51 views

Uploaded by Kamill Batabaevv

- Testing Sample Write Up
- Host Country Benefits
- 14093_Ch01 [Compatibility Mode]
- Exhaustive Regression An Exploration of Regression-Based Data Mining Techniques Using Super Computation
- regression analysis
- Excellent Market Share and Import File
- Eco No Metrics Painless Assignment
- Regression Tutorial 101 With NumXL
- CORRELATION OF THE PHYSICAL PROPERTIES OF ETHANE
- 1
- fox - intro to struct eqn modelling
- Robust Two-Stage Least Squares: Some Monte Carlo Experiments
- Paper 3 December 2008
- 10.1.1.153
- Progress and Poverty in Early Modern Europe
- The BUGS Project Examples Volume 1
- Obes 12012
- Hasil Sas Ral Non Fakt
- Impact of Blast Fragmentation on Excavator Dig Time1
- cu stat3008 Assignment 1

You are on page 1of 9

ECO 7427

ANSWER KEY

Prof. Sarah Hamersma

1. For your data work this week, I would like you to do exercise 5.4 in Wooldridge (all

parts). The data are available on the shared drive in the Economics department or on

Wooldridges website at: http://www.msu.edu/~ec/faculty/wooldridge/book2.htm

The paper you will replicate is Card 1995. It was published in a book, but there are copies

of the working paper version online (NBER # 4483) for your reference.

Wooldridge Ch 5-4

Here is my Stata code, with the verbal answers to the questions embedded in it.

------------------------------------------------------------------------------log: G:\Wooldridge5-4.log

log type: text

opened on: 16 Feb 2005, 11:03:07

.

.

. * Sarah Hamersma

. * 2/15/05

. * program name: Wooldridge5-4.do

.

. * This program provides an answer key to Wooldridge question 5.4

.

. #delimit ;

delimiter now ;

. use "H:\Wooldridge Data\CARD.DTA", clear;

. * Part a ;

. gen logwage = log(wage);

. regress logwage educ exper expersq black south smsa reg661 reg662

>

reg663 reg664 reg665 reg666 reg667 reg668 smsa66;

Source |

SS

df

MS

-------------+-----------------------------Model | 177.695591

15 11.8463727

Residual | 414.946054 2994 .138592536

-------------+-----------------------------Total | 592.641645 3009 .196956346

Number of obs

F( 15, 2994)

Prob > F

R-squared

Adj R-squared

Root MSE

=

=

=

=

=

=

3010

85.48

0.0000

0.2998

0.2963

.37228

-----------------------------------------------------------------------------logwage |

Coef.

Std. Err.

t

P>|t|

[95% Conf. Interval]

-------------+---------------------------------------------------------------educ |

.0746933

.0034983

21.35

0.000

.0678339

.0815527

exper |

.084832

.0066242

12.81

0.000

.0718435

.0978205

expersq |

-.002287

.0003166

-7.22

0.000

-.0029079

-.0016662

black | -.1990123

.0182483

-10.91

0.000

-.2347927

-.1632318

south |

-.147955

.0259799

-5.69

0.000

-.1988952

-.0970148

smsa |

.1363845

.0201005

6.79

0.000

.0969724

.1757967

reg661 | -.1185698

.0388301

-3.05

0.002

-.194706

-.0424335

reg662 | -.0222026

.0282575

-0.79

0.432

-.0776088

.0332036

reg663 |

.0259703

.0273644

0.95

0.343

-.0276846

.0796251

reg664 | -.0634942

.0356803

-1.78

0.075

-.1334546

.0064662

reg665 |

.0094551

.0361174

0.26

0.794

-.0613623

.0802725

reg666 |

.0219476

.0400984

0.55

0.584

-.0566755

.1005708

reg667 | -.0005887

.0393793

-0.01

0.988

-.077802

.0766245

reg668 | -.1750058

.0463394

-3.78

0.000

-.265866

-.0841456

smsa66 |

.0262417

.0194477

1.35

0.177

-.0118905

.0643739

_cons |

4.739377

.0715282

66.26

0.000

4.599127

4.879626

-----------------------------------------------------------------------------.

