## Abstract

Using a representative online panel from the United States, we examine how individuals' macroeconomic expectations causally affect their personal economic prospects and their behavior. To exogenously vary respondents' expectations, we provide them with different professional forecasts about the likelihood of a recession. Respondents update their macroeconomic outlook in response to the forecasts, extrapolate to expectations about their personal economic circumstances, and adjust their consumption plans and stock purchases. Extrapolation to expectations about personal unemployment is driven by individuals with higher exposure to macroeconomic risk, consistent with macroeconomic models of imperfect information in which people are inattentive but understand how the economy works.

## I. Introduction

HOUSEHOLDS' expectations about their future income affect their consumption and financial behavior and should be shaped by perceptions of both idiosyncratic and aggregate risk. Policymakers attach an important role to the macroeconomic outlook of households, and low consumer confidence about the aggregate economy is central to many accounts of the slow recovery of consumption after the Great Recession. However, aggregate risk accounts for only a small fraction of the total income risk that households face. Macroeconomic models of imperfect information therefore predict that households are typically uninformed about news that is relevant for the macroeconomic outlook (Maćkowiak & Wiederholt, 2015; Reis, 2006; Sims, 2003). This raises two questions. First, are relevant pieces of news about the macroeconomy, such as professional forecasts about economic growth, part of households' information sets? Second, do households adjust their expectations about their own economic situation and their behavior in response to changes in their expectations about aggregate economic growth?

In this paper, we use experimental methods to test for the causal effects of households' expectations about future macroeconomic conditions on their personal economic prospects and behavior.1 We propose a randomized information experiment embedded in an online survey on a sample that is representative of the portion of the US population that is employed full time. Our experiment proceeds as follows. First, we elicit our respondents' prior beliefs about the likelihood of a recession. We define a recession as a fall in US real GDP around three months after the time of the survey. Subsequently, we provide our respondents with one of two truthful professional forecasts about the likelihood of a recession taken from the microdata of the Survey of Professional Forecasters (SPF). Respondents in the “high recession treatment” receive information from a very pessimistic forecaster, and respondents in the “low recession treatment” receive a prediction from a very optimistic forecaster. Thereafter, we measure our respondents' expectations about the evolution of aggregate unemployment and their personal economic situation over the twelve months after the survey and elicit both their consumption plans and their posterior beliefs about the likelihood of a recession. We reinterview a subset of our respondents in a follow-up survey two weeks after the information provision.

Our experimental design allows us to study whether people adjust their personal job loss and earnings expectations and their economic behavior in response to changes in their macroeconomic outlook. Moreover, the setup enables us to shed light on different predictions of macroeconomic models of imperfect information, which parsimoniously explain many stylized facts in macroeconomics (Carroll et al., 2018; Maćkowiak & Wiederholt, 2015) and dramatically change policy predictions relative to standard models (Wiederholt, 2015). In such models, people are imperfectly informed about the state of the economy due to either infrequent updating of information sets (Carroll, 2003; Mankiw & Reis, 2006; Reis, 2006) or receiving noisy signals (Maćkowiak & Wiederholt, 2015; Sims, 2003; Woodford, 2003). For example, an adjustment of our respondents' beliefs in response to the information implies that they are imperfectly informed about the professional forecasts, even though those forecasts are relevant for their economic outlook.

We document several findings on people's recession expectations and their relationship to their personal economic outlook and behavior. First, we find that our respondents have much more pessimistic and dispersed prior beliefs about the likelihood of a recession compared with professional forecasters. Respondents update their beliefs about the likelihood of a recession in the direction of the forecasts, putting a weight of around one-third on the forecast. Among those with a college degree, learning rates are significantly higher for respondents who are less confident in their prior beliefs, in line with Bayesian updating. For those without a college degree, there is no such heterogeneity. The findings for highly educated respondents are in line with models of imperfect information in which people are initially inattentive but update rationally after receiving new information.

Second, we explore the degree of extrapolation from recession expectations to personal economic expectations. We find that a negative macroeconomic outlook has a negative causal effect on people's subjective financial prospects for their household and increases their perceived chance of becoming personally unemployed. A back-of-the-envelope calculation suggests that 11.3% of our respondents would need to become unemployed in case of a recession for their expectations to be accurate on average. This effect is large but still relatively close to the increase in the job loss rate by 7 percentage points during the previous recession. However, there is no significant average effect on people's expected earnings growth conditional on keeping their job. In the two-week follow-up survey, differences in expectations decrease in size, but mostly remain economically and statistically significant.

Third, we characterize heterogeneity in the effect of recession expectations on personal expectations. The negative effect on perceived job security is driven by individuals with a higher exposure to past recessions, such as people with lower education and lower earnings, as well as men. Individuals who are more strongly exposed to macroeconomic risk (e.g., those with previous unemployment experience, those living in counties with higher unemployment, or those working in more cyclical industries) more strongly update their expectations about personal unemployment. Similarly, we provide evidence of updating of earnings expectations conditional on working in the same job for groups that should not be constrained by downward rigidity in wages. Thus, the updating of personal expectations is data-consistent in terms of size and heterogeneity, indicating that our respondents have an understanding of their actual exposure to recessions. The assumption that people understand the true model of the economy is a key feature of imperfect information models.

Fourth, we provide evidence of adjustments in behavior in response to the information. We find that a more pessimistic macroeconomic outlook causes a significantly lower planned consumption growth, in line with recent evidence that recessions can entail shocks to permanent income (Krueger, Mitman, & Perri, 2016; Yagan, 2018). Furthermore, we document surprisingly large effects of our treatment on active adjustments in people's stockholdings between the main intervention and the follow-up survey, measured with self-reports.

Finally, we provide causal evidence on the relationship between people's expectations about economic growth and inflation.2 There was substantial disinflation during most past recessions (Coibion & Gorodnichenko, 2015b), and many macroeconomic models predict a comovement of inflation and unemployment in response to shocks. However, our fifth main finding is that exogenous changes in beliefs about the likelihood of a recession do not decrease people's inflation expectations.

We contribute to a growing literature that uses survey experiments to study the expectation formation process and the importance of information rigidities. This literature has mainly focused on expectations about inflation (Armantier et al., 2016, 2015; Binder & Rodrigue, 2018; Cavallo, Cruces, & Perez-Truglia, 2017; Coibion, Gorodnichenko, & Kumar, 2018) and house prices (Armona, Fuster, & Zafar, 2018; Fuster et al., 2018), documenting that consumers and firms update their expectations in response to the provision of publicly available information. Our paper is the first to exogenously shift households' expectations about future GDP growth to assess whether people extrapolate from expectations about aggregate conditions to their personal economic outlook and whether these expectations causally affect consumer and financial behavior. A key contribution of our paper is to document that updating of personal expectations in response to a revised macroeconomic outlook is driven by groups that are actually more strongly exposed to macroeconomic risk, suggesting that households have a basic understanding of their exposure to business cycle fluctuations.

A larger literature uses observational data to study how people's macroeconomic expectations are formed (Das, Kuhnen, & Nagel, 2020; Goldfayn-Frank & Wohlfart, 2018; Kuchler & Zafar, 2019; Malmendier & Nagel, 2011, 2016; Manski, 2017; Mian, Sufi, & Khoshkhou, 2018; Tortorice, 2012) and how these expectations shape household behavior, such as the effect of home price expectations on housing-related behavior (Bailey, Dávila et al., 2018; Bailey, Cao et al., 2018) or the effect of inflation expectations on consumption behavior (Bachmann, Berg, & Sims, 2015; D'Acunto, Hoang, & Weber, 2018). A literature in finance uses survey data to study the extent to which optimism and pessimism about stock returns and the macroeconomic outlook can explain households' investment behavior (Das et al., 2020; Greenwood & Shleifer, 2014; Malmendier & Nagel, 2011; Vissing-Jorgensen, 2004). In a different context, managerial decision making, Coibion, Gorodnichenko, and Ropele (2020) provide causal evidence showing that higher inflation expectations lead firms to raise their prices, increase their utilization of credit, and reduce their employment.

Our paper also contributes to a literature that uses observational data to study the importance of information rigidities in macroeconomics (Carroll, 2003; Coibion & Gorodnichenko, 2012, 2015a; Mankiw, Reis, & Wolfers, 2003).

The rest of the paper is structured as follows: Section II describes the design of the main experiment and provides details on the data collection. In section III, we present evidence on belief updating in response to the professional forecasts. Section IV presents the results on the causal effect of expectations about a recession on people's personal economic outlook, behavior, and other macroeconomic expectations. We provide various robustness checks in section V. Section VI concludes.

## II. Experimental Design

In this section we describe the survey administration, present our experimental design, and explain the structure of the main survey and the follow-up survey. The full experimental instructions for all experiments (including robustness experiments 1, 2, and 3) are available at https://goo.gl/1C9vLK. Figures A.1 and A.2 show detailed timelines of the experiment and the relevant reference periods for behavioral outcomes and expectations.

### A. Survey

We collect a sample of 1,124 respondents that is representative of the full-time employed US population in terms of gender, age, region, and total household income through the widely used market research company Research Now. We invite only people who have a paid job and work full time. The data were collected in summer 2017. We conducted the follow-up survey approximately two weeks after the main survey was administered and managed to recontact 737 respondents, which corresponds to a recontact rate of 65%.

### B. Baseline Experiment

#### Prior beliefs: Likelihood of a recession.

First, we ask subjects to complete a questionnaire on demographics, which includes questions on gender, age, income, education, and region of residence. Subsequently, we give our respondents a brief introduction on how to probabilistically express expectations about future outcomes, and we explain several relevant economic concepts, such as “recession” and “GDP.” Then we ask our respondents to estimate the likelihood that there will be a fall in US real GDP in the fourth quarter of 2017 compared to the third quarter of 2017. The survey was conducted in summer 2017, so this corresponds to a fall in real GDP three to six months after the survey.3 Thereafter, we ask our respondents how confident they are in their estimate.

#### Information treatment: Professional forecasters.

The Federal Reserve Bank of Philadelphia regularly collects and publishes predictions by professional forecasters about a range of macroeconomic variables in their Survey of Professional Forecasters (SPF) (Croushore, 1993). The SPF is conducted in the middle of each calendar quarter, and forecasters have to estimate the likelihood of a decline in real GDP in the quarter of the survey, as well as each of the four following quarters. The average probability assigned to a drop in GDP in the quarter after the survey has had high predictive power for actual recessions in the past. In our survey, we randomly assign our respondents to receive one of two forecasts taken from the microdata of the wave of the SPF conducted in the second quarter of 2017, the most recent wave of the SPF available at the time of our survey. To make the forecast more meaningful to respondents, we tell them that it comes from a financial services provider that regularly participates in a survey of professional forecasters conducted by the Federal Reserve Bank of Philadelphia.

