How Effective is Formalization of Land Rental Agreement for Dispute Prevention? Evidence from Randomized Control Trial in Uganda’s Refugee Camp

The number of refugees and internally-displaced people (IDPs) has been increasing in the world. According to United Nations High Commissioner for Refugees (UNHCR),117.2 million people will be forcibly displaced or stateless in 2023. Especially, the refugee population in sub-Saharan Africa increased by 1.1 million (22 percent) in 2017 mainly due to the crisis in South Sudan, from where more than 1 million people fled primarily to Sudan and Uganda (UNHCR 2019).

Uganda is considered one of the most tolerant countries for refugees. As of 2019, more than 1.4 million refugees and asylum seekers reside in the country (UNHCR 2019). Historically, Uganda has taken an open-door policy towards new refugee arrivals which gives refugees freedom of movement, the right to work and establish a business, and access to social services, such as healthcare and education. The Government of Uganda has allocated a small piece of land within the camp to new refugees. Local communities nearby the refugee settlements (called host communities) share resources and services with refugees, such as water, schools, and health facilities. Some of local Ugandans also rent or borrow out their agricultural land to refugees.

Given that most of the refugees rely on agriculture in their home countries, enhancing access to agricultural land is important to improve their livelihood and to promote refugees’ economic independence. At the same time, developing a peaceful relationship between refugees and host communities is crucial for the welfare of both parties.

The main purpose of this study is to investigate the impact of the formalization of the land rental agreement on the access to land for refugees and conflict prevention and resolution between refugees and hosts. To do so, we take the case of Rhino Camp, Arua district, Uganda where refugees have rented farmland from host communities but some of them have experienced land-related disputes. As there is no effective medication mechanism, NGO called Norweigian Refugee Council (NRC) has recently introduced and implemented a formalized land rental agreement in a part of the camp. Both tenants (refugees) and landlords (Ugandans) sign an agreement that states the rental fee and period in front of the village chiefs and the refugee block leaders who play the role of witnesses and mediators. Although this seems to reduce land disputes, this is not compulsory. As there is a transaction cost of signing an agreement, the landlords may refuse to rent out land to refugees who demand signing a formal agreement. Another methodological challenge rises as demand for a formal agreement can be affected by tenants’ experience of land disputes and decree of risk aversion, which makes us difficult to identify the causal impact on the prevention of future land disputes.

To overcome this challenge, we conduct a randomized control trial. First of all, we conduct a baseline household interview with 308 refugees in blocks where a formalization agreement has not been introduced. Then, we randomly assign refugees into two groups: (1) the control group which does not receive any intervention (105 refugees); (2) the treatment group which receives detailed information on formal land rental agreements and the support for making a formal land rental agreement (203 refugees). By using this randomly assigned treatment status as an instrumental variable (IV), we can investigate the impact of the formal land rental agreement on the access to land for refugees and conflict prevention and resolution between refugees and hosts.

2022-12-19

2023-03-31

Dummy variable of experiencing any land disputes, Dummy variable of experiencing unresolved land disputes, Dummy variable,experiencing threats, property damage, or violence in land disputes, Length of land conflict (Conditional on land disputes,Resolved land disputes (Conditional on land disputes), Land rental fee (Ush/acre)

Yield or value of output per hectare (for refugees), Income per hectare (for refugees), Household income (for refugees)

This study consists of a baseline survey, intervention, and follow-up survey in two populations. The first population is refugee households that are engaged in agricultural production (we exclude refugee households who do not engage in crop production but according to the pretest and interview for informants, most of the households engage in crop production). The second population is Ugandan households that reside in villages near the refugee camp and are potential landlords who own at least 5 acres of cultivable land and have rented out land to refugees. The sampling is done by stratified random sampling where the strata are the village.

First, we conduct household survey for 308 refugee households in 7 villages (44 refugee households in 7 villages) in Rhino refugee camp and 65 Ugandan households in the surrounding 13 villages. During the baseline survey, we collect the information on the experience of land disputes, access to land, and their income sources. Before the baseline survey, we will get a consent form from the interviewees by explaining our intention and that our survey is purely for academic purposes.

The intervention for refugee households includes information provision about the formal land rental agreement (its benefits and procedure) and supporting services to make formal agreements such as translation of the form, and arrangements among all people who sign the form (landlord, LC1 chairman, and refugee block leader). The intervention for Ugandan households is information provision about the formal land rental agreement (its benefits and procedure).

In each village, we select randomly 29 refugee households in our sample. These households are invited to the intervention (treatment group). This is just an invitation but is considered as recruitment to participate in the intervention. Fifteen sample households in each village do not receive any interventions (control group). Similarly,we randomly select half of the surrounding Ugandan villages and invite potential land loards to participate in the intervention (treatment group) while the other land load in the control villages are not invited (control group). The Ugandan household data are mainly used for descriptive analyses while the refugee household data are used for regression analyses.

Randomization done in office by a computer

For refugee, the randomization is done at household level, while for Ugandans at village (LC1) level.

No

seven refugee villages in the camp and 13 surrounding Ugandan villages (LC1s)

308 refugee households and 65 local Ugandans

For refugees, 105 households in the control group and 203 households in the treatment group. For Ugandans, 30 households in six control villages (LC1s) and 35 households in seven treated villages.

In this study, we will interview 308 refugees and 65 Ugandans in selected villages in the refugee camp and surrounding local Ugandan villages. Regarding the sample size, we did preliminary interviews in some zones in refugee camps. Although we did not cover the same villages, the situation is comparable to our sample villages according to our local research assistance. Among 18 refugee households whom we interviewed, 12 refugees have rented-in land so far. Among those who rented the land, 6 refugees have experienced conflicts with the landlord, suggesting that about 50 percent of tenants have experienced conflicts. Thus, we assume that the probability of experiencing conflicts without a formal contract is 50%. Theoretically speaking, people would not experience any conflict after they sign the contract. Although we cannot predict how many refugees agree to sign the contract, assume that 50 % of treated farmers would sign the formal land rental contract, while 10% of farmers in the control group do so. The probability of conflicts for control group is 0.9 x 0.5=0.45. On the other hand, the probability of conflicts for the treatment group is 0.5 x 0.5=0.25. For us to detect the impact of treatment to reduce the probability of experience from 0.45 to 0.25, we need a sample size of 100 for each control and treatment group. (We use STATA commend “power twomeans 0.45 0.25, sd (0.5)” to calculate the minimum sample size). Since our intervention is invitation and participation is up to the subjects. As assumed above, we set that 50% of treatment households make a formal land rental agreement. Thus, we should sample 200 treatment households. Given that we will conduct our survey in 7 villages, we will select the same number of households from each village, so that our sample size of treated households is 203 (29 households in each village) and that of control households is 105 (15 households in each village). By using this sample size, the minimun detectable effects is -0.168 ( calculated by using STATA command "power twomeans 0.45, n1(203) n2(105) power(0.8) sd(0.5) direction(lower)"). For Ugandan households, we will analyse the data desctiptively.