Data challenges in estimating the HDI: The cases of Cuba, Palau and the Occupied Palestinian Territory

This post first appeared on UNDP's Let's Talk Human Development

By Francisco Rodríguez and Clara García
Head of Research, and Statistics and Research Analyst (respectively),
Human Development Report Office, UNDP

There are many statistical and methodological challenges in estimating the annual Human Development Index (HDI). The goal of the Human Development Report has always been to make national HDI coverage as extensive as possible, based on a common and consistent set of indicators for the HDI's health, education and income dimensions.

In recent months the Human Development Report Office has been working with national and international statistical authorities to expand the range of national indicators available to calculate the HDI, using methods we describe here and, in greater technical detail, in a forthcoming Human Development research paper. As a result, we are now able to show how Cuba, Palau and the Occupied Palestinian Territory would have been ranked in the 2010 HDI had these figures then been available.

Cuba, for example, would have been placed at #53 in the 2010 HDI country rankings, in the High Human Development Index category. Palau, also omitted from the 2010 HDI due to data unavailability at the time of its calculation last year, would have been ranked #54, also in the High Human Development category, while the Occupied Palestinian Territory would have been #97, in the Medium Human Development group. (See  2010 HDI update, with Cuba, Occupied Palestinian Territory and Palau included.)

One of the biggest challenges in constructing a composite measure of development like the HDI is to balance comprehensiveness against country coverage. The choices are not easy. For example, we all would agree that it would be better to include a measure of educational quality in the HDI instead of just the measures of years of school attendance currently used. But the indicator of educational quality with the broadest country coverage currently available, compiled by the OECD, covers only 65 countries, far fewer than the 173 countries for which we were able provide HDI education indicators in 2010, using expected years of schooling for children entering primary school and years of schooling completed by the adult population.

One key challenge relates to cross-country comparability. Take the case of income: In order to compare levels of income across countries, it is necessary to use a common currency - one can’t meaningfully compare zlotys with dirhams, for example – but even after one converts into (say) US dollars, we need to take into account that a dollar does not purchase the same amount of goods in Morocco as it does in Poland. For this reason, the World Bank collects data that allows estimates of exchange rates adjusted for these differences in purchasing power for 181 countries. Unfortunately, 13 countries are not included. Three of these – Cuba, Palau and the Occupied Palestinian Territory – have data on all other HDI components, so the estimation of GNI is the only constraint that we need to address to calculate an HDI value.

Indeed, for Cuba, we know that performance on the non-income dimensions has been stellar. As highlighted in an earlier Let’s Talk Human Development blog posting (Subtracting GNI from the HDI: A non-income HDI), Cuba is the only Latin American country in the top ten non-income-HDI movers over the past decade, with life expectancy increasing by two years and expected years of schooling increasing by five years. These are remarkable improvements for a country that already had very high health and education indicators at the outset of the decade. Cuba is in fact the best performing developing country in terms of the non-income HDI, as that previous article demonstrated.

Problems with cross-country comparability led to reduced national coverage in the 2010 HDI – a trend that the Human Development Report Office was keen to reverse. Recent efforts have allowed us to generate estimates that will allow us to calculate the HDI in the 2011 Report for at least 13 new countries, raising the HDI’s coverage from 169 to more than 180 countries, representing 99 percent of the world’s population.

This effort has been undertaken in close consultation with the international statistical community to address the problems of countries for which we did not have enough data to calculate an HDI last year. This dialogue has led to the creation of a Statistical Advisory Panel, a Working Group of Countries without an HDI, and been reflected in continuous consultations with global and regional statistical associations such as the United Nations Statistics Commission and the Economic Commission for Latin America and the Caribbean.

Cuba is an interesting case, because two currencies are in circulation – the Peso Cubano (CUP), which is traded at an implicit exchange rate of 24 to the dollar, and the Peso Cubano Convertible (CUC), which is traded at an exchange rate of 1 to the dollar.[i] Estimating a purchasing-power-parity (PPP) adjusted exchange rate for this country requires knowing what fraction of spending is carried out in each currency, information which is not available in official Cuban statistics. In background research carried out by HDRO, we found that the estimated level of GNI per capita in 2008 PPP$ in Cuba could vary hugely – between $264 and $9225 – using methods developed by other international organizations, depending on the exchange rate used in the calculation.

