A well-known phenomenon with Ph.D. students is that they often push so hard on writing something new, that they forget the basics. A similar thing has happened to us during our research. Each of us focuses on the tiny bump we plan to create on the surface of the existing knowledge, but in the meanwhile, some of us neglected to study the core of our field. In our case, a bunch of Ph.D. students and young scholars interested in the philosophy of economics, the neglected core is the philosophical examination of choice theory, and especially the philosophically problematic notions of ‘choice’ and ‘preference.’
Choices and preferences are the building blocks of modern economic research. They are fundamental elements in some of the principal questions regarding the meaning of rationality, the essence of the utility function, the epistemic scopes and limits of economists’ work, and more. In other words, understanding what are choices and preferences – their ontic and epistemic status and how economists use them – is vital for extending our understanding of current economic research. Being the first post on the subject in this blog (and the first in the blog!), it will attempt to articulate a primary difference in the definition of choice and preference between economists and philosophers of economics. The posts to follow will use this distinction to present questions, ideas and thoughts that came up during our discussions in the Khavruta.
The genealogical origin of the modern use of “preferences” and “choices” is Paul Samuelson’s seminal 1938 paper “A note on the Pure Theory of Consumer’s Behaviour.” Although some work in this line was done before him (for example in his famous “An essay on the nature and significance of Economic Science” Lionel Robbins denounced hedonic psychology of preferences), Samuelson paved the way for economics to depart from a utilitarian concept of utility, understood in the Millian sense of pursuing happiness and avoiding pain*. Why was that so important for Samuelson? Because the hedonic notion of utility is formulated by mental states, meaning that it is a subjective, non-empirical and normative formulation. In other words, under a positivist-behaviorist project for rigorous economic science, a utilitarian view of choice theory cannot be justified.
According to Samuelson’s suggestion, ordinal utility and the utility function can be built on the agent’s consumption choices as evident from her revealed preferences. In other words, empirical data on the choices the agent makes when faced with alternatives is equivalent to her preferences, and these preferences can be formally translated to form her utility function (actually, Samuelson wanted to eliminate any psychological terms like utility or preference, but that did not work and the discipline continues to use them).
Allegedly, the above paragraph could have been written without the use of ‘preferences,’ meaning that the agent’s choices can be formally phrased as her utility function without the need to refer to any notion of inner preferences. However, current choice theory tacitly assumes that the agent’s revealed preferences reflect her subjective context-independent preference**. Why the theory needs these fixed preferences? Well, one explanation is that from the revealed preference of choosing to consume α over β we can deduce that the agent would choose α over β whatever the case and background of a future choice. This assumption is the one that allows the extrapolation of the utility function that was built according to data collected in time t1 to the agent’s choices in t2, and so building the expected utility – and the change in well-being – in cases of policy change.
In other words, despite the attempt to eliminate non-empirical psychological components from economic science, mental states were just swiped under the carpet. Mental states may not be used directly, or openly, to express the agent’s utility, but they are not abolished. They still roam under the rhetoric surface as tacit assumptions.
The immediate text to look in for interesting philosophical questions is Daniel Hausman’s 2012 “Preference, Value, Choice and Welfare.” A particularly interesting piece in this work is Hausman’s definition of choice as the outcome of both preference and belief (under certain factual constraints of feasibility). According to Hausman, a choice is the product of the combination of preferences – set in an ordinal form between different alternatives – and beliefs on the compatibility of the alternatives to the preference. His account set to describe what economists do, even though they often insist that they leave aside whatever lays deeper than the revealed preference. Therefore, even though economists insist on staying in the zone of the observable, they cannot do that as long as they want to be able to calculate the expected utility from a particular shift in the agent’s consumption.
In a nutshell, for an economist to be able to predict expected utility, she needs to have some explanatory account on how choices are made. Given that a belief is part of the explanation why in a specific context α was chosen over β, then dismissing belief prevents a justified extrapolation of current utility to the expected one: if you cannot give a full explanation for a choice in scenario A, how can you predict what will be the agent’s choice in scenario B? This is one reason why Hausman argues for the importance of embedding directly unobservable mental states such as belief into economics’ account on revealed preference (for the broader and detailed argument see chapters 4 & 5 in his book, or wait to the next posts that will dive deeper into Hausman’s definition of preferences and his critiques).
However, it is not certain that Hausman gave a suitable account of the practice of economists. In a relatively concise summary of the history of choice theory, DW Hands’ “Foundations of Contemporary Revealed Preference Theory” argues that the current choice theory denies the need (and the possibility) to formulate an explanation for consumer behavior. This is the “causal utility fallacy”, in which Hausman intuitively assumes that choice is causally explained by preferences. However, according to revealed preference theory, this is a fallacy because it is more accurate to say that the explanation for the agent’s preference of α over β is the observation that the agent chose to consume α instead of β. Therefore, there is no need for preferences at all, and the philosophical problems with choice theory might be laying somewhere else.
These points on the definition of preference are nothing but an appetizer that for us became a gate to a vast field of good questions about contemporary choice theory and its reciprocal relations with behavioral economics. The significant bulk of these questions remain, at least for now, unresolved and we are still discussing them. Here are some, for example:
- What are the primitives of the current revealed preference theory? What are the epistemic and ontic foundations of the theory? Why should economists care about it?
- Can a utility function based on the agent’s preferences in time t1 can be extracted for explaining/predicting consumption in t2? A different question of the same sort is whether extrapolation of the utility function to a policy welfare statement (e.g., the agent is better/worse-off after policy Z is applied) is justified?
- Despite its positivistic ambitions, does current choice theory is in some sense normative or prescriptional?
- How, if at all, has the research in behavioral economics changed the answers to the above questions?
We will probably not arrive at a unanimous answer, but then again this blog is about sharing our travels in the philosophy of economics, and traveling is about the ride, not just the final destination.
* More on the history of the concept of utility can be found in this highly appraised book (others did, we didn’t read it thoroughly enough yet) by Ivan Moscati “Measuring Utility: From the Marginal Revolution to Behavioral Economics”.
** as was explained by Amartya Sen, choices are assumed to coincide with preferences, and that in itself is not trivial at all: see “Behaviour and the Concept of Preference.”