(AGENPARL) – Wed 12 November 2025 Mercati, infrastrutture, sistemi di pagamento
(Markets, Infrastructures, Payment Systems)
Demand and supply of Italian government bonds
during the exit from expansionary monetary policy
Number
November 2025
by Fabio Capasso, Francesco Musto, Michele Pagano, Onofrio Panzarino,
Alfonso Puorro, Vittorio Siracusa
Mercati, infrastrutture, sistemi di pagamento
(Markets, Infrastructures, Payment Systems)
Demand and supply of Italian government bonds
during the exit from expansionary monetary policy
by Fabio Capasso, Francesco Musto, Michele Pagano, Onofrio Panzarino,
Alfonso Puorro, Vittorio Siracusa
Number 71 – November 2025
The papers published in the ‘Markets, Infrastructures, Payment Systems’ series provide
information and analysis on aspects regarding the institutional duties of the Bank of
Italy in relation to the monitoring of financial markets and payment systems and the
development and management of the corresponding infrastructures in order to foster
a better understanding of these issues and stimulate discussion among institutions,
economic actors and citizens.
The views expressed in the papers are those of the authors and do not necessarily reflect
those of the Bank of Italy.
The series is available online at http://www.bancaditalia.it.
Printed copies can be requested from the Paolo Baffi Library:
Editorial Board: Stefano Siviero, Paolo Del Giovane, Massimo Doria,
Giuseppe Zingrillo, Paolo Libri, Guerino Ardizzi, Paolo Bramini, Francesco Columba,
Luca Filidi, Tiziana Pietraforte, Alfonso Puorro, Antonio Sparacino.
Secretariat: Yi Teresa Wu.
ISSN 2724-6418 (online)
ISSN 2724-640X (print)
Banca d’Italia
Via Nazionale, 91 – 00184 Rome – Italy
Designed and printing by the Printing and Publishing Division of the Bank of Italy
DEMAND AND SUPPLY OF ITALIAN GOVERNMENT
BONDS DURING THE EXIT FROM EXPANSIONARY
MONETARY POLICY
by Fabio Capasso,* Francesco Musto,* Michele Pagano,* Onofrio Panzarino,**
Alfonso Puorro,* Vittorio Siracusa*
Abstract
In response to 2022-23 inflationary pressures, the portfolio of assets held by euro-area central
banks for monetary policy purposes has gradually shrunk (quantitative tightening), after years of
rapid expansion in the Eurosystem’s balance sheet. In these circumstances, the capacity of financial
markets to continue to absorb the supply of government bonds in an orderly and efficient manner,
without any significant impact on prices, is important. Drawing on granular data from the primary
and secondary markets for Italian government bonds, this paper investigates Market Absorption
Capacity (MAC), defined as a market’s ability to respond to supply and demand shocks with limited
price impact. Our analysis shows that when Eurosystem purchases were reduced or suspended,
private investment in sovereign securities remained stable or increased, with modest and statistically
non-significant impacts on issuance costs. However, during periods of market stress, demand from
end investors tends to weaken, and issuance costs at auctions rise accordingly.
JEL Classification: E52, E58, G12, G14.
Keywords: Unconventional monetary policy, Public debt, Supply effects, Dealer intermediation,
Market stability.
Sintesi
In risposta alle pressioni inflazionistiche del 2022-2023, dopo anni di forte espansione del bilancio
dell’Eurosistema, il portafoglio di attività detenute delle banche centrali dell’area per finalità di
politica monetaria è stato gradualmente ridotto (quantitative tightening). In tali circostanze, assume
rilevanza la capacità dei mercati finanziari di continuare ad assorbire in maniera ordinata ed
efficiente l’offerta di titoli pubblici, senza che si verifichino impatti significativi sulle quotazioni.
Basandosi su dati granulari relativi ai mercati primario e secondario dei titoli di Stato italiani, questo
lavoro analizza la cosiddetta capacità di assorbimento del mercato (Market Absorption Capacity,
MAC), definita come la capacità di un mercato di fare fronte a shock di offerta e di domanda con
un impatto contenuto sui prezzi. L’analisi mostra che, nei periodi in cui gli acquisti dell’Eurosistema
sono stati ridotti o sospesi, gli investimenti privati in titoli sovrani sono rimasti stabili o sono
aumentati, con impatti modesti e statisticamente non significativi sui costi di emissione. Tuttavia,
durante fasi di tensione sui mercati, la domanda da parte degli investitori finali tende a indebolirsi,
con conseguente aumento dei costi di emissione nelle aste.
Banca d’Italia, Directorate General for Markets and monetary policy operations.
Banca d’Italia, Directorate General for Payments and market infrastructures.
INDEX
1. Introduction
2. Related literature
3. The demand for sovereign bonds:
an analysis based on trading flows on the secondary market
4. Market absorption capacity, measurement and costs
5. Robustness
6. Conclusions
References
ANNEX
1. Introduction
The outburst of inflationary pressures in 2022 and 2023 called for a massive monetary
tightening in several advanced economies, with rapid and significant increases in policy rates
and a shift from quantitative easing to quantitative tightening. Uncertainty about the
inflationary outlook and the required pace of monetary restriction generated volatility and
hindered liquidity in government bond markets, including US Treasuries and German Bunds.
At the same time, the halt in public bond net purchases by central banks radically changed the
landscape on the demand side of the market, while supply remained generally sustained. With
regard to the Italian government bond market, after several years of negative net supply, the
amounts to be absorbed by the market experienced a considerable upward shift.
Managing public debt in this context can be challenging. The capacity of sovereign markets to
absorb bond supply (Market absorption capacity, MAC) in such a rapidly changing
environment is key. MAC can be defined as the ability of a market to withstand shifts in the
supply and demand of a certain asset class with limited price impact, i.e. in an efficient and
cost-effective way. In other words, new information is incorporated in an appropriate, prompt
and smooth manner into prices, thus allowing them to generate meaningful signals and
converge in an orderly manner to new equilibrium levels, so that market participants with
diverse trading interests can adjust, share and redistribute financial exposures.
This work contributes to a strand of literature on the topic which is still building up. While
many studies focus on the demand for government debt securities in auctions and examine the
impact of bond issuance on market yields, few have attempted to shed light on the potential
implications for the MAC stemming from central banks’ balance sheet normalisation. Our
findings are intended to support and complement those of some recent studies analysing how
marginal buyers change over time, along with changes in central banks’ purchasing regimes
(Cordes and Ferris, 2024), as well as the costs associated to the issuance of government
securities in primary markets (Albuquerque et al., 2023) and stemming from secondary market
dynamics (Ferrara, 2024).
On the back of the above, this paper looks into the MAC of the Italian government bond market,
based on primary and secondary market data from the last 10 years. In particular, we focus on
the potential effects that may stem from a lighter presence (or a full absence) of the Eurosystem
on the market in a context of unchanged or increased supply, as well as from phases of
accentuated volatility.
The work is organized as follows. Section 2 reviews the relevant academic literature. Section
3 focuses on the demand for sovereign debt and analyses the trading flows on the secondary
market in stylized scenarios characterized by different degrees of Eurosystem presence and
different volatility conditions. Section 4 investigates the costs related to market capacity to
absorb new issuance on the primary market, by analysing the price-elasticity of demand in a
sample of Italian sovereign bond auctions, as well as the issuance costs stemming from
secondary market dynamics around the auction time. Section 5 reports the results of some
robustness exercises, and Section 6 concludes.
2. Related literature
MAC is related to a rather recent strand of literature, which is becoming increasingly
established and mostly consists of empirical studies focused on the dynamics of primary or
secondary market demand. Some of them provide valuable insights into investors’ behavior in
the context of sovereign bonds auctions, central bank policy shifts, or financial stress.
