In 2027, the economic plans of candidates running for the French presidency will be analyzed and compared using macroeconomic models. These models help predict the outcomes of policy proposals, but they rely on assumptions that are often not made public. This raises a democratic concern: before any political discussion even begins, these assumptions can shape what is seen as "serious" or "untenable" in policy terms. In 2022 and 2024, two macroeconomic models were used to evaluate similar left-wing economic programs, and they reached very different results. One predicted a rise in public debt by more than five percentage points of GDP, while the other suggested a decrease of ten percentage points. The difference came largely from the models themselves and their underlying assumptions, which were not discussed in public. In 2027, similar scrutiny will be applied to presidential programs, and the assumptions behind the models will again influence public perception of what is feasible. Macroeconomic models are tools that simplify complex economic realities. They use data on employment, GDP, and inflation to simulate the effects of policy changes. In France, these models are used by government agencies, research institutions, and financial organizations. However, they are not neutral. The way they are built influences which economic problems are considered and how solutions are approached. A chapter of the roadmap submitted to the United Nations Human Rights Council in June 2026 highlights how the assumptions used in models can affect their outcomes. These assumptions are rarely discussed publicly, yet they can shape what is seen as realistic or impossible before any political decisions are made. Macroeconomic models focus on specific aspects of the economy and make assumptions about how people and businesses behave. For example, many models used in France simplify all households and companies into a single "representative agent," assuming that everyone responds the same way to economic changes. In reality, the effects of policies can be very different for rich and poor households. These models also make assumptions about how people and companies adapt to changes in their economic situation. For example, many models used in environmental policy assume that higher fossil fuel prices will lead companies to switch to cleaner energy sources. This depends on a parameter called "elasticity of substitution," which determines how easily and cheaply the transition can happen. The models also need to decide which economic variables are fixed and which can change. This is called the "closure of the model." For instance, some models assume that investment creates savings, while others say savings limit investment. These choices affect who bears the impact of economic shocks. One model might predict falling wages in a crisis, while another might predict job losses. To illustrate the importance of these assumptions, consider the European Commission’s evaluation of carbon neutrality by 2050. Different models produced contrasting results for the same scenarios. One model, GEM-E3, assumed that the economy is already operating at full capacity, meaning that investing in green energy would take resources away from other sectors. Another model, E3ME, assumed that unused resources, such as idle factories and unemployed workers, were available to support the transition. The first model predicted a 1.3% drop in GDP by 2050, while the second predicted a 2.2% increase. The same scenario, two very different outcomes. There is no perfect model. Each is better suited to different questions. The challenge is to make the assumptions behind the models clear and allow for comparison between results. This requires transparency: institutions should keep a public record of the models they use and explain which ones were used in each evaluation. There is also a need for diversity in modeling approaches. Evaluations would benefit from using at least two different modeling traditions. While using a single model ensures consistency over time, this should not prevent the use of alternative models for comparison. This kind of approach is already used in the work of the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services (IPBES), which uses thirteen different models to assess trends in biodiversity and ecosystem services. These include models of ecological processes, statistical methods, and ecosystem service evaluations. Just as models used before the 2008 financial crisis failed to predict the crisis because they did not account for the complexity of the financial sector, models that ignore health, inequalities, or environmental conditions cannot fully assess the impact of policies. Evaluating the feasibility of the transition should take these factors into account. Documenting and comparing models will not make them neutral, but it will make their assumptions clear and open to discussion. Behind every number is a representation of how the economy works: Is finance just a middleman? Do inequalities affect growth? Do ecological limits change how the economy functions? Understanding how economic expertise is built helps to remove technical barriers and support democratic debate.