20.10.2026

Can science be free from ideology? Economists have long had a ready answer. Since Lionel Robbins and Milton Friedman, the discipline has distinguished positive economics, which describes how the economy works, from normative economics, which says how it should work. The first is science; the second is left to ethics and politics. This is the picture I was trained with, and it remains, by and large, the discipline’s official account of itself.

Philosophers of science have mostly given up on the idea behind it: that science can be value-free. The argument is not that scientists are biased, or that science is ideology in disguise. It is that doing science requires choices the evidence alone cannot settle. How much evidence is enough to accept a hypothesis depends on what is at stake if one is wrong. Building a model requires deciding what to represent and what to leave out. In both cases values are needed to make the choice, and they are not a defect that more care could remove. The question is therefore no longer whether values enter science, but which ones legitimately may, and how. Philosophers call this the new demarcation problem.

Climate economics is a particularly interesting case for this question. Climate policy relies on models that couple a simplified climate to a simplified economy. Their outputs, such as the social cost of carbon, enter policy directly. The field’s best-known disagreement shows how values enter them. In the 2000s, Nicholas Stern and William Nordhaus reached very different conclusions about how fast emissions should be cut, from broadly similar climate science. The difference came largely from the discount rate, that is, from how much the welfare of future generations should count against ours. In other words, the disagreement was about ethics, and both sides knew it.

This is in fact the easy case: The value judgment is visible and can be argued about. Many others are not. Take a decision every modeller faces: how many agents the model contains. A single representative consumer makes the model much easier to solve. Yet a model with one consumer cannot express the idea that a policy helps some people and harms others. The distributional question does not get a bad answer; it simply cannot be asked. The choice is made for technical reasons, but it has ethical consequences. Similarly, the damage function, which links temperature to economic losses, is estimated on the narrow range of warming observed so far and extrapolated to temperatures for which there are no data. Robert Pindyck has argued that, at the warming levels that matter most, these functions are close to arbitrary. Where the evidence runs out, the modelling assumption does the work, together with whatever judgment it carries.

Philosophers of economics have long argued that concepts such as efficiency or welfare describe and evaluate at the same time. What has received much less attention is the next question: Once we accept that values are there, how are they handled?

Climate science provides a useful comparison. The value judgments involved in modelling the climate and communicating uncertainty about it have been discussed at length, with effects: The calibrated language the IPCC now uses to report confidence (“likely”, “very likely”) can be seen in part as an institutional response to it. Criticism of the economic side has been at least as intense, from inside economics as much as from outside. Yet the standard models and their core parameters have changed little.

Why this difference? I see two possible explanations. The first concerns the epistemic virtues economics prioritises, that is, the qualities that make a model good as a piece of science: fitting the data, being simple, explaining why things happen, and so on. These virtues often pull in different directions, and every discipline has to rank them. Mainstream economics gives a high value to models it can solve because it provides a deep understanding of the results it produces. One can, for instance, trace a conclusion back to the assumptions that produce it, and see what would change if an assumption changed. Economists tend to value this kind of understanding more than a close match with the data. A model that reproduces the world well but whose workings remain opaque is, by these standards, less valued.

This preference is not unreasonable, but whether it is appropriate depends on what the model is for. If the aim is to understand a mechanism, for instance how a carbon tax changes incentives, a simple model that can be solved may be what is needed. But if the aim is to estimate how much damage a given amount of warming will cause, a model that is easy to understand but misrepresents a non-linear climate is a poor tool. Philosophers of science call this adequacy for purpose: A model is not good or bad in general, but good or bad for something. Climate-economic models are often used for the second kind of task while being judged by the standards of the first. Models that are more faithful to a complex physical system but harder to solve may then look like weaker science, rather than science with different priorities.

The second explanation concerns who is in charge of the field. The concepts used in climate-economic models come from economics, but many of those who build these models are not academic economists, and academic economists have engaged comparatively little with climate assessment. Critics then attack “neoclassical economics” as embodied in these models, and economists reply that this is not how their discipline uses the concepts. Both may be right, and criticism ends up with no clear addressee.

If values cannot be removed, they have to be managed. Some institutions already try. Germany’s Federal Environment Agency, for instance, recommends a cost of €350 per tonne of CO₂ emitted in 2026, but reports alongside it a value of €1,000 that weights present and future generations equally, making the ethical choice visible.

This is a good start, but transparency does not always suffice. Disclosing a discount rate answers the objection that it was hidden, not the objection that it encodes a view of intergenerational justice that those who will bear its consequences had no part in setting. In the 1990s, the IPCC’s Second Assessment Report was criticised for valuing statistical lives lost in poor countries at a fraction of those lost in rich ones. The problem was not that the assumption was hidden, but that it took a position on whose losses count, and on what terms; the question is still debated today. Different value choices seem to call for different remedies: sometimes disclosure, sometimes broader participation, sometimes a community able to criticise the choice at all.

So, can science be free from ideology? If ideology simply means values, the answer is no, and economics is no exception. But this does not make all science ideological in the pejorative sense. What distinguishes a legitimate use of values from an illegitimate one is not their absence but how they are handled: whether they are visible, whether those affected have a say, and whether criticism of them can be heard. Economics built much of its identity on the separation between positive and normative questions. It has not yet systematically asked how it handles the values this separation leaves in place. Climate economics, where the stakes are high and the evidence is often thin, seems a good place to start.

Authors

References →