Kandel et al. (1999) pioneered this research by analyzing Initial Public Offering (IPO) data to
identify key factors influencing the demand for stocks. Duffie (2010) discusses asset price
dynamics caused by slow-moving capital, presenting an illustrative model showing that
secondary market prices react sharply to supply or demand shocks due to a limited risk-bearing
capacity by investors. Lou et al. (2013) focus on anticipated and repeated shocks in secondary,
liquid markets, using market data to analyze the significant effects of these shocks on prices
and liquidity. Beetsma et al. (2016) examine the price effects of sovereign debt auctions in the
secondary market during the Eurozone crisis, finding that new public debt issuance in Italy
turns into stylized yield movements described as an auction cycle, in contrast to the evidence
obtained for Germany. Fleming and Liu (2016) examine the intraday effects of US Treasury
auctions on prices and liquidity, finding that prices decrease in the hours leading up to the
auction and recover in the hours following it, with better liquidity before the auction. Logan
and Bindseil (2019) consider the impact of a sizeable central bank balance sheet increase on
market functioning, concluding it can significantly influence market liquidity and stability.
Building on the understanding of demand shocks in liquid markets, Koijen and Yogo (2019)
advance the literature by developing a comprehensive demand system approach to asset
pricing. Their framework departs from the representative agent paradigm and explicitly models
how asset prices reflect the heterogeneous, partially inelastic demand of different investor
sectors. By calibrating this model to detailed flow-of-funds and holdings data, they show that
deviations from the representative agent benchmark can meaningfully explain cross-sectional
return predictability and amplify the price impact of large balance sheet shifts, including those
triggered by central bank interventions or regulatory changes. Their empirical approach is
particularly relevant for sovereign bond markets, where the interplay between heterogeneous
investor appetites and balance sheet capacity constraints can drive persistent yield dynamics.
Bellia (2018) studies how the supply of bonds through primary auctions affects prices and
liquidity in the secondary market, finding a general inverted ‘V-shaped’ effect on yields, quite
pronounced for Italian bonds. Bouveret et al. (2022) analyze flash crashes in EU sovereign
bond markets, finding that liquidity vanishes before crashes, causing a significant and longlasting price impact. Fleming et al. (2022) revise a previous study (Fleming and Rosenberg,
2007) to analyze how US Treasury dealers manage their positions, using a large span of dealer
position data and concluding that Treasury issuance is the main driver of weekly changes in
dealer inventory. Spronsen and Beetsma (2022) confirm the impact of unconventional
monetary policy on auction cycles by using evidence for Eurozone sovereign debt issuance.
Albuquerque et al. (2023) highlight the price elasticity of demand and risk-bearing capacity in
Portuguese sovereign bond auctions, finding that demand elasticity is strongly correlated with
yield volatility in the secondary market. Holm-Hadulla et al. (2023) examine the balance sheet
responses of investors to monetary policy shocks, finding significant variations between banks
and non-banks. Panzarino (2023) investigates Italian sovereign bonds investors behavior under
market stress, reporting a tendency to reduce exposure during stress periods which negatively
affects prices and liquidity. Shida (2023) analyzes the issuance for German government bonds
in the primary market, identifying key factors capturing auction-specific, institutional,
regulatory and financial market conditions that influence demand. Cordes and Ferris (2024) set
up a framework to point out US Treasuries marginal buyer shifts when the Fed reduces its
securities holdings, finding that with the post-pandemic balance sheet reduction, dealers,
private insurances, and foreign investors increased their purchases, while this was not the case
for hedge funds. Finally, Ferrara (2024) investigates the relationship between unconventional
monetary policy and auction cycles in the euro area, focusing on the impact of central bank
asset purchases on government bond yields in secondary markets around public debt auction
dates and finding that Eurosystem asset purchase flows on medium-term maturities help
dampen yield cycles around public debt auctions, counteracting the amplification effects of
market volatility.
Altogether, these works highlight the presence of several key factors influencing market
absorption capacity. Primary dealers’ risk-bearing capacity, the timing and frequency of debt
issuances and the behavior of investors during periods of market stress all play significant roles
in determining how well markets can absorb new supply and ultimately new information.
Moreover, the impact of large central bank balance sheets and unconventional monetary
policies on market liquidity and stability underscores the importance of coordinated policy
measures to enhance market resilience. Understanding these dynamics seems essential for
policymakers and market participants to ensure efficient market functioning and mitigate the
risks of market disruptions.
However, so far there has been limited exploration into the potential implications for MAC
arising from the normalization of central banks’ balance sheets. This work is aimed at
integrating the existing research by broadening the investigation in this direction. With this
purpose, within the scope of Italian government bonds, it examines the evolution over the last
ten years of the appetite of multiple investors and the resulting issuance costs in primary and
secondary markets, subject to the occurred shifts in the Eurosystem’s purchasing programmes.
More specific evidence from the aforementioned studies, concerned with either trading flows,
price-elasticity of primary market demand, or secondary market yield patterns around auctions,
is recalled in the respective sections of this work dedicated to these matters.
3. The demand for sovereign bonds: an analysis based on trading flows on the secondary
market
Italian sovereign bonds are traded on a multitude of trading venues and over-the-counter
(OTC). Trading is facilitated by the presence of Primary Dealers, which play a key role in
intermediating bond supply and demand. With orders arriving in large lots at irregular times,
liquidity in government bond markets is essentially determined by dealers’ ability to match
buyers and sellers and to temporarily absorb order imbalances.
This section analyzes the net trading flows intermediated by dealers on the secondary market
to investigate the buying and selling behavior of bond investors in Italian sovereign securities
in two types of periods: (i) periods with ‘lighter’ Eurosystem presence; (ii) periods
characterized by high volatility. We use data reported by primary dealers under the European
Market Activity Report (EMAR),1 which covers trades in Italian sovereign securities
negotiated by dealers with bond investors in the secondary market, i.e. within the so-called
dealer-to-customer segment. Investors are classified into six categories: non-dealer banks, asset
managers, hedge funds, insurance companies and pension funds, non-financials (including
corporate and retail investors) and public entities. To investigate changes in investors’
purchasing behavior we adopt the following specification:
𝑁𝑒𝑡𝐵𝑢𝑦𝑖𝑡 = ∑ 𝛽1,𝑖 𝟙𝑖 × 𝑁𝑜𝐶𝐵𝑡 + ∑ 𝛽2,𝑖 𝟙𝑖 × 𝐻𝑖𝑔ℎ𝑉𝑜𝑙𝑡 + 𝑋𝑖𝑡 + µ𝑖 + µ𝑞 + 𝜀𝑖𝑡
where the dependent variable, NetBuy, is the difference between the amount of securities
bought and sold by sector i from primary dealers on day t (and expressed in euro billions).2
NoCB is a dummy that takes the value of one if it is a period with reduced or no central bank
purchases. This variable identifies three sub-periods in which Eurosystem purchases were: (i)
completely absent (i.e. before March 2015); (ii) ‘virtually’, or close to, zero (i.e. the periods
January 2019–October 2019 and July 2022–February 2023, when there were no net purchases
but only reinvestments of redemptions); (iii) particularly low and negative (i.e. from March
2023 onwards, when there were no net purchases and only partial reinvestments of
redemptions). HighVol identifies periods characterised by high volatility (i.e. if above the 90th
percentile of the sample distribution). X includes the daily series of gross issues of Italian
securities on the primary market, as well as lags of the same series, which are added as control
variables since secondary market trading flows may be affected by Treasury issuance. 3 We
include sector and quarter fixed effects to control for time-invariant sector characteristics and
time trends. The sample period spans from January 2014 to January 2024.
The dataset contains all transactions involving the primary dealers in the Italian sovereign bond market, thus
providing an extensive, albeit not full, picture of the secondary-market activity. It provides a sector classification
for each counterparty in a trade, which enables to disentangle trading activity among different type of investors,
regardless of whether it is a resident or a foreign entity. The reporting scheme is consolidated at the European
level. See Euro Market Activity Report (EMAR) for more details.
The trading activity is reported at the nominal amount (par value) therefore the net purchases variable does not
reflect valuation effects related to changes in bond prices, but only the difference between actual sales and
purchases made by market participants.
Bonds maturing in investors’ portfolios do not constitute sales transactions and are thus not observed in the
primary dealer reporting.
