Quality-of-life indicators at different scales: Theoretical background

Quality-of-life indicators at different scales: Theoretical background

ecological indicators 8 (2008) 854–862 available at www.sciencedirect.com journal homepage: www.elsevier.com/locate/ecolind Quality-of-life indicat...

333KB Sizes 40 Downloads 108 Views

ecological indicators 8 (2008) 854–862

available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/ecolind

Quality-of-life indicators at different scales: Theoretical background Irina G. Malkina-Pykh *, Yuri A. Pykh Research Center for Interdisciplinary Environmental Cooperation INENCO, Russian Academy of Sciences, St.-Petersburg 191187, nab. Kutuzova 14, Russia

article info

abstract

Article history:

The article is presenting the general analysis of the systems approach and model

Received 18 April 2006

approaches for the development of QoL indicators and indices. In our study we propose

Received in revised form

the method of response function as a method of the construction of purposeful, credible

27 January 2007

integrated models from data and prior knowledge or information. The method of response

Accepted 30 January 2007

function implies credible models in the sense that they are identifiable, and, hopefully, explains system output behaviour satisfactorily.Using response function method for the development of QoL models, we are able to obtain QoL indices as the direct output of the

Keywords:

models. # 2007 Elsevier Ltd. All rights reserved.

Quality of life Indices Method of response function

1.

Introduction

One of the most important political and societal problems of nowadays is how to improve and secure the quality of life of mankind while living within the carrying capacity of the environment and without compromising the long-term human, economical, and ecological capital of the future. The term quality of life (QoL) has been widely used in a number of disciplines to express the idea of personal wellbeing in a framework, which goes beyond the simple economist equation of well-being with income. Quality of life is generally used as the overarching concept, which encompasses income (and therefore consumption) but also includes other factors, which contribute to well-being such as satisfaction with the living environment or a greater sense of happiness or joy (Jacobs, http://www.compenvironment. lancs.ac.uk/sociology/esf/papers.htm). There are two related but separate concepts: individual QoL and social QoL. This is particularly important in relation to

envirtonmental goods. Some environmental goods and costs directly affect individual QoL—air quality, for example, or traffic congestion. But many do not. Natural habitats do not make me better off personally, nor does reducing the risk of global warming to future generations. These contribute rather to the health or quality of society. The same is true for of many social or shared goods, including cultural goods which many people do not use themselves, such as universities and public service broadcasting. Of course social QoL contributes to individual QoL: (some) individuals feel better off when they live in a better society. But this is not the justification for pursuing social QoL. They are logically separate. Conversely, individual QoL should contribute to social QoL: a society would not be very good or healthy if its natural habitats were preserved and inequality eradicated but its people were all stressed at work and going through divorce. If people feel that social QoL contributes to their own personal QoL this indicates a self-identification with, or feeling of membership of, society. Politically this

* Corresponding author. E-mail addresses: [email protected], [email protected] (I.G. Malkina-Pykh). 1470-160X/$ – see front matter # 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.ecolind.2007.01.008

ecological indicators 8 (2008) 854–862

would appear to be an important prerequisite for defending social goods whose contribution is to social QoL. The starting place for most QoL studies has been the subjective experience of well-being of the individual. However the attempt to measure this has involved an inexorable slide towards a non-individual perspective. People’s subjective perceptions of their well-being are so clearly non-comparable, and affected by expectation and social comparison, that attention quickly turned to the identification of objective conditions which influence subjective experience: people’s objective state of health, for example, rather than their feelings of wellness. But many of these objective conditions are not (or cannot be measured as) peculiar to the individual at all. The quality of air, the level of education or indeed the level of employment, all requires collective or aggregate measurement. So the quality of life gradually became, for many researchers, a description of the collectively experienced conditions of a society or place, with only an indirect and contingent relationship to the subjective experience of wellbeing of individuals. The simplest definition of individual QoL is the subjective feeling that one’s life overall is going well. Note that this differentiates QoL from ‘‘happiness’’, which tends to connote too transitory and emotional a condition. ‘‘Overall’’ is intended to define QoL as the overarching judgement of how all the different elements of one’s life combine together. There are three problems with this definition, however. The first is that it can only be measured subjectively, by asking people about their own QoL. This raises all the familiar problems of subjective measurement, its reliability and comparability. The second is that QoL in this definition relies heavily on the character and dispositions of the individual. A person may be rich, successful in their job, healthy and happily married and still not feel their life is going well, perhaps because they have unfulfilled personal goals or simply because they have a depressive personality. If we say such a person does not have a good QoL, as we will have to on this definition, the concept becomes more or less meaningless in terms of public policy and research. The third problem is the converse of this. Subjective satisfaction with one’s life is strongly related to one’s expectations of it. Expectations in turn are related to social position: people compare themselves to others in their selfperceived social position. Low expectations achieved lead to higher subjective reporting of QoL than high achievement that fails to meet expectations. This leads to the apparent conclusion that one way to increase QoL is to reduce people’s expectations. Yet this fails to account for the desirability of personal growth and development, of the accomplishment of challenging individual life goals. These problems suggest a definition of QoL not in terms of overall subjective experience, but as a set of conditions relating to an individual’s life that would appear to indicate, from outside, that it is going well. This definition accepts that it may not, in fact, capture the subjective perception of overall well-being, but makes a generalized claim that – if these conditions obtain – in most cases it will. Quality of life is a complex concept that is difficult to operationalise. Nevertheless, it is possible to establish one

