1.1. Operationalization of hierarchical structural model variables

1.1.1.     Endogenous latent variables: capabilities as dimensions of human development

Selecting capabilities or functionings is not an end in itself. It must be done based on a set of criteria (Robeyns, 2005; Alkire, 2013) to find a compromise between the theoretical ambitions and the empirical constraints (Sen, 1992). The model combines five fundamental dimensions of human development namely health, education, shelter, employment, and mobility and communication. These capabilities are latent, unobservable, and endogenous in our model. However, functionings, which are the achievements in each dimension, are observable and directly measurable through statistical indicators. Only one indicator can be used, but it is more appropriate to use a set of available indicators to measure the performance achieved in each dimension (Krishnakumar, 2007; Bhatti & Akram, 2020; Choudhury, 2019)

Thus, in the field of education, three indicators are proposed: the gross enrollment ratio (Enrolment), the adult literacy rate (Literacy) and the average years of total schooling (Schooling). Three indicators are also selected in the field of health: healthy life expectancy at birth (Expectancy), survival to age 65 (Survival) and disability-adjusted life expectancy (Disability). The level of achievement in the employment field is assessed by three indicators: the employment-to-population ratio (Employment), the labor force participation rate (Laborforce) and the female labor force (Femalabor). In the field of housing, two indicators are selected: access to electricity (Electricity) and access to an improved water source (Water). Finally, in the field of mobility and communication, three indicators are selected: 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 variables associated with these dimensions reflect the national level attained in each of the dimensions of human development. To have an overall view of the level of development, the model introduces a second-order construct formed from the five selected dimensions and named “generic capability” (Figure 1).

1.1.2.     Exogenous latent variables: instrumental freedoms

Three of the five instrumental freedoms mentioned above are retained in the model: political freedoms, economic facilities, and transparency guarantees. To measure the “political freedoms” latent variable, four statistical indicators are proposed: the plurality and quality of the electoral process (Plurality), the level of political participation (Participation), democratic culture (Democulture) and civil liberties (Civiliberties). To assess the “economic facilities” construct, the indicators of economic freedom are suggested: the degree of freedom of trade (Trade), the degree of financial freedom (Finance), the degree of monetary freedom (Currency), and the degree of freedom to invest (Investment). To assess the "transparency guarantees" construct, the following governance indicators are selected: the control of corruption (Corruption), the authority of the law (Lawauthority), the quality of regulation (Regulation), and government effectiveness (Effectiveness). The combination of the three categories of freedoms constitutes the exogenous second-order variable namely “instrumental freedoms” (Figure 1).

1.2. Estimation method

The framework taken in this paper is based on the structural equation models (SEM). For estimating the model, the partial least squares (PLS) approach is used. This choice is explained by several reasons: its statistical flexibility that does not require strict statistical conditions on model variables, its compatibility with small samples (Lacroux, 2009), its adaptability with often imperfect and overly correlated data (Sosik, Kahai & Piovoso, 2009; Jakobowicz, 2007), and its ability to calculate scores of latent variables to predict their levels and to evaluate structural relationships between them.

According to Chin (1998) and Law et al. (1998) (as quoted in Becker, Klein & Wetzels, 2012), hierarchical latent models or higher-order constructs are an explicit representation of multidimensional concepts with a high level of abstraction. 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 have already been used to estimate the first-order constructs. To overcome this limit, three solutions have been proposed, according to Becker, Klein and Wetzels (2012) namely: (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, three reasons are sufficient to favor the approach of the repeated indicators. The first advantage is that the upper-level latent variable is constructed from all the items of the lower-level constructs. The second advantage comes from the fact that this approach simultaneously estimates both the lower level and higher level constructs, which allows all parts of the model to be taken into account and thus producing a better interpretation of the results (Wilson & Henseler 2007). The third 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 & Nitti 2010).

1.3. Conceptual model

The structural equation model of the study is designed in such a way that it allows to measure the direct effects of instrumental freedoms on the five selected human capabilities. But it will also capture the indirect effects between all the latent variables (Figure 1).

Figure 1:

Conceptual model

Source: Author's computation /SmartPLS (Version 3.3.2) Output

1.4. Data source

This empirical study is a cross-section of 60 countries for the year 2010. The main source of data is the World Bank Group (World Development Indicators) excluding health indicators which are from the World Health Organization.

3.   RESULTS/FINDINGS

To validate the model, it is recommended by Hult, Sarstedt, Ringle and Hair (2016) to go through three steps: 1) examination of the statistical indicators chosen (manifest variables), 2) evaluation of the measurement model (relationships between the manifest variables and the latent variables with which they are associated) to ensure the relevance of the different blocks of items, and 3) evaluation of the internal or structural model (relationships between the latent variables).

