Systemic stress, sovereign health and household deposit behavior in the euro area

(1)

1 | P a g e

SYSTEMIC STRESS, SOVEREIGN HEALTH

AND HOUSEHOLD DEPOSIT BEHAVIOR IN

THE EURO AREA

Master thesis in Economics, Monetary Policy and Banking

Faculty of Economics and Business, University of Amsterdam

Author: Sander Mulders, 10203877

Supervisor: Dr. K. Mavromatis

Version: 14 august 2017

Word count: 12500 (excl. references and appendices)

Key words:

(2)

2 | P a g e

ABSTRACT

This research investigates the effect of sovereign financial stress on household deposit balances during periods of high systemic stress in the financial markets. The results show that the lagged changes in financial stress indicators of several governments have significant relationships with changes in household deposits if high systemic stress is present in the financial markets. Countries that experienced high macro-financial imbalances – such as Greece, Spain, Italy and Ireland – displayed a negative relationship while Finland demonstrated a positive relationship. The policy implication following from this research is of utmost importance for countries that experience future high systemic stress in their financial markets, as democratically elected administrations have an electoral incentive to interfere in the financial markets during periods of systemic stress, taking up extra risk. This thesis provides evidence that this possibly aggravates bank funding problems through stimulation of flight to safety behavior of household deposits. Given that the negative relationships are primarily found for countries that experienced high macro-financial imbalances, our findings suggest that these imbalances played an important role. However, more research is needed to be conclusive on that point.

STATEMENT OF ORIGINALITY

This document is written by Sander Louis Mulders, who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and no sources other than those mentioned in the text and those in references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3 | P a g e

DEFINITIONS

SSS = Systemic stress in financial markets at the sovereign level EUROSS = Systemic stress in financial markets at the Euro-area level

CISS = Composite indicator of systemic stress.

Core-countries = Germany, Netherlands and Finland PIIGS-countries = Portugal, Ireland, Italy, Greece and Spain Non defined class- countries = Austria, France and Belgium

(4)

4 | P a g e

1. INTRODUCTION

This thesis analyzes research into the effect of changes in sovereign financial stress to household deposits balances for eleven euro-area countries during periods of high systemic stress in financial markets. The countries of interest are Germany, the Netherlands, Finland, France, Belgium, Austria, Portugal, Ireland, Italy, Greece and Spain. The research question of interest is: To what extent did preceding changes in sovereign financial stress exhibit a significant relation with changes in household deposits, during periods of high systemic stress in financial markets?

Household deposits are an important source of liquidity and funding for banks. Literature on the subject indicates that 75% of bank funding depends on deposits and are hence of paramount importance to the liquidity needs and stability (Demsetz and Strahan, 1997; Demirgüç-Kunt et al., 2013). However, in the aftermath of the financial crisis, the euro area has experienced financial fragmentation.1_{The fall of Lehman Brothers in the United States of America led to a substantial} hoarding of liquidity in the EU banking system. In order to mitigate the effects, the EU states

engaged in the bailout of systemically important banks to preserve their financial system. This action led to a deterioration in the sovereign fiscal stances. Consequently, sovereign risk premiums rose, and interest rates increased substantially for the vulnerable countries (Knot, 2018). Knot (2018) further reasoned that these uncertain developments might incentivize a certain households and business behavior, known as “flight to safety” .2_{This behavior could have further exacerbated bank} funding problems in the vulnerable countries.

Besides that, flight to safety of household deposits reinforce financial fragmentation among Euro area members.3_{Fragmented financial markets can be a hurdle for the implementation of} monetary policy. As the governing council of the European Central Bank (ECB) sets the monetary policy for the entire euro area, their policy decisions are based on the macroeconomic condition of its member states. As long as these member states show a certain degree of homogeneity, the optimal monetary policy affects each economy moderately in a homogeneous way. However, if the member states are heterogeneous, decision making in the governing council becomes more difficult. As the optimal monetary policy is based on the average economic condition of the member states. Consequently, no monetary policy is considered optimal for all members. Graph 1 provides insight

1_{Financial fragmentation is considered a process of disintegration. Integration in financial markets ought to be}

strengthened in case there are identical rules and agents have equal access to financial instruments or services in these markets. (Baele,L., et al., 2004). An example of financial fragmentation could be the price spread between securities, while equal in terms of maturity and risky aspects. (Horny, G., et al., 2016).

2_{Flight to safety is the action of investors moving their capital away from riskier investments to safer ones.} Uncertainty in the financial or international markets usually causes this herd-like shift.

(6)

6 | P a g e into heterogeneous capital flows among euro area members. It displays TARGET24_{balances for the} PIIGS- and core-countries 5

Graph 1

To address the research question, two time series models are constructed. The first time series model measures the effect of lagged changes in sovereign financial stress indicators (i.e., the change in the long-term interest rate, short-term interest rate and debt-to-GDP) on the growth rate in deposit balances, during periods of systemic stress in financial markets at the country level. The second time series model measures the effect of lagged changes (in the same indicators) on the growth rate in deposit balances, during periods of systemic stress in financial markets at the

European level. Both time series models use a category variable that distinguishes between periods of “high systemic stress,” “mediocre systemic stress” and “low systemic stress” periods in the financial markets. The category variables are constructed by the author and extract their value from a Composite Indicator of Systemic Stress (CISS) index provided by the European Central Bank (ECB).

The results indicate that changes in household deposit balances in Greece, Spain, Ireland and Italy had a significant inverse relationship with lagged changes in sovereign financial stress indicators during periods of high systemic stress in their domestic financial markets. This infers that, besides the effect of systemic stress in the financial markets on household deposit behavior, previous changes in the financial resilience of the sovereign financial health also contributed to household deposit inflows or outflows. The policy implication for these findings is of utmost importance for these countries, as democratically elected administrations have an electoral incentive to interfere

4_{TARGET2 is the real-time gross settlement (RTGS) system owned and operated by the Eurosystem.} 5_{The core countries consist of Germany, Netherlands and Finland. The PIIGS-Countries consist of Portugal,} Ireland, Italy, Greece and Spain.

-1500000,00 -1000000,00 -500000,00 0,00 500000,00 1000000,00 1500000,00 20 01J an 20 01De c 20 02No v 20 03Oct 20 04S e p 20 05Aug ₂₀06J u l 20 07J u n 20 08 May 20 09Apr 20 10Mar 20 11F e b 20 12J an 20 12De c 20 13No v 20 14Oct 20 15S e p 20 16Aug ₂₀17J u l B ill io n s o f e u ro s

Target2

Core-countries PIIGS - Countries

(7)

7 | P a g e during periods of systemic stress for their financial markets, taking up more risk. Our results prove that these actions are like carrying water in a danaid’s jar for several countries,6_{as these}

interferences led to a worsening fiscal health and perceptions of increased default risk on the financial markets. They further stimulated flight to safety behavior among household depositors and aggravated domestic bank funding problems.

Furthermore, our results find a positive significant relation between changes in household deposits and changes in sovereign financial stress indicators for Finland during periods of high systemic stress in financial markets at the European level. These results suggest that Finland was considered a safe haven by many depositors in the euro area. Therefore, they experienced increases in household deposit balances while sovereign stress indicators worsened.

To put our findings in context provided by existing literature, the countries that saw highest macro-financial imbalances in the euro area were likely to provide significant inverse results except for Portugal. As literature (Smaga 2014) indicates that macro-financial imbalances contribute to the build-up of systemic risk and increase the potential impact of that risk. This research strongly

suggests that existing macro-financial imbalances played an important role in our findings. However, more research is needed to be conclusive on that point. But it would be interesting to further investigate in order to strengthen the policy implication following this research.

This thesis is structured as follows. Section 2 discusses existing literature on capital flows and macro-financial imbalances in Europe, systemic risk, switching deposit behavior and a historic

overview on the period of the financial crisis and sovereign debt crisis. Section 3 presents the time series models and hypothesis. Section 4 describes the data that has been gathered and provides a detailed description of the construction of the categorical variables that represent the periods of high, mediocre and low systemic stress. Section 5 presents the results. Section 6 interprets the significant findings of the two models and discusses them in the appropriate context. Section 7 concludes on the research. The appendices provide extra information on statistical assumptions and mathematical calculations used for this thesis.

