7 views

Uploaded by Eason Saint

Vector Autoregressive introduction

- The Measurement of Inequality of Opportunity Theory and Application to Latin America
- Time Series Econometrics
- Toward a Unified Theory of Causality
- cointegration
- Short Dissertation
- ETEC 500 Full Proposal
- cointegration
- US Federal Reserve: 200403pap
- line plots - intro
- Neill2008_WritingUpAnANOVA
- VAR_VECM_Toda-Yamamoto Models by Afees Salisu
- Chap.10
- Midterm Sample
- Geo Experiments Final Version
- 3rd Test Result
- Output
- Paper
- Analytics Training Proposal_SVIMS.PDF
- sem
- SQQS2013_Assg_2

You are on page 1of 34

TOPIC 7:

VECTOR

AUTOREGRESSIVE

MODELS AND ITS

APPLICATION

By:

Assoc. Prof. Dr. Sallahuddin Hassan

SEEQ5133

Applied Econometrics

INTRODUCTION

2

explanatory variables for a given

dependent variable, but they are

also explained by the variable that

they are used to determined.

Model of simultaneous equations

exogenous, endogenous and

predetermined.

SEEQ5133

INTRODUCTION

simultaneity among a number of

variables, then all these variables should

be treated in the same way.

Therefore, there should be no distinction

between endogenous and exogenous

variables. All variables should be

treated as endogenous variable.

SEEQ5133

INTRODUCTION

4

development of the VAR model.

Why we need VAR:

to describe the dynamic interrelationship

between stationary variables.

We are not really confident that a

variable is actually exogenous.

Performing forecasting analysis.

SEEQ5133

FEATURES OF A VAR

MODEL

Can be estimated by OLS and get consistent

estimators.

All variables are endogenous. Only lagged

endogenous variables on RHS.

All variables are assumed stationary.

Coefficient in reduced form not structural

parameter.

Contemporaneous effect captured by residuals.

are uncorrelated white-noise error terms.

1t & 2t

SEEQ5133

VAR MODEL

6

The regressors are lagged values of

all k series.

variables:

yt

xt

and

SEEQ5133

VAR MODEL

7

variables yield a system of equations:

yt 10 11 yt 1 12 xt 1

y

t

x

t

xt 20 21 yt 1 22 xt 1

lag and the lag of the other variable

in the system.

SEEQ5133

VAR MODEL

8

variables,

VAR model is:

y

yt 10 11 yt 1 12 xt 1 t

xt 20 21 yt 1 22 xt 1

x

t

estimated using OLS.

SEEQ5133

I 0

VAR MODEL

9

I 1

variables, and not cointegrated,

we work with the first difference.

VAR model is:

yt 11yt 1 12 xt 1 ty

xt 11yt 1 12 xt 1 tx

. The

I 1

system can be estimated

using OLS

SEEQ5133

VECM MODEL

10

1

Ivariables,

and cointegrated

We need to modify the system of

equations to allow for the

cointegrating relationship

between the nonstationary

variables.

SEEQ5133

VECM MODEL

11

Why we do this?

about the cointegrating relationship.

To ensure the best technique that take

into account the properties of time series

data.

special form of the VAR for nonstationary

variables that are cointegrated.

SEEQ5133

VECM MODEL

12

Model:

Yt 1 11 Yt 1 0 1 X t 1 v1t

X t 2 21 Yt 1 0 1 X t 1 v 2t

how much

will change in response to

yt

a change in the explanatory variable (the

cointegration part, ), as well as the speed

of the change (the error correction part,

ECTt 1

)

SEEQ5133

VECM MODEL

13

Yt 1 ECTt 1 X t t

is the impact multiplier (the short-run

effect) that measures the immediate

impact that

X t a current change in Yt will have

on a change in .

effect, and shows the speed of adjustment

1

or how much of the disequilibrium is being

corrected.

To ensure stability.

SEEQ5133

SPECIFICATION ISSUES

14

Logs or no logs?

How many variables?

equation is proportional to the number

of variables.

Keep the number of variables small:

variables.

to avoid estimation error

forecasting

accuracy.

SEEQ5133

SPECIFICATION ISSUES

15

If some I(1) but cointegrated,

then level of ECM.

