Is Macroeconomics a Science?

TL;DR

• Macroeconomics is a science - but a fragile, conditional one. It is a social science, not a natural science like physics or chemistry.
• The right test isn't Popper's falsifiability but Lakatos's question: is the dominant research programme generating new, empirically confirmed insights - or just patching failures?
• We should trust macroeconomists on mechanisms and trade-offs, be sceptical of precise forecasts, and be honest about where technical analysis ends and political judgement begins.

Introduction

Macroeconomics claims scientific authority over the most powerful lever in modern societies: the management of money, employment, and the business cycle. Central banks set interest rates that determine the cost of every mortgage and every business loan. Treasuries frame budgets that shape the life chances of millions. International institutions impose policy frameworks on sovereign nations as conditions for financial survival. These decisions are justified, explicitly or implicitly, by appeal to macroeconomic science. Whether that appeal is warranted is not a question of academic taxonomy. It is a question about the legitimacy of power.

This essay argues that macroeconomics is a social science, a designation that carries genuine intellectual weight but also hard limitations. It is a discipline that produces real knowledge about aggregate economic behaviour, but knowledge that is contingent, institutionally embedded, and far less precise than either its practitioners or the public typically acknowledge. Its scientific status is not a settled achievement but a conditional and fragile one, sustained only to the extent that the discipline maintains rigorous empirical standards, tolerates productive dissent, and resists the institutional pressures that can degrade inquiry into ideology.

To make this case, the essay must first establish what science means and which philosophical framework best captures macroeconomic practice. It must distinguish macroeconomics from microeconomics, assess the discipline’s empirical methodology and mathematical tools, examine the competing roles of orthodoxy and heterodoxy, and confront macroeconomics’ central function as a form of social engineering. It must also reckon honestly with the discipline’s forecasting record, its international track record, and the institutional forces that shape what counts as legitimate knowledge.

What Do We Mean by Science?

Three Frameworks and a Methodological Defence

The question ‘is macroeconomics a science?’ requires a prior question: what makes any discipline scientific? Three philosophical frameworks dominate this debate, and they yield different verdicts when applied to macroeconomics. Choosing between them is not a neutral exercise; it determines what standard the discipline must meet.

Karl Popper’s criterion of falsifiability provides the most stringent test. A theory is scientific if and only if it makes predictions that could, in principle, be shown to be wrong. By this standard, macroeconomics has a mixed record. Some macroeconomic propositions are clearly falsifiable: the quantity theory of money predicts a relationship between money supply growth and inflation (more money chasing the same goods eventually means higher prices) that can be tested against data (it is not always the case in the short-run – and it was precisely this unreliability that led central banks to abandon money supply targeting in the 1980s and 90s in favour of interest rate instruments – but the relationship appears to hold across nations over decades).

However, many of the discipline’s most important claims involve counterfactuals that are intrinsically difficult to test. The assertion that a stimulus package ‘prevented a deeper recession’ cannot be directly falsified because the counterfactual, what would have happened without the stimulus, is unobservable. If we applied Popper’s criterion strictly, large portions of macroeconomic reasoning would fail the test.

Thomas Kuhn’s framework, developed in The Structure of Scientific Revolutions (1962), shifts attention from individual theories to the sociological structure of scientific communities. For Kuhn, science alternates between periods of ‘normal science,’ conducted within an accepted paradigm, and dramatic revolutions in which the old paradigm is overthrown. Kuhn captures something important about macroeconomics: the discipline does exhibit paradigm-like structures, and its history is punctuated by crisis-driven shifts, from classical to Keynesian economics after the Great Depression, from Keynesianism to monetarism and rational expectations after 1970s stagflation, and eventually to the New Keynesian synthesis in the 1990s that absorbed elements of both. But Kuhn’s framework is better at describing the dynamics of macroeconomic thought than at evaluating its epistemic quality. It tells us that paradigm shifts happen; it does not tell us whether the discipline is progressing or merely cycling.

Imre Lakatos in The Methodology of Scientific Research Programmes (1978) offers the framework that fits macroeconomics most precisely. His concept of research programmes posits that science is organised around a ‘hard core’ of foundational commitments, protected by a ‘protective belt’ of auxiliary hypotheses that can be modified to accommodate anomalous evidence. The critical distinction is between progressive programmes, whose modifications generate novel predictions that are subsequently confirmed, and degenerating programmes, whose modifications are merely ad hoc defences of an embattled core. This framework allows us to ask the most penetrating question about macroeconomics: not whether it is scientific in some abstract sense, but whether its dominant research programmes are currently progressive or degenerating.