(8)

8 | P a g e

2. LITERATURE REVIEW

This section provides the literature review, in order to give background information on the possible drivers of cross-border deposit flows among nations in the EMU-area. The primary focus for the literature review is on macro-financial imbalances within the monetary union and how they contribute to the build-up of systemic risk. Some attention is given to former research on deposit behavior. The literature review ends with an overview of the period of the great financial crisis and sovereign debt crisis.

2.2 Euro area capital flows and imbalances

Country specific imbalances grew in the first decade after the start of the euro in the EMU-area. A variety of researchers have published about those imbalances. Houben and Kakes (2013) report on the financial imbalances between the PIIGS-countries and Germany and relating it to the need for new macro-prudential policies for the EMU-area. Buiter, W.H. (2011) reports on the possible implications of intra-euro area imbalances in credit flows. In 2007, Sergio Rossi warns of economic divergence within the EMU-area as a consequence of free capital mobility within a currency area. He explains that capital–in the form of bank deposits–is mobile within a currency area, but immobile between two monetary spaces, as a consequence of the bookkeeping nature of bank deposits. Within the EMU, investments are made in those countries where revenues are highest, leading to an outflow of capital. These findings are of paramount importance towards our research as country specific imbalances indicators may provide explanations for deposit switching behavior within the euro area. The next paragraph discusses different empirics with respect to financial imbalances in the EMU-area.

2.2.1 Empirics of macro-financial imbalances.

Houben and Kakes (2013) provide a coherent overview of capital imbalances that have emerged since the formation of the EMU. They show that in the pre-period (1990-1998) and EMU-period (1998 till 2007), a large difference emerged. Financial imbalances among euro-members are increasingly related to the domestic monetary conditions. To examine the developments of the financial imbalances more closely, they presented empirics on the PIIGS-countries that experienced a steep decline in interest rates, rising asset prices, increased private and public debt and

deteriorating competitiveness. Summed up, these developments suggest that a too low interest rate created economic growth in the PIIGS-countries, financed by rising debts ratios. This expansion was not sustainable, as it was accompanied by growing current account deficits and loss of

(9)

9 | P a g e Another measure for intra euro capital flows is that of the Trans-European Automated Real-Time Gross Settlement Express Transfer System, abbreviated as TARGET2. As shown in Graph 3 on the next page, it is a clearing, recording and settlement system used by both public and private market participants and operated by the ECB. National Central Banks (NCBs) can build up claims and liabilities vis-à-vis TARGET2 over time. An increase in the balance of the European core countries exhibits a net capital outflow from the PIIGS countries (Buiter. et al, 2011). The following example clarify this: if an Italian account holder pays a Dutch account holder for its services, a claim is created from the Dutchman to the Italian. The Dutchman has a claim on his commercial bank, the

commercial bank has a claim on the De Nederlandsche Bank (DNB), the DNB has a claim on the ECB, the ECB on Banca d’Italia (BdI), BdI on a Italian commercial bank and the Italian commercial bank on the Italian. Commercial Banks settle their accounts with clients through their accounts at the commercial bank. NCBs settle with commercial banks through their account at the NCB. NCBs register these claims and liabilities in Target 2 (CPB, 2012). There are a variety of transaction that might be the driving force behind intra euro capital imbalances. Buiter(2011) discusses mainly two other economic indicators that might give insight into changes in TARGET2: the current account imbalances and deposits flows.

(10)

10 | P a g e Graph2 Source: www.ecb.europ -200000,00 0,00 200000,00 400000,00 600000,00 800000,00 1000000,00 20 05J an 20 05J u l 20 06J an 20 06J u l 20 07J an 20 07J u l 20 08J an 20 08J u l 20 09J an 20 09J u l 20 10J an 20 10J u l 20 11J an 20 11J u l 20 12J an 20 12J u l 20 13J an 20 13J u l 20 14J an 20 14J u l 20 15J an 20 15J u l 20 16J an 20 16J u l 20 17J an 20 17J u l 20 18J an M ill io n s o f E u ro

Target 2 core-countries

Germany Finland Netherlands -160000,00 -140000,00 -120000,00 -100000,00 -80000,00 -60000,00 -40000,00 -20000,00 0,00 20000,00 40000,00 20 05J an 20 05J u l 20 06J an 20 06J u l 20 07 Jan 20 07J u l 20 08J an 20 08J u l 20 09J an 20 09J u l 20 10J an 20 10J u l 20 11J an 20 11J u l 20 12J an 20 12J u l 20 13J an 20 13J u l 20 14J an 20 14J u l 20 15J an 20 15J u l 20 16J an 20 16J u l 20 17J an 20 17 Ju l 20 18J an M ill io n s o f E u ro

Target 2 non-defined class-countries

Austria Belgium France -500000,00 -400000,00 -300000,00 -200000,00 -100000,00 0,00 100000,00 20 05J an 20 05J u l 20 06J an 20 06J u l 20 07J an 20 07J u l 20 08J an 20 08J u l 20 09 Jan 20 09J u l 20 10J an 20 10 Ju l 20 11J an 20 11J u l 20 12J an 20 12J u l 20 13J an 20 13J u l 20 14J an 20 14J u l 20 15J an 20 15J u l 20 16J an 20 16J u l 20 17J an 20 17J u l 20 18J an M ill io n s o f E u ro

Target 2 PIIGS-countries

Spain Greece Ireland Italy Portugal

(11)

11 | P a g e Current account surplus can be defined as the increase in a nation’s net foreign claims.7_A current account deficit is the opposite (Buiter, 2011). The International Monetary Fund uses the following definition: “the difference between the value of exports of goods and services and the value of imports of goods and services.” A deficit then means that the nation is importing more goods and services than it is exporting; the current account also includes net income (such as interest and dividends) and transfers from abroad (such as foreign aid), which are usually a small fraction of the total. A current account deficit could be funded by capital outflow that is the result of transactions in financial assets between domestic, foreign, private and public entities., Schmits and von Hagen (2011) state that current account imbalances in the euro area have widened significantly since the start of EMU. Lane and Milesi-ferretti (2011) show that the pre-crisis current account deficit and rate of domestic credit expansion correlate with the scale of the decline in output and expenditure between 2007 and 2009. Additionally, they show that above-normal current account deficits during 2005 to 2008 were associated with sharp current account reversals and expenditure reductions after 2008. Graph 3 displays the current account for the countries of interest of this research.

7_{The current account surplus is the value of the net change in claims on the rest of the world, not the change in} the value of net claims on the rest of the world.

(12)

12 | P a g e Graph 3 Source: OECD -20,0 -15,0 -10,0 -5,0 0,0 5,0 10,0 15,0 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 % o f GDP

Current account PIIGS-countries

Spain Greece Ireland Italy Portugal -2 -1 0 1 2 3 4 5 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 % o f GDP

Current account non defined class-countries

Austria Belgium France -4,0 -2,0 0,0 2,0 4,0 6,0 8,0 10,0 12,0 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 % o f GDP

Current account core-countries

Germany Finland Netherlands

(13)

13 | P a g e 2.2.2 Macro-financial imbalances and systemic risk

Smaga (2014) reviews a variety of systemic risk definitions in the literature. He proposes defining systemic risk as a risk that a shock will result in such a significant materialization of (e.g., macro-financial) imbalances that the spread, impairs the functioning of the financial system to the extent that it adversely affects the real economy (such as by reducing economic growth) . Following the paper created by the BIS, IMF and FSB (2009) for the G20, systemic risk can be defined as a “risk of disruption to financial services that are caused by an impairment of all or parts of the financial system and has the potential to have serious negative consequences for the real economy”.

