Obstacles to Constructing a Multidimensional Index of National Wellbeing Julia E. Heilig Department of Economics, University of York

Transcription

1 RESEARCH ARTICLE Obstacles to Constructing a Multidimensional Index of National Wellbeing Julia E. Heilig Department of Economics, University of York A BSTRACT This paper explains that the construction of alternative measures of national wellbeing is facing a major obstacle: lack of internationally comparable statistics in many of the dimensions that can be considered important for national wellbeing. For this paper data have been collected on 27 indicators and indexes related to public health, educational achievements, environmental conditions, political freedom and participation, demographic conditions, economic growth, and income disparity. The initial aim was to conduct multivariate statistical analyses of welfare based on these indicators for 197 countries. While this list of indicators initially appeared to be rather promising, detailed analysis showed that (1) most indicators are not available for the majority of developing countries and (2) some of the most interesting indicators are not even available for all developed countries. The list of indicators that is available for a majority of both developed and developing countries is shockingly short. Major efforts in generating internationally comparable statistics are necessary, before alternative measures of national wellbeing can be considered. With the current deficits of comparable international statistics, alternative measures of national wellbeing are unrealistic. A RTICLE I NFO Correspondence regarding this article should be sent to jh1437@york.ac.uk Cite this article as: Heilig, J. E. (2013) Obstacles to Constructing a Multidimensional Index of National Wellbeing. Human Welfare, 2(2), Introduction The use of per capita Gross Domestic Product (GDP) as a measure of national wellbeing has been frequently criticized of generating misleading and distorted international comparisons. Many authors have argued that the GDP is biased towards measuring economic growth while ignoring social, political, health-related, cultural, and especially environmental dimensions of wellbeing. For this paper a large number of statistical data has been collected to identify possible measures for some of these other dimensions of wellbeing. The paper commences with an explanation of the indicators used and explains on what principles they were selected. Further the methodology for the analysis of wellbeing is discussed. Literature The pitfalls of GDP as a measure of wellbeing (as discussed by Dean, 2007; Van den Bergh, 2007; Victor, 2010) have created numerous new methods, some of which will be outlined below. There are two different approaches: one might study subjective wellbeing as measured by self-assessment surveys, or one might analyze various objective (aggregate) indicators relating to wellbeing, such as employment levels or health expenditure in a country. Examples of subjective wellbeing surveys include Ravallion s Economic Ladder Question (1999), Blanchflower s and Oswald s study on Hypertension and Happiness across Nations (2008), or the Day Reconstruction Method (DRM) ISSN: Human Welfare. Volume 2. Issue 2. Spring