Applied to the New Keynesian programme, the answer is genuinely uncertain. After 2008, the mainstream rapidly incorporated financial frictions, the zero lower bound for interest rates, heterogeneous agents, and unconventional monetary policy into its models. Were these modifications progressive, generating new empirical insights, or degenerating, patching a framework that had failed to predict the most important macroeconomic event in eighty years? The honest answer is: partly both. Some modifications, such as the incorporation of financial accelerator mechanisms (whereby deteriorating balance sheets amplify and propagate economic shocks through the credit system), have proved empirically productive. Others look more like retrofitting. The discipline’s scientific status, evaluated by Lakatos’s standard, is genuinely in the balance, and will take time to fully assess.

To these three frameworks we must add Milton Friedman’s profoundly influential 1953 essay, The Methodology of Positive Economics, which provided the methodological self-understanding that most practising economists operate with. Friedman argued that the realism of a theory’s assumptions is irrelevant; what matters is whether the theory generates accurate predictions. This instrumentalist position permits the use of highly simplified assumptions – representative agents, rational expectations, frictionless markets – in the name of predictive tractability. Friedman’s methodology has been enormously productive, but it contains a deep vulnerability: if a model’s assumptions are systematically unrealistic, it provides no reliable basis for predicting the effects of policy changes that move the economy outside the model’s original domain. This is precisely the kind of counterfactual policy analysis that macroeconomics is most needed for, and most trusted to deliver. In Lakatos’s terms, Friedman’s defence is vulnerable precisely where models are asked to support large counterfactual leaps beyond their original domain. This unresolved tension between predictive tractability and structural realism runs through the discipline and, as we shall see, drives much of the debate over empirical methodology and the role of structural models.

r* and the Taylor Rule: A Lakatosian Case Study

The debate over the neutral interest rate, r* – the rate at which monetary policy neither stimulates nor restrains the economy – provides a concrete illustration of the progressive/degenerating distinction in real time. Growth theory predicts that r* should reflect real GDP growth over the long-run. Through the 2010s, however, r* estimated from growth-theoretic foundations diverged persistently from what central banks were observing: rates remained stubbornly low in ways the theory could not explain. The mainstream response was not to interrogate the theory but to replace it with the Holston-Laubach-Williams (HLW) model, a statistical filter that extracts a trend for r* directly from the data without any structural economic model generating it. The unobservable variable was, in effect, estimated from the data it was supposed to explain.

This is a textbook degenerating move in Lakatos’s sense: rather than confronting the anomaly – why doesn’t r* behave as growth theory predicts? – the profession protected its hard core by substituting an atheoretical statistical estimate while continuing to use the same language and policy framework as though the theoretical foundations were intact. That r* estimates from HLW carried enormous confidence intervals, and that a subsequent coding error materially affected the published estimates, only reinforced the concern that mathematical sophistication was creating an illusion of knowledge.

The return of the Phillips curve after the pandemic raises a complementary question. The Taylor Rule, a theoretically grounded monetary policy rule developed by John Taylor in 1993, prescribes that central banks should set interest rates in response to deviations of inflation from target and output from potential. It is simple, transparent, and for a long period described central bank behaviour remarkably well. It fell out of practical relevance during the 2010s when interest rates hit the zero lower bound and the Phillips curve appeared flat: the rule kept prescribing rates that could not be implemented in conditions it was not designed for. Central banks turned instead to unconventional tools – quantitative easing, forward guidance – that the rule could not capture.

Now that inflation has returned, the Phillips curve is active, and interest rates are well above zero, the environment once again resembles the one the Taylor Rule was designed for. If the rule returns to descriptive and prescriptive relevance, it would represent a genuinely progressive development: a theoretically grounded framework reasserting itself after a period of atheoretical improvisation. If it does not, if the structural relationships have shifted permanently, it would suggest that the pre-crisis theoretical architecture requires more fundamental revision than the mainstream has yet undertaken. Either way, the outcome will be a real-time test of whether the discipline’s dominant research programme is progressive or degenerating.

The Natural Science Standard and the Social Science Reality

A final distinction is between the natural sciences and the social sciences. The natural sciences study phenomena that are law-governed, repeatable, and amenable to controlled experimentation. The social sciences study human behaviour, which is reflexive, contingent over multiple domains and resistant to experimental control. Macroeconomics sits firmly on the social science side of this divide. Its objects of study, aggregate output, employment, inflation, are not natural kinds but social constructs, produced by institutions that vary across time and place, and influenced by the very theories that purport to describe them.

This does not make macroeconomics unscientific. It makes it a different kind of science, one whose knowledge is real but bounded, whose predictions are probabilistic rather than deterministic, and whose practitioners must live with a degree of uncertainty that would be unacceptable in physics. The question is whether the discipline acknowledges these limitations honestly or papers over them with false precision.