855

principal characteristic: its multidimensionality. Like life itself, quality of life has multiple ingredients. International research into the quality of life tends to divide life into a number of domains, which are then studied separately. These domains might be physical, psychological and social; physical and mental health; emotional and cognitive dimensions, for example, happiness and satisfaction with life; the ability to function bodily, sexually, socially and occupationally; objective status in terms of finances, working conditions, family conditions, etc. The subject answers a number of questions on how well he or she is doing in these various aspects of life. The responses are scored, weighted and combined in various ways, thus giving us a quality-of-life rating scale. Researchers seldom stop to contemplate why certain life domains are included while others are not. Establishing what domains are and are not relevant presupposes an overall theory of what the quality of life is, what makes a good life and what life is all about. The lack of a theoretical framework is a decisive weakness in much of the empirical research on the quality of life. Many of the leading researchers within the field have pointed this out. If practical research is to move forwards, it will have to rest on a sound theory: nothing is as practical as a good theory.

2.

Quality-of-life indicators and indices

One of the possibilities to operationalise the QoL concept is to design QoL indicators and indices. The last 30 years have seen a great many attempts to measure quality of life (QoL) in many parts of the world (Hagerty et al., 2001). Various indices of QoL have been proposed by public policy institutes, government agencies, and news media. However, the advantages and liabilities of each have not been systematically evaluated. Many QoL scholars point to the need for both objective and subjective indicators (Cummins, 1996, 1997; Cummins et al., 1994; Firat and Karafakioglu, 1990; Samli, 1995). Typically, measuring QoL overall or within a specific life domain (at any level of analysis) has been done through either subjective or objective indicators. Subjective indicators are mostly based on psychological responses, such as life satisfaction, job satisfaction, and personal happiness, among others. Objective indicators are measures based on frequency or physical quantity. Examples include standard of living, physical health status, and personal income, among others. Twenty-two of the most-used QoL indices from around the world were reviewed using 14 criteria for determining the validity and usefulness of such QoL indices to public policy (Hagerty et al., 2001). The authors collected QoL indices with any of three criteria: They received attention from researchers in the field, they received attention from the press, or they had public policy applications. The list of the reviewed indices includes: CDC’s Health-Related Quality of Life, WHOQOL, Consumer Confidence Index (CCI), Money’s ‘‘Best Places’’, Index of Economic Well-Being (IEWB), Genuine Progress Index (GPI), American Demographics Index of Well-Being, Johnston’s QOL Index, Eurobarometer, Veenhoven’s Happy Life-Expectancy Scale (HLE), International Living Index, U.N. Human Development Index (HDI), Miringoffs’ Index of Social Health,

