefore, fixed effects and random effects models are typically an improvement over pooled models since they can capture the variability among subjects using varying intercepts. In panel data models, a pooling test (partial F-test) is used to examine this improvement (Hill et al., 2010). The null hypothesis of this test is that all intercepts between subjects are equal. If the p-value associated with the test statistic is below the range of accepting the null hypothesis (i.e. 0.05), it can be concluded that the panel estimators (i.e. fixed and random effects) are preferred to the pooled model. In order to choose an appropriate method between fixed effects and random effects models, a Hausman test is used (Hill et al., 2010; Wooldridge, 2012). The null hypothesis of this test is that, if there are no omitted variables, the random effects model is more efficient (Polebitski & Palmer, 2010). This means that if the null hypothesis of the test is not rejected, the random effects model is preferred. The random effects model has a useful feature over the fixed effects because it can recover parameter estimates for time invariant variables as well (Fenrick & Getachew, 2012).

In this study, the dependent variable is the annual average daily water consumption per apartment at the meshblock level. In order to calculate this, the average annual apartment water use in each meshblock (calculated by adding average monthly consumption) was divided by the number of days in each year. The developed dataset of study included apartment average water consumption across 126 meshblock units over 6 years (i.e. August 2008 to July 2014). The water consumption data was estimated on the annual basis because the water price in Auckland changed annually (i.e. in July each year). Thus, it can better reflect the overall effects of price changes across the years.

The independent variables in this study are household socioeconomic and housing characteristics, water price and air temperature. Socioeconomic variables were estimated from censuses 2006 and 2013 microdata. A yearly estimate of census variables is used for the panel data analysis. The housing characteristics include average number of units in the apartment building, average number of bedrooms in apartments, age of building, lot size of building (i.e. section size of building excludes structure size) and two dummy variables representing the presence of outdoor swimming pool and garden in the buildings. In order to investigate the effects of water pricing, both volumetric and fixed charges of water and wastewater were included in the model. Since in Auckland the wastewater price is calculated based on the metered water use, the study summed up the charges of water and wastewater. This helps to evaluate the overall effect of volumetric and fixed charges. A dummy variable distinguishing the within and out-of-CBD apartments is also included in the model in order to examine if the water consumption among these two groups of apartments is different.

Table 1 provides a list of variables used for the demand analysis. In this study the prices and income were deflated into real 2013 terms using the consumer price index (CPI) (Statistics-NZ, 2015).

Table 1.

Description, mean and standard deviation (SD) of the variables used in this study.

VariablesDefinitionMeanSDVariable unit
DWU Average daily water use 379 96 Litre/apartment/day 
HhSize Average household size 1.9 0.3 People 
Income Household median income 70,600 33,500 NZ dollars/year 
AgeUR Median age of usual residents 33.4 11.0 Year 
Owner Percentage of households owned the dwelling 33.3 18.9 
BRooms Average number of bedrooms 1.8 0.5 Bedrooms 
Units Average number of units in the apartment building 84 70 Apartments 
AgeBld Average age of apartment building in 2014 16 Year 
LotSize Average lot size of apartment building 855 1,093 m2 
DumPool Dummy variables representing the apartment building with pool 0.17 0.37 N/A 
DumGarden Dummy variables representing the apartment building with garden 0.16 0.36 N/A 
Temp Mean daily maximum air temperature 19.3 0.4 °C 
VPrice Average volumetric price of water and wastewater 4.1 0.8 NZ dollars/m3 water 
FPrice Average fixed price of water and wastewater 126 88 NZ dollars/year 
DumCBD Dummy variables representing houses within CBD 0.59 0.49 N/A 
VariablesDefinitionMeanSDVariable unit
DWU Average daily water use 379 96 Litre/apartment/day 
HhSize Average household size 1.9 0.3 People 
Income Household median income 70,600 33,500 NZ dollars/year 
AgeUR Median age of usual residents 33.4 11.0 Year 
Owner Percentage of households owned the dwelling 33.3 18.9 
BRooms Average number of bedrooms 1.8 0.5 Bedrooms 
Units Average number of units in the apartment building 84 70 Apartments 
AgeBld Average age of apartment building in 2014 16 Year 
LotSize Average lot size of apartment building 855 1,093 m2 
DumPool Dummy variables representing the apartment building with pool 0.17 0.37 N/A 
DumGarden Dummy variables representing the apartment building with garden 0.16 0.36 N/A 
Temp Mean daily maximum air temperature 19.3 0.4 °C 
VPrice Average volumetric price of water and wastewater 4.1 0.8 NZ dollars/m3 water 
FPrice Average fixed price of water and wastewater 126 88 NZ dollars/year 
DumCBD Dummy variables representing houses within CBD 0.59 0.49 N/A 

