69749 Quality perception in utility provision A conceptual and empirical approach Research team General coordinator: Marcelo Celani† Experts in survey design, field studies and statistics: Luis Acosta†† and María José De Gregorio††† Sector and regulatory expert: Martín Rodríguez Pardina†††† General contributors, data processing and econometrics: Ariel Melamud*, Germán Caruso** and Germán Sember*** † Universidad Di Tella, Buenos Aires mcelani@utdt.edu †† FLACSO (Latin-American School of Social Sciences) lacosta@fibertel.com.ar. ††† UBA (University of Buenos Aires); Cualitativo y Cuantitativo S.A. mjdegregorio@yahoo.com.ar †††† Macroconsulting S.A, Buenos Aires, Argentina. marp@macroconsulting.net * World Bank. arimelamud@yahoo.com.ar ** Universidad de San Andrés gcaruso@udesa.com.ar *** Macroconsulting S.A, Buenos Aires, Argentina gsember@macroconsulting.net 1 Contents The importance of measuring the satisfaction level of utility demand ................................ 3 Objectives ................................................................................................................................... 6 Motivation for this study................................................................................................................. 3 Theoretical Approach ................................................................................................................... 8 Methodological approach to quality assessment........................................................................... 22 The survey................................................................................................................................. 22 Sectors analyzed........................................................................................................................ 24 Areas covered............................................................................................................................ 24 Associated answers ................................................................................................................... 26 Unit of Analysis ........................................................................................................................ 27 Sample selection ....................................................................................................................... 27 Socioeconomic stratification of surveyed households .............................................................. 29 Selection of Census Radii and Sample Points .......................................................................... 35 Empirical approach: Results ......................................................................................................... 37 Determinants of Overall Satisfaction with the Service ............................................................. 37 Quality Perception and Characteristics of Customers .............................................................. 55 Overall Satisfaction and Partial Indicators ............................................................................... 80 Company benchmarking ........................................................................................................... 93 Joint assessment of both waves............................................................................................... 106 Conclusions and Recommendations ........................................................................................... 118 Annex 1. .................................................................................................................................. 121 Bibliography ........................................................................................................................... 126 2 Motivation for this study The importance of measuring the satisfaction level of utility demand From a general perspective, customer satisfaction is a measure of the quality perceived by them and this aspect is becoming more and more relevant for the design and assessment of social policies worldwide and, in particular, of policies related to utilities, because it shows an approach from the demand’s point of view, contrary to the traditional quality indexes that show the supply’s standpoint. Indeed, service quality is becoming an essential requirement of modern regulation, and its improvement is today a strategic objective of regulatory agencies and even of international development agencies. However, the desired improvement of quality can only be reached through the understanding of its determinants and its evaluation from the demand’s point of view. In a certain sense, as the old business adage goes: “you can’t improve what you don’t measure� (Reichheld and Sasser, 1990). Obviously, the discussion on the possibility of measuring satisfaction is relevant in any sector of the economy; however, when considering utilities, it gets a different importance since their “public� nature tends to guaranty the “common� welfare of the population. This means that the public nature of the service plays a kind of leveling role for each individual. The importance of properly measuring satisfaction lies in the fact that it would offer a parameter to identify the aspects of the service that need to be improved for a more efficient and fair provision, thereby increasing companies’ benefits and, at the same time, increasing households’ access and general wellbeing. The distinctive features of utilities determine that the universe of study is the individual and that subjective issues prevail in the formation of expectations and perceptions, with an important role of human resources, as well as equipment and infrastructure, in the process of quality and satisfaction assessment. On the contrary, in the goods sector, more objective assessments can be 3 performed before the sale of the goods and then, after the sale, we find out consumer needs and demands. The regular use of a service by the customer turns it into a citizen-customer. Taking into account the importance of utilities and their general interest in terms of poverty reduction, it is necessary to supplement the prevailing paradigm in the studies that deal with measuring satisfaction by introducing in these studies some indicator of the access to a certain minimum quality level, generally determined by regulatory agencies. The introduction of price cap regulation, which creates incentives to productive efficiency –cost reduction-, turns service quality review and control particularly important. If there are not specific restrictions in this sense, one of the ways to reduce costs can be the reduction of service quality. Therefore, price cap regulation takes into account a group of issues that include not only prices but also service obligations consisting of minimum operation and quality standards. This has two direct implications for the importance of this study. First, the features discussed above make it necessary to define a quality level for the network as a whole. Consequently, the role of the regulator, who evaluates and defines the quality standard to be met and sets the rules and incentives for the company to comply with the necessary service level, becomes crucial. The study of customers’ willingness and ability to pay according to different quality levels is necessary to determine the contractual conditions of the service since it supplements quality studies from the supply standpoint and allows getting a broader picture of the quality issue. Second, it is necessary to carry out a continuous and periodic monitoring of the service in order to verify the compliance with the standards set. Therefore, the generation of information through surveys represents an important tool in every efficient regulatory system. This is the reason why, 4 many regulatory frameworks include explicit obligations for the performance of periodic service satisfaction surveys. 1 For these reasons, the quantification of satisfaction in the utilities sector –which results particularly complex because of the intangible nature of the services, on the one hand, and the subjective feature entailed by quality, on the other hand- appears to be necessary to supplement price cap regulation. Quality has historically been a concept with strong engineering connotations and, therefore, it has always been associated with production technology and service distribution. From a technical or engineering point of view, quality is a concept with a certain degree of objectivity and the different industries have standards to measure it. Regulatory agencies and providing companies generally use these standards to perform the respective measuring. However, the perceived quality considered as a distinctive element of customer satisfaction represents a subjective concept and this explains its non-universal nature. In general, the number of customers of a service can be large and quite heterogeneous in terms of needs and, therefore, in terms of demand. This frequently leads to the conclusion that there is not only one type of customer but several groups with expectations that may be equivalent to or different from each other and, therefore, it is very difficult to satisfy all the people through the production of standardized services. Although there are many ways of assessing customer satisfaction, most of them are based on a multidimensional approach that, in general, includes different aspects of the service and characteristics of the customers that make it possible to estimate representative statistics about quality perception. 1 However, surveys are only one of the necessary tools since there are quality dimensions that may not be perceived by customers but may significantly impact the service. A typical example of this type of problem is the chemical component of water. 5 The different service dimensions that can be identified do not usually have the same importance or the same priority for the different customers (Bañón and Carrillo, 1996). Disagreement as to the kind of evaluation (or valuation) criteria that are significant, fair or acceptable in a given situation is unavoidable. In sum, it is difficult to reach an agreement about the criteria when different values or interests are involved. In addition, the demands for higher service quality may be quite significant, especially if there is some kind of “demonstration effect� caused by the provision of the services (or what it is supposed to be the service provision) in other countries or in the private sector (Beltrán Villalva, 2000). In infrastructure sectors, the technological features also affect the aspects related to quality. In competitive sectors, the market typically offers different combinations of price and quality, allowing customers to choose the service of the utility company that best matches their budgetary restrictions. However, in the utility sector, the natural monopoly features make this selection among the different utility companies impossible. Moreover, in these infrastructure industries, the quality level is, to a great extent, determined by network design and management and therefore, there is no physical possibility for the utility company to render different quality services to the different customers. Getting to know the customer expectations in more detail is then a way of guiding and designing supply in accordance with market requirements. If we also measure the adjustments or gaps between customer expectations and perceptions, then we will get a more reliable quality indicator. The integration of the traditional approaches for quality evaluation and this new approach of customer satisfaction allows a broader and more precise assessment of operators’ performance, thereby improving the regulatory system and contributing to poverty reduction as a public policy. Objectives 6 The perspective of this study is not only academic but also practical. As regards the academic aspect, this work will explore the professional literature related to quality perception and, particularly, the conceptual discussion about the object of satisfaction measurement in a market research involving utility customers and their perception of service quality. As the immediate consequence of this theoretical review, the study intends to develop the general guidelines of a survey meeting the best standards from the point of view of the theory. In the second place, this work seeks to carry out some measuring so as to empirically evaluate the principles analyzed in the previous section. In light of these purposes, the work is structured as follows: chapter one deals with the theme from a theoretical perspective and reviews literature arising from disciplines such as marketing and economics. Chapter two develops the general structure of a survey with suggestions about its construction and design that will be taken into account when performing the appropriate field studies; chapter three presents the results of the field studies. Finally, the study shows the conclusions drawn. Acknowledgments The authors wish to thank Tomás Serebrisky and the World Bank for their support, suggestions and constant contributions to the development and completion of the project. This project could not have been completed without the economic support of the Department of International Development (DFID) through the Markets and International Trade Program (LAMIT) and Public Sector and Political Systems (PSPS). 7 Theoretical Approach The discussion presented below deals with the initial problem faced by every applied quantitative research regarding the definition of the variables representing the theoretical concepts to be assessed. From a conceptual point of view, quality is defined as the totality of features and characteristics, whether expressed or not, of an entity (product or service) that bears on its ability to satisfy consumers’ stated or implied needs. The concepts of quality and satisfaction have been the focus of the debate among researchers. Some authors suggest that the perceived quality is a precursor of satisfaction, while others state that satisfaction is the antecedent of quality. The approach that we consider to be the most appropriate deals with a dynamic relationship where the perceived quality is the antecedent of satisfaction and a greater level of customer satisfaction will reinforce their perceptions of quality (Berné, Múgica and Yagüe, 1996). The studies about quality and satisfaction have proliferated mainly in the disciplines of management and marketing. The models that deal with these constructions also analyze the causal relationship and the different aspects that distinguish these concepts from the firm’s reputation, customer loyalty or retention and the value of a service or product. A market research aimed at analyzing these aspects needs to start from the idea that customer aspirations, likes and motivation are influenced by an important number of variables, very difficult to define and, consequently, to measure statistically. This is the reason why lately, perception studies have been developed through the use of opinion polls which allow gathering quantitative information. Particularly, the field work of this study reflects not only subjective aspects of the overall and partial satisfaction of water, electricity and gas customers with technical service, the customer service and the billing system but also some objective aspects of the service quality such as service interruptions and the adjustment of service pressure or voltage. 8 The perceived quality considers the customer as an evaluator and it is described as the degree and direction of discrepancy between the perceived quality and consumer expectations of the service (Oliver, 1980). Quality analysis unit is the individual, who has a certain attitude towards the service which depends on the initial attitude at the moment of consumption and the degree of satisfaction with the service provision. Both subjective (cognitive, psychological and emotional) and objective elements (distance between perception and expectations and past experiences) are combined in the formation of the attitude towards the utility company. In this sense, quality assessment is not easy since it considers the performance of service general attributes and functions, including the social approval of the company (reputation) and the costs and efforts sacrificed in it consumption (perceived value). Satisfaction can relate to the assessment of a specific transaction or the overall experience with the service. The first one focuses on how satisfied a customer is with the provision of the service at a given time, while the second includes the whole consumption experience and represents the performance indicator that more impact causes on the company’s reputation and benefits. Today’s interest lies in the cumulative satisfaction which is a better predictor of consumers’ behavior and the performance of the utility company or a sector of the economy (Johnson and Fornel 1991). The company’s reputation is also an attitude that customers have towards the providing company and, in this case, it relates to the aspects that consumers associate with the name or the image of the company, that in turn impact on quality perception. Reputation involves an overall evaluation of the service longer than the construction of satisfaction, because reputation is built on the basis of the continuity of satisfaction with the service. However, the importance of enjoying a good reputation lies in the fact that it significantly influences consumers’ selection when the attributes of the provision are difficult to notice, as it is the case of utilities. Customer loyalty or retention expresses future behavioral intentions of users of a certain service, indicates the probability of changing current provider and depends on emotional 9 factors (trust, involvement) as well as on quantitative factors (substitution costs, current performance and reputation). In certain contexts, loyalty is attributed to the fact that customers are really satisfied with the quality obtained, whereas in other cases there may be a predominance of technical, economic or psychological factors raising more barriers to exit. Companies aim at generating satisfaction as a retention strategy, because barriers are supposed to disappear in the long term. To sum up, retention looks at the way consumers act ahead while satisfaction looks at their past behavior. The service value perceived is a comparison made post-consumption based on the price or rate, quality-price relation, assessment on whether the results comply with the expectations and are appropriate to the sacrifices made. The value positively depends on personal values, usefulness and quality, whereas perception of monetary and non- monetary sacrifice (time and effort) has a negative impact on the value. Bolton and Drew (1991) sustain that quality and satisfaction can be measured through the value perceived, comparing the cost assumed by customers with the benefits obtained from the service provision. Quality and satisfaction are antecedents of value, although the most valuable services generate more satisfaction. Likewise, value perception significantly affects buying intention (loyalty) and a company’s reputation. The growth experienced by the service sector created the need to further analyze customer satisfaction and quality measurement, to identify its determinants and quantification methodology, and to determine if such determinants differ from those of the product sector. Quality management is important because it is the main tool used by successful companies to keep their customers satisfied, retain them and increase their share of the market. Improvements in service provision and consumer satisfaction start with the identification of customer needs. The measurement of subjective and objective components of service quality provides executives and regulatory agencies with information on the quality perceived by customers, in order to improve the most deficient aspects of the service and pay attention 10 to those attributes with greater value for and level of impact on satisfaction. 2 In cases where providers and regulators conduct analyses to estimate consumer expectations, it will be possible, from the public and private sectors, to work on the maximization of consumer satisfaction and improvement of their quality of life, causing company return and number of customers to increase. Latu and Everett (2000) developed the basic characteristics of services, which are different from products due to the former’s intangibleness, inseparability, heterogeneity, provider-consumer interaction, and perishable nature. On the one hand, quality assessment is not equal for products and services since the latter involves a greater number of intangible elements. On the other hand, services are usually associated with what economic literature called “experience goods�, whose main characteristic is that their quality assessment, and consequently their provision of satisfaction, is subsequent to their consumption or use. In the case of utilities, the object of assessment is the service flow quality, i.e. continuous service provision over time, which enables customers to make their assessments through experience, quality and degree of satisfaction obtained from consumption. In the case of products, or rather in some of them, there are more possibilities of assessing intrinsic characteristics beforehand because these are visible, may be included in inventories, tested or explored before purchasing them (Gronroos, 1984). Service intangibility makes reference to the fact that services are immaterial, they are not palpable objects but performances. Most services may not be measured, counted, tested and verified. All there is to see are equipments, personnel, facilities, advertising, communication, images, etc. Particularly, the more intangible the service is, the more word-of-mouth communications, prices and image, the most noticeable elements of service provision, are considered for service assessment. In this regard, the degree of objectivity in quality assessment becomes more complex, and the subjective dimension and cognitive aspects of each individual are greatly involved during service provision. 2 Thus, the greater the impact of an individual variable on overall customer satisfaction with the service, the greater should be the attention paid to such variable from the point of view of regulatory framework control and design. 11 This question turns subjectivity into the most relevant dimension for customer satisfaction and quality assessment (Bateson, 1985). However, it is possible to include in satisfaction studies indicators that allow for an objective assessment of product and service technical quality. 