5000 words or more

SEES0095

Advanced Quantitative Methods 2019/2020

Project Description and Q & A

Svetlana Makarova room 525 (16 Taviton) [email protected]

2 SEES0095 AQM Project Description and Q & A

Brief Extraction from the SEESG46 Advanced Quantitative Methods Course Description

The course will be assessed through 100% coursework assignment. The assignment will be in the form of a project. For your project are you encouraged to use data related to emerging markets and transition economies – for example, on Demographic and Health indicators, voting patterns and determinants, regional GDP, Foreign Direct Investment, Growth, inflation, labour markets etc – and carry out an appropriate statistical analysis. Your results should be written up as a research project incorporating an introduction, a contextual or literature background, a methodology, data description and interpretation, results and conclusions. The practical exercises throughout the course will prepare you for this assignment. You will be expected to use Stata as the only computational tool. No other statistical packages will be allowed. The assignment should be no more than 5,000 words. There will be progress check on the first Monday after the reading week. The initial project proposal deadline is during the week after the reading week. The assignment should be produced in the style of an academic journal publication and submitted electronically via Turnitin link provided on Moodle

3 SEES0095 AQM Project Description and Q & A

Deadline:

Thursday, 30 April, by 3 PM

4 SEES0095 AQM Project Description and Q & A

The aim of the Project

 To develop specific skills in applied quantitative research by means of supervised independent computer-based econometric project work.

 To develop the general transferable skills of academic report writing. Intended learning outcomes

 Appropriate use of data related, to emerging & transition economies and the economic and econometric/statistical knowledge for specifying and estimating an applied economic model.

 Improving skills in data collection, Internet literature search, descriptive data analysis and report writing.

 Demonstrating subject specific skills related to:  econometric model formulation and estimation & selection,

 testing statistical hypotheses in single-equation econometric models,

 evaluation and interpretation of the results,

 use of Stata for empirical econometric modelling.

5 SEES0095 AQM Project Description and Q & A

How to structure the CW We have discussed: 1. Multiple regression analysis with cross section data (Lectures 1-4 & 9).

2. Time series models (Lectures 5-7).

3. Panel data models (Lectures 7-8 & 9).

The mark of your Coursework won’t depend on the topic you choose.

It will only depend on the quality of your essay and its presentation.

6 SEES0095 AQM Project Description and Q & A

More Hints: 1. Specific of this CW: focus on econometrics.

 You need to demonstrate all details and steps of empirical econometrics modelling (as we did in the lectures & examples).

 This is different from the CW on majority of other courses (including your dissertation where the focus is on socio-economic analysis).

 However, be mindful and reasonable. Support your modelling by short discussion/description explaining socio-economics behind your models.

 You may use data for emerging as well as developed markets. This is, most likely, different from what you might be required to do for other courses or your dissertation.

7 SEES0095 AQM Project Description and Q & A

2. Start empirical part with data and variables descriptions.

3. List and explain all variables’ names.

4. Clearly describe all re-calculations and transformations that you performed to your variables.

5. General structure of empirical part (what should not be missed)

 In all tests state clearly the null and the alternative hypothesis. Explain your conclusions.

 Name all graphs (as Figure 1, Figure 2, etc) and tables (as Table 1, Table 2, etc) and make explicit references and comments on all included graphs and tables.

 …continues on the next page…

8 SEES0095 AQM Project Description and Q & A

Cross Section Data Time Series Data Panel Data Provisional visual analysis:

 Scatter diagram(s).

 Functional form (hence, possible transformation of variables).

 Comment on possible heteroscedasticity.

 Time series graphs.

 Comment on possible non- stationarity and, if applicable, cointegration.

 Use ACF and PACF to comment for stationarity (and AR order (if needed).

 Visual analysis for stationarity.

 Analysis of ‘averages’ over time and over cross section units (see Lecture 7)

Provisional testing:  ADF test for stationarity

 Engle-Granger test for cointegration.

 If non-stationary variables are not- cointegrated, then transform them into stationary form.

 Testing for stationarity in PD (Lecture 8).

 Transform non-stationary variables into stationary form.

Formulate and estimate the model(s) following General-to-Specific approach  If non-stationary variables are

cointegrated, then start with ECM model(s).

 Start with FE model(s).

 Test for Pooled model vs FE.

 Test RE vs FE and Polled vs RE.

