We tracked 20 years of crop cycles and protests in Kenya – unrest follows a clear seasonal rhythm

- Source:

Agricultural cycles structure much of rural economic life in developing economies, where smallholder farming remains a primary source of employment and income. Smallholder farmers typically live on the edge of subsistence, often experiencing fluctuations in income, liquidity and labour demand over the agricultural year.

In Kenya, farmers have tended to complain about low commodity prices, delayed payments, land disputes, competition from cheaper imports, and inadequate infrastructure. This is particularly the case in the sugar, tea and coffee growing sectors.

Although these grievances often originate in rural areas, the resulting protests do not necessarily take place on farms or within rural communities. Protests often target the institutions and private actors governing the rural economy. Rather than demonstrating on their own land, farmers may gather at factory gates, along major roads, or outside county assemblies to press their demands.

These protests are distinct from demonstrations over food prices, taxation, land disputes, flawed elections and public services. Approximately 7% of all protests over the past two decades (my own calculations using the Armed Conflict Location and Events Dataset) relate to the agrarian setting in Kenya.

For a majority of rural households in Kenya, the year is structured around planting, crop management and harvesting. These phases shape when people are busy in the fields, when they have money and food, and when they face scarcity.

I'm a PhD student in economics studying how environmental conditions and development interventions shape economic behaviour and human welfare. In a recent paper, behavioural economics researcher Sophia Seelos and I examined whether the agricultural cycle influenced the timing of protests in Kenya.

Practically, if unrest follows a predictable seasonal pattern, governments, humanitarian actors and donors could anticipate periods of elevated risk rather than react to them.

To establish this, we combined two sources of high-resolution data for Kenya over the period 2000 to 2019. The first measures the changing greenness of plants and indicates crop development. We used this to determine the crop cycle. The second was the Armed Conflict Location Events Data (ACLED), which records the precise geographical locations of conflict events.

Combining the two data sources enabled us to:

• determine the timing of the harvest at the ward level, the smallest administrative unit in the country

• identify locally specific agricultural calendars rather than a single national season (the country has a lot of agro-ecological variation)

• analyse how protest behaviour varies throughout the crop cycle.

We found that protest activity in Kenya's agrarian areas followed a predictable agricultural rhythm. Protests increased two to three months after the harvest and were concentrated around sowing, at the beginning of the next growing season. This period was potentially marked by financial stress.

Our main conclusion is that social unrest in agrarian settings has a seasonal structure that can be anticipated. This has implications for policy. Early warning systems for conflict and unrest typically rely on prices, weather anomalies and political events. Incorporating local agricultural calendars could improve the timing of these predictions at little cost.

In addition, social protection programmes, food assistance and public works schemes could be scheduled to support households during the months between the depletion of harvest stocks and the start of the next season. This is when our results show the risk of unrest is highest. Doing so would address both the welfare needs of households and a period of heightened political tension.

Patterns of protest

The motivation of our study was twofold. First, we wanted to know whether people protested less during the harvest season, when they were busy in their fields and about to earn income, and more when food ran short and they faced liquidity constraints. Second, we wanted to know whether this made unrest predictable enough for governments and organisations to prepare for it.

Conceptually, the harvest cycle allows us to separate two mechanisms that are central to the literature on agricultural income and conflict but are rarely disentangled for protest behaviour. The first mechanism is what drives the choice between time allocated to work or protest – which economists refer to as opportunity cost of time. During labour-intensive farming periods, participating in a protest means forgoing work in the fields, so mobilisation should be less likely. The second mechanism is economic hardship. As stocks from the last harvest run down and households enter the lean season, grievances and food insecurity rise. This could make protest more likely.

To establish a causal relationship between the crop cycle and protest behaviour we used the variation in the timing of crop cycles, given that climatic conditions can shift major events such as harvest periods. We subjected the data to further analysis to account for persistent differences in agricultural cycles across locations and common shocks over time.

Because national elections are a major driver of mobilisation in Kenya, we also excluded election-related protest waves to minimise the possibility that those protests drove the results.

We concluded that protests increased two to three months after the harvest and were concentrated around sowing, at the beginning of the next growing season. This timing is informative about the above mentioned underlying mechanisms. Around and directly after the harvest, households have food and income and are occupied with harvesting, storing and selling crops. Both grievances and the time available for protest are limited. In the following months, stocks decline and the lean season approaches. At the same time, farmers have to buy seeds and fertiliser for the new season before any new income comes in. Many delay these purchases until just before planting because cash is short. This squeeze on household finances adds to economic grievances, while the period before planting also leaves more time for other activities.

The peak at the start of the new cycle is therefore consistent with a combination of reduced opportunity costs of time and rising economic pressure.

How the findings matter

These findings contribute to the existing literature in three main ways. First, they show that the economics of unrest are not only about unexpected shocks but also about predictable, recurring variation in living conditions.

Second, by exploiting locally specific calendars measured from space, the study demonstrates that remote sensing can capture economically meaningful variation – complete with location and timing – where survey data on income or food security are not available.

Third, the findings help reconcile seemingly contradictory results in the literature, where higher incomes are sometimes associated with more and sometimes with less conflict. The timing within the agricultural year matters for which effect dominates.

Nicole Robiro does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.

By Nicole Robiro, PhD Candidate in Economics at the Chair of Development Economics, Goethe University Frankfurt am Main