Predicting Taliban attacks: Q&A with Jason Lyall

Winning the “hearts and minds” of Afghan civilians is a key component of American strategy in Afghanistan, where war rages more than 16 years after NATO-backed forces knocked the Taliban from power.

A recent study co-authored by Yale political scientist Jason Lyall provides evidence that “hearts-and-minds” aid programs have an unintended consequence: Taliban insurgents target villages where aid projects have gained traction.

Lyall, associate professor of political science, and his coauthors, Kentaro Hirose, assistant professor at Waseda University, and Kosuke Imai, professor of politics at Princeton University, published their findings in the Journal of Peace Research. 

Lyall, director of the Political Violence FieldLab at the MacMillan Center at Yale, spoke with YaleNews about his latest study. An edited version of the conversation follows.

How did this study originate?

I’ve been working in Afghanistan since 2009 and part of what I’ve been trying to do is figure out ways to measure people’s attitudes to really sensitive questions. I work on designing surveys that measure people’s support for the Taliban or the Afghan government using experimental techniques. These are sensitive issues and you can’t ask questions about them in a direct way without endangering respondents or having them lie to you.

In this paper, we tried to take the next step: Now that we know how to measure people’s attitudes, can it help to explain patterns of violence?  In particular, we wanted to learn whether knowing these attitudes could help predict where insurgents are going to attack six months or a year into the future.

It turns out that by adding our survey data into models of forecasting attacks, you can accurately predict the location and timing of an attack down to the village level for thousands of villages in Afghanistan.

What is the connection you found between aid programs and insurgent attacks?

One of our key findings was that he hearts-and-minds programs create targeting cues for insurgents. The hearts-and-minds campaigns in the areas we studied were actually starting to work. There was an increase in support for the Afghan government and a loss of support for the Taliban, but the Taliban wasn’t going to sit idly by and watch this happen. What they did essentially was watch where the attitudes were changing and use that information to inform their attacks. The areas that show the biggest increase in support for the government were the places that the Taliban decided to target.

It is not just that they attack those areas; they also shift their tactics. In areas that show the biggest increase in support for the government, the Taliban becomes most indiscriminate in its attacks. In areas with Taliban support, they tend to take precautions to protect civilians. They’ll enter a house at night and shoot the one person they have targeted. They will focus attacks exclusively on military targets. They will engage a NATO patrol or a patrol by Afghan forces. They will do so away from a village so as not to accidentally hit civilians.

In areas that are more pro-government, they are more willing to put improvised explosive devices on the roads and take the chance that it will hit a civilian bus instead of a NATO patrol. They use far more explosives in the IEDs. You see more daytime violence. Essentially, they relax prohibitions on targeting civilians as a way of punishing these villages for tipping toward the government.

How do your findings relate to previous research on aid programs in conflict areas?

It’s filling in a gap. Most of the work we do on assessing the impact of aid doesn’t really examine what happens in the aftermath of an aid program. We typically look at whether the services promised were delivered and if the aid recipients seem reasonably happy. Then we move on to the next site. There isn’t a sense of what happens at the sites six months or a year later.

The underlying assumption is that if you win hearts and minds, the people become hardened against the insurgents and either the insurgents won’t go into these areas anymore because their support has dried up or the population will actively resist and force the insurgents to move to easier areas to control. 

There is a sense that hearts-and-minds efforts tamp down the violence, and we’re finding just the opposite: In fact, they increase violence, at least in the short and medium term.  

What our study shows is that if you’re going launch aid programs in these environments, you have to think about the months and the years following the completion of the program because you are increasing people’s vulnerability. It’s not that the Taliban is staying away from these areas. The aid programs are actually attracting them. This is a piece of the puzzle that we have ignored in our impact evaluations.

Is there a need to rethink aid programs in these areas?

The tighter integration of security and aid is needed for sure. If you’re in areas where you cannot credibly promise security in the aftermath of an aid program, then you shouldn’t be programming there. You probably want to be choosing easier areas to work in.

It could also affect the types of programs you implement. USAID likes labor-intensive infrastructure projects where they’ll build a road or improve a canal system — projects that get young men out and doing productive work instead of engaging with the insurgency. Those projects also make visible targets for insurgents to hit.

On the other hand, another form of aid that USAID sometimes provides is cash payouts wired directly into people’s accounts. This kind of program doesn’t provide a visible target or easily visible evidence that a village has begun shifting its loyalty to the government.

In my assessment, if you still want to program in these environment, you might want to stay away from the big infrastructure projects that require the villages support and instead send cash aid directly to people’s bank accounts. That would lower the risk profile in these areas.

What is it like to perform fieldwork in Afghanistan?

The field research is the best part of the job. We partner with local survey firms to conduct random samples at the village and household level using local enumerators. We also work in partnership with local elders, commanders, and the Taliban, to secure access to the village we want to survey. Our teams enter under the protection of local power holders, including the Taliban, who give us permission to move within the village and conduct the survey.

We also draw on data provided by the U.S. military, especially declassified insurgent attack data known as “SIGACTS.” We augment it with secondary data from non-governmental organizations I have partnered with that have people in the villages collecting violence-related data. It is an amalgam of data from the thousands and thousands of attacks against NATO forces as well as the NGO data. Everything is pegged to the location, the day, the type of incident, the target, and casualties. It is very fine-grained and precise.  

You get the Taliban’s permission to perform surveys?

Most of the surveys I’ve done in Afghanistan occurred in highly contested areas so I really couldn’t perform the work with the Taliban’s permission. Oftentimes, the villages are in Taliban hands. The question is: What does the Taliban get out of letting me into a village? They want you in the village because they believe that if the villagers have an unvarnished way of sharing their stories, then the Taliban would be better understood. There is a sense that the Taliban struggles to spread its message and that allowing these surveys is a way of having the villagers express satisfaction with life under the Taliban.

In this particular project, we had no intention originally of doing predictions, so we certainly didn’t tell that we would use the information to predict their attacks.

Is it difficult to get civilians to participate?

There are certainly difficulties in getting civilians to participate. We take a number of precautions in our logistics and survey design to minimize risk to participants. And we’ve found that in may cases villagers are actually quite willing to speak to us, in part because they see this as a way of getting the government’s attention. These villages are often remote and people there feel largely forgotten. Anytime, they get to share their stories and talk about their daily lives, they’re happy to do it. They like the opportunity to tell their side of the story, and we’re there to listen to them. I’ve performed 150,000 surveys in Afghanistan and I’ve never had a response rate lower than 80%.

Two factors explain this high response: The first is Afghan hospitality norms, particularly in Pashtun areas. There is a tribal code called Pashtunwali that requires people to welcome guests. Secondly, the unemployment rate is off the charts. Normally, you miss people because they’re off at jobs. Where we work in Afghanistan, almost everybody is home almost all the time. You don’t have to chase respondents.

What is the next phase for this research?

The next stage is to see how things develop over time. This study is based on one survey, and we are only predicting nine months outward. We want to capture the dynamics of attitudinal changes over time and capture the currents as people’s attitudes shift over time in response to the dynamics of the war itself. If we’re right, we should be able to use this panel data — cross-sectional surveys over a long period of time — to track changing attitudes and to predict insurgent attacks across a much broader area of Afghanistan.

A second question we want to pursue is whether all aid programs attract insurgent violence or if, on the other hand, some types of program are better suited for these environments than others. Some aid programs may only have a temporary effect on civilian attitudes; others may have more lasting effects. Each type of aid program would therefore shape the kinds of Taliban violence we might expect to see in the days and months after the aid project concludes.

Written by Mike Cummings for YaleNews.