Find the question you need
Choose a route or read from the beginning. All teaching chapters stay open; the grouped index below is expandable.
Combine the rules
Chapters 6–7. Test overlapping withdrawals and assessment periods.
Practise and continue
Chapters 11–12. Use changed cases, questions and connected reading.
Full chapter index · Practice and worked checks · Public Policy subject root
A support programme can help a household at one income and leave it with fewer resources after a small pay rise. To understand that possibility, inspect how earnings and support change together. The eligibility label alone does not tell you the size or direction of the change.
This guide develops that calculation from a simple threshold to overlapping programmes. Every income, award, household and rule below is invented. These are policy-design exercises, not descriptions of Singapore benefits or advice about anyone’s eligibility. The Public Policy subject root explains how this numerical question fits public purpose, delivery and evaluation.
Chapter index
See the threshold · Chapters 1–3
1. Begin with a resource definition
2. Specify exactly which side owns the threshold
3. Calculate the cliff across a small earnings change
Build a taper · Chapters 4–5
4. Replace the jump with an explicit taper
5. Check the taper endpoint and its tradeoff
Combine the rules · Chapters 6–7
6. Combine programmes on the same earnings base
7. Put time and assessment units back into the model
Judge the design · Chapters 8–10
8. Compare public cost using a stated distribution
9. Turn a formula into a usable service
10. Distinguish a resource curve from behavioural evidence
Practise and continue · Chapters 11–12
11. Practise with changed rules and worked checks
12. Questions, connected routes and sources
Let e mean monthly earnings and B(e) mean monthly cash support. Start with the deliberately narrow resource measure R(e) = e + B(e). All amounts use the same fictional currency. There are initially no taxes, work expenses, other awards or changes in household composition.
That boundary matters. If a programme pays a school directly, its award is not necessarily cash available for food. If an award reimburses a cost, count the cost and reimbursement consistently. A useful analysis states what resources mean before adding numbers.
Our first question is mechanical: how much does R change when e changes? Later questions concern who receives support, what the policy costs and whether it changes decisions. Answering the first does not answer all the others.
CHAPTER 2 OF 12 · See the threshold
2. Specify exactly which side owns the threshold
Suppose the fictional rule pays 400 whenever monthly earnings are at most 2,000, and pays zero above 2,000. The words “at most” include exactly 2,000. At 2,001, the award disappears.
Write the two branches in words before using them. For e ≤ 2,000, R = e + 400. For e > 2,000, R = e. A graph would contain two rising line segments separated by a downward jump. Joining their endpoints with an ordinary sloping line would invent awards the rule does not provide.
The Federal Reserve Bank of Atlanta’s explanation of benefits cliffs describes how losing assistance can offset earnings gains. Here we isolate one threshold so its arithmetic can be inspected without implying that a real programme has this rule.
CHAPTER 3 OF 12 · See the threshold
3. Calculate the cliff across a small earnings change
At earnings of 1,990, support is 400 and resources are 2,390. At earnings of 2,010, support is zero and resources are 2,010. Earnings rise by 20 while measured resources fall by 380.
The calculation is ΔR = Δe + ΔB = 20 − 400 = −380. The error to avoid is reporting the 20-unit pay rise as a 20-unit resource gain. Another error is calling the loss 400: that omits the simultaneous earnings increase.
| Monthly earnings | Cliff support | Earnings plus support |
|---|---|---|
| 1,990 | 400 | 2,390 |
| 2,000 | 400 | 2,400 |
| 2,001 | 0 | 2,001 |
| 2,010 | 0 | 2,010 |
Across 2,000 to 2,001, the resource change is −399. Across 1,990 to 2,010 it is −380. The award jump is the same; the earnings interval differs. Always attach a finite-change result to its two endpoints.
Consider an alternative rule. Support remains 400 through earnings of 2,000. Above that point it falls by 0.50 for each extra unit earned, until it reaches zero. Thus, within the taper, B(e) = 400 − 0.50(e − 2,000).