Macroeconomics and Microeconomics: A Necessary Distinction

Any assessment of economics’ scientific credentials must distinguish between its two principal branches, because they face different epistemic challenges. Microeconomics studies the behaviour of individual agents, consumers, firms, and markets, and can test many of its predictions in controlled or quasi-controlled settings. Experimental economics and behavioural economics have brought something approaching laboratory conditions to microeconomic hypothesis-testing. Macroeconomics, studying the economy as a whole, enjoys no such luxury; its ‘experiments’ are single-shot historical events embedded in unique institutional contexts.

Three features make macroeconomics epistemically harder than microeconomics. First, macroeconomic phenomena are emergent: the behaviour of an entire economy cannot be fully or straightforwardly deduced from the behaviour of its individual agents (we will come back to this contention when we discuss DSGE models below). What is rational for an individual, saving more during a recession, may be catastrophic in the aggregate: if everyone saves fearing a recession, the resulting fall in spending can bring on the very recession they feared, as Keynes’s paradox of thrift illustrates. Second, macroeconomists cannot run controlled experiments. Every macroeconomic ‘experiment’ is a one-off event embedded in a unique historical context, and the number of independent observations is inherently small. Third, the data is limited, noisy, and subject to substantial revision: GDP figures are estimates, initial releases are routinely revised, and measurement error pervades the discipline’s empirical foundations.

The Microfoundations Debate

The relationship between micro and macroeconomics has itself been one of the discipline’s most contentious methodological battlegrounds. Following the Lucas critique of the 1970s, which argued that econometric relationships estimated under one policy regime cannot be relied upon under a different regime because agents adjust their behaviour, the mainstream insisted that macroeconomic models must be built on explicit microfoundations: aggregate behaviour must be derived from the optimising decisions of individual agents.

This insistence produced the Dynamic Stochastic General Equilibrium (DSGE) framework, which became the workhorse of mainstream macroeconomics (although not at the Reserve Bank of Australia). DSGE models are mathematically sophisticated, built around representative agents (for people, firms, etc.) that solve intertemporal optimisation problems using Bellman equations and dynamic programming. They offer internal consistency, logical rigour, and a framework for counterfactual policy experiments.

The criticism is equally powerful. Representative agents suppress the heterogeneity and distributional dynamics that many economists believe are central to macroeconomic phenomena. Rational expectations may be unrealistic in conditions of genuine Knightian uncertainty (when the range of possible outcomes is unknown) such that agents cannot assign meaningful probabilities to all possible states. The pre-crisis generation of DSGE models largely excluded the financial sector, rendering them unable to account for the dynamics that drove the 2008 crisis. The question is whether the microfoundations programme achieved rigour at the cost of realism, and if so, whether the post-crisis modifications have restored the balance or merely disguised the problem.

Macroeconomics as a Social Science: What It Knows and What It Cannot

To classify macroeconomics as a social science is to make two claims simultaneously. The first is affirmative: macroeconomics produces genuine knowledge. It has established that sustained monetary expansion tends to produce inflation, that severe financial crises produce prolonged recessions, that expectations matter for outcomes, and that countercyclical policy can moderate the severity of downturns. These are not trivial achievements.

The second claim is limiting, and the limitations are severe. Macroeconomic knowledge is contingent: the relationships between variables shift as institutions, technologies, and social norms change. It is reflexive: economic agents respond to macroeconomic theories and forecasts, changing the phenomena being studied. And it is inescapably value-laden: the choice of what to measure, what to optimise, and how to frame trade-offs reflects political and moral commitments that precede the analysis.

The Phillips Curve: A Case Study in Contingency and Persistence

The Phillips curve provides the single most instructive case study of what macroeconomic knowledge looks like in practice. In its original form, it posited a stable trade-off between inflation and unemployment, and appeared to hold in the 1950s and 1960s. It broke down spectacularly in the stagflation of the 1970s, when high inflation and high unemployment coexisted, prompting the Friedman-Phelps reformulation in terms of expectations and contributing to the rational expectations revolution. For the next several decades, the empirical relationship appeared to flatten dramatically: unemployment fell to historic lows in the late 2010s with barely a murmur from inflation, leading many economists to question whether the Phillips curve retained any empirical relevance.

Then the post-pandemic period brought the Phillips curve roaring back. The combination of massive fiscal and monetary stimulus, severely disrupted supply chains, and labour markets tightening to historic levels produced exactly the inflationary surge that the original Phillips intuition would have predicted. The relationship that had appeared dead turned out to have been dormant, suppressed by a particular configuration of globalisation, anchored expectations, and central bank credibility that had temporarily flattened the curve. When those conditions were disrupted by a sufficiently large shock, the underlying relationship reasserted itself with a vengeance.

