The Rise of E-Wallets and Buy-Now-Pay-Later: Payment Competition, Credit Expansion, and Consumer Behavior
Bian · Cong · Ji — NBER Working Paper 31202, 2023
Thesis
The biggest BNPL experiment on Earth runs inside Chinese super-app wallets serving a billion-plus users. The pattern rhymes with the West — BNPL access measurably boosts spending — but with one sharp difference: users who rely on the wallet as their only credit source mostly refuse to revolve, moderating borrowing the moment interest appears. BNPL here is less a debt trap, more digital cash with training wheels — and its real effect is widening who gets credit at all.
Method
- Transaction-level data from a world-leading e-wallet provider (1B+ users), matched to merchant and consumer information; e-wallet payment shares 2019–2021.
- Payment-choice model across BNPL, cards, and balance payments; merchant-side acceptance analysis (~43.5% of online merchants accept BNPL).
- Randomized credit-access experiment run by the provider (Jan–Jul 2017) + difference-in-differences on spending, 700k user-months; single-access vs dual-access (card + BNPL) heterogeneity.
Key findings
- BNPL dominates e-wallet payments online and on-site; it crowds out other wallet options but expands credit to the underserved: 78–84% of sampled users have BNPL access vs ~30% credit-card access; 68%+ lack cards entirely.
- The randomized shock: BNPL access lifts spending +RMB ¥86/month (≈5.6% of average consumption); dual-access users +¥204 — and card spending also rises for dual-access users, i.e. BNPL complements rather than substitutes big-ticket card purchases.
- Default stays under 1% (below 2% among all BNPL users population-wide); about two-thirds of single-access users repay in full every cycle — they borrow as a payment rail, not a rolling balance.
- Where BNPL reaches users banks wouldn't touch, it does so at delinquency APRs of 13–16% — cheaper than Western BNPL penalties, costlier than it looks.
Limitations
- One provider's ecosystem in a cash-to-cashless transition economy — the institutional context (no cheap revolving cards for most users, super-app lock-in) doesn't transfer cleanly to the US/Canada.
- Spending measured inside the wallet; off-wallet spending is invisible, so the +5.6% is a floor, not a total.
- The 2017 experiment predates COVID e-commerce distortion, but also predates current product features.
- NBER working paper — not peer-reviewed at this version.
So what for Nick
- The counterfactual worth noticing: where BNPL is the only credit in town, most users treat it like cash and default stays rare. The harm case in North America is arguably about stacking it on top of cards — not the installments themselves.
- Credit expansion is the quiet story: BNPL is how millions of underbanked users enter formal credit. That's genuinely useful — until it's stacked.
- Watch the household analogy: if pay-in-4 stays a short, fully-repaid rail for someone with no other credit, fine. The Di Maggio/Guttman-Kenney damage shows up when it's a third liquidity source layered on cards.