U.N. Secretary General Ban Ki-moon announced last week that the 2015 Paris climate change accords will, in all likelihood, go into effect this year. These accords commit the U.S. and dozens of other nations to try to lower greenhouse gas emissions to mitigate climate change.

But devising multilateral agreements is painstaking work. Typically, negotiators from hundreds of states deliberate for many years. Any agreement they reach often involves major compromises, which means provisions that experts consider important tend to be watered down

Nowhere is this more apparent than the case of the Paris Climate Change Accords. To break decades of multilateral gridlock on climate change, negotiators followed the mantra don’t let the perfect be the enemy of the good.

But is there a cost to being flexible?

Some researchers, including Rachel Brewster, caution that flexible agreement language can fail to commit states to anything substantial, turning stepping-stones into stumbling blocks. In contrast, Anne-Marie Slaughter and David Victor celebrate the more flexible strategy in the Paris climate change agreement.

It is hard to tell who is right here. Typically, states reserve flexible language for all the burdensome agreement provisions — the points that they expect will be hard to implement. So when states fail to comply with a flexibly worded provision, it is difficult to say whether this is because the language was flexible, or because the provision required burdensome changes in state behavior.

Here’s how we tested the impact of flexible language

In a recent article in International Organization, we put these ideas to the test by studying a different Paris accord — the Paris Principles on national human rights institutions. The Paris Principles offer a rare opportunity to assess the impact flexible language can have on compliance. Drafted in a haphazard and highly unusual way over just a couple of days in 1991 by a handful of practitioners with very limited information, the Paris Principles came into effect without governments watering down the final text. In 1993, governments adopted these standards wholesale in a nonbinding General Assembly resolution.

The Paris Principles call on all countries to establish national human rights institutions (NHRIs) and specify how to design these institutions. In the human rights field, where some question whether international agreements have any meaningful effect on the actions of states, it would be surprising to find that the principles had any effect, let alone effects traceable to the language of particular provisions.

But this is exactly what we do find. NHRIs across the globe are more likely to incorporate a provision when the Paris Principles use firm, as opposed to flexible, language.

Strong recommendations worked better

Despite the Paris Principles being nonbinding, we found a significant difference between strong and weak recommendations — i.e., provisions that state that NHRIs “shall” be given certain powers, as opposed to provisions that NHRIs “may” be given other prerogatives.

For example, a strong recommendation states that NHRIs “shall be given as broad as mandate as possible … set forth in a constitutional or legislative text.” A weak recommendation, in contrast, states that NHRIs “may be authorized to hear and consider complaints.”

We studied 22 such provisions in over 100 NHRIs, using a team of law students to help code NHRI charters, as well as diverse other sources. We found the probability that a country would adopt a particular provision increased greatly (from 55 percent to 74 percent) if the provision was strongly recommended. In contrast, the probability that a country would adopt a weakly recommended provisions either stayed constant or declined.

Some of our analyses are based on before/after comparisons; we found, for example, that the strongly recommended mandate to advise on legislation is included in 63 percent of NHRIs established before Paris and 79 percent of NHRIs established after Paris. But we were also able to run more robust fixed effects models to corroborate these findings.

Since 1993, politicians used slight variations in the language of the Paris Principles to justify both the inclusion and the omission of key safeguards. For example, Chile has one of the few NHRIs in Latin America that lacks the ability to hear individual complaints. In justifying this omission, Antonio Viera Gallo, the minister introducing the Chilean NHRI legislation, argued that, “according to the Paris Principles, it is not indispensable that institutions of this kind have the faculty to receive complaints.”

So our research showed that variation in linguistic flexibility had a significant effect on state behavior, even in a nonbinding standard.

Some governments set up NHRIs with limited authority

Some authoritarian governments that wanted to be seen to be doing something set up national institutions with no more formal powers than absolutely necessary — to signal technical compliance with the baseline standard. Authoritarian states exploited the flexible language of some of the Paris Principles’ provisions to cut back on safeguards they would otherwise create, and deny their citizens a more robust NHRI. As Matthew Stephenson cautions, our findings highlight risks as much as opportunities in pursuing compliance through flexible international standards.

What does our study say about the future of the Paris climate change agreements?

The 2015 Paris agreement under the UNFCCC pays close attention to language. A central concern to negotiators, especially from the United States, was which provisions to make firm (the “shall” statements) and which to make flexible (written as “should” or “encourage”).

The 2015 Paris agreement expresses provisions on cutting long-term emissions and transparency in firm terms but couches provisions on monitoring and adaptation (building resilience and reducing vulnerability to climate change) in flexible language.

Observers hailed the provisions on adaptation, alongside mitigation, as a victory for developing countries, who had long sought this concession from developed nations. However, these provisions explain that states will engage in adaptation planning “as appropriate” but do not include firm transparency and accountability directives. Their flexible language has left experts “unclear how effective they will be.”

So the good news is that firmly specified standards, even in a nonbinding agreement, can positively shift states’ behavior. But there’s a cautionary note: Compromises made through weak recommendations on controversial items will, at best, not help your cause. At worst, these compromises may actually undercut your objectives.

There is a real risk that the states will focus their environmental efforts on the clearly specified and firmly stated mitigation obligations — at the expense of flexibly worded, but no less vital adaptation tasks. As always, the devil is in the details.

Katerina Linos is a professor of international law at the University of California at Berkeley, and author of “The Democratic Foundations of Policy Diffusion.”

Tom Pegram is senior lecturer in global governance at University College London and deputy director of the UCL Global Governance Institute.