+
Bristol Business School
Academic Year: 09/10
Assessment Period: August
Assessment Type: Referral Coursework
Module Leader: Tony Flegg
Module Number: UMECRW-20-3
Module Name: Econometrics
Word Limit: ….
Coursework Submission Date and Time:
Assignments are to be submitted by 2pm Monday 16 August 2010 at the
assignment boxes near Cribs coffee shop (2B corridor). Please be aware
that there is NO 24hr or 10 day window this year.
If you wish to post your assignment, we recommend that you post it via
Recorded Delivery and obtain proof of the date and time of posting. Your
work must be in the post by the deadline.
Deadline: Monday 16 August 2010 14:00
Assignment Instruction:
UMECRW-20-3
ECONOMETRICS REFERRED ASSIGNMENT AUGUST 2010: Part I (60
marks)
Introduction
This part of the referred coursework is concerned with using regression
analysis to analyse the determinants of demand for butter and margarine
in Great Britain. You will be making use of a database specially
constructed for this purpose. It contains annual data for the period
1956–96 on the prices and quantities consumed of butter and margarine,
the retail price index for all food and an index of real personal disposable
income per head. Your main task is to develop an appropriate
econometric demand function for both of these commodities, using data
for the subperiod 1970–94. After examining your results in detail, you are
asked to forecast the consumption of both commodities up to 1996.
Section 1: The Estimation of Alternative Demand Functions (20 marks)
Use Microfit and the data file BUT96.FIT to construct alternative linear
demand functions for both butter and margarine, using OLS and data for
the subperiod 1970–94 (see page 3). Choose no more than four of your
regressions for purposes of discussion (two models for each
commodity). A summary of your results must be presented in tabular
form in the body of your report, with the full results of your modelling
included in an appendix. You should make full use of the available
Microfit commands to produce a correlation matrix for the estimation
period, plots of residuals, fitted and observed values of the regressand,
etc. Cross-reference the results in the appendix with those in your
summary table. In this table, show standard errors in parentheses
beneath the regression coefficients, followed by t ratios in bold type. At
the bottom of the table, report R2, R 2 , the standard error of the
regression, the D.W. statistic, Durbin's h (if applicable), along with the
four χ2 diagnostic test statistics. To avoid charges of spurious accuracy,
round all figures appropriately. Include a glossary giving precise
definitions of the variables (including units of measurement).
Provide a rationale for the inclusion of the various regressors in your
demand functions and comment briefly on the process whereby you
arrived at these particular regressions. Are the signs of the regression
coefficients in accordance with theoretical expectations?
Section 2: Goodness of Fit (5 marks)
Explain the rationale and derivation of Theil's adjusted R2 and Amemiya's
modified R2. Are they in agreement as to the best model? If not, why not?
Does the standard error of the regression give the same ranking of
models as Theil's R 2 ? If so, is this merely a coincidence?
Section 3: Other Considerations (15 marks)
Apart from goodness of fit, what additional considerations should be
borne in mind when selecting a regression equation? In the light of these
considerations, is the equation yielding the best fit in this instance still
the most satisfactory one?
Section 4: Using the Wrong Explanatory Variables (15 marks)
Carefully explain the theoretical consequences of (i) excluding a relevant
regressor and (ii) including an irrelevant one. Illustrate your answer by
referring to your regression results.
Section 5: Alternative Functional Forms (15 marks)
Re-estimate your preferred linear demand function in log-linear,
exponential and hyperbolic forms. Present your results in tabular form
as in Section 1. Obtain estimates of the price and income elasticities of
demand in 1980 and 1994 for each non-linear model and compare the
results with the corresponding figures for your best linear model. Which
model, if any, do you regard as the most satisfactory for each
commodity?
Section 6: Interpretation of Regression Coefficients (10 marks)
Using, for each commodity, your preferred linear or non-linear
regression equation, carefully explain how the results can be interpreted.
Illustrate your answer diagrammatically. Comment on the statistical
reliability of the results.
Section 7: Intertemporal Stability of Coefficients (10 marks)
Now re-estimate your preferred demand functions for each commodity,
using data for two appropriate subperiods. Perform an appropriate
statistical test to see whether it is reasonable to treat the period 1970–94
as a whole. Carefully explain the rationale of the test you decide to
employ. Comment on your findings.
Section 8: Forecasting (10 marks)
Use your preferred regression models to forecast consumption of butter
and margarine in the two post-estimation years. How well do your
models perform? Try to explain any discrepancies between forecasted
and observed consumption. With hindsight, how could you have
improved your models?