3.1   Examination of statistical indicators

The robustness of the measurement instruments depends on the internal consistency reliability and the unidimensionality of the blocks of items. These two preliminary conditions are verified through the calculation of the Cronbach's alpha and the application of the principal component analysis (PCA) to each block of items. The significance of the two calculated normality tests namely the Kolmogorov-Smirnov test (K-S) and the Shapiro-Wilk test (S-W) proves that the variables retained do not follow a normal distribution (Table 1).

Table 1

Examination of statistical variables

Source: Author's calculation/SmartPLS (Version 3.3.2) Output

3.2   Validation of measurement model

According to Hult et al. (2016), the validity of the measurement model is determined through a procedure of three important steps: (1) evaluation of the internal consistency of the measurement instruments, (2) assessment of the convergent validity, and (3) assessment of the discriminant validity.

Reliability of indicators and validity of constructs

Table 2 shows that all the latent variables have good composite reliability (CR) for exceeding the threshold value of 0.7 which is commonly recommended (Henseler, Ringle & Sinkovics, 2009). The loadings of items are consolidated by analyzing their statistical significance using the bootstrapping technique (Table 2).

Table 2:

Reliability of indicators and validity of constructs

Source: Author's calculation/SmartPLS (Version 3.3.2) Output

Convergent validity

The convergent validity of the constructs is checked using the average variance extracted (AVE) (Fornell & Larcker, 1981; Picot-Coupey, 2009). Table 2 shows that each latent variable shares more than 50% of the variance with its own items (AVE > 0.5).

Discriminant validity of constructs

Discriminant validity is proven when each latent variable shares more variance with its items than with those of the other latent variables (Chin, 1998). Table 3 shows that the factorial contributions of each item are higher than its cross-loadings.

Table 3:

Discriminant validity of constructs

Source: Author's calculation/SmartPLS (Version 3.3.2) Output

3.3   Validation of structural model

As with the measurement model, the validation of the structural model requires a series of tests. Hult et al. (2016) summarized the procedure for validating the structural model in five important steps: (1) evaluation of the collinearity level of the model, (2) evaluation of the coefficient of determination levels, (3) evaluation of the relevance and significance of structural relationships, (4) evaluation of the effect size, and (5) evaluation of the predictive relevance of the model and its total quality.

Collinearity assessment

The tool conventionally used to judge the level of collinearity i.e. whether tolerable or not is the variance inflation factor (VIF) (Henseler et al., 2009). The commonly accepted threshold is a VIF value of less than 10. In other words, a VIF greater than 10 reveals a critical collinearity level for model estimation, whereas a VIF of less than 3 is generally considered to be excellent. Table 4 shows that the calculated VIF is below the recommended thresholds.

Table 4:

Collinearity assessment

Source: Author's calculation/SmartPLS (Version 3.3.2) Output

The relevance and significance of the structural model path coefficients

In this study, the path coefficients are greater than 0.5. The analysis of the relevance of structural relationships is supplemented by evaluating the significance levels of the different structural model path coefficients obtained using the bootstrapping procedure (Table 5).

Table 5:

Relevance and significance of the structural model path coefficients

Source: Author's calculation/SmartPLS (Version 3.3.2) Output

Evaluation of coefficients of determination and effect size

Referring to Chin (1998) and Hult et al. (2016), we can interpret the R² value for mobility and communication capability (R² = 0.747) as very high, and for health (R² = 0.526), education (R² = 0.409) and employment (R² = 0.310) as moderate whilst for housing capability (R² = 0.175) as low. Based on the evaluation of the R² changes following the omission of an exogenous variable, the effect size f² is used to evaluate whether the omitted exogenous variable has a high, medium, or low impact on the endogenous variables. According to the criteria of the PLS approach (Hult et al., 2016), we can interpret the effect of the capabilities of mobility and communication (f² = 2.957), health (f² = 1.110), education (f² = 0.691), and employment (f² = 0.449) as very strong, and that of housing capability (f²= 0.212) as moderate.

Table 6:

Coefficients of determination and effect size

Source: Author's calculation/SmartPLS (Version 3.3.2) Output

Testing the predictive relevance of the model 

The blindfolding procedure is used to generate the Stone-Geisser Q² which is a commonly accepted indicator of the predictive relevance of models. A Q² of Stone-Geisser greater than 0 indicates a predictive relevance of the model (Henseler et al., 2009; Hult et al., 2016). Table 7 presents the results of the Stone and Geisser test. The cross-validation test of the Stone-Geisser Q² calculated for the hierarchical model is much greater than 0. This result proves that the model has significant predictive relevance.

Table 7:

The predictive relevance of the model

Source: Author's calculation/SmartPLS (Version 3.3.2) Output

3.4   The multi-group analysis

The Partial Least Squares Multi-Group Analysis (PLS-MGA) is a specific method (Hult et al., 2016) to determine if the model changes significantly depending on whether it is in the context of developed or developing countries. Table 8 shows that the level of development does not moderate the effect of instrumental freedoms on the various capabilities.

