Network Models

Question 1

Role of connectedness in market risk assessment

Answer 1

separates risk of a ptf from risk of its components -> likelohood of extreme mkt movements (most assets moving in same direction) depends on connectedness, see: counterparty risk)

Question 2

Spillover/cross variance definition from variance decomposition

Answer 2

Spillover are the fractions of 1-ste-ahead error variances in forecasting variable i due to shock to the other variables j=1,…N

Question 3

Diebold Yilmaz limitations

Answer 3

1) VD relies on orthogonal innovations -> relies on cholesky identification -> dependent on ordering. Solution: use generalized VAR framework where VAR innovations are comtemporaneously correlated.
2) Only estimates total spillovers, but NOT directional net spillovers

Question 4

GVAR/KPPS approach

Answer 4

Does not attempt to orthogonalize shocks, instead allows correlated shocks but accounts for them using the historically observed corelation patterns of errors UNDER NORMALITY.

-> we can learn about direction of volatility spillovers and net volatility spillover

Question 5

How to read network table

Answer 5

Entire row x_1: variance FROM others

entire column x_1: variance TO others

Question 6

Erdos-Renyi networks

Answer 6

independent bernoulli trials (-> degree distribution is binomial). network diameter grows only as a function of ln(N). poor descriptions of real-world networks