In the high recession treatment, respondents receive a forecast from the most pessimistic panelist in the SPF, who assigns a 35% probability to a fall in US real GDP in the fourth quarter compared to the third quarter of 2017. In the low recession treatment, respondents receive information from one of the most optimistic forecasters, who expects a fall in US real GDP with a probability of 5%.4 In order to make the treatment more meaningful to our respondents, we provide them with a figure that contrasts their prior belief with the prediction from the professional forecaster (see figure A.3 for an illustration of the treatment screen).

#### Personal expectations, economic behavior, and macroeconomic expectations.

After the information provision, all respondents are asked to estimate the likelihood that the unemployment rate in the United States will increase over the twelve months after the survey, as well as a qualitative question on how they expect unemployment to change. This is followed by questions on personal economic expectations, other macroeconomic expectations, and their consumption plans. While we elicit most expectations probabilistically, we also include some qualitative questions with categorical answer options.5

We first ask our respondents whether they think that their family will be better or worse off twelve months after the survey. Subsequently, we elicit people's density forecast about their earnings growth conditional on working at the same place where they currently work. We ask our respondents to assign probabilities to ten brackets of earnings growth over the next twelve months, which are mutually exclusive and collectively exhaustive. Respondents could not continue to the next screen if their entered probabilities did not sum up to 100%. The elicitation of a subjective probability distribution allows us to measure both mean expected earnings growth and uncertainty about earnings growth.6 Thereafter, respondents estimate their subjective probability of job loss and their subjective probability of finding a new job within three months in case they lose their job over the next twelve months. In addition, we elicit density forecasts of inflation over the next twelve months using the same methodology as for earnings expectations.7

Subsequently, we ask our respondents some qualitative questions related to their consumption behavior. First, we ask them whether they think that it is a good time to buy major durable goods. Second, we ask them how they plan to adjust their consumption expenditures on food at home, food away from home, and leisure activities during the four weeks after the survey compared to the four weeks prior to the survey. Thereafter, our respondents answer a qualitative question on how they expect firm profits to change over the next twelve months, and they estimate the percent chance that unemployment in their county of residence will increase over the next twelve months. Finally, we reelicit beliefs about the likelihood of a fall in real US GDP in the fourth quarter of 2017 compared to the third quarter of 2017. At the end of the survey, our respondents complete a series of additional questions on the combined dollar value of their spending on food at home, food away from home, clothing, and leisure activities over the seven days before the survey, the industry in which they work, and their tenure at their employer, as well as a set of questions measuring their financial literacy (Lusardi & Mitchell, 2014). Moreover, we ask them a series of questions on their assets, their political affiliation, and their postal code of residence.

### C. Follow-Up Survey

We designed our main survey to minimize concerns about numerical anchoring and experimenter demand. First, instead of eliciting posterior beliefs about the likelihood of a recession immediately after the information provision, respondents answer a range of other questions and only report posteriors at the end of the survey, roughly ten minutes after the information. Second, we elicit both probabilistic and qualitative expectations to ensure the robustness of our findings to different question framing and numerical anchoring. While we believe that these design features already address some concerns regarding numerical anchoring, we further mitigate such concerns by conducting a two-week follow-up survey in which no additional information is provided. We chose to have a two-week gap between the main study and the follow-up to balance the trade-off between testing for persistence and maximizing the recontact rate in the follow-up.

In the follow-up survey, we reelicit some of the key outcomes from the main survey, such as the likelihood of increases in national- and county-level unemployment, expectations about firm profits, and personal economic expectations, such as subjective job security and earnings expectations. We reelicit our respondents' estimated likelihood of a fall in real GDP in the fourth quarter of 2017 compared to the third quarter of that year. Moreover, we collect data on our respondents' consumer and financial behavior in the time between the main intervention and the follow-up. First, we ask our respondents about their combined spending on food at home, food away from home, clothing, and leisure activities over the seven days before the follow-up.8 Second, we ask them whether they bought any major durable goods and whether they actively increased or reduced their stockholdings during the fourteen days prior to the follow-up. Finally, we elicit our respondents' beliefs their employers' exposure to aggregate risk and about the most likely causes of a potential recession, as well as their personal unemployment history.

### D. Discussion of the Experimental Design

In our experiment, we provide respondents with different individual professional forecasters' assessments of the likelihood of a recession. An alternative experimental design would provide the average professional forecast to respondents in the treatment group while giving no information to individuals in a pure control group. We believe that our design provides important advantages for studying the causal effect of recession expectations on personal economic expectations and behavior.

The variation in recession expectations in the alternative design would stem from differences between individuals whose beliefs have been shifted and those who still hold their prior beliefs. Thus, the alternative design identifies the causal effect of recession expectations on outcomes of individuals who hold unrealistic priors ex ante, as the treatment shifts beliefs only for this group. This could threaten the external validity of results obtained under the alternative design. By contrast, our design also generates variation in recession expectations among individuals with more realistic priors, and therefore it identifies average causal effects of recession expectations for a broader population. In addition, receiving a forecast may not only shift the level of individuals' beliefs but may also have side effects, such as reducing the uncertainty surrounding the level of their beliefs or priming respondents on recessions and professional forecasts. In our design, the only difference between the two treatment arms is the percent chance assigned to the event of a recession by the professional forecast our respondents receive, while side effects of receiving a forecast should be common across treatment arms.9

Not having a pure control group has two disadvantages. First, a pure control group would allow us to assess whether the questions and procedures of the experiment induce a change in subjects' beliefs about a potential upcoming recession. While this would give an indication of the external validity of our findings, we note that such changes in expectations should be common across treatment arms and do not threaten the internal validity of our results. Second, a pure control group would provide us with a potentially more meaningful benchmark to interpret the magnitudes of the experimentally estimated causal effects of subjects' recession expectations on their macroeconomic and personal expectations, as well as their behavior.

Under which conditions will our experimental design generate variation in respondents' recession expectations? As discussed in online appendix D.2, we require that respondents do not fully “debias” the signals and thereby perfectly learn about the average professional forecast and that they believe that the professional forecasts provide a relevant signal about the future state of the economy that is not yet fully incorporated into their information sets.

### E. Data

#### Representativeness.

Table A2 in the online appendix provides summary statistics for our sample. Around 80% of our respondents indicate that they are the main earner in their household. Moreover, table A3 displays the distributions of a range of individual characteristics among respondents in full-time employment in the 2015 American Community Survey (ACS) and in our data. Our sample matches the distributions of gender, age, region, and total household income precisely. In addition, the composition of our sample is quite close to the composition of the population in full-time employment along nontargeted dimensions, such as industry and hours worked. One caveat is that our sample has higher labor earnings and is more educated than the US population in full-time employment, similar to the New York Fed Survey of Consumer Expectations. We address this issue by conducting heterogeneity analyses according to education and demonstrating the robustness of our results to reweighting (section V).

#### Definition of variables.

First, we generate a variable measuring the perceived chance of becoming personally unemployed over the next twelve months as the product of people's perceived probability of losing their main job within the next twelve months and their perceived probability of not finding a new job within the following three months. For each respondent, we calculate the mean and standard deviation of expected inflation and expected earnings growth using the midpoints of the bins to which the respondent has assigned probabilities.10 Moreover, we create an index of people's planned change in nondurable consumption from the four weeks prior to the main survey to the four weeks after the survey, using their qualitative spending plans for food at home, food away from home, and leisure activities. Finally, we create a measure of people's actual changes in spending on food at home, food away from home, clothing, and leisure based on their self-reported spending during the seven days before the main survey and the seven days before the follow-up survey.11 The questions on expected firm profits, the expected financial situation of the household, and the change in stockholdings between main survey and follow-up were elicited on 5- and 7-point scales. We code these variables such that higher values refer to “increase” or “improve” and lower values refer to “decrease” or “worsen.” These qualitative outcome variables are normalized using the mean and standard deviation separately for the main survey and the follow-up survey. For the quantitative measures, we do not normalize outcome variables, as they have a natural interpretation.

#### Integrity of the randomization.

Our sample is well balanced for a set of key characteristics and pretreatment beliefs about the likelihood of a recession (table A5). The means do not differ significantly across treatment arms for any of these variables, and we cannot reject the null hypothesis that the partial correlations of the variables with a dummy for being in the high recession treatment are jointly 0. Moreover, we observe no differential attrition in our main survey across treatment arms, and participation in the follow-up survey is not related to treatment status in the main experiment. The sample of individuals in the follow-up is balanced across the two treatment arms in terms of key covariates (table A6). There are marginally significantly more individuals with a college degree and more men in the low recession treatment arm in the follow-up sample, but we cannot reject the null hypothesis that the correlations of the covariates with the high recession dummy are jointly 0. To rule out any concerns, we include a set of control variables in all of our estimations.

#### Data quality.

We provide evidence that our expectations data on earnings and inflation are of high quality by comparing our data with a panel survey by the New York Fed that was launched as a predecessor to the Survey of Consumer Expectations (SCE) (Armantier et al., 2013). For example, for inflation expectations, 80% of our respondents assign positive probability to more than one bin (89.4% in the Fed survey) and the average number of bins with positive probability is 4.24 (3.83 in the Fed survey). Although a larger share of our respondents assign positive probability to noncontiguous bins (6.9% versus 0.9%), this still accounts for a very small fraction of our sample. Only 0.4%, 6.5%, and 0.3% of our respondents enter a prior probability of a fall in real GDP of 0%, 50%, and 100%, respectively, which may indicate mental overload (de Bruin et al., 2000; Manski, 2017).

## III. Updating of Recession Expectations

In this section, we provide descriptive evidence on our respondents' prior beliefs about the likelihood of a recession and study how respondents update these beliefs in response to information.

### A. Prior Beliefs

#### Stylized facts.

Respondents in our sample have a much more pessimistic macroeconomic outlook than experts do (figures 1 and A.4 and table A4). The median professional forecaster in the second quarter of 2017 reports a likelihood of a recession in the fourth quarter of 2017 of just 15%. By contrast, our median respondent assigns a probability of 40%, as pessimistic as professional forecasters were for the last time in the second quarter of 2009. While there is a large dispersion in beliefs about the likelihood of a recession among consumers, the dispersion of beliefs is much smaller in the sample of professional forecasters, ranging from four professional forecasters estimating a 5% chance of a recession to one forecaster assigning a 35% chance.
Figure 1.

Prior and Posterior Beliefs about the Likelihood of a Recession

This figure displays the distributions of prior and posterior beliefs in the two treatment arms.

Figure 1.

Prior and Posterior Beliefs about the Likelihood of a Recession

This figure displays the distributions of prior and posterior beliefs in the two treatment arms.

We confirm these patterns using robustness experiment 1 (described in more detail in table A1), which was conducted with an online convenience sample from the online labor market Amazon Mechanical Turk (MTurk), which is widely used in experimental economics research (Cavallo et al., 2017; D'Acunto, 2018; Kuziemko et al., 2015; Roth, Settele, & Wohlfart, 2019). We discuss the advantages and disadvantages of MTurk samples in appendix section C.1. The median professional forecaster in the second quarter of 2018 reports a likelihood of a recession in the fourth quarter of 2018 of 10%, while the median respondent in our MTurk sample assigns a probability of 45% (figure A.8). The distribution of recession expectations in the MTurk sample is remarkably robust to incentivizing the consumers' forecast using a quadratic scoring rule (see figure A.9).12 A Kolmogorov-Smirnov test confirms that the distributions of incentivized and unincentivized beliefs are not statistically distinguishable ($p=0.319$).

