ECB : Inflation Expectations

 The core focus of the study is investigating how disagreement in inflation narratives between general-audience and specialized newspapers contributes to the absolute gap in inflation expectations between households and experts. This investigation is situated within the broader analysis of inflation narratives and the expectation gap.

The Expectation Gap and Inflation Narratives

The "expectation gap" refers to the divergence between the inflation expectations of households and those of experts, a gap that is sometimes large and volatile. Central banks consider the anchoring of private-sector inflation expectations highly important, as unanchored expectations can undermine credibility and interfere with the goal of price stability.

While expectations held by professionals are typically "well anchored," those of households often diverge.

The core objective of the research is to determine whether the absolute expectation gap widens when demand-supply narrative disagreement increases between general and specialized newspapers. The findings confirm this central testable hypothesis at both the aggregate and individual levels.

The Role of Narrative Disagreement

The study uses "inflation narratives" to mean the perceived triggers of inflation, which are extracted using a Causality Extraction algorithm that identifies causal relationships between events mentioned in text. These narratives are then classified into demand narratives (e.g., strong consumer spending, government spending, monetary policy) and supply narratives (e.g., energy price increases, supply chain, labor).

The demand–supply narrative disagreement measures the extent to which general-audience newspapers (The New York Times, USA Today, The Washington Post) and specialized newspapers (The Wall Street Journal) differ in their attention to demand-side versus supply-side attributions of inflation causes.

Key findings regarding narrative disagreement and the expectation gap include:

• Narrative Disagreement Widens the Gap: The absolute expectation gap (between U.S. households and experts) increases when narrative disagreement between general and specialized newspapers increases.
• Vulnerability of Specific Groups: This relationship is stronger for specific household demographics, namely non-college-educated and older households. This is explained by the higher likelihood of older individuals reading newspapers, and the college-educated being more likely to read specialized newspapers (whose narratives align more closely with experts' views).
• Incentives to Be Informed: The positive relationship between the expectation gap and narrative disagreement strengthens when the level and persistence of inflation rise, which is when the costs of being uninformed about inflation increase.
• Consistency of Narratives: The narratives presented in general newspapers, which households are more likely to read, are found to incorrectly align with experts' demand-supply views and macroeconomic data dynamics. In contrast, specialized newspapers' narratives correctly align with experts' economic views.

Implications for Policy and Media Analysis

The results suggest that policymakers cannot rely solely on increasing media coverage to bridge the expectation gap. Although greater newspaper coverage of inflation might, in theory, narrow the gap by lowering information costs for households, empirical evidence often suggests the opposite.

The issue lies not just in the volume of coverage, but in the consistency and accuracy of the explanation of inflation drivers. Efforts to reduce the gap require ensuring that clear and consistent explanations of inflation drivers reach a broad audience through multiple channels, including general-audience outlets.

The study suggests that the narratives of general newspapers differ in how they capture the views of households versus experts, possibly conveying incorrect narratives to households.

General newspapers generally communicate demand–supply stories consistent with households’ views of the economy, but inconsistent with experts’ economic views and macroeconomic data.

The investigation into the expectation gap through the lens of narrative disagreement highlights that the gap shrinks with inflation press coverage only when media disagreement is minimal. This emphasis on narrative alignment underscores the complexity of central bank communication aimed at reducing the dispersion of inflation forecasts across different groups.

The study employs a sophisticated methodology, rooted in text analysis and natural language processing (NLP), to examine how differences in media coverage contribute to the volatile gap between the inflation expectations of households and experts.

Data Sources

The analysis relies on two primary types of data: newspaper articles, which serve as the source of inflation narratives, and established surveys for measuring inflation expectations.

1. Newspaper Data (Inflation Narratives)

The research utilizes a corpus of over 180,000 U.S. newspaper articles on inflation published between 1991 and 2022. These articles were identified by mentioning specific keywords (e.g., "inflation," "cpi," "consumer price," "ppi," or "producer price"). The articles are strategically divided into two groups, aligning with different audience segments:

• General-Audience Newspapers: These include The New York Times (NYT), USA Today (USAT), and The Washington Post (WaPo). These are classified together because their readership demographics align more closely with the general public and are sources traditionally used in models of household expectation formation.
• Specialized Newspaper: The Wall Street Journal (WSJ) is classified as the specialized outlet, recognizing its focus on business leaders, investors, and affluent consumers, whose views are likely closer to those of economic experts.

In total, the final corpus used for narrative analysis comprises 157,130 inflation articles, with the WSJ publishing the majority (92,974).

