[Last Name] 1

UNIVERSITY OF ESSEX

DEPARTMENT OF ECONOMICS

Using financial network analysis tools from RBI App; develop

visualizations of interconnectedness, systemic risk analytics for cross border

financial flows, calling upon BIS data & regulatory capital data from

Bankscope.

Supervisor: Professor Sheri Markose - University of Essex

 [Last Name] 2

Written by: John Ajagbe

Registration Number: 1706314

Word Count: 0
 [Last Name] 3

TABLE OF CONTENT

1 Abstract.……………………………………………………......……………....…...0

2 Introduction ……………………………………….………………............................0

3 Literature review……………………………………………….……………………….0

7 Conclusions & Summary ………………...............................................................17

8 References ………………………………………………………………….………0

9 Appendix……………………………………………………………………….…...0

10 Glossary………………………………………………………………………..……0
[Last Name] 4
 [Last Name] 5

Introduction to the paper and your motivation of the paper,

[Last Name] 6
 [Last Name] 7

The 2007 Global Financial Crisis got transformed into the Eurozone financial crisis.

Explain what Basel 2 is and what does it stipulate

Acharaya explain why the Euro zone crisis was the largest carry trade caused by zero risk weights stipulated by Basel

II on Government bonds.
[Last Name] 8
[Last Name] 9
 [Last Name] 10

Describe RBI App and its Role within the Economy

 Use RBI App and create financial network using BIS data for Eurozone countries and

also include USA and UK. Explain the components of the App and how they work.

o Pick only 3 specific, parts of the data e.g 2006 before crises 2008, during crises

and 2012 post crises.

Describe what the RBI App is and its role. (Reword)

Using the RBI App requires two input files The Gross Bilateral Matrix X, It is an (N+1)x(N+1)

matrix (N=number of banks) where the first column and row must be the row headers

and column headers respectively

and the Bank Data, which contains It is a matrix of banks characteristics, (N+1)xM where

M is the number of characteristics the actual version of the app requires. This is the

updated list of variables:

1. 1. Banks (name of the Bank)

1. 2. Total assets
 [Last Name] 11

a. 3. Total borrowing

i. 4. Tier II Capital

1. 5. Tier I Capital

a. 6. RWA

Once the data is imported, you have four different filters for banks:

b. All Banks: all banks will be shown based on the list of banks in the input files

1. Top 20 Banks: Only top 20 banks based on Eigen Vector

Centrality (EVC) will be shown, the other banks will be merge in one

object called “Others”

1. Groups: banks will be grouped following the Group Id and named

by Group Name. Those with group id = -1 will not be grouped and

left as individual banks

a. Custom Selection: custom filtering can be done by picking

the banks to show from the scrollable list below. The banks

that are not picked will be merge in one object called

“Others”

The filters will be applied to both the network and the matrix flows (Gross and Net).

Banks Info tab will not be affected by the choosing filter.

By pressing the button GROSS, the network plot will switch to Gross flows. By

pressing NET the network will switch back to the net flows network.
 [Last Name] 12

You can zoom in and zoom out in the Maps to see the Maps. The two buttons “Zoom In” and “Zoom

Out” allow the user to magnify or minimize the overall area of the network.

We can also use the second scroll pane allows the user to use the mouse to manually

move the network plot.

Transforming: to move the entire network

Picking: to move a single node

[Last Name] 13
 [Last Name] 14

What evidence can we gather from our results of the RBI App

Give evidence for and explain why USA is the most systemically node.

Could the systemic importance/risk of the Euro area periphery countries (Portugal,

Ireland and Spain) have been identified well before 2007 in terms of network properties relating

to their cross border banking liabilities relative to their own country capital base?
 [Last Name] 15

Conduct contagion analysis for the top 3 eigen-vector central countries and explain the

results.

It has been said that in the crisis and post crisis period of 2007-2009 cross border the

maximum row sum of capital adjusted liabilities were quite distinct in size. Compare this with

the stability metric, viz. the maximum eigenvalue of the cross border matrix of financial claims

adjusted for capital.

Case Study: Spain

•Focus only on the matrix of gross financial liabilities and assets to Spain over the period 2005-2013; comment on the

capital adequacy situation in Greece. Greece is found to have negative equity.

 [Last Name] 16

Notes

Lecture Notes

Multi-Agent Global Macro-net Models: New Tools for Macro-Prudential PolicySheri Markose (Economics Dept.

University of Essex) scher@essex.ac.uk

Global Macro-Nets:Cross Border Imbalances and Within Country Sectoral Imbalances : Why Do We

Need This ?

• 2007 financial crisis highlighted need to model interconnectedness of financial firms (Haldane,2009), non-financial private and public

units, both nationally and globally, also transmission channels between financial and real side of economy.

• Equation driven mainstream models only have reduced form feedback methods to characterize global imbalances to do with current

account, credit and capital account, carry trades and offshoring of production.

•Global consequences of large scale QE in US, UK and Euro-zone: Many BRICs and Switzerland, in particular, have countervailing

strategies for currency appreciation from capital flight from advanced countries (see, Volz, 2012, Agenór and Pereira da Silva , 2011). Estimated at

$1.040 -$0.9 Trillion

 [Last Name] 17

International financial integration has led to cross-border banking to expand dramatically. Particularly, the EU crisis,

especially within the Euro area, is highly attributable to the increasing cross-border financial activity in the pre-crisis period (Lane

and Pels, 2012).

Interbank loans between Euro area banks and holdings of debt securities issued by other Euro area banks grew from

15.5% and 12.1% in 1997 to 23.5% and 31.3% in 2008 respectively.

Hence, the scale of the increasing interconnectedness within the banking system in the Euro Area is quite evident. (Allen

et al, 2011).

However, cross-border external financial linkages, such as those with the US and other EU countries like the UK, are also

really important (Lane, 2013).

[Last Name] 18
 [Last Name] 19

o (i) Capacity of SR metrics to provide early warning of potential system failure and the capacity to signal

sudden shift in systemic importance of Fis

 (ii)The Role of regulatory capital and liquidity thresholds and SRIs: we recommend relevant

points at which system distress is evaluated is before insolvency and system stability is determined

at points when capital and/or liquidity thresholds are breached

 (iii)Significance of fixed point mappings for SRIs and systemic importance of FIs: In an

important observation first made by Gauthier et al (2010), SRIs in relation to macro-

prudential capital (and liquidity buffers) should be modelled in a fixed point framework.

Systemic importance depends on other SI nodes in a fixed point; ditto for SRI itself.

o (iv)Analytical ex ante SRI metrics and ex post simulation based capital

losses : Ex post SRI metrics based on the average (simulated) losses of capital

from financial contagion process or algorithm typically like the Furfine (2003)

or the Eisenberg-Noe ones. An ex ante SRI an application of an analytical result

that can give prior conditions of system stability in terms of relevant structural

parameters.
 [Last Name] 20

22 National Banking systems exposures to debtor countries (BIS 2010Q4)

 Stability of Networks and Eigen-Pair Analysis: Markose et. al. (2012)

 3 main questions of macro-prudential regulation :

(i)Is financial system more or less stable?

(ii)Who contributes to Systemic Risk ?

(iii)How to internalize costs of systemic risk of ‘super-spreaders’ using Pigou tax based on eigenvector centrality: Management of moral hazard,

Bail in vs Bail out : How to Stabilize system using EIG Algorithm ?

 Superspreader Lite Escrow Fund

 Generalization to multi-layer networks

● Conclusions

 Markose et. al. (2012a,b) Eigen Pair Analysis : Endogenous to contractual financial

obligations and not on external shocks etc

o Stress Test and Systemic Risk Metrics

 Monitoring Systemic Risk : Is the financial system becoming more or

less stable ?

 Monitor maximum Eigen-value of the ratio of net liabilities

to Tier 1 capital matrix

o Cause for concern if max eigen value is greater

than the fixed threshold/ratio of prefunded

capital : Focus on policy relevant regulatory

variable

 Advantages: Certifiable and transparent

contractual obligations; I do not think a

 [Last Name] 21

FI can be held culpable for pre existing

macro conditions or for unknowable

losses from fire sales.

 [Last Name] 22

Conclusion :Systems are unstable and operationalizing of Eigen-vector Pigou Tax aim to mitigate instability

o Too interconnected to fail addressed only if systemic risk from individual banks can be rectified with a price

or tax reflecting the negative externalities of their connectivity

 Lessons to be learnt : Disease Transmission in scale free networks (May and Lloyd (1998),

Barthelemy et. al : With higher probabilities that a node is connected to highly connected nodes

means disease spread follows a hierarchical order. Knowledge of financial interconnectivity

essential for targetted interventions

 Highly connected nodes become infected first and epidemic dying out fast and often

contained in first two tiers

o Innoculate a few rather than whole population; Strengthen hub; Reduce

variance of node strength in dominant eigenvalue formula

[Last Name] 23
 [Last Name] 24

Using the key parts in Markose et. al (2012) on the eigen-pair method to explain maximum eigen value as systemic

risk index and the use of the right eigenvector for systemic importance and left eigenvector for systemic vulnerability.

Systemic Risk Analytics: A Data Driven Multi-Agent Financial Network (MAFN) Approach

Sheri M. Markose

The 2007 financial crisis has undoubtedly exposed shortcomings of monetary and macro-

economics [2,3]
and of the regulatory framework of Basel II . Macroeconomic models and their
[4-7]

use in policy analysis have come under severe criticism. Critics have accused macroeconomists

of heavy reliance on a particular class of macroeconomic models that has abstracted away

institutional details and financial interconnections in the provision of liquidity, capital adequacy

and solvency [8,9]

. Consequently, the paradigm shift needed and skills gap that has to be fixed
 [Last Name] 25

among academic economists and their regulatory counterparts is quite considerable in order to

keep abreast of the institutional and technological innovations in monetary and financial sectors.

These innovations have created at least three challenges: (i) Unprecedented volumes of ‘inside’

money via securitization and other forms of procyclical collateralized private credit [10,11]
; (ii) A

shrinking of state supplied cash in circulation in low cash based economies with an IT based

payments technology which has changed payments habits and transactions demand for money

irrevocably and also may have vitiated the monetary transmission of inflation in the CPI index [12,13]

; (iii) A vast interconnected system of digital transference of financial liquidity in real time with

very low latency, along with ‘algo’ based hyper high frequency financial markets , often

leveraged by instruments such as contracts for differences.

On the eve of the collapse of Lehman Brothers in September 2008 when the American

Insurance Group (AIG) also stood imperilled due to its inability to make good on collateral calls

for credit default swap (CDS) guarantees on mortgage backed securities of large FIs, a lack of

data and models on the possible knock on effects in the US and globally, forced US Treasury and

Federal Reserve officials to fly blind at the critical juncture. The moral hazard problem inherent

to US, UK and European tax payer bailout of key FIs that ranged from full and partial

nationalization to financial guarantees reached unprecedented amounts of over $14 trillion,

Alessandri and Haldane , has aptly been called ‘too interconnected to fail’. A prominent
[19]

example of this was the US bailout package of $85 bn for AIG which was geared toward

averting substantial losses to its major counterparties.

 [Last Name] 26

Post 2007, systemic risk from financial activities is being viewed as a negative externality

analogous to environmental pollution . In this case, over use and degradation of resources
[20,21]

occurs as clean up costs are not internalized by the polluting economic agents due to a missing

market. Likewise, over supply of leverage and excessive risk taking by financial agents follow

because costs of their failure on others are not borne by them in the spirit of a Pigou tax . The [22]

state of play in both the control of pollution and of financial leverage is that institutions have not

yet been designed or evolved to adequately address the problem of aligning the interests of the

individual actor and system wide stability. To ‘see’ why seemingly rational behaviour at the

level of an individual FI contributes to system wide instability is a non-trivial exercise which

requires holistic visualization and modelling techniques. Haldane [23]

has proposed the use of

financial network models for the analysis of systemic risk from financial contagion that is driven

by interconnected balance sheets. Clearly, the absence of such a quantitative modelling

framework of the financial system has impeded progress in the monitoring and management of

financial systemic risk.

The US Office of Financial Research was set up in 2012 to overcome problems of

balkanization of financial and banking data and to have better models to provide quantitative

oversight of the financial system nationally or globally. Institutions like the European Central

Bank, International Monetary Fund and newly set up financial stability divisions in different

countries with the Financial Stability Board (FSB) for international coordination have intensified

efforts to develop modelling tools such as financial network analysis, ‘big’ financial data

facilities , which require Information and Communication Technology (ICT), and

[24]

corresponding systemic risk analytics . [25]

 [Last Name] 27

Steps are now underway to design ‘bail in’ arrangements at the time of failure of FIs as

part of resolution procedures [26,27] 6

and those that are paid for by FIs before failure to alleviate

unacceptable socialization of losses from them. The focus of this paper is on the second of these.

The FSB has identified a list of global SIFIs and seek to impose capital surcharges on a sliding

scale from 1% to 2.5% depending on factors such as large size, prominence in markets or

functions (non-substitutability), complexity, global activity and interconnectedness. Will such

generic surcharges help mitigate specific excesses such as the activities of SIFIs in derivatives

markets? To answer this and other related questions, it is important to develop a framework to

quantify the surcharges on FIs and to see whether they can mitigate the moral hazard problem

entailed in tax payer bailouts.