Results and discussion

This study examined pooled, fixed and random effects models to select the most appropriate panel data method for the demand analysis. The result of the pooling test showed that the panel models (i.e. fixed and random effects models) are an improvement over the pooled model. The Hausman test also revealed that the random effect model is more efficient than the fixed effect model and can better produce consistent parameter estimates. Table 2 shows the results of the random effects model as the best estimator. However, in general the results of all three models were consistent in terms of water consumption and its determinants. The time trend was included in the model in order to accommodate the nonlinearities in the underlying data. All the variables (except FPrice which contains zero values) were transferred by natural logarithm thus the coefficients can be interpreted as the elasticity.

Table 2.

Random effects water demand model.

VariablesEstimate
Const 6.14*** 
HhSize 0.49*** 
Income −0.03 
AgeUR −0.13 
Owner 0.03 
BRooms 0.12* 
Units −0.01 
AgeBld 0.07 
LotSize −0.02 
DumPool 0.02 
DumGarden 0.07 
Temp −0.03 
VPrice 0.02 
FPrice −0.00003 
DumCBD 0.06 
time 0.04*** 
time2 −0.005*** 
Partial F-test 35.76*** 
Hausman test 15.36 
Number of meshblock units 126 
Overall adjusted-r2 0.5 
VariablesEstimate
Const 6.14*** 
HhSize 0.49*** 
Income −0.03 
AgeUR −0.13 
Owner 0.03 
BRooms 0.12* 
Units −0.01 
AgeBld 0.07 
LotSize −0.02 
DumPool 0.02 
DumGarden 0.07 
Temp −0.03 
VPrice 0.02 
FPrice −0.00003 
DumCBD 0.06 
time 0.04*** 
time2 −0.005*** 
Partial F-test 35.76*** 
Hausman test 15.36 
Number of meshblock units 126 
Overall adjusted-r2 0.5 

***, ** and * denote the level of significance at 1%, 5% and 10%, respectively.

The random effects model provided satisfactory results as the estimated variables had the expected signs and significance. Moreover, the adjusted R2-value of 0.50 is on the high end of the range presented in earlier studies of apartment water demand (Agthe & Billings, 2002; Zhang & Brown, 2005; Mayer et al., 2006; Wentz et al., 2014).

The results of the study showed that household size is the most influential factor on the apartment water use. The estimated coefficient for the household size in the random effects model is around 0.5, implying that a 10% increase in the household size would result in a 5% increase in the apartment water consumption. This result is in agreement with many other water demand studies, where it was argued that, owing to economies of scale in the use of water, the increase in water consumption is less than proportional to the increase in household size (Arbués et al., 2003, 2004; Hoffmann et al., 2006; Schleich & Hillenbrand, 2009).

The study also revealed that household income was not significantly correlated with the apartment water consumption. This result was expected in the case of Auckland high-rise apartments, where the majority of water consumption is in the form of indoor usage (i.e. water is used for basic needs). In general, the income variable mainly influences the household outdoor water consumption in the single-unit housing. That is because the higher income households are more likely to own water-using capital stock such as larger lawns and gardens, and swimming pools (Hoffmann et al., 2006; Schleich & Hillenbrand, 2009; Mieno & Braden, 2011).

