The theory of rational inattention supports the development of trading strategies by providing a model of how market participants manage the scarcity of attention. In general, people cannot continuously process and act upon all information, but they can set priorities and choose the mistakes they are willing to accept. Rational inattention explains why agents pay disproportionate attention to popular variables, simplify the world into a small set of indicators, pay more attention in times of uncertainty, and limit their range of actions. In macroeconomics, rational inattention elucidates why forecasters underreact to shocks and why pure nominal variables, such as money and interest rates have persistent real effects. In finance, rational inattention explains why markets ignore a wide range of relevant data, leave pockets of information advantage, exaggerate price volatility, and propagate financial contagion.
Maćkowiak, Bartosz, Filip Matějka, and Mirko Wiederholt (2021), “Rational inattention: a review”.
The below post is based mainly on quotes from the above paper, with some added selected notes and emphasis.
This post ties in with this site’s summary of macro information inefficiency in financial markets.
The theory of rational inattention
“Every decision situation comes with a choice of attention. Agents face the fundamental trade-off between processing more information to improve decisions and saving on the mental effort of doing so. Humans cannot process all available information, yet they can choose how to deal with this cognitive limitation. The theory of rational inattention…formalizes this idea.”
“Economics is about adjustments to scarcity. Rational inattention studies adjustments to scarcity of attention. Understanding how people select, summarize, and digest the abundant available information is key to understanding many phenomena in economics…More forms of information than ever before are available due to new technologies, and yet we are able to digest little of it. Which form of imperfect information we possess and act upon is thus largely not determined by which information is given to us, but by which information we choose to attend to.”
“The theory of rational inattention provides a model of how cognitively limited people simplify and summarize available information. It describes behavior that seems error-prone, yet the form of mistakes in final actions is subject to agents’ choice; it is driven by agents’ preferences and the stochastic properties of the environment. Rational inattention is motivated by the observation that people often cannot avoid mistakes due to lack of information, but they can choose what to think about, what to pay attention to, and to what level of detail, i.e., what type of mistakes to minimize.”
“A large class of models describes information acquisition, i.e. agents’ choices of costly signals. Rational inattention belongs to this family of models, but its distinctive feature is that agents can choose to acquire signals of any form. Any information is available, yet costly to process.”
Why does rational inattention matter for trading strategies?
“Many insights from rational inattention have implications across fields. For instance, rational inattention makes it precise why and in what circumstances private agents and policymakers under-prepare for a rare event, be it a global financial crisis or a deadly pandemic, which leads to lessons for how the society can prepare for and respond to a rare event.”
Put simply, rational inattention theory helps to derive good hypotheses on what types of information financial markets may neglect, what types of events may shock the market, and under what circumstance attention is increasing or decreasing.
“The purest form of rational inattention treats prices just as any other piece of available information. Agents can choose to look at them in more or less detail, or not at all. If market prices move continuously, we do not follow them perfectly every millisecond, even if the numbers are right in front of us on a computer screen.”
This supports the use of automated systems for the evaluation of price and economic data and related trading actions. Even if underlying models are simplistic and inflexible relative to human reasoning, they can still create value through sheer consistency and speed of response.
Basics of the rational inattention model
“Christopher Sims put forth two main cornerstones of rational inattention as a model of processing available information:
- The idea of selective and costly attention: Available information is not internalized information. In principle, we can have the whole Internet at our disposal, yet we choose to process only a very limited amount of this information; we choose what questions we ask our friends, or what to read about in the news.
- A convenient modelling framework: A combination of the flexible choice of information with a specific form of an entropy-based cost function. Sims formulates a dynamic model where a single agent chooses how much information to process about different Gaussian shocks [and] emphasizes that in practice it is not only the amount of information that agents choose but also the nature of information, both of which can be modelled.”
“The agent faces a two-stage decision problem:
- What to pay attention to: The agent selects an information strategy to refine her belief about the state…
- What action y to take: This is a standard choice under uncertainty with the beliefs generated in the first stage via Bayesian updating. The objective is to maximize [expected utility] less the cost of information, which is a function of the information strategy.”
“The timing is as follows:
- The agent chooses the information strategy to maximize the expectation of utility less the cost of information while considering the action strategy she applies later.
- The agent receives a signal, the cost of information is incurred, and her posterior is formed.
- The agent chooses an action…to maximize the expectation of [utility] given the posterior.”