3. Digital Mapping of the Financial System From Databases

3.1 A Multi-Agent Financial Network (MAFN) Model

Recently, many have emphasized the uses of agent based computational economics

(ACE) simulation platforms for digital mapping and monitoring of the financial system, stress

testing and for institutional design (see [52-55, ]

). These artificial environments can depict real time

orientation, institutional rules, and also complex interactions. For the simulation framework to

be useful for the assessment of policy, financial firm level responses must be modelled in the

context of prevalent market and regulatory conditions and with automated access to balance

sheet and off- balance sheet data of FIs to anchor their financial decisions.
 [Last Name] 28

From the vantage of 21 century ICT based tools, a non-economist may be forgiven for
st

painting the following picture of how regulators manage systemic risk problems in the financial

system. Mark Buchanan [52]

in a paper in Nature gives an account of what advanced IT based tools

can deliver : “A screen on the wall maps the world's largest financial players — banks,

governments and hedge funds — as well as the web of loans … and other legal claims that link

them. High-powered computers have been using these enormous volumes of data to run through

scenarios that flush out unexpected risks. And this morning they have triggered an alarm....

Flashing orange alerts on the screen show that a cluster of US-based hedge funds has

unknowingly taken large ownership positions in similar assets. If one of the funds should have to

sell assets to raise cash, the computers warn, its action could drive down the assets' value and

force others to start selling their own holdings in a self-amplifying downward spiral. Many of

the funds could be bankrupt within 30 minutes, creating a threat to the entire financial system.

Armed with this information, financial authorities step in to orchestrate a controlled elimination

of the dangerous tangle.” Needless to say, such web based visualization of financial data and

real time operations relating to financial crisis management is far from being implemented. The

fundamental computational methodology for web based visualization of complex data sets is

object oriented programming (OOP) and multi-agent modelling. The technological ICT aids of

the ‘zoom’ that can navigate between the coarse grained bird’s eye view and the fine grained

ones can mitigate the well known befuddling aspects of not being able to see ‘the woods for the

trees’. The ‘probe’ can automate and highlight behind the scenes hidden links of each FI in

multiple markets. Unfortunately, such enabling technologies of advanced ICT economies have

yet to be harnessed for economic analysis and systemic risk monitoring.

 [Last Name] 29

Agent based computational economics or ACE using the acronym coined by Leigh

Tesfatsion [56,57]
is based on object oriented programming that can produce agents that are both

inanimate (eg. repositories of data bases) as well as behavioural agents capable of varying

degrees of computational intelligence. These range from fixed rules to fully adaptive agents

representing real world entities (such as banks, consumers and regulators) in artificial computer

environments which can be replicas of, for instance, the financial system. Unlike conventional

programming in which a program entails a lists of tasks or subroutines, in ACE and OOP, each

agent which is an instance of a class is capable of interacting with other agents by receiving and

sending ‘messages’, processing data, and producing outputs on the basis of their computational

intelligence. The outputs can be accessed by the experimenter and the agents themselves using

‘probes’. There is considerable literature on ACE models that have produced qualitative insights

into classic non-deducible self-organized outcomes. These range from the Schelling [58]
model on

racial segregation to the Santa-Fe Institute stock market model of Arthur et. al. [59]
which showed

the endogenous generation of boom bust cycles on account of the contrarian payoff structures in

stock markets. Thurner et. al. [60]

give a recent example of how an agent based simulation model

can show how leverage exacerbates boom bust cycles in an artificial stock market model.

However, the empirical data resolution end of these ACE models is low for macro-prudential

regulatory purposes.

In financial networks, nodes stand for financial agents such as banks, non-bank

intermediaries, the final end users and central banks. The edges or connective links represent

directed inflows (in degrees) of liquidity or receivables, and outflows (out degrees) represent

obligations to make payments. By data base driven MAFNs is meant that disaggregated data at
 [Last Name] 30

the level of individual FIs with regard to bilateral flows to each of their counterparties will have

to be accessed electronically to provide ‘as is’ quantitative characteristics.

Modelling of FIs and their Interconnections For Purposes of Macro-Prudential

Policy

Figure 1 gives a stylized graph for the scope, discussed in section 2, of a MAFN model

which incorporates the classes of financial agents (depository and non-depository FIs, items II,

III, and IV), types of financial products/markets (RMBS, Repo, Derivatives, Sovereign Bonds,

Equities, items I, VI, VII, VIII) and the complex interconnections between them (the arrows and

item XI). The financial interconnections at a bilateral level for all depository and non-depository

FIs can be embedded into a more aggregative framework for the sectoral uses and sources of

funds involving the household sector, government, non-financial corporate sector, monetary and

financial sector and global flows. The item (V) in Figure 1 on Peer-Peer lending indicates the

need to include new financial sectors as they evolve.

In principle, each FI is a vector of financial activities operating in a multi-layer system of

markets for different financial products, each of which has its own network topology,

institutional incentives and constraints. This is illustrated in Figure 2. In the multi-layer

networks in Figure 2 (RHS), the broken vertical lines show the FIs that are common to the
 [Last Name] 31

different networks for financial products and hence they can become the conduit by which an

exogenous shock in one market can be propagated across other markets. However, to date, most

financial networks are modelled as single layer ones, either of a single market or one which

represents an aggregation of several products. Full scale developments of MAFN models as

hyper-networks [61,62]
are only in their infancy. Nevertheless, for purposes of regulatory monitoring,

as will be seen, even quarter by quarter snap shots of network visualizations and analytics of

bilateral financial flows data whether for specific financial products or as a single layer network

for flows aggregated over products, are useful to gauge instability of the system and of the

centrality or systemic importance of FIs in it.

 [Last Name] 32

Single Layer Network (LHS) and Multi-Layer Networks (RHS)

Integration and automation of financial data bases in a MAFN framework, therefore, aims

to transform the data from a document or record view of the world to an object-centric view [63]

where multiple facts about the same real-world financial entity are accessed to give a composite

visualization of their interactions with other such entities in a scalable way. Without powerful

integrative tools for system wide visualization of firm level data pertaining to all sectors of the

financial system, in an increasingly complex environment where size of nodes or parts of

networks alter and new subnets form as new financial instruments come on stream, it will be

hard to ‘see’ or quantify systemic risk impacts of units such as key broker-dealers, a sector such

as a centralized clearing platform or a market for unfunded claims such as credit derivatives. As

in the Buchanan’s excerpt above, orange alerts can be assigned to threat factors such as

overleveraged positions and the pro-cyclicality of underlying assets to the same macro-variable

that include house prices or debt of a specific sovereign.

Till recently, the IBM MIDAS project [63,64]

and the EC grant FP6 -034270-2 project of

Markose and Giansante currently being implemented at the Reserve Bank of India (see
 [Last Name] 33

ACEfinmod.com) are the only known software technologies being developed for large scale firm

level financial database driven models for systemic risk analysis. Since mid 2010, the Financial

Stability Unit of the Reserve Bank of India has started mandating all depository institutions

(including foreign banks operating in India) and a large majority of the non-depository FIs to

submit bilateral financial data for non-electronically cleared products in the funded and unfunded

derivatives markets. Central Banks of Brazil and Mexico are also mandating bilateral financial

data from their FIs.

The EURACE project aims to develop a methodology for large scale data base driven

multi-agent macro-economic models for the Euro zone . However, though some agent based
[65]

exercises have been conducted with simulated data, the EURACE project has not produced any

implementations of large scale data base driven agent based macro-economics models. The

RAMSI (Risk Assessment Model for Systemic Risk Institutions) model of the Bank of England

is based on the balance sheet data of the 10 core UK banks for 650 balance sheet entries. In the

absence of bilateral data, the balance sheet bilateral interconnections are modelled using the

Entropy maximization method, which, as will be discussed, is known to introduce model risk.

Aikman et. al [66]

stress test the model for asset and liability side shocks which precipitate non-

linear feedbacks like those that arise from deleveraging and fire sales. Calibrations are used to

determine loss of capital from fire sales. The counterparty failures add to losses via the matrix of

interbank exposures for the top UK banks. Banks are each assigned points on the basis of

structural imbalances such as reliance on short term money markets, maturity mismatch etc and

those that score points in excess of 35 are judged to be in the danger zone. The absence of off

balance sheet items and non-bank FIs, clearly make the RAMSI model less than comprehensive

for purposes of systemic risk monitoring. Further, the lack of a publicly available electronic data
 [Last Name] 34

warehouse for quarterly data on the financial statements of depository institutions operating in

the UK, let alone those for non-depository FIs, implies that there is some way to go before a

MAFN model with the scope of Figure 1 can be developed for the UK even using empirical

calibrations for reconstructing financial interconnections.

Finally, how does ACE compare with estimation based macro-econometric models for

policy analysis? In ACE, interactions of agents produce system wide dynamics that are not

restricted to pre-specified equations which have to be estimated using past data in econometric or

time series approaches. In ACE, each agent follows explicit rules or evolves strategies under

specified market conditions and a ‘probe’ monitors causal internal workings and also aggregates

outcomes. In contrast, the main draw-back of estimation based equation analyses is their

susceptibility to the Lucas Critique in that structure changes from strategic behaviour and tracing

of causal links are almost impossible to do. The idea that nodes in the network which constitute

FIs and other financial actors are themselves intelligent ‘agents’ autonomously evolving

strategies while operating within constraints and incentives provided by the markets and

regulations has not been fully operationalized yet for purposes of policy design. In response to

the riposte by many [6,69,70]

that a great source of systemic risk is perverse incentives from policy,

viz, policy fails precisely when FIs comply with it, Markose et. al. [41]
give an exemplar on how

this can be monitored on an ongoing basis using MAFNs. All FIs from the US FDIC Call

Reports data set from 2003 onwards that were involved in credit default swaps were made to

comply, by programming this in the ACE model, with the Basel II incentives in synthetic

securitization to reduce capital from 8% to 1.6% by keeping RMBS assets on balance sheets and

acquiring credit default swap guarantees. The build up of what became toxic RMBS assets on
 [Last Name] 35

major US banks’ balance sheets along with CDS purchases which were shown to belong to a

topological unstable network structure, followed as if according to a gory script !

In agent based models, rule following behaviour as in complying with the regulation and

the conduct of carry trade activity are relatively easy to implement. This is because unlike fully

fledged adaptive behaviour, agents’ strategies, intelligence and autonomy are limited to

following the letter of the law and strictly verifying conditions for which the most profitable

arbitrage applies. The modeller, however, faces the challenge of understanding the regulation,

provide market conditions for the triggers that need to be followed in a carry trade and then

implement the agents’ strategies in an algorithm. Markose et. al. [41]

argue that as stress tests for

perverse incentives of policy is among the easiest of MAFN exercises, it must be de rigueur in

macro-prudential policy in order that flawed policies do not get perpetuated.

Financial Network Topology and Propagation of Contagion : Network Topology

Matters

The reason why it is important to map the actual interconnections between FIs is because

network topology is a major determinant in how contagion propagates and the system fails.

Interventions and stabilization crucially depend on knowing who is linked to whom. In the

absence of actual bilateral financial data, the bulk of the pre 2007 simulated financial network

models assumed that they were random graphs or used the entropy maximization method for

network formation which aims to maximize the homogeneity of financial flows between a FI and

its counterparties. Many have discussed [72,73]

why networks produced by the entropy method or as

random graphs are not suited to characterize real world financial networks. Some important

aspects of these discussions will be illustrated here and also in Section 4.

 [Last Name] 36

The following Figure 3 shows a random graph (left) with no specific structure or

clustering and a highly tiered core-periphery graph (right) that characterizes OTC derivatives and

to a lesser extent interbank credit markets. FIs in the periphery have no links between

themselves while those in the core have dense interconnections amongst themselves . [1,74]

Graphical representation: Random network (left), Tiered network with core-

periphery structure (right)

Markose et. al [7, 40]

show how failure of a node, the one placed in the centre of Figure 4,

propagates contagion in a random network structure (right) and that in a core-periphery sparse

network (left). The latter depicts what it means to be too interconnected to fail, typical of the

CDS [32,40,41]
and other derivatives markets . The highly tiered network has a central core of large
[1]

banks which directly take a hit when a similarly connected bank collapses. The contagion stops

at this point as the network loses connectivity with the demise of the super-spreaders. But in the

spirit of being too interconnected to fail, 4 top banks are brought down, Figure 4 (left). It is of

course cold comfort that there are no second order failures spreading to the whole system when

the first order shock wipes out the top 4 banks and some 70% of Tier 1 capital of the system. In

contrast, the random network with no tiered structure and no bank is too interconnected, suffers

as many as 17 (out of the 26) bank failures in a series of cascades which cannot be predicted,

Figure 4 (right). Thus, in the context of controlling epidemics, the clustered network allows

easier solutions in terms of inoculating the few super-spreaders, while in the random network the

whole population has to be inoculated. Haldane [23]

calls such hub banks ‘super-spreaders’ and he

recommends that super-spreaders should have larger buffers. The vulnerability of the tiered

network to failures of any member of the core, as stated by Haldane , requires that steps should
[23]

be taken to reverse the current practice of more lenient reserves and collateral requirements for
 [Last Name] 37

large broker dealers than their less connected counterparts. There is as yet no consensus as to

how to operationalize the ‘bail in’ arrangements paid for by the FIs themselves for their systemic

risk impact on others for being too interconnected.