3 Consumers obtain resources and information to assess service quality in the simultaneous process of interaction with providers. Therefore, the performance of the human resources used is a key factor for quality generation and measurement. Continuing with the distinction between the objective and subjective aspects involved in provision assessment, there are two fields typical of the characteristics of the service sector, differentiated by the technical and functional quality of the productive process. The first field relates to the instrumental service performance which may be objectively measured such as any technical dimension of the productive process, while the second one includes subjectively measured qualitative aspects as regards “customer assistance� provided by human resources (Meuter and Bitner, 1997). In those services with a more intensive use of labor, there is greater heterogeneity of performance among different providers, consumers and in the inter-temporal aspect, this is the reason why more attention is paid to functional quality. Likewise, Gronroos (1984) classifies the variables that affect satisfaction to a certain extent in internal variables, on which the company may exercise higher control, and external variables, for those which are beyond such control. External variables may be found in customers’ tradition, values, ideology and word-of-mouth communications. Expressive service performance involves consumers’ psychological sphere, which depends on the contact made with the personnel and with other customers, and on which companies have little impact. However, companies have tools to influence in the technical quality of its products and services, commercial service and human resources 3 Some objective measures of technical service quality are number and duration of water and electric power service interruptions, while technical product quality may be measured through the number and frequency of inadequate water and gas pressure, of voltage variations, and through bad taste, smell and color of water. 12 performance. 4 For such purpose, employee training and coaching programs may be implemented, and control and quality tests may be run on plants, machinery and services. Technical quality is an essential though insufficient prerequisite to satisfaction, since customers will not be satisfied with technical quality, deficient assistance and an unfavorable atmosphere in their relation with employees and other customers. Inherent service attributes determine that consumer satisfaction assessment depends on certain aspects unconnected with their provision, causing satisfaction measurement to be more complicated in connection with goods. Woodruff, Cadotte and Jenkins (1983) sustain that the assessment process is complex because it pertains to an abstract reaction that involves emotional, cognitive and psychological aspects which may vary according to different situations and are different for each individual, and which also depend on the underlying cultural context, rules and customs of the community. Personal attitudes are unique and it is impossible to make generalizations through the study of individuals’ behavior. In this regard, Gustafsson, Johnson and Roos (2005) assert that there may be changes in the parameters used by customers in their assessment because their needs and preferences change, together with service quality. The satisfaction measurement structure changes due to the existence of objective situations where performance is deteriorated/perfected, and also to subjective reactions resulting from economic, demographic, technical (obsolescence) and labor changes. The inherent characteristics of each service, or their heterogeneity, determine the need for a different degree of consumer involvement with the productive process. In cases such as Education and Health, active consumer-provider communication is recommended, since it has been proven to produce positive externalities on service quality and level of satisfaction. In such services, a small customer participation will not allow for the identification of those aspects which might be improved nor of particular needs and requirements, thereby reducing the general well-being of all users. In more standardized services, such as utilities, the need for consumers to intervene in production has not been 4 Commercial service may be measured through service interruption announcements, billing features and complaint-handling systems. 13 proven (Zuloaga, 2002). Nevertheless, utility customer satisfaction studies make it possible to identify those dimensions regarded by users as the most relevant for the service, and possibly bring about improvements in regulation, service provision and quality of life of citizens. On the other hand, consumers have information on the attributes the service must meet in order to fulfill their requirements. Satisfaction may be generally measured with an overall index contemplating all attributes that constitute the service, or partially measured with indexes that only contemplate certain attributes and are weighed according to the relative significance allocated to each by consumers in determining overall satisfaction. 5 The subjective nature of satisfaction and the difficulty of its measurement derive from the fact that the significance of each dimension differs not only among consumers due to their preferences, needs and availability of resources, but also among customers’ different situations and social and cultural contexts. In this regard, satisfaction studies allow for, among other things, the identification of the existing relation between overall satisfaction indexes and partial satisfaction indexes, as well as the assessment on whether satisfaction levels are connected with customers’ socioeconomic level, the company providing the service, the company information available to customers, and the level of satisfaction with the provision of another service. 6 Satisfaction determinants Some authors have tried to classify the attributes that generally characterize overall and/or partial satisfaction indexes. Kano and others (1996) propose three types of 5 Partial satisfaction indexes may measure, among other things, company performance during service interruptions and/or pressure/voltage variations depending on each sector, complaint-handling system, and detailed and easy-to-understand/calculate bills. 6 Beforehand, it is expected that: (i) the level of income will affect supply and demand for quality simultaneously; (ii) an individual’s degree of information on the service will affect their demand for quality and, consequently, their satisfaction; and (iii) the bad quality of a service will affect the quality of other services. Last, the comparison of quality dimensions among companies or benchmarking is one of the main instruments for reducing information asymmetry, which is a characteristic of the regulator-company relation. 14 attributes that may be tangible or intangible and affect consumer satisfaction at different degrees: • Basic factors: consumers take these for granted and therefore do not deem it necessary to express them. Although their presence does not increase satisfaction, their absence causes great dissatisfaction; • Performance factors: those expressed by consumers when asked what they expect from the service; depending on their fulfillment, such attributes will cause greater or lesser satisfaction; and • Excitement factors: customers do not expect them so their absence does not cause dissatisfaction but their presence significantly increases satisfaction. In infrastructure sectors, the quality-satisfaction relation is markedly asymmetric. To some extent, there are no “excitement� factors that may significantly increase customer satisfaction. On the contrary, small negative deviations from the “basic� requirements have a highly marked negative impact on customer satisfaction. The combination of excitement and performance attributes is what provides the company with a greater competitiveness level in the market, improving its reputation and image. However, the company must make sure it complies with the basic attributes due to the cost that may result from their deficiency. The “disconfirmation paradigm� is the first instrument used for measuring customer satisfaction with regard to a given product or service, and refers to the subjective emotional and cognitive process through which an individual compares their expectations regarding service provision with the effective performance, causing satisfaction or dissatisfaction. This methodology assumes the existence of expectations as regards the service prior to consumption, after which these are compared with perceived performance to finally assess whether the experience was better or worse than expected, thereby confirming or disconfirming such expectations (Woodruff, Cadotte and Jenkins; 1983). The comparison is positive when performance exceeds expectations, negative when expectations exceed performance and confirmative when expectations are approximately 15 equal to performance. Authors suggest the existence of three areas in a “Gaussian bell shape curve�: indifferent or confirmative, positive comparison and negative comparison. Oliver (1980) asserts that individuals may make different types of performance comparisons, according to expected, deserved, ideal and minimum tolerable provision. Although all of them are expectations, the last three refer to aspects broader than mere predictions. The notion of deserved and ideal provision implies regulatory standards required for quality provision, generally considered by regulatory agencies. Likewise, the universe of comparison used by consumers for building their expectations covers more aspects than the previous experience with service provision, since it includes other companies’ provision quality and quality perception by other users. In this sense, society’s prevailing cultural rules as to how the service should be provided influence the creation of expectations. The adaptation-level theory is applied to the satisfaction assessment process because it suggests that the individual perceives stimuli in connection with an adopted standard that depends on context and perceptions. Parasuraman, Zeithaml and Berry (1985) propose an instrumental methodology named SERVQUAL that estimates the “gap� between customer expectations and service perception. This is a general methodology and represents an interesting contribution since it aims at assessing and clarifying the concept of service quality and satisfaction. Literature argues whether the satisfaction and “overall service quality� concepts represent the same phenomenon. Bitner (1990), Parasuraman, Zeithaml and Berry (1985) and Bolton and Drew (1991) affirm that the difference between such concepts lies in that overall service quality entails a long-term assessment, where consumers compare ideal to actual provision, whereas satisfaction is a specific and exact measurement that takes into account what consumers effectively expect at a given time. Although most field studies assume that overall service quality is a precedent to customer satisfaction, we consider a dynamic relation between these two concepts for the purpose of addressing their measurement and assessment methodology. 16 In the overall service quality approach, the reason for gaps between expectations and perception may be based on: i) consumers’ expectations and executives’ perception of what customers expect; ii) executives’ perception of what customers want and service specifications (involving restrictions on resources, executives’ indifference and market conditions); iii) the specified service quality and the service provided (involves personnel that does not always develop a standardized conduct in spite of having procedure and function manuals); and iv) current service performance and external communication (if inappropriate, it may increase expectations as regards service provision possibilities). The model introduces five dimensions (tangibility, reliability, responsiveness, security and empathy) to measure the four gaps. It should be borne in mind that in infrastructure sectors, due to the technological issues discussed above, part of these relations is regulated by regulatory frameworks. Thus, in several dimensions that form the quality aspect, the executives’ perception of what customers want (ii) and the specified service quality (iii) are embodied in service contractual rules to be complied with by the company. The gap between customer expectations and executives’ perception (i) is also beyond market forces and must be reflected by regulations. This model has been widely used, criticized, modified and improved as regards its theoretical as well as operational aspects. The greatest progress was made in the area of design and registration format of expectations. Cadotte, Woodruff and Jenkins (1987) incorporate market signs for determining expectations, arguing that the latter do not depend solely on rules but that there is also a dynamic component involved in their determination, created by the competitiveness of other companies of the sector and the rest of consumers. Customers compare the service received with what other companies provide and, in turn, with what other customers receive, in order to assess provision equity and quality. 17 Carman (1990) uses the question of previous service experiences to propose that expectations equal to 0 should be adopted whenever there is no service experience, thereby calculating quality based only on perception, thus avoiding assessment distortion. In the case of frequently used services, such as utilities and durable goods, experiences indicate that expectations are stable and homogeneous over time. Therefore, Carman suggests that in any such cases expectations should be considered as constant. However, changes in structural factors (context, preferences, customs, etc.) affect expectations, the importance given to each attribute, and determine the need to gather data and assess service quality at certain intervals. On the other hand, Selnes (1993) analyzes the creation of expectations and the role of the company’s image in customer satisfaction and quality perception. The company’s reputation generates certain expectations in customers causing them to establish a “rule� on the acceptable level of service provision. This rule may also be created based on the average performance of a similar group of companies within a specific service sector or category. The assessment shall be determined with a higher level of objectivity in cases where physical evidence containing clear and reliable information on the service is made available to customers. On the contrary, consumers lack information for making their assessment in the case of services that are too technical and specialized, where the company’s reputation and image have a greater effect on quality perception. This is the reason why it is suggested that a reliable reputation should be built wherever an objective quality measurement is unlikely. Teas (1994) identifies that variability in consumers’ attitudes, assessments and responses in satisfaction surveys results from the different interpretations made of each question. The difficulty this presents for measuring quality and satisfaction should be considered not only when designing the questions but also when analyzing the results. The author also questions the use of “deserved or ideal� expectations since this involves assuming a perfect context for service provision, which does not match the truth. Another problem identified with the use of “unreal� expectations is that satisfaction may be low or high regardless the provision of a high- or low-quality service. 18 Reliability on expectations and the ambiguity of some services concerning the objective assessment of the degree of satisfaction are aspects addressed by Yi (1993). Particularly, service provision quality may be estimated with a high, medium or low probability based on the degree of credibility of expectations, which, in turn, depend on the source of information, previous service experience and service provision variability. Where the service does not fulfill the expectations, individuals with more reliable expectations will have a greater dissatisfaction than those who did not rely on their expectations. Additionally, this model suggests that satisfaction measurement should only be made through perception where services are highly ambiguous as to provision assessment, whereas expectations should be included in services whose performance is not ambiguous. Devlin, Gwynne and Ennew (2002) introduce communication, promises, advertising, image and reputation as antecedents for determining expectations. In addition, authors distinguish the factors that have an impact on simple expectations from those affecting anticipated expectations. While simple expectations depend on previous experience, promises and word-of-mouth communication, anticipated expectations are determined based on personal needs (physical, social, psychological, etc.). Although companies are able to influence simple expectations, they may not control anticipated expectations. The models that incorporated the structural relation defined in SERVQUAL have allowed for considerable progress in understanding how users assess the service received, although there are some weak points in the quality of predictor variables. Peter, Churchill and Brown (1993) noted the problems of multicollinearity, homoscedasticity and spurious correlations resulting from the use of “difference scores�. The authors sustain that the models used to measure quality and satisfaction include expectations, performance and their comparison as independent variables, and due to their correlation they arrive at biased estimators. Therefore, it is recommended that the use of the variable measuring the expectation-perception gap be avoided and only perception be used instead. 19 However, in spite of these weak points, which may partially be mitigated with the correct structure of field studies supporting these approaches, it is possible to affirm that the main contributions of this literature are the definition of the object to be measured, i.e. customer satisfaction, and the methodologies used to determine such estimate. Cronin and Taylor (1992) propose a methodological alternative -the SERVPERF model- more practical than the SERVQUAL model as regards its implementation. The SERVPERF model is applied to this study on service customer satisfaction. Such model uses only perceptions to estimate service quality and satisfaction, since it is assumed that expectations are included in said perceptions, based on the fact that when users answer about their degree of satisfaction, they implicitly include what they expected from the service. 7 The approach complements the SERVQUAL method by simplifying its operation, obtaining more consistent estimators, supported solely by attitudes, and abandoning the “disconfirmation paradigm�. In this sense, the model proposes to ask the battery of questions once, and not in terms of expectations and perceptions. The authors mentioned in the last place conclude that the suggested SERVPERF method is the one that better explains the causal relation among consumer loyalty, satisfaction and overall service quality. The instrumental rationality offered by the SERVQUAL model was improved with the SERVPERF model, which is currently the prevailing approach in satisfaction and quality measurement practices, and has been applied to the measurements of this study. Although the most significant advances of the academic field have been on the building of expectations and how they impact on the individuals’ overall service assessment, the design of the survey must aim at: simplifying satisfaction measurement including only service perception, without expectation assessment; preparing clear questions that do not 7 As mentioned above, service features and expectations are generally determined by the applicable regulatory framework. 20 lead to multiple interpretations; and including those attributes regarded by consumers as the most important for measuring service quality and satisfaction. The results obtained for such measurement must be carefully interpreted, since surveys would not reflect whether users obtain a basic quality level, regardless of their socioeconomic status. It has been said that individuals build their perceptions in different ways and based on their needs, preferences, cultural context, consumption experience and availability of resources. For studies on satisfaction to measure equity and equality in service provision, surveys include objective attributes (technical, commercial) of the service and some characteristics of family members (level of income, social and socioeconomic indicators) complementing quality assessment and facilitating an analysis on whether all segments of population have access to certain minimum quality levels. Therefore, one of the lessons of this literature as regards the object of the present study is that satisfaction measurement for utilities must have a different perspective and probably a greater scope or dimension than private services or goods. 21 Methodological approach to quality assessment Conceptual aspects arousing discussion at theoretical level on how to define quality perception having been detailed, it is necessary to move on methodological aspects that contribute to the measurement of such concepts. The survey Since no uniform procedures have been agreed upon to measure utilities quality level due to the latter’s heterogeneous, intangible and perishable nature, and whose production interacts with and may not be separated from their consumption, a customer satisfaction survey is usually used as an alternative. Customer satisfaction assessment for users of a certain utility is related both to assessing the result of the service itself –continuous, without interruptions, and uniform, without significant variations-, and to assessing the users-provider interactive relation, especially on periodic billing and collection, or as a consequence of company performance in the events of service interruptions and/or variations. For such reasons, a customer satisfaction survey requires defining the variables that turn the three abovementioned concepts operational: satisfaction, result and users-provider interactive relation. In the first place, only the result of a particular service may be objectively assessed using numerical scales: number, frequency and duration of service interruptions and/or variations. On the contrary, users-service provider interactive relation and customer satisfaction assessment require a subjective scale although numerically expressed. A 0-10 scale was used in this investigation, where 0 equals completely dissatisfied and 10 completely satisfied. Although including 0 in the scale would make it possible to regard it as, for example, a ratio scale and conveniently use it for regression adjustments, it is worth noting that there 22 is no uniform agreement in this regard. An alternative would be to use a 1-9 or 1-7 interval scale. The variables surveyed in this investigation –knowledge, attitudes and practices- are as follows: • Knowledge of the service provider’s name • Global satisfaction with the service • Continuous service or with interruptions • Uniform service or with significant variations • Number, average duration and frequency of interruptions and variations • Damages resulting from interruptions and variations • Service provider performance assessment in the events of interruptions and variations • User’s communication with service provider • Means of communication used • Reason for the communication • Assessment of customer service rendered by the service provider when contacted for a specific reason • Service provider announcements to users • Assessment of service provider announcements to users • Number of service units used per period of time • Rate: $ per service unit • Assessment of billing and collection methods used by the service provider • Knowledge of consumer protection associations and bureaus • Use of services provided by consumer protection associations and bureaus Socioeconomic, demographic and geographical variables were also surveyed and used as information classification criteria. 