Can Behavioural Nudges Improve Social Protection Uptake Among Informal Workers? Evidence from a Field Experiment

This study tested whether simple mobile phone messages could encourage more informal gig workers in India to register for e-Shram, a government program that provides social security benefits. The intervention targeted app-based workers like delivery partners and drivers, who often lack formal protection but may be eligible for government support.

Over a three-week period in June 2025, participants received one message per week through SMS or WhatsApp, in their preferred language (Odia, Hindi, or English). The messages aimed to reduce confusion, improve awareness, and motivate action. Three types of messages were used:

One group received general information about the benefits of registering.

Another group received messages tailored to their job type, emphasising that they were eligible.

A third group received peer-based messages highlighting that many similar workers had already registered.

A fourth group did not receive any messages and served as a comparison. The study measured whether these messages increased registration rates, improved understanding of the program, and reached workers who otherwise might have been left out.

This field experiment evaluates the effectiveness of behavioural digital nudges in increasing registration for e-Shram, India’s national social security platform for informal and gig economy workers. The study sample consists of 1,740 platform-based workers across two cities in Odisha, India, recruited through a screening and baseline survey conducted in April–May 2025.

Participants were randomly assigned to one of four groups:

T0 (Control): No intervention.

T1 (Generic Information Nudge): Received three weekly SMS/WhatsApp messages emphasising the general benefits of e-Shram registration (e.g., accident insurance, social security linkage), using simplification and salience framing.

T2 (Personalised Eligibility Nudge): Received three weekly messages that referenced the respondent’s occupation (e.g., driver, delivery agent) and stated their likely eligibility. These used personalisation and relevance cues.

T3 (Peer Comparison Nudge): Received three weekly messages stating that "many workers like you have already registered," invoking social norm and framing effects.

All messages were designed in three languages (Odia, Hindi, English) and pre-tested for clarity and tone. Messages were sent during 2nd, 3rd, and 4th week of June 2025), and follow-up surveys were conducted in July 2025 to assess outcomes including e-Shram registration (self-reported and optionally verified), recall of the message, and intermediate behavioural mechanisms (e.g., trust, motivation, perceived eligibility, peer spillovers).

The study tests whether digital nudges delivered at scale can overcome behavioural and informational barriers to accessing voluntary welfare schemes. It also explores heterogeneous effects by gender, digital access, trust in government, and platform type.

2025-06-09

2025-06-29

e-Shram Registration Status (binary):
Whether the respondent registered for the e-Shram platform during or after the intervention period (self-reported in follow-up survey, with optional verification through screenshot or ID).

Message Recall (categorical):
Whether the respondent remembers receiving a message, and which type (generic, personalised, peer-based). Measured in the follow-up survey.

Perceived Eligibility (binary):
Whether the respondent believes they are eligible to register for e-Shram.

Registration Intention (ordinal scale):
Willingness to register in the next 30 days if not yet registered (on a 4-point Likert scale).

e-Shram Registration Status will be captured through a direct follow-up survey question: “Have you registered for e-Shram since June 1, 2025?” (Yes/No). Respondents who answer “Yes” will be asked to provide optional verification (e.g., screenshot of confirmation or copy of e-Shram ID). Self-reports will be the primary measure, with verification used in robustness checks.

Message Recall will be measured by asking:
(a) “Do you recall receiving any SMS or WhatsApp messages in June 2025 about e-Shram?” (Yes/No); and
(b) “What did the message say?” with pre-coded options corresponding to T1 (generic), T2 (personalised), and T3 (peer). Accuracy of recall will be cross-checked against group assignment.

Perceived Eligibility will be derived from the question: “Do you believe you are eligible to register for e-Shram?” (Yes/No/Not sure). A binary variable will be constructed coding “Yes” as 1 and all other responses as 0.

Registration Intention will be constructed from the follow-up question: “If not already registered, how likely are you to register for e-Shram in the next 30 days?” Responses on a 4-point Likert scale (“Very likely,” “Somewhat likely,” “Not likely,” “Definitely not”) will be coded from 3 to 0 for analysis.

Message recall

Intention to register (stated)

Trust in government (Likert scale index)

Perceived eligibility for e-Shram

Motivation to act (proxied by action planning or goal setting)

Peer registration awareness (social spillovers)

Use of official digital platforms or help from others

Self-reported digital fluency

Message recall is assessed through a multiple-choice question about message content.

Intention to register is measured via a direct question at endline asking whether the respondent plans to register within two weeks.

Trust in government is constructed as a standardised index from responses to four Likert-scale items covering trust in welfare schemes, institutions, and digital platforms.

Perceived eligibility is a binary indicator based on whether respondents believe they qualify for e-Shram.

Motivation to act is proxied by whether respondents report setting a goal or timeline for registration.

Peer registration awareness is a binary variable indicating whether respondents know someone in their network who registered recently.

Use of official platforms/help tracks whether respondents accessed e-Shram services via UMANG or CSC centres, or received help from peers.