Table 8:

PLS-MGA results

Source: Author's calculation/SmartPLS (Version 3.3.2) Output

4.       DISCUSSION

5.     CONCLUSION

6.     LIMITATIONS AND FUTURE RECOMMENDATIONS

Several questions remain open at the end of this work. Among these, it is important to know how political freedoms, economic facilities, and transparency guarantees are mutually reinforcing. In other words, further empirical studies should be conducted to measure the effect of each kind of instrumental freedom on the others, while capturing the effect of their interconnection on the level of the main human capabilities like health and education.

REFERENCES

Aguenane, N. E. (2020). Regional Disparities In Human Development: The Case Of Moroccan Regions. International Journal of Social Sciences and Economic Review, 2(2), 28-34. https://doi.org/10.36923/ijsser.v2i2.57

Alkire, S. (2010). Instrumental freedoms and human capabilities. In Esquith, Stephen L.; Gifford (eds.), Capabilities, Power, and Institutions. Towards a More Critical Development Ethics, 18-32, Pennsylvania Park : The Pennsylvania State University Press.

Alkire, S. (2013). Choosing dimensions: The capability approach and multidimensional poverty. In The many dimensions of poverty, (pp. 89-119). Palgrave Macmillan, London.

Ayalew, N. A. (2019). Duty Of Confidentiality In Arbitration Process In Ethiopia. International Journal of Social Sciences and Economic Review, 1(2), 92-100. doi:https://doi.org/10.36923/ijsser.v1i3.42

Becker, J. M., Klein, K., & Wetzels, M. (2012). Hierarchical latent variable models in PLS-SEM: guidelines for using reflective-formative type models. Long range planning, 45(5-6), 359-394.

Bhatti, A., & Akram, H. (2020). The Moderating Role Of Subjective Norms Between Online Shopping Behaviour And Its Determinants. International Journal of Social Sciences and Economic Review, 2(2), 1-09. doi:https://doi.org/10.36923/ijsser.v2i2.52

Bradley, W., & Henseler, J. (2007). Modeling reflective higher-order constructs using three approaches with PLS path modeling: a Monte Carlo comparison. In Conference Proceedings ANZMAC 2007, edited by M. Thyne, and K. R. Deans, 791–800. Dunedin: ANZMAC.

Chin, W. W. (1998). The partial least squares approach to structural equation modeling. Modern methods for business research, 295(2), 295-336.

Choudhury, A. H. (2019). Historical background of the Rohingya refugee crisis and the implication of their statelessness. International Journal of Social Sciences and Economic Review, 1(1), 8-15. doi: https://doi.org/10.36923/ijsser.v1i1.23

Ciavolino, E., and M. Nitti. 2010, September). High-order Constructs for the Structural Equation Model. Presented at the V Meeting on Dynamics of Social and Economic Systems, Benevento.

Henseler, J., C. M., Ringle, and R. R. Sinkovics. (2009). The Use of Partial Least Squares Path Modeling in International Marketing.” In Advances in International Marketing, 277-319, edited by R. R. Rudolf and P. N. Ghauri,  Bingley: Emerald.

Henseler, J., Wilson, B., Götz, O., & Hautvast, C. (2007). Investigating the moderating role of fit on sports sponsoring and brand equity: A structural model. 8(4), 1-10.

Hult, G., M. Sarstedt, M. C. Ringle, and J. Hair. (2016). A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM). Thousand Oaks : SAGE Publications.

Jacobowicz, E. (2007). Contributions aux modèles d'équations structurelles à variables latentes. Thèse de doctorat, Conservatoire national des arts et métiers-CNAM.

Krishnakumar, J. (2007). Going Beyond Functionings to Capabilities: An Econometric Model to Explain and Estimate Capabilities. Journal of Human Development and Capabilities, Taylor & Francis Journals, 8(1), 39-63.

Lacroux, A. (2009) : L’analyse des modèles de relations structurelles par la méthode PLS : une approche émergente dans la recherche quantitative en GRH. Presented at the XXème congrès de l’AGRH, Toulouse.

Law, M., Stewart, D., Letts, L., Pollock, N., Bosch, J., & Westmorland, M. (1998). Guidelines for critical review of qualitative studies. McMaster University occupational therapy evidence-based practice research Group, 1-9.

Robeyns, I. (2005). Selecting capabilities for quality of life measurement. Social indicators research, 74(1), 191-215.

Sen, A. (1999b). Development as freedom. Oxford: Oxford University Press.

Sen, A. K. (1992). Inequality Re-examined. Oxford: Oxford University Press.

Sen, A. K. (1999a). Commodities and Capabilities. Oxford: Oxford University Press

Sen, A. K. (2009). The Idea of Justice. Londres: Penguin Books.

Sosik, J. J., kahai, S. S., piovoso, M. J. (2009). Silver bullet or Voodoo statistics? A primer for using the partial least squares data analytic technique in group and organization research. Group & Organization Management, 34(1), 5-36.

International journal of Social Sciences and Economic Review, V2 | I3 | 2020 | PP: 12-21 https://doi.org/10.36923/ijsser.v2i3.65

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License