From a conceptual point of view, systemic risk has two dimensions. First is the

cross-sectional dimension; this relates to the interconnectedness and common exposures to shocks of the financial system. Second, there is the time dimension; This relates to the build-up of risk over time and is endogenous to the financial cycle of the economy. The dynamic forces of the real economy and the financial system reinforce each other. The amplitude of booms and busts increases, which undermines the macroeconomic and financial stability (Caruanan, J 2010)

Smaga (2014) discusses financial system characteristics and argues that macro-financial imbalances contribute to the build-up of systemic risk and increase its potential impact when the risk materializes. Furthermore, he claims that home bias 8_{with respect to sovereign bond holdings also} increases the cross-sectional exposure of the financial system and the sovereign.

Ang and Longstaf (2014) provide strong support for the view that systemic sovereign risk has its roots in financial markets. They state that sovereign systemic risk is highly correlated with

financial market variables. However, more research is needed to understand the causal relation between the two.

8_{Home bias is the tendency for investors to invest in a large number of domestic assets, despite the purported} benefits of diversifying into foreign assets.

(14)

14 | P a g e

2.3 Switching deposit behavior

To give more insight into what drives deposits account switching behavior, this section includes a coherent overview of the literature around what drives switching behavior between bank accounts. Most of the literature on these matters is related to individuals rather than the flows within countries. However, in order to have a better understanding of the stability of deposits, it is crucial to understand the determinants. Recent literature focusses primarily on the switching behavior of an individual’s main bank. Kisser (2002) and Brunetti et al. (2014) consider actual bank switching behavior. Brunetti et al. (2014) find evidence that the use of more than a single bank, as well as the number of services used (intensity), and the services used (scope relationship) at the main bank are significantly influential in shaping the households’ decision to switch. Chakravarty et al. (2004), Manrai and Manrai (2007) and van der Cruijsen and Diepstraaten (2015) focus more on the propensity to switch accounts. Cruijsen and Diepstraaten (2015) claim that the primary motives to stay at one’s bank are practical barriers, good bank-customer relationship, and the perception that there is not much benefit in switching. On the contrary, Gerritsen, Bikker and Brandsen (2017) research towards bank switching behavior of individuals’ accounts, instead of only individuals’ main accounts. They match the degree of switching with the deposit differential for several banks. After controlling for demographics, they found that switching is positively related with the differential in deposit rates between banks. While the fraction of deposits switched increased during the financial crisis, deposit rates were unrelated to bank switching during that period. These findings suggest that there were other external factors at play. They suggest that the increased switching behavior during the financial crisis was due to a flight-to-safety among depositors and consider this as evidence that a deposit guarantee scheme did not provide sufficient stability once oversized banks experience several shocks. Graph 4 on the next page provides household deposit balances for the 11 euro area countries of interest in this research.

(15)

15 | P a g e Graph 4 Source: data.europe.eu 80 100 120 140 160 180 200 20 04J an 20 04J u l 20 05J an 20 05J u l 20 06J an 20 06J u l 20 07J an 20 07J u l 20 08J an 20 08J u l 20 09J an 20 09J u l 20 10J an 20 10J u l 20 11 Jan 20 11 Ju l 20 12 Jan 20 12 Ju l 20 13 Jan 20 13J u l 20 14J an 20 14J u l 20 15J an 20 15J u l 20 16J an 20 16J u l 20 17J an 20 17J u l

Household deposits core-countries (Jan 2004 = 100)

Germany Finland Netherlands 80 100 120 140 160 180 200 220 20 04J an 20 04J u l 20 05J an 20 05 Ju l 20 06J an 20 06J u l 20 07J an 20 07J u l 20 08J an 20 08J u l 20 09J an 20 09J u l 20 10J an 20 10J u l 20 11J an 20 11J u l 20 12 Jan 20 12J u l 20 13J an 20 13J u l 20 14J an 20 14J u l 20 15J an 20 15J u l 20 16J an 20 16J u l 20 17J an 20 17J u l

Household deposits non-defined class-countries

(Jan 2004 = 100)

Austria Belgium France 80 100 120 140 160 180 200 220 240 20 04J an 20 04J u l 20 05J an 20 05 Ju l 20 06J an 20 06J u l 20 07J an 20 07J u l 20 08J an 20 08J u l 20 09J an 20 09J u l 20 10J an 20 10J u l 20 11J an 20 11J u l 20 12 Jan 20 12J u l 20 13J an 20 13J u l 20 14J an 20 14J u l 20 15J an 20 15J u l 20 16J an 20 16J u l 20 17J an 20 17J u l

Household deposits PIIGS-countries (Jan 2004 = 100)

Spain Ireland Portugal Greece Italy

(16)

16 | P a g e

2.4 Period of the financial crisis, sovereign debt crisis and response ECB, EU.

The advent of the global financial crisis was in August 2007. The crisis started with the major losses of asset-backed securities in the U.S. market. Because European banks had relatively high exposure to the U.S. market, the crisis entered Europe through cross sectional contagion. The global crisis became more critical in September 2008 with the fall of Lehman Brothers. Initially, the

authorities focused less on the sovereign debt of the European nations. Throughout the early years of the crisis, the focus was primarily on the monetary actions of the ECB so as to address the impairments in the financial system (Lane 2012). The shock brought forward asymmetric shocks across the euro-area. Cross-border financial flows dried up in late 2008, with investors repatriating capital to home markets and reevaluating their transnational exposure levels (Milesi-Ferretti and Tille 2011). This process affected countries in a disproportionate way. Countries that were heavily relying on external funding, especially in short term debt markets, were most severely hit.

Additionally, the crisis provoked a reassessment of asset prices and growth prospects, especially for those countries that displayed macro-financial imbalances. However, relatively low pre-crisis public debt gave countries sufficient comfort that they could likely absorb the costs of a banking crisis. Demand for government debt countries was increased by the banking sector, by means of sovereign bonds that were attractively rated as collateral by the Basel regulatory capital requirements (Lane 2012).

In the winter of 2009, the crisis in Europe reached a new stage. There was a variety of countries that reported a larger than expected fiscal deficit as a percentage of GDP that year. The most appalling news came from Greece. After the elections held in 2009, the new government revised the budget deficit forecast to a stunning 12.7 percent of GDP, more than doubling the previous estimate. Politicians attributed these developments to the irresponsible fiscal behavior of the peripheral countries, but the financial imbalances were also important factors in the worsening fiscal stance. These adverse developments were reflected in rising spreads on long-term sovereign bonds. The spread on these bonds diverged massively between PIIGS-countries and Germany (which is considered risk-free). Before the crisis, the difference was close to zero. As all countries share the same currency, these divergences thus predominantly represent perceived credit risk and

differences in volatility. Greece was the first country to be shut out of the bond market in May 2010, followed closely by Ireland and Portugal, in November 2010 and April 2011, respectively. Spain and Cyprus followed later (Lain 2012). Though, the public debt crisis followed the banking crisis, the reversed is not unthinkable. A public debt crisis entails potential risk for the banking sector. This relation is often shown in the literature by the existing correlations between banks and sovereign

(17)

17 | P a g e Credit Default Swaps (CDS) (Blundell-Wignall & Slovik, 2011). The primary reason this occurs is that the value of bank assets falls due to their holdings of over indebted countries’ public debt, whose market value decreases (Roman and Bilan, 2012). Jeanne and Bolton (2011) show that, in highly integrated areas like the EMU, additional contagion effects may occur. Therefore, through the banking sector, a public debt crisis might affect the other countries. These potential threats to for the overall stability of the financial sector in the EMU, led to a strong monetary policy response in the ECB.

2.4.1 Response ECB

The literature that is available on the response of the ECB on the financial crisis and the sovereign debt crisis splits the time frame since the fall of Lehman brothers into several phases. Each phase is characterized by an economic event or cause and a remedy implemented by the ECB. The first phase stretches from the fall of Lehman Brothers until April 2010. The second phase starts with the dawn of the sovereign debt crisis in May 2010, which lasted until Augusts 2011. During this phase, Greece, Ireland and Portugal were bailed out. The third phase is a re-intensification of the euro-area sovereign debt crisis, which carried on between August 2011 and January 2013. Phase four is epitomized by deflationary risks for the euro area. This section discusses the several phases, causes and responses (ECB.europe.eu).