If I(1) but not cointegrated, then

difference to I(0).

SEEQ5133

SPECIFICATION ISSUES

16

as few as possible.

Too many lags consume degree of

freedom and multicollinearity

( and

x1

x2

are linearly dependent)

Too few lags specification error (omitting

influential or including non-influential

explanatory variables and omission of

relevant or inclusion of irrelevant

irrelevant variables)

SEEQ5133

17

SPECIFICATION OF VAR

MODEL

1) Testing for stationarity

stationary or non stationary.

Performing unit root tests DF, ADF,

or PP tests.

Double click on the series and

choose View/Unit Root

Test/Perform test

specification/OK

SEEQ5133

18

SEEQ5133

19

SPECIFICATION OF VAR

MODEL

H0: Unit root/non stationary.

H1: Stationary.

(

Reject H0 if the ADF statistics

C value ( )

)< the critical

then PRI ~I(n); n>0.

PRI of

PRI

differences

PRI

as t 1

t PRI

SEEQ5133

20

SPECIFICATION OF VAR

MODEL

2) Testing for cointegration

Steps:

Setting the appropriate lag length of the model.

Choosing the appropriate model.

Determining the number of cointegrating vector.

statistics/Johansen Cointegration

test/OK

SEEQ5133

21

SPECIFICATION OF VAR

MODEL

Step 1:Testing the order of integration

of the variables.

Step 2: Setting the appropriate lag

length of the model

levels

Inspect the values of the AIC, SBC and do

diagnostic checking (autocorrelation,

heteroscedasticity, normality, possible ARCH

effect).

SEEQ5133

22

SPECIFICATION OF VAR

MODEL

VAR Estimation

SEEQ5133

Normality Test

23

normality.

It based on two measures:

the residuals are around zero.

Kurtosis refers to the peakedness

of the distribution. For a normal

distribution, the kurtosis value is 3.

SEEQ5133

Normality Test

24

N

JB

6

3

2

S

Sample size

We reject the hypothesis of normally

distributed error if a calculated value of the

statistics exceeds a critical value selected

from the chi-squared distribution.

SEEQ5133

25

SPECIFICATION OF VAR

MODEL

regarding the deterministic components in

the multivariate system.

SEEQ5133

26

SPECIFICATION OF VAR

MODEL

Inform ation Criteria by Rank and Model

Data Trend:

Rank or

No. of CEs

None

No Intercept

No Trend

None

Intercept

No Trend

Linear

Intercept

No Trend

Linear

Intercept

Trend

0

1

2

3

Log Likelihood

-2412.003

-2400.054

-2393.478

-2392.300

-2412.003

-2408.563

-2408.563

-2399.586

-2398.039

-2397.803

-2392.181

-2391.979

-2391.155

-2390.937

-2390.937

-2389.406

0

1

2

3

121.5001

121.5001

121.4782

121.4782

121.2434

121.2027

121.2293

121.2519

121.2901

121.0948*

121.1739

121.2091

121.2490

121.3077

121.2839

121.4150

121.4969

121.4969

121.5703

121.5703

0

1

2

3

122.2601

122.2601

122.3648

122.2160*

122.2848

122.3919

122.4405

122.5602

122.6423

122.9350

123.1435

123.1435

(colum ns )

122.3648

122.4724

122.7855

123.3436

SEEQ5133

Quadratic

Intercept

Trend

-2400.869

-2391.895

-2389.677

-2389.406

122.2568

122.3614

122.8039

123.3436

27

SPECIFICATION OF VAR

MODEL

of

Step 4: Determining the rank

or

the number of cointegrating vector.

Using two tests:

H0

cointegrating relationship)

H1: (r + 1) vector.

Maximal eigenvalue statistic to test how

many of the number of the characteristic roots

are significantly different from zero.

max r , r 1 T ln 1 r 1

SEEQ5133

28

SPECIFICATION OF VAR

MODEL

H0

than of equal to r.

H1: The number of cointegrating vectors is more

than r.

Trace statistic:

n

trace r T ln 1

r 1

i r 1

If

value; so the model does not show cointegration.