23 Therefore, the proposed approach focuses on detailing how the conducted surveys were based on the best practices for sample design and field methodologies, as well as on revealing the questioning structure used (see Annex 1 for detailed survey). For such reason, the purpose of this chapter is to offer –although implicitly- a series of recommendations on how to conduct this investigation from a statistical point of view. Sectors analyzed On this occasion, the water, gas and electric power sectors were analyzed as utility leading cases. More specifically, the services included are electric power distribution, drinking water distribution and sewage services and gas distribution. These three sectors account for a large part of family expenditure and, in this sense, their availability and quality have an important impact on the quality of life, particularly of the poorest sectors of population, which allocate a greater portion of their income to the payment of utilities. The lack of economic alternatives, i.e. service offer is limited to one provider per area, together with the essential nature of such services turn satisfaction with the service into a core element of users’ quality of life and give regulators a highly important role for improving access to and equity in utility provision. Although within the universe of utilities that might be analyzed under the present study, the telecommunications and transportation sectors were not assessed. In spite of their great importance, the fact that they are not household-connected services makes their inclusion incompatible with the methodological approach proposed through household surveys. Transportation and, to a lesser extent, mobile telephone services (cell phones) are provided at an individual level instead of a household level. In this sense, such services’ satisfaction and quality measurement determines that individual surveys be conducted in geographical scenarios other than households. Likewise, different members of the same family unit may have markedly different levels of satisfaction. Areas covered 24 The field work and the surveys included in this study were conducted in the Greater Area of Buenos Aires, comprised of the Autonomous City of Buenos Aires (CABA) and the 17 districts of the province of Buenos Aires (PBA). One of the aspects taken into account when choosing the area to be analyzed is related to the services under study and the existence of several companies providing each of such services. In other words, the services assessed in the abovementioned area are provided by the 5 companies detailed below, and the further one goes from the Greater Area of Buenos Aires, the more providers there are 8, which would hinder the conduction of this analysis. Within the analyzed area, there are currently two electric power distribution companies (Edenor S.A. and Edesur S.A.), two gas distribution companies (Metrogas S.A. and Gas Natural BAN S.A.) and one water distribution company (AySA S.A. 9). All of the following companies have different coverage areas, which expands the object analyzed in this study. Chart 1 Coverage Areas of the Utility Companies Assessed Gas Natural BAN Metrogas Edenor Edesur AySA Vicente López, San Isidro, Pilar, Malvinas Capital City Capital City, San Capital City, San Fernando, Tigre, San Argentinas, José C. (1), Avellaneda, Fernando, San Avellaneda, Martín, Tres de Febrero, Paz, San Miguel, Lanús, Lomas Martín, San Lanús, Morón, Hurlingham, Moreno, Gral. de Zamora, Isidro, Tigre, Quilmes, Ituzaingó, Moreno, Merlo, Rodríguez, Escobar, Quilmes, Vicente López, Berazategui, Marcos Paz, Gral. Las Heras, Tigre, San Almirante Hurlingham, Almirante San Miguel, José C. Paz, Fernando, San Brown, Ezeiza, Ituzaingó, La Brown, Malvinas Argentinas, Pilar, Isidro, Vicente, San Esteban Matanza, Morón, Florencio Gral. Rodríguez, Luján, Martín, Tres de Echeverría, Tres de Febrero, Varela, Mercedes, San Andrés de Febrero, Capital Florencio Almirante Brown, Esteban Giles, Suipacha, Carmen de City (*), Varela, Esteban Echeverría, Areco, La Matanza, Escobar, Hurlingham, Berazategui, Echeverría, Ezeiza, Campana, Exaltación de la Ituzaingó, Morón, Presidente Ezeiza, Lanús, Presidente Cruz, Zarate, San Antonio de Merlo, La Matanza, Perón, San Lomas de Perón and Areco and Capitán Marcos Paz and Vicente and Zamora, Quilmes San Vicente. Sarmiento. Gral. Las Heras. Cañuelas. and Avellaneda. 8 Cooperatives, mutual companies, municipalities, water distribution companies, partially state-owned companies, etcetera. 9 The Government constituted the company on March 21, 2006; 90% of shares belongs to the Federal State and the remaining 10% to its employees. Formerly, the company Aguas Argentinas S.A. was responsible for drinking water distribution and sewage services. 25 Source: Own elaboration based on data provided by utility companies. Observation: (1) The boundaries of the Capital City are as follows: Dock "D", unidentified street, lay out of the future Coastal Highway, extension Pueyrredón Avenue, Pueyrredón Avenue, Córdoba Avenue, Ferrocarril San Martín tracks, General San Martín Avenue, Zamudio, Tinogasta, General San Martín Avenue, General Paz Avenue and Río de La Plata. Associated answers Since customer satisfaction was sequentially investigated as regards three services in this survey, in order to prevent interviewees’ assessments from being associated with the order in which services were surveyed, random procedures were used in connection with the numbering of questionnaires so as to determine the order in which each of these three services were inquired about to every interviewee. Thus, service assessments were not associated with the order in which each service was tested as confirmed by the variance analysis test which led to the non-rejection of the hypothesis of overall satisfaction mean equality for the three services, according to the different survey orders, with the usual level in social sciences of α=0.05: Chart 2 Overall Satisfaction Means according to Survey Order Survey order Service p-value WEG WGE EWG EGW GWE GEW Water 7.98 7.93 7.78 7.66 7.44 7.58 0.309 Electric power 7.85 7.70 7.56 8.07 7.70 7.63 0.241 Gas 8.57 8.38 8.29 8.54 8.11 8.16 0.101 A similar problem was posed by the fact that the geographical coverage of this survey was shared by a drinking water distribution company (AYSA), two electric power distribution companies and two gas distribution companies, suggesting the possibility that the assessments of one service might be associated with those of another service. The situation is illustrated as follows: Diagram 1 Possible Survey Combinations Electric power Edenor Edesur Metrogas X X Gas Gas Natural Ban X Source: Own elaboration 26 As in the previous case, the variance analysis test for overall satisfaction led to the non- rejection of the hypotheses of overall satisfaction mean equality for the electric power and gas services, according to the different combinations of the companies providing such services, with the usual level in social sciences of α=0.05. On the other hand, the relatively high scoring of the drinking water distribution service in the area of Edenor Metrogas led to the rejection of the null hypothesis of mean equality: Chart 3 Overall Satisfaction Means according to Companies Electric power and gas distribution companies Service p-value Edenor Gas Natural Ban Edenor Metrogas Edesur Metrogas Water 7.53 8.15 7.51 0.035 Electric power 7.74 7.70 7.67 0.919 Gas 8.26 8.23 8.27 0.974 Source: Own elaboration Unit of Analysis Another issue to take into account before designing sample features is the unit of analysis to be considered. To begin with, the analysis of the level of satisfaction with the service should be carried out for each particular company. Having more than one company in the area under analysis, the average satisfaction level of the area would not make much sense since it is possible that the companies’ performance will differ greatly. In this sense, it is necessary for the sample to be representative 10 of every company providing services in each area, which would involve considering a necessary observation size of each company in order to infer its population behavior. Sample selection The size of the random sample selected was of n=500 households located in the Greater Area of Buenos Aires, comprised of the CABA 11 and the 17 districts 12 of the PBA13, which constitute the first ring of the metropolitan area of Buenos Aires 14: 10 Sample representativeness is achieved only through probability sampling methods, which imply that all population members have the same probability of being selected to be part of a sample and, consequently, all possible size samples have the same probability of being selected. 11 CABA stands for Autonomous City of Buenos Aires. 27 Chart 4 Population and Areas of the Districts of the PBA Area in Density District Population km2 inhabit/km2 Total 13,827,203 307,571 45.0 17 Districts of the Greater Buenos Aires 6,272,012 2,491 2,517.9 Avellaneda 328,980 55 5,981.5 Esteban Echeverría 243,974 120 2,033.1 General San Martín 403,107 56 7,198.3 Hurlingham 172,245 36 4,784.6 Ituzaingó 158,121 39 4,054.4 José C. Paz 230,208 50 4,604.2 La Matanza 1,255,288 323 3,886.3 Lanús 453,082 45 10,068.5 Lomas de Zamora 591,345 89 6,644.3 Morón 309,380 56 5,524.6 Quilmes 518,788 125 4 150.3 San Fernando 151,131 924 163.6 San Isidro 291,505 48 6,073.0 San Miguel 253,086 80 3,163.6 Tigre 301,223 360 836.7 Tres de Febrero 336,467 46 7,314.5 Vicente López 274,082 39 7,027.7 8 Districts of the second ring (1) 2,412,425 1,136 2,123.6 Rest of the Province of Buenos Aires (2) 5,142,766 303,944 16.9 City of Buenos Aires (3) 2,776,138 203 13,679.6 Source: INDEC and Military Geographical Institute [Instituto Geográfico Militar]. Observations: (1) A. Brown, Berazategui, Ezeiza, F. Varela, Gral. Sarmiento, Malvinas Argentinas, Merlo and Moreno. (2) The information pertains to the remaining 110 districts of the PBA. (3) The CABA has 21 School Districts 15, whose breakdown is excluded from the chart for understandability purposes. The proposed area includes 2,835,413 households to be surveyed according to the 2001 National Census of Population, Households and Dwellings (CNPHV) prepared by the INDEC. The CABA contains 1,024,540 households, while the number in the 117 districts 12 Districts are political areas used by the National Institute of Statistics and Censuses (INDEC) for conducting the 2001 National Census of Population, Households and Dwellings (CNPHV). The 17 districts surveyed group the 6065 census radii and each of them may contain between 103 and 1093 census radii. 13 PBA stands for Province of Buenos Aires. 14 The first ring of the metropolitan area is composed of the Districts that border on the City of Buenos Aires. The Province of Buenos Aires contains 135 districts; the 8 nearest ones represent the second ring of the metropolitan area and the remaining 110 districts, the rest of the Province of Buenos Aires. 15 School districts are geographical areas used by the INDEC to conduct the 2001 CNPHV. Such areas group the 3,406 census radii of the City of Buenos Aires, and each school district includes between 63 and 315 census radii. 28 assessed in the PBA amounts to 1,810,873. Thus, the sample size of n = 500 households to be surveyed guaranteed that under simple random sampling without replacement, the maximum range of the confidence interval of 95% for a population proportion is +/- 0.0438 16. In other words, once the estimate was complete, it could be affirmed with a 95% significance degree that the population mean was within the interval determined by the estimated mean of +/- 0.0438. On the other hand, the three main phases for sample selection are as follows: • Selection of a stratified sample of 100 Census Radii (CR), considering socioeconomic level; • Random selection of a Sample Point (SP) in each selected Census Radius (CR); • Selection of a random sample of 5 households in each selected SP. The CR are geographical areas defined by the INDEC in which 2001 CNPHV data was gathered. Such areas include about 300 households to be surveyed each. Particularly, the area in this study included 9,471 CR totaling 2,835,413 households to be surveyed. On the other hand, SP is the 9-block set constituted by a central block and 8 surrounding ones. Socioeconomic stratification of surveyed households Under the methodological section, it was concluded that household satisfaction and quality assessment of services depend greatly on the subjective perception of each household regarding service provision and that this is, in turn, a function of the socioeconomic stratum to which the household belongs. 17 In this sense, households were divided into stratums depending on their socioeconomic level using as proxy the 16 The range of the confidence interval is calculated considering that under simple random sampling without replacement, the sampling mean has a distribution that converges on a normal distribution focused on the population mean with variance (1-n/N)S2/n. In this way, the range of the confidence interval of 100(1-�)% for estimating population mean equals +/- e, where: e = z�/2 √[(1-n/N)S2/n]. It is known that in the case of population proportion: S2 = Nπ (1-π)/(N-1) (1). Considering in our case a level of confidence of 95%, this implies that zα/2 ≈ 1.96;; N = 2,835,413; n = 500; π = 0.50, π value that leads to the maximum of π (1-π); it results in e = 0.0438. 17 The households pertaining to the lowest socioeconomic stratums probably have more tolerance for lower levels of service provision. This indicates that the parameters used for assessing quality and satisfaction differ among the different socioeconomic stratums. 29 distribution of households with material privation elaborated under the 2001 CNPHV for the CABA and the 17 districts of the PBA (GBA). Material privation of households is measured taking into account not only the lack of common resources but also financial privation. 18 The following graphic shows the distribution of the households to be surveyed in the CABA and the 17 districts selected in the Province of Buenos Aires, according to the percentage of households with material privation within the ranges of the CR detailed for each of the two areas. Graphic 1: Hogares con Privación Material En % del Radio Censal 50% Ciudad de Buenos Aires 40% 17 Partidos del conurbano 30% Bonaerense 20% 10% 0% 0 ≤5 5 ≤ 10 10 15 ≤ 15 20 ≤ 20 25 ≤ 25 30 ≤ 30 35 ≤ 35 40 ≤ 40 45 ≤ 45 50 ≤ 50 55 ≤ 55 60 ≤ 60 ≤ 6565 ≤ 7070 7575 ≤ ≤ 8080 ≤ 8585 ≤ a 9090 95 a 100 ≤ 95 Promedio Fuente: Elaboración propia en base a datos del INDEC. The results indicate that the above distributions differ significantly between the two geographical areas analyzed. In the first place, the average material privation of the CR pertaining to the CABA (14%) is significantly lower than that of the 17 districts of the metropolitan area of the Province of Buenos Aires (36%). In turn, households with material privation have a greater concentration in the CABA as compared with the 17 districts of the PBA. This could be an indicator of a higher level of socioeconomic inequality among the different CR in the CABA in comparison with the analyzed districts of the PBA. All things considered, the results of divergence at socioeconomic level support the idea of considering both geographical areas at an individual level for measuring service quality and household satisfaction. 18 The former may vary in the short term and is connected with economy fluctuations, while the latter affects households in a more stable manner (level of savings) and, due to its persistent nature, is regarded structural or chronic. 30 The chart below contains an itemized list of the location of households with material privation in the 21 school districts of the CABA for the purpose of identifying the geographical location of those households with material and financial deficiencies within such geographical area. The following chart includes information on the cumulative percentages of households and the cumulative percentages of households with material privation, from lower to higher according to the percentage of households with material privation within each school district. Thus, it is possible to divide school districts into 3 socioeconomic stratums containing an approximately equal number of households to be surveyed. This allows for the identification of those school districts and stratums with a percentage of households with material privation lower, equal or higher not only than the average of each stratum but also than the average of the CABA. Chart 5 Households with Material Privation per School District of the CABA Households with Average of School Cumulative Cumulative Households Stratum Privation / Households with District Households with Privation Households Privation 10 9% 8% 5% 8 9% 13% 9% I 7 10% 19% 13% 9% 9 10% 28% 20% 17 10% 33% 23% 16 10% 35% 25% 15 10% 39% 27% 2 11% 49% 35% 18 12% 52% 38% II 12% 12 12% 56% 42% 11 12% 59% 45% 14% 6 12% 65% 50% 14 13% 68% 53% 1 14% 78% 62% 3 14% 83% 68% 13 18% 86% 72% 4 19% 89% 76% III 20% 20 22% 92% 81% 5 27% 95% 87% 21 32% 97% 92% 19 40% 100% 100% Source: Own elaboration based on data provided by the INDEC From the data shown in chart 5, it might be inferred that there is a strong divergence as to material privation of households among the school districts of the CABA. While Stratum 31 I is characterized by containing the districts located in the northern area of the CABA and has, on average, approximately half the number of households with privation in comparison with Stratum III, this is composed of the districts with greater economic deficiencies located in the southern area. Likewise, the information contained in graphic 2 indicates that a great part of the total number of households with material privation is concentrated in a small number of households. It is noticeable that 20% of the households in the poorest district represents 40% of the total number of households with privation. In this sense, the heterogeneity of CABA’s socioeconomic structure was taken into account when creating a sample that would represent and infer through surveys the behavior of all of its population as regards their satisfaction with service quality. Graphic 2 Households with Material Privation in the CABA Gini = 0,21 100% 80% % Cu HWP 60% 40% 20% 0% 0% 20% 40% 60% 80% 100% % Cu HOU Source: Own elaboration based on data provided by the INDEC. Observation: Cu HOU refers to cumulative households and Cu HWP to cumulative households with material privation. To continue with the same analysis, chapter 6 contains the cumulative percentages of households and the cumulative percentages of households with material privation pertaining to the 17 districts surveyed in the metropolitan area of the PBA, from lower to higher according to the percentage of households with privation in each district. In addition, this data allowed for the division of the districts into 3 socioeconomic stratums 32 with an approximately equal number of households to be surveyed. Likewise, it is also possible to identify those districts and stratums with a percentage of households with material privation lower, equal or higher not only than the average of each stratum but also than the average of the 17 districts taken into account. Chart 6 Households with Material Privation per District of the PBA Households with Cumulative Average of Cumulative Stratum District Privation / Households with Households with Households Households Privation Privation Vicente López 13% 5% 2% San Isidro 18% 10% 4% Morón 23% 15% 8% IV 23% Tres de Febrero 24% 21% 12% Avellaneda 28% 26% 16% Gral. San Martín 30% 33% 21% Ituzaingo 31% 35% 23% Lanus 31% 43% 30% Hurlingham 33% 45% 32% 36% V 36% San Fernando 34% 48% 35% Quilmas 40% 56% 44% Lomas de Zamora 40% 65% 54% Tigre 43% 69% 59% Almirante Brown 46% 76% 68% VI La Matanza 47% 95% 93% 47% Esteban Echeverría 49% 98% 97% Ezeiza 57% 100% 100% Source: Own elaboration based on data provided by the INDEC. In the stratification of material privation of households per district, it is also possible to infer that the northern districts have less privation than those of the southern area. This heterogeneity in the socioeconomic structure is evidenced by the fact that the districts of Stratum IV have, on average, half the number of households with material privation as compared with the districts of Stratum VI. Likewise, graph 3 shows that the concentration of households with material privation in the districts of the PBA is lower than that of the school districts of the CABA, since in the former 20% of the poorest households represent 30% of the total number of households with material privation in all 17 districts. In this regard, it is necessary to consider the existing socioeconomic divergence among the districts analyzed in the metropolitan area of the Province of 33 Buenos Aires in order to create a representative sample of customer satisfaction and service quality. Graph 3 Households with Material Privation in the PBA Gini = 0,17 100% 80% % Cu HWP 60% 40% 20% 0% 0% 20% 40% 60% 80% 100% % Cu HOU Source: Own elaboration Source: Own elaboration based on data provided by the INDEC. Observation: Cu HOU refers to cumulative households and Cu HWP to cumulative households with material privation. Six stratums were obtained as a result of the classification of the school districts of the CABA and districts of the metropolitan area of the Province of Buenos Aires per socioeconomic stratums, and their main characteristics are detailed below. Chart 7 contains not only the allocation of the sample of 100 CR to classified stratums but also a breakdown of the number of households to be surveyed in each stratum. The methodology for allocating the sample (CR and Households) was applied proportionally to the product of the size of the stratum (Number of Households) according to the standard deviation 19 of each stratum. Chart 7 Size of the CR and Households surveyed per Stratum 19 The standard deviation estimate is contained in formula (1) of observation 9. 34 Population Sample Households Stratum Census Number of with Standard Census Number of Radius Households Privation / Deviation Radius Households Households I CABA 1,068 334,094 9% 31% 7 35 II CABA 1,222 363,986 12% 34% 9 45 III CABA 1,116 326,460 20% 45% 10 50 IV PBA 2,023 595,625 23% 48% 20 100 V PBA 1,975 578,011 36% 60% 24 120 VI PBA 2,067 637,237 47% 68% 30 150 Total 9,471 2,835,413 28% 100 500 Source: Own elaboration based on data provided by the INDEC. Observation: CABA refers to CABA and PBA to 17 districts of the PBA. The field study of the 17 surveyed districts of the PBA contains more CR and, in turn, more households than the school districts of the CABA. Likewise, the stratums with higher and lower privation of the PBA are more than twice the value of the pertaining stratums in the CABA. On the other hand, standard deviations of the better positioned stratums, in terms of socioeconomic level, of both geographical areas are lower than those of the stratums with greater privation. The procedure followed for allocating sample size to each stratum resulted not only in a greater number of surveys in the PBA than in the CABA but also in a greater number of surveys in the stratums of the PBA and the CABA with lower socioeconomic development. Selection of Census Radii and Sample Points The systematic selection of the CR pertaining to each stratum addressed in the field study was made through two random starts and probabilities proportional to population size, starting from the CR list arranged from lower to higher value according to the proportion of households with material privation. In accordance with the addressed theoretical guidelines, it could be affirmed that the procedures established for CR selection led to estimates with less variability than that which would result from a simple random sample. 20 Once CR were selected, it was possible to choose at random one SP for each of them. As explained above, a SP is a set of 9 blocks constituted by a main block and the 8 20 A simple random sample establishes that each population component is selected at random having the same probabilities of being selected. 