Figure 4: Instability propagation in Clustered Empirical CDS Network (left) and in Equivalent Random Network

(right) NB: Black nodes denote failed banks with successive concentric circles denoting the q-steps of

4.3 Markose [1]

Eigen- Pair Method For Systemic Risk Analytics

For Macro-prudential Policy we need to assess 3 questions:

I. Is the financial system becoming unstable

a. Who is causing systemic risk and who is vulnerable

i. How do you stabilize system and penalize those who causing systemic risk

In order to simultaneously determine the stability of the financial network system and the contribution of each FI to

this, it is useful to model the dynamics of failures as in the epidemiology literature, viz. as the spread of ‘disease’ from other

failed FIs. In the classic Furfine [30] contagion stress test, any arbitrarily selected trigger bank is monitored for the direct and

indirect failures it brings about on counterparties via their exposures to the trigger bank, relative to their own capital.
 [Last Name] 38

The causal direction of the contagion and hence systemic risk of a FI, follows from the

‘trigger’ FI, i, owing its counterparty j more than what j owes i, relative to j’s Tier1 capital. This

is denoted by the positive entries net liabilities to i to j relative to j’s capital given by elements θ ij

= (x - x ) / C in matrix (1) for those pairs of FIs which have a direct financial links. Here, C is
ij ji
+
j0 j0

j’s initial capital. Hence, the matrix Θ that is crucial for the contagion analysis will have

elements given as follows:

Failure of a FI is usually determined by the criteria that losses exceed a predetermined buffer ratio, ρ, of Tier 1

capital. In the epidemiology literature , ρ is the common cure rate and (1 - ρ) is the rate of not surviving in the worst case
[90]

scenario. The dynamics relating to the probability of failure of each ith FI at a given time step q+1 denoted by u iq+1, given j

counterparties of i have failed at the previous time step q. As shown in the Appendix, this is determined by (i) i’s own survival

probability given by the capital C iq it has remaining relative to initial capital C i0 (ii) and the sum of ‘infection rates’ defined by

the sum of net liabilities of its j failed counterparties relative to its own capital is given by the term . Note the

‘infection rate’ or how counterparties impact on an FI is pair wise heterogeneous.

It can be shown in Markose and as outlined in the Appendix, the stability of the network
[1]

system involving matrix Θ in (1) requires that its maximum eigenvalue is less than the

homogeneous Tier 1 capital threshold, ρ:

λmax(Θ´) < ρ . (2)

If this condition is violated, a negative shock, in the absence of outside interventions, can

propagate through the networked system as a whole and cause system failure.
 [Last Name] 39

The dynamics characterizing transmission of ‘infection’ in a financial networked system can be given by

uiq+1 = (1 - ρ) uiq + (A.1)

Here, we have a FI’s own metric of failure at q which is given by u iq= (1- Ciq/Ci0 ), where Ciq/Ci0 is the ratio of capital at

q and capital at initial date. The second term in (A.1) involves the losses from counterparties,j, that fail at q and these are

denoted by the an indicator function which is set equal to 1. The sum of ‘infection rates’ is defined by the sum of net liabilities

of its j failed counterparties relative to its own capital is given by the term .

In order for the eigen-pair stability analysis to be used, in matrix notation the dynamics of financial contagion takes

the following form:

Uq +1 = [(1 - ρ)I + Θ´]Uq = QUq . (A.2)

Here, Θ ´ is the transpose of the matrix in (1) with each element θij´= and I is the identity matrix.

The system stability of (A2) will be evaluated on the basis of the power iteration of the matrix Q. From (A2), Uq takes

the form:

U q= Q q U 0 . (A.3)

Markose (2012) shows how the stability of the system in (A.3) as q tends to infinity, requires that the maximum eigen-

value, λmax, is less than the common threshold on capital, ρ.

Eigen-vector Centralities: Systemic Importance and Systemic Vulnerable

 [Last Name] 40

In the management of contagion and the design of the super-spreader tax or inoculation

measures, it is important to understand the relationship between the maximum eigenvalue of the

Θ matrix and its corresponding, respectively, right and left eigenvectors and v . Note, a
1

matrix (with non-negative values) and its transpose have the same maximum eigenvalue.

However, right and left eigenvectors corresponding to the λ , are different. Here, the right
max

eigenvector denoted by corresponds to the systemic risk metric for FIs, relating to the

damage a FI can inflict on others as the direction of impact in elements θ in matrix (1) arise from
ij

the liabilities of i to j. In contrast, the left eigenvector denoted by v relates to the exposure index
1

of FIs, viz. the damage a FI faces from others from elements θ of the matrix transpose Θ´ of (1).
ji

The power iteration algorithm that determines the maximum eigenvalue of the matrix Θ, also

determines its corresponding eigen-vector, , which yields the rank order of the

centrality of the FIs vis-à-vis the instability metric of the system given by λ . What is important
max

to note, as discussed in Markose (2012), is how the power iteration algorithm yields a simple

relationship between the upper bound of λ and the maximum row (column) sum of the matrix
max

Θ´ (Θ in (1)). Further, high eigen-vector central, EVC, nodes score highly because their

connections to high-scoring nodes contribute more to the EVC score of the node in question than

equal connections to low-scoring nodes. Denoting as the right eigenvector centrality

for the i node for matrix Θ, the centrality score is proportional to the sum of the centrality scores
th

of all nodes to which it is connected. Hence,

 [Last Name] 41

= (3)

Using vector notation, the eigenvalue equation for the matrix in (1) for the eigen-pair ( λ , max

) is given as:

Θ = λ max . (4)

Thus, high EVC FIs with high connectivity to a large number of highly connected counterparties can contribute greatly

to the instability of the system when λmax > ρ. Note, using the eigenvalue equation, the left eigenvector is defined as

v1Θ = Θ´v1 = λ max v1 . (5)

In Θ´v as the matrix transpose of (1) is involved with elements θ denoting exposure of i
1 ji

to j (rather than the case of the impact of i’s liabilities on j in θ in (4)), the rank order of the left ij

eigen-vector v gives the measure of those that are most vulnerable or exposed to others in the
1

system.

In what follows, the role of row sums in the stability of the system Θ´ will be exploited to determine the Pigou tax for

a FI as a function of its eigen-vector centrality, denoted by . Given the relative simplicity in the determination of

the above systemic risk metrics for the matrix in (1) which yields the appropriate dynamical system for the demise of FIs from

failing counterparties, the eigen-pair method was used on the bilateral financial data for a large Asian interbank market for

quarters from mid 2010 to end of 2011. Remarkably, a situation reminiscent of the aggressive borrowing on the interbank, short

term money markets done by UK banks that demised in the 2007 crisis was observed. From mid-2011, a bank that was ranked

number 5 or 6 in terms of eigenvector centrality in mid 2010 was seen to have catapulted to the most eigen-vector central bank

within a few quarters. A combination of increased connectivity of the FI and its large liabilities relative to the distribution of

capital in the system accounts for its dominant eigen-vector central position. Clearly, what is rational/profitable for this bank

that enabled it to increase its loan market share led to an adverse loss of stability for the interbank system as a whole. System

wide capital losses from a Furfine [30]

type stress test, with this bank as the trigger, jumped to 29.4% from more modest levels of

6%-14% in previous quarters when other banks were dominant in terms of eigen-vector centrality. This real world exercise shows

that it is not sensible to have a priori lists for SIFIs in macro-prudential policy and that sudden jumps in eigen-vector centrality of

a bank should give cause for concern.

 [Last Name] 42

4.4 A Progressive Pigou tax for Eigen-vector Central FIs

Each FI is taxed according to its right EVC in order for the FI to internalize the costs that they inflict on

others by their failure and to mitigate their contribution to network instability as given by λmax. The progressive nature of the tax

justifies the moniker ‘super-spreader’ tax. The rationale behind the application of the right eigenvector centrality of a node as

the basis of the Pigou tax is to enable a FI to provide a buffer proportional to its own capacity to propagate contagion.

The network stabilization algorithm has been called the EIG algorithm, Markose [1], in keeping with Giakkoupis et. al [91]
.

The main objective of the EIG algorithm is to apply to each i a tax denoted by in order to bring down the maximum

eigenvalue of the Θ matrix to the desired threshold ρ. The details of this can be found in Markose [1] and Markose et. al .
[40]

The nature of the systemic risk stabilization super-spreader fund is that it operates like an escrow fund. The funds

commensurate to each FI’s surcharge as a proportion of its Tier 1 capital are collected at some initial point whether within a

centralized clearing platform (CCP) context or by the regulatory authority. The funds are deployed at the time of potential

failure of a FI for the collective good to mitigate tax payer bailouts of the failing bank in order to prevent a financial contagion.

Some backtesting of this has been done in Markose [1]

and Markose et. al [40]
. Further, the right EVC of a FI as a metric for its

systemic importance has been empirically validated as a good proxy of the actual losses of capital that it can bring about as a

trigger in a Furfine contagion stress test for the Indian financial system. The correlation in the rank order of the EVC of FIs and

that for the capital losses they bring about as a trigger in the Furfine stress test was over 98% in all the 4 quarters for which it

was analysed.

Many of the financial network models for systemic risk modelling [92,73]
use the network framework to study the impact of

one or more of the factors that are known to accompany and exacerbate a financial crisis. These factors are strictly extraneous

to the weighted network of financial flows that represent contractual obligations. These factors include : (i) The impact of

common macro-economic shocks such as a rise in interest rates or a fall in house prices that depress balance sheets, (ii) Second

order effects from fire sales and deleveraging [21]

. (iii) Probabilistic considerations of future losses arising from the course of a

contagion. These are interesting stress tests that can inform regulators of the extent of losses under different scenarios.

However, going by the spirit of financial market laws, a FI cannot be held culpable for damage to others from pre-existing

macro-economic conditions such as loose monetary conditions or future market conditions that may arise, for example, during

deleveraging that are unknowable at the time of contracting. The eigen-pair method has the advantage that it is based only on

extant bilateral contractual financial obligations of FIs and their Tier 1 capital and the network topology that is implied by the

certified bilateral data submissions.

The mathematics of networks

 [Last Name] 43

M. E. J. Newman

In much of economic theory it is assumed that economic agents interact, directly or indirectly, with all others, or

at least that they have the opportunity to do so in order to achieve a desired outcome for themselves. In reality, as

common sense tells us, things are quite different. Traders in a market have preferred trading partners, perhaps because of

an established history of trust, or simply for convenience. Buyers and sellers have preferred suppliers and customers.

Consumers have preferred brands and outlets. And most individuals limit their interactions, economic or otherwise, to a

select circle of partners or acquaintances. In many cases partners are chosen not on economic grounds but for social

reasons: individuals tend overwhelmingly to deal with others who revolve in the same circles as they do, socially,

intellectually or culturally.

The patterns of connections between agents form a social network (Fig. 1) and it is intuitively clear that the

structure of such networks must affect the pattern of economic transactions, not to mention essentially every other type of

social interaction amongst human beings. Any theory of interaction that ignores these networks is necessarily incomplete,

and may in fact be missing some important and crucial phenomena. In the last few decades, therefore, a researchers have

conducted extensive investigations of networks in economics, mathematics, sociology and a number of other fields, in an

effort to understand and explain network effects.

The study of social (and other) networks has three primary components. First, empirical studies of networks probe

network structure using a variety of techniques such as interviews, question-

1 Figure 1: An example of a social network, in this case of collaborative links. The nodes (squares) represent

people and the edges (lines) social ties between them.

 [Last Name] 44

naires, direct observation of individuals, use of archival records, and specialist tools like “snowball sampling” and

“ego-centred” studies. The goal of such studies is to create a picture of the connections between individuals, of the type

shown in Fig. 1. Since there are many different kinds of possible connections between people—business relationships,

personal relationships, and so forth—studies must be designed appropriately to measure the particular connections of

interest to the experimenter.

Second, once one has empirical data on a network, one can answer questions about the community the network

represents using mathematical or statistical analyses. This is the domain of classical social network analysis, which focuses

on issues such as: Who are the most central mem-

2 bers of a network and who are the most peripheral? Which people have most influence over others? Does the

community break down into smaller groups and if so what are they? Which connections are most crucial to the functioning

of a group?

And third, building on the insights obtained from observational data and its quantitative analy-sis, one can create

models, such as mathematical models or computer models, of processes taking place in networked systems—the interactions

of traders, for example, or the diffusion of information or innovations through a community. Modelling work of this type

allows us to make predictions about the behaviour of a community as a function of the parameters affecting the system.

This article reviews the mathematical techniques involved in the second and third of these three components: the

quantitative analysis of network data and the mathematical modelling of networked systems. Necessarily this review is

short. Much more substantial coverage can be found in the many books and review articles in the field [1–8].

Let us begin with some simple definitions. A network—also called a graph in the mathematics literature—is made

up of points, usually called nodes or vertices, and lines connecting them, usually called edges. Mathematically, a network
 [Last Name] 45

can be represented by a matrix called the adjacency matrix A, which in the simplest case is an n×n symmetric matrix,

where n is the number of vertices in the network. The adjacency matrix has elements    1 if there is an edge between

vertices i and j, A ij = (1)   0 otherwise.

The matrix is symmetric since if there is an edge between i and j then clearly there is also an edge between j and

i. Thus A ij = A ji .

