The honest diagnostic test
In quality improvement, the first move when a system is stable at the wrong level is to ask: what level of fix does this problem require? The question matters because a fix applied at Level 1 to a Level 3 problem will not produce a structural change point. It will produce a temporary fluctuation — sometimes large enough to generate a positive press release — followed by a return to the previous trajectory.
UK economic growth has been on a 60-year structural deceleration. The Bootstrap CUSUM analysis of cumulative UK GDP data finds eight distinct structural stages from 1949 to the present, each delivering less growth than the last: +24.1% → +20.5% → +12.5% → +13.3% → +9.9% → +8.2% → +6.0%. The deceleration is continuous across governments of every political persuasion. It predates 2008, predates Brexit, predates Covid. It is a structural feature of the system, not a response to any particular event.
Bootstrap CUSUM finds structural change points at 1958, 1966, 1976, 1986, 1996, 2003, and 2014 — moments of institutional and policy change, not financial crises. The 2008 financial crisis is absorbed entirely within Stage 7 (2003–2014). This means understanding why the structural breaks occurred at those dates is more useful than relitigating 2008, Brexit, or Covid. It is also what Joiner’s framework is designed to investigate: which level of intervention produced each structural shift?
Level 1 fixes — what Chancellors always reach for
A Level 1 fix addresses the output directly. It is reactive, reversible, and changes nothing structural. It is also the overwhelming default of UK economic policy, because it is fast, legible, and generates immediate headlines. The bond markets know this. Gilt yields move on Budget days not because markets expect structural change but because they are watching to see whether a Level 1 adjustment has been made credibly within existing constraints.
| Intervention | Level | Why it is Level 1 |
|---|---|---|
| Income tax rate changes | Level 1 | Adjusts the output (tax take / incentive at the margin). Reversed by the next government. No structural change in what the system is capable of producing. |
| National Insurance adjustments | Level 1 | Changes the cost of employment at the margin. Does not change the supply of labour, skills, or productive capacity. The Reeves 2024 employer NI increase is a textbook Level 1: revenue-raising without structural redesign. |
| Interest rate adjustments (Bank of England) | Level 1 | Modulates demand by changing the cost of borrowing. Addresses the symptom of inflation or low growth. Does not change the productive capacity of the economy. |
| Energy price caps | Level 1 | Protects consumers from the output of an energy price shock. Does not change the structure of energy supply, demand, or pricing. |
| Fuel duty freezes | Level 1 | Adjusts a price signal. No structural effect on transport, logistics, or carbon trajectory. |
| Cost of living payments (2022–23) | Level 1 | Transfers income to offset a supply-side shock. Disappears when payments stop. Changes nothing about why living costs rose. |
A tax cut produces an immediate measurable effect: more money in pockets, higher consumer confidence surveys, a positive growth quarter. This is the improvement equivalent of treating the symptom — the patient feels better, the underlying condition continues. In run chart terms: the metric moves temporarily, then returns to its previous trajectory. No structural change point. The Bootstrap CUSUM line does not inflect.
Level 2 fixes — durable mechanisms that changed behaviour
A Level 2 fix creates a durable mechanism that allows people to act differently within the existing system. It persists across governments because it is embedded in an institution or a legal structure. It does not change what the system is fundamentally capable of, but it does change what individuals can do within it. A Bootstrap CUSUM change point may be detectable in specific sub-metrics (savings rates, participation rates) even where the aggregate GDP trajectory is unchanged.
