Top

Financial Markets and Portfolio Management

Published in:

Open Access 25-11-2024

Political uncertainty and sovereign bond markets

Authors: Lukas Handler, Rainer Jankowitsch

Published in: Financial Markets and Portfolio Management | Issue 1/2025

Activate our intelligent search to find suitable subject content or patents.

search-config

AI-assisted search

Patentsearch

Off

Abstract

We analyze the effects of political uncertainty on prices and liquidity of sovereign bonds. Specifically, we investigate Italian government bonds during the European sovereign debt crisis and focus on political summits and elections. We find a significant drop in prices in combination with high illiquidity and sell-side pressure before the events. The event returns are significantly positive and followed by a positive price trend. The effects are stronger when uncertainty, as measured by the EPU index, is high and economic conditions are weak. In addition, political uncertainty also affects the primary market and we find significant costs associated with issuing sovereign bonds in highly uncertain times.

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

1 Introduction

Political news about government policies and reactions represents important information for investors around the world, suggesting that uncertainty about the outcomes of the political decision-making process is relevant for asset prices and trading activity. The recent theoretical literature provides guidance to understand these asset pricing implications. In particular, the model presented in Pastor and Veronesi (2013) shows that investors demand a risk premium for political uncertainty, i.e., a compensation for being exposed to the stream of political news around important events, such as elections. So far, researchers have focused on potential policy changes affecting firms and have primarily modeled stock price reactions. However, government policies affect many risk factors, determining the general economic conditions. Thus, the effects are not limited to stock markets and, in particular, will be very relevant for financial instruments directly linked to the activities of governments, such as sovereign bonds.

In this paper, we study whether political uncertainty indeed affects prices and liquidity in government bond markets. We explore the time period of the European sovereign debt crisis, as it is an ideal laboratory for such an analysis because investors were exposed to significant political uncertainty due to several political events focusing, almost exclusively, on how to handle the crisis. In addition, during this episode, sovereign bonds represented the most directly affected financial instruments and understanding their price reactions offers significant insights, enriching the results presented for stock and option markets in the empirical literature so far, see e.g., Kelly et al. (2016). Furthermore, these results allow us to quantify the additional financing costs for sovereign debt resulting from the degree of political uncertainty during the issuance process.

In our empirical study, we mainly focus on the Italian sovereign bond market. With a notional amount outstanding of more than two trillion EUR the Italian market is a major European bond market, comparable in size with Germany and France. Moreover, Italian government bonds are traded on the MTS platform and, thus, detailed transaction data are available. This provides us with representative high-frequency data on quotes and transactions, which enables us to calculate a variety of liquidity measures and conduct an in-depth investigation of price movements. However, based on less detailed data, we also present results for other highly affected countries, i.e., Greece, Ireland, Portugal and Spain. During the European sovereign debt crisis, sovereign bonds issued by all these countries offer a unique opportunity to study political uncertainty. First, all countries experienced severe distress during the crisis which led to significant default risk associated with their sovereign debt and, in turn, made the prices of the government bonds very susceptible to changes in the economic and political framework. In contrast, bonds issued by the core eurozone countries did not show a similar development and remained relatively stable throughout the crisis. In addition, the European sovereign debt crisis was very political in nature. All participants were constantly facing the possible exit of one or more countries from the eurozone while trying to coordinate a response to the crisis on a multinational level and several key political decisions marked the crisis, e.g., the creation of the European Financial Stability Facility (EFSF) or the 2012 Greek elections.

We define a list of relevant political events similar to Kelly et al. (2016), covering the peak of the European sovereign debt crisis from 2010 to 2013. The list includes political summits (Euro, G8 and G20) and relevant elections. We analyze the uncertainty around these events and its effects on prices and liquidity in the spirit of Pastor and Veronesi (2013). In this context, there exists uncertainty before the event about which policy will be implemented and what the impact of these policies will be. Investors can form expectations about the optimal policy and its impact based on economic fundamentals. However, investors are aware that decision-makers are influenced by non-economic objectives (political costs) in their decisions, as well. These political costs are not known to investors, but they can learn about these costs from political news before the event. Political news does not provide any information about economic fundamentals or the impact of policies, but still leads to the revision of the likelihood of the implementation of the different policies and, thus, affects prices. Therefore, being exposed to a stream of political news before the event adds to the overall uncertainty, for which investors demand a risk premium.¹ At the event, the policy decision is revealed and the uncertainty related to this decision resolves. However, in Pastor and Veronesi (2013) the impact of the policy is still uncertain and investors can only learn about the policy after its implementation over time.

We are aware that during the sovereign debt crisis, the resolution of uncertainties might not be as clear cut as theory suggests. In particular, policy uncertainty including political risk might not fully resolve at the individual events as negotiations and political discussions continued thereafter. In addition, some events might potentially not only focus on the resolution of the sovereign debt crisis. Nonetheless, it is interesting whether general market reactions are in line with the theoretical guidance. Thus, in our analysis we focus on price and liquidity effects before, around and after the events focusing on a time window covering 60 trading days before and after the event day. Additionally, we provide a regression analysis exploring whether the observed effects can be linked to the EPU index measuring political uncertainty based on newspaper articles (see Baker et al. 2016) and to the economic conditions at the time of the event, while controlling for bond characteristics.

We find a negative pre-event return, i.e., strong negative price reactions of Italian government coupon bonds before the events, indicating that political uncertainty affects prices. On average, prices fall by around $1\%$ in the time window 20 days before the event compared to the beginning of the corresponding 60 day window. This effect is highly significant in statistical and economic terms. We observe negative price reactions whether we calculate returns or abnormal returns with respect to maturity-matched German government bonds. Directly around the event (i.e., three days before/after the event) we find significant positive price impacts of around $0.4\%$, i.e., a positive event return. Thus, we observe a positive effect as soon as the policy decision is revealed. However, prices do not further recover for the next 20 trading days. Within 60 days after the events, prices basically fully recover to pre-event levels indicating that impact uncertainty resolves to certain degree. We find similar effects for government bonds issued by Greece, Ireland, Portugal and Spain. Overall, these results are in line with the discussed theoretical literature.

In addition, we study trading activity measures representing liquidity measures related to transaction costs and volume based measures. The liquidity measures cover bid-ask spreads, as well as, the Roll, Amihud and price dispersion measure. We find that all these measures indicate higher illiquidity shortly before the event when prices tend to fall, with the highest illiquidity being observed directly around the event, e.g., the price dispersion measure increases from 15 bp to 20 bp. The trading costs return to previous levels after 60 days in line with the observed price recovery. Analyzing the trading volume, we observe a significant higher volume of around $5\%$ before the event. This coincides with a much higher sell-side trading activity, in particular three days before the events. Thus, we see that the observed price movements coincide with sell-side pressure. Combining the insights on liquidity with the price results above, reveals that a short-term trading strategy around the event might not be profitable given the high transaction costs. However, a strategy buying the bond before the event and holding it 60 days would be profitable even after considering transaction costs.

In our regression analysis, we relate the observed pre-event, event and post-event returns to variables measuring political uncertainty as well as market and economic conditions at the time of the event. As a direct measure of political uncertainty we use the Italian Economic Policy Uncertainty (EPU) index by Baker et al. (2016). We consider the VSTOXX index for general market volatility and include the Italian CDS spread as a market-based measure of credit risk. The economic conditions are represented by Italian GDP growth and inflation. In addition, we include the 3 M Euribor, representing short-term interest rates, bond-specific characteristics and liquidity measures. We find a significant effect of the EPU index for pre-event returns where a higher increase in uncertainty indicates a stronger fall in bond prices. In addition, weaker economic conditions are related to stronger price reductions. Interestingly, the effect of the EPU index fades out and is not present in the post-event returns, indicating that prices are only related to political uncertainty in the context of upcoming events and not unconditionally at all times.

In addition, we find that political uncertainty also affects the primary market. In the time period covering the 20 days before the events, the Italian government issued 193 billion EUR of notional volume, thereof 155 billion EUR as seasoned bond offerings. Extending our analysis to these seasoned bond offerings, we find significant direct costs of 2.70 billion EUR for issuing bonds during these periods and thereby exposing primary dealers and investors to political uncertainty. Thus, political uncertainty related to political events should be considered in the issuance policy of debt management offices (DMO), as such costs could potentially be avoided by more active cash management. In fact, when we consider opportunity costs for deviating from the issuing strategy (e.g., cost for adjusting the auction calendar or higher administrative costs), we still find overall costs of 1.44 billion EUR that could have been saved. However, as we are not modelling the DMO’s utility function, we cannot comment on whether the observed issuance policy was optimal or not.

Overall, the results in this paper foster our understanding of the effect of political uncertainty. We present price effects for sovereign debt markets, which have not been considered by the empirical literature so far. The price reactions of these financial instruments are especially important when analyzing political uncertainty, as these instruments are directly linked to the activities of governments and, thus, are closely related to political uncertainty. This is particularly true for our considered time period, in which political events were almost exclusively focusing on policies handling the European debt crisis. Furthermore, we quantify and analyze the trading activity around the events, which further increases our understanding of the price reactions. The remainder of the paper is structured in the following way: Sect. 2 reviews the literature. Section 3 explains the data set and Sect. 4 presents our hypotheses and methodology. Our results are discussed in Sect. 5. Section 6 concludes.

2 Literature

This paper is related to various strands of the literature. First, we consider the recent theoretical literature studying political uncertainty and providing guidance on its asset pricing implications. In addition, we review the corresponding empirical literature, which is focused on stock and option markets, as well as related empirical studies investigating the impact on bond yields. Further, our paper also relates to the literature studying European sovereign bonds based on MTS data and analyzing bond market liquidity in general.

Concerning the effects of political uncertainty in the context of asset pricing, Pastor and Veronesi (2012) and Pastor and Veronesi (2013) make important theoretical contributions. In Pastor and Veronesi (2012) the uncertainty about government policies and its effects on stock prices are modeled in a general equilibrium framework. Investors are exposed to potential policy changes by the government and, although the government maximizes investors’ welfare, it also takes into account non-economic objectives, i.e., political costs (or benefits) incurred by changing the policy. Investors do not know this cost and, thus, cannot fully anticipate policy changes. After a policy change, investors’ beliefs are reset and they are exposed to uncertainty concerning the impact of the new policy. Note, they can only learn about the impact after the implementation of the policy. Pastor and Veronesi (2012) focus on announcement returns and show that the unconditional risk premium is positive. However, returns conditional on a policy change can be positive or negative and tend to be positive after long and severe economic downturns.

Building on these results, Pastor and Veronesi (2013) introduce the possibility for investors to learn about the political costs associated with the introduction of potential new policies. This introduces political shocks triggered by political news and leads to investors revising their beliefs concerning the probability of policy changes. Note that political news is unrelated to economic fundamentals and the potential impacts of policies. Pastor and Veronesi (2013) show that investors additionally demand a risk premium for being exposed to the stream of purely political news around major events. Again, this risk premium is larger when there is more uncertainty and in severe economic downturns. Thus, these theoretical results establish an additional source of price risk before the actual announcement of the policy decision.² We base our empirical analysis on this theoretical guidance. Thus, we expect finding negative price effects before the event, when investors are exposed to political news adding to the uncertainty about which policy will be implemented. The event returns is expected to be positive as this uncertainty resolves. However, the impact uncertainty about the implemented policy remains and its resolution should lead to positive price effects after the event. We expect these effects to be more pronounced when uncertainty is high and economic conditions are weak.

Analyzing the empirical results presented in the relevant literature, Pastor and Veronesi (2013) use the S&P 500 index to show that investors indeed demand a risk premium for political uncertainty, by relating the observed returns to the EPU index. In addition, the results show that the effect is stronger in weak economic conditions, measured by various proxies, e.g., recession period dummies and credit spreads. However, they do not directly relate the returns to political events in this exercise. Kelly et al. (2016) empirically analyze the pricing of political uncertainty in the option market. They isolate political uncertainty by studying its variation for important political events, such as national elections and global summits. They find that options that mature after such events are on average more expensive since they protect investors against the risk caused by these events. Similar to the previous paper, this effect is more pronounced when the economy is in a weak condition and the political uncertainty is higher.

Considering bond markets, Gao and Qi (2013) isolate uncertainty around US gubernatorial elections and study how it affects the offering yield of newly issued municipal bonds. They find that yields increase before elections and that stronger movements can be observed during an economic downturn and for elections with low predictability. In addition, they provide first evidence that the trading activity is affected by political uncertainty by reporting a lower number of net buy orders before the elections. Regarding international political uncertainty, Hunag et al. (2015) analyze the impact of international political crisis and find increasing end-of-month yields. Brogaard et al. (2020) investigate global political uncertainty measured by the US election cycle. They mainly analyze non-US equity markets and find a fall in returns before these events. In an additional analysis, they also investigate sovereign bond returns which tend to increase before these elections indicating a flight-to-safety effect from equity to bond markets.³ A more recent empirical paper related to our analysis is Pierluigi et al. (2020). This paper explores the impact of political events and policy announcements on financial markets and real economic activity in Italy from 2013 to 2019. Using an instrumental variable approach based on changes in CDS spreads, their analysis shows significant effects of these events, in particular, on government bond and stock prices. This paper discusses also results based on different CDS clauses providing measures of redenomination and redenomination-free credit risk. In addition, they show international spillover effects. Overall, the paper documents the important impact of political risk. Our paper differs in several important aspects. In contrast to all the mentioned papers, we focus on the height of the sovereign debt crisis and use detailed transaction data that allows to explore liquidity effects, as well. When comparing to Pierluigi et al. (2020), in addition our choice of events is significantly different, as they focus on the impact of strengthening populist movements and its impact on the economy. Thus, our paper provides results on different aspects of political uncertainty compared to the existing literature.

Papers measuring political uncertainty either make use of poll data to identify elections with uncertain outcomes or make use of uncertainty indices. Baker et al. (2016) provide such an important index called economic policy uncertainty (EPU) index. This index is based on the relative frequency of newspaper articles containing keywords representing policy uncertainty. They focus on the USA, but they also provide indices for France, Germany, Italy and Spain among others. The paper analyzes the relation of the EPU index to firm-level data and finds that policy uncertainty is associated with greater stock price volatility and leads to a reduction of investment and employment. They further show that the index is related to macro-economic variables in the USA and in 12 other major economies. There is a growing empirical literature using the EPU index: Manzo (2013) shows that an increase in the European EPU index leads to higher sovereign credit spreads in Europe. Brogaard and Detzel (2015) also use this index and show that it positively forecasts log excess market returns. Hardouvelis et al. (2018) construct an EPU index for Greece and show that positive shocks are associated with an increase in bond yields. Similarly, Paule-Vianez et al. (2021) find a positive impact of economic policy uncertainty on the Spanish 10-year-bond yield.⁴

There are several papers dealing with the MTS dataset and European sovereign bonds. Most of these papers explore the relation of credit and liquidity risk and analyze the impact of ECB interventions during the financial and European sovereign debt crisis. Dufour and Skinner (2010) provide a summary of the dataset, discussing specific aspects of the market structure. Beber et al. (2009) analyze whether bond investors demand credit quality or liquidity, showing that in times of market stress investors focus on liquidity. Pelizzon et al. (2016) show that during the European sovereign debt crisis starting in 2010 credit risk drives the liquidity of the overall market and that this effect was weakened after the ECB interventions. Eser and Schwaab (2016) study the effect of the securities markets program in the eurozone from 2010 to 2011. In their study, they find large announcement effects and lower bond yield volatility on intervention days. Finally, Schlepper (2017) quantifies the price impact of the public sector purchase program in the German bond market. They find 10-minute price impacts that are larger when market liquidity is low.

Regarding liquidity measures, various proxies are commonly used in the context of fixed-income markets. Following Schestag et al. (2016), one can distinguish between transaction cost measures and price impact measures. The most well-known transaction cost measure is the Roll (1984) measure. It uses the negative covariance of consecutive daily returns, caused by bid-ask bounce, to proxy for liquidity.⁵ A price impact measure was first proposed by Amihud (2002). The Amihud measure is the daily ratio of the absolute return to the trading volume. Subsequently with the availability of detailed transaction data, these originally low-frequency proxies were adapted to high-frequency measures. For example, in their analysis Dick-Nielsen et al. 2012 use an intraday version of the Roll measure and also the Amihud measure can be adapted similarly, as was done by Goyenko et al. (2009), for example. A high-frequency measure specific for dealer markets is the price dispersion measure proposed by Jankowitsch et al. (2011). It is based on the idea that dealers face inventory risk and, thus, prices might deviate from the expected market valuation to compensate them for the risk. Hence, the dispersion around the market valuation can be seen as a transaction cost measure. Another measure for dealer markets is the imputed round-trip measure by Feldhütter (2012). The measure analyzes prices of trades that happened in short succession and have the same volume indicating a matched trade and, thus, the difference is a proxy for round-trip costs.⁶

3 Data

3.1 Political events

We hand-collect a list of events strongly related to political uncertainty during the European sovereign debt crisis, similar to Kelly et al. (2016). The list includes summits and elections from the beginning of 2010 to the end of 2013, covering the height of the European sovereign debt crisis.

In the aftermath of the 2007/08 financial crisis, the structural problems of the euro area became evident. A monetary union consisting of member states with independent fiscal authorities and a system of central banks focusing purely on inflation targets without a mandate to purchase government debt exposed countries of the periphery to the risk of significantly increasing refinancing costs in crisis periods. The European sovereign debt crisis started in late 2009 with the newly elected government in Greece revising the deficit forecast to 12.7% and increasing the Greek debt to $113\%$ of GDP.⁷

In the case of Italy, investors became very nervous in mid-2011 given a debt level of $120\%$ and a budget deficit second only to Greece in the eurozone, resulting in a downgrade in September 2011 and a negative outlook for the third largest economy in the eurozone. This situation worsened in November 2011 when Berlusconi lost his parliamentary majority on a budget vote. In March 2012 Greece defaulted on its sovereign debt erasing around 100 billion EUR. In 2012 two elections followed increasing the political uncertainty concerning the future of the eurozone. The whole crisis was accompanied by many rescuing initiatives, resulting in the European Stability Mechanism (ESM) and European Financial Stability Facility (EFSF), which were officially established by the end of 2012. In August 2013 the eurozone emerged from recession, marking the end of the most severe period of the crisis for most countries. In particular, Italy started out with 10-year bond yields of around $4\%$ in 2009, which spiked at $7.5\%$ in November 2011 and after reducing to $5\%$ returned to $6.5\%$ by mid-2012. By the end of 2013, yields returned to $4\%$.