The finding of greater pessimism and a higher dispersion of beliefs among consumers than among professional forecasters is in line with previous findings on inflation expectations (Armantier et al., 2013; Mankiw et al., 2003) and with qualitative expectations on aggregate economic conditions over a longer time period from the Michigan Survey of Consumers (Das et al., 2020).13

#### Correlates of recession expectations.

Neither education nor age is related to people's recession expectations, but women have a significantly more pessimistic macroeconomic outlook than men do (table A7). Interestingly, Democrats are much more pessimistic compared to Independents, while Republicans are much more optimistic, consistent with evidence on partisan bias in economic expectations (Mian et al., 2018). People who have been personally unemployed in the past are significantly more pessimistic about aggregate economic conditions, in line with Kuchler and Zafar (2019), who find that individuals who lose their jobs become significantly less optimistic about the aggregate economy. Taken together, it is reassuring that the correlations between covariates and recession expectations are in line with previous literature.

### B. Updating of Recession Expectations

Do our respondents update their recession expectations upon receiving the professional forecasts? Figure 1 shows our first main result:

Result 1.

The information provision strongly shifts expectations toward the professional forecast in both treatment arms, and cross-sectional disagreement within the treatment arms declines. This implies that the respondents were initially not fully informed about the forecasts and that the forecasts are relevant to the respondents' economic outlook.

Figure A.5 displays scatter plots of prior and posterior beliefs. Observations along the horizontal lines indicate full updating of beliefs toward the professional forecast, while respondents along the 45 degree line do not update at all. We observe more updating of beliefs among respondents in the low recession treatment, where the average absolute distance of prior beliefs to the signal of 5% is greater than in the high recession treatment, which provides a forecast of 35%. In the low recession treatment, 11.5% of respondents and in the high recession treatment, 19.5 percent of respondents do not update their beliefs at all, while 68.6% (47.8%) of respondents either fully or partially update their beliefs toward the signal (see table A8). The remaining respondents either “overextrapolate” from the signal or update in the opposite direction. However, some of these observed changes in beliefs could be caused by typos or by respondents changing their beliefs because taking a survey on macroeconomic topics makes them think more carefully about the question. Finally, the cross-sectional disagreement in posterior beliefs as measured through the interquartile range and standard deviation declines within both treatment arms compared to prior beliefs (table A4).

#### Magnitudes.

We quantify the degree of updating of recession expectations by estimating a Bayesian learning rule that we derive in online appendix D.1. We define $updatingi$ as the difference in people's posterior and prior expectations, and the “shock” as the difference between the professional forecast and the prior belief: (35 $-$$priori$) for people in the high recession treatment and (5 $-$$priori$) for people in the low recession treatment. We assume that people's prior beliefs about the probability of a recession follow a beta distribution and that the loss function is quadratic. Under these assumptions, people should follow a linear learning rule, $updatingi=α1shocki$, where $α1$ lies in the interval [0,1] and depends negatively on the strength of the respondent's prior belief.

The individual-level shock depends on the respondent's prior, which introduces two problems. First, the prior is measured with error, thereby leading to attenuation bias in the estimated learning rate $α1$. Second, self-reported expectations could differ between the prior and the posterior for reasons that are unrelated to the treatment but potentially correlated with the prior. Most important, people who hold higher priors and are subject to a more negative shock should mechanically display more negative changes in their expectations since the probability of a recession is bounded to be in the interval [0,100], leading to an upward bias in the estimate of $α1$. Controlling linearly for people's prior belief removes attenuation bias and mechanical correlations between people's updating and the shock, while not changing the interpretation of the estimated coefficient $α1$ as the learning rate. Moreover, we include a vector of additional control variables $Xi$, which increases our power to precisely estimate treatment effects and allows us to control for the slight imbalance in the follow-up sample.14 Specifically, we estimate the following equation using OLS:
$updatingi=α0+α1shocki+α2priori+ΠTXi+ɛi,$
(1)
where $ɛi$ is an idiosyncratic error. We report robust standard errors throughout the paper.

We estimate a highly significant learning rate equal to about one-third of the shock to individual beliefs (see table 1). Our estimated learning rate from professional forecasts is in the range of estimates in related literature (Armantier et al., 2016; Coibion, Gorodnichenko, & Kumar, 2018; Fuster et al., 2018). Thus, our information treatment generates a difference of about 10 percentage points in people's average posterior beliefs across treatment arms. The size and significance of the estimated learning rate imply that the respondents found that the forecasts contain some relevant information that was not already incorporated into their priors. Online appendix D.2 provides a more detailed discussion of the estimated learning rate in the context of different corner cases and estimates in related literature.

Table 1.
Main Results: Learning Rates
 Updating (Main Survey) Updating (Follow-Up) (1) (2) (3) (4) (5) (6) (7) Shock 0.318*** 0.417*** 0.436*** 0.358*** 0.128** 0.263*** 0.371*** (0.034) (0.047) (0.076) (0.041) (0.050) (0.073) (0.117) Shock $×$ Confident −0.187*** −0.273*** (0.068) (0.101) Confident 0.752 0.453 (2.258) (3.658) Shock $×$ Follow news −0.150* −0.293** (0.085) (0.130) Follow news 1.798 2.176 (2.740) (4.178) Prior −0.247*** −0.242*** −0.198** −0.205*** −0.640*** −0.549*** −0.450*** (0.038) (0.057) (0.081) (0.048) (0.059) (0.093) (0.125) Observations 1,124 1,124 1,124 736 736 736 736 Sample Baseline Baseline Baseline Basel. (compl. Follow-up Follow-up Follow-up follow-up)
 Updating (Main Survey) Updating (Follow-Up) (1) (2) (3) (4) (5) (6) (7) Shock 0.318*** 0.417*** 0.436*** 0.358*** 0.128** 0.263*** 0.371*** (0.034) (0.047) (0.076) (0.041) (0.050) (0.073) (0.117) Shock $×$ Confident −0.187*** −0.273*** (0.068) (0.101) Confident 0.752 0.453 (2.258) (3.658) Shock $×$ Follow news −0.150* −0.293** (0.085) (0.130) Follow news 1.798 2.176 (2.740) (4.178) Prior −0.247*** −0.242*** −0.198** −0.205*** −0.640*** −0.549*** −0.450*** (0.038) (0.057) (0.081) (0.048) (0.059) (0.093) (0.125) Observations 1,124 1,124 1,124 736 736 736 736 Sample Baseline Baseline Baseline Basel. (compl. Follow-up Follow-up Follow-up follow-up)

The table shows OLS estimates of the learning rate from the professional forecasts based on specification 1. All specifications control for the respondent's prior belief, age, age squared, a dummy for females, log income, a dummy for respondents with at least a bachelor's degree, dummies for the respondent's Census region of residence, a measure of the respondent's financial literacy, as well as a dummy for Republicans and a dummy for Democrats. Specifications 2, 3, 6, and 7 also control for interactions of the prior with the dimension of heterogeneity. The outcome in columns 1 to 4 is the difference between the posterior belief measured in the main study and the prior belief. The outcome in columns 5 to 7 is the difference between the posterior measured in the follow-up study and the prior belief. “Confident” takes value 1 for respondents saying that they are “very sure” or “sure” about their estimate of the likelihood of a recession. “Follow news” takes value 0 if respondents somewhat or strongly disagree with the statement, “I usually follow news on the national economy” and value 1 otherwise. Robust standard errors are in parentheses. Significant at $*$10%, $**$5%, and $***$1%.

#### Are changes in expectations consistent with Bayesian updating?

First, Bayesian updating predicts that respondents should adjust their expectations partially or fully toward new signals that they find informative—learning rates should lie in the interval [0,1]. Our estimated learning rate of one-third is in line with this prediction. Second, Bayesian updating implies that respondents who are less confident in their prior belief should react more strongly to new signals. We examine this prediction by constructing a dummy indicating whether the respondent is at least “sure” about his or her prior estimate. Consistent with Bayesian updating, the estimated learning rate is significantly lower for respondents who are more confident in their prior belief (table 1, column 2). Moreover, respondents who report that they usually do not follow news on the national economy place significantly higher weight on the signal (column 3), consistent with the idea that information acquisition prior to the experiment increases the strength of people's prior belief.15 In robustness experiment 3 described later, we also find support of two more predictions of Bayesian updating: (a) receiving a forecast makes respondents more confident in their beliefs and (b) changes in confidence are positively related to the individual-level learning rate (table A19).

#### Heterogeneous updating across demographic groups.

Individuals with lower education update more strongly from the forecasts, while there are no significant differences according to income, gender, industry, personal unemployment experiences, the unemployment rate in the county of residence, and financial literacy (table A10). Heterogeneity in learning rates could stem from differences in trust toward experts, differential ex ante informedness about the professional forecasts across groups,16 or different learning rules.

One way in which learning rules could differ across individuals is that less sophisticated individuals could find it more difficult to rationally learn from the information. As shown in table A11, the heterogeneity in learning rates according to confidence in the prior is fully driven by individuals with a college degree, while those without a college degree weigh the new information independent of their confidence in their priors. The coefficients on the interaction terms between the shock and confidence in the prior are significantly different between the two groups ($p<0.01$). Thus, while learning from information is consistent with Bayesian updating for more sophisticated individuals, less sophisticated individuals seem to follow simpler learning rules. Similarly, heterogeneity in learning rates by news consumption is fully driven by highly educated respondents.

#### Do changes in recession expectations persist?

Following Cavallo et al. (2017), we employ a two-week follow-up survey in which no treatment information is administered. The medium-run learning rate (calculated using the follow-up) amounts to about 40% of the short-run learning rate (table 1, column 5), in line with respondents receiving new relevant signals about the macroeconomy between the two surveys or imperfect memory (see also figures 1 and A.6). Moreover, learning rates still differ significantly by confidence in the prior and news consumption prior to the main survey.

#### Implications for macroeconomic models.

Our results presented in this section have several implications for macroeconomic models. The finding that respondents use the professional forecasts to persistently update their beliefs implies that (a) the professional forecasts were not fully incorporated into our respondents' information sets before the survey and (b) our respondents consider the information relevant for their expectations about the future. This finding suggests that there exist costs of information acquisition, as in microfounded sticky information models (Reis, 2006).17 Our experiment sets these costs to 0 for a particular piece of news about aggregate economic growth.

