By Julia Giese and Lucas Fuhrer
The yield curve is an important barometer of market sentiment and reflects interest rate expectations as well as different risk premia. In this post, we show how changes in demand for UK government bonds, also called gilts, may affect the shape of the yield curve. We find that demand shocks have persistent local effects on the yield curve, in particular at longer maturities and during volatile market conditions. These findings therefore indicate that investors in longer-term gilts tend to be less price-sensitive. Moreover, we find that demand shocks for one bond transmit to neighbouring bonds, while the transmission to other bonds declines with the difference in the residual maturity.
Importance of bond markets. Government bond markets play a key role for public finance and for a well-functioning financial system. In particular, yields on sovereign debt serve as an important reference for market participants to price other financial instruments and consequently also reveal relevant information for monetary policy.
Investors. Demand for government debt comes from a variety of investors and may change over time. In particular, over recent years the investor base for government bonds changed due to central banks’ asset purchase programmes. For UK government bonds, the largest investors in 2018 were insurance companies and pension funds with about 30%, followed by overseas investors with about 28% and the Bank of England with 25%. This differs considerably from pre-crisis holdings, when insurance companies and pension funds held about 75% and overseas investors about 25% (see DMO data). In the gilt market, some market participants are thought to view bonds with different characteristics as not perfectly substitutable, so-called ‘preferred habitats’ investors (eg pension funds and insurance companies with preferences for longer-dated bonds; see DMO and Vayanos and Vila (2009)). To adjust their portfolios away from their preferred allocation, preferred habitat investors require compensation.
Auctions. For the United Kingdom, the Debt Management Office (DMO) regularly sells gilts on behalf of Her Majesty’s Treasury (conventional gilts and index-linked gilts) in the so-called primary market. This is done by either issuing new securities or increasing the outstanding volume of existing securities and occurs mostly via auctions. All details of the auctioned bond (including issuance volume) are typically announced one week in advance and in the auction, so-called Gilt-Edged Market Makers can bid for the auctioned bond based on their own interest or on behalf of clients. Gilt-edged Market Makers submit competitive bids and successful bidders are allotted gilts on a bid-price basis, paying the price they bid. The highest, lowest and average prices of accepted bids are published immediately following the auction.
Empirical analysis. To identify the impact of changes in investor demand for gilts (ie demand shocks), we analyse market prices around conventional gilt auctions between 2003 and 2019, following a similar analysis for US Treasuries by Gorodnichenko and Ray (2018). Doing so offers an ideal setup to identify demand shocks for two reasons. First, all characteristics of the bond that will be issued are known by market participants in advance of the auction, in particular the issuance volume and therefore supply. Second, auctions offer an opportunity for investors to buy large quantities in one go. Consequently, price movements in the secondary market directly after the auction reflect information about demand by investors revealed in, and the prices they were willing to pay, in the auction. Such price movements may therefore be interpreted as the market reaction to demand shocks. When investor demand is strong (weak), the price for the bond will increase (decrease) and yields fall (increase). A popular and widely used indicator to measure demand by investors is the bid-to-cover ratio, which is the proportion of bids received versus the amount sold in an auction (see, for example, Beetsma, Giuliodori, Hanson and Jong (2017) and Reuters Article: “Benchmark 5-year gilt sale finds strong demand”).
In our empirical analysis, we regress daily yield changes in the auctioned bond on surprises in the bid-to-cover ratio (defined as deviations in the bid-to-cover ratio from its long-term mean) which is our measure for demand shocks. We show that positive demand shocks are associated with falling gilt yields (equivalent to an increase in gilt prices), while negative demand shocks go along with increasing gilt yields (see regression coefficients depicted in Figure 1). This result is in line with previous studies of the gilt market using older datasets (see Breedon and Ganley (2000) and Ahmad and Steeley (2008)). Moreover, we find that the results are particularly pronounced for long gilts, and to a lesser extent for short gilts, indicating that investors in these parts of the curve are less price sensitive than in the medium bucket. Put differently, as investors at the long-end of the yield curve may have preferred habitats, prices in this segment react more strongly without causing a reallocation of portfolios. In our analysis, we also differentiate between more and less volatile market conditions, and find that gilt yields are more affected by demand shocks during more volatile market conditions, as recently also documented for the European sovereign debt market (Beetsma, Giuliodori, Hanson and Jong (2017)).