>

>

>

>

>

>

>

>

>

*

*

*

*

*

*

*

*

difference is that Card uses "expersq/100" as his regressor so

his coefficient is exactly 100 times the size of ours (but this

affects the std error the same way, so the significance level

is identical). This is a useful place to note that it can be easier

for the reader if you scale variables for which the coefficient

would be very very small, to make it easier to interpret, which

is what Card did.

> * Part b ;

. regress educ exper expersq black south smsa reg661 reg662

>

reg663 reg664 reg665 reg666 reg667 reg668 smsa66 nearc4;

Source |

SS

df

MS

-------------+-----------------------------Model | 10287.6179

15 685.841194

Residual | 11274.4622 2994 3.76568542

-------------+-----------------------------Total | 21562.0801 3009 7.16586243

Number of obs

F( 15, 2994)

Prob > F

R-squared

Adj R-squared

Root MSE

=

=

=

=

=

=

3010

182.13

0.0000

0.4771

0.4745

1.9405

-----------------------------------------------------------------------------educ |

Coef.

Std. Err.

t

P>|t|

[95% Conf. Interval]

-------------+---------------------------------------------------------------exper | -.4125334

.0336996

-12.24

0.000

-.4786101

-.3464566

expersq |

.0008686

.0016504

0.53

0.599

-.0023674

.0041046

black | -.9355287

.0937348

-9.98

0.000

-1.11932

-.7517377

south | -.0516126

.1354284

-0.38

0.703

-.3171548

.2139296

smsa |

.4021825

.1048112

3.84

0.000

.1966732

.6076918

reg661 |

-.210271

.2024568

-1.04

0.299

-.6072395

.1866975

reg662 | -.2889073

.1473395

-1.96

0.050

-.5778042

-.0000105

reg663 | -.2382099

.1426357

-1.67

0.095

-.5178838

.0414639

reg664 |

-.093089

.1859827

-0.50

0.617

-.4577559

.2715779

reg665 | -.4828875

.1881872

-2.57

0.010

-.8518767

-.1138982

reg666 | -.5130857

.2096352

-2.45

0.014

-.9241293

-.1020421

reg667 | -.4270887

.2056208

-2.08

0.038

-.8302611

-.0239163

reg668 |

.3136204

.2416739

1.30

0.194

-.1602433

.7874841

smsa66 |

.0254805

.1057692

0.24

0.810

-.1819071

.2328682

nearc4 |

.3198989

.0878638

3.64

0.000

.1476194

.4921785

_cons |

16.84852

.2111222

79.80

0.000

16.43456

17.26248

-----------------------------------------------------------------------------.

>

>

>

*

*

*

*

large - about 1/3 year added education (on a base of about 13 years)

for those living near a college and it contributes to the regression meaningfully. One way to assert

>

.

>

>

>

>

*

*

*

*

*

since there is only one instrument, we can simply look at a t-test of

the instrument in the first-stage regression and we can see that it

has statistical explanatory power (see Wooldridge top of p. 105). The

size and significance suggest a reasonably strong instrument.

> * Part c ;

. ivreg logwage exper expersq black south smsa reg661 reg662

>

reg663 reg664 reg665 reg666 reg667 reg668 smsa66 (educ = nearc4);

Instrumental variables (2SLS) regression

Source |

SS

df

MS

-------------+-----------------------------Model | 141.146813

15 9.40978752

Residual | 451.494832 2994 .150799877

-------------+-----------------------------Total | 592.641645 3009 .196956346

Number of obs

F( 15, 2994)

Prob > F

R-squared

Adj R-squared

Root MSE

=

=

=

=

=

=

3010

51.01

0.0000

0.2382

0.2343

.38833

-----------------------------------------------------------------------------logwage |

Coef.