Conditional on acquiring information, we observe heterogeneity in learning rates across groups. This is in line with the idea that our respondents perceive the piece of information with individual-specific noise, as in models of noisy information (Maćkowiak & Wiederholt, 2015; Sims, 2003). In addition, we also observe heterogeneity in learning rules. On the one hand, highly educated respondents put lower weight on the information when they hold stronger priors, in line with the predictions of Bayesian updating. Rational learning from new information is a key feature of both sticky and noisy information models, so these models may be able to proxy the expectation formation of more sophisticated individuals in a reasonable manner. On the other hand, less highly educated respondents' learning from the information does not seem to be well captured by Bayesian updating, highlighting a role for cognitive limitations and heterogeneity in belief formation mechanisms in macroeconomic models. These findings are consistent with recent evidence that individuals with cognitive limitations display larger biases in their expectation formation (D'Acunto et al., 2019a, 2019b, 2019c). Our findings are inconsistent with more traditional models of full-information rational expectations (Muth, 1961) or models with no heterogeneity.18

Finally, in line with the model and time series evidence in Carroll (2003), our findings imply that households exhibit some trust toward experts in the context of expectations about general economic conditions.

## IV. The Causal Effect of Recession Expectations

### A. Empirical Specification

We have established that our respondents durably update their beliefs about the likelihood of a recession in response to professional forecasts. This provides us with a first stage to examine the causal effect of recession expectations on expectations about personal economic outcomes. Specifically, we examine whether people's subjective economic model, as measured through the size and heterogeneity of extrapolation to expectations about their personal situation, is in line with empirical facts. As a first step, we examine how these expectations, $expi$, are correlated with our respondents' posterior beliefs about the likelihood of a recession, $posteriori$,
$expi=β0+β1posteriori+ΠTXi+ɛi,$
(2)
where $Xi$ is a vector of the same control variables that we included in our previous estimations. The OLS estimate of $β1$ cannot be given a causal interpretation. For example, it is possible that people who are generally more optimistic or pessimistic respond differently to both the question on the posterior as well as the questions related to the evolution of other economic outcomes. It is also conceivable that the direction of causality runs from the personal situation to macroeconomic expectations, as suggested by recent evidence in Kuchler and Zafar (2019). Finally, the estimate of $β1$ could be biased toward 0 because of measurement error in the posterior belief. To deal with omitted variable bias, reverse causality and measurement error, we instrument our respondents' posterior beliefs with the random assignment to the different professional forecasts, where $highrecessioni$ is an indicator taking value 1 for individuals who received the pessimistic professional forecast and value 0 for respondents receiving the optimistic forecast. Specifically, we use two-stage least squares and estimate the following equation,
$expi=β0+β1posteriori^+ΠTXi+ɛi,$
(3)
where
$posteriori^=α0^+α1^highrecessioni+Θ^TXi.$

We have a strong and highly significant first stage on people's posttreatment beliefs about the likelihood of a recession based on the random assignment of the different professional forecasts ($F$-statistic $=$ 75.16; see table 2).

Table 2.
Main Results: Macroeconomic and Personal Economic Expectations
 National Unemployment (percent) National Unemployment (categorical) County Unemployment (percent) Household Financial Prospects Earnings Growth: Mean Earnings Growth: Uncertainty Personal Unemployment (percent) Inflation: Mean Inflation: Uncertainty Firm Profits (categorical) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) OLS Posterior: 0.528*** 0.014*** 0.508*** −0.011*** −0.010** 0.010*** 0.112*** 0.013*** 0.013*** −0.009*** Recession (0.033) (0.001) (0.035) (0.002) (0.005) (0.004) (0.018) (0.004) (0.004) (0.002) IV Posterior: 0.895*** 0.030*** 0.536*** −0.012** −0.013 0.002 0.113* 0.014 0.006 −0.013** Recession (0.131) (0.006) (0.118) (0.006) (0.020) (0.013) (0.066) (0.018) (0.015) (0.005) Observations 1,124 1,124 1,124 1,124 1,118 1,118 1,123 1,121 1,121 1,124 Mean dependent 32.09 0.01 29.55 −0.01 2.64 1.79 6.61 2.60 2.74 −0.01 variable SD dependent 24.18 1.00 23.20 1.00 3.42 2.41 11.47 3.05 2.71 1.00 variable First stage 75.16 75.16 75.16 75.16 74.56 74.56 75.25 75.71 75.71 75.16 $F$-statistic
 National Unemployment (percent) National Unemployment (categorical) County Unemployment (percent) Household Financial Prospects Earnings Growth: Mean Earnings Growth: Uncertainty Personal Unemployment (percent) Inflation: Mean Inflation: Uncertainty Firm Profits (categorical) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) OLS Posterior: 0.528*** 0.014*** 0.508*** −0.011*** −0.010** 0.010*** 0.112*** 0.013*** 0.013*** −0.009*** Recession (0.033) (0.001) (0.035) (0.002) (0.005) (0.004) (0.018) (0.004) (0.004) (0.002) IV Posterior: 0.895*** 0.030*** 0.536*** −0.012** −0.013 0.002 0.113* 0.014 0.006 −0.013** Recession (0.131) (0.006) (0.118) (0.006) (0.020) (0.013) (0.066) (0.018) (0.015) (0.005) Observations 1,124 1,124 1,124 1,124 1,118 1,118 1,123 1,121 1,121 1,124 Mean dependent 32.09 0.01 29.55 −0.01 2.64 1.79 6.61 2.60 2.74 −0.01 variable SD dependent 24.18 1.00 23.20 1.00 3.42 2.41 11.47 3.05 2.71 1.00 variable First stage 75.16 75.16 75.16 75.16 74.56 74.56 75.25 75.71 75.71 75.16 $F$-statistic

The table shows OLS estimates based on specification 2 (panel A) and IV estimates based on specification 3 (panel B) of the effect of recession expectations on expectations about macroeconomic and personal outcomes. All specifications control for age, age squared, a dummy for females, log income, a dummy for respondents with at least a bachelor's degree, dummies for the respondent's census region of residence, a measure of the respondent's financial literacy, as well as a dummy for Republicans and a dummy for Democrats. The outcomes in columns 2, 4, and 10 are $z$-scored using the mean and standard deviation of our sample. Robust standard errors are in parentheses. Significant at $*$10%, $**$5%, and $***$1%.

### B. Do Recession Expectations Affect Personal Expectations?

Consistent with the evidence on updating of recession expectations, the experimental variation successfully shifts the respondents' expectations about aggregate unemployment. Posterior beliefs about a recession significantly affect people's subjective probability that the national unemployment rate will increase. In the IV specification, a 1 percentage point higher likelihood of a recession causes a 0.895 and 0.536 percentage point increases in the perceived chance that national (county-level) unemployment will increase (panel B of table 2, columns 1 and 3). We find similar effects using the categorical measure, which is immune to numerical anchoring (column 2). The results on national and county-level unemployment expectations are significant and of similar size in OLS and IV estimations.

Do recession expectations affect people's beliefs about their personal economic outcomes? Table 2 shows our second main result:

Result 2.

People extrapolate from their recession expectations to their own households' financial prospects and to expectations about personal unemployment. The estimated effect sizes are large but still close to job transitions during the last recession.

People who think that a recession is more probable are also more likely to hold pessimistic beliefs about their own household's financial prospects and expect lower earnings growth in their job. They also report lower levels of subjective job security. The estimated effects in the IV specifications are very similar in size to the OLS estimates, but the effects on expected earnings growth become statistically insignificant (panel B). The effect size on subjective job security is substantial, yet in line with job losses during the last recession: a 1 percentage point increase in the likelihood of a recession leads to an increase in subjective unemployment risk of 0.113%. To illustrate the magnitude of this effect, consider moving from a situation with zero risk of a recession to a situation in which a recession will happen with certainty: 11.3% of our respondents would then need to become unemployed for their expectations to be accurate on average. For comparison, the job loss rate increased by 7 percentage points during the Great Recession 2007–2009, and most laid-off workers remained unemployed for several months (Farber, 2011). Thus, although the magnitude of our estimated effect is relatively large, it is still close to the increase in unemployment during the last recession.19

### C. Heterogeneous Extrapolation to Personal Expectations

#### Actual differences in risk exposure across groups.

Actual exposure to macroeconomic risk should affect the extent to which people extrapolate from news about the macroeconomy to their personal expectations. Therefore, we examine changes in unemployment rates over the Great Recession for different demographic groups using data from the Merged Outgoing Rotation Groups of the Current Population Survey (CPS). The unemployment rate increased much more strongly among individuals without a college degree and among males (figure A.10), consistent with previous literature (Hoynes, Miller, & Schaller, 2012). There were similar changes in unemployment rates for individuals aged 25 to 44 and those aged 45 to 55. Moreover, the increase in unemployment during the Great Recession was concentrated among workers who were previously employed in cyclical industries such as manufacturing, construction, and services, while industries such as health and education were less affected (Takhtamanova & Sierminska, 2016). Therefore, we expect respondents with lower education, male respondents, and respondents working in more cyclical industries to update their expectations regarding personal unemployment more strongly in response to a change in their macroeconomic outlook.

#### Who extrapolates from macroeconomic to personal expectations?

In order to test whether extrapolation to expectations about personal economic outcomes is driven by respondents who are more strongly exposed to macroeconomic risk, we interact the posterior belief with dummies for several dimensions of heterogeneity, $heti$. Specifically, we estimate the following IV specification:20
$expi=β0+β1posteriori^+β2posteriori×heti^+β3heti+ΠTXi+ɛi,$
(4)
where
$posteriori^=γ0^+γ1^highrecessioni+γ2^highrecessioni×heti+γ3^heti+Θ^TXi,posteriori×heti^=δ0^+δ1^highrecessioni+δ2^highrecessioni×heti+δ3^heti+Ξ^TXi.$
Our third main result is as follows:
Result 3.

People extrapolate from their macroeconomic outlook to their expected chance of personal unemployment. These effects are driven by individuals most strongly exposed to macroeconomic risk. Thus, updating of personal expectations is data-consistent in terms of size and heterogeneity, indicating that households have an understanding of their exposure to macroeconomic risk.

For example, the perceived chance of becoming unemployed responds strongly for people with lower education, while there is no such effect for people with high education (see figure 2 and table A12). We find qualitatively similar patterns if we instead examine heterogeneity according to the level of earnings. While these differences across groups are large in terms of magnitudes, they are statistically insignificant, potentially due to low power. We find no strong differential response across age groups or gender.
Figure 2.

