Christian and Muslim regions are similar.

The shutters are painted blue, and someone has written Bureau de vote N° 2 by hand on a sheet of paper, in case the printed banner above the door comes down. A cartoon poster on the right-hand shutter carries a line in capitals: Ne vend pas ta voix, vote selon ta conscience. Don’t sell your vote; vote with your conscience.

This is Ouagadougou on 29 November 2015, the day Burkina Faso held its first national elections since a popular uprising ended Blaise Compaoré’s 27 years in power. In September the presidential guard had attempted a coup, which collapsed within a week. About 60 percent of registered voters turned out, and Roch Marc Christian Kaboré won the presidency in the first round with 53.5 percent of the vote.

Burkina Faso is a majority-Muslim country. In the survey I use, 58 percent of Burkinabè respondents in the fifth round gave their religion as Islam. By the most familiar finding about religion in comparative politics, that should have made a day like this one less likely.

Is there something about Islam that makes democracy harder?

That question is a sharp version of a larger one, which runs through the dissertation this paper belongs to. Are the political habits of a people carried in its culture, so that some societies are better suited to democracy than others? Or are they shaped by the institutions people live under? Religion is a hard test for the second view, because religious identity is among the most durable things a person inherits. Here I move the question down in scale, from the world to countries, to the states of one country, to regions and finally to individual people, to see where the gap between Muslim and Christian societies actually sits.

The pattern is one of the most robust in cross-national research. Countries with larger Muslim shares of the population score lower on the standard measures of democracy, and the gap survives when researchers account for income, oil and gas, education, ethnic diversity, colonial history, gender equality and religious freedom. Five years after the Arab Spring, when I wrote the paper, scholars of democratisation still did not agree on why.

Most political scientists are sceptical of cultural explanations. Many observers, intellectuals and policymakers are not. Francis Fukuyama, writing about Algeria and Iran in 1992, called Islam a “major barrier to democratization”, and Samuel Huntington’s The Clash of Civilizations made a broader version of the case. Those who hold the view point to passages in the Qur’an, and to the popular support that shari’a and Islamist parties hostile to democratic institutions command in many Muslim-majority countries.

The cultural account deserves to be taken seriously, and some evidence seems to fit it. In the studies I draw on, the Muslim-majority states that regulate religion most strictly, such as Saudi Arabia, Sudan, Turkmenistan, Uzbekistan and Qatar, also allowed the least political competition, while more secular or religiously moderate societies such as Turkey, Indonesia, Senegal and Lebanon allowed more. Christianity offers a precedent for religion changing politics. After the Second Vatican Council in the early 1960s, the Catholic Church embraced religious liberty and human rights, and became a force for democratisation in Spain, Poland, Brazil and the Philippines.

Africa has its own version. In Senegal, Mali and Niger, the few Muslim-majority African countries that liberalised or democratised in the early 1990s, the movements for change were led by secular, French-speaking elites in the cities: students, teachers, lawyers and journalists. Clerics and religious groups in Mali and Niger protested against the new regimes, and argued at first that democracy did not suit their societies and was incompatible with Islam. Organised religion in the Muslim world has rarely been the democratising force that many Catholic and Protestant churches were in the third wave of democratisation, in countries such as Malawi, South Africa and Ghana.

So the hypothesis is a serious one. If Muslims are less likely to press for democracy, or more likely to support movements that oppose it, then places with more Muslims should be governed less democratically. Africa shows the familiar pattern between countries. The Afrobarometer, the survey I rely on, asks people how democratic they think their country is.

Comparing countries cannot settle the argument, because religion arrives together with other things that also shape political regimes, and I discuss three of them.

The first is oil and gas. Many Muslim-majority countries are rich in petroleum, and not only in the Middle East: Nigeria, Malaysia and Kazakhstan are large exporters too. Oil money pays for large armies and police forces, buys off critics and local elites, and cushions rulers against economic shocks without political concessions. Non-Muslim oil states such as Angola, Venezuela and Russia follow the same pattern.

The second is history. Many Muslim-majority countries were governed for centuries by Islamic empires with unusually durable, well-organised states. Some scholars argue that these states kept civil society weak, held off European colonisation until late in the nineteenth century, and left institutions that resist democratisation to this day. Others point to the waqf, the Islamic religious endowment, and its effect on how institutions developed over the long run.