Although most of the discussions were focused on the Greek crisis, this situation triggered the need for policy decisions that were very important for other countries as well, especially Italy. This involved agreeing on a European level how to deal with countries with financial problems and an acceptance of these measures by the affected countries. Thus, there is a spillover due to necessary policy decisions across countries. In addition, there is a potential spillover of uncertainty. When the uncertainty concerning the political-decision making process is high in certain countries (e.g., in Greece), i.e., when it is unclear how the country will react to potential European measures, because the political costs of accepting demands for financial help might be high, then most likely also the uncertainty and political costs in Italy might be high. Although it is clear that there are important structural differences between countries, there can be a significant spillover of uncertainty.

Thus, our list of relevant events covers all Euro summits, i.e., the meetings of the heads of governments of the eurozone members. The main topic of these summits during those years was to coordinate a common crisis response within the eurozone, making them the main set of events for our analysis. In addition, we include G8 and G20 summits that were held during that time period if the European sovereign debt crisis or another relevant economic topic was the main part on the agenda of the respective summits.⁸ As we focus on Italy, we include Italian parliamentary elections and, in addition, the Greek 2012 elections as there was substantial risk that the election results could lead to Greece leaving the eurozone with severe consequences for all other countries. Table 1 shows the full list of events with the main topics on the agenda and an indicator if it is included or was omitted due to not focusing on an economic topic or being too close to an already included event, i.e., within seven days of a previous event.⁹ We consider 17 events resulting in roughly 4 events per year.

Table 1

List of political events

Date	Type	Topics	Included
2010-03-25	Euro Summit	Greek crisis, Europe 2020	Yes
2010-05-07	Euro Summit	Greek crisis	Yes
2010-06-25	G8 Summit	Global recession, European debt crisis	Yes
2010-06-26	G20 Summit	Global recession, European debt crisis	No
2010-11-11	G20 Summit	Global economic recovery, financial regulation, global financial safety nets	Yes
2011-03-11	Euro Summit	Pact for the Euro, ESM	Yes
2011-05-26	G8 Summit	Internet, innovation, green growth, nuclear safety, Arab Spring	No
2011-07-21	Euro Summit	Greek crisis	Yes
2011-10-23	Euro Summit	Economic policy, banking package	Yes
2011-11-03	G20 Summit	International monetary system, strengthen financial regulation	Yes
2011-12-08	Euro Summit	Economic policy, fiscal compact	Yes
2012-01-30	Euro Summit	Stimulating employment, completing the single market	Yes
2012-03-01	Euro Summit	Economic policy, Treaty on stability	Yes
2012-05-07	Election	Greek election	Yes
2012-05-18	G8 Summit	European sovereign debt crisis	Yes
2012-06-17	Election	Greek election	Yes
2012-06-18	G20 Summit	European sovereign debt crisis	No
2012-06-29	Euro Summit	EMU, multinational financial framework, compact for growth and jobs	Yes
2013-02-25	Election	Italian election	Yes
2013-03-14	Euro Summit	Economic and social policy	Yes
2013-06-18	G8 Summit	Syrian civil war	No
2013-09-05	G20 Summit	Syrian civil war	No

This table provides all Euro, G8 and G20 summits as well as Italian and Greek parliamentary elections during the European sovereign debt crisis represented by the time period 2010 to 2013. For each event the date, type and main topics of the agenda are provided. The last column provides an indicator whether the event is included in our analysis. We exclude events if it is held within seven days of a previous event or does not focus on economic topics

This set of events offers an ideal laboratory to study political uncertainty. In most circumstances, political events are considered with many different aspects concerning various policies, e.g., in the case of regular elections. However, in the case of the European sovereign debt crisis most political events at this time were almost exclusively focusing on policies handling the crisis. This allowed investors to receive more precise signals compared to regular times. This is especially true for the Euro summits, which, in addition, were held frequently compared to regular elections. In addition, the level of political uncertainty was high and all countries experienced severe distress during this crisis leading to significant default risk associated with their sovereign debt, which made the prices of the bonds very susceptible to changes in the economic or political framework.

3.2 Bond data

Our bond dataset represents MTS transaction data for Italian sovereign bonds from the beginning of 2010 to the end of 2013. This dataset contains high-frequency data on transactions, i.e., traded price (clean price) and volume with time stamp and buy/sell indicator, as well as intra-day quotations (best-bid and best-ask). MTS covers bonds issued by several different European countries and similar issuers. The overall dataset contains various MTS interdealer markets, i.e., EuroMTS, EuroCredit MTS and the domestic MTS markets. The EuroMTS serves as a reference market for European benchmark bonds as well as bonds with an outstanding amount greater than 5 billion. Other bonds are covered by EuroCredit MTS or one of the domestic MTS markets. The dataset was established in April 2003 and originally contained information on bonds of 11 different eurozone countries: Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, The Netherlands, Portugal and Spain. We focus on the Italian sovereign bond market as, in contrast to most other countries, the vast majority of transactions happens via the MTS system.¹⁰ This provides us with representative high-frequency data on quotes and transactions.

In our analysis we focus on the regular coupon bonds issued by the Italian government: Buoni del Tesoro Poliennali (BTP) are long-term bonds with a maturity of 3, 5, 10, 15 or 30 years.¹¹ We exclude index-linked securities and floating rate notes from our sample, which represent only $12\%$ of the outstanding volume and are likely to show different price reactions than regular bonds. Overall, we cover 79 bonds. Table 2 shows summary statistics of our sample, i.e., information on the amount issued, coupon and maturity. We find an average issuing size of 20 billion EUR with an average maturity of around 11 years and an average coupon of $4.2\%$.

Table 2

Descriptive statistics of bond characteristics

Panel A: bonds
Observations	Amount issued			Coupon			Time-to-maturity
	$Q_{10}$	Mean	$Q_{90}$	$Q_{10}$	Mean	$Q_{90}$	$Q_{10}$	Mean	$Q_{90}$
79	15.16	20.25	25.90	2.75	4.24	5.50	3.01	10.90	30.22

Panel B: bond-event combinations
Filter	Obs	Transactions		Volume
Filter	Obs	Sum	Daily bond average	Sum	Daily bond average
No	865	548, 860	5	2, 828, 422	27
Yes	860	547, 489	5	2, 822, 172	27

This table shows two types of summary statistics. Panel A presents information on the Italian government bonds used in our sample, i.e., the number of bonds, the amount issued in billions of EUR, the coupon in percent and time-to-maturity given in years. Panel B provides information on the trading activity across all bond-event observations. The trading activity is represented by the number of transactions and trading volume in millions of EUR. We present the overall sum and daily bond-specific average calculated within the time window of 60 trading days before to 60 trading days after the considered political event. The results are shown with and without the data filters we apply to obtain our final sample. Our data set covers the European sovereign debt crisis represented by the time period 2010 to 2013, and the political events include Euro, G8 and G20 summits as well as relevant elections during this time period

We filter out erroneous entries for both transactions and quotes. Concerning the transactions, we apply similar filters as Dick-Nielsen (2009), but adapted to the context of the MTS dataset.¹² In particular, we control for high intra-day price variation, i.e., if any price shows a $10\%$ percent deviation compared to the median price on that day, it is considered as erroneous. In addition, we eliminate transactions if any price changes between two consecutive trades are greater than $5\%$.¹³ Concerning quotations, we first apply filters based on the observed intra-day spreads between best-ask and best-bid quote. We winsorize the intra-day spreads of each bond on every day, by setting the top and bottom $2.5\%$ percent to the respective quantiles.¹⁴ In addition, we filter out the highest percent of daily average bid-ask spreads in our sample period across bonds.¹⁵

Based on this filtered dataset, we consider the transactions and quotations in a time window starting 60 days before and ending 60 days after the events (see Sect. 3.1). Overall, we obtain 865 bond-event observations (see Table 2). To ensure a minimum level of trading activity of the bonds around the events, we consider only observations with at least five trades before and after the event with a total trading volume of at least 100 million EUR. These restrictions result in 860 remaining bond-event observations. Table 2 shows that in the time window around the events the considered bonds were traded 547, 489 times which corresponds to an average of 5 trades per day and bond. Similarly, the total trading volume is 2.8 trillion EUR or 27 million EUR per day and bond. Figure 1 shows the average prices and transaction costs in the considered time window.

In addition, to extend our analysis of price effects to other countries that were highly affected by the European sovereign debt crisis, i.e., Greece, Ireland, Portugal and Spain, we obtain zero-coupon yield curves from Bloomberg for these countries for our sample period. The yield curves include rates from 3 months to 30 years, allowing us to study synthetic zero-bond prices with matched durations compared to the Italian bonds. In addition, we obtain zero-coupon yields for Germany from Bloomberg, which we use as a eurozone benchmark in our analysis.

3.3 Uncertainty index and economic variables

In our analysis, we consider the effects of political uncertainty and economic conditions. As a relevant measure of political uncertainty we obtain the Italian version of the EPU index from the homepage of Baker, Bloom and Davis, which provides us with monthly data (see http://www.policyuncertainty.com/).¹⁶ Figure 2 shows the Italian EPU Index for our sample period. In addition, we consider general market volatility by including the VSTOXX index as a daily measure retrieved from Bloomberg and use the 5 Year Italian CDS spread, as a measure of credit risk.¹⁷ To measure the economic condition, we consider the 3-month Euribor rate, the monthly year-to-year inflation in Italy and quarterly Italian year-to-year GDP growth data. All these variables are downloaded from Bloomberg.

4 Methodology

4.1 Price and trading activity variables

In order to measure the price and trading activity effects, we focus on daily measures. Thus, we calculate the daily average price for each bond and we measure its daily trading volume and net-bid volume, representing the difference between sell-side and buy-side initiated transactions. In addition, we study several daily bond-specific transaction cost measures. In particular, we provide the bid-ask spread based on quotations, as well as three transaction-based liquidity measures: the Roll, Amihud and price dispersion measure.

Volume-Weighted Average Daily Price: We calculate a volume-weighted price, placing more weight on transactions with higher volume and, thus, reducing the potential noise of small unrepresentative trades, see e.g., Bessembinder et al. (2009). The daily volume-weighted price for bond i on day t is given by:

$$\begin{aligned} P_{i,t} = \frac{\sum _{k=1}^{K_{i,t}} p_{i,t,k}v_{i,t,k}}{\sum _{k=1}^{K_{i,t}} v_{i,t,k}}, \end{aligned}$$

where $K_{i,t}$ denotes the number of trades, $p_{i,t,k}$ the prices of these trades and $v_{i,t,k}$ the respective volumes.

Trading Activity: We measure trading activity by the daily trading volumes of the bonds. The cumulative trading volume for bond i on day t is given by:

$$\begin{aligned} V_{i,t} = \sum _{k=1}^{K_{i,t}} v_{i,t,k}. \end{aligned}$$

In addition, we calculate the difference between sell-side and buy-side initiated transactions. The net-bid for bond i on day t is defined as:

$$\begin{aligned} \text {Net-Bid}_{i,t} = \sum _{b=1}^{B_{i,t}} v_{i,t,b} - \sum _{a=1}^{A_{i,t}} v_{i,t,a}, \end{aligned}$$

where $B_{i,t}$ denotes the number of trades where the initiator wanted to sell and $A_{i,t}$ the number of trades where the initiator wanted to buy.

Transaction Costs: We estimate transaction costs based on quotations by calculating the daily bid-ask spread of bond i on day t and relate this spread to the average observed price:

$$\begin{aligned} \text {Bid-Ask Spread}_{i,t} = \frac{\sum _{q=1}^{Q_{i,t}} \text {Best-Ask}_{i,t,q} - \text {Best-Bid}_{i,t,q}}{Q_{i,t} \cdot \text {P}_{i,t}} \end{aligned}$$

where ${Q_{i,t}}$ is the number of intra-day quotations represented by a best-ask and best-bid quote. In addition, we employ the Roll measure (see Roll 1984) based on intra-day transaction similar to Dick-Nielsen et al. (2012). The Roll measure for bond i on day t is defined as:

$$\begin{aligned} \text {Roll}_{i,t} = 2 \cdot \sqrt{- Cov(\Delta p_{i,t,k}; \Delta p_{i,t,k-1})} / P_{i,t} \end{aligned}$$

where $\Delta p_{i,t,k}$ denotes the intra-day change of the consecutive prices k and $k-1$. We estimate this measure for every bond and day with at least four transactions and standardize the measure with the respective average price. Furthermore, we calculate an intra-day version of the Amihud measure (see Amihud 2002), which relates the absolute return of consecutive transactions to the observed trading volume. The measure is defined as

$$\begin{aligned} \text {Amihud}_{i,t} = \frac{1}{K_{i,t}} \cdot \sum _{k=1}^{K_{i,t}} \frac{|r_{i,t,k}|}{v_{i,t,k}}, \end{aligned}$$

where $r_{i,t,k}$ is the return of trade k, $v_{i,t,k}$ is the respective volume, and $K_{i,t}$ is the number of trades on day t for bond i. The measure allows us to analyze by how much consecutive prices change given a certain trading volume. Thus, small price changes after a high volume transaction indicates high liquidity. We estimate this measure if at least two transactions are available. As a third transaction-based measure, we calculate the price dispersion measure by Jankowitsch et al. (2011) by taking the square root of the mean squared differences between the traded prices of a bond and its market valuation. The dispersion measure for bond i is:

$$\begin{aligned} \text {Price Dispersion}_{i,t} = \sqrt{\frac{1}{\sum _{k=1}^{K_{i,t}} v_{i,t,k}} \sum _{k=1}^{K_{i,t}} (p_{i,t,k} - m_{i,t})^2 \cdot v_{i,t,k}} \end{aligned}$$

This measure assumes that price fluctuations around the fundamental price represent deviations caused by trading costs. The fundamental price is approximated by the average mid-price $m_{i,t}$ based on the observed quotations. Thus, this measure can be computed even with only one available transaction.

4.2 Price and trading activity effects

We measure changes in price and trading activity variables across various time windows to analyze effects in the pre-event, event and post-event period. In particular, we consider the following time windows: $(-60,-21)$, $(-20,-1)$, $(-3,-1)$, (0, 3), (1, 20) and (21, 60), where 0 is the event day and all other numbers represent the trading days relative to the event day. For each time window, we calculate the bond-specific average value of the price and trading activity variables. Based on these averages, we compute the price return and the change in trading activity variables across various combinations of two different time windows. In particular, we compare the time windows $(-60,-21)$ and $(-20,-1)$ to measure the pre-event effect, $(-3,-1)$ and (0, 3) to explore the event effect and (1, 20) and (21, 60) to analyze the post-event period. Regarding the pre-event effect, we focus on a time window starting 20 trading days before the event, because in most cases the agenda of the summits is made public four weeks before the event. The three-day period around the event is simply chosen to consider enough trading information, which would otherwise not be available based on a shorter time window for all bonds. The post-event analysis again focuses on 20 trading days after the event, because many press releases commenting the summit results are published in this time period. The periods from 21 up to 60 days are chosen to assure that these time windows are not affected by the events.¹⁸ Overall, our pre- and post-event periods cover three months each and it is clear that in such long periods price changes are not only driven by information related to our events. Thus, when interpreting our descriptive statistics, we have to assume that price effects due to underlying risk factors cancel out across events, on average. However in our regression analysis, we control for these factors, see Sect. 4.3.

Concerning prices, we estimate three different price returns. First, we simply calculate returns based on the average reported (clean) prices, i.e., representing a return without deterministic coupon effects. In addition, we add accrued interest and coupon payments when calculating the average prices and, thus, provide returns based on the total (dirty) prices. Furthermore, we calculate abnormal returns by adjusting these price returns for market returns, represented by duration-matched returns based on synthetic German zero-coupon bonds. Concerning the trading activity variables, we provide the changes across time periods based on the calculated averages, to analyze potential effects. As trading costs can be very low for very liquid instruments, we do not employ relative changes, as small changes in absolute terms could be translated into unreasonably high relative changes.

4.3 Regression analysis

We run pooled regressions to explore whether the observed price changes across event windows can be related to variables measuring political uncertainty and economic conditions.¹⁹ The dependent variable represents the total price returns based on all bond-event combinations for either the pre-event, event or post-event period.²⁰ The regression specification modelling the returns y for bond i in the event t for a particular period is given by

$$\begin{aligned} y_{i,t} = \alpha&+ \beta (\Delta \text {Uncertainty Measure})_t + \gamma (\Delta \text {Market Variables})_t \\&+ \delta (\Delta \text {Economic Variables})_t + \zeta (\Delta \text {Bond Liquidity})_{i,t} \\&+ \eta (\text {Bond Characteristics})_{i,t} + \epsilon _{i,t}, \end{aligned}$$

where we employ the EPU as uncertainty measure, the VSTOXX index and Italian 5Y CDS spread as market-based variables, and inflation, GDP growth and 3 M Euribor rate as characteristics measuring the economic condition. In addition, we include liquidity, represented by the price dispersion measure, and bond-specific characteristics, i.e., the coupon and time-to-maturity.²¹ Some measures are not available on a daily basis, e.g., GDP growth. Thus, we define the employed variables in the following way: if the variables are observable in the relevant two time windows of the considered period (e.g., $(-60,-21)$ and $(-20,-1)$ in the case of the pre-event period), we take the average across all available observations within the two individual time windows. If no observations are available we take the closest value before the first time window and the closest value after the second time window. Based on these two values, we calculate the change in the considered variables across time windows. As bond prices tend to behave similarly around a specific event we follow the approach by Petersen (2009) and calculate standard errors corrected for heteroscedasticity and clustered at event level.²²

5 Results

In this section we present our results starting with a graphical analysis and descriptive statistics of the price and trading activity effects around the political events. The main analysis provides tests concerning price returns and changes in trading activity in the pre-event, event and post-event period. In addition, we present our regression analysis relating the price and liquidity effects to measures of political uncertainty and economic conditions.

5.1 Impact of political uncertainty

We provide a first overview of price and trading activity effects around political events in Fig. 1, based on the reported prices and price dispersion measures around the events, based on all Italian bonds in our sample. We find that prices start at around $100.5\%$ and begin a downward trend roughly 40 days before the event. In the 20 days before the event, prices drop significantly and reach a low of around $99.5\%$ right before the event. The price effects show a short-term reversal on the event day. However, prices stay on a low level and are quite volatile. Event prices start to increase again 20 days after the event and reach roughly their pre-event level 60 days after the event. Concerning the liquidity of the bonds, we find that the price dispersion measure is quite volatile in the whole time series, increasing before the event and peaking directly around the event. Thereafter, the trading cost measure is very high for the next 20 days and then slowly returns to its pre-event level.