Heterogeneity in Extrapolation to Personal Expectations

This figure displays IV estimates of the effect of posterior beliefs about the likelihood of a recession on people's (a) subjective chance of being unemployed (panel A) and (b) expected mean earnings growth conditional on working at the same job (panel B) for different demographic groups, including 90% confidence bands. Individuals with above-median earnings ($54,800) are classified as having high earnings. Panel C displays IV estimates of the effect of posterior beliefs about the likelihood of a recession on people's subjective chance of being unemployed for groups with different exposure to risk, including 90% confidence bands. Health, education, and public administration are classified as noncyclical industries, while construction, manufacturing, services, retail and wholesale, transportation, and finance are classified as cyclical industries. “High county unemployment” indicates living in a county with an above-median unemployment rate (4.5%). The estimates are based on IV estimations, where the posterior likelihood of a recession is interacted with the dimension of heterogeneity of interest. These results are also shown in table A12 in the online appendix. Figure 2. Heterogeneity in Extrapolation to Personal Expectations This figure displays IV estimates of the effect of posterior beliefs about the likelihood of a recession on people's (a) subjective chance of being unemployed (panel A) and (b) expected mean earnings growth conditional on working at the same job (panel B) for different demographic groups, including 90% confidence bands. Individuals with above-median earnings ($54,800) are classified as having high earnings. Panel C displays IV estimates of the effect of posterior beliefs about the likelihood of a recession on people's subjective chance of being unemployed for groups with different exposure to risk, including 90% confidence bands. Health, education, and public administration are classified as noncyclical industries, while construction, manufacturing, services, retail and wholesale, transportation, and finance are classified as cyclical industries. “High county unemployment” indicates living in a county with an above-median unemployment rate (4.5%). The estimates are based on IV estimations, where the posterior likelihood of a recession is interacted with the dimension of heterogeneity of interest. These results are also shown in table A12 in the online appendix.

Moreover, the effects of an expected economic downturn on personal unemployment expectations are driven by individuals working in cyclical industries, those with previous unemployment experiences, and those living in counties with higher unemployment (figure 2 and table A13).21 These differences are large in magnitude and statistically significant for previous unemployment experiences ($p<0.1$) and county-level unemployment ($p<0.1$) and insignificant for industry of employment ($p=0.18$). The effects are driven by job loss expectations for individuals with a personal unemployment history and by conditional job finding expectations for those living in areas with high unemployment (table A13). This is in line with the idea that high county-level unemployment could make it more difficult to find reemployment in case of job loss, while a personal unemployment history could proxy for being marginal. Overall, the effects on personal unemployment expectations are driven by those with larger exposure to macroeconomic shocks.

Further, individuals with higher earnings, older individuals, and men expect a reduced earnings growth conditional on keeping their jobs as a result of an economic downturn (figure 2 and table A14). The effects for these subgroups are significantly different from 0 and significantly larger than the effects on individuals with lower earnings ($p<0.05$), younger individuals ($p<0.1$), and women ($p<0.05$). These patterns are in line with higher trend growth in earnings among men and individuals with higher earnings, as well as downward rigidity in wages. Accordingly, an economic downturn could lead to lower but still nonnegative earnings growth at the top of the distribution, while individuals at the bottom of the distribution are affected through job loss, potentially because their wages cannot fall (e.g. due to binding minimum wages).

Finally, there is no heterogeneity in the effect of beliefs about the likelihood of a recession on the perceived chance that national unemployment will increase (table A15). Hence, while more exposed groups drive the results on extrapolation from recession expectations to their personal economic outlook, they expect changes in aggregate unemployment similar to less exposed groups.

### D. Do the Effects Persist over Time?

Table A9 shows that most of our results on updating of expectations decrease in size but remain economically and statistically significant in the two-week follow-up survey. The table shows reduced-form estimates obtained from regressing the different outcome variables on an indicator for the high recession treatment and controls.22 People who receive more pessimistic forecasts about the likelihood of a recession still report a significantly higher probability of an increase in unemployment. For expectations about national- and county-level unemployment, the effect sizes in the follow-up are about 50% and about 42% of the original effect sizes in the main study, respectively. The treatment effects for all personal outcomes are not statistically distinguishable from the treatment effects in the main experiment. However, the coefficients are less precisely estimated in the follow-up and are about 50% (financial prospects) and about 25% (personal unemployment expectations) smaller than in the main study.

This reflects a substantial degree of persistence, given that our intervention was mild and that people likely received other relevant signals about macroeconomic conditions and their personal situation between the two surveys. Indeed, 65% of our respondents agree that they followed news about the economy in the time between the main survey and the follow-up survey.23 An alternative explanation for the reduced effect sizes is that survey respondents could forget about the forecasts and revert back to a default level of their expectations. In addition, we are naturally less powered to detect significant treatment effects in the smaller sample of respondents who completed the follow-up. Taken together, the persistence of the treatment effects suggests that our information treatment leads to true belief updating, while concerns about numerical anchoring, short-lived emotional responses to the treatment, or experimenter demand are mitigated.

### E. Do Macroeconomic Expectations Affect Behavior?

According to a standard Euler equation, an innovation to expected future economic resources should induce households to immediately adjust their consumption. Recent evidence indicates that earnings reductions during recessions are large (Farber, 2011), that recessions can accelerate preexisting adverse trends in the labor market situation of subgroups (Charles et al., 2016; Hershbein & Kahn, 2018), and that recessions can have scarring effects that induce workers to permanently drop out of the labor force (Yagan, 2018). Combined, these findings suggest that economic downturns can entail substantial shocks to people's permanent income. Therefore, we expect individuals to revise their consumption plans when they change their expectations regarding a recession.

In this section, we examine whether updating of recession expectations leads people to adjust their behavior. First, we examine whether updating of these expectations affects our measures of planned and actual changes in nondurable spending around the main intervention. We focus on nondurables for this category because consumption plausibly equals expenditure. Second, we examine whether updating of recession expectations leads our respondents to report a more negative climate for durables purchases or to postpone the actual adjustment in their stock of durables (Bertola, Guiso, & Pistaferri, 2005). Third, we analyze whether updating of recession expectations leads households to actively adjust their stockholdings. Given the well-documented inertia in household portfolios (Bilias, Georgarakos, & Haliassos, 2010; Calvet, Campbell, & Sodini, 2009), the reaction of stock purchases should be small.

We estimate the same IV specification 3 as for our previous analysis, except that our independent variable is now the difference between posterior and prior beliefs about the likelihood of a recession, as our outcome variables refer to changes in individual behavior instead of levels of expectations. In addition we control for people's prior belief. Table 3 shows our fourth main result:

Table 3.
Behavioral Outcomes (IV)
 Consumption Growth Consumption Growth Durable Purchase Durable Purchase Stocks Net Purchases Stocks Net Purchases Stocks Net Sales (planned) (actual) Climate (actual) (scale) (dummy) (dummy) (1) (2) (3) (4) (5) (6) (7) Updating: Recession −0.013** −0.007 −0.006 −0.001 −0.014** −0.005** 0.001 (0.006) (0.005) (0.006) (0.002) (0.007) (0.002) (0.001) Prior −0.013*** −0.003 −0.012*** 0.000 −0.003 −0.001 0.001 (0.004) (0.003) (0.004) (0.001) (0.004) (0.001) (0.001) Observations 1,124 706 1,124 732 732 732 732 Mean dependent variable −0.01 −0.02 −0.00 0.13 −0.01 0.13 0.02 SD dependent variable 1.00 0.73 1.00 0.34 0.99 0.34 0.13 First-stage $F$-statistic 85.76 79.20 85.76 73.45 73.45 73.45 73.45
 Consumption Growth Consumption Growth Durable Purchase Durable Purchase Stocks Net Purchases Stocks Net Purchases Stocks Net Sales (planned) (actual) Climate (actual) (scale) (dummy) (dummy) (1) (2) (3) (4) (5) (6) (7) Updating: Recession −0.013** −0.007 −0.006 −0.001 −0.014** −0.005** 0.001 (0.006) (0.005) (0.006) (0.002) (0.007) (0.002) (0.001) Prior −0.013*** −0.003 −0.012*** 0.000 −0.003 −0.001 0.001 (0.004) (0.003) (0.004) (0.001) (0.004) (0.001) (0.001) Observations 1,124 706 1,124 732 732 732 732 Mean dependent variable −0.01 −0.02 −0.00 0.13 −0.01 0.13 0.02 SD dependent variable 1.00 0.73 1.00 0.34 0.99 0.34 0.13 First-stage $F$-statistic 85.76 79.20 85.76 73.45 73.45 73.45 73.45

The table shows IV estimates of the effect of updating of recession expectations on changes in people's behavior. All specifications control for age, age squared, a dummy for females, log income, a dummy for respondents with at least a bachelor's degree, dummies for the respondent's Census region of residence, a measure of the respondent's financial literacy, as well as a dummy for Republicans and a dummy for Democrats. The outcomes in columns 1, 2, 3, and 5 are $z$-scored using the mean and standard deviation in our sample. Robust standard errors are in parentheses. Significant at $*$10%, $**$5%, and $***$1%.

Result 4.

People's macroeconomic outlook causally affects their consumption plans and stock purchases.

Specifically, becoming more pessimistic about the aggregate economy has a significantly negative effect on our respondents' consumption plans for nondurable goods (column 1). A 10 percentage point increase in the perceived likelihood of a recession leads to a decrease in planned consumption growth by 13% of a standard deviation. This is in line with the effect size of 11% of a standard deviation on the expected change in the financial situation of the household in table 2, column 4. While there seem to be increases in actual consumption growth, these effects are noisily measured and statistically insignificant (column 2). This noisy measurement could arise from (a) the fact that our measure includes some categories that are quite lumpy and at the weekly frequency vary a lot across individuals (such as clothing), (b) imperfect memory about actual spending, and (c) the smaller sample size in the follow-up compared to the main survey. We find no evidence that macroeconomic expectations affect people's assessment of the consumption climate for durable goods (column 3) or their actual durables purchases (column 4).

Moreover, increased pessimism about the economy strongly affects people's self-reported net purchases of stocks between the main survey and the follow-up (column 5 of table 3). The large reaction despite inertia in household portfolios may be due to the fact that respondents in both treatment arms were extremely pessimistic before the treatment, and thus the information provision implied a shift toward a lower subjective probability of a recession that was strong enough to trigger adjustments in portfolios. Consistent with this explanation, the effect is fully driven by higher net purchases of stocks in the treatment arm that received the more optimistic forecast, while there is no significant difference for net sales of stocks (columns 6 and 7).24 A 10 percentage point increase in the likelihood of a recession reduces the likelihood of purchasing stocks by 5 percentage points.

Thus, a higher expected probability of a recession reduces planned consumption growth and should, for a given income, increase saving. Higher saving and lower net purchases of stocks should be reflected in a reduction of the risky portfolio share.25 Survey measures of consumers' expected stock returns behave procyclically and co-move with expectations about general economic conditions, even though this is at odds with theory, market measures of expected returns, and the actual equity premium in the United States (Amromin & Sharpe, 2013; Greenwood & Shleifer, 2014). Moreover, consumers' subjective risk surrounding future returns behaves countercyclically. This suggests that higher and less uncertain expected returns could be driving our results. Alternatively, a higher perceived probability of a recession could increase perceived consumption risk or reduce the expected level of consumption, both of which lead to a lower risky portfolio share in standard portfolio choice problems with CRRA utility.

### F. Expectations and News Consumption

An increase in macroeconomic risk should lead rationally (in-)attentive economic agents to allocate more of their attention to macroeconomic news (Maćkowiak & Wiederholt, 2015; Sims, 2003). However, we find no evidence that expectations about the likelihood of a recession causally affect people's consumption of news about the general economy as measured in the follow-up survey (table A22).26

We also study whether updating of recession expectations between the main survey and the follow-up survey is affected by people's consumption of news between the surveys. Columns 2 and 3 show that news consumption between the two surveys is uncorrelated with people's updating, defined as the difference between the posterior belief in the follow-up and the prior belief in the main survey.