Figure 2 shows the monthly variations of average apartment water use and air temperature in Auckland. In general, in the residential sector, the difference between summer and winter water consumption can be attributed to outdoor use (Billings & Jones, 2008). In the case of Auckland apartments, the variation of water use between summer (i.e. February) and winter (i.e. July) is very limited, implying that indoor water use is predominant at the apartments.

Fig. 2.

Monthly variations of apartment water consumption (left axis) and air temperature (right axis) in Auckland.

Fig. 2.

Monthly variations of apartment water consumption (left axis) and air temperature (right axis) in Auckland.

Figure 2 also shows that in the summer the average apartment water consumption is relatively lower than in the winter. This is in contrast to the common pattern of single-unit housing water use, where the consumption is typically higher in summers rather than winters (Billings & Jones, 2008; Polebitski & Palmer, 2010). In general, the higher summer water demand in single-unit housings can be attributed to the higher water use arising from outdoor activities such as lawn watering, gardening and filling swimming pools (Billings & Jones, 2008). Conversely, in multi-unit housings, where the indoor usage is predominant, water use is likely to remain relatively stable across the different seasons (Domene & Saurí, 2006). However, in the case of Auckland apartments, the water consumption increases each year around the end of February, stays relatively constant until November, and then declines (Figure 2). This pattern closely follows the tertiary academic calendar in New Zealand rather than the usual summer and winter seasons. In a time-series study of apartment water demand, Ghavidelfar et al. (2016b) showed that during the months of the academic year the average water use of apartments in CBD increases by around 10%. They attributed this increase to the higher number of occupants in the apartments during the academic months.

In this study, the age of residents and property ownership were found to be insignificantly related to the apartment water use. This result was expected since these two variables had a direct relationship with the household income (i.e. older residents had higher income and subsequently a higher chance of owning the property). Thus, similar to the income variable which did not significantly influence indoor water consumption, the effects of these two variables on the apartment water consumption were also limited.

The study also revealed that the lot size of apartment buildings and the presence of outdoor swimming pool and garden did not significantly affect the average water consumption in the Auckland apartments. This is because in general the outdoor space of apartment buildings in Auckland is limited, mainly used as the car parks. The size of gardens is also small and the vegetated landscaping is limited to the planting of shrubs and trees which basically do not require much water. Moreover, the year-around precipitation in Auckland reduces the need for irrigation for this type of landscaping.

The study showed that the number of bedrooms in apartments had a positive correlation with apartment water use. However, the estimated coefficient was not statistically significant at the 0.05 level. In general, number of bedrooms is a proxy for the number of residents in apartments, which has a positive correlation with the water use (Agthe & Billings, 2002; Mayer et al., 2006).

In this study, the number of units in the buildings was found to be insignificantly correlated with the apartment water consumption. This result implies that the economies of scale for the shared water use (i.e. water is used for the building maintenance, cleaning, etc.) do not play a significant role in the average apartment water use.

The results also showed that there was no statistically significant difference between the older and the newer apartment buildings in terms of water use. This finding was expected since the majority of apartment buildings in Auckland have been constructed in the last decade. In addition, the old buildings are likely to be renovated. This result is in contrast to the finding of Agthe & Billings (2002), who indicated that older apartments used more water than newer apartments, but is in agreement with the finding of Wentz et al. (2014).

The study revealed that the volumetric price of water had insignificant correlation with the apartment water consumption. This result is not surprising since the water generally is used for basic needs at the apartments. In general, the indoor water use is unlikely to exhibit high price sensitivity (Arbués et al., 2003; Mieno & Braden, 2011). The fixed price of water also had insignificant impact on the water consumption. In general, the only effect of the fixed charge on water consumption would be through its effect on reducing disposable income. Since the water costs usually comprise a small share of household expenditure, it is expected that the effect of fixed price becomes insignificant (Mieno & Braden, 2011).