Results are reported in Table 1. Columns (a) and (b) display the coefficients 𝛽1,𝑖 and 𝛽2,𝑖
associated with changes in sectors’ net purchases during periods of limited central bank’s
presence and high volatility, respectively. The coefficients on the NoCB dummy (Table 1,
column a) are not statistically significant or positive, suggesting that market participants tend
not to significantly change, or at most to slightly increase, their (net) purchases of securities in
months with limited presence of the Eurosystem. In particular, the types of investors tending
to increase net purchases in these months are asset managers and (non-dealer) banks. By
contrast, during periods of high volatility private investor trading flows show some marked
divergences. As shown in Table 1 (column b), for some sectors the coefficients on the HighVol
dummy tend to be statistically different from zero and show opposite signs. In times of turmoil,
some non-bank investors, such as hedge funds and asset managers, become net sellers, while
banks significantly increase net purchases. These findings underscore that, in stressed
conditions, government bond markets are not exempt from significant liquidity imbalances and
large one-sided flows (FSB, 2022), which could put considerable pressure on dealers’
intermediation capacity and adversely affect MAC
Table 1: Net purchases from primary dealers
Changes in net purchases from dealers, bn
(Non-dealer) Banks
Asset managers
Hedge funds
(a) periods with limited, or no,
central banks’ purchases
(b) stressed/high volatility periods
(>90th perc.)
0.0768*
(1.6702)
0.1165***
(2.6701)
0.0337
(0.8478)
0.3434***
(5.6779)
-0.1650***
(-3.2655)
-0.1829***
(-3.8696)
Insurances and pension
funds
0.0026
-0.0334
Non-financials
(0.0721)
-0.0122
(-0.3450)
(-1.2717)
-0.0159
(-0.6824)
Public entities (inc.
foreign CBs)
-0.0535
-0.0393
(-1.3953)
Sector FE
Quarter FE
Control variables
R-squared
Observations
(-1.1941)
0.1682
15443
The table reports the estimates of the coefficients β1,i and β2,i from specification (1), which are associated with
changes in the net purchases of the sectors during periods of limited central bank presence and high volatility,
respectively. Newey-West heteroskedasticity consistent standard errors are shown in parentheses. Data begins
in January 2014 and ends in January 2024. *, **, *** indicate significance at the 10, 5 and 1 percent level,
respectively.
The findings are robust to different specifications and a number of robustness checks (see
Section 5). The use of different percentiles to identify high volatility periods also yields
consistent results. Interestingly, when a lower percentile is chosen to identify high volatility
periods (i.e. if the threshold is set at the 80th, 70th, 60th percentile, instead of 90th as in the
baseline specification) investors’ trading flows become less divergent. These findings are in
line with related studies showing that the trading behavior of fixed-income investors may differ
across sectors and in response to past changes in yields. Sectoral heterogeneity may even
increase during crisis periods, also reflecting ‘ different exposure to liquidity risk across firms
(see, e.g., Timmer, 2018, Czech and Robert-Sklar, 2019, Panzarino, 2023).4
As anticipated, the sample period of our analysis covers three different monetary policy phases
where the Eurosystem net purchases on the secondary market were low or completely absent.
In order to capture the differential effect of distinct monetary policy phases, we run an
alternative specification where we include three indicator variables to uniquely identify these
periods (and account for potential differences in trading activity by market participants).
Specifically, we include three dummies to identify the following periods: (i) January 2014 –
March 2015 (no central bank purchases), (ii) January 2019–October 2019 and July 2022–
February 2023 (zero net purchases, full reinvestment of redemptions), (iii) March 2023–
January 2024 (negative net purchases, due to partial reinvestment of redemptions).
Table 2: Net purchases in periods with limited central banks’ presence
Changes in net purchases from dealers in periods with limited central banks’ purchases, bn
(d) negative
(b) no CB
(c) zero net
(a) all
purchases
purchases
purchases
(Non-dealer) Banks
0.0768*
0.0786
-0.0003
0.2649***
(1.6702)
(1.0024)
(-0.0053)
(2.7627)
Asset managers
0.1165***
0.0722
0.1267**
0.2264**
(2.6701)
(1.1217)
(2.1915)
(2.4200)
Hedge funds
0.0337
-0.0454
0.1275**
0.0570
(0.8478)
(-0.8148)
(2.4095)
(0.6149)
Insurances and pension funds
0.0026
-0.0003
0.0097
0.0642
(0.0721)
(-0.0064)
(0.2166)
(0.7982)
Non-financials
-0.0122
0.0113
-0.0154
0.0318
(-0.3450)
(0.2094)
(-0.3462)
(0.3952)
Public entities (inc. foreign CBs)
-0.0535
0.0932
-0.1295***
-0.0525
(-1.3953)
(1.5376)
(-2.8134)
(-0.6313)
Sector FE
Quarter FE
Control variables
R-squared
0.1742
Observations
15443
The table reports, in column (a), the estimated coefficients β1,i from specification (1), which are associated to
changes in net purchases by the sectors during periods of limited central bank presence. Columns (b), (c) and
(d) report results from an alternative specification that incorporates three indicator variables to differentiate
among distinct monetary policy phases. Newey-West heteroskedasticity consistent standard errors are shown
in parentheses. Data begins in January 2014 and ends in January 2024. *, **, *** indicate significance at the
10, 5 and 1 percent level, respectively.
The results are presented in Table 2 and show that the main findings are generally confirmed
across all sub-periods, with market participants tending, on average, not to change
Recent studies on the trading activity of financial institutions in fixed-income assets shows that the demand for
securities is generally elastic to changes in yields and that the response of fixed income investors to past returns
may differ across sectors. More generally not all investors have access to the same information, follow the same
trading strategy, take the same investment horizon, or have the same balance sheet structure. See also Panzarino
(2023) for a review of the relevant studies focusing on the trading behavior of different types of bond investors.
significantly, or at most to increase slightly, their net purchases of securities. The results also
show that the participation of bond investors in the secondary market was not the same in all
sub-periods. Bond purchases were generally higher in the most recent phase, from March 2023
onwards (column d), suggesting that various types of private investors revived their appetite
for securities after the Eurosystem started to shrink its balance sheet. Net purchases were higher
for banks and investment funds in particular, although the investor base may be broader than
the data suggest. For example, the figure for banks may also reflect purchases made by the
sector on behalf of their retail clients,5 whose demand for government bonds has been
particularly strong in the last year (see Bank of Italy, 2024).
The results also provide further evidence of the great diversity and importance of non-bank
investors in government bond markets, whose footprint has increased significantly in recent
years globally (see, e.g., Eren and Wooldridge, 2021). Asset managers and hedge funds, which
are traditionally highly sensitive to changes in yields (Panzarino, 2023), are among the most
active investors within the non-bank sector, although net purchases by the latter have been
more muted in the latter part of the sample than in earlier periods. Such findings highlight that
the role of marginal buyers may change over time, in line with evidence from other markets.
Cordes and Ferris (2024), for example, recently showed that during the recent ’balance sheet
reduction phase’ initiated by the Federal Reserve (FED) in June 2022, households purchased a
large share of US Treasury securities no longer held by the FED, while hedge funds did not
engage in such purchases to the same extent as observed in the past (e.g., during the balance
sheet reduction phase in 2017-19). Moreover, our findings corroborate the procyclical behavior
exhibited by certain market participants, notably hedge funds, whose flows can shift, also
abruptly, in response to prevailing market conditions.6 Monitoring hedge funds flows holds
particular significance from a central banking perspective, given the increasingly prominent
role hedge funds play in the liquidity and overall functioning of global sovereign bond
markets.7
One feature of our dataset is captures who is executing the trade, while not necessarily who the beneficial owner
is. For example, a bank might execute a trade on behalf of a retail client. For this reason, we cannot rule out the
possibility that the large purchases of Italian government bonds by the retail sector in recent months may have
been reflected in the figure for banks in our database.
See, e.g., Brandt et al. (2017).
See Kolokolova et al. (2018).
