1 an Introduction to Panel Data Methods

Embed Size (px)

Citation preview

  • 7/28/2019 1 an Introduction to Panel Data Methods

    1/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    An introduction to Panel datamethods

  • 7/28/2019 1 an Introduction to Panel Data Methods

    2/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Contents

    Preliminary background

    Static panel data

    Dynamic panels

    Panel time series

  • 7/28/2019 1 an Introduction to Panel Data Methods

    3/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    What is Panel data?Year Firm Investment Value of the firm

    1950 1 642.9 3755.6

    1951 1 755.9 4833

    1952 1 891.2 4924.9

    1953 1 1304.4 6241.71954 1 1486.7 5593.6

    1950 2 100.66 693.5

    1951 2 160.62 809

    1952 2 145 727

    1953 2 174.93 1001.5

    1954 2 172.49 703.2

    1950 3 93.5 1610.51951 3 135.2 1819.4

    1952 3 157.3 2079.7

    1953 3 179.5 2371.6

    1954 3 189.6 2759.9

    1950 4 32.24 635.2

    1951 4 54.38 723.8

    1952 4 71.78 864.1

    1953 4 90.08 1193.5

    1954 4 68.6 1188.9

    1950 5 418.8 1677.4

    1951 5 588.2 2289.5

    1952 5 645.2 2159.4

    1953 5 641 2031.31954 5 459.3 2115.5

  • 7/28/2019 1 an Introduction to Panel Data Methods

    4/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    What is Panel Data?

    Year Firm Investment Value of the firm

    1950 1 642.9 3755.6

    1951 1 755.9 4833

    1952 1 891.2 4924.91953 1 1304.4 6241.7

    1954 1 1486.7 5593.6

    1950 2 100.66 693.5

    1951 2 160.62 809

    1952 2 145 727

    1953 2 174.93 1001.5

    1954 2 172.49 703.2

    1950 3 93.5 1610.5

    1951 3 135.2 1819.4

    1952 3 157.3 2079.7

    1953 3 179.5 2371.6

    1954 3 189.6 2759.9

    1950 4 32.24 635.21951 4 54.38 723.8

    1952 4 71.78 864.1

    1953 4 90.08 1193.5

    1954 4 68.6 1188.9

    1950 5 418.8 1677.4

    1951 5 588.2 2289.5

    1952 5 645.2 2159.41953 5 641 2031.3

    1954 5 459.3 2115.5

    Time dimension

    Cross section dimension

  • 7/28/2019 1 an Introduction to Panel Data Methods

    5/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    What is panel data?

    Year Firm Investment Value of the firm

    1950 1 642.9 3755.6

    1951 1 755.9 4833

    1952 1 891.2 4924.91953 1 1304.4 6241.7

    1954 1 1486.7 5593.6

    1950 2 100.66 693.5

    1951 2 160.62 809

    1952 2 145 727

    1953 2 174.93 1001.5

    1954 2 172.49 703.2

    1950 3 93.5 1610.5

    1951 3 135.2 1819.4

    1952 3 157.3 2079.7

    1953 3 179.5 2371.6

    1954 3 189.6 2759.9

    1950 4 32.24 635.21951 4 54.38 723.8

    1952 4 71.78 864.1

    1953 4 90.08 1193.5

    1954 4 68.6 1188.9

    1950 5 418.8 1677.4

    1951 5 588.2 2289.5

    1952 5 645.2 2159.41953 5 641 2031.3

    1954 5 459.3 2115.5

  • 7/28/2019 1 an Introduction to Panel Data Methods

    6/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    What is panel data?

    Data for many individuals (cross sections)

    across time

    Previous slides show the data in stackedformat.

    Wh t i th l ti hi b t

  • 7/28/2019 1 an Introduction to Panel Data Methods

    7/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    What is the relationship betweeninvestment and value of the

    firm? Plot Investment and

    value as a scatterplot.

    Data is stacked whichimplies that this is therelationship acrosstime and across firms

    0

    1000

    2000

    3000

    4000

    5000

    6000

    7000

    0 200 400 600 8001000 1400

    Investment

    Valu

    e

    Wh t i th l ti hi b t

  • 7/28/2019 1 an Introduction to Panel Data Methods

    8/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    What is the relationship betweeninvestment and value of the

    firm? The relationship is

    positive.

    Is this the case for

    each firm in ourpanel?

    0

    1000

    2000

    3000

    4000

    5000

    6000

    7000

    8000

    0 200 400 600 8001000 1400

    Investment

    Value

    Wh t i th l ti hi b t

  • 7/28/2019 1 an Introduction to Panel Data Methods

    9/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    What is the relationship betweeninvestment and value of the

    firm?

