Risk sharing within the firm

How to cite

Shukla, Aruj; Vitali, Anna. 2026. "Risk sharing within the firm." AEA RCT Registry. AEARCTR-0019155. https://doi.org/10.1257/rct.19155-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Aruj Shukla* , Anna Vitali

* Primary investigator

Registry ID
AEARCTR-0019155
DOI
https://doi.org/10.1257/rct.19155-1.0
First registered
Last updated
Status
In development
Trial dates
2026-09-15 – 2027-08-31
Countries
Private
JEL codes
J24, J46, J63, L22, L23, D22
Registry record
http://www.socialscienceregistry.org/trials/19155

Abstract

In many low-income settings, wage employment at small firms does not fully insure workers against income risk: workers are often paid piece rates or through other output-contingent arrangements, leaving them exposed to firms’ demand risk. This may make wage employment less attractive, increase turnover, and reduce firms’ incentives to train and retain workers. This study tests these hypotheses through a randomized experiment with 900 tailoring, welding, and carpentry firms in Kampala, Uganda. Using a mystery-shopper approach, the research team will place standardized product orders with firms, randomly varying whether induced demand is stable or variable over time and, among firms facing variable demand, whether the order schedule is known in advance. By holding total induced demand constant across the treatment arms, this design isolates the roles of volatility and predictability on a range of firm- and worker-level outcomes, which are measured through high-frequency follow-up surveys during and after the intervention. The study will estimate impacts on firm performance, worker earnings and earnings stability, risk-sharing, worker retention and upgrading, and firms’ organization of production. In doing so, it sheds light on how demand volatility shapes worker turnover and how within-firm labor dynamics may constrain small-firm growth.

Intervention

The core intervention is a demand shock that experimentally varies the level, volatility and predictability of demand faced by small manufacturing firms in Kampala, Uganda.
Firms will be randomly assigned to one of three interventions:

Stable demand: the firm receives the same quantity of experimentally-generated orders each week, summing to a fixed total induced quantity over the intervention period.
Variable predictable demand: the firm receives a variable sequence of weekly orders that sums to the same total induced quantity, and is told the full order schedule upfront.
Variable unpredictable demand: the firm receives a variable sequence of weekly orders that sums to the same total induced quantity, but does not know the future order schedule in advance.

The interventions will be implemented through a mystery-shopper experiment where enumerators will anonymously place standardized product orders with firms, using standardized scripts and order protocols, in the tailoring, welding, and carpentry sectors.

Experimental design

The study uses a four-arm randomized controlled trial with approximately 900 firms. Firms will be randomly assigned to:

1. Control (C): firms receive no induced orders.
2. Stable demand (S): firms receive the same quantity of induced orders each week, summing to Q over the intervention period.
3. Variable unpredictable demand (VU): firms receive a variable sequence of weekly orders summing to Q, but do not know the future order schedule in advance.
4. Variable predictable demand (VP): firms receive a variable sequence of weekly orders summing to Q, and know the full order schedule upfront.

The pooled variable-demand arm V consists of VU and VP. The main comparisons are:

1. S vs. C: effect of receiving a smooth, stable demand shock.
2. V vs. C: effect of receiving a variable demand shock.
3. V vs. S: effect of demand volatility holding total induced quantity fixed.
4. VP vs. VU: effect of advance information within the variable-demand environment.

Secondary comparisons include VP vs. S and VU vs. S, which decomposes the effect of variable demand into predictable and unpredictable components relative to the stable-demand benchmark.

Design details

Randomization method
Randomization will be done in office using a computer
Randomization unit
The unit of randomization is the firm.
Clusters
900 firms.
Observations
The study targets approximately 900 firms and includes both firm owners and workers in the study firms.
Arms
The planned allocation is: 1. Control (C): 250 firms. 2. Stable demand (S): 250 firms. 3. Variable unpredictable demand (VU): 200 firms. 4. Variable predictable demand (VP): 200 firms. The pooled variable-demand arm contains 400 firms.
Primary outcomes
The key outcomes of interest are at the firm or worker level. Specifically, we study: Firm performance: Sales, costs, and profits will be first order measures of performance. We will also construct measures of sales and profit volatility by using the coefficient of variation (CV) of the respective base measures. Whether the firm accepted and delivered the order, the delivery time, product quality, and output per unit of input will also be included in this family of outcomes. Product quality will be assessed by independent experts in each sector based on a scoring rubric. Worker outcomes: Earnings, earnings volatility, hours worked, employment status, wage contract type, payment timing, and bonuses are the key worker-level outcomes. Organization of production: This family of outcomes assess changes in the organization of production at the firm. Task allocation, markups, owners’ time use, use of temporary workers, hiring or retention of permanent workers, renting or borrowing machines, and sharing orders or collaborating with other firms are some key outcomes of interest. Retention, upgrading, and human capital accumulation: worker exit, training, specialization, supervisory and customer-facing responsibilities, and worker willingness to accept outside opportunities will be primary outcomes of interest. The willingness to accept outside opportunities will be measured through an incentivized exercise in which we elicit workers’ willingness to pay/accept for a short-term self-employment opportunity. In addition to self-reported measures of training received by the worker, we will also capture the owners’ willingness to pay/accept for a week-long training program for a randomly selected worker at their firm. As the total size of the experimental order is fixed, but firms are heterogeneous in their size, we expect the experiment to induce differential demand shocks across firms of different sizes. For this reason, we will conduct heterogeneity analysis by baseline firm size, measured in terms of number of workers, revenues, and profits. Measures of risk aversion, for both owners and workers, as well as credit constraints (operationalized as prior/baseline access to credit) will also be used for heterogeneity analysis. Additionally, the impact of the induced demand experiment can be mediated by the geographical proximity to other firms in the same sector. Consequently, firm density—the number of other firms within a 500m radius—is another dimension of heterogeneity that is of particular interest to our analysis.

Cited by

No citation data available for this registration yet.

Keywords