2. Inflation Expectation Data (The Expectation Gap)

The gap that the methodology seeks to explain is measured using standard survey data:

• Household Expectations: These are derived from the monthly University of Michigan Survey of Consumers (MSC). The analysis uses both aggregate mean expectations and individual household expectations, incorporating demographic characteristics like age and education.
• Expert Expectations: These come from the Survey of Professional Forecasters (SPF), conducted by the Federal Reserve Bank of Philadelphia, which collects forecasts from private firms. Since the SPF is quarterly, its data is linearly interpolated to obtain monthly estimates for comparison against the household data.

Methodology for Measuring Narrative Disagreement

The core of the methodology lies in transforming unstructured text from newspapers into quantifiable measures of inflation narratives and then measuring the divergence between media types.

1. Causality Extraction (CE) Algorithm

A Causality Extraction (CE) algorithm, an NLP tool, is employed to identify the perceived triggers of inflation in the text, defining them as inflation "narratives". This approach is utilized because it can specifically identify explicit causal relationships ("cause" and "effect") mentioned within a sentence, allowing for the extraction of inflation drivers, which cannot be captured by simpler methods like dictionary searches or topic models.

The CE algorithm operates by:

1. Identifying causal relations expressed via explicit causal keywords (e.g., "because," "trigger").
2. Checking that an inflation expression is the specified "effect".
3. Extracting the corresponding cause (the inflation driver) as the inflation narrative.

2. Classification into Demand and Supply Narratives

The resulting extracted narratives are classified using a dictionary method into two fundamental categories based on the perceived drivers of inflation:

• Demand Narratives: Attributing inflation to factors like consumer spending, monetary policy, or government spending/deficits.
• Supply Narratives: Attributing inflation to factors like supply chain issues, labor markets, or commodity/energy price increases.

This classification determines whether an article is predominantly a demand or supply article.

3. Measuring Narrative Disagreement

The central variable of interest, demand–supply narrative disagreement ($NetDemand_{G-S}$), quantifies the extent to which general and specialized newspapers differ in their attention to these classified narratives.

• First, the relative attention of each newspaper type is calculated as $NetDemand_{n,t}$, which is the difference between the monthly volume of demand and supply articles published, scaled to range between -1 and 1.
• The final disagreement measure is the difference between the relative attention of general and specialized newspapers ($NetDemand_{G, t} - NetDemand_{S, t}$).
• The key variable used in testing the hypothesis is the absolute value of this measure, $|NetDemand_{G-S, t-1}|$, which captures the quantity of disagreement, regardless of whether general or specialized newspapers emphasized demand or supply more heavily.

This methodical framework allows the study to test the hypothesis that the absolute expectation gap between households and experts widens when the measured demand–supply narrative disagreement in the media increases.

The empirical findings center on establishing a link between demand-supply narrative disagreement in the media and the absolute gap in inflation expectations between U.S. households and experts. The study confirms its central hypothesis and yields several specific insights regarding this relationship and its moderators.

1. Narrative Disagreement Widens the Expectation Gap

The core finding is the confirmation of the central testable hypothesis (H2): the absolute expectation gap widens when demand–supply narrative disagreement increases between general and specialized newspapers. This result is confirmed at both the aggregate and individual household levels.

• Magnitude of Disagreement Matters: The expectation gap widens with the absolute value of the disagreement measure ($|NetDemand_{G-S, t-1}|$), which captures the quantity of disagreement regardless of whether general or specialized newspapers emphasized demand or supply more heavily.
• Other Dimensions of Disagreement: The individual absolute expectation gap also widens significantly with disagreement related to the hawkish/dovish nature of the narratives ($|NetHawkish_{G-S}|$) and whether the narratives discuss observed or expected inflation episodes ($|NetObserved_{G-S}|$).

2. Moderating Role of Household Demographics

The relationship between the expectation gap and narrative disagreement varies significantly across different household demographics (H3 is broadly confirmed):

• Non-College-Educated and Older Households: The positive association between the expectation gap and narrative disagreement is stronger for individuals without a college degree and for older individuals.
  • This result is intuitive, as non-college-educated households are less likely to read specialized newspapers, and older people are more likely to read newspapers overall.
• Education and Income: The relationship weakens with college education and tends to weaken for individuals in the second and fifth income quintiles compared to the middle quintile.
• Sex: The study found no significant change based on the sex of the respondent, aligning with evidence that men and women do not differ in their news readership.

3. Impact of Inflation Levels and Persistence

The study examined how the incentives for gathering information about inflation affect the relationship between the expectation gap and narrative disagreement:

• Inflation Level: The individual absolute expectation gap widens with narrative disagreement only when the level of inflation is above its mean.
• Inflation Persistence: The relationship strengthens when the persistence of inflation rises (i.e., when persistence exceeds its mean).
• Incentive Rationale: These results suggest that narrative disagreement is particularly important for the absolute expectation gap when inflation is high or persistent, which are periods when the costs of being uninformed about inflation increase.

