Methodology
Author | Maccaferri, Sara; Bellia, Mario |
Pages | 4-15 |
4
1 Introduction
In 2017 the Financial Stability Board (FSB) established a network of experts (Evaluation Working
Group, EWG) to examine the s et of regulatory reforms addressing the too -big-to-fail (TBTF)
expe ctations. In particular, the goal of the EWG has been to assess to what extent TBTF reforms
are reducing the systemic and moral hazard risks associated with (Global) Systemically Important
Banks ((G-) SIBs) and to analyse the broader effects (positive or negative) of these reforms on
the financial system.
The working group has divided its work into four subgroups: (1) implicit funding subsidies; (2)
bank behavior and structure; (3) res olution and re solvability; and (4) bro ader effects of TBTF
reforms .
The EWG is mad e of experts coming from worldwide countries and international organizations.
The Europ ean Commission (EC) is part of this working group and DG FISMA asked the JRC to
supp ort the activities of the subgroup on resolution and resolvability by undertaking event studies
to me asure the credibility of the set o f fin ancial regulatory reforms. In pa rticular, DG FISMA
reque sted the JRC to replicate the methodology presented in the 2017 paper Expecting Bail-in?
Eviden ce from European Banks by A. Schä fer, I. Schnabel and B. Weder di Mauro (ref erred to
Schäfer et. al. ( 2017) hereinafter) and to extend the analysis considering market reactions to
more rec ent events.
The purpose of the study is to estimate the short term reaction of stock and Credit Default Swap
(CDS) prices of European banks to various events a nd announcements, such as b ail-ins,
recap italizations, and the proposal and final agreement of the EU reform package of prudential
and reso lution rules in banking (“banking package”). The paper replicates the work of Schfer et
al (2017), focusing on bail-in events, but also considers and assesses the impact of a broader
and m ore recent se t of even ts an d ann ouncements (e.g. res olution of Sp anish bank Banco
Popula r).
2 Methodology
The e mpirical a nalysis applied in this report closely follows the one ap plied in Sc häfer et al
(2016a ), Schäfer et al (2016b) and Schäfer et al (2017), and consists of estimating a system of
equa tions with a technique called Seemingly Unrelated Regression (SUR), introduced by Zellner
(1962). The advantage of this technique is to assume that the error terms are correlated across
equa tions or, in other terms, the errors associated with the dependent variables are correlated.
Following Zellner (1962), the general SUR model consists of m=1 … M linear regression equations
for j=1 … N banks. The mth equation for the bank i can be written as:
Where is a ve ctor of observations for the jth dependent variable, a m atrix of
obse rvations for the independent (explanatory) variables, is a vector o f regression
coef ficient and a vector or error term, with zero mean. The full SUR model can be written
as:
The mode l can be es timated via FGLS or Maximum Likelihood. The same coefficients can be
obtained estimating a system of simultaneous equations, or a multivariate regression. In some
case s, the covariance matrix of error terms appears to be singular using the SUR method, thus
the analysis is run using a multivariate regression framework. Once the model is estimated, it is
poss ible to impose restrictions on the coefficient, or run comparisons across groups of banks.
However, it worth mentioning that since the explanatory variables are the same across the N
banks in the sample, the estimated coefficients of the SUR are identical compared to OLS. One
To continue reading
Request your trial