SEEQ5133

29

SPECIFICATION OF VAR

MODEL

Unres tricted Cointegration Rank Tes t (Trace)

Hypothes ized

No. of CE(s )

None *

At m os t 1 *

At m os t 2

Eigenvalue

Trace

Statis tic

0.05

Critical Value

Prob.**

0.449779

0.280235

0.057191

39.40629

15.50888

2.355655

24.27596

12.32090

4.129906

0.0003

0.0141

0.1474

* denotes rejection of the hypothes is at the 0.05 level

**MacKinnon-Haug-Michelis (1999) p-values

Unres tricted Cointegration Rank Tes t (Maxim um Eigenvalue)

Hypothes ized

No. of CE(s )

None *

At m os t 1 *

At m os t 2

Eigenvalue

Max-Eigen

Statis tic

0.05

Critical Value

Prob.**

0.449779

0.280235

0.057191

23.89741

13.15323

2.355655

17.79730

11.22480

4.129906

0.0053

0.0226

0.1474

* denotes rejection of the hypothes is at the 0.05 level

**MacKinnon-Haug-Michelis (1999) p-values

SEEQ5133

30

SPECIFICATION OF VAR

MODEL

Both the trace and the maximal

eigenvalue statistics suggest the

existence of two cointegrating

vectors.

Eviews then reports results regarding

the coefficients of the speed of

adjustment coefficients

( ) and the

matrix of the long-run

coefficients

( ).

SEEQ5133

31

SPECIFICATION OF VAR

MODEL

cointegrating vectors, we

proceed with the estimation of

the ECM.

SEEQ5133

32

VECM MODEL

ESTIMATION

estimate the VECM.

SEEQ5133

33

VECM MODEL

ESTIMATION

Cointegrating Eq:

CointEq1

REXP_M(-1)

1.000000

RFDI_M(-1)

-10.52989

(10.3874)

[-1.01372]

RGDP_M(-1)

-0.019252

(0.00352)

[-5.47628]

Error Correction:

D(REXP_M)

D(RFDI_M)

D(RGDP_M)

CointEq1

-0.039037

(0.01669)

[-2.33941]

-0.011229

(0.00529)

[-2.12306]

-4.457555

(0.95264)

[-4.67917]

D(REXP_M(-1))

0.457965

(0.18590)

[ 2.46349]

-0.026410

(0.05892)

[-0.44823]

7.924660

(10.6129)

[ 0.74670]

D(RFDI_M(-1))

-0.831488

(0.63688)

[-1.30557]

-0.043661

(0.20186)

[-0.21629]

-3.312521

(36.3589)

[-0.09111]

D(RGDP_M(-1))

-0.005673

(0.00410)

[-1.38363]

-0.002730

(0.00130)

[-2.10112]

-0.242330

(0.23406)

[-1.03532]

SEEQ5133

34

can estimate the unrestricted VAR.

SEEQ5133

- The Measurement of Inequality of Opportunity Theory and Application to Latin AmericaUploaded byChristian Portella
- Time Series EconometricsUploaded byBlue Moon
- Toward a Unified Theory of CausalityUploaded byBlenda Lara
- cointegrationUploaded byPrashant Sharma
- Short DissertationUploaded byMadhu Suresh
- ETEC 500 Full ProposalUploaded byquinlan8
- cointegrationUploaded bySamia Nasreen
- US Federal Reserve: 200403papUploaded byThe Fed
- line plots - introUploaded byapi-270878578
- Neill2008_WritingUpAnANOVAUploaded bymadhu24mummy
- VAR_VECM_Toda-Yamamoto Models by Afees SalisuUploaded byzamir
- Chap.10Uploaded byTúlio Igor Soares Pereira
- Midterm SampleUploaded byBurak
- Geo Experiments Final VersionUploaded bymanuelomanuelo
- 3rd Test ResultUploaded byDare Quimada
- OutputUploaded byAnnisa Nanda Yusri
- PaperUploaded byViver
- Analytics Training Proposal_SVIMS.PDFUploaded byChandan Kokane
- semUploaded byrudden
- SQQS2013_Assg_2Uploaded byHafiz Razali
- chapter4laurendallasUploaded byapi-345780702
- p 04 Pls Formative ModelsUploaded bycdnader
- analisis multivariat rudiUploaded byegha
- Counter fUploaded byAlejandra Parra
- When You'Re StuckUploaded byHoang_Hon
- Statistical Package for Social ScienceUploaded byGopinath Yadula
- Management Accounting and Erp SystemUploaded bysulthanhakim
- Practicum Data AnalysisUploaded byksu10
- Project ReportUploaded byFaheem Firdous Qureshi
- BrmUploaded byAhmed Warraich