4. Alignment of Narratives with Expectations and Macro Data

A crucial set of findings relates to which narratives align with which audience (H5 is partially confirmed):

• Household Alignment: The narratives presented in both general and specialized newspapers align correctly with the expectations of households (specifically, how households expect inflation and unemployment to co-move, suggesting that supply articles lead to the expectation of inflation and unemployment moving in the same direction).
• Expert Alignment: Only the narratives of specialized newspapers correctly align with experts’ expectations regarding inflation and unemployment co-movement.
• Misalignment of General Newspapers: The narratives of general newspapers are found to incorrectly align with experts’ economic views and with actual macroeconomic data dynamics. Specifically, general newspapers mistakenly publish relatively more demand narratives when inflation and unemployment move in the same direction. This suggests general newspapers may convey "incorrect narratives" to households.

5. Findings Related to Press Coverage and Forecast Errors

The study provides additional context regarding the role of media volume and forecast accuracy:

• Press Coverage vs. Gap: The evidence rejects the simple hypothesis (H1) that the absolute expectation gap narrows with inflation press coverage. Instead, the individual absolute expectation gap generally rises with inflation press coverage, supporting the findings of earlier work by Pfajfar and Santoro (2013). The study suggests that the issue is not the volume of news, but the disagreement in the narrative content.
• Causal Press Coverage: When focusing only on articles containing explicit inflation narratives ("causal inflation articles"), the individual absolute expectation gap still widens with causal inflation press coverage from both general and specialized newspapers.
• Forecast Errors: The individual absolute forecast errors made by households widen with narrative disagreement, although this relationship weakens when controlling for the level and volatility of inflation.

In sum, the sources demonstrate that the divergence in media narratives, particularly between general and specialized outlets regarding demand-supply drivers, is a statistically significant factor explaining why household and expert inflation expectations drift apart, especially for those demographics most reliant on general news. This indicates that clarity and consistency in the explanation of inflation drivers across media channels are crucial for effective expectation management by policymakers.

The sources provide specific implications for monetary policymakers regarding how to effectively manage inflation expectations, particularly by addressing the role of inflation narratives and media disagreement within the expectation gap analysis.

Rethinking Central Bank Communication Strategy

The core implication for central banks is that they cannot rely solely on increasing media coverage of inflation to bridge the gap between household and expert expectations. While increasing coverage theoretically lowers information costs, empirical evidence suggests that the absolute expectation gap generally rises with inflation press coverage,,,.

The key issue is the content and consistency of the message:

1. Focus on Narrative Consistency: Policymakers must ensure that clear and consistent explanations of the inflation drivers reach a broad audience through multiple communication channels, especially general-audience outlets. The sources find that the expectation gap shrinks with inflation press coverage only when media disagreement is minimal,.
2. Addressing Narrative Misalignment: The study highlights a major challenge: the narratives conveyed by general newspapers incorrectly align with experts’ economic views and macroeconomic data, even though these are the outlets households are more likely to read,,. In contrast, specialized newspapers' narratives correctly align with experts' views,. This misalignment suggests that information consumed by the general public may be distorted or imprecise concerning the actual economic drivers of inflation,.
3. Reducing Forecast Dispersion: If central bank communication aims to reduce the dispersion of inflation forecasts among different groups of individuals, it must disseminate its inflation narratives across a broad range of channels. The goal is to ensure consistency so that the clear explanation of inflation drivers reaches the general public.

Understanding the Volatility of the Expectation Gap

Central banks attach great importance to anchoring private-sector inflation expectations because unanchored expectations weaken credibility and hinder price stability. Understanding the fluctuations in the expectation gap is critical for them. The study offers policymakers insights into when the expectation gap is most vulnerable to narrative disagreement:

• Heightened Vigilance During High Inflation: The relationship between the absolute expectation gap and narrative disagreement strengthens when the level and persistence of inflation rise,. This occurs during periods when the costs for households of being uninformed about inflation are higher,,. Policymakers should be particularly concerned with narrative consistency when inflation is high or persistent.
• Targeting Vulnerable Demographics: The expectation gap widens more significantly with narrative disagreement for non-college-educated and older households,,. These groups are more likely to rely on general-audience newspapers,. This suggests communication strategies need to be tailored to ensure these key demographics receive accurate information.
• Focusing on Key Narratives: Disagreement about monetary policy narratives is found to widen the expectation gap the most,,. This is particularly relevant given that experts' expectations are more reactive to central bank communication.

In essence, the findings suggest that policymakers must actively track and respond not only to the volume of inflation news but also to the quality and homogeneity of the inflation narratives being disseminated to the public,. The introduction of simple measures of demand and supply narratives can serve as real-time proxies for tracking the consistency of economic views across households and professionals.