Digital fluency is measured by a composite score of self-reported ability to read/send SMS/WhatsApp messages, use government apps, and navigate online forms.

Motivation to Register will be derived from responses to: “How important is it for you to register for government schemes like e-Shram?” (scale of 1 to 5). A continuous motivation index will be constructed if appropriate.

Digital Access Constraints will be measured through binary and categorical questions on smartphone access, internet reliability, and ability to navigate digital forms (e.g., “Do you have regular access to the internet on your phone?” Yes/No).

Peer Spillovers will be measured by asking: “Did you discuss e-Shram registration with anyone after receiving the message(s)?” Responses include friends, family, coworkers, or no one. A binary variable (1 = discussed with anyone, 0 = did not) will be used.

Use of Government Platforms will capture whether the respondent used any official digital portals (e.g., UMANG, DigiLocker) in the past 30 days. Coded as binary (Yes/No), with follow-up on frequency.

Self-Reported Skill Loss will be constructed from responses to: “Compared to before the pandemic, how confident are you in your professional or academic skills?” Responses will be on a 5-point scale and dichotomised as needed for analysis.

This study evaluates the impact of behavioural digital nudges on the uptake of e-Shram registration among gig and informal workers in Odisha, India. The trial uses a four-arm parallel design with individual-level random assignment. A total of 1,740 eligible respondents were randomly allocated into four groups:

T0 (Control): No messages

T1 (Generic Nudge): Salience- and simplification-based message

T2 (Personalised Eligibility Nudge): Message highlighting individual eligibility

T3 (Peer Comparison Nudge): Message using social norms and framing

Messages were sent weekly via SMS or WhatsApp over a three-week period in June 2025. Outcomes were measured via baseline and follow-up surveys, with primary focus on self-reported e-Shram registration status and behavioural mechanisms such as recall, perceived eligibility, and registration intent. The design accounts for stratification by platform type, gender, and prior awareness to ensure balanced groups.

This study employs a randomized controlled trial (RCT) to test the impact of behavioural digital nudges on uptake of e-Shram—India’s national social security platform—among gig and informal workers in Odisha. The experimental sample consists of 1,740 respondents recruited from platform-based and informal gig work networks. Eligibility criteria included being aged 16–59, working at least 15 hours per week, and not having registered for e-Shram at baseline.

Participants were stratified by gender, city tier, platform type (e.g., ride-hailing, delivery, domestic work), and digital access. They were randomly assigned to one of four groups:

T0 (Control) – Received no messages.

T1 (Generic Information Nudge) – A concise SMS/WhatsApp message explaining the benefits and process of e-Shram registration, based on salience and simplification principles.

T2 (Personalised Eligibility Nudge) – A message framed to highlight the respondent’s likely eligibility, leveraging personalisation and perceived fit.

T3 (Peer Comparison Nudge) – A message stating that “others like you” in similar jobs or localities had registered, drawing on social norm and framing principles.

All treatment arms (T1–T3) received three nudges over three consecutive weeks in June 2025. Messaging content was pilot-tested for linguistic clarity and sent in the participant’s preferred language (Odia, Hindi, or English). Delivery occurred through SMS and WhatsApp using automated tools, and timestamped delivery logs were maintained.

Baseline data were collected during April–May 2025 and included demographics, employment status, digital access, awareness of e-Shram, and behavioural indicators such as trust in government and perceived eligibility. The follow-up survey was conducted in July 2025, approximately two weeks after the final message. Primary outcomes included self-reported e-Shram registration, recall of the message, and intent to register. Secondary outcomes included digital constraints, use of government platforms, trust in government, and social spillovers.

To reduce attrition and measurement error, follow-up surveys included verification prompts (e.g., screenshot submission of e-Shram card) and attention checks. The analysis plan pre-specifies subgroup analyses by gender, trust levels, platform type, and digital fluency, along with estimation of cost-effectiveness and behavioural pathways. Power calculations (α = 0.05, power = 0.8) were conducted for detecting a minimum detectable effect size (MDE) of 6–8 percentage points on the primary outcome, assuming 85% follow-up rate.

Stratified randomization using computer-generated random numbers in Stata, implemented in office.

Individual-level randomization stratified by gender, platform type, and city tier.

No

Not applicable (individual-level randomisation)

1,740 individuals (gig workers registered on digital platforms)

435 individuals – Control (T0), 435 – Generic Info Nudge (T1), 435 – Personalised Eligibility Nudge (T2), 435 – Peer Comparison Nudge (T3)

We calculated the minimum detectable effect size (MDE) for binary outcomes (e.g., e-Shram registration) assuming α = 0.05, power = 0.80, equal allocation across four arms (N = 435 per group), and no clustering (individual-level randomisation). The MDE for pairwise comparisons between treatment and control arms is approximately 6.5–7.0 percentage points (assuming a control group mean of 20% and a standard deviation of 0.40). This reflects the minimum effect size we can detect with sufficient power for our primary outcome.