4. Market absorption capacity, measurement and costs
In this section we look into the costs associated to the MAC of Italian government securities,
by focusing on the placement of securities on the primary market and their impact on market
prices. We first follow a price–elasticity approach to study primary dealers’ demand in auctions
of Italian sovereign bonds. The analysis exploits granular data related to the bids submitted by
dealer banks in auctions and studies their demand curves under different market scenarios. We
then focus on the secondary market and examine the impact of the Italian Treasury auctions on
bond yields. Consistently with previous studies, we find that prices decrease in the hours
preceding auction and recover afterwards, suggesting that supply shocks generate price
pressures, which might put a strain on the MAC.
4.1 The price-elasticity of demand in auctions of Italian sovereign securities
Issuing bonds via auctions is by far the most important financing method for the governments
of advanced economies (Shida, 2023). We focus on the primary market demand for Italian
government bonds and study the behavior of primary dealers in auctions, by analyzing the
elasticity of their bids in accommodating a supply shock (in the short run). Following an
approach similar to Albuquerque et al. (2023), our elasticity measure is based on the slope of
the empirical demand curves. With a particular focus on the market absorption capacity, the
elasticity measure is computed as the marginal decrease in the bond’s price demanded by
investors due to a marginal increase in the demanded quantity8. A common theoretical
assumption is that the price elasticity of demand for an asset is infinite in a perfectly
competitive market, which indicates that investors can absorb any supply shock at the
equilibrium price (Albuquerque et al., 2023). Existing research based on the reaction of market
prices to supply and demand shocks has however questioned this assumption and documented
that the demand for financial assets is not perfectly elastic (Duffie, 2010; Albuquerque et al.,
2023). According to these studies, the evidence suggests the existence of an implicit cost that
primary dealers charge to the issuer to absorb the bond supply, which is connected to their
limited risk-bearing capacity. The price–elasticity of demand in auctions could therefore
provide an indication of the additional issuance costs borne by the government for the
placement of securities on the primary market and, more broadly, of the markets’ capacity to
absorb new issuances. Italian government medium and long-term bonds are mainly placed via
public, marginal price auctions, in which all allocated bids are awarded at the same price (the
marginal one). Auctions are carried out by the Bank of Italy – on behalf of the Italian Ministry
of Economy and Finance – and are restricted to market makers (primary dealers). In marginal
price auctions, the amount placed is determined discretionally by the Italian Treasury, within a
minimum and maximum amount announced in a press release some days before the auction.
The lowest price among the awarded bids is the auction (marginal) price, which is then applied
to all the allotted bids9.
In the context of bond auctions, low elasticity (as defined below) implies that primary dealers
are able to absorb an increase in the quantity of the bond supplied by asking only a slight
The choice in computing the elasticity as shown above is to emphasize the price effects produced by an additional
quantity. Applying the common definition of elasticity to the context of government bond auctions, elasticity is
usually computed as the marginal increase in the quantity demanded by investors for a marginal decrease in the
bond’s price, as in Albuquerque et al. (2023).
The framework ruling Italian bond auctions execution is described in detail in the Appendix.
decrease in the price at issuance. This makes low elasticity a desirable feature for the bond
issuer. Indeed, a low (absolute) value of the elasticity means that only small price decreases
are associated with (relatively) large increases in offered quantities; in other words, a supply
shock is absorbed by demand without much of a price decrease. The use of the indicator of
price elasticity to study primary dealers’ behavior in government bond auctions follows an
approach similar to Albuquerque et al. (2023), who studied the primary market for Portuguese
government bonds. More specifically, the elasticity measure is defined as the percentage
decrease in the cut-off price that one would observe for a marginal increase in quantity10:
𝐸𝑙𝑎𝑠𝑡. =
𝛥𝑃𝑟𝑖𝑐𝑒 𝑜𝑓𝑓𝑒𝑟𝑒𝑑 𝑏𝑦 𝑃𝑟𝑖𝑚𝑎𝑟𝑦 𝐷𝑒𝑎𝑙𝑒𝑟𝑠 (𝑎𝑠 % 𝑜𝑓 𝑡ℎ𝑒 𝑐𝑢𝑡 − 𝑜𝑓𝑓 𝑝𝑟𝑖𝑐𝑒)
𝛥𝑄𝑢𝑎𝑛𝑡𝑖𝑡𝑦 𝑑𝑒𝑚𝑎𝑛𝑑𝑒𝑑 𝑏𝑦 𝑃𝑟𝑖𝑚𝑎𝑟𝑦 𝐷𝑒𝑎𝑙𝑒𝑟𝑠 (𝑎𝑠 % 𝑜𝑓 𝑡ℎ𝑒 𝑎𝑙𝑙𝑜𝑡𝑡𝑒𝑑 𝑞𝑢𝑎𝑛𝑡𝑖𝑡𝑦)
In marginal price auctions, as is the case for Italy, elasticity is mostly relevant when it is
measured near the so-called ‘cut-off price’, since the latter is the only price determining the
cost at issuance for the whole quantity supplied. Therefore, for each auction of interest, we
compute this measure of elasticity using the auction demand curve and focusing, in particular,
on the primary dealers’ bids placed around the allotted quantity. Specifically we consider the
bids corresponding to a cumulative demand positioned in the range between –10% (satisfied
bids) and +10% (unsatisfied bids) of the quantity allotted by the Treasury (Figure 1). An
alternative section of the demand curve for calculating the demand elasticity has also been
taken into account for robustness check purposes, since the choice of one methodology over
another might provide different information (see Section 5).
Figure 1: Illustrative auction demand curve
The figure reports an illustrative empirical auction demand curve, obtained by sorting the auction bids in
decreasing order of price and plotting them in a chart with the price on the vertical axis and the cumulative
quantities requested for each price on the horizontal axis. Source: authors’ calculations on Bank of Italy data.
We rescaled the value by multiplying by 10, to facilitate the presentation of results
For each auction, we obtain elasticity by multiplying the slope of the selected section of the
curve – derived from a linear regression model of Price (P) on Quantity (Q) – by the ratio of
the quantity determined in auction to the cut-off and the cut-off price. In this way we obtain an
a-dimensional measure of elasticity that can be compared across time and across securities.
We analyse the elasticity of demand for Italian 5- and 10-year benchmark government bonds
using data derived from 228 auctions conducted between December 2013 and November 2023.
We focus on these specific maturities, as they exhibit the highest number of auctions and the
lowest volatility of elasticity over the sample period. To obtain a monthly elasticity value, the
simple average of the elasticity of the 5-year and 10-year bonds auctions was calculated. The
dataset includes all the bids submitted by each primary dealer in each auction.
Figure 2: Elasticity over time
Quantitative expansionary
monetary policy
Elasticity
Elasticity
Oct-23
Dec-23
Jun-23
Aug-23
Feb-23
Apr-23
Dec-22
Jun-22
Oct-22
Aug-22
Feb-22
Apr-22
Dec-21
Jun-21
Oct-21
Apr-21
Aug-21
Oct-20
Feb-21
Dec-20
Jun-20
Aug-20
Feb-20
Apr-20
Dec-19
Jun-19
Oct-19
Aug-19
Apr-19
Oct-18
Feb-19
Dec-18
Jun-18
Apr-18
Aug-18
Oct-17
Feb-18
Dec-17
Jun-17
Aug-17
Feb-17
Apr-17
Dec-16
Jun-16
Oct-16
Apr-16
Aug-16
Oct-15
Feb-16
Dec-15
Jun-15
Aug-15
Feb-15
Apr-15
Dec-14
Jun-14
Oct-14
Aug-14
Feb-14
Apr-14
Dec-13
The values in the figure are obtained as a tenfold increase in the monthly elasticity computed according to
equation (2). The time series thus proxies the percentage decrease in the cut-off price that would have been
observed if the allotted quantity increased by 10% in each auction. Source: authors’ calculations on Bank of
Italy data.
