Fixed effects models in social science

acceptablepeasΑσφάλεια

30 Νοε 2013 (πριν από 3 χρόνια και 8 μήνες)

73 εμφανίσεις

Fixed effects models in social science



This course covers the motivation and introduction to fixed effect modeling in both the time domain
and in other domains.

Fixed effect modeling is an alternative to random effects modeling. Both
approached provide estimators that deal with cluster based samples. In random effects models the
aim is to provide correct standard errors when data is nested within clusters. In

f
ix
ed effect
modeling
the
aim

is

in addition
to

remov
e

unobserved within cluster effects. For individuals
observed across time this would be unobserved time
-
invariant individual effects. For instance, when
analyzing students tests
-
scores within individual eff
ects could be unobserved time
-
invariant

personal
traits
.

For cross sectional units, e.g.
school
classes, countries or families this could
be
teacher effects such as cross class differences in unobserved teacher competencies, cross country
differences in un
observed institutional differences and unobserved family invariant variables.
The
benefit from using fixed effect estimators is that any spurious effects from unobserved variables on
observed variables of interest are removed. Hence fixed effects estimator
s are often closer to a
causal interpretation that ordinary cross sectional or random effects estimators.
Using the fixed
effect estimator is
often very straight forward, easy to implement and
facilitated in many canned
routines in statistical packages, such as
xtreg

in
stata
.
The fixed effects estimator is

also available
for non
-
linear models, e.g.
xtlogit

in
stata
. However, interpretation is often crucial depending on the
application at hand.
This course therefore aims at discussing at length several different
research
topics where fixed effects models are used

in order to inspire wider application of the fixed effect
technique.



I
ntroduction:

Allison,
Fixed Effecs regression models
, Sage
(2009)


Ejernaes and Holm (2006) "Comparing fixed effect and covariance structure estimators",
Sociological Methods and Research
, vol 35(1), 61
-
83.


Fixed effect in the time domain:

Halaby,

Panel Models in Sociological research: Theory and Practice

,
Annu
al

Rev
iew of

Sociol
ogy
,

30:507

44
.


F
ixed effect in other domains:

S
chwerdt

and

Wuppermann


Is traditional teaching really all that bad? A within
-
student

between
-
subject approach

,

Economics of Education Review

30 (2011) 365

379.


Black, Devereux and
Salvanes (2010) “Under Pressure? The Effect of Peers on Outcomes of Young
Adults”,
U
npublished WP.


McGue, Osler and Christensen (2010) “Causal inference and Observational Research: The utility of
Twins,
Perspectives on Psychological Science
, 5: 546
-
556.


Holm and Trolle (2011) “What did you learn in school today?


S
tudent achievement across time
using siblings and twin data”
,

Unpublished WP.


Mixing time and other domains:

Jæger (2011)
Does Cultural Capital Really Affect Academic Achievement? New Evidence from
Combined Sibling and Panel Data,
Sociology of Education,
84: 281
-
298


Todd and Smith (2003) On

the specification and estimation of the production function for cognitive
achievement,
Economic journal
, 113, 3
-
33.


Fletcher (2010) “
Spillover Effects of Inclusion of Classmates with Emotional Problems on Test
Scores in Early Elementary School,
Journal
of Policy Analysis and Management
, Vol. 29, No. 1,
69

83
.


Hanushek, Kain, Markman and Rivkin (2003) Does Peer Ability Affect Student Achievement?
Journal of Applied Econometrics
,
18
: 527

544 (2003)


Supplementary material:

Lancaster (2000) “The incidental parameter problem since 1948”,
Journal of Econometrics
, 95,
391
-
413


Neuhaus and Kalbfleisch (1998) “Bettween
-

and Within
-
Cluster Covariate Effects in the Analysis
of Clustered Data”,
Biometrics
, 54: 638
-
645.