Household survey data may not contain precise ages for all household members. Age heaping, an unusually high share of ages ending in 0 and 5, is especially common in survey data from developing countries. Age heaping can be caused by uncertainty of survey respondents about their own age or the age of other household members, intentional misreporting, or errors during data collection and processing. Errors in age data can affect the estimation of education indicators from household survey data because these indicators are often calculated for specific age groups. Examples include the youth literacy rate and school attendance rates for the population of primary and secondary school age.
An article on age distribution in household survey data on this site demonstrated age heaping in survey data from India, Nigeria and to a lesser extent Indonesia. Data for Brazil showed little to no age heaping. To investigate whether age heaping is more common among certain segments of the population, the survey samples can be disaggregated by household wealth quintile. For this purpose, the households in the sample are first ranked by wealth, from poorest to richest. The population is then divided into five equally sized groups with 20 percent each of all household members in the sample.
Figure 1 shows the age distribution by single year of age and wealth quintile in data from Brazil. The data were collected in 2006 with a Pesquisa Nacional por Amostra de Domicílios (PNAD) or National Household Sample Survey. No preference for ages ending in 0 and 5 could be observed for the entire survey sample combined and disaggregation does not change the result. The age distribution in each quintile is smooth, with no peaks at ages ending in 0 and 5. The only obvious difference between the population in the different quintiles is that poorer families tend to have more children, indicated by a peak in the age distribution in the younger age groups.
Figure 1: Age distribution in household survey data by single-year age group and household wealth quintile, Brazil
Data source: Brazil PNAD 2006.
Figure 2 shows the age distribution in Demographic and Health Survey (DHS) data from India. The data were collected in 2005-06. In contrast to Brazil, there is considerable age heaping in the Indian data. However, peaks around ages ending in 0 and 5 are more pronounced among poorer households. Increasing household wealth is associated with a decrease in age heaping.
Figure 2: Age distribution in household survey data by single-year age group and household wealth quintile, India
Data source: India DHS 2005-06.
Data from Indonesia, collected with a Demographic and Health Survey in 2007, are shown in Figure 3. At the aggregate level, the survey data from Indonesia exhibit little age heaping. However, disaggregation by wealth quintile reveals that reported ages ending in 0 and 5 are more common among poorer households.
Figure 3: Age distribution in household survey data by single-year age group and household wealth quintile, Indonesia
Data source: Indonesia DHS 2007.
Finally, Figure 4 displays data from a 2008 Demographic and Health Survey in Nigeria. Similar to India, there is a high percentage of ages ending in 0 and 5 in the combined survey sample. The disaggregated data show that age heaping occurs more frequently among poorer households but also exists in the richest wealth quintile.
Figure 4: Age distribution in household survey data by single-year age group and household wealth quintile, Nigeria
Data source: Nigeria DHS 2008.
Disaggregation of household survey data from Brazil, India, Indonesia and Nigeria has shown that age heaping occurs more frequently in data collected from poorer households. Wealthier households may have more access to birth registration and therefore may be able to verify their ages with birth certificates. Wealthier households are also likely to be smaller and survey respondents would therefore have to know and report the ages of fewer persons than respondents from larger households.
Age heaping in survey data reduces the accuracy of education indicators that are calculated for single years of age, for example for all children of primary school entrance or graduation age. However, indicator estimates for larger age groups, for example all children of primary or secondary school age, are less likely to be affected by errors in age data.
Related articles
Permanent URL: http://huebler.blogspot.com/2010/04/age.html
An article on age distribution in household survey data on this site demonstrated age heaping in survey data from India, Nigeria and to a lesser extent Indonesia. Data for Brazil showed little to no age heaping. To investigate whether age heaping is more common among certain segments of the population, the survey samples can be disaggregated by household wealth quintile. For this purpose, the households in the sample are first ranked by wealth, from poorest to richest. The population is then divided into five equally sized groups with 20 percent each of all household members in the sample.
Figure 1 shows the age distribution by single year of age and wealth quintile in data from Brazil. The data were collected in 2006 with a Pesquisa Nacional por Amostra de Domicílios (PNAD) or National Household Sample Survey. No preference for ages ending in 0 and 5 could be observed for the entire survey sample combined and disaggregation does not change the result. The age distribution in each quintile is smooth, with no peaks at ages ending in 0 and 5. The only obvious difference between the population in the different quintiles is that poorer families tend to have more children, indicated by a peak in the age distribution in the younger age groups.
Figure 1: Age distribution in household survey data by single-year age group and household wealth quintile, Brazil
Data source: Brazil PNAD 2006.
Figure 2 shows the age distribution in Demographic and Health Survey (DHS) data from India. The data were collected in 2005-06. In contrast to Brazil, there is considerable age heaping in the Indian data. However, peaks around ages ending in 0 and 5 are more pronounced among poorer households. Increasing household wealth is associated with a decrease in age heaping.
Figure 2: Age distribution in household survey data by single-year age group and household wealth quintile, India
Data source: India DHS 2005-06.
Data from Indonesia, collected with a Demographic and Health Survey in 2007, are shown in Figure 3. At the aggregate level, the survey data from Indonesia exhibit little age heaping. However, disaggregation by wealth quintile reveals that reported ages ending in 0 and 5 are more common among poorer households.
Figure 3: Age distribution in household survey data by single-year age group and household wealth quintile, Indonesia
Data source: Indonesia DHS 2007.
Finally, Figure 4 displays data from a 2008 Demographic and Health Survey in Nigeria. Similar to India, there is a high percentage of ages ending in 0 and 5 in the combined survey sample. The disaggregated data show that age heaping occurs more frequently among poorer households but also exists in the richest wealth quintile.
Figure 4: Age distribution in household survey data by single-year age group and household wealth quintile, Nigeria
Data source: Nigeria DHS 2008.
Disaggregation of household survey data from Brazil, India, Indonesia and Nigeria has shown that age heaping occurs more frequently in data collected from poorer households. Wealthier households may have more access to birth registration and therefore may be able to verify their ages with birth certificates. Wealthier households are also likely to be smaller and survey respondents would therefore have to know and report the ages of fewer persons than respondents from larger households.
Age heaping in survey data reduces the accuracy of education indicators that are calculated for single years of age, for example for all children of primary school entrance or graduation age. However, indicator estimates for larger age groups, for example all children of primary or secondary school age, are less likely to be affected by errors in age data.
Related articles
- Age distribution in household survey data
- Educational attainment in Brazil since 1920
- School attendance in Brazil
- Educational attainment in India, 1950-2000
- Primary school completion in India, 1950-2000
- Age and level of education in Nigeria
- Adult literacy in Nigeria
- Youth literacy rates
- Official school ages: primary, secondary, and compulsory education
- Demographic and Health Surveys (DHS)
- Brazil National Household Sample Survey 2006
- Education glossary by the UNESCO Institute for Statistics
Permanent URL: http://huebler.blogspot.com/2010/04/age.html
A new 