2 developed by Kahnemann et al. (2006). For an elaborate discussion on the consistency of subjective measures see Veenhoven (1996), Noll (2004, p. 159), Halliwell (2002), Krueger and Schkade (2008), and Erikson (1993, p. 77). Objective measurements of wellbeing combine a set of reproducible social indicators based on empirical evidence such as life expectancy at birth, level of education, and per capita GDP in the case of the United Nations Human Development Index (HDI). The Global Peace Index (GPI) can also be considered an objective measure of wellbeing. This index, designed by the Institute for Economics and Peace, is made up of 23 individual indicators. For a critique on the ambiguity of objective and subjective measures of wellbeing see Erikson (1993, p. 77). Diener and Suh (1997, pp ) give an in-depth analysis of pros and cons of using social indicators to measure wellbeing within or across countries. For the history and possible future Ferriss, Easterlin, Patrick, & Pavot (2006). Some countries are actively searching for alternative methods of analyzing the wellbeing of their population (Domitrovic, 2009). Especially France has made widely discussed efforts towards developing a gross national satisfaction index, which President Sarkozy wanted to implement as an indicator for economic development and growth (Samuel, 2009). However, critics have argued that the hidden agenda of measuring happiness and satisfaction is to disguise the increasingly slow GDP growth in Europe and especially in France. Another country which pursued interests in measuring happiness is Great Britain. Under Tony Blair an unofficial Department of Happiness was created, in order to aid in the process of designing policies (Weiner, 2009). Selection of Indicators/Indices The indicators and indices selected cover the trends of social indicators research see Sirgy, Michalos, following dimensions: health, economic development, Indicators/Indices Method Year Source Countries with data CO2 emissions (metric tons per capita) Latest available 2007 World Bank Data 196 Corruption perceptions index Latest available 2010 Transparency international 177 Democracy index Average 2006/8/10 Economist Intelligence Unit 167 Difference between female and male life expectancy at birth Latest available 2010 UNPD 229 Ease of doing business, ranking Latest available 2010 World Bank Data 182 GEF benefits index for biodiversity Latest available 2008 World Bank Data 203 Gini Index Latest available World Bank Data 152 Hazardous waste generation (tons) Average UNSD 83 Healthy life expectancy at birth (total population) Latest available 2004 WHO 192 Industrial/ Transport accidents (number of affected people) Total OFDA/CRED International 146 Disaster Database Inflation, average consumer prices (% change) Latest available 2008 IMF 181 Intentional homicide, rate per 100,000 population Latest available UNODC 151 Level of violent crime Latest available 2010 Economist Intelligence Unit 149 Malnutrition prevalence, weight (% children under 5) Average World Bank Data 126 Natural Disasters (number of affected people) Total OFDA/CRED International 198 Disaster Database Overall Life expectancy Latest available 2010 UNPD 229 Overall mathematics scale (PISA) Latest available 2009 OECD (PISA) 63 Overall Reading Scale (PISA) Latest available 2009 OECD (PISA) 63 Overall science scale (PISA) Latest available 2009 OECD (PISA) 63 Patent applications per 1000 population Average / WIPO 181 pop Per capita GDP Latest available 2009 World Bank Data 174 Prison Population rate (per 100,000 of population) Latest available International center for 209 prison studies Road traffic death rate per 100,000 population Latest available WHO 177 School enrollment, secondary (percentage net) Latest available World Bank Data 158 Unemployment, total (% of total labor force) Average World Bank Data 160 Willingness to fight Latest available 2010 Economist Intelligence Unit 148 ISSN: Human Welfare. Volume 2. Issue 1. Spring 2013

3 education, environment, and political conditions. One can assume that welfare progresses within a country if these elements of wellbeing improve. More specifically data for the following indicators were collected (see Table 1). In addition, the natural logs were calculated for the collected data. Security and the Political Situation within a Country The democracy index was used since it already comprises several indicators related to the degree of political freedom and the functioning of governments. An average from the years 2006, 2008, and 2010 was used to avoid random single-year fluctuations in the index value. For the other indicators, the latest available statistics were used since there was enough country data, such as the intentional homicide rate per 100,000 population, the corruption perception index, the prison population rate (per 100,000 of population), the willingness to fight (a qualitative assessment of the population s preparedness to fight in a war), and the level of violent crime, since they all relate to security, prevalence of crimes, and quality of life in general. Public Health Overall life expectancy at birth is arguably the most significant indicator regarding health within a country. Simply taking data on the frequently used health expenditure per capita can be misleading, since it is not clear if higher health expenditures are better or if they just measure the costs of unhealthy lifestyles in a population. Therefore, data on the healthy life expectancy at birth was included in the analyses. The difference between female and male life expectancy at birth is also contained within the health dimension. This can be interpreted as a measure of gender disparity in health. Data on the prevalence of malnutrition by age (% of children under 5) is only relevant for developing countries since there is no data available for most developed countries even if there may be malnutrition among poor children in some developed countries. Because of the missing data an average of malnutrition over the years was used. Environmental Quality The Global Environmental Facility (GEF) provides an index for biodiversity which measures the extent of endangerment of species and the diversity of habitats within a country. CO 2 emissions (metric tons per capita) are created by various industrial processes, including manufacturing cement or by burning fossil fuels for transportation. For the natural disasters indicator (measured by the number of people affected) including droughts, earthquakes, extreme temperatures, floods, storms, volcano eruptions, and wildfires an average over the years 1990 to 2011 was used. This 21- year average was calculated to avoid that a singular event, such as a flood in a particular year, has too much weight. One can assume that frequent natural disasters over a longer period lower the national wellbeing. Industrial and transport accidents (also measured in total number of people affected) were also included. The number of affected people is calculated based on those injured, killed, or displaced. Hazardous waste generation (measured in 1,000 tons) was included since waste can greatly reduce the quality of the environment particularly if it includes toxic, infectious, radioactive, or flammable waste. An average from 1995 to 2009 was calculated. The reason for going so far back in time is that firstly there would have been many countries with missing data if only the last 2 years would have been used and secondly because the waste with infectious or radioactive properties which was created in 1995 can still negatively affect the wellbeing of people today. Quality and Provision of Education Secondary school enrollment was used to measure the advanced levels of education. The Program for 100