The response of the ECB during phase one was mainly focused on providing extra credit to the financial system. Therefore, the ECB implemented the “enhanced credit support” (ECS) program. This program involved some non-standard measures. First the ECB provided a fixed rate system with full allotment tenders. Many financial institutions faced a liquidity shortage. Through this measure, financial institutions could refinance through the ECB at a fixed rate for a known period. Additionally, the ECB cut the main refinancing rate (MRR) from 4.25 to 1.00. Moreover, within the ECS-program the loan conditions were altered. The ECB lowered the threshold for collateral and implemented three, six and 12 month long-term refinancing operations (LTROs) (Rodriquez & Carrasco 2014). Finally, the ECB decided to implement their first corporate bond purchase program (CBPP1) in order to ease funding conditions and encourage institutions to expand and maintain their lending to clients. The size of this purchase program amounted to 61 billion euros (ECB.europe.eu). During the second phase, the ECB initiated the securities market program (SMP). This program had reached a size of 100 billion euros by august 2011. In this period, secondary markets showed less interest in the bonds of bailed out countries. According to the press release of the ECB, the rationale of the SMP was the following: “conduct interventions in the euro area public and private debt securities markets to ensure depth and liquidity in those segments which are dysfunctional”. This was all done

(18)

18 | P a g e with the objective: “to address the malfunctioning of securities markets and restore an appropriate monetary policy transmission mechanism” (Governing Council of the ECB, 2010) . During the third phase, the ECB announced that the non-standard measures as implemented under the ECS-program would remain in place as long as necessary. Additionally, the ECB cut the MRR to 0.75 basepoints and re-activated the SMP in august 2011. The SMP reached a final amount of 220 billion euros in sovereign bonds from stressed countries in February 2012. The ECB also implemented the second covered bond purchase program (CBPP2). This program reached 16 billion euros. Finally, the ECB provided two new, very long-term refinancing operations (VLTROs). The first one, in December 2011 amounted to 489 billion euros and the second, in February 2012, 529 billion euro. In the face of serious deflationary risk and weak recovery of the Eurozone economy, the ECB used a variety of instruments that had been used earlier to reach their objectives. Among these are lowering the MRR and the use of LTROs. However the response to the economic conditions in the fourth phase

included some new instruments. The first is the use of forward guidance9_{on the future policy path} of the ECB (Rodriquez & Carrasco 2014). The second was the implementation of the extended asset purchase program (APP). This program is comprised of several other programs and includes: a third covered bond purchase program (CBPP3); an asset-backed security program (ABSP); a public sector purchase program (PSPP) and a Corporate sector purchase program (PSPP) (ECB.europe.eu). These programs are still in effect at the time of writing. The euro system has purchased around €60 billion worth of securities every month from March 2015; from March 2016 to March 2017, the monthly purchasing volume totaled €80 billion, and fell again to €60 billion afterwards. In 2018 the monthly purchasing volume totals €30 billion (bundesbank.de).

2.4.2 Response EU member states and IMF

Besides the effort of the ECB, the member states of the European Union provided help to countries that were in distress during the second phase. This process started with the exclusion of Greece from the financial markets in May 2010. Consequently, Greece needed financial support. European leaders understood that no support would likely deteriorate the situation for Europe, as contagion might arise. Therefore, the first loan to Greece was given through bilateral support of the other Eurozone members and the IMF. In June 2010, the European Financial Stability Facility (EFSF) was created to function as a temporary backstop for sovereigns that risked exclusion from the financial

9_{If a central bank gives forward guidance, it means it is providing information about its future monetary policy} intentions, based on its assessment of the outlook for price stability.

The ECB began using forward guidance in July 2013 when the ECB’s Governing Council said that it expected interest rates to remain low for an extended period of time (ECB.europe.eu)

(19)

19 | P a g e markets. The EFSF provided three loans: The first was for Ireland in February 2011; the second was for Portugal in June 2011; and third was the second effort at support for Greece in March 2012. After these disbursements, the EU replaced the EFSF with the European Stability Mechanism (ESM) in October 2012. Spain was the first country to receive a loan from the ESM to recapitalize its banking system. In May 2013 ,Cyprus was granted a loan from the ESM. In July/August 2015, a country, Greece, received support from the ESM for the final time. It is currently the only ESM program that is still active (ESM.europe.eu)

table 2.1: Overview implemented purchase programs ECB. Announcement

date

Phase Global financial crisi september 2008

response i Coverd bond purchase program (CBPP) 07/05/2009

Phase Euro sovereign debt crisis may 2010

response i Securities Market Program (SMP) 10/05/2010

Phase Re-intsensification of sovereign debt crisis and banking sector strains

response i re-implementation of SMP 04/08/2011

ii Covered bond purchase program (CBPP2) 06/10/2011

Forecast of low growth and decling inflation

Phase Extended asset purchase program

response i covered bond purchase program 3 (CBPP3) 04/09/2014

ii Asset backed securities program (ABSP) 04/09/2014

iii Public sector purchase program (PSPP) 22/01/2015

iv Coporate sector purchase program (CSPP) 10/03/2016

Source: authors cut

Table 2.2: Overview financial support EU, EFSF, ESM and IMF Country Size loan (euros)

Greece 100 billion may-10 may-10 to jun-13

Ireland 67.5 billion nov-10 feb-11 to dec-13 Portugal 78 billion may-11 jun-11 to may-14

Greece 130 billion mrt-12 mrt-12 to jun-15

Spain 100 billion jul-12 jul-12 to dec-13

Greece 86 billion jul-15 aug-15 to aug-18

source: https://ec.europa.eu

https://www.esm.europa.eu/assistance

(20)

20 | P a g e

3. METHODOLOGY

This section discusses the econometric methodology and verifies whether certain hypotheses are falsifiable. This thesis opts for two time series models with multiple predictors (Stock & Watson 2007) and uses OLS-methodology to obtain the coefficients. The first model is used to gain a deeper understanding on the effect of previous period changes in sovereign financial stress indicators on the growth rate in household deposit balances during periods of high systemic stress in financial markets at the sovereign level (SSS). The second model verifies whether previous periods changes in

sovereign financial stress indicators had an effect on the growth rate of household deposit balances during periods of high systemic stress in financial markets at the European level (EUROSS). More explanation on why this thesis opts for a second model is explained in subsection 3.3 Model 2: Systemic stress at the European level Within both models binary variables, lagged values of multiple predictors and their interaction terms are regressed on the dependent variable. The interaction terms are of specific interest to the research. Both regressions are run using robust standard errors.

3.1 Baseline model & hypotheses

The two time series models are based on a baseline model. Due to limitations in the data and minimizing of collinearity issues, the final results for some countries have fewer variables. The limitations are further discussed in section 4. The transformations of the model with regard to multicollinearity presented in the results are discussed in APPENDIX C . The baseline model is as follows: ∆𝑯𝑯𝑫𝒕,𝒌= 𝛼𝑖,𝑘+ 𝛿𝑖1,𝑘∗ 𝑆𝑆𝐼𝐻𝑖,𝑘 + 𝛿𝑖2,𝑘∗ 𝑆𝑆𝐼𝑀𝑖,𝑘 + 𝛽1∆𝑰𝑹. 𝑺𝑻𝒕−𝟏,𝒌+ 𝛽2∆𝑰𝑹. 𝑳𝑻𝒕−𝟏,𝒌 + 𝛽3∆ ( 𝐃𝐄𝐁𝐓 𝑮𝑫𝑷𝒕−𝟏,𝒌) + 𝛾𝑖1,𝑘(𝑆𝑆𝐼𝐻𝑖,𝑘∗ ∆𝑰𝑹. 𝑺𝑻𝒕−𝟏,𝒌) + 𝛾𝑖2,𝑘(𝑆𝑆𝐼𝐻𝑖,𝑘∗ ∆𝑰𝑹. 𝑳𝑻𝒕−𝟏,𝒌) + 𝛾𝑖3,𝑘(𝑆𝑆𝐼𝐻𝑖,𝑘∗ ∆ ( 𝐃𝐄𝐁𝐓 𝑮𝑫𝑷𝒕−𝟏,𝒌 )) + 𝜀𝑡