### G. Subjective Beliefs about the Macroeconomy

Our design also allows us to shed light on how expectations about different macroeconomic variables are causally related. Many macroeconomic models incorporate a Phillips curve, a negative relationship between unemployment and inflation. An implicit assumption in most models is that agents form their expectations according to the true model. Moreover, there was substantial disinflation during most recessions in the past (Coibion & Gorodnichenko, 2015b). Thus, a higher likelihood of a recession could lower people's inflation expectations. Columns 8 and 9 of table 2 show our fifth main result:

Result 5.

There is no significantly negative causal effect of people's expectations about the likelihood of a recession on their inflation expectations.

While mean expected inflation is positively correlated with people's recession expectations (panel A, column 8), this relationship is statistically insignificant in an IV specification (panel B). In the IV specification, we can reject effects below $-$0.015 at a significance level of 10%. Recession expectations are positively correlated with inflation uncertainty, but again this effect vanishes in the IV specification (column 9). These results mirror the findings by Coibion, Gorodnichenko, and Kumar (2018), who show that firms do not update their expectations about GDP growth and unemployment when their inflation expectations are shocked.

There are several potential explanations for why we do not find a negative effect of recession expectations on expected inflation. First, the reference time horizon of twelve months for our expectations questions may be too short. Second, our respondents could think that a potential recession is caused by a negative technology shock or a cost-push shock, both of which entail a negative comovement of the output gap and inflation in standard NewKeynesian models. In our data on beliefs about likely causes of a recession, collected in the follow-up, a decline in consumer confidence and political turmoil are the most frequently mentioned causes, while supply-side factors, such as an oil price increase, are not mentioned as frequently (figure A.11).27 Third, consumers may not be sufficiently sophisticated to account for complex relations between macroeconomic variables in their belief formation. Indeed, the high standard errors of our estimates suggest that there is a lot of disagreement among respondents on how a recession will affect inflation.

Finally, recession expectations causally affect our respondents' expectations regarding firm profits (column 10). A 10 percentage point increase in the likelihood of a recession leads to a decrease in expected firm profits by 13% of a standard deviation. The fact that our respondents expect part of an economic downturn to be absorbed by firm profits is in line with recent empirical evidence that firms partially insure their workers against negative shocks (Fagereng, Guiso, & Pistaferri, 2017a, 2017b). In appendix section E.2, we provide additional results on our respondents' subjective beliefs about insurance within the firm.

## V. Robustness

### A. Experimenter Demand Effects

Treatment effects in experiments that shift respondents' expectations could be biased as a result of experimenter demand effects. Specifically, respondents in the different treatment groups may form different beliefs about the experimenter's expectations and try to conform to these expectations.28 We provide several pieces of evidence against the relevance of experimenter demand effects.

First, we assess the sensitivity of our respondents' economic expectations to “demand treatments” (de Quidt, Haushofer, & Roth, 2018), through which we try to deliberately shift our respondents' beliefs about the experimenters' hypothesis about the participants' responses. We conducted an additional experiment on MTurk (robustness experiment 2, described in more detail in table A1) in which a random subset of our respondents is assigned to receive a “demand treatment,” while a control group does not receive any information or signal. Neither of the two groups is shown a professional forecast. In the demand treatment, respondents are provided with the following message: “In this experiment people are randomly assigned to receive different instructions. We hypothesize that participants who are shown the same instructions as you report more optimistic expectations about the US economy.” Afterward we elicit all respondents' recession expectations, their qualitative household financial prospects, and their consumption plans. The demand treatment has very small and insignificant effects on the outcome measures (table A21). This suggests that respondents' self-reported expectations in a setting close to ours are not responsive to explicit signals about the experimental hypothesis.

Second, the patterns of heterogeneity in extrapolation from macroeconomic to personal expectations that we documented in section IVC imply that our findings could only be explained by experimenter demand effects that are systematically related to people's actual exposure to aggregate risk. In addition, the heterogeneity in updating of recession expectations in response to the professional forecasts documented in section IIIB (e.g., by people's confidence in their prior) is also consistent only with differential experimenter demand effects across these groups, which we find unlikely.

Third, within-designs may induce stronger experimenter demand effects than between-designs. In our main experiment, we elicit priors only for recession expectations, while we rely on a between-design for all other outcomes. In robustness experiment 3 (described in more detail in table A1), we also examine whether updating of recession expectations depends on whether a within-design or a between-design is employed. For this purpose, we cross-randomize whether people are asked about their prior belief on top of the random assignment of the professional forecast predicting a 5% probability of a recession. We find no significant difference in updating of recession expectations in response to a 5% forecast regardless of the design employed (table A18, column 5). This suggests that it is unlikely that the within-design employed induces strong demand effects.

Fourth, in our initial experiment, we displayed people's prior belief using a red bar and the professional forecast using a yellow bar. In robustness experiment 3, we instead display people's prior with a blue bar and the professional forecast with a yellow bar, which potentially helps to avoid giving respondents the impression that their priors are the “wrong” beliefs (Bazley, Cronqvist, & Mormann, 2018). Our estimated learning rates from professional forecasts of 5% and 30% from robustness experiment 3 are remarkably close to our estimates from the main experiment (table A18, column 3). This suggests that our findings are robust to the exact graphical illustration of the treatment information.

Finally, in robustness experiment 3, we also modified our experimental instructions to provide an even briefer explanation of economic concepts to our respondents at the beginning of the survey. Next to the quantitative similarity in learning rates from the forecasts, the estimated causal effects of recession expectations on expectations about national unemployment and firm profits in IV estimations remain highly significant (table A20). Taken together, despite the changes in the experimental instructions, different colors of bars, the different survey populations, and the different times, we find very similar effects of our intervention on respondents' expectations.

### B. Numerical Anchoring

An additional methodological concern for our quantitative outcome measures, such as posterior beliefs about the likelihood of a recession, is unconscious numerical anchoring. We alleviate concerns about numerical anchoring in several ways. First, we follow the approach of providing irrelevant numerical anchors suggested by Coibion, Gorodnichenko, Kumar, and Pedemonte (2018) and Cavallo et al. (2017). In robustness experiment 3, we randomly assign our respondents to receiving (a) a professional forecast predicting a 5% probability of a recession, (b) an irrelevant numerical anchor stating that “according to official statistics, 5% of the total U.S. population in 1970 were legal immigrants,” or (c) no information. While provision of the professional forecast strongly shifts respondents' expectations, the effect of the irrelevant numerical anchor is a precisely estimated 0 and the difference in learning rates is significant at the 1% level (table A18). In robustness experiment 2, we show that the provision of a different irrelevant numerical anchor does not significantly shift respondents' recession expectations, their household financial prospects, or their consumption plans (table A21).29

Second, our treatment has significant and strong effects on categorical measures of expectations about national unemployment, firm profits, and the household's financial situation, all of which are naturally immune to numerical anchoring. This suggests that changes in quantitative measures of expectations are not driven by numerical anchoring.

Finally, as documented in section IVD, changes in beliefs remain economically and statistically significant in the two-week follow-up. Since numerical anchoring is a very short-lived phenomenon by definition, this provides additional evidence against the possibility that our treatment effects are driven by numerical anchoring.

### C. External Validity

Our sample is representative of the full-time employed US population in terms of age, gender, region, and income but not in terms of education. In order to check the external validity of our findings, we use the 2015 American Community Survey to create weights that make our sample also representative in terms of education. Specifically, we create weights based on the following 64 cells: gender (2) $×$ aged above 42 (2) $×$ above median income (2) $×$ at least college degree (2) $×$ region of residence (west, south, northeast, midwest) (4). Reweighting has no appreciable effects on our main findings (table A23). In our main analysis we therefore focus on unweighted results, which should be less sensitive to outliers.

## VI. Conclusion

We conduct an information experiment in which we provide respondents with different professional forecasters' assessment of the probability of a fall in real GDP. We use the exogenous variation generated by the information treatment to examine the causal effect of recession expectations on expectations about personal outcomes and behavior. Respondents extrapolate to their perceived chance of becoming personally unemployed in a data-consistent manner. The magnitude of the effect is consistent with job losses during the last recession, and there is heterogeneity in line with proxies for actual exposure to risk. Finally, we provide evidence that people's expectations about the macroeconomy causally affect their consumption plans and stock purchases.

Overall, our findings are consistent with macroeconomic models of imperfect information (Maćkowiak & Wiederholt, 2015; Reis, 2006; Sims, 2003; Woodford, 2003). First, we find that consumers are initially uninformed about relevant signals about the macroeconomy. Second, respondents update their economic expectations in response to news about the macroeconomic environment in line with the predictions of Bayesian updating, although this is not the case for individuals with lower education. Third, updating of personal expectations is data-consistent in terms of size and heterogeneity, indicating that our respondents have an understanding of their own exposure to macroeconomic risk. At a practical level, our findings identify specific groups that policymakers can expect to react to an improved macroeconomic outlook. Specifically, groups with the largest exposure to aggregate risk, such as individuals working in cyclical industries, are most likely to respond to an improved macroeconomic outlook, while a large fraction of the population is unlikely to react. Policymakers could maximize the effectiveness of their communication strategies by targeting these groups.

## Notes

1

Identifying this causal channel is important, as research shows that people's personal situation affects their macroeconomic expectations (see Kuchler and Zafar, 2019). Further, omitted variables could affect both macroeconomic and personal expectations.

2

We build on work examining how beliefs about unemployment correlate with beliefs about interest rates and inflation (Carvalho & Nechio, 2014; Dräger, Lamla, & Pfajfar, 2016; Kuchler & Zafar, 2019). Andre et al. (2019) measure respondents' beliefs about how unemployment and inflation change in response to different macroeconomic shocks.

3

One concern could be that quarterly GDP also fell outside actual recessions in the past, so eliciting beliefs about this outcome could not really capture beliefs about the likelihood of a recession. However, a fall in US real GDP in the fourth quarter has happened only during actual recessions since World War II.

4

The professional forecasts correspond to SPF panelists' beliefs about a drop in real GDP two quarters after this wave of the SPF was conducted.

5

The question framing we use to elicit people's expectations closely follows the New York Fed's Survey of Consumer Expectations (SCE). The question framing was optimized after extensive testing (Armantier et al., 2017) and follows Manski's (2017) guidelines on the measurement of subjective expectations.

6

Means of density forecasts are easy to interpret, while point forecasts could capture mean, mode, or some other moment of our respondents' subjective probability distributions (Engelberg, Manski, & Williams, 2009).

7

We ask our respondents about inflation, as done in the New York Fed's Survey of Consumer Expectations, instead of changes in the general price level, as done in the Michigan Survey of Consumers. Asking consumers to think about prices results in more extreme and disagreeing self-reported inflation expectations (de Bruin, Van der Klaauw, & Topa, 2011).