856

ecological indicators 8 (2008) 854–862

State-Level QOL Surveys, Estes’ Index of Social Progress (ISP), Diener’s Basic and Advanced QOL Indexes, Cummins’ Comprehensive Quality of Life Scale (ComQol), Michalos’ North American Social Report, Philippines’ Weather Station, Netherlands Living Conditions Index (LCI), German System of Social Indicators, Swedish ULF System. A few indices were not reviewed because they were not applicable to public policy or were published too late to review (Calvert–Henderson QOL by Henderson et al., 2000). It was established that QoL indices should satisfy the following requirements: 1. The index must have a clear practical purpose, i.e., a public policy purpose. 2. The index should help public policymakers develop and assess programs at all levels of aggregation. This begins at the individual level of aggregation (e.g., physicians and counselors helping individuals in need) and continues to the family or household level (e.g., social workers helping families in need), community level (e.g., town governments developing policies and programs that can enhance community QoL), state (or province) level (e.g., state bodies developing policies and programs that can assist residents of the entire state or province), the country level (e.g., national agencies developing policies and programs that can assist citizens of that country), and the international level (e.g., international agencies developing policies and programs that can assist the world’s citizens and the planet at large). 3. The index should be based on time series to allow periodic monitoring and control. This is crucial for public policy in order to assess whether conditions are improving for targeted populations and to forecast future conditions. 4. The index should be grounded in well-established theory. By ‘‘theory’’ it means the ‘‘nomological net’’ of concepts and causal paths that specify how QoL is related to exogenous and endogenous variables. By ‘‘well established’’, it means that its parts have been subjected to empirical test. In particular for public policy applications, the paths and mediating variables by which policy variables will affect different domains of QOL must be specified so that policy-makers can predict the effects of new programs. 5. The components of the index should be reliable, valid, and sensitive. As in any measurement system, components must be shown to be reliable and valid. By ‘‘sensitive’’ it means that the index should show changes in response to public policy inputs. 6. The index should be reported as a single number, but can be broken down into components. 7. The domains in aggregate must encompass the totality of life experience. QoL is a term that implies the quality of a person’s whole life, not just some component part. It therefore follows that if QoL is to be segmented into its component domains, those domains in aggregate must represent the total construct. 8. Each domain must encompass a substantial but discrete portion of the QoL construct. The number of possible

9. 10.

11.

12. 13.

14.

domains is infinite if one regards each aspect of life as a putative domain, so parsimony is essential in order to define a small number of domains that fulfill the requirements of 7 above. Each domain must have the potential to be measured in both objective and subjective dimensions. Each domain within a generic QoL instrument must have relevance for most people. QoL measures designed with a specific target population in mind, in a specific social context, may not capture the totality of life experience for other populations in different contexts and settings. Hence, the validity of a generic measure of QoL has to be demonstrated across a variety of populations in different contexts. If a specific domain is proposed for a non-generic instrument, it must be demonstrated to contribute unique variance to the QoL construct beyond the generic domains for the target group. Domains must be potentially neutral, positive, or negative in their contribution to the QoL construct. Domains differ from the dimensions of personality (e.g., extraversion, self-esteem), cognitive processes (e.g., cognitive dissonance), and affect (e.g., joy) in that they cannot be measured objectively. This criterion is related selfevidentially to criterion 9. It also acts to separate the cognitive and affective processes that lead to subjective QoL from QoL as an outcome variable. The subjective dimension of each domain has both a cognitive and an affective component. They are measured by questions concerning ‘‘satisfaction’’. It is widely considered that the perception of QoL is a result of multiple comparative processes. These processes compute the ‘‘gap’’ or discrepancy between one’s perceived current circumstance and imagined other circumstances that may refer to other people, the past, etc. The response to questions in terms of satisfaction are considered the most parsimonious measure of such discrepancies and therefore of QoL (Cummins, 1997).

We added the results by Hagerty et al. (2001) with the similar analysis of the recent Calvert–Henderson QOL (Henderson et al., 2000; Flynn et al., 2002), using the same criteria. Our results didn’t change in any way the results presented by Hagerty et al. (2001). The general conclusion shows that the indices (on average) achieved some criteria quite well, but that they failed to incorporate other criteria. The criteria that were achieved best were: (1) The index has a clear practical purpose for public policy, (2) the index should be based on time series to allow periodic monitoring and control, and (3) the index should be reported as a single number, but can be broken down into components. In contrast, the criteria that were not well achieved were: (1) The domains must encompass a substantial, but discrete, portion of QOL, (2) the index must be grounded in well-established theory, and (3) the composite index should be reliable, valid, and sensitive. 1. Domains must encompass a substantial, but discrete, portion of QOL

