Yemen Situation Update: September 2023

Are peace talks impacting the level of political violence in Yemen?

In September, the sixth round of high-level peace talks between the Houthis and Saudi Arabia to end the conflict in Yemen began in the Saudi capital of Riyadh. Amid this new diplomatic push, armed clashes broke out in several areas of Yemen. While there was a slight decrease in the levels of violence overall,  an escalation in hostilities occurred before the talks. This pattern, often referred to as ‘escalating to de-escalate,’ has been observed in previous rounds of negotiations, occasionally resulting in an upsurge of violence after the talks.

This report examines confrontations between Houthi and anti-Houthi forces, seeking correlations between political violence and talks. Drawing from historical insights, we contend that talks with anticipated follow-up discussions are less prone to culminate in significant escalations of violence.

The Riyadh talks

On 14 September, a Houthi delegation made its first official visit to Riyadh since the start of the conflict in 2015. The five-day talks with Saudi officials were described as positive by both sides, and a second round of talks is expected to take place next month. 

The Riyadh talks took place within a backdrop of reduced hostilities influenced by the April 2022 truce and the subsequent ‘unofficial’ truce that began on 2 October 2022. Between April and October 2022, incidents between Houthi and anti-Houthi forces averaged 430 events per month. However, in the past 11 months, this figure has decreased to an average of 125 monthly events. Critically, this drop in the overall number of events has coincided with a nearly seven times increase in the lethality of violence (see graph below). This drastic increase may signal a more targeted use of violence compared to the less lethal ‘territorial-marking’ incidents that characterized the official truce period.

September 2023 saw among the fewest incidents between Houthi and anti-Houthi forces, as well as the second-lowest number of reported fatalities from political violence during the unofficial truce period. Arguably, the Riyadh talks, which lasted from 14 to 19 September, had an impact on reducing the monthly level of events and fatalities.

In the days preceding the talks, there was minimal fluctuation in the number of events. Yet, fatalities spiked between 9 and 12 September with 54 reported deaths — a quarter of the monthly total (see graph below). Most of these were recorded in Taizz, al-Dali, and Abyan, where anti-Houthi forces rebuffed Houthi assaults and the sides exchanged shellfire. This spike was followed by a sudden drop, with no fatalities for either combatants or civilians, recorded between 13 and 18 September. 

This trend mirrors an ‘escalate to de-escalate’ pattern that entails an escalation of hostilities before talks, which usually implies two primary aims: leveraging improved negotiating conditions and emphasizing the cessation of hostilities. However, in September 2023, the localized escalation near established frontlines suggests spoiler activity by Internationally Recognized Government forces, highlighting their exclusion from Houthi-Saudi negotiations. Without political inclusion, this escalation pattern may endure before future talks.

During the talks, there was little discernible change in the number of incidents, which remained largely consistent with the rest of the month, albeit less lethal. On 19 September, however, violence erupted in al-Dali, resulting in a number of Houthi fighters being killed. These were the only reported combatant fatalities during the talks.

The end of the Riyadh talks was marked by a resumption of hostilities, with political violence events recorded on 20 September across Taizz, Saada, and al-Hudayda, marking the highest daily total recorded in the month. In the remainder of the month, fighting slowed down and remained at low levels, even after a Houthi drone strike that killed several Bahraini soldiers in southern Saudi Arabia on 25 September. Despite being the first attack on Saudi soil since the April 2022 truce, the incident did not provoke a military response from the Saudi-led Coalition and did not result in any escalation on the frontlines, somehow demonstrating Riyadh’s determination to continue the peace process.

Trends similar to those observed in September 2023 occurred around talks in the past, particularly ahead of the Geneva peace talks in 2015 and Kuwait talks in 2016. The continued pattern of escalating to de-escalate during the latest round of negotiations suggests that this may be the shape of political violence during any future negotiations while the unofficial truce remains in place.

Key violence patterns around previous talks

The September 2023 talks represent the sixth round of high-level peace negotiations since 2015 (see image below). Over eight years, despite contextual variations, notable patterns have emerged in how the talks have related to military activities.

Spikes of activity before talks

Several rounds of peace talks were preceded by an uptick in battles and airstrikes as warring parties aimed to consolidate their territorial control or make gains ahead of potential ceasefire agreements. This occurred notably in 2015 and 2016.

For example, before the December 2015 Switzerland talks, the Houthis initiated an aerial campaign involving a barrage of missile and rocket attacks targeting Saudi Arabia that ceased in coincidence with the talks. Concurrently, the majority of the frontlines saw a brief intensification in fighting. In 2016, the Kuwait talks, which lasted from 21 April to 6 August, were also preceded by a significant spike in violence, most notable in the form of armed clashes and airstrikes in Taizz, Marib, and al-Jawf, as well as armed clashes in al-Bayda (see graph below).

Escalation does not always follow a pattern of abrupt surges followed by sudden drops during negotiations. At times, the decrease occurs more gradually. In 2023, the 8 to 14 April talks were foreshadowed by spikes of activity in al-Dali and Marib. Battles, in particular, escalated in March 2023, leading to a record-high number of fatalities in Marib since the beginning of the truce. Clashes dragged on until the first day of negotiations and then halted. Similarly, the most recent 14 to 19 September talks came at a time when incidents were starting to pick up in several parts of the country. This included Marib, where clashes saw a slight uptick from mid-July, having been almost non-existent since the April talks.