Based on the defined time windows (see Sect. 4), we analyze the price returns in the pre-event, event and post-event periods, see Table 3. In addition, we analyze the results of other affected sovereign bond markets, based on less detailed data. Analyzing the results for the Italian coupon bonds, we find statistical significant negative returns when comparing the time windows $(-60,-21)$ and $(-20,-1)$. We find an average return of $-0.94\%$ based on the reported prices, the total price return (including accrued interest and coupon payments) is $-0.53\%$ and the abnormal return with respect to the German government bond market is even $-1.20\%$, on average.²³ Thus, we find a significant price drop before the event indicating a risk premium for uncertainty as discussed in Sect. 2.²⁴ Analyzing the time periods directly around the event, i.e., $(-3,-1)$ and (0, 3), we find a statistical significant price increase of around $0.4\%$ in all three specifications.²⁵ Thus, the policy decisions lead to a positive price effect, on average. After the event, represented by the returns between the windows (1, 20) and (21, 60), we find statistical significant positive returns, e.g., $0.85\%$ based on reported prices.²⁶ Thus, prices stay on the event level for 20 days and return to pre-event levels after 60 days. In particular, the abnormal returns with respect to German bonds indicate that the results of the Euro summits and elections are not affecting price factors relevant for all eurozone countries, but concern rescue policies relevant for the most affected countries. Thus, in particular decisions affecting the general level of short-term interest rates, e.g., by the ECB monetary policy, cannot be the driver of these results, as such effects would impact all government bond markets.²⁷ In addition, we show in Fig. 3 the distribution of bond returns directly around the event. Although event returns are significantly positive on average, individual bond returns of particular events can be well below $-3\%$, supporting our interpretation of the existence of a risk premium before the event.²⁸

Table 3

Price effects of bonds around political events

	Price returns	Total price returns	Abnormal returns w.r.t. Germany	Returns of other affected markets	Abn. returns of other affected markets
Panel A: returns
$(-60,-21);(-20,-1)$	$-0.936$	$-0.528$	$-1.198$	$-0.542$	$-1.211$
$(-3,-1);(0,3)$	0.362	0.410	0.311	0.404	0.305
(1, 20); (21, 60)	0.849	1.198	0.527	1.728	1.057
Panel B: T-test: t-values
$(-60,-21);(-20,-1)$	$-9.316^{***}$	$-5.184^{***}$	$-10.186^{***}$	$-3.243^{***}$	$-6.640^{***}$
$(-3,-1);(0,3)$	$6.794^{***}$	$7.637^{***}$	$5.266^{***}$	$4.099^{***}$	$2.852^{***}$
(1, 20); (21, 60)	$8.993^{***}$	$12.485^{***}$	$4.704^{***}$	$8.946^{***}$	$4.986^{***}$
Panel C: Wilcoxon test: p-values
$(-60,-21);(-20,-1)$	0.000$^{***}$	0.000$^{***}$	0.000$^{***}$	0.008$^{***}$	0.000$^{***}$
$(-3,-1);(0,3)$	0.000$^{***}$	0.000$^{***}$	0.000$^{***}$	0.242	0.028$^{**}$
(1, 20); (21, 60)	0.000$^{***}$	0.000$^{***}$	0.000$^{***}$	0.000$^{***}$	0.000$^{***}$

This table shows the price returns around political events of Italian government coupon bonds (BTP). The following time windows around the events are considered: $(-60,-21)$, $(-20, -1)$, $(-3,-1)$, (0, 3), (1, 20) and (21, 60), where 0 is the event day and all other numbers represent the trading days relative to the event day. For each time window, the bond-specific average values of the prices are calculated and the returns across combinations of two time windows are computed, measuring price effects before, during and after the events. The returns are based on reported (clean) prices and total (dirty) prices. In addition, the abnormal returns with respect to German government bond prices are presented based on total prices. Furthermore, returns and abnormal returns for other highly affected sovereign bond markets, i.e., Greece, Ireland, Portugal and Spain, are presented based on less detailed data. Panel A provides the average price effects measured in percent, Panel B t-values of one-sample t-tests and Panel C p-values of one-sample Wilcoxon signed rank tests. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013.

The significance is indicated as follows: * $< 0.1$, ** $< 0.05$, *** $< 0.01$

These findings are strengthened by the results concerning other highly affected countries (i.e., Greece, Ireland, Portugal and Spain in our analysis). As indicated in Sect. 3, these results are not based on detailed trading data, but represent the returns of theoretical bond prices based on the observed time series of zero-coupon yields. To make the results directly comparable to the Italian government bond returns, we study synthetic zero-bond prices with matched maturities compared to the Italian bonds. Thus, for all available Italian bonds synthetic prices are calculated in the considered time windows for each day. In Table 3, we present the returns and abnormal returns based on equally weighting the returns of the four countries for the defined time windows. Focusing, on the abnormal returns, we find very similar results compared to the Italian bonds, i.e., returns of $-1.21\%$ before the event, $0.31\%$ during the event and $1.06\%$ afterward. All these results are highly statistically significant.

In our second analysis, we investigate the changes in liquidity by studying our various trading activity variables in the specific time windows. Table 4 presents the descriptive statistics for the liquidity measures based on the Italian government coupon bonds. We find that most measures are lowest in the window $(-60,-21)$ before the event. The measures are slightly higher closer to the event based on $(-20,-1)$ and $(-3,-1)$, have their highest value right at the event represented by the time windows (0, 3) and stay on a higher level for the next 20 days (1, 20). Thereafter the liquidity measures return to lower levels in the period (21, 60). The transaction costs measures based on trading data (i.e., Roll, Amihud and price dispersion measure) indicate the same magnitude of around 10 to 20 bp for a round-trip transaction. For example, the price dispersion measure starts with 15.4 bp, has its maximum directly at the event with 20.5 bp and returns to 17 bp at the end of our time window. The trading costs represented by bid-ask spreads based on quotations show a similar behavior, but are much higher in magnitude with around 50 bp. Additionally, the period before the event shows an especially high trading activity of around 38 million EUR. Interestingly, this high trading activity comes together with increasing sell-side activity, measured by net-bid, which peaks just before the event in the window $(-3,-1)$. Directly after the event, the trading volume reduces to 36 million EUR and we find on average more buy-side activity. Thus, some investors might sell to avoid political risk before the event and this trend is reversed after the event. Thereafter, the trading volume returns to lower levels with a more balanced trading activity in the time window (21, 60). Table 5 presents the differences between the individual periods and provides statistical tests for these difference. As discussed, we find that the measures representing trading costs based on transactions slightly increase in the period right before the event with an increase of sell-side activity. Given the high volatility of the measures in general, we find only marginal statistical significance for these effects. Directly around the event, we find a sharp increase of transaction costs, with significantly more buy-side activity. These differences are all highly significant. After the event, transaction costs and trading activity return to pre-event levels, also these changes are highly significant. Overall, these results are in line with our price effects. The price drop before the event is associated with higher illiquidity and more sell-side activity. The price reversal around the event itself is related to high transaction costs and buy-side activity. Thereafter, prices and liquidity return to pre-event levels.

Table 4

Liquidity of bonds around political events

	Bid-ask	Roll	Amihud	Price dispersion	Volume	Net-bid
$(-60,-21)$	49.873	12.462	8.251	15.379	38.541	1.588
$(-20,-1)$	49.716	12.488	9.032	16.181	38.284	1.988
$(-3,-1)$	54.024	12.226	9.991	16.353	38.244	2.891
(0,3)	57.769	14.589	12.064	20.526	36.038	$-0.516$
(1,20)	58.696	14.540	11.315	18.998	33.751	1.928
(21,60)	50.414	13.650	10.170	17.068	32.870	0.413

This table shows the averages of the trading cost measures, trading volume and net-bid measure in various subperiods around political events of Italian government coupon bonds (BTP). The following time windows are considered: $(-60,-21)$, $(-20,-1)$, $(-3,-1)$, (0, 3), (1, 20) and (21, 60), where 0 is the event day and all other numbers represent the trading days relative to the event day. The liquidity measures are represented by the quoted bid-ask spread, as well as, the Roll, Amihud and price dispersion measure based on transaction data. The liquidity proxies are given in basis points, in the case of the Amihud measure for a trade of one million EUR, and the trading volume and the net-bid measure are measured in millions of EUR. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013

Table 5

Liquidity effects of bonds around political events

	Bid-ask	Roll	Amihud	Price disp	Volume	Net-bid
Panel A: differences
$(-60,-21);(-20,-1)$	$-0.036$	1.250	0.866	0.817	$-1.145$	0.492
$(-3,-1);(0,3)$	4.074	2.727	1.491	3.845	$-1.869$	$-3.372$
(1, 20); (21, 60)	$-8.462$	$-1.190$	$-1.197$	$-2.009$	$-0.687$	$-1.543$
Panel B: T-test: t-values
$(-60,-21);(-20,-1)$	$-0.039$	$2.087^{**}$	$2.497^{**}$	$2.322^{**}$	$-1.548$	0.939
$(-3,-1);(0,3)$	$6.636^{***}$	$2.680^{***}$	$1.775^{*}$	$6.318^{***}$	−1.369	$-2.804^{***}$
(1, 20); (21, 60)	$-11.575^{***}$	$-2.118^{**}$	$-2.946^{***}$	$-5.878^{***}$	$-1.089$	$-3.249^{***}$
Panel C: Wilcoxon test: p-values
$(-60,-21);(-20,-1)$	0.772	0.160	$0.139^{**}$	$0.045^{**}$	$0.086^{**}$	0.619
$(-3,-1);(0,3)$	$0.000^{***}$	$0.011^{**}$	$0.001^{***}$	$0.000^{***}$	0.672	$0.002^{***}$
(1, 20); (21, 60)	$0.000^{***}$	$0.000^{***}$	$0.001^{***}$	$0.000^{***}$	0.535	$0.000^{***}$

This table shows the liquidity effects based on the trading cost measures, trading volume and net-bid measure in various subperiods around political events of Italian government coupon bonds (BTP). The following time windows are considered: $(-60,-21)$, $(-20,-1)$, $(-3,-1)$, (0, 3), (1, 20) and (21, 60), where 0 is the event day and all other numbers represent the trading days relative to the event day. For each time window, the bond-specific average values of the measures are calculated and the changes across combinations of two time windows are computed, measuring liquidity effects before, during and after the events. The transaction cost measures are represented by the quoted bid-ask spread, as well as, the Roll, Amihud and price dispersion measure based on transaction data. These liquidity proxies are given in basis points, in the case of the Amihud measure for a trade of one million EUR, and the trading volume and the net-bid measure are measured in millions of EUR. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013. Panel A provides the average liquidity effects, Panel B t-values of one-sample t-tests and Panel C p-values of one-sample Wilcoxon signed rank tests. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013

The significance is indicated as follows: * $< 0.1$, ** $< 0.05$, *** $< 0.01$

5.2 Regression analysis

In this section, we present the results of our regression analysis. We are particularly interested whether measures of political uncertainty and economic variables are related to the observed price effects. Table 6 shows detailed regression results for the pre-event returns. We focus on the following groups of explanatory variables: political uncertainty measures, market variables, economic variables, liquidity measures and bond characteristics. We present different sets of regressions, each based on different subsets of the variables. The first regression shows the results based on changes in the EPU, VSTOXX index and CDS spread. The second regression represents changes in inflation, GDP growth and 3 M Euribor. The third regression includes liquidity, time-to-maturity and coupon of the bonds. In Regression 4 all variables are included. We focus our discussion on Regression 4. The resulting parameters and significance levels are very similar compared to Regression 1 to 3, indicating that multi-collinearity is of minor importance. The market-based CDS spread is significant at the $1\%$ level and shows the strongest economic effect of all variables with a one-standard deviation move resulting in a return of $-1.92\%$. Focusing on the effect of political uncertainty, we find that the coefficient of the EPU Index is negative and significant at the $1\%$ level, i.e., if uncertainty increases before a political event, this is related to a price decline. The effect is also very strong in economic terms; a one-standard deviation increase in political uncertainty reduces the prices by $0.91\%$. Compared to the observed average price reductions of roughly $1\%$, this effect is very significant, showing that political uncertainty is an important variable linked to pre-event returns.²⁹ Our results also show that the EPU index is not simply picking up an increase in credit risk.³⁰ Our measure for general market volatility represented by the VSTOXX index is only marginally significant in this specification and shows only a minor economic effect. Analyzing the economic variables, we find insignificant results for inflation and GPD growth. However, we find significant results for the 3 M Euribor. Thus, this result shows that the EPU index is not simply picking up changes in the economic variables, but has an effect over and above changes in the risk-free rate. The 3 M Euribor has an economic effect based on a one-standard deviation move of $-0.602\%$. Our bond characteristics also confirm the importance of liquidity, showing a return of $-0.601\%$ for a one-standard deviation increase in illiquidity. In addition, we find a marginally significant coupon effect. Overall, these results show that our price effects can be related to changes in the political uncertainty and in the underlying economic conditions.

Table 6

Pre-event return regressions for bonds

	Dependent variable
	Pre-event return
	(1)	(2)	(3)	(4)
$\Delta \text {EPU}$	−0.033$^{***}$			−0.029$^{***}$
	(0.007)			(0.005)
$\Delta \text {VSTOXX}$	0.143			0.117$^{*}$
	(0.106)			(0.065)
$\Delta \text {CDS}$	−0.044$^{***}$			−0.036$^{***}$
	(0.007)			(0.005)
$\Delta \text {Inflation}$		0.078		−0.340
		(0.918)		(0.398)
$\Delta \text {GDP change}$		0.269		0.146
		(0.280)		(0.103)
$\Delta \text {Euribor}$		−11.235		−5.289$^{***}$
		(7.554)		(1.854)
$\Delta \text {Liquidity}$			−15.248$^{***}$	−5.802$^{***}$
			(3.971)	(1.865)
$\text {TTM}$			−0.0001	−0.0001
			(0.0001)	(0.0001)
$\text {Coupon}$			−0.099	−0.125$^{*}$
			(0.077)	(0.069)
Constant	0.704$^{**}$	−0.842$^{*}$	0.243	1.237$^{**}$
	(0.334)	(0.485)	(0.348)	(0.557)
Observations	838	838	838	838
R$^{2}$	0.628	0.181	0.298	0.738
Adjusted R$^{2}$	0.626	0.178	0.296	0.735
Residual standard error	1.763 (df = 834)	2.615 (df = 834)	2.420 (df = 834)	1.485 (df = 828)

This table shows the results of different regression models, where the dependent variable is the pre-event return of Italian government coupon bonds (BTP). The pre-event time windows are given by $(-60,-21)$ and $(-20,-1)$, where 0 is the event day and all other numbers represent the trading days relative to the event day. For each time window, the bond-specific average values of the total prices are calculated and the returns across combinations of two time windows are computed. The explanatory variables are represented by EPU index, market variables (VSTOXX Index, 5Y CDS spread), economic variables (inflation, GDP growth, 3 M Euribor), liquidity (price dispersion measure) and bond characteristics (TTM and coupon). For the uncertainty measure, market variables, economic variables and liquidity the change of these variables across periods is considered, whereas bond characteristics are represented by their level. The standard errors are given in parentheses and are corrected for heteroscedasticity and clustered at the event level. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013

The significance is indicated as follows: * $< 0.1$, ** $< 0.05$, *** $< 0.01$

In addition, we provide regressions based on all variables for the event and post-event returns. These results are presented in Table 7.³¹ For the event returns, we find that the two most important variables are again the EPU index and CDS spread. The economic effect based on a one standard deviation move is $-0.32\%$ and $-0.83\%$, respectively. Thus, we find that political uncertainty is still important during the event. Considering that the event will reduce the political uncertainty, this effect basically represents an average price increase given a reduction in the EPU index. Additionally, credit risk also remains an important price factor during the event. Analyzing the post-event returns reveals interesting results as well. The parameter of political uncertainty is now insignificant. Thus, after the event the effect of changes in uncertainty is not related to prices, as no immediate decisions are made. However, the effect of credit risk is now much stronger with an economic effect of $-2.11\%$, indicating that resolving impact uncertainty reveals the consequence of potential policy changes on the credit risk of the considered bonds.³² Thus, we find that the results for the event and post-event returns are in line with the theoretical literature.

Table 7

Subperiod regressions for bonds

	Dependent variable
	Pre-event return	Event return	Post-event return
	(1)	(2)	(3)
$\Delta \text {EPU}$	−0.029$^{***}$	−0.010$^{**}$	−0.004
	(0.005)	(0.004)	(0.009)
$\Delta \text {VSTOXX}$	0.117$^{*}$	−0.042	0.145$^{**}$
	(0.065)	(0.032)	(0.068)
$\Delta \text {CDS}$	−0.036$^{***}$	−0.022$^{***}$	−0.039$^{***}$
	(0.005)	(0.002)	(0.013)
$\Delta \text {Inflation}$	−0.340	0.433	−1.050
	(0.398)	(0.355)	(1.213)
$\Delta \text {GDP change}$	0.146	−0.244$^{***}$	−0.992$^{**}$
	(0.103)	(0.090)	(0.450)
$\Delta \text {Euribor}$	−5.289$^{***}$	10.309$^{*}$	3.589
	(1.854)	(5.837)	(2.988)
$\Delta \text {Liquidity}$	−5.802$^{***}$	0.066	−5.368$^{***}$
	(1.865)	(0.395)	(1.431)
$\text {TTM}$	−0.0001	0.00005	0.0001
	(0.0001)	(0.0001)	(0.0001)
$\text {Coupon}$	−0.125$^{*}$	0.061	−0.031
	(0.069)	(0.037)	(0.074)
Constant	1.237$^{**}$	0.046	0.828
	(0.557)	(0.281)	(0.549)
Observations	838	733	827
R$^{2}$	0.738	0.555	0.508
Adjusted R$^{2}$	0.735	0.550	0.503
Residual standard error	1.485 (df = 828)	0.975 (df = 723)	1.906 (df = 817)

This table shows the results of different regression models, where the dependent variable is the return of Italian government coupon bonds (BTP) across various subperiods. The pre-event return is defined by the time windows $(-60,-21)$ and $(-20,-1)$, the event return by $(-3,-1)$ and (0, 3), and the post-event return by (1, 20) and (21, 60), where 0 is the event day and all other numbers represent the trading days relative to the event day. For each set of time windows, the bond-specific average values of the total prices are calculated and the returns across combinations of two time windows are computed. The explanatory variables are represented by EPU index, market variables (VSTOXX index, 5Y CDS spread), economic variables (inflation, GDP growth, CDS spreads, 3 M Euribor), liquidity (price dispersion measure) and bond characteristics (TTM and coupon). For the uncertainty measure, market variables, economic variables and liquidity the change of these variables across periods is considered, whereas bond characteristics are represented by their level. The standard errors are given in parentheses and are corrected for heteroscedasticity and clustered at the event level. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013

The significance is indicated as follows: * $< 0.1$, ** $< 0.05$, *** $< 0.01$

Analyzing the pre-event returns in more detail, we explore events that are potentially more closely related to Italy. Here, we focus on events with topics more important to Italy, e.g., banking resolution, and consider events with descriptions like “Greek Crisis” and “Greek Election” as less related (see Table 1). We include a dummy variable for more closely related events and add an interaction term between this dummy and the EPU index in our regression analysis. The results are presented in Table 8. The coefficients of the dummy variable and the interaction term are both insignificant, indicating that all our considered events are relevant in the context of the European sovereign debt crisis. Thus, also uncertainty from events not directly related to Italy, like the Greek elections, is impacting Italian government bond prices, documenting policy uncertainty spillovers.