In some networks the edges are weighted, meaning that some edges represent stronger connections than others,

in which case the nonzero elements of the adjacency matrix can be generalized to values other than unity to represent

stronger and weaker connections. Another variant is the directed network, in which edges point in a particular direction

between two vertices. For instance,

3 in a network of cash sales between buyers and sellers the directions of edges might represent the direction of

the flow of goods (or conversely of money) between individuals. Directed networks can be represented by an asymmetric

adjacency matrix in which A ij = 1 implies the existence (conven-tionally) of an edge pointing from j to i (note the

direction), which will in general be independent of the existence of an edge from i to j.

Networks may also have multiedges (repeated edges between the same pair of vertices), self-edges (edges

connecting a vertex to itself), hyperedges (edges that connect more than two vertices together) and many other features.

We here concentrate however primarily on the simplest networks having undirected, unweighted single edges between

pairs of vertices.

Turning to the analysis of network data, we start by looking at centrality measures, which are some of the most

fundamental and frequently used measures of network structure. Centrality measures address the question, “Who is the

most important or central person in this network?” There are many answers to this question, depending on what we mean
 [Last Name] 46

by important. Perhaps the simplest of centrality measures is degree centrality, also called simply degree. The degree of a

vertex in a network is the number of edges attached to it. In mathematical terms, the degree k i of a vertex i is n k i = ∑ A ij

. (2)

j=1

Though simple, degree is often a highly effective measure of the influence or importance of a node: in many social

settings people with more connections tend to have more power.

A more sophisticated version of the same idea is the so-called eigenvector centrality. Where degree centrality

gives a simple count of the number of connections a vertex has, eigenvector centrality acknowledges that not all

connections are equal. In general, connections to people who are themselves influential will lend a person more influence

than connections to less influential people. If we denote the centrality of vertex i by x i , then we can allow for this effect

by making x i

4 proportional to the average of the centralities of i’s network neighbours:

n 1 x i = ∑ A ij x j , λ j=1 where λ is a constant. Defining the vector of centralities x = (x 1 ,x 2 ,...), we can

rewrite this equation in matrix form as λx = A·x, (4)

(3)
 [Last Name] 47

and hence we see that x is an eigenvector of the adjacency matrix with eigenvalue λ. Assuming that we wish the

centralities to be non-negative, it can be shown (using the Perron–Frobenius theorem) that λ must be the largest eigenvalue

of the adjacency matrix and x the corresponding eigenvector.

The eigenvector centrality defined in this way accords each vertex a centrality that depends both on the number

and the quality of its connections: having a large number of connections still counts for something, but a vertex with a

smaller number of high-quality contacts may outrank one with a larger number of mediocre contacts. Eigenvector centrality

turns out to be a revealing measure in many situations. For example, a variant of eigenvector centrality is employed by the

well-known Web search engine Google to rank Web pages, and works well in that context.

Two other useful centrality measures are closeness centrality and betweenness centrality. Both are based upon on

the concept of network paths. A path in a network is a sequence of vertices traversed by following edges from one to

another across the network. A geodesic path is the shortest path, in terms of number of edges traversed, between a

specified pair of vertices. (Geodesic paths need not be unique; there is no reason why there should not be two paths that

tie for the title of shortest.) The closeness centrality of vertex i is the mean geodesic distance (i.e., the mean length of a

geodesic path) from vertex i to every other vertex. Closeness centrality is lower for vertices that are more central in the

sense of having a shorter network distance on average to other vertices. (Some writers define closeness centrality to be the

reciprocal of the average so that higher numbers indicate greater centrality. Also, some vertices may not be reachable from

vertex i—two vertices

5 can lie in separate “components” of a network, with no connection between the components at all. In this case

closeness as above is not well defined. The usual solution to this problem is simply to define closeness to be the average

geodesic distance to all reachable vertices, excluding those to which no path exists.)
 [Last Name] 48

The betweenness centrality of vertex i is the fraction of geodesic paths between other vertices that i falls on.

That is, we find the shortest path (or paths) between every pair of vertices, and ask on what fraction of those paths vertex i

lies. Betweenness is a crude measure of the control i exerts over the flow of information (or any other commodity) between

others. If we imagine information flowing between individuals in the network and always taking the shortest possible path,

then betweenness centrality measures the fraction of that information that will flow through i on its way to wherever it is

going. In many social contexts a vertex with high betweenness will exert substantial influence by virtue not of being in the

middle of the network (although it may be) but of lying “between” other vertices in this way. It is in most cases only an

approximation to assume that information flows along geodesic paths; normally it will not, and variations of betweenness

centrality such as “flow betweenness” and “random walk betweenness” have been proposed to allow for this. In many

practical cases however, the simple (geodesic path) betweenness centrality gives quite informative answers.

The study of shortest paths on networks also leads to another interesting network concept, the small-world effect.

It is found that in most networks the mean geodesic distance between vertex pairs is small compared to the size of the

network as a whole. In a famous experiment conducted in the 1960s, the psychologist Stanley Milgram asked participants to

get a message to a specified target person elsewhere in the country by passing it from one acquaintance to another,

stepwise through the population. Milgram’s remarkable finding that the typical message passed though just six people on its

journey between (roughly) randomly chosen initial and final individuals has been immortalized in popular culture in the

phrase “six degrees of separation”, which was the title of

6 a 1990 Broadway play by John Guare in which one of the characters discusses the small-world effect. Since

Milgram’s experiment, the small-world effect has been confirmed experimentally in many other networks, both social and

nonsocial.

Other network properties that have attracted the attention of researchers in recent years include network

transitivity or clustering (the tendency for triangles of connections to appear frequently in networks—in common parlance,

“the friend of my friend is also my friend”), vertex similarity (the extent to which two given vertices do or do not occupy

similar positions in the network), communities or groups within networks and methods for their detection, and crucially,

the distribution of vertex degrees, a topic discussed in more detail below.

 [Last Name] 49

Turning to models of networks and of the behaviour of networked systems, perhaps the simplest useful model of a

network (and one of the oldest) is the Bernoulli random graph, often called just the random graph for short [9–11]. In this

model one takes a certain number of vertices n and creates edges between them with independent probability p for each

vertex pair. When p is small there are only a few edges in the network, and most vertices exist in isolation or in small

groups of connected vertices. Conversely, for large p almost every possible edge is present between the
 [Last Name] 50

EUROPEAN ECONOMY

Economic Papers 497 | April 2013

Capital Flows in the Euro Area

Philip R. Lane

Gross Capital Flows

It is important to appreciate that the implications of any particular level of aggregate net flows and net positions

crucially depend on the composition of the underlying gross flows and gross positions. In particular, the mix of debt and

equity in foreign assets and foreign liabilities matters, as does the maturity structure and currency composition of debt and

the sectoral identities of participants in cross-border financial trade (banks, governments, non-financial corporates, and

households).
 [Last Name] 51

More generally, gross asset trade affects the macroeconomic and financial equilibrium of all participating

countries, even those with zero net imbalances. Indeed, since the scale of gross flows far exceeds net flows, understanding

the full matrix of capital inflows and capital outflows and the level and composition of the international balance sheet is

essential for monitoring and surveillance purposes.

In principle, high gross levels of capital outflows and capital inflows can be stabilising by supporting international

risk diversification. State-contingent foreign liabilities allow domestic economic risks to be shared with foreign investors,

while holding foreign assets can provide some insulation for domestic investors. In addition, high gross flows may improve

the efficiency of financial intermediation by supporting the growth of international financial centres (to the extent that

agglomeration externalities and scale economies are important).

However, gross flows can also raise macroeconomic and financial risks. For instance, a domestic credit boom may

be amplified by cross-border debt inflows into the domestic banking system, allowing an expansion in domestic lending

(Borio et al 2011, Bruno and Shin 2012, Lane and McQuade 2012). Moreover, domestic financial risks can be amplified even

if capital inflows are fully recycled into capital outflows. For instance, the funds that Icelandic banks

3 borrowed overseas were largely used to fund foreign acquisitions by Icelandic entrepreneurs, while the Irish

banking crisis was deepened by the external financial activities of Irish speculators that were aggressive investors in foreign

property markets as well as in the local property market.

Issuing foreign liabilities to fund foreign claims may also fuel the international expansion of banks. While this

could provide risk diversification, it may also facilitate poorly-managed banks to take on excessive risks in particular sectors

(global real estate) or enter new activities in which it does not have a comparative advantage (eg US subprime,

poorlyunderstood local lending markets) . In turn, foreign loan losses may threaten domestic financial stability and the
 [Last Name] 52

scaling up of bank balance sheets through internationalisation may contribute to “too big to fail” problems (Broadbent

2012).

Finally, some types of gross flows may just be motivated by tax and regulatory arbitrage, especially in relation to

“round-tripping” arrangements. A byproduct of such flows is that it adds to financial complexity, making it difficult to

identify and track the distribution of risk exposures across countries.

2.3 Capital Flows and Monetary Union

The analysis of international capital flows takes on special resonance in relation to the euro area. Large external

imbalances of individual member countries pose special adjustment challenges, since the elimination of national currencies

means that real exchange rate adjustment is in part dependent on the external evolution of the euro and in part on

differential price and wage dynamics inside the euro area. In relation to the former, surplus and deficit countries within the

euro area will have conflicting views on the appropriate direction for the external value of the euro. Moreover, even if the

euro area were running a collective imbalance, the volatility of currency markets means that the euro cannot be relied

upon to move in a helpful direction over any near-term time scale.

Moreover, in the presence of nominal and real rigidities, engineering bilateral real depreciations inside a

monetary union is especially problematic. Procyclical real exchange rate behaviour inside a monetary union is a

destabilising force in relation to nominal debt and real interest rate dynamics. A positive differential in wage and price

inflation during current account deficit phases improves capacity to take on extra nominal debt by boosting nominal

incomes while also providing an incentive to bring forward spending plans in the face of a common area-wide nominal
 [Last Name] 53

interest rate. These forces work in the opposite direction during adjustment phases, with a negative inflation differential

raising the real value of nominal debt liabilities and encouraging the deferral of spending plans.

The absence of national currencies also affects the payoff structure on nominal assets and liabilities. During the

crisis, several advanced economies with independent currencies obtained net external wealth gains through currency

depreciation, which raised the local currency value of foreign-currency assets relative to domestic-currency liabilities. This

mechanism is not available to individual countries inside a monetary union. More generally, national policymakers cannot

deploy inflation and currency depreciation to alter the returns on local-currency instruments relative to foreign-currency

instruments.

In terms of accumulated net positions, bilateral creditor-debtor relations inside the euro area may give rise to

stark conflicts of interest during periods of financial distress, in terms of striking the balance between fostering debt

payment discipline and debt restructuring. At the same time, strong political and institutional ties between creditor and

debtor economies also facilitate additional policy options, such as the provision of official financing at below market rates,

even if the design of the associated policy conditionality programme provides further room for dispute

See also CGFS (2010), Allen et al (2011) and CIEPR (2012).

4 between creditors and debtors.

 [Last Name] 54

In relation to liquidity provision in the event of market disruption or rollover risk, the euro denomination of cross-

border debt liabilities means that the eurosystem can provide cross-border liquidity to banks. This is in sharp contrast to

the environment facing emerging market economies that have foreign-currency liabilities, which must rely on international

organisations or foreign central banks (through swap lines) to provide foreign-currency liquidity. In this way, membership of

monetary union provides a “safe harbour,” at least relative to similarly-indebted open economies that can only obtain

foreign-currency funding. Liquidity provision is also an issue in the sovereign debt market. Until recently, the fear was that

individual sovereigns within the euro area could not rely on central bank support to counter liquidity runs. However, the

OMT programme announced by the ECB de facto acts to forestall such runs where the solvency of vulnerable governments is

underpinned by adhering to policy conditionality under an official ESM programme.

2.4 Global Capital Flows

While bilateral capital flows and bilateral positions within the euro area are the major proportion of total cross-

border financial linkages, it is also important to recognise that external financial linkages are also important. Within

Europe, the United Kingdom plays a special role as an international financial centre, with high two-way flows vis-a-vis the

euro area. Bilateral financial links are also strong with other European advanced economies, while the euro area is also a

significant net investor in Central and Eastern Europe.

Globally, two-way financial trade with the United States is especially important. This was underlined during the

2008-2009 global financial crisis, with European banks incurring significant losses in the ABS market in the United States,

while the high reliance of European banks on dollar funding markets left these banks vulnerable to the freezing of these

markets. The group of emerging economies in Asia and Latin America are also an increasingly important source and

destination for capital flows. Finally, offshore financial centres (such as the Caribbean islands) are another important global

counter-party for euro area investors.

These external financial linkages matter for several reasons. Although the aggregate current account balance of

the euro area has been relatively small in recent years, net ex- ternal financial flows can allow the euro area to run
 [Last Name] 55

collective current account imbalances, providing scope for smoothing in the event of area-wide shocks. In addition, gross

external financial positions provide room for risk sharing with the rest of the world, which is especially relevant for area-

wide shocks. At the same time, as vividly illustrated by the US financial shock in 2008, it also means the euro area is

exposed to external financial shocks.