    2500

    3000

    3500

    4000

    4500

    5000

    5500

    6000

    6500

    200 400 600 800 1000 1200 1400 1600

    Investment

    Value

    1

    400

    500

    600

    700

    800

    900

    1000

    1100

    20 40 60 80 100 120 140 160 180

    Investment

    Value

    2

    1000

    1500

    2000

    2500

    3000

    0 40 80 120 160 200

    Investment

    Value

    3

    0

    200

    400

    600

    800

    1000

    1200

    1400

    10 20 30 40 50 60 70 80 90 100

    Investment

    Value

    4

    1200

    1400

    1600

    1800

    2000

    2200

    2400

    2600

    2800

    200 300 400 500 600 700

    Investment

    Value

    5

    Positiveslope

    But differentStarting points

  • 7/28/2019 1 an Introduction to Panel Data Methods

    10/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Panel data methods

    Most of panel data analysis is concerned

    with measuring these two aspects of paneldata relationships:

    i,t i Vi,t i,t

    Slope

    Intercept

    i=15 denotes firms

    T=1950is time

  • 7/28/2019 1 an Introduction to Panel Data Methods

    11/56

  • 7/28/2019 1 an Introduction to Panel Data Methods

    12/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Pooled OLS

    Model:

    To estimate:

    Step 1: Stack data

    Step 2: Apply OLS formula

    i,t Vi,t i,t

    Same intercept

  • 7/28/2019 1 an Introduction to Panel Data Methods

    13/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Step1

    Year Firm Investment Value of the firm Constant

    1950 1 642.9 3755.6 1

    1951 1 755.9 4833 1

    1952 1 891.2 4924.9 1

    1953 1 1304.4 6241.7 1

    1954 1 1486.7 5593.6 1

  • 7/28/2019 1 an Introduction to Panel Data Methods

    14/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Step 1

    Year Firm Investment Value of the firm Constant

    1950 1 642.9 3755.6 1

    1951 1 755.9 4833 1

    1952 1 891.2 4924.9 1

    1953 1 1304.4 6241.7 1

    1954 1 1486.7 5593.6 1

    1950 2 100.66 693.5 1

    1951 2 160.62 809 11952 2 145 727 1

    1953 2 174.93 1001.5 1

    1954 2 172.49 703.2 1

    Data for firms on top of one another

  • 7/28/2019 1 an Introduction to Panel Data Methods

    15/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Step 1

    Year Firm Investment Value of the firm Constant

    1950 1 642.9 3755.6 1

    1951 1 755.9 4833 1

    1952 1 891.2 4924.9 1

    1953 1 1304.4 6241.7 1

    1954 1 1486.7 5593.6 1

    1950 2 100.66 693.5 1

    1951 2 160.62 809 1

    1952 2 145 727 1

    1953 2 174.93 1001.5 1

    1954 2 172.49 703.2 1

    1950 3 93.5 1610.5 11951 3 135.2 1819.4 1

    1952 3 157.3 2079.7 1

    1953 3 179.5 2371.6 1

    1954 3 189.6 2759.9 1

    1950 4 32.24 635.2 1

    1951 4 54.38 723.8 1

    1952 4 71.78 864.1 1

    1953 4 90.08 1193.5 11954 4 68.6 1188.9 1

    1950 5 418.8 1677.4 1

    1951 5 588.2 2289.5 1

    1952 5 645.2 2159.4 1

    1953 5 641 2031.3 1

    1954 5 459.3 2115.5 1

    Y variable

    Xvariables

  • 7/28/2019 1 an Introduction to Panel Data Methods

    16/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Step 2

    Apply OLS formula

    , XX1XY

    Y Ii,tX Vi,t,constant

  • 7/28/2019 1 an Introduction to Panel Data Methods

    17/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Fixed Effects

    Model

    To estimate

    (a) Stack data(b) Use OLS formula

    i,t i Vi,t i,t

    Different intercept for each firm

  • 7/28/2019 1 an Introduction to Panel Data Methods

    18/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Step 1: the data

    Year Firm Investment Value of the firm Constant for firm 1 Constant for firm2 Constant for firm 3 Constant for firm 4 Constant for firm 5