Table 8

Pre-event return regressions for bonds—Italian events

	Dependent variable
	Pre-event return
	(1)	(2)	(3)	(4)
$\Delta \text {EPU}$	−0.029$^{***}$	−0.028$^{***}$	−0.027$^{***}$	−0.027$^{***}$
	(0.005)	(0.004)	(0.007)	(0.008)
Italian-event dummy		−0.268		−0.232
		(0.847)		(0.813)
$\Delta \text {EPU}$ $\cdot$ dummy			−0.005	−0.002
			(0.022)	(0.023)
$\Delta \text {VSTOXX}$	0.117$^{*}$	0.103	0.111	0.102
	(0.065)	(0.083)	(0.076)	(0.086)
$\Delta \text {CDS}$	−0.036$^{***}$	−0.038$^{***}$	−0.036$^{***}$	−0.037$^{***}$
	(0.005)	(0.004)	(0.006)	(0.005)
$\Delta \text {Inflation}$	−0.340	−0.290	−0.358	−0.305
	(0.398)	(0.488)	(0.393)	(0.494)
$\Delta \text {GDP change}$	0.146	0.111	0.137	0.112
	(0.103)	(0.094)	(0.102)	(0.089)
$\Delta \text {EURIBOR}$	−5.289$^{***}$	−4.981$^{***}$	−4.832$^{*}$	−4.823$^{*}$
	(1.854)	(1.556)	(2.470)	(2.459)
$\Delta \text {Liquidity}$	−5.802$^{***}$	−5.773$^{***}$	−5.677$^{***}$	−5.722$^{***}$
	(1.865)	(1.776)	(1.896)	(1.841)
TTM	−0.0001	−0.0001	−0.0001	−0.0001
	(0.0001)	(0.0001)	(0.0001)	(0.0001)
Coupon	−0.125$^{*}$	−0.125$^{*}$	−0.125$^{*}$	−0.125$^{*}$
	(0.069)	(0.069)	(0.068)	(0.069)
Constant	1.237$^{**}$	1.379$^{**}$	1.241$^{**}$	1.361$^{**}$
	(0.557)	(0.687)	(0.557)	(0.671)
Observations	838	838	838	838
R$^{2}$	0.740	0.741	0.740	0.741
Adjusted R$^{2}$	0.737	0.737	0.737	0.737
Residual standard error	1.513 (df = 828)	1.512 (df = 827)	1.513 (df = 827)	1.513 (df = 826)

This table shows the results of different regression models, where the dependent variable is the pre-event return of Italian government coupon bonds (BTP). The pre-event time windows are given by $(-60,-21)$ and $(-20,-1)$, where 0 is the event day and all other numbers represent the trading days relative to the event day. For each time window, the bond-specific average values of the total prices are calculated and the returns across combinations of two time windows are computed. The explanatory variables are represented by the EPU index, market variables (VSTOXX Index, 5Y CDS spread), economic variables (inflation, GDP growth, 3 M Euribor), liquidity (price dispersion measure) and bond characteristics (TTM and coupon). For the uncertainty measure, market variables, economic variables and liquidity the change of these variables across periods is considered, whereas bond characteristics are represented by their level. In addition, a dummy indicating whether the event had a particularly strong focus on Italy, as well as its interactions with the EPU index is added. The standard errors are given in parentheses and are corrected for heteroscedasticity and clustered at the event level. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013

The significance is indicated as follows: * $< 0.1$, ** $< 0.05$, *** $< 0.01$

5.3 Robustness tests

In this section, we discuss our set of robustness tests analyzing different factors and considering different methodologies to confirm our findings. The tables presenting the results of these checks are reported in Appendix to conserve space. We focus on the price impact of political uncertainty in these robustness tests and, thus, mainly report how different specifications affect the results for the pre-event period. Overall, we find that all our main results hold within these robustness tests.³³

Event Selection: In the original event selection, we exclude events that would have overlapping time windows in $(-3, 3)$ avoiding that the event return itself is affected. However, overlaps in the remaining time windows are still possible. In this robustness test, we provide different event selections to analyze the potential impact of these overlaps on our results. However, a simple exclusion of events by avoiding overlaps in the entire considered windows of 121 trading days would result in too few events, i.e., less than five events. Thus, we eliminate overlapping events within 41 trading days, avoiding, in particular, overlaps in the pre-event period, where we find the most severe price decrease. Table 9 provides the price effects for this set of events. We find basically the same results, e.g., the price effect of the pre-event period is $-81.4$ bp compared to $-93.6$ bp in the original specification.³⁴ Table 10 shows the corresponding regression analysis and, again, we find identical results. In addition, we also analyze whether our results depend on certain event types. We first focus on Euro summits, as this group represents the most important and consistent group compared to elections or G8/G20 summits. Furthermore, we analyze the two combinations of Euro summits with G8/G20 summits and Euro summits with elections, as G8/G20 summits and elections cannot be meaningfully analyzed on a stand-alone basis as they have too few observations. The results for all these subsets are shown in Table 10 and confirm our main results.

Time Windows: In this robustness test, we analyze the impact of different time window definitions on the pre-event returns. In all specifications, we use the time window $(-20, -1)$ as the agenda of summits is released four weeks in advance and, thus, we see no reason to widen this window, as we would potentially include trading days, where even the agenda is not known yet. As the price trend is negative, shortening this window would only make the results more pronounced. However, we provide different specifications for the time window $(-60, -21)$ using instead the windows $(-60, -41)$ and $(-40, -21)$. The price effect of the pre-event period based on these two windows is $-97$ bp and $-86$ bp, respectively. Both returns are significant at the $1\%$ level. The results of the regression analysis based on these windows are presented in Table 11. The results are basically identical compared to the original specification. Only the impact of liquidity is slightly smaller for $(-40, -21)$, as the bond market in this time window is slightly less liquid than in $(-60,-41)$.

Regression Analysis: In our regression analysis, we focus on pooled regressions explaining the observed price return across various defined periods. As an alternative specification, we employ a regression analysis, where we explore the bond-specific daily return series by stacking these observations across all bond-event combinations. We use time dummy variables in addition to our other explanatory variables to analyze whether the returns are significantly different in certain time periods. In particular, we define three sets of regressions where we consider dummies for the time windows $(-20, -1)$, (0, 3) and (1, 20). As we analyze daily returns, we only include explanatory variables based on market data, for which we have daily observations available. Table 12 presents the results. We find similar results compared to our original specification. There is a price drop of 1.8 bp in the pre-event window and a increase of 3.3 bp in the event window. Note that the magnitudes are much smaller compared to the original analysis as these values represent daily changes in the time window and not the overall change between two windows.

Short-term zero-coupon Bonds: The Italian government also issues short-term discount bonds with a maturity of up to one year (Buoni Ordinari del Tesoro) and of two years (Certificato del Tesoro Zero-coupon). In this robustness test, we analyze this set of bonds. We expect to find much smaller effects given the short maturities of these bonds. However, the general impact of political uncertainty on their prices should be similar. Table 13 presents these price effects, Table 14 and 15 provide the liquidity measures, and Tables 16 and 17 show the regression results. We find negative price returns in the pre-event period of 4.5 bp and positive effects in the event and post-event period. Thus, the results are similar compared to the original specification, however, on a much smaller magnitude, as expected. The liquidity measures confirm a significant increase of sell-side activity before the event. The regression results show a relation between prices and political uncertainty, but we find only marginal statistical significance. Overall, the results are similar for short-term bonds, but not all results can be replicated given the small magnitude of the price effects.

Reverse Causality: Ideally, the events for our analysis should be planned long in advance with no subsequent changes to the agenda. However, short-term adjustments of the event schedule are possible. Thus, our analysis could be subject to reverse causality concerns if the yields are not affected by the events, but instead the events are scheduled in response to changing (and in particular rising) yields. To address this concern, we study several probit models trying to predict the likelihood of an event. To perform this analysis, we create an indicator variable that is equal to one in weeks in which an event occurred and zero otherwise.³⁵ We use the spread between the 5-year Italian and German zero-coupon yield as our explanatory variable and test whether averages or changes in certain time windows before the events have power in predicting the likelihood of an event.³⁶ Specifically, we analyze the effect of the averages in the time windows $(-60, -41)$, $(-40, -21)$ and $(-20, -1)$ as well as the differences between these time windows. We find that the effect of the yield spread is insignificant in all specifications, suggesting that the events are not scheduled in direct response to sovereign yields.³⁷ The full results are shown in Appendix in Table 23.

5.4 Implications for issuing costs

In this section, we investigate the implications of our results on the issuing activities in the primary market. The price levels reported in the various time windows for the secondary market will most likely be important for investors in the primary market, especially because many dealers active in this market sell off their positions in the secondary market. Thus, issuing bonds in times of high political uncertainty might result in lower prices, representing additional costs. The Italian government coupon bonds are particularly well suited for the analysis of such effects, as much of the issuance activity is organized as seasoned bond offerings. This allows us to observe secondary market prices even before the (re-)issuance of the given bonds.

In our analysis, we compare the prices of seasoned primary market auctions occurring directly before our set of political events, i.e., in the time period $(-20, -1)$, with the prices of the same bonds during the period $(-60,-21)$. Thus, we directly estimate the costs of issuing in times of uncertainty, assuming that these costs could be avoided by issuing before the political uncertainty affects prices. We find that 155 billion EUR were issued in the period $(-20, -1)$ in our sample by seasoned offerings. The overall issuance volume in the considered time period is 193 billion EUR. Thus, seasoned offerings represent the vast majority of bond issuances in our case. Given the frequency of our events (roughly four per year, i.e., the considered time windows cover 80 trading days per year), we find that the periods before important events show the same average issuing activity as regular periods. Thus, the debt management office seems not to avoid (or focus) on issuing bonds in these periods. In our analysis, we find an average price effect of $-1.74\%$ when comparing the prices in the period $(-60, 21 )$ to the auction prices. Note that this effect is stronger than in our previous analysis, indicating that bonds with issuance activity are even more affected by uncertainty, on average.

Given the price effects of $-1.74\%$, we assume that this amount represents the potential magnitude lost by issuing a bond in times of political uncertainty.³⁸ Thus, we find direct costs of roughly 2.70 billion EUR that could be avoided by considering important political events in the issuing activity within our sample period. Thus, a more active cash and debt management by countries could save substantial amounts, as investors demand significant risk premiums for political uncertainty.

However, avoiding the affected time periods could cause opportunity costs as the debt management office has to deviate from a planned issuing strategy. This deviation would often be on short notice, as the agenda of summits is not known much in advance and there can be early elections. In this context, Eisl et al. (2019) discuss issuance strategies and cash buffers of debt management offices. In their model, increasing the cash buffer, e.g., by issuing before the planned issuance date, comes with opportunity costs representing costs for adjusting the auction calendar, higher administrative costs and other costs, e.g., fees compensating market makers for funding constraints and inventory risk. The paper provides an estimate of $0.39\%$ for these costs for developed countries, based on a literature review. In addition, issuing a bond earlier makes it necessary to increases the cash buffer. Eisl et al. (2019) indicate that that these buffers are normally invested in the money market. Thus, an additional opportunity costs stem from the difference between the money market and coupon rate for the period from the issuance to optimal issuance date. Issuing the affected bonds during the period $(-60,-21)$ instead of $(-20,-1)$ results on average in an 30 days earlier issuance. Thus, we calculate the difference between the relevant 1 M Euribor and coupon rate for this period. This results in average additional opportunity costs of $0.42\%$ in our sample and, thus, the potential reduction of issuing costs would be $-0.93\%$. Thus, even after considering the potential effects of deviating from the planned issuing strategy, we find costs of 1.44 billion EUR related to political uncertainty. However, even though this analysis highlights that the DMOs should take these uncertainty effects into account, this does not necessarily mean that it would be optimal to avoid issuing debt during these times. In case the effect on the utility of the DMO is similar in comparison with the bond investors with respect to the event outcomes, then the issuing policy could still be optimal regardless of the timing.

6 Conclusion

In this paper, we study whether political uncertainty affects prices and liquidity in government bond markets. We explore the time period of the European sovereign debt crisis, as it is an ideal laboratory for such an analysis, because investors were exposed to significant political uncertainty and many political events at this time were almost exclusively focusing on policies handling this crisis. We cover the peak of the crisis from 2010 to 2013, including political summits (Euro, G8 and G20) and relevant elections. We analyze the Italian sovereign bond market, which represents a major European bond market with a notional amount outstanding of around two trillion EUR, comparable in size with Germany and France. Furthermore, Italian government bonds are traded on the MTS platform and, thus, detailed transaction data are available. This provides us with representative high-frequency data on quotes and transactions which allows us to calculate a variety of liquidity measures and conduct an in-depth investigation of price movements.

We analyze price and liquidity effects before the events, directly around the events and after the events. This allows us to analyze the effect of uncertainty around policy decisions. We find strong negative price reactions of Italian government bonds before the events. On average, prices fall by around $1\%$ in the time window 20 days before the event. Directly around the event (i.e., three days before/after the event) we find significant positive price impacts of around $0.4\%$. Thus, we observe a positive effect as soon as the uncertainty about the policy decision is resolved to certain degree. However, prices do not further recover for the next 20 trading days showing that a significant uncertainty concerning the impact and precise implementation of policies remains. Within 60 days after the events, prices fully recover to pre-event levels. In an additional analysis based on less detailed data, we find similar effects for government bonds issued by Greece, Ireland, Portugal and Spain.

In addition, we study trading activity measures representing liquidity measures related to transaction costs and volume-based measures for the Italian government bond market. We find that our liquidity measures indicate higher illiquidity shortly before the event when prices tend to fall, with the highest illiquidity being observed directly around the event, e.g., the price dispersion measure increases from 15 bp to 20 bp. The trading costs return to previous levels after 60 days in line with the observed price recovery. Analyzing the trading volume, we observe a significant higher volume of around $5\%$ before the event. This coincides with a much higher sell-side trading activity, in particular three days before the events. Thus, we can show that the observed price movements coincide with sell-side pressure.

In our regression analysis, we relate the observed pre-event, event and post-event returns to variables measuring political uncertainty and representing the economic conditions at the time of the event. We find a significant effect of the EPU index, measuring uncertainty based on newspaper articles, for pre-event returns where an increase in uncertainty is related to falling bond prices. In addition, weaker economic conditions are related to stronger price reductions. Interestingly, the effect of the EPU index fades out and is not present in the post-event returns indicating that prices only react to political uncertainty in the context of upcoming events and not unconditionally at all times.

Based on our results, we explore how much notional volume was issued by the Italian government in the period 20 days before the events, when prices are severely affected. In these periods, 193 billion EUR of notional volume were issued, thereof 155 billion EUR as seasoned bond offerings. In our analysis, we find costs of 1.44 billion EUR for exposing primary dealers and investors to uncertainty. Thus, we find significant costs stemming from these issuing activities. These results show that uncertainty related to political events should be considered in the issuance policy of debt management offices, as such costs could potentially be avoided by more active cash management. Overall, the results in this paper foster our understanding of the effect of political uncertainty, enriching the results presented for stock and option markets in the empirical literature so far.

Acknowledgements

We thank two anonymous referees, the editor Markus Schmid, Pedro Barroso, Stefan Bogner, Nataliya Gerasimova, Adam Golinski, Yalin Gunduz, Kurt Hornik, Ruggero Jappelli, Alexander Pasler, Florian Pauer, Stefan Pichler, Otto Randl, Pietro Veronesi, Patrick Weiss and participants at the Portuguese Finance Network 2018, DGF 2018, AWG 2018, EFA 2019, EFMA 2019, and SFA 2022 for helpful comments and suggestions. All errors remain our own.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

previous article Financial knowledge acquisition and trading behavior: empirical evidence from an online information tool

next article Second hand or second generation? The performance of secondary buyouts

Appendix

See Tables 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24.