This trajectory illustrates several things at once. Macroeconomic relationships are regime-dependent, sensitive to the institutional context in which they are embedded. But regime dependence is not the same as arbitrariness: underlying economic forces, the relationship between resource utilisation and price pressure, persist even when their surface manifestations change. Apparent structural breaks can be regime-specific rather than permanent. And orthodoxy sometimes has more staying power than its critics assume. The Phillips curve’s resurrection is a reminder that the discipline’s established insights, while neither universal nor eternal, are not lightly discarded.

The Forecasting Record: Structural Understanding versus Point Prediction

If macroeconomics is a science, its predictive record deserves honest examination, and the examination is sobering. Studies of central bank and IMF forecasts have consistently found that macroeconomic forecasters perform poorly at predicting turning points: recessions are almost never forecast in advance, and growth forecasts display a systematic optimism bias. The IMF’s own internal evaluations have acknowledged that its forecasts failed to anticipate the vast majority of recessions in its member countries. As the IMF economist Prakash Loungani observed in a landmark study, ‘the record of failure to predict recessions is virtually unblemished’ – a finding that subsequent IMF research, covering 63 countries from 1992 to 2014, confirmed: forecasters missed the magnitude of downturns by wide margins until the year was almost over.

But how much should this matter? The answer depends on what kind of predictive success we demand. An illuminating analogy is with climate science. Climate models are poor at predicting the weather on a specific day in a specific city, but they are scientifically valuable because they capture the structural dynamics of the climate system: the relationships between greenhouse gas concentrations, radiative forcing, and temperature trajectories over time. Climate science’s authority rests on structural understanding, not point forecasting.

Macroeconomics may be in an analogous position. It is poor at predicting when the next recession will occur, but it has a reasonable structural understanding of why recessions occur, what propagation mechanisms amplify shocks, and what policy interventions can moderate their severity. The discipline’s value lies not in forecasting specific outcomes but in illuminating the causal architecture of economic systems, and in providing frameworks for thinking about the consequences of policy choices under uncertainty.

This distinction between structural understanding and point prediction is crucial for calibrating the authority we grant to macroeconomic pronouncements. Macroeconomists should be listened to when they reason about mechanisms, trade-offs, and the likely direction of effects. They should be treated with considerable scepticism when they offer precise numerical forecasts. The discipline’s failure to make this distinction clearly, and the public’s demand for a precision it cannot deliver, are major sources of the frustration that surrounds macroeconomic expertise.

Empirical Methodology and the Role of Mathematics

The Econometric Toolkit

Given the impossibility of controlled experiments at the macroeconomic level, the discipline has developed sophisticated empirical techniques to identify causal relationships from observational data. Time-series analysis, vector autoregression (VAR) models, and cointegration techniques allow researchers to study dynamic relationships between macroeconomic variables over time. Identification strategies, such as instrumental variables, difference-in-differences estimation, and regression discontinuity designs, attempt to isolate causal effects by exploiting natural experiments or quasi-random variation.

The credibility revolution in empirical economics has had a profound impact on microeconomics but has been harder to apply at the macroeconomic level, where natural experiments are rarer and independent observations inherently limited. Macroeconomists studying fiscal multipliers have exploited cross-country variation in fiscal responses to the 2008 crisis, or within-country variation in military spending driven by geopolitical events, to estimate causal effects. These represent genuine methodological progress, but the identifying assumptions are often debatable, results are sensitive to specification, and external validity – whether findings travel beyond the specific context in which they were estimated – is always uncertain. The toolkit is real and improving; it is not, and may never be, sufficient to resolve the discipline’s deepest empirical disagreements.

Structural Models and Counterfactual Analysis

One of the most important functions of mathematical formalism in macroeconomics is the construction of structural models: models that specify the underlying economic mechanisms, the preferences of agents, the production technologies of firms, and the constraints imposed by institutions, and derive aggregate outcomes from these primitives. Structural models, including DSGE models built around Bellman equations and dynamic programming, are not curve-fitting exercises. They are theoretical constructs that embody specific hypotheses about how the economy works.

Their great virtue is enabling counterfactual policy analysis. Because they specify the mechanisms through which policy affects the economy, they can, in principle, answer ‘what if’ questions that reduced-form empirical analysis cannot: what would happen under a different interest rate rule, or a different fiscal regime? This capacity is essential for macroeconomics’ social engineering function. Without structural models, policy analysis degenerates into extrapolation from historical correlations, precisely the error the Lucas critique identified.

The limitations are equally clear. Structural models are only as reliable as their assumptions, and the assumptions required for tractability are always severe simplifications. Functional forms, parameter calibrations, and shock specifications all involve judgements that materially affect conclusions. Structural models are tools for disciplined counterfactual reasoning, not oracles. They are most useful when their users are explicit about what they assume and honest about what they do not know.

