Does writing a constitution lead to democracy? Sometimes—but not always. Post-Arab Spring, citizens of Egypt and Morocco ended up with “hybrid” regimes that mix some features of democracy with authoritarianism, as has been happening around the world. But in neighboring Tunisia, the new constitution did usher in profound democratic change.

That difference exemplifies a bifurcated global trend: Some constitutions unleash democracy—and others enable autocrats.

In a study we just published in The American Political Science Review, we studied 138 countries that had adopted new constitutions between 1974 and 2011. Three years after those constitutions were in place, 62 countries were more democratic—but 70 others were either less so or no better than before. In getting to democracy, we found, how the new constitution is crafted is as (or more) important as the language in its text. The process matters at least as much as the parchment.

Previous studies, including seminal work by Zachary Elkins, Tom Ginsburg and James Melton, have focused either on constitutional endurance or on the text itself. But our in-depth analysis shows that the level and depth of popular participation consistently makes the difference in whether that country ends up a democracy.

The ideal of constitutions is that they are governing documents that mediate differences, lay out rights, and establish common trust. But that vision has fallen on hard times.  Yes, after the fall of the Berlin Wall, Eurasia produced a series of iconic constitutions that underpin real democracies. But since then, we’ve been seeing more cynical “window dressing” constitutions in Africa, Latin America, and the Middle East. These constitutions focus as much on preserving dictators’ prerogatives as on planting the seeds of democratic norms and institutions. Even when the concepts are clear, the practice can be iffy: Consider efforts by presidents in Burundi and Burkina Faso to extend their term limits, signs of constitutionalism’s setbacks.

How do you write a constitution that can hold off dictators? 

There’s a reason we titled our “When Talk Trumps Text: The Democratizing Effects of Deliberation during Constitution-Making, 1974–2011.” Some of the best insurance against authoritarian backsliding is a broad-based, participatory constitution-making process with extensive public consultation.  That comes in three phases: drafting, debating, and ratification.

Democracy grows most fully, we found, when many different kinds of people help in the first phase: drafting a constitution. Bringing everyone in—from the business elites to the pushcart vendors, from conservatives to liberals, from constitutional lawyers to illiterates—while actually writing something is critical.

Our statistical tests show that democracy is much less likely to grow if citizens aren’t involved until the debate phase, when legislators and assemblies are discussing document drafts that have already been submitted, or during ratification, when the public either approves or rejects a new constitution in a referendum. What matters is who is in the room when the constitution is initially conceived and crafted—so long as all prospective drafters are allowed to express themselves and contribute to the debate.

For example, Venezuela’s first draft of a constitution was written by a 1999 constituent assembly stacked with Hugo Chavez cronies. Chavez called for the constituent assembly on the day he took office, established the electoral rules for picking the assembly members, and hand-picked more than 90 percent of them himself.  The result: no democracy.

By contrast, Tunisia did the opposite. “Right after the revolution, our main concern was how to avoid a return to authoritarian rule,” one of Tunisia’s lead drafters told us. “We decided to open the [constitutional] process from its initial stages because only an open, participatory process can secure a democratic transition.” Tunisia elected a constituent assembly composed of members from across society—from Islamists to secular liberals to leftists—who discussed and debated meaningfully, leading to a constitution that balances provisions advocated by each of these groups. Tunisia’s democracy has thrived as a result. According to the most recent Freedom House report, Tunisia now is the only Arab democracy with a political rights score similar to those of Western democracies.

An inclusive drafting process has lasting benefits for democracy. We believe that the drafting stage matters more because incumbents at the time of drafting know in advance whether they want to grab power by codifying benefits for themselves, or whether they must concede to citizen involvement that will fortify the nation’s democratic fabric.  Constitution writers’ intentions can be evidenced by the company they keep, and by the type of constitutional convocation that they convene.

Our findings emphasize that it is very difficult for popular debate at the ratification stage to fix a draft that was written at the top and only later opened to wider discussion and voting. By then, the basics of what’s on offer have already been decided: citizens can say yes or no, but cannot guide the substance. After the fact, it’s hard for society to make up for an early failure to include a wide range of societal interests, citizens, and political groups.

Requiring a referendum on a new constitution cannot substitute for real democratic involvement in writing it

Our findings should help policymakers consider what and where they should  support when considering nations’ new constitutions. Ratification through referendums has increased dramatically over the last few decades. According to data from the Centre for Research on Direct Democracy, between 1974 and 2012, there were more referendums (192) on new constitutions than there had been during two previous centuries (117). International donors have pushed these referenda, which do look thrilling, offering ceremonial images of democracy as smiling citizens hold inked-thumbs up at voting urns. But donors and policymakers should reconsider whether these votes are the best use of their investments.

Our empirical doubts about these referenda are consistent with the new thinking about “deliberative democracy,” a theory made prominent by scholars such as John Dryzek, John Parkinson and Jane Mansbridge. Those three and others argue that increased opportunities to vote are an inadequate substitute for more meaningful participation. Our findings decisively show that development assistance might be better spent helping convince constitution-drafting governments to welcome opposition parties, unions, human rights organizations, and interest groups into a robust debate.

Domestic and international energies should be devoted to avoiding the “original sin” of exclusionary deliberation in the drafting of constitutions, rather than investing in the civic religion of voting for whatever document results.

Todd Eisenstadt, professor of government at American University, is the author of “Politics, Identity, and Mexico’s Indigenous Rights Movements” and “Courting Democracy in Mexico: Party Strategies and Electoral Institutions.”  A. Carl LeVan, assistant professor in American University’s School of International Service, is the author of “Dictators and Democracy in African Development: the Political Economy of Good Governance in Nigeria” and can be found on Twitter @Dev4Security. Tofigh Maboudi is a Ph.D. candidate in political science at American University, where he is writing his dissertation on the rise of constitutionalism in the Middle East.