Figure 2 illustrates the evolution of the price elasticity measure over the sample period. Given
the focus of our analysis, the chart also reports the periods characterized by the expansionary
interventions of the Eurosystem. Between 2015 and 2017, the elasticity stabilized around
subdued levels. This period coincides with the onset of the Eurosystem’s Public Sector
Purchase Programme, which signalled an extraordinary accommodative monetary policy along
with a gradual improvement of the Italian macroeconomic outlook. The descriptive analysis
seems also to reveal few spikes coinciding with specific exogenous events. The highest
elasticity value (around 0.4) in the series is observed in May 2018. This outlier11 can be
attributed to the rise in the political uncertainty in Italy, which was associated to heightened
volatility and a sharp increase in the sovereign bonds’ risk premia.12 The period between
The evidence that the observation relating to May 2018 qualifies as a structural breakpoint of the series is
provided by the test reported in the Table A.3 of the Annex.
Between the end of May and the start of June 2018, coinciding with uncertainty around the formation of a new
government in Italy, tensions on the Italian government securities market heightened, driving up yields and
following a material deterioration in market liquidity conditions (Bouveret et al., 2022).
January and October 2019 witnessed a reduction in central banks’ interventions. The second
spike in the series occurs in July 2019, when elasticity reaches a value equalling 1.5 times the
series average, coinciding with an exogenous shock (government crisis) that led to an increase
in volatility and risk premia for Italian bonds. Then, in February-April 2020, a quick upsurge
in elasticity coincided with the first wave of the COVID-19 pandemic, after which a return to
the mean has been observed. Remarkably, the first talks and the final allocation of Next
Generation EU funds, occurred respectively in July 2020 and July 2021. Starting from July
2022, the cycle of ECB rate hikes did not exert a substantial influence on the trajectory of
elasticity. A qualitative analysis of the elasticity values suggests that the metric has been
particularly sensitive to exogenous shocks over the sample period, while remaining almost
invariant overall following monetary policy tightening, when large–scale asset purchase
programs have been scaled back or suspended.
On the back of the variability shown by elasticity across time, we also conduct a regression
analysis to check whether changes of the estimated elasticity indicator are associated to periods
with: (i) limited central bank purchases and (ii) higher volatility. As outlined in previous
studies, price elasticity is strictly connected to primary dealers’ capacity to absorb new
issuance. This ability is likely influenced by several factors, like the composition of their
existing inventory as well as the capability (and willingness) to warehouse the issued securities
in their portfolios before they are absorbed by the broader financial system (Lou et al., 2013,
Fleming et al., 2022) (see Section 3.2). These components are dealer-specific and difficult to
be captured without access to individual trading books, which are generally unobservable.
However, previous research suggests that the ability of dealers to absorb supply shocks tends
to be driven to some extent by common factors affecting their risk appetite. In particular, when
volatility is low, dealers tend to have a greater capacity to warehouse securities (Logan and
Bindseil, 2019, Holm-Hadulla et al., 2023).
To investigate whether the exit from monetary policy purchase programs could be associated
with a higher cost in the placement of government bonds on the primary market, we adopt the
following specification:
𝐸𝑡 = 𝛼 + (𝛽 ∗ 𝑁𝑜𝐶𝐵𝑡 ) + (𝛾 ∗ 𝐻𝑖𝑔ℎ𝑉𝑜𝑙𝑡 ) + 𝜀𝑡
Where the dependent variable, 𝐸𝑡, is our elasticity measure, 𝑁𝑜𝐶𝐵𝑡 is a dummy variable which
is equal to one in months of reduced or no central bank purchases under PSPP and PEPP
programs (and zero otherwise), and 𝐻𝑖𝑔ℎ𝑉𝑜𝑙𝑡 is a dummy variable which identifies periods of
high volatility. Data are expressed on a monthly basis and the sample period runs from
December 2013 to November 2023. The aim of the analysis is to identify changes in the price
elasticity of demand, by distinguishing phases with reduced central bank purchases of
government bonds on the secondary market and heightened volatility.
The reduction or absence of central banks’ purchases on the secondary market tends to have a
negligible effect on the price elasticity of demand in auctions. 13 As represented in Table 3,
during the periods of reduced or absent Eurosystem’s purchases, independently of the
percentile chosen to capture the level of market volatility, there are no (statistically) significant
changes to the elasticity. This pattern is overall consistent with the governance of central bank
As represented in Table 3 by the low value of the estimate for the coefficient 𝛽.
purchase programs, aimed at preserving market neutrality, in particular during public auction
events.
Table 3. Regression results.
Volatility > 50th perc.
HighVol (50 )
Volatility > 70th perc.
Volatility > 90th perc.
HighVol (70th)
HighVol (90 )
R-squared:
0.053673
R-squared:
0.081815
R-squared:
0.090609
The table provides the results of the regression model Et = α + (β ∗ NoCBt ) + (γ ∗ HighVolt ) + εt. Et is the
elasticity series. NoCBt is a dummy variable which is equal to one in months of reduced or no central bank
purchases under PSPP and PEPP programs (and zero otherwise). HighVolt is a dummy variable which is equal to
one when volatility on the secondary market, measured as the standard deviation of the yields of the ten-year
Italian benchmark bond, exceeds the 50th, 70th or 90th percentile of its daily volatility distribution (and zero
otherwise). Data are expressed on a monthly basis. The sample period runs from January 2014 to November 2023.
*, **, *** denote significance at 1%, 5%, and 10% confidence level, respectively.
By contrast, periods of higher volatility are associated to an appreciable increase in the
elasticity, consistently with previous studies (Albuquerque et al., 2023). Table 3 shows the
regression using alternative volatility measures as regressors, based on different percentiles of
the volatility distribution. The impact of volatility on our estimates magnifies as volatility
increases, with higher levels of significance of the coefficient associated with volatility. The
results are robust to different specifications (See Section 5).
4.2 An analysis based on yield patterns around auctions
Another strand of the literature focusing on debt issuance costs examines how the bond supply,
via auctions on the primary market, impacts prices on the secondary market. A number of
empirical works14 shows that supply shocks in government securities, even when fully
anticipated,15 have temporary effects on price dynamics in the secondary market. These studies
document the existence of a so-called ‘auction cycle’, where bond yields tend to increase before
auctions and to decline afterwards, following an inverted ‘V-shaped’ pattern. A pattern of the
kind has been demonstrated for several advanced economies, including the United States (Lou
et al., 2013, Fleming and Liu, 2016) and the euro area countries (Beetsma et al., 2016).
Previous research suggests that these patterns are mainly attributable to two factors: the limited
risk-bearing capacity of dealers and the ‘imperfect’ capital mobility of final investors. Primary
dealers are expected to actively participate in all Treasury auctions, by submitting competitive
See, for example, van Spronsen and Beetsma (2022), Fleming and Rosenberg (2007), Lou et al. (2013), Beetsma
et al. (2016), Fleming and Liu (2016).
Auctions’ timing and issuances’ size are typically known days in advance; in an efficient market one would
expect no predictable bond price or yield movements around auctions (Beetsma et al., 2016).
bids. In turn dealers require to be compensated for the risks associated to the auction-driven
inventory changes (affecting their trading portfolio), given that they are risk adverse and that
their capital is costly.16Additionally, the magnitude of auction cycles might be linked to the
demand of end-investors. Beetsma et al. (2016) suggest that auction cycles would be smaller
if it is easier for primary dealers to unload their inventory of the newly issued security. Issuers
do indeed rely on the capability (and willingness) of primary banks to warehouse bonds before
they are ‘absorbed’ by the broader financial system. The investor base is therefore a factor that
influence the behavior of dealers in government bond auctions, with implications for yield
dynamics on the secondary market.17 The magnitude of the auction cycle is then indicative of
the markets’ capacity to absorb new issuances, which is the focus of this analysis.
Figure 3: Yield movements before and after auction
The figure reports the average of 𝑦𝑡 − 𝑦0, where 𝑦𝑡 is the yield of the ten-year BTP (onthe-run) t minutes away from the auction, and 𝑦0 is the yield at the time of the auction
(11:00 a.m.). Yield differences are measured in the three-hour window surrounding the
auction time, for both auction (orange) and non-auction (blue) days, and expressed in
basis points. Shaded areas are 90% confidence intervals. The sample period runs from
January 1, 2014 to January 31, 2024 and includes 110 auctions.