ecological indicators 8 (2008) 854–862

QOL instruments can be designed at four levels, as follows: (a) No domains. There are two versions of such scales. The first and simplest is the single question, ‘‘How satisfied are you with your life as a whole?’’ (Andrews and Withey, 1976). The second is those scales with multiple questions that are intended to measure the single latent construct of ‘‘life as a whole’’. The advantage of such scales is their simplicity and brevity. Their disadvantage is that they provide no comparative information on the component parts of the QoL experience and they can only be constructed to measure the subjective domain. (b) Single domain. Here the intention is to generate a multiple-item scale that measures a latent construct envisaged as a single QoL domain. Such scales may be either objective or subjective. (c) Multiple, intentionally overlapping domains. These are the diagnostic instruments that measure the details of life within some narrow specific context. They may be either objective or subjective. (d) Multiple, maximally independent domains. The construction of such instruments poses the greatest challenge of test developers in this area. In addition to the usual psychometric requirements of reliability, validity, and sensitivity, such instruments must be maximally parsimonious while also encompassing the entire QoL construct. So, one way to judge the adequacy of such instruments is to seek the ones with the smallest number of domains and then ask whether those domains are adequate, in aggregate, to define QoL. The optimal number and character of domains in this context is not yet fixed. Hagerty et al. (2001) concluded that seven domains can adequately span the space of perceived QOL: relationships with family and friends, emotional wellbeing, material well-being, health, work and productive activity, feeling part of one’s local community, and personal safety. 2. The composite index should be reliable, valid, and sensitive. Most of the indices have been well tested for reliability. The components that are objective measures (e.g., GDP/ person) have long been judged reliable, while the components that are subjective measures (e.g., satisfaction) have shown acceptable validity in this review. Further, most of the indices have demonstrated some convergent validity (they correlate with other measures of the same concept). However, all of the indices fall short on sensitivity. By ‘‘sensitivity’’, it means their ability to predict future ‘‘outputs of the system’’, sometimes termed predictive validity. A QoL index for public policy should be subjected to predictive validity tests, especially validating how the inputs of public policy affect the outputs. 3. The index should weight domains appropriately.

857

QoL indicators can be presented individually or they can be mathematically aggregated in some fashion to form of ‘‘QoL index’’. The methodology of indicators aggregation into an index is crucial for the question of indices development. The methodology underlying indices development is valuable because it gives greater credibility to indices application. In general, if the index is designed properly it can present a simplified picture of the state of the QoL, but broader vision for decision-makers, then the single indicator or even the set of indicators. The two problems which then arise are (Shults and Beauchamp, 1972): scaling (normalisation) the factors or components down to a single scale and rendering the units of measurement dimensionless; and weighing the various factors or components of the indices. Various scaling methods exist. The most widely accepted method is to use a simple mathematical function in order to translate the values of all components of an index to values between 0 and 1. Weighing provides a method with which to arrive at a ranking order of importance for the index components. An important topic related to reliability and validity of QoL indices is how to define importance weights for each domain. Such weighting would be used in the computation of a weighted average of the domains to produce a final composite score for QoL. Not surprisingly, weighting domains in the computation of composite indicators is a debated topic. The need for a weighting system to attach importance to specific physical, social, and economic characteristics of human well-being has been recognised for many years. Despite this, little progress has been made in establishing such a weighting system. Even with the development of direct monitoring of quality of life via survey research, no definitive list of criteria has been developed yet, nor has a weighting system been proposed which could be combined with the ‘‘objective’’ indicators. Indeed, it is questionable whether such a definitive list of criteria can be produced for all quality-of-life studies given that perceptions of life quality and the forces which impinge upon them are in a state of constant flux and operate in different ways at the variety of scales which are studied under the umbrella of quality-of-life research. The main approach which has been used to derive weightings has been to consider ‘‘expert opinion’’ as a means of determining the list of criteria and their significance. The weighing mechanism can also be based on delphi techniques, multi-criteria analysis or public opinion polls (Hope and Parker, 1993; Hope et al., 1992). Some QOL indices do not weight and do not provide an explanation for this approach. However, no weighting is still a form of weighting—equal weighting. The general controversy about weighting is joined by some researchers who claim equal weighting is not far behind the performance of purportedly optimal weighting schemes (Andrews and Withey, 1976). Hagerty et al. (2001) proposed using two advanced statistical methods to improve weighting—two-stage factor analysis and conjoint analysis. Two-stage factor analysis is a structural equation model that could be used to better