Std. Err.

t

P>|t|

[95% Conf. Interval]

-------------+---------------------------------------------------------------educ |

.1315038

.0549637

2.39

0.017

.0237335

.2392742

exper |

.1082711

.0236586

4.58

0.000

.0618824

.1546598

expersq | -.0023349

.0003335

-7.00

0.000

-.0029888

-.001681

black | -.1467757

.0538999

-2.72

0.007

-.2524603

-.0410912

south | -.1446715

.0272846

-5.30

0.000

-.19817

-.091173

smsa |

.1118083

.031662

3.53

0.000

.0497269

.1738898

reg661 | -.1078142

.0418137

-2.58

0.010

-.1898007

-.0258278

reg662 | -.0070465

.0329073

-0.21

0.830

-.0715696

.0574767

reg663 |

.0404445

.0317806

1.27

0.203

-.0218694

.1027585

reg664 | -.0579172

.0376059

-1.54

0.124

-.1316532

.0158189

reg665 |

.0384577

.0469387

0.82

0.413

-.0535777

.130493

reg666 |

.0550887

.0526597

1.05

0.296

-.0481642

.1583416

reg667 |

.026758

.0488287

0.55

0.584

-.0689832

.1224992

reg668 | -.1908912

.0507113

-3.76

0.000

-.2903238

-.0914586

smsa66 |

.0185311

.0216086

0.86

0.391

-.0238381

.0609003

_cons |

3.773965

.934947

4.04

0.000

1.940762

5.607169

-----------------------------------------------------------------------------Instrumented: educ

Instruments:

exper expersq black south smsa reg661 reg662 reg663 reg664

reg665 reg666 reg667 reg668 smsa66 nearc4

-----------------------------------------------------------------------------.

>

>

>

>

>

>

>

*

*

*

*

*

*

*

*

The new estimate of the return to education is almost twice as high (13%

vs. 7.5% before). The 95% confidence interval here is (.024, .239). The

earlier one was (.068,.082). We have a lot less precision with the IV

procedure. However, this lack of precision is appropriate. The precision

is part (a) is false in the sense that the estimates are not even

consistent since educ is endogenous (plus they are biased). When we

account for the endogeneity, we use a two-stage IV procedure that will

result in less precision but consistency and unbiasedness of the estimate.;

. * Part d ;

. regress educ exper expersq black south smsa reg661 reg662

>

reg663 reg664 reg665 reg666 reg667 reg668 smsa66 nearc4 nearc2;

Source |

SS

df

MS

-------------+-----------------------------Model | 10297.1164

16 643.569774

Number of obs =

F( 16, 2993) =

Prob > F

=

3010

170.99

0.0000

-------------+-----------------------------Total | 21562.0801 3009 7.16586243

R-squared

=

Adj R-squared =

Root MSE

=

0.4776

0.4748

1.94

-----------------------------------------------------------------------------educ |

Coef.

Std. Err.

t

P>|t|

[95% Conf. Interval]

-------------+---------------------------------------------------------------exper | -.4122915