The air temperature variable was found to be insignificantly related to the apartment water use. This result was expected since the weather variables typically affect outdoor water demand rather than indoor (Arbués et al., 2003).

Finally, the CBD dummy variable was statistically insignificant, implying that there is no difference between the within and out-of-CBD apartments in terms of water consumption, when controlling for all other variables.

Management implications

This study showed that the household size is the most influential determinant of high-rise apartment water consumption in Auckland, while other socioeconomic factors, property characteristics, water pricing and weather variables were not significant determinants. In general, this result is in agreement with the findings of the study of low-rise apartment water use in Auckland (Ghavidelfar et al., 2016a). However, in the low-rise apartments the air temperature had a positive correlation with water consumption and the effect of water pricing, although very small, was statistically significant. These subtle differences can be attributed to the fact that in the low-rise apartments there is more opportunity for the outdoor water use (e.g. gardening and swimming pools) in comparison with the high-rise apartments. Therefore, the low-rise apartment sector showed more sensitivity to the weather variable and water pricing. In general, in the apartments the seasonal variation of water use is very limited (less than 10%), stressing that the indoor water use is the predominant usage in this sector. This result suggests that in cases where water conservation would be required in this sector, the conservation programs should concentrate more on the methods associated with regulating indoor use such as correcting household water use habits, for example through running education campaigns or by increasing the efficiency of water appliances, rather than more conventional methods such as water pricing.

From the perspective of urban planning and water management, the findings of this study are also important. Through comparison of water use in the low-rise and high-rise apartments, the study revealed that densifying central city areas, through developing higher density housing, may not necessarily lead to a reduction in water use. The results of the study showed that the average daily water use in the high-rise apartments is around 3% higher than in the low-rise apartments (the average daily water consumption in the high-rise and low-rise apartments was 379 and 367 L per day, respectively). More interestingly, considering the average household size of 1.9 and 2.3 persons in the high-rise and the low-rise apartments, respectively, on a per capita basis the water consumption in the high-rise apartments was around 24% higher than in the low-rise apartments (i.e. per capita water consumption in the high-rise and low-rise apartments was 199 and 160 L per person per day, respectively). The higher per capita water consumption in the high-rise apartments can be mainly attributed to the smaller household size in this sector. In general, household size can exert an important effect on the per capita domestic water consumption (Domene & Saurí, 2006). By decreasing the household size, the per capita water consumption typically increases since economies of scale cannot be achieved in the smaller households (for instance, full loads in washing machines, dishwashers, etc.) (Arbués et al., 2003; Domene & Saurí, 2006; Hummel & Lux, 2007). These results imply that although housing intensification may have some social, financial and environmental benefits (Randolph, 2006; Haarhoff et al., 2012), in respect of water consumption it may not substantially help to reduce water demand since the higher density housing may lead to smaller household size which can make the efficient use of water more difficult.

Housing intensification (i.e. transition from large single houses to the more intensified multi-unit housings) is an ongoing phenomenon in many major cities around the world. The findings of this research in conjunction with other studies that investigated low-rise and single house water use in Auckland (Ghavidelfar et al., 2016a, 2017b) can help water infrastructure/water companies to better understand the implications of housing intensification for water use and to evaluate the effects of urban development policy on future water demand (Ghavidelfar et al., 2017a).

Conclusions

Over the past decade, intensive housing development, in the form of multi-unit housing, has been extensively promoted in Auckland in order to mitigate the social, economic and environmental concerns arising from low density residential development, while addressing the increasing need for dwellings by the growing population.

In order to take into account the implications of increasing high-rise apartment living on the future water and wastewater planning in Auckland, this study thoroughly investigated water demand and its determinants in this sector. To accomplish this, the study utilized GIS to integrate apartment water consumption data with the census microdata distinguishing sociodemographic characteristics of households living in the high-rise apartments, apartment physical characteristics, water pricing and weather variables. This rich dataset provided a unique opportunity to fully evaluate the effects of a wide range of variables on the apartment water use.