…continues on the next page…

9 SEES0095 AQM Project Description and Q & A

Cross Section Data Time Series Data Panel Data …continues from the previous page…

Post-estimation check and Residual analysis Discuss magnitude, sign and significance of the estimates, comment on OVB, etc.

 Discuss endogeneity and apply IV-GMM if needed Plot residuals graph(s) and comment on homo- / heteroscedasticity.

 Perform tests on heteroscedasticity. Address if needed, comment.

 Test for restrictions. Re- formulate the model(s) if needed; repeat the analysis, if needed.

 Test for normality and functional form, address if needed.

 Compare models by AIC and BIC.

 Discuss endogeneity and apply IV-GMM if needed.

 Plot residual vs lagged residuals, comment on autocorrelation.

 Test for autocorrelation (tests, ACF).

 If autocorrelation is detected, reformulate your model.

 In case of ECM, comment on appropriateness.

 In case of AR(p) models explain the choice of the order lags.

 Discuss endogeneity and apply IV-GMM if needed.

 If Pooled model is more appropriate, analyse residuals as for cross section data.

Otherwise, for FE:

 Test for restrictions.

 Test for autocorrelation and cross-section dependency (Lecture 8).

 Address by using robust or cluster options if needed.

 Re-formulate your models, if needed.

 Discuss endogeneity and apply IV-GMM if needed.

…continues on the next page…

10 SEES0095 AQM Project Description and Q & A

Cross Section Data Time Series Data Panel Data …continues from the previous page…

 Compare all admissible models.

 Choose your preferable model, explain, comment, analyse, discuss limitations and advantages.

Some common Q & A

1. Some of widely-used models (e.g. GARCH models for time series or Logit & Probit for binary depended variables) have not been taught. Can I use them?

I DO NOT advise you to do this. In the first instance you have to demonstrate what you have learned within the course. There might be much for the CW if you continue to more advanced models and approaches.

2. What is the ideal number of references for this coursework?

There are no upper or lower limits for references.

11 SEES0095 AQM Project Description and Q & A

More on how to structure your CW

1. Executive summary, giving main features and conclusions of the study in not more than 300 words. [app. 3%]

2. Clear account of what problems the project addresses. [app. 7%]

3. Explanation of the economic theory and literature. [app. 7%]

4. Description of data and their properties. [app. 20%]

5. Building, estimation of the model, post-estimation analysis, application of the model. [app. 48%]

6. Conclusions. [app. 15%]

12 SEES0095 AQM Project Description and Q & A

Structure and contents of the project 1. Executive summary (app. 1/4 page)

 Aim.

 Theory.

 Data.

 Methods.

 Conclusions regarding the quality of the results (comparative and policy analysis and/or forecasts).

2. Basic theory for your modelling (app. 1-2 pages) Good use of references: to the theory (lecture reading) and (ideally) to other empirical work.

E.g. introduction to the basic Solow growth model and conditional beta convergence.

13 SEES0095 AQM Project Description and Q & A

3. The development of estimable model (app. 1-2 pages)

 How to replace in the theoretical model the micro/macro-economic notions by statistical data (see e.g. Sala-i-Martin, 1996).

 Problem of unobservable variables and/or missing observations.  How to solve out unobservable variables (e.g. permanent income, expected

inflation, etc.).

 How (and why) to simplify.

14 SEES0095 AQM Project Description and Q & A

4. Description of data quality and sources (app. 1-3 pages)

 Clear indication of data source. Possible Data Sources for your Project: Moodle, Internet, other.

 Choice of data period and /or cross section units (coutris, regions etc.).

 Clear definitions of variables and their names.

 Recalculation tricks (if appropriate), e.g.:  related to changes in definition and measurements of variables,

 related to changes in definition of the base for index variables,

 approximation of gaps (missing observations) in the series,

 recalculation of the nominal into real variables.

 Problems with outliers.

 Your assessment of the data quality.

Important: write down about data deficiencies. There is no need to use

perfect data!

15 SEES0095 AQM Project Description and Q & A

Some common mistakes here:

1. Mistaking nominal data and real (e.g. real and nominal income, real and nominal money).

2. Mistaking inflation with CPI index: CPI index:

0 /

t t CPI P P .

CPI inflation:

1 ln( ) ln( )t t t t t t t t

t t

P P CPI P P p p

P CPI  

 

 

        ,

where ln( ) t t

p P .

