Scaling Capital for Small Businesses: Experimental Evidence on Differentiated Cash Grants to Formalized Microentrepreneurs in Chile

The intervention consists of two components: a non-reimbursable cash grant and access to a digital business training platform.
Cash grant (Financing). Participants in the two treatment arms receive a one-time, non-reimbursable cash transfer disbursed in two installments: 70% of the total amount approximately two months after selection (late January 2026), and the remaining 30% approximately four months after selection (late March 2026). Recipients may use the funds at their discretion for their business; the program imposes no restrictions on specific uses. The grant amount differs across treatment arms:

Treatment 1: CLP 1,500,000 (≈ USD 1,500)
Treatment 2: CLP 3,500,000 (≈ USD 3,500)

La Brújula ("The Compass"). All participants in the study — including the control group — are offered access to La Brújula, an asynchronous digital training platform developed by Fundación Luksic. The platform delivers content through text and video on topics relevant to small business management, including value proposition, sales channels, formalization (e.g., sanitary regulations, types of companies), people management, and accounting. Content is personalized for each user based on a diagnostic questionnaire (~50 items) completed during the application stage, which identifies the entrepreneur's strengths and weaknesses.
Treatment arms.

Treatment 1 (n = 235): cash grant of CLP 1,500,000 + access to La Brújula
Treatment 2 (n = 235): cash grant of CLP 3,500,000 + access to La Brújula
Control (n = 235): access to La Brújula only

2026-01-05

2026-04-01

All primary outcomes are measured at endline (October–November 2026), approximately 12 months after baseline:

Monthly business sales. Self-reported monthly sales for July, August, and September 2026. Each month will be analyzed separately, and a quarterly average (July–September) will also be reported.
Business costs. Direct and indirect business costs for September 2026, reported by component and aggregated as total costs.
Business profits (September 2026). Two measures: (i) self-reported profit, and (ii) constructed profit, calculated as monthly sales minus total costs.
Number of workers. Total number of workers employed in the business in September 2026.
Business survival. Binary indicator equal to 1 if the business is still operating 12 months after baseline, and 0 otherwise.

All primary outcomes are measured through the endline phone survey (October–November 2026). Parallel measures are collected at baseline (October–November 2025) using the same definitions, to allow the inclusion of lagged outcomes in the estimation as in de Mel, McKenzie & Woodruff (2014).
1. Monthly business sales. Self-reported total business revenue for each of the three months prior to endline (July, August, and September 2026), in Chilean pesos (CLP), measured through direct survey questions about each month separately. Four outcomes are constructed from these variables: (i) July sales, (ii) August sales, (iii) September sales, and (iv) the simple arithmetic average of sales across July–September. The same definitions are used at baseline.
2. Business costs. Self-reported business costs for September 2026, disaggregated by component into direct costs (e.g., inputs, supplies, raw materials) and indirect costs (e.g., rent, utilities). The exact list of cost components is defined in the survey instrument and matches the baseline questionnaire. Total costs are constructed as the sum of all reported components. Both individual components and total costs are analyzed.
3. Business profits (September 2026). Two parallel measures are constructed:

Self-reported profit: the entrepreneur's direct response to a survey question asking for the net business profit for September. Following the microenterprise literature (de Mel, McKenzie & Woodruff, 2009), self-reported profit tends to yield less noisy estimates than constructed measures and is reported as the leading profit measure.
Constructed profit: September sales minus total September costs (i.e., outcome 1 for September minus outcome 2).

Both measures are reported.
4. Number of workers. Self-reported total number of workers employed in the business in September 2026. The definition (e.g., inclusion of the entrepreneur, family workers, part-time vs. full-time staff) follows the baseline definition to ensure comparability across waves.
5. Business survival. Binary indicator equal to 1 if the business is reported as still operating at the time of the endline survey (approximately 12 months after baseline), and 0 otherwise. Operationalization is based on the entrepreneur's self-reported business activity status at endline.

All secondary outcomes are measured at endline (October–November 2026):

Labor protection. Three binary indicators capturing whether the entrepreneur contributes to (i) health insurance, (ii) pension, and (iii) unemployment insurance at the time of the endline survey.
Credit and debt. Indicators of (i) loan application, (ii) loan rejection, and (iii) outstanding debt, with amount and source reported when applicable.
Business management practices. Seven items measuring the adoption of management practices in the business (e.g., recording of sales and costs, goal-setting and monitoring, business planning). Each item is analyzed separately, and an aggregated index is constructed to summarize overall adoption.