At e = 2,200, support is 300 and resources are 2,500. Relative to e = 2,000, earnings increase by 200, support falls by 100 and resources increase by 100. The household retains half the additional earnings within this simplified range.
Revisit the earlier 1,990-to-2,010 change. The new award at 2,010 is 395, giving resources of 2,405. Resources rise by 15 from 2,390. Only ten of the twenty extra earnings units lie inside the taper; applying the 50% withdrawal to all twenty would give the wrong answer.
This illustrates why an average change across two points can differ from the local rule operating between nearby points. Divide the earnings interval at every place the formula changes.
Solve 400 − 0.50(e − 2,000) = 0. The award reaches zero at e = 2,800. Beyond that, keep support at zero. Continuing the line would create a negative award, which is absent from this fictional design.
At 2,800, resources equal 2,800. At 2,801, they equal 2,801. There is no downward jump at this endpoint because the taper has already reduced the award to zero. There is a change in slope: after withdrawal ends, an extra earnings unit adds a whole unit to our resource measure.
A slower taper retains more of each additional earnings unit, but extends support farther up the earnings range if the starting award is unchanged. For a withdrawal of 0.25 per unit, the same 400 award takes 1,600 additional earnings units to disappear. Its endpoint becomes 3,600. That design choice needs a distributional and budget explanation, rather than an assumption that smoothing is costless.
CHAPTER 6 OF 12 · Combine the rules
6. Combine programmes on the same earnings base
Suppose, within a particular interval, the first award withdraws at 0.50 per extra earnings unit, a second at 0.25, and tax rises by 0.10. Assume all three use the same additional earnings, with no offsets or interactions. An extra 100 raises earnings by 100, reduces support by 75 and increases tax by 10. Net resources rise by 15.
The combined withdrawal-and-tax fraction is 0.85; the retained fraction is 0.15. Two individually gradual programmes can therefore produce a very shallow combined resource gain.
If the relevant fractions were 0.50, 0.40 and 0.20, their sum would be 1.10. An extra 100 would reduce net resources by 10 over that interval. A discontinuous eligibility jump is therefore one way, but not the only way, additional earnings can coincide with lower resources.
Do not add published percentages until their bases are checked. One scheme might assess income after another deduction; another might use household income per person. Their combined calculation may require a sequence of rules rather than a simple sum.
CHAPTER 7 OF 12 · Combine the rules
7. Put time and assessment units back into the model
A monthly rule and an annual rule can treat variable earnings differently even when the annual total is identical. In a new fictional example, compare earnings of 1,000 in each of two months with earnings of 0 followed by 2,000. Both total 2,000.
Suppose a monthly award pays 300 only when that month’s earnings are at most 1,200. The steady pattern receives 600 across the two months. The uneven pattern receives 300. The earnings total alone cannot recover the support total under that rule.
This does not establish what any real programme should do. It identifies an assessment-period question. Averaging, temporary disregards, renewal dates, payment delays and later corrections can each change the relationship between entitlement and cash received.
For an actual policy, record the assessment unit, income definition, period, payment schedule and revision process from current official documents. Keep an entitlement calculation separate from the date money reaches someone’s account.
CHAPTER 8 OF 12 · Judge the design
8. Compare public cost using a stated distribution
Take four fictional households, with earnings of 1,900, 2,000, 2,200 and 2,800. Hold their earnings fixed while comparing our original cliff with the 0.50 taper. This is a static exercise: nobody changes work or take-up.
| Earnings | Cliff award | Tapered award |
|---|---|---|
| 1,900 | 400 | 400 |
| 2,000 | 400 | 400 |
| 2,200 | 0 | 300 |
| 2,800 | 0 | 0 |
| Total | 800 | 1,100 |
The additional award cost is 300 for these four households in this month. It is not a national budget forecast. A population with many households inside the taper could produce a different cost change from one with almost none there.