This study applied panel data models to evaluate water demand in a sample of 11,000 apartments, within 126 census meshblock units, over a period of 6 years. Through examining three common panel data methods (i.e. pooled, fixed and random effects), the study found that the random effects is the most appropriate model for the developed dataset. The random effects model was capable of evaluating the impacts of both time-varying (e.g. household sociodemographics, weather and water pricing) and time invariant (e.g. housing characteristics) variables on apartment water demand, while controlling for the heterogeneity among different apartments.

The study revealed that the number of people in the household (i.e. household size) is the most important determinant of water demand in the Auckland apartments. The results showed that other socioeconomic variables such as household income, age of residents and property ownership did not have significant correlation with the apartment water use. This is because at the Auckland apartments most of the water is used for basic needs (i.e. indoor use) and the outdoor usage, which is more sensitive to the income-related variables, generally comprises a negligible share of household water use. This characteristic of apartment water demand in Auckland can also explain the insignificant impact of property outdoor characteristics (i.e. lot size, garden and swimming pools), water pricing and weather conditions on the water consumption. In general, all of these variables typically affect outdoor water use rather than indoor use. The study also showed that the age of the dwelling and number of apartments in the building did not have significant correlation with the apartment water consumption.

This study also compared the high-rise apartment water use and its determinants with the low-rise apartments. This enabled the study to examine the effects of housing densification in the central city areas on the water use where the higher density housing is considered as a sustainable urban form in contemporary urban planning. The results of the study showed that regarding the determinants of water use, there is not a considerable difference between the high-rise and low-rise apartments. In both sectors the household size was the major driver of water use and the effects of other socioeconomic variables, housing characteristics, weather and water pricing were marginal. The seasonal variation of water use was also limited in both sectors, implying that the majority of water is used for basic needs (i.e. indoor use) in the apartments. The comparison also revealed that creating higher density housing in the central city areas may not lead to the reduction of water consumption. This is because in the higher density housing the size of households can be smaller. This may result in the less efficient use of water and subsequently an increase in per capita water consumption.

This study demonstrated how the data integration technique can be used to utilize the full potential of available GIS-based data to evaluate water consumption in the apartment sector, where data scarcity is a significant issue. Through data integration, this study managed to reveal the major determinants of water consumption in the high-rise apartments in Auckland. However, the study was limited by the availability of water consumption and socioeconomic data at the individual apartment level. This would be a domain for further research, where the water consumption, property information and socioeconomic variables can be integrated at the apartment level to provide better understanding of apartment water use in the compact city environment.

Acknowledgements

The funding for this research was provided by Watercare Services Limited. Access to the census data used in this study was provided by Statistics New Zealand under conditions designed to give effect to the security and confidentiality provisions of the Statistics Act 1975. The results presented in this study are the work of the authors, not Statistics NZ.

