Physics models can have a wide variety of different simplifying assumptions. Models in some areas will include a full description of relativistic behaviour, while others will not. Some models describe objects based on the quantum mechanical interactions between particles, some treat gas molecules as essentially point particles that don’t interact. These different models and different simplifying assumptions are necessary. Even though physicists are confident that objects in the real world always behave according to quantum mechanics and relativity, it is simply impossible to understand many real world physical phenomena without abstracting away from the complications introduced by the most advanced theories in physics.
Macroeconomists understand that making simplifying assumptions is often necessary to make models work. In fact, macroeconomists often use assumptions that seem ridiculous to anyone outside the field. For example when modelling housing markets in the most common type of macroeconomic models (DSGE models), it is common to assume that there is a chance that a house will randomly decrease in size or quality. Other examples of typically made assumptions include that the economy can be described using a representative consumer, that people live forever, and that everyone is taxed an amount that is not based on income.
Economists are often criticised for making simplifying assumptions, and they are correct when they say all models must make simplifications. Critics of economics that simply point to the existence of simplifying assumptions are thus misguided: simplifying assumptions are necessary to understand the world, and physicists would not be able to do much of the work they do without them.
However there is one important respect in which macroeconomics is different from physics. In modern macroeconomics there is one simplifying assumption that can never be made. In order to be published in a top macro journal your paper must assume that consumers optimise utility over their lifetimes, and explicitly include the calculations they make in your model.
This makes very little sense. Even if you believe that on a fundamental level consumers are maximizing utility over their lifetimes it doesn’t make sense that every model must explicitly include their calculations. Even physicists, who are very certain their most advanced descriptions of reality are close to correct do not insist that all models explicitly include calculations that start from those models. Model assumptions should be chosen based on the particular modelling situation involved.
Macroeconomists might argue that all models must incorporate consumers who optimise consumption over their lifetime by arguing that such consumers are so important to macroeconomics that it does not make sense to ever abstract away from them. Unfortunately, arguments demonstrating that have not been provided. In fact, explicitly modelling consumers as optimising is required even when it is known to give the wrong results.
For example, up until recently most macro models have had a feature known as Ricardian Equivalence (RE). Under Ricardian equivalence, government debt does not change consumer behaviour, because consumers anticipate having to pay the debt back, and so simply save it in their bank accounts in order to pay the future taxes. In the earlier models where everyone was taxed an amount that didn’t depend on income RE holds exactly, while in more modern models where taxes are based on income RE holds in an approximate sense.
However, it has been known since 1986 that consumers are much less Ricardian when taxes are based on income if their incomes are uncertain. The intuition behind this result is that even if the government is going to pay back its debt, if incomes are uncertain then you do not know that you are going to be the one facing higher taxes. Thus it does not make sense to save in anticipation of future tax increases. Similar departures from RE can be found if consumers do not live forever.
The most advanced modern macro models (Heterogeneous Agent New Keynesian, or HANK models) do incorporate uncertainty about future incomes, and so are finally starting to get results that depart from Ricardian Equivalence substantially. However, for 30 years all macroeconomic models have been incorporating consumer behaviour that simply does not match how optimising consumers would behave in the real world. Clearly, the decision that no macro model can ever abstract away from optimising behaviour is not based on whether doing so actually helps macroeconomists get the right answers.
If you are familiar with physics you will understand how it would distort the results of the field if every model had to have a full quantum mechanical description of the world. Research in many areas would simply be impossible, so physics would basically consist entirely of research into quantum mechanics. For economists, insight into how the macroeconomic research agenda is distorted by the fact that every model must include optimising consumers can be found by considering how macroeconomics would look if different features absolutely had to be included in macroeconomic models. For example, if all models had to incorporate heterogeneous agents and income uncertainty, basically nothing in macro would have been publishable until very recently, since HANK models are so difficult to solve. Similarly, if all models had to incorporate finite life spans then most current macro papers would be unpublishable. Insisting that every model incorporate certain features distorts the research in a field.
Indeed, currently academic macro literature has very little to say about many of the most important macro questions. Questions like whether a certain level of debt is sustainable or not, what fiscal policy should be during recessions and the like are barely touched on by the mainstream literature. As a result, macroeconomic policies are often formed based on the unarticulated and uncriticized mental models of macroeconomists who reached their positions by showing expertise on models that have no bearing on the questions at hand. Since these models are not formally articulated or critiqued in the academic literature, they are often bad, so macro policy is decided based on little more than gut feelings (for example Larry summers infamous ⅓ chance of inflation comment). There are models that do not include optimising behaviour that are used by governments, simply because they need to have a tool for certain tasks, but the fact that those models are not discussed in the academic literature means they are undoubtedly less developed than they should be. Overall, the determination to never abstract away from optimising behaviour means macro policy is usually based on little more than gut feelings.
How did macroeconomics get this way? I am not a historian of economics, and would be very interested in what a historian of economics would say. None of the arguments that supposedly brought about this change seem very convincing to me, so I suspect a large part of the reason was political. It is much easier to say that the economy is efficient and that the government should do nothing within the prevailing macro paradigm than to say anything else. The models that existed prior to this undoubtedly had their empirical and theoretical failings, which helped support the radical paradigm shift, but these do not seem to justify the totality of the shift in paradigms.
Since then there have undoubtedly been selection effects going on, with the people going into macro being those that find the prevailing macro paradigm appealing. Individual macroeconomists also benefit from excluding other approaches from the top journals. Less competition to publish in top journals because it means more space in those journals for their papers, and a better chance that they obtain one of a limited number of academic positions. In addition, they get to suggest policies based on their gut feelings without ever really articulating good arguments for them, which might no doubt be appealing to some.
However, I do believe that macroeconomists today are not necessarily as against government intervention today as the ones who created the first macro models based on optimising behaviour are. There seem to be many attempts to say more realistic things in the prevailing macroeconomic framework. Unfortunately, the tools they are required to use mean their results are biased in a certain way, and so the results of the field are much more against government intervention than they might otherwise be.
Currently, academic macro models are pretty irrelevant to many practical economic debates. Given the state of the field I would argue that is correct. However, I do think we should pay less attention to the views of macroeconomists in general. Due to the limited nature of academic macro research their views on many issues are in practice based on mental models that have not been critiqued or even properly articulated, and there is no reason to suspect their mental models are better than anyone else’s. To the extent anyone outside the field can influence macro research I think we should encourage macro to explore models that use a wider variety of simplifying assumptions.
If macroeconomists want their field to be more relevant in encouraging them to learn about different classes of models. I would suggest learning about the Godley-Lavoie SFC models to start with, or even that they just take some time and think about the Kalecki profit equation. Failing that actually articulating the mental models they use to make policy recommendations would allow those models to be properly critiqued, and would greatly improve the prevailing macro discourse. Doing so would even enable the most common macro models to be better understood, as whether or not particular results depend on optimizing behaviour could be better recognized.
Models outside the mainstream will undoubtedly be less sophisticated, since 30 years of macro research have been spent exclusively developing a certain class of models. However, that does not mean that they will give less correct results. Even if you think consumers do optimise consumption over their lifetimes, in some cases abstracting away from that can give a better description of how optimizers would behave in the real world than explicitly modelling agents as optimising while leaving out other important real world effects.