Digital Monetary Salience: A Field Experiment on How Payment Interfaces Shape Inflation Expectations and Consumer Spending in Emerging Economies

We implement a digital payment interface intervention with two conditions:

1. HIGH SALIENCE INTERFACE (Treatment Group):
- Display of cumulative spending before payment confirmation
- Visual budget progress bar (biweekly/monthly limits)
- Prominent warnings about future costs (BNPL installments, deferred payments)
- Alerts when approaching spending thresholds

2. LOW SALIENCE INTERFACE (Control Group):
- Standard payment interface (current market design)
- Prominent balance/credit availability display
- Minimal friction in payment process
- No future cost warnings or budget tracking

The intervention is delivered through a simulated payment app interface that participants use for 6 biweekly waves. All participants continue using their regular payment methods, but report transactions through the experimental interface which applies the salience modifications.

TECHNICAL IMPLEMENTATION DETAILS:

Platform: Custom-built payment simulation interface integrated with Qualtrics
Delivery: Mobile-responsive web application accessible via smartphone

High Salience Features:
- Real-time calculation of cumulative spending from all reported transactions
- Color-coded budget bar (green/yellow/red based on spending relative to income)
- Pop-up modal for BNPL transactions: "This purchase will result in 6 payments of $X over 3 months. Total cost: $Y"
- Monthly spending summary with category breakdown
- Push notifications for 80% and 100% budget utilization

Low Salience Features:
- Simple transaction logging form
- Current balance display (simulated based on reported income)
- "Credit available" indicator always visible
- One-click payment confirmation with no additional information

Both groups receive identical transaction tracking functionality but differ in the salience of financial information presented. The intervention runs for 5 months (July-November 2025) with 6 measurement waves.

PARTICIPANT INSTRUCTIONS:
"Use this app to record ALL your digital payments over the next 5 months. The app will help you track your spending and may provide additional information about your financial decisions."

ETHICAL CONSIDERATIONS:
- No actual financial transactions processed through the app
- All spending data is self-reported or from uploaded receipts
- Interface modifications are informational only (no financial risk induced)
- Clear disclosure that this is a research tool, not a banking app

2026-01-05

2026-04-30

1. Non-durable expenditure amount (biweekly, continuous, MXN)
2. 12-month inflation expectation (percentage point estimate)
3. Inflation perception gap (difference between perceived and official inflation)

1. NON-DURABLE EXPENDITURE:
- Constructed from: Self-reported spending on food, transportation, utilities, personal care
- Measurement: Biweekly sum across all non-durable categories
- Unit: Mexican Pesos (MXN)
- Source: Transaction logs in experimental app + receipt uploads

2. 12-MONTH INFLATION EXPECTATION:
- Constructed from: Response to "What do you expect the annual inflation rate to be over the next 12 months?"
- Measurement: Continuous percentage (0-100%)
- Validation: Cross-checked with Bank of Mexico (Banxico) survey format
- Collected: Each biweekly wave

3. INFLATION PERCEPTION GAP:
- Constructed as: |Perceived current inflation - Official inflation rate|
- Perceived inflation: Response to "How much do you think prices have changed in the last 2 weeks?" (7-point scale converted to percentage)
- Official inflation: Biweekly consumer price index changes from INEGI
- Interpretation: Lower gap = better inflation perception anchoring

1. Shadow Liquidity Index (SLI) score (0-3 scale)
2. Price checking frequency (categorical: Never/Sometimes/Always)
3. Durable goods expenditure (biweekly, MXN)
4. Financial stress index (5-point Likert scale composite)
5. Treatment effect heterogeneity by age, income, financial literacy

1. SHADOW LIQUIDITY INDEX (SLI):
- Composite measure: SLI = (balance_visibility + price_checking + payment_friction) / 3
- Components:
* balance_visibility: "Does the app show your available balance?" (0=No, 1=Yes)
* price_checking: "How often do you check unit prices?" (0=Never, 0.5=Sometimes, 1=Always)
* payment_friction: Inverted "How easy was the payment?" (1=Very easy → 5=Very difficult, rescaled to 0-1)
- Interpretation: Higher SLI = greater financial awareness