For these reasons, HDRO proposed a method for directly estimating GNI in PPP-adjusted dollars based on a regression model. Two key pre-conditions are that the explanatory variables were not originally constructed in domestic currency (as then we would still require an exchange rate to allow conversion) and that they be directly linked to the state of the economy (as we are estimating GNI, not the level of human development). In light of the results of extensive robustness and sensitivity analysis, we settled on a model which explained GNI per capita as a function of international trade, per capita energy use, the share of the population with internet access and a set of regional variables.

The cases of Palau and the Palestinian Occupied Territory are more straightforward, since we can apply existing estimation models, which use information on GDP and other macroeconomic variables valued at market exchange rates to infer PPP-adjusted exchange rates.

These methods will be used to estimate GNI per capita for these three countries in the upcoming 2011 Report. In the meantime, it is relevant to ask what would have happened to the HDI rankings last year if we had used these methods then to estimate the Index for the countries in question. The table below provides the answers, showing where these countries would have placed in the HDI 2010 rankings.[ii]


2010 HDI update for Cuba, Occupied Palestinian Territory and Palau [iii]

	Cuba	Occupied Palestinian Territory	Palau
Ranked above	Croatia Uruguay	Gabon Suriname	Uruguay Cuba
HDI rank	53	97	54
HDI value	0.760	0.645	0.757
Ranked below	Palau Libyan Arab Jamahiriya	Bolivia, Plurinational State of Paraguay	Libyan Arab Jamahiriya Panama
Model used	HDRO	ICP 2005	Sun & Swanson 2008

Sources: Rodriguez and Garcia, "Estimating Purchasing Power Parities" 2011. Human Development Research Paper (forthcoming); "International Comparison Programme, Global Purchasing Power Parities and Real Expenditures" (Washington, DC: The World Bank, 2008) http://siteresources.worldbank.org/ICPINT/Resources/icp-final.pdf; Sun and Swanson, "Estimation of PPPs for non-benchmark economies for the 2005 ICP round", http://web.worldbank.org/WBSITE/EXTERNAL/DATASTATISTICS/ICPEXT.html

Using this newly developed method, it is clear that Cuba would have ranked well in the 2010 HDR – indeed at number 53 in the global rankings, and thus among the High HDI group and sixth in the Latin America and the Caribbean region. Globally, in terms of the HDI, its nearest neighbours would be Uruguay and Croatia (just above) and Libya and Palau just below. This reflects the combined achievements of Cubans averaging ten years of schooling among adults, almost 18 expected years of schooling among new entrants and 79 years of life expectancy, and a model based estimate of approximately $5747 in income per capita in PPP terms.

Palau would rank slightly below, at number 54, in the high human development group between Cuba and Libya. Its estimated income per capita is $9376. The Occupied Palestinian Territory falls in the medium human development group, between Suriname and Bolivia, with an estimated income per capita of $3933.

These types of model-based estimates allow us to meeting the objective of maximizing HDI country coverage. The disadvantage is that modeled data are typically of lesser quality than information collected directly from censuses, surveys, or national registers. Mindful of this caveat, and heeding the recommendations of the statistical community of using caution in such cases, HDRO does not estimate more than one indicator of the HDI for any given country. Moreover, as noted above, extensive robustness and sensitivity analysis are adopted before adopting any estimation model. The methods used for these new country HDI calculations satisfy these strict criteria, and thus allow us to expand the coverage of the HDI without sacrificing methodological rigour. A fuller description of these challenges and approaches will be provided in our forthcoming Human Development Research Paper (Rodriguez and Garcia 2011), which will be available on our website.

[i] Strictly speaking, CUPs are not convertible to dollars but are convertible to CUCs at a rate of 24 to 1.
[ii] Note that mean years of schooling was missing for Occupied Palestinian Territory in 2010, so that our rule of not imputing more than one component of the index would have precluded us from producing an HDI. However, we now have mean years of schooling for 2010.
[iii] The 2010 HDI value and ranking reflect the inclusion of one or more HDI indicators that were not available at the time of the preparation of the 2010 Human Development Report. See table below.

	HDI value*	HDI rank*	GNI per capita (PPP)*
Cuba	0.760	53	$5747
Occupied Palestinian Territory	0.645	97	$3933
Palau	0.757	54	$9376

2010 HDI update including Cuba, Occupied Palestinian Territory and Palau [75 KB]