.0336914

-12.24

0.000

-.4783521

-.3462309

expersq |

.0008479

.00165

0.51

0.607

-.0023874

.0040832

black | -.9451729

.0939073

-10.06

0.000

-1.129302

-.7610434

south | -.0419115

.1355316

-0.31

0.757

-.3076561

.2238331

smsa |

.4013708

.1047858

3.83

0.000

.1959113

.6068303

reg661 | -.1687829

.2040832

-0.83

0.408

-.5689404

.2313747

reg662 |

-.269031

.1478324

-1.82

0.069

-.5588944

.0208325

reg663 | -.1902114

.1457652

-1.30

0.192

-.4760216

.0955987

reg664 |

-.037715

.1891745

-0.20

0.842

-.4086403

.3332102

reg665 | -.4371387

.1903306

-2.30

0.022

-.8103307

-.0639467

reg666 | -.5022265

.2096933

-2.40

0.017

-.9133841

-.0910688

reg667 | -.3775317

.207922

-1.82

0.070

-.7852162

.0301529

reg668 |

.3820043

.2454171

1.56

0.120

-.0991991

.8632076

smsa66 |

.0000782

.1069445

0.00

0.999

-.2096139

.2097704

nearc4 |

.3205819

.0878425

3.65

0.000

.148344

.4928197

nearc2 |

.1229986

.0774256

1.59

0.112

-.0288142

.2748114

_cons |

16.77306

.2163481

77.53

0.000

16.34885

17.19727

-----------------------------------------------------------------------------.

>

.

>

>

* more precisely estimated for nearc4;

ivreg logwage exper expersq black south smsa reg661 reg662

reg663 reg664 reg665 reg666 reg667 reg668 smsa66 (educ = nearc4 nearc

2);

Source |

SS

df

MS

-------------+-----------------------------Model |

100.869

15 6.72459998

Residual | 491.772645 2994

.16425272

-------------+-----------------------------Total | 592.641645 3009 .196956346

Number of obs

F( 15, 2994)

Prob > F

R-squared

Adj R-squared

Root MSE

=

=

=

=

=

=

3010

47.07

0.0000

0.1702

0.1660

.40528

-----------------------------------------------------------------------------logwage |

Coef.

Std. Err.

t

P>|t|

[95% Conf. Interval]

-------------+---------------------------------------------------------------educ |

.1570594

.0525782

2.99

0.003

.0539662

.2601525

exper |

.1188149

.0228061

5.21

0.000

.0740977

.163532

expersq | -.0023565

.0003475

-6.78

0.000

-.0030379

-.0016751

black | -.1232778

.05215

-2.36

0.018

-.2255313

-.0210243

south | -.1431945

.0284448

-5.03

0.000

-.1989678

-.0874212

smsa |

.100753

.0315193

3.20

0.001

.0389512

.1625548

reg661 |

-.102976

.0434224

-2.37

0.018

-.1881167

-.0178353

reg662 | -.0002286

.0337943

-0.01

0.995

-.066491

.0660337

reg663 |

.0469556

.032649

1.44

0.150

-.0170612

.1109724

reg664 | -.0554084

.0391828

-1.41

0.157

-.1322364

.0214196

reg665 |

.0515041

.0475678

1.08

0.279

-.0417647

.144773

reg666 |

.0699968

.0533049

1.31

0.189

-.0345212

.1745148

reg667 |

.0390596

.0497499

0.79

0.432

-.0584878

.136607

reg668 | -.1980371

.052535

-3.77

0.000

-.3010454

-.0950287

smsa66 |

.0150626

.022336

0.67

0.500

-.0287328

.058858

_cons |

3.339687

.8945377

3.73

0.000

1.585716

5.093658

-----------------------------------------------------------------------------Instrumented: educ

Instruments:

exper expersq black south smsa reg661 reg662 reg663 reg664

-----------------------------------------------------------------------------.

>

>

>

>

>

>

>

*

*

*

*

*

*

*

*

The new estimate of the returns to educ is even higher, at 15.7%. There

is again sufficient precision to say with confidence that there is a

positive effect of education on earnings, but the confidence interval is

still fairly wide at (.054,.260). It's notable, though, that the bottom

end of the interval is not that much lower (in magnitude) than the point

estimate in the OLS where we ignored endogeneity. Although we cannot be

certain, this seems to suggest that the uncorrected OLS estimate was likely

an underestimate of the real return to education. ;

. * Part e ;

. regress iq nearc4;

Source |

SS

df

MS

-------------+-----------------------------Model | 2869.62905

1 2869.62905

Residual | 487188.423 2059 236.614096

-------------+-----------------------------Total | 490058.052 2060 237.892258

Number of obs

F( 1, 2059)

Prob > F

R-squared

Adj R-squared

Root MSE

=

=

=

=

=

=

2061

12.13

0.0005

0.0059

0.0054

15.382

-----------------------------------------------------------------------------iq |

Coef.