| Intervention | Level | Why it is Level 2 — and its limits |
|---|---|---|
| ISAs (1999) | Level 2 | Created a durable tax-advantaged savings mechanism. Changes how individuals save within the existing financial system. Does not change what the economy produces or how productively it uses capital. |
| Premium Bonds (1956) | Level 2 | A persistent savings vehicle that has outlasted dozens of governments. Genuine Level 2: changed savings behaviour without requiring ongoing political maintenance. Still operating 70 years later. |
| Auto-enrolment pensions (2012) | Level 2 | The strongest Level 2 in recent decades. Changed the default (opt-in to opt-out), producing a measurable structural change point in pension participation rates. Applies behavioural architecture at system scale. Does not address where pension funds are invested or whether investment produces productive capacity. |
| Help to Buy (2013) | Level 2 — adverse outcome | Created a mechanism for first-time buyers to access mortgages. A genuine Level 2. But the mechanism was aimed at demand rather than supply: it inflated house prices rather than building productive capacity. A Level 2 fix aimed at the wrong constraint. |
| Minimum wage / National Living Wage (1999 / 2016) | Level 2 | A durable process change: sets a floor that persists and is uprated. Has provably changed the distribution of low wages without the employment catastrophe that Level 1 thinking predicted. A durable mechanism with genuine structural effect on the income floor. |
Level 3 fixes — the interventions that shifted structural possibility
A Level 3 fix changes what is structurally possible — who has authority, what resources exist, what institutions operate, which constraints are binding. A genuine Level 3 should be detectable as a structural change point in Bootstrap CUSUM analysis of the relevant sub-metric, even where aggregate GDP trajectory is slow to respond. The structural breaks identified in the UK GDP data at 1966, 1976, 1986, 1996, and 2003 correspond broadly to Level 3 institutional changes, not to tax rate adjustments.
| Intervention | Level | Why it is Level 3 — and the honest assessment |
|---|---|---|
| The Open University (1969) | Level 3 | Created access to higher education that structurally did not exist before. Distance learning, no entry requirements, part-time study. Changed who could participate in the knowledge economy. A genuine structural expansion of possibility. Still operating and still structurally distinctive 55 years later. |
| The National Lottery (1994) | Level 3 | Created a sustained, non-tax funding mechanism for arts, sports, and heritage outside the annual spending review. A genuine Level 3 design: changes what can be funded and how, independently of government fiscal cycles. Has survived multiple governments unchanged in structure because it is genuinely self-sustaining. |
| Bank of England independence (1997) | Level 3 | Changed who has authority over monetary policy. Removed inflation-targeting from the political cycle. A structural redesign of the decision-making system. The Bootstrap CUSUM change point in UK inflation stability at 1997–2003 is partly attributable to this structural change. |
| Scottish Parliament / Welsh Assembly / Devolution (1999) | Level 3 | Changed who has authority over significant areas of policy. Created new decision-making systems with genuine structural independence. The Greater Manchester Combined Authority model (2017 onwards) — including the mayoral model that Andy Burnham developed — is the strongest recent proof case: a Level 3 structural redesign that changed what was locally possible in health, transport, and economic development. |
| Polytechnics becoming universities (1992) | Level 2 masquerading as Level 3 | Raised the credential level of a large number of institutions without matching structural demand for graduate-level skills. Produced credential inflation rather than structural economic capacity. A Level 2 mechanism (new institutions, new titles) that did not address the Level 3 constraint (what the economy needed to absorb graduates productively). |
| Student tuition fees (1998 / 2012) | Level 3 in intent — adverse structural outcome | A genuine Level 3 in intent: changed who pays for higher education at scale, shifting from public to individual funding. But the structural outcome was a debt burden that distorted graduate career choices, location decisions, and risk appetite. A Level 3 change that produced its own Level 3 constraint. |
Level 3 Deep — mindset shifts that changed what government is for
A Level 3 Deep fix changes the fundamental belief about what the system exists to do. These are rare, politically seismic, and — once embedded — nearly impossible to reverse because they change the assumptions that all subsequent decisions are made within. Two examples in UK postwar history are unambiguous.
The two unambiguous Level 3 Deep shifts in postwar UK economic history
Notice the asymmetry: Level 3 Deep changes constrain the options of those who follow, regardless of political persuasion. A genuine Level 3 Deep change is recognisable because subsequent governments inherit its structural constraints even when they disagree with its premises.
The squandered Level 3: North Sea oil
The most instructive case study in UK economic history is the one Level 3 opportunity that was treated as Level 1 revenue. North Sea oil revenues from the late 1970s through the 1980s — estimated at £170 billion at 2023 prices — were deployed as current spending to fund tax cuts and manage the costs of deindustrialisation. They were not invested in a sovereign wealth fund.