Four of the six rounds of talks have been preceded by a surge in political violence by Houthi forces in a recurrent pattern of escalation as part of political messaging to the other parties to the conflict. Based on this trend, any forthcoming talks may be pre-dated by an increase in political violence driven by the Houthis.

De-escalation during talks

Pre-talk escalations have not frequently carried over into the talks, which have often been accompanied by a decrease in political violence, despite the heightened military activity in the lead-up.

During the 2016 Kuwait talks, the negotiating period also corresponded with a notable decline in political violence. Notably, the talks coincided with the longest sustained ceasefire agreement prior to the 2022 truce, albeit daily fighting continued as a constant backdrop to the meetings, eventually contributing to their collapse. In May, during the first month of the talks, the number of armed clashes dipped to a monthly total of 71 events — a low point that would not be seen again until March 2022. Furthermore, April and May 2016, saw the lowest monthly fatality counts since the start of the conflict, only reaching similar levels once the United Nations-mediated truce began in April 2022. 

During the Kuwait talks, there was also a significant decrease in territorial changes between the warring parties. However, this trend abruptly reversed in August when the negotiations collapsed, resulting in the highest monthly total of territorial changes recorded in the conflict up to that point.

Talks also had a decisive impact on Coalition airstrikes. The first three months of the Kuwait talks — April, May, and June 2016 — saw the lowest number of airstrike events since the start of the conflict, with the numbers rising again as the talks concluded. A comparable pattern emerged during the Stockholm talks (6 to 13 December 2018), with a notable increase in airstrike events prior to the negotiations. This was followed by a gradual reduction during the talks, culminating in a complete cessation on the day of the signing of the Stockholm Agreement.

The April 2023 talks were supported by a more gradual decrease in hostilities, partly a result of the unofficial truce that remained in place despite its official expiration in October 2022. Taizz, in particular, saw a notable decline in political violence over the course of 2023, which reached its lowest point following the Sanaa talks. Political violence in Marib and al-Dali also dropped off during the meetings, despite seeing an initial spike at the start of the talks.

Overall, with the exception of the early talks in 2015, political violence has decreased during the subsequent rounds of talks, including the two Houthi-Saudi meetings this year. However, the de-escalation has not been absolute during any of the talks, even those supported by a concurrent ceasefire have still witnessed ongoing incidents throughout the negotiations.

Resumption in activity as a backlash to failed talks

The unsuccessful conclusion of five of the six rounds of talks has led to a surge in political violence in their immediate aftermath as the lack of material progress resulted in the parties to the conflict pivoting away from the political tracks back to a military-driven approach.

Despite a slowdown in hostilities at the beginning of the three months of the Kuwait talks, the parties’ confidence in the process appeared to wane as the discussions dragged on, translating directly into a re-escalation across almost all of the main frontline areas from late July. As the talks ended, the situation had deteriorated in several governorates. Clashes intensified in Taizz from late June; from mid-July, airstrikes surged in Saada, Hajjah, and al-Jawf, while the Houthis resumed cross-border missile attacks. Violence further escalated after the conclusion of the talks. 

Instead, the impact of the 2018 Stockholm talks was largely limited to al-Hudayda. Reports of artillery fire surged as soon as the talks concluded, remaining high for at least the next six months. Meanwhile, airstrikes picked up in Hajjah in the aftermath of the talks, as the frozen fighting on the West Coast moved further north and anti-Houthi forces supported a tribal uprising against the Houthis. As secondary conflicts sprung up in the wake of the al-Hudayda Agreement, and clashes also renewed in al-Jawf and al-Dali.

For other rounds of talks, the uptick in armed violence in the immediate aftermath of unsuccessful negotiation periods has been more acute. This was particularly evident in the aftermath of the first round of Geneva talks in June 2015, which saw a short-lived surge in armed clashes on the ground, with a 45% increase in armed clashes the week following the talks compared to the week of the talks. Similarly, airstrikes gradually picked up after the talks, increasing by nearly 12 times by early August.

Failed peace talks typically lead to a significant rise in political violence. Truce periods and ceasefires offer the warring parties a chance to regroup, recruit new fighters, and restock their weapons, contributing to heightened conflict levels. Furthermore, the collapse of institutional political initiatives points toward military strategies as the sole viable option. Even successful talks — such as the Stockholm ones that produced a halt to the West Coast offensive — were followed by the emergence of secondary conflicts that dragged in Houthi and anti-Houthi forces.

Looking forward

Based on the analysis in this report, we can draw general lessons regarding the influence of peace talks on conflict violence:

• Ceasefires during talks tend to be effective when parties have confidence in negotiation progress, as seen in the Kuwait talks and 2023 Houthi-Saudi meetings. If talks lack credibility, their impact on the ground is limited. The recent Riyadh talks have sustained the unofficial truce and low violence levels, even without significant progress in negotiations.
• After past talks, violence typically resumed between Houthi and anti-Houthi forces, often surpassing pre-talk levels due to waning confidence in the negotiations. The Stockholm talks were a partial exception, leading to the al-Hudayda Agreement and a broader but temporary de-escalation. Talks with an expected second round did not lead to a major violence spike, as seen in the 2023 meetings.
• The impact of future talks on violence in Yemen remains uncertain. Previous negotiation records are mixed, but recent trends show potential for reduced violence if talks occur alongside the unofficial truce. However, if talks fail, the risk of increased violence remains high.