AMARTYA SEN’S CAPABILITY APPROACH TO HUMAN DEVELOPMENT: AN EMPIRICAL OPERATIONALIZATION USING PARTIAL LEAST SQUARES MODELING

Dr. Nour Eddine Aguenane Assistant Professor, Department of Economics, Faculty of Law, Economics and Social Sciences, Ibn Zohr University of Agadir, Morocco. ...................................................................................................................... Manuscript Info Abstract ......................... ........................................................................ Manuscript History Received: 23 November 2019 Final Accepted: 25 December 2019 Published: January 2020

From the point of view of the capability approach, human development is defined as a process of expanding human capabilities. In other words, well-being must be assessed by the different things that a person can aspire to be or to do or what Amartya Sen calls functionings. The set of functionings vectors that are possible for the individual represents his capability. The latter is, therefore, a substantial freedom which indicates that an individual is free to lead the kind of life he values. Therefore, this paper aims to operationalize these concepts by modeling the capabilities as latent constructs and the functionings as manifest variables. The estimation of the hierarchical model built using the partial least squares approach made it possible to calculate dimensional composite indicators reflecting the level of five human capabilities retained as fundamental dimensions of human development; namely, health, education, shelter, employment and mobility and communication. An overall composite indicator reflecting the level of human development measured in terms of capabilities is also estimated at the level of the sixty sample countries. This paper explores the structural relationships between capabilities, classifies countries according to the level of capabilities and compares the index obtained with the human development index and the gross domestic product per capita.

…………………………………………………………………………………………………….... Introduction:-
"Human development is about enlarging freedoms so that all human beings can pursue choices that they value" (UNDP, 2016). This definition of the United Nations Development Programme places the freedom of people to fulfill their well-being at the heart of human development. Development is "empowering" when it expands the range of choices available to the individual so that he can lead the life project he values (Sen, 1999b). This range of choices constitutes its capability set (Sen, 1992). When the capability set is rich the individual is sufficiently equipped with the assets facilitating him to take full advantage of the opportunities available to him (Tovar, 2008). In short, it is a development process that improves the richness of human lives as well as the economic richness of the countries (UNDP, 2016).
It is in this sense that the capability approach offers both economists and public decision-makers a promising alternative framework for understanding human development. One of the reasons for this success is its recognition of plurality in various ways (Sen, 1992(Sen, , 1999a(Sen, , 1999b(Sen, , 2009 223 valuable "beings and doings" the perspective of capabilities appropriates two fundamental heterogeneities: the diversity of human beings and the multiplicity of variables according to which one can carry out an evaluation of the level of well-being (Sen, 1992). The point of view of the capability approach is to widen the information base of well-being to contain other factors whose direct impact on the quality of life level is no less decisive than income.
Nevertheless, the ambition to integrate a wide range of constitutive domains of human life increases the abstraction of the concept of capability (Aguenane, 2019a(Aguenane, , 2019b. Several methodological challenges arise from this high level of abstraction and the applicability of the approach becomes a complex task. Among the various difficulties hindering the operationalization of Sen's approach its "intrinsic complexity" (Chiappero-Martinetti, 2008). Indeed, for Chiappero-Martinetti and Roche (2009), the plurality of the capability approach is due to a diversity of the evaluation spaces, dimensions, units of measurement and contexts. A second difficulty stems from the "incompleteness" of the approach (Sen, 1992;Qizilbash, 2008). Sen, in opposition to Martha Nussbaum, intentionally dismissed the idea of a list of relevant capabilities or functionings. Even more, he did not provide an operational framework that would guide researchers in their evaluation exercises (Frediani, 2010). A third difficulty arises from the counterfactual character of the capability analysis. The question is when a person chooses specific freedoms instead of other available freedoms or when specific circumstances compel him to choose a particular lifestyle (Comim, 2001). A fourth difficulty relates to the dual nature of Sen's approach. According to Alkire (2008), the researcher must take into account the evaluative dimension of the approach but also its prospective vision. It is therefore necessary to go beyond a simple comparison of social states to an understanding of policies, activities and recommendations that at all times improve the level of well-being. The fifth and final difficulty that I will mention briefly here is related to the increased information requirement of the capability approach. For Amartya Sen to adopt areas of valuation such as wealth or income takes into consideration only means of well-being and not the extent of well-being. While the concentration on the space of the functionings makes it possible to have a real idea on all the ways of "beings and doings" of the individuals (Sen, 1992). But this requires statistical data that is rarely available.
Therefore, this paper seeks to reduce these methodological and technical constraints by connecting the concepts of functionings and capabilities in a network and fit this network of constructs to data through a hierarchical model. The objective is to measure the level of capabilities (considered as the main dimensions of human development) using the level of functionings (considered as the achievements in each dimension).