The dependent variable ∆𝑯𝑯𝑫𝒕,𝒌 represents the growth rate in household deposits as a percentage

of the former period for a specific quarterly date “t” and country “k”. The variable 𝛼𝑖,𝑘 represents a

constant for every country “k” and for 𝑖 {EUROSS

SSS . The variable 𝑆𝑆𝐼𝐻𝑖,𝑘 is a binary variable systemic stress indicator. 𝑆𝑆𝐼𝐻𝑖,𝑘 is equivalent to a value of one during a period of high systemic stress in

financial markets and zero in all other cases, for 𝑖 {EUROSS

SSS and every country or euro-area “k”. The variable 𝑆𝑆𝐼𝑀𝑖,𝑘 is a binary variable systemic stress indicator. 𝑆𝑆𝐼𝑀𝑖,𝑘 is equivalent to a value of

(21)

21 | P a g e 𝑖 {EUROSS

SSS and every country or euro-area “k”. For more information on the definitions and used source information for the creation of the dummy variables 𝑆𝑆𝐼𝑀𝑖,𝑘 and 𝑆𝑆𝐼𝐻𝑖,𝑘 , refer to

subsection 4.1 A composite indicator for systemic stress.. The variable ∆𝑰𝑹. 𝑺𝑻𝒕−𝟏,𝒌 represents the

lagged change in the short-term interest on 3-month treasury bills, measured as a percentage, for a specific quarterly date “t” and country “k”. The variable ∆𝑰𝑹. 𝑳𝑻𝒕−𝟏,𝒌 represents the lagged change

in the long-term interest rate on 10-year government bonds, measured as a percentage, for a specific quarterly date “t” and country “k”. The variable ∆ ( 𝐃𝐄𝐁𝐓_𝑮𝑫𝑷)

𝒕−𝟏,𝒌 represents the lagged change

in the amount of debt outstanding, as a percentage of GDP, for a specific quarterly date ‘𝑡’ and country “k”. The interaction variable (𝑆𝑆𝐼𝐻𝑖,𝑘∗ ∆𝑰𝑹. 𝑺𝑻𝒕−𝟏,𝒌) represents the interaction effect of

the lagged change in short-term interest rate and a period of high systemic stress for 𝑖 {EUROSS SSS , every country “k” and period “t”. The interaction variable (𝑆𝑆𝐼𝐻𝑖,𝑘∗ ∆𝑰𝑹. 𝑳𝑻𝒕−𝟏,𝒌) represents the

interaction effect of the lagged change in long-term interest rate and a period of high systemic stress for 𝑖 {EUROSS

SSS , every country “k” and period “𝑡”. The interaction variable (𝑆𝑆𝐼𝐻𝑖,𝑘∗ ∆ ( 𝐃𝐄𝐁𝐓

𝑮𝑫𝑷 𝒕−𝟏,𝒌)) represents the interaction effect of the lagged change in debt to GDP ratio and a

period of high systemic stress for 𝑖 {EUROSS

SSS , every country “k” and period “𝑡”. The residual 𝜀𝑡 is the error term of the regression.

The contribution of the thesis is to verify the following hypotheses.

“previous period changes in (1) the short-term interest rate, (2) the long-term interest rate and (3) debt to GDP of a sovereign during periods of high systemic stress do have a significant impact on household deposits behavior for the countries of interest.”

Accordingly, this thesis verifies whether the interaction variable (𝑆𝑆𝐼𝐻𝑖,𝑘∗ ∆𝑰𝑹. 𝑺𝑻𝒕−𝟏,𝒌) had a

significant impact on household deposits or not. Therefore, a test was performed to evaluate whether there is sufficient evidence to reject 𝐻0: 𝛾𝑖1,𝑘= 0 , and accept 𝐻1: 𝛾𝑖1,𝑘 ≠ 0 for all

countries “k” and 𝑖 {EUROSS

SSS . Second, this thesis verifies whether the interaction variable (𝑆𝑆𝐼𝐻𝑖,𝑘∗ ∆𝑰𝑹. 𝑳𝑻𝒕−𝟏,𝒌) had a significant impact on household deposits or not. Therefore a test

was performed to evaluate whether there is sufficient evidence to reject 𝐻2: 𝛾𝑖2,𝑘 = 0 , and accept

𝐻3: 𝛾𝑖2,𝑘≠ 0 for all countries “k” and 𝑖 {EUROSS_SSS . Third, this thesis verifies whether the interaction

variable (𝑆𝑆𝐼𝐻𝑖,𝑘∗ ∆ ( 𝐃𝐄𝐁𝐓

(22)

22 | P a g e Therefore, a test was performed to evaluate whether there is sufficient evidence to reject

(23)

23 | P a g e

3.2 Model 1: Systemic stress at the sovereign level.

The first model verifies the effect of the lagged change in the long-term interest rate, the short-term interest rate and the debt to GDP ratio on the growth rate of household deposits during periods of high systemic stress in domestic financial markets. This regression is executed on

Portugal, Ireland, Italy, Greece and Spain. These five countries are categorized into a group called the PIIGS-countries. Due to no availability of periods of high sovereign systemic stress10_{for the other} countries of interest, the regression is only run for the PIIGS-countries. The regression equation reads: ∆𝑯𝑯𝑫𝒕,𝒌= 𝛼𝑆𝑆𝑆,𝑘+ 𝛿𝑆𝑆𝑆1,𝑘∗ 𝑆𝑆𝐼𝐻𝑠𝑠𝑠,𝑘 + 𝛿𝑠𝑠𝑠2,𝑘∗ 𝑆𝑆𝐼𝑀𝑠𝑠𝑠,𝑘 + 𝛽1∆𝑰𝑹. 𝑺𝑻𝒕−𝟏,𝒌+ 𝛽2∆𝑰𝑹. 𝑳𝑻𝒕−𝟏,𝒌 + 𝛽3∆ ( 𝐃𝐄𝐁𝐓 𝑮𝑫𝑷𝒕−𝟏,𝒌) + 𝛾𝑠𝑠𝑠1,𝑘(𝑆𝑆𝐼𝐻𝑠𝑠𝑠,𝑘∗ ∆𝑰𝑹. 𝑺𝑻𝒕−𝟏,𝒌) + 𝛾𝑠𝑠𝑠2,𝑘(𝑆𝑆𝐼𝐻𝑠𝑠𝑠,𝑘∗ ∆𝑰𝑹. 𝑳𝑻𝒕−𝟏,𝒌) + 𝛾𝑠𝑠𝑠3,𝑘(𝑆𝑆𝐼𝐻𝑠𝑠𝑠,𝑘∗ ∆ ( 𝐃𝐄𝐁𝐓 𝑮𝑫𝑷𝒕−𝟏,𝒌 )) + 𝜀𝑡 10_{This means that the binary variable 𝑆𝑆𝐼𝐻}

𝑘 only takes a value of 0 for the mentioned countries. There are no observations were the variable is equivalent to 1. For more information on the construction of the binary variable see subsection 4.1.1

(24)

24 | P a g e

3.3 Model 2: Systemic stress at the European level

The second model verifies the effect of the lagged change in the long-term interest rate, the short-term interest rate and the debt to GDP ratio on the growth rate of household deposits during periods of high systemic stress in European financial markets for 11 euro area countries. Note that this model is nearly the same as the first model. The only difference is that the data for the category variables are different. The primary reason that this thesis opts for a second model is to illustrate more accurately the relationship between the aforementioned variables and household deposits for Germany, Netherlands, Finland (i.e., core countries) and Austria, France and Belgium ( i.e., non-defined class-countries). The country-specific CISS-index on systemic stress did not provide sufficient data to construct a binary variable that indicated periods of high systemic stress for these countries. However, there might also be a causal relationship between changes in sovereign financials stress indicators during periods of high systemic stress in the euro area. The following example clarifies on a possible relationship.