Norway applied a Level 3 fix to North Sea oil revenue: created the Government Pension Fund Global (now worth over $1.7 trillion), which invests oil revenue in global equities and bonds and distributes only the investment returns. The structural rule — spend the returns, never the principal — is the Level 3 design. It changes what is permanently possible for Norwegian public finances regardless of the oil price. The UK applied a Level 1 fix to the same resource: spent the revenue as it arrived. The structural opportunity to change what the UK economy was capable of — permanently — was converted into a temporary reduction in the rate of deindustrialisation pain. No structural change point. No lasting difference to the trajectory.
What a new government should try — genuine Level 3 options
With a new Prime Minister and Chancellor expected in late 2026, the question Joiner’s framework asks is simple: which options on the table are Level 3, and which are Level 1 dressed as reform?
The pressure on any incoming Chancellor is to demonstrate fiscal credibility by maintaining the existing fiscal rules (a Level 1 stability signal) while promising structural change (a Level 3 ambition). These are not incompatible — but the Level 3 options require accepting that structural change takes longer to appear in the data than a Budget headline. The Bootstrap CUSUM test of whether an intervention has worked requires data collected after the intervention, at the pre-specified confidence threshold, over the timeframe predicted in advance.
| Option | Level | What would confirm it worked |
|---|---|---|
| Sovereign wealth fund from future energy revenue The proposal: ring-fence revenues from future North Sea oil, offshore wind levies, or carbon pricing into a dedicated fund that invests in global assets and distributes only the returns — never the principal. The structural rule (invest, do not spend) is the mechanism that makes it Level 3: it changes permanently what future public finances can draw on, regardless of which party is in government. |
Level 3 | The intended mechanism: a fund compounding at 4–5% annually for 20 years creates a structural income stream that does not depend on tax rates or economic cycles. Detectable as a Bootstrap CUSUM change point in UK public investment capacity over a 10–15 year horizon. The Norwegian model is the proof case: the Government Pension Fund Global (now over $1.7 trillion) was built on exactly this structural rule applied to North Sea revenue from 1990 onwards. |
| Investment rules that separate productive infrastructure from current spending The proposal: change the fiscal accounting rules so that long-lived productive assets (roads, rail, grid, broadband, housing) are treated like a business investment rather than a government expense. Currently, the UK fiscal rules count infrastructure spending the same as paying wages — it all comes off the same deficit limit. The intention is to unlock investment that the current rules prohibit without the appearance of borrowing. |
Level 3 | Changes what can be invested without breaching fiscal rules — a structural redesign of the accounting boundary between capital and current. The intended mechanism: if £50bn of infrastructure investment no longer counts against the deficit, it can proceed without a tax rise or a spending cut elsewhere. Detectable as a Bootstrap CUSUM change point in UK infrastructure investment as a % of GDP, and eventually in productivity growth. Balancing measure: gilt yields — if markets treat the rule change as fiscal loosening, borrowing costs rise and the benefit is eroded. |
| Real fiscal devolution — tax-raising powers to regions The proposal: give combined authorities genuine fiscal levers — local income tax variation, full business rate retention, borrowing authority against future tax receipts — not just spending authority over a block grant set in Whitehall. The distinction matters: a mayor with a block grant is executing national policy locally (Level 2). A mayor with tax-raising authority can respond to local economic conditions and fund locally-designed interventions (Level 3). |
Level 3 | Changes who has authority over fiscal decisions. The intended mechanism: regions with different economic conditions (skills shortages, housing constraints, transport gaps) can design and fund solutions matched to their specific constraint rather than waiting for national programmes that fit average conditions. Detectable as a Bootstrap CUSUM change point in regional productivity divergence — specifically, convergence between the highest and lowest performing regions. Balancing measure: whether fiscal devolution widens inequality between devolved regions (those with stronger tax bases gain more) rather than reducing it. |
| Graduate tax replacing tuition fees The proposal: replace the current student loan system (where graduates repay a fixed debt linked to their degree cost) with a graduate tax (an additional income tax rate applied after graduation, proportional to earnings, with no fixed debt to repay). Burnham’s stated policy. The intention is to remove the debt-aversion structural constraint on graduate behaviour: currently, graduates from lower-income families are more likely to choose high-salary careers over vocation, more likely to avoid postgraduate study, and more likely to leave regions with lower average salaries to service their debt. |