Methods:-
Human capabilities as latent constructs and functionings as manifest variables:-Our model retains five human capabilities as fundamental dimensions of human development (Sen, 1992;Robeyns, 2005;Alkire, 2007); namely, health, education, shelter, employment and mobility and communication. These capabilities are latent, because they cannot be observed and measured directly. However, functionings, which are the achievements in each dimension, are observable and directly measurable through statistical indicators. This empirical study uses a set of available indicators to measure the performance achieved in each dimension (Di Tommaso, 2007;Krishnakumar, 2007;Krishnakumar and Ballon, 2008): 1. Health capability: The level of achievement in the field of health is assessed by three indicators: healthy life expectancy at birth (Expectancy), survival to age 65 (Survival) and disability-adjusted life expectancy (Disability). 2. Education capability: In the field of education, three indicators are proposed: the gross enrolment ratio (Enrolment), the adult literacy rate (Literacy) and the average years of total schooling (Schooling). 3. Employment capability: Employment field is assessed by three indicators: the employment-to-population ratio (Employment), the labour force participation rate (Laborforce) and the female labour force (Femalabor). 4. Shelter capability: In the field of housing, two indicators are selected: access to electricity (Electricity) and access to an improved water source (Water). 5. Mobility and communication capability: Finally, in the field of mobility and communication, three indicators are chosen: the number of fixed telephone subscriptions (Telephone), the number of internet users (Internet) and the average pump price for gasoline (Gasoline).
It can be said, thus, that the latent constructs scores reflect the national level attained in each of the dimensions of human development. While the second-order construct (general capability: GC), formed by the five dimensions, gives a general view of the level of human development achieved by the sixty countries of the sample.

Estimation Method:-
The method used for the estimation of the structural equation model is the Partial Least Squares (PLS) approach. It is known for its statistical flexibility that does not require strict statistical conditions on model variables, its compatibility with small samples (Lacroux, 2009), and its adaptability with often imperfect and overly correlated data (Sosik et al., 2009). It is able to provide scores of latent variables (capabilities and general capability), to predict their levels in various countries and to shed light on the structural relationships between them. The structural equation model is composed of two sub-models. The first sub-model, called . However, the classical problem that arises for the estimation of hierarchical models is that the items necessary for the estimation of the constructs of higher levels no longer exist since they are already used to estimate first-order constructs. To overcome this limit, three solutions have been proposed, according to Becker et al (2012): (1) the repeated indicator approach, (2) the two-step approach and (3) the hybrid approach. Without giving in to a long comparison between these approaches, we cite two sufficient reasons to favor the approach of the repeated indicators in this empirical study. The first advantage comes from the fact that the upper-level latent variable is constructed from all the items of the lower-level constructs. Consequently, this approach simultaneously estimates the constructs of lower levels and those of higher levels; which allows all parts of the model to be taken into account and a better interpretation of the results (Wilson and Henseler, 2007). The second advantage is that this method makes it possible to evaluate the effect of the manifest variables not only on the latent variables of the first level but also on those of higher levels (Ciavolino and Nitti, 2010).
The relevance of a model requires the validation of three distinct but closely linked stages: (

Validation of Measurement Model:-
To build relevant measurement instruments, it is necessary to ensure the internal consistency reliability and the unidimensionality of the blocks of items. The Cronbach's alpha calculated for each block, exceeding 0.82, proves an internal consistency of the items. Whereas the principal components analysis suggests a single component restoring at least 74% of the total variances explained. The normality analysis of the data is done on the basis of two tests: the Kolmogorov-Smirnov test (K-S) and the Shapiro-Wilk test (S-W). According to these two tests, the distributions of the variables retained are significantly different from a normal distribution since both tests are significant. Significance *. This is a lower bound of the true significance. Table 1:-Homogeneity and unidimensionality of the measurement blocks. Table 2 evaluates the items by analyzing the relevance and the significance of their loadings. All correlations between the items and the latent variables to which they are attached exceeds significantly 0.7, which allows us to say that a large part of the variances is explained. We use the average variance extracted (AVE) to make sure that the block of manifest items represents a single latent variable (Fornell and Larcker, 1981

Discriminant validity:
The discriminant validity is established when a latent variable shares more variance with its own items than with other latent variables. The Fornell and Larcker test shows that the average variance extracted of each latent variable is greater than the square of the correlation with the other latent variables (

Results and Discussion:-
To ensure the relevance of the hierarchical structural model, it was necessary to check: 1. The level of multi-collinearity of the model using the variance inflation factor (VIF) ( 0.000 a Standard deviation, T value and P value are generated by the bootstrapping procedure (n = 5000) b T > 1.58: significance at the 10% level (*); T> 1.96: significance at the 5% level (**); T> 2,58: significance at the 1% level (***) Table 5:-Significance testing results of the structural path model coefficients.

