The Benefits Cliff: When Earning More Leaves You Worse Off

Support for lower-income households is means-tested, so it falls as income rises. That's the point of targeting. But where a benefit ends abruptly, or where several phase out over the same income range, a raise can cost more than it pays — on the worked example below, $2,000 of extra earnings leaves the household $700 worse off. Asset limits do the same thing to savings. That makes the one protection that reliably holds in a crisis — cash in the bank — something a household can be penalized for building, and it pushes every shock that savings would have absorbed onto credit instead.

The effective marginal rate

The share of an extra dollar a household loses to taxes and reduced benefits combined. No single programme sets it; it emerges from how programmes stack.

Source of loss on an extra dollar	Stylized share
Payroll and income taxes	15¢
One benefit tapering at 30%	30¢
Combined	45¢

A household keeping 55 cents of each extra dollar faces a higher effective rate than most high earners pay in income tax, without any single rate being high. That's the first observation: effective rates at the bottom of the income distribution can exceed those near the top, as a by-product of targeting.

The figures are illustrative. Actual rates depend on which programmes a household receives, where it lives, its size, and how each programme measures income — which is itself part of the problem, since a household can rarely work out its own effective rate in advance.

A raise that costs money

Tapering is one thing. An abrupt end is another. A stylized household receives a $2,000 raise that happens to cross the eligibility threshold for one benefit:

	Change
Extra annual earnings	+$2,000
Additional taxes	−$300
Benefit A, tapering at 30%	−$600
Benefit B, ending at a threshold	−$1,800
Net change in resources	−$700
Effective marginal rate on the raise	135%

The household is worse off for having earned more. Not by a rounding error — by a third of the raise. The numbers are stylized; cliffs of this shape exist wherever support ends at a line rather than tapering to zero.

Per our household finance analysis, a firm facing a 135% marginal rate on extra revenue would decline the business. A household that declines the raise, the extra shift, or the promotion is making the same calculation — and is frequently described as lacking ambition for doing so.

Asset limits and the buffer

The same structure applied to savings, and it's where this reaches the core of household resilience.

Some programmes limit how much a household can hold in savings and remain eligible. Where the limit is low and the benefit is valuable:

• A household with savings just below the limit gains nothing by saving more.
• Saving past the limit costs the benefit — which may be worth more per year than the savings.
• The rational choice is to keep savings below the line.

Now connect it to this desk's earlier work. Per our correlation analysis, liquid savings are the one household protection that doesn't fail when everything else does — credit lines are withdrawn, family help is under the same pressure, assets fall in value, but cash stays cash. Per our savings analysis, a modest buffer does more than large amounts of contingent access.

An asset limit caps precisely that protection, for precisely the households with the fewest other protections. The design meant to target help to those without resources can prevent them from accumulating the resource that would most reduce their need for help.

Rules vary considerably by programme and jurisdiction, and some places have raised or removed asset limits in particular programmes — which is itself useful, because it creates the comparisons that let the effect be measured.

The design meant to direct help to people without resources can stop them building the one resource that would reduce their need for it.

Where it reaches credit

The link to this library's core subject.

A household that can't hold a buffer meets every shock with credit. Per our liquidity analysis, the car repair, the medical bill, and the short month all become borrowing — and per our time preference analysis, borrowing under acute pressure is frequently expensive and rational at the same time.

So the chain runs:

1. Asset limit caps savings.
2. Shocks go to credit instead of to a buffer.
3. Credit costs money, which reduces the resources a benefit was meant to support.
4. Repeated borrowing affects the credit file, which affects housing and future borrowing.

The public cost saved by the asset limit reappears as private cost in the credit market — paid by the household, and partly captured by lenders. Per our pricing analysis, small emergency loans are among the most expensive credit there is, so the conversion rate is poor.

There's a second channel. Income cliffs discourage the earnings growth that would improve a household's credit position — per our affordability analysis, residual income is what lenders should care about, and a cliff can hold it flat even as gross income rises.

Volatile income makes it worse

The interaction with irregular work, and it's sharp.

Per our irregular income analysis, many lower-income households have earnings that swing month to month. Where eligibility is assessed over short periods, a single good month can cross a threshold.

• An overtime month pushes income over a line and support is lost or reduced.
• The next month income falls back, but reinstating support takes time and paperwork.
• The household is worse off across the two months than if the overtime had been refused.

