Upload
crucozbaur
View
7
Download
0
Embed Size (px)
DESCRIPTION
econometrics
Citation preview
1. Cross sectional dependence, times series dependence, panel structure
Memory, no memory, short memory, long memory, white noise process, random walk, law of large
numbers, central limit theorem, consistency, information asymmetry, adverse selection, moral hazard
AR process, MA process, ARMA process
Contagion: when is full connectedness is desirable/ undesirable?
Summary:
1.
Cross sectional dependence refers to the relationship/dependence between units at a given point in
time. For example the dependence between countries A, B, C in period t.
Times series dependence refers to the dependence between one unit and more time periods, or, in
other words the behavior of one unit during time, over a particular period. Example: A over period t,
t+1, t+2, ... t+n.
Panel structure is a cross sectional dependence over time, meaning, dependence between more units
over a period of time. Ex: A, B, C over period t, t+1, … t+n.
2.
Memory – (dependence over time) another way of saying to which extend the past is useful in the
future. Ex: price has memory means that the price from the past is useful in predicting the price in the
future.
No memory (white noise) - there is no dependence over time, the past doesn`t help in the future.
Short memory –
Long memory – everything is important
White noise process -refers to the fact that there is no correlation between the random variables.
Random walk – the random variables have the same distribution and they don`t depend one on each
other, so the past cannot be used in order to predict the future. Ex: the future price of a stock cannot be
determined with the help of the past price. “The value today should not be different from yesterday If it
is different, there must be new information”(Lecture)
Law of large numbers – as the sample of the population grows, its mean will get closer to the population
average.
Central limit theorem – the average of the sample tends to be normally distributed. The sum of the
variables will have a normal distribution, regardless of how variables change.
Importance-
Consistency – if �̂� converges to β, it means that �̂� is a consistent estimator. If it converges to something
else, then is an inconsistent estimator.
Information asymmetry – a situation in which one party knows more than the other, or has more and
better information about a common interest, so that the one that has more information can benefit
from the one that doesn`t. Ex: I sell you a non-functional sewing machine. I benefit from your lack of
knowledge. I generate unnecessary loss on your side. I am better off and you are worse off. Can take
two forms: adverse selection and moral hazard.
Adverse selection – the case in which one person is insured and he/she behave differently. It is prior to
the transaction, so sometimes it might prevent the transaction from occurring.
Moral hazard -
AR process – current values are influenced by the past values. There exists memory that is important for
explaining the future. (1) the past value is important, (2) the past two values are important
MA process – refers to the fact that the shock generated cannot be absorbed in one single period, so it
has to be distributed on more periods. For example a hock that occurred in period t-3 will affect t-2 but
also t-1. The shock has to be distributed between a wider time period
ARMA process – the past information influences the present and, by splitting the shocks over multiple
periods. Takes the characteristic of AR and MA process.
3. I would say that full connectedness is desirable in the case in which there is full transparency and the
information is equally known by all the participants of the network. This way, the entities will trust each
other and the level of the confidence is high. ( I trust you that you don`t have today the money that you
owe me, but I know for sure that tomorrow you will pay me back.) the risk burden is shared by all the
participants. Helps weak entities to stay in.
I would say that full connectedness is not desirable in the case is which there does not exist full
transparency, and the information is not known by everyone. The entities will not trust each other. (You
tell me that you will pay me back tomorrow but you don`t, so I will never borrow you money again. ) The
risk burden affects only one particular entity that eventually will collapse.