The third is geopolitics. Fear of revolutions like Iran’s in 1979, and later of Islamist terrorism, led American and European governments to back friendly authoritarian regimes in Egypt, Saudi Arabia, Jordan and Kuwait. Elsewhere the end of the Cold War brought Western pressure for free elections. In much of the Muslim world it brought less of it.

Each of these travels with religion across countries, and there are only a few dozen Muslim-majority states to compare. Statistical controls can do only so much with so few cases.

The way out is to change the unit of comparison. Some explanations are tied to the national state. A foreign government supports a regime, not a province, and an imperial history belongs to the whole state. Other explanations can vary inside a country. Oil fields lie in particular regions, and at the regional level in Africa, oil and gas production is almost unrelated to religion, with a correlation of 0.06. Political culture, if it exists, lives where people live. It should be visible in the regions where Muslims or Christians are the majority.

There is a stronger reason to look at regions. Theories of political culture work from the bottom up. Citizens shape their governments in two ways. One is what they want from government, which I call the preference side. The other is how hard they work to hold officials to account, by following politics, voting, protesting and refusing to cooperate, which I call the accountability side.

Both operate most directly close to home. Most of what citizens experience of the state, in Africa as elsewhere, comes through its most ordinary functions, such as health, roads, schools and policing, delivered locally under imperfect control from the capital. Robert Putnam’s classic study of civic culture in Italy found its effects by comparing regional governments. If an Islamic political culture weakens democracy, it should show up in the regions where Muslims live at least as clearly as it does between countries.

A reasonable objection is that most African regions do not elect their own governments. Only a handful of countries, among them Nigeria, South Africa, Ethiopia and, from 2013, Kenya, have elected regional executives, and about a third have elected regional councils that share power with an appointed governor.

But appointment does not cut the link between citizens and rulers. Some appointed governors depend on strong local coalitions. National elections are administered, and manipulated, locally, and how freely people can meet, speak and organise depends on how coercion is applied on the ground. Cultural theories of democracy have never depended on elections in any case: their proponents apply them to China and the Gulf monarchies.

Before the full test, one country shows what the comparison looks like. Nigeria, Africa’s most populous country, is divided between a mostly Muslim north and a mostly Christian south, and its 36 states each elect their own governor. Start with the whole continent.

This fits what students of Nigerian elections had described before I wrote the paper: the northern Muslim states were generally regarded as more democratic on average than the Christian south, and especially than the Niger Delta, where vote buying, violence and political “godfathers” were notorious.

But Nigeria alone proves little. The Delta has oil, which brings its own politics, and in Nigeria, as in Mali, survey teams sometimes had to leave out areas that were too insecure to visit. One country can show what the comparison looks like. The test needs every country at once, comparing each region only with others in the same country.

I chose Africa for three reasons. The first is the range of religious settings. About 450 million Muslims lived in Africa when I wrote the paper, just under a third of the world’s Muslims and a little under half of all Africans. Some live in Muslim-majority countries in North and West Africa and the Horn. Others form large national minorities in West and East Africa. Most live in communities where they are the local majority, and some are minorities in capital cities and elsewhere.

The second is that African states vary a great deal inside their own borders. Many governments have struggled to project power into their hinterlands, and after independence they ruled different regions in sharply different ways. The experts who code the Varieties of Democracy data agree: of 54 African countries, only 6 have more even election quality across their territory than the world average, and only 14 have more even protection of civil liberties. If religion shaped regional politics, Africa is where it would have room to show.

The third is data. The Afrobarometer interviews nationally representative samples face to face, in the language of the respondent’s choice, with samples drawn region by region. Rounds 2 to 6, from 2002 to 2015, give 189,000 interviews in 36 countries, an average of about 400 people in each of up to 492 regions. Religion matters to the people interviewed. In the sixth round, the average Christian and the average Muslim respondent prayed, read the Bible or the Qur’an, or attended a religious service close to once a day.

If an Islamic political culture weakens democracy, regions with more Muslims should be rated less democratic than other regions of the same country, where the national government, the national history and the foreign alliances are all the same.

These are pictures rather than estimates, since regions differ in size, wealth and education, and the same region appears in several survey rounds. My models handle all of that, and add controls for each region’s distance from the capital and the border, its size, whether it contains the capital, its share of the national population, and approval of the president. Here is what they find, one step at a time.