- Money creation in the modern economy Bank EnglandUploaded bybreakingthesilence
- KMDC CFP Early Bird Scheme 2019 Sem 1Uploaded byEason Saint
- Exercise 3_2012018Uploaded byEason Saint
- FIMMF Manual (Eng)Uploaded byEason Saint
- Consumer Protection Act 1999 _Act 599Uploaded bypeterparker
- 60.1.126-142Uploaded byprashantss2001
- Lecture 4 ARIMAUploaded byEason Saint
- Lecture 6 Stationarity and CointegrationUploaded byEason Saint
- Book of Proceedings Esd Bangkok 2016 OnlineUploaded byEason Saint
- Applied Econometrics - Exercise 3Uploaded byEason Saint
- Lecture 8 Application of VAR ModelUploaded byEason Saint
- 1.0 International Trade - Theory and PolicyUploaded byNoel Mahung Melendez
- GST Letter to SupplierUploaded byEason Saint
- SENARAI TERKINI TENDER 2~2014Uploaded byEason Saint
- Macroeconomics exerciseUploaded byEason Saint
- Garch notesUploaded byEason Saint
- Economic Theory of Bank CreditUploaded byEason Saint
- 25 Theories to Get You StartedUploaded byShivaji_Saha_6955
- ΤΑ ΧΡΗΜΑΤΑ ΣΤΗΝ ΜΟΝΤΕΡΝΑ ΟΙΚΟΝΟΜΙΑUploaded byellhngreek
- Geography & Economic DevelopmentUploaded byEason Saint
- The Dynamics of Commodity Prices - AClustering ApproachUploaded byEason Saint
- PR 0308 EntertainmentUploaded bymichelle_tann
- Steven SimUploaded byEason Saint

- GeoStat Course AMECUploaded byjmelor_74294880
- GLUE TechniqueUploaded bySuman Bhowmick
- spss_readme.txtUploaded byAlberto
- IBM SPSS Bootstrapping 19Uploaded bySomaskhandan Chinnasamy
- Regression TechniquesUploaded byRitesh Raman
- Non Linear RelationshipsUploaded bySunil Karamchandani
- Case 7.1Uploaded bySammy Ben Menahem
- Robust Commands StataUploaded byUNLV234
- Reserves estimation methods for prospect evaluation with 3D CSEM data Copyright First Break.pdfUploaded byskywalk189
- CHE 555 Curve FittingUploaded byNorizzatul Akmal
- (4.4) (140)Uploaded byFherry Leonheart
- Maximum LikelihoodUploaded byShaibal Barua
- Mediation and Moderation AnalysisUploaded byTrinitarian Theophany
- model good 2.pdfUploaded bySumi Srianand
- GARCHUploaded byOisín Ó Cionaoith
- Statistics PrelimsUploaded byசுப.தமிழினியன்
- 061220ppt1 - Logistic RegressionUploaded byShashank Magulur
- OPRE 6301-SYSM 6303 Chapter 10 Slides_studentsUploaded bySrinivas Reddy
- Output SpssUploaded byYuli Andriyani Halim
- Core StatiCore Statistics new Simon Woods Bookstics New Simon Woods BookUploaded byJosé Manuel Carranza
- mlogitUploaded byk_ij9658
- 1-5 Basic Statistics for ResearchUploaded byAsif Jamil
- Chapter6 IVUploaded byJawhar Bacha
- Voltage Flicker Parameters Estimation UsingShuffled Frog Leaping Algorithm (SFLA) andImperialistic Competitive Algorithm (ICA)Uploaded byChief Editor
- STA5328_Ramin_Shamshiri_HW3Uploaded byRaminShamshiri
- Msda3 NotesUploaded byatleti2
- ReportUploaded byRancid Remunerator
- DSP 2015.pdfUploaded byPraveen
- BR Assignment Report FullUploaded byThanobol Cenphakdee
- Proc UCMUploaded byNirmal Roy