150 Years of Ten Year Treasury Yield, 100 Years of the 10yr-3mo Spread

Reader Steven Kopits opines on the CBO projection: “by historical standards [1982-2007], we might expect the 10 year rate around 5.0% for the next decade”. I thought it useful to look at the data:

First the ten year Treasury yield.

Figure 1: Ten year Treasury yields, % (blue). NBER defined peak to trough recession dates shaded gray. Source: Treasury via FRED, Shiller database, CBO (January 2025), and author’s calculations.

I’d say it ill-advised to take the sample mean from 1982-2007 as representative. Here’s the spread.

Figure 2: 10yr-3mo Treasury spread, % (blue). NBER defined peak to trough recession dates shaded gray. Source: Treasury via FRED, NBER historical series, Shiller database, CBO (January 2025), and author’s calculations.

 

 

 

3 thoughts on “150 Years of Ten Year Treasury Yield, 100 Years of the 10yr-3mo Spread

  1. New Deal democrat

    While I think comparing short and long term Treasury rates is a useful tool, it is not the be all and end all.

    In the first place, there was no inverted yield curve between the Discount or Fed Funds rates and long term government bonds at any time between 1931 and 1958, and yet there were a number of recessions during that time including the deep 1938 recession primarily caused by fiscal tightening.

    Secondly, the Fed is a human actor with concentrated economic power. That means it can be foolish or smart, and later Fed’s can learn from the mistakes of earlier Fed’s. Hence, one of the best forecasting headlines I’ve ever read: “The Fed wants to make your recession forecast wrong.” This is what Yuval Harari’s calls second order chaos, I.e., when you observe the humans, they always observe back.

    The difficulties with the yield curve is why it is not an element of ECRI’s forecasting system. Rather, they look broadly at the level of interest rates, historically found in the (former) Dow Jones Bond Index.

    In the past a correspondent gave me a heads up about historical commercial paper rates. These can be found in data going all the way back to the Civil War. Sure enough, a pronounced increase in commercial paper rates was a very good indicator for recessions ahead.

    I have not been able to trace long term treasuries back before 1924. I would be interested to see what a graph of commercial paper vs. 10 year treasuries looks like before then, or else a link to the location of the source data for that period, if Prof. Chinn is willing to do so. Because it looks like there were extended periods of inversions between 1870 and 1900 in particular – and lots of recessions too.

    Here is a link to the graph of the ancestral interest rates I have referred to in this comment:
    https://fred.stlouisfed.org/graph/?g=1D6pc

    1. Macroduck

      Exactly right about the Fed. If we are to use spread averages as a forecasting tool, we might want to consider whether the Fed has changed its behavior in managing interest rates, and when changes have occurred. Some pretty smart people think that the Fed, and some of the Fed’s relatives, went from a system of scarce reserves to abundant reserves in response to the mortgage crash:

      https://cepr.org/voxeu/columns/double-faced-demand-bank-reserves-and-its-implications

      The transition to an abundant reserve system occurred right after Kopits 1982-2007 reference period. Not only did Kopits insist on using a period of high inflation to forecast yield spreads in a period of expected low inflation, but also a period of scarce reserves to forecast yield spreads in a period of expected abundant reserves. There have been at least two structural changes in the determination in long-end rates between Kopits’s reference period and now.

  2. Steven Kopits

    I think you are highlighting some of the problems intrinsic to statistical analysis. In this case, the question is: What is a comparable period? It can certainly be debated.

    In your previous post, you went back to 1960, implying that you thought the post-1960 period was representative. That’s the sample you chose.

    I chose the 1980-2007 period for a number of reasons. First, I think that’s how far that particular time series went back. Second, I wanted to include the oil shock period, which really starts in 1973. Almost every recession after 1973, bar that of 2001 and 2020, are linked to oil shocks. So I think that’s the most comparable period. I left out the post-2007 period because short term rates were at rock-bottom, and I consider that the China Depression period, as you know. I don’t think that’s comparable to today. I also left out the post-2019 period due to the pandemic, which I also think is not entirely comparable to today. That’s leaves 1980 to 2007, but I could certainly have gone back to 1973.

    We can certainly debate the issue. And that highlights, to my mind, one of the great weaknesses of statistics for historical and trend analysis: the problem of selecting samples and comparables. Doing so involves judgements of the validity of precedent, that is, it involves a view of fundamentals, not just technicals. To do proper statistics, you need to understand the historical narrative and the associated datasets.

    Now, to return to the particulars. If I take the period 1960-2007, the 10 yr – 3 mo rate averages 1.4%. If I take 1960 to the latest data, the difference is 1.52%. And further, if I simply look at the 10-year constant rate, it was above 5% most of the time from 1965 until 2007. Moreover, the 3 mo rate from 1965 to 2007 was mostly above 3.5%.

    The CBO sees the budget deficit averaging 5.8% of GDP to 2035. I don’t see Trump cutting any deficits, so my expectation is in line with the CBOs, but I could imagine worse.

    Therefore, the CBO’s short rate forecast of 3.3% over the next ten years looks materially low compared to a historical average around 3.5%. By extension, I think it’s pretty reasonable, based on the historical record, to assume the 10-year rate will be in the 4.8-5.1% range, not 3.86% as the CBO forecasts through 2035.

    All of this begs the question of crowding out. In the run-up to the GFC, China’s holdings of US treasuries were soaring, presumably leading to low interest rates and a search for yield. By contrast, Beijing has been cutting its holdings of treasuries for the last several years, bringing the share of incremental US debt purchased by US citizens back to early 2000s levels. Thus, rather than China depressing US interest rates, it now seems to be elevating them. All this puts more pressure on the US public and other foreign entities to purchase US treasuries, thereby supporting higher interest rates. This in turn would seem to crowd out US longer term investment, notably mortgages.

    Therefore, the administration would seem to be facing some difficult choices: continued high deficits with reduced housing affordability; lower deficits with higher (corporate) taxes; lower deficits with lower spending (“Hands off my social security!”); or pressure on the Fed to print some money to monetize the debt, which would lead to inflation. Unlike the 2017-2021 period, the US now appears to be facing materially constrained fiscal options.

    https://econbrowser.com/archives/2025/01/a-recession-is-coming-yield-curve-indication
    https://fred.stlouisfed.org/series/FDHBFIN
    https://fred.stlouisfed.org/series/TB3MS#0
    https://fred.stlouisfed.org/series/DGS10/

Comments are closed.