Abstract
We use latent class analysis to create an indicator of human development that integrates a broad range of country characteristics including per capita income, health, inequality, environmental performance, and life satisfaction. We show that each of these characteristics is important in distinguishing across development experiences. Because latent class analysis is model-based and allows us to test the significance of each aspect of development and the number and size of the groupings, this approach is superior to an index that divide countries into ad hoc equally-sized groups. Our results suggest that income per capita does not, by itself, capture all of the important dimensions of development and that a rank-ordering of development experiences may not always be appropriate.
Similar content being viewed by others
Notes
Although the HDI is widely used, other indices also attempt to rank countries based on broader measures of economic performance. For example, the Happy Planet Index ranks countries based on ecological footprints, life satisfaction and life expectancy and the Prosperity Index ranks countries based on income per capita and life satisfaction. These indices use different methods of construction from the HDI, however, they are subject to similar criticisms.
Since the writing of, Wolff et al. (2011) the HDI methodology was changed and a fourth grouping of countries was added in order to reduce the potential for misclassification of countries.
Although researchers can use several rules to determine the number of clusters in cluster analysis, these rules do not depend on the models’ log-likelihood as is the case in latent class models.
Based on the idea that social capital is an unobservable multidimensional construct, Owen and Videras (2009) use latent class analysis to measure social capital.
See Vermunt and Magidson (2005) for a more detailed discussion of latent class models and their estimation.
We use Latent GOLD to perform the estimation.
UNDP also releases a supplemental inequality-adjusted HDI. However, our expanded model contains more than just inequality so we do not make a direct comparison to the supplemental HDI index.
Booysen (2002) points out that while indicators of development may focus on outcomes, inputs, or both, variables like educational attainment measure both developmental ends and inputs into development.
The BIC = −2LL + log(N)J and the CAIC = −2LL + log(N + 1)J, where LL is the value of the log likelihood, N is the sample size, and J is the number of parameters estimated.
The five class model is also supported by a bootstrapped likelihood ratio test comparing the four and five class models. Because the four-class model cannot be obtained simply by imposing a restriction on the five-class model, in order to implement this test, data from the four-class model is simulated and then both the four and five class models are estimated to compute the likelihood ratio statistic. (Skrondal and Rabe-Hasketh 2004). The bootstrapped p value for this test is 0.000.
In fact, if we were to force the four class model in spite of the fit statistics reported in Table 4, we find that the four classes would not be of equal sizes even then, with the largest class being 45 % of the sample and the smallest class being only 8 % of the sample.
Note that some high-income oil-producing countries are in Class 5, producing a high average income level for that class, even though other human development indicators are at more moderate levels.
The Czech Republic has a 35 % probability of being in Class 1. Other lower probability classified countries also have nontrivial probabilities of being in another class. For example, Brunei has a 20 percent probability of being in Class 5, Jamaica has a 48 % probability of being in Class 1, and Tonga has a 46 % probability of Class 1. As will become apparent in the next set of estimation results, with additional indicators, countries are classified with higher probabilities.
Booysen (2002) argues that cardinal indices are in fact ordinal to the extent that the numerical difference in the index between any two countries has no meaningful interpretation.
The results we describe in this section are available upon request.
For example, when we add a variable measuring the rule of law from the World Bank governance indicators, we do not find that it is statistically significant in predicting class membership once education and income are used as predictors.
References
Acemoglu, D., Johnson, S., & Robinson, J. (2001). Colonial origins of comparative development: An empirical investigation. American Economic Review, 91(5), 1369–1401.
Alfo, M., Giovanni, T., & Robert, J. W. (2008). Testing for country heterogeneity in growth models using a finite mixture aproach. Journal of Applied Econometrics, 23, 487–514.
Berenger, V., & Verdier-Chouchane, A. (2007). Multidimensional measures of well-being: Standard of living and quality of life across countries. World Development, 35(7), 1259–1276.
Bloom, D. E., David, C., & Jaypee, S. (2003). Geography and poverty traps. Journal of Economic Growth, 8, 355–378.
Booysen, F. (2002). An overview and evaluation of composite indices of development. Social Indicators Research, 59, 115–151.
Easterly, W. (1999). Life during growth. Journal of Economic Growth, 4(3), 239–276.
Easterly, W, & Levine, R. (1997). Africa’s growth tragedy: Policies and ethnic divisions. Quarterly Journal of Economics, 112(4), 1203–1250.
Hirschberg, J. G., Moussoumi, E., & Slottje, D. J. (1991). Cluster analysis for measuring welfare and quality of life across countries. Journal of Econometrics, 50(1–2), 131–150.
Kerekes, M. (2012). Growth miracles and failures in a markov switching classificaiton model of growth. Journal of Development Economics, 98, 167–177.
Knack, S., & Keefer, P. (1997). Does social capital have an economic payoff? A cross-country investigation. Quarterly Journal of Economics, 112(4), 1251–1288.
Levine, R., & Renelt, D. (1992). A sensitivity analysis of cross-country growth regressions. American Economic Review, 82(4), 942–963.
Lucas, R. E. (1988). On the mechanics of economic development. Journal of Monetary Economics, 22(1), 3–42.
Mankiw, N. G., Romer, D., & Weil, D. N. (1992). A contribution to the empirics of economic growth. Quarterly Journal of Economics, 107(2), 407–437.
McGillivray, M. (2005). Measuring non-economic well-being achievement. Review of Income and Wealth., 51(2), 337–364.
Owen, A. L., & Videras, J. (2009). Reconsidering social capital: A latent class approach. Empirical Economics, 37(3), 555–582.
Owen, A. L., Videras, J., & Davis, L. (2009). Do all countries follow the same growth process? Journal of Economic Growth, 14(4), 265–286.
Paap, R., Franses, P. H., & van Dijk, D. (2005). Does africa grow slower than asia, latin america and the middle east? Evidence from a new data-based classification method. Journal of Development Economics, 77, 553–570.
Sen, A. (1985). Commodities and capabilities. Amsterdam: North Holland.
Slottje, D. J. (1991). Measuring the quality of life across countries. Review of Economics and Statistics, 73(4), 684–693.
Skrondal, A., & Rabe-Hesketh, S. (2004). Generalized latent variable modeling: Multilevel, longitudinal and structural equation models. Boca Raton, FL: Chapman & Hall/CRC.
Vermunt, J. K., & Magidson, J. (2005). Latent GOLD 4.0 user’s guide. Belmont, MA: Statistical Innovations.
Wolff, H., Chong, H., & Auffhammer, M. (2011). Classification, detection and consequences of data error: Evidence from the human development index. The Economic Journal, 121, 843–870.
Acknowledgments
We are grateful for helpful comments from Lewis Davis.
Author information
Authors and Affiliations
Corresponding author
Appendix
Appendix
See Table 12.
Rights and permissions
About this article
Cite this article
Owen, A.L., Videras, J. Classifying Human Development with Latent Class Analysis. Soc Indic Res 127, 959–981 (2016). https://doi.org/10.1007/s11205-015-0992-8
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11205-015-0992-8