Mathematics: Formalisation, Not Decoration

Mathematics is indispensable in macroeconomics. It provides a language for stating assumptions precisely, deriving their logical implications rigorously, and identifying inconsistencies that verbal reasoning would miss. The Bellman equation provides a mathematically precise framework for analysing intertemporal decision-making under uncertainty, a problem virtually impossible to reason about coherently without formal methods. General equilibrium theory, growth theory, and the study of dynamic systems all depend on mathematical formalism in ways that are substantive rather than decorative.

This is not to deny that mathematics can be misused. Paul Romer’s critique of ‘mathiness,’ the use of formalism to lend an aura of rigour to arguments that are imprecise or tendentious, identifies a real pathology. So does Deirdre McCloskey’s argument that economists sometimes mistake mathematical sophistication for scientific achievement. But the correct response to the misuse of mathematics is not to abandon formalism; it is to insist that formalism earn its keep. Mathematics is essential when it clarifies logical structure, enables testable predictions, or makes counterfactual analysis possible. It is problematic only when it becomes an end in itself, disconnected from empirical reality. The discipline’s challenge is not too much mathematics but too little insistence that mathematical models demonstrate their empirical relevance.

The Role of Orthodoxy: Normal Science and Its Risks

At any given time, macroeconomics is dominated by an orthodox paradigm that defines accepted models, methods, and policy prescriptions. This orthodoxy provides essential infrastructure: a shared language, cumulative research traditions, training pipelines, and coherent frameworks for policy action. The post-war Keynesian consensus, the monetarist and rational expectations revolution that succeeded it, and the New Keynesian synthesis that emerged from the 1990s onward each provided the intellectual architecture for their era’s dominant policy framework.

The mainstream economic community, centred in elite departments, central banks, international institutions, and leading journals, guards this orthodoxy through peer review, professional norms, and career incentive structures. In Lakatos’s terms, the community protects the hard core of its research programme while allowing the protective belt to be modified in response to evidence. This is not a deficiency; it is how all scientific communities function, and the discipline would be incoherent without it.

But the sociology of the profession creates real risks. Groupthink can entrench. Anomalies that do not fit the paradigm can be explained away rather than confronted. Heterodox voices can be marginalised not because their ideas have been rigorously tested and found wanting but because they challenge assumptions on which careers and reputations rest. The profession’s failure to anticipate the 2008 financial crisis owed something to this dynamic: financial fragility and the possibility of systemic collapse were understudied not because they were unknowable but because they were outside the paradigm. The dominant models assumed efficient financial markets and focused on technology shocks as drivers of business cycles, leaving the profession intellectually unprepared for a crisis driven by leverage, interconnection, and panic.

The critical question, in Lakatos’s framework, is whether the mainstream’s response to such failures is progressive or degenerating. A progressive response generates genuine new insights; a degenerating response merely patches the existing framework to immunise it from criticism. The post-2008 modifications to mainstream macro, incorporating financial frictions, heterogeneous agents, and unconventional policy tools, contain elements of both. The discipline’s scientific credibility depends on which tendency prevails.

The Role of Heterodoxy: Challenge, Innovation, and Revolution

If orthodoxy represents normal science, heterodoxy represents the ferment from which paradigm renewal can emerge. Heterodox economics encompasses post-Keynesian, Modern Monetary Theory (MMT), Austrian, Marxian, ecological, and institutional schools, among others. They share a defining characteristic: each challenges one or more foundational assumptions of the mainstream.

The value of heterodoxy lies not in the assumption that dissenters are right but in their structural function within the discipline’s ecosystem. Heterodox thinkers ask questions the mainstream neglects, develop alternative frameworks that may prove essential when the paradigm encounters anomalies it cannot resolve, and keep the discipline honest by refusing to treat current orthodoxy as the final word.

Heterodox Ideas and Their Trajectories

The history of macroeconomics can be read as a series of Kuhnian revolutions precipitated by crises the prevailing orthodoxy could not explain. The Great Depression discredited classical economics and ushered in the Keynesian revolution. Stagflation discredited crude Keynesianism and ushered in monetarism and rational expectations. In each case, the revolutionary ideas had been developed within heterodox communities for years or decades before the mainstream was ready to take them seriously.

Hyman Minsky’s financial instability hypothesis illustrates the pattern precisely. Minsky argued that financial stability breeds risk-taking that endogenously generates instability, a claim the mainstream dismissed for decades. After 2008, ‘Minsky moments’ entered the vocabulary of central bankers, and financial frictions were rapidly incorporated into DSGE models. The trajectory, from heterodox insight to mainstream absorption, illustrates both the value of heterodoxy and the capacity of a healthy orthodoxy to evolve.

Modern Monetary Theory offers a more contested recent example. MMT’s core claims, that a sovereign currency issuer cannot run out of money and that the binding constraint on government spending is inflation rather than solvency, were long treated as eccentric. Yet during the pandemic, governments engaged in fiscal expansions that bore more resemblance to MMT prescriptions than to orthodox fiscal frameworks. Whether this vindicated MMT, represented pragmatic crisis response, or served as a cautionary tale about inflationary consequences remains a live debate. But it illustrates how heterodox ideas can shape policy even without formal mainstream acceptance.

