In the past decade, economists such as Daron Acemoglu, Abhijit Banerjee, Nathan Nunn, and James Robinson have empirically validated the primacy of ‘good’ institutions in driving beneficial political and economic outcomes. While this has been a great leap for academic economics, the applicability to policy is debatable. Specifically, as the empirical techniques employed generally exclude components of institutional variation that change over the short- to medium-run (see Rohini Pande and Christopher Udry), the respective findings potentially don't have much to say about what can be expected from deliberate attempts to generate 'good' institutions.
Serious empirical investigation of the effects of institutional reform remains scant, and for good reason. Rigorously identifying the effects of democratization – or any other specific reform – is extremely difficult, particularly at the national-level. When and where societies enact democratic reforms (such as in Eastern Europe in 1989), such reforms go part in parcel with sweeping changes in economic policy, institutional frameworks, and political actors (in the technical lexicon, such reforms are ‘endogenous’). This makes it almost impossible to isolate the effects of the reform itself from the effects of the multitude of other contemporaneous changes.
Opportunities do, though, occasionally present themselves at the local level. That is, there are, from time-to-time, instances whereby specific reforms are introduced asymmetrically across sub-national units and which are essentially uncorrelated with other factors (that is, ‘exogenous’). One such case occurred in Afghanistan in 2007, when the creation of democratically-elected, gender-balanced village development councils was randomized across 500 villages in 6 provinces. We had the opportunity to study this event and, in a recent paper, we report back on what we found.
We were particularly interested in understanding how the creation of the councils affected the most direct potential outcome: governance quality. But finding an appropriate measure of "governance quality" presented additional headaches. Measures based on on perceptions (e.g., “do village leaders act in the interests of the all villagers or for themselves?”) are often imprecise as respondents are often reticent to discuss sensitive local matters with outsiders. Such measures can also be statistically biased if a reform affects how respondents engage with outsiders or causes enumerators to be inclined to elicit particular responses (i.e., Clever Hans Effects). Alternatively, to proxy governance quality with the outcomes of specific governance activities (such as the resolution of disputes), one needs to be able to assign objective values to outcomes of such activities, a treacherous undertaking.
In rural Afghanistan, however, we got a bit lucky in stumbling across one common local governance activity which seemed amenable to quantification: distributions of externally-supplied food aid by village leaders to village households. That these distributions are often problematic is what made them particularly attractive from a research perspective. That is, they give village leaders a good degree of latitude (that tends to be exercised) between retaining aid for personal or familial benefit or distributing it to the indigent villagers for whom it is intended. Thus, by gauging whether leaders retain or equitably distribute, we thought we could obtain a decent proxy for the behavior of village leaders and thereby of the quality of local governance.
What was also nice about such distributions is that they avoided a couple of problems that sometimes befall similar efforts to generate ‘behavioral’ outcomes. For example, behavioral outcomes that are generated by abnormal events (e.g., public goods games) sometimes capture behaviors quite different from how people act in ‘real life’ (i.e., Hawthorne Effects). On other occasions, such outcomes can mix the effects of institutional change with capacity building, such as when measures overlap with the functions of newly-created bodies. By contrast, food aid distributions occur perennially in rural Afghanistan and are generally the domain of customary local authorities (such as headmen), thereby providing a measure of the effect on the equilibrium behavior of de facto village leaders.
An additional challenge was to isolate the different effects of the creation of elected councils. In addition to potentially changing the incentives which govern the behavior of de facto village leaders, councils can increase female participation in local governance (via the rule that they must be gender-balanced) and bestow de jure authority for governance functions to a new body (the council). Each of these effects is interesting per se. To separate them, we induced randomized variation in distribution procedures. In half of the treatment villages (which had elected councils), the council was mandated to manage the distribution, while the de facto village leaders managed the distribution in the other half. In half of the control villages (which did not have elected councils), senior village women were invited to participate, while the other half featured no such provision. By comparing outcomes across these four groups, we infer the individual effects of: (i) creating councils; (ii) mandating that councils manage the distribution; and (iii) involving women.
The results ended up being quite intriguing. The mandated management of distributions by elected councils improved the probability of aid reaching households that were objectively worse off. However, the mere presence of councils without an accompanying mandate that they manage the distribution resulted in increased leakage of aid and reduced villager participation in decision-making. Mandating female participation also increased leakage. Thus, in cases where responsibility for the distribution was not well-defined (i.e., in villages in which elected councils exist and it was not clear whether councils or customary leaders were responsible) or was otherwise shared between groups (i.e., between customary leaders and senior village women), governance outcomes tended to be worse than in villagers in which no councils existed and in which customary leaders oversaw the distribution.
So can democratization improve governance? The devil, per our results, is in the details. If institutional changes such as the creation of elected councils obfuscate chains of accountability, they can spur opportunistic behavior by local elites and produce a degradation of governance quality. However, if representative institutions generated by such reforms are given a well-defined mandate, outcomes do improve over functions for which such bodies are explicitly responsible.