Also separate awards from administration and other effects. A comparison might include implementation, error correction and the consequences of changed participation. State which costs are counted and who bears them. A smoother resource curve supplies one design benefit; it does not determine the best use of a fixed public budget by itself.
Imagine a resident asking, “What happens if my monthly earnings vary?” A service needs more than a calculator that accepts one number. It needs to explain the assessment period, which income counts, when changes must be reported and how a mistaken decision can be reviewed.
Test the boundary cases in plain language: exactly at the threshold, immediately above it, at the taper endpoint and after a corrected income record. Check that the letter, online form and assisted channel apply the same version of the rule.
For delivery questions, continue into public-sector plain-language communication and appeals management. The practical test is whether a person can understand the next action and obtain a reasoned correction when needed.
CHAPTER 10 OF 12 · Judge the design
10. Distinguish a resource curve from behavioural evidence
Our calculations establish consequences of invented rules at specified earnings. They do not establish how many people reduce hours, decline promotions or change applications. Those are empirical questions involving preferences, constraints, information and expectations.
A cluster of reported incomes near a threshold could prompt investigation, but it is not sufficient proof of a particular motive. Earnings may be rounded, measured imperfectly or generated by common pay schedules. Conversely, an absence of visible clustering does not establish an absence of household difficulty.
A useful evaluation asks whether people understand the rule, whether the calculated awards match payments and whether the proposed change affects the intended outcomes. Follow the Public Policy root’s evaluation sequence for comparisons, uncertainty and delivery evidence. Keep the size of the behavioural claim within the evidence actually collected.
CHAPTER 11 OF 12 · Practise and continue
11. Practise with changed rules and worked checks
Check one: find a fresh cliff
An invented award is 250 at earnings of at most 1,500 and zero above. Earnings rise from 1,490 to 1,530. Resources change from 1,740 to 1,530, a fall of 210. The earnings rise of 40 offsets part of the 250 award loss.
Check two: locate a new taper endpoint
An award of 360 begins withdrawing at earnings of 1,800, at 0.30 per extra unit. It reaches zero after 360 ÷ 0.30 = 1,200 additional earnings units, so the endpoint is 3,000. At 2,000, support is 300 and earnings plus support equal 2,300.
Check three: explain a plateau
Within a stated interval, two awards withdraw by 0.60 and 0.30 per additional earnings unit, and tax rises by 0.10. With a common base and no interactions, the total is 1.00. An extra 80 leaves the resource measure unchanged. That is a plateau in this model, even though neither award ends abruptly.
Before looking at the answers, write the resource boundary, applicable formula branch and two endpoints. If a result looks surprising, inspect those three choices before changing the arithmetic.
CHAPTER 12 OF 12 · Practise and continue
12. Questions, connected routes and sources
Is every threshold a benefits cliff?
No. A threshold can change a rate without creating a downward jump. Inspect the resulting resource change. The label “means-tested” does not specify the shape of the award formula.
Does a taper guarantee that extra earnings help?
Only under the full set of relevant rules and costs. Our single 0.50 taper produces rising earnings-plus-support. Overlapping withdrawals or new expenses can change a broader measure.
Can these examples establish someone’s entitlement?
No. Their rules and amounts are fictional. Real entitlement requires the applicable official rules, assessment period and individual circumstances.
Where should I continue?
For the wider decision, use Public Policy and Economics. For education access, the existing student financial aid and means-testing guide connects thresholds with applications, costs and timing. The Workfare and inclusive labour economy article provides a separate programme context; verify current official scheme details before using its figures.
For the learning connection, return to The Gold Standard of Education: learning that transfers into life or explore the How X Works subject map. The transferable habit is to follow a change through every rule affecting the outcome.
The primary background source is the Federal Reserve Bank of Atlanta’s What Are Benefits Cliffs?. Its examples concern the United States. All formulas, tables and practice cases in this article are original fictional teaching models.
Sources checked 21 October 2026. Worked examples are educational models with the assumptions stated above.