The Limits of Heterodoxy: Signal, Noise, and the Stopped Clock

A necessary caution. The foregoing risks romanticising heterodoxy, and romanticism is as dangerous as complacency.

The reality is that the overwhelming majority of heterodox predictions turn out to be wrong. For every Minsky whose warnings were vindicated, there are dozens of commentators who predicted crises that never materialised, hyperinflation that never arrived, or collapses that stubbornly refused to occur. Every decision of the Reserve Bank of Australia on interest rates, every Federal Reserve announcement, every fiscal budget is met by a chorus of critics predicting that the economy is about to go to hell in a handbasket. Most of the time, it does not. The economy muddles through, the predicted disaster fails to arrive, and the critics move on without accounting for their prior errors.

There is a deep selection bias at work. We remember the contrarians who were right and construct narratives of heroic dissent. We forget the vastly larger number who were wrong, because failed predictions of catastrophe are not newsworthy. A stopped clock is right twice a day, and a permanent prediction of crisis will eventually be vindicated, but this does not validate the analytical framework that produced the prediction. The question is whether a framework provides a systematic and reliable basis for understanding when and why crises occur, a far more demanding standard that few heterodox approaches meet.

Moreover, the practical demands of economic governance require a functioning orthodoxy in a way that is easy to underestimate from the outside. Central banks must make decisions at regular intervals. Treasuries must frame budgets. Regulators must set capital requirements. These decisions cannot wait for paradigm debates to be resolved. They require a shared analytical framework, however imperfect, that allows policymakers to assess trade-offs, communicate reasoning, and be held accountable. A discipline composed entirely of heterodox dissenters, each operating from different premises, would not provide this.

The correct posture is neither deference to orthodoxy nor reflexive sympathy for dissent. It is a discriminating scepticism that takes the mainstream seriously as the best available working framework while remaining open to challenges that meet a high evidential bar. Heterodoxy earns its place not by the volume of its objections but by the quality of its alternatives: their internal coherence, their empirical support, and their capacity to explain phenomena the mainstream cannot.

Macroeconomics as Social Engineering

The feature that most clearly distinguishes macroeconomics from the natural sciences is its intimate, constitutive relationship with policy. Macroeconomics is not merely a descriptive enterprise; it is a form of social engineering. Its central purpose, at least since Keynes, has been to provide governments and central banks with the intellectual tools to manage national economies. This is both its greatest practical contribution and its deepest epistemological complication: macroeconomic theories do not merely describe the economy; they reshape it.

Smoothing the Business Cycle

Capitalist economies are inherently cyclical. The Keynesian insight, now broadly accepted even by economists who would not call themselves Keynesians, is that governments can and should use fiscal and monetary policy to moderate these cycles: stimulating demand during recessions and restraining it during overheating. The tools have evolved, from discretionary fiscal policy to monetary policy conducted through interest rate adjustments and, since 2008, unconventional measures such as quantitative easing and forward guidance, but the basic principle of countercyclical intervention remains one of macroeconomics’ most important practical contributions. The precise mechanisms, timing, and magnitudes remain subjects of intense debate, but the post-war record of reduced output volatility relative to the pre-war era suggests the principle has substance.

Maximising Employment

The second objective is achieving the highest sustainable level of employment consistent with price stability. The concept of the Non-Accelerating Inflation Rate of Unemployment (NAIRU) attempts to define this technically, positing a level of unemployment below which inflationary pressures build. Both the concept and its estimation remain contentious: the NAIRU is not directly observable, estimates vary widely, and critics argue that treating it as a structural constant ignores the ways in which policy itself can influence productive capacity. The hysteresis hypothesis, which holds that prolonged unemployment permanently damages potential output by eroding skills and discouraging participation, strengthens the case for aggressive early intervention and complicates any clean separation between cyclical and structural unemployment.

The human costs of unemployment, not only lost output but diminished wellbeing, deteriorating health, and social dislocation, make this more than a technical question. The weight assigned to unemployment versus inflation reflects moral commitment, not merely scientific judgement.

Minimising Inflation

Sustained high inflation erodes purchasing power, distorts price signals, redistributes wealth arbitrarily, and undermines the social contract. The hard-won lesson of the 1970s and 1980s was that inflation, once established, is costly to bring under control. Modern central banking, with its emphasis on credible inflation targets, operational independence, and forward guidance, represents the institutional embodiment of this insight.