35 surrounding it. The following diagram shows the way in which a SP is expressed. The indicated numbers determine the order to be followed by interviewers to go through the SP, in order to locate the potential households to be interviewed. The points and arrows shown in each block indicate the starting corner and the traveling direction to be followed by each interviewer in each block. Diagram 2 Sample Point and Interviewer’s Route 9 2 3 8 1 4 7 6 5 To continue with the analysis, the main block of the SP was selected choosing at random a number of blocks to constitute the selected CR that were previously numbered correlatively. In this regard, the plan of the Sample Point that was part of the elements provided to the interviewer was designed based on such main block. After selecting and preparing the SP, a random sample had to be obtained from 5 households, included in each point. For selecting the interviewed households, there were a series of procedures to be followed in the field under the exclusive responsibility of the interviewer. This specially designed methodology allowed for a combination of the list with the random selection of households and dwellings. Additionally, the established procedures facilitated the supervision of interviewers’ activities. In this regard, Annex 1 includes a flow chart that helps visualizing these procedures and actions to be followed by each interviewer and also a brief explanation on the procedures used in the field study. 36 Empirical approach: Results Determinants of Overall Satisfaction with the Service This section contains a first analysis of determinants of customer satisfaction in general as well as with the specific aspects of each of the services included in the sample. 1. Water sector Satisfaction and Technical Service Level In the first place, an analysis is conducted on the extent to which service continuity affects customers’ general degree of satisfaction. For this purpose, we compared the relative distribution of the overall degree of satisfaction of the customers who reported having a continuous service as against those who alleged having experienced a service with continuity issues. The results are shown in Graphic 4. Graphic 4 Overall Satisfaction and Service Continuity 37 Survey Flag_Water_ServiceWith Water (All) Count of Questionnaire 35,00% 30,00% 25,00% 20,00% PA03a_Gral_Servic_continuous? With interruptions Continuous 15,00% 10,00% 5,00% 0,00% 0 1 2 3 4 5 6 7 8 9 10 PA02_Gral_Ov_Satisf Source: Own elaboration From the analysis of the graphic, it can be affirmed that service continuity has a significant impact on water customers’ overall satisfaction. Customers whose service is discontinuous (light blue bars of the graphic) show a 5.5 average degree of satisfaction while those with a continuous service present an 8.1 value. The relative importance of service continuity on overall satisfaction is further shown by the comparison of the end values of the scale. The probability of being completely dissatisfied with the service is 11 times higher for users affected by service interruptions than for those with a continuous service. At the other end, the probability of being completely satisfied is 8.5 times higher for those users who did not inform any issues regarding service continuity. Due to the importance of service continuity in overall satisfaction, it may be analyzed whether there is a “tolerable� number of service interruptions for users of the water provision service. To such effect, we can compare the answer on continuous or discontinuous service (P03a - dichotomic) to the number of service interruptions during 38 the last six months reported by the users in the survey. The distribution is shown in Graphic 5. Graphic 5 Service Continuity and Number of Interruptions Survey Flag_Water_ServiceWith Water (All) Count of Questionnaire 100% 90% 80% 70% 60% PA03a_Gral_Servic_continuous? 50% With interruptions Continuous 40% 30% 20% 10% 0% 0 1 2 3 4 5 6 7 10 12 15 20 30 40 60 100 200 PA04_Interr_q_Interruptions Source: Own elaboration The dark bar indicates the number of interruptions during the last semester reported by users who claimed to have a continuous service. As expected, the great majority did not experience service interruptions during such period. It is worth noting, however, that a significant number of users who experienced one interruption (over 5%) also declared having a continuous service. On the other hand, the light blue bar shows the distribution of the number of interruptions experienced by users who regarded the service as discontinuous. The graphic shows that there is a significant number of users (7.8% of the total) who informed there were no service interruptions but still considered the service to be discontinuous. It is also worth mentioning that the greatest concentration is between two and three interruptions. 39 There may be two reasons for such concentration. First, it could be that this is actually the real number of interruptions evidencing the greatest probability of occurrence in the service. Alternatively, users who experience one or two interruptions may not regard such situation as discontinuous service. In other words, there may be a certain degree of tolerance due to which one or two interruptions do not affect customer satisfaction with service quality. The same analysis developed so far for service continuity may be applied to pressure, which is the other parameter defining technical service quality. Thus, we analyzed the extent to which pressure impacts on customers’ overall degree of satisfaction. For this purpose, we compared the relative distribution of the general degree of satisfaction of the customers who reported having good pressure as against those who alleged having experienced a service with pressure issues. The results are shown in Graphic 6. Graphic 6 Overall Satisfaction and Pressure Flag_Water_ServiceWith waterSurvey (All) Count of Questionnaire 35,00% 30,00% 25,00% 20,00% PA03b_Gral_Adequate_Pressure? With adequate pressure With inadequate pressure 15,00% 10,00% 5,00% 0,00% 0 1 2 3 4 5 6 7 8 9 10 PA02_Gral_Ov_Satisf Source: Own elaboration 40 From the analysis of the graphic, it can be affirmed that adequate pressure has a significant impact on water customers’ overall satisfaction. Customers with low-pressure water service (dark bars of the graphic) show a 5.4 average degree of satisfaction while those with adequate pressure present an 8.5 value. As it was the case with service continuity, the importance of water pressure on overall satisfaction is further shown by the comparison of the end values of the scale. The probability of being completely dissatisfied with the service is 56 times higher for users who experienced low-pressure than for those with adequate pressure. At the other end, the probability of being completely satisfied is 10.2 times higher for those users who did not inform any issues regarding water pressure. As we did in the case of service continuity, it may be analyzed whether there is a “tolerable� number of low-pressure events for users of the water service. To such effect, we can compare the answer about adequate or inadequate pressure service (P03a - dichotomic) to the number of low-pressure events during the last six months reported by the users in the survey. The distribution is shown in Graphic 7. Graphic 7 Adequate Pressure and Low-Pressure Events 41 Survey Flag_Water_ServiceWith Water (All) Count of Questionnaire 100,00% 90,00% 80,00% 70,00% 60,00% PA03b_Gral_Adequate_Pressure? 50,00% With adequate pressure With inadequate pressure 40,00% 30,00% 20,00% 10,00% 0,00% 0 1 2 3 4 5 6 7 8 10 12 14 15 18 20 24 25 30 40 48 50 60 80 90 99 100120150160180 PA06_Interr_q_Low_Pressure_Last_6_Months Source: Own elaboration The light blue bar indicates the number of low-pressure events during the last semester reported by users who claimed to have an adequate water pressure service. As expected, the great majority did not experience low pressure during such period and, unlike the previous case, the number of users who experienced one event and declared having an adequate pressure level is small (0.9%). On the other hand, the dark bar shows the distribution of the number of low-pressure events experienced by users who regarded the service as having low water pressure. The graphic shows that there is a significant number of users (8.3% of the total) who informed there were no low-pressure events but still considered the service to have inadequate water pressure. It is also worth mentioning that the greatest concentration is between two and six events. As in the service continuity case, there may be two reasons for such concentration. First, it could be that this is actually the real number of low-pressure events evidencing the greatest probability of occurrence in the service. Alternatively, users who experience 42 between one and six low-pressure events may not regard such situation as inadequate water pressure service. Determining the most appropriate explanation for the two characteristics of technical service insofar analyzed – continuity and pressure – is essential from the regulatory point of view. Therefore, and in order to determine which one is the valid explanation, it is necessary to supplement the survey with effective information on service interruptions and low-pressure events experienced in the areas of the customers included in the sample. Such information, which should at first be available in the utility company, would allow for a comparison of customer “perception� with the “effective� quality of the service provided. Overall Satisfaction and Technical Product Quality This section contains an analysis of the impact of the different elements contributing to technical product quality – i.e. water characteristics as to color, taste and smell – on the degree of overall satisfaction with the service. Therefore, in the first place, we compared the degree of satisfaction. For this purpose, we compared the overall degree of satisfaction of the customers who did not experience any problems regarding water quality as against those who had problems with at least one of the three dimensions considered in the survey (color, taste and smell). The relative distribution of each customer group is shown in Graphic 8. Graphic 8 Overall Satisfaction and Water Quality 43 40,0% 35,0% 30,0% 25,0% At least 1 20,0% None 15,0% 10,0% 5,0% 0,0% 0 1 2 3 4 5 6 7 8 9 10 Source: Own elaboration From the analysis of the graphic, it can be affirmed that water quality has a significant impact on water customers’ overall satisfaction. Customers who experienced water quality issues regarding at least one of the three surveyed parameters (light blue bars of the graphic) show a 7.1 average degree of satisfaction while those who experienced no issues whatsoever present an 8.4 value. As in the technical service case analyzed above, the importance of technical product quality on overall satisfaction is further shown by the comparison of the end values of the scale. The probability of being completely dissatisfied with the service is 3.5 times higher for customers who experienced water quality issues than for those with a satisfactory water quality. At the other end, the probability of being completely satisfied is 1.9 times higher for those customers who did not inform any issues regarding water quality. It is important to note that the analysis indicates that customers give more importance to technical service quality (continuity and pressure) than to technical product quality (observable water physical characteristics). The second phase consists in determining whether there are any differences in the impact of each of the three characteristics taken into account in the survey on customers’ level of satisfaction. Therefore, we compared the degree of overall satisfaction of those customers 44 who experienced problems with each of these dimensions: color, smell and taste. The comparative analysis is shown in Graphic 9. Graphic 9 Overall Satisfaction and Water Characteristics 30% 25% 20% Color 15% Smell Taste 10% 5% 0% 0 1 2 3 4 5 6 7 8 9 10 Source: Own elaboration To begin with, it is worth noting that having issues with only one of the attributes does not result in complete (nor unitary) dissatisfaction in any of the sample cases. From all users who experienced problems with only one of the attributes, those who had water color issues have lesser overall satisfaction. At the other end, water taste issues seem to have less effect on customer satisfaction. When considering mean satisfaction, the values are quite similar for users with only color or smell problems (7.55 and 7.56, respectively) while satisfaction of customers who only experienced taste problems is slightly higher (8.32). The following phase consists in verifying how overall satisfaction with the service affects the simultaneous occurrence of two or more of such issues. The different possible 45 combinations and degree of overall satisfaction associated to each of them are shown in Graphic 10. Graphic 10 Overall Satisfaction and Product Quality Simultaneous Issues 40% 35% 0 30% 1 2 25% 3 4 20% 5 6 15% 7 8 10% 9 10 5% 0% 3 Issues 2 Issues 1 Issue None Source: Own elaboration As expected, the graphic clearly shows that low overall satisfaction cases increase consistently with the accumulation of issues regarding the quality of water obtained. Indeed, overall mean satisfaction goes from 8.4 in customers who did not experience any issues to 7.7 for those who only had problems with one of the measured attributes. When there are problems with two or three attributes, the mean falls to 7.3 and 6.1, respectively. It is also worth highlighting that the existence of completely dissatisfied customers (0 or 1 in the scale used) results from the coexistence of problems with at least two water attributes. 46 Conclusions for the Water Sector It may be concluded from the analysis above that, as a general rule, customers consider service quality dimensions (continuity and pressure) to be more important than product quality dimensions (physical characteristics of water). There also seems to be a minimum acceptable level of interruptions and low pressure events which would be tolerable by customers to the extent that they do not seem to affect their overall level of satisfaction with the service provided. In order to confirm this hypothesis, the analysis conducted must be deepened, on the basis of the information about interruptions and actual low pressure events which took place at each of the areas where the data have been compiled. Among the features of the product received, taste seems to be the least important factor, while smell and color show similar values as to their effects on the overall satisfaction of customers. B. Electric Power Sector Satisfaction and Technical Service Level In the first place, an analysis is conducted on the extent to which service continuity affects customers’ general degree of satisfaction. For this purpose, we compared the relative distribution of the overall degree of satisfaction of the customers who reported having a continuous service as against those who alleged having experienced a service with continuity issues. The results are shown in Graphic 11. Graphic 11 Overall Satisfaction and Service Continuity 47 Flag_Service_ELEC With ELEC Survey (All) Count of Questionnaire 30,00% 25,00% 20,00% PE03a_Gral_Servic_continuous? 15,00% With outages Continuous 10,00% 5,00% 0,00% 0 1 2 3 4 5 6 7 8 9 10 PE02_Gral_Satisf_Gral Source: Own elaboration On the basis of the information disclosed in graphic 11 above, it may be observed that service continuity causes a significant impact on electric power customers’ general satisfaction. Customers whose service is discontinuous (light blue bars of the graphic) show a 6.6 average degree of satisfaction while those with a continuous service present an 8.3 value. The relative importance of service continuity on overall satisfaction is further shown by the comparison of the end values of the scale. The probability of being completely dissatisfied with the service is 5.7 times higher for customers affected by outages than for those with a continuous service. At the other end, the probability of being completely satisfied is 4.4 times higher for those users who did not inform any issues regarding service continuity. Due to the importance of service continuity in overall satisfaction, it may be analyzed whether there is a “tolerable� number of outages for electric power customers. To such effect, we can compare the answer on continuous or discontinuous service (P03a - dichotomic) to the number of outages during the last six months reported by the customers in the survey. The distribution is shown in Graphic 12. 48 Graphic 12 Service Continuity and Number of Outages Flag_Service_ELEC Con ELEC Survey (All) Count of Questionnaire 70,00% 60,00% 50,00% 40,00% PE03a_Gral_Servic_continuous? With outages Continuous 30,00% 20,00% 10,00% 0,00% 0 1 2 3 4 5 6 7 8 10 12 15 17 20 24 25 27 28 30 50 99 180 PE04_Interr_q_Interr Source: Own elaboration The dark bar indicates the number of outages during the last semester reported by customers who claimed to have a continuous service. As expected, the great majority did not experience outages during such period. It is worth noting, however, that a significant number of customers who experienced one (12%) or two (10%) outages also declared having a continuous service. On the other hand, the light blue bar shows the distribution of the number of outages experienced by customers who regarded the service as discontinuous. The graphic shows that, unlike the water sector, the number of customers who declared that there were no outages but still considered the service to be discontinuous is insignificant (0.6%). The greatest concentration of customers who declare the service to be discontinuous is between two and three outages per semester. In fact, the number of customers that classify the service as continuous with only one outage is almost the same as the one that considers it to be discontinuous. 49 As discussed in the previous section in relation to the water sector, there may be two reasons for such concentration. First, it could be that this is actually the real number of outages evidencing the greatest probability of occurrence in the service. Alternatively, customers who experience one or two outages may not regard such situation as discontinuous service. In other words, there may be a certain degree of tolerance due to which one or two outages do not affect customer satisfaction with service quality. The same analysis developed so far for service continuity may be applied to voltage level, which is the other parameter defining technical service quality. Thus, we analyzed the extent to which voltage impacts on customers’ overall degree of satisfaction. For this purpose, we compared the relative distribution of the general degree of satisfaction of the customers who reported having a good voltage level as against those who alleged having experienced a service with voltage issues. The results are shown in Graphic 13. Graphic 13 Overall Satisfaction and Voltage Level Flag_Service_ELEC With ELEC Survey (All) Count of Questionnaire 30,00% 25,00% 20,00% PE03b_Gral_Volt_Appropriat? 15,00% With adequate voltage With inadequate voltage 10,00% 5,00% 0,00% 0 1 2 3 4 5 6 7 8 9 10 PE02_Gral_Satisf_Gral Source: Own elaboration The adequate voltage level has a significant impact on electric power customers’ overall satisfaction. Customers with low-voltage problems in the service (dark bars of the 50 graphic) show a 6.2 average degree of satisfaction while those with no voltage problems in the service present an 8.2 value. As it was the case with service continuity, the importance of voltage on overall satisfaction is further shown by the comparison of the end values of the scale. The probability of being completely dissatisfied with the service is 11.7 times higher for customers who experienced low-voltage than for those with adequate voltage level. At the other end, the probability of being completely satisfied is 7.5 times higher for those customers who did not inform of any issues regarding voltage level. As we did in the case of service continuity, it may be analyzed whether there is a “tolerable� number of voltage level variation events for customers of the electric power service. To such effect, we can compare the answer about adequate or inadequate voltage service (P03a - dichotomic) to the number of voltage variation events during the last six months reported by the users in the survey. The distribution is shown in Graphic 14. Graphic 14 Adequate Voltage and Voltage Level Variation Events Flag_Service_ELEC With ELECSurvey (All) Count of Questionnaire 100,00% 90,00% 80,00% 70,00% 60,00% PE03b_Gral_Volt_Adequate? 50,00% With adequate voltage With inadequate voltage 40,00% 30,00% 20,00% 10,00% 0,00% 0 1 2 3 4 5 6 7 8 10 12 14 15 18 20 24 25 30 40 48 50 60 62 90 99 100 120 150 180 700 PE06_Interr_q_Fluct_Volt Source: Own elaboration 51 The light blue bar indicates the number of voltage level variation events during the last semester reported by customers who claimed to have an adequate voltage level service. As expected, the great majority did not experience voltage problems during such period and, unlike the previous case, the number of customers who experienced one event and declared having an adequate voltage level is relatively small (2%). On the other hand, the dark bar shows the distribution of the number of voltage level variation events experienced by customers who regarded the service as having inadequate voltage level. The graphic shows that there is a significant number of customers (2.9% of the total) who informed there were no voltage variation events but still considered the service to have inadequate voltage level. It is also worth mentioning that the greatest concentration is between two and six events. As in the service continuity case, there may be two reasons for such concentration. First, it could be that this is actually the real number of low voltage events evidencing the greatest probability of occurrence in the service. Alternatively, customers who experience between one and two low-voltage events may not regard such situation as inadequate voltage level service. Determining the most appropriate explanation for the two characteristics of technical service insofar analyzed –continuity and voltage variation– is essential from the regulatory point of view. Therefore, and in order to determine which one is the valid explanation, it is necessary to supplement the survey with effective information on outages and voltage variation events per customer of the customers included in the sample. This information, which is available in the utility company because it is so required in the regulatory framework, would allow for a comparison of customer “perception� and/or demand with the “effective� quality of the service provided and/or supply. C. Gas Sector Satisfaction and Technical Service Level 52 In the first place, an analysis is conducted on the extent to which technical service quality impacts on the degree of overall customer satisfaction with the gas service. Given the technological particularities of this sector –in which service interruptions are almost non- existent – the only dimension analyzed is pressure. For this purpose, we compared the relative distribution of the overall degree of satisfaction of the customers who reported having a service with adequate pressure as against those who alleged having experienced a service with pressure issues. The results are shown in Graphic 15. Graphic 15 Overall Satisfaction and Technical Service Quality (Pressure) Flag_Service_Gas With Gas Survey (All) Count of Questionnaire 40,00% 35,00% 30,00% 25,00% PG03_Gral_Servic_Pressure_Adequate? 