Signals from the Fed's Research Agenda: Inflation, Labor Markets, and the Boundaries of AI Prediction

Introduction: Why Central Bank Working Papers Deserve Attention from the Business Community

The Federal Reserve Bank of San Francisco (SF Fed) regularly publishes working papers showcasing frontier research by its economists and external scholars. Although academic in purpose, their topic selection often reflects the core concerns of the Fed's decision-making circle. Through the series of papers from the first half of 2026, we can observe a clear shift in the central bank's research agenda: from traditional aggregate analysis toward micro-level heterogeneity, innovation in data methods, and the long-term effects of structural trends. For businesses and investors, understanding these changes in research direction helps anticipate macro policy turning points and industry evolution paths.

Key Observation 1: The Contest Between Soft and Hard Data Is Reshaping Economic Forecasting

The paper *Do Vibes Predict Recessions?* cuts directly into a classic debate: can "soft data" built on sentiment, expectations, and narratives predict recessions better than hard data? The researchers use a real-time forecasting framework to compare the predictive power of the two data types for recessions 1 to 12 months ahead. The findings show that soft data respond more sensitively to rising recession risk, but can also emit noisy signals. This reminds business decision-makers: in highly uncertain environments, shifts in market sentiment and expectations may lead official statistics, but should not be used alone as a basis for decisions.

Key Observation 2: Heterogeneity in Financial Constraints Determines the "Temperature Gap" in Monetary Policy Transmission

*Prices and Monetary Policy: The Role of Financial Constraints* uses Swedish firm-level micro data and finds that smaller, more financially constrained firms adjust prices significantly less than large firms when monetary policy changes. This breaks the simplifying assumption that "price stickiness is homogeneous." For the North American market, this finding implies that in tightening cycles, small and medium-sized enterprises may face greater profit margin compression, while large firms find it easier to pass on costs. When investors assess corporate earnings resilience, they need to incorporate firm balance sheet structure and pricing power into the same framework.

Key Observation 3: Long-Term Trends in the Labor Market—Declining Participation Rates and the Remote Work Premium*Trends in Labor Force Participation and Unemployment, 1976-2024* uses CPS microdata to estimate the trend and cyclical components of labor force participation for 44 age-sex-education groups. The research reveals long-term structural changes in labor supply, with particularly pronounced trend declines in participation among less-educated groups. Meanwhile, *The Work-from-home Wage Premium*, based on French administrative data, finds that within the same occupation, industry, and commuting zone, remote workers earn on average 12% higher hourly wages than on-site workers. About half of this premium can be explained by observable characteristics; the other half may be related to productivity or work flexibility. Together, these two studies show that remote work is not simply a change in workplace location, but is redefining skill premiums and labor market stratification. If companies continue to set compensation based on traditional attendance logic, they may miss a key variable in talent competition.

Key Observation 4: The Realistic Boundaries of AI in Macroeconomic Forecasting

At a time when the market generally has high hopes for generative AI, the SF Fed's *ChatMacro* study offers a sobering piece of evidence: real-time inflation forecasts obtained through ChatGPT prompts are "basically inaccurate and stale" out of sample. This shows that although large language models can mimic human writing and logic, they still cannot replace structured economic models in time-series forecasting of macro variables. The implication for financial institutions and businesses is that AI tools are suitable for processing unstructured text and constructing alternative indicators, but using them directly as core business forecasting engines remains dangerous. Research institutions should clarify AI's supporting role rather than treating it as a "crystal ball."

Key Observation 5: New Measures of Inflation Shock Momentum and Policy Transmission

*Measuring Inflation Shock Momentum* proposes a nonparametric filtering method to identify sustained directional runs in monthly inflation shocks, and constructs an "inflation shock momentum" indicator based on more than 100 PCE components. This framework helps distinguish temporary price fluctuations from persistent inflation trends. At the same time, *Financial Conditions and Capital Investment Choices*, using 150 years of data from 17 countries, points out that tight financial conditions prompt firms to shift toward cheaper but less energy-efficient capital goods. This means high interest rates not only suppress the total volume of investment, but may also distort the structure of investment, affecting the pace of energy transition in the long run. Policymakers and ESG investors should pay attention to this hidden cost.

Three Implications for the North American Business Environment1. The uneven transmission of monetary policy demands more refined corporate risk management. Large enterprises and SMEs, as well as high-cash-flow and low-cash-flow firms, differ significantly in their sensitivity to interest rates. Financial hedging strategies should not only focus on the direction of rates, but also assess the relative level of their own financing constraints.

2. The structural decline in labor supply will intensify competition for talent. Long-term trends in labor force participation and the remote-work premium indicate that workers now demand greater flexibility and location independence. Northern tech companies have been the first to adapt; traditional manufacturing and services, if they do not follow suit, will face greater hiring difficulties.

3. Data and AI applications must distinguish between "narrative" and "prediction." Soft data, market sentiment, and AI-generated content can serve as early signals, but final decisions must still return to rigorous economic models and stress tests. Enterprises that rely too heavily on a single source of information will pay the price when recession signals are mixed.

Long-Term Outlook

Over the next 3–5 years, the Federal Reserve's research system may tilt further toward quantized and heterogeneous models. Interdisciplinary "big micro data + machine learning" will replace pure time-series statistics as the mainstream of policy analysis. At the same time, the limitations of AI prediction will give rise to a hybrid decision-making paradigm of "human judgment + machine computation." At the industry level, the long-term distortion of capital efficiency in a tightening environment may incentivize policy adjustments for energy-saving technologies, while the normalization of remote work will drive interstate labor mobility and pay transparency in the United States.

For North American business observers, Federal Reserve working papers are not merely academic archives; they record policy hypotheses and market signals that may be validated in the next economic cycle. Reading the research agenda is reading future policy in advance.

Key Conclusions

• Combined forecasts of soft and hard data will become the standard, but decision-makers must be wary of noise.
• Differences in financial constraints lead to "layered" monetary policy transmission, with SME profit margins under greater pressure.
• Remote work is creating a persistent wage premium, and HR strategies need to be repriced.
• Generative AI performs poorly in macroeconomic forecasting; companies should clearly define the boundaries of AI capabilities.
• Research on inflation shock momentum indicators and the energy efficiency of capital provides new tools for inflation tracking and ESG investing.

References

San Francisco Fed working papers page: https://www.frbsf.org/research-and-insights/publications/working-papers