All secondary outcomes are measured at endline (October–November 2026).
1. Labor protection. Three binary indicators capturing whether the entrepreneur contributes to (i) health insurance, (ii) pension, and (iii) unemployment insurance at the time of the endline survey.
2. Credit and debt. Indicators of (i) loan application, (ii) loan rejection, and (iii) outstanding debt, with amount and source reported when applicable.
3. Business management practices. Seven items measuring the adoption of management practices in the business (e.g., recording of sales and costs, goal-setting and monitoring, planning). Each item is analyzed separately, and an aggregated index is constructed to summarize the adoption of practices.

This study is a randomized controlled trial (RCT) with three arms. Random assignment was conducted on November 14, 2025, on 705 finalists selected from the broader pool of applicants to the seventh edition of Impulso Chileno (ICH7). Finalists were assigned in equal proportions to one of three arms:

Treatment 1 (n = 235): non-reimbursable cash grant of CLP 1,500,000 (~USD 1,500) plus access to La Brújula digital training platform.
Treatment 2 (n = 235): non-reimbursable cash grant of CLP 3,500,000 (~USD 3,500) plus access to La Brújula digital training platform.
Control (n = 235): access to La Brújula digital training platform only.

Assignment followed a stratified design with 12 strata defined by the interaction of three pre-treatment variables: gender (2 categories), macro-region (3 categories: Norte, Metropolitana, and Sur), and baseline sales level (binary, split at the sample median).
Data collection comprises three survey waves:

Baseline (October–November 2025): administered via phone prior to random assignment, covering business outcomes (sales, costs, profits, workers, survival) and entrepreneur characteristics (labor wellbeing, business practices).
Midline (June 2026): self-administered online survey covering sales and the number of workers, applied to the two treatment groups.
Endline (October–November 2026): administered via phone to the 705 finalists, replicating the baseline questionnaire to ensure comparability.

The design enables estimation of: (i) the effect of receiving a cash grant in addition to digital training (Treatment 1 vs. Control and Treatment 2 vs. Control); and (ii) the differential effect of a larger versus a smaller grant (Treatment 2 vs. Treatment 1).

Not available

Randomization was performed in the office by computer on November 14, 2025, using Stata's randtreat command (Carril, 2017), with the following specification:

Stratification by: gender (2 categories), macro-region (3 categories: Norte / Metropolitana / Sur), and baseline sales level (binary indicator split at the sample median, ≈ CLP 1,066,000), yielding 12 strata in total.
Equal-probability assignment to each of three arms (1/3, 1/3, 1/3) within each stratum.
Misfits (units that cannot be allocated evenly within a stratum) were handled with the global method; the procedure produced 18 misfits, which were reallocated globally.
Seed fixed for reproducibility (setseed(0)).

The final assignment yielded 235 individuals per arm. The randomization do-file and log are retained for verification.

Individual entrepreneur. Each of the 705 finalists was independently assigned to one of three arms (Treatment 1, Treatment 2, or Control) within their stratum. There is a single level of randomization; no clustering.

No

Not applicable — randomization is at the individual level (705 entrepreneurs). No clustering.

705 individual entrepreneurs.

235 entrepreneurs Control (La Brújula only), 235 entrepreneurs Treatment 1 (CLP 1,500,000 grant + La Brújula), 235 entrepreneurs Treatment 2 (CLP 3,500,000 grant + La Brújula).

Minimum Detectable Effect Size for Main Outcomes Power calculations are based on the main outcome variable, monthly business sales (CLP). Inputs were estimated from a sub-sample of 750 applicants from ICH5, selected by applying the ICH7 eligibility and ranking criteria: Mean monthly sales: CLP 1,082,416 Standard deviation: CLP 434,123 With 235 entrepreneurs per arm and 80% statistical power, the minimum detectable effect size (MDE) on monthly sales is reported under three attrition scenarios: Optimistic (0% attrition in control / 0% in treatment): MDE = CLP 112,432 = 0.259 SD = 10.4% increase in monthly sales. Intermediate (10% / 10%): MDE = CLP 118,400 = 0.273 SD = 10.9% increase in monthly sales. Pessimistic (20% / 10%): MDE = CLP 122,138 = 0.281 SD = 11.3% increase in monthly sales. Across scenarios, the MDE ranges from approximately 0.26 to 0.28 standard deviations, equivalent to a 10–11% increase in monthly sales.