2.5 The Drivers of Capital Flows

In relation to the underlying drivers of capital flows, the literature has traditionally been organised around “push”

and “pull” factors, where the former refers to the determination of outward flows from investor economies and the latter

refers to the characteristics of those economies receiving capital inflows. However, this distinction has more limited

relevance in understanding the general levels of gross flows, since capital inflows and capital outflows for individual

economies are very highly correlated.

In respect of net capital flows, there is a vast literature on the determination of current account balances. In

addition to cyclical and fiscal factors, this literature has also highlighted the contributions of country characteristics such as

demographic structures, the level of development and natural resource endowments in explaining persistent current

account imbalances (see Lane and Milesi-Ferretti 2012 for a recent overview).

As highlighted by Forbes and Warnock (2012a, 2012b), there is a striking global factor in

5 gross capital flow patterns, with common waves of higher or lower gross capital flows affecting all countries. In

turn, this global factor can be linked to the general financial environment, with a strong correlation with indicators of

expected financial market volatility (such as the VIX index)

 [Last Name] 56

. In addition, there is also considerable cross-country variation in the gross scale of capital flows. As shown by

Lane and Milesi-Ferretti (2008a), countries with larger domestic financial systems, higher output per capita, greater trade

openness and smaller populations typically exhibit higher levels of international financial integration. Furthermore, the

composition of capital flows differs across different country groups, with advanced economies typically showing a higher

equity share in foreign assets and a higher debt share in foreign liabilities than emerging or developing economies (Lane

and Milesi-Ferretti 2007).

Europe has been to the forefront of international financial integration. In addition to having the basic country

characteristics favouring high capital flows, the abolition of capital controls in the 1980s and early 1990s, the

harmonisation of financial regulations at EU level and the introduction of the single currency have all promoted levels of

capital flows in excess of other advanced economies (Lane 2006, Lane 2009, Lane and Milesi-Ferretti 2008). These factors

especially stimulated financial trade within the euro area by reducing transaction costs and increasing the elasticity of

substitution between assets issued by the individual member countries (Coeurdacier and Martin 2009, Spiegel 2008a,

Spiegel 2008b). However, asset trade between the euro area and the rest of the world was also stimulated by the creation

of a deeper, more liquid financial market.

Importantly, the creation of the euro had a bigger impact on debt-type flows than on equity-type flows. The

commodity-type nature of wholesale debt products and the high perceived substitutability of common-currency bonds in a

low-risk environment fuelled a rapid expansion in cross-border debt flows. As surveyed by Lane (2006, 2009), the creation

of the euro also promoted cross-border equity/FDI trade. However, exchange rate risk is a relatively minor factor in the

valuation of equity-type assets, so that the euro effect was necessarily smaller than for the debt category.
 [Last Name] 57

In terms of the debt-equity composition of capital flows, Faria et al (2007) find that larger, more open economies

with a better institutional quality score have a greater equity share in external liabilities. Moreover, these authors find that

equity financing is stronger among those countries that have undertaken a greater degree of domestic financial reform.

Pg 12

Capital Flows and the Crisis

In terms of crisis dynamics, Figure 12 shows the strong correlation between current account balances in 2007 and

the subsequent adjustment process. High-deficit countries experienced a contraction in the size of current account

imbalances and much larger recessions than other euro area countries. Although the cross-sectional patterns in real

exchange rates over 2007-2012 are correlated with the size of initial imbalances, the correlations are quite low and the

magnitudes of the shifts in real exchange rates are quite small (see also Table 4). These patterns for the euro area are in

line with the evidence for a much larger sample of countries reported in Lane and Milesi-Ferretti (2012).

13

Table 5 provides further insight by reporting the shifts in savings and investment rates that have accompanied

current account adjustment. The most striking pattern is that high-deficit countries experienced extraordinary declines in

investment rates. For Greece, Ireland and Spain, this investment slump was predominantly in the labour-intensive

construction sector, which was associated with a sharp reduction in employment.

Figure 12 and Tables 2-5 illustrate the negative macroeconomic impact of the boom- bust cycle in net capital

flows. The very high pre-crisis current account deficits in the euro periphery meant that these countries were especially

exposed to a sudden adverse shift in financial markets, in view of the close correlation between general financial sentiment

and the scale of capital flows.

 [Last Name] 58

14

In turn, the rapid reversal in capital flows was associated with large-scale expenditure reduction. Since there was

only minor movement in real exchange rates, there was little by way of expenditure switching such that the net outcome

was severe output declines in the high-deficit countries.

15

The scale of current account adjustment would surely have been larger in the absence of cross-border ESCB

liquidity flows (as reflected in Target 2 balances) and official EU/IMF funding to Greece, Ireland and Portugal (Cecchetti et al

2012, Lane and Milesi-Ferretti 2012, Merler and Pisani-Ferry 2012, Auer 2013, Whelan 2013). The changes in Target 2

balances between 2008.Q3 and 2012.Q2 are shown in Table 4 and show large increases in net Target 2 liabilities in the

high-deficit countries and in net Target 2 claims in the high-surplus countries.

Large official gross flows also allowed private-sector foreign investors in creditor countries to exit from positions

in the high-deficit countries by declining to rollover expiring claims. In the absence of large-scale official flows, foreign

investors would plausibly have incurred larger valuation losses through sharper declines in asset values and more extensive

debt writedowns.

In relation to gross capital flows, Milesi-Ferretti and Tille (2011) show that the spec- tacular

contraction during the crisis was a global phenomenon. However, it hit the euro area especially hard since the

gross scale of capital flows was so much bigger than in other advanced economies or in emerging markets. Since the

freezing of credit markets was at the centre of the original phase of the crisis during 2008-2009, it is not surprising that

bank-related debt flows fell the most.

 [Last Name] 59

Still, although the reversal in net capital flows was surely destabilising for the high- deficit countries, some types

of capital flows have acted as a buffer during the crisis. For instance, the evidence provided by Forbes (2012) suggests that

high stocks of foreign port-folio assets may have mitigated contagion effects for some countries.

More generally, the ability to repatriate foreign assets has provided much needed liquidity to distressed entities,

especially where foreign assets maintained more value than domestic assets. This has been important for some multi-

country banks that were able to extract capital from foreign affiliates in order to shore up domestic operations. At a

national level, the liquidation of the foreign assets in Ireland’s sovereign wealth fund has been an important source of

funding in addressing its domestic banking crisis.

16

Still, it must be acknowledged that another potentially stabilising role for capital flows has had only limited impact

for euro area countries. That is, for a country with an independent currency, exchange rate depreciation during a crisis

might stimulate capital inflows since the decline in the foreign-currency value of domestic assets should encourage bargain

seekers. This mechanism is switched off for individual members of the euro area.

4.2 Macro-Financial Stabilisation Policy Framework

The volatility of capital flows should also inform design of macro-financial stabilisation policies at both national

and European levels. Taking first a national-level perspective, the high costs of boom-bust cycles in international debt flows

reinforce the case for a macro- prudential framework, which encompasses both financial stability policy and fiscal policy.

Indeed, this perspective underpins the recently-introduced “macroeconomic imbalance procedure” that is now a key

component of the European policy framework.

 [Last Name] 60

4.2.1 Macro-Prudential Policies

PG 18

A macro-prudential regulatory system can mitigate the risks of excessive cross-border debt flows by influencing

the composition of both the asset side and the liability side of the balance sheets of banking systems. On the asset side,

regulations that that guard against excessive geographic concentration in loan portfolios would limit the expansion of

regionally-specialised banks and, indirectly, stimulate the relative market share of larger, geographically-diversified banks.

In related fashion, regulations that limit sectoral concentration in loan portfolios would reduce the risk of excessive

exposure to individual sectors (such as construction or asset-backed securities). In turn, this would help to limit the

amplification dynamics that fuel regional property boom-bust cycles. Similarly, regulators could mandate that the bond

holdings of banks be sufficiently diversified, including limits on the exposure to individual sovereign governments. Such

regulations should be more effective under the new single supervisory mechanism, since it is important to take a

consolidated view of the activities of each banking group, fully incorporating the correlations across the asset holdings of

individual subsidiaries of the parent bank.

On the liability side, a regulatory approach that dissuades excessive wholesale funding would in effect reduce the

scale of cross-border debt inflows during boom periods; given that cross-border debt flows are primarily inter-bank flows

(see also Committee on International Economic Policy and Reform 2012). Although, this in part may be offset by an

expansion in direct cross-border lending, this could also be mitigated through cooperation with home-country supervisors of

foreign banks, as is facilitated by Basel III (Borio et al 2011). Indeed, the recently-agreed single supervisory mechanism

should eliminate the risk of such regulatory arbitrage within the European system (see also Coeure 2013).

Structural Reforms

The shifts in sectoral economic activity associated with fluctuations in net capital flows also reinforce the

importance of labour market institutions (and ancillary policies) that can facilitate the mobility of workers across sectors.
 [Last Name] 61

This fundamental principle should inform policy choices across a wide spectrum of policy issues, including retraining,

housing, pensions and wage flexibility.

Resilience in the face of sectoral volatility is also enhanced by strong financial and legal systems that can facilitate

the entry and growth of firms in expanding sectors and efficiently manage the decline and exit of firms in contracting

sectors. The value of a robust banking system in managing reversal episodes underlines the importance of ensuring the

resilience of the banking system in the face of capital flow volatility. In relation to the legal system, efficient mechanisms

for debt restructuring (households, corporates, and banks) are an important element in exiting from crisis episodes (Laeven

and Laryea 2009, Laryea 2010, Brown and Lane 2011).

 [Last Name] 62

DomeStic creDit groWth anD international caPital FloWS

Philip R. Lane and Peter McQuade

NON-TECHNICAL SUMMARY

Understanding the origins of the various waves of the global financial crisis (especially the current European crisis)

is a high priority for researchers and policy-makers. Such diagnostic work is essential both in designing policy solutions to

resolve the current crisis and in improving preventive frameworks to mitigate the risk of future crises.

Two key contributory factors in the current crisis have been the balance sheet problems associated with rapid

credit growth in some countries (most obviously, Ireland and Spain) during the pre-crisis period and excessive external

imbalances. For instance, Lane and Milesi-Ferretti (2011) have documented that the variation in the size of recessions

during 2008-2009 was significantly related to the scale of domestic credit growth during the 2003-2008 period and the size

of outstanding current account imbalances. In related fashion, Lane and Milesi-Ferretti (2012) show that above-normal

current account deficits during the pre-crisis period was significantly associated with major declines in domestic demand

and sharp reversals in private capital flows over 2008-2010.

The importance of these twin factors raises the question of whether there are important interactions between

domestic credit growth and international capital flows. If these variables are jointly determined and/or interact in

economically-interesting ways, this should frame the analytical framework guiding theoretical and policy analysis. Along one

dimension, it would indicate that international capital flows should be a central theme in the rapidly-growing macro-

prudential literature that seeks to understand the dynamics of domestic credit growth (and the associated risk factors).

Along another dimension, it would indicate the domestic credit channel is a key channel in understanding the relation

between international capital flows and domestic macroeconomic and financial variables.

In this paper, our main focus is on the relation between domestic credit growth and international capital flows for

a sample of European countries. In particular, we focus on the EU27, plus Norway, Switzerland and Iceland. (Taken

together, we label these countries as the E30 group.) Europe is an important testing ground for exploring the inter-relation
 [Last Name] 63

between credit and capital flows, in view of the remarkable dispersion in domestic credit patterns during the pre-crisis

period and the very high level of cross-border capital flows.

Moreover, the large and persistent intra-European external imbalances provide an additional layer of complexity

(Giavazzi and Spaventa 2010, Lane and Pels 2012). In particular, net capital ows and domestic credit growth have been

separately identified as important sources of macroeconomic imbalances, such that it is highly relevant to understand any

inter-connections between these variables.

A major trend in European banking systems during the pre-crisis period was the divergence between domestic

deposit growth and credit growth. In order to finance credit growth that was more rapid than deposit growth, banks raised

funds by borrowing short term on international interbank and money markets and by issuing bonds. These shifts in bank

funding patterns in Europe and the associated growth in cross-border bank-related financial flows are suggestive that a

systematic relation might exist between international capital flows and domestic credit growth.

The primary focus is on the E30 group over the period 2003-2008, although additional analysis is conducted on

annual data for an extended sample of 54 countries for the period 1994-2008. Using these data we present stylised facts

and run a series of OLS regressions to explain credit growth. As there may be two-way causality effects between domestic

credit growth and international capital flows, we also report IV estimates, where international financial flows are

instrumented by their lagged values.

Our analysis confirms that the current account balance is a misleading indicator in understanding the relation

between international capital flows and domestic credit growth, in view of the striking differences in the co-variation of

domestic credit growth with net debt flows and net equity flows. Moreover, it is striking that net debt flows appears to be

the relevant measure, with no apparent gain to splitting net debt flows between gross debt inflows and gross debt outflows.

This may have to do with the nature of international trade in debt claims, with many types of gross inflows and outflows

essentially cancelling each other out. Furthermore, while our primary motivation is to study the European experience, it is

striking that the results are quite similar in the extended sample.

The apparent empirical connection between net international debt flows and domestic credit growth calls for

analytical models that can capture this relation. In particular, it is important to understand better both the direct relation

between international debt flows and domestic credit growth (for instance, through the international funding activities of

domestic banks) and the indirect relation (the impact of international debt flows on domestic macroeconomic and financial

variables that can affect both supply and demand factors influencing domestic credit growth).

