Published on Development Impact

Go Set a Watchman: Fear of crime and agricultural development. Guest post by Julian Dyer

This page in:

This is the third in our series of PhD students on the job market this year.

In early 2015, residents of Kiambu, Kenya caught a man stealing arrowroots from a farm. Had the chief not intervened, they would have set him ablaze to send a warning to other potential thieves. Indeed, fear of theft among farmers is pervasive, as is the perception that institutions are unable to protect them from crime. The repercussions of theft risk are substantial in developing countries, where we know that firms spend a non-negligible amount on unproductive security labour, and that farmers give gifts to neighbours to invest in relationships and protect themselves from theft. But could the direct costs of deterring anticipated theft be just the tip of the iceberg? If risk of theft influences other production decisions that farmers make, there could be substantial indirect effects on agricultural development.

In my job market paper, I randomly mobilised two hundred and thirty watchmen to six hundred farmers across seventy-six villages in Migori County, Kenya as part of a field experiment designed to show that fear of crime influences a number of farm production decisions. To understand if improved security can facilitate agricultural development, I randomized an intervention where farmers were matched with trusted outsider watchmen to guard their farms, at heavily subsidized wages, in order to alleviate their fear of crime during the 2018-2019 short rains season. To minimize the chance of crime being diverted from treatment households to those in the control group, I randomized the intervention at the village level, with approximately eight households in treated villages being assigned watchmen. I built a team of local coordinators who made sure the arriving watchmen were placed at the correct households (despite an inconvenient government crackdown on matatu minibuses) and a team of coordinators who made sure the watchmen were not diverted to other labour.

So, did the intervention work?

There was high (about eighty-five percent) take-up of watchmen among the intervention group, who were paying roughly 2.50 USD per week, and whose fear of theft decreased dramatically. Qualitative work indicated that crime was believed to target certain crops more than others, and farmers are particularly afraid of crime when they pursue agricultural practices that are high-value or different from the norm. Additionally, leaving the farm to pursue off-farm economic activity is seen to increase risk of theft. This means farmers might be missing out on marketing opportunities for their crops, as off-farm markets tend to give farmers higher sale prices.

The improved security led to behaviour changes - ones which might have implications for long-run agricultural development. Treated farmers were significantly more likely to have started growing a new crop, or allocated more land to a crop they were already growing, where they indicated that improved security was the reason for this change. This led to treated farmers reallocating a larger share of their land to such crops. In addition to cropping decisions, farmers in the intervention group also reported that they increased their crop sales at off-farm markets. These types of discouraged economic activity are ones that might be important for long-run transformation of rural economies, with farmers shifting to new crops and marketing strategies.

Did farm productivity actually increase?

In addition, the productivity of farms increased during the intervention season. The value of agricultural production per farmed acre was higher for the intervention group by about fifteen percent of the control mean. To understand where this yield increase came from, I split the crops into categories based on their objective characteristics. These categories capture the way farmers described anticipated crop theft in qualitative discussions, and which characteristics matter for the risk of a particular crop being stolen. Surprisingly, the strongest effect on yield comes from crops that aren’t highly vulnerable to being stolen by an opportunistic thief. There are a few mechanisms that could explain this. First, farmers could simply be misinformed about which crops were being stolen, and these yield effects simply indicate theft that is no longer occurring. This is perhaps unlikely, as the characteristics of these crops (such as cassava, which is a root crop with an attached shrub, and requires effort to harvest) mean they are not easily stolen. Another potential explanation is that reduced theft risk on the high expected theft crops allowed farmers to reallocate their time use. If farmers are able to reallocate their time towards other crops, this could explain why these crops see a yield increase.

It should also be noted that these results from a short-term intervention may be different from what we might observe with a long-term improvement in security. There are a number of farming investments that have long maturity periods, and would only be made if farmers know their plots would be secure for the long-run. Therefore, with a long-run intervention we might see larger changes in cropping patterns, and yield effects on different crops.

But what about other households nearby?

For an intervention like this, it’s important to think about the impact on other nearby households. Most farmers think the likely thieves who would steal from their farm are mostly within the village, and that suspicion of theft led to conflict among neighbours. I find that the intervention reduced the degree to which treated farmers were suspicious of their neighbours and others within the village committing theft while they were away from their farm. This is consistent with the perception that off-farm activities expose farmers to greater risk of theft. In addition, I find that the intervention group had about half as many arguments with their neighbours that were a result of them interfering on their farm. What’s more, these weren’t just minor disputes, as the intervention group had only forty percent as many disputes as the mean for the control group, when I restrict only to arguments that involved some form of threat or aggression.

From other work on crime interventions, including policing crime hotspots and automotive theft deterrents, we know that there is a possibility that crime is largely displaced to nearby households or villages. In this case, the pattern of spillovers is actually the opposite. Nearby farmers within the village reported less theft experienced than those near the control households, and a decrease relative to previous seasons. I also find no evidence of spillovers to the closest control villages compared to control villages further away. This is likely a context-specific result stemming from the fact that plot sizes in my sample are quite small, and the fact that there may not be sufficient variation in distance within the sample to capture all spillovers.


Why does this matter?

All this suggests that the security of farms has important effects on agricultural development. Improving protection from theft helps improve short-run productivity as well as facilitating behaviour that may lead to long-run development and agricultural transformation. This tells us that improving land tenure is not the full story for the importance of property rights institutions for agricultural development. The value of agricultural production increased by approximately 5,000 Kenyan Shillings (KSH) per acre, which is 15,000 KSHfor the mean farm size. The wages paid to watchmen during the intervention, however, were 18,000 KSH. Therefore, it is not clear that the private benefit to farmers justifies the cost of this intervention. As a result, the evidence suggests a clear policy case for improved formal security in order to facilitate agricultural development.


Julian Dyer is a PhD Student at the University of Toronto.

Join the Conversation

The content of this field is kept private and will not be shown publicly
Remaining characters: 1000