Table 9

Price effects of bonds around non-overlapping political events

	Price returns	Total price returns	Abnormal returns w.r.t. Germany	Returns of other affected markets	Abn. returns of other affected markets
Panel A: returns
$(-60,-21);(-20,-1)$	$-0.813$	$-0.395$	$-0.943$	$-1.104$	$-1.651$
$(-3,-1);(0,3)$	0.556	0.589	0.540	0.794	0.745
(1, 20); (21, 60)	0.814	1.079	0.251	0.917	0.089
Panel B: T-test: t-values
$(-60,-21);(-20,-1)$	$-5.163^{***}$	$-2.500^{**}$	$-5.328^{***}$	$-5.741^{***}$	$-7.636^{***}$
$(-3,-1);(0,3)$	$6.985^{***}$	$7.361^{***}$	$6.291^{***}$	$5.807^{***}$	$4.983^{***}$
(1, 20); (21, 60)	$5.455^{***}$	$7.071^{***}$	1.304	$4.524^{***}$	0.348
Panel C: Wilcoxon test: p-values
$(-60,-21);(-20,-1)$	0.000$^{***}$	0.000$^{***}$	0.000$^{***}$	0.006$^{***}$	0.003$^{***}$
$(-3,-1);(0,3)$	0.000$^{***}$	0.002$^{***}$	0.000$^{***}$	0.602	0.732
(1,20);(21,60)	0.000$^{***}$	0.000$^{***}$	0.356	0.000$^{***}$	0.153

This table shows the price returns around political events of Italian government coupon bonds (BTP) excluding overlapping events within 41 trading days. The following time windows around the events are considered: $(-60,-21)$, $(-20,-1)$, $(-3,-1)$, (0, 3), (1, 20) and (21, 60), where 0 is the event day and all other numbers represent the trading days relative to the event day. For each time window, the bond-specific average values of the prices are calculated and the returns across combinations of two time windows are computed, measuring price effects before, during and after the events. The returns are based on reported (clean) prices and total (dirty) prices. In addition, the abnormal returns with respect to German government bond prices are presented based on total prices. Furthermore, returns and abnormal returns for other highly affected sovereign bond markets, i.e., Greece, Ireland, Portugal and Spain, are presented based on less detailed data. Panel A provides the average price effects measured in percent, Panel B t-values of one-sample t-tests and Panel C p-values of one-sample Wilcoxon signed rank tests. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013

The significance is indicated as follows: * $< 0.1$, ** $< 0.05$, *** $< 0.01$

Table 10

Regressions for various event subsets

	Dependent variable
	Pre_Event_Return
	(No overlap)	(Euro-summits)	(Summits)	(Euro summits & elections)	(All events)
$\Delta \text {EPU}$	−0.045$^{***}$	−0.038$^{***}$	−0.034$^{***}$	−0.029$^{***}$	−0.029$^{***}$
	(0.013)	(0.009)	(0.005)	(0.007)	(0.005)
$\Delta \text {VSTOXX}$	−0.097	0.111	0.092	0.126$^{*}$	0.117$^{*}$
	(0.210)	(0.108)	(0.057)	(0.073)	(0.065)
$\Delta \text {CDS}$	−0.035$^{***}$	−0.032$^{***}$	−0.034$^{***}$	−0.035$^{***}$	−0.036$^{***}$
	(0.003)	(0.007)	(0.004)	(0.005)	(0.005)
$\Delta \text {Inflation}$	−2.040	−0.154	−0.117	−0.352	−0.340
	(2.279)	(0.349)	(0.373)	(0.367)	(0.398)
$\Delta \text {GDP change}$	0.046	0.243	0.201$^{**}$	0.140	0.146
	(0.157)	(0.180)	(0.093)	(0.178)	(0.103)
$\Delta \text {Euribor}$	0.312	−7.169	−6.004$^{***}$	−5.901$^{**}$	−5.289$^{***}$
	(3.977)	(5.036)	(1.999)	(2.635)	(1.854)
$\Delta \text {Liquidity}$	−7.104$^{***}$	−6.100$^{***}$	−5.618$^{***}$	−6.323$^{***}$	−5.802$^{***}$
	(1.345)	(1.700)	(1.843)	(1.891)	(1.865)
$\text {TTM}$	−0.00003	−0.00005	−0.0001	−0.00005	−0.0001
	(0.0001)	(0.0001)	(0.0001)	(0.0001)	(0.0001)
$\text {Coupon}$	−0.036	−0.069	−0.147$^{*}$	−0.074	−0.125$^{*}$
	(0.092)	(0.091)	(0.076)	(0.073)	(0.069)
Constant	0.214	0.519	1.130$^{*}$	0.838	1.237$^{**}$
	(0.736)	(0.800)	(0.613)	(0.647)	(0.557)
Observations	441	494	736	646	838
R$^{2}$	0.779	0.768	0.756	0.749	0.740
Adjusted R$^{2}$	0.774	0.764	0.753	0.745	0.737
Residual standard error	1.577 (df = 431)	1.702 (df = 484)	1.541 (df = 726)	1.609 (df = 636)	1.513 (df = 828)

This table shows the results of different regression models, where the dependent variable is the pre-event return of Italian government coupon bonds (BTP). The pre-event time windows are given by $(-60,-21)$ and $(-20,-1)$, where 0 is the event day and all other numbers represent the trading days relative to the event day. For each time window, the bond-specific average values of the total prices are calculated and the returns across combinations of two time windows are computed. The explanatory variables are represented by EPU index, market variables (VSTOXX index, 5Y CDS spread), economic variables (inflation, GDP growth, CDS spreads, 3 M Euribor), liquidity (price dispersion measure) and bond characteristics (TTM and coupon). For the uncertainty measure, market variables, economic variables and liquidity the change of these variables across periods is considered, whereas bond characteristics are represented by their level. The standard errors are given in parentheses and are corrected for heteroscedasticity and clustered at the event level. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013. The table shows the results for various event subsets separately, i.e., for non-overlapping events, Euro-Summits, all summits as well as Euro-Summits and elections. The last column includes all events for comparison

The significance is indicated as follows: * $< 0.1$, ** $< 0.05$, *** $< 0.01$

Table 11

Regressions for different pre-event windows

	Dependent variable
	Pre-event return
	$(-60,-41)$	$(-40,-21)$	$(-60,-21)$
$\Delta \text {EPU}$	−0.023$^{**}$	−0.025$^{***}$	−0.029$^{***}$
	(0.010)	(0.004)	(0.005)
$\Delta \text {VSTOXX}$	0.053	0.103	0.117$^{*}$
	(0.105)	(0.063)	(0.065)
$\Delta \text {CDS}$	−0.031$^{***}$	−0.035$^{***}$	−0.036$^{***}$
	(0.008)	(0.006)	(0.005)
$\Delta \text {Inflation}$	−0.113	−0.142	−0.340
	(0.460)	(0.497)	(0.398)
$\Delta \text {GDP change}$	0.198	0.312	0.146
	(0.154)	(0.202)	(0.103)
$\Delta \text {Euribor}$	−5.517$^{**}$	−6.453$^{***}$	−5.289$^{***}$
	(2.242)	(2.108)	(1.854)
$\Delta \text {Liquidity}$	−8.209$^{***}$	−3.568$^{**}$	−5.802$^{***}$
	(1.889)	(1.514)	(1.865)
$\text {TTM}$	−0.0001	−0.0001	−0.0001
	(0.0001)	(0.0001)	(0.0001)
$\text {Coupon}$	−0.162	−0.084	−0.125$^{*}$
	(0.098)	(0.059)	(0.069)
Constant	1.603$^{**}$	0.734$^{*}$	1.237$^{**}$
	(0.735)	(0.412)	(0.557)
Observations	829	838	838
R$^{2}$	0.738	0.649	0.740
Adjusted R$^{2}$	0.735	0.645	0.737
Residual standard error	1.934 (df = 819)	1.410 (df = 828)	1.513 (df = 828)

This table shows the results of different regression models, where the dependent variable is the pre-event return of Italian government coupon bonds (BTP). The different pre-event time windows are given by $(-60,-41)$, $(-40,-21)$ and $(-60,-21)$ versus $(-20,-1)$, where 0 is the event day and all other numbers represent the trading days relative to the event day. For each time window, the bond-specific average values of the total prices are calculated and the returns across combinations of two time windows are computed. The explanatory variables are represented by EPU index, market variables (VSTOXX index, 5Y CDS spread), economic variables (inflation, GDP growth, CDS spreads, 3 M Euribor), liquidity (price dispersion measure) and bond characteristics (TTM and coupon). For the uncertainty measure, market variables, economic variables and liquidity the change of these variables across periods is considered, whereas bond characteristics are represented by their level. The standard errors are given in parentheses and are corrected for heteroscedasticity and clustered at the event level. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013

The significance is indicated as follows: * $< 0.1$, ** $< 0.05$, *** $< 0.01$

Table 12

Multivariate regression with time dummies

	Dependent variable
	Returns
	(1)	(2)	(3)	(4)
$\Delta \text {VSTOXX}$	0.013$^{***}$	0.013$^{***}$	0.013$^{***}$	0.013$^{***}$
	(0.001)	(0.001)	(0.001)	(0.001)
$\Delta \text {CDS}$	$-$0.024$^{***}$	$-$0.024$^{***}$	$-$0.024$^{***}$	$-$0.024$^{***}$
	(0.0002)	(0.0002)	(0.0002)	(0.0002)
$\Delta \text {Euribor}$	$-$1.068$^{***}$	$-$0.994$^{***}$	$-$1.078$^{***}$	$-$1.060$^{***}$
	(0.299)	(0.300)	(0.301)	(0.302)
$(-20,-1)$	$-$0.018$^{***}$			$-$0.018$^{***}$
	(0.005)			(0.005)
$(0,3)$		0.027$^{**}$		0.033$^{***}$
		(0.012)		(0.013)
$(1,20)$			$-$0.004	$-$0.009
			(0.006)	(0.006)
Observations	50,891	50,891	50,891	50,891
R$^{2}$	0.282	0.282	0.282	0.282
Adjusted R$^{2}$	0.282	0.282	0.282	0.282
Residual standard error	0.512 (df = 50887)	0.512 (df = 50887)	0.512 (df = 50887)	0.512 (df = 50885)
F Statistic	5,005.619$^{***}$ (df = 4; 50887)	5,003.677$^{***}$ (df = 4; 50887)	5,002.063$^{***}$ (df = 4; 50887)	3,338.627$^{***}$ (df = 6; 50885)

This table shows the results of different regression models, where the dependent variable is the daily return of Italian government coupon bonds (BTP) in the time window of 60 trading days before to 60 trading days after the various political events. The explanatory variables are the daily changes in the VSTOXX, CDS spreads and 3M Euribor as well as time dummies for the windows $(\text {-}20,\text {-}1)$, (0, 3) and (1, 20). The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013

The standard errors are given in parentheses and the significance is indicated as follows: * $< 0.1$, ** $< 0.05$, *** $< 0.01$

Table 13

Price effects of zero-coupon bonds around political events

	Price returns	Total price returns	Abnormal returns w.r.t. Germany
Panel A: returns
$(-60,-21);(-20,-1)$	$-0.045$	$-0.045$	$-0.075$
$(-3,-1);(0,3)$	0.046	0.046	0.035
(1, 20); (21, 60)	0.176	0.176	0.144
Panel B: T-test: t-values
$(-60,-21);(-20,-1)$	$-1.418$	$-1.418$	$-2.221^{**}$
$(-3,-1);(0,3)$	$3.598^{***}$	$3.598^{***}$	$2.659^{***}$
(1,20);(21,60)	$7.155^{***}$	$7.155^{***}$	$5.675^{***}$
Panel C: Wilcoxon test: p-values
$(-60,-21);(-20,-1)$	0.000$^{***}$	0.000$^{***}$	0.000$^{**}$
$(-3,-1);(0,3)$	0.000$^{***}$	0.000$^{***}$	0.003$^{***}$
(1,20);(21,60)	0.000$^{***}$	0.000$^{***}$	0.000$^{***}$

This table shows the price returns around political events of Italian government zero-coupon bonds (BOT/CTZ). The following time windows around the events are considered: $(-60,-21)$, $(-20,-1)$, $(-3,-1)$, (0, 3), (1, 20) and (21, 60), where 0 is the event day and all other numbers represent the trading days relative to the event day. For each time window, the bond-specific average values of the prices are calculated and the returns across combinations of two time windows are computed, measuring price effects before, during and after the events. The returns are based on reported (clean) prices and total (dirty) prices. In addition, the abnormal returns with respect to German government bond prices are presented based on total prices. Panel A provides the average price effects measured in percent, Panel B t-values of one-sample t-tests and Panel C p-values of one-sample Wilcoxon signed rank tests. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013

The significance is indicated as follows: * $< 0.1$, ** $< 0.05$, *** $< 0.01$

Table 14

Liquidity of zero-coupon bonds around political events

	Bid-ask	Roll	Amihud	Price dispersion	Volume	Net-bid
$(-60,-21)$	15.257	2.160	0.780	3.384	65.226	$-4.704$
$(-20,-1)$	14.111	2.019	0.813	3.207	53.647	$-1.546$
$(-3,-1)$	13.906	2.490	0.944	3.321	46.835	4.817
(0,3)	16.245	2.732	1.076	3.738	46.317	$-3.053$
(1,20)	17.465	2.393	1.127	3.704	43.506	$-1.755$
(21,60)	12.744	1.810	0.736	2.579	48.196	$-7.835$

This table shows the averages of the transaction cost measures, trading volume and net-bid measure in various subperiods around political events of Italian government zero-coupon bonds (BOT/CTZ). The following time windows are considered: $(-60,-21)$, $(-20,-1)$, $(-3,-1)$, (0, 3), (1, 20) and (21, 60), where 0 is the event day and all other numbers represent the trading days relative to the event day. The liquidity measures are represented by the quoted bid-ask spread, as well as the Roll, Amihud and price dispersion measure based on transaction data. The liquidity proxies are given in basis points, in the case of the Amihud measure for a trade of one million EUR, and the trading volume and the net-bid measure are measured in millions of EUR. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013

Table 15

Liquidity effects of zero-coupon bonds around political events

	Bid-ask	Roll	Amihud	Price disp	Volume	Net-bid
Panel A: differences
$(-60,-21);(-20,-1)$	$-0.933$	$-0.106$	0.065	$-0.174$	$-17.783$	3.001
$(-3,-1);(0,3)$	2.339	0.457	0.101	0.460	$-0.487$	$-8.548$
(1,20);(21,60)	$-5.940$	$-0.804$	$-0.476$	$-1.391$	5.986	$-6.691$
Panel B: T-test: t-values
$(-60,-21);(-20,-1)$	$-1.127$	$-0.594$	0.827	$-1.007$	$-7.596$	$2.308^{**}$
$(-3,-1);(0,3)$	$5.463^{***}$	0.785	2.375	$2.375^{**}$	$-0.172$	$-3.522^{***}$
(1,20);(21,60)	$-8.944^{***}$	$-4.335^{***}$	$-6.526^{***}$	$-8.678^{***}$	$2.670^{***}$	$-3.172^{***}$
Panel C: Wilcoxon test: p-values
$(-60,-21);(-20,-1)$	$0.002^{***}$	0.110	$0.505^{*}$	$0.009^{***}$	$0.000^{***}$	$0.096^{*}$
$(-3,-1);(0,3)$	$0.000^{***}$	0.902	$0.033^{**}$	$0.022^{**}$	0.730	$0.001^{**}$
(1,20);(21,60)	$0.000^{***}$	$0.000^{***}$	$0.000^{***}$	$0.003^{***}$	$0.000^{***}$	$0.001^{***}$

This table shows the liquidity effects based on the trading cost measures, trading volume and net-bid measure in various subperiods around political events of Italian government zero-coupon bonds (BOT/CTZ). The following time windows are considered: $(-60,-21)$, $(-20,-1)$, $(-3,-1)$, (0, 3), (1, 20) and (21, 60), where 0 is the event day and all other numbers represent the trading days relative to the event day. For each time window, the bond-specific average values of the measures are calculated and the changes across combinations of two time windows are computed, measuring liquidity effects before, during and after the events. The transaction cost measures are represented by the quoted bid-ask spread, as well as the Roll, Amihud and price dispersion measure based on transaction data. These liquidity proxies are given in basis points, in the case of the Amihud measure for a trade of one million EUR, and the trading volume and the net-bid measure are measured in millions of EUR. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013. Panel A provides the average liquidity effects, Panel B t-values of one-sample t-tests and Panel C p-values of one-sample Wilcoxon signed rank tests. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013

The significance is indicated as follows: * $< 0.1$, ** $< 0.05$, *** $< 0.01$

Table 16

Pre-event return regressions for zero-coupon bonds

	Dependent variable
	Pre-event return
	(1)	(2)	(3)	(4)
$\Delta \text {EPU}$	−0.005$^{*}$			−0.002
	(0.003)			(0.002)
$\Delta \text {VSTOXX}$	−0.017			−0.021
	(0.040)			(0.016)
$\Delta \text {CDS}$	−0.006$^{***}$			−0.003$^{***}$
	(0.002)			(0.001)
$\Delta \text {Inflation}$		0.001		−0.058
		(0.230)		(0.101)
$\Delta \text {GDP change}$		−0.021		−0.042
		(0.073)		(0.027)
$\Delta \text {Euribor}$		−1.991$^{*}$		−0.573
		(1.083)		(0.423)
$\Delta \text {Liquidity}$			−14.274$^{***}$	−10.371$^{***}$
			(2.992)	(1.399)
$\text {TTM}$			0.0002	0.0001
			(0.0004)	(0.0004)
Constant	0.330$^{***}$	0.091	0.096	0.173
	(0.080)	(0.111)	(0.083)	(0.135)
Observations	389	389	389	389
R$^{2}$	0.373	0.138	0.515	0.646
Adjusted R$^{2}$	0.368	0.131	0.513	0.639
Residual standard error	0.523 (df = 385)	0.614 (df = 385)	0.459 (df = 386)	0.396 (df = 380)

This table shows the results of different regression models, where the dependent variable is the pre-event return of Italian government zero-coupon bonds (BOT/CTZ). The pre-event time windows are given by $(-60,-21)$ and $(-20,-1)$, where 0 is the event day and all other numbers represent the trading days relative to the event day. For each time window, the bond-specific average values of the total prices are calculated and the returns across combinations of two time windows are computed. The explanatory variables are represented by EPU index, market variables (VSTOXX Index, 5Y CDS spread), economic variables (inflation, GDP growth, 3 M Euribor), liquidity (price dispersion measure) and bond characteristics (TTM). For the uncertainty measure, market variables, economic variables and liquidity the change of these variables across periods is considered, whereas bond characteristics are represented by their level. The standard errors are given in parentheses and are corrected for heteroscedasticity and clustered at the event level. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013