In turn, these findings have implications for macro-prudential policy frameworks and the monitoring of excessive

imbalances. In particular, our analysis indicates that there is a strong international dimension to the determination of

national credit growth rates and that domestic credit growth and external imbalances should be interpreted in an

integrated, joint framework.

 [Last Name] 64

Gross or Net International Financial Flows

Understanding the Financial Crisis

Karen H. Johnson July 2009

It has been argued by many analysts, including Chairman Ben Bernanke of the Federal Reserve, that the global external

imbalances, which have been a feature of the world economy for more than a dec-ade, are an important, causal factor

underlying the global financial crisis of the past two years. 1 The claim is that net financial flows, which are the counterpart of

net current account balances, resulted in capital flowing from surplus countries, such as China, to the major deficit countries,

especially the United States, and contributed to low dollar interest rates, unrealistic risk premia, high U.S. asset prices, and
 [Last Name] 65

excessive expansion of credit. Once credit quality problems emerged, the bust soon fol-lowed. But it is not the net external

imbalances (or net financial flows) of the surplus countries that move across borders and must be intermediated by the global

financial sector; it is the gross financial flows of all countries that require this. Moreover, it is the gross stocks of assets held

cross-border (far larger than any one year’s gross flows) that must be managed for risk. This paper will explain in detail the

differences between the net external imbalances and measured gross financial flows. This comparison will reveal that data on

net external imbalances do not identify which countries are most actively involved in large cross-border financial transactions

and, hence, whose behavior is most critical for global financial stability.

The data for multilateral current account balances are readily available and are most often what is meant when the

term “external imbalances” is used. The multilateral nature of this concept is frequently useful because it reflects the complex

and multifaceted trade links among countries in the global economy. And this concept reports the change (either positive or

negative) in the net claims on the rest of the world of the country in question. Hence it cumulates to a measure of the net cross-

border wealth holdings (or indebtedness) of the country. 2 Although current account data are most often derived from the trade

side of balance-of-payments data, there is a counterpart set of financial flow accounts, which, apart from a statistical

discrepancy, give an identical result for the net imbal-ance.

Insights from Multilateral External Imbalances

Table 1 provides current account balances for the major surplus and deficit countries for 2007. Although losses in U.S.

subprime mortgages began to emerge in 2007, data for that year record the last full year before the crisis hit in a major way.

The table contains data in U.S. dollars, so larger economies tend to be at the top of the lists of surplus or deficit countries. In

analyzing the recent crisis, this approach, in contrast to expressing the balances as a share of gross domestic product (GDP), for

ex-ample, reflects the potential for the given country’s deficit or surplus to be of importance for global financial markets. For

broad global market developments, the size of the imbalance—and counterpart net financial flow—is what matters.

In order for there to be some countries experiencing deficit, there must be others with surplus, as the balance for the

entire global economy must be zero. So, in general, one should look both to countries with large deficits and large surpluses for

underlying factors that contributed to the global financial crisis. Those countries with current account deficits in a given year

received net financial inflows 4

from the rest of the world; those with surpluses supplied financial capital—that is, had net financial outflows—to the

rest of the world.

What can be learned from the actual data for 2007? First, the United States clearly ran the largest deficit and so

absorbed the largest net inflow of financial flows from abroad. The United Kingdom was second, with a net inflow of about $80

billion, a bit more than 10 percent of that of the United States. After these two countries, the deficits reported on the list
 [Last Name] 66

rapidly get smaller in absolute size. Issues of absorbing sizable net inflows of capital from the rest of the world would seem to be

largely confined to these two countries, with Australia running third. With respect to surplus countries, those providing net

financial flows to be absorbed elsewhere, China reported the largest figure at more than $370 billion. Japan was second, with

about $210 billion, roughly two-thirds of the Chinese surplus. Together their surpluses are almost equal to the U.S. deficit. The

next three surplus countries benefited from substantial earnings from the sale of crude oil. The subsequent two, Switzerland and

Singapore, are global financial centers with small domestic economies.

Based on these data, the following implications have been pointed out by many analysts:

– The large net inflows of financial capital into the United States added to the financial excesses that occurred in these

markets. They exerted downward pressure on dollar interest rates and upward pressure on U.S. asset prices. Risk management

practices were allowed to slacken.

Many of the same excesses were experienced in London markets. Net inflows of foreign financial capital contributed to

the excessive expansion of the financial sector, to the measurement of substantial output in the services sector of the economy

despite weakness elsewhere, and to lower interest rates and higher asset prices than would otherwise have occurred.

China, Japan, and the oil-exporting countries had high domestic saving relative to investment and sought attractive

returns elsewhere in global markets. The capital markets of New York and London are the most developed and safest such

markets and attracted capital inflows. The surpluses of these high-saving countries contributed (along with many other factors)

to the outcome of large deficits in the United States and United Kingdom as global saving flowed into their markets. The

resulting imbalances thus fed the global boom that ultimately collapsed in crisis.

Countries that were nearly in balance had an essentially neutral effect through global financial flows on the

development and then bursting of the global financial bubble/crisis. (They may have had domestic financial vulnerabilities that

have succumbed to global pressures.) In particular, the euro area, with a small recorded surplus in 2007, was not a factor in

adding to the pressures created by the global flow of financial capital. Similarly, Canada did not contribute importantly to these

net financial flows and resulting pressures.

A worrisome part of the story is that over time, as China, Japan, and others have accumulated surpluses for several

years, they have built up large holdings of international assets (some are referred to as sovereign wealth funds), which include

large dollar balances. The risk of a reallocation of these funds out of dollar assets into others adds to the uncertainty now

troubling financial mar-kets.

Insights from Gross Cross-border Financial Flows

So much for the lessons from net cross-border financial flows. What can be learned from gross financial inflows and

outflows? Because the global financial crisis is seen as arising in the United States, an examination of U.S. data can provide new

information and insight. Clearly, the total U.S. financial inflow in 2007 of more than $2 trillion, shown in Table 2, was many

times the U.S. current account balance of about $730 billion. Of this total, about 75 percent were inflows from private investors.
 [Last Name] 67

The argument is often made that the economic incentives that drive direct investment inflows are quite different from

those that drive the portfolio investment into securities or the flows among banks. Direct investment is seen as having a longer-

term focus. The flows are often between a firm and its foreign subsidiaries or foreign partners and may be the result of earnings

of the foreign subsidiary that are retained abroad. For these reasons, it probably makes sense to concentrate on private

securities transactions and bank flows in looking for insights into the financial crisis.

Private inflows into U.S. securities were the largest single category and were about 40 percent of the total private

inflow. U.S. Treasury security purchases were about 10 percent, with other securities such as stocks and private bonds

accounting for the remaining 90 percent. These securities were traded on U.S. capital markets, rated by the major rating

agencies, held by a wide range of investors, and figured prominently in the crisis. The increase in liabilities to private foreign

investors reported by U.S. banks, about $510 billion, was also large. Many bank transactions are with other banks. Such interbank

transactions can result in both banks increasing their claims on each other, and hence increasing the gross flows reported in the

balance of payments, for business reasons that are technical

and may not be significant. Shifting of funds by an international bank among its branches is one such example. But

because the global banks are at the Center of this financial crisis, it seems wise to look closely at all the data, including that of

substantial increases in the liabilities of U.S. banks to foreign private investors.

A look at the changes from 2003 to 2007 in the main components of U.S. private financial inflows is enlightening. Table

3 shows that the total gross financial inflow to the United States increased $1.3 trillion over this period, whereas the current

account deficit rose in size by only $200 billion. Of the $1.1 trillion increase in private inflows, net private securities purchases

by foreign investors increased by $360 billion and the net increase in liabilities of U.S. banks to foreigners rose by about $410

billion—again, much larger than the increase in the current account deficit.

Do U.S. Gross Financial Flows Paint a Different Picture Than Net Current Account Balances?
 [Last Name] 68

To Summarize the differences in implications for 2007:

– Current account data lead to the conclusion that the major financial flows were from China and Japan to the United

States. But the data on financial flows show that in fact the two countries/currency areas with the largest gross financial inflows

into the United States in 2007 were the euro area—with an essentially neutral current account position—and the United Kingdom,

which had a current account deficit.

– A large amount of two-way trade in assets occurred in 2007 between the United States and the United Kingdom and

between the United States and the euro area. This was not the case for China and Japan, where the U.S. balance-of-payments

data record significant acquisitions by Chinese investors of U.S. assets but net reduction by U.S. investors of claims on China. In

the case of Japan, acquisition of U.S. assets was limited, and again U.S. investors actually reduced their financial claims.

– The components of gross financial inflows and outflows for the United States show that the risk characteristics of U.S.

and foreign investors differed. Although the detailed data are limited, foreign inflows from the other industrial countries and

regions showed large acquisition of U.S. securities, including those other than U.S. Treasury securities. Foreign investors were

clearly contributing to the demand for securities trading on the capital markets. The increase in U.S. claims on these

counterparties tended rather to be claims reported by banks. These “mismatches” provide insight into where concentrated risks

developed. They may also cast light on where regulatory differences were being exploited in a cross-border fashion.

Conclusions for the larger question of the links between external imbalances and the global financial crisis:

Current account balances do not move over time in a predictable, stable fashion with gross cross-border financial

flows.

Current account balances may signal problems that require policy response and are the answer to some questions. But

they are not a problem because they engender large cross-border financial flows. Those flows occur anyway in today’s world.

Small surpluses or deficits do not signal that a country is refraining from engaging in large cross-border asset transactions that

may be risky, and large surpluses or deficits do not necessarily indicate that a country is not managing its international portfolio

well.

Policies based on current account balances, such as ceilings that would trigger remedial policy response or IMF-

enhanced surveillance, cannot be justified because of the risks posed by global financial flows to global financial stability.

Current account data are too partial (because of netting in the statistics) to give useful information on the associated financial

flows. The netting embed.

ded in current account data can be misleading because the assets acquired by some domestic residents are not

necessarily available to meet the liabilities of other domestic residents and because the composition of claims acquired on the

rest of the world, and hence their risk characteristics, may be quite different from the composition of claims on the domestic

economy acquired by foreign residents.

 [Last Name] 69

– There is no substitute for careful supervision and regulation of the major international financial institutions and asset

markets. Gross financial flow data can indicate where rapid expansion of particular forms of credit is occurring, where leverage

may be becoming excessive, where regulatory inconsistencies are being exploited, and where heightened systemic linkages may

be found. Although the data on gross flows may not provide a complete picture of the countries of residence of ultimate

borrowers and lenders, it does provide a picture of which countries are important in the infrastructure of global finance.

– For understanding the factors behind the global financial crisis and the policy measures revealed as necessary by it,

one needs to look at the role of the various countries and institutions in the financial markets, not in the markets for goods and

services. Macroeconomic policies designed to narrow current account imbalances will not necessarily strengthen the global

financial network.
 [Last Name] 70

Causes and Consequences of the Spanish Economic Crisis: Why the Recovery is Taken so Long?

Francisco Carballo-Cruz
 [Last Name] 71

The Crisis’s Impact

The Spanish economy’s downward phase of the expansion cycle began in 2007. Four years later an economic recovery is

yet to be seen. During 2007 and the first quarter of 2008, there was a growth slowdown, and from the second quarter of 2008

until the last quarter of 2009, the economy was in recession. This recession period of seven consecutive quarters was unusually

long, since in a normal cycle, the recession does not usually extend beyond three or four quarters. The hardest stage of the crisis

in terms of product breakdown and job destruction coincided with the first quarter of 2009, when GDP fell 6.3% and

unemployment increased by around 800,000 people.

The Spanish economy began to recover in 2010, with a slight growth in output in every quarter except in the third. The

output’s growth of that year, 0.6%, was very modest when compared with the accumulated loss of production in the previous two

years (almost 5%). The 2011 data show the weakness of the recovery, with growth rates of 0.3% and 0.2% in the first and second

quarters, respectively.

As far as unemployment is concerned, the available data is very alarming. The unemployment rate rose from 8.3%

(1,834,000 unemployed), in late 2007, to 20.1% (4,632,000 unemployed) in late 2010. The years of higher job destruction were

2008 (growth of 41%) and especially 2009 (growth of 60.2%), in which the unemployment rate increased to 18.0% of the total

workforce (4,150,000 unemployed). Unemployment increased mainly among younger workers (41% in the range of 16 to 25

years), particularly those with lower qualifications (between 25 and 45%, depending on education levels) and among foreigners

(30%). Throughout 2010 there was a strong growth in long-term unemployment, which represents 42.5% of the total, particularly

among workers between the ages of 45 and 64 years old (52.5%). Unemployment growth is a differentiating aspect of the crisis in

Spain. Between 2007 and 2010, the unemployment rate in Spain rose from 8.3% to 20.1% (11.8 percentage points), whereas in the

euro zone the increase was less accentuated merely from 7.5% to 10.1% (2.6 percentage points).

In addition to the high unemployment rates, there are other differentiating aspects when compared to the eurozone.

Although the product’s fall is similar in terms of magnitude, its composition presents significant differences. Between 2008 and

2010, domestic demand in Spain fell 7.6%, whereas in the eurozone it fell merely

1.6%. Investment in housing was the most affected component, which in these three years decreased by 41%.