Structure of the general human capability:-
The correlations between the human general capability as a global score with capabilities as partial scores were quite satisfactory ( Table 6). All the structural relationships linking the five human capabilities (first order latent variables) to the general capability (second order variable) are important according to the standards of the PLS approach. Indeed, at a significance level of 1%, we note the strong contribution of health to capabilistic well-being (0.295; t = 13.075; P = 0.000). This expected result confirms that widening the scope of health freedom is at the top of the priorities once the improvement of the structure of human capabilities is in question. Our choice to use statistical indicators such as the healthy life expectancy at birth, survival to age 65 and disability-adjusted life expectancy is to show that freedom of health does not just depend on health infrastructures. But it must be appreciated, much more, by the degree of accomplishment in functionings like " living as long as possible in good health", "be able to avoid an unnecessary suffering" or "to be able to avoid a premature death" (Sen, 1992). The estimation of the model reveals the importance of the capacities relating to communication and mobility (0.279; t = 10.425; p = 0.000) which even precede the education capability (0.265; t = 14.605; p = 0.000). We understand, therefore, that the structure of capabilities is evolving since functionings such as "communicating with others" and "being mobile" become strong determinants of the level of well-being. This is confirmed by the bottom ranks of material capabilities such as housing (0.195; t = 9.826; p = 0.000) and employment (0.183; t = 5.195; p = 0.000).
General Capability (GC) Education 0.869 Employment 0.643 Health 0.866 Mobility-Communication 0.881 Shelter 0.775 Table 6:-Correlation between partial scores and global score

Composite index of well-being measured in terms of capabilities:
Let us now turn to the question of the calculation of the composite index of well-being based on capabilities or what it can be called the "general capability index" Index (GCI). A composite index, according to Saisana and Tarantola (2002), is a mathematical aggregation of a set of indicators that must reflect a compromise between scientific accuracy and data availability. Indeed, one of the reasons why the latent variable structural equation models are chosen and more precisely the PLS estimation method is that this approach allows to build a composite indicator system (Trinchera et al., 2008). In other words, once the measurement and the structural models are well specified, the PLS approach estimates scores for the latent variables of the global model. This allows to have dimensional (or partial) composite indicators reflecting capabilities and an overall composite indicator reflecting the general capability at the level of each country in the sample.
However, as with all composite index calculations, the problem of normalization arises. This is the method of normalization which must be mobilized to calculate dimensional indices and the global index of well-being. In the 228 case of this study the scores obtained using the PLS approach are normalized to the [0;1] scale. Thus, the normalized scores will constitute the GCI values as shown in Table 7.

The human capability index versus the HDI and per capita GDP:-
Theoretically the index based on functionings and capabilities complements the GDP and enriches the HDI. These indicators, even though they are intelligently designed, leave areas of shadow in the evaluation of human development that need to be illuminated by broader information-based indicators such as the GCI. This should in principle be translated statistically into strong correlations between these indicators. Table 8 shows the presence of a strong positive and significant correlation between GCI and HDI (0.950) and in the same proportions, but somewhat lower between GCI and GDP (0.902). Furthermore, if the link between HDI and GDP (0.936) is introduced in this correlation analysis, it can be concluded that the GCI composite index is closer to HDI than to GDP. These results empirically agree Amartya Sen's multidimensional conception of functionings and capabilities.  (2-tailed) 0,000 0,000 N 60 60 60 **. Correlation is significant at the 0.01 level (2-tailed). Table 8:-Correlation between CWBI, HDI and GDP.

Correlations
However, behind the strong correlation between the HDI and GCI, there are considerable differences for some countries. For example, the United Kingdom, Slovenia, Cyprus, Croatia and China gain more than 9 places in the capability ranking compared to that using the HDI. While the situation is reversed for other countries like Ireland (-10), Mexico (-13) and the United States (-19). Comparing the capability index to GDP rankings also reveals particular differences. For instance, Israel, Bosnia, Georgia, Ukraine and Armenia are better ranked in terms of the capability index than that using the GDP. While other countries with high GDP per capita are not doing well with the capability index. The most striking cases are those of the United States of America (-18) and Luxembourg (-17). This is explained by the low scores recorded by these countries at the level of several human capabilities (Table 7).

Conclusion:-
This paper highlights several points. Firstly, we understand that more and more we are integrating indicators that touch the domains of human life, more and more we are shedding light on the shadows that characterize the current assessment of the well-being in particular and human development in general. Secondly, the high scores recorded by 230 capabilities such as mobility and communication suggest that they should be considered by researchers and policymakers as priority like health and education. It also shows that the structure of human capabilities is changing. Tirthly, the study demonstrated the great contribution that latent variable structural models can make to the operationalization of the capability approach. Especially because the estimation under the partial least squares approach using the latest techniques of path modeling offers the possibility of understanding the latent and simultaneous nature of the capabilities.