Local education spending mandates: Indonesia’s 20 percent rule

ABSTRACT

Education spending mandates are popular around the world, especially among central governments in large, decentralized nations, who find that they may be useful in influencing varied local fiscal behavior. I evaluate Indonesia’s education spending mandate policy, which insists that district governments allocate at least 20 percent of their budgets to education. I find that the mandate has indeed induced some districts to spend more on education than they otherwise might have, thereby meeting the official target. However, the evidence indicates that the local spending mandate has had no significant effect on improving school participation and learning outcomes.

KEYWORDS:

• Spending mandates
• enrollments
• test scores
• regression discontinuity
• developing countries
• Indonesia

JEL Classification:

• H72
• I21
• I28

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Some subordinate agencies might already have satisfied the mandate target. i.e. even before it was officially instituted. In this case, a mandate’s main objective is to encourage those not yet meeting the target to do so.

2 See Kurniawati et al (2019) for an analysis of the extent to which government has so far been successful in improving education quality in Indonesia, with a focus on numeracy and literacy. The authors argue that education outcomes achieved so far do not justify the significant investments that government has made in the sector.

3 In 2001 there were only 348 districts. The number of districts has grown significantly since decentralization began. New districts are created when a single jurisdiction splits into two or more administrative units. In Indonesia, the process is known as pemekaran, and it is a common phenomenon. For an up-to-date analysis of the determinants of district splitting, see Pierskalla (2016), and for recent investigations of the (largely deleterious) effects of pemekaran, see Burgess et al. (2012), Lewis (2017) and Bazzi and Gudgeon (2021).

4 Islamic schools comprise about 15 percent of the total number of pre-tertiary schools. They are under the direct authority of the Ministry of Religion.

5 For an examination of the impact of a one-off increase in teacher salaries resulting from the implementation of government’s teacher certification program see de Ree et al (2018). The authors find that the salary increase had a significant impact on teacher welfare but no effect on student learning outcomes.

6 Currently the central government estimates that 37 percent of the general-purpose grant is used by provinces and districts collectively for education activities. This figure is not possible to confirm since subnationals have total discretion over how the general-purpose grant is spent and there is no way to discern from subnational budgets on which functions those funds are spent. In any case, the central government takes this 37 percent as part of its own education spending, along with all education-specific transfers to subnationals.

7 Provincial governments appear to be significantly less likely to adhere to the 20 percent rule (Lewis 2021). I do not discuss further provincial education budget shares in this paper.

8 Noncompliance rates are calculated as the number of districts with education budget shares less than 20 percent divided by the total number of districts in the data set in the relevant year.

9 This range of donut hole radiuses leads to a similar number of dropped observations as in Cattaneo, Idrobo, and Titiunik (2019).

10 Statistical power is estimated for an effect size equivalent to one half the standard deviation of the outcome variable for the control group (Cattaneo, Titiunik, and Vazquez-Bare 2019).

11 Note that the sum of the number of districts in Table 7 (572) exceeds the total number of districts in the sample (489). This is because some districts fall in both treatment and control groups, i.e. for some years they satisfy the mandate and for some years they don’t.

12 Since there may be some doubt about the exogeneity of intergovernmental transfers I interpret the estimated effects as associations and not causal impacts.