Cash Transfers, Training, and Demand Spillovers in the Refugee Economy

At the core of our study are two interventions:

Unconditional cash transfers: In 65 percent of the sample, GiveDirectly is providing an unconditional cash transfer of USD 725 (KES 100,000). This is provided as a lump-sum payment, in two transfers. The direct theory of change for this intervention, as established in a large and ever-growing literature, is that transfers of this magnitude and timing enable investments in income-generating activities. This is particularly relevant in a context of credit market failure, wherein individuals have profitable, "lumpy" opportunities but are unable to borrow to finance these. As noted above, however, at scale such transfers may also have indirect effects: some of the transfer and the income it induces are consumed, and this creates demand for local goods and services -- a fiscal multiplier (Egger et al., 2022). This may in turn raise the profits of other locally embedded entrepreneurs, and may itself raise the return to cash-transfer interventions.

Entrepreneurship and financial literacy training. For half of those individuals receiving cash, the Equity Group Foundation will provide training in financial & entrepreneurship literacy and digital skills. The theory of change for such training presupposes that refugees face human and managerial capital constraints that prevent them from profitably expanding their enterprises; these may result in part from disruptions to education or losses of context-specific knowledge that arise from climate or conflict-induced migration.

2025-06-01

2025-10-14

We organize our primary outcomes into three families. Together with specific dimensions of variation and test statistics, these are invoked in our primary hypotheses.

First, Beneficiary income-generating activities. These are outcomes reflecting study participants' economic success, business performance, and investments of labor and capital. They comprise the following:

- Self-employment status: a binary indicator for engagement in an income-generating activity in the 7 days preceding the survey.
- Wage employment status: a binary indicator for spending positive hours in economic activities other than self-employment.
- Hours worked: the sum of hours worked across all income-generating activities. This excludes housework, childcare, study, and unpaid apprenticeship.
- Monthly self-employment business revenues: sum of revenues across all self-employment activities.
- Monthly business costs: sum of material input costs, labor costs, licenses, informal payments, insurance, utilities, and rent in the last 30 days.
- Business non-land capital: value of movable business assets owned by all self-employment enterprises of the respondent.
- Beneficiary wage income: the sum of income from all sources excluding self-employment enterprises.
- Beneficiary total income: an aggregate measure of business revenues minus business costs plus monthly wage income.

Across these outcomes, we define self-employment as when a respondent owns or operates a business or sells any other goods or other services for profit.

Second, Economic integration. For each of the following business relationship types, we measure (a) a total count and (b) a count of other-group members:

- Retail customers: total count and other-group count.
- Wholesale customers: total count and other-group count.
- Suppliers: total count and other-group count.
- Other business partners, mentors, investors, and authorities: total count and other-group count.

Third, Social integration.

- Refugee sense of belonging [refugees only]: index of four items measured on a 5-point Likert scale capturing refugees' sense of connection and acceptance in Kenya.
- Zero-sum thinking index [hosts only]: index of 12 items measured on a 5-point Likert scale, coded so that higher values indicate more zero-sum thinking about the relationship between refugees and Kenyans.
- Feelings thermometer: rating toward out-groups on a scale from 0 (cold/very negative) to 100 (warm/very positive).
- Intergroup interaction frequency: index of social interactions between refugees and hosts in everyday life, including:
- Frequency of contact with people from other countries.
- Count of people from other countries contacted for any social reason in the past 30 days.
- Attendance at religious services with people from other countries.
- Children attending school with children from other countries.
- Receipt of help from non-household members in the past month.
- Provision of help to others in the past month.

Economic integration outcomes: For each business relationship type, we measure a total count and a count of other-group members. Where the survey instrument measures the other-group share as a coarsened percentage range rather than a direct count, we take the midpoint of the range and multiply by the total count to obtain an estimated count.

Zero-sum thinking index: Index of 12 Likert-scaled items coded so that higher values indicate more zero-sum thinking about the relationship between refugees and Kenyans.

Intergroup interaction frequency: Index combining frequency of contact, count of out-group contacts, attendance at religious services with out-group members, children attending school with out-group children, receipt and provision of help.

In addition to the above primary outcomes, we will examine several secondary outcomes. We organize these into the following families.

First, Household economic welfare. These outcomes capture measures of economic opportunity and well-being at the household level, unless indicated otherwise.

- Total expenditures: annualized sum of food expenditure and non-food expenditure in the last 7 days, frequent non-food purchases in the last month, and infrequent purchases in the last 30 days.
- Food expenditure: household expenditure amounts in the last 7 days on rice, flour, bread, noodles/pasta, maize, beans/green grams/lentils, vegetables, fruits, milk and milk products, oil/cooking fat, sugar/sweeteners, tea/coffee, dry spices, eggs, meat, and fish.
- Outside home food expenditure: expenditure in the last 7 days on food consumed by the respondent outside of the household.
- Non-food expenditure in the last 7 days: individual expenditure on pre-paid mobile phone airtime, post-paid mobile phone service, transport (matatu, bus, tuktuk, or train), boda boda transport, water, electricity for home or personal use, and cigarettes or tobacco.
- Non-food expenditure in the last 30 days: individual expenditure on charcoal; kerosene, gas, petrol; soap; personal care items; household cleaning products; internet; newspapers and magazines; clothing or shoes; water and electricity.
- Annualized infrequent non-food expenditure: expenses in the last 30 days on donations, funerals, repairs or maintenance to home dwelling, repairs to household and personal items, healthcare expenses, and education expenses.
- Education expenditure: expenses in the last 30 days on education for self and others.
- Health expenditure: expenses in the last 30 days on health care for self and others.
- Food insecurity: Sum of the frequency-of-occurrence during the past 30 days for the 9 food insecurity-related conditions on the Household Food Insecurity Access Scale (HFIAS).