To assess the impact of Treasury auctions on secondary market yields, we present the results
of an event study analysis.18 Figure 3 reports intraday yield movements of the 10 year BTP
(on-the-run) in the three-hour19 window surrounding the auction, for both auction and nonauction days. Specifically, the figure reports yield differences calculated as the simple
difference between the yield of the bond t minutes away from the auction, and the yield
To hedge the risk they are about to acquire, dealers (short) sell securities in advance in the secondary market
(i.e. before auctions), exerting downward pressure on bond prices. The compensation comes in the form of higher
auction yields from which the dealers generate trading profits (see, for instance, Fleming and Rosenberg, 2007).
According to Lou et al. (2016), for instance, a large fraction of potential end-investors in U.S. public debt are
passive investors that do not stand ready to absorb new debt issues.
In the vein of Fleming and Liu (2016).
Fleming and Liu (2016) use a larger time window, that goes from minus four hours to plus four hours of the
auction time. In line with Bellia (2018) we focus on a smaller window for two reasons: the first is related to the
auction time in Italy (11 a.m.); the second is related to the high price volatility and bid-ask spread at the beginning
of the day. This effect is since not all the dealers’ quotes are present immediately after market opening.
observed on the same bond at the time of the auction (11:00); t ranges from -90 (one hour and
half before auction) to +90 (one hour and half after auction). Yield differences are expressed
in basis points and computed for each five-minute interval within the time window. Shaded
areas are 90% confidence intervals. Data are from the MTS market, which is the most liquid
trading venue in the interdealer segment for the Italian sovereign securities. The sample period
runs from January 1, 2014 to January 31, 2024.
The results show the presence of an intraday yield pattern around the auction time. On auction
days, bond yields tend to rise in the run-up to the auction and to fall back around their original
level after the auction. By contrast, on non-auction days, no clear patterns seem to emerge:
yield differences are miniscule over the same window (and not significantly different from
zero). The pricing patterns we observe are thus unique to auction days and centred around the
auction time.
We complement our event study analysis by conducting regressions to better control for the
potential presence of confounding factors occurring during the event window. We adopt the
following specification:
∆𝑦𝑑𝑡 = 𝛽1 𝐴𝑈𝐶𝑑 × 𝐵𝑒𝑓𝑜𝑟𝑒𝑡 + 𝛽2 𝐴𝑈𝐶𝑑 × 𝐴𝑓𝑡𝑒𝑟𝑡 + µ𝑡 + µ𝑞 + 𝜀𝑑𝑡
where the dependent variable (∆𝑦) is the difference between on-the-run yields quoted t minutes
from auction and at the time of auction (11 am) on day d; Before (After) is a dummy variable
equal to 1 in the one-hour time window before (after) the auction, i.e. from 10:00 am (11:00
am) to 11:00 am (12:00 am); AUC is a dummy variable indicating auction days. In order to
investigate whether auction cycles change during phases of reduced Eurosystem presence
and/or in periods of higher volatility, we expand our baseline specification by adding two
dummy variables, NoCB and 𝐻𝑖𝑔ℎ𝑉𝑜𝑙, which take the value of 1 if it is a period of limited
presence of central banks’ purchases or higher volatility, respectively. We also add hour and
year-quarter fixed effects to control for intraday dynamics as well as other time-varying factors
that may affect debt issuance costs. The sample period runs from January 1, 2014 to January
31, 2024. The results are reported in Table 4.
Regression estimates corroborate the results of the event study analysis and document the
existence of an auction cycle in the Italian sovereign bond market. There is a clear and
statistically significant downward price pressure around auction time: ten-year yields tend to
be half a basis point lower before and after the auction.20 The effect is temporary and slightly
asymmetric: after the auction, yields revert to levels that are slightly higher than their original
ones. The results also imply a ‘hidden’ issuance cost for the Italian Treasury, estimated to be
around 130 million euros for the issuance size in 2023.21 As outlined in previous studies,
Results are robust to the choice of different time interval before and after auction time (e.g., 10:30-11:00, 10:1510:45; see Section 5).
We compute the auction-induced additional issuance cost for the Italian Treasury as in Beetsma et al. (2016).
Based on the findings of our analysis, the estimated average price pressure effect (around auctions time) for the
10-year securities is approximately 0.5 basis points. Hence, referring to the total amount issued in 2023 for these
securities (around 45 billion), we compute an additional annual interest payment of around 2.26 million (i.e. total
amount allotted, 45 billion, times 0.5 basis points). We then multiply this number by the modified duration of a
10-year benchmark bond at the end date of our sample period. The additional debt issuance component that arise
purely because of the auction cycle is than equal to 18 million euros. Assuming that all the debt issued in 2023 by
although not being the major components of the total financing cost borne by governments,
issuance costs associated with auction cycles are not (economically) negligible and may
provide insights on the market capacity to ‘absorb’ new issuances.
Table 4. Auction effects on yields.
Yield differences around
auction time, bps
AUC × Before
AUC × After
-0.5374***
(-9.3720)
-0.4463***
(-7.4564)
-0.5521***
(-9.8383)
-0.3989***
(-6.1025)
-0.3762***
(-7.1335)
-0.4722***
(-9.9133)
-0.4075***
(-7.5940)
-0.4396***
(-8.2438)
-0.3838***
(-7.0090)
-0.4159***
(-7.8412)
0.0567
(1.4787)
0.2165
(1.5671)
-0.1349
(-1.0169)
0.0493
(1.3173)
-0.4115***
(-3.7440)
0.0715
(0.6485)
AUC × Before × NoCB
0.0445
(0.3286)
-0.1433
(-1.1096)
AUC × After × NoCB
0.1721
(1.2192)
-0.1792
(-1.3113)
HighVol
AUC × Before × HighVol
AUC × After × HighVol
Hours FE
Quarter FE
R-squared
Observations
0.0003
95599
0.0003
95599
-0.3255***
(-2.9503)
0.0522
(0.4699)
0.0003
95599
-0.3771***
(-3.3080)
0.1059
(0.9006)
0.0003
95599
0.0003
95599
The table report the results of regression (4). The dependent variable is the difference between on-the-run yields
quoted t minutes from auction and at the time of auction (11 am) on day d; t ranges from -90, or one hour and
half before auction, to 90, or one hour and half after the auction; yield differences are from the interdealer market
MTS Italy and are expressed in basis points. Before (After) is a dummy variable equal to 1 in the 1-hour time
window from 10:00 a.m. (11:00 a.m.) to 11:00 a.m. (12:00 a.m.); AUC is a dummy variable indicating auction
days; NoCB (HighVol) is a dummy that takes the value of one if it is a month of ‘reduced’, or no, central banks’
purchases (period of high volatility, i.e. if above the 90th percentile). Data begins in January 2014 and ends in
January 2024. Newey-West heteroskedasticity consistent standard errors are shown in parentheses. *, **, ***
indicate significance at the 10, 5 and 1 percent level, respectively.
Such evidence aligns with the existing literature and is common to other government bond
markets. The impact of government security supply shock on secondary market yields has been
demonstrated both in the US (Fleming and Liu, 2016, Lou, Yan, and Zhang, 2013) and in the
euro area (Shida, 2023, Spronsen and Beetsma, 2022, Beetsma et al., 2016, Lou et al., 2013).
In most cases previous research is based on lower frequency (i.e. daily) data, which make the
magnitude of the cycles not directly comparable with our study (because of differing time
intervals of the analysis). To the best of our knowledge, the only empirical contributions that
use intraday22 data are the works of Fleming and Liu (2016) and Bellia (2018), which are
the Italian Treasury has been subject to the same additional issuance cost (of 0.5 basis points), the total cost borne
by the Italian Treasury would have been equal to around 130 million euros (500 billion times 0.5 basis points
times the average modified duration of the government securities).
