|
Econometrics literally
means 'economic
measurement'. It is a combination
of
mathematical economics and
statistics.
The two main purposes of econometrics
are to give
empirical content to economic theory
and to subject economic theory to potentially
falsifying tests. For example, economic
theory may predict that a given demand
curve should slope down. Econometric
estimates can either verify or falsify
that prediction, and shed light on the
magnitude of the effect.
The most important statistical method
in econometrics is
regression analysis. For an overview
of a linear implementation of this framework,
see
linear regression. Regression methods
are important in econometrics because
economists typically cannot use controlled
experiments. Observational data are
often subject to
lurking variable and other problems
which must be addressed statistically
using regression models. Econometricians
often seek illuminating
natural experiments in the absence
of evidence from controlled experiments.
Econometric analysis is divided into
time-series analysis and
cross-sectional analysis. Time-series
analysis examines variables over time,
such as the effects of population growth
on a nation's GDP. Cross-sectional analysis
examines the relationship between different
variables at a point in time; for instance,
the relationship between individuals'
income and food expenditures. When time-series
analysis and cross-sectional analysis
are conducted simultaneously on the
same
sample, it is called
panel analysis. If the sample is
different each time, it is called repeated
cross section data. Multi-dimensional
panel data analysis is conducted on
data sets that have more than two dimensions.
For example, some forecast data sets
provide forecasts for multiple target
periods, conducted by multiple forecasters,
and made at multiple horizons. The three
dimensions provide more information
than can be gleaned from two dimensional
panel data sets.
Econometric analysis may also be classified
on the basis of the number of relationships
modelled.
Single equation methods model a
single variable (the
dependent variable) as a function
of one or more explanatory variables.
In many econometric contexts such single
equation methods may not be able to
recover estimates of causal relationships
because either the dependent variable
causes changes in one of the explanatory
variables or because variables not included
in the model cause both the dependent
and at least one of the independent
variables.
Simultaneous equation methods have
been developed as one means of addressing
these problems. Many of these methods
use variants of
instrumental variables models to
make estimates.
Much larger
econometric models are used in an
attempt to explain or predict the behavior
of national economies.
A simple example of a relationship in
econometrics is:
-
wage = constant + (rate of return
to education) * education + random
error
In this equation, a person's wage is
a linear function of the number of years
of education he has. The econometric
goal is to estimate the expected change
in wages a person would receive if she
obtained one more year of education.
If the researcher could randomly assign
people to differing levels of education,
the correlation between education and
wages would reveal the causal effect
of education on wages. But it is not
feasible to conduct such experiments.
Instead the econometrician only observes
how many years of education people obtain,
and the wages they receive. The correlation
between wages and education reflects
both the effect of education on wages
and unobserved variables which may affect
both outcomes. For example, more intelligent
people may tend to obtain more education
and may also earn more at any level
of education than less intelligent people.
Econometric methods could be used to
overcome these problems and estimate
the underlying causal effect of education
on wages.
|
