The FRED® Blog

## The many ways to calculate how much the U.S. federal government owes

What’s the debt level of the U.S. federal government? The answer isn’t as straightforward as it may seem. A quick search on FRED for “federal debt” delivers the graph above, which shows the total level of the federal debt, in millions of dollars, at a quarterly frequency since first quarter 1966. The latest figure, as of the writing of this post, corresponds to second quarter 2019 and amounts to over \$22 trillion. We can also express the federal debt as a percentage of GDP, like so:

The federal debt reached 103% of GDP in second quarter 2019. These numbers, however, don’t properly reflect the amount owed by the federal government to private bondholders, since certain federal agencies (primarily, the Social Security trust funds) also hold federal debt. These agency bond holdings are liabilities the federal government owes to itself and therefore should be netted out. This adjustment is made in a series called “Federal Debt Held by the Public,” which FRED has both in millions of dollars and as a percentage of GDP. The latter is below:

As we can see, these adjusted amounts are substantially lower than the ones previously shown. Federal debt held by the public amounted to roughly \$16 trillion or 76% of GDP in second quarter 2019. However, since the Federal Reserve Banks are actually private banks, they’re included in the government’s definition of “public.” Since Federal Reserve Banks remit their profits to the Treasury, any interest earned on their federal debt is rebated to the federal government. Thus, debt held by Federal Reserve Banks constitutes liabilities that the federal government owes to itself. FRED has a series called “Federal Debt Held by Federal Reserve Banks”:

If we deduct this value above from the federal debt level, we can create a more accurate series of federal debt held by the public, excluding the holdings by Federal Reserve Banks. (Simply subtract “Federal Debt Held by Federal Reserve Banks” from “Federal Debt Held by the Public” after making sure they’re both expressed in the same units.)

So, as of second quarter 2019, federal debt is \$13.7 trillion or 64% of GDP, much smaller than the figures we started with before netting out the holdings of federal agencies and Federal Reserve Banks.

How these graphs were created: For all but the last one, search for the series name and click on the relevant result. For the last, take the next-to-last graph, click on “Edit Graph,” add a series by searching for “federal debt held by the public as percent of GDP,” and apply formula b-a.

Suggested by Fernando Martin.

View on FRED, series used in this post: FYGFGDQ188S, GFDEBTN, GFDEGDQ188S, HBFRGDQ188S

## What's behind two different responses in the housing market?

Monetary policy affects interest rates, which affect mortgages, which affect decisions in the housing market. That may be easy to understand, but the housing data may not have such clear-cut patterns. Let’s see what FRED has to show us.

The red line in the graph is the average 30-year fixed-rate mortgage (right axis) from the early 1970s to 2019. The blue line in the graph is the ratio of housing starts built by contractors over housing starts built by owners (left axis) for the same period.

From 1985 to 2007, this ratio was generally flat, around 1.5, implying contractors built approximately 60% of housing starts. But during episodes of macroeconomic turbulence, the ratio has diverged from its historical average. But not in a consistent way… In the late 1970s and early 1980s, this ratio declined sharply, to below 1.0. This implies individual owners built more housing starts than developers during this period. But during the Great Recession, this ratio increased sharply, to over 2.0, peaking at 2.6 in 2016, which implies contractors built 72% of housing starts.

Why would the ratio plummet in the late 1970s and rise sharply in the late 2000s? In both cases, GDP declined and unemployment rose, but this housing measure behaved differently.

Maybe you’ve already seen it, but a clear difference between these two episodes is the level of mortgage rates: Rates were much higher in the 1970s and 80s and much lower leading up to and through the Great Recession. As mortgage rates go up, the ratio goes down and vice versa. A potential reason is that, as the price of mortgages increases, the cost of purchasing a new home from a contractor increases relative to the cost of building one’s own home. And, if the costs are basically the same, many would-be homeowners might choose to build their own home rather than purchase one that someone else built.

The late 1970s was a period of high inflation; in an effort to reduce inflation, the Federal Reserve imposed higher interest interest rates, which included mortgage rates. In contrast, during the Great Recession, the Federal Reserve slashed interest rates and, by extension, mortgage rates. It looks like homeowners respond to changes in these interest rates: building their own houses to try to economize on the financing during periods of high rates and purchasing new houses from contractors during periods of low rates.

How this graph was created: Search for “New Privately Owned Housing Starts” and select “Contractor-Built-One-Family Units, Thousands of Units, Seasonally Adjusted.” From the “Edit Graph” panel, under the box “Modify frequency,” select “Semiannual.” Use the “Customize Data” option to search for “New Privately Owned Housing Starts in the United States by Purpose of Construction, Owner-Built One-Family Units” and select “Thousands of Units, Seasonally Adjusted.” This latter series is now labeled as b. Under “Customize data,” type a/b into the “Formula” box and select “Apply” to get the ratio of the two series. Now select “Add Line” and search for “30-Year Fixed Rate Mortgage Average in the United States, Percent, Semiannual, Not Seasonally Adjusted.” Under the box “Modify frequency,” select “Semiannual.” Under “Format,” under the option for “LINE 2,” select “Y-Axis Positon” as “Right.”

Suggested by Matthew Famiglietti and Carlos Garriga.

View on FRED, series used in this post: HOUSTCB1FQ, HOUSTOB1FQ, MORTGAGE30US

## Take note: FRED has updated some series names

The FRED Team has just automated the process of how it names many of its data series. Because FRED aggregates data from 89 different sources, choosing the right name for any of the 627,000 data series is no small matter. Yes, the Bard wrote “A rose by any other name would smell as sweet.” But in the world of data, a confounding name can be a thorny problem.

Let’s choose a common example. The data series for the unemployment rate in the U.S. is collected by the Bureau of Labor Statistics (BLS). But the media can choose to report the data with a variety of names: national unemployment rate, civilian unemployment rate, official unemployment rate, harmonized unemployment rate, or U3.

The FRED graph below shows two series: the unemployment rate (from the BLS) and the harmonized unemployment rate (from the OECD). Why do we see only one line? Because the series are one and the same. So, what is the correct name for the unemployment rate data series? The answer depends on the source of the data. So, FRED will now display the series name as reported by the source of the data from the most comprehensive machine-readable location.

In the case of the BLS, that location is series LNS14000000. The series is accessible through the LABSTAT public database, which contains current and historical surveys and press releases. For the BLS series LNS14000000, the name of the data series is “unemployment rate,” so FRED will call it simply that: unemployment rate.

Although the FRED data series identifiers have not changed, there are 2,782 data series names that have changed. For a complete list, see this CSV file. You’ll notice that many data series in FRED related to the consumer price index now have updated names.

Suggested by Diego Mendez-Carbajo and Maria Arias.

View on FRED, series used in this post: LRHUTTTTUSM156S, UNRATE