Households reduced their savings rate to historically low

pp. 309-328

levels and increased their fixed capital investment to maximum levels. In 2007, this behaviour decisively contributed to

the increase of the households’ debt up to 130% of their gross disposable income (GDI). Currently, the ratio of debt lies close to

125%, higher than the eurozone average (98%), similar to the United States (118%) and lower than the United Kingdom (151%).

The slow pace of households’ deleveraging is mainly due to the weakness of household income and the long term debt

amortization for house purchase, which hinder the rapid liabilities depletion.
 [Last Name] 72

Non-financial firms’ debt, in terms of gross operating surplus, is around 750%, surpassing those of the eurozone (~

550%), the UK (~ 650%) and the U.S. (~ 350%). Since the peaks reached in 2008, the ratio of firm debt has only been reduced by

1% due to the scant growth of firms’ results. It should be noted, however, that in terms of debt, there are considerable

differences at the sectoral level. For instance in the construction and property development sectors, the level of debt greatly

exceeds that of other sectors of the economy, in spite of the fact that over the past two years, the deleveraging process of non-

financial firms has been mainly focused on these sectors. The high amount of debt in these sectors reflects the enormous stock

of unsold real estate assets.

In 2009, the high growth of public investment (11.2%) served to offset the impact of the strong adjustment of private

investment. However, in 2010, the need to control the growth of public deficit led public investment to contract by more than

17%. The Spanish economy went from having a surplus in public accounts of 1.9% in 2007, to presenting a deficit of 9.2% in 2010,

having peaked at 11.1% in 2009. Public debt increased from 36.1% in 2007 to 60.1% in 2010.

In 2010 a process of fiscal consolidation was initiated. The Updated Stability Programme of February 2010 established a

procedure for gradually reducing the deficit to 3% of the GDP by 2013. In 2010, fiscal consolidation was achieved thanks to a one

percentage point increase in the ratio of tax revenues on GDP and a fall of eight tenths in the ratio of public expenditure on

GDP. Table 1 presents the main budget adjustments approved by the Spanish government since September 2009. In 2010,

revenues increased primarily due to the increased VAT collection, which offset the less dynamic collection of direct taxation.

The cut in spending has focused mostly on capital expenses and intermediate consumption. Social benefits increased due to the

inertial evolution of pension expenditure and the high incidence of unemployment benefits. Interests on debt followed an

upward trend and already account for 1.9% of GDP. In the coming years due to the likely evolution of public debt and interest

rates, it is expected that the interest burden of public debt will increase its weight in public expenditure.

In April 2011, the government submitted a new version of the Stability Pro-gramme, for the 2011-2014 period.

According to the set out objectives, public deficit will be reduced to 3% of the GDP by the end of 2013 and the public debt ratio

will stabilize just below 70% of the GDP, in the 2012-2013 biennium. 65% of the adjustment in the 2009-2013 period is based on

spending cuts, including a deep reduction of public consumption (35% of total).

In light of current macroeconomic situation there is great uncertainty regarding compliance with the objectives of last

April’s Stability Programme. Firstly, because they are based on a macroeconomic scenario, that assumes a relatively high rate of

economic growth. Secondly, because a significant proportion of Spain’s public expenditure has, traditionally, an incremental

inertia, namely expenditures in health, education and pensions. Thirdly, due to the significant expense of unemployment

benefits (equivalent to 3% of the GDP in 2010), caused by the persistence of high unemployment levels (over 20%), whose

reduction is not foreseeable in the near future.

Regarding prices’ evolution, the sharp drop in consumption and in fixed capital investment has allowed to partially

correct the huge imbalances accumulated in the pre-crisis expansion period, in terms of inflation differential with the eurozone.
 [Last Name] 73

As far as wages are concerned, in the early years of the crisis, remuneration per employee continued to increase as a result of

the inertia of collective bargaining and the sharp increase in non-wage costs, such as compensations for dismissal. In 2010,

remuneration per employee grew only by 1.4% (2.3 percentage points less than in 2009), in part due to the Agreement for

Employment and Collective Bargaining (AENC), signed by the social agents in February 2010. The expected wage moderation

indicates the beginning of the process to correct the Spanish Economy’s real exchange rate.

As previously mentioned, the construction and property development sectors in Spain had an essential role in the

detonation and extension of the current economic crisis. The disproportionate growth of housing prices has led to a housing

bubble of enormous proportions. There are three factors that have contributed to its emergence and development. First, the

monetary policy followed by the European Central Bank, since 2001, which kept the reference interest rate at very low levels for

the cyclical position of the Spanish economy. Secondly, the fiscal policy followed by the Spanish government, which promoted

the acquisition of housing instead of other alternatives, such as renting, and encouraged the purchase of real estate assets

(including housing) in detriment of other investment assets. Thirdly, the advantages of a model of economic growth based on

construction and property development activities, from the political economy point of view i) reduction of unemployment, given

that these are labour-intensive activities (favouring politicians), ii) increase in housing value (fa-vouring the median voter, who is

usually a home owner), and iii) generation of large tax revenues (particularly real estate) for the different public administrations

(favour-ing politicians) (Celia Bilbao Terol, María A. García Valiñas, and Javier Suárez Pendiello 2006). Therefore, in this context,

it can be stated that for many years, there was no political interest in halting the excessive growth of construction activities and

property development.

The initial price increases derived from favourable market conditions for mortgages, followed by additional increases

resulting from the contagion of positive expectations about price evolution, resulting in a bubble of enormous proportions. The

bursting of the bubble led to a severe fall in demand, which in turn resulted in an adjustment in supply, either via prices or via

quantities.

Credit for construction and property development activities was the mean of transmitting the housing crisis to the

banking sector. In the pre-crisis expansion period, 2004-2007, credit to the construction sector experienced an average annual

growth of 24.6%, whereas credit to the real estate sector grew at an average annual rate of 43.1%. In 2007 loans to both sectors

accounted for almost 45% of the Spanish GDP (14.5% to construction and 30% to property development), when their overall

weight in product was less than 20%.

This unbalanced growth of credit has resulted in a high concentration of risks in the construction and property

development sectors, both on the supply and on the demand side, resulting from the inadequate risk policy of the banking
 [Last Name] 74

system and the Bank of Spain’s insufficient supervision during this period. The huge stock of real estate assets, whose

construction or purchase was financed with bank loans, remained in the financial entities balance sheets and, in a recession

period with a demand deficit like the current one, produces losses by default and assets depreciation and adversely affects

banks’ turnovers. This situation is at the basis of the solvency problems of an important part of the banking system, namely the

Savings Banks –Cajas de Ahorros, whose weight in the sector, measured by various indicators, is close to 50%. To tackle the

solvency problems, throughout 2010 an intense restructuring process took place in the banking sector in Spain, which up to now

is producing positive results in terms of reorganization, losses recognition and recapitalization.

Currently, the program of fiscal consolidation, the elevated rates of unemployment and the high indebtedness levels

explain the low dynamism of internal de-mand. The recovery of economic growth is also aggravated by the sovereign debt crisis.

Spain, Italy and the three rescued countries (Greece, Ireland, and Portugal) 1 are facing financing problems as a result of the

eurozone institutional design’s problems (Paul De Grauwe 2011a). In these circumstances, be a member of the eurozone can

contribute to the deepening of its own crisis (Paul Krugman 2011). This is particularly evident in the Spanish case, as their fiscal

problems do not stem from irre-

1. For excellent reviews on the crisis in these countries see Georgios P. Kouretas and Prodromos Vlamis (2010),

Constantin Gurdgiev et al. (2011), and João Sousa Andrade and Adelaide Duarte (2011), respec-tively.

2. The Two Key Sectors

2.1 Evolution and Current Situation of the Real Estate Sector

Between 1997 and 2007, there was a long cycle of housing expansion in Spain. This cycle is different from others due to

the extraordinary construction volume and its exceptional duration (eleven years). During this period, the average annual growth

in the construction sector was higher than 5%. In late 2007, the construction sector concentrated almost 14% of employment and

16% of the Spanish GDP. By including demand related sectors, output and employment dependent on the construction sector

achieved respectively, in that same year, 25% of the GDP and 23% of the overall employment (Ramón Tamames 2009).
 [Last Name] 75

The demand for housing has been stimulated by several factors. The main ones are the strong economic expansion seen

in this period (in part due to the real estate’s boom) and the reduced interest rate on housing loans after the Euro integration

(the reference rate for loans of this type decreased from 9.6% in 1997 to 3.3% in 2007). Other factors explaining the strong

dynamics of demand are the greater competition in the banking sector, the growth in the number of households, largely due to

the massive influx of immigrants (about 4.5 million in the 1997-2007 period), and the housing purchase by non-residents, as

second homes.

The strong demand for housing found a very dynamic response on the supply side. In the 1997-2007 period almost 5.3

million dwellings were finished in Spain and, in several years of the expansion cycle, the number of finished dwellings per year

surpassed half a million. The net increase in the housing stock between 2001 and 2008 was of 4.3 million homes. The stock rose

from 20.8 million in 2001 to 25.1 million in 2008, representing an increase of almost 21%. The extraordinary growth in demand

has resulted in an increase in housing investment (from 4.7% of GDP in 1997 to 9.7% in 2007).

The massive housing acquisition stimulated by these demand factors has spurred an extraordinary demand for credit.

Between 1997 and 2007, housing loans as a percentage of GDP increased from 28.4% to 102.9%. The widespread use of credit for

housing exceptionally increased households’ private debt. This debt rose from 52.7% of disposable income in 1997 to a maximum

of 132.1% in 2007. As a result, the effort of individuals to acquire a dwelling rose from 4.3 years of salary at the beginning of the

cycle to 9.1 years at the end of it.

In the demand’s growth initial phase, the inability to adjust supply automati-cally, due to the specific characteristics

of the good’s production process, led to the emergence of tensions in prices (Prakash Loungani 2008). The increases in real

estate prices became more intense when expectations of the price’s future growth affected their own demand, inducing a spiral

of growth in demand, supply and prices. This situation led the average price growth to come close to 20%, in several years.

According to the Bank of Spain (the Spanish Central Bank), between 1997 and 2007, the average housing price in Spain

rose by 115% in real terms, while in Ireland the revaluations were of 160%, in the UK of 140%, in the United States of 80% and in

the eurozone merely 40% (Banco de España 2011a). In Spain, the economic fundamentals of price growth are not the only factors

explaining the high inflation in the sector. The observed increase is also explained by the demand’s significant over-reaction. The

extraordinary revaluation of property assets until 2007, followed by a sharp decrease, reveals that there was a great bubble in

the Spanish real estate sector, which caused strong overvaluation of residential real estate.

From late 2007, prices began to slightly fall and sales experienced a very strong decrease. In other words, in the

months following the bubble’s burst, adjustments were made through quantities rather than prices. Later on, a price adjustment

was made in order to approach market prices to the real economic value of the assets. In the Spanish case, the reduction in
 [Last Name] 76

prices from peak levels was very gradual; albeit in cumulative terms it is already beginning to reflect a significant adjustment.

According to the Ministry of Public Works, the fall in prices between the maximum of 2007 and the first quarter of 2011 was 15%

in nominal terms and 20% in real terms (Banco de España 2011a). Tinsa (the main real estate valuation company in Spain)

estimates that the adjustment in prices was even greater. In nominal terms, the cumulative fall in prices, between the maximum

at the end of 2007 and July of 2011 was slightly above 22% (Tinsa 2011). This decrease in prices is justified by the decline in

employment, the increase in the cost of capital, the growth in the housing stock and the presence of significant levels of credit

rationing. Despite the continuous declines in prices, the decrease in housing sales in cumulative terms between 2007 and 2010

was higher than 43% (Banco de España 2010).

The supply’s adjustment has also been very important. In 2006, the number of dwellings started surpassed 850,000.

Two years later only one-third were actually begun. In 2010, the number of works started included only 90,000 dwellings, 20,000

less than in the previous year (Julio Rodríguez 2011). According to the evolution of the number of dwellings completed in the

2008-2010 period, the housing stock in late 2010 could have reached 25.7 million homes. In that year, the estimates on the stock

of completed unsold homes ranged between 670,000 and 1,100,000 (Rodríguez 2011). In the real estate’s expansive cycle an

average of nearly 500,000 dwellings a year were built, when according to the demographic structure of the country, the annual

potential of homes’ creation is of approximately 350,000 per year (David Martínez, Tomás Riestra, and Ignacio San Martín 2006);

however, that potential, in the Spanish case, depends largely on the effective evolution of employment.

The sectoral adjustment will be long and slow. According to the transaction volumes over the last two years, the

absorption of the stock of new unsold housing can last four to five years. In 2010, the volume of transactions grew approximately

6%, countering the drops of previous years. However, the total number of transactions amounts to little more than half of those

held in the most dynamic exercise (2006). This growth in transactions is primarily justified by the anticipation of purchases to

save taxes, the fall in asset prices and the low interest rates (there was, nevertheless, a tightening of the credit criteria).

Last August, the government announced a temporary VAT reduction for new housing purchases by four percentage

points (from 8% to 4%) until the end of the year, in order to stimulate housing demand and accelerate the reduction of the

accumulated stock. It is unlikely that this conjectural measure will boost the Spanish housing market. Its main impact will be the

anticipation of housing acquisition by a limited number of households to take advantage of this VAT reduction.