20,00% Adequate Pressure Inadequate Pressure 15,00% 10,00% 5,00% 0,00% 0 1 2 3 4 5 6 7 8 9 10 PG02_Gral_Satisf_Gral Source: Own elaboration From the analysis of the graphic, it can be affirmed that service pressure has a significant impact on the overall satisfaction of gas sector customers. Customers with pressure problems in the gas service (dark bars of the graphic) show a 6.4 average degree of satisfaction while those with adequate pressure present an 8.7 value. 53 The relative importance of service pressure on overall satisfaction is further shown by the comparison of the end values of the scale. The probability of showing a low satisfaction21 with the service is 10.6 times higher for customers affected by pressure problems than for those with adequate pressure. At the other end, the probability of being completely satisfied is 11.8 times higher for those customers who did not inform of any issues in the service provision. Due to the importance of service pressure in overall satisfaction, it may be analyzed whether there is a “tolerable� number of low pressure events for users of the gas service. To such effect, we can compare the answer on adequate or inadequate service pressure (P03a - dichotomic) to the number of low pressure events during the last six months reported by the customers in the survey. The distribution is shown in Graphic 16. Graphic 16 Adequate Pressure and Number of Low Pressure Events Flag_Servicio_Gas Con Gas Encuesta (All) Count of Cuestionario 20,00% 18,00% 16,00% 14,00% 12,00% PG03_Gral_Servic_Presion_Adecuada? 10,00% Presión adecuada Presión inadecuada 8,00% 6,00% 4,00% 2,00% 0,00% 0 1 2 3 4 5 6 7 8 10 12 15 18 20 24 25 30 40 48 50 60 80 90 91 99 100 120 150 180 PG04_Baja_Presion_q_Ult_6_Meses Source: Own elaboration Este cuadro no era editable. Éstas son las traducciones que van a necesitar para reproducirlo en inglés: 21 In this case the comparison is made on the basis of customers stating a 1 satisfaction, since there are no cero-satisfaction cases and service with adequate pressure. 54 Flag_Service_Gas[With Gas]Survey(All) Count of questionnaire PG03 _ Gral _ Service _ Adequate _ Pressure? [celeste] Adequate Pressure [color oscuro] Inadequate Pressure PG04 _ Low _ Pressure _ q _ last_ 6 _ Months The light blue bar indicates the number of low pressure events during the last semester reported by customers who claimed to have an adequate service. As expected, the great majority did not experience low pressure problems during such period. The dark bar, in turn, shows the distribution of the number of low pressure events experienced by customers who regarded the service as inadequate. The graphic shows that there is a significant number of customers (3.7%) who informed there were no low pressure problems but still considered the service pressure to be inadequate. One of the highest concentrations of customers defining the service as discontinuous took place upon occurrence of the three low pressure events in that six-month period. Unlike other distribution services, this case shows other important peaks at the level of 10 and 30 low pressure events in the semester (11.1% and 6.7%, respectively). Quality Perception and Characteristics of Customers The purpose of this section is to assess whether different characteristics gathered from the households or the individuals answering the survey affect quality perception. For that purpose, the socioeconomic variables included in the questionnaire are used, on the one hand. On the other hand, the knowledge that customers have –or not- about the service provider is used as a proxy of their degree of information and knowledge of the sector. Therefore, whether the degree of satisfaction varies from those customers who know which the providing company is with respect to those who do not know it is first assessed. In principle, it may be expected that the degree of information that an individual 55 has about the service may affect their demand for quality, and therefore, their satisfaction with the service provided. In the second place, this section assesses whether the socioeconomic index, either with or without the customer income scale, the income scale and the disaggregated socioeconomic variables impact on the degree of overall customer satisfaction. 22 A priori, it is expected that the income level may affect both supply and demand for quality. With respect to supply, it is reasonable to expect that low-income areas show higher infrastructure deficits adversely affecting service quality. 23 As regards demand, it is reasonable to expect that low-income customers are less sensitive to failures in the quality of the service rendered. 24 Consequently, the result on quality perception would then be uncertain a priori. It may be concluded from the analysis of the survey that this group of customers generally shows higher satisfaction levels than customers from a higher socioeconomic level. Therefore, a lower demand for quality seems to prevail. In order to simplify exposure, the results arising from the estimates have been structured by sectors and regressions, in accordance with the description of subjects set out above. All the regressions have been analyzed in this study. The idea is to select the most robust model for future versions, setting aside non-significant variables, and verifying the structure of the model remainders. a) Water Regressions 1 and 4 estimate whether there are any differences in the degree of overall satisfaction between customers who know the water company and customers who don’t. The results obtained from regressions 1.1, 1.2 and 4.2 do not enable to affirm that the fact of knowing information about the water company is a significant variable to explain the 22 The socioeconomic index aggregates the set of socioeconomic variables plus the income scale. 23 This entails assuming that low socioeconomic level customers are located in low socioeconomic level areas, which is not necessarily the case in this sample. 24 This, in part, could be the consequence of a smaller stock of electrical and gas devices as well as of the lack of some devices which are highly sensitive to quality, as it is the case of computers. 56 behavior of overall customer satisfaction with the service. However, if we assess its significance with the perception variables for the first wave of the survey (regression 4.1), it is concluded that those customers that can identify the providing company show, in average, 0.40 more satisfaction points. st a = α + β1 ∗ rsni + β 2 ∗ iii + β 3 ∗ adii + β 4 ∗ fat i + β 5 ∗ dcf i + β 6 ∗ fli + β 7 ∗ sai + µ st a = α + β1 ∗ rsni + β 2 ∗ iii + β 3 ∗ adii + β 4 ∗ fat i + β 5 ∗ dcf i + β 6 ∗ fli + β 7 ∗ isi + µ st a = α + β1 ∗ rsni + β 2 ∗ iii + β 3 ∗ adii + β 4 ∗ fat i + β 5 ∗ dcf i + β 6 ∗ fli + β 7 ∗ sai + β 8 ∗ isi + µ sta: overall satisfaction with water service (PA02) rsni: satisfaction with ordinary service restoration (PA09) iii : satisfaction with announcements of service interruptions or low gas pressure events (PA09) adii : satisfaction with company’s response during service interruptions or low gas pressure events (PA09) fati : satisfaction with punctuality of bill (PA21) dcfi : satisfaction with itemized billing (PA21) fli: satisfaction with easy-to understand bills (PA21) sai: knows water providing company (PA01) isi: socioeconomic index Chart 8 Overall Satisfaction versus Partial Indexes, Knowledge of Provider and Socioeconomic Index 57 Source: Own elaboration. Note: Customer knowledge means whether the customer is aware of which supplying company they have. Furthermore, regressions 2 and 5 estimate if the socioeconomic level of customers is related to the degree of overall satisfaction with the water service. In turn, regressions 3 and 6 estimate whether the socioeconomic level of customers is significant to explain the overall satisfaction level, including if the customer knows the water company. The majority of the cases analyzed indicate that it cannot be asserted that the socioeconomic level of customers influences their degree of satisfaction with the service rendered. However, some results arising from the second phase of the survey (regressions 2.2 and 3.2) show that a variation in one unit in the socioeconomic level generates an average increase of 1/4 point (0.26 and 0.27) in overall satisfaction. While, as it has been mentioned above for the first phase of the survey, the fact of knowing the utility company causes an increase in satisfaction equivalent to 0.43 (regression 6.1). st a = α + β 1 ∗ sac i + β 2 ∗ paa i + β 3 ∗ ia i + β 4 ∗ mo i + β 5 ∗ mc i + β 6 ∗ ms i + β 7 ∗ sa i + µ st a = α + β 1 ∗ sac i + β 2 ∗ paa i + β 3 ∗ ia i + β 4 ∗ mo i + β 5 ∗ mc i + β 6 ∗ ms i + β 7 ∗ is i + µ st a = α + β 1 ∗ sac i + β 2 ∗ paa i + β 3 ∗ ia i + β 4 ∗ moi + β 5 ∗ mc i + β 6 ∗ ms i + β 7 ∗ sa i + β 8 ∗ is i + µ i: 1,2,3….500 58 Where sta: overall satisfaction with water service (PA02) saci: uninterrupted water service (PA03) check question. paai : adequate pressure of water (PA03) idem iai : water interruptions (PA04) moi : bad smelling water (PA10) mci : bad color of water (PA10) msi : bad tasting water (PA10) sai: knows water company (PA01) isi: socioeconomic index Chart 9 Overall Satisfaction versus Service Quality and Technical Product, Knowledge of Provider and Socioeconomic Index Source: Own elaboration In turn, regressions 7 and 9 estimate whether the customer income scale is a variable that explains the behavior of overall customer satisfaction. Regressions 8 and 10 estimate if the income scale is significant to explain the behavior of overall satisfaction, including if the customer knows the utility company. In most cases, it is not possible to affirm that the 59 income scale and the fact of knowing the providing company is significant. However, the results of regression 10.1 indicate that an increase in the income scale reduces overall satisfaction by 0.15 points in average. In addition, regression 10.2 shows that satisfaction increases by about half a point in the cases in which the customer knows the providing company. st a = α + β1 ∗ rsni + β 2 ∗ iii + β 3 ∗ adii + β 4 ∗ fat i + β 5 ∗ dcf i + β 6 ∗ fli + β 7 ∗ eii + µ st a = α + β1 ∗ rsni + β 2 ∗ iii + β 3 ∗ adii + β 4 ∗ fat i + β 5 ∗ dcf i + β 6 ∗ fli + β 7 ∗ sai + β 8 ∗ eii + µ eii: Income Scale Chart 10 Overall Satisfaction versus Partial Indexes, Knowledge of Provider and Income Scale Source: Own elaboration st a = α + β 1 ∗ sac i + β 2 ∗ paa i + β 3 ∗ ia i + β 4 ∗ mo i + β 5 ∗ mc i + β 6 ∗ ms i + β 7 ∗ eii + µ st a = α + β 1 ∗ sac i + β 2 ∗ paa i + β 3 ∗ ia i + β 4 ∗ moi + β 5 ∗ mc i + β 6 ∗ ms i + β 7 ∗ sa i + β 8 ∗ eii + µ Chart 11 60 Overall Satisfaction versus Service Quality and Technical Product, Knowledge of Provider and Income Scale Source: Own elaboration Following with the analysis on the relation between household income level and satisfaction, regressions 11, 12, 13 and 14 estimate whether the index that aggregates all the socioeconomic variables (other than the income scale), could be related to water customers’ overall satisfaction. In all the cases, it is not possible to affirm that there is a significant relation from the statistical point of view between this socioeconomic index and the level of overall customer satisfaction. st a = α + β1 ∗ rsni + β 2 ∗ iii + β 3 ∗ adii + β 4 ∗ fat i + β 5 ∗ dcf i + β 6 ∗ fli + β 7 ∗ isoi + µ isi: Social Index st a = α + β1 ∗ rsni + β 2 ∗ iii + β 3 ∗ adii + β 4 ∗ fat i + β 5 ∗ dcf i + β 6 ∗ fli + β 7 ∗ sai + β 8 ∗ isoi + µ Chart 12 Overall Satisfaction versus Partial Indexes, Knowledge of Provider and Social Index 61 Source: Own elaboration st a = α + β1 ∗ saci + β 2 ∗ paai + β 3 ∗ iai + β 4 ∗ moi + β 5 ∗ mci + β 6 ∗ msi + β 7 ∗ isoi + µ st a = α + β 1 ∗ sac i + β 2 ∗ paa i + β 3 ∗ ia i + β 4 ∗ moi + β 5 ∗ mc i + β 6 ∗ ms i + β 7 ∗ sa i + β 8 ∗ iso i + µ Chart 13 Overall Satisfaction versus Service Quality and Technical Product, Knowledge of Provider and Social Index 62 Source: Own elaboration In turn, regressions 15 and 16 estimate whether the socioeconomic variables would be individually related to overall satisfaction with the water service. The results arising from the first exercise of the survey shown in 15.1 indicate that the fact there is a bathroom in the household or that the head of the household holds a postgraduate degree are significant variables with a probability of 90% and 95%, respectively. In particular, the strong link between these two variables and overall satisfaction is evidenced by the fact that the latter increases about 4 points if there is a bathroom in the household and 5 points if the head of the household has completed a postgraduate course of study. However, it is not possible to affirm that these variables are significant in the exercise of the second wave of the survey (15.2); although there actually is a negative relation (-0.30) between the number of household members and the level of satisfaction with the service, with a 95% probability. Finally, if the same estimate is made but, this time, including the variables related to the perception of service technical quality (regressions 17 and 18), the variables that reflect if the head of the household lives on welfare, has not completed the elementary or the secondary school, are significant in the first wave of the survey (17.1 and 18.1); with a 95% probability for the first two variables and 90% for the last 63 variable. 25 In this sense, overall satisfaction increases by one point if the head of the household lives on welfare, by about 1.6 if the head of the household has not completed the elementary school and by 1.1 if the head of the household has not completed the secondary school. In turn, the results arising from the second phase of the survey (regressions 17.2 and 18.2) indicate that it is possible to affirm, with a probability equal to or higher than 95%, that there exists a relation between overall satisfaction and the variables that measure household members (-0.13), the availability of a bathroom (-1.2) and the head of the household’s welfare plan (-0.8). 25 Therefore, it is not possible to affirm that this variable is still significant in 24.1. 64 Chart 14 Overall Satisfaction versus Partial Indexes, Knowledge of Provider and Socioeconomic Indicators 65 Chart 15 Overall Satisfaction versus Service Quality and Technical Product, Knowledge of Provider and Socioeconomic Indicators Source: Own elaboration 66 b) Electric Power Regressions 19 and 22 estimate whether, in principle, it may be affirmed that there are differences in the degree of overall satisfaction between customers who know the electric power company and those who don’t. In both cases, the results show that it is not possible to affirm that this variable is related to overall satisfaction, both in relation to the results of the first wave of the survey (19.1 and 22.1) and the results of the second wave of the survey (19.2 and 22.2). st e = α + β1 ∗ rsni + β 2 ∗ iii + β 3 ∗ adii + β 4 ∗ dcf i + β 5 ∗ fli + β 6 ∗ sei + µ se: knows the electric power company (PE01) st e = α + β1 ∗ rsni + β 2 ∗ iii + β 3 ∗ adii + β 4 ∗ dcf i + β 5 ∗ fli + β 6 ∗ isi + µ se: socioeconomic index st e = α + β1 ∗ rsni + β 2 ∗ iii + β 3 ∗ adii + β 4 ∗ dcf i + β 5 ∗ fli + β 6 ∗ sei + β 7 ∗ isi + µ Chart 16 Overall Satisfaction versus Partial Indexes, Knowledge of Provider and Socioeconomic Index Source: Own elaboration 67 On the other hand, regressions 20 and 23 estimate if the socioeconomic level of electric power customers would be influencing, to some extent, the level of overall satisfaction with the service. Furthermore, regressions 21 and 24 reflect if the socioeconomic level of customers as well as their knowledge of the utility company are related to overall satisfaction. In all cases, on the basis of the first and second phases of the survey, it cannot be asserted that these variables may be related to the level of overall customer satisfaction. st e = α + β1 ∗ iei + β 2 ∗ tai + β 3 ∗ sei + µ st e = α + β1 ∗ iei + β 2 ∗ tai + β 3 ∗ isi + µ st e = α + β1 ∗ iei + β 2 ∗ tai + β 3 ∗ sei + β 4 ∗ isi + µ Chart 17 Overall Satisfaction versus Service Quality and Technical Product, Knowledge of Provider and Socioeconomic Index Source: Own elaboration On the basis of the relation between income and satisfaction, regressions 25 and 27 estimate whether the customers’ income scale is significant to explain overall satisfaction with the electric power service. In accordance with the results arising from the first wave of the survey (25.1), it is possible to affirm with a 95% probability that they are related. 68 This arises as a result of the fact that any increase in the income scale generates an average overall customer satisfaction increase equivalent to 0.21. Nevertheless, it is not possible to affirm that the income scale is related to overall satisfaction when its significance was measured together with the objective perception variables (27.1). However, pursuant to the information arising from the second wave (25.2), it is not possible to affirm that there is any link between the income scale and the level of satisfaction of customers. The income scale variable gains significance, equivalent to - 0.1, upon assessment of its relation with satisfaction together with the variables of service technical quality with a 90% probability (27.2). Furthermore, regressions 26 and 28 estimate whether the income scale is significant with the variable that reflects if the customer knows the providing company. There are no differences in the results, and just as it was mentioned above, it is not possible to assert that the fact of knowing the providing company is related to the degree of customer satisfaction. st e = α + β1 ∗ rsni + β 2 ∗ iii + β 3 ∗ adii + β 4 ∗ dcf i + β 5 ∗ fli + β 6 ∗ eii + µ st e = α + β1 ∗ rsni + β 2 ∗ iii + β 3 ∗ adii + β 4 ∗ dcf i + β 5 ∗ fli + β 6 ∗ sei + β 7 ∗ eii + µ Chart 18 Overall Satisfaction versus Partial Indexes, Knowledge of Provider and Income Scale 69 Source: Own elaboration st e = α + β1 ∗ iei + β 2 ∗ tai + β 3 ∗ eii + µ st e = α + β1 ∗ iei + β 2 ∗ tai + β 3 ∗ eii + β 4 ∗ sei + µ Chart 19 Overall Satisfaction versus Service Quality and Technical Product, Knowledge of Provider and Income Scale Source: Own elaboration Following this analysis, the results arising from regressions 29, 30, 31 and 32 show that it is not possible to assert that the socioeconomic index which encompasses all the socioeconomic variables (other than income scale) may be significant to explain the overall customer satisfaction with the electric power service. Notwithstanding that, it is worth highlighting that the variable reflecting knowledge of the electric power company becomes significance with a 95% probability, according to the results arising from the second wave of the survey (32.2). That is to say, satisfaction is reduced by 0.6 in average in the cases in which customers know the providing company. st e = α + β1 ∗ rsni + β 2 ∗ iii + β 3 ∗ adii + β 4 ∗ dcf i + β 5 ∗ fli + β 6 ∗ iso i + µ st e = α + β1 ∗ iei + β 2 ∗ tai + β 3 ∗ isoi + µ st e = α + β1 ∗ rsni + β 2 ∗ iii + β 3 ∗ adii + β 4 ∗ dcf i + β 5 ∗ fli + β 6 ∗ se i + β 7 ∗ isoi + µ st e = α + β1 ∗ iei + β 2 ∗ tai + β 3 ∗ sei + β 4 ∗ isoi + µ 70 Chart 20 Overall Satisfaction versus Partial Indexes, Service Quality and Technical Product, Knowledge of Provider and Social Index Source: Own elaboration Finally, it is estimated whether the disaggregated socioeconomic variables are individually related to the overall satisfaction of electric power customers. The results arising from the first wave of the survey (33.1) reflect that the variables that indicate the level of education of the head of the household (incomplete postsecondary studies, complete university course of study or postgraduate degree) explain the behavior of overall satisfaction with a probability equal to 95% in the first two cases and 90% in the last case. 26 Overall satisfaction increases an average of 3 points in the cases where the head of the household has completed post-secondary, university or postgraduate studies. In turn, the results obtained from the second wave of the survey (34.2) reflect, with a probability of 95%, that satisfaction decreases by 0.63 points and 2.5 points in the cases in which the head of the household lives on welfare and holds a postgraduate degree, respectively. Regressions 35 and 36 estimate whether it is possible to sustain that there is 26 Only the variable “incomplete post-secondary studies� keeps significant in regression 36.1. 71 a connection between the disaggregated socioeconomic variables and overall satisfaction, including whether customers know the providing company. The results obtained for both waves of the survey do not vary. To sum up, results reflect that the level of education of the head of the household impacts, to a great extent, on the overall satisfaction of electric power customers. 72 Chart 21 Overall Satisfaction versus Partial Indexes, Service Quality and Technical Product, Knowledge of Provider and Socioeconomic Indicators Source: Own elaboration 73 c) Gas Regressions 37 and 40 estimate if satisfaction may be explained on the grounds that customers know the providing company. The results show that it is not possible to affirm that such a variable is linked to the levels of overall customer satisfaction. st g = α + β1 ∗ rpni + β 2 ∗ ibpi + β 3 ∗ adbpi + β 4 ∗ dcf i + β 5 ∗ fli + β 6 ∗ sg i + µ sgi: knows the gas company (PG01) st g = α + β1 ∗ rpni + β 2 ∗ ibpi + β 3 ∗ adbpi + β 4 ∗ dcf i + β 5 ∗ fli + β 6 ∗ isi + µ st g = α + β1 ∗ rpni + β 2 ∗ ibpi + β 3 ∗ adbpi + β 4 ∗ dcf i + β 5 ∗ fli + β 6 ∗ sg i + β 7 ∗ isi + µ Chart 22 Overall Satisfaction versus Partial Indexes, Knowledge of Provider and Socioeconomic index Source: Own elaboration Regressions 38 and 41 estimate whether the socioeconomic level of customers is related to their overall satisfaction. In addition, regressions 39 and 42 estimate whether the socioeconomic level is significant to explain the behavior of overall satisfaction, 74 including if the customer knows the gas company. The results arising from the first phase of the survey (38.1 and 39.1) indicate that it may be asserted with a 90% probability, that the degree of satisfaction of gas customers is related to their socioeconomic level, and would not depend on the degree of information they have about the company. In this sense, it is expected that a one-unit increase in the socioeconomic index would generate an average increase in satisfaction equivalent to 0.26. It is worth stating that it is not possible to assert that this variable is significant in the second stage of the survey (38.2 and 39.2). Notwithstanding that, after assessing the relevance of customers’ socioeconomic level, together with service quality perception variables (41 and 42), it is not possible to sustain that the socioeconomic level variable is significant to explain the behavior of customer satisfaction with the gas service. st g = α + β1 ∗ pag i + β 2 ∗ sg i + µ st g = α + β1 ∗ pag i + β 2 ∗ isi + µ st g = α + β1 ∗ pag i + β 2 ∗ sg i + β 3 ∗ isi + µ Chart 23 Overall Satisfaction versus Quality of the Technical Product, Knowledge of Provider and Socioeconomic index Source: Own elaboration Regressions 43 and 44 estimate whether the income scale of gas customers is related to their overall satisfaction. Furthermore, regressions 45 and 46 estimate whether the 75 income scale is related to satisfaction, including the variable that indicates if the customer knows the providing company. In all the cases of the first wave of the survey, it is observed that the customers’ income scale is significant to explain the behavior of satisfaction with a 90% probability. This becomes evident by the fact that a higher income scale generates, in average, a 0.29 increase of overall satisfaction, when its significance is estimated with the subjective satisfaction variables (43.1 and 45.1). At the same time, it produces a 0.1 decrease in overall satisfaction, when its significance is estimated with the variables related to service technical quality (44.1 and 46.1). As it has been mentioned above, the fact of knowing or not knowing the providing company does not influence satisfaction results. In accordance with the results of the second wave of the survey (44.2, 45.2 and 46.2), it is not possible to affirm that there is a relation between the income scale and the level of satisfaction with the service. st g = α + β1 ∗ rpni + β 2 ∗ ibpi + β 3 ∗ adbpi + β 4 ∗ dcf i + β 5 ∗ fli + β 6 ∗ eii + µ st g = α + β1 ∗ pag i + β 2 ∗ eii + µ st g = α + β1 ∗ rpni + β 2 ∗ ibpi + β 3 ∗ adbpi + β 4 ∗ dcf i + β 5 ∗ fli + β 6 ∗ sg i + β 7 ∗ eii + µ st g = α + β1 ∗ pag i + β 2 ∗ sg i + β 3 ∗ eii + µ 76 Chart 24 Overall Satisfaction versus Partial Indexes, Quality of Technical Product, Knowledge of Provider and Income Scale Source: Own elaboration Following with this analysis of the income-satisfaction relation, regressions 47, 48, 49 and 50 estimate whether the socioeconomic index that aggregates all socioeconomic variables, other than the income scale, is related to the levels of gas customers’ overall satisfaction. Once again, it is not possible to sustain that this index may have some type of relation with the overall satisfaction of gas customers, both for the results of the first wave of the survey and the second one. st g = α + β1 ∗ rpni + β 2 ∗ ibpi + β 3 ∗ adbpi + β 4 ∗ dcf i + β 5 ∗ fli + β 6 ∗ isoi + µ st g = α + β1 ∗ pag i + β 2 ∗ isoi + µ st g = α + β1 ∗ rpni + β 2 ∗ ibpi + β 3 ∗ adbpi + β 4 ∗ dcf i + β 5 ∗ fli + β 6 ∗ sg i + β 7 ∗ isoi + µ st g = α + β1 ∗ pag i + β 2 ∗ sg i + β 3 ∗ isoi + µ 77 Chart 25 Overall Satisfaction versus Partial Indexes, Quality of Technical Product, Knowledge of Provider and Social Index Source: Own elaboration Finally, through the observation -in regressions 51 and 53- of the individual relation between socioeconomic variables and overall satisfaction with the gas service, together with partial satisfaction variables, it may be concluded that none of them is significant. However, if the same estimate is made, but this time including service quality perception variables for the first phase of the survey (52.1 and 54.1), it can be sustained, with a 95% significance level, that overall satisfaction is related to the variable that shows if the head of the household has completed university studies. It is expected that overall satisfaction be reduced one point in average (1.12) if the head of the household holds a university degree. 