Second, Household net wealth.

- Household net wealth: sum of business assets, non-business assets (including land and buildings), savings, less debt.
- Value of household net non-business, non-land assets.
- Stock of savings: total value of savings across bank accounts, mobile banking, SACCOs, informal savings groups, merry-go-rounds, and other locations.
- Stock of debt: total value currently owed, by creditor source.
- Net land acquisition values.
- Household and business building investments (KShs).

Third, Financial integration and behavior.

- Uses a debit or credit card issued by their bank.
- Is member of a SACCO.
- Is member of a ROSCA, USLA, or other informal financial institution.
- Has an MPESA or other mobile banking account.
- Has a business insurance policy (including assets, income, or damages, e.g., vehicle insurance).
- Household has any form of insurance (including hospitalization or other health expenditure, income loss, funeral).
- Incoming transfers from extended family, friends, and neighbors (KShs).
- Outgoing transfers to extended family, friends, and neighbors (KShs).
- Stock of savings at Equity (end-of-quarter stocks, derived from Equity Bank ledger information). Note: quarters are defined as three-month intervals after the assignment to treatment.
- Cumulative flow of deposits into bank account at Equity (quarterly aggregates, derived from Equity Bank ledger information, excluding transfers from GiveDirectly).

Fourth, Psychological well-being. These are outcomes pertaining to the mental and psychological health of respondents.

- Depression and anxiety: PHQ-4 screening instrument.
- Life satisfaction: Cantril ladder of life.
- Happiness: self-reported happiness on a 4-point scale.

Fifth, Price mechanisms. These are outcomes pertaining to pecuniary effects through which cash transfers and training may operate on the market.

- Input prices: for firms established at baseline, we track a panel of inputs they purchase; we define this outcome as the log price of each input.
- Output prices: for firms established at baseline, we track a panel of goods they sell; we define this outcome as the log price of each good.
- Consumer prices: in each geographic neighborhood, we identify stores at which refugees and hosts purchase household consumption items. We impanel a set of firm-goods as the unit of analysis and measure log prices on a monthly basis.

Sixth, Geographic mobility and integration.

- Indicator for whether the respondent moves outside of their baseline neighborhood.
- Indicator for whether the respondent has a business outside of their baseline neighborhood.
- Indicator for location of a respondent business in a given neighborhood, analyzed at the respondent-by-neighborhood level, controlling for neighborhood of baseline residence and distance to it, and testing for impacts of saturation in own and other neighborhoods.

Seventh, Mechanisms of the Equity Training.

- Financial literacy index.
- Business practices index.

The study targets a sample size of 4000 individuals. Of these individuals 70% (approx. 2800) will be refugees, with the remaining 30% (approx. 1200) drawn from the host community. Both are recruited through local organizations serving the local populations, and subject to an income ceiling. The study design provides for the ability to answer the general equilibrium research questions while maintaining individual-level assignment of cash transfers and training to refugee and host study participants. To do so, we first assign clusters defined by neighborhood and nationality to a saturation rate; then, conditional on this assigned saturation rate, we assign individuals within those locations to direct treatment conditions of control, cash, or cash + training.

Not available

Randomization was done in the office using a computer

Assignment is conducted at the "neighborhood" level for saturation and "individual" for treatment assignment

We identify a set of 62 paired clusters of neighborhood and nationality, with each cluster-pair consisting of a refugee sample and a (potential) host-community sample who reside in the same physical neighborhood. Neighborhoods are then assigned to one of five saturation conditions, where these consist of a treatment frequency for both refugees and host-community members: H0, L0, LL, HL, or LH. Here, the first character denotes the refugee saturation rate, and the second denotes the host-community saturation rate; these are assigned to either H - High Saturation, L - Low Saturation, or 0 - no host-community sample in the neighborhood. When the host community is assigned a saturation rate of "zero", we do not sample host-community members at all. Sampled study participants are assigned to receive cash transfers with probability 86% (high-saturation neighborhoods) or 41% (low-saturation neighborhoods).

Yes

There are a set of 62 paired clusters of neighborhood and nationality, with each cluster-pair consisting of a refugee sample and a (potential) host-community sample who reside in the same physical neighborhood.

The study targets a sample size of 4000 individuals. Of these individuals 70% (approx. 2800) will be refugees, with the remaining 30% (approx. 1200) drawn from the host community.

- 1300 Individuals will be assigned to receive cash only
- 1300 Individuals will be assigned to receive cash + training
- 1400 Individuals will be assigned to the comparison - will receive cash only a year after