Using intraday data substantially reduces the potential for confounding effects arising from unrelated events
during the day. For instance, by focusing on intervals immediately before and after the auction, our approach
avoids other timeframes that are usually linked to significant information releases, such as macroeconomic
focused, respectively, on the US treasury market and on the Italian and German sovereign bond
markets. Consistent with our findings, the authors provide evidence of a price pressure effect
around auctions, which is not present in non-auction days, with a maximum intraday yield
movement of about 0.5–1 basis points. Notably, despite differing markets and sample periods,
the results are consistent with our findings (and estimates are broadly comparable in terms of
magnitude).
Figure 4: Yield movements before and after auctions in high volatility periods
(>perc.)
Yields reported as difference between on-the-run mid-quotes t minutes from/after auction and at the time
of auction (11 am); t ranges from -90 to 90 (minutes). Yield quotes are from the interdealer MTS market
and are expressed in basis points. Shaded areas are 90% confidence intervals. The sample period runs from
January 1, 2014 to January 31, 2024 and includes 110 auctions.
announcements. However, previous studies have suggested that price pressure effects may unfold over longer
timeframes, often spanning several days, implying that our intraday framework may only capture a portion of the
total effect.
As shown in Table 4, the absence (or reduced presence) of central bank purchases is not
associated to significant changes in yield patterns around auctions (the coefficients are not
statistically different from zero). By contrast, yield patterns exhibit a different behavior during
periods of heightened uncertainty: bond yields tend to increase slightly more on average before
the auction, and they have a tendency to remain elevated for a longer period after the auction
(see Figure 4). Dealers are likely demanding higher premia when volatility is higher and the
impact of treasury auctions on bond yields is more persistent within the examined time window.
5. Robustness
This section investigates the robustness of our findings. We conduct a series of robustness
checks, with results summarized in Tables A.1–A.6. Our key conclusions remain stable across
alternative model specifications, different thresholds used to define high-volatility periods, and
various measures of elasticity. In the following subsections, we explore two aspects in greater
detail: (i) the potential interaction between central bank purchases and market volatility, and
(ii) the robustness of our estimates on auction elasticity across different maturities.
5.1 Central bank purchases and market volatility
This study analyses issuance costs under two distinct market conditions: (i) periods with limited
or no central bank purchases, and (ii) episodes of heightened market volatility. A potential
concern is that these two regimes may not be fully independent. For example, reduced central
bank interventions might coincide with calmer market conditions, such as declining volatility
or restored market functioning, potentially biasing our estimates.
To address this potential issue, we test for multicollinearity between the variables that identify
high-volatility periods and phases of reduced central bank purchases in our setting. The results
(see Table A.4 in the Appendix) indicate no significant correlation, suggesting that reduced
ECB activity does not systematically align with any volatility regime.
The institutional framework of asset purchase programs during our sample period corroborates
this finding. Monetary policy programmes, particularly the Asset Purchase Programme (APP),
were designed to meet broader monetary policy objectives, rather than to stabilise the market
in the short-term. Although instruments explicitly linked to market functioning – such as the
Transmission Protection Instrument (TPI) or the flexibility embedded in the Pandemic
Emergency Purchase Programme (PEPPflex), were announced during the sample period (June
2022) -, they had no material interference with purchase flows (excluding the “announcement
effect”). These factors reduce the likelihood of a structural link between asset purchases
intensity and prevailing market conditions during the period under study. Furthermore, our
analysis employs binary regime indicators (i.e. dummies) rather than continuous measures of
purchase flows or volatility, which helps to further reduce the risk of multicollinearity. As a
result, any component of central bank purchase activity potentially linked to volatility would
not be captured by a framework based on binary indicators since the regime describing central
bank purchases would not shift. It is also important to note that our analysis specifically focuses
on periods of non-intervention.
As an additional robustness check, we regress our measure of issuance costs (i.e. auction
demand elasticity) on both central bank net purchases and the orthogonal component of market
volatility, with the latter defined as the residual from a regression of volatility on ECB
purchases. This exogenous volatility measure retains significant explanatory power in the
model (see Table A.5), which supports the conclusion that market volatility influences issuance
costs beyond central bank activity.
5.2 Measure of price-elasticity (across different tenors)
One of the key challenges in estimating price elasticity is selecting which segment of the
demand curve to focus on. In fact, the slope of the curve can be measured in multiple stretches,
for instance considering a neighbourhood of the allocated quantity (chosen measure), as
illustrated above, or rather using: (i) all the demand price points, or (ii) only the allotted demand
price points, or (iii) the right tail of the curve, made up of the unallotted bids. The first two
alternative measures would misrepresent the quantity increase that would have been observed
if the cut-off price dropped. The third measure, instead, could potentially capture consistently
how much the price would have to decline if the Treasury were to increase the quantity sold
into the untapped liquidity. Nevertheless, since it focuses on a right-skewed neighbourhood of
the allocated quantity, the measurement may lack robustness. This is especially the case when
auctions exhibit slope clustering, e.g. when the slope (in absolute value) is quite low in the
allotted portion of demand and very high in the unallotted portion, due to particularly
opportunistic tail-bids. However, due to the popularity of this alternative measure, used by
Albuquerque et al. (2023) – who used a representative portion of this part of the curve – and
similar to that used by Kandel et al. (1999), as a further control, we have also computed
elasticity over the whole of the demand curve that lies at the right of the cut-off price. The
results are summarized in Table A.1 and A.2. They show the same pattern, albeit with different
coefficients, as those obtained with the chosen measure and, thus, do not deserve any
discussion.
Figure 5: Elasticity measures for different tenors
Quantitative expansionary monetary policy
Elasticity 5-10y
Elasticity 3-7y
Elasticity
Jul-23
Sep-23
Nov-23
May-23
Jan-23
Mar-23
Nov-22
Jul-22
Sep-22
Jan-22
Mar-22
May-22
Nov-21
Jul-21
Sep-21
May-21
Jan-21
Mar-21
Jul-20
Sep-20
Nov-20
May-20
Jan-20
Mar-20
Nov-19
Jul-19
Sep-19
Jan-19
Mar-19
May-19
Jul-18
Sep-18
Nov-18
May-18
Jan-18
Mar-18
Jul-17
Sep-17
Nov-17
Jan-17
Mar-17
May-17
Nov-16
Jul-16
Sep-16
Jan-16
Mar-16
May-16
Jul-15
Sep-15
Nov-15
May-15
Jan-15
Mar-15
Nov-14
Jul-14
Sep-14
Jan-14
Mar-14
May-14
The figure presents auction elasticity measures, as defined in equation (2), for the 3- and 7-year tenors, along with
the same indicator for 5- and 10-year BTP auctions. The series show the percentage decline in the cut-off price
that would have resulted from a 10% increase in the allotted quantity in each auction. Source: authors’ calculations
based on Bank of Italy data.
As a further analysis, the data at our disposal allow us to determine the elasticity series for
other maturities as well. In this study, we have focused on the 5- and 10-year BTP auctions,
also considering the indicative value of the 10-year maturity: both securities are continuously
offered by the Treasury (end-of-month auctions) and on the same auction day. Elasticity has
also been calculated for other tenors. Similar to the 5- and 10-year BTPs, the 3- and 7-year
BTP auctions are conducted on a continuous basis (mid-month auctions), whereas the Treasury
offers securities with maturities longer than 10 years by alternating maturities at its discretion
(15, 20, 30, and 50 years), which in this latter case makes the study of elasticity less
straightforward. The elasticity series determined as the average elasticity of the 3- and 7-year
BTP auctions shows a trend similar to that obtained for the 5- and 10-year BTP auctions (Figure
5), albeit with some differences due to the time gap between mid-month and end-of-month
auctions.
6. Conclusions
The analysis unveils that, at least so far, the issuance of Italian sovereign debt has been
smoothly absorbed by the market, despite reduced asset purchases by the Eurosystem as a result
of quantitative tightening.
