Search Results
Working Paper
Total Recall? Evaluating the Macroeconomic Knowledge of Large Language Models
We evaluate the ability of large language models (LLMs) to estimate historical macroeconomic variables and data release dates. We find that LLMs have precise knowledge of some recent statistics, but performance degrades as we go farther back in history. We highlight two particularly important kinds of recall errors: mixing together first print data with subsequent revisions (i.e., smoothing across vintages) and mixing data for past and future reference periods (i.e., smoothing within vintages). We also find that LLMs can often recall individual data release dates accurately, but aggregating ...
Journal Article
Economic data—appearances can be deceiving
Working Paper
Why Have Long-term Treasury Yields Fallen Since the 1980s? Expected Short Rates and Term Premiums in (Quasi-) Real Time
Treasury yields have fallen since the 1980s. Standard decompositions of Treasury yields into expected short-term interest rates and term premiums suggest term premiums account for much of the decline. In an alternative real-time decomposition, term premiums have fluctuated in a stable range, while long-run expected short-term interest rates have fallen. For example, a real-time decomposition of the 10-yr. Treasury yield shows term premiums essentially equal in late 2013 and 2023, while the long-run value of expected short-term interest rates is estimated to have fallen in a manner similar to ...
Working Paper
Implications of real-time data for forecasting and modeling expectations
This note extends the analysis in Stark and Croushore (2001) with an emphasis on the importance of data vintage for survey forecasts and modeling expectations. For both of these types of empirical exercises, results suggest that the choice of latest available or real-time data is critical for variables subject to large level revisions, but almost irrelevant for variables subject to only small revisions. Other forecasting practices were examined, with some surprising results.
Working Paper
Evaluating real-time VAR forecasts with an informative democratic prior
This paper proposes Bayesian forecasting in a vector autoregression using a democratic prior. This prior is chosen to match the predictions of survey respondents. In particular, the unconditional mean for each series in the vector autoregression is centered around long-horizon survey forecasts. Heavy shrinkage toward the democratic prior is found to give good real-time predictions of a range of macroeconomic variables, as these survey projections are good at quickly capturing endpoint-shifts.
Journal Article
Real-time performance of GDPplus and alternative model-based measures of GDP: 2005—2014
Like most macroeconomic variables, real gross domestic product is subject to measurement error. Because the U.S. Bureau of Economic Analysis lacks complete information at the time it publishes its initial GDP estimates, revisions are often substantial. Analysts concerned about the accuracy of these early estimates for expenditure GDP could focus instead on gross domestic income, the BEA?s measure of U.S. output on the income side of the national accounts. Conceptually, GDP on the expenditure side should equal GDP on the income side, and there should be no choice to make between the two ...
Journal Article
The mismeasured personal saving rate is still useful: using real-time data to improve forecasting
People make decisions based on information. Often, with hindsight, they could have made better choices. Economics faces a similar problem: Economic data, when first released, are often inaccurate and may subsequently be revised. In "The Mismeasured Personal Saving Rate Is Still Useful: Using Real-Time Data to Improve Forecasting," Leonard Nakamura uses the U.S. personal saving rate - a statistic that has often been initially low, then substantially revised upward - to discuss how modern economic statistical techniques can improve forecasting.
Working Paper
Tests of equal predictive ability with real-time data
This paper examines the asymptotic and finite-sample properties of tests of equal forecast accuracy applied to direct, multi-step predictions from both non-nested and nested linear regression models. In contrast to earlier work in the literature, our asymptotics take account of the real-time, revised nature of the data. Monte Carlo simulations indicate that our asymptotic approximations yield reasonable size and power properties in most circumstances. The paper concludes with an examination of the real-time predictive content of various measures of economic activity for inflation.
Working Paper
Learning and Shifts in Long-Run Productivity Growth
Shifts in the long-run rate of productivity growth–such as those experienced by the U.S. economy in the 1970s and 1990s–are difficult, in real time, to distinguish from transitory fluctuations. In this paper, we analyze the evolution of forecasts of long-run productivity growth during the 1970s and 1990s and examine in the context of a dynamic general equilibrium model the consequences of gradual real-time learning on the responses to shifts in the long-run productivity growth rate. We find that a simple updating rule based on an estimated Kalman filter model using real-time data ...
Report
Forecasting through the rear-view mirror: data revisions and bond return predictability
Real-time macroeconomic data reflect the information available to market participants, whereas final data?containing revisions and released with a delay?overstate the information set available to them. We document that the in-sample and out-of-sample Treasury return predictability is significantly diminished when real-time as opposed to revised macroeconomic data are used. In fact, much of the predictive information in macroeconomic time series is due to the data revision and publication lag components.