Std. Err.

t

P>|t|

[95% Conf. Interval]

-------------+---------------------------------------------------------------nearc4 |

2.5962

.7454966

3.48

0.001

1.134195

4.058206

_cons |

100.6106

.6274557

160.35

0.000

99.38014

101.8412

-----------------------------------------------------------------------------.

>

>

>

>

>

>

*

*

*

*

*

They are correlated - probably not that shocking, given profs' smart kids!

We might be concerned about this correlation because IQ could affect

wages directly, so nearc4 could be picking up IQ effects, which would make

nearc4 correlated with the error in the outcome equation (that would be

very bad!).

. * Part f ;

. regress iq nearc4 smsa66 reg661 reg662 reg669;

Source |

SS

df

MS

-------------+-----------------------------Model | 14792.5727

5 2958.51453

Residual |

475265.48 2055

231.27274

-------------+-----------------------------Total | 490058.052 2060 237.892258

Number of obs

F( 5, 2055)

Prob > F

R-squared

Adj R-squared

Root MSE

=

=

=

=

=

=

2061

12.79

0.0000

0.0302

0.0278

15.208

-----------------------------------------------------------------------------iq |

Coef.

Std. Err.

t

P>|t|

[95% Conf. Interval]

-------------+---------------------------------------------------------------nearc4 |

.8680808

.8216913

1.06

0.291

-.7433537

2.479515

smsa66 |

1.354527

.8027961

1.69

0.092

-.2198513

2.928906

reg661 |

4.768099

1.546809

3.08

0.002

1.734623

7.801576

reg662 |

5.80812

.9017539

6.44

0.000

4.039673

7.576566

reg669 |

1.844655

1.151703

1.60

0.109

-.4139708

4.103281

_cons |

99.38472

.7016631

141.64

0.000

98.00868

100.7608

-----------------------------------------------------------------------------. * I'm not sure why we didn't use the whole set of dummies here...anyway,

> * this is good - IQ and nearc4 no longer appear to be partially correlated.

> * Or, at least, there is not a strong enough correlation for us to be

>

>

>

>

>

.

* for us to control for 1966 location and regional dummies in the outcome

* equation because these soak up the effects of IQ in a way that allows

* the instrument to end up uncorrelated with the error in the outcome

* equation (which is required in order for us to legitimately use IV).;

log close;

log: G:\Wooldridge5-4.log

log type: text

closed on: 16 Feb 2005, 11:03:08

instrument. Specifically, tell me about differences in the assumptions required for each to

be valid and differences in the type of situation in which it would be useful to use such a

variable.

Proxy:

Use a proxy when you need a representative for an omitted variable for which you

dont have direct data (such as using IQ to proxy for ability in a returns-toeducation context). You place it directly into the regression to represent the

variable you dont have data for. The interpretation of the coefficient is the

predictive power of the proxy on the outcome. Note that we still cannot measure

the effect of the omitted variable a proxy just acts as a control so that the other

coefficients in the regression arent biased.

Assumptions:

a) uncorrelated with the error in the outcome equation (this is also called

redundant in the structural equation if we had the real variable, this

one would be redundant)

b) correlated with the omitted variable

- more specifically, it should be closely related enough to the omitted

variable that the other Xs have no power for predicting the omitted

variable once the proxy is taken into account (though theres no way

to check this exactly, since we dont have data on the omitted

variable)

Instrument:

Use an instrument when you have data on an endogenous variable that you think

is correlated with some other omitted variable in your outcome equation, causing

the regression estimates to be biased. An instrument is used to represent the

endogenous variable (NOT the omitted one) and if we consider the singlevariable case, the instrument is put into the outcome equation directly. While we

could look at the coefficient on the instrument, we are typically interested in the

effect of the endogenous variable, which we get by dividing the coefficient on the

instrument in the outcome equation by the coefficient on the instrument from the

first-stage.