References

References
Agthe
D. E.
&
Billings
R. B.
,
2002
.
Water price influence on apartment complex water use
.
Journal of Water Resources Planning and Management
128
(
5
),
366
369
.
Arbués
F.
,
García-Valiñas
M. A.
&
Martínez-Espiñeira
R.
,
2003
.
Estimation of residential water demand: a state-of-the-art review
.
Journal of Socio-Economics
32
(
1
),
81
102
.
Arbués
F.
,
Barberán
R.
&
Villanúa
I.
,
2004
.
Price impact on urban residential water demand: a dynamic panel data approach
.
Water Resources Research
40
(
11
),
W1140201
W1140209
.
Arbués
F.
,
Villanúa
I.
&
Barberán
R.
,
2010
.
Household size and residential water demand: an empirical approach
.
Australian Journal of Agricultural and Resource Economics
54
(
1
),
61
80
.
Auckland Council
2015
.
Property Information
.
Auckland
,
New Zealand
. [accessed 12 October 2017].
Billings
R. B.
&
Jones
C. V.
,
2008
.
Forecasting Urban Water Demand
, 2nd edition.
American Water Works Association
,
Denver
,
CO
.
Boon
J.
,
2010
.
The interplay of market forces and government action in the achievement of urban intensification: the case of Auckland, New Zealand
.
Journal of Urbanism
3
(
3
),
295
310
.
Chang
H.
,
Parandvash
G.
&
Shandas
V.
,
2010
.
Spatial variations of single-family residential water consumption in Portland, Oregon
.
Urban Geography
31
(
7
),
953
972
.
CliFlo
2015
.
New Zealand's National Climate Database
.
National Institute of Water and Atmospheric Research
,
New Zealand
.
Dziedzic
R.
,
Margerm
K.
,
Evenson
J.
&
Karney
B. W.
,
2015
.
Building an integrated water-land use database for defining benchmarks, conservation targets, and user clusters
.
Journal of Water Resources Planning and Management
141
(
4
),
1
9
.
Ghavidelfar
S.
,
Shamseldin
A. Y.
&
Melville
B. W.
, (
2016a
).
A multi-scale analysis of low-rise apartment water demand through integration of water consumption, land use, and demographic data
.
Journal of the American Water Resources Association
52
(
5
),
1056
1067
.
Ghavidelfar
S.
,
Shamseldin
A. Y.
&
Melville
B. W.
, (
2016b
).
Estimation of the effects of price on apartment water demand using cointegration and error correction techniques
.
Applied Economics
48
(
6
),
461
470
.
Ghavidelfar
S.
,
Shamseldin
A. Y.
&
Melville
B. W.
, (
2017a
).
Future implications of urban intensification on residential water demand
.
Journal of Environmental Planning and Management
60
,
1809
1824
.
doi:10.1080/09640568.2016.1257976
.
Ghavidelfar
S.
,
Shamseldin
A. Y.
&
Melville
B. W.
, (
2017b
).
A multi-scale analysis of single-unit housing water demand through integration of water consumption, land use and demographic data
.
Water Resources Management
31
(
7
),
2173
2186
.
Goodyear
R.
&
Fabian
A.
,
2014
.
Housing in Auckland: Trends in Housing From the Census of Population and Dwellings 1991 to 2013
.
Available from http://www.stats.govt.nz/ [accessed 12 October 2017]
.
Haarhoff
E.
,
Beattie
L.
,
Dixon
J.
,
Dupuis
A.
,
Lysnar
P.
,
Murphy
L.
&
Solomon
R.
,
2012
.
Future Intensive Insights for Auckland Housing
. In:
A Research Report Prepared for the Auckland Council, Published by Transforming Cities: Innovations for Sustainable Futures, National Institute of Creative Arts and Industries
.
The University of Auckland
,
Auckland
,
New Zealand
, pp.
1
296
.
Hill
R. C.
,
Griffiths
W. E.
&
Lim
G. C.
,
2010
.
Principles of Econometrics
, 4th edn.
John Wiley & Sons
,
New York
.
Hoffmann
M.
,
Worthington
A.
&
Higgs
H.
,
2006
.
Urban water demand with fixed volumetric charging in a large municipality: the case of Brisbane, Australia
.
Australian Journal of Agricultural and Resource Economics
50
(
3
),
347
359
.
House-Peters
L. A.
&
Chang
H.
,
2011
.
Urban water demand modeling: review of concepts, methods, and organizing principles
.
Water Resources Research
47
(
5
),
1
15
.