    1950 1 642.9 3755.6 1 0 0 0 0

    1951 1 755.9 4833 1 0 0 0 01952 1 891.2 4924.9 1 0 0 0 0

    1953 1 1304.4 6241.7 1 0 0 0 0

    1954 1 1486.7 5593.6 1 0 0 0 0

    1950 2 100.66 693.5 0 1 0 0 0

    1951 2 160.62 809 0 1 0 0 0

    1952 2 145 727 0 1 0 0 0

    1953 2 174.93 1001.5 0 1 0 0 0

    1954 2 172.49 703.2 0 1 0 0 0

    1950 3 93.5 1610.5 0 0 1 0 01951 3 135.2 1819.4 0 0 1 0 0

    1952 3 157.3 2079.7 0 0 1 0 0

    1953 3 179.5 2371.6 0 0 1 0 0

    1954 3 189.6 2759.9 0 0 1 0 0

    1950 4 32.24 635.2 0 0 0 1 0

    1951 4 54.38 723.8 0 0 0 1 0

    1952 4 71.78 864.1 0 0 0 1 0

    1953 4 90.08 1193.5 0 0 0 1 0

    1954 4 68.6 1188.9 0 0 0 1 01950 5 418.8 1677.4 0 0 0 0 1

    1951 5 588.2 2289.5 0 0 0 0 1

    1952 5 645.2 2159.4 0 0 0 0 1

    1953 5 641 2031.3 0 0 0 0 1

    1954 5 459.3 2115.5 0 0 0 0 1

    Y variable

    X variables

    Dummy for each firm

  • 7/28/2019 1 an Introduction to Panel Data Methods

    19/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Step 2: Use OLS formula

    i,t i Vi,t i,t

    , 1, 2, 3, 4, 5 X

    X1

    X

    YY Ii,tX Vi,t,5 dummies

    We get estimates of the (common) slope

    And estimates of the constant for each firm. These areReferred to as fixed effects or individual effects

    Thi i i i i l

  • 7/28/2019 1 an Introduction to Panel Data Methods

    20/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    This estimation is equivalent

    to..(a) Transforming each variable by subtracting the cross-sectional

    mean

    (b) Use OLS formula on this de-meaned data. This is usedwhen the number of cross sections is very large

    (c) Note that this is called the within transformation and

    the resulting estimator is called the within estimator(d) Uses only variation through time of each unit, not

    between units

    Y Ii,tiX Vi,tV i

  • 7/28/2019 1 an Introduction to Panel Data Methods

    21/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Transformation?

    Year Firm Investment Cross Section Mean of Investme

    1950 1 642.9 1016.22

    1951 1 755.9 1016.22

    1952 1 891.2 1016.22

    1953 1 1304.4 1016.22

    1954 1 1486.7 1016.22

    1950 2 100.66 150.74

    1951 2 160.62 150.74

    1952 2 145 150.74

    1953 2 174.93 150.74

    1954 2 172.49 150.74

    1950 3 93.5 151.02

    1951 3 135.2 151.021952 3 157.3 151.02

    1953 3 179.5 151.02

    1954 3 189.6 151.02

    1950 4 32.24 63.416

    1951 4 54.38 63.416

    1952 4 71.78 63.4161953 4 90.08 63.416

    1954 4 68.6 63.416

    1950 5 418.8 550.5

    1951 5 588.2 550.5

    1952 5 645.2 550.5

    1953 5 641 550.51954 5 459.3 550.5

    MeanFor firm i acrosstime

  • 7/28/2019 1 an Introduction to Panel Data Methods

    22/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Before transformation

    -3 -2 -1 0 1 2 3

    0

    20

    40

    60

    80

    100

    Y

    X

    Firm 1

    -1 0 1 2 3 4 5

    0

    20

    40

    60

    80

    100

    Y

    X

    Firm 2

    -1 0 1 2 3 4

    0

    20

    40

    60

    80

    100

    Y

    X

    Firm 3

    -4 -3 -2 -1 0 1 2 3

    0

    20

    40

    60

    80

    100

    Y

    X

    Firm 4

    Sameslope

    DifferentIntercepts!

  • 7/28/2019 1 an Introduction to Panel Data Methods

    23/56

  • 7/28/2019 1 an Introduction to Panel Data Methods

    24/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Between estimator

    The fixed effects estimator uses

    information on variation through time The between estimator uses only

    information across cross-sections OLS regression on averages i.e.

