The relevance of wastewater treatment service has increased in recent years, since it has a significant impact on the natural environment. A treatment plant facilitates energy generation, the recovery of products from waste, and the reuse of wastewater for industrial and irrigation purposes. An indirect environmental effect is the high energy consumption for pumping water and for tank aeration. The objective of this research is to develop a tool for measuring the energy costs of wastewater

more »

... ment plants and identifying how they can be reduced. The method adopted is double-bootstrap data envelopment analysis. The results show that the variables with a significant influence on efficiency are the chemical oxygen demand concentration; plant capacity; rate of used capacity, which positively affects efficiency; weight of industrial customers, which exerts a negative impact; and aeration system, with a negative impact for turbines. This paper suggests the adoption of an effective control tool to monitor the costs drivers and energy expenditure of water utilities. As reported by some authors, the higher consumption of electric energy is required by pumps (79%) to pump and treat wastewater [11] . The active sludge treatment, with the biological oxidation of pollutants, absorbs approximately 50% to 65% of the total consumption energy, in addition to the 11% required for the primary treatment, for grit, sand, and oil removal as well as sedimentation [12] . However, plant managers have the opportunity to significantly reduce energy costs through preliminary energy audits followed by process modifications. As described in [13] , only an optimization of the aeration and pumping activities allows for annual savings ranging from 547 to 1057 million kWh, reducing the energy consumption by 6%. Prior research on energy efficiency in WWTPs shows that several variables should be constantly monitored by the plant manager because they exert an influence on efficiency trends. According to [14] , large variations in the quality of wastewater inflows, measured in terms of the five-day biochemical oxygen demand/nitrate as nitrogen (BOD5/NO3-N) ratio, reduce the efficiency achieved in biological denitrification. This result successfully illustrates the trade-off related to the high BOD5 concentration: this pollutant feeds microorganisms and allows the digestion of nitrogen, but, at the same time, it contributes to generating sludge through the fast growth of bacteria. The extant technical literature provides an optimal value of this ratio, which should be around 100:17-100:19 [15] . Recently, the use of a bootstrap approach found that if the value of the chemical oxygen demand (COD)/BOD5 ratio does not respect the standard of scientific literature, normally, the plant managers add more carbon elements and begin other chemical treatment processes, despite the incremental costs per m 3 of wastewater treated of these alternatives and, consequently, this reduces the level of efficiency [16] . Further, it has been demonstrated how seasonality actually influences the energy efficiency of WWTPs, especially for activated sludge technology [17] ; similarly, it has been shown that the consumption of energy in a sample of 177 Spanish WWTPs, using extended aeration technology, is 0.82 kWh/m 3 , with better efficiency recorded for non-seasonal plants [17, 18] . Further, in terms of the technology used to aerate the oxidation tanks, it has been shown that diffusers are more efficient than turbines, since they allow for a higher removal rate of COD, but, at the same time, they require greater energy consumption per m 3 of wastewater treated [18] . Plant size is another key performance driver of energy costs in wastewater treatment. Several studies [18] [19] [20] have shown that cost savings are achieved by larger plants in terms of population equivalent (PE), while the m 3 is not a relevant factor to capture increasing return to scale. Moreover, it has been demonstrated that the consumption of energy per kg of COD removed significantly decreases from plants with a capacity of less than 2000 PE (3.21 kWh/kg COD) to those with a capacity of more than 100,000 PE (0.85 kWh/kg COD) [20] . Conversely, [21] show that diseconomies of scale can significantly affect wastewater treatment if an efficiency measure that includes greenhouse gas emissions is considered [21] . In addition, energy consumption is related to the load factor, the ratio between the load of wastewater inflows received to the design value of the plant. Further, two particular studies have confirmed that undersized plants work better than do oversized plants, and energy savings increase when the load factor approaches 100% [22, 23] . In light of this scarce and quite recent literature, the current article aims to provide further insights into the energy efficiency of WWTP, observing the effect exerted by several previously examined variables as well as by new environmental factors, such as the rate of wastewater coming from non-domestic customers, rate of production capacity used, and rate of sludge disposed in agriculture. Different from other researches, this article studies the energy efficiency of the whole treatment process, including wastewater and sludge handling, of 127 WWTPs. The prior literature adopts several measurements for energy efficiency. Traditionally, energy consumption is reported in terms of kwh/m 3 or per unit of population (kwh/PE). This measure can have some drawbacks for benchmarking purposes when COD concentration varies among plants [22] ; thus, the ratio of kWh to kg of COD removed (or other pollutants) is estimated [24] . Considering that the COD concentrations of inflows are quite similar for the plants observed, this article examines two measures of energy efficiency: (i) the cost of energy per m 3 of wastewater treated and (ii) DEA score, based on a set on inputs and outputs.