The setting: Sierra Leone, where the bottom layer of government is the chiefs. These chiefs can raise taxes, have a strong role in resolving disputes and allocate land (as they can in a number of African countries). In 1896, the British put in place a system of paramount chiefs to be elected by a "tribal authority" made up of ruling families designated by the British. This system (without the British part) persists into the present day and the paper has a great discussion of the institution and how it evolved. As you might imagine, some communities have more ruling families than others -- and it's this variation that's going to let Acemoglu and co. take a look at how political competition plays out.
Before explaining how they convince us that this variation in political competition is exogenous, let me rave a bit about the data. In order to capture a complete record of who the ruling families and chiefs were, they start with the archival records. But civil war and inadequate preservation mean that these are incomplete. So they set up a survey, sending out enumerators trained in qualitative techniques to the 149 chiefdoms in Sierra Leone. The first port of call for the enumerators are the "encyclopedias" - the local folks who are designated to keep the oral history. Enumerators take information from them and then go to each of the ruling families to check things out. This is a pretty nice example of how to capture institutions in some detail. They then complement this with a bunch of contemporary surveys which will give them the outcomes they are looking for.
OK, so we all also know that politics is one big endogenous mess. So to convince us that this is not a concern in this case, Acemoglu and co. start with pre-colonial development. Sure, we don't have local poverty numbers for Sierra Leone in the 1800s, but Acemoglu and co. use tax records (and this was a "hut tax" back in those days) denominated by area and 1963 population figures. This is not correlated with the number of ruling families. They also take a look for potentially selective migration. No correlation. What about economic shocks during the colonial period (Sierra Leone discovered gold in the 1930s). Here they get creative again, using mining permissions from the government. Not an issue. And they proceed to work through another couple of issues. One interesting result: distance to the 1907 railroad is in fact correlated (negatively) with the number of ruling families -- but it’s very small. Given some potential for geography to matter though, they're going to control for geography in their regressions.
So what do they find? In a nutshell, more political competition leads to better human development and economic outcomes (more on social capital in a minute). In terms of education, moving from the bottom to the top quartile of the number of ruling families gets you a 7 percentage point bump in literacy, primary school attainment and secondary school attainment. They also find higher weight for height z-scores (although this is only significant at 10 percent) and sharply lower anemia in more political competitive communities. They then turn to look at economic outcomes and find that folks in more competitive areas are more likely to have non-agricultural employment and score higher on a wealth index. And, in something close to my heart, they find that more competition among ruling families means that citizens are less likely to have to ask the chief for permission to use the land.
So it looks like political competition is good for some dimensions of both human capital and economic activity. What about the social dimensions? Here they come up with three sets of indicators: 1) bridging social capital (attending community meetings, meetings with the chief), 2) bonding activities between those of similar social status (group membership - e.g. ROSCAs, communal labor groups), and 3)collective action (participation in community road maintenance and other community projects). And lo and behold, these are all negatively correlated with more political competition. That is, more political competition depresses these indicators. This was a surprise for me. However, Acemoglu and co. offer an explanation: the chiefs who face less political competition (and are producing worse development outcomes), use these mechanisms to exert control and monitor what is going on. As they point out from the anthropological literature: "community meetings...are often used as a form of social control, and are used by elites to construct the appearance of governance based on community consensus, when in fact consensus has little to do with their decisions."
These countervailing effects show the messy interplay of political institutions and development. They also show the value of exerting some serious effort to map out institutions and how they evolve – the understanding of which is critical for effective long-term development.