The significance is indicated as follows: * $< 0.1$, ** $< 0.05$, *** $< 0.01$

Table 17

Subperiod regressions for zero-coupon bonds

	Dependent variable
	Pre-event return	Event return	post-event return
	(1)	(2)	(3)
$\Delta \text {EPU}$	−0.002	−0.002	−0.001
	(0.002)	(0.001)	(0.001)
$\Delta \text {VSTOXX}$	−0.021	−0.003	0.007
	(0.016)	(0.005)	(0.008)
$\Delta \text {CDS}$	−0.003$^{***}$	−0.002$^{*}$	−0.004$^{**}$
	(0.001)	(0.001)	(0.002)
$\Delta \text {Inflation}$	−0.058	0.019	0.127
	(0.101)	(0.062)	(0.139)
$\Delta \text {GDP change}$	−0.042	−0.018	−0.090$^{**}$
	(0.027)	(0.014)	(0.046)
$\Delta \text {Euribor}$	−0.573	1.238	−0.372
	(0.423)	(1.026)	(0.295)
$\Delta \text {Liquidity}$	−10.371$^{***}$	−0.422	−8.480$^{***}$
	(1.399)	(0.551)	(1.118)
$\text {TTM}$	0.0001	0.0004	0.001$^{***}$
	(0.0004)	(0.0002)	(0.0005)
Constant	0.173	−0.011	−0.078
	(0.135)	(0.059)	(0.109)
Observations	389	365	391
R$^{2}$	0.646	0.277	0.637
Adjusted R$^{2}$	0.639	0.260	0.629
Residual standard error	0.396 (df = 380)	0.213 (df = 356)	0.365 (df = 382)

This table shows the results of different regression models, where the dependent variable is the return of Italian government zero-coupon bonds (BOT/CTZ) across various subperiods. The pre-event return is defined by the time windows $(-60,-21)$ and $(-20,-1)$, the event return by $(-3,-1)$ and (0, 3), and the post-event return by (1, 20) and (21, 60), where 0 is the event day and all other numbers represent the trading days relative to the event day. For each set of time windows, the bond-specific average values of the total prices are calculated and the returns across combinations of two time windows are computed. The explanatory variables are represented by EPU index, market variables (VSTOXX index, 5Y CDS spread), economic variables (inflation, GDP growth, CDS spreads, 3 M Euribor), liquidity (price dispersion measure) and bond characteristics (TTM). For the uncertainty measure, market variables, economic variables and liquidity the change of these variables across periods is considered, whereas bond characteristics are represented by their level. The standard errors are given in parentheses and are corrected for heteroscedasticity and clustered at the event level. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013

The significance is indicated as follows: * $< 0.1$, ** $< 0.05$, *** $< 0.01$

Table 18

Pre-event return regressions VSTOXX / VIX

	Dependent variable
	Pre-event return
	(1)	(2)
$\Delta \text {EPU}$	−0.025$^{***}$	−0.029$^{***}$
	(0.006)	(0.005)
$\Delta \text {VSTOXX}$	0.028
	(0.093)
$\Delta \text {VIX}$		0.117$^{*}$
		(0.065)
$\Delta \text {CDS}$	−0.032$^{***}$	−0.036$^{***}$
	(0.006)	(0.005)
$\Delta \text {Inflation}$	−0.495	−0.340
	(0.419)	(0.398)
$\Delta \text {GDP change}$	0.166	0.146
	(0.158)	(0.103)
$\Delta \text {Euribor}$	−5.281$^{***}$	−5.289$^{***}$
	(2.002)	(1.854)
$\Delta \text {Liquidity}$	−5.952$^{***}$	−5.802$^{***}$
	(2.266)	(1.865)
$\text {TTM}$	−0.0001	−0.0001
	(0.0001)	(0.0001)
$\text {Coupon}$	−0.121$^{*}$	−0.125$^{*}$
	(0.068)	(0.069)
Constant	1.146$^{**}$	1.237$^{**}$
	(0.541)	(0.557)
Observations	838	838
R$^{2}$	0.727	0.740
Adjusted R$^{2}$	0.724	0.737
Residual standard error (df = 828)	1.549	1.513

Comparison This table shows the results of different regression models, where the dependent variable is the pre-event return of Italian government coupon bonds (BTP). The pre-event time windows are given by $(-60,-21)$ and $(-20,-1)$, where 0 is the event day and all other numbers represent the trading days relative to the event day. For each time window, the bond-specific average values of the total prices are calculated and the returns across combinations of two time windows are computed. The explanatory variables are represented by EPU index, market variables (VSTOXX Index or VIX Index, 5Y CDS spread), economic variables (inflation, GDP growth, 3 M Euribor), liquidity (price dispersion measure) and bond characteristics (TTM and coupon). For the uncertainty measure, market variables, economic variables and liquidity the change of these variables across periods is considered, whereas bond characteristics are represented by their level. The standard errors are given in parentheses and are corrected for heteroscedasticity and clustered at the event level. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013

The significance is indicated as follows: * $< 0.1$, ** $< 0.05$, *** $< 0.01$

Table 19

Pre-event return regressions based on clean prices

	Dependent variable
	Pre-event return
	(1)	(2)	(3)	(4)
$\Delta \text {EPU}$	−0.033$^{***}$			−0.029$^{***}$
	(0.008)			(0.005)
$\Delta \text {VSTOXX}$	0.150			0.122$^{*}$
	(0.107)			(0.066)
$\Delta \text {CDS}$	−0.044$^{***}$			−0.036$^{***}$
	(0.007)			(0.005)
$\Delta \text {Inflation}$		0.090		−0.313
		(0.906)		(0.392)
$\Delta \text {GDP change}$		0.288		0.168
		(0.274)		(0.103)
$\Delta \text {Euribor}$		−11.578		−5.710$^{***}$
		(7.484)		(1.867)
$\Delta \text {Liquidity}$			−15.169$^{***}$	−5.828$^{***}$
			(3.996)	(2.056)
$\text {TTM}$			−0.0001	−0.0001
			(0.0001)	(0.0001)
$\text {Coupon}$			−0.116$^{*}$	−0.143$^{**}$
			(0.069)	(0.065)
Constant	0.299	−1.248$^{***}$	0.026	0.998$^{*}$
	(0.346)	(0.477)	(0.358)	(0.555)
Observations	838	838	838	838
R$^{2}$	0.629	0.161	0.305	0.738
Adjusted R$^{2}$	0.628	0.158	0.302	0.735
Residual std. error	1.804 (df = 834)	2.713 (df = 834)	2.469 (df = 834)	1.520 (df = 828)

This table shows the results of different regression models, where the dependent variable is the pre-event return of Italian government coupon bonds (BTP) based on clean prices. The pre-event time windows are given by $(-60,-21)$ and $(-20,-1)$, where 0 is the event day and all other numbers represent the trading days relative to the event day. For each time window, the bond-specific average values of the total prices are calculated and the returns across combinations of two time windows are computed. The explanatory variables are represented by EPU index, market variables (VSTOXX Index, 5Y CDS spread), economic variables (inflation, GDP growth, 3 M Euribor), liquidity (price dispersion measure) and bond characteristics (TTM and coupon). For the uncertainty measure, market variables, economic variables and liquidity the change of these variables across periods is considered, whereas bond characteristics are represented by their level. The standard errors are given in parentheses and are corrected for heteroscedasticity and clustered at the event level. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013

The significance is indicated as follows: * $< 0.1$, ** $< 0.05$, *** $< 0.01$

Table 20

Subperiod regressions based on clean prices

	Dependent variable
	Pre-Event Return	Event Return	Post-Event Return
	(1)	(2)	(3)
$\Delta \text {EPU}$	−0.029$^{***}$	−0.010$^{**}$	−0.006
	(0.005)	(0.004)	(0.010)
$\Delta \text {VSTOXX}$	0.122$^{*}$	−0.043	0.166$^{**}$
	(0.066)	(0.033)	(0.071)
$\Delta \text {CDS}$	−0.036$^{***}$	−0.022$^{***}$	−0.041$^{***}$
	(0.005)	(0.002)	(0.013)
$\Delta \text {Inflation}$	−0.313	0.445	−1.061
	(0.392)	(0.365)	(1.257)
$\Delta \text {GDP change}$	0.168	−0.247$^{***}$	−1.033$^{**}$
	(0.103)	(0.092)	(0.473)
$\Delta \text {Euribor}$	−5.710$^{***}$	10.465$^{*}$	4.785
	(1.867)	(5.995)	(3.254)
$\Delta \text {Liquidity}$	−5.828$^{***}$	0.085	−5.088$^{***}$
	(2.056)	(0.401)	(1.515)
$\text {TTM}$	−0.0001	0.00005	0.0001
	(0.0001)	(0.0001)	(0.0001)
$\text {Coupon}$	−0.143$^{**}$	0.052	−0.033
	(0.065)	(0.038)	(0.074)
Constant	0.998$^{*}$	0.039	0.653
	(0.555)	(0.287)	(0.577)
Observations	838	733	827
R$^{2}$	0.738	0.556	0.510
Adjusted R$^{2}$	0.735	0.551	0.505
Residual standard error	1.520 (df = 828)	0.990 (df = 723)	1.955 (df = 817)

This table shows the results of different regression models, where the dependent variable is the return of Italian government coupon bonds (BTP) across various subperiods based on clean prices. The pre-event return is defined by the time windows $(-60,-21)$ and $(-20,-1)$, the event return by $(-3,-1)$ and (0, 3), and the post-event return by (1, 20) and (21, 60), where 0 is the event day and all other numbers represent the trading days relative to the event day. For each set of time windows, the bond-specific average values of the total prices are calculated and the returns across combinations of two time windows are computed. The explanatory variables are represented by EPU index, market variables (VSTOXX index, 5Y CDS spread), economic variables (inflation, GDP growth, CDS spreads, 3 M Euribor), liquidity (price dispersion measure) and bond characteristics (TTM and coupon). For the uncertainty measure, market variables, economic variables and liquidity the change of these variables across periods is considered, whereas bond characteristics are represented by their level. The standard errors are given in parentheses and are corrected for heteroscedasticity and clustered at the event level. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013

The significance is indicated as follows: * $< 0.1$, ** $< 0.05$, *** $< 0.01$

Table 21

Pre-event return regressions based on abnormal returns

	Dependent variable
	Pre-event return
	(1)	(2)	(3)	(4)
$\Delta \text {EPU}$	−0.024$^{***}$			−0.021$^{***}$
	(0.006)			(0.005)
$\Delta \text {VSTOXX}$	0.114			0.118$^{**}$
	(0.071)			(0.051)
$\Delta \text {CDS}$	−0.053$^{***}$			−0.049$^{***}$
	(0.005)			(0.004)
$\Delta \text {Inflation}$		0.130		−0.360
		(1.126)		(0.402)
$\Delta \text {GDP change}$		0.020		−0.139
		(0.330)		(0.098)
$\Delta \text {Euribor}$		−7.949		−1.009
		(9.295)		(1.691)
$\Delta \text {Liquidity}$			−16.145$^{***}$	−4.848$^{***}$
			(4.121)	(1.314)
$\text {TTM}$			−0.0002$^{**}$	−0.0002$^{**}$
			(0.0001)	(0.0001)
$\text {Coupon}$			−0.120	−0.142$^{*}$
			(0.089)	(0.073)
Constant	0.196	−1.472$^{**}$	0.163	1.466$^{**}$
	(0.252)	(0.627)	(0.381)	(0.604)
Observations	838	838	838	838
R$^{2}$	0.660	0.067	0.295	0.745
Adjusted R$^{2}$	0.659	0.064	0.292	0.742
Residual std. error	1.989 (df = 834)	3.294 (df = 834)	2.863 (df = 834)	1.728 (df = 828)

This table shows the results of different regression models, where the dependent variable is the abnormal pre-event return of Italian government coupon bonds (BTP) w.r.t. the German market. The pre-event time windows are given by $(-60,-21)$ and $(-20,-1)$, where 0 is the event day and all other numbers represent the trading days relative to the event day. For each time window, the bond-specific average values of the total prices are calculated and the returns across combinations of two time windows are computed. The explanatory variables are represented by EPU index, market variables (VSTOXX Index, 5Y CDS spread), economic variables (inflation, GDP growth, 3 M Euribor), liquidity (price dispersion measure) and bond characteristics (TTM and coupon). For the uncertainty measure, market variables, economic variables and liquidity the change of these variables across periods is considered, whereas bond characteristics are represented by their level. The standard errors are given in parentheses and are corrected for heteroscedasticity and clustered at the event level. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013

The significance is indicated as follows: * $< 0.1$, ** $< 0.05$, *** $< 0.01$

Table 22

Subperiod regressions based on abnormal returns

	Dependent variable
	Pre-event return	Event return	Post-event return
	(1)	(2)	(3)
$\Delta \text {EPU}$	−0.021$^{***}$	−0.014$^{***}$	−0.012$^{**}$
	(0.005)	(0.004)	(0.006)
$\Delta \text {VSTOXX}$	0.118$^{**}$	−0.092$^{***}$	0.062
	(0.051)	(0.024)	(0.046)
$\Delta \text {CDS}$	−0.049$^{***}$	−0.025$^{***}$	−0.039$^{***}$
	(0.004)	(0.002)	(0.009)
$\Delta \text {Inflation}$	−0.360	−0.085	−1.442
	(0.402)	(0.334)	(1.016)
$\Delta \text {GDP change}$	−0.139	−0.058	−0.769$^{**}$
	(0.098)	(0.063)	(0.318)
$\Delta \text {Euribor}$	−1.009	8.891	2.838
	(1.691)	(5.418)	(2.410)
$\Delta \text {Liquidity}$	−4.848$^{***}$	0.097	−6.112$^{***}$
	(1.314)	(0.416)	(1.444)
$\text {TTM}$	−0.0002$^{**}$	0.00003	−0.0001
	(0.0001)	(0.0001)	(0.0001)
$\text {Coupon}$	−0.142$^{*}$	0.056	−0.052
	(0.073)	(0.044)	(0.073)
Constant	1.466$^{**}$	0.034	0.558
	(0.604)	(0.332)	(0.633)
Observations	838	733	827
R$^{2}$	0.745	0.607	0.629
Adjusted R$^{2}$	0.742	0.602	0.625
Residual standard error	1.728 (df = 828)	1.008 (df = 723)	1.972 (df = 817)

This table shows the results of different regression models, where the dependent variable is the abnormal return of Italian government coupon bonds (BTP) w.r.t. the German market across various subperiods. The pre-event return is defined by the time windows $(-60,-21)$ and $(-20,-1)$, the event return by $(-3,-1)$ and (0, 3), and the post-event return by (1, 20) and (21, 60), where 0 is the event day and all other numbers represent the trading days relative to the event day. For each set of time windows, the bond-specific average values of the total prices are calculated and the returns across combinations of two time windows are computed. The explanatory variables are represented by EPU index, market variables (VSTOXX index, 5Y CDS spread), economic variables (inflation, GDP growth, CDS spreads, 3 M Euribor), liquidity (price dispersion measure) and bond characteristics (TTM and coupon). For the uncertainty measure, market variables, economic variables and liquidity the change of these variables across periods is considered, whereas bond characteristics are represented by their level. The standard errors are given in parentheses and are corrected for heteroscedasticity and clustered at the event level. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013

The significance is indicated as follows: * $< 0.1$, ** $< 0.05$, *** $< 0.01$

Table 23

Event prediction regressions

	Dependent variable
	Event indicator
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Yield spread (−60, −41)	0.040
	(0.095)
Yield spread (−40, −21)		0.064
		(0.099)
Yield spread (−20, −1)			0.132
			(0.112)
Spread change (−60, −41);(-40, −21)				0.209
				(0.286)
Spread change (-60, −41);(−20, −1)					0.298
					(0.231)
Spread change (−40, −21);(−20, −1)						0.565
						(0.377)
Spread change (−60, −21);(−20, −1)							0.428
							(0.312)
Constant	−1.471$^{***}$	−1.538$^{***}$	−1.732$^{***}$	−1.376$^{***}$	−1.413$^{***}$	−1.421$^{***}$	−1.421$^{***}$
	(0.281)	(0.292)	(0.341)	(0.135)	(0.143)	(0.144)	(0.145)
Observations	197	197	197	197	197	197	197
Log likelihood	−57.798	−57.661	−56.923	−57.636	−56.428	−55.788	−56.051
Akaike Inf. Crit	119.596	119.322	117.847	119.271	116.856	115.576	116.101

This table shows the results of different probit models, where the dependent variable is an indicator variable which is equal to one in weeks in which an event occurred and zero otherwise. The independent variables represent the spread between the 5 year Italian and German zero-coupon yield in certain time windows before the event and changes in the spread across these windows. The standard errors are given in parentheses and are corrected for heteroscedasticity and autocorrelation. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013

The significance is indicated as follows: * $< 0.1$, ** $< 0.05$, *** $< 0.01$

Table 24

Pre-event return regressions for bonds—various EPU indices

	Dependent variable
	Pre-Event Return
	(1)	(2)	(3)	(4)	(5)	(6)
$\Delta \text {EPU Italy}$	−0.029$^{***}$	−0.029$^{***}$	−0.029$^{***}$	−0.028$^{***}$	−0.029$^{***}$	−0.024$^{*}$
	(0.005)	(0.004)	(0.006)	(0.005)	(0.006)	(0.013)
$\Delta \text {EPU Europe}$		0.0002				−0.009
		(0.013)				(0.021)
$\Delta \text {EPU Germany}$			0.001			0.0003
			(0.005)			(0.004)
$\Delta \text {EPU Spain}$				0.003		0.009
				(0.008)		(0.014)
$\Delta \text {EPU Greece}$					−0.0003	0.002
					(0.003)	(0.006)
$\Delta \text {VSTOXX}$	0.117$^{*}$	0.117$^{*}$	0.113	0.105	0.118$^{*}$	0.083
	(0.065)	(0.068)	(0.084)	(0.081)	(0.064)	(0.092)
$\Delta \text {CDS}$	−0.036$^{***}$	−0.037$^{***}$	−0.037$^{***}$	−0.037$^{***}$	−0.036$^{***}$	−0.035$^{***}$
	(0.005)	(0.006)	(0.004)	(0.005)	(0.005)	(0.006)
$\Delta \text {Inflation}$	−0.340	−0.348	−0.374	−0.426	−0.345	−0.252
	(0.398)	(0.651)	(0.528)	(0.517)	(0.427)	(0.746)
$\Delta \text {GDP change}$	0.146	0.144	0.140	0.100	0.145	0.109
	(0.103)	(0.159)	(0.111)	(0.170)	(0.105)	(0.170)
$\Delta \text {EURIBOR}$	−5.289$^{***}$	−5.282$^{***}$	−5.288$^{***}$	−4.281	−5.312$^{***}$	−2.807
	(1.854)	(1.889)	(1.822)	(3.101)	(1.840)	(4.820)
$\Delta \text {Liquidity}$	−5.802$^{***}$	−5.816$^{***}$	−5.868$^{***}$	−6.073$^{***}$	−5.766$^{***}$	−6.211$^{***}$
	(1.865)	(1.942)	(1.873)	(1.931)	(1.738)	(1.616)
TTM	−0.0001	−0.0001	−0.0001	−0.0001	−0.0001	−0.0001
	(0.0001)	(0.0001)	(0.0001)	(0.0001)	(0.0001)	(0.0001)
Coupon	−0.125$^{*}$	−0.125$^{*}$	−0.125$^{*}$	−0.125$^{*}$	−0.125$^{*}$	−0.125$^{*}$
	(0.069)	(0.069)	(0.068)	(0.069)	(0.069)	(0.068)
Constant	1.237$^{**}$	1.239$^{**}$	1.248$^{**}$	1.289$^{**}$	1.236$^{**}$	1.331$^{**}$
	(0.557)	(0.546)	(0.542)	(0.549)	(0.557)	(0.563)
Observations	838	838	838	838	838	838
R$^{2}$	0.740	0.740	0.740	0.741	0.740	0.742
Adjusted R$^{2}$	0.737	0.737	0.737	0.738	0.737	0.738
Residual standard error	1.513 (df = 828)	1.513 (df = 827)	1.513 (df = 827)	1.512 (df = 827)	1.513 (df = 827)	1.512 (df = 824)

This table shows the results of different regression models, where the dependent variable is the pre-event return of Italian government coupon bonds (BTP). The pre-event time windows are given by $(-60,-21)$ and $(-20,-1)$, where 0 is the event day and all other numbers represent the trading days relative to the event day. For each time window, the bond-specific average values of the total prices are calculated and the returns across combinations of two time windows are computed. The explanatory variables are represented by various EPU indices, market variables (VSTOXX Index, 5Y CDS spread), economic variables (inflation, GDP growth, 3 M Euribor), liquidity (price dispersion measure) and bond characteristics (TTM and coupon). For the uncertainty measures, market variables, economic variables and liquidity the change of these variables across periods is considered, whereas bond characteristics are represented by their level. The standard errors are given in parentheses and are corrected for heteroscedasticity and clustered at the event level. The political events include Euro, G8 and G20 summits and relevant elections during the European sovereign debt crisis represented by the time period 2010 to 2013

The significance is indicated as follows: * $< 0.1$, ** $< 0.05$, *** $< 0.01$

Note that we consider political uncertainty as the main driver of policy uncertainty before the events in line with the theoretical guidance and, thus, we do not distinguish between these two in the discussion of the results.