16 SEES0095 AQM Project Description and Q & A

 Describe static (cross section) and/or dynamic (TS) properties of the data (whatever is appropriate, 6-8 pages):

 Visual, descriptive analysis of data.

 Cross-sectional, panel data: graphs, analysis of heteroscedasticity (original and transformed variables).

 Time series, panel data: original and transformed data for the modelling (e.g. income and money in logs).

 Visual analysis of stationarity (TS) and normality of data (if relevant).

 Are there any outliers (structural breaks) in the data/series? If so, why did they happen? (Look back to the literature).

 Stationarity properties for TS data (Dickey-Fuller test etc.)

 Analysis of heteroscedasticity, outliers

 Analysis of (stationary) transformations of data (autocorrelation function, normality, outliers), if relevant.

See more in further lectures on how to adjust the stutter of analysis if you use time series (TS) or panel data (PD).

17 SEES0095 AQM Project Description and Q & A

5. Estimation and economic interpretation (app. 4-5 pages)

 Explaining the economic rationale for the relationship. What are the expected signs (and values) of the parameters?

 Was your estimation procedure applied correctly?

 Justification of correct model estimation (references to omitted variables bias, stochastic regressors, nonstationarity etc.).

 Hypothesis testing.

 Post-estimation check: further statistical analysis, graphical and analytical (e.g. heteroscedasticity, cointegration, long-run relationships etc. whatever is appropriate).

See more in further lectures on how to adjust the stutter of analysis if you use time series (TS) or panel data (PD).

18 SEES0095 AQM Project Description and Q & A

6. Conclusions (1page)

What has been achieved; what is good and what is bad about the model?

a) Are assumptions realistic?

b) Are data adequate (long enough, of a good quality, etc.)?

c) Is any important variabes missing or approximated?

d) Are estimates interpretable?

e) Is model good enough for forecasting/policy analysis?

f) How it can be further developed in the future?

19 SEES0095 AQM Project Description and Q & A

Organisation Details

Submit your Progress Check Questionnaire through Moodle by Monday 18 February 2019 (just after the Reading Week). You can find the Progress Check Questionnaire on the Moodle in the “Project / Coursework Information” section.

1. Submit an electronic version of your Project into Turnitin through the link provided on Moodle by 3pm, Thursday 30 April

2. Your data (as student-number.dta Stata file, e.g.1111111.dta) must be also submitted through separate link on Moodle by 3pm, Thursday 30 April.

20 SEES0095 AQM Project Description and Q & A

 Use Times New Roman 12 pt., singe spaced, for writing the Project.

 Do not forget about page numbering!

 References template examples

Books:

Charemza, W.W. and Deadman, D.F. (1997). New directions in econometric practice, 2nd edition. Cheltenham: Edward Elgar.

Chapter in a book:

Hildreth, A. and Pudney, S.E. (1999). Linked cross section employer-worker surveys. In Biffignandi, S. (ed.) Micro- and Macrodata of Firms, Heidelberg: Physica Verlag, pp. 509-540.

21 SEES0095 AQM Project Description and Q & A

Journal articles:

Lee, K.C., Pesaran, M.H. and Smith, R. (1997). Growth and convergence in a multi- country empirical stochastic Solow model, Journal of Applied Econometrics, vol 12, pp. 357-392.

Internet article:

Soper, J.B. (1997). ‘Integrating interactive media in courses: the WinEcon software with workbook approach’, Journal of Interactive Media in Education http://www.jime.open.ac.uk/97/2

Data publication:

Office for National Statistics (1999). Family Spending. Report of the 1998 Family Expenditure Survey. London: HMSO

Internet source:

Million, N. (2001), ‘Monetary policies, the oil crisis and the Fisher effect hypothesis, University Paris I’ – Eurequa, paper presented at the European Meeting of the Econometric Society, Venice, 2002, http://www.eea-esem.com/papers/eea- esem/esem2002/1596/million.pdf

22 SEES0095 AQM Project Description and Q & A

In order to achieve high mark

 State clearly all your steps in modelling (data, model, methods etc.) and possible remedies for detected problems in modelling.

 Explain clearly all steps of econometric modelling procedure.

 Cleary explain model selection procedure (post-estimation check).

 Put the most important empirical findings and testing outcomes into the main part of the text.

 Give titles to all tables and graphs and make good reference to them.