The regional result could still mislead in two ways. Muslims might grade the same government more generously, perhaps because they compare it with governments elsewhere in the Muslim world. Or the theory could be right about people and wrong about places. The interviews can check both, by comparing Muslims and Christians who live in the same region, under the same local officials.

Put together, the theory of political culture that motivated the test predicts, if anything, that Muslim-majority regions should be slightly more democratic, because their residents watch their rulers at least as closely and want the same kind of government. The reason there is no regional democracy gap is simple. On the measures the theory says matter, Muslims do not carry a distinct political culture.

A regional average is mostly the majority’s view. The people best placed to see whether a region is governed democratically for everyone may be those who do not share its majority religion. So I rebuild the rating from local religious minorities alone: Christians in Muslim-majority regions, Muslims in Christian-majority ones, and people of other faiths or none wherever they are outnumbered.

Pooling all rounds, 274 regions in 31 countries have more than 15 such respondents. Almost all of North Africa drops out, because its regions are so uniformly Muslim. The minorities’ ratings track everyone else’s closely, with a correlation of .90.

Minorities’ ratings of electoral integrity and political freedom show the same pattern. Their reports of local corruption and bribes do not. I read this as a possible problem with the democratic accommodation of religious minorities, likely connected to the much greater religious uniformity of Muslim-majority regions.

In northern Nigeria and on the semi-autonomous Zanzibar islands of Tanzania, shari’a has been applied to family law, criminal law, inheritance, gambling, alcohol and dress. There have been reports of persecution of Christians in Kenya, Mali, Senegal and The Gambia, and more widely in the Horn of Africa and North Africa. And as the last chart showed, Muslims are slightly less accepting of neighbours of another faith.

There are reasons for caution. Minority samples are small, so their ratings are noisier, and the comparison is between minorities living under different majorities, not between minorities and majorities. But the result should not be explained away. A region can be governed as democratically as its neighbours for most of its residents, and less so for some.

As for the question I started with, where ordinary citizens meet the state, regions with more Muslims are governed about as democratically as regions with fewer, by the judgement of the people who live there. Muslims judge their governments as critically as Christians do, watch them at least as closely, and want democracy as much. Cross-country comparisons, and many policymakers, attribute part of the democracy gap to the political culture of ordinary Muslims. The evidence from African regions suggests that direct effect is overestimated.

That does not make religion irrelevant to democracy in Muslim-majority countries. If the national pattern is not spurious, it seems to operate at the level of the state. It may run through the legacy of rule by Islamic empires, through Western support for authoritarian governments that are not Islamist, or through national religious organisations.

The Christian contrast may be organisational more than cultural. Churches are hierarchical organisations that can concentrate their efforts on national political reform, as the Catholic Church did after the Second Vatican Council, whatever the values of individual believers. And the Sykes–Picot Agreement of 1916, the partition of the Ottoman Empire and the territorial gains of the Islamic State are reminders that many states in the Muslim world have artificial borders, and an uncertain hold on loyalties that also run to the wider community of Muslims, the ummah.

Africa may not generalise. Islam spread in much of Africa through Sufi brotherhoods organised around individual preachers, which blended Islamic teaching with older local practice, and in Europe and the United States Sufism is often seen as a tolerant form of the faith. But the evidence from South and Southeast Asia points the same way, and the Middle East and North Africa, where Salafism has the most sway, are home to less than a fifth of the world’s Muslims. The African experience may be closer to the typical Muslim experience than the Middle Eastern one is.

This matters now because the belief that Muslims as a group carry a political culture unsuited to democracy still shapes arguments about immigration in Europe and North America, and about policy towards the Middle East and the Sahel. The paper cannot settle those arguments, since it covers perceptions in African regions between 2002 and 2015. What it shows is narrower and more useful. When Muslims and Christians live under the same local institutions, they judge those institutions alike, want the same kind of government, and know at least as much about who governs them.

For anyone who wants more democracy in Muslim-majority countries, the evidence points to states: their histories, their resources, their alliances, and how they treat the people who do not share the majority faith. Elsewhere in the dissertation, culture follows institutions in the same way. On the Danish–German border, trust converged within two generations of a change of state. In African regions, the most inherited of identities makes little difference to how people judge, watch and want the governments closest to them.