Figure 1: Regression results
Note: The figure depicts the estimated daily yield change in the auctioneered gilt (in basis points) to a one standard deviation shock in the bid-to-cover ratio using different regression specifications. The regression coefficients are depicted in blue bars, while 95% significance levels (Newey-West) are illustrated in black lines. Volatile market conditions are defined as days with a CBOE Volatility Index (VIX) level above 20 percentage points. Auctions are classified according to maturity buckets: short: 0-7 years; medium: 7-15 years; long: 15 years or more. The sample period lasts from 1.1.2003 to 31.12.2018.
In a second step, we analyse whether demand shocks have a persistent effect on yields. Figure 2 illustrates the relationship between surprises in the bid-to-cover ratio and changes gilt yields over different time windows, starting from a one-day change up to a five-day change. The figure illustrates that the effects documented above are persistent or long-lasting in short and long gilts. Consistent with our previous findings, demand shocks in medium gilts are less strong and diminish after only three days.
Figure 2: Persistence of demand shocks
Note: The figure shows yield changes (in basis points) to a one standard deviation shock in the bid-to-cover ratio of a short, medium and long gilt auction using different window sizes (from one day to 5 days). Auctions are classified according to maturity buckets: short: 0-7 years; medium: 7-15 years; long: 15 years or more. One standard error (Newey-West) confidence intervals are depicted with dotted lines. The sample period lasts from 1.1.2003 to 31.12.2018.
Finally, similar to Gorodnichenko and Ray (2018) we find that demand shocks for a specific bond transmit across the yield curve. Figure 3 illustrates the response of the gilt curve at different benchmark points, ie the 1, 2, 5, 10, 15, 20 and 30-year bond, to local demand shocks in a short, medium or long gilt auction. We find that demand shocks transmit to close-by benchmark maturities almost one-to-one, whereas the transmission to other maturities is smaller. For example, a one standard deviation shock in the long end of the curve transmits by almost the same amount back to the 10-year benchmark, while the transmission to short maturities becomes considerably smaller. Vice versa, a shock to the short end of the yield curve transmits to bonds with a maturity of up to ten years, while bonds with a longer residual maturity are hardly affected.
Figure 3: Demand shocks and reactions in the yield curve
Note: The figure shows the reaction of the 1, 2, 5, 10, 15, 20 and 30Y benchmark gilt (in basis points) to a one standard deviation shock in the bid-to-cover ratio of a short, medium and long gilts auction. Auctions are classified according to maturity buckets: short: 0-7 years; medium: 7-15 years; long: 15 years or more. One standard error (Newey-West) confidence intervals are depicted with dotted lines. The sample period lasts from 1.1.2003 to 31.12.2018.
Policy implications. From a policy perspective, our results are interesting for the following three reasons. First, the analysis contributes to our understanding of price dynamics in government bond markets, in particular around government bond auctions. We show that auctions may reveal new information about investor demand, and demand shocks can have a significant effect on gilt yields. Second, central banks’ asset purchase programmes are thought to work partly through a portfolio rebalancing channel (see eg Haldane, Roberts-Sklar, Wieladek and Young (2016)). This channel relies on the existence of preferred habitat investors as they require compensation for substituting their portfolio. Our results are consistent with the existence of these investors in particular in the long end of the gilt market and hence support this transmission channel of central banks’ asset purchase programmes. Third, the results indicate that asset purchase programmes and their potential unwind might have a larger effect on the yield curve during volatile market conditions than during normal times.
Julia Giese works in the Bank’s International Surveillance Division and Lucas Fuhrer is on secondment from the Swiss National Bank.
If you want to get in touch, please email us at email@example.com or leave a comment below.
Comments will only appear once approved by a moderator, and are only published where a full name is supplied.Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.