Developing country governments routinely use administrative data to estimate two key indicators, Gross Enrolment Rate (GER) and Net Enrolment Rate (NER), to measure changes in school participation and attainment of children. In Bangladesh, the Directorate of Primary Education (DPE) uses Annual School Census (ASC) data for policy planning purpose as well as calculating the GER and NER. Every year, DPE publishes the Annual Sector Performance Report (ASPR) which compiles data on GER and NER and many other indicators by gender. Based on such estimates, Bangladesh has done exceptionally well in bringing children to primary schools compared to other countries (Asadullah et al. 2014). A notable aspect of Bangladesh’s educational progress has been the reversal of the gender gap in school enrolment rates (Asadullah and Chaudhury 2009). However, a recent edition of the ASPR also acknowledges an important limitation of administrative data, namely, inability to directly identify children who are out of school.
This is because enrolment rate figures based on the 2010 Household Income and Expenditures Survey (HIES 2010) show that NER figure is considerably lower compared to what ASC data shows. However, being a sample survey, HIES data is prone to errors in reporting enrolment status of children; the accuracy of response rate may vary depending on who provided the data. Fewer errors are expected when the primary care-giver, the mother, is interviewed. Moreover, questions may be raised about the veracity of the respondents, whether data was actually collected from the sample households. Yet no study has formally attempted to verify the reliability of data on enrolment rates based on HIES.
In this note, we provide an updated picture on primary school enrolment status of boys and girls in rural and urban (non-metropolitan) Bangladesh based on a new nationally representative household survey data set. At the same time, this document will serve as a reality check for two reasons. First, we cross-validate the data on GER and NER as presented in ASPR by revisiting HIES households in a later year and independently gathering information on children's school enrolment and completion status. This approach has the added advantage of producing other policy relevant parameters with longitudinal information on a sub-sample of children. To this end, we exploit a unique survey experiment where about 5,560 of the HIES 2010 rural households where resurveyed. In the second part of the experiment, a sub-sample of the urban PSU’s in the 2010 HIES (but not necessarily the same households) were independently re-surveyed. Second, we examine enrolment status in heavily industrialised urban locations that are not adequately covered by national surveys such as HIES.
Sample and survey description: For a study on women’s life choices and social attitudes in Bangladesh, the 2010 HIES was used to develop a nationwide sample of households, which were visited and interviewed between May and July 2014. The sample for the 2014 study (henceforth called 2014 WiLCAS) included (i) all rural households in the 2010 HIES which had at least one female household member in the age-group 16-35 years; (ii) a random 50% of rural households in the 2010 HIES which did not have women in this age group; (iii) a stratified sample of urban households based on a census carried out in 87 non-metropolitan urban primary sampling units (PSU’s) throughout the country.This procedure yielded a sample of 7,974 households (1,436 in urban areas) and 6,293 individual interviews with women in the age group 20-39 years (1,557 in urban areas).
The household survey included data-collection on children’s schooling, including age of first enrolment, current enrolment status, and reasons for drop-out or non-enrolment. Mothers were asked to provide the same information for their own children during the individual interviews.
The 2014 WiLCAS can be combined with the 2010 HIES to create a panel of rural households from all 64 districts in Bangladesh to examine how school enrolment rates have evolved during this four-year period. In addition, for urban areas, 59 PSU’s in the 2014 WiLCAS are identical to those in the 2010 HIES (one from each district except Dhaka, Narayanganj and the Chittagong Hill Tracks). This allows us to construct a PSU-level panel in urban areas. A census was conducted in each of the urban PSU’s, including basic information on child schooling, which allows us to construct more precise figures on child school enrolment in these areas compared to what is possible using a household survey.
From the rural households in the 2014 WiLCAS, we obtain a sample of 3,745 children of primary-school age (i.e. 6-10 years), and 4,211 children of secondary-school age (i.e. 11-16 years). In the 2010 HIES, the same households included 4,423 children of primary-school age and 3,955 children of secondary-school age. From the urban census conducted for the 59 PSU’s common to the 2014 WiLCAS and the 2010 HIES, we obtain 7,312 children of primary-school age and 7,686 children of secondary-school age. The households included in the 2010 HIES from the same PSU’s include 1,789 children of primary-school age and 2,007 children of secondary-school age.
Main findings: Different estimates of GER and NER for boys and girls are presented in Table 1. Estimates based on the original HIES 2010 sample and the restricted HIES-WilCAS subsample are identical and expected as the samples are very similar. This therefore implies that we can compare estimates of GER and NER based on HIES with WiLCAS to assess whether (a) there’s progress over the last 4 years and/or (b) the HIES data under-estimates GER/NER. For rural Bangladesh, comparison with the HIES shows that by 2014, primary GER increased significantly and gender parity was maintained. In case of primary NER, the situation has slightly improved for boys by 2014, closing the 5 percentage point gender gap observed in 2010. Progress is also notable for girls in secondary education irrespective of whether we look at GER or NER.
Compared to rural locations, the GER and NER figures are much higher for urban non-metropolitan areas. Once again, our estimates of primary GER and NER are remarkably similar to that of HIES. This is also true for secondary NER though a slight decline is notable in case of secondary GER for girls. Overall, the estimates based on WiLCAS 2014 validates enrolment data from HIES 2010, and once again, confirms that GER and NER figures based on ASC data overstate the actual state of school participation and attainment in Bangladesh. This is also clear when we look at the NER figure for boys and girls (92% and 97% respectively) based on ASC data (see the bottom panel of Table 1).
Table 1: Estimates of Enrolment Rates in Rural and Urban Bangladesh by Gender
|2010 HIES-WiLCAS Subsample|
|Urban non-metropolitan areas|
Note: (a) N refers to the number of children in the primary/secondary school-going age (i.e. 6-10 years/11-16 years). (b) country-wide figures presented in the bottom panel are based on school census data. Figures for primary education are based on Annual School Census data whilst those for secondary education are obtained from BANBEIS http://www.banbeis.gov.bd/webnew/index.php?option=com_content&view=article&id=333:gross-and-net-enrolment-rate-&catid=60:key-performance-indicators&Itemid=179
The much lower NER figure (compared to ASC data) in our data indicates that a significant proportion of children leave primary school early. To formally investigate the issue, Table 2 presents crude estimates of dropout rate for rural Bangladesh. For children who were aged 6-10 years and in primary grades 1-5 in the WiLCAS-HIES 2010 sample, we re-examined their enrolment status in 2014. Only 46% of boys and 60% of girls were in grades 5-9 in 2014 suggesting a dropout rate of 32% for boys and 24% for girls.
Table 2: Estimates of Dropout Rates, Rural Bangladesh
|2010 NER||N in 2010||2014 NER||N in 2014||Dropout Rate|
Note: (1) Estimates based on children aged 6-10 years in HIES 2010 and 10-14 in WiLCAS 2014. (2) Only children from households present in both surveys (HIES 2010 and WiLCAS 2014) are considered.
M. Niaz Asadullah and Zaki Wahhaj