78 Chart 26 Overall Satisfaction versus Partial Indexes, Quality of Technical Product, Knowledge of Provider and Socioeconomic Indicators Source: Own elaboration 79 Overall Satisfaction and Partial Indicators The purpose of this section is to assess the relation between partial satisfaction indexes and overall satisfaction index for each sector in order to identify the relative importance of each dimension of overall satisfaction. Overall indexes reflect the degree of customer satisfaction with the service provision, while partial indexes express the degree of customer satisfaction in relation to certain attributes that are inherent to the service and contribute to the quality of the technical service, the technical product and the commercial service, respectively. Chart 27 below details the aspects considered for each of these dimensions with respect to each of the services included in the sample. Chart 27 Service Attributes considered in the Survey WATER ELECTRIC GAS POWER Technical Service Number (PA04) and Number (PA04) and duration (PA05) of duration (PA05) of service interruptions outages Technical Product Number (PA06) and Number (PE06) and Number (PG04) and frequency (PA07) of frequency (PE07) of frequency (PG05) of inadequate pressure voltage variation low pressure events. events. events Features of supplied water (color, smell, taste, etc.) (PA10) Commercial Service Announcement of Announcement of Announcement of service interruption outages (PE09) service interruption (PA09). Features of Bills (PG07) Features of Bills (PE20) Features of Bills (PA21) Complaint-handling (PG18) Complaint-handling system (PE14) Complaint-handling system (PA14) system (PG14) Source: Own elaboration The aim is to estimate the relation between the quality perception of each of these service dimensions and overall satisfaction for the purpose of measuring the extent to which 80 service quality variables impact on satisfaction. Quality perception is measured with objective variables that, in the particular case of the water sector, may be identified through the following characteristics: perception in respect of service continuity, adequate water pressure, service interruptions, and bad smell, color or taste of water. As regards the electric power sector, these indexes indicate if the service shows any interruptions and the adequacy of voltage. In the gas sector, only the adequacy of pressure is measured. This analysis is extremely important from the point of view of these utilities’ regulation since it contributes essential information for the design and control of the respective regulatory systems. As it was discussed in the first report, the imposition of quality standards constitutes an integral part of every price cap regulatory system. This makes it essential to identify the service quality dimensions valued by customers since, in principle, control should be focused on quality dimensions that the company can control and that, at the same time, are regards by customers as important. 27 In other words: the dimensions that may affect overall customer satisfaction with the service. So, the higher the impact of an individual variable on overall customer satisfaction, the higher the attention that it should be paid from the point of view of regulation. This attention comprises two aspects: design and control. As regards the design of quality standards, the most significant variables should be included first within the rules, upon definition of the quality system (for example, at the commencement of a concession) and the reviews (typically during tariff reviews). In the second place, the design of the quality system should reflect the importance that customers give to these dimensions, and therefore create clear incentives (such as fines) on the basis of each dimension’s relative importance in customers’ perception. 27 In addition, any dimension which may affect the health of people or the environment must be controlled even if they are not easily perceived by customers. This is particularly interesting for the water sector. 81 As regards control, the higher the relevance of one dimension, the greater the regulatory effort that should be made to ensure that the company complies with the provisions set out in relation to it. Clearly, this analysis should be supplemented, from the regulatory point of view, by the customers’ willingness to pay for each of the relevant quality dimensions. This clearly exceeds the scope of the survey conducted. The results of each of the three sectors considered in the surveys are shown below. General model: The estimate for each sector is as follows: st = α + β1 ∗ sp1 + β 2 ∗ sp2 + ... + β n ∗ spn + µ Where: α: constant st: overall satisfaction sp1: partial satisfaction 1 sp2: partial satisfaction 2 ........................................ spn: partial satisfaction n μ: random error a) Water The following specification is considered for the water sector model: st a = α + β1 ∗ rsni + β 2 ∗ iii + β 3 ∗ adii + β 4 ∗ fat i + β 5 ∗ dcf i + β 6 ∗ fli + µ i: 1,2,3….500 Where sta: overall satisfaction with water service (PA02) rsni: satisfaction with ordinary service restoration (PA09) 82 iii: satisfaction with announcements of service interruptions or low water pressure events (PA09) Adii : satisfaction with company’s response during service interruptions or low water pressure events (PA09) fati: satisfaction with punctuality of bill (PA21) dcfi : satisfaction with itemized billing (PA21) fli : satisfaction with easy-to understand bills (PA21) Chart 28 below shows the results arising from regression 55 that estimates the weight of each partial satisfaction index on overall satisfaction in the water sector. 28 Chart 28 Overall Satisfaction versus Partial Indexes Source: Own elaboration The estimates were made on the basis of 83 and 91 observations and the coefficients of determination (R-squared) amounted to 0.48 and 0.62, thereby indicating that between 48 28 From this point onwards, in order to make reading easier, when making reference to regression “n�, reference is made to regression “n.1� and regression “n.2� pertaining to the first and second wave of the survey, respectively. 83 and 62% of the variations in overall satisfaction are explained by the partial satisfaction indexes included in the regression. In the first wave of the survey (55.1), all the partial satisfaction indexes were significant to explain the overall customer satisfaction with the water service, except for the partial index that measures satisfaction with itemized billing. Therefore, in principle, it may not be asserted that satisfaction with itemized billing is related to overall customer satisfaction. Furthermore, on the basis of the results arising from the second wave of the survey (55.2), it cannot be asserted either, that the partial satisfaction index regarding the information on interruptions or low pressure events is significant to explain the behavior of overall satisfaction with the water service. The null hypothesis (Ho) that states that the coefficients of partial satisfaction with the company’s response during service interruptions or low pressure events and with punctuality of the bill are equivalent to zero, is rejected with a probability equal to or higher than 95%. In turn, the results in 55.1 reject, with a 90% probability, that overall satisfaction is related to the performance of companies to promptly restore service provision, announce long-lasting interruptions or low pressure events, and offer an easy- to-understand bill. The significance levels of the first and third variable in 55.2 increased to 99% and 95%, respectively. Among the variables measuring the quality of the commercial service, the results of 55.1 reflect that the punctuality of the bill would be the most important dimension of overall satisfaction. A one-point increase in this partial satisfaction generates an average increase of 0.39 in overall satisfaction. In turn, partial satisfaction related to the company’s response during service interruptions or low pressure events also causes a significant impact on overall satisfaction since a one-point increase in this satisfaction entails an average increase of 0.28 in overall customer satisfaction. However, the variations in partial satisfaction regarding the prompt restoration of service and the announcement of long-lasting interruptions or low pressure events have a lower impact on overall satisfaction; 0.20 and 0.16 respectively. Finally, the relation between partial satisfaction 84 with respect to easy-to-understand bills and overall satisfaction is negative, since a one- point increase in this attribute of satisfaction generates a sharp fall in overall satisfaction equivalent to 0.48, in average. In contrast, the results of 55.2 reflect a relevant variation in the relation between overall satisfaction and partial satisfaction indexes regarding the prompt restoration of service, punctuality of the bill and easy-to-understand bills. In particular, a one-unit increase in partial satisfaction regarding prompt restoration of the service causes an average variation of almost half a point (0.42) in overall satisfaction. Even though the relation between punctuality of the bill and satisfaction with the service used to be positive, it has now become negative, thus getting a greater relevance (0.68). Furthermore, the impact on the other dimension related to the features of the bill (easy-to-understand bills) has also changed significantly from being negative to producing a positive variation equivalent to 0.31. Regression 56 – its results are shown in Chart 29– estimates the relation between the variables that measure technical service and technical product’s quality perception and overall satisfaction in the water sector. The estimated model is the following: st a = α + β 1 ∗ sac i + β 2 ∗ paa i + β 3 ∗ ia i + β 4 ∗ mo i + β 5 ∗ mc i + β 6 ∗ ms i + µ i: 1,2,3….500 Where sta: overall satisfaction with water service (PA02) saci: uninterrupted water service (PA03) check question. paai : adequate pressure of water (PA03) idem iai : water interruptions (PA04) moi : bad smelling water (PA10) mci : bad color of water (PA10) msi : bad tasting water (PA10) 85 Chart 29 Overall Satisfaction versus Service Quality and Technical Product Source: Own elaboration In the first wave of the survey (56.1), the null hypothesis (Ho) that the coefficient of the variable that measures if the service is continuous equals zero is rejected with a 90% probability, while in the case of the variables that measure pressure and bad smell of water, the Ho is rejected with a high probability of 99%. Notwithstanding that, it is not possible to assert that service interruptions and bad color and taste of water are related to the overall customer satisfaction. In the second wave of the survey (56.1), the null hypothesis (Ho) regarding the non-significance of service continuity and bad color of water is rejected, with a probability of 99%, while on the basis of the information available, it is not possible to assert that the bad smell of water is still significant to explain the behavior of overall satisfaction. The estimates were conducted on the basis of 426 and 441 observations and the coefficients of determination (R-squared) amounted to 0.42 and 0.44, thereby indicating that between 42% and 44% of the variations in overall satisfaction are explained by the service quality perception variables included in the regression. 86 In comparative terms, it is worth highlighting that the perception variables of service technical quality have a greater impact on overall customer satisfaction than the commercial quality variables on partial satisfaction. The strong incidence of these objective variables becomes evident by the fact that the results of 56.1 reflect that overall satisfaction increases more than 2 points, in average, if the pressure is adequate; it is reduced by almost one point in the cases of bad smell of water, and it is increased by 0.67 when the service is uninterrupted. In turn, the information arising from 56.2 shows that overall satisfaction increases, in average, more than one point in the cases of uninterrupted service, and it is reduced by about half a point in the cases of bad color of the water. b) Electric power The specific electric power sector model is the following: st e = α + β1 ∗ rsni + β 2 ∗ iii + β 3 ∗ adii + β 4 ∗ dcf i + β 5 ∗ fli + µ i: 1,2,3….500 where ste: overall satisfaction with electric power service (PE02) rsni: satisfaction with ordinary service restoration (PE09) iii : satisfaction with announcements of outages and low voltage events (PE09) adii : satisfaction with company’s response during outages or low voltage events (PE09) dcfi : satisfaction with itemized billing (PE20) fli : satisfaction with easy-to-understand bills (PE20) Regression 57 estimates the weight of each partial satisfaction index in relation to overall satisfaction in the electric power sector for both waves. The results are shown in Chart 30. 87 Chart 30 Overall Satisfaction versus Partial Indexes Source: own elaboration The estimates were made on the basis of 209 and 227 observations and the coefficients of determination (R-squared) amounted to 0.41 and 0.49, which indicates that between 41 and 49% of overall satisfaction variations are determined by the explanatory variables in the regression. Based on the results of the first phase of the survey (57.1), the partial satisfaction indexes which are significant to explain electric power customers’ overall satisfaction are: the electric power company’s efforts to restore the service; announce long-lasting outages or low voltage events; and the satisfaction with the itemized billing. However, it cannot be asserted, a priori, that the company’s response during outages or low voltage events and easy-to-understand bills are related to overall satisfaction. Moreover, the data collected during the second phase of the survey (57.2) does not enable to assert that the partial satisfaction index regarding the announcement of outages is still significant enough to explain the behavior of overall satisfaction with the service. On the other hand, it can be observed that satisfaction with the company’s response during outages has become significant. 88 Based on the foregoing, the hypothesis that partial customer satisfaction coefficients regarding the company’s efforts to restore the service and satisfaction with the itemized billing equal zero is rejected with a significance level equal to or greater than 95%. Moreover, the results in 57.1 and 57.2 reflect that partial satisfaction in connection with the announcements of outages and the company’s response during such outages is relevant to explain the behavior of overall satisfaction with a probability of 90% and 95%, respectively. Thus, if we consider the electric power company’s efforts to restore the service, we conclude that it is the satisfaction dimension that has the greatest influence on overall satisfaction. This is explained by the fact that a one-point increase in this partial satisfaction variable causes an average increase of about ½ point in overall satisfaction. Furthermore, partial satisfaction with itemized billing is also strongly linked to overall satisfaction. A one-point increase in this satisfaction variable causes an average increase of almost 1/4 point in overall satisfaction. However, it can be noted in 57.1 and 57.2 that partial satisfaction variations related to announcements of long-lasting outages or low voltage events and the company’s response during outages have a lower impact (of about 0.1) on overall satisfaction. Moreover, regression 58 –whose results are shown in chart 31– estimates the relationship between service quality perception and overall satisfaction in the electric power sector. The functional form is as follows: st e = α + β 1 ∗ ie i + β 2 ∗ ta i + µ i: 1,2,3….500 where ste: overall satisfaction with electric power service (PE02) iei: satisfaction with outages (PE04) check question. tai : satisfaction with adequate voltage (PE06) idem. 89 Chart 31 Overall Satisfaction versus Service Quality and Technical Product Source: own elaboration The estimates were made on the basis of 500 observations and the coefficients of determination (R-squared) amounted to 0.24 and 0.27, thereby indicating that objective perception variables collectively explain only between 24 and 27% of overall satisfaction variations. Consequently, the hypothesis that the variable reflecting outages and the variable indicating if the voltage is inadequate are not relevant is rejected. These two variables explain the behavior of overall satisfaction with a high probability of 99%. Thus, results show, as well as in the water sector, that objective variables affecting technical quality perception have a greater impact on overall satisfaction than subjective satisfaction variables. The strong link between these variables and overall customer satisfaction is evidenced by the fact that the latter decreases between 1.3 and 1.7 when the voltage is inadequate and about 1.0 during outages. It is worth highlighting that according to the results of the second wave of the survey (58.2), inadequate voltage causes an impact 27% lower on overall satisfaction, in relation to the data collected during the initial phase of the survey (58.1). c) Gas 90 The specific gas model was estimated as follows: st g = α + β1 ∗ rpni + β 2 ∗ ibpi + β 3 ∗ adbpi + β 4 ∗ dcf i + β 5 ∗ fli + µ i: 1,2,3….500 where stg: overall satisfaction with gas service (PG02) rpni: satisfaction with restoration of ordinary pressure (PG07) ibpi : satisfaction with announcements of low gas pressure events (PG07) adbpi satisfaction with company’s response during low gas pressure events (PG07) dcfi : satisfaction with itemized billing (PG18) fli : satisfaction with easy-to-understand bills (PG18) In regression 59, the impact of each partial customer satisfaction index on overall customer satisfaction is estimated. The results are reported on chart 32. Chart 32 Overall Satisfaction versus Partial Indexes Source: own elaboration The estimates were made on the basis of 61 and 72 observations and the coefficients of determination (R-squared) amounted to 0.41 and 0.47, thereby showing that 41- 47% of 91 overall satisfaction variations are collectively explained by the partial satisfaction indexes included in the regression. According to the results of the first phase of the survey (591), the indexes which are significant to explain the behavior of overall satisfaction with the gas service are satisfaction with itemized billing and the company’s efforts to restore the service. Although, a priori, it should be expected otherwise, it may not be concluded that satisfaction with the company’s announcements of low gas pressure events, its response during low gas pressure events and easy-to-understand bills is related to gas customers’ overall satisfaction. In addition, the results obtained from the data collected during the second phase of the survey (59.2) do not show that the partial customer satisfaction index regarding itemized billing remains significant. Particularly, it is shown in 59.1 that the null hypothesis regarding the non-significance of partial satisfaction in connection with itemized billing is rejected on the basis of a high probability of 99%. Whereas, both results evidence that the hypothesis stating that the partial satisfaction coefficient regarding the company’s efforts to restore the service equals zero is rejected based on a 95% level of probability. It is worth stressing that costumer satisfaction in connection with itemized billing is the most important dimension of overall satisfaction with the gas service, according to 59.1. This is evidenced by the fact that a one-point increase in this satisfaction dimension causes, on average, an increase of almost half a point (0.44) in overall satisfaction. As regards the relation between partial customer satisfaction in connection with the company’s efforts to restore the service and overall customer satisfaction, it is shown that it increased from 0.21 to 0.32. Moreover, regression 60 – whose results are shown on chart 33 – estimates the relation between service quality perception and overall satisfaction in the gas sector based on the following form: st g = α + β 1 ∗ pag i + µ 92 i: 1,2,3….500 Where stg: overall satisfaction with gas service pagi: adequate gas pressure (PG03) check question. Chart 33 Overall Satisfaction versus Technical Product Quality Source: own elaboration The estimates were made on the basis of 423 and 426 observations and the coefficients of determination (R-squared) amounted to 0.20 and 0.29, thereby showing that over 70% of satisfaction variations are due to variables which were not included in the model. The objective service quality variable indicating if the gas pressure is adequate explains gas customers’ overall satisfaction with a high probability of 99%. As in the water and electric power