Why do workers choose informal work arrangements? Evidence from Latin America

Participants completed an online survey that included a discrete choice experiment. In each choice task, respondents were asked to choose between hypothetical job offers that varied along multiple attributes, including wage level, job formality status, work schedule flexibility, and degree of independence. The attributes and their levels were experimentally varied across respondents to estimate willingness to pay for formal employment and other job characteristics.

In each phases of the DCE, respondents are presented with a series of ten pairs of choices that they must choose between. Choice tasks are generated using Netquest's planned experimental design approach rather than pure random generation. In each phase, Netquest uses a D-efficient experimental design algorithm (Kuhfeld/Fedorov-style exchange procedure) to select 20 sequences of job pairs. This algorithm enumerates a candidate list of profiles, samples an initial random design, computes normalized D-efficiency, iteratively swaps profiles to improve D-efficiency, repeats from multiple starting points, and keeps the best design. Each respondent is randomly assigned to one design version and completes all jobs choices within that version.
The overall experimental design structure is common across countries; however, country-specific adaptations are implemented where necessary. In particular, salary ranges are adjusted to reflect local currencies and appropriate upper and lower bounds, and the definitions and labels of attributes are adapted to match country-specific institutional contexts.
In both phases of the DCE, the gross pay for hypothetical jobs are anchored to the respondents actual hourly earnings rate reported at the beginning of the survey in order to facilitate realistic choices. Specifically, we ask respondents to report their gross monthly labor income and the number of hours they work per week. For respondents who do not know their gross income and instead report net monthly earnings, gross income is recovered using country-specific mandatory taxes and social security contributions. All income information is collected and recorded in the local currency of each country.
We do not let the total working hours vary across jobs in the DCEs--both job options always have the same hours per week. The hours per week in the job options is set following one of two rule: 1) 40 hours a week for all job options; or, 2) all job options have the same usual working hours as the respondents' current job. To test for sensitivity to framing, we randomly assign half of respondents to evaluate job options based on their self-reported current working hours (across \textit{all} DCE comparisons), while the other half evaluate job options based on a standardized 40-hour work week.

2025-10-01

2026-03-02

The primary outcome variable is a binary indicator for whether a given job alternative is selected in each discrete choice task. Choices are analyzed within a random utility framework to estimate marginal utilities and implied willingness to pay (WTP), measured as trade-offs with gross salary.
In Phase 1, we estimate WTP for the following experimentally varied job attributes: (i) formal employment status (relative to informal employment), (ii) self-employment status (relative to informal employment), (iii) schedule flexibility, and (iv) autonomy.
In Phase 2, we estimate WTP for specific benefits associated with formal employment, including: (i) access to pension benefits, (ii) dismissal protections, (iii) paid vacation and sick leave, (iv) service bonuses, (v) employer-provided health insurance, and (vi) taxation parameters, including both willingness to pay to avoid taxation and preferences over a linear tax rate.
WTP measures are derived from estimated coefficients on non-wage attributes relative to the gross salary coefficient.

The study consists of a survey-based discrete choice experiment administered to adult respondents. Participants are presented with a series of hypothetical job choice tasks in which they select between alternative employment offers that vary along experimentally manipulated attributes.
The experiment is implemented in two phases. In Phase 1, job offers vary by employment type (formal employment, informal employment, and self-employment) and non-wage attributes such as schedule flexibility and autonomy. In Phase 2, job offers vary by specific benefits associated with formal employment, including pension access, dismissal protections, paid leave, service bonuses, employer-provided health insurance, and taxation parameters.
Attribute levels are randomly assigned across choice tasks according to a pre-specified experimental design.

The study is based on an online survey administered by NetQuest to nationally representative samples in Argentina, Brazil, Chile, Colombia, Ecuador, Mexico, and Peru. The target sample consists of 17,500 workers aged 18–65 who performed any paid work in the past month, including formal employees, informal workers, and the self-employed. Fieldwork is conducted simultaneously across countries and concludes on March 2, 2026.
After an initial module on demographics, employment characteristics, and access to benefits, participants are randomly assigned (1:1), with country-level stratification by education, employment formality, gender, age, and socioeconomic status, to:
(i) a treatment group receiving detailed country‑specific information on legal benefits of formal employment, along with comprehension checks; or
(ii) a control group, which does not receive this information until after Phase 1.
The study includes two discrete choice experiments (DCEs), each consisting of 10 binary choices. NetQuest generates 20 experimental designs using a D‑efficient Kuhfeld/Fedorov algorithm, and respondents are randomly assigned to one version. In both phases, hypothetical salaries are anchored to respondents’ own hourly earnings, and weekly hours are held constant—set either to 40 hours or to the respondent’s usual hours, assigned at random.
Phase 1: Respondents choose between jobs varying in contract type, schedule flexibility, autonomy, and salary (ranging from −50% to +50% of their base wage).
Phase 2: Jobs vary in gross salary, tax rate (0%, 10%, 20%, 30%), and eligibility for typical formal‑sector benefits: pension coverage, health insurance, paid leave and vacations, dismissal protections, and legally mandated bonuses. Institutional labels and definitions are adapted to each country.

Randomization is implemented automatically by the survey software. Attribute levels within each choice task are randomly assigned according to a pre-programmed D-efficient experimental design. Respondents are also randomly assigned to blocks of choice tasks by the survey platform’s built-in randomization algorithm. No manual intervention occurs in the randomization process.

The unit of randomization is the individual respondent. Participants are randomly assigned to receive the information treatment or to a control condition. In addition, attribute levels within each choice task are randomized by the survey software

No

N/A

A total sample of 17,500 participants is planned, approximately distributed as follows by country: 3,000 respondents in each country except Peru with 2,000 and Ecuador with 500.

Control arm: 50% of individual respondents (no information treatment)
Treatment arm: 50% of individual respondents (information treatment)