2. PRICE CHECKING FREQUENCY:
- Direct survey item: "Before making a purchase, how often do you check the price per unit/quantity?"
- Response: Never (0), Sometimes (1), Always (2)
- Collected: Post-transaction ESM surveys

3. DURABLE GOODS EXPENDITURE:
- Sum of reported spending on electronics, appliances, furniture, vehicles
- Timeframe: Biweekly, but expected to be less frequent than non-durables
- Validation: Cross-checked with payment method (credit vs debit)

4. FINANCIAL STRESS INDEX:
- Composite of 3 items (5-point Likert each):
* "How worried are you about your current financial situation?"
* "How difficult is it to cover your monthly expenses?"
* "How confident are you in your ability to handle unexpected expenses?" (reverse-coded)
- Scale: 3-15, higher = more stress

STUDY DESIGN: Longitudinal Randomized Controlled Trial (RCT) with A/B testing

TIMELINE: 5 months (July-November 2025) with 6 biweekly measurement waves

PARTICIPANTS: 1,000 adults in Aguascalientes, Mexico, recruited from 4 shopping centers representing different socioeconomic strata

RANDOMIZATION: 1:1 allocation to High Salience vs Low Salience interfaces

MEASUREMENT:
- Baseline survey: Demographics, financial habits, initial expectations
- Biweekly surveys: Expenditure tracking, inflation expectations, financial behaviors
- Experience Sampling (ESM): Short surveys triggered after reported transactions
- Event measurement: Response to exogenous price shocks (fuel, food price changes)

ANALYTIC APPROACH:
1. Hierarchical Bayesian models for primary outcomes
2. Causal Forest for heterogeneous treatment effects
3. Event-study analysis around price shocks
4. Fixed effects panel models for robustness

ETHICS: Approved by Institutional Review Board, informed consent, data anonymization

RECRUITMENT PROCEDURE DETAILS:

1. SITE SELECTION: Four shopping centers in Aguascalientes:
- Altaria: High/middle-high SES (AB, C+)
- Galerías: Middle-high SES (C+)
- Espacio Aguascalientes: Middle SES (C, C-)
- El Parián: Middle-low/popular SES (D+, C-)

2. RECRUITMENT PROTOCOL:
- Intercept approach at payment points (cashiers, food courts)
- Screening: ≥3 digital payments/week, smartphone ownership, willingness to share expenditure data
- Compensation: Non-monetary (points redeemable for shopping vouchers)
- Retention strategy: Weekly reminders, technical support, progress reports

3. QUOTA MANAGEMENT:
Target distribution (N=1,000):
- Income: <$10K (40%), $10-20K (35%), >$20K (25%)
- Gender: 50% women, 50% men
- Age: 18-29 (35%), 30-44 (40%), 45-60 (20%), >60 (5%)
- Financial education: Low (40%), Medium (40%), High (20%)

DATA COLLECTION PROTOCOL:

1. WAVE 0 (Baseline - July 2025):
- In-person onboarding at recruitment sites
- Installation of experimental app on participants' smartphones
- Training on transaction logging and survey completion
- Baseline survey (15-20 minutes)

2. WAVES 1-6 (Biweekly - August-October 2025):
- Automated survey delivery via Qualtrics (SMS/email)
- ESM triggers: After each reported transaction (>$50 MXN or daily first transaction)
- Compliance monitoring: Weekly check-ins for non-responders

3. QUALITY CONTROL:
- Receipt verification: 10% random sample asked to upload receipts
- Attention checks: Embedded in surveys
- Data validation: Range checks, consistency across waves

ANALYSIS PLAN DETAILS:

PRIMARY ANALYSIS (Model 1 - Hierarchical Bayesian):
brm(spend_nd_t ~ salience + income + educ_fin + (1 | id), ...)