House-Peters
L.
,
Pratt
B.
&
Chang
H.
,
2010
.
Effects of urban spatial structure, sociodemographics, and climate on residential water consumption in Hillsboro, Oregon
.
Journal of the American Water Resources Association
46
(
3
),
461
472
.
Hummel
D.
&
Lux
A.
,
2007
.
Population decline and infrastructure: the case of the German water supply system
.
Vienna Yearbook of Population Research
, pp.
167
191
.
Kenney
D. S.
,
Goemans
C.
,
Klein
R.
,
Lowrey
J.
&
Reidy
K.
,
2008
.
Residential water demand management: lessons from Aurora, Colorado
.
Journal of the American Water Resources Association
44
(
1
),
192
207
.
LINZ
2015
.
LINZ Data Service
.
Land Information New Zealand
,
Wellington
,
New Zealand
.
Available from http://www.linz.govt.nz/about-linz/linz-data-service [accessed 12 October 2017]
.
Loh
M.
,
Coghlan
P.
&
Australia
W.
,
2003
.
Domestic water use study in Perth, Western Australia 1998–2001
.
Water Corporation, Perth, Western Australia
, pp.
1
235
.
Martinez-Espiñeira
R.
,
2002
.
Residential water demand in the Northwest of Spain
.
Environmental and Resource Economics
21
(
2
),
161
187
.
Mayer
P. W.
,
Bennett
D.
,
Deoreo
W. B.
&
Towler
E.
,
2006
.
Third-party billing of multifamily customers presents new challenges to water providers
.
Journal of the American Water Works Association
98
(
8
),
74
82
.
Mieno
T.
&
Braden
J. B.
,
2011
.
Residential demand for water in the Chicago metropolitan area
.
Journal of the American Water Resources Association
47
(
4
),
713
723
.
NIWA
2015
.
Climate Summaries
.
National Institute of Water and Atmospheric Research
,
New Zealand
.
Polebitski
A. S.
&
Palmer
R. N.
,
2010
.
Seasonal residential water demand forecasting for census tracts
.
Journal of Water Resources Planning and Management
136
(
1
),
27
36
.
Rockaway
T. D.
,
Coomes
P. A.
,
Joshua
R.
&
Barry
K.
,
2011
.
Residential water use trends in North America
.
Journal of the American Water Works Association
103
(
2
),
76
89
.
Russac
D. A. V.
,
Rushton
K. R.
&
Simpson
R. J.
,
1991
.
Insights into domestic demand from a metering trial
.
Journal of the Institution of Water and Environmental Management
5
(
3
),
342
351
.
Schleich
J.
&
Hillenbrand
T.
,
2009
.
Determinants of residential water demand in Germany
.
Ecological Economics
68
(
6
),
1756
1769
.
Shandas
V.
&
Parandvash
G. H.
,
2010
.
Integrating urban form and demographics in water-demand management: an empirical case study of Portland, Oregon
.
Environment and Planning B: Planning and Design
37
(
1
),
112
128
.
Statistics-NZ
2010
.
Apartment Dwellers: 2006 Census
.
Statistics New Zealand
,
Wellington
.
Statistics-NZ
2015
.
Statistics New Zealand. New Zealand. Available from http://www.stats.govt.nz/ [accessed 12 October 2017]
.
Troy
P.
&
Holloway
D.
,
2004
.
The use of residential water consumption as an urban planning tool: a pilot study in Adelaide
.
Journal of Environmental Planning and Management
47
(
1
),
97
114
.
Watercare
2014
.
Auckland Regional Water Demand Management Plan 2013–2016
.
Watercare Services Limited
,
Auckland
,
New Zealand
.
Watercare
2017
.
Water Sources
.
Watercare Services Limited
,
Auckland
,
New Zealand
.
Wentz
E. A.
&
Gober
P.
,
2007
.
Determinants of small-area water consumption for the City of Phoenix, Arizona
.
Water Resources Management
21
(
11
),
1849
1863
.
Wentz
E. A.
,
Wills
A. J.
,
Kim
W. K.
,
Myint
S. W.
,
Gober
P.
&
Balling
R. C.
Jr
,
2014
.
Factors influencing water consumption in multifamily housing in Tempe, Arizona
.
The Professional Geographer
66
(
3
),
501
510
.
Wooldridge
J.
,
2012
.
Introductory Econometrics: A Modern Approach
, 5th edn.
South-Western Cengage Learning
,
Mason, OH
.
Zhang
H. H.
&
Brown
D. F.
,
2005
.
Understanding urban residential water use in Beijing and Tianjin, China
.
Habitat International
29
(
3
),
469
491
.