What Household Surveys Tell Us About Education

This post originally appeared on the Global Partnership for Education's Education for All blog on September 29, 2014. You can see the original post here.

Household surveys provide critical information on countries’ education systems, but what can we expect going forward?

The post-2015 development discussions have increasingly focused on the role of data, with many calling for a “data revolution” to address development challenges. On this blog in particular, specific attention has been paid to the role of household surveys in generating information about pressing educational issues (see here, here and here).

At the Education Policy and Data Center at FHI 360, we welcome this increased focus on household surveys. Since 2004, we have published data extracted from two of the most comprehensive household survey sources, Demographic and Health Surveys (DHS) and Multiple Indicator Cluster Surveys (MICS), along with traditional sources such as the UNESCO Institute for Statistics and education ministries. These surveys have long provided a wealth of important information on educational levels within countries, and have contributed to policies and interventions to address common schooling challenges.

While the blog posts mentioned above are rich with ideas about the future of education data, we thought the discussion would benefit from a closer look at where we stand in terms of the type and breadth of information on educational systems generated from household surveys.

Household surveys provide useful insights for the education sector

As part of our standard data extractions, we document the information that is pertinent to our understanding of the education data coming from a specific survey.  The types of education data that come from these sources can generally be grouped into four categories:

• educational participation, including attendance and out-of-school rates
• attainment
• schooling efficiency, including dropout and repetition
• literacy rates.

For the present discussion, we looked at 73 DHS and MICS administered since 2007 to reveal important trends. [1]

Figure 1 Country coverage of the DHS/MICS used in our analysis (not comprehensive)

Of the surveys we looked at, all but one MICS contain data corresponding to the four major types of education questions, while only 38% (20 of 52) of the DHS include data pertaining to all four types.

Moreover, since 2011 only 6 out of 22 DHS include data pertaining to all four types, leading to speculation that DHS datasets will include less information on education as the program enters the seventh phase of its long history of survey administration.

What is left out? Data on dropout and repetition

It is clear that MICS, when compared to DHS, tend to offer the broadest look at a country’s educational system. The type of data that gets left out is on previous years’ attendance, meaning complete information on school dropout and repetition is more difficult to obtain in countries where DHS are administered. This is important because a particular advantage of household surveys is in showing subnational and group-level inequalities, such as those by urban and rural area or by household wealth.

With primary dropout levels remaining virtually unchanged since 1999, information from household surveys becomes all the more crucial (as a complement to official data from administrative sources) in crafting appropriate policy interventions to address schooling retention, as well as measuring progress on key EFA indicators.

Figure 2 Repetition rates at the subnational level, Ghana

With that brief overview in mind, here are a few points to consider moving forward:

• Don’t undervalue what we already have. Some of the previous posts contained valid points about the limitations of education data coming from household surveys, including that many are designed to measure health characteristics and only contain limited education questionnaire modules. However, it’s important also not to undervalue the information that we currently have—household survey data provides a wealth of information that can’t be obtained from other sources, and are used to inform policies and interventions, particularly to address issues around out-of-school children.
• Reach a consensus about the “bare minimum”.  Knowing that we currently obtain important information from household surveys, a valid concern is how to improve on this going forward. DHS surveys are likely moving to a more limited education module; an important consideration is the “bare minimum” level of information that donors, policymakers and those supporting education interventions should be happy with. A move away from including information on previous years’ attendance, for example, is a step in the wrong direction.
• A role for a dedicated education household survey? An interesting subset of this discussion has been around the merits of an education-focused household survey, versus surveys containing less data but administered more frequently. We would enthusiastically support an initiative such as a dedicated education survey, and believe the wealth of data that could be generated would be very useful. However, as others have noted, there are valid questions about the extent to which it would actually be used. An important component of any  “data revolution” would be to find ways to ensure that the data are used to inform policy.
• Creation of an international “education data” task force. One way to promote the use of data would be to create a task force, suggested in an earlier post and a welcome idea. In regards to household surveys, the task force could recommend the type of data to be collected, standardize indicator definitions and data sources, and create awareness around the release and findings of new household survey datasets. Such a task force should include a broad-based collection of stakeholders, which would ensure that datasets do not remain under the custodianship of a single donor or organization, and instead are disseminated widely for analysis.
• Supporting the statistical capacity of developing countries.  Finally, as important as our discussion about new sources of data has been, it is important not to forget about the lack of capacity in national statistical offices, particularly in sub-Saharan Africa. This is important for the simple reason that household surveys are built on sampling frames and population information from national censuses, but also because EMIS systems produce the bulk of data that are useful to policymakers working in education. Providing disaggregated EMIS data, linking to household and community-level characteristics (some surveys such as ASER already link household, community and school-level characteristics), and building the statistical capacity of country governments should be top priorities going forward.

[1] Please note that this is a rough analysis which does not correspond to the ‘phases’ of DHS and MICS, which are typically administered every five years. More information on the different phases of these surveys can be found on their respective websites.