An analysis of investors’ trading flows on the secondary market shows that the periods
(although limited in our sample) of ‘lighter’ Eurosystem presence did not come along with
significant changes in investors’ purchasing behaviors. During these periods, market
participants have on average kept stable or increased their bonds purchases, maintaining their
appetite for debt securities. The findings remain valid also in the latter part of our sample, from
March 2023 onwards, when the Eurosystem started to shrink its balance sheet and private
investors stepped in. By contrast, we observe significant and “asymmetric” changes in
investors’ purchasing behaviors in times of stress, when asset managers and hedge funds
quickly shy away and positive flows only stem from banks’ demand. In such context trading
imbalances are pronounced, generating high pressure on liquidity providers and potentially
hampering market absorption capacity.
Consistently with these results, the investigation of the costs associated with the market’s
capacity to absorb bond supply confirms that the lack of Eurosystem purchases has not
significantly affected issuance costs so far. The reduction or absence of central bank purchases
in the secondary market tends to have a negligible effect on the price elasticity of demand in
auctions. Primary dealers can absorb increased net issuance of securities without demanding
significantly higher yields. However, a statistically significant impact is observed during
periods of heightened volatility, when primary dealers’ activity in auctions may become less
supportive, and demand from final investors weakens. This conclusion is further supported by
evidence from an analysis of secondary market movements around auctions (auction cycles).
The absence or reduction of central bank purchases is not associated with significant changes
in yield patterns around auctions. In contrast, yield patterns exhibit a distinct behavior during
periods of heightened uncertainty; not only bond yields tend to increase slightly more on
average before the auction, but they have a tendency to remain elevated for a longer period
afterward. When volatility is elevated, dealers are likely demanding higher premia and the
impact of treasury auctions on bond yields is more visible and persistent.
The analysis also underscores the key role of primary dealers in intermediating supply and
demand for bonds in the market, thereby supporting absorption capacity by end-investors, and
ultimately market stability and liquidity. Uncertainty and volatility are factors that can trigger
material trading imbalances from end investors and threaten the capacity of market makers to
provide immediacy services.
The gradualism and predictability that have so far characterised the Eurosystem’s approach in
the current tightening cycle have been crucial, as they have helped to avoid sudden spikes in
volatility that prove detrimental to market absorption capacity.
References
Albuquerque, R., Cardoso-Costa, J. M., Faias J. A. (2023), “Price elasticity of demand and
risk-bearing capacity in sovereign bond auctions”, Banco de Portugal, Working Papers, 02
2023.
Bank of Italy (2024), Financial Stability Report, No. 1.
Beetsma, R., Giuliodori, M., de Jong, F., Widijanto, D. (2016), “Price effects of sovereign debt
auctions in the euro-zone: the role of the crisis”, Journal of Financial Intermediation, Vol. 25,
pp. 30-53.
Bellia, M. (2018), “Intraday Pricing and Liquidity of Italian and German Treasury Auctions”,
Brandt, M. W., Nucera, F., Valente, G. (2017), “Can Hedge Funds Time the Market?”,
International Review of Finance, 17(1), 1-28.
Bouveret, A., Haferkorn, M., Marseglia, G., Panzarino, O. (2022), “Flash crashes on sovereign
bond markets – EU evidence”, Markets, Infrastructures, Payment Systems, Bank of Italy, No.
Cordes, L., Ferris, E. (2024), “Who Buys Treasuries When the Fed Reduces its Holdings”,
FEDS Notes. 06 2024.
Czech, R., Roberts-Sklar, M. (2019), “Investor behavior and reaching for yield: evidence from
the sterling corporate bond market”, Financial Markets, Institutions & Instruments, Vol. 28,
Issue 5.
Duffie, D. (2010), “Presidential Address: Asset Price Dynamics with SlowMoving Capital”,
The Journal of Finance, 65, 1237–1267.
Eren E., Wooldridge P. (2021), “Non-bank financial institutions and the functioning of
government bond markets”, BIS Papers n. 119
Ferrara, F. (2024), “Central bank asset purchases and auction cycles revisited: new evidence
from the euro area”, ECB Working Paper Series No 2927.
Financial Stability Board (FSB, 2022), “Liquidity in Core Government Bond Markets”.
Fleming, M.J., Liu, W. (2016), “Intraday pricing and liquidity effects of US Treasury
auctions”, Federal Reserve Bank of New York Working Paper.
Fleming, M.J., Rosenberg, J.V. (2007), “How do treasury dealers manage their positions?”,
Staff Reports, Federal Reserve Bank of New York, Staff Report No. 299.
Fleming, M.J., Nguyen, G., Rosenberg, J.V. (2022), “How do Treasury dealers manage their
positions?”, Working Paper Federal Reserve Bank of New York.
Holm-Hadulla, F., Falk, M., Rast, S. (2023), “Bank and non-bank balance sheet responses to
monetary policy shocks”, Economics Letters, Volume 222, 2023.
Italian Ministry of Economy and Finance (MEF), Department of the Treasury (2024),
“Government Bond Auctions”, http://www.dt.mef.gov.it. Last accessed: March 2024.
Kandel, S., Sarig, O., Wohl, A. (1999), “The Demand for Stocks: An Analysis of IPO Auctions”,
Review of Financial Studies, 12, 227–247.
Koijen, R., Yogo, M. (2019), “A Demand System Approach to Asset Pricing”, Journal of
Political Economy, Vol. 127, Issue 4, pp. 1475-1992.
Kolokolova, O., Lin, M.-T., Poon, S.-H. (2018), “Too big to ignore? Hedge fund flows and bond yields”,
Journal of Banking and Finance, 87, 1-15
Logan, L., Bindseil, U. (2019), “Large central bank balance sheets and market functioning”,
Bank for International Settlements, Markets Committee.
Lou, D., Yan, H., Zhang, J. (2013), “Anticipated and Repeated Shocks in Liquid Markets”,
Review of Financial Studies, Vol. 26, pp. 1891-1912.
Panzarino, O. (2023), “Investor behavior under market stress: evidence from the Italian
sovereign bond market”, Markets, Infrastructures, Payment Systems, Bank of Italy, No. 33.
Spronsen, J.v., Beetsma, R., (2022), “Unconventional monetary policy and auction cycles of
eurozone sovereign debt”, Journal of Money, Credit and Banking, Vol. 54, pp. 169-202.
Shida, J. (2023), “Primary market demand for German government bonds”, Journal of
International Money and Finance, Vol. 137.
Timmer, Y. (2018), “Cyclical investment behavior across financial institutions”, Journal of
Financial Economics, 129, 268-356.
ANNEX
Background on Italian auctions
The information relating to the rules governing government bond auctions and the related
issuance procedures are detailed by the Ministry of Economy and Finance (MEF, 2024).
Italian government medium and long-term bonds are mostly placed via public auction to
guarantee access by an extensive group of investors and maintain a high level of competition
and transparency. At the end of every calendar year, the Italian Treasury publishes the annual
auction calendar for the next year, along with its public debt management guidelines. These
documents inform dealers well beforehand about the frequency of auction placements and the
quality profile that will guide the issuance policy during the year. In particular, the Calendar
contains all the dates of the issue press releases, of the auctions and their settlements, grouped
by bond category. Moreover, an “issue programme” is published quarterly to disclose all the
information regarding new bonds to be placed via auction and re-proposed regularly during the
quarter, together with the information regarding the offer of outstanding bonds. Prominently, a
press release is made before each auction: aside from announcing the bonds to be issued and
their characteristics, it indicates the precise minimum and maximum quantities offered in the
auction as well as all relevant dates, including the bond settlement date. The settlement date
for all government bonds is typically two business days following the auction date (t + 2).
When the settlement date of medium/long-term bonds does not coincide with the date in which
the bond’s interest begins to accrue (the interest commencement date), subscribers pay the
Treasury the corresponding accrued interest.
Government bond auctions are carried out by the Bank of Italy. Authorized dealers that are
market makers (primary dealers) have obligations, as to subscriptions in government bond
auctions and trading volumes on the secondary market, that give rise to some privileges in a
variety of other operations. Primary dealers’ bids to participate in the auctions are sent online.
Dealers can place bids for each bond offered until 11 am of the auction day. The system
automatically rejects bids beyond the deadline. Dealers can repeatedly adjust their bids,
substituting the previous ones, since the system only considers the final bid made within the