If there was mediocre systemic stress in financial markets at the country level in a large economy core-country (e.g., Germany) it is not unlikely that there was also stress in the neighboring countries in a monetary union11_{. It might also be the case that systemic stress was higher in the} neighboring countries or that the mediocre systemic stress in Germany is caused by a high systemic stress in a neighboring country. Consequently, systemic stress at the state level in the core-country of interest would be high relative to former periods, but low relative to other countries systemic stress at that period in time. So Germany might be a relative stress-less country instead, which could explain a deposit inflow for such countries. This author is not necessarily stating that this causal relation is factual, but rather that there is the possibility of a causal relationship. Therefore, it is interesting to investigate the results of a model that incorporates systemic stress at the European level instead of the state level.

11_{Markets in a Monetary Union are usually highly integrated therefore financial or sovereign stress in one} market might affect other markets through cross-sectional exposures.

(25)

25 | P a g e The second model was created to verify the effects of sovereign financial stress indicators during periods of high systemic stress at the European level for the country of interest. This regression is executed on 11 countries that can be divided into three groups: the PIIGS-countries (Portugal, Ireland, Italy, Spain and Greece), the core countries (Germany, the Netherlands and Finland) and three undefined-class countries (Austria, France and Belgium). The regression equation reads as follows: ∆𝑯𝑯𝑫𝒕,𝒌= 𝛼𝐸𝑈𝑅𝑂𝑆𝑆,𝑘+ 𝛿𝐸𝑈𝑅𝑂𝑆𝑆1,𝑘∗ 𝑆𝑆𝐼𝐻𝐸𝑈𝑅𝑂𝑆𝑆,𝑘 + 𝛿𝐸𝑈𝑅𝑂𝑆𝑆2,𝑘∗ 𝑆𝑆𝐼𝑀𝐸𝑈𝑅𝑂𝑆𝑆,𝑘 + 𝛽1∆𝑰𝑹. 𝑺𝑻𝒕−𝟏,𝒌+ 𝛽2∆𝑰𝑹. 𝑳𝑻𝒕−𝟏,𝒌+ 𝛽3∆ ( 𝐃𝐄𝐁𝐓 𝑮𝑫𝑷𝒕−𝟏,𝒌)_{𝒕−𝟏,𝒌} + 𝛾𝐸𝑈𝑅𝑂𝑆𝑆1,𝑘(𝑆𝑆𝐼𝐻𝐸𝑈𝑅𝑂𝑆𝑆,𝑘∗ ∆𝑰𝑹. 𝑺𝑻𝒕−𝟏,𝒌) + 𝛾𝐸𝑈𝑅𝑂𝑆𝑆2,𝑘(𝑆𝑆𝐼𝐻𝐸𝑈𝑅𝑂𝑆𝑆,𝑘∗ ∆𝑰𝑹. 𝑳𝑻𝒕−𝟏,𝒌) + 𝛾𝐸𝑈𝑅𝑂𝑆𝑆3,𝑘(𝑆𝑆𝐼𝐻𝐸𝑈𝑅𝑂𝑆𝑆,𝑘∗ ∆ ( 𝐃𝐄𝐁𝐓 𝑮𝑫𝑷𝒕−𝟏,𝒌)) +

(26)

26 | P a g e

4. DATA

This section discusses the data gathered for this research. Furthermore, it describes in more detail the construction of the systemic stress binary variable categories in Subsection 4.1. In

Subsection 4.2, the mathematical relationship between bond prices and interest rates is presented. In Subsection 4.3 the relation between general deficit and debt is presented and why this variable is a relevant indicators as a measure of sovereign stress.

All data is quarterly and ranges from the second quarter in 2003 until the fourth quarter in 2017. The data was gathered for 11 euro area countries. The eleven countries are categorized in three different groups. The first group are the PIIGS-countries and consist of Portugal, Ireland, Italy, Greece and Spain. The second group are the core-countries and consist of Germany, the Netherlands and Finland. The third group are the undefined-class-countries and consist of France, Austria and Belgium. Other EMU-countries are left out of the research due to limitations in the dataset of the household deposits. The dependent variable comprises of a time series of the stock balances of household deposits for every country. The growth rate is calculated by the author. The data is obtained from the database of the European Union,12_{whose sources are the balance sheets of the} country specific monetary financial institutions. By using the quarterly growth rate in household deposits the stationarity assumption holds. A Dicky-Fuller test is employed in APPENDIX A to verify this postulation for all countries. Data is gathered for three continuous independent variables in order to explain the dependent variable. These variables concern the quarterly change in the short-term interest rate, the long-short-term interest rate and the debt-to-GDP ratio. A Dicky- Fuller test is employed in APPENDIX A to verify stationarity for all three independent variables. Data on the short-term interest rate is obtained through the OECD.13_{The short-term interest rate are the rates at} which short-term borrowings are effected between financial institutions or the rate at which short term government paper is issued or traded in the market. The short-term interest rates are usually averages of daily rates, measured as a percentage. Typical standardized names are money market rate and treasury bill rate. (OECD, 2018). Data on the long-term interest rate is obtained through the OECD14_{. Long-term interest rates refer to government bonds maturing in ten years. Rates are mainly} determined by the price charged by the lender, the risk from the borrower and the fall or rise in the value of the bonds. Long-term interest rates are usually averages of daily rates, measured as a percentage. These interest rates are implied by the prices at which the government bonds are traded on financial markets, not the interest rates at which the loans were issued. In all cases, the

12_{Source: https://data.europa.eu/euodp/data/dataset/bank-balance-sheet-deposits-stocks} 13_{https://data.oecd.org/interest/short-term-interest-rates.htm}

(27)

27 | P a g e interest rates refer to bonds with capital reimbursement guaranteed by governments. Long-term interest rates are one of the determinants of business investment. Low long-term interest rates encourage investment in new equipment and high interest rates discourage it. Investment is, in turn, a major source of economic growth (OECD, 2018).

The causal relationships between changes in the interest rate and deposit behavior are twofold. The first relates to changes in the interest rate and default risk of government bonds. The rationale is as follows: investors are willing to pay a higher price for bonds that bear low default risk and a lower price for bonds that bear a high default risk in the bond market. Subsequently, this affects the interest or yield to maturity gain on the bonds. Changes in the long-term interest rate on 10-year government bonds do inhibit information on the amount of default-risk and stress at the sovereign level15_{, given that the interest rate set by the ECB is constant}16_{. The same rationale applies to the} relationship between 3-month treasury bills and the short-term interest rate. However the causal relationship between changes in the short-term interest rate and default risk of the sovereign is less strong. The short-term interest rate is more strongly defined by monetary policy17_{in comparison to} the long-term interest rate. This is due to the time to maturity differences between long-term bonds and 3-month treasury bills. According to the term structure theory, increased uncertainty with respect to the default of a government affects long-term bond interest rates more severely than short-term interest rates on treasury bills in normal times. More information and a detailed

mathematical derivation of the relation between bonds and interest rates is given in Subsection 4.2 and APPENDIX B. The second argument is related to the non-satiation assumption of the axioms of rational choice theory. Generally, this microeconomic theory states that more of the same is always better. Generally depositors move their deposits where the ratio of interest to risk is highest, given equal risk preferences18_{. Note that this relationship might be stronger for the short-term interest} rate than for the long-term interest rate. Deposits are liquid instruments that can be transferred easily to another account with a higher (short-term) interest rate. The latter relation was noted previously in the literature review; Gerritsen, Bikker and Brandsen (2017) provided evidence that

15_{Note that it is a financial marked based measure. So changes are perceptions of financial markets.} 16_{Through the term-structure of the interest rate. Changes in the main refinancing rate of the ECB do also} influence the long-term interest rate. Therefore a change in the interest does not necessarily reflect a change in the amount of risk or stress at the sovereign level.

17_{The ECB sets the main refinancing rate. This is part of their monetary policy.}

18_{Note that we made a simplification to explain the causal relation between interest rates and deposit} behavior. In real life every depositor has its own risk appetite. Therefore this statement is not true for every situation, as depositors with risk-averse preferences might choose a lower interest/risk ratio as his preference is to store his deposits were risk is lower.

(28)

28 | P a g e interest rate differentials among banks explain deposit switching behavior. Higher interest rates should attract more deposits in non-crisis years.