    Y iX V i

  • 7/28/2019 1 an Introduction to Panel Data Methods

    25/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Generalised Least Squares

    Note that for any heteroskedastic model:

    Estimation by GLS exploits the fact that:

    VEXy

    2][

    =

    +=

    VVV

    VVEE

    Xy

    VXVyV

    22/12/12

    2/12/1**

    ***

    2/12/12/1

    ][]'[

    ==

    =

    +=

    +=

    Square

    rootofcovariancematrix

  • 7/28/2019 1 an Introduction to Panel Data Methods

    26/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Random Effects

    Model:

    Variation across cross sections is random

    i,t Vi,t i,t i

    Unit specific shock

    Ei,t 0,Ei 0

    Ei,t2 2,Ei2 2

    Ei,t,j 0 i,j 0

  • 7/28/2019 1 an Introduction to Panel Data Methods

    27/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Random Effects

    Lets consider the ith cross-section

    The covariance matrix of w is

    It Vt t i

    Wt t i

    EWt2

    2

    2

    EWtWs 2 , t s

  • 7/28/2019 1 an Introduction to Panel Data Methods

    28/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Random Effects

    introduces a particular (heteroscedastic)

    structure

    +

    ++

    =

    2222

    2222

    2222

    uuu

    uuu

    uuu

    L

    MOMM

    L

    L

    VARWt

  • 7/28/2019 1 an Introduction to Panel Data Methods

    29/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Random Effects

    Covariance matrix block diagonal

    =

    = I

    L

    MOMML

    L

    00

    00

    00

  • 7/28/2019 1 an Introduction to Panel Data Methods

    30/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Estimation

    As with FE, estimation is carried out by

    (a) transforming the variables to remove thecovariance structure.

    (b) Using OLS on the transformed variables

    This is an example of generalised leastsquares

  • 7/28/2019 1 an Introduction to Panel Data Methods

    31/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Estimation

    We need

    This is a function of

    GLS is equivalent to the following

    transformation of Y and X

    2/12/1 ][ = I

    2,

    2

    1/2Yit 1

    Yit Yi

    1

    T u2

    2

    If this is equal to zeroFE and RE the same

  • 7/28/2019 1 an Introduction to Panel Data Methods

    32/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Estimation

    We need estimates of

    Consider

    And

    Deviations removes heterogeneity

    ,

    it Vit it i

    i Vi i i

    Iit i Vit Vi it i

  • 7/28/2019 1 an Introduction to Panel Data Methods

    33/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Estimation

    So FE can be used to estimate

    Consider the between regression

    e i i Vi i i

    vare i

    2

    T 2

    2 varei

    2

    T

  • 7/28/2019 1 an Introduction to Panel Data Methods

    34/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Fixed or Random

    Type of dataset

    Hausman test. Tests the assumptions

    needed for random effects to be consistent

    If null hypothesis is rejected use fixedeffects

    i,t,j 0

  • 7/28/2019 1 an Introduction to Panel Data Methods

    35/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Hausman test

    Call the FE estimator with estimatedcovariance matrix ; and the RE estimatorwith

    If the RE model is correct, is consistent and

    efficient, and so > If the RE model is wrong, then the RE estimates

    are inconsistent, but the FE estimates are still

    consistent

    The Hausman test uses the test statistic:

    FR)(

    FV

    )( RV

    )(~)

    ()]

    ()

    ([)

    (

    21

    kVV

    RFRFRF

    R

    )(F

    V )(R

    V

  • 7/28/2019 1 an Introduction to Panel Data Methods

    36/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Pooled OLS?

    Can test to see if fixed effects are

    significant and should be included.

    Pooled estimator not appropriate if fixedeffects important

  • 7/28/2019 1 an Introduction to Panel Data Methods

    37/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Dynamic Panel data

    Model

    Lets estimate using FE

    This is because the average contains

    i,t i Vi,t i,t1 i,t

    i,t i Vi,t V i i,t1 t1,i i,t i

    Correlated !Implies inconsistency

    it it1

  • 7/28/2019 1 an Introduction to Panel Data Methods

    38/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Anderson Hsiao IV estimator

    Step 1: Take first differences

    Step 2: use instruments for the RHSvariables

    Valid instruments include

    Level instruments generally preferred

    i,t i,t1 Vi,t Vi,t1 i,t1 Ii,t2 i,t i,t1

    Zit Ii,t2 , Ii,t2

    Zit Zit1Zit Iit

    d i

  • 7/28/2019 1 an Introduction to Panel Data Methods

    39/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Anderson Hsiao

    Consistent estimator, but does not exploit allmoment conditions

    In period 3 our equation is:

    Instruments

    In period 4

    Instruments and so on

    i,3 i,2 Vi,3 Vi,2 i,2 i,1 i,3 i,2Vi,1, i,1

    i,4 i,3 Vi,4 Vi,3 i,3 i,2 i,4 i,3

    Vi,2, i,2,Vi,1, i,1

    A l d B d E i

  • 7/28/2019 1 an Introduction to Panel Data Methods

    40/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Arrelano and Bond Estimator

    AB propose using these extra moment

    conditions

    W

    Zi1Zi1 ,Zi2

    .

