AI Inflation Expectations (AIIE)

Select Country:

The AI Inflation Expectations (AIIE) website delivers monthly AI-generated forecasts for the seasonally adjusted year-over-year Consumer Price Index (CPI) inflation in the United States, utilizing OpenAI’s Large Language Model, ChatGPT-4. These forecasts aim to provide a novel perspective on inflation expectations by integrating traditional economic indicators with cutting-edge artificial intelligence insights. Operating continuously, the AIIE model updates its inflation forecasts every hour, enabling real-time adjustments to reflect changes in economic conditions and new data inputs.


Figure 1: AI Inflation Expectations


Note: The AIIE model utilizes a predefined prompt to generate monthly inflation forecasts for the United States. This prompt, designed for precision and consistency, instructs the AI model as follows:

“Assume that you are in {current_date}. Please give me your best forecast of year-over-year seasonally adjusted Consumer Price Index (CPI) inflation in the United States for the current month and the next 12 months. Please give me numeric values for these forecasts. Do not use any information that was not available to you as of {current_date} to formulate these forecasts. Give me a single estimate for each month in the following format and also provide the information sources for the forecast.

Example:
Jan2024: 3.2%
...
Information Sources:”


Additional Forecasts


Figure 2: AI Unemployment Rate Expectations


Note: The AIIE model utilizes a predefined prompt to generate monthly unemployment rate forecasts for the United States. This prompt, designed for precision and consistency, instructs the AI model as follows:

“Assume that you are in {current_date}. Please give me your best forecast of the year-over-year seasonally adjusted unemployment rate in the United States for the current month and the next 12 months. Please provide numeric values for these forecasts. Do not use any information that was not available to you as of {current_date} to formulate these forecasts. Give me a single estimate for each month in the following format and also provide the information sources for the forecast.

Example:
Jan2024: 3.2%
...
Information Sources:”



Figure 3: AI Federal Funds Rate Expectations


Note: The AIIE model utilizes a predefined prompt to generate monthly Federal Funds rate forecasts for the United States. This prompt, designed for precision and consistency, instructs the AI model as follows:

“Assume that you are in {current_date}. Please provide your best forecast for the interest rates measured by the Federal Funds rate in the United States for the current month and the subsequent 12 months. Provide numeric values for these forecasts without using any information that was not available as of {current_date} . Give a single estimate for each month in the following format and include the sources of information for these forecasts.

Example:
Apr2024: 5.5%
...
Information Sources:”



Figure 4: AI GDP Growth Rate Expectations


Note: The AIIE model utilizes a predefined prompt to generate quarterly GDP growth rates forecasts for the United States. This prompt, designed for precision and consistency, instructs the AI model as follows:

“Assume that you are in {current_date}. Please give me your best forecast of the quarterly GDP growth rates (annualized) in the United States for the current quarter and the next four quarters. Please provide numeric values for these forecasts. Do not use any information that was not available to you as of {current_date} to formulate these forecasts. Give me a single estimate for each quarter in the following format and also provide the information sources for the forecast.

Example:
2024Q1: 3.2%
...
Information Sources:”



Methodology

Forecast Generation Process

The process of generating forecasts includes the following steps:

  1. Date Specification: The model recognizes the placeholder {current_date} which is dynamically replaced with the actual date when the forecast is generated.
  2. Data Constraint: The model adheres strictly to the instruction of excluding any data not available as of {current_date}, ensuring the forecast relies solely on data up to that date.
  3. Forecast Output: The AI provides a numeric estimate of CPI inflation for the current month and each of the following twelve months, formatted as specified.
  4. Information Sources Disclosure: Accompanying each forecast, the model lists all utilized sources of information.

Example Forecast

An example of a forecast generated by the AI model might look like this:

        2024Jan: 3.2%
        2024Feb: 3.1%
        2024Mar: 3.0%
        ...
        Information Sources: U.S. Bureau of Labor Statistics, Federal Reserve Economic Data
        

Data Availability

Data Downloads

For researchers, economists, and the public interested in detailed analysis, we offer the following downloadable data options:

  • Original Data: Complete dataset of hourly inflation forecasts as generated by the AI model.
  • Aggregate Data: Monthly aggregates of the forecasts, providing a more macroscopic view of trends.