2.2 Evolution and Current Situation of the Banking Sector

From 2007 onwards, the evolution of the Spanish banking sector is highly conditioned by the extent of the international

crisis and the bursting of the housing bubble in Spain. The generalization of the international crisis and the lack of confidence in

the financial markets, have restricted Spanish banks’ access to financing in the international markets. In recent years, the

internal factors that have mostly affected the sector’s profits are the banks’ high exposure to real estate, together with the

economic downturn and the consequent unemployment increase. Note that these factors decisively contributed to the rising of

the default rate, with consequences on the solvency levels of the sector.
 [Last Name] 77

At the end of 2010, the Spanish banking sector was among the firsts in the ranking of the eurozone in terms of return

on assets (0.47%) and return on own resources (7.9%) (Banco de España 2011b). On the contrary, in terms of solvency, its position

is not as favourable. The overall solvency of the Spanish banking sector is among the worst of the eurozone, 11.9%, almost two

percentage points below the average. The use of other indicators of solvency, for example, Tier 1, also shows the weakness of

the Spanish banking sector (9.6%, well below the eurozone average).

This overview of the banking sector conceals the existence of significant asymmetries between the two main types of

entities, particularly in terms of sol-vency. In the Spanish banking sector two major classes of entities coexist: traditional banks

and Savings Banks – Cajas de Ahorros. The latter assume a central role in the Spanish banking system, since they concentrate

more than 48% of deposits and more than 46% of the loans of the banking sector (Confederación Española de Cajas de Ahorros

2011). During the long expansion period before the crisis, a significant part of the Savings Bank sub-sector accumulated financial

imbalances of various kinds, which became evident after the change of the macroeconomic conditions.

One of the most troubling aspects of the balance sheets of these entities is the high risk concentration in construction

activities and property development, both on the supply and the demand side. In late 2009, before the beginning of the

restructuring process of the banking sector, credit to the construction and property development sectors granted by the Savings

Banks accounted for 56.3% of the total financing for productive activities and 27.7% of its loan portfolio to the resident private

sector. In that date, credit for the acquisition and rehabilitation of housing represented 41% of the total number of loans

granted. Other problematic aspects of this sub-sector are the excessive dependence on wholesale financial markets, the excess

of capacity installed by the intense growth of their retail distribution networks, the significant sectoral fragmentation and the

loss of profitability derived from the structures’ oversize, the increase in unprofitable assets and the rising financing cost.

In late 2010, due to the potential problems that exposure to the construction and property development sectors could

generate to the entire banking sector, the Bank of Spain promoted an informative transparency exercise. The goal was for banks

to publish, in early 2011 and according to predefined standards, additional information about their exposure to these sectors,

identifying the percentage of bad loans, guarantees at their disposal and the provisions made to cope with possible as-sets’

deterioration. The results of this exercise revealed a significant degree of heterogeneity in the sector, in terms of levels of

solvency, rates of bad loans, degree of exposure to the construction and property development sectors, magnitude of the assets

awarded and the degree of coverage with provisions, among others.

In late 2010, the credit concentration in the construction sector, real estate activities and house purchases reached

nearly 60% of the credit to the resident private sector. Credit to construction and property development accounted for around

20% of the loan portfolio to the resident private sector (22% of the Savings banks’ loan portfolio and 17% from the Banks’), while
 [Last Name] 78

housing loans accounted for approximately 39% of it. As mentioned before, this high concentration of risks is extremely worry-

ing, due to the continuous growth of default rates and the high value of real estate assets that passed into the hands of banks, as

a consequence of high failure rates. In late 2010, the default rate for credits granted to real estate developers reached 14%,

while those granted to construction companies neared 11%. Contrarily, the default rate for housing loans was significantly lower,

falling below 2.5%.

In late 2010, Savings Banks continued to report higher levels of exposure to construction and property development.

Specifically, their total exposure was of €217 billion, of which €173 billion (80%) were related to investment credit and the

remaining €44 billion (20%) to awarded properties. At that time, the potentially problematic investment which included bad

debts, substandard loans and awarded properties amounted to €100 billion (46% of the total). The specific provisions made

represent 31% of these exposures (38% if general provisions are accounted for).

Regardless of the various challenges that the banking sector is currently fac-ing, the most worrying from an economic

point of view as a whole is its inability to fulfil its basic function of financing economic growth. With the intensification of the

crisis there was a sharp contraction of the annual credit growth rate, which became negative due to the prolonged crisis. The

type of credit that experienced a greater contraction during the crisis was the credit to companies. Credit to households has also

declined, albeit to a lesser extent. With regard to bank credit to companies, it presents a clear pro-cyclical behaviour. In early

2008, credit to companies in Spain grew at annual rates exceeding 30% (twice the eurozone average). In late 2009, the growth of

credit to companies became negative, reaching its maximum in the first half of 2010 (-4.2%). Since that time, it has remained

virtually stagnant. The construction industry was the sector most affected by the crisis, showing a downward trend since late

2008, which reached the two digits by mid-2009. In the case of households’ credit the rates falling has been less pronounced and,

since early 2009 they have remained close to zero.

Tensions experienced during the last year and a half in the eurozone’s financial markets, resulting from the sovereign

debt crisis, brought about an increase in the state’s and the banks’ financing costs, and has made access to markets more

difficult for the latter. Risk premiums from resident issuers have been increasing. Last August, they went slightly beyond the 400

basic points. The risk premiums remain high and subject to the fluctuating market perception on the situation of public finances

and the macroeconomic framework of various countries, as well as the solvency of their financial systems in a context of high

volatility. In this context, banks’ activity is subject to pressures of a different nature. Firstly, because the sovereign debt crisis

has resulted in a significant increase in financing costs for these entities, and secondly, because they continue to bear the losses

arising from the deterioration of real estate assets, due to the excessive exposure of its risks’ portfolio to a sector where the

process of price adjustment has not yet been completed.

 [Last Name] 79

The Two Big Problems

3.1 The Huge Debt Growth

In the medium term, the most important problem of the Spanish economy is not the public debt’s size, as in other

European economies. The main source of problems is the high private debt, resulting from a historically high amount of liabilities

of businesses and households. In Spain, since 2004, the increase in the private sector’s debt was five times higher than in the

eurozone. Credit growth rates to the private sector followed an upward trend until 2006, when the annual growth rate

approached 30%, remaining at levels higher than the nominal growth in GDP until the end of 2008.

The Spanish public debt was at the end of 2010 equivalent to 60.1% of the GDP, 25 percentage points lower than the

eurozone’s average (85.1%), on the other hand private debt represented around 224% of the product. Households’ debt amounts

to €902 billion (~ 85% of the GDP), while firms’ debt reaches €1.477 billion (~ 139% of the GDP). Since 2004, the debt’s total

increased by over 71% since in that year it was only of €1,800 billion compared to the €3,085 billion in late 2010.

The private sector’s indebtedness has been mainly growing during the pre-crisis period. In the 2004-2007 period, the

credit’s average annual growth in Spain was of 21.8%, while in the eurozone it was only of 8.9%. This particular credit evolution

can be explained by the strong growth in housing loans to households and especially by the excessive credit growth for

productive activities, in particular, for construction and real estate activities. The remaining productive sectors registered credit

increases in line with the eurozone.

The economic recession caused a significant drop in households and firms’ incomes that refrain private debt growth.

Simultaneously, economy’s private sectors began a process of debt reduction, albeit at an extremely reduced pace. Since its

peak in 2008, the debt ratio of non-financial firms fell by only 1%, while the decrease in the ratio of households was even more

modest.

The pressure that markets exert on the Spanish economy, since the beginning of 2010, results in part from the fact that

a very significant proportion of public and private debt is computed as external debt. In late 2010, for example, in the case of

public debt, foreign investors owned almost half (47%) of the Spanish sovereign debt. This is a differential aspect of the Spanish

economy over other highly indebted economies such as Italy, Belgium or Japan, since in these cases the debt is mostly financed

by domestic savings.

In late 2010, the Spanish economy’s external debt was of €1,740 billion, equivalent to 164% of the GDP. In March of

that year it had reached its historical peak at €1,790 billion, representing 167% of the GDP. The distinctive feature of the Spanish

external debt is not its relative size; the most worrying aspect of its recent evolution is its rapid growth, since between 2002 and

2010 it nearly tripled, rising from €600 billion to €1,740 billion today. During this period, two thirds of the funds used to finance

the Spanish economy’s growth came from abroad.

The deleveraging of the private sector should intensify in the coming years. The best way to further this process is

through recovering income, and that is why all policies aimed at improving the capacity to generate economic growth are the
 [Last Name] 80

best contribution to complete it successfully. Otherwise, debt will continue to be a burden on spending, especially in an

environment where interest rates may rise again in response to inflationary pressures in the core countries of the eurozone.

The structural deleveraging of the private sector is extremely important for the banking sector, since it reduces their

financing needs. According to the Bank of America Merrill Lynch, each basic point

reduction in leverage implies a decrease of €15 billion in the financing needs of Spanish banking (Centro del Sector

Financiero PwC and IE Business School 2011). Thus, a credits’ reduction and a zero growth in deposits would reduce the financing

needs of Spanish banking by €37 billion.

3.2 Unemployment’s Uncontrolled Growth

The crisis’ impact on employment in the Spanish economy was devastating. Since the crisis began until the end of

2010, the number of people unemployed increased by 152%, up to 4,632,000 workers (20.1% of the workforce), of which 4 out of

10 are long-term unemployed persons, 1 out of 4 have a temporary contract, 1 out of 3 are less than 29 years old and 6 out of 10

are low-skilled. The current number of unemployed people is higher than in early 1994 (just over 3.5 million), when the

unemployment rate peaked in the time series (24% of the workforce). These data reveal that there are institutional aspects in

the Spanish labour market that give rise to structural unemployment, that unemployment particularly affects well-defined groups

and that the problem of high unemployment is not a new phenomenon, probably because some of its main causes have been

persisting for decades.

The current high unemployment rate has its origin in factors of a different na-ture. The first factor is the impact of the

housing crisis on employment levels. In the years of greater intensity, between 2008 and 2010, job losses in the Spanish

construction sector (including real estate) were higher than 36% (Carlos Alvarez Aledo 2011). Job losses in construction were

higher only in the Baltic countries and Ireland (around 50%), whereas in other countries affected by housing crisis, such as the

United States and Denmark, were considerably lower (around 25%). The reduction in the employment’s percentage is amplified in

the Spanish case by the relatively high weight that employment in the construction sector had in the employment’s total at the
 [Last Name] 81

beginning of the crisis (13.2% versus 8.4% in the eurozone). To this loss we must add those of the construction activities’

industrial and service suppliers.

Another factor that justifies the high unemployment growth since the beginning of the crisis is the sharp appreciation

of the real exchange rate observed from

 [Last Name] 82

With the RBI App

Extract Time Series Data

Show Gross Liabilities between the US and the UK

The Key issue was that the cross-border liabilities between the counties was extremely high, usually the US is the

Lender to other nations in this case the US was the borrower in mass quantities. Gross Liabilities for the US was high.

Plot the Capital of the Banking System

Tier 1 Capital went close to Zero (Show on the Graph)

With the RBI App and the Counties show PRE-CRISES (2006). / CRISES (2007/2008) / POST CRISES 2015 – with this we

can assess contagion on 3 different Levels.

Use RBI App to construct a tired structure

RED is the borrower it should be in the Middle ‘

K in & K out mean borrowing lending

Eigen Vector

What does Eigen Value Mean Left & Right?

Some nodes are more systemically important than others.

Others are on the periphery? Left Systemically important

Right System Vulnerability

Add Time series Stuff from ORB.

Instability of the EuroZone (Network) --->> Each Student of the Quarters

Pigovian tax = is a corrective tax, it is a tax that recoups the marginal external cost.

In gragh theory,

Eigen Vector connectivity (is a measure of the influence of a node in a network).

In gragh theory
 [Last Name] 83

Clustering co efficient is a measure of the degree to which nodes in graph tend to cluster together.

A high score eigenvector score means that a node is connected to too many nodes who themselves have high scores/..

You can use adjacency matrix to find the Eigen vector connectivity.

Basel 2 is recommendations of Banking Laws and Regulations issued by the Basel Committee on Banking Supervisor.

Network Measures of Centrality

Degree: the number of Edges connected to the Node

Between : extent to which a particular node lies on the shortest path between other nodes.

Closeness: the average of the shortest distances to all other nodes in the graph

Eigen Vector: a measure of the extent to which a node is connected to influential other nodes. I.e. Google Page Rank

uses this measure.

2007 financial crises highlighted the need to model interconnectedness of financial firms,

International financial integration has led to cross border banking to expand dramatically.
[Last Name] 84
 [Last Name] 85

Notes and suggestions

Plot the

o Analyse quarter by quarter. And every three years – in case year by year

changes are too small

 Compare the very first with the last quarter for a snapshot compare the

trend for the us ; uk and rest of the world: how much they

borrowed

lent

Derive the leontief matrix and Ghosh matrix

Play around with the parameters in the network diagram to change the tiering of the

network

Structure according to pre-crisis , crisis and post-crisis

Who's the superspreader

o Left eigen value– vulnerable

 Right – systemically important

[Last Name] 86
 [Last Name] 87

This above is the Network Diagram for 2008

This above is the Network Diagram for 2005

This above is the Network Diagram for 2013