858

ecological indicators 8 (2008) 854–862

gauge the relative importances of indicators to each domain and each domain to overall QoL. Conjoint analysis is a decompositional modeling technique based on the premise that humans evaluate objects based on the separate amounts of value provided by each attribute. If domains of subjective QoL could be kept to a set of 6–9 variables, a hybrid conjoint modeling approach could be employed to derive part-worths or relative importances for the variables and their respective levels (‘‘satisfied’’ – ‘‘tolerable’’ – ‘‘worse than tolerable’ ’– ‘‘unacceptable’’). Once the weighing procedure has been completed, the question arises as to how the weighted index components should be aggregated up to a single integrated index, or, if that is not feasible, up to an index with at most three dimensions. Ott (1978) provides a survey of all sorts of aggregation functions, but advocates the use of the nonlinear aggregation root-sum-power function. 4. The index must be grounded in well-established theory. Most of the indices reviewed failed to specify any wellestablished theory behind the index. Generally, indices that are currently used to report about the QoL are organised and valued without using any conceptual or integrated modelling framework. This implies that they do not yield information about linkages between causes and effects (vertical integration), nor do they address cross-linkages between various causes and various effects (horizontal integration). Since the existing indices performed poorly on these four criteria, we propose a solution that is a systems-theory approach to QoL. One of the main experiences in quality of life research thus far, is that the reductionistic approach based on aspect-compartment oriented research methods has failed in analysing adequately complex, multidisciplinary, largescale quality of life phenomena. A more promising way seems to be the holistic, integrated approach, based on a systemsoriented analysis, which concentrates on the interactions and feedback mechanisms between the different subsystems of cause-effect chains rather than focusing on each subsystem in isolation. Systems analysis approach with models as a primary tool is the most appropriate way to investigate QOL phenomena in general and its complex systems in particular. The object of systems analysis is not only to study the particular system structure and to classify and describe the entities (components) of the system, but also to understand the processes, interactions and feedback mechanisms within the system that generate changes in its dynamics and structure. In enables a synoptic approach that addresses the interdependencies between the cause-effect chains. Given the complexity of the system under consideration, and the relative ignorance about the basic processes and interactions that determine its dynamics, the systems approach can help to foster understanding of the causal relationships that are responsible for changes in the structure and dynamics of the system. Therefore, the systems approach seems to be an appropriate method to capture the complexity of the interrelationships between the various subsystems of the complex QoL system. A prerequisite for such a systems approach is inter- and multidisciplinary, based on the integration of

knowledge gleaned from a variety of scientific disciplines (Rotmans and de Vries, 1997). We use the following general definition of what is quality of life (http://www.kesgrave.suffolk.sch.uk/Curric/geog/QoL. htm): ‘‘Quality of life is a measure of how positively or negatively we perceive our lives—a measure of well-being.’’ This measure is affected by three main ‘‘environments QoLs’’ (explained below) and is a relative concept. In other words, each and every individual has a different perception of life, even within a group of people living in the same area of the world. Experiences and desires within the three environments QoLs affect it: built environment QoL, social environment QoL and economic environment QoL. The built QoL is essentially about where you live: your house, the surroundings and the facilities and amenities available to you. In addition, it encompasses the availability of infrastructure, such as electricity supplies, telephone lines, running water and sewerage systems, etc. Compare for instance, life in a rural area of Brazil and a town in the UK. The differences in the built environment QoL would be extreme, and would affect the quality of life in the areas markedly. The social environment QoL involves friends and family and the people you interact with. It also involves entertainment, health (in terms of actual physical health and the healthcare systems available to you) and personal education/ literacy levels. The economic environment QoL looks at money and how money is spent and the employment (or lack of it), that leads to the money being earned or created in the first place. It can also look at how the money is made, including the formal and informal economies. Thus, we see that the quality of life concept has the same components as the concept of sustainable development. From the other hand, quality of life (QoL) is, by definition, a subjective concept, dependent on cultural perspectives and values. Values, the attributes of our world that we believe are functionally important, morally good, or personally desirable, are derived from our individual perspectives. Our age, ethnicity, gender, socio-economic status, education, health, religion, occupation, etc shape our perspectives. These differences in personal experience lead us to different beliefs about what is important, good, or desirable. These beliefs, or values, determine not only what we believe makes good quality of life, but also what conditions represent a quality of life problem. In our research on QoL indices we use the framework proposed for SD indicators selection. It allows to combine objective and subjective measures (domains) of QoL. At the moment, four types of conceptual frameworks for sustainable development (SD) indicators selection are widely accepted: the ‘‘pressure-state-response’’ (PSR) framework used by the OECD (Hammond et al., 1995), ‘‘pressure-stateimpact-response’’ (PSIR) framework used by UNEP and RIVM (Hardi and Zdan, 1997), the ‘‘driving force-state-response’’ (DSR) framework adopted by the UN Commission on Sustainable Development (CSD) in 1995 as a tool for organising information on sustainable development and for developing, presenting and analysing indicators of sustainable development. In recent publication of European Environment

