Buy Now, Pay Later Credit: User Characteristics and Effects on Spending Patterns

Di Maggio · Williams · Katz — NBER WP 30508, 2022

Thesis

BNPL isn't just another payment rail — it changes how much people spend and on what. Access to point-of-sale installment credit raises total spending by amounts standard economics can't explain, pushes spending toward retail, and does so for liquidity-constrained and unconstrained users alike. The authors' label: a "liquidity flypaper effect" — the credit sticks where it hits and spending follows.

Method

De-identified bank + credit-card transaction data from a US data aggregator → an unbalanced panel of 400,000 unique users / 39 million transactions (200k ever-BNPL + 200k never-BNPL), with demographics.
Difference-in-differences around first BNPL use vs prior spending episodes.
IV design: instrument BNPL use with heterogeneity in retailer BNPL adoption timing in each consumer's historical shopping basket — plausibly exogenous to the consumer's own spending shock.
Calibration exercises testing whether a standard lifecycle model with liquidity constraints can generate the observed magnitudes (it can't).

Key findings

First BNPL use: total spending jumps ~$130 on impact (≈$30 of it on BNPL itself, the rest spillover to non-BNPL retail, essentials, discretionary) — and stays elevated for the 24 weeks observed.
IV estimate: +$60/week permanent increase in total spending, almost entirely in non-BNPL retail — a ~12% shift of the spending basket toward retail at impact, still +1.5% weeks later.
Fees: overdraft fee probability +0.47 pp (a 20% relative increase), low-balance fee +0.48 pp (17% relative) — BNPL use drains liquidity elsewhere, not just at checkout.
The spending surge shows up even for users with liquidity buffers and credit cards — binding constraints can't explain it; neither can reasonable discount rates given most BNPL is repaid within 6 weeks.
BNPL smooths consumption for lower-income users (reduced spending sensitivity to income) — the one clearly positive dimension.
Survey facts: users skew young, cash-constrained; motivations include "purchase wouldn't fit budget" (38.4%), "borrow without credit check" (24.7%), "couldn't get approved for a credit card" (14%).

Limitations

Single bank's transaction data: misses spending/credit at other institutions; external validity to the full population uncertain.
BNPL adoption overlaps COVID-era spending shifts (paper excludes early adopters and shows robustness, but the confound is real).
IV exclusion: retailer adoption timing could correlate with local/sector spending trends — arguable, though well-defended.
NBER working paper — not peer-reviewed at this version.
Observer can't see total BNPL debt stacking across multiple platforms (a known industry blind spot).

So what for Nick

"Pay in 4" is engineered spending, not deferred payment. The clean data result: the checkout choice changes the total basket, weeks later, in categories BNPL never touched.
The +$60/week is the number to remember — ~$3,100/yr of behavior change from a "0% interest" button. If a checkout offers installments, treat it as spending today that merely feels like future-you's problem.
Watch it in family/household context: BNPL adoption among younger adults he may advise is highest exactly where liquidity is thinnest, and the fee spillover (overdrafts) is where it silently gets expensive.