ES5611 Introduction to Econometrics

1
ES5611Introduction to Econometrics

• Introductory remarks
• What is Econometrics?
• What is this course like?

Slides by Ken Clark, adapted from P. Anderson,
2004.
2
Introductory Remarks

• Lecturer Ken Clark, N.5.6, Dover Street
• email ken.clark_at_man.ac.uk
• http//www.ses.man.ac.uk/clark/es561/
• Office Hours Wednesday 10-12

3
What is Econometrics?

• Some definitions
• Why study econometrics?
• The Econometric Process
• Example wages and productivity
• Types of data
• Causality
• Examples

4
Definition (outputs)

• Wooldridge statistical methods for estimating
  economic relationships, testing economic
  theories, and evaluating and implementing
  government and business policy (p.1).
• Ramanathan (1) estimating economic
  relationships, (2) confronting economic theory
  with facts and testing hypotheses involving
  economic behaviour, and (3) forecasting the
  behaviour of economic variables

5
Definition (outputs)

• Estimation/Measurement
• Inference/Hypothesis testing
• Forecasting/Prediction
• Evaluation

6
Definition (inputs)

• Ingredients of an econometric exercise
• Economic Theory
• Mathematics
• Statistical Theory
• Data
• Computing Power
• Interpretation/Economic Knowledge/Common
  Sense.

7
Why study Econometrics?

• It is rare in economics to have experimental
  data
• Econometrics uses nonexperimental, or
  observational, data to draw conclusions about the
  real world
• This enables us to apply economic theory to real
  world data

8
Why study Econometrics?

• Econometrics can test and refine economic theory
• Theory may be ambiguous about impact of a policy
  change econometrics can evaluate the policy
  program
• Econometric analysis is useful to decision
  makers.

9
Econometrics as a process
10
Example wages and productivity

• wagef(educ, exper, training)
• deterministic economic model
• wageß0 ß1educ ß2 exper
• ß3training u
• econometric (statistical model)
• u random error term
• ßs parameters.

11
Example

• use computer to estimate the parameters
• what is the ceteris paribus effect of educ on
  the wage? what are the returns to education
• and to test hypotheses (inference)
• is ß30? (a more subtle question than it seems)
• could also forecast wages for workers with given
  characteristics (e.g. to predict how much an
  accident victim would have earned in future).

12
Types of Data Cross Sectional

• Cross-sectional data are usually a random sample
  from some population
• Each observation is a new individual, firm,
  household, etc. with information at a point in
  time
• If the data are not a random sample, we have a
  sample-selection problem

13
Types of Data Time Series

• Time series data has a separate observation for
  each time period e.g. stock prices
• Since not a random sample, different problems to
  consider
• Trends and seasonality will be important

14
Types of Data Panel

• Can follow the same random individual
  observations over time known as panel data or
  longitudinal data
• ES5622 covers this.

15
The Question of Causality

• Simply establishing a relationship between
  variables is rarely sufficient in economics
• Want to the effect to be considered causal
• If weve truly controlled for enough other
  variables, then the estimated ceteris paribus
  effect can often be considered to be causal
• Can be difficult to establish causality problem
  of endogeneity

16
Example 1 Returns to Education

• A model of human capital investment implies
  getting more education should lead to higher
  earnings
• A simple econometric model

• Is ß1 truly the returns to education?
• Does more education cause higher earnings?

17
Example 1 (continued)

• Problem Suppose more able people have (a)
  higher earnings and (b) more education.
• Observed relationship between education and
  earnings could actually reflect impact of ability
  model gives the wrong answer
• Technically E(ueduc) ? E(u). Education and
  error are correlated. Endogeneity.

18
Example 1 (continued)

• Two situations where were OK
• No relationship between education and ability
• We observe ability and include it in the model
  (control for ability)

• Frequently one or both does not hold.

19
Example 2 Policing and Crime

• Do more police reduce crime?
• Does ß1 reflect causal influence of police on
  crime?
• But cities with high crime rates may employ more
  police

20
Causality Roundup

• The key question is Have enough other factors
  been held fixed to make a case for causality?
• When carefully applied, econometric methods can
  simulate a ceteris paribus experiment
• (Wooldridge, p. 14).

21
What is this course like?

• 10 two hour lectures (Mondays 2-4pm)
• 9/10 Examples Classes (START 11/10/04)
• allocation and details to follow
• exercise sheets to follow
• students should attempt exercises prior to
  examples class
• Assessment Two hour examination in January 2005.
  
• Note course changed substantially in 03/04.

22
What is this course like?

• Computing we will use Stata and Microfit.
• Microfit available from all networked PCs
• Stata available in O.G.10, Dover St. and
  Williamson 3.59.
• Find under Programs/Faculty/FSSL/
• Stata is called Intercooled Stata8
• Microfit is called Microfit 4 on the Econometric
  Software sub menu

23
What is this course like?

• Stata Tutorial There will be a one hour
  Introduction to Stata tutorial in Room O.G.10
  THIS WEEK (w/b 27/09/04).
• You should sign up for this tutorial on the
  sheets on the wall outside Room N.5.3, Dover St.)
  Sign up for one only and note there is a maximum
  of 20 places in each session.

24
What is this course like?

• Reading All students should purchase 
• Wooldridge, Jeff, Introductory Econometrics A
  Modern Approach, Second Edition, South-Western
  Thomson, 2003.
• Available in Blackwells, Precinct Centre.

25
Course Objectives

•  An introductory econometrics course
• Assumes no previous knowledge of econometrics.
• On completion of the course students should be
  able to understand the results of econometric
  procedures which they read about in applied
  economics research and to use basic econometric
  techniques in their own work.

26
Course Objectives

•  Students who have studied introductory
  econometrics before will benefit from taking
  either ES5521 Time Series Econometrics or ES5501
  Advanced Econometric Theory rather than ES5611.
  Students should consult the lecturers of these
  courses or their course director if unsure which
  to choose.

27
Student Background

•  Assume some familiarity with random variables,
  population vs. sample, expectation, correlation,
  independence, variance, sampling, estimation,
  hypothesis testing.
• Important material from pre-session maths
  course linear functions, differential calculus,

28
Warning

•  There will be some revision of key concepts next
  week but it is YOUR responsibility to ensure your
  background is sufficient.
• Sample supplementary reference on introductory
  statistics
• Wonnacott and Wonnacott, Introductory Statistics
  for Business and Economics, 4th ed., 1990, Wiley.

29
Advice

•  Econometrics is a mixture of many ingredients
  maths, stats, economics, computing.
• Break the problem down.
• Keep your eye on the

30
(No Transcript)
31
Topics

•  See course outline
• based around estimation and testing of multiple
  regression model
• follows Wooldridge closely.
• Next week review of key concepts
• Reading Wooldridge, Appendices A, B, C1, C2,
  C5, C6.