Data of the debt-to-GDP ratio is obtained through EUROSTAT.19_{The general government} debt-to-GDP ratio is the amount of a state’s total gross government debt as a percentage of its debt-to-GDP. It is an indicator of an economy's strength and a key factor for the sustainability of government finance. A highly indebted country loses its ability and credibility to depositors to nationalize or bail out financial institutions. Debt is obtained as the sum of the following liability categories: currency and deposits; securities other than shares, except financial derivatives; loans; insurance technical reserves; and other accounts payable. Changes in government debt over time reflect the impact of government deficits. This indicator is measured as a percentage of GDP (OECD, 2018). Subsection 4.3

Mathematical derivation: The relation between general deficit and government debt provides a mathematical derivation on the relationship between government deficits and

government debt and further explains the relationship to systemic concerns and deposit behavior. Lastly, the binary variables that indicate whether a certain period is characterized by high systemic stress, mediocre systemic stress or low systemic stress extract their value from a composite indicator of systemic stress (CISS) index. This indicator is obtained through the statistical data warehouse of the ECB20_{. More detailed information on the construction of the dummy variables and} why three category variables are used is presented in the next section, Section 4.1 A

composite indicator for systemic stress.

19_{http://ec.europa.eu/eurostat/tgm/table.do?tab=table&init=1&language=en&pcode=teina230&plugin=1} 20_{https://sdw.ecb.europa.eu/browse.do?node=9689686}

(29)

29 | P a g e

4.1 A composite indicator for systemic stress.

This subsection discusses the dataset used to construct the binary variables systemic stress indicators 𝑆𝑆𝐼𝐻𝑖,𝑘 and 𝑆𝑆𝐼𝑀𝑖,𝑘 . It encompasses a composite indicator of systemic stress (CISS)

index of the financial system for every euro area country and the euro-area as a whole. The originators of the indicator are Dàniel Holló, Manfred Kremer and Marco Lo Duca (2012), who constructed the indicator in a working paper for the ECB. First, the definition of the CISS is discussed, and the technique that the authors used for the construction. Second, this part further expounds on the method employed to create the binary variables used in the regression analysis of this thesis. The binary variables extract their value from the CISS index.

The CISS index aims to quantify the current state of instability in the financial system entirely, which is roughly equivalent to systemic stress. It is a continuous variable stretching from a value of 0 (very low systemic stress) to 1 (very high systemic stress). Systemic stress is interpreted as the amount of systemic risk that has already materialized. Systemic risk, in turn, can be defined as the risk that financial instability will be so extensive that it impairs the functioning of a financial system to the point where economic growth and welfare suffer substantially (Holló, Kremer and Lo Duca, 2012).

The main distinguishing feature of the CISS-index in comparison to other financial stress indicators (FSI) is that it focusses on the systemic dimension of financial stress, an indicator

developed through a specific statistical design that incorporates standard definitions of systemic risk. The CISS contains 15 mostly marked based financial stress measures equally split into five categories. The categories are the financial intermediaries sector, money markets, equity markets, bond

markets and foreign exchange markets. A separate financial stress sub index is created for each of these five market segments after appropriate transformation of the individual stress measures. The foremost technical innovation of the CISS, in comparison to other FSIs, is the application of the basic portfolio theory to the aggregation of the sub-indices into a composite indicator. This methodology integrates the time-varying cross-correlations between the sub-indices. As a result, the CISS places relatively more weight on circumstances in which stress predominates in several market segments simultaneously. Therefore it comprehensively includes the thinking that systemic stress is more substantial if financial instability is spread broadly across the entire financial system. The second element of the aggregations scheme featuring systemic risk is the fact that the portfolio weights attached to each of the five sub-indices are calibrated on the basis of the relative strength of their dynamic impact on a measure of economic activity. (Holló, Kremer and Lo Duca, 2012).

(30)

30 | P a g e The variables 𝑆𝑆𝐼𝐻𝑖,𝑘 and 𝑆𝑆𝐼𝑀𝑖,𝑘 extract their value from the CISS index for every country

“k” and 𝑖 {EUROSS

SSS . The systemic stress binary variable 𝑆𝑆𝐼𝐻𝑖,𝑘 represent periods of high systemic stress while 𝑆𝑆𝐼𝑀𝑖,𝑘 represents periods of mediocre systemic stress. The reason for the creation of

the three binary variables is twofold21_.

First, this research is interested in showing the effect of lagged changes in sovereign financial stress indicators during periods of high systemic stress - relative to periods of low systemic stress - on the growth rate in household deposit balances. Therefore, it is convenient for the

interpretation of the results to construct binary variables that capture periods of high systemic stress and low systemic stress; to model the difference between these periods more accurately, the third category variable 𝑆𝑆𝐼𝑀𝑖,𝑘 is constructed. Binary variables provide a difference in mean analysis

(Stock & Watson, 2007). The difference of the mean between two sample periods of high and low systemic stress are likely to provide economically more significant coefficients, if the time interval is split up in three categories of low, mediocre and high periods of systemic stress than if only low and high systemic stress categories are used. Note that this only holds if there is a difference in the sample of the growth rate in household deposit balances between high and low periods of systemic stress. Therefore, If the difference in the mean of the growth rate in household deposits balances rises gradually with the increases in the CISS index, the difference between high and low periods is better captured through incorporation of a third binary variable that equals to 1 for periods of mediocre systemic stress.

Second, the CISS index is an index and hence a continuous variable with a minimum of 0 and a maximum of 1 that has non-stationary features. Through creation of a category of variables that extract their value from this continuous index, the non-stationarity is no longer present. This is done through the creation of thresholds for the CISS index. The next paragraph expounds on the

thresholds chosen to construct the binary variable

21_{Note that the constant represents periods of low systemic stress. Therefore the author mentions three binary} variables.

(31)

31 | P a g e 4.1.1 Construction binary sovereign systemic stress variable: Model 1

Within the first model, the binary systemic stress variables extract their value from the country specific CISS index created by Holló, Kremer and Lo Duca (2012).22_{The methodology that is} employed to create the sovereign systemic stress binary variables 𝑆𝑆𝐼𝐻𝑆𝑆𝑆,𝑘 and 𝑆𝑆𝐼𝑀𝑆𝑆𝑆,𝑘 is simple.

For the first model, two absolute thresholds are created to divide the continuous CISS index into three classes. As the CISS is an index that stretches from zero to one, the thresholds are chosen in such a way that it splits the range of the index in three classes23_{. One class for data points that are} considered high; one class for data points that are considered mediocre; and one class for data points that are considered low. The thresholds for the high systemic stress binary variable 𝑆𝑆𝐼𝐻𝑆𝑆𝑆,𝑘

lie within the following range:

0.7 ≤ 𝐶𝐼𝑆𝑆𝑆𝑆𝑆,𝑘 ≤ 1 ( 1 )

That means that the binary variable 𝑆𝑆𝐼𝐻𝑆𝑆𝑆,𝑘 has a value of 1 if the 𝐶𝐼𝑆𝑆𝑆𝑆𝑆,𝑘 is larger than or equal

to 0.7 and smaller or equal to 1 for every country “k”. In all other cases the binary variable is zero. The thresholds for the mediocre systemic stress binary variable 𝑆𝑆𝐼𝑀𝑆𝑆𝑆,𝑘 lie within the following

range:

0.35 ≤ 𝐶𝐼𝑆𝑆𝑆𝑆𝑆,𝑘 < 0.7 ( 2 )

That means that the binary variable 𝑆𝑆𝐼𝑀𝑆𝑆𝑆,𝑘 has a value of 1 if the 𝐶𝐼𝑆𝑆𝑆𝑆𝑆,𝑘 is larger than or equal

to 0.35 and smaller than 0.7 for every country “k”. In all other cases, the binary variable is 0. The thresholds for the low systemic stress periods lie within the following range:

0 ≤ 𝐶𝐼𝑆𝑆𝑆𝑆𝑆,𝑘< 0.35 ( 3 )

Note that this effect will be captured by the constant 𝛼𝑆𝑆𝑆,𝑘 in Model 1 for every country “k”. The

next pages provide a graphs with the SSS indicator per PIIGS-country for Model 1 and the thresholds.