6

Assumptions:

a) uncorrelated with the error in the outcome equation (this is also called

redundant in the structural equation if we had a clean version of the

endogenous regressor (without its implicit correlation with some other

omitted variable) then the instrument would be redundant)

b) correlated with the endogenous variable (and NOT with the omitted

variable that is causing the endogenous regressor to be endogenous if it is

correlated with the omitted variable, it will fail to meet assumption (a).)

In terms of comparing the two one clear similarity is in the assumptions

(particularly the first one). However, a clear difference is that they are used to fix

different problems. In one case (instrument) we have a variable of interest but it

is correlated with some omitted variable, preventing us from estimating the effect

properly. We want a representative that will get rid of the endogenous part of the

variable of interest. In the case of a proxy, controlling for the omitted variable

itself is of interest, and we are looking for a way to do this with some substitute

because the data are not available. In a very practical sense, these are distinct in

that there can be no first stage in a proxy setting because we do not have data

on the variable we are trying to represent (and if we did, we wouldnt need the

proxy!).

3. Regarding Lotts work: I would like you to tell me if you think this (his website

defense) is a sufficient argument for choosing not to use clustering in the analysis.

Do your best to convince me of your position by explaining why the analysis does or

does not need clustering.

This was a hard question. I gave substantial partial credit for wrong answers that were well

thought-out. But please do make sure you read this so you know the right answer.

Outline of answer:

a) when clustering standard errors is still needed, even with dummies

b) explanation of what dummies can and cannot successfully fix

c) explanation of why clustering will make SEs bigger even if its unneeded

John Lotts analysis uses county-level data from several states and looks at the impact of

state-level treatments. Note that he does not use individual data at all the unit of

observation is the county. This means when he refers to using county fixed effects, this is

equivalent to an individual fixed effect from the perspective of his sample where each

observation is a county. He argues that including county fixed effects implicitly includes

state fixed effects. This argument is correct. However, this only moves us one step closer to

the real question: Does including state fixed-effects mean you dont need clustering at the

state level? The answer is that you still may need clustering.

State fixed effects are an important component of an analysis that uses state-level treatments.

There may be correlated outcomes Y within a state that are not picked up by observable Xs.

This can be thought of as an omitted variables (endogeneity) problem so if this is the case,

and we do not include state fixed effects, our estimates of the treatment effect will be biased

and inconsistent (not to mention the standard errors!). Including a state fixed effect allows

us to explain some of this variation. Econometrically, it will force the expected value of the

residuals within each state to be zero (if they averaged something else, this would have been

incorporated into the estimate of the fixed effect by construction).

Suppose that these state fixed effects properly fix the point estimates (i.e. there is no longer

an omitted variables problem). What does the error structure look like now? Well, within

each state there are several counties. We can estimate a regression and look at the residuals

within each state they will average zero (as noted above) but depending on the state they

might be spread widely or distributed narrowly around zero. This is a heteroskedasticity

problem solve it with the robust function to fix your standard errors.

Where does the clustering come in? It is worth noting that the clustering problem would

have been HUGE if we ignored the fixed effects to start with, and so including them does

make the problem smaller (which is why some of our intuition suggested that it could fix the

problem). However, it may still remain. The issue is that we have controlled only for a very

specific form of correlation among observations within a state we have controlled for a

form of correlation in which every observation in the state has a common (state-level)

component of variance and a random component that is individual-specific (or, in Lotts

case, county-specific). We have assumed all states have this same within-state correlation

structure. It is conceivable, though, that there are other correlations among counties in a

state that are not picked up by this very simple model of correlation. Wooldridge, in his

paper Cluster-Sample Methods in Applied Econometrics, says that an example would be

something that is somehow related to the other Xs in the regression...such as if people

within certain states tend to have certain Xs that are related to certain error-term patterns,

which could cause a complication in the relationships among the errors within a state.