ecological indicators 8 (2008) 854–862

QoL Agency the new ‘‘driving force – pressure-state-impactresponse’’ (DPSIR) framework is presented (Smeets and Weterings, 1999). Fig. 1 displays the path analysis diagram of our approach. We combined the system theory structure of QoL concepts suggested by Hagerty et al. (2001) with PSIR (Hardi and Zdan, 1997) conceptual framework. The first column of the figure contains pressure variables, which denote exogenous environmental variables affecting citizens’ QoL. Common measures of this are ‘‘objective indicators’’ of QoL. Many of these indicators can be controlled by public policy to improve QoL and are much studied by policy analysts to learn how to improve them. The second column of the figure contains state variables, which describe the individual’s response to this environment (e.g., education achieved, marriage choice). These are also often measured as ‘‘objective indicators’’, but reflect the individual’s choice in response to the environment and to public policy. The third column contains impact variable, which is the result of pressure and state. Modelling is a principal tool – perhaps the primary tool – of systems approach for studying the behaviour of large, complex systems such as QoL phenomena. Thus we propose that the development of QoL indices should be combined with or based on formal, dynamic modelling. The co-evolution of indices and models will not only help to identify trouble spots in the ‘‘global system,’’ but can help to gauge extent and intensity, and time proper corrective actions. Models can help (1) to relate indicators to each other, (2) to analyse consequences of policies and changes, (3) to find critical aspects of the system (that could be useful indicators), (4) to make world views explicit, and (5) to put indicators in an interactive context (Rotmans and de Vries, 1997). One of the possible approach to modelling of QoL phenomena is the integrated modelling approach. Integration means capturing as much as possible of cause-effect relationships and describing them with an operator of transition, or ‘‘input–output’’ function. The theory of the method of response function and its application for environmental model’s design has been described in several articles and monograph (Malkina-Pykh and Pykh, 2003; Pykh and MalkinaPykh, 1994, 1997, 2000). In our study we propose the method of response functions as a method of the construction of purposeful, credible, integrated QoL model from data and prior knowledge or information. The data are usually time series observations of system inputs and outputs, and sometimes of internal states. The method of response

859

functions implies credible models in the sense that they are identifiable, and, hopefully, explains system output behaviour satisfactorily. Using response function method for the development of QoL models, we are able to obtain indices as the direct output of the models. It means that the problems of scaling, weighing and aggregation are solved automatically during the process of model’s parameters identification.

3. The method of response function for the development of QoL indices A fundamental system characteristic upon which the method of response functions (MRF) is constructed is that every complex system has a response to each possible combination of conditions of the local environment which impinges on it. We define these conditions as ‘‘environmental factors’’. In case of QoL phenomena ‘‘environmental’’ means economic, social, environmental, etc. Given the system, the magnitude (and kind) of its response depends on the levels of environmental factors at that instant in time. The resultant relationship between environmental factors and QoL system response is described with the response functions. By pressure factors, we mean the factors that directly affect processes or characteristics under study. Let us designate the pressure factors as vector x = (x1, x2, . . ., xn). It is clear that each xi has a real value, i.e. xi 2 R1 ; i ¼ 1; . . . ; n. The interval ; xmax Þ, limited with the maximal and minimal values xi ¼ ðxmin i i of the factor xi, is called the tolerant interval on the given factor. Then, the space of all pressure factors can be designated as a set Xn = (x1, x2, . . ., xn) where Xn consists of all kinds of sequences of the pressure factors (x1, x2, . . ., xn). Thus it is clear that Xn is a multidimensional cube in Euclidean opt in which the space Rn. The point (or the interval) xi characteristic under study reaches the maximal value is called the optimal point (interval or zone) on the given factor. Similar to the set of pressure factors, the characteristics and processes studied have their domains of definition. For simplicity let us restrict our consideration to one characteristic. In this case let I represent the interval on the real axis limited by minimum and maximum values of the characteristic under study. Then the function F, which maps the space Xn onto the interval I by associating each point (x1, x2, . . ., xn) of the space Xn with the number F(x1, . . ., xn) on the scale I (i.e. F: Xn ! I), can be called the response function of the system characteristic for the whole set of pressure ecological factors

Fig. 1 – PSIR structure of quality of life concepts and causes.

860

ecological indicators 8 (2008) 854–862

(x1, x2, . . ., xn) This definition does not imply a statistical or functional independence of the active factors. The main problem with this approach is the choice of the parametric form of the function F(x1, . . ., xn), because, in many instances, the form of the response function F(x1, . . ., xn) cannot be determined a priori. Thus the problem is usually divided into a set of subproblems, taking into account the definition of the partial response functions fi to every pressure factor xi. Then, by partial response function of the characteristic or the process we mean a function which depends on a single active factor, i.e. the function of a single variable fi(xi)—by generalised response function we mean a function F(x1, . . ., xn) which accounts for all pressure factors considered and presented as a combination of partial response functions fi(xi). Let us consider for simplicity one impacting factor and denote it as x (without index i—the number of the factor). The most popular types of partial response functions are as follows:   a xg f ðxÞ ¼ wðx þ dÞb exp c xmax  x

a

f ðxÞ ¼ w  bð1  expða  cxÞg Þ

( f ðxÞ ¼ w

b þ expða  cxÞg

0

d

i¼1

where n is the number of pressure factors under study, ai is a vector of parameters, the values of which we have to determine in the process of identification. We introduce also the additional restriction in the identification procedure: max f i ðai ; xi Þ ¼ 11:0 xi