The post-pandemic inflation of 2021–2023 tested these frameworks anew. The initial debate over whether inflation was ‘transitory’ or persistent exposed deep divisions. The subsequent tightening cycle illustrated both the power and the bluntness of monetary policy: effective at reducing inflation but at the cost of slower growth and higher unemployment, with distributional burdens falling disproportionately on borrowers, the young, and interest-sensitive sectors. The episode confirmed that the trade-offs macroeconomics identifies are real, even if the discipline cannot resolve them with the precision it sometimes implies.

Science, Values, and the Limits of Technical Authority

The social engineering function of macroeconomics exposes a boundary that the discipline frequently acknowledges in theory and routinely violates in practice: the boundary between positive economics and normative economics. Positive economics asks what is: what are the causal relationships between variables, what are the likely consequences of a given policy? Normative economics asks what ought to be: which outcomes should we pursue, and how should we weigh competing objectives against one another? Friedman’s 1953 essay was explicitly about the methodology of positive economics, and he was clear that normative questions belonged to a separate domain. The distinction is philosophically sound. The problem is that macroeconomic policy advice, the discipline’s primary public function, inherently requires both.

Science can estimate the likely effect of raising interest rates on inflation and unemployment. It cannot tell us whether reducing inflation by two percentage points is worth the additional unemployment it causes. That is a distributional and moral question: who bears the costs, who reaps the benefits, and how do we weigh the welfare of borrowers against the welfare of savers, the young against the old, the employed against those who lose their jobs? These are value judgements, and by definition, value judgements are not scientific. No amount of econometric sophistication can derive an ‘ought’ from an ‘is.’

Yet macroeconomic practice routinely embeds normative choices within ostensibly technical frameworks. When a central bank targets 2% inflation rather than 4%, that is a value judgement about the acceptable trade-off between price stability and other objectives, presented as a technical calibration. When the IMF prescribes fiscal austerity as a condition for assistance, the distributional consequences, which sectors contract, which workers lose their jobs, which public services are cut, are normative choices, not scientific findings. When policymakers invoke the NAIRU to argue against further stimulus, they are making a judgement about the acceptable level of unemployment, a judgement that falls hardest on those least able to bear it.

The pretence that these choices are merely technical, that they follow inevitably from the science, is one of the most corrosive forms of the false precision this essay warns against. It allows political decisions to be depoliticised, removed from democratic contestation and placed in the hands of technocrats whose authority rests on a claimed scientific objectivity that, at the normative level, does not and cannot exist. This is not an argument against technocratic expertise; central banks and treasuries need skilled economists. It is an argument for honesty about where the science ends and the politics begins.

A macroeconomics that is honest about this boundary is stronger, not weaker. It can say with scientific authority: ‘if you raise interest rates, unemployment will likely increase by this range.’ It cannot say with scientific authority: ‘therefore you should raise interest rates.’ The ‘therefore’ requires a value judgement about the relative importance of inflation and employment that belongs to democratic deliberation, not to economic science. Acknowledging this does not diminish macroeconomics; it clarifies what the discipline can legitimately claim and protects its scientific credibility from being eroded by overreach.

The Inescapable Trade-offs

These three objectives, cycle smoothing, employment maximisation, and inflation control, exist in irreducible tension. Policies that reduce unemployment may increase inflation. Policies that control inflation may increase unemployment. Policies that smooth the cycle in the short run may create moral hazard or asset bubbles that amplify instability in the long run. These trade-offs would be difficult enough even with precise knowledge; in a discipline operating under deep uncertainty, they are unavoidable and politically charged. Managing them is not a technical exercise; it requires political judgement, distributional choices, and an acceptance of uncertainty that sits uneasily with the aspiration to scientific precision.

The International Dimension and the Politics of Knowledge

Macroeconomics Beyond the Advanced Economies

The argument so far has been implicitly centred on advanced-economy experience. Any assessment of macroeconomics’ scientific credentials must also reckon with its application in the developing world, where the discipline’s prescriptions have been tested under very different conditions and with deeply contested results.

The structural adjustment programmes imposed by the IMF and World Bank from the 1980s onward, demanding fiscal austerity, trade liberalisation, privatisation, and deregulation as conditions for assistance, represented the most ambitious exercise in macroeconomic social engineering ever attempted. The results were, at best, mixed. In many countries, structural adjustment was associated with deindustrialisation, rising inequality, and prolonged stagnation. The East Asian crisis of 1997–1998, in which IMF prescriptions were widely seen as exacerbating the crisis, prompted fundamental reassessment of the orthodox policy framework.

The developing-country experience raises uncomfortable questions about the universality of macroeconomic knowledge. If relationships identified in advanced economies do not hold in different institutional contexts, the discipline’s claim to produce generalisable knowledge is weakened. A science that works only under specific institutional conditions is not thereby unscientific, but it must be transparent about the boundaries of its applicability, and the record suggests that this transparency has often been lacking.