sector, objective technical quality variables have a larger impact on overall satisfaction than subjective satisfaction variables. The strong link between this quality perception variable and overall satisfaction, and therefore its relevance, is evidenced by the fact that the latter increases by about 2 points on average if the gas pressure is adequate. Company benchmarking Benchmarking is one of the main tools for reducing information asymmetry in the relationship between the regulator and the utility company. The use of this tool is 93 widespread among regulators around the world, particularly, in connection with setting productivity factors (X) in price cap regulatory regimes. As there are several utility companies providing the service in adjacent areas even when they constitute a monopoly, it is possible to compare their relative performance. Although this comparison is usually performed on the basis of cost or production functions, it may be easily applied to quality dimensions and it may provide useful information from the regulatory point of view. Thus, this section compares the answers provided by the customers of the two gas companies (Gas Ban and Metrogas) and the two electric power companies (Edenor and Edesur) for the purpose of determining if they are significant differences in overall and partial satisfaction indexes and service quality perception variables. This comparison shows the performance of the companies offering the same service in different geographical areas. It cannot be performed in the water sector, as there is only one company in the sector. a) Electric power A difference of means test was performed regarding overall satisfaction, partial satisfaction and quality service perceptions among electricity companies and gas companies as well. The null hypothesis is that means are equal. The test is aimed at determining if the satisfaction indexes differ among customers of the utility companies. In the case of the electric power service, the results obtained from the first wave of the survey reflect that it may not be concluded that the means of satisfaction with the companies’ announcements of outages or low voltage events and their response during such events, are equal with a 90% and 95% level of probability, respectively. Chart 34 shows that the Edenor customers’ mean satisfaction with these two attributes exceeds the mean satisfaction of Edesur customers by 15-20%. According to the results of the second wave of the survey, it may not be concluded, with a 95% level of probability, that the mean of the variables determining if the voltage level is adequate and customer 94 satisfaction with the company’s announcements of outages or low voltage events are statistically equal for both utility companies. Based on the figures reported on chart 34, Edesur’s performance exceeded Edenor’s in these two satisfaction dimensions by 17.5 and 36.4%. It is not possible to conclude that the mean of the remaining variables show statistically significant differences, however, it is worth mentioning that the average expenditure of Edenor customers exceeded by 6% the average expenditure of Edesur customers in the first wave of the survey and by 19% in the second wave. Besides, Edenor customers showed, on average, higher satisfaction with itemized billing and easy-to-understand bills. Whereas, while in the first wave Edesur customers showed higher satisfaction levels with the company’s efforts to restore the service, Edenor customers showed higher satisfaction in this dimension in the second wave of the survey. Chart 34 Means of Overall Satisfaction, Partial Indexes, Service Quality and Technical Product and Expenditure Source: own elaboration The distribution of the overall customer satisfaction function for the companies in the electric power sector – whose results are shown in Graphic 17 – reflects a marked bias towards the right, with a large number of high level responses for both companies. Edenor customers showed, on average, more medium and medium-high satisfaction levels than those of Edesur in the first wave, and more responses evidencing a high 95 satisfaction level in the second wave. Whereas, Edesur customers provided more responses evidencing a low satisfaction level and also reported the maximum level in the first wave, and medium and medium-high satisfaction levels in the second wave. To sum up, the overall customer satisfaction of electric power customers is high for both companies. Graphic 17 Overall Satisfaction Histogram First wave Second wave Source: Own elaboration As regards satisfaction with electric power service restoration – whose results are shown in graphic 18 – it may be concluded that for both companies most responses reflect medium-high levels. However, in the first wave results show that Edenor customers provided more medium-level responses than Edesur customers, who in turn provided more low-level responses; whereas, in the second wave of the survey Edenor customers provided more medium-high level responses. Thus, satisfaction with this attribute was higher for Edesur customers in the first wave, while Edenor customers showed higher satisfaction levels in the second wave. 96 Graphic 18 Histogram of Satisfaction with Service Restoration First wave Second wave Source: Own elaboration Furthermore, the distribution of customers’ responses in connection with the company’s announcement of long-lasting outages or low voltage events is displayed in graphic 19. In both waves, a bias to the left may be noticed for both companies and most responses are associated with minimum satisfaction, thereby evidencing serious weaknesses in this satisfaction dimension. However, in the first wave Edesur customers were more likely to express this kind of satisfaction than Edenor customers, whereas in the second wave the opposite is shown. Therefore, satisfaction with the company’s announcement of technical failures tends to be low for both companies. Graphic 19 Histogram of Satisfaction with Announcements of Outages or Low voltage Events First wave Second wave 97 Source: Own elaboration Graphic 20 shows customer satisfaction with the company’s response during outages or low voltage events. In the first wave, response distribution shows a bias to the right in the case of Edenor and reflects mostly medium-high levels. In the second wave, the number of low-level responses is increased thus, such bias is slighter. In the case of Edesur, there were more cases showing low levels in the first wave, with a slighter bias to the right. Though satisfaction with this dimension does not follow a trend and a clear bias towards certain distribution values, there are more medium-high levels. Graphic 20 Histogram of Satisfaction with Company’s response during Outages or Low Voltage Events First wave Second wave Source: Own elaboration The distribution of satisfaction with itemized billing – whose results are reported in graphic 21 – shows a sharp bias towards the right, reflecting that most responses indicated high satisfaction levels. This means that, in general, the customers of both companies are clearly satisfied with itemized billing. 98 Graphic 21 Histogram of Satisfaction with Itemized Billing First wave Second wave Source: Own elaboration Satisfaction with easy-to-understand bills – whose results are reported in graphic 22 – also shows a sharp bias towards the right for both companies and in both waves, thereby reaching in most cases satisfaction levels of 8 - 10 units. To sum up, the costumers of both companies report high levels of satisfaction with this overall satisfaction dimension, and generally, the features of the bill. Graphic 22 Histogram of Satisfaction with Easy-to-understand Bills First wave Second wave Source: Own elaboration 99 Finally, below it is shown that the distribution of electric power expenditure strictly ranges from $0-75 to $75-100 for both companies in both waves of the survey. However, it is shown that Edesur customers provided more responses indicating greater expenditure than Edenor customers, which is evidenced by the high expenditure segments, from $200 upwards. Thus, Edesur customers tend to spend more than Edenor customers. Graphic 23 Customer Expenditure Histogram First wave Second wave Source: Own elaboration b) Gas In the gas sector, according to the results of the first wave of the survey, it cannot be concluded that the means of adequate pressure and service expenditure are statistically equal for both utility companies with a 95% and 99% probability, respectively. Thus, according to the statements of Gas Ban costumers, their average expenditure exceeds the average expenditure reported by Metrogas customers by 37.2%. In addition, the first group of customers reported a less adequate gas pressure than the gas pressure reported by the second group. Whereas, based on the results of the second wave, it cannot be concluded that the means of the variables measuring satisfaction, perception and customer expenditure statistically differ between both gas companies. 100 Chart 35 Means of Overall Satisfaction, Partial Indexes, Service Quality and Technical Product and Expenditure Source: Own elaboration It cannot be affirmed that the company means differ, however, the responses furnished by Gas Ban costumers in the first wave show higher satisfaction levels than the responses of Metrogas costumers in connection with most variables, other than itemized billing. Furthermore, Metrogas costumers have reported a higher level of overall satisfaction on average, as well as 17.2% higher partial satisfaction with the announcements of low pressure events and higher satisfaction with pressure adequacy. The distribution of responses regarding overall satisfaction with the sector – whose results are displayed in graphic 24 – reflects a sharp bias towards the highest units, both for Gas Ban and Metrogas, in both waves of the survey. Practically, neither company reported low satisfaction cases. However, the graphic shows that the bias towards the right in the first wave was slighter for Metrogas than for Gas Ban, whereas Metrogas reported higher medium levels in the second wave. Thus, it is highly probable that the costumers of both companies report a high level of overall satisfaction with the service. 101 Graphic 24 Overall Satisfaction Histogram First wave Second wave Source: Own elaboration As regards satisfaction with gas pressure restoration, graphic 25 shows a bias towards the right of the satisfaction responses furnished by the customers of both companies, which deviation is slighter in the second wave than in the first wave of the survey. Notwithstanding the foregoing, Metrogas customers reported, on average, a higher satisfaction level than Gas Ban costumers, who reported to a large extent a very low satisfaction level in the second wave. Thus, customers reported a medium and medium- high level of satisfaction with this dimension. Graphic 25 Histogram of Satisfaction with Service Restoration First wave Second wave Source: Own elaboration 102 The distribution of customers’ responses regarding announcements of long-lasting low pressure events – whose results are reported in graphic 26–, reflects a bias towards the left with lower values. Thus, both companies report a high probability of occurrence of the minimum satisfaction range in both waves of the survey. However, the probability of a minimum satisfaction score in this attribute is higher for Metrogas in the first wave, and for Gas Ban in the second wave. Thus, as in the electric power sector, this kind of satisfaction is to a large extent low for both gas companies. Graphic 26 Histogram of Satisfaction with Announcements of Low Pressure Events First wave Second wave Source: Own elaboration As regards the distribution of customers’ responses regarding satisfaction with the company’s response during low pressure events, the data reported in graphic 27 shows no bias. However, in the case of Gas Ban it is noteworthy that there is a high probability of occurrence of a medium score in the first wave, as well as a low and high score in the second wave. Moreover, Metrogas customers chose the minimum score in both waves. Therefore, the distribution of customers’ responses in this satisfaction dimension is quite irregular. 103 Graphic 27 Histogram of Satisfaction with Company’s Response during Low Pressure Events First wave Second wave Source: Own elaboration Furthermore, the distribution of satisfaction with itemized billing – whose results are reported in graphic 28 – shows a sharp bias towards the highest scores for customers of both companies in both waves of the survey. The foregoing indicates that customers are, in general, very pleased with itemized billing. Graphic 28 Histogram of Satisfaction with Itemized Billing First wave Second wave Source: Own elaboration 104 Likewise, the data shown in graphic 29 indicates that satisfaction with easy-to-understand bills also had a sharp bias towards the right for both companies in both waves, thus reaching in most cases a 9-10 satisfaction level. The foregoing reinforces the conclusion that gas company customers as electric power company customers are very pleased with the features of bills in general and, particularly, their understandability. Graphic 29 Histogram of Satisfaction with Easy-to-understand Bills First wave Second wave Source: Own elaboration Finally, in graphic 30 it is shown, as in the case of the electric power service, a bias towards the left in the gas expenditure distribution for both companies in both waves of the survey. However, Metrogas costumers provided more responses within the range $0- 100 than Gas Ban costumers, who furnished largely responses reporting expenditure over $100. 105 Graphic 30 Customers Expenditure Histogram First wave Second wave Source: Own elaboration Joint assessment of both waves In this section, we analyze if the order of the survey modules influences to some extent the interviewees’ responses in connection with their overall satisfaction with the service. This issue is purely practical in terms of the design and implementation of quality surveys. In principle, it is convenient to cover several sectors with a single survey. Then, it is important to analyze if the joint assessment influences in any way whatsoever the responses provided by customers. Two factors influencing responses may be identified a priori. First, by surveying two sectors jointly there is a risk that the survey might be too long and therefore the interviewee’s responses might be biased towards the last sector. A priori, this bias might be positive (e.g., “I am tired, I’ll tell the surveyor everything is ok, so he leaves soon�) or negative (“I am tired, everything seems worse than it actually is�). Secondly, the poor quality of a service may affect the quality perception of other services resulting from a spillover effect caused by identifying “utilities� as a whole (particularly in connection with privatized companies). Again, a priori the impact of the poor quality of a utility on other utilities may be either positive or negative. As the quality of a utility 106 is poor, thus, in comparison, the quality of the other utilities is better or otherwise the very poor quality of a utility leads to a negative perception of utilities as a whole. Chart 36 Means and Standard Deviation based on the Sequence of Modules Source: Own elaboration In order to study the first issue, it is analyzed if the service satisfaction and perception levels tend to be different when the respective module is in a different position in the survey. The test for mean equality is designed to compare both means. The null hypothesis indicates that both means are equal in three potential situations. The first situation involves dependant observations, i.e., when the two variables analyzed are measured on the same individuals. The second and third situations involve independent observations, i.e., the same variable is measured on different individuals and the variables have the same or a different variance. The ANOVA measures the same variable in individuals from two subpopulations defined 107 by the two categories of the classification criteria. The F statistic is used to test the null hypothesis indicating that the means for all subpopulations are equal. This test requires that the distributions of subpopulations be normal and have an identical variance. The null hypothesis that the means for the six sequences are equal is tested as follows: S ij = m + a j + eij Model i = 1,2,…,nj j = AEG,AGE,EAG,EGA,GAE,GEA H0: αAEG=αAGE=…=αGEA Chart 37 ANOVA test for the First Wave of the Survey Source: Based on own data. Note: A refers water, E electricity and G gas. As mentioned above, one of the ANOVA requirements is the equality of variances between sequences. This requirement is not met by G (Gas), whereas E (Electric power) is on the verge of acceptance, considering a 0.05 significance level: Chart 38 108 Source: Based on own data. However, this case is a typical case where ANOVA is applied, as the size of the samples and the magnitude relation of standard deviations meet the literature requirements. 29 Chart 39 ANOVA Test for the Second Wave of the Survey Source: Own elaboration According to the p-values of the ANOVA test for the second wave, the null hypothesis that the six sequences are equal is not rejected. The following box diagrams show that the medians, the maximum and minimum values of each module, do not differ significantly, considering their order in the survey. In the first wave of the survey (a), the order of the module does not substantially alter the responses in the water sector, as the medians and the minimum values coincide in 1-2-3-5 and 4-6, whereas the maximum values coincide in 1-3-5 and 2-4-6. Furthermore, in the electric power sector it is shown that the medians and the minimum values are equal in 1- 3-4-5-6 and the maximum values in 1-3-4-6. Finally, the gas sector also shows significant coincidences in the responses even though the module has a different position within the survey. The medians coincide in 2-4-5-6 and 1-3, while the maximum values coincide in 1-3-6 and 2-4-5. Also, the responses provided by the customers in the second wave (b) 29 The size of the sample must exceed 30 observations and the highest deviation must be less than twice the lowest deviation. 109 support the hypothesis that the means, maximum and minimum values coincide or tend to be practically equal between the different sequences of the survey. For the water sector, it is shown that the medians, maximum and minimum values are equal in groups 1-2-3 and 5, and in the cases where they differ (4 and 6), such difference is minor. For the electric power sector, the medians coincide in 1-3-5-6 and 2-4, whereas the maximum and minimum values do not differ significantly between the different sequences. Finally, in the gas sector, the medians coincide with the groups 1-3-6 and 2-4-5, and as in the rest of the sectors the extreme values (maximum and minimum) are similar. Graphic 31 Box Diagram by each Module Sequence (a) First wave (b) Second wave . Source: Own elaboration. Note: SEC means sequence. Next, it is shown that the satisfaction means for each sector do not differ significantly depending on the order of the module in the survey in both waves. Particularly, the sequences of each sector do not have particular features distinguishing them from the others. For this reason, it may not be concluded that responses about customer satisfaction in each sector depend on the order of the module. This indicates that no sequence has more low or high satisfaction responses in relation to other sequences. For the water sector – whose results are reported in graphic 32 – it is shown that satisfaction tends to be high in all sequences. Regarding satisfaction in the 8th and 10th 110 places there are more differences in the first wave – the first sequence having more responses in the first case and the second sequence having fewer responses in the second case. Moreover, in the second wave it is shown that the third sequence had fewer satisfaction responses in the 7th place, in relation to other sequences. Notwithstanding the foregoing, the behavior towards a certain satisfaction level in each sequence are not regular in both waves of the survey. Graphic 32 Overall Satisfaction by Module Order – Water Sector (a) First wave (b) Second wave 40 45 35 40 30 35 First 30 25 First 25 20 Second Second 20 15 Third Third 15 10 10 5 5 0 0 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 Source: Own elaboration For the electric power sector, as for the water sector, customers showed high satisfaction in the three module sequences. However, it is shown that the customers interviewed in the first wave showed higher satisfaction levels than the customers interviewed in the second wave. Particularly, the third sequence had a low number of responses in the 7th place in the first wave. Thus, this sequence shows more satisfaction responses in the 8th and 10th places. As in the case of the water sector, in both waves of the survey there is no evidence that any of the sequences has more responses at a specific satisfaction level. 111 Graphic 33 Overall Satisfaction by Module Order – Electric Power Sector (a) First wave (b) Second wave 60 60 50 50 40 40 First First 30 Second 30 Second Third Third 20 20 10 10 0 0 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 Source: Own elaboration Moreover, the responses about the gas sector do not show a trend based on the module sequence in the survey. That is to say, all of them tend to show high satisfaction levels, and none of them are noteworthy, as is the case of the sectors analyzed above. The only aspect that is worth mentioning is that in the second wave the first sequence had more satisfaction responses in the 7th and 8th places and therefore fewer responses in the 9th place. Graphic 34 Overall Satisfaction by Module Order – Gas Sector (a) First wave (b) Second wave 60 60 50 50 40 40 First First 30 Second 30 Second Third Third 20 20 10 10 0 0 0 1 2 3 4 5 6 7 8 9 10 0 1 2 4 5 6 7 8 9 10 Source: Own elaboration In order to analyze if bad quality perception influences the quality of the other utilities, sector independence is studied. Thus, correlation matrices are shown below for both waves of the survey, determining the sign and the degree of relation between overall satisfaction levels for each sector. Besides, the degree of relation between sectors is also tested and the specifics about service quality of each sector are included. 112 Chart 40 Overall Satisfaction Correlation Matrix (a) First wave (b) Second wave Source: Own elaboration It is shown that sectors are positively related, the closest relation being between the gas and water sectors (0.41 and 0.29), and between the gas and electric power sectors (0.42 and 0.32). However, the relation between the electric power and the water sector is not very close (0.24 and 0.19). It should be highlighted that the correlation coefficients obtained in the second wave of the survey are lower than those obtained in the first wave. That is to say, customer satisfaction with the service in each sector is more independent in the second wave than in the first. The analysis is completed with an estimate of a set of econometric models including the quality reported regarding the other utilities within the explanatory variables of each service overall satisfaction. A. Water The tested model is the following: st a = α + β 1 ∗ sac i + β 2 ∗ paa i + β 3 ∗ ia i + β 4 ∗ moi + β 5 ∗ mc i + β 6 ∗ ms i + β 7 ∗ sat g + β 8 ∗ sat e + µ This model is used to check if overall satisfaction with electric power and gas services, including the perception variables for the water service quality, are statistically significant to explain overall satisfaction with the water service. The results of the first wave (61.1) indicate that overall satisfaction with the water service is related with overall satisfaction with the gas service, water pressure and the bad smell of water. In the first two cases, the significance levels are high (99%), and in the latter they amount to 95%. 