Greetings from Jakarta, where I am visiting for a whirlwind 7 days of presentations, meetings, and the grand opening of the Jakarta office of the American Institute for Indonesian Studies.
While I’m here, I will be doing a presentation for the Australian Agency for International Development on local budgeting in Indonesia, joint work with two economists, one at ANU and the other at U of Sydney. We are interested in this funny feature of local government spending in Indonesia: local governments get lots of funds transferred directly to their accounts from the central government, and they are told to spend it on local infrastructure and things like that, but a lot of them don’t actually do this. They save it. It’s good for governments to save, yes, but these are saving rates that far exceed even IMF/World Bank benchmarks for how much is the proper fiscal surplus to maintain in good times. So here’s a puzzle: why so much saving? Especially given the obvious need for infrastructure spending, and the obvious incentives to target public funds to your cronies? (Well, one point is, these surpluses sit in relatively unsupervised accounts, which means that the funds–and the interest they earn–can be used for various less-than-aboveboard things.)
Our approach is political: we think that where there is robust political competition among local parties and elites, you are less likely to find these surpluses. For more on why that is, and how we can show it, you’ll have to read the paper; for you to read the paper, we’ll have to write it first! But in the course of this research we examined the factors that affect political competition, and one such factor is social cleavages.
So here is an interesting, and frankly puzzling, scatterplot of social cleavages (as measured by ethnolinguistic heterogeneity) and political competition (as measured by party fractionalization within local DPRDs).
When you look at this graph, you’ll conclude that, as you might expect, when ethnolinguistic heterogeneity is high, political fractionalization is high. But you’ll also find that when ethnolinguistic heterogeneity is very low, political fractionalization is high. A non-parametric lowess fit (the red line) illustrates this, and a polynomial regression (shown below the graph) is consistent with this interpretation.
I don’t have a ready explanation for why this is. One of the purposes of my presentation is to get some feedback from on-the-ground experts, but I’d also be happy to take suggestions from anyone who happens across this.