    .

    Zi1 ,Zi2 . . .Zit2

    W

    it i 0

    A l d B d E ti t

  • 7/28/2019 1 an Introduction to Panel Data Methods

    41/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Arrelano and Bond Estimator

    One and two step estimator

    Second step standard errors biaseddownwards so inference based on firststep

    1step XitWV01WXit1XitWV0

    1WIit

    Preliminary estimate of weighting matrix

    2step Xit

    WV11

    WXit1

    Xit

    WV11

    WIit

    Based on residuals from the first step

    V lid I t t

  • 7/28/2019 1 an Introduction to Panel Data Methods

    42/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Valid Instruments

    Test for second order serial correlation for

    validity of GMM Sargan test

    itit2 0

    21

    1

    ~)())((' Kp

    asy

    i

    N

    iiii WWWs

    =

    Second step residuals Number of columns of W

    P l Ti S i

  • 7/28/2019 1 an Introduction to Panel Data Methods

    43/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Panel Time Series

    Panels with long T readily available.

    Raises some special issues for panel data

    analysis.

    Heterogeneity

    H t l

  • 7/28/2019 1 an Introduction to Panel Data Methods

    44/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Heterogeneous panels

    it i Bix i,t iyit1 v t

    Change across

    Cross-sections

    Bi B i

    i i

    y it i Bx it yit1 e te t ix it iy it1 v it

    If homogeneity assumed Serially correlated error term

    Heterogeneous panels

  • 7/28/2019 1 an Introduction to Panel Data Methods

    45/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Heterogeneous panels

    Fixed effects and GMM biased and

    inconsistent as they impose slopehomogeneity

    A valid model is the mean group model

    Estimate

    Heterogeneous panels

  • 7/28/2019 1 an Introduction to Panel Data Methods

    46/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Heterogeneous panels

    Estimate

    y1t 1 B1x1,t 1y1t1 v1t

    y2t 2 B2x2,t 2y2t1 v2t

    .

    .

    yNt N BNxN,t NyNt1 vNt

    1Ni

    B 1NBi

    1

    N

    i

    Means

    Panel unit root tests

  • 7/28/2019 1 an Introduction to Panel Data Methods

    47/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Panel unit root tests

    Levin and Lin (1992) and Levin Lin and Chu(2002). Model is:

    Test

    The equation can be augmented to accountfor serial correlation

    Assume that all series have the sameunderthe alternative hypothesis

    y it i t y it1 c it v it

    H0 : 0 vsH0 : 0

    Panel unit root tests

  • 7/28/2019 1 an Introduction to Panel Data Methods

    48/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Panel unit root tests

    Instead, Im, Pesaran and Shin (1992-

    2003) Model is:

    Testbased on the averageT-ratio across i

    Maddala and Wu 1999 suggest the teststatistic

    y it i t iy it1 c it v it

    0 : i 0 vs H0 : i 0

    2i1N lnp i~22N

    P-value for ith ADF regression

    Panel Unit Root tests

  • 7/28/2019 1 an Introduction to Panel Data Methods

    49/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Panel Unit Root tests

    These unit root tests assume cross

    sections are uncorrelated. Pesaran (2007) proposes the following

    modification to the IPS test

    yit i t y it1 y t1 c it v it

    Average across cross sectionsProxy for common component

    Panel unit root tests

  • 7/28/2019 1 an Introduction to Panel Data Methods

    50/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Panel unit root tests

    Interpretation can be difficult

    If the null hypothesis is rejected it does notmean that all series are stationary

    It may simply mean some of the series inthe panel are stationary

    Panel data in Eviews

  • 7/28/2019 1 an Introduction to Panel Data Methods

    51/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Panel data in Eviews

    Organising data

    Date and

    id

  • 7/28/2019 1 an Introduction to Panel Data Methods

    52/56

    Panel data in Eviews: reshaping

  • 7/28/2019 1 an Introduction to Panel Data Methods

    53/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    data

    Re-shaping data into a panel

  • 7/28/2019 1 an Introduction to Panel Data Methods

    54/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Re-shaping data into a panel

    Re-shaping data into a panel

  • 7/28/2019 1 an Introduction to Panel Data Methods

    55/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Re-shaping data into a panel

    Estimation

  • 7/28/2019 1 an Introduction to Panel Data Methods

    56/56

    The Bank of England does not accept any liability for misleading or inaccurate information or omissions in the information provided.

    [email protected]@bankofengland.co.uk Inflation TargetingInflation Targeting Bank of EnglandBank of England

    Estimation

    Once this is done standard equation

    commands apply. Goto->quickestimate equation to

    estimate model