Clustering the standard errors, along with making them robust (Stata does this

automatically), will address this problem. (However, let me note that in the case described

by Wooldridge there it seems there may also be an endogeneity problem if there is some

nonrandom relationship between Xs and error terms).

Mitch Petersons paper Estimating Standard Errors in Finance Panel Data Sets:

Comparing Approaches also addresses this issue and gives a nice example of a situation in

which clustering is still needed in the presence of fixed effects. He examines the use of

various standard error corrections in the presence of different types of error correlation.

Some key insights are on pages 6-8, Section IV (starting page 23), and Section V (starting

page 26). I have pasted the most transparent part of the paper for our purposes below.

Petersons point is that adding firm (or in our case county) fixed effects will fix everything IF

the only correlation among counties is a fixed, time-invariant component. (This echoes

Wooldridge). He notes that this will fail if there is a gradually-changing firm effect. I would

add that the same may be true for a geographically-based correlations (nearby counties may

be more correlated with each other than distant counties, even within a state).

Once we include the firm effects, the OLS standard errors are unbiased .... The clustered standard errors

are unbiased with and without the fixed effects (see Kezdi, 2004, for examples where the clustered standard

errors are too large in a fixed effect model). This conclusion, however, depends on the firm effect being fixed. If

the firm effect decays over time, the firm dummies no longer fully capture the within cluster dependence and

OLS standard errors are still biased (see Table 5 - Panel A, columns II-IV). In these simulations, the firm

effect decays over time (in column II, 61 percent of the firm effect dissipates after 9 years). Once the firm

effect is temporary, the OLS standard errors again underestimate the true standard

errors even when firm dummies are included in the regression (Wooldridge, 2003, Baker,

Stein, and Wurgler, 2003). (p. 28)

If it happens that you are certain that the error structure of your data is perfectly picked up

with fixed effects, you will not need to cluster. Moreover, you will not want to cluster.

Why? Your standard errors will get unnecessarily larger. But why should they change,

especially given the way Moulton (1990) presented the formula for the adjustment (which

seems to imply that if there is no correlation, no adjustment is made)? The intuition here is

that anytime we allow for more flexibility of estimation, it costs us something. Estimating

these flexible standard errors causes a loss of efficiency you can think of it as using up

some of our observations (degrees of freedom) to calculate these special standard errors.

This would lead us to want to KNOW whether we need clustering, since we wouldnt want

to use it unnecessarily.

A newer (2004) paper by Lott posted on his website contains an appendix with his argument

for why any correlation in his errors is taken care of with his fixed effects (though he now

includes clustering throughout the paper, for comparability with other work and to be

conservative about his estimates). He does a test for correlation of errors to argue his point

(which myself and another econometrics colleague have found to be weak at best). This

seems a step in the right direction the idea being that one must still make some kind of

argument for choosing NOT to cluster, even when fixed effects are included. The

argument that fixed effects are included is not itself a sufficient reason to avoid

clustering.