Suggested References
Lecture notes and handouts plus:
Dougherty, Introduction to Econometrics, 2nd ed., pp. 60–2, 105–9, 128–
32, 149–60, 170–4, 191–3, 196–208, 219, 334–42, 346–8, 355–62
Gujarati, Essentials of Econometrics, 2nd ed., pp. 218–28, 239–50, 313–
36, 377–91, 405–24
Kennedy, A Guide to Econometrics, 4th ed., pp. 73–83, 121–4, 183–9, 221–
6, 288–92
Maddala, Introduction to Econometrics, 2nd ed., especially pp. 88–101,
161–77, 229–32, 269–80
Maddala, Introduction to Econometrics, 3rd ed., especially pp. 159–76,
228–30, 245–7, 479–88
Studenmund, Using Econometrics: A Practical Guide, 5th ed., especially
chapters 3, 69, 11
Thomas, Modern Econometrics, pp. 237–44, 260–9, 296–307, 339–48,
354–9, 361–8
Verbeek, A Guide to Modern Econometrics, 2nd ed., chapter 3
Notes of Guidance: Please see page 6 of the module handbook.
Deadline: late August 2010 (exact date not yet determined). Submit your
work via the BBS assignment boxes near Cribbs café.
Tony Flegg May 2010
APPENDIX
Names and description of variables included in BUT96.FIT
1 QB Quantity of butter
2 QM Quantity of margarine
3 PB Average yearly price of butter
4 PM Average yearly price of margarine
5 RPIAF Retail price index for all food (1980 = 100)
6 INC Index of real personal disposable income per head (1980 =
100)
7 RPB Real price of butter [(PB RPIAF) 100]
8 RPM Real price of margarine [(PM RPIAF) 100]
9 A Intercept (column of ones)
10 TIME linear trend starting in 1970
Note: Quantities are grams per person per week; prices are in pence per
kilogram.
Note: The data file can be accessed via Tony’s folder on the network and
on Blackboard under Assignments.
UMECRW-20-3
ECONOMETRICS REFERRED ASSIGNMENT AUGUST 2010: Part II (40
marks)
The Microfit file UKCON.FIT, to be found in the Microfit TUTOR datasets and in the
Assignment folder on Blackboard, contains 161 quarterly observations on consumer
expenditure and disposable income in the UK for the period 1956Q1 to 1995Q1. You
are required to investigate and model the behaviour of real consumers’ expenditure
(SEASONALLY ADJUSTED) and to provide forecasts of this expenditure for a four-
quarter period. The relevant dependent variable name in this dataset is C.
You should aim to develop optimal forms of TWO unconstrained models, optimality
being determined by the standard statistical diagnostics as well as forecasting
performance. The first is to be a linear regression model using only (as required)
time trends and lagged values of the dependent variable as possible regressors, i.e.
essentially an AR model. The second model will incorporate further behavioural
explanatory variables (as included in the dataset) in current and/or lagged form, i.e.
essentially an ARDL model. In both instances, you may wish to consider functional
transformations of some or all of the variables used and you will need to experiment
to determine the optimal model of each type.
You should not use any form of alternative estimation method e.g. Koyck or Almon
approaches.
You should write a brief summary report outlining the exploratory, modelling and
analytical work undertaken. An essential part of this work is a clear explanation of
the way in which you have developed your optimal model. You should provide a
comparative appraisal of your two optimal models, indicating which you think
performs best and for what reason(s).
Assessment will be based upon content, structure, clarity of explanation, use of
language and presentation. You should include a summary table of your results in the
text and supporting Microfit printout, plots, etc. in the appendices. ALL pages must
be numbered and cross-referencing used where necessary. The report should not
exceed 1000 words (excluding footnotes, bibliography and appendices). It is essential
that your report is independently written and represents your understanding and
perception of the models used. Be aware of the university’s rules relating to
plagiarism.
COURSEWORK HELP, ADVICE AND MORAL SUPPORT?
I will be available for all but a week or so in July and August by email and
phone to answer queries about progress and direction. I can be
contacted by email on chris.webber@uwe.ac.uk or on 07979-547932
(mobile), including evenings up to 7.00 pm but not weekends, please.
Chris Webber May 2010
Deadline: late August 2010 (exact date not yet determined). Submit your
work via the BBS assignment boxes near Cribbs café.