Power, Institutions, and Ideology

The sociology of economic knowledge production deserves more scrutiny than it typically receives. Who funds economic research? How do revolving doors between academia, central banks, and the private financial sector shape what counts as legitimate knowledge? How do hiring practices at elite departments and editorial policies at leading journals determine which questions are asked and which methods considered acceptable?

These are not idle questions, and they go beyond the generic observation that all disciplines have sociological biases. In macroeconomics, the potential for institutional capture is unusually direct. Central banks employ academic economists and fund academic research. Financial institutions sponsor conferences and endow chairs. The career incentives facing young macroeconomists, the need to publish in a small number of gatekeeping journals, to secure positions at a small number of elite departments, create conformity pressures that may be subtle but are real. The discipline’s overwhelming consensus on the benefits of financial liberalisation in the decades before 2008, a consensus now acknowledged to have been at least partly mistaken, was shaped not only by theory and evidence but by the institutional environment in which that theory and evidence were produced.

The question this raises is pointed: does institutional capture actively degrade epistemic standards? Does consensus become endogenous to power structures rather than to evidence? Does the discipline systematically underprice tail risks because acknowledging them would be inconvenient for the institutions that fund and employ its practitioners? The pre-2008 consensus on financial liberalisation offers a concrete case. The mainstream view that deregulated financial markets were self-correcting and self-stabilising was not merely a theoretical conclusion; it was sustained by a professional ecosystem in which the financial sector funded research, employed economists in lucrative consulting roles, and populated the advisory boards of the institutions producing the research. When the consensus proved catastrophically wrong, the intellectual failure could not be cleanly separated from the institutional environment that had produced and sustained it. None of this means macroeconomics is merely ideology dressed as science. But a discipline that aspires to scientific status must be willing to subject its own knowledge-production processes to the same critical scrutiny it applies to the phenomena it studies. On this score, the discipline’s record is uneven.

Conclusion

Is macroeconomics a science? Yes. But its scientific status is not a settled achievement. It is a conditional and precarious standing, maintained only through ongoing effort and constantly at risk of erosion.

Macroeconomics is a social science that produces genuine knowledge about aggregate economic behaviour. Its mathematical formalism is not decoration but an essential tool for formalising relationships, maintaining logical consistency, and conducting the counterfactual policy analysis on which its social engineering function depends. Its empirical toolkit, while operating under constraints that limit certainty, represents real and improving methodological sophistication. Its accumulated findings, on the dynamics of inflation, the propagation of financial crises, the mechanisms of business cycle stabilisation, constitute substantive intellectual achievement.

But macroeconomics cannot be, and should not pretend to be, a science in the sense that physics is a science. It cannot conduct controlled experiments. Its predictions are probabilistic and frequently wrong. Its knowledge is regime-dependent: relationships that hold under one institutional configuration may not hold under another. Its practitioners are embedded in the political and financial systems they study, creating risks of institutional capture that the discipline has not adequately confronted.

Evaluated by Lakatos’s criterion, the most demanding and most appropriate standard, the discipline’s status is genuinely uncertain. Its dominant research programmes contain both progressive elements, genuine extensions that have generated new empirical insights, and degenerating elements, ad hoc modifications that serve to protect an embattled core rather than advance understanding. The balance between these tendencies is not fixed; it depends on choices that the profession makes about what to study, how to evaluate evidence, and how to treat dissent.

The discipline’s health requires maintaining a productive tension between orthodoxy and heterodoxy. Orthodoxy provides the institutional and intellectual infrastructure that economic governance requires: the shared frameworks, the cumulative research traditions, the common language that allows central banks to make decisions and be held accountable. Heterodoxy provides the creative disruption from which paradigm renewal can emerge. But heterodoxy must be evaluated with the same rigour as orthodoxy; the mere fact of dissent confers no epistemic privilege. The stopped clock is no more scientific than the consensus it criticises.

What macroeconomics owes its practitioners and the societies they serve is honesty about the boundaries of its knowledge. It should be listened to when it reasons about mechanisms and trade-offs. It should be treated with scepticism when it offers precise forecasts. It should be challenged when its institutional structures create incentives for conformity over inquiry. And it should be held to the standard it claims for itself: that its conclusions follow from evidence and logic, not from convenience, convention, or power.

Macroeconomics at its best is a powerful tool for understanding and managing the complex systems upon which human prosperity depends. At its worst, it is a source of false certainty and ideological cover. Under Lakatos’s framework, the discipline remains scientific if and only if its modifications generate novel, empirically confirmed predictions rather than post hoc rationalisations. The difference lies not in whether we call it a science, but in whether its practitioners have the intellectual courage to behave as scientists: to follow evidence where it leads, to acknowledge what they do not know, and to treat the authority society grants them not as an entitlement but as a trust that must be continuously earned, if it is to retain any claim to the legitimacy of power.