Analysis of Panel Data, Second Edition
Author: 
Cheng Hsiao 
Publisher: 
Cambridge University Press 
Copyright: 
2003 
ISBN13: 
9780521522717 
Pages: 
366; paperback 
Price: 
$35.75 



Comment from the Stata technical group
The second edition of Hsiao’s Analysis of Panel Data is a
complete revision of the 1986 version that has served as an essential
reference on paneldata models. With the increased availability of panel
datasets, this book will be welcomed not only by economists but also by
researchers in other social sciences, business, and the natural sciences.
The judicious combination of theory and applications makes Hsiao’s
monograph useful both as a textbook for students and as a reference manual
for practitioners.
The first three chapters of the book provide a detailed introduction to
static random and fixedeffects models, including model estimation,
specification testing, and treatment of heteroskedasticity and correlation.
Chapter 4 is a concise overview of dynamic paneldata models, including the
GMM estimators that have become quite popular in recent years, and Hsiao
does a superb job of bringing the literature together into one cohesive
discussion. Simultaneousequation estimation is covered in chapter 5.
Chapter 6 is devoted to variablecoefficient models, including Swamy’s
random coefficients model and its relatives: fixedcoefficients models and
their dynamic counterparts. Chapters 7 and 8 discuss parametric and
semiparametric approaches to limited dependentvariable models and provide
many references to the growing literature in this field; chapter 7 also
includes a clear presentation of the incidental parameters problem.
Chapters 9 and 10 present a smorgasbord of further topics, including
incomplete panels, pseudopanels, largeN and largeT
asymptotics, and multilevel models. The sections in these two chapters are
rather brief, but they do provide many references to the relevant
literature. Chapter 11 summarizes the use of paneldata models.
In short, the second edition of Analysis of Panel Data will prove to
be an invaluable reference to users of panel data, just as the first edition
has for 14 years.
Table of contents
Preface to the Second Edition
Preface to the First Edition
1. Introduction
1.1 Advantages of Panel Data
1.2 Issues Involved in Utilizing Panel Data
1.2.1 Heterogeneity Bias
1.2.2 Selectivity Bias
1.3 Outline of the Monograph
2. Analysis of Covariance
2.1 Introduction
2.2 Analysis of Covariance
2.3 An Example
3. Simple Regression with Variable Intercepts
3.1 Introduction
3.2 FixedEffects Models: LeastSquares DummyVariable Approach
3.3 RandomEffects Models: Estimation of VarianceComponents Models
3.3.1 Covariance Estimation
3.3.2 GeneralizedLeastSquares Estimation
3.3.3 Maximum Likelihood Estimation
3.4 Fixed Effects or Random Effects
3.4.1 An Example
3.4.2 Conditional Inference or Unconditional (Marginal) Inference
3.4.2.a Mundlak’s Formulation
3.4.2.b Conditional and Unconditional Inferences in the
Presence or Absence of Correlation between Individual
Effects and Attributes
3.5 Tests for Misspecification
3.6 Models with Specific Variables and Both Individual and
TimeSpecific Effects
3.6.1 Estimation of Models with IndividualSpecific Variables
3.6.2 Estimation of Models with Both Individual and Time Effects
3.7 Heteroscedasticity
3.8 Models with Serially Correlated Errors
3.9 Models with Arbitrary Error Structure  Chamberlain π Approach
Appendix 3A: Consistency and Asymptotic Normality of the
MinimumDistance Estimator
Appendix 3B: Characteristic Vectors and the Inverse of the
Variance–Covariance Matrix of a ThreeComponent Model
4. Dynamic Models with Variable Intercepts
4.1 Introduction
4.2 The Covariance Estimator
4.3 The RandomEffects Models
4.3.1 Bias in the OLS Estimator
4.3.2 Model Formulation
4.3.3 Estimation of RandomEffects Models
4.3.3.a Maximum Likelihood Estimator
4.3.3.b GeneralizedLeastSquares Estimator
4.3.3.c InstrumentalVariable Estimator
4.3.3.d Generalized Methods of Moments Estimator
4.3.4 Testing Some Maintained Hypotheses on Initial Conditions
4.3.5 Simulation Evidence
4.4 An Example
4.5 FixedEffects Models
4.5.1 Transformed Likelihood Approach
4.5.2 MinimumDistance Estimator
4.5.3 Relations between the LikelihoodBased Estimator and the
Generalized Method of Moments Estimator (GMM)
4.5.4 Random versus FixedEffects Specification
4.6 Estimation of Dynamic Models with Arbitrary Correlations in the
Residuals