Level 3 | Removes the debt-aversion structural constraint on graduate career and location choices. The intended mechanism: if graduates pay a proportion of income rather than a fixed debt, the financial penalty for choosing public service careers, regional locations, or further study is reduced. Detectable as a Bootstrap CUSUM change point in graduate migration patterns, subject choice distribution (vocational vs financial), and public sector graduate recruitment rates. Balancing measure: total HE funding — whether a graduate tax generates equivalent revenue to the loan system, or creates a funding gap that reduces university capacity. |
| Maintaining or adjusting fiscal rules The proposal: keep Rachel Reeves’ fiscal framework (current budget in balance, debt falling as a % of GDP) with minor adjustments to headroom. The intention is to signal credibility to bond markets and avoid the gilt crisis that followed the Truss mini-budget of 2022. |
Level 1 | A credibility signal to bond markets. Important for stability — without it, borrowing costs rise and crowd out any Level 3 investment. But it does not change what the economy is structurally capable of producing. The gilt market reaction on Budget days confirms this is being watched as Level 1 management, not Level 3 transformation. Necessary condition for Level 3 fixes to be affordable. Not sufficient to produce them. |
| Levying a new tax (wealth tax, mansion tax, etc.) The proposal: introduce a new tax on wealth, property values above a threshold, or financial assets. The intention is typically redistribution — reducing inequality — or revenue-raising to fund public services without raising income tax. |
Level 1 | Changes the distribution of the existing output. Does not change the size or productive capacity of the system generating that output. A wealth tax that raises £10bn and spends it on current services leaves productive capacity unchanged. Only if the revenue funds a Level 3 structural investment — and the investment is evaluated against a pre-specified outcome measure — does the tax become part of a Level 3 fix. |
The intention: the Greater Manchester Combined Authority (2014 onwards), with an elected metro mayor from 2017, was designed as a Level 3 structural fix. The theory of change: devolving authority over transport, housing, health integration, and economic development to a single elected leader with a mandate and real levers would allow locally-designed interventions matched to Greater Manchester’s specific constraints — rather than national programmes fitted to average conditions.
What the data shows: Greater Manchester has recorded a 31% increase in productivity over the devolution period, becoming the most productive large English city outside London — a result attributed in part to sustained investment in public transport and infrastructure under mayoral authority. The Oxford Road innovation corridor, Metrolink expansion, and integrated health and care budget are cited as structural changes that would not have been possible under the previous fragmented governance model.
What remains contested: a 2026 peer-reviewed study (Sweeney) found that when controlling for initial economic conditions, the measured effect of devolution on growth in Greater Manchester was negative — suggesting the productivity gains may reflect pre-existing advantages rather than the structural effect of devolution itself. The Productivity Institute (2026) notes that commentators disagree on how much of the gain is genuinely attributable to the devolution model versus agglomeration effects that would have occurred anyway.
The Bootstrap CUSUM question: the honest test would be a pre-specified Bootstrap CUSUM analysis of Greater Manchester GVA per hour worked against a synthetic control of comparable non-devolved city regions — identifying whether a structural change point appeared at the point of devolution, or whether the trajectory was already established. That analysis has not been published. The Greater Manchester case is the strongest available evidence for mayoral devolution — but it is not yet confirmed evidence in the Bootstrap CUSUM sense. It is the best available proof of concept, not a confirmed structural change point.
International Bright Spots — Level 3 and Level 3 Deep fixes that produced structural change
A Bright Spot is not a country that is simply richer or luckier than the UK. It is a country that faced a comparable starting constraint — scarce resources, low productivity base, post-war or post-colonial starting position — made a different structural decision, and produced a measurably different outcome trajectory. The Bright Spots question applied at national scale: who is already doing it better, what structural decision did they make differently, and is that decision replicable?
Each case below is framed against Joiner’s levels: what level of fix did they apply, what was the structural mechanism, and what would Bootstrap CUSUM find if applied to the relevant outcome series? See Bright Spots and Positive Deviance for the full investigation framework.