4 International Student Assessment (PISA) surveys are the only reliable and internationally comparable measure of the knowledge that students have in a particular grade. The students are tested in reading, mathematics, and science. Contrary to the adult literacy rate, the PISA reading scores are much more accurate since literacy in some countries is defined as someone who declares that s/he can read or write. Economic Development Per capita GDP was used since it can provide an initial approximation to material wellbeing across countries. The ease of doing business is also an important indicator with respect to economic progress. Per capita patent applications correlate with per capita GDP and also explain the state of innovation within an Indicator CO2 emissions (metric tons per capita) Corruption perceptions index Difference between female and male life expectancy at birth Ease of doing business, ranking Gini Index Healthy life expectancy at birth (total population) Index for biodiversity Industrial/ Transport accidents (total number of affected) Intentional homicide, rate per 100,000 population Level of violent crime Log Gini Index Log: CO2 emissions (metric tons per capita) Log: Corruption perceptions index Log: Difference between female and male life expectancy at birth Log: Healthy life expectancy at birth (total population) Log: Index for biodiversity Log: Industrial/ Transport accidents (total number of affected) Log: Intentional homicide, rate per 100,000 population Log: Natural Disasters (number of affected) Log: Overall Life expectancy Log: Patent Applications rate per 100,000 Log: Per capita GDP Log: Prison Population rate (per 100,000 of population) Log: Road traffic death rate (per 100,000 population) Log: School enrollment, secondary (% net) Natural Disasters (number of affected) Overall Life expectancy Patent Applications rate per 100,000 Per capita GDP Prison Population rate (per 100,000 of population) Road traffic death rate (per 100,000 population) School enrollment, secondary (% net) Willingness to fight economy. Unemployment not only reduces individual disposable income but can lead to unstable living conditions and exclusion from society, reducing wellbeing within a nation. The Gini Index was used to include a measure of inequality. Although per capita GDP can be high within a country, the wealth may be distributed unevenly among the population. Finally, inflation (in percentage-change of average consumer prices) was included. Results Principal Component Analysis A principal component analysis was used since it provides a small number of components which can explain the variability within a large number of Notation CO2E CPI DBFAMLE EODB GI HLE IFB ITA IH LOVC LOG_GI LOG_CO2E LOG_CPI LOG_DBFAMLE LOG_HLE LOG_IFB LOG_ITA LOG_IH LOG_ND LOG_OLE LOG_PAR LOG_PCGDP LOG_PPR LOG_RTDR LOG_SE ND OLE PAR PCGDP PPR RTDR SE WTF 101

5 measures. This method is termed data reduction. It is usually performed when one wants to exclude some of the original measures from the analyses, however still wants to use the information these measures contain. Thus the principal components are less than or equal to the number of original variables. The rotation method was varimax with Kaiser Normalization. The reason for using a varimax method was that the interpretation of the results is simplified. The missing values were deleted pairwise. Several indicators needed to be excluded for the factor analysis. The Program for International Student Assessment (PISA) surveys all had to be neglected since data was only available for 63 countries. Similarly, malnutrition prevalence and hazardous waste generation had too much missing data to be included in the factor analysis. In addition the initial communalities showed that the following indicators did not adequately explain the factor results and were thus excluded: democracy index, inflation, and unemployment. The individual indicators were collected for 197 countries. These countries all have a population of more than 100,000. Countries with less population typically have much missing data and in addition are very difficult to compare to large countries such as China, India, or America. For the factor analysis the natural logs were used, with the exception of the indicators willingness to fight, level of violent crime, and ease of doing business, since they are all rankings. For the notation of the indicators in SPSS see Table 2. KMO and Bartlett s Test These tests evaluate whether the correlations in the data are arranged in such a way that a factor analysis is appropriate. The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy ranges from 0 to 1, the higher values signify a greater suitability. Since the KMO is greater than 0.5 one can proceed with the principal component analysis. As one can see in Table 3., the KMO measure is 0.831, which suggests that the section of indicators is highly suitable to conduct a principal component analysis. Table 3. KMO and Bartlett s Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy..831 Bartlett's Test of Sphericity Approx. Chi-Square Communalities Df Sig..000 The high values in the communalities (Table 4) show the indicators/indices that will to a large extent explain the component results. Those indicators with less than.5 values don t adequately explain the components, and could be omitted in the further analysis. However, in this case all indicators are suitable for the principal component analysis. Table 4. Communalities 102 Initial Extraction LOG_IH LOG_CPI LOG_PPR WTF LOVC LOG_HLE LOG_OLE LOG_DBFAMLE LOG_RTDR LOG_IFB LOG_CO2E LOG_ND LOG_ITA LOG_SE LOG_PAR LOG_PCGDP EODB LOG_GI Total Variance There are as many components extracted during a