4.7 FixedEffects Vector Autoregressive Models
4.7.1 Model Formulation
4.7.2 Generalized Method of Moments (GMM) Estimation
4.7.3 (Transformed) Maximum Likelihood Estimator
4.7.4 MinimumDistance Estimator
Appendix 4A: Derivation of the Asymptotic Covariance Matrix of the
Feasible MDE
5. SimultaneousEquations Models
5.1 Introduction
5.2 Joint GeneralizedLeastSquares Estimation Technique
5.3 Estimation of Structural Equations
5.3.1 Estimation of a Single Equation in the Structural Model
5.3.2 Estimation of the Complete Structural System
5.4 Triangular System
5.4.1 Identification
5.4.2 Estimation
5.4.2.a InstrumentalVariable Method
5.4.2.b MaximumLikelihood Method
5.4.3 An Example
Appendix 5A
6. VariableCoefficient Models
6.1 Introduction
6.2 Coefficients That Vary over CrossSectional Units
6.2.1 FixedCoefficient Model
6.2.2 RandomCoefficient Model
6.2.2.a The Model
6.2.2.b Estimation
6.2.2.c Predicting Individual Coefficients
6.2.2.d Testing for Coefficient Variation
6.2.2.e Fixed or Random Coefficients
6.2.2.f An Example
6.3 Coefficients That Vary over Time and CrossSectional Units
6.3.1 The Model
6.3.2 FixedCoefficient Model
6.3.3 RandomCoefficient Model
6.4 Coefficients That Evolve over Time
6.4.1 The Model
6.4.2 Predicting B_{t} by the Kalman Filter
6.4.3 Maximum Likelihood Estimation
6.4.4 Tests for Parameter Constancy
6.5 Coefficients That Are Functions of Other Exogenous Variables
6.6 A Mixed Fixed and RandomCoefficients Model
6.6.1 Model Formulation
6.6.2 A Bayes Solution
6.6.3 An Example
6.6.4 Random or Fixed Parameters
6.6.4.a An Example
6.6.4.b Model Selection
6.7 Dynamic RandomCoefficient Models
6.8 An Example — Liquidity Constraints and Firm Investment Expenditure
Appendix 6A: Combination of Two Normal Distributions
7. Discrete Data
7.1 Introduction
7.2 Some DiscreteResponse Models
7.3 Parametric Approach to Static Models with Heterogeneity
7.3.1 FixedEffects Models
7.3.1.a Maximum Likelihood Estimator
7.3.1.b Conditions for the Existence of a Consistent Estimator
7.3.1.c Some Monte Carlo Evidence
7.3.2 RandomEffects Model
7.4 Semiparametric Approach to Static Models
7.4.1 Maximum Score Estimator
7.4.2 A RootN Consistent Semiparametric Estimator
7.5 Dynamic Models
7.5.1 The General Model
7.5.2 Initial Conditions
7.5.3 A Conditional Approach
7.5.4 State Dependence versus Heterogeneity
7.5.5 Two Examples
7.5.5.a Female Employment
7.5.5.b Household Brand Choices
8. Truncated and Censored Data
8.1 Introduction
8.2 An Example — Nonrandomly Missing Data
8.2.1 Introduction
8.2.2 A Probability Model of Attrition and Selection Bias
8.2.3 Attrition in the Gary IncomeMaintenance Experiment
8.3 Tobit Models with Random Individual Effects
8.4 FixedEffects Estimator
8.4.1 Pairwise Trimmed LeastSquares and LeastAbsoluteDeviation
Estimators for Truncated and Censored Regressions
8.4.1.a Truncated Regression
8.4.1.b Censored Regressions
8.4.2 A Semiparametric TwoStep Estimator for the Endogenously
Determined Sample Selection Model
8.5 An Example: Housing Expenditure
8.6 Dynamic Tobit Models
8.6.1 Dynamic Censored Models
8.6.2 Dynamic Sample Selection Models
9. Incomplete Panel Data
9.1 Estimating Distributed Lags in Short Panels
9.1.1 Introduction
9.1.2 Common Assumptions
9.1.3 Identification Using Prior Structure of the Process of the
Exogenous Variable
9.1.4 Identification Using Prior Structure of the Lag Coefficients
9.1.5 Estimation and Testing
9.2 Rotating or Randomly Missing Data
9.3 Pseudopanels (or Repeated CrossSectional Data)
9.4 Pooling of a Single CrossSectional and a Single TimeSeries
Data Set
9.4.1 Introduction
9.4.2 The Likelihood Approach to Pooling CrossSectional and
TimeSeries Data
9.4.3 An Example
10. Miscellaneous Topics
10.1 Simulation Methods
10.2 Panels with Large N and T
10.3 UnitRoot Tests
10.4 Data with Multilevel Structures
10.5 Errors of Measurement
10.6 Modeling CrossSectional Dependence
11. A Summary View
11.1 Introduction
11.2 Benefits and Limitations of Panel Data
11.2.1 Increasing Degrees of Freedom and Lessening the Problem
of Multicollinearity
11.2.2 Identification and Discrimination between Competing
Hypotheses
11.2.3 Reducing Estimation Bias
11.2.3.a OmittedVariable Bias
11.2.3.b Bias Induced by the Dynamic Structure of a Model
11.2.3.c Simultaneity Bias
11.2.3.d Bias Induced by Measurement Errors
11.2.4 Providing Micro Foundations for Aggregate Data Analysis
11.3 Efficiency of the Estimates
Notes
References
Author Index
Subject Index