Working Paper · Rev 4.2 · May 2026

The tax refund
disappears.
We stop that.

A behavioral architecture designed to prevent $6.2 billion in charitable tax incentives from dissolving into personal consumption every year.
Open science. No commercial interest.

$6.2B Social Dividend / year
Federal + state refund via §170(p)
$1.4B Recapturable today
No legislation required
$7–10B Legislative horizon
Single Form 1040 checkbox
GIVING TAX DAY May 15 The moment a GIVER's December commitment becomes May reality
Why Now

The same law that created the problem
created the opportunity.

For decades, 86% of American households — those who take the standard deduction — had no meaningful tax incentive to give. The One Big Beautiful Bill Act changed that permanently. §170(p) is the largest expansion of charitable tax access in a generation.

But the incentive arrives in May. The giving decision happened in December. Without behavioral architecture, the government's co-investment dissolves before it reaches the cause.

This is the window. The deduction is new. The behavioral infrastructure doesn't exist yet. What gets built now becomes the standard.
§170(p) · Effective January 1, 2026
+
$74B over 10 years

New incentives for non-itemizers

$81B over 10 years

New disincentives — 0.5% floor + 35% cap

−$7B net sector impact

Without behavioral architecture to capture the refund flow

Source: National Council of Nonprofits · Indiana University Research

2026

§170(p) takes effect. First filing season. Infrastructure gap is visible.

NOW

Behavioral architecture must be designed before the pattern locks in.

2028+

Absorption becomes habitual. The window narrows. Cost of inaction compounds.

The Challenge

The refund black hole.

The One Big Beautiful Bill Act (§170(p), effective January 2026) created a permanent above-the-line deduction for the 86% of Americans who don't itemize. The intent was to expand charitable giving.

What wasn't designed: what happens when that benefit arrives months after the giving decision. The research is unambiguous — tax refunds dissolve.

60–75% of refunds absorbed into discretionary spending within 90 days Souleles 1999 · Parker et al. 2013
30 days absorption is fastest in the first month after receipt JPMorgan Chase Institute 2019

This is a design problem, not a generosity problem. GIVERs gave in December. The question is whether the government's co-investment reaches the cause — or disappears into an unrelated expense.

Three saboteurs of charitable intent

01
Mental Accounting

A refund entering a bank account is reclassified as "personal money." Redirection then requires active, costly effort.

Thaler 1985 · Arkes et al. 1994
02
Endowment Effect

Once money is perceived as personal property, redirecting it feels like a loss — approximately 2× heavier than an equivalent gain.

Kahneman, Knetsch & Thaler 1991
03
Salience Decay

Tax incentives shape behavior only when salient at the point of decision. By May, December motivation has long decayed.

Chetty, Looney & Kroft 2009
The Mechanism

One deduction. One moment.
Two interventions.

The GIVER architecture inserts a circuit-breaker before the causal chain completes. The key is timing, not willpower.

DEC

Ulysses Pact

GIVER authorizes an ACH transfer at peak motivation — before the refund exists as psychological property. No second decision required in May.

Behavioral pre-commitment
JAN

GIVER Identity Shift

Acknowledgment includes Social Dividend estimate. The donor becomes a Civic Investor — responsible for the government's co-investment reaching the cause.

+15–25% follow-through (Bem 1972 · Cialdini 2009)
MAY
15

Giving Tax Day

ACH executes automatically. 92% of refunds have settled. The federal co-investment lands in the cause's account — visible, measurable, shareable.

The moment everything converges

Three product branches — one principle

Same net cost to the GIVER. More impact to the cause.

SEED

First-time GIVERs

Converts a one-time donor into a GIVER. Targets the sector's most costly retention problem: 77–81% first-gift churn.

Expected ρ: 6–12%
COMPOUND

Recurring GIVERs

Reveals the government's hidden co-investment to committed donors. Adds salary-linked escalation over time.

Expected ρ: 8–15%
Evidence Base

Triangulated. Transparent. Testable.

ρ

The Redonation Rate

The most important — and most carefully triangulated — parameter in the model.

Central estimate: 12%  (range 6–22%)

Conservative prior for a first pilot in a nascent US behavioral infrastructure. Triangulated from four anchors, then discounted for first-year novelty and absence of employer-match framing.

