Loggerheads

Government Debt Management at the Zero Lower Bound is a very nice and interesting paper by Robin Greenwood, Sam Hanson, Josh Rudolph, and Larry Summers.

Figure 1. Comparing Quantitative Easing and Treasury Maturity Extension, 2007–2014 ...the cumulative change in 10-year equivalents (scaled as a percentage of GDP) associated with the respective balance sheet policies undertaken by the Federal Reserve and the Treasury. Positive values increase the interest rate risk placed in public hands (Treasury policies), while negative values decrease it (typically Fed QE, but also Treasury maturity shortening in 2008–2009).

First point, what the Fed taketh away, the Treasury giveth. (Hence the title of this post). The Fed bought lots of long-term debt, with the idea that this would raise the price, lower the interest rate on the long-term debt, and thus stimulate the economy.

At the same time, however, the Treasury was selling lots of long-term debt. Interest rates are very low, and debts are high, so it's a great time to lock in low-rate financing. Homeowners and businesses are doing the same thing.


Alas, the Treasury and Fed are part of one budget constraint, so you can't have it both ways. As it turns out, the Treasury sold even more than the Fed bought, so by their calculations, during the period of QE, the private sector absorbed more long-term government debt!

...despite successive rounds of QE, the stock of government debt with a maturity over 5 years that is held by the public (excluding the Fed’s holdings) has risen from 8 percent of GDP at the end of 2007 to 15 percent at the middle of 2014. Said differently, the volume of 10-year duration equivalent debt has doubled from 13 percent of GDP to 26 percent of GDP over the same interval. Pressure to absorb long-term government debt has actually increased rather than decreased over the last six years!

The closing section, and main point, I think, is a proposal for how  Fed and Treasury should divide responsibility to avoid such loggerheads.

QE

Obviously, if the public has been holding more, not less, long term debt overall, that calls in to question if or how QE "worked.''

If  QE works, does it work down some price-pressure, portfolio-balance, segmented-market, demand-for-maturity curves? Or does it work by signaling Fed intentions about interest rates, and not at all per se? The fact that Treasury also changes maturity structure in private hands can let us sort out the two stories

But if the direct supply effects of QE have been offset by the massive expansion in outstanding government debt and the Treasury’s decision to extend the debt maturity, then what explains the large market impact of QE announcements documented in so many studies, as well as the fact that estimates of term premia on long-term bonds have been steadily driven into negative territory and remain miniscule today, as shown in Figure 3? The most natural explanation is that the Fed’s announcements about its intended asset purchases also conveyed information about its future policies, including both the likely path of future short-term rates and the Fed’s willingness to undertake further asset purchases in response to evolving economic conditions...

... there are reasons to think that announcements of Fed asset purchases may have a greater impact on term premia than comparably sized Treasury supply announcements. Consistent with this, Rudolph (2014) provides event-study evidence suggesting that Fed announcements have about twice the impact as Treasury announcements of a similar size...

I might add that event studies around announcements of future purchases tell us what the market thinks the effect of those purchases will be. Rational expectations is not a solid guide to events outside of all historical experience.

I've long wondered, if the Fed wanted to move long term rates, why did it not just do it -- buy and sell as required to nail the long term bond rate to 2%, say? It turns out this has a precedent:

A few months after the U.S. entered World War II, and in the midst of a rapid increase in government spending, the Fed and the Treasury agreed to fix the entire yield curve of Treasury securities. Three-month bill yields were limited to 0.375% and bond yields were held at 2.5%. The Fed stood ready to buy or sell any amount of treasury securities necessary to maintain this positively sloped yield curve

Optimal maturity structure: Liquidity premium vs. insurance against rate rises 

They argue for a much shorter maturity structure overall, trading the liquidity and low interest rate benefits of issuing short term debt against the insurance against interest rate rises benefits of issuing longer debt.

the main messages we take from these counterfactual exercises are (1) that the additional budgetary volatility incurred by shifting the government debt into short-term securities is less than is commonly supposed, and (2) that doing this would have allowed the government to capture liquidity premia on an ongoing basis

Here's the simulation. They compare surplus/deficit and total debt under existing debt and what would have happened if the Treasury issued only 3 month bills.

Figure 8 Debt and Deficits under Counterfactual Debt Management Plans. The counterfactual exercise measures the path of deficits and debt supposing that the U.S. Treasury had financed itself using rolling 3-month Treasury bills starting in 1952. ....

You can see with 3 month bills that interest-induced changes in deficits come sooner. But there isn't a big difference. From this they conclude that the interest-rate-insurance that comes from long term debt isn't a big deal.

I've been worrying about the interest-costs that a rise in interest rates would imply for some time now,  and advocating longer maturity structures on that basis. (For example in an earlier blog post  here and a longer paper here and most recently in "monetary policy with interest on reserves" ). So what's the difference? I'd make a two little complaints

1. The  US maturity structure is quite short. Last time I put together the numbers, the US rolls over half our debt every two years. And historically, it's been much shorter. So shortening down to three months doesn't change things a lot. How would lengthening to perpetuities work here?
2. The danger is a large debt to GDP ratio and the risk of a rate rise. Now we have $18 trillion of debt, so interest rates rising to 4% means $760 billion more deficits.  The graphs show two important data points really. At the end of WWII we had big debt/DGP. And interest rates stayed low until the 1970s. At the end of the 1980s, we had a big rise in real rates. And a low debt/GDP ratio.  So, Russian roulette, the gun clicked twice, doesn't mean we're safe. This isn't about averages, it's about risk management. 

Still, it's a challenging calculation, and to answer it properly requires a simulation of possible interest rate paths and debt stocks. 

They opine on real vs. nominal debt, too, arguing for more nominal debt, and much else. The whole thing is a good read.