SECONDARY ANALYSES:
1. CATE Estimation: causal_forest(X, Y, W) with covariates: age, income, SLI, financial literacy
2. Event Study: feols(outcome ~ i(time, treatment) + shocks | id + time)
3. Mediation: SLI as mediator between treatment and expenditure
4. Subgroup: Effects by payment method (credit vs debit vs BNPL)

ATTENTION & COMPLIANCE:
- Minimum compliance: Complete ≥4 of 6 waves for inclusion in primary analysis
- Attrition management: Replacement pool of 20% extra recruits
- Treatment fidelity: Weekly checks that intervention features are functioning

BLINDING:
- Participants: Single-blind (not told which group they're in)
- Analysts: Blinded during initial analysis (treatment coded as Group A/B)
- Field team: Not blinded (necessary for technical support)

Computer-generated random assignment using R's randomizr package with block randomization within quotas.

PROCEDURE:
1. After baseline survey completion, participant enters randomization pool
2. Software assigns to High/Low salience based on:
- Pre-specified 1:1 allocation ratio
- Stratification by: recruitment site, income bracket, age group
- Balancing across: gender, financial education level
3. Assignment immediately triggers appropriate app configuration
4. Allocation concealment: Field staff cannot influence assignment

RANDOMIZATION CODE:
library(randomizr)
assignment <- block_ra(blocks = strata_vars,
conditions = c("high_salience", "low_salience"))

Individual participant level.

Each participant is independently randomized to either High Salience or Low Salience interface.

No clustering at household/shop level to avoid contamination, though we record household composition for post-hoc analysis of spillovers.

No

Not applicable (individual-level randomization, no clustering)

6,000 participant-wave observations (1,000 participants × 6 waves) With expected attrition of 20%, we anticipate approximately: - 800 participants completing all 6 waves = 4,800 observations - Additional 200 participants completing partial waves = ~600 observations - Total: ~5,400 analyzable observations Minimum requirement for primary analysis: 302 participants (pre-registered power calculation)

High Salience Group: 500 participants (50% of total)
Low Salience Group: 500 participants (50% of total)

Within each treatment arm, quota distribution:
- Income: <$10K (200), $10-20K (175), >$20K (125)
- Gender: 250 women, 250 men
- Age: 18-29 (175), 30-44 (200), 45-60 (100), >60 (25)
- Financial education: Low (200), Medium (200), High (100)

PRIMARY OUTCOME: Non-durable expenditure Parameters: - Significance level (α): 0.05 (two-tailed) - Power (1-β): 0.80 - Intraclass correlation (ICC): 0.50 (estimated from pilot) - Number of waves: 6 - Effective sample after attrition/compliance: 675 (from initial 1,000) Minimum Detectable Effect Size (MDES): - Standardized (Cohen's d): 0.30 - In raw units: 0.30 × SD(spend_nd_t) - Based on pilot SD = 48.2 MXN biweekly → MDES = 14.46 MXN biweekly - Percentage change: From baseline mean of 216.8 MXN → 6.7% reduction SECONDARY OUTCOME: Inflation expectation gap Parameters: - SD from pilot: 2.3 percentage points - MDES (d=0.30): 0.69 percentage points - Baseline mean gap: 3.1 pp → 22% reduction detectable JUSTIFICATION: - Effect size d=0.30 chosen as "moderate" per Cohen conventions - Consistent with behavioral economics literature on salience interventions - Achievable given intervention intensity (5 months, multiple touchpoints) - Policy relevant: 6-7% spending reduction meaningful for household finance SENSITIVITY ANALYSIS: If attrition higher (30%) and compliance lower (85%): - Effective N = 595 - MDES increases to d=0.32 (15.4 MXN biweekly) - Still sufficient for policy-relevant effects