6 principal components analysis as there are variables that are put into it. Since 18 variables were used, there are 18 components (see Table 5) The eigenvalues are the discrepancies between the principal components. The first component will always explain the most variance and thus have the highest total eigenvalue. The following component will try to explain the remaining variance, and so on. In the rotation sums of squared loadings the last value in the cumulative % column is , which means that the first four components collectively explain % of the total variance. Rotated component matrix The results of the rotated component matrix (see Table 6) seemed promising. They showed that component 1 was mostly explained by the following indicators (listed in order of their importance): Healthy life expectancy, secondary school enrollment, overall life expectancy, per capita GDP, patent applications, and ease of doing business. In addition CO2 emissions (in metric tons per capita), corruption perception, and road traffic death rate loaded high on the first factor. This factor is representative of overall economic and human development. The second component is mainly described by intentional homicide, the Gini index, and the level of violent crime within a country. It clearly deals with the dimension of social violence and equity. The index for biodiversity, natural, and industrial disasters determine component three, which can be termed as environmental quality and frequency of natural disasters. The final component relates to the perceived political stability and security within a country. The indicator that loaded high was the perception of the population of the willingness to fight. Clearly, this dimension would benefit from additional statistical measures related to the political stability or the quality of the political system. Initial Eigenvalues Extraction Sums of Squared Loadings Rotation Sums of Squared Loadings Component Total % of Variance Cumulative % Total % of Variance Cumulative % Total % of Variance Cumulative %

7 Component LOG_IH LOG_CPI LOG_PPR WTF LOVC LOG_HLE LOG_OLE LOG_DBFAMLE LOG_RTDR LOG_IFB LOG_CO2E LOG_ND LOG_ITA LOG_SE LOG_PAR LOG_PCGDP EODB LOG_GI Country Scores Based on the analysis one can calculate values for the countries. After saving as variables, it became evident that scores had only been calculated for a small amount of countries. Table 7 shows 72 countries for which a score could be calculated. Table 7 displays the final results in a combined form as an overall ranking. For that purpose, each dimension was sorted according to its factor scores. The resulting four rankings were averaged for each country, and the countries were finally sorted according to this average ranking. As can be seen from Table 7, the results are mostly plausible: the highest overall welfare, according to this combined four-dimensional welfare index, was assigned to Norway, Belgium, Finland, and Switzerland. The lowest welfare was measured for Columbia, Venezuela, Philippines, and Cambodia. Some other results are less obvious: highly developed countries, such as the United States of America, Israel, or Chile have received only medium ranks in this combined index primarily, because they had low values in one or two of the dimensions. The United States of America, for instance, ranked poorly in the dimension of environmental quality and natural disasters. Israel had a low score in social violence (which is perhaps explainable), and a high negative score in dimension four which is rather strange, because it should measure willingness to fight for one s country. Chile had a positive score only in the overall dimension of economic and health development but negative scores in all other dimensions. Missing Country Information Although the principal component analysis displayed results which suggest that the indicators used explain several dimensions of welfare within a country, the scores for each country were difficult to obtain. It was only possible to gather data on educational attainment, environmental conditions, demographic 104

8 Norway Belgium Finland Switzerland Netherlands New Zealand Estonia Japan Australia Portugal United Kingdom Hungary France Malaysia Croatia Uruguay Sri Lanka Burkina Faso Republic of Korea Poland TFYR Macedonia Swaziland Israel Italy Greece Ghana Tunisia Jordan Singapore United States of America Belarus Namibia Mozambique Serbia Panama Azerbaijan Egypt Bangladesh Pakistan Albania Yemen Argentina Chile Mongolia Morocco Turkey Paraguay Dominican Republic Nigeria Nicaragua Syrian Arab Republic Bolivia (Plurinational State of) Uganda Kazakhstan Ukraine El Salvador Mexico Peru Spain Kenya Viet Nam Uzbekistan Ecuador Ethiopia Indonesia Guatemala Thailand Brazil Colombia Venezuela (Bolivarian Republic of) Philippines Cambodia Health, education and economic development Violence and equality Natural and industrial disasters and environment Willingness to fight conditions, economic development, and health for 72 countries, which makes an international comparison unrealistic. The countries where there is no available data have already been excluded by only using 197 countries with a population larger than 100,000. The main problematic is that the missing data is distributed among the countries. However, there are many countries that have missing data in 3 to 4 out of 18 indicators. This is why there are no scores available for these countries (see Table 8). 105

