A Comment on: “Presidential Address: Economics and Measurement: New Measures to Model Decision Making” by Ingvild Almås, Orazio Attanasio, and Pamela Jervis

Abstract

ALMÅS, ATTANASIO, AND JERVIS (this issue) make the case for economists going beyond standard choice data in a highly model-based manner due to inherent limits of traditional forms of behavioral and administrative data. Their central claim is that "Measurement is not a substitute for rigorous theory, it is an important complement to it, and should be developed in parallel to it". They trace the history of attempts to convince the profession of the proposition that standard administrative and survey data on choices and outcomes alone is far too restrictive to identify modern economic models, a proposition with which I heartily agree (Caplin ( 2024)). They provide several broad categories related to the decision-making structure of households, separation of preferences and beliefs, and impacts of the broader environment. They also make note of the importance going forward of additional research aimed at understanding income-relevant skills, both cognitive and non-cognitive. In this supporting note, I follow up on measurement of beliefs and skills. I reiterate their main points: that little can be known without model-based data innovation; and the high potential to know more with particular forms of such innovation. These measurement innovations are very different in the two cases. For beliefs, the key is to introduce appropriate new survey architectures linked with administrative data on standard outcomes and choices. For skills, an additional challenge is to measure them in a precise model-based manner. The first road is increasingly well-traveled, while the second is in a more nascent phase. SURVEY MEASUREMENT OF SUBJECTIVE BELIEFS ABOUT FUTURE EARNINGSEarnings from work are of fundamental economic importance. They are central drivers of consumption, savings, wealth, inequality, etc. In recent years, researchers have made massive progress in characterizing patterns of earnings over the life cycle in administrative data. For example, Guvenen, Karahan, Ozkan, and Song (2021) used rich administrative data to characterize the distribution of earnings growth in the United States. In the course of doing this, they provided information on just how uncertain earnings appear to be. They documented that higher-order moments, skewness and kurtosis, in addition to mean and variance, are important for describing the distribution of earnings growth in the population. They then showed how these moments vary with age and the level of earnings to help characterize labor market risks that workers face.Insightful as it is to identify patterns in administrative data, there are important reasons to dig deeper. In particular, objective data on outcomes is not the same as subjective data on beliefs. It is the latter that drive key decisions, such as whether or not to go to college, and if so, what subjects to study, what skills to hone, what type of work to engage in, where to apply for jobs, as well as such day-to-day decisions as whether or not to search