In a related theoretical contribution, Sialm (2006) studies the effect of stochastic taxation on asset prices in a dynamic general equilibrium model, showing that tax changes result in price adjustments and that the effect is stronger for assets with a longer duration. Concerning bond markets, Ulrich (2013) models the uncertainty about future government spending and shows that this uncertainty is a first-order risk factor in the bond market.

In other empirical work, Erb et al. (1996) show that the International Country Risk index, which includes political uncertainty, is correlated with future equity returns. In addition, Pantzalis et al. (2000) and Li and Born (2006) find abnormal stock returns prior to elections and Boutchkova et al. (2012) link global and political risk to industry stock return volatility.

Other indices include the index by Azzimonti (2014) which measures the degree of partisan conflict or the state-level measure by Shoag and Veuger (2016), for example. However, both are only US specific.

Other proxies in this category are the Bayesian Gibbs measure proposed by Hasbrouck (2009) and the high-low estimator by Corwin and Schultz (2012), who argue that the high-low ratio reflects the variance and the bid-ask spread.

Friewald et al. (2017) provide a comprehensive comparison of these various liquidity measures.

For a detailed discussion on what factors led to the European sovereign debt crisis, see Lane (2012).

We exclude EU summits as most were held in combination with a Euro summit or were held very close to these summits.

We consider different specifications to avoid potential effects from overlapping event windows in our robustness tests, see Sect. 5.3.

Primary dealers estimate that around $75\%$ to $80\%$ of all transactions are covered by MTS in the case of Italian government bonds, whereas this percentage is much lower for other countries.

We present results for Italian short-term bonds in our robustness tests, see Sect. 5.3.

Dick-Nielsen (2009) provides filters for the TRACE dataset representing transactions in the US corporate bond market.

Note that the MTS transaction dataset in the case of Italian government bonds is much cleaner compared to the US corporate bond market as a significantly lower number of instruments and more liquid bonds are involved. Thus, we had to eliminate less than ten observations based on these transaction filters.

Dealers in this market quote very high spreads at certain trading times, e.g., at the end of day, when they are not interested in attracting trades. Furthermore, stale quotes can lead to negative spreads. Thus, we adjust these unrepresentative quotes.

For particular bonds on days with no trading activity only very high spreads are observable which are not representing real potential trading costs, but rather indicate that dealers do not want to attract any trades.

Note that we assume that the EPU index also covers the political uncertainty concerning redenomination risk. Klose and Weigert (2014) discuss this risk. However, their measure based on tradeable bets is not available for our whole sample period. Other authors, e.g., De Santis (2019) and Kremens (2022), discuss measures based on CDS prices. However, given the increased counterparty and liquidity risk during our sample period, we are not employing these measures.

In an additional analysis, we use the VIX index instead of the VSTOXX index. However, the two time series are highly correlated and lead to basically the same results, see Appendix Table 18.

Our results are robust to alternative specifications of these time windows, see Sect. 5.3.

We present results based on multi-variate regressions as robustness tests, see Sect. 5.3.

In our main analysis we focus on total returns controlling for the maturity, coupon and risk-free rate, among other variables. However, we obtain very similar results if we either consider returns based on clean prices or abnormal returns, see Tables 19, 20, 21 and 22 in Appendix.

Using one of the other available liquidity variables provides the same results. However, we include only one metric as the different measures are highly correlated.

In an additional analysis, we also included bond fixed effects which lead to very similar results.

All the results are significant at the 1% level based on t-values of paired t-tests and pseudomedians of paired Wilcoxon tests.

Note that we test for reverse causality, i.e., whether falling bond prices potentially triggered events, in the robustness section, see Sect. 5.3. However, we cannot completely rule out that falling bond prices affected policy decisions, in general.

These results are significant at the $1\%$ level.

This result is significant at the 1% level.

In an additional analysis (not reported in detail) we consider various sample splits to explore whether the price effects are consistent over time or are reduced due to learning by investors about the decision-making process in this crisis. We do not find any time-related effects. Furthermore, we separately analyze on- and off-the-run bonds and find no significant differences in the results, indicating that slow-moving capital is not driving our results.

Note that a qualitative analysis of the events with negative returns indicates that these events are generally associated with outcomes of gridlock fostering the risk premium interpretation. For example, the Italian election on February 25, 2013, did not produce a clear parliamentary majority and lead Italy politically into a deadlock.

In an additional analysis we include alternative EPU indices to this regression. In particular, we explored the potential impact of the Greek, Spanish, German and European EPU index. None of these indices provide significant results in addition to the Italian EPU index, indicating that a potential spillover of uncertainty is directly contained in the Italian EPU index. We report the details of this analysis in Table 24 of Appendix.

Note that in an additional analysis (not reported in detail), we use the pre-event level of the CDS spread instead of the contemporaneous CDS spread change. We find basically identical results compared to the original specification.

The regression result for the pre-event period based on all variables is included as well, to allow a direct comparison.

Note that in an additional analysis (not reported in detail), we add the pre-event and event return to the post-return regression, exploring whether price effects in these stages have a significant impact on the post-event period. However, we find insignificant results for both variables.

Note that the results of the event period are driven by the sharp price increase at the event day and, thus, different specifications are not impacting these results. In the post-event period we do not find any relation between prices and political uncertainty in any of the specifications. Thus, these results are not reported in detail.

This price return is again significant at the $1\%$ level.

We use a weekly indicator since our event return calculation is also effectively focused on a horizon of one week. In our main specification we use all observed weeks. However, in alternative specifications, we are eliminating a certain number of weeks before and after the event weeks to avoid overlaps in the independent variables and obtain the same results.

Additionally, we repeated the analysis using different maturities (2 and 10 years) and using the Italian zero-coupon yields directly as an explanatory variable. We find the same results in both cases.

Note that in an additional robustness test (not reported in detail), we explore whether Greek yield spreads can predict our events, especially in the time window $(-20, -1)$. Also in this analysis we find insignificant results.

In an additional analysis (not reported in detail), we analyze the secondary market prices of the bonds with seasoned offerings in the affected period $(-20,-1)$ instead of their respective auction prices. We find similar magnitudes, i.e., the bonds are traded at much lower prices in the affected period compared to the observed prices before the events.

Amihud, Y.: Illiquidity and stock returns: cross-section and time-series effects. J. Financ. Mark. 5(1), 31–56 (2002)CrossRef

Azzimonti, M.: Partisan conflict. FRB of Philadelphia Working Paper, (2014)

Baker, S., Bloom, N., Davis, S.: Measuring economic policy uncertainty. Q. J. Econ. 131(4), 1593–1636 (2016)CrossRef

Beber, A., Brandt, M., Kavajecz, K.: Flight-to-quality or flight-to-liquidity? Evidence from the euro-area bond market. Rev. Financ. Stud. 22(3), 925–957 (2009)CrossRef

Bessembinder, H., Kahle, K.M., Maxwell, W.F., Danielle, X.: Measuring abnormal bond performance. Rev. Financ. Stud. 22(10), 4219–4258 (2009)CrossRef

Boutchkova, M., Doshi, H., Durnev, A., Molchanov, A.: Precarious politics and return volatility. Rev. Financ. Stud. 25(4), 2417–2480 (2012)CrossRef

Brogaard, J., Detzel, A.: The asset-pricing implications of government economic policy uncertainty. Manag. Sci. 61(1), 3–18 (2015)CrossRef

Brogaard, J., Dai, L., Ngo, P.T.H., Zhang, B.: Global political uncertainty and asset prices. Rev. Financ. Stud. 33, 1737–1780 (2020)CrossRef

Corwin, S., Schultz, P.: A simple way to estimate bid-ask spreads from daily high and low prices. J. Financ. 67(2), 719–760 (2012)CrossRef

De Santis, R.A.: Redenomination risk. J. Money Credit Bank. 51(8), 2173–2206 (2019)CrossRef

Dick-Nielsen, J.: Liquidity biases in TRACE. J. Fixed Income 19(2), 43–55 (2009)CrossRef

Dick-Nielsen, J., Feldhütter, P., Lando, D.: Corporate bond liquidity before and after the onset of the subprime crisis. J. Financ. Econ. 103(3), 471–492 (2012)CrossRef

Dufour, A., Skinner, F.: MTS time series: market and data description for the European bond and repo database. Technical Report Version 5, ICMA Centre, University of Reading, (2010)

Eisl, A., Ochs, C., Pichler, S.: Optimal cash buffers of sovereign debt management offices. Working Paper, (2019)

Erb, Claude B., Harvey, C.R., Viskanta, T.E.: Political risk, economic risk and financial risk. Fuqua School of Business Working Paper, (1996)

Eser, F., Schwaab, B.: Evaluating the impact of unconventional monetary policy measures: empirical evidence from the ECB’s securities markets programme. J. Financ. Econ. 119(1), 147–167 (2016)CrossRef

Feldhütter, P.: The same bond at different prices: identifying search frictions and selling pressures. Rev. Financ. Stud. 25(4), 1155–1206 (2012)CrossRef

Friewald, N., Jankowitsch, R., Subrahmanyam, M.G.: Transparency and liquidity in the structured product market. Rev. Asset Pricing Stud. 7(2), 316–348 (2017)CrossRef

Gao, P., Qi, Y.: Political uncertainty and public financing costs: evidence from U.S. gubernatorial elections and municipal bond markets. Working Paper, (2013)

Goyenko, R., Holden, C., Trzcinka, C.: Do liquidity measures measure liquidity? J. Financ. Econ. 92(2), 153–181 (2009)CrossRef

Hardouvelis, G.A., Karalas, G.I., Karanastasis, D.I., Samartzis, P.K.: Economic policy uncertainty, political uncertainty and the greek economic crisis. Working Paper, (2018)

Hasbrouck, J.: Trading costs and returns for U.S. equities: estimating effective costs from daily data. J. Financ. 64(3), 1445–1477 (2009)CrossRef

Hunag, T., Fei, W., Ying, Yu., Zhang, B.: International political risk and government bond pricing. J. Bank. Financ. 55, 393–405 (2015)CrossRef

Jankowitsch, R., Nashikkar, A., Subrahmanyam, M.G.: Price dispersion in OTC markets: a new measure of liquidity. J. Bank. Financ. 35(2), 343–357 (2011)CrossRef

Kelly, B., Pastor, L., Veronesi, P.: The price of political uncertainty: theory and evidence from the option market. J. Financ. 71(5), 2417–2480 (2016)CrossRef

Klose, J., Weigert, B.: Sovereign yield spreads during the euro crisis: fundamental factors versus redenomination risk. Int. Financ. 17(1), 25–50 (2014)CrossRef

Kremens, L.: Currency redenomination risk. Working Paper, (2022)

Lane, P.R.: The European sovereign debt crisis. J. Econ. Persp. Paper 26(3), 49–68 (2012)CrossRef

Li, J., Born, J.A.: Presidential election uncertainty and common stock returns in the United States. J. Financ. Res. 29(4), 609–622 (2006)CrossRef

Manzo, G.: Political uncertainty, credit risk premium and default risk. Working Paper, (2013)

Pantzalis, C., Stangeland, D.A., Turtle, H.: Political elections and the resolution of uncertainty: the international evidence. J. Bank. Financ. 24(10), 1575–1604 (2000)CrossRef

Pastor, L., Veronesi, P.: Uncertainty about government policy and stock prices. J. Financ. 67(4), 1219–1264 (2012)CrossRef

Pastor, L., Veronesi, P.: Political uncertainty and risk premia. J. Financ. Econ. 110(3), 520–545 (2013)CrossRef

Paule-Vianez, J., Escamilla-Solano, S., Martinez-Navalon, J.G.: The effect of economic policy uncertainty on the Spanish bond yield. (2021)

Pelizzon, L., Subrahmanyam, M.G., Tomio, D., Uno, J.: Sovereign credit risk, liquidity, and European Central Bank intervention: Deus ex machina? J. Financ. Econ. 122(1), 86–115 (2016)CrossRef

Petersen, M.A.: Estimating standard errors in finance panel data sets: comparing approaches. Rev. Financ. Stud. 22(1), 435–480 (2009)CrossRef

Pierluigi, B., Emanuele, B., Marco, B., Fabio S.: Populism, political risk and the economy: lessons from Italy. Working Paper (2020)

Roll, R.: A simple implicit measure of the effective bid-ask spread in an efficient market. J. Financ. 39(4), 1127–1139 (1984)

Schestag, R., Schuster, P., Uhrig-Homburg, M.: Measuring liquidity in bond markets. Rev. Financ. Stud. 29(5), 1170–1219 (2016)CrossRef

Schlepper, K.: Heiko, H., Ryan, R., Andreas S.: Scarcity effects of QE: a transaction-level analysis in the bund market. BIS Working Papers, (2017)

Shoag, D., Veuger, S.: Uncertainty and the geography of the great recession. J. Monet. Econ. 84, 84–93 (2016)CrossRef

Sialm, C.: Stochastic taxation and asset pricing in dynamic general equilibrium. J. Econ. Dyn. Contr. 30(3), 511–540 (2006)CrossRef

Ulrich, M.: How does the bond market perceive government interventions? Columbia Business School Research Paper, (2013)

Title: Political uncertainty and sovereign bond markets
Authors: Lukas Handler
Rainer Jankowitsch
Publication date: 25-11-2024
Publisher: Springer Berlin Heidelberg
Published in: Financial Markets and Portfolio Management / Issue 1/2025
Print ISSN: 1934-4554
Electronic ISSN: 2373-8529
DOI: https://doi.org/10.1007/s11408-024-00461-6

Report of the co-editors 2024

Ben Mezrich, Dumb Money (Harper Collins Publishers, 2023)

Book Review

	Price returns	Total price returns	Abnormal returns w.r.t. Germany	Returns of other affected markets	Abn. returns of other affected markets
Panel A: returns
\((-60,-21);(-20,-1)\)	\(-0.936\)	\(-0.528\)	\(-1.198\)	\(-0.542\)	\(-1.211\)
\((-3,-1);(0,3)\)	0.362	0.410	0.311	0.404	0.305
(1, 20); (21, 60)	0.849	1.198	0.527	1.728	1.057
Panel B: T-test: t-values
\((-60,-21);(-20,-1)\)	\(-9.316^{***}\)	\(-5.184^{***}\)	\(-10.186^{***}\)	\(-3.243^{***}\)	\(-6.640^{***}\)
\((-3,-1);(0,3)\)	\(6.794^{***}\)	\(7.637^{***}\)	\(5.266^{***}\)	\(4.099^{***}\)	\(2.852^{***}\)
(1, 20); (21, 60)	\(8.993^{***}\)	\(12.485^{***}\)	\(4.704^{***}\)	\(8.946^{***}\)	\(4.986^{***}\)
Panel C: Wilcoxon test: p-values
\((-60,-21);(-20,-1)\)	0.000\(^{***}\)	0.000\(^{***}\)	0.000\(^{***}\)	0.008\(^{***}\)	0.000\(^{***}\)
\((-3,-1);(0,3)\)	0.000\(^{***}\)	0.000\(^{***}\)	0.000\(^{***}\)	0.242	0.028\(^{**}\)
(1, 20); (21, 60)	0.000\(^{***}\)	0.000\(^{***}\)	0.000\(^{***}\)	0.000\(^{***}\)	0.000\(^{***}\)

	Bid-ask	Roll	Amihud	Price disp	Volume	Net-bid
Panel A: differences
\((-60,-21);(-20,-1)\)	\(-0.036\)	1.250	0.866	0.817	\(-1.145\)	0.492
\((-3,-1);(0,3)\)	4.074	2.727	1.491	3.845	\(-1.869\)	\(-3.372\)
(1, 20); (21, 60)	\(-8.462\)	\(-1.190\)	\(-1.197\)	\(-2.009\)	\(-0.687\)	\(-1.543\)
Panel B: T-test: t-values
\((-60,-21);(-20,-1)\)	\(-0.039\)	\(2.087^{**}\)	\(2.497^{**}\)	\(2.322^{**}\)	\(-1.548\)	0.939
\((-3,-1);(0,3)\)	\(6.636^{***}\)	\(2.680^{***}\)	\(1.775^{*}\)	\(6.318^{***}\)	−1.369	\(-2.804^{***}\)
(1, 20); (21, 60)	\(-11.575^{***}\)	\(-2.118^{**}\)	\(-2.946^{***}\)	\(-5.878^{***}\)	\(-1.089\)	\(-3.249^{***}\)
Panel C: Wilcoxon test: p-values
\((-60,-21);(-20,-1)\)	0.772	0.160	\(0.139^{**}\)	\(0.045^{**}\)	\(0.086^{**}\)	0.619
\((-3,-1);(0,3)\)	\(0.000^{***}\)	\(0.011^{**}\)	\(0.001^{***}\)	\(0.000^{***}\)	0.672	\(0.002^{***}\)
(1, 20); (21, 60)	\(0.000^{***}\)	\(0.000^{***}\)	\(0.001^{***}\)	\(0.000^{***}\)	0.535	\(0.000^{***}\)

