XN = (X – Worst)/(Best – Worst) Where: XN is the normalized value X is the actual value Worst is the highest observed value for the indicator Best is the lowest possible value for the indicator This technique ensures scores range between 0 and 1 and that all indicators are coded positively (i.e., higher scores indicate better performance). All indicators are weighted equally. The overall index score was calculated by summing across the normalized scores for each indicator and then dividing by eight. Scores were multiplied by 1,000 and rounded to three digits. Countries were sorted from high to low by this overall index score (with 1,000 being the best possible score) and then ranked from 1 to 172. To help ensure index scores are comparable from year to year and that countries can track progress over time, the best (i.e., minimum) and worst (i.e., maximum) values and the indicator weights will be fixed from this inaugural year onwards as follows: Ender Indicator Best Worst Weight Child dies Under-5 mortality rate 0.0 156.9 1/8 0.0 57.5 1/8 Child is severely malnourished Child stunting (%) Child is out of school Out-of-school children, adolescents and youth (%) 0.0 67.5 1/8 Child begins work life Child labor (%) 0.0 55.8 1/8 Child marries Adolescents currently married or in union (%) 0.0 59.8 1/8 Child has a child Adolescent birth rate 0.0 201.2 1/8 Child is a victim of extreme violence Population forcibly displaced by conflict (%) 0.0 65.4 1/8 Child is a victim of extreme violence Child homicide rate 0.0 32.8 1/8 Missing Values The End of Childhood Index includes all 186 countries and areas with recent data (from 2005 and onwards) for at least five of the eight indicators, but not all of them are included in the index rankings. Countries missing homicide data were dropped from the rankings.74 Countries missing up to any two of the other seven indicators were allowed to remain in the index. This was the only way to ensure a sufficient number of high-income OECD countries were included, as most do not collect or publish data on stunting or child labor. In total, 172 countries had sufficient data to be ranked. 92 countries have data for all eight indicators. 34 countries are missing one indicator. 46 countries are missing two indicators, 33 are high-income countries.75 74 75 76 77 Page 38 For countries lacking stunting, out-of-school or child labor data, the normalized score for their under-5 mortality rate was used in place of the missing value(s). Under-5 mortality is the leading indicator of child well-being and is strongly correlated with these three indicators.76 For countries lacking child marriage data, the normalized score for their adolescent birth rate was ascribed. These two indicators are also high correlated.77 Ascribing fillers for these missing values helped ensure countries weren't being rewarded for the lack of data and that data imputations for missing values were based on a country's performance on another, strongly correlated, childhood ender. Nauru and Tuvalu The End of Childhood Index table indicates the indicator(s) missing for these 80 countries. Correlation coefficients are 0.68, 0.76 and 0.74, respectively. The correlation coefficient is 0.81. Indicator Very low Low Moderate High Very high Under-5 mortality rate (per 1,000) < 10 10 to < 25 25 to < 50 50 to < 100 ≥ 100 Child stunting (%) <5 5 to < 20 20 to < 30 30 to < 40 ≥ 40 Out-of-school children and youth (%) <5 5 to < 20 20 to < 30 30 to < 40 ≥ 40 Child labor (%) <5 5 to < 20 20 to < 30 30 to < 40 ≥ 40 Child marriage (%) <5 5 to < 20 20 to < 30 30 to < 40 ≥ 40 Adolescent birth rate (per 1,000) < 15 15 to < 50 50 to < 100 100 to < 150 ≥ 150 Population displaced by conflict (%) <1 1 to < 2 2 to < 5 5 to < 20 ≥ 20 Child homicide rate (per 100,000) <1 1 to < 5 5 to < 10 10 to < 20 ≥ 20 ≥ 940 760 – 939 600 – 759 380 – 599 ≤ 379 Performance Band Prevalence Thresholds and Performance Bands Country-level performance on each indicator was assessed according to the thresholds in the table above. Countries with "moderate," "high" or "very high" prevalence of enders were color-coded as shown. To the greatest extent possible, indicator thresholds were based on international standards. The classification schemes used for U5MR and stunting are established.78 The same breakdowns for stunting were used for out-ofschool, child labor and child marriage. The breakdowns for adolescent births