8

We chose to have a one-week time horizon because this mitigates concerns about measurement error due to imperfect memory and because we were constrained by the time window between the main survey and the follow-up. One caveat is that our measure includes categories that are quite lumpy, such as clothing, and therefore may vary greatly across individuals at the weekly frequency, which could lead to noisier estimates.

9

Moreover, since in the alternative design the treatment intensity is correlated with the level of the prior belief, heterogeneous effects across groups would conflate differences in priors and differential extrapolation from macroeconomic to personal expectations.

10

We elicit probabilities over eight closed bins between $-$12% and 12% and two open bins to which we assign $-$14% and 14%.

11

We take the difference in log spending from the follow-up and the baseline survey, so this variable measures the percent change in spending. We deal with outliers by setting spending growth to missing for respondents in the top and bottom 2% of observed spending growth. We obtain qualitatively similar results if we instead use 1% or 5% as cutoff or if we winsorize the variable.

12

Respondents in the incentive condition are told that they can earn up to \$1 depending on the accuracy of their forecast.

13

In section E.1 in the online appendix, we confirm the external validity of these findings using data from the New York Fed's Survey of Consumer Expectations.

14

The controls are as follows: age, age squared, a dummy for females, log income, a dummy for respondents with at least a bachelor's degree, dummies for the respondent's Census region of residence, a measure of the respondent's financial literacy, as well as a dummy for Republicans and a dummy for Democrats.

15

We examine whether individuals put differential weight on signals that are more optimistic or more pessimistic than their prior belief. We interact the individual-specific shock with a dummy variable taking value 1 if $shock<0$ and 0 otherwise. There is no asymmetric updating from relatively positive and relatively negative signals. Similarly, the weight put on the prior does not differ systematically between the two treatment arms ($p=0.443$), indicating that our respondents do not differentially weigh signals that are more or less positive in absolute terms. Finally, we find no significant differences in learning rates according to the prior. Results are available on request.

16

According to theories of rational inattention, individuals with greater exposure to macroeconomic risk and individuals with lower cost of acquiring information should hold stronger prior beliefs about the likelihood of a recession. We cannot disentangle these two forces in our data.

17

Our evidence on information acquisition costs in the context of expected economic growth complements findings from experimental studies of households' expectations about inflation (Armantier et al., 2016; Cavallo et al., 2017) and house prices (Armona et al., 2018; Fuster et al., 2018) or firm expectations (Coibion, Gorodnichenko, & Kumar, 2018).

18

Our findings suggest that in a setting where individuals observe one specific piece of information once, more highly educated respondents' learning from information may be well approximated by Bayesian updating. However, in general, salience could also matter for how much weight individuals put on information. For instance, D'Acunto et al. (2018) document that the price changes of more frequently purchased goods matter more for the formation of inflation and interest rate expectations.

19

Figure A.18 displays local polynomial regressions of people's expectations about personal economic circumstances on their prior beliefs about the likelihood of a recession. The correlations are all strong and go into the expected directions.

20

The IV specifications account for differential first-stage effects of the high recession treatment on posterior recession expectations across groups and are able to isolate differential second-stage effects of posterior beliefs on personal outcomes. Reduced form specifications would conflate differential first- and second-stage effects across groups.

21

We classify health, education, and “other industries” (mostly public administration) as noncyclical industries, while construction, manufacturing, services, retail and wholesale, transportation, and finance are classified as cyclical industries, in line with empirical evidence (Guvenen et al., 2017; Takhtamanova & Sierminska, 2016).

22

We present reduced-form results rather than instrumental variable estimates as the first stage for an IV regression where we instrument posterior beliefs with random treatment assignment would suffer from weak instrument problems in the smaller follow-up sample, that is, the first-stage $F$-statistic is 6.37, below 10.

23

If all respondents received the same perfectly informative signal between the main survey and the follow-up, they would put a weight of 100% on the new signal, leading to identical follow-up beliefs in the two treatment arms.

24

Only 12 individuals in our sample report net sales of stocks, while 54 individuals (41 individuals) in the more optimistic (pessimistic) treatment report net purchases of stocks. This corresponds to a 36% higher fraction of respondents with net purchases in the low recession treatment compared to the high recession treatment.

25

Given that our variables on consumption plans and stock purchases are categorical, this is not guaranteed and depends on the fractions of people changing their behavior and the conditional amounts by which people adjust.

26

We find no significant heterogeneity in treatment effects on respondents' news consumption by proxies for people's exposure to macroeconomic risk or proxies for people's information acquisition costs. Results are available upon request.

27

We find no heterogeneous responses of inflation expectations dependent on whether respondents think that the recession will be caused by supply- or demand-side factors.

28

Evidence suggests that respondents in online surveys respond only very moderately to explicit signals about the experimenter's wishes (de Quidt et al., 2018).

29

We tell our respondents, “We would like to provide you with some information about the share of illegal immigrants in the United States. According to the Department of Homeland Security, 3 percent of the total U.S. population are illegal immigrants.”