22_{Holló, Kremer and Lo Duca (2012) created CISS index for every euro-area country.}

23_{The thresholds are rounded up to the nearest 0.05 decimal. Therefore the thresholds are not 0.33 and 0.66 but} rather 0.35 and 0.7

(32)

32 | P a g e 0 0,2 0,4 0,6 0,8 1 1,2 20 03Q2 20 03Q4 20 04Q2 20 04Q4 20 05Q2 20 05Q4 20 06Q2 20 06Q4 20 07Q2 20 07Q4 20 08Q2 20 08Q4 20 09Q2 20 09Q4 20 10Q2 20 10Q4 20 11Q2 20 11Q4 20 12Q2 20 12Q4 20 13Q2 20 13Q4 20 14Q2 20 14Q4 20 15Q2 20 15Q4 20 16Q2 20 16Q4 20 17Q2 20 17Q4

Sovereign Systemic Stress Indicator PIIGS- Countries

Spain Greece Italy Portugal Ireland SSIH SSIM

(33)

33 | P a g e 4.1.2 Construction binary Euro-area systemic stress variable: Model 2

Within the second model the binary systemic stress variable extract its value from the euro area CISS index created by Holló, Kremer and Lo Duca (2012). The methodology employed to create the binary variable for high euro area systemic stress 𝑆𝑆𝐼𝐻𝐸𝑈𝑅𝑂𝑆𝑆,𝑘 is slightly different than for the

first model. The reason why the methodology is different is the following. The absolute thresholds method used to obtain the category variables for the first model do not provide sufficient data points. Therefore, the threshold of high systemic stress in the euro area is not based on an absolute value but on a relative value. The choice for the relative threshold is further explained in APPENDIX D. Through this, we obtain sufficient data points on relatively high systemic stress periods in the euro-area .The thresholds for the high systemic stress binary variable 𝑆𝑆𝐼𝐻𝐸𝑈𝑅𝑂𝑆𝑆 lie within the

following range:

90𝑡ℎ 𝑝𝑒𝑟𝑐𝑒𝑛𝑡𝑖𝑙𝑒 < 𝐶𝐼𝑆𝑆𝐸𝑈𝑅𝑂𝑆𝑆≤ 1 ( 4 )

That means that the binary variable 𝑆𝑆𝐼𝐻𝐸𝑈𝑅𝑂𝑆𝑆,𝑘 has a value of 1 if the 𝐶𝐼𝑆𝑆𝐸𝑈𝑅𝑂𝑆𝑆,𝑘 is larger than

the 90th_{percentile of the dataset of the euro area CISS index and smaller or equal to 1. In all other} cases, the binary variable is 0. The thresholds for the mediocre systemic stress binary variable

𝑆𝑆𝐼𝑀𝐸𝑈𝑅𝑂𝑆𝑆 lie within the following range:

70𝑡ℎ 𝑝𝑒𝑟𝑐𝑒𝑛𝑡𝑖𝑙𝑒 ≤ 𝐶𝐼𝑆𝑆𝐸𝑈𝑅𝑂𝑆𝑆≤ 90𝑡ℎ 𝑝𝑒𝑟𝑐𝑒𝑛𝑡𝑖𝑙𝑒 ( 5 )

That means that the binary variable 𝑆𝑆𝐼𝑀𝐸𝑈𝑅𝑂𝑆𝑆 has a value of 1 if the 𝐶𝐼𝑆𝑆𝑆𝑆𝑆,𝑘 is smaller than or

equal to 90th_{percentile and larger or equal to the 70}th_{percentile of the dataset the of euro area CISS} index. In all other cases, the binary variable is 0. The thresholds for the low systemic stress periods lies within the following range:

0 ≤ 𝐶𝐼𝑆𝑆𝐸𝑈𝑅𝑂𝑆𝑆< 70𝑡ℎ 𝑝𝑒𝑟𝑐𝑒𝑛𝑡𝑖𝑙𝑒 ( 6 )

Note that this effect will be captured by the constant 𝛼𝐸𝑈𝑅𝑂𝑆𝑆,𝑘 in Model 2 for every country “k”.

(34)

34 | P a g e 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 20 03Q2 20 03Q4 20 04Q2 20 04Q4 20 05Q2 20 05Q4 20 06Q2 20 06Q4 20 07Q2 20 07Q4 20 08Q2 20 08Q4 20 09Q2 20 09Q4 20 10Q2 20 10Q4 20 11Q2 20 11Q4 20 12Q2 20 12Q4 20 13Q2 20 13Q4 20 14Q2 20 14Q4 20 15Q2 20 15Q4 20 16Q2 20 16Q4 20 17Q2 20 17Q4

Euro-area Systemic Stress Indicator

EUROSSIM (RELATIVE THRESHOLD) Euro-area

(35)

35 | P a g e

4.2 Mathematical derivation: the relation between bonds and interest rates

This part expounds on the relation between the yield of long-term government bonds, treasury bills and interest rates. The principle behind bond valuation is that the present value of the expected future cash flow of a bond is equal to the value of the bond today. Mathematically this is explained with the following equation:

𝑃𝑏𝑜𝑛𝑑 = ∑ ( 𝐶𝑛 (1+𝑖)𝑛) + 𝑀 (1+𝑖)𝑛 𝑁 𝑛=1 ( 7 )

From this equation, the relationship between bond prices and interest rates can be calculated. Called the duration of a bond, the mathematical derivation of the duration is extended in APPENDIX . The formula for the duration reads:

∆𝑃

𝑃 = −𝐷 ( ∆𝑖

1+𝑖) ( 8 )

This equation clearly shows the negative relationship between prices and interest rates of the bonds. If the prices increase the interest rate decreases. A government in economic distress has more difficulty acquiring funds, as investors require significant compensation for their risk. Therefore, prices of bonds and treasury bills fall and interest rate will rises. Assuming that the efficient market hypothesis holds (EMH), there is no room for arbitrage and interest rates on private debts and deposits for similar maturities and risk will rise. Which in turn affects household deposit behavior.

(36)

36 | P a g e

4.3 Mathematical derivation: The relation between general deficit and government debt

This section expounds on the relationships between government debt , deficit and household deposits. The intertemporal government budget constraint captures the relationship between the general deficit to GDP and the total amount of debt-to-GDP. The following derivation provides a better understanding of the measurement of these variables, and their relationship tothe dependent variable.

𝐷𝑡+1= (1 + 𝑟)(𝐷𝑡+ 𝐺𝑡− 𝑇𝑡) ( 9 )

Rewriting equation (7) and forwarding the substitution provides the following equation:

∑ 1 (1+𝑟)𝑡 ∞ 𝑡=0 (𝑇𝑡− 𝐺𝑡) + lim 𝑠→∞ 𝐷_𝑠 (1+𝑟)𝑠= 𝐷0 (10)

Assuming no Ponzi-game condition:

Ds (1+r)s= 0 (11) Equation (2) becomes: ∑ 1 (1+𝑟)𝑡 ∞ 𝑡=1 (𝑇𝑡− 𝐺𝑡) = 𝐷0 (12)

Equation (12) shows that the present value of future government balances is equal to the initial debt. The amount of total debt provides information on the future tax burden/government spending constraints for the national economy. It is, consequently, a good measure for the health of the national economy and the ability of the government to assist the domestic economy in times of economic distress. Expectations are that countries with high initial debt are more sensitive towards deposit flow exits during economic downturn. The build-up of an oversized sovereign debt (financial imbalance) is therefore likely to raise systemic stress. However, this thesis does not include debt-to-GDP data due to non-stationarity of these time series. The change in debt-to-debt-to-GDP does provide a good measurement, however, for the build-up of sovereign stress in the short run. Debt payoff is likely to ease systemic stress, while an increase in debt is likely to increase systemic stress in the short-term. The capture of the systemic stress over time due to financial imbalances and cross-sectional exposures is done through the binary variables in our model. More information on the measurement of periods of systemic stress is provided in Subsection 4.1 A composite indicator for systemic stress..

Systemic stress, sovereign health and household deposit behavior in the euro area