were adapted from those used by the World Bank and the WHO.79 Those for displacement were based loosely on categories used by UNHCR for a related indicator.80 The homicide strata were adapted from UNICEF and UNODC.81 To establish tiers or "performance bands," the boundary points between threshold levels were normalized for each indicator and then indexed. Index scores were rounded to two decimal places and then multiplied by 1,000 to give the cut-off points for each tier. Countries were placed into one of 5 bands according to their index score: Share of children missing out on childhood Index scores 1 Relatively few children ≥ 940 2 Some children 760 to 939 3 Many children 600 to 759 4 Most children 380 to 599 5 Nearly all children ≤ 379 Band Calculations As indicators are measured on different scales, each was first normalized using a linear scaling technique: Country Classification The End of Childhood Index presents data for the world as a whole and for various country groupings. Regions are classified as "developing" or "developed," as consistent with the United Nations system.82 Developing regions are further broken down into the subregions listed at the bottom of the index table on page 34. These groupings are based on UNICEF's nomenclature and regional classification. For a list of countries and territories in each region, see: UNICEF, The State of the World's Children 2016, p. 112. Regional data were sourced from the UN,83 with the exception of out-of-school, forced displacement and child homicide rates. Global and regional averages for these three indicators were calculated by Save the Children. Each was calculated as a weighted average, with the relevant population used as the weights.84 All UNICEF countries with available data were included in these estimates.85 Population coverage was at or above 89 percent for all region-indicator pairs.86 The designations employed in this report do not imply the expression of any opinion whatsoever on the part of Save the Children concerning the legal status of any country or territory or concerning the delimitation of its frontiers or boundaries. See, for example, data.unicef.org/topic/child-survival/under-five-mortality/ and data.unicef.org/topic/nutrition/malnutrition/ 79 See gamapserver.who.int/mapLibrary/Files/ Maps/Global_AdolescentBirthRate_2015.png and data.worldbank.org/indicator/SP.ADO.TFRT?view=map 80 UNHCR. Mid-Year Trends 2016. (Geneva: 2017), p.11 81 See UNICEF. Hidden in Plain Sight: A Statistical Analysis of Violence Against Children. (New York: 2014), p.39 and UNODC. Global Study on Homicide 2013. (Vienna: 2014), p.12 82 This designation is made for the purpose of statistical analysis only. For the list of countries belonging to "developed" and "developing" regions, see: unstats.un.org/sdgs/indicators/ regional-groups/ 83 U5MR and stunting data were pulled from data.unicef.org. Data for developed regions were pulled from the SDG Indicators Global Database (unstats.un.org/sdgs/indicators/ database/). All other world and developing sub-regional averages were pulled from SOWC 2016, with the exception of the global average for child labor, the CEE/CIS average for child labor and child labor and child marriage rates for South Asia, which were all pulled from SOWC 2015; results for 2016 were not available. 84 The official primary + secondary school-aged population in 2015 (UIS.Stat), total national population in 2016 (UN DESA) and child population aged 0-19 in 2015 (UN DESA) were used as weights for out-of-school, displacement and child homicide averages, respectively. The only exceptions were school-aged populations for Brazil (2014), Ukraine (2014) and South Korea (2016). Global aggregates were estimated in the same way for out-of-school and child homicide (i.e., as weighted averages of country rates). But the rate of forced displacement worldwide was calculated directly from the latest global count from UNHCR (unhcr.org/en-us/figures-at-a-glance.html) and world population data from UN DESA, both for the start of 2016. 85 Even if a country didn't have sufficient data to be included in the End of Childhood Index or if it had indicator-level data, it was included in global and regional rates for that indicator. 86 Coverage exceeded 99% for all estimates except for school-age populations in developed regions (89% coverage) and the Middle East and North Africa (90% coverage). 78 Page 39