113 Particularly, the close relation between the gas and water sector is evidenced by the fact that a one-unit increase in gas satisfaction causes an average increase of 0.36 in water satisfaction. According to the results of the second wave (61.2), satisfaction with the electric power service is significant to explain the variations of satisfaction with the water service with a 90% probability, though its impact is quite low (0.08). In addition, the impact of satisfaction with the gas service on satisfaction with the water service decreases significantly (58%), as the coefficient is reduced from 0.36 to 0.15. Chart 41 Overall Satisfaction versus Overall Satisfaction with other Utilities and Service Quality and Technical Product Source: Own elaboration The estimates were made on the basis of 381 and 391 observations and the coefficients of determination (R-squared) amounted to 0.45 and 0.52, thereby indicating that the independent variables of the regression explain collectively between 45% and 52% of the variations of overall satisfaction with the water service. 114 B. Electric power The following model was estimated: st e = α + β1 ∗ iei + β 2 ∗ tai + β 3 ∗ sat a + β 4 ∗ sat g + µ This regression – whose results are reported in chart 42 – analyzes if overall satisfaction with the water and gas sectors, plus the perception variables of electric power service quality are relevant to explain the behavior of overall satisfaction with the electric power service. The results of the first wave indicate that it may be affirmed with a 99% probability that there is a relation among overall customer satisfaction with the gas service, outages and an adequate voltage. The close relation between the gas and electric power sectors is evidenced by the fact that a one-point increase in overall satisfaction with the gas service causes, on average, an increase of 0.39 in overall satisfaction with the electric power service. However, the dependence of customer satisfaction in these two sectors decreased by 48%. According to the results of the second wave, a one-unit increase in satisfaction with the gas service causes an average increase of 1/4 point (0.24) in satisfaction with the electric power service. Chart 42 Overall Satisfaction versus Overall Satisfaction with other Utilities and Service Quality and Technical Product Source: Own elaboration 115 The estimates were made on the basis of 381 and 391 observations and the coefficients of determination (R-squared) amounted to 0.28 and 0.38, indicating that the independent variables of the regression collectively explain 28% and 38% of the variations in overall satisfaction with the electric power service. C. Gas The following model was estimated: st g = α + β 1 ∗ pag i + β 2 ∗ sat a + β 3 ∗ sat e + µ In this regression – whose results are reported in chart 43 – the significance of overall satisfaction with the electric power and water sectors, plus the perception variables of the water service quality, are analyzed to explain the variation of overall satisfaction with the gas service. The results of the first wave indicate that overall satisfaction with the electric power sector and the water sector, as well as an adequate gas pressure influence the degree of overall satisfaction with the gas service, with a 99% probability in all cases. Thus, a one-point increase in overall satisfaction with the electric power service and water service causes an average increase of 0.28 and 0.17 in overall satisfaction with the gas service, respectively. However, in the second wave a slightest relation between the sectors is evidenced by the decrease of the satisfaction coefficients for the electric power and gas sectors. That is to say, a one-unit variation in satisfaction with the electric power and water sectors causes, on average, an increase of 0.22 and 0.17 in gas satisfaction. 116 Chart 43 Overall Satisfaction versus Overall Satisfaction with other Utilities and Service Quality and Technical Product Source: Own elaboration The estimates were made on the basis of 381 y 391 observations and the coefficients of determination (R=squared) amounted to 0.38 and 0.40, thereby indicating that the independent variables for the regression collectively explain 38 and 40% of overall satisfaction variations. 117 Conclusions and Recommendations In this study, basically three tasks have been developed and from each of them some recommendations for regulators have been formulated. On the one hand, it was proposed that the literature associated with quality be revised from the point of view of demand, i.e. focusing on the issue from the customers’ point of view. The main merit of this task is that it was proposed that regulation be supported by quality assessments from the point of view of both supply and costumers. The traditional view is based on objective quality measures related with supply performance indicators. This kind of analysis underestimates customer perception as a determinant of their willingness to pay, which has a long-term impact. Moreover, this part provided an academic point of view on quality as it highlighted several possible interpretations of quality and described the proposed models for its analysis. A simple conclusion that may be drawn from this literature is that there is no current consensus as to how to link expectations with quality perception and, therefore, there is still room for further research. In this study, a SERVPERF methodology was used, since it was deemed more adequate than the SERVQUAL model. The second task assigned was to design a survey methodology supported by the literature on quality and, of course, by the best international practices on field measurements. From this part, some recommendations of interest for regulators and researchers may also be formulated. One of these recommendations suggests that samples should be designed on the basis of a socioeconomic variable since perception may vary depending on this condition due to the possibility of using substitutes, the market basket composition, among other factors. If the sample is stratified, proportional allocation should be used to simplify future database use by others. 118 Another recommendation related to the sample has been formulated. The sample should be as large and widely spread as possible depending on the available financing, whereas the unit of record, i.e. the unit of analysis, should be determined in agreement with the utility meter used to determine the consumed amount to be billed. Additionally, it is suggested that the person who is to answer the questionnaire within the unit of record, particularly if it is self-administered, should be determined following a simple procedure and in accordance with the customs and usages of the population to be surveyed. Finally, the following recommendations related to questionnaire design and processing should be added:  If the database published contains blank fields, an explanation of such fields should be included.  The database should state how to enter “Don’t know� responses.  The answers to opinion questions should have the same scale ranging from “totally disagree� to “totally agree�. The score for this scale: five, seven, nine, eleven, should be determined according to the customs and usages of the population to be surveyed.  If the questionnaire is self-administered, the scale should exclude the “Don’t know� option to encourage as many responses as possible.  The questionnaire should avoid open-ended questions.  As regards trial questions, in order to prepare a set of preprinted answers, several pilot surveys should be used until the list is satisfactorily fine- tuned. In these cases, the “Others. Please, specify.� option should be added. Finally, the purpose of the third task developed was to perform an empirical study, i.e. a field assessment to analyze the model and draw conclusions from the data gathered. 119 In this regard, the experience was successful as it was proven that the methodology used was effective and could provide an insight of the quality issue from the point of view of demand. As a result of this successful task, we may suggest that surveying should be a systematic task, i.e., surveys should be made periodically. Said surveys provide useful information both for design and control of the utilities quality levels. For example, many variables show a tolerance level as regards the required quality level – for example, a service interruption within a six-month term does not affect overall quality perception. During this study, periodicity was limited by the initial project schedule, which was not necessarily meant to coincide with the ideal field schedule. Two surveys on the row showed low variation and statistically served as one. More research should be done to learn about ideal periodicity, though there should probably be an interval of six months between each survey so as to capture seasonal variations. From this part, it is also concluded that utilities surveys should be performed jointly. The overlapping scope of the surveys for the different utilities and their results show that in principle its results do not depend on the order or the quality of other utilities. From the study, it may be clearly inferred that information on the service quality actually offered by the utility companies should be attached to the surveys (service interruptions, low pressure events, etc.) so as to determine if there are any differences between the “objective� quality and customer perception. Ideally, the customer number of each interviewee should be available so as to obtain the consumption and quality records from the utility companies, so that the analysis might include more objective variables in the customer consumption pattern. Last but not least, we have concluded that, in order to include socioeconomic features in the analysis more attention should be paid to the characteristics of the dwelling and/or neighborhood, as the responses on income are not reliable enough to base the population stratification on them. 120 Annex 1. Flowchart and Steps followed by Interviewers Surveying procedures for an SP Start First address/ring bell of SP K=0 End Draw line on IS What’s the value Revisit questionnaires of K? DA, HA and IA Take questionnaire # Following address/ Take house questionnaire number # IS row marked with yes @? Eligible address/ Record ANE on house yes IS number? one How many More dwellings? than one Dwelling selection sheet Ring bell yes Answer the Keep Record DA on IS bell? questionnaire yes Agrees to be Record RD on IS interviewed? one More How manythan one households? Household selection sheet Eligible Record HNE on yes household? Record HA on IS yes and Anybody questionnaire present? Record IA on IS yes and Interviewee questionnaire present? yes Agrees to be Record RI on IS interviewed ? Next address Keep Record EI on IS and Perform the questionnaire questionnaire interview 121 Steps to be followed by Interviewers (Dwelling and Household Selection) First, interviewers will start visiting a SP with the following tools and elements: • A map of the SP; • Several numbered itinerary sheets (IS) (see Flowchart); • Dwelling selection sheet for the SP (see Flowchart); • Household selection sheet for the SP (see Flowchart). Interviewers will visit the blocks in the order stated on the SP map. The direction in which they should walk will be previously indicated. Then, the interviewers will start walking from the intersections marked with a dot on the diagram in the direction shown by the arrow in each case. The IS sheets contain 50 rows, 10 of which will be marked with an at sign (@).This character will be independently and randomly allocated to the rows on each page of the IS at the office. Moreover, the dwelling and household selection sheets will be specific for each SP and will also be completed randomly at the office. The questionnaires for the interviews will include three set of questions, plus the classification data, one set for each issue analyzed in this study. According to the foregoing, the questionnaires will be prepared considering the six possible orders for the set of questions, for the purpose of eliminating any distortions that might cause the order of the questions. Thus, the orders of the five questionnaires for each interviewer’s SP will differ. These points having been straighten out; we may describe the procedures to be followed by each interviewer while performing their duties. They must start the SP itinerary assigned with the first questionnaire available, from the starting intersection of block 1. As they follow the SP itinerary, they must record the addresses on the IS, using the street name and house number stated by the municipality. 122 In the event the IS row is not marked with an at sign indicating the address of the household to be surveyed, interviewers will continue with the following address. However, if the IS row is marked with an at sign, they will have to determine whether it is eligible or not 30 following the instructions given. If the address is not eligible (ANE), the interviewer will indicate it on the IS and then will go on with the following address as stated in (a). If the address is eligible and there is only one household, interviewers will proceed. However, if the household is eligible and there is more than one household, the interviewer will select the household to be surveyed randomly using the dwelling selection sheet. The dwelling selected shall be the one whose number 31 matches the last digit of the questionnaire, and then interviewers will proceed. Once the dwelling has been selected, the interviewers will ring the bell. If there is no answer, they will do as follows: • Write dwelling absent (DA) and will record the number on the IS questionnaire; • Write DA on the margin, keep it and will continue the following questionnaire and the following address, as stated in (a). On the other hand, if they answer but refuse to be surveyed, then the interviewers will write rejected by dwelling (RD) on the IS and will go on with the following address, according to (a). If they answer and there is only one household, then the interviewers will go on with the procedure. However, if they answer but there is more than one household, the interviewers will select the household to be surveyed using the household selection sheet. This mechanism, as well as the dwelling selection mechanism, is based on the number of 30 It was deemed convenient that dwellings containing collective households (e.g. hotels, pensions, convents, quarters, etc.), public buildings (schools, hospitals, police stations, etc.), stores, office buildings, warehouses, buildings under construction, , etcetera are not eligible. 31 Usually, the janitor’s home is not considered a dwelling. 123 households existing on the dwelling and the last digit of the questionnaire. Then, the interviewers will proceed as stated in (j). 32 Once the household has been selected, the interviewers must determine if it is eligible.33 If the household is not eligible, they will write not eligible (HNE) on the IS and will continue with the following address, according to (a). Notwithstanding, if the household is eligible but all its members are absent, the interviewer will perform the following procedure: • Write household absent (HA) and record the questionnaire number on the IS; • Record HA on the margin, keep it and will continue with the following address according to (a). Moreover, if the household is eligible and there is at least one member present, but they refused to be surveyed, interviewers will write rejected by household (RH) and record the questionnaire number on the IS, then they will continue with the following address, according to (a). If the household is eligible, there is at least one member present and they agree to be surveyed the interviewers will determine if there is at least one eligible member, as previously instructed. In such cases where the person to be interviewed is absent, the interviewers will follow this procedure: • Write interviewee absent (IA) and record the questionnaire number on the IS; • Record IA on the questionnaire margin, keep it, and continue with the following questionnaire and the next address, as stated in (a). When the interviewee is present but refuses to be interviewed, the interviewers must write rejected by interviewee (RI) on the IS and continue with the following address, as 32 Housheold selection must be performed only if all households contribute to the payment of utilities. If only one household must pay the bills, then this is the household to be interviewed. 33 Usually a household is deemed as not eligible if there are reasons that prevent the interview, e.g., if the members do not speak Spanish, are older than the age determined in the instructions, etcetera. 124 stated in (a). Finally, if the interviewee is present and agrees to be interviewed, the interviewer may start the interview, as follows: • Write effective interviewee (EI) and record the questionnaire number on the IS; • Record EI on the margin, keep it, and will continue with the following questionnaire and the next address, as stated in (a). Once the 5 questionnaires have been completed, interviewers will draw a line on the IS. The final phase will be completed once the five questionnaires for the SP bear on their margins the following: DA, HA, IA, EI. To conclude the tasks, interviewers will take the questionnaires marked DA, HA and IA and will revisit such addresses on the dates specified and as many times as instructed. If any questionnaires that do not bear EI on its margin remain, once the instructions stated above have been fulfilled, the interviewers will answer the questions stated there. 125 Bibliography Bañon i Martinez, Rafael and Ernesto Carrillo Barroso (1996): “Evaluación de la calidad de los servicios públicos� [Public Utilities Quality Assessment], en Rafael Bañon i Martinez (director): La productividad y la calidad en la gestión pública [Public Administration Productivity and Quality], Xunta de Galicia, Escola Galega de Administración Pública (EGAP), Santiago de Compostela, page 171. Bateson, J. (1985): “Self-Service consumer: An exploratory study�, Journal of retailing, Vol. 61, No. 3, pgs. 46-76. Beltráan Villalva, M. (1996): “De la reforma de la administración al control de calidad de los servicios públicos� [From Administration Reforms to Public Utilities Quality Control], Instituto Nacional de Administración Pública [National Institute of Public Administration], Gestión y Análisis de Políticas Públicas [Public Policy Management and Analysis] No. 5-6, pgs. 5-16. Berné, C.; Múgica J. M. y Yagüe M. J. (1996): “La gestión estratégica y los conceptos de calidad percibida, satisfacción del cliente y lealtad� [Strategic management, perceived quality, customer satisfaction and customer loyalty] Economía Industrial, nº 307, January -February, pgs. 63-74. Bitner, M. (1990): “Evaluating service encounters: The effects of physical surroundings and employee responses�, Journal of Marketing, Vol. 54 Issue 2, pgs. 69-82. Bolton, R. and Drew, J. (1991): “A multistage model of customer: Assessments of service quality and value�, Journal of Consumer Research, Vol. 17 Issue 4, p. 375. Cadotte, E., Woodruff, R. and Jenkins, R “(1987): “Expectations and norms in models of consumer expectations�, Journal of Marketing Research 24, pgs. 305 - 314. Carman, J. (1990): “Consumer perceptions of service quality: An assessment of the SERVQUAL dimensions, Journal of Retailing, Vol. 66, pgs 33-55. Cronin, J. and Taylor, S. (1992): “Measuring service quality: A reexamination and extension�, Journal of Marketing, Vol. 56, pgs. 55-68. Cronin, J. and Taylor, S. (1994): “SERVPERF versus SERVQUAL. Reconciling performance-based and perceptions-minus-expectations�, Journal of Marketing, Vol. 58 Issue 1, p. 125. Dabholkar, P., Shepherd, C. and Thorpe, D. (2000): “A comprehensive framework for service quality: An investigation of critical conceptual and measurement issues through a longitudinal study�, Journal of Retailing, Vol. 76, No. 2, pgs 139-173. 126 Devlin, J., Gwynne, A. and Ennew, C (2002): “The antecedents of service expectations�, Service Industries Journal, Vol. 22 Issue 4, pgs. 117-136. Grönroos, C. (1984): “A service quality model and its marketing implications�, European Journal of Marketing, Vol. 18 No. 4, pgs. 37-44. Gustafsson, A., Johnson, M. and Roos, I. (2005): “The effects of customer satisfaction, relationship commitment dimensions, and triggers on customer retention�, Journal of Marketing, Vol. 69 Issue 4, pgs. 210-218. Johnson, M., Fornell, C. (1991): “A framework for comparing customer satisfaction across individuals and product categories�, Journal of economic psychology, Vol. 12, pgs 267-286. Kano, N., Seraku, N., Takahashi, F. and Tsuji, S. (1996): “Attractive quality and must-be quality. In The best on quality, edited by John D. Hromi, Vol. 7, Milwaukee: ASQC Quality Press. Langeard, E., Bateson, J., Lovelock, C. and Eiglier, P. (1981): “Services marketing: New insights from consumers and managers�, Marketing science institute report, No. 81-104. Latu, T. and Everett, A. (2000): “Review of satisfaction research and measurement approaches�, New Zeland Department of Conservation - DOC Science Publishing. Meuter, M. and Bitner, M. (1997): “The new service encounter: customer usage and satisfaction with serf-service technologies�, Vanderbitt University Press. Oliver, R. (1980): “A cognitive model of antecedents and consequences of satisfaction decisions�, Journal of Marketing Research, Vol. 17, pgs. 460-469. Oliver, R. (1994): “Conceptual issues in the structural analysis of consumption emotion, satisfaction, and quality: Evidence in a service setting�, Advances in Consumer Research, Vol. 21 Issue 1, pgs. 16-22. Parasuraman, A., Zeithaml, V. and Berry, L. (1985): “A conceptual model of service quality and its implications for future research�, Journal of Marketing, Vol. 49 Issue 4, pgs. 41-50. Peter, J., Churchill, G. and Brown, T. (1993): “Caution in the use of difference scores in consumer research�, Journal of Consumer Research, Vol 19 No. 4. Reicheld, F. and Sasser, W. (1990): “Zero defections: Quality comes to services�, Harvard Business Review (68:5), pgs 2-9. 127 Rust, R. and Oliver, R. (1994): “Service quality: Insights and managerial implications from the frontier�, en Service Quality: New Directions in Theory and Practice, Rust y Oliver eds., Sage Publications. Selnes, F. (1993): “An examination of the effect of product performance on brand reputation, satisfaction and royalty�, European Journal of Marketing, Vol. 27 (9), pgs. 19-35. Teas, R (1994): “Expectations as a comparison standard in measuring service quality: An assessment of a reassessment�, Journal of Marketing, Vol. 58 Issue 1, p. 132. Westbrook, R. and Oliver, R. (1984): “The dimensions of consumption emotion patterns and consumer satisfaction�, Journal of Consumer Research; Vol. 18 Issue 1, p. 84. Woodruff, R., Cadotte, E. and Jenkins, R. (1983): “Modeling consumer satisfaction processes using experience-based norms�, Journal of Marketing Research, Vol. 20 Issue 3, pgs. 296-304. Yi, Y. (1993): “The determinants of consumer satisfaction: The moderating role of ambiguity�, Advances in Consumer Research, Vol. 20 Issue 1, pgs. 502-506. Zuluaga, L (2002): “Public services user satisfaction: An application to higher education�, University of Barcelona, Centre for Research in Welfare Economy. 128