- Testing Sample Write UpUploaded bySteven Troxler
- Host Country BenefitsUploaded byMohamed Salem
- 14093_Ch01 [Compatibility Mode]Uploaded byMarsha Mae
- Exhaustive Regression An Exploration of Regression-Based Data Mining Techniques Using Super ComputationUploaded byStopSpyingOnMe
- regression analysisUploaded byayu7kaji
- Excellent Market Share and Import FileUploaded byHaroon Sajjad
- Eco No Metrics Painless AssignmentUploaded byGlen O'Donoghue
- Regression Tutorial 101 With NumXLUploaded bySpider Financial
- CORRELATION OF THE PHYSICAL PROPERTIES OF ETHANEUploaded byMirko Granese
- 1Uploaded bySambit Mishra
- fox - intro to struct eqn modellingUploaded bykhalkhalesh
- Robust Two-Stage Least Squares: Some Monte Carlo ExperimentsUploaded bySudhanshu K Mishra
- Paper 3 December 2008Uploaded byecobalas7
- 10.1.1.153Uploaded byhandegokdeniz
- Progress and Poverty in Early Modern EuropeUploaded byjeni_paraschiv
- The BUGS Project Examples Volume 1Uploaded byferalaes
- Obes 12012Uploaded byKevin Yang
- Hasil Sas Ral Non FaktUploaded byDiego Martin
- Impact of Blast Fragmentation on Excavator Dig Time1Uploaded bytkujun
- cu stat3008 Assignment 1Uploaded byJim Hack
- Regression in DepthUploaded bypratik kumar
- HW3Uploaded byrogervalen5049
- 10-31Uploaded bybunsinh
- regression.pdfUploaded byapriani_herni
- daun bawahUploaded byDanar Fahmi Sudarsono
- Time Series TestUploaded byscribd438
- UTS 2Uploaded byhendri
- Econometrics 1Uploaded byleonardmakuvaza
- Chapter1 Labor EcoUploaded bykokabm
- Ch03 PPT Hongren CostAccounting 2eUploaded byLiu Qi

- Manual Sony FST-ZX80DUploaded by̶D̶j̶a̶l̶m̶a̶ ̶M̶o̶r̶e̶i̶r̶a̶
- IRJET-Performance Analysis of Electric Bike using Battery and SupercapacitorUploaded byIRJET Journal
- Hero Honda PptUploaded byVaishnavi Suthrai
- Pure Pro User Manual _V1.30_PMODEUploaded byErick DC
- Methods of Resisting Hydro Static Uplift in SubstructuresUploaded byIsabelle Wong
- Transbay comUploaded byAdiba Mahroze Hoque
- Seismic Analysis and Design of Hospital Building-libreUploaded bypaarth93
- Airport Curbside and Terminal Area Roadway Operations ACRPUploaded byGioPed19
- Radiology Final QuestionsUploaded byRashed Shatnawi
- Ah Lao Sac Hotel Assessment Form 2014Uploaded byyank
- Gai-Tronic Phone ManualUploaded byLeigh Smith
- 33rd Report ID PPPUploaded bySanjay Bose
- Erosion Abrasion Attrition and Abfrcation1Erosion Abrasion Attrition and Abfrcation1Uploaded byMadalina A Dobre
- U5_DragisicUploaded byalex ali
- A BRIEF REVIEW ON GENE THERAPYUploaded byiajps
- CostsUploaded byMahrukh Malik
- MoynihanKatrina.pdfUploaded byJohn Jackson
- Rural Road Investment Efficiency: Lessons from Burkina Faso, Cameroon, and UgandaUploaded byWorld Bank Publications
- Basis of Structural Analysis 0910Uploaded bylufias2009
- AA Skema Paper 3Uploaded byShahrul Anis
- ASME B16.1-89Uploaded byAntonio Carlos Rossini
- Welding ImperfectionsUploaded byAnonymous
- 2n6509 data sheetUploaded bypapasky
- Management of Riga Fede Disease a Case ReportUploaded bykermink
- ch02Uploaded bykrypton21
- BelkundChi Bakhar Marathi Jungle Story by Dr.v.TUploaded byapi-3777175
- oslUploaded byKaushikGarkoti
- 15 Spk-GenUploaded byRavi Shankar
- Plant as a Bio RobotUploaded byRajesh Kesapragada
- Husam's Reconstruction of Schaefer's Formula of Visibility of Crescent Of Luna [Revised]Uploaded byAHLUSSUNNAH WAL JAMAA"AH [ASHA:'IRAH & MATURIDIYAH]