🌎 Four national Bright Spots — Level 3 and Level 3 Deep
Different structural decision: From 1990, Norway legislated that all oil revenue flows into the Government Pension Fund Global and only the investment returns (capped at 3% annually) may be spent. The structural rule — invest the principal, never spend it — is the Level 3 mechanism. It changes permanently what Norwegian public finances can draw on, independent of the oil price or which party governs.
What Bootstrap CUSUM would find: A structural change point in Norwegian public investment capacity at approximately 1996–2000, as fund assets crossed the threshold where returns became material. The fund now exceeds $1.7 trillion — equivalent to approximately $330,000 per Norwegian citizen. UK North Sea revenue over the same period: spent as current income, no structural change point in UK public investment capacity.
The replication question: The structural rule is simple and legible. The UK has future renewable energy revenue (offshore wind, carbon pricing) that could be subject to the same rule. The constraint is political: the structural rule requires a government willing to accept that the revenue cannot be spent in the current Parliament. That is a Level 3 Deep constraint — a belief about what government revenue is for.
Different structural decision: Singapore’s government made a Level 3 Deep decision: the fundamental belief that the only productive asset Singapore possessed was its people. From this belief flowed a 60-year sequence of structural interventions — a five-year education plan (1960) redesigning the curriculum around mathematics, science, and technical skills matched to the economy’s next developmental stage; the Economic Development Board (1961) to attract foreign capital aligned to the skills being built; and continuous curriculum redesign as the economy moved from low-skill manufacturing to electronics to financial services to biomedical research. Each stage anticipated the next constraint rather than reacting to it.
What Bootstrap CUSUM would find: A structural change point in Singapore’s GDP growth trajectory at approximately 1965–1973, when average annual real GDP growth reached 12.7%. By 2024, GDP per capita exceeded $82,000 — among the highest in the world, from a base comparable to the UK’s poorest regions in 1965. Manufacturing’s share of GDP climbed from 14% to 22% within a decade of the structural decisions.
The replication question: The structural practice is anticipatory curriculum design — redesigning the skills pipeline for the economy’s next stage before that stage arrives. The UK consistently redesigns education reactively, after the skills gap is visible in productivity data. The lag between curriculum change and workforce change is 15–20 years. A Bootstrap CUSUM change point in UK productivity would not be detectable for two decades after the curriculum structural change. That lag is why UK governments consistently abandon the intervention before the evidence can appear.
Different structural decision: Estonia legislated digital infrastructure as a constitutional matter from 1997. The Digital Signatures Act (2000) gave digital signatures full legal equivalence with handwritten ones — the structural foundation that made everything else possible. X-Road, a decentralised data exchange system allowing public and private entities to interoperate securely, was built as shared infrastructure rather than siloed departmental systems. By 2024, X-Road handled over 2.7 billion data queries annually. The structural rule: citizen data is entered once and never again; no government agency may ask a citizen for information another agency already holds.
What Bootstrap CUSUM would find: A structural change point in Estonian administrative productivity (measured as transactions per civil servant) at approximately 2000–2005, as X-Road reached critical mass. Estonia ranks second globally in the UN e-government index, 34% above the Central and Eastern European average on digital public services. The OECD notes the productivity potential from digital adoption in Estonia is substantially above the OECD average — though converting digital government efficiency into aggregate productivity growth remains a longer-term lag.
The replication question: The UK has spent decades and billions on digital government without producing equivalent structural change — because it has built departmental systems rather than shared interoperable infrastructure. The structural difference is not the technology; it is the rule that data is entered once. That is a Level 3 design decision about system architecture, not a technology procurement decision.
Different structural decision: South Korea’s government used state-directed industrial policy — a Level 3 structural redesign of which sectors received capital, infrastructure, and protection at which stage of development. Steel in the 1960s, petrochemicals and shipbuilding in the 1970s, semiconductors and electronics in the 1980s, and advanced manufacturing in the 1990s and beyond. At each stage, the state made a pre-specified structural bet on the next constraint the economy would face, built the infrastructure and skills pipeline in advance, and then withdrew support once the sector was self-sustaining.