“Rational inattention models best describe repetitive decisions with a great deal of available information. In such decision situations not only can agents choose any pieces of information that they wish, but they are also more likely to use the optimal information strategy. On the other hand, decision situations that the model fits less clearly are new and quick one-time decisions, because they probably feature violations of [the assumption that] agents choose information optimally.”
General findings of rational inattention theory
“Solutions to rational inattention problems thus take a convenient analytical form and yet can exhibit rich behavioral properties…The implied features of the behaviors of agents are the following.
- Stochastic choice: Agents make random mistakes. The mapping from the state [of the world] to the action is not deterministic.
- Stakes and lower cost of information increase responsiveness: Scaling up the utility function or down the cost of information implies that action becomes more responsive.
- Magnified relative elasticities: Agents pay attention to important variables, which in relative terms makes them even more important than under perfect information.
- More attention to more volatile variables: Agents pay more attention when prior uncertainty is larger.
- Categorization, discreteness, and consideration sets: Agents most often find it optimal to contemplate a low number of actions only…If the agent chooses to focus on some actions only, she does not waste information capacity on small movements and is thus less likely to make larger errors. For a sufficiently high cost of information or a sufficiently strong prior, the agent may even consider only a single action.
- Multi-dimensional simplification – indexation: If agents need to pay attention to several shocks and choose multiple actions, then rational inattention models what kind of simplified representation of this high-dimensional environment the agents use. [For example] an investor pays attention to a particular linear combination of the asset prices only, i.e., to an….index, and then purchases or sells the whole portfolio given by the index.”
How rational inattention affects macro and finance
“[Christopher Sims] concluded that multiple sources of slow adjustment were necessary for the model to match the inertia in macroeconomic data…The inertia in the data could instead be understood as the result of a single new source of slow adjustment, rational inattention. Sims’s hypothesis has defined a research agenda.”
“The average forecast across agents of various macroeconomic variables underreacts to shocks to the economy. If a shock raises inflation for some time, the average inflation forecast of agents increases by less than actual future inflation. Relatedly, the ex-post average forecast error is predictable with the ex-ante average forecast revision. If inflation is rising and forecasts are being revised up, the subsequent average forecast error tends to be positive. Rational inattention implies exactly the systematic deviations from full information rational expectations found in the data.”
“Price setters observe nominal aggregate demand with idiosyncratic noise [with] the noise resulting from limited attention… [As a result,] nominal shocks have strong and persistent real effects. The idea is that macroeconomic data is publicly available with little delay, but most agents presumably have little incentive to track it carefully; as a result, prices respond slowly to nominal shocks (interest rate shocks or money supply shocks) and nominal shocks have real effects.”
“Firms set prices subject to rational inattention… [There is] evidence that firms use coarse pricing policies that are updated infrequently and consist of a small menu of prices… There is a fixed cost of reviewing a pricing policy and the choice of a price within a pricing policy is made under rational inattention.”
“Households make consumption decisions subject to rational inattention. In equilibrium, households pay little attention to the real interest rate, because fluctuations in the real interest rate are modest and small deviations from the consumption.”
“In today’s globalized financial markets, investors can choose among a wide array of assets and a large amount of information relevant for a portfolio decision arrives continuously. Rational inattention makes the plausible assumption that investors cannot keep track of all information necessary for an optimal portfolio choice. Rationally inattentive investors develop strategies for processing information that leave them systematically oblivious to some data.”
“Rational inattention suggests why information advantages can persist even though the information is publicly available… [For example] investors endowed with a small home information advantage specialize in home information and tilt their portfolios to home assets (“home bias”).”
“When investors are rationally inattentive, asset prices can co-move excessively, relative to the covariance of their fundamentals… Investors choose to track a linear combination of asset payoffs rather than individual asset payoffs [for example the performance of an economic sector], which magnifies the co-movement of asset prices because good news about one asset is partly attributed to the other asset.”
“[In] a model of financial contagion and test two key predictions of the model. Investors hold shares in two markets with uncorrelated fundamentals. If returns in market A become more volatile, investors divert attention to it from market B. Uncertainty rises and prices drop in market B. Financial contagion arises due to attention reallocation.”
“[Professional investment] managers…attend to macroeconomic information more carefully in recessions than in booms. The elevated price of risk in recessions magnifies this effect. Managers choose their portfolios accordingly, focusing on ‘market timing’ in recessions and on ‘stock picking’ in booms. By allocating attention rationally, managers outperform “unskilled investors” more in recessions than in booms.”