Usage of Data

The downloadable data can be used for academic research, economic analysis, policy making, and educational purposes. We encourage users to cite the source as follows when using the data in publications or presentations:

Jahangir Alam, Huiyu Li, and Tatevik Sekhposyan, Inflation Expectations Through the Lens of Artificial Intelligence, [Accessed Date], Retrieved from https://aiinflationexpectations.org/.

Chidambaram Ganesan provided excellent research assistance and technical support for this project.

Model Limitations

While the AIIE model provides some predictions for economic indicators, users should be aware of its limitations, including:

  • Dependency on the quality and timeliness of input data.
  • Inherent challenges in interpreting AI-generated predictions due to the “black-box” nature of neural networks.
  • Potential biases in the training data that may affect forecast outcomes.






AI Inflation Expectations with RAG

This experimental approach uses Retrieval-Augmented Generation (RAG) to leverage the top 10 Google search results for providing context to the AIIE model. By summarizing these search results with ChatGPT, we remove unnecessary content and integrate the summaries as context for the AIIE model.


Figure 5: AI Inflation Expectations with RAG


Note: The AIIE model utilizes a predefined prompt to generate monthly Federal Funds rate forecasts for the United States . This prompt, designed for precision and consistency, instructs the AI model as follows:

“Assume that you are in {current_date}. Please give me your best forecast of year-over-year seasonally adjusted Consumer Price Index (CPI) inflation in the United States for the current month and the next 12 months. Please give me numeric values for these forecasts. Do not use any information that was not available to you as of {current_date} to formulate these forecasts. Give me a single estimate for each month in the following format and also provide the information sources for the forecast.

Example:
Jan2024: 3.2%
...
Information Sources:”

You can refer to the following context to improve your answer:
{Context}



Figure 6: AI Unemployment Rate Expectations with RAG


Note: The AIIE model utilizes a predefined prompt to generate monthly Federal Fundsrate forecasts for the United States This prompt, designed for precision and consistency, instructs the AI model as follows:

“Assume that you are in {current_date}. Please give me your best forecast of the year-over-year seasonally adjusted unemployment rate in the United States for the current month and the next 12 months. Please provide numeric values for these forecasts. Do not use any information that was not available to you as of {current_date} to formulate these forecasts. Give me a single estimate for each month in the following format and also provide the information sources for the forecast.

Example:
Jan2024: 3.2%
...
Information Sources:”

You can refer to the following context to improve your answer:
{Context}



Figure 7: AI Federal Funds Rate Expectations with RAG


Note: The AIIE model utilizes a predefined prompt to generate monthly Federal Funds rate forecasts for the United States. This prompt, designed for precision and consistency, instructs the AI model as follows:

“Assume that you are in {current_date}. Please provide your best forecast for the interest rates measured by the Federal Funds rate in the United States for the current month and the subsequent 12 months. Provide numeric values for these forecasts without using any information that was not available as of {current_date} . Give a single estimate for each month in the following format and include the sources of information for these forecasts.

Example:
Apr2024: 5.5%
...
Information Sources:”

You can refer to the following context to improve your answer:
{Context}



Figure 8: AI GDP Growth Rate Expectations


Note: The AIIE model utilizes a predefined prompt to generate monthly Federal Funds rate forecasts for the United States. This prompt, designed for precision and consistency, instructs the AI model as follows:

“Assume that you are in {current_date}. Please give me your best forecast of the quarterly GDP growth rates (annualized) in the United States for the current quarter and the next four quarters. Please provide numeric values for these forecasts. Do not use any information that was not available to you as of {current_date} to formulate these forecasts. Give me a single estimate for each quarter in the following format and also provide the information sources for the forecast.

Example:
2024Q1: 3.2%
...
Information Sources:”

You can refer to the following context to improve your answer:
{Context}