What Bootstrap CUSUM would find: Multiple structural change points in South Korean GDP per capita: at approximately 1963–1966 (industrialisation begins), 1972–1976 (heavy and chemical industries), and 1986–1990 (electronics exports reach scale). South Korea’s GDP per capita grew from approximately $100 in 1960 to over $33,000 by 2023 — one of the fastest structural transformations in economic history.
The replication question: The structural practice is not "pick winners" — it is "identify the next productive constraint the economy will face, build the infrastructure and skills for it in advance, and pre-specify the Bootstrap CUSUM change point that would confirm the bet was correct." That is a Level 3 discipline the UK has not yet applied systematically.
Every international comparison risks the same trap as the dominant mechanism problem in Disaggregate: attributing the outcome to the visible structural decision rather than the conditions that made it possible. Singapore’s success required a particular political structure (a dominant party with a long planning horizon) that the UK does not have. South Korea’s industrial policy operated in a context of authoritarian government that is not replicable in a liberal democracy. Norway’s sovereign wealth fund required a political consensus about resource revenue that the UK has not achieved. Estonia’s digital infrastructure was built on a clean-slate post-Soviet system without legacy IT to overcome.
The Bright Spots discipline applies here as it does anywhere: separate the structural practice from the context. The replication hypothesis is built on what is structurally replicable, not on wholesale adoption of another country’s system. Norway’s sovereign wealth fund rule is replicable. Norway’s political consensus is not — it must be built separately.
The Bootstrap CUSUM test: has anything worked since 2008?
The honest answer from the data: no structural change point in UK productivity growth is detectable since 2014. Stage 8 of the Bootstrap CUSUM analysis of cumulative UK GDP runs from 2014 to the present. The 2008 financial crisis, the 2010–19 austerity period, the 2020 Covid shock, and the 2021–23 inflation crisis are all absorbed within stages that do not show a fundamental change in the UK’s growth trajectory. The system is behaving consistently with a 60-year deceleration.
This is not an argument for despair. It is an argument for precision. Joiner’s framework says: if the current trajectory is not good enough, and Level 1 and Level 2 fixes have not moved it, the question is what Level 3 structural change has not yet been tried — or has been tried at the wrong scale, in the wrong place, or without the pre-specified prediction that would confirm whether it worked.
The same discipline that applies to an NHS improvement programme applies here. A Chancellor who uses GDP growth as the only measure is using a lag outcome measure with a 12–18 month reporting delay, significant attribution problems, and no leading indicators. The measurement system for any Level 3 economic intervention needs all three types:
- Outcome measure (lag): UK productivity growth per hour worked, or regional productivity convergence — what the intervention ultimately exists to move. Slow to respond, hard to attribute, but the only valid basis for claiming it worked.
- Process measure: Infrastructure investment as % of GDP, pension fund allocation to productive UK assets, or graduate retention rates by region — whether the structural change is actually being implemented as designed.
- Balancing measures: Gilt yields, inflation, regional inequality, and public service capacity — what might be getting worse in another part of the system while the target measure improves. Pre-specify at least two before the intervention begins.
The Bootstrap CUSUM prediction must be made against the outcome measure, not the process measure. Improved infrastructure spending (process) with unchanged productivity (outcome) is not success — it is evidence that the theory connecting the two may be wrong. See Types of Measures: Outcome, Process, and Balancing for the full framework.
Apply the improvement discipline to economic policy
Any Chancellor serious about structural change should be willing to state, in writing, before implementation:
- What will change: the specific structural intervention, not the fiscal adjustment.
- What metric will move: UK productivity growth rate, regional productivity convergence, or investment as % of GDP — not the quarterly GDP headline.
- The direction: up or down.
- The timing: within how many years do you expect a Bootstrap CUSUM change point to be detectable?
- The confidence threshold: what level of evidence would confirm the intervention worked?
- The balancing measures: what could get worse — and you will monitor those simultaneously.
If the prediction cannot be stated before implementation, the intervention is a guess. Level 3 fixes are not inherently better than Level 1 fixes if they are implemented without a testable prediction. The full pre-specified experiment framework — including how to write the prediction, collect the data, and run the Bootstrap CUSUM Study step — is at Stratify, Experiment, Disaggregate: How to Run a PDSA Cycle That Gives You an Honest Answer.