(1)

(2)

)a

1

Fig. 2 represents the typical graphs of the basic partial response functions mentioned above. Various modifications of these functions are also widely used. Eq. (1) is the generalised Gamma function, (2) is the generalised Weibull function and (3) is the generalised logistic function. All of these functions are flexible and non-linear in the parameters; therefore, non-linear regression techniques are required to estimate the parameters. For example, function (1) is used do describe Now we propose to present the general response function in the form: n Y f i ðai ; xi Þ (4) Fðx1 ; . . . ; xn Þ ¼

(3)

where a; b; c; d; w; a; b; and g are non-negative parameters for evaluation.

(5)

It is evident that standardisation condition (5) gives us a possibility to compare the impact of different pressure factors on the process under study. It seems evident that the main objective of MRF is to describe the resulting changes of systems’ characteristic under study on the impact of changes of pressure factors (x1, x2, . . ., xn). Thus we enter the operator (designate it as Htt ) that determines the evolution (or reaction) of the system (characteristic) under consideration as resulting from changes (spatial or temporal) of pressure factors, i.e.: FðxðtÞÞ ¼ Htt ðFðxðtÞÞ

Fig. 2 – Main types of pSfartial response functions used in QoL researches.

(6)

ecological indicators 8 (2008) 854–862

After the selection of the forms of generalised response functions F(x1, . . ., xn) and operator Htt we solve the problem of the mutual evaluation of all parameters taken together. The estimation of parameters is provided for operator Htt using the whole data set of responses and combinations of pressure factors. The parameterisation of the generalised response functions of the model is such that the jointly dependent observations can be fitted with the existing computer programs for non-linear optimisation. This approach to parameter estimation allows us to consider the mutual influence of the factors on the system’s characteristic (general response) under study, as well as to reveal the information and relationships ‘‘hidden’’ in the time series observations. In considering this approach, we are stressing relatively datarich or data-intensive situations where time series measurements are available for the modelling exercise in question. The requirement for large amounts of data of kinds not now readily available is one major current problem of the MRF approach, although without these data the approach can be very useful when appropriate hypothetical structures are used for modelling and for stimulation of our mental processes. The MRF paradigm does, however, provide a useful framework as a guide in collection, organisation, and reporting of useful data. It appears that QoL research ecology, to progress far beyond its current stage of development, will have to become a much more data rich science, and that the problem of effectively handling of the useful and necessary data and of providing the very important integrative data summaries which will be required may be at least pertially met through the utilisation of this paradigm. This may seem restrictive but data-intensive problems are important to model for many reasons. They can be used more easily to assess the success of models, developed for different systems. They facilitate selection among alternative model families and model components, as well as generating insights into system behaviour. They motivate the development of modelling techniques including numerical algorithms which are robust or at least in tune with the nature of the problem being studied. They derive knowledge (e.g. on system dynamics and processes) or information (e.g. output data) that is transferable either to other system types or at least to similar systems where there is a shortage of data. Data-intensive models can be used to assess the information content in data and consequently provide a guide as to the maximum level of detail that should be built into models of analogous processes, where data may not exist. Through case studies they can indicate the predictive uncertainties to be expected for decision makers, and where best to concentrate the collection of additional information, and indeed research, in order to reduce uncertainties. In short, data-intensive modelling studies are needed to provide advances in knowledge of QoL phenomena, and in an increasingly data-rich world such studies will continue to have more direct relevance (Jakeman and Post, 1994). The MRF approach may be used either to build relatively narrow purpose but often successful models, or, as part of a wider strategy, to develop regional model structures which may or may not ultimately be used to infer a more generic model.

4.

861

Conclusions

The term quality of life (QoL) has been widely used in a number of disciplines to express the idea of personal well-being in a framework, which goes beyond the simple economist equation of well-being with income. There are thus two related but separate concepts: individual QoL and social QoL. International research into the quality of life tends to divide life into a number of domains, which are then studied separately. One of the main experiences in quality of life research thus far, is that the reductionistic approach based on aspectcompartment oriented research methods has failed in analysing adequately complex, multidisciplinary, large-scale quality of life phenomena. A more promising way seems to be the holistic, integrated approach, based on a systems-oriented analysis, which concentrates on the interactions and feedback mechanisms between the different subsystems of cause-effect chains rather than focusing on each subsystem in isolation. Systems analysis approach with models as a primary tool is the most appropriate way to investigate QOL phenomena in general and its complex systems in particular. Indicators and indices as a communication tool between scientists and decision-makers should be linked to QOL models. We propose the method of response functions as a method of the construction of purposeful, credible integrated models from data and prior knowledge or information.

