You are on page 1of 35

125.

785 Research Methods in


Finance

Seminar One
Monday 17 July

1
Honest politicians make the other
95% look bad
-- Mark Twain

2
Overview

 Administrative Issues
– Timetable
– Labs
– Textbook
– Assessment
 Aims and Objectives
 Introduction
– Eviews
 Readings: Chapters 1-3, Chapter 16 optional

3
Administration

 The general format will be for 1st half


– A 2 hour seminar
– A 2 hour lab in either CLQB4 or IIMS5/6
– Finish approximately 7pm.
Textbook is Studenmund
Using Econometrics: A Practical Guide. 5th Ed.
– 4th Edition can also be used.

4
Assessment

 1 Assignment Due September 1: 20%


 Quiz 1: 31 July (10%)
 Quiz 2: 14 August (10%)
 Quiz 3: To Be Advised (10%)
– Probably 28 August

5
Web Support

 Web CT should be available for students


 In the interim, the following website also will
have material:
 http://www.massey.ac.nz/~bjmoyle/mu/teach.html

6
Computer Labs

 Your user-name is your student ID


 Your password is your 4 digit pin number
 You will benefit from bringing
– A floppy disk OR
– A USB drive (preferred)
 We will use Eviews for this section of the
course

7
Learning Objectives

 Develop your skills at estimating economic


relationships.
– This skill cannot be memorised from a textbook or
lectures
– The textbook and seminars are to assist and
guide you.
 Increase your familiarity with statistical
software.

8
Learning Outcome

 To give you a sufficient background that you


can:
– Attempt a research project with some of the skills
you have learned; or
– Can progress on to advanced techniques used in
financial econometrics without difficulty.
 It is impossible to teach you all the tools you
might use in the constraints of this paper.

9
The Unreliability of Textbooks

 This is an applied paper, not a theory paper.


– Every data set you model, will have ‘different’
problems present.
– It is impossible to memorise all the permutations
of problems that you will encounter.
– Skilled researchers are those with good problem-
solving strategies, not recall of textbook stylised
facts.
– Most of this skill must be developed with practical
work.
10
Introduction to Research

 A research project involves three stages


– Choosing a Topic
– Analysis
– Writing Report

11
Choosing a Topic

 Ideally choose something you are interested


in for motivation
 Make sure you can get enough data
 Make sure there is some substance to topic
– Not purely descriptive
– Not tautological (so obvious to be uninteresting).
 E.g. does an increase in the number of bidders raise
prices?

12
Analysis

 Develop your theoretical model first


– May involve reading literature
 Specify the model
– What causes what?
 Hypothesise the effects you expect
– This must be done before you run any models
 Collect the data

13
Analysis 2

 Estimate the Equation


– This should take the least amount of effort
 Document the results
– There must be enough information given, that
someone else could replicate your results.

14
Report Writing

 This is an important step


– The purpose of research is often to generate
information for a decision-maker.
– Hopefully, a manager or policy-maker could read
your report and learn something new.
 A box of computer printouts, neither informs
nor impresses.

15
Report Writing 2

 A report should not gloss over or ignore


results that you did not expect.
– It is a common mistake to not discuss results that
contradict your prior beliefs.
 Keeping a research journal can assist
– Record your hypotheses, regression results,
statistical tests etc.

16
Practical Advice

 Use common sense and economic theory


– E.g. Real variables should not be matched with
nominal.
 Ask the right questions
– Sometimes regression problems are a
consequence of the wrong specification
 Know the context
– Understand the problem, not just the statistics

17
Practical Advice 2

 Inspect the Data


– Graphs or summary statistics can reveal missing
variables, outliers or other anomalies.
 Keep it sensibly simple
– Complexity is not ‘good’ for its own sake
– Consider Occam’s Razor.
 Look long and hard at your results
– Does it make sense? You have to explain this to
others.

18
Practical Guide 3

 Practice data-mining with care


– Exhaustive experimentation to ‘get’ the ‘right’
results shows you’re biased…
 Be prepared to compromise
– Trying to find the perfect model will drive you
crazy.
– Real data tends to throw up intractable problems.

19
Practical Guide 4

 Do not confuse statistical significance with


meaningful magnitude.
– Trivial variables may have very small effects, but
are highly significant.
– It is tempting to use statistical significance as a
measure of a model’s performance.
 Report a sensitivity analysis
– Do results vary of you change the sample period
etc?

20
Basic Stats

 We use statistical tools  We signal our


in this paper uncertainty about the
– But it is not a course in parameter with a type
statistics of ‘spread’.
 We will estimate the  E.g.
value of many – Variance
parameters – Standard Deviation
 E.g.  These uncertainty
– A mean (average) measures form the
– A regression coefficient basis of statistical tests.

21
Recap

 The main difference between statistics and


other maths, is answers will have 2
dimensions
 In normal algebra, variables combine to
produce an explicit solution.
 In statistics, we think in 2 dimensions
– What we think the value of something is
– How confident we are in that estimate

22
Correlation

 Quantifies the  Correlations do not


relationship between 2 prove causality
variables.
 Correlations can be
 -1 ≤ r ≤ 1 shown graphically
 Correlations imply
– Relationships or
associations
– General tendencies

23
Correlation between GDP and G

25000
CGDP

20000

15000
CGDP

10000

5000

0
0 10 20 30 40 50 60 70
CG
24
Regression

 Suppose we wanted to  Regression Models can


– Forecast prices for an be used for several
asset. purposes.
– Determine causes of – Forecasting
unemployment – Testing hypotheses
 A regression “models” – Detecting influential
finance or economic variables
data

25
Simple OLS Model

 We begin with the  The line ‘approximates’


Ordinary Least the relationship
Squares (OLS) between the two
regression model variables
 This generates a  The variables are
‘straight line’ between 2 – Dependent (Y)
variables. – Independent or
explanatory (X).

26
Regression Example

 Y is GDP per capita


 X is Govt spending
 We ‘explain’ Y in terms
of X
 We can estimate Y if
we know
– Intercept of line- constant
– Slope of line

27
Note on Regression

 Researchers can (potentially) use many


different regression techniques.
 OLS is a convenient starting point.
– But not all regression models use least-squares
methods.
– If certain assumptions are met, OLS is the best
method to use.

28
Intuition

 OLS is based on Cartesian Geometry.


– The line we estimate (with intercept and slope),
comes closer to all the observations than any
other line.
– We minimise the (sum of) distance between the
line and the observations (squared). This is an
idea that draws on geometry.
– As a minimisation problem, it can be readily
solved with differential calculus.

29
Eviews

 Eviews is a popular (and powerful)


econometrics program.
– It is the software most students use for their
graduate research reports

30
Where menu to estimate
equations is located

31
Equation Estimation Menu

32
Dependent Explanatory
Variable Variable

Constant

33
Output

34
Readings

 Studenmund
– Chapter 16 Statistical Principles
– Chapter 1-3
 WebCT
– Guide to Eviews- Introduction

35

You might also like