As the era of the Millenium Development Goals draws to a close and the world turns to discussions of what comes next, localizing the development agenda is one key area for dialogue raised by the UN Secretary General’s report on the sustainable development goals. Listening to local voices in shaping global priorities is is no doubt key for the effective and sustainable allocation of development funds. But this also begs the question: If being responsive to those most in need is an increasingly local task, how do we measure progress in closing the gap without granular data? The need for subnational development data that is actionable and open has never been more apparent. In this post we drill down into why this matters in one particular example, education in Timor Leste, to demonstrate a broader point.
Timor Leste: A Case-in-Point
In Timor Leste, children may speak one of over 16 native languages at home, but at school they must learn Tetum, Portuguese, Indonesian or English in the classroom. The implication is that these kids face a double burden in the classroom - they must not only learn new subject matter, but also an entirely new language. The implications of this are serious for Timor Leste in achieving gains in eradicating illiteracy. For example, a 2009 World Bank study of early grade reading acquisition found that more than 70 percent of first graders could not read a single word of simple passages in Portuguese or Tetum, the official languages of Timor Leste.
So, what’s happened since the World Bank brought Timor’s silent illiteracy crisis to light? Have these new insights changed how donors coordinate their education dollars? Using AidData’s recently published geocoded data for Timor Leste, we can take a closer look into the development efforts of all donors reporting to the country’s Aid Management Platform (AMP) since 2010.
Visualizing the Problem: The Tetum Language and Illiteracy
To understand the issue in a broader sense, we’ll look at the allocation of education aid since 2010 as it relates to where those categorized as illiterate, were living at that time, according to the Timor Leste Census. In this case, illiterates are defined as people who speak a given language, but cannot read or write it (rather than a person who cannot speak, read or write that language). For this analysis, I’ve focused on the Tetum language, as according to the Census, there is a much larger population of people who can speak but cannot read or write Tetum (272,313 or 30% of the population) compared to Portuguese (36,346 or 4% of the population). It is interesting to note that nearly 25% of these 272,313 Tetum illiterates are between the ages 5 and 9.
Total number of illiterates by district in Tetum according to dot density.
In the map above, I assume that literacy provides a good indication of how much an education intervention is needed, and thus should be factored into where projects are sited at appraisal. Though you could extend this same type of analysis to include enrollment rates or even a multi-dimensional educational achievement index, it is simpler in this case to use literacy numbers for context.
Proportion of education aid dollars flowing into sub-districts since 2010: a total of $124 million in education aid.
The second map above visualizes the proportion of education aid dollars flowing to each sub-district since 2010 (all donors that reported to the AMP). This represents a total of $124 million in education aid. For example, the urban Dili sub-district of Vera Cruz receives a large portion of this bucket of aid--over 18% or about $23 million.
Efficiency of education aid flowing to 2010 Tetum illiterates.
In the final map above, I’ve compared the distribution of illiterates to the distribution of funding through the following equation: for each district, the difference in the proportion of total education funding received and the proportion of Timorese Illiterates living there.
Each district is actually represented according to standard deviation from the mean value. In this simplified, uni-dimensional world, perfect allocation would mean that the percentage of education funding should match the percentage of illiterates. Specifically, we would expect to see that areas with high levels of illiteracy are receiving greater amounts of education aid, while areas with lower levels of illiteracy are receiving less aid.
Observations and Takeaways
It is important to keep in mind that Timor Leste’s AMP does not represent 100% of the foreign funding, or even the total resource envelope of public spending in Timor. In an ideal world, I would use the combination of geocoded aid and geocoded domestic budgets to assess allocative efficiency. Despite this gap, even a limited analysis of the aid landscape can be used to facilitate discussion about how we can leverage spatial data to allocate development interventions more effectively and perhaps even inspire greater transparency for other types of funding for development.
It is therefore revealing to see that for a handful of districts in Timor Leste, the difference in funding and literacy need lies far from the mean. In particular, the Hatolia sub-district has the largest gap in proportional funding versus the proportional illiterate population. In 2010, Hatolia was home to 6.7% of Timor’s Tetum-speaking illiterates, but has only received 1.1% of the education aid since.
There appears to be a bias in funding allocated towards the the capital of Dili and the surrounding Dili district, which accounts for about 37% of the total bucket of education aid. There may be many reasons for this overfunding, such as the relative ease and lower operating costs of delivering educational public goods in an urban and semi-urban setting. Furthermore, it should be noted that the westernmost part of Timor-Leste is likely not conducive to this kind of analysis, as upwards of 90% of it’s inhabitants speak a different native language, Baikeno. But it’s certainly interesting to see the gap in funding in the sub-districts that lie to the south and west of the capital district.
A deeper, more robust, dive into the data might include teasing out rationales for the uneven distribution of funds in certain areas -- considering age ranges of the illiterate population as well as the specific activities of education interventions. For instance, given a dataset that is geocoded and activity coded, we could separate buckets of funding that seek to improve primary, secondary, tertiary, or non-formal education. Comparing the disaggregated activities funding to illiterates in different age ranges would offer an even more refined look at effectiveness and targeting.
Even in the context of a simplified, uni-dimensional analysis of allocative efficiency, these maps leveraging rich, sub-national data on development help illuminate issues about stakeholder coordination--particularly in the siting of projects and in the allocation of new resources against need. In a post-2015 world, a concerted effort on the part of donors, recipients, aid trackers, and open data champions is needed to make granular, geographic data more available and actionable. If we want to localize the post-2015 agenda, we should also be localizing the data used for benchmarking those goals.