## REFERENCES

Amromin
,
Gene
, and
Steven A.
Sharpe
, “
From the Horse's Mouth: Economic Conditions and Investor Expectations of Risk and Return,
Management Science
60
(
2013
),
845
866
.
Andre
,
Peter
,
Carlo
Pizzinelli
,
Christopher
Roth
, and
Johannes
Wohlfart
, “
Subjective Models of the Macroeconomy: Evidence from Experts and a Representative Sample
,”
CESifo working paper
7850
(
2019
).
Armantier
,
Olivier
,
Wändi Bruine
de Bruin
,
Giorgio
Topa
,
Wilbert
Klaauw
, and
Basit
Zafar
, “
Inflation Expectations and Behavior: Do Survey Respondents Act on Their Beliefs?
International Economic Review
56
(
2015
),
505
536
.
Armantier
,
Olivier
,
Simon
Potter
,
Giorgio
Topa
,
Wilbert
van der Klaauw
, and
Basit
Zafar
, “
Measuring Inflation Expectations,
Annual Review of Economics
5
(
2013
),
273
301
.
Armantier
,
Olivier
,
Giorgio
Topa
,
Wilbert
van der Klaauw
, and
Basit
Zafar
, “
An Overview of the Survey of Consumer Expectations
,”
Economic Policy Review
2017
,
23
:
2
(
2017
).
Armantier
,
Olivier
,
Giorgio
Topa
,
Wilbert
van der Klaauw
, and
Basit
Zafar
The Price Is Right: Updating Inflation Expectations in a Randomized Price Information Experiment
,” this review
98
(
2016
),
503
523
.
Armona
,
Luis C.
,
Andreas
Fuster
, and
Basit
Zafar
, “
Home Price Expectations and Behavior: Evidence from a Randomized Information Experiment,
Review of Economic Studies
86
(
2018
),
1371
1410
.
Bachmann,
Rüdiger
,
Tim O.
Berg
, and
Eric R.
Sims
, “
Inflation Expectations and Readiness to Spend: Cross-Sectional Evidence,
American Economic Journal: Economic Policy
9
(
2015
),
1
35
.
Bailey
,
Michael
,
Ruiqing
Cao
,
Theresa
Kuchler
, and
Johannes
Stroebel
, “
The Economic Effects of Social Networks: Evidence from the Housing Market,
Journal of Political Economy
126
(
2018
),
2224
2276
.
Bailey
,
Michael
,
Eduardo
Dávila
,
Theresa
Kuchler
, and
Johannes
Stroebel
, “
House Price Beliefs and Mortgage Leverage Choice,
Review of Economic Studies
86
(
2018
),
2403
2452
.
Bazley
,
W.
,
H.
Cronqvist
, and
M.
Mormann
, “
In the Red: The Effects of Color on Investment Behavior
,” Swedish House of Finance research paper
17
(
2018
).
Berinsky
,
,
Michele F.
Margolis
, and
Michael W.
Sances
, “
Separating the Shirkers from the Workers? Making Sure Respondents Pay Attention on Self-Administered Surveys,
American Journal of Political Science
58
(
2014
),
739
753
.
Bertola
,
Giuseppe
,
Luigi
Guiso
, and
Luigi
Pistaferri
, “
Review of Economic Studies
72
(
2005
),
973
1007
.
Bilias
,
Yannis
,
Dimitris
Georgarakos
, and
Michael
Haliassos
, “
Portfolio Inertia and Stock Market Fluctuations,
Journal of Money, Credit and Banking
42
(
2010
),
715
742
.
Binder
,
Carola
, and
Alex
Rodrigue
, “
Household Informedness and Long-Run Inflation Expectations: Experimental Evidence,
Southern Economic Journal
85
(
2018
),
580
598
.
Calvet
,
Laurent E.
,
John Y.
Campbell
, and
Paolo
Sodini
, “
Fight or Flight? Portfolio Rebalancing by Individual Investors,
Quarterly Journal of Economics
124
(
2009
),
301
348
.
Carroll
,
Christopher D.
, “
Macroeconomic Expectations of Households and Professional Forecasters
,”
Quarterly Journal of Economics
118
:
1
(
2003
),
269
298
.
Carroll
,
Christopher D.
,
Edmund
Crawley
,
Jiri
Slacalek
,
Kiichi
Tokuoka
, and
Matthew N.
White
, “
Sticky Expectations and Consumption Dynamics
,”
NBER working paper
24377
(
2018
).
Carvalho
,
Carlos
, and
Fernanda
Nechio
, “
Do People Understand Monetary Policy?
Journal of Monetary Economics
66
(
2014
),
108
123
.
Cavallo
,
Alberto
,
Guillermo
Cruces
, and
Ricardo
Perez-Truglia
, “
Inflation Expectations, Learning and Supermarket Prices: Evidence from Field Experiments,
American Economic Journal: Macroeconomics
9
(
2017
),
1
35
.
Charles
,
Kerwin Kofi
,
Erik
Hurst
, and
Matthew J.
Notowidigdo
, “
The Masking of the Decline in Manufacturing Employment by the Housing Bubble,
Journal of Economic Perspectives
30
(
2016
),
179
200
.
Coibion
,
Olivier
, and
Yuriy
Gorodnichenko
, “
What Can Survey Forecasts Tell Us about Information Rigidities?
Journal of Political Economy
120
(
2012
),
116
159
.
Coibion
,
Olivier
, and
Yuriy
Gorodnichenko
Information Rigidity and the Expectations Formation Process: A Simple Framework and New Facts,
American Economic Review
105
(
2015a
),
2644
2678
.
Coibion
,
Olivier
, and
Yuriy
Gorodnichenko
Is the Phillips Curve Alive and Well After All? Inflation Expectations and the Missing Disinflation,
American Economic Journal: Macroeconomics
7
(
2015b
),
197
232
.
Coibion
,
Olivier
,
Yuriy
Gorodnichenko
, and
Saten
Kumar
, “
How Do Firms Form Their Expectations? New Survey Evidence,
American Economic Review
108
(
2018
),
2671
2713
.
Coibion
,
Olivier
,
Yuriy
Gorodnichenko
,
Saten
Kumar
, and
Mathieu
Pedemonte
, “
Inflation Expectations as a Policy Tool?
” NBER working paper 24788 (
2018
).
Coibion
,
Olivier
,
Yuriy
Gorodnichenko
, and
Tiziano
Ropele
, “
Inflation Expectations and Firm Decisions: New Causal Evidence,
Quarterly Journal of Economics
135
(
2020
),
165
219
.
Croushore
,
Dean D.
, “
Introducing: The Survey of Professional Forecasters
,”
(
1993
),
6
,
3
.
D'Acunto
,
Francesco
, “
Tear Down This Wall Street: Anti-Market Rhetoric, Motivated Beliefs, and Investment
,” University of Chicago, Booth School of Business working paper (
2018
).
D'Acunto
,
Francesco
,
Daniel
Hoang
,
Maritta
Paloviita
, and
Michael
Weber
, “
Cognitive Abilities and Inflation Expectations,
American Economic Review Papers and Proceedings
109
(
2019a
),
562
566
.
D'Acunto
,
Francesco
,
Daniel
Hoang
,
Maritta
Paloviita
, and
Michael
Weber
Human Frictions in the Transmission of Economic Policy
,”
Becker Friedman Institute for Economics working paper
2019-07
(
2019b
).
D'Acunto
,
Francesco
,
Daniel
Hoang
,
Maritta
Paloviita
, and
Michael
Weber
IQ, Expectations, and Choice
,”
NBER working paper
25496
(
2019c
).
D'Acunto
,
Francesco
,
Daniel
Hoang
, and
Michael
Weber
, “
The Effect of Unconventional Fiscal Policy on Consumption Expenditure
,”
Becker Friedman Institute for Research in Economics working paper
2018-17
(
2018
).
D'Acunto
,
Francesco
,
Ulrike
Malmendier
,
Juan
Ospina
, and
Michael
Weber
, “
Salient Price Changes, Inflation Expectations, and Household Behavior
,”
SSRN
(
2018
).
Das
,
Sreyoshi
,
Camelia M.
Kuhnen
, and
Stefan
Nagel
, “
Socioeconomic Status and Macroeconomic Expectations
,”
Review of Financial Studies
33
(
2020
),
395
432
).
de Bruin
,
Wändi Bruine
,
Baruch
Fischhoff
,
Susan G.
Millstein
, and
Bonnie L.
Halpern-Felsher
, “
Verbal and Numerical Expressions of Probability: ‘It's a Fifty–Fifty Chance,'
Organizational Behavior and Human Decision Processes
81
(
2000
),
115
131
.
de Bruin
,
Wändi Bruine
,
Wilbert
Van der Klaauw
, and
Giorgio
Topa
, “
Expectations of Inflation: The Biasing Effect of Thoughts about Specific Prices,
Journal of Economic Psychology
32
(
2011
),
834
845
.
de Quidt
,
Jonathan
,
Johannes
Haushofer
, and
Christopher
Roth
, “
Measuring and Bounding Experimenter Demand,
American Economic Review
108
(
2018
),
3266
3302
.
Dräger
,
Lena
,
Michael J.
Lamla
, and
Damjan
Pfajfar
, “
Are Survey Expectations Theory- Consistent? The Role of Central Bank Communication and News
,”
European Economic Review
85
(
2016
),
85
,
84
111
.
Engelberg
,
Joseph
,
Charles F.
Manski
, and
Jared
Williams
, “
Comparing the Point Predictions and Subjective Probability Distributions of Professional Forecasters,
Journal of Business and Economic Statistics
27
(
2009
),
30
41
.
Fagereng
,
Andreas
,
Luigi
Guiso
, and
Luigi
Pistaferri
, “
Firm-Related Risk and Precautionary Saving Response,
American Economic Review Papers and Proceedings
107
(
2017a
),
393
397
.
Fagereng
,
Andreas
,
Luigi
Guiso
, and
Luigi
Pistaferri
Portfolio Choices, Firm Shocks, and Uninsurable Wage Risk,
Review of Economic Studies
85
(
2017b
),
437
474
.
Farber
,
Henry S.
, “
Job Loss in the Great Recession: Historical Perspective from the Displaced Workers Survey, 1984–2010
,”
NBER working paper
17040
(
2011
).
Fuster
,
Andreas
,
Ricardo
Perez-Truglia
,
Mirko
Wiederholt
, and
Basit
Zafar
, “
Expectations with Endogenous Information Acquisition: An Experimental Investigation
,”
NBER working paper w24767
(
2018
).
Goldfayn-Frank
,
Olga
, and
Johannes
Wohlfart
, “
How Do Consumers Adapt to a New Environment in their Economic Forecasting? Evidence from the German Reunification
,”
IMFS working paper
129
(
2018
).
Greenwood
,
Robin
, and
Andrei
Shleifer
, “
Expectations of Returns and Expected Returns,
Review of Financial Studies
27
(
2014
),
714
746
.
Guvenen
,
Fatih
,
Sam
Schulhofer-Wohl
,
Jae
Song
, and
Motohiro
Yogo
, “
Worker Betas: Five Facts about Systematic Earnings Risk,
American Economic Review Papers and Proceedings
107
(
2017
),
398
403
.
Hershbein
,
, and
Lisa B.
Kahn
, “
Do Recessions Accelerate Routine-Biased Technological Change? Evidence from Vacancy Postings,
American Economic Review
108
(
2018
),
1737
1772
.
Hoynes
,
Hilary
,
Douglas L.
Miller
, and
Jessamyn
Schaller
, “
Who Suffers during Recessions?
Journal of Economic Perspectives
26
(
2012
),
27
47
.
Krueger
,
Dirk
,
Kurt
Mitman
, and
Fabrizio
Perri
, “
On the Distribution of the Welfare Losses of Large Recessions
,” in
B.
Honoré
,
A.
Pakes
,
M.
Piazzesi
, and
L.
Samuelson
, eds.,
, vol.
2
(
Cambridge
:
Cambridge University Press
,
2016
).
Kuchler
,
Theresa
, and
Basit
Zafar
, “
Personal Experiences and Expectations about Aggregate Outcomes,
Journal of Finance
74
(
2019
),
2491
2542
.
Kuziemko
,
Ilyana
,
Michael I.
Norton
,
Emmanuel
Saez
, and
Stefanie
Stantcheva
, “
How Elastic Are Preferences for Redistribution? Evidence from Randomized Survey Experiments,
American Economic Review
105
(
2015
),
1478
1508
.
Lusardi
,
Annamaria
, and
Olivia S.
Mitchell
, “
The Economic Importance of Financial Literacy: Theory and Evidence,
Journal of Economic Literature
52
(
2014
),
5
44
.
Maćkowiak
,
Bartosz
, and
Mirko
Wiederholt
, “
Business Cycle Dynamics under Rational Inattention,
Review of Economic Studies
82
(
2015
),
1502
1532
.
Malmendier
,
Ulrike
, and
Stefan
Nagel
, “
Depression Babies: Do Macroeconomic Experiences Affect Risk-Taking?
Quarterly Journal of Economics
126
(
2011
),
373
416
.
Malmendier
,
Ulrike
, and
Stefan
Nagel
Learning from Inflation Experiences,
Quarterly Journal of Economics
131
(
2016
),
53
87
.
Mankiw
,
N. Gregory
, and
Ricardo
Reis
, “
Pervasive Stickiness,
American Economic Review
96
(
2006
),
164
169
.
Mankiw
,
N.
Gregory
,
Ricardo Reis
, and
Justin
Wolfers
, “
” (pp.
209
248
), in
Mark
Gertler
and
Kenneth
Rogoff
, eds.,
NBER Macroeconomics Annual
,
18
(
Cambridge, MA
:
MIT Press
,
2003
,
18
,
209
248
.
Manski
,
Charles F.
, “
Survey Measurement of Probabilistic Macroeconomic Expectations: Progress and Promise
,” in
Martin
Eichenbaum
and
Jonathan A.
Parker
, eds.,
NBER Macroeconomics Annual 2017
(
Chicago
:
University of Chicago Press
,
2017
).
Mian
,
Atif
,
Amir
Sufi
, and
Nasim
Khoshkhou
, “
Partisan Bias, Economic Expectations, and Household Spending
,”
Fama-Miller working paper
(
2018
).
Muth
,
John F.
, “
Rational Expectations and the Theory of Price Movements
,”
Econometrica
(
1961
),
315
335
.
Reis
,
Ricardo
, “
Inattentive Consumers,
Journal of Monetary Economics
53
(
2006
),
1761
1800
.
Roth
,
Christopher
,
Sonja
Settele
, and
Johannes
Wohlfart
, “
Beliefs about Public Debt and the Demand for Government Spending
,”
CESifo working paper
8087
(
2019
).
Sims
,
Christopher A.
, “
Implications of Rational Inattention,
Journal of Monetary Economics
50
(
2003
),
665
690
.
Takhtamanova
,
Yelena F.
, and
Eva
Sierminska
, “
Impact of the Great Recession on Industry Unemployment: A 1976–2011 Comparison
,”
IZA discussion paper
10340
(
2016
).
Tortorice
,
Daniel Louis
, “
Unemployment Expectations and the Business Cycle
,”
BE Journal of Macroeconomics
12:1
(
2012
),
art. 1
.
Vissing-Jorgensen
,
Annette
, “Perspectives on Behavioral Finance: Does ‘Irrationality’ Disappear with Wealth? Evidence from Expectations and Actions” (pp.
139
194
), in
Mark
Gertler
and
Kenneth
Rogoff
, eds.,
NBER Macroeconomics Annual
18
(
Cambridge, MA
:
MIT Press
,
2004
).
Wiederholt
,
Mirko
, “
Empirical Properties of Inflation Expectations and the Zero Lower Bound
,”
Goethe University mimeograph
(
2015
).
Woodford
,
Michael
, “
Imperfect Common Knowledge and the Effects of Monetary Policy
,” in
Philippe
Aghion
,
Roman
Frydman
,
Joseph E.
Stiglitz
, and
Michael
Woodford
, eds.,
Knowledge, Information, and Expectations in Modern Macroeconomics: In Honor of Edmund S. Phelps
(
Princeton, NJ
:
Princeton University Press
,
2003
).
Yagan
Danny
, “
Employment Hysteresis from the Great Recession,
Journal of Political Economy
127
(
2018
),
2505
2558
.

## Author notes

We thank the editor, Olivier Coibion, as well as two anonymous referees for thoughtful comments that improved the paper considerably. We thank Goethe University Frankfurt and Vereinigung von Freunden und Förderern der Goethe Universität for financial support. J.W. thanks support through the DFG project “Implications of Financial Market Imperfections for Wealth and Debt Accumulation in the Household Sector.” The activities of the Center for Economic Behavior and Inequality (CEBI) are funded by the Danish National Research Foundation. We received ethics approval from the University of Oxford.

A supplemental appendix and the experimental instructions are available online at http://www.mitpressjournals.org/doi/suppl/10.1162/rest_a_00867.