Per our time analysis, the reinstatement process costs hours during business hours — the same hours that shift workers least control. Volatility converts a cliff that a stable earner crosses once into one an irregular earner crosses repeatedly, with a loss and a reapplication each time.

The trade-off nobody escapes

The balanced core of the argument, because this is a contested policy question and the difficulty is real.

Programme design is trying to achieve three things at once:

Goal	What it favours	What it costs
Target those most in need	Fast phase-outs, asset tests	High effective rates, cliffs
Keep public cost bounded	Fast phase-outs, low thresholds	High effective rates, cliffs
Make earning and saving always pay	Slow phase-outs, no asset tests	Higher cost, support reaching higher incomes

You can have any two more easily than all three. Slowing a phase-out removes the cliff but extends support further up the income scale, which costs more. Removing an asset test improves saving incentives but may extend support to households with substantial resources.

The arguments on each side deserve their strongest form:

• For tight targeting: public resources are limited, support should go to those with the least, and asset tests prevent support flowing to households with savings to draw on.
• For preserving incentives: cliffs trap households below thresholds, asset limits prevent the resilience that reduces future need, and the long-run cost of discouraging work and saving may exceed the short-run savings.

This report doesn't settle which weighting is right — that's a value judgment people reasonably make differently. Its narrower claim is that cliffs, as opposed to gradual phase-outs, are a design choice with costs on nearly every dimension, and that asset limits have a specific cost through the credit channel that's rarely counted.

What this means for lenders

The practitioner reading.

• Benefit income can change abruptly for reasons unrelated to the borrower's behaviour — a raise can reduce total resources. Affordability assessment should allow for it.
• A small savings balance may be a rational response to an asset limit, not evidence of poor financial habits. Reading thin savings as a behavioural signal misreads the constraint.
• Income growth may not translate into capacity growth near thresholds — so extending more credit to a borrower whose wages just rose can be mistaken.
• Products that hold savings outside what counts as an asset — where rules permit — are genuinely valuable to these households, and their permissibility is a question for counsel.

Per our record retention analysis and framing analysis, the general lesson recurs: behaviour that looks like a personal failing is frequently a rational response to a constraint the observer can't see.

The strongest objections

"Most households don't face cliffs this steep." Plausibly true — many receive one programme with a gradual taper. The report's claim is about the households that do face overlapping programmes, which are disproportionately those with children and the lowest incomes. How many face effective rates above 100% is an empirical question and depends heavily on location.

"People don't know their effective rates, so they can't respond to them." Partly — the complexity itself means responses are imperfect. But households learn from experience, particularly after losing support once, and advice spreads within communities. The response doesn't require a calculation, only a memory of what happened last time.

"You're making a political argument." The report tries not to. It describes a mechanism and its credit consequences, and sets out the trade-off with the strongest case on each side. Where to land on it is a political judgment this report deliberately leaves to the reader.

Testable implications

1. Earnings should bunch just below benefit thresholds, more than smooth income distributions would predict.
2. Savings should bunch just below asset limits, and should rise where limits are raised or removed.
3. Where asset limits are removed, reliance on high-cost credit should fall among affected households — the credit-channel prediction.
4. Enrolment should rise less than feared when asset limits are removed, if few households hold savings near the limit.
5. Households with volatile income near thresholds should show more benefit churn — loss and reapplication — than stable earners at the same average income.
6. Borrowers whose income rose across a threshold should not show improved repayment relative to those whose income stayed below it.

The third and fourth are testable where jurisdictions have changed asset rules, which creates natural comparisons. If removing an asset limit raises savings and lowers expensive borrowing without a large enrolment increase, the limit was costing households more through the credit market than it saved the programme — and that's a finding people on both sides of the targeting debate would want.

The conclusion we'd hold: means testing is necessary and cliffs are a design choice, and the asset-limit version of the cliff has a cost that runs straight through household credit. A buffer is the protection that holds when everything else fails, and a rule that caps it moves risk from the programme's budget to the household's credit file.

Frequently asked questions

---

This report presents an analytical framework and the authors' interpretation; it is not legal, tax, or benefits advice, and it takes no position on how public programmes should be designed. Worked figures are stylized illustrations rather than descriptions of any specific programme. Eligibility rules, income and asset tests, and phase-out schedules vary by programme and jurisdiction and change over time; anyone assessing their own position should consult the programme or a qualified adviser. The implications identified as testable are hypotheses.