Anchor Implied ρ Bias
Save More Tomorrow voluntary commitment
Thaler & Benartzi 2004
25–35% ↑ Upward
Opt-in consent rates across contexts
Johnson & Goldstein 2003
15–25% → Neutral
Pre-income charitable field experiments
Altmann et al. 2019 · Messer et al. 2016
18–28% ↓ Slight down
UK Gift Aid real-world opt-in
HMRC 2025
28–35% ↑ Upward

A conservative discount of ~55% is applied to the anchor midpoint (27%) to account for: nascent US behavioral infrastructure, absence of employer-match framing, first-year novelty effects, and the absence of direct prior evidence in this context. The pilot exists to replace this prior with direct measurement. If ρ exceeds 12%, the model improves.

Monte Carlo Results

10,000 iterations · Conservative pilot prior (ρ: 6–22%, mode 12%)

P10
$380M
P50
$840M
P90
$1.6B

Conservative pilot scenario. If ρ validates above 12%, every percentage point adds ~$70M to the central estimate.

International Precedents

🇬🇧 UK
Gift Aid (1990)

£1.7B/year · Active 35 years · Never repealed

🇮🇹 Italy
5x1000 (2005)

€520M/year · ~18M taxpayers · Active 20 years, expanded

🇪🇸 Spain
Ley 49/2002 + Redonada

Design phase with UNICEF España · StockCrowd platform · Origin of this model

Confidence by Claim

§170(p) permanent deduction from TY 2026 High
Refund dissolves within 90 days High
GIVER identity adds 15–25% follow-through Moderate
ρ = 12% conservative prior (6–22% pilot range) Moderate
Monte Carlo P50 ≈ $840M capturable (conservative pilot) Moderate (conditional)

The RCT closes the gap

A pre-registered 4-arm field experiment across Minnesota, Arizona, and Texas (N≈2,000, filing season 2027) will directly measure ρ in the US tax context. Pre-registered on the AEA Registry before first enrollment. All results published regardless of direction.

Explore the RCT Design →
Arm A

Control — standard acknowledgment only

Arm B

Reframe — GIVER identity framing, no ACH

Arm D

Tax-software integration at filing moment

The Ask

Three immediate actions.

This is a testable $1–10B opportunity. The cost of testing it is low. The cost of ignoring it is not.

01

Form a Steering Committee

Academia + nonprofit + tech. Pre-register the RCT protocol. Coordinate on a Code of Practice.

Target: June 2026
02

Fund the Tri-State RCT

Minnesota + Arizona + Texas. Modest scale: N≈2,000. Low cost, high signal.

Close: December 2026
Arnold Ventures · Open Philanthropy · Gates · Walton
03

Technical Conversations

Begin conversations with top-3 tax-software vendors and top-5 CRM vendors on the shared Redonation Checkbox standard.

Target: Q4 2026

Academic Partners

Behavioral economists to co-design and validate the RCT. Target journals: JPE · AEJ Applied.

Pioneering NGOs

3–5 organizations ready for a manual MVP in December 2026. Fewer than 2 hours/month commitment.

Tech Platforms

CRMs and payment gateways interested in co-designing the GIVER UX. 6–10 weeks of commitment.

Frequently Asked Questions

No. This is an open science and open source initiative. No founder or commercial entity holds a financial stake. All code is open source. All findings are published regardless of direction.
Yes. The architecture is designed conservatively under Regulation E and NACHA, including mandatory 14-day pre-notification and one-click cancellation through May 1. UDAAP compliance requires explicit opt-in only — never pre-checked.
We are not seeking venture capital. We are looking for philanthropic funds to cover the empirical RCT, institutional academic partnerships, and nonprofit pilot organizations. The model is intentionally low-cost to test.
We triangulate from four independent behavioral anchors (Save More Tomorrow, opt-in consent research, charitable field experiments, UK Gift Aid) with an implied midpoint of ~27%. We then apply a conservative discount to ρ = 12% (range 6–22%) for a first pilot in a nascent US context. Spain's Redonada mechanism — the direct predecessor — is in design phase with UNICEF España and StockCrowd. The US pilot RCT will replace this prior with direct measurement.
We publish our falsification criteria openly. The model would be invalidated if: §170(p) take-up falls to ≤15%, if the pilot shows no significant effect in any arm, or if CBO/JCT scores the Redirect Box above $5B behavioral cost. We consider transparency about failure conditions a prerequisite for credibility.
Giving Tax Day · May 15, 2027

On May 15, 2027,
every GIVER's commitment
executes automatically.

Not hopes. Not pledges. Automatic.
That is the cultural moment we are building toward.

Contact — jesusrivasg@hotmail.com

Nonprofit · Platform · Academic