references

Andrews, F.M., Withey, S.B., 1976. Social Indicators of Wellbeing: American’s Perceptions of Life Quality. Plenum Press, New York. Cummins, R.A., McCabe, M.P., Romeo, Y., Gullone, E., 1994. The comprehensive quality of life scale (ComQoL): instrument development and psychometric evaluation on college staff and students. Educ. Psychol. Meas. 54 (2), 372–382. Cummins, R.A., 1996. The Domains of Life Satisfaction: An Attempt to Order Chaos, vol. 38. Social Indicators Research, pp. 303–328. Cummins, R.A., 1997. Comprehensive Quality of Life Scale. Adult Manual, 5th ed. School of Psychology, Deakin University, Melbourne. Firat, A.F., Karafakioglu, A.S., 1990. Quality of life issues and marketing as a social change agent. In: Meadow, H.L., Sirgy, M.J. (Eds.), Quality-of-Life Studies in Marketing and Management. Virginia Tech, International Society for Quality-of-Life Studies, Blacksburg, VA, pp. 623–632. Flynn, P., Berry, D., Heintz, T., 2002. Sustainability and Quality of Life indicators: toward the integration of economic, social and QOLAL measures. Indicators: J. Social Health 1 (4), 23–36. Hagerty, M.R., Cummins, R.A., Ferriss, A.L., et al., 2001. Quality of life indexes for national policy: review and agenda for research. A report of the Committee for Societal QOL Indexes, ISQOLS. Hammond, A., Adriaanse, A., Rodenburg, E., Bryant, D., Woodward, R., 1995. Environmental Indicators: A Systematic Approach to Measuring and Reporting on

862

ecological indicators 8 (2008) 854–862

Environmental Policy Performance in the Context of Sustainable Development. World Resource Institute. Hardi, P., Zdan, T. (Eds.), 1997. Assessing Sustainable Development: Principles in Practice. IISD, Winnipeg. Henderson, H., Lickerman, J., Flynn, P., 2000. Calvert–Henderson Quality of Life Indicators: A New Tool for Assessing National Trends. Calvert Group, Bethesda, MD. Hope, C., Parker, J., 1993. Forum: sharpening the environmental debate. Energy Policy 11, 1075–1076. Hope, C., Parker, J., Peake, S., 1992. A pilot environmental index for the U.K. in the 1980s. Energy Policy 4, 335–343. Jakeman, A.J., Post, D.A., 1994. From data and theory to environmental model: the case of rainfall runoff. Environmetrics 5, 297–314. Malkina-Pykh, I.G., Pykh, Yu.A., 2003. Application of the method of response function for the development of the quality of life indices. In: Tiezzi, E., Brebbia, C.A., Uso, J.L. (Eds.), Ecosystems and Sustainable Development. IV. Advances in Ecological Studies, vol. 2. Computational Mechanics Publications (CMP), Southampton Boston, pp. 939–948. Ott, W.R., 1978. Environmental Indices: Theory and Practice. Ann Arbor Science Publ. Inc.. Pykh, Yu.A., Malkina-Pykh, I.G., 2000. The Method of Response Function in Ecology. WIT Press, Southampton Boston, p. 275.

Pykh, Yu.A., Malkina-Pykh, I.G., 1994. Environmental indicators and their applications (trends of activity and development). IIASA Working Paper WP-94–127. Pykh, Yu.A., Malkina-Pykh, I.G., 1997. Environmental indices: systems analysis approach and application. In: Uso, J.L., Patten, B.C., Brebbia, C.A. (Eds.), Ecosystems and Sustainable Development, Advances in Ecological Studies, vol. 1. Computational Mechanics Publications (CMP), Southampton Boston, pp. 156–165. Rotmans, J., de Vries, B., 1997. Perspectives on Global Change. The TARGET Approach. Cambridge University Press, Cambridge. Samli, A.C., 1995. QOL research and measurement: some parameters and dimensions. In: Meadow, H.L., Sirgy, M.J., Rahtz, D. (Eds.), Developments in Quality-of-Life Studies in Marketing, vol. 5. Academy of Marketing Science and the International Society for Quality-of-Life Studies, DeKalb, IL, pp. 173–180. Shults, W.D., Beauchamp, J.J., 1972. Statistically based air-quality indicators. In: Thomas, W.A. (Ed.), Indicators of Environmental Quality. Plenum Press, London, UK. Smeets, E., Weterings, R., 1999. Environmental indicators: typology and overview. Technical report no. 25.