9 Countries Indicators Conclusions This paper lead to three main conclusions: 1. A systematic effort to collect welfare-related indicators in various dimensions for as many countries as possible has yielded a list of only 18 indicators or indexes, for which recent statistical information was available for a reasonable number of countries. Many interesting indicators, such as the scores of the PISA educational attainment study, are available for only a small number of (developed) nations. 2. The factor analysis carried out on these 18 indicators resulted in 4 significant dimensions, that could be described as (a) an overall development and health dimension, (b) a dimension related to social equality and violence, (c) a dimension related to environmental quality and the frequency of natural disasters, and (d) a dimension capturing the willingness to fight for one s country (which could be interpreted as identification with one s country). Combining the rankings in these four dimensions the 72 countries analyzed could be ordered into a plausible ranking, with Norway, Belgium, and Switzerland having the highest national welfare and the Philippines, Venezuela, and Cambodia the lowest. However, these results are plagued by a relatively high number of missing values for particular country-indicator combinations, which existed even in the reduced list of only 18 indicators. 3. Currently, it is unrealistic to develop an alternative international welfare index that could replace per-capita GDP, because statistical data on many relevant dimensions of human wellbeing are not available for a large number of countries or are of questionable reliability. The situation is particularly dismal concerning education statistics. The available indicators are either meaningless (such as the literacy rate), or they are available for only a small number of countries (PISA scores). Since education is a major dimension of human welfare, significant improvement in international educational statistics is necessary before alternative measures of human welfare can be constructed. References Blanchflower D., Oswald A., (2008), Hypertension and happiness across nations, National Bureau of Economic Research, Working Paper Dean J., (2007) National welfare and individual happiness: income distribution and beyond, Journal of Policy Modeling, 28 Paper Diener, E., & Suh, E. (1997). Measuring quality of life: Economic, social, and subjective indicators. Social Indicators Research, 40, Domitrovic B. (September 28, 2009) Gross Domestic Happiness?, The Wall Street Journal Erikson R. (1993) Description of inequality: The Swedish approach to welfare research (pp ).Oxford University Press 106

10 Halliwell J. (July 2002), How s Life? Combining Individual and National variables to explain Subjective Wellbeing, National Bureau of Economic Research Working Paper 9065 Weiner E. (September 9, 2009) From Bhutan to France: Gross National Happiness, World Hum Kahnemann D., et al (May, 2006), Would you be happier if you were richer? A focus illusion, CEPS, Working Paper 125 Krueger A., Schkade D. (2008), The reliability of subjective wellbeing measures, Journal of Public Economics Noll, H. H. (2004). Social indicators and quality of life research: Background, achievements and current trends in Advances in sociological knowledge: over half a century (Nikolaĭ Genov, 2004), Verlag fuer Sozialwissenshaften, p Ravallion M., Lokshin, M. (April 1999), Subjective Economic Welfare, The World Bank, Policy Research Working Paper 2106 Samuel H., (September 14, 2009) Nicolas Sarkozy wants to measure economic success in happiness, Telegraph Media Group Limited Sirgy, M. J., Michalos, A. C., Ferriss, A. L., Easterlin, R. A., Patrick, D., & Pavot, W. (2006). The Quality- Of-Life (QOL) research movement: Past, present, and future. Social Indicators Research, 76, Van den Bergh J. (February, 2007) Abolishing GDP, Tinbergen Institute Discussion Paper Van den Bergh J. (February, 2007) Abolishing GDP, Tinbergen Institute Discussion Paper Veenhoven, R. (1996). Happy life-expectancy: A comprehensive measure of quality-of-life in nations. Social Indicators Research 39, 1 5. Victor P. (18 November, 2010), Questioning economic growth, Nature, Volume 468, Page , Macmillan Publishers Limited 107