	Price returns	Total price returns	Abnormal returns w.r.t. Germany	Returns of other affected markets	Abn. returns of other affected markets
Panel A: returns
\((-60,-21);(-20,-1)\)	\(-0.813\)	\(-0.395\)	\(-0.943\)	\(-1.104\)	\(-1.651\)
\((-3,-1);(0,3)\)	0.556	0.589	0.540	0.794	0.745
(1, 20); (21, 60)	0.814	1.079	0.251	0.917	0.089
Panel B: T-test: t-values
\((-60,-21);(-20,-1)\)	\(-5.163^{***}\)	\(-2.500^{**}\)	\(-5.328^{***}\)	\(-5.741^{***}\)	\(-7.636^{***}\)
\((-3,-1);(0,3)\)	\(6.985^{***}\)	\(7.361^{***}\)	\(6.291^{***}\)	\(5.807^{***}\)	\(4.983^{***}\)
(1, 20); (21, 60)	\(5.455^{***}\)	\(7.071^{***}\)	1.304	\(4.524^{***}\)	0.348
Panel C: Wilcoxon test: p-values
\((-60,-21);(-20,-1)\)	0.000\(^{***}\)	0.000\(^{***}\)	0.000\(^{***}\)	0.006\(^{***}\)	0.003\(^{***}\)
\((-3,-1);(0,3)\)	0.000\(^{***}\)	0.002\(^{***}\)	0.000\(^{***}\)	0.602	0.732
(1,20);(21,60)	0.000\(^{***}\)	0.000\(^{***}\)	0.356	0.000\(^{***}\)	0.153

	Price returns	Total price returns	Abnormal returns w.r.t. Germany
Panel A: returns
\((-60,-21);(-20,-1)\)	\(-0.045\)	\(-0.045\)	\(-0.075\)
\((-3,-1);(0,3)\)	0.046	0.046	0.035
(1, 20); (21, 60)	0.176	0.176	0.144
Panel B: T-test: t-values
\((-60,-21);(-20,-1)\)	\(-1.418\)	\(-1.418\)	\(-2.221^{**}\)
\((-3,-1);(0,3)\)	\(3.598^{***}\)	\(3.598^{***}\)	\(2.659^{***}\)
(1,20);(21,60)	\(7.155^{***}\)	\(7.155^{***}\)	\(5.675^{***}\)
Panel C: Wilcoxon test: p-values
\((-60,-21);(-20,-1)\)	0.000\(^{***}\)	0.000\(^{***}\)	0.000\(^{**}\)
\((-3,-1);(0,3)\)	0.000\(^{***}\)	0.000\(^{***}\)	0.003\(^{***}\)
(1,20);(21,60)	0.000\(^{***}\)	0.000\(^{***}\)	0.000\(^{***}\)

	Bid-ask	Roll	Amihud	Price disp	Volume	Net-bid
Panel A: differences
\((-60,-21);(-20,-1)\)	\(-0.933\)	\(-0.106\)	0.065	\(-0.174\)	\(-17.783\)	3.001
\((-3,-1);(0,3)\)	2.339	0.457	0.101	0.460	\(-0.487\)	\(-8.548\)
(1,20);(21,60)	\(-5.940\)	\(-0.804\)	\(-0.476\)	\(-1.391\)	5.986	\(-6.691\)
Panel B: T-test: t-values
\((-60,-21);(-20,-1)\)	\(-1.127\)	\(-0.594\)	0.827	\(-1.007\)	\(-7.596\)	\(2.308^{**}\)
\((-3,-1);(0,3)\)	\(5.463^{***}\)	0.785	2.375	\(2.375^{**}\)	\(-0.172\)	\(-3.522^{***}\)
(1,20);(21,60)	\(-8.944^{***}\)	\(-4.335^{***}\)	\(-6.526^{***}\)	\(-8.678^{***}\)	\(2.670^{***}\)	\(-3.172^{***}\)
Panel C: Wilcoxon test: p-values
\((-60,-21);(-20,-1)\)	\(0.002^{***}\)	0.110	\(0.505^{*}\)	\(0.009^{***}\)	\(0.000^{***}\)	\(0.096^{*}\)
\((-3,-1);(0,3)\)	\(0.000^{***}\)	0.902	\(0.033^{**}\)	\(0.022^{**}\)	0.730	\(0.001^{**}\)
(1,20);(21,60)	\(0.000^{***}\)	\(0.000^{***}\)	\(0.000^{***}\)	\(0.003^{***}\)	\(0.000^{***}\)	\(0.001^{***}\)

Springer Professional

Political uncertainty and sovereign bond markets

Abstract

Publisher's Note

1 Introduction

2 Literature

3 Data

3.1 Political events

3.2 Bond data

3.3 Uncertainty index and economic variables

4 Methodology

4.1 Price and trading activity variables

4.2 Price and trading activity effects

4.3 Regression analysis

5 Results

5.1 Impact of political uncertainty

5.2 Regression analysis

5.3 Robustness tests

5.4 Implications for issuing costs

6 Conclusion

Acknowledgements

Publisher's Note

Appendix

Other articles of this Issue 1/2025

Report of the co-editors 2024

Ben Mezrich, Dumb Money (Harper Collins Publishers, 2023)

Financial knowledge acquisition and trading behavior: empirical evidence from an online information tool

Second hand or second generation? The performance of secondary buyouts

Extending the demand system approach to asset pricing

Premium Partner

	Bid-ask	Roll	Amihud	Price dispersion	Volume	Net-bid
\((-60,-21)\)	49.873	12.462	8.251	15.379	38.541	1.588
\((-20,-1)\)	49.716	12.488	9.032	16.181	38.284	1.988
\((-3,-1)\)	54.024	12.226	9.991	16.353	38.244	2.891
(0,3)	57.769	14.589	12.064	20.526	36.038	\(-0.516\)
(1,20)	58.696	14.540	11.315	18.998	33.751	1.928
(21,60)	50.414	13.650	10.170	17.068	32.870	0.413

	Dependent variable
	Pre-event return
	(1)	(2)	(3)	(4)
\(\Delta \text {EPU}\)	−0.033\(^{***}\)			−0.029\(^{***}\)
	(0.007)			(0.005)
\(\Delta \text {VSTOXX}\)	0.143			0.117\(^{*}\)
	(0.106)			(0.065)
\(\Delta \text {CDS}\)	−0.044\(^{***}\)			−0.036\(^{***}\)
	(0.007)			(0.005)
\(\Delta \text {Inflation}\)		0.078		−0.340
		(0.918)		(0.398)
\(\Delta \text {GDP change}\)		0.269		0.146
		(0.280)		(0.103)
\(\Delta \text {Euribor}\)		−11.235		−5.289\(^{***}\)
		(7.554)		(1.854)
\(\Delta \text {Liquidity}\)			−15.248\(^{***}\)	−5.802\(^{***}\)
			(3.971)	(1.865)
\(\text {TTM}\)			−0.0001	−0.0001
			(0.0001)	(0.0001)
\(\text {Coupon}\)			−0.099	−0.125\(^{*}\)
			(0.077)	(0.069)
Constant	0.704\(^{**}\)	−0.842\(^{*}\)	0.243	1.237\(^{**}\)
	(0.334)	(0.485)	(0.348)	(0.557)
Observations	838	838	838	838
R\(^{2}\)	0.628	0.181	0.298	0.738
Adjusted R\(^{2}\)	0.626	0.178	0.296	0.735
Residual standard error	1.763 (df = 834)	2.615 (df = 834)	2.420 (df = 834)	1.485 (df = 828)

	Dependent variable
	Pre-event return
	\((-60,-41)\)	\((-40,-21)\)	\((-60,-21)\)
\(\Delta \text {EPU}\)	−0.023\(^{**}\)	−0.025\(^{***}\)	−0.029\(^{***}\)
	(0.010)	(0.004)	(0.005)
\(\Delta \text {VSTOXX}\)	0.053	0.103	0.117\(^{*}\)
	(0.105)	(0.063)	(0.065)
\(\Delta \text {CDS}\)	−0.031\(^{***}\)	−0.035\(^{***}\)	−0.036\(^{***}\)
	(0.008)	(0.006)	(0.005)
\(\Delta \text {Inflation}\)	−0.113	−0.142	−0.340
	(0.460)	(0.497)	(0.398)
\(\Delta \text {GDP change}\)	0.198	0.312	0.146
	(0.154)	(0.202)	(0.103)
\(\Delta \text {Euribor}\)	−5.517\(^{**}\)	−6.453\(^{***}\)	−5.289\(^{***}\)
	(2.242)	(2.108)	(1.854)
\(\Delta \text {Liquidity}\)	−8.209\(^{***}\)	−3.568\(^{**}\)	−5.802\(^{***}\)
	(1.889)	(1.514)	(1.865)
\(\text {TTM}\)	−0.0001	−0.0001	−0.0001
	(0.0001)	(0.0001)	(0.0001)
\(\text {Coupon}\)	−0.162	−0.084	−0.125\(^{*}\)
	(0.098)	(0.059)	(0.069)
Constant	1.603\(^{**}\)	0.734\(^{*}\)	1.237\(^{**}\)
	(0.735)	(0.412)	(0.557)
Observations	829	838	838
R\(^{2}\)	0.738	0.649	0.740
Adjusted R\(^{2}\)	0.735	0.645	0.737
Residual standard error	1.934 (df = 819)	1.410 (df = 828)	1.513 (df = 828)

	Dependent variable
	Returns
	(1)	(2)	(3)	(4)
\(\Delta \text {VSTOXX}\)	0.013\(^{***}\)	0.013\(^{***}\)	0.013\(^{***}\)	0.013\(^{***}\)
	(0.001)	(0.001)	(0.001)	(0.001)
\(\Delta \text {CDS}\)	\(-\)0.024\(^{***}\)	\(-\)0.024\(^{***}\)	\(-\)0.024\(^{***}\)	\(-\)0.024\(^{***}\)
	(0.0002)	(0.0002)	(0.0002)	(0.0002)
\(\Delta \text {Euribor}\)	\(-\)1.068\(^{***}\)	\(-\)0.994\(^{***}\)	\(-\)1.078\(^{***}\)	\(-\)1.060\(^{***}\)
	(0.299)	(0.300)	(0.301)	(0.302)
\((-20,-1)\)	\(-\)0.018\(^{***}\)			\(-\)0.018\(^{***}\)
	(0.005)			(0.005)
\((0,3)\)		0.027\(^{**}\)		0.033\(^{***}\)
		(0.012)		(0.013)
\((1,20)\)			\(-\)0.004	\(-\)0.009
			(0.006)	(0.006)
Observations	50,891	50,891	50,891	50,891
R\(^{2}\)	0.282	0.282	0.282	0.282
Adjusted R\(^{2}\)	0.282	0.282	0.282	0.282
Residual standard error	0.512 (df = 50887)	0.512 (df = 50887)	0.512 (df = 50887)	0.512 (df = 50885)
F Statistic	5,005.619\(^{***}\) (df = 4; 50887)	5,003.677\(^{***}\) (df = 4; 50887)	5,002.063\(^{***}\) (df = 4; 50887)	3,338.627\(^{***}\) (df = 6; 50885)

	Dependent variable
	Pre-event return
	(1)	(2)	(3)	(4)
\(\Delta \text {EPU}\)	−0.005\(^{*}\)			−0.002
	(0.003)			(0.002)
\(\Delta \text {VSTOXX}\)	−0.017			−0.021
	(0.040)			(0.016)
\(\Delta \text {CDS}\)	−0.006\(^{***}\)			−0.003\(^{***}\)
	(0.002)			(0.001)
\(\Delta \text {Inflation}\)		0.001		−0.058
		(0.230)		(0.101)
\(\Delta \text {GDP change}\)		−0.021		−0.042
		(0.073)		(0.027)
\(\Delta \text {Euribor}\)		−1.991\(^{*}\)		−0.573
		(1.083)		(0.423)
\(\Delta \text {Liquidity}\)			−14.274\(^{***}\)	−10.371\(^{***}\)
			(2.992)	(1.399)
\(\text {TTM}\)			0.0002	0.0001
			(0.0004)	(0.0004)
Constant	0.330\(^{***}\)	0.091	0.096	0.173
	(0.080)	(0.111)	(0.083)	(0.135)
Observations	389	389	389	389
R\(^{2}\)	0.373	0.138	0.515	0.646
Adjusted R\(^{2}\)	0.368	0.131	0.513	0.639
Residual standard error	0.523 (df = 385)	0.614 (df = 385)	0.459 (df = 386)	0.396 (df = 380)

	Dependent variable
	Pre-event return
	(1)	(2)	(3)	(4)
\(\Delta \text {EPU}\)	−0.024\(^{***}\)			−0.021\(^{***}\)
	(0.006)			(0.005)
\(\Delta \text {VSTOXX}\)	0.114			0.118\(^{**}\)
	(0.071)			(0.051)
\(\Delta \text {CDS}\)	−0.053\(^{***}\)			−0.049\(^{***}\)
	(0.005)			(0.004)
\(\Delta \text {Inflation}\)		0.130		−0.360
		(1.126)		(0.402)
\(\Delta \text {GDP change}\)		0.020		−0.139
		(0.330)		(0.098)
\(\Delta \text {Euribor}\)		−7.949		−1.009
		(9.295)		(1.691)
\(\Delta \text {Liquidity}\)			−16.145\(^{***}\)	−4.848\(^{***}\)
			(4.121)	(1.314)
\(\text {TTM}\)			−0.0002\(^{**}\)	−0.0002\(^{**}\)
			(0.0001)	(0.0001)
\(\text {Coupon}\)			−0.120	−0.142\(^{*}\)
			(0.089)	(0.073)
Constant	0.196	−1.472\(^{**}\)	0.163	1.466\(^{**}\)
	(0.252)	(0.627)	(0.381)	(0.604)
Observations	838	838	838	838
R\(^{2}\)	0.660	0.067	0.295	0.745
Adjusted R\(^{2}\)	0.659	0.064	0.292	0.742
Residual std. error	1.989 (df = 834)	3.294 (df = 834)	2.863 (df = 834)	1.728 (df = 828)

	Dependent variable
	Pre-Event Return
	(1)	(2)	(3)	(4)	(5)	(6)
\(\Delta \text {EPU Italy}\)	−0.029\(^{***}\)	−0.029\(^{***}\)	−0.029\(^{***}\)	−0.028\(^{***}\)	−0.029\(^{***}\)	−0.024\(^{*}\)
	(0.005)	(0.004)	(0.006)	(0.005)	(0.006)	(0.013)
\(\Delta \text {EPU Europe}\)		0.0002				−0.009
		(0.013)				(0.021)
\(\Delta \text {EPU Germany}\)			0.001			0.0003
			(0.005)			(0.004)
\(\Delta \text {EPU Spain}\)				0.003		0.009
				(0.008)		(0.014)
\(\Delta \text {EPU Greece}\)					−0.0003	0.002
					(0.003)	(0.006)
\(\Delta \text {VSTOXX}\)	0.117\(^{*}\)	0.117\(^{*}\)	0.113	0.105	0.118\(^{*}\)	0.083
	(0.065)	(0.068)	(0.084)	(0.081)	(0.064)	(0.092)
\(\Delta \text {CDS}\)	−0.036\(^{***}\)	−0.037\(^{***}\)	−0.037\(^{***}\)	−0.037\(^{***}\)	−0.036\(^{***}\)	−0.035\(^{***}\)
	(0.005)	(0.006)	(0.004)	(0.005)	(0.005)	(0.006)
\(\Delta \text {Inflation}\)	−0.340	−0.348	−0.374	−0.426	−0.345	−0.252
	(0.398)	(0.651)	(0.528)	(0.517)	(0.427)	(0.746)
\(\Delta \text {GDP change}\)	0.146	0.144	0.140	0.100	0.145	0.109
	(0.103)	(0.159)	(0.111)	(0.170)	(0.105)	(0.170)
\(\Delta \text {EURIBOR}\)	−5.289\(^{***}\)	−5.282\(^{***}\)	−5.288\(^{***}\)	−4.281	−5.312\(^{***}\)	−2.807
	(1.854)	(1.889)	(1.822)	(3.101)	(1.840)	(4.820)
\(\Delta \text {Liquidity}\)	−5.802\(^{***}\)	−5.816\(^{***}\)	−5.868\(^{***}\)	−6.073\(^{***}\)	−5.766\(^{***}\)	−6.211\(^{***}\)
	(1.865)	(1.942)	(1.873)	(1.931)	(1.738)	(1.616)
TTM	−0.0001	−0.0001	−0.0001	−0.0001	−0.0001	−0.0001
	(0.0001)	(0.0001)	(0.0001)	(0.0001)	(0.0001)	(0.0001)
Coupon	−0.125\(^{*}\)	−0.125\(^{*}\)	−0.125\(^{*}\)	−0.125\(^{*}\)	−0.125\(^{*}\)	−0.125\(^{*}\)
	(0.069)	(0.069)	(0.068)	(0.069)	(0.069)	(0.068)
Constant	1.237\(^{**}\)	1.239\(^{**}\)	1.248\(^{**}\)	1.289\(^{**}\)	1.236\(^{**}\)	1.331\(^{**}\)
	(0.557)	(0.546)	(0.542)	(0.549)	(0.557)	(0.563)
Observations	838	838	838	838	838	838
R\(^{2}\)	0.740	0.740	0.740	0.741	0.740	0.742
Adjusted R\(^{2}\)	0.737	0.737	0.737	0.738	0.737	0.738
Residual standard error	1.513 (df = 828)	1.513 (df = 827)	1.513 (df = 827)	1.512 (df = 827)	1.513 (df = 827)	1.512 (df = 824)