Tip:
Highlight text to annotate it
X
In this video I want to talk about what actually do we mean
by econometrics. So econometrics
is in general a statistical tool set which helps us to evaluate
some sort of relationship of interest.
An example might be we're interested in, for individuals,
what is the effect of an individual's level of education
on the average wage which that individual
might expect to obtain. So if an individual's level of education increases
we might expect that the level of wages
which an individual obtains on average might increase.
So if I was to plot a graph of the
level of education of individuals on the
x-axis, against the wages
which a group of individuals have obtained on the y-axis,
then we might hope to see some sort of positive correlation between these two
variables.
That's not to say that is necessarily a causal relationship
only that there is some sort of positive relationship between these two variables.
Econometrics help us to quantify
this degree of correlation by in a sense drawing
a line through the centre of all those points.
And by drawing a line through the centre of all those points
we are hoping to capture what is the average effect
of education on wages. So
on average an individual who has 10 years of education,
might expect to obtain a wage which is let's say
four hundred dollars. Whereas an individual
perhaps if they had 11 years worth of education
might expect back their wages to go up
by a hundred dollars. So they now earn five hundred dollars.
Well econometrics is a toolset
for finding out what the strength of this relationship is. So
how much do wages actually go up by. And
this type of relationship here where we're concerned with the relationship
for individual people or individual firms is
the subject of microeconometrics.
And it's called microeconometrics for analogy with
microeconomics. Another sort of microeconometric
relationship we might be interested in might be, 'What is the
effect
of TV advertising on
a company's level of sales. So
if I was to draw a graph of a
company's level of sales over time
then we might have something which looks something like this. If there is some sort of
seasonality.
Perhaps this is coffee sales or ice cream sales.
And we might be interested in do these peaks which we observe
in the data - are they caused by TV advertising?
And TV advertising might look something
like these bars that I have drawn below. So
econometrics is a way of understanding for this time series data
does this TV advertising here
cause sales to go up? And similarly does this TV
advertising here cause sales to go up. So
this is slightly different to the previous example in that we are dealing
with what we call time series data, whereas the original data was what we
call
cross-sectional data. But it's still actually what we call Microeconometric
data because we are dealing with
data for a particular firm. Another type
of econometrics is the subject of
macroeconometrics and
macroeconometrics as its name suggests
is to deal with macro relationships. So
an example here might be what is the effect
of interest rate falls on inflation.
So traditional economic theory suggests
that if the interest rate falls then
the inflation rate should increase because of
increases in aggregate demand. So econometrics is a way
of quantifying that particular relationship. So all
of these relationships are variations
on the same sort of theme which we see time and time again in
econometrics. So the idea with econometrics
is that there is some sort of population or in the case
of time series data we actually call this a data generating process
but you can kind of think about it for now at least in terms of a population.
And within this population there is some sort of
true relationship between the variables which are interested in.
So there might be some sort of true relationship between
an individual's level of wages and let's say
their level of education. So
in this example here beta
quantifies the effects of one year of education
on an individual's level of average wages.
But there are other factors which also determine wages,
which we group together in our population
error 'u', which are all these other idiosyncratic factors which affect
an individual's level of
wages. So an example might be 'where does the individual
live?' Are they based in OECD countries for example?
What are their interests? Are they interested in
pursuing a career where they earn lots of money like an investment banker
or are they interested in going into the civil service for example?
Those are the sorts of things which are contained within the population
error 'u'. But the idea with econometrics
is that we don't actually have the entirety of the population
in terms of the dataset. In general we only have
a sample from that population, because
we deal with non-experimental data. So we just get or we
hope to get a randomised sample from the population.
The idea with econometrics is that by applying some sort of
tool to this sample
dataset we hope to estimate
these parameters. So in this case here we hope to come up with
estimates beta and perhaps alpha.
So that's what we hope to get out by our econometric
analysis on this particular sample. But in general
these sample estimates will differ from the actual population true values
because of the fact that our sample isn't necessarily
perfectly indicative of what's going on in the population,
and we actually have a name for this effect which we call
sampling error. And the idea with econometrics
is we want to use tools which
actually reduce this sampling error
as much as possible. So they make our estimates which we
produce on the sample as close to the true population
values as possible. And also
on doing so that will allow us to make some sort of
inference about what's going on in the population. So
is the true effect of education on wages
let's say a hundred dollars per extra year of
education? In the next video I want to talk about
the difference between econometrics and hard science
where in hard science we can actually carry out experiments when
let's say we which only increase an individual's level of
education holding all other factors constant.
In reality we can't actually do that, so econometrics is
a way of dealing with non-experimentally
generated data. I hope to see you in the next video.