The FRED^® Blog

The takeaway

Most economic time series reflect what we know today. That is, they provide the most current, most accurate versions of the data. But what if we want to look back at what we thought was true yesterday? That’s where ALFRED can help.

FRED’s archival data service has turned 20

ALFRED, launched in 2006, has helped many researchers better understand economic history by taking them back in time to see the data that were available on specific dates. ALFRED illuminates windows in time to help you accomplish many tasks:

• Identify and correct data errors. The January 6, 2021, release of the Economic Policy Uncertainty Index contained a 300% spike—an error that was corrected the next day.
• Compare different methods to collect data. In November 2021, Realtor.com changed how it reported the number of houses for sale. This change could impact the analysis of trends and cycles.
• Track periodic comprehensive updates. Major economic indicators such as real GDP undergo periodic comprehensive updates to reflect best measurement practices. Between 1991 and 2023 there were eight updates, each adjusting reference years for inflation.
• Improve data accuracy and completeness. Data collection is complex. Reporting timely data creates a tradeoff between accuracy and speed. Employment figures undergo two regular types of revisions: twice when additional information from employers becomes available and two additional annual benchmarking revisions to realign those data with state administrative records. These revisions make the data more accurate.

An example of data revisions captured in ALFRED

Our ALFRED graph above shows how a natural disaster resulted in large data revisions. Hurricane Harvey first made landfall on Friday, August 25, 2017, near Rockport, Texas. Hurricane Irma hit the lower Florida Keys on Sunday, September 10. A total of 87 counties, representing 7.7% of national employment, were declared federal disaster areas. Although data collection wasn’t severely disrupted, a steep decline in food services and drinking places and below-trend growth in some industries likely reflected the impact of both hurricanes.

The first release of employment data for September 2017 (blue bar) showed a decrease in employment of 33,000 persons. The second (red bar) and third (green bar) releases showed increases in employment of as many as 38,000 persons. Keeping track of data vintages is important for storytelling and analysis.

To learn more about putting ALFRED to good use, see this essay.

How this graph was created: Search ALFRED for and select “All Employees: Total Nonfarm, Monthly, Seasonally Adjusted.” Click on “Edit Graph” and select the “Edit Bars” tab. For Bar 1, change the “As-of date” to “2017-10-06.” For Bar 2, change the “As-of date” to “2017-11-03.” Adjust the date range to “2017-01-01” to “2017-11-01.” Click “Edit Graph” and change the “Units” to “Change from Year Ago, Thousands of Persons” and click “Copy to all.”

Suggested by Diego Mendez-Carbajo.

Headlines earlier this year reported that federal US debt exceeded the size of the US economy. Another way to put that: The US debt-to-GDP ratio is over 100%.

Opinions and economic analysis vary about what a sustainable fiscal path looks like in the United States. And the multiple data series in FRED that track the US debt-to-GDP ratio also vary. This post provides some clarification and guidance on understanding this ratio in its many forms and expands the descriptions from a 2025 blog post on this topic.

US public debt and intragovernmental transactions

Our FRED graph above shows three data series that show US debt as a share of GDP: two with 2025 values well above 100% and one with a 2025 value that’s noticeably lower, slightly below 100%. The names of these series are similar, so why are there differences?

Let’s start with intragovernmental transactions. To do their work, various federal government entities engage in financial activities with each other. These transactions affect certain measures of the debt-to-GDP ratio. So we need to disaggregate the components of these measures. Specifically, the two lines with higher values look much more like the third line when we remove federal debt held by agencies and trusts.

Our second FRED graph, above, shows the same three series, after subtracting that component. Here, they look much more similar. In fact, two of them are nearly identical. But the Gross federal debt as percent of GDP series is still noticeably different from the other two. So, our investigation continues…

Fiscal year versus calendar year

FRED provides notes for each series, under the graph. If you read the notes for Gross federal debt as percent of GDP, you’ll see this ratio is sourced from the gross federal debt series, which comes to FRED from the Council of Economic Advisers and follows the federal government’s fiscal year, not calendar year. (The federal fiscal year is Oct. 1 through Sept. 30 and has been for around half a century.) But the other two debt-to-GDP ratio series (GFDEBTN and FYGFDPUN) collect their debt figures from US Treasury Bureau of the Fiscal Service datasets, which follow the calendar year. Hence, the slight difference.

More debt-to-GDP pitfalls

In economics, US federal government debt is called a stock variable, whereas GDP is a flow variable. Stock variables are defined at a singular point in time, such as the end of the month. Some stock variables may actually be measured at the beginning of the period.

Flow variables like GDP provide observations per unit in time, such as per quarter or per year. So, when creating ratios that combine both these variables, one must be careful to understand precisely what the numerator and denominator are capturing. For example, the Gross federal debt as percent of GDP series has a fiscal-year numerator and a calendar-year denominator. In short, reading the notes matters!

How these graphs were created: Search FRED for series ID GFDEGDQ188S and click on the result. Select “Edit Graph” and modify the frequency to “Annual.” Use the “Add Line” tab to search for series ID GFDGDPA188S and click “Add data series.” Repeat to add FYGFGDQ188S. Return to the “Edit Lines” tab, select Line 3 (FYGFGDQ188S), then modify the frequency to “Annual.” From the “Format” tab, open the “Customize” box for all three lines. Adjust line style, line color, and other settings as you wish. Update the time range to begin 1981-01-01 and end 2025-01-01. For the second graph, start with the first graph. From “Edit Graph,” select Line 1 and enter HBATGDQ188S into the search bar under “Customize data.” Click “Add.” In the “Formula” bar, enter a-b and click “Apply Formula.” Switch to Line 2 under the “Select Lines” dropdown and repeat the subtraction process.

Suggested by Scott St. Louis and Christian Zimmermann.

The Federal Reserve has a dual mandate from Congress: stable prices and full employment.

For the first objective, what prices should the Fed keep stable? There are many to choose from. Although they’re obviously correlated, they do deviate from each other, especially in the short run.

The Fed and specifically the FOMC look at many price indexes, but their preferred measure is the personal consumption expenditures (PCE) price index. The inflation rate from this index is published monthly by the Bureau of Economic Analysis. Its advantage over the consumer price index (CPI) and the producer price index (PPI), for example, is that the PCE covers a broader set of household expenses.

Although monetary policy aims at price stability, it does not have an instantaneous effect on prices. Policy is believed to follow long and variable lags, on average, over a couple of years. So it’s really important to have a measure of prices that can be well predicted, as the FOMC is trying to influence future prices. For this reason, the FOMC primarily focuses on core PCE inflation. Core PCE inflation excludes food and energy because those two types of prices can fluctuate dramatically, because of seasonal factors or the high volatility of markets. Given the lags that the FOMC has to work with, this kind of volatility makes forecasting the path of prices much more difficult. And the FOMC is not in a position to react to short-term price fluctuations anyway.

The FRED graph above shows three series: core PCE inflation, PCE food inflation, and PCE energy inflation. It’s clear that core PCE is much more stable, while the other two fluctuate widely around it. One could modify this graph to show month-to-month inflation instead of year-to-year inflation, and that picture is even starker. In engineering parlance, the signal-to-noise ratio is much better with core PCE inflation.

How this graph was created: On FRED, find the release table for PCE price indexes by major type of product. Check the three series and click “Add to Graph.” Then